User login
Overanticoagulation in AF boosts dementia risk
CHICAGO – Patients with atrial fibrillation who frequently have a supratherapeutic international normalized ratio are at sharply increased risk for developing dementia, according to a large observational study.
“We postulate that the mechanism is an accumulation of microbleeds in the brain,” Dr. T. Jared Bunch said at the American Heart Association Scientific Sessions.
“In patients with hypertension, a condition that’s extremely common with atrial fibrillation, these repetitive small bleeds are preferentially in the hippocampus, where memory is stored,” added Dr. Bunch, who is medical director for heart rhythm services at the Intermountain Medical Center Heart Institute in Salt Lake City.
He presented a study of 1,031 patients with atrial fibrillation (AF) in Intermountain’s centralized anticoagulation service. All were on dual therapy with warfarin plus aspirin or, much less commonly, another antiplatelet agent. At baseline, their average age was in the early- to mid-70s, and none of the subjects had a history of stroke or any notes in their medical record suggestive of early cognitive decline. At this dedicated anticoagulation center, their INR was measured on a weekly or biweekly basis, as a result of which their average time spent in the therapeutic range of 2.0-3.0 was relatively high at nearly 70%.
The increased risk of dementia in patients with AF has previously been recognized. The association is stronger in patients under age 75 than in those who are older. But the mechanism has been unknown. Dr. Bunch and coinvestigators decided to test their hypothesis that the mechanism involves microbleeds secondary to long-term overanticoagulation by dividing the patients into three groups based upon their percentage of INR measurements above 3.0 during a mean follow-up of more than 4 and up to a maximum of 10 years: 240 patients had a supratherapeutic INR 25% of the time or more; 374 did so less than 10% of the time; and 417 had an elevated INR 10%-24% of the time.
The incidence of dementia diagnosed by a consultant neurologist during follow-up was 5.8% in the group with an INR above 3.0 at least 25% of the time, more than twice the 2.7% rate in patients with a high INR less than 10% of the time. In the middle group, the incidence of dementia was 4.1%. In a multivariate Cox regression analysis, having an INR above 3.0 on at least 25% of occasions was independently associated with a 2.59-fold increased risk of developing dementia, making it by far the most potent risk factor in their analysis.
The next step in their research will be to perform serial brain imaging and volumetric scans, Dr. Bunch said. Also, he and his coworkers are 3 years into an ongoing study looking at the incidence of dementia in AF patients on the various novel oral anticoagulants, where INR is a nonissue. Their hypothesis is the dementia risk will be lower than in patients on warfarin. Dr. Bunch has particularly high hopes for AF patients on apixaban (Eliquis) because it’s known to have a reduced risk of large bleeds, stroke, and GI bleeding; the hope is it will cause fewer cerebral microbleeds as well.
In an interview, the cardiologist said he believes his study showing an increased risk of dementia in AF patients with supratherapeutic INRs on warfarin plus antiplatelet therapy holds several important lessons for AF patients and physicians alike.
For patients, the message is don’t just start taking aspirin on your own because you’ve read it’s good for your heart or may reduce cancer risk; consult your physician.
And for physicians, it’s important to ask all patients on warfarin if they’re using aspirin; many don’t ask. Also, periodically reconsider the need for dual therapy with warfarin and aspirin.
“We find the risks of stroke and bleeding change dynamically over time, so the initial therapy for stroke prevention may not be the ideal therapy after 5-10 years,” Dr. Bunch said.
Lastly, for patients who are overanticoagulated on a substantial percentage of their INR measurements, it’s essential to consider a change in strategy. Either follow their INRs more closely and adjust warfarin dosing accordingly, or switch to one of the novel, more predictable oral anticoagulants, he concluded.
This study was funded internally by Intermountain Healthcare. Dr. Bunch reported having no financial conflicts.
CHICAGO – Patients with atrial fibrillation who frequently have a supratherapeutic international normalized ratio are at sharply increased risk for developing dementia, according to a large observational study.
“We postulate that the mechanism is an accumulation of microbleeds in the brain,” Dr. T. Jared Bunch said at the American Heart Association Scientific Sessions.
“In patients with hypertension, a condition that’s extremely common with atrial fibrillation, these repetitive small bleeds are preferentially in the hippocampus, where memory is stored,” added Dr. Bunch, who is medical director for heart rhythm services at the Intermountain Medical Center Heart Institute in Salt Lake City.
He presented a study of 1,031 patients with atrial fibrillation (AF) in Intermountain’s centralized anticoagulation service. All were on dual therapy with warfarin plus aspirin or, much less commonly, another antiplatelet agent. At baseline, their average age was in the early- to mid-70s, and none of the subjects had a history of stroke or any notes in their medical record suggestive of early cognitive decline. At this dedicated anticoagulation center, their INR was measured on a weekly or biweekly basis, as a result of which their average time spent in the therapeutic range of 2.0-3.0 was relatively high at nearly 70%.
The increased risk of dementia in patients with AF has previously been recognized. The association is stronger in patients under age 75 than in those who are older. But the mechanism has been unknown. Dr. Bunch and coinvestigators decided to test their hypothesis that the mechanism involves microbleeds secondary to long-term overanticoagulation by dividing the patients into three groups based upon their percentage of INR measurements above 3.0 during a mean follow-up of more than 4 and up to a maximum of 10 years: 240 patients had a supratherapeutic INR 25% of the time or more; 374 did so less than 10% of the time; and 417 had an elevated INR 10%-24% of the time.
The incidence of dementia diagnosed by a consultant neurologist during follow-up was 5.8% in the group with an INR above 3.0 at least 25% of the time, more than twice the 2.7% rate in patients with a high INR less than 10% of the time. In the middle group, the incidence of dementia was 4.1%. In a multivariate Cox regression analysis, having an INR above 3.0 on at least 25% of occasions was independently associated with a 2.59-fold increased risk of developing dementia, making it by far the most potent risk factor in their analysis.
The next step in their research will be to perform serial brain imaging and volumetric scans, Dr. Bunch said. Also, he and his coworkers are 3 years into an ongoing study looking at the incidence of dementia in AF patients on the various novel oral anticoagulants, where INR is a nonissue. Their hypothesis is the dementia risk will be lower than in patients on warfarin. Dr. Bunch has particularly high hopes for AF patients on apixaban (Eliquis) because it’s known to have a reduced risk of large bleeds, stroke, and GI bleeding; the hope is it will cause fewer cerebral microbleeds as well.
In an interview, the cardiologist said he believes his study showing an increased risk of dementia in AF patients with supratherapeutic INRs on warfarin plus antiplatelet therapy holds several important lessons for AF patients and physicians alike.
For patients, the message is don’t just start taking aspirin on your own because you’ve read it’s good for your heart or may reduce cancer risk; consult your physician.
And for physicians, it’s important to ask all patients on warfarin if they’re using aspirin; many don’t ask. Also, periodically reconsider the need for dual therapy with warfarin and aspirin.
“We find the risks of stroke and bleeding change dynamically over time, so the initial therapy for stroke prevention may not be the ideal therapy after 5-10 years,” Dr. Bunch said.
Lastly, for patients who are overanticoagulated on a substantial percentage of their INR measurements, it’s essential to consider a change in strategy. Either follow their INRs more closely and adjust warfarin dosing accordingly, or switch to one of the novel, more predictable oral anticoagulants, he concluded.
This study was funded internally by Intermountain Healthcare. Dr. Bunch reported having no financial conflicts.
CHICAGO – Patients with atrial fibrillation who frequently have a supratherapeutic international normalized ratio are at sharply increased risk for developing dementia, according to a large observational study.
“We postulate that the mechanism is an accumulation of microbleeds in the brain,” Dr. T. Jared Bunch said at the American Heart Association Scientific Sessions.
“In patients with hypertension, a condition that’s extremely common with atrial fibrillation, these repetitive small bleeds are preferentially in the hippocampus, where memory is stored,” added Dr. Bunch, who is medical director for heart rhythm services at the Intermountain Medical Center Heart Institute in Salt Lake City.
He presented a study of 1,031 patients with atrial fibrillation (AF) in Intermountain’s centralized anticoagulation service. All were on dual therapy with warfarin plus aspirin or, much less commonly, another antiplatelet agent. At baseline, their average age was in the early- to mid-70s, and none of the subjects had a history of stroke or any notes in their medical record suggestive of early cognitive decline. At this dedicated anticoagulation center, their INR was measured on a weekly or biweekly basis, as a result of which their average time spent in the therapeutic range of 2.0-3.0 was relatively high at nearly 70%.
The increased risk of dementia in patients with AF has previously been recognized. The association is stronger in patients under age 75 than in those who are older. But the mechanism has been unknown. Dr. Bunch and coinvestigators decided to test their hypothesis that the mechanism involves microbleeds secondary to long-term overanticoagulation by dividing the patients into three groups based upon their percentage of INR measurements above 3.0 during a mean follow-up of more than 4 and up to a maximum of 10 years: 240 patients had a supratherapeutic INR 25% of the time or more; 374 did so less than 10% of the time; and 417 had an elevated INR 10%-24% of the time.
The incidence of dementia diagnosed by a consultant neurologist during follow-up was 5.8% in the group with an INR above 3.0 at least 25% of the time, more than twice the 2.7% rate in patients with a high INR less than 10% of the time. In the middle group, the incidence of dementia was 4.1%. In a multivariate Cox regression analysis, having an INR above 3.0 on at least 25% of occasions was independently associated with a 2.59-fold increased risk of developing dementia, making it by far the most potent risk factor in their analysis.
The next step in their research will be to perform serial brain imaging and volumetric scans, Dr. Bunch said. Also, he and his coworkers are 3 years into an ongoing study looking at the incidence of dementia in AF patients on the various novel oral anticoagulants, where INR is a nonissue. Their hypothesis is the dementia risk will be lower than in patients on warfarin. Dr. Bunch has particularly high hopes for AF patients on apixaban (Eliquis) because it’s known to have a reduced risk of large bleeds, stroke, and GI bleeding; the hope is it will cause fewer cerebral microbleeds as well.
In an interview, the cardiologist said he believes his study showing an increased risk of dementia in AF patients with supratherapeutic INRs on warfarin plus antiplatelet therapy holds several important lessons for AF patients and physicians alike.
For patients, the message is don’t just start taking aspirin on your own because you’ve read it’s good for your heart or may reduce cancer risk; consult your physician.
And for physicians, it’s important to ask all patients on warfarin if they’re using aspirin; many don’t ask. Also, periodically reconsider the need for dual therapy with warfarin and aspirin.
“We find the risks of stroke and bleeding change dynamically over time, so the initial therapy for stroke prevention may not be the ideal therapy after 5-10 years,” Dr. Bunch said.
Lastly, for patients who are overanticoagulated on a substantial percentage of their INR measurements, it’s essential to consider a change in strategy. Either follow their INRs more closely and adjust warfarin dosing accordingly, or switch to one of the novel, more predictable oral anticoagulants, he concluded.
This study was funded internally by Intermountain Healthcare. Dr. Bunch reported having no financial conflicts.
AT THE AHA SCIENTIFIC SESSIONS
Key clinical point: The more time patients with atrial fibrillation who are on warfarin plus aspirin spend with a supratherapeutic INR, the greater their risk of developing dementia.
Major finding: Atrial fibrillation patients on warfarin plus an antiplatelet agent who had an INR above 3.0 on at least 25% of occasions had a 5.8% incidence of dementia during follow-up, compared with a 2.7% incidence in those with a high INR less than 10% of the time.
Data source: This was a retrospective, case-control study involving 1,031 patients with atrial fibrillation on warfarin plus aspirin or another antiplatelet agent followed for a mean of more than 4 years.
Disclosures: This study was funded internally by Intermountain Healthcare. Dr. Bunch reported having no financial conflicts.
LISTEN NOW: The Doctor as Patient
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/Doctor-As-Patient.mp3"][/audio]
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/Doctor-As-Patient.mp3"][/audio]
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/Doctor-As-Patient.mp3"][/audio]
LISTEN NOW: Highlights of the January 2015 issue of The Hospitalist
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/2015-January-Highlights.mp3"][/audio]
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/2015-January-Highlights.mp3"][/audio]
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/2015-January-Highlights.mp3"][/audio]
LISTEN NOW: Co-Management in Hospital Medicine
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/Hospital-Medicine-co-management-Jan2015.mp3"][/audio]
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/Hospital-Medicine-co-management-Jan2015.mp3"][/audio]
[audio mp3="http://www.the-hospitalist.org/wp-content/uploads/2015/01/Hospital-Medicine-co-management-Jan2015.mp3"][/audio]
Older patients benefit from brentuximab treatment
Credit: NIH
SAN FRANCISCO—Younger patients with Hodgkin lymphoma fare well on brentuximab vedotin, experiencing an overall objective response rate (ORR) of 75% and a complete response (CR) rate of 34% in the pivotal phase 2 study of patients with relapsed/refractory disease.
And a retrospective study of patients older than 60 years showed that single-agent therapy was well tolerated, prompting an ORR of 53% and a CR rate of 40% in a relapsed or refractory population.
So investigators decided to explore in a prospective study whether patients 60 years or older could benefit from up-front treatment with brentuximab as a single agent or in combination.
Andres Forero-Torres, MD, of the University of Alabama in Birmingham, presented the results of this trial at the 2014 ASH Annual Meeting (abstract 294).*
Enrolled patients had classic Hodgkin lymphoma, were treatment-naïve, and were ineligible for or declined conventional front-line treatment. The primary endpoint was ORR.
The study is being conducted in 3 parts—brentuximab as a single agent, brentuximab plus dacarbazine, and brentuximab plus bendamustine. At the time of the ASH presentation, data for the brentuximab-bendamustine combination were not available.
Single-agent brentuximab
Twenty-seven patients on the single-agent arm were evaluable for efficacy and safety. They were a median age of 78 (range, 64 to 92). About half (52%) were male, and 78% had an ECOG performance status of grade 0 or 1.
Forty-four percent had moderate renal function impairment with a creatinine clearance between 30 and 60 mL/min. Thirty percent had B symptoms, 22% had bulky disease, and 52% had extra-nodal involvement.
Patients received 1.8 mg/kg of brentuximab intravenously on day 1 of a 21-day cycle. Response was assessed by CT scan during cycles 2, 4, 8, and 16, and by CT plus PET scan during cycles 2 and 8.
The median follow-up was 8.7 months. Dr Forero-Torres pointed out that, initially, “there were no progressions,” and all patients achieved tumor reduction.
The ORR was 93%, the CR rate was 70%, the partial response rate was 22%, and the rate of stable disease was 7%.
The median duration of response was 9.1 months (range, 0.03 to 13.14), and the median progression-free survival was 10.5 months (range, 2.6 to 14.3). For patients who had a CR, the median progression-free survival was about 12 months, Dr Forero-Torres said.
The median number of treatment cycles administered per patient was 8 (range, 3 to 23). Patients discontinued treatment primarily because of progressive disease (41%) or adverse events (AEs, 37%).
AEs occurring in 20% or more of patients were constipation, decreased appetite, diarrhea, peripheral edema, nausea, fatigue, and peripheral sensory neuropathy. All were grade 1 or 2, except for peripheral sensory neuropathy, which also had about 20% grade 3 events.
Grade 3 or higher treatment-related AEs included peripheral sensory neuropathy (n=7), peripheral motor neuropathy (n=2), rash (n=2), and 1 patient each with anemia, increased aspartate aminotransferase, asthenia, neutropenia, orthostatic hypotension, generalized rash, and maculopapular rash.
Serious AEs (SAEs) were minimal, Dr Forero-Torres said, and included 1 patient each with pyrexia, orthostatic hypotension, asthenia and rash, and deep vein thrombosis.
Seven patients discontinued treatment due to peripheral sensory neuropathy, 2 due to peripheral motor neuropathy, and 1 due to orthostatic hypotension.
Dr Forero-Torres emphasized that there were no grade 4 AEs, no AE-related deaths, and no deaths within 30 days of the last dose of medication.
Brentuximab plus dacarbazine
Fourteen of 18 patients in the combination arm were evaluable for efficacy and safety. Their median age was 72.5 (range, 62 to 87), 72% were male, 67% had an ECOG status of grade 0 or 1, and 56% had normal renal function with a creatinine clearance greater than 80 mL/min.
Forty-four percent exhibited B symptoms, 11% had bulky disease, and 50% had extra-nodal involvement.
They received brentuximab at 1.8 mg/kg plus dacarbazine at 375 mg/m2 for cycles 1-12, followed by monotherapy for cycles 13-16.
At the time of the interim analysis, 83% of patients were still on treatment, “so this is very early preliminary data,” Dr Forero-Torres noted.
All of the patients achieved tumor reduction, and 4 patients achieved a CR.
They had a median treatment duration of 16.7 weeks (range, 3 to 36), received a median of 5.5 cycles (range, 1 to 12), and had a median follow-up time of 19.1 weeks (range, 6.1 to 36.1).
The most common grade 1 or 2 AEs were peripheral sensory neuropathy (33%), nausea (33%), diarrhea (28%), constipation (28%), fatigue (22%), alopecia (22%), arthralgia (22%), and headache (22%).
Grade 3 AEs or SAEs, with 1 patient each, were C difficile colitis (SAE), hypotension (SAE), and hyperglycemia.
Dr Forero-Torres noted that investigators observed “robust antitumor activity” among these older patients receiving front-line brentuximab.
The cohort combining brentuximab with bendamustine is currently enrolling patients.
The study is sponsored by Seattle Genetics, Inc., developer of brentuximab vedotin (Adcetris).
*Information in the abstract differs from that presented at the meeting.
Credit: NIH
SAN FRANCISCO—Younger patients with Hodgkin lymphoma fare well on brentuximab vedotin, experiencing an overall objective response rate (ORR) of 75% and a complete response (CR) rate of 34% in the pivotal phase 2 study of patients with relapsed/refractory disease.
And a retrospective study of patients older than 60 years showed that single-agent therapy was well tolerated, prompting an ORR of 53% and a CR rate of 40% in a relapsed or refractory population.
So investigators decided to explore in a prospective study whether patients 60 years or older could benefit from up-front treatment with brentuximab as a single agent or in combination.
Andres Forero-Torres, MD, of the University of Alabama in Birmingham, presented the results of this trial at the 2014 ASH Annual Meeting (abstract 294).*
Enrolled patients had classic Hodgkin lymphoma, were treatment-naïve, and were ineligible for or declined conventional front-line treatment. The primary endpoint was ORR.
The study is being conducted in 3 parts—brentuximab as a single agent, brentuximab plus dacarbazine, and brentuximab plus bendamustine. At the time of the ASH presentation, data for the brentuximab-bendamustine combination were not available.
Single-agent brentuximab
Twenty-seven patients on the single-agent arm were evaluable for efficacy and safety. They were a median age of 78 (range, 64 to 92). About half (52%) were male, and 78% had an ECOG performance status of grade 0 or 1.
Forty-four percent had moderate renal function impairment with a creatinine clearance between 30 and 60 mL/min. Thirty percent had B symptoms, 22% had bulky disease, and 52% had extra-nodal involvement.
Patients received 1.8 mg/kg of brentuximab intravenously on day 1 of a 21-day cycle. Response was assessed by CT scan during cycles 2, 4, 8, and 16, and by CT plus PET scan during cycles 2 and 8.
The median follow-up was 8.7 months. Dr Forero-Torres pointed out that, initially, “there were no progressions,” and all patients achieved tumor reduction.
The ORR was 93%, the CR rate was 70%, the partial response rate was 22%, and the rate of stable disease was 7%.
The median duration of response was 9.1 months (range, 0.03 to 13.14), and the median progression-free survival was 10.5 months (range, 2.6 to 14.3). For patients who had a CR, the median progression-free survival was about 12 months, Dr Forero-Torres said.
The median number of treatment cycles administered per patient was 8 (range, 3 to 23). Patients discontinued treatment primarily because of progressive disease (41%) or adverse events (AEs, 37%).
AEs occurring in 20% or more of patients were constipation, decreased appetite, diarrhea, peripheral edema, nausea, fatigue, and peripheral sensory neuropathy. All were grade 1 or 2, except for peripheral sensory neuropathy, which also had about 20% grade 3 events.
Grade 3 or higher treatment-related AEs included peripheral sensory neuropathy (n=7), peripheral motor neuropathy (n=2), rash (n=2), and 1 patient each with anemia, increased aspartate aminotransferase, asthenia, neutropenia, orthostatic hypotension, generalized rash, and maculopapular rash.
Serious AEs (SAEs) were minimal, Dr Forero-Torres said, and included 1 patient each with pyrexia, orthostatic hypotension, asthenia and rash, and deep vein thrombosis.
Seven patients discontinued treatment due to peripheral sensory neuropathy, 2 due to peripheral motor neuropathy, and 1 due to orthostatic hypotension.
Dr Forero-Torres emphasized that there were no grade 4 AEs, no AE-related deaths, and no deaths within 30 days of the last dose of medication.
Brentuximab plus dacarbazine
Fourteen of 18 patients in the combination arm were evaluable for efficacy and safety. Their median age was 72.5 (range, 62 to 87), 72% were male, 67% had an ECOG status of grade 0 or 1, and 56% had normal renal function with a creatinine clearance greater than 80 mL/min.
Forty-four percent exhibited B symptoms, 11% had bulky disease, and 50% had extra-nodal involvement.
They received brentuximab at 1.8 mg/kg plus dacarbazine at 375 mg/m2 for cycles 1-12, followed by monotherapy for cycles 13-16.
At the time of the interim analysis, 83% of patients were still on treatment, “so this is very early preliminary data,” Dr Forero-Torres noted.
All of the patients achieved tumor reduction, and 4 patients achieved a CR.
They had a median treatment duration of 16.7 weeks (range, 3 to 36), received a median of 5.5 cycles (range, 1 to 12), and had a median follow-up time of 19.1 weeks (range, 6.1 to 36.1).
The most common grade 1 or 2 AEs were peripheral sensory neuropathy (33%), nausea (33%), diarrhea (28%), constipation (28%), fatigue (22%), alopecia (22%), arthralgia (22%), and headache (22%).
Grade 3 AEs or SAEs, with 1 patient each, were C difficile colitis (SAE), hypotension (SAE), and hyperglycemia.
Dr Forero-Torres noted that investigators observed “robust antitumor activity” among these older patients receiving front-line brentuximab.
The cohort combining brentuximab with bendamustine is currently enrolling patients.
The study is sponsored by Seattle Genetics, Inc., developer of brentuximab vedotin (Adcetris).
*Information in the abstract differs from that presented at the meeting.
Credit: NIH
SAN FRANCISCO—Younger patients with Hodgkin lymphoma fare well on brentuximab vedotin, experiencing an overall objective response rate (ORR) of 75% and a complete response (CR) rate of 34% in the pivotal phase 2 study of patients with relapsed/refractory disease.
And a retrospective study of patients older than 60 years showed that single-agent therapy was well tolerated, prompting an ORR of 53% and a CR rate of 40% in a relapsed or refractory population.
So investigators decided to explore in a prospective study whether patients 60 years or older could benefit from up-front treatment with brentuximab as a single agent or in combination.
Andres Forero-Torres, MD, of the University of Alabama in Birmingham, presented the results of this trial at the 2014 ASH Annual Meeting (abstract 294).*
Enrolled patients had classic Hodgkin lymphoma, were treatment-naïve, and were ineligible for or declined conventional front-line treatment. The primary endpoint was ORR.
The study is being conducted in 3 parts—brentuximab as a single agent, brentuximab plus dacarbazine, and brentuximab plus bendamustine. At the time of the ASH presentation, data for the brentuximab-bendamustine combination were not available.
Single-agent brentuximab
Twenty-seven patients on the single-agent arm were evaluable for efficacy and safety. They were a median age of 78 (range, 64 to 92). About half (52%) were male, and 78% had an ECOG performance status of grade 0 or 1.
Forty-four percent had moderate renal function impairment with a creatinine clearance between 30 and 60 mL/min. Thirty percent had B symptoms, 22% had bulky disease, and 52% had extra-nodal involvement.
Patients received 1.8 mg/kg of brentuximab intravenously on day 1 of a 21-day cycle. Response was assessed by CT scan during cycles 2, 4, 8, and 16, and by CT plus PET scan during cycles 2 and 8.
The median follow-up was 8.7 months. Dr Forero-Torres pointed out that, initially, “there were no progressions,” and all patients achieved tumor reduction.
The ORR was 93%, the CR rate was 70%, the partial response rate was 22%, and the rate of stable disease was 7%.
The median duration of response was 9.1 months (range, 0.03 to 13.14), and the median progression-free survival was 10.5 months (range, 2.6 to 14.3). For patients who had a CR, the median progression-free survival was about 12 months, Dr Forero-Torres said.
The median number of treatment cycles administered per patient was 8 (range, 3 to 23). Patients discontinued treatment primarily because of progressive disease (41%) or adverse events (AEs, 37%).
AEs occurring in 20% or more of patients were constipation, decreased appetite, diarrhea, peripheral edema, nausea, fatigue, and peripheral sensory neuropathy. All were grade 1 or 2, except for peripheral sensory neuropathy, which also had about 20% grade 3 events.
Grade 3 or higher treatment-related AEs included peripheral sensory neuropathy (n=7), peripheral motor neuropathy (n=2), rash (n=2), and 1 patient each with anemia, increased aspartate aminotransferase, asthenia, neutropenia, orthostatic hypotension, generalized rash, and maculopapular rash.
Serious AEs (SAEs) were minimal, Dr Forero-Torres said, and included 1 patient each with pyrexia, orthostatic hypotension, asthenia and rash, and deep vein thrombosis.
Seven patients discontinued treatment due to peripheral sensory neuropathy, 2 due to peripheral motor neuropathy, and 1 due to orthostatic hypotension.
Dr Forero-Torres emphasized that there were no grade 4 AEs, no AE-related deaths, and no deaths within 30 days of the last dose of medication.
Brentuximab plus dacarbazine
Fourteen of 18 patients in the combination arm were evaluable for efficacy and safety. Their median age was 72.5 (range, 62 to 87), 72% were male, 67% had an ECOG status of grade 0 or 1, and 56% had normal renal function with a creatinine clearance greater than 80 mL/min.
Forty-four percent exhibited B symptoms, 11% had bulky disease, and 50% had extra-nodal involvement.
They received brentuximab at 1.8 mg/kg plus dacarbazine at 375 mg/m2 for cycles 1-12, followed by monotherapy for cycles 13-16.
At the time of the interim analysis, 83% of patients were still on treatment, “so this is very early preliminary data,” Dr Forero-Torres noted.
All of the patients achieved tumor reduction, and 4 patients achieved a CR.
They had a median treatment duration of 16.7 weeks (range, 3 to 36), received a median of 5.5 cycles (range, 1 to 12), and had a median follow-up time of 19.1 weeks (range, 6.1 to 36.1).
The most common grade 1 or 2 AEs were peripheral sensory neuropathy (33%), nausea (33%), diarrhea (28%), constipation (28%), fatigue (22%), alopecia (22%), arthralgia (22%), and headache (22%).
Grade 3 AEs or SAEs, with 1 patient each, were C difficile colitis (SAE), hypotension (SAE), and hyperglycemia.
Dr Forero-Torres noted that investigators observed “robust antitumor activity” among these older patients receiving front-line brentuximab.
The cohort combining brentuximab with bendamustine is currently enrolling patients.
The study is sponsored by Seattle Genetics, Inc., developer of brentuximab vedotin (Adcetris).
*Information in the abstract differs from that presented at the meeting.
CRISPR bests TALEN in iPSCs
Credit: Salk Institute
The gene-editing technology CRISPR can precisely and efficiently alter human stem cells, according to research published in Molecular Therapy.
Using JAK2 and other genes as models, researchers showed that CRISPR offers advantages over TALEN, another gene-editing technique, for manipulating induced pluripotent stem cells (iPSCs).
And, unlike in a previous study, CRISPR did not produce any off-target effects.
The team believes their findings could streamline and speed up efforts to modify human iPSCs for use as treatments or in the development of model systems to study diseases and test drugs.
“Stem cell technology is quickly advancing, and we think that the days when we can use iPSCs for human therapy aren’t that far away,” said study author Zhaohui Ye, PhD, of the Johns Hopkins University School of Medicine in Baltimore, Maryland.
“This is one of the first studies to detail the use of CRISPR in human iPSCs, showcasing its potential in these cells.”
CRISPR originated from a microbial immune system that contains DNA segments known as “clustered regularly interspaced short palindromic repeats.” The system makes use of an enzyme that nicks together DNA with a piece of small RNA that guides the tool to where researchers want to introduce cuts or other changes in the genome.
Previous research has shown that CRISPR can generate genomic changes or mutations through these interventions more efficiently than other gene-editing techniques, such as TALEN, which is short for “transcription activator-like effector nuclease.”
Despite CRISPR’s advantages, a recent study suggested it might also produce a large number of off-target effects in human cancer cell lines; specifically, modification of genes that researchers didn’t mean to change.
To see if this unwanted effect occurred in other human cell types, Dr Ye and his colleagues pitted CRISPR against TALEN in human iPSCs. The researchers compared the ability of both techniques to either cut out pieces of known genes in iPSCs or cut out a piece of these genes and replace it with another.
As model genes, the researchers used JAK2, a gene that, when mutated, causes myeloproliferative neoplasms; SERPINA1, a gene that, when mutated, causes alpha1-antitrypsin deficiency, an inherited disorder that may cause lung and liver disease; and AAVS1, a gene that’s been recently discovered to be a “safe harbor” in the human genome for inserting foreign genes.
The comparison showed that, when simply cutting out portions of genes, the CRISPR system was significantly more efficient than TALEN in all 3 gene systems, inducing up to 100 times more cuts.
However, when using these genome-editing tools for replacing portions of the genes, such as the disease-causing mutations in JAK2 and SERPINA1 genes, CRISPR and TALEN showed about the same efficiency in patient-derived iPSCs.
Contrary to results of the human cancer cell line study, both CRISPR and TALEN had the same targeting specificity in human iPSCs, hitting only the genes they were designed to affect.
The researchers also found that the CRISPR system has a second advantage over TALEN. It can be designed to target only the mutation-containing gene without affecting the healthy gene in patients where only one copy of a gene is affected.
These findings, according to the researchers, offer reassurance that CRISPR will be a useful tool for editing the genes of human iPSCs with little risk of off-target effects.
Credit: Salk Institute
The gene-editing technology CRISPR can precisely and efficiently alter human stem cells, according to research published in Molecular Therapy.
Using JAK2 and other genes as models, researchers showed that CRISPR offers advantages over TALEN, another gene-editing technique, for manipulating induced pluripotent stem cells (iPSCs).
And, unlike in a previous study, CRISPR did not produce any off-target effects.
The team believes their findings could streamline and speed up efforts to modify human iPSCs for use as treatments or in the development of model systems to study diseases and test drugs.
“Stem cell technology is quickly advancing, and we think that the days when we can use iPSCs for human therapy aren’t that far away,” said study author Zhaohui Ye, PhD, of the Johns Hopkins University School of Medicine in Baltimore, Maryland.
“This is one of the first studies to detail the use of CRISPR in human iPSCs, showcasing its potential in these cells.”
CRISPR originated from a microbial immune system that contains DNA segments known as “clustered regularly interspaced short palindromic repeats.” The system makes use of an enzyme that nicks together DNA with a piece of small RNA that guides the tool to where researchers want to introduce cuts or other changes in the genome.
Previous research has shown that CRISPR can generate genomic changes or mutations through these interventions more efficiently than other gene-editing techniques, such as TALEN, which is short for “transcription activator-like effector nuclease.”
Despite CRISPR’s advantages, a recent study suggested it might also produce a large number of off-target effects in human cancer cell lines; specifically, modification of genes that researchers didn’t mean to change.
To see if this unwanted effect occurred in other human cell types, Dr Ye and his colleagues pitted CRISPR against TALEN in human iPSCs. The researchers compared the ability of both techniques to either cut out pieces of known genes in iPSCs or cut out a piece of these genes and replace it with another.
As model genes, the researchers used JAK2, a gene that, when mutated, causes myeloproliferative neoplasms; SERPINA1, a gene that, when mutated, causes alpha1-antitrypsin deficiency, an inherited disorder that may cause lung and liver disease; and AAVS1, a gene that’s been recently discovered to be a “safe harbor” in the human genome for inserting foreign genes.
The comparison showed that, when simply cutting out portions of genes, the CRISPR system was significantly more efficient than TALEN in all 3 gene systems, inducing up to 100 times more cuts.
However, when using these genome-editing tools for replacing portions of the genes, such as the disease-causing mutations in JAK2 and SERPINA1 genes, CRISPR and TALEN showed about the same efficiency in patient-derived iPSCs.
Contrary to results of the human cancer cell line study, both CRISPR and TALEN had the same targeting specificity in human iPSCs, hitting only the genes they were designed to affect.
The researchers also found that the CRISPR system has a second advantage over TALEN. It can be designed to target only the mutation-containing gene without affecting the healthy gene in patients where only one copy of a gene is affected.
These findings, according to the researchers, offer reassurance that CRISPR will be a useful tool for editing the genes of human iPSCs with little risk of off-target effects.
Credit: Salk Institute
The gene-editing technology CRISPR can precisely and efficiently alter human stem cells, according to research published in Molecular Therapy.
Using JAK2 and other genes as models, researchers showed that CRISPR offers advantages over TALEN, another gene-editing technique, for manipulating induced pluripotent stem cells (iPSCs).
And, unlike in a previous study, CRISPR did not produce any off-target effects.
The team believes their findings could streamline and speed up efforts to modify human iPSCs for use as treatments or in the development of model systems to study diseases and test drugs.
“Stem cell technology is quickly advancing, and we think that the days when we can use iPSCs for human therapy aren’t that far away,” said study author Zhaohui Ye, PhD, of the Johns Hopkins University School of Medicine in Baltimore, Maryland.
“This is one of the first studies to detail the use of CRISPR in human iPSCs, showcasing its potential in these cells.”
CRISPR originated from a microbial immune system that contains DNA segments known as “clustered regularly interspaced short palindromic repeats.” The system makes use of an enzyme that nicks together DNA with a piece of small RNA that guides the tool to where researchers want to introduce cuts or other changes in the genome.
Previous research has shown that CRISPR can generate genomic changes or mutations through these interventions more efficiently than other gene-editing techniques, such as TALEN, which is short for “transcription activator-like effector nuclease.”
Despite CRISPR’s advantages, a recent study suggested it might also produce a large number of off-target effects in human cancer cell lines; specifically, modification of genes that researchers didn’t mean to change.
To see if this unwanted effect occurred in other human cell types, Dr Ye and his colleagues pitted CRISPR against TALEN in human iPSCs. The researchers compared the ability of both techniques to either cut out pieces of known genes in iPSCs or cut out a piece of these genes and replace it with another.
As model genes, the researchers used JAK2, a gene that, when mutated, causes myeloproliferative neoplasms; SERPINA1, a gene that, when mutated, causes alpha1-antitrypsin deficiency, an inherited disorder that may cause lung and liver disease; and AAVS1, a gene that’s been recently discovered to be a “safe harbor” in the human genome for inserting foreign genes.
The comparison showed that, when simply cutting out portions of genes, the CRISPR system was significantly more efficient than TALEN in all 3 gene systems, inducing up to 100 times more cuts.
However, when using these genome-editing tools for replacing portions of the genes, such as the disease-causing mutations in JAK2 and SERPINA1 genes, CRISPR and TALEN showed about the same efficiency in patient-derived iPSCs.
Contrary to results of the human cancer cell line study, both CRISPR and TALEN had the same targeting specificity in human iPSCs, hitting only the genes they were designed to affect.
The researchers also found that the CRISPR system has a second advantage over TALEN. It can be designed to target only the mutation-containing gene without affecting the healthy gene in patients where only one copy of a gene is affected.
These findings, according to the researchers, offer reassurance that CRISPR will be a useful tool for editing the genes of human iPSCs with little risk of off-target effects.
Ibrutinib proves active in high-risk CLL
Credit: Mary Ann Thompson
Single-agent ibrutinib can elicit a high response rate in patients with high-risk chronic lymphocytic leukemia (CLL), results of a phase 2 trial suggest.
The Bruton’s tyrosine kinase inhibitor prompted a 92% objective response rate in patients who had previously untreated or relapsed/refractory CLL with either 17p deletion (del 17p) or tumor protein 53 (TP53) aberrations.
Researchers reported this and other results of the trial in The Lancet Oncology.
“Ibrutinib treatment results observed in CLL patients with del 17p or TP53 aberrations are very encouraging given that these patients have a high relapse rate after chemotherapy and are in need of tolerable, effective, and durable treatment options,” said study author Mohammed Farooqui, DO, of the National Heart, Lung, and Blood Institute in Bethesda, Maryland.
He and his colleagues studied 51 patients in this trial, 35 with previously untreated CLL and 16 with relapsed or refractory CLL. Forty-seven of the patients (92%) had del 17p, and 4 patients carried the TP53 aberration but did not have del 17p.
The study’s primary endpoint was overall response rate after 24 weeks. Secondary endpoints included safety, overall survival, progression-free survival, best response, and nodal response.
The median follow-up for all patients was 24 months (15 months for the previously untreated cohort). At 24 weeks, 48 patients were evaluable for response, assessed according to the modified IWCLL 2008 criteria.
Response rates
At 24 weeks, 92% (n=44) of the 48 evaluable patients achieved an objective response. Fifty percent of all evaluable patients achieved a partial response (n=24)—55% of previously untreated patients (n=18) and 40% of relapsed/refractory patients (n=6).
As for best response, 10% of all patients achieved a complete response (n=5)—12% of previously untreated patients (n=4) and 7% of relapsed/refractory patients (n=1). And 67% of patients had a partial response (n=32)—70% of previously untreated patients (n=23) and 60% of relapsed/refractory patients (n=9).
After 8 weeks on therapy, ibrutinib was associated with a more than 50% mean reduction in tumor burden in the bone marrow (44%), lymph nodes (70%), and spleen (79%). After 24 weeks of therapy, the rates of tumor burden reduction (> 50%) increased to 83%, 93%, and 95%, respectively.
Survival and safety
The estimated progression-free survival at 24 months for all patients on an intention-to-treat basis was 82%. Forty-two of the 51 patients (82%) continued on ibrutinib treatment without disease progression.
The estimated overall survival at 24 months was 80% for all patients—84% for previously untreated patients and 74% for patients with relapsed or refractory disease.
At the final follow-up, 8 (16%) patients had died—5 (10%) from progressive disease, 2 (4%) from infection, and 1 (2%) patient with a sudden, unexplained death that may have been treatment-related.
The most common adverse events (occurring in more than 30% of all patients) potentially related to ibrutinib were arthralgia (59%), diarrhea (51%), rash (47%), nail ridging (43%), bruising (33%), and muscle spasms (31%).
The most frequent grade 3 or 4 hematologic adverse events were neutropenia (24%), anemia (14%), and thrombocytopenia (10%). The most common nonhematologic grade 3 adverse event was pneumonia, which occurred in 3 patients (6%).
Nine patients (18%) discontinued treatment. The reasons for discontinuation included disease progression in 5 patients (10%) and death for 3 patients (6%).
This research was sponsored by the Intramural Research Program of the National Heart, Lung, and Blood Institute and the National Cancer Institute; Danish Cancer Society; Novo Nordisk Foundation; National Institutes of Health Medical Research Scholars Program; and Pharmacyclics Inc.
Ibrutinib is jointly developed and commercialized by Pharmacyclics and Janssen Biotech, Inc.
Credit: Mary Ann Thompson
Single-agent ibrutinib can elicit a high response rate in patients with high-risk chronic lymphocytic leukemia (CLL), results of a phase 2 trial suggest.
The Bruton’s tyrosine kinase inhibitor prompted a 92% objective response rate in patients who had previously untreated or relapsed/refractory CLL with either 17p deletion (del 17p) or tumor protein 53 (TP53) aberrations.
Researchers reported this and other results of the trial in The Lancet Oncology.
“Ibrutinib treatment results observed in CLL patients with del 17p or TP53 aberrations are very encouraging given that these patients have a high relapse rate after chemotherapy and are in need of tolerable, effective, and durable treatment options,” said study author Mohammed Farooqui, DO, of the National Heart, Lung, and Blood Institute in Bethesda, Maryland.
He and his colleagues studied 51 patients in this trial, 35 with previously untreated CLL and 16 with relapsed or refractory CLL. Forty-seven of the patients (92%) had del 17p, and 4 patients carried the TP53 aberration but did not have del 17p.
The study’s primary endpoint was overall response rate after 24 weeks. Secondary endpoints included safety, overall survival, progression-free survival, best response, and nodal response.
The median follow-up for all patients was 24 months (15 months for the previously untreated cohort). At 24 weeks, 48 patients were evaluable for response, assessed according to the modified IWCLL 2008 criteria.
Response rates
At 24 weeks, 92% (n=44) of the 48 evaluable patients achieved an objective response. Fifty percent of all evaluable patients achieved a partial response (n=24)—55% of previously untreated patients (n=18) and 40% of relapsed/refractory patients (n=6).
As for best response, 10% of all patients achieved a complete response (n=5)—12% of previously untreated patients (n=4) and 7% of relapsed/refractory patients (n=1). And 67% of patients had a partial response (n=32)—70% of previously untreated patients (n=23) and 60% of relapsed/refractory patients (n=9).
After 8 weeks on therapy, ibrutinib was associated with a more than 50% mean reduction in tumor burden in the bone marrow (44%), lymph nodes (70%), and spleen (79%). After 24 weeks of therapy, the rates of tumor burden reduction (> 50%) increased to 83%, 93%, and 95%, respectively.
Survival and safety
The estimated progression-free survival at 24 months for all patients on an intention-to-treat basis was 82%. Forty-two of the 51 patients (82%) continued on ibrutinib treatment without disease progression.
The estimated overall survival at 24 months was 80% for all patients—84% for previously untreated patients and 74% for patients with relapsed or refractory disease.
At the final follow-up, 8 (16%) patients had died—5 (10%) from progressive disease, 2 (4%) from infection, and 1 (2%) patient with a sudden, unexplained death that may have been treatment-related.
The most common adverse events (occurring in more than 30% of all patients) potentially related to ibrutinib were arthralgia (59%), diarrhea (51%), rash (47%), nail ridging (43%), bruising (33%), and muscle spasms (31%).
The most frequent grade 3 or 4 hematologic adverse events were neutropenia (24%), anemia (14%), and thrombocytopenia (10%). The most common nonhematologic grade 3 adverse event was pneumonia, which occurred in 3 patients (6%).
Nine patients (18%) discontinued treatment. The reasons for discontinuation included disease progression in 5 patients (10%) and death for 3 patients (6%).
This research was sponsored by the Intramural Research Program of the National Heart, Lung, and Blood Institute and the National Cancer Institute; Danish Cancer Society; Novo Nordisk Foundation; National Institutes of Health Medical Research Scholars Program; and Pharmacyclics Inc.
Ibrutinib is jointly developed and commercialized by Pharmacyclics and Janssen Biotech, Inc.
Credit: Mary Ann Thompson
Single-agent ibrutinib can elicit a high response rate in patients with high-risk chronic lymphocytic leukemia (CLL), results of a phase 2 trial suggest.
The Bruton’s tyrosine kinase inhibitor prompted a 92% objective response rate in patients who had previously untreated or relapsed/refractory CLL with either 17p deletion (del 17p) or tumor protein 53 (TP53) aberrations.
Researchers reported this and other results of the trial in The Lancet Oncology.
“Ibrutinib treatment results observed in CLL patients with del 17p or TP53 aberrations are very encouraging given that these patients have a high relapse rate after chemotherapy and are in need of tolerable, effective, and durable treatment options,” said study author Mohammed Farooqui, DO, of the National Heart, Lung, and Blood Institute in Bethesda, Maryland.
He and his colleagues studied 51 patients in this trial, 35 with previously untreated CLL and 16 with relapsed or refractory CLL. Forty-seven of the patients (92%) had del 17p, and 4 patients carried the TP53 aberration but did not have del 17p.
The study’s primary endpoint was overall response rate after 24 weeks. Secondary endpoints included safety, overall survival, progression-free survival, best response, and nodal response.
The median follow-up for all patients was 24 months (15 months for the previously untreated cohort). At 24 weeks, 48 patients were evaluable for response, assessed according to the modified IWCLL 2008 criteria.
Response rates
At 24 weeks, 92% (n=44) of the 48 evaluable patients achieved an objective response. Fifty percent of all evaluable patients achieved a partial response (n=24)—55% of previously untreated patients (n=18) and 40% of relapsed/refractory patients (n=6).
As for best response, 10% of all patients achieved a complete response (n=5)—12% of previously untreated patients (n=4) and 7% of relapsed/refractory patients (n=1). And 67% of patients had a partial response (n=32)—70% of previously untreated patients (n=23) and 60% of relapsed/refractory patients (n=9).
After 8 weeks on therapy, ibrutinib was associated with a more than 50% mean reduction in tumor burden in the bone marrow (44%), lymph nodes (70%), and spleen (79%). After 24 weeks of therapy, the rates of tumor burden reduction (> 50%) increased to 83%, 93%, and 95%, respectively.
Survival and safety
The estimated progression-free survival at 24 months for all patients on an intention-to-treat basis was 82%. Forty-two of the 51 patients (82%) continued on ibrutinib treatment without disease progression.
The estimated overall survival at 24 months was 80% for all patients—84% for previously untreated patients and 74% for patients with relapsed or refractory disease.
At the final follow-up, 8 (16%) patients had died—5 (10%) from progressive disease, 2 (4%) from infection, and 1 (2%) patient with a sudden, unexplained death that may have been treatment-related.
The most common adverse events (occurring in more than 30% of all patients) potentially related to ibrutinib were arthralgia (59%), diarrhea (51%), rash (47%), nail ridging (43%), bruising (33%), and muscle spasms (31%).
The most frequent grade 3 or 4 hematologic adverse events were neutropenia (24%), anemia (14%), and thrombocytopenia (10%). The most common nonhematologic grade 3 adverse event was pneumonia, which occurred in 3 patients (6%).
Nine patients (18%) discontinued treatment. The reasons for discontinuation included disease progression in 5 patients (10%) and death for 3 patients (6%).
This research was sponsored by the Intramural Research Program of the National Heart, Lung, and Blood Institute and the National Cancer Institute; Danish Cancer Society; Novo Nordisk Foundation; National Institutes of Health Medical Research Scholars Program; and Pharmacyclics Inc.
Ibrutinib is jointly developed and commercialized by Pharmacyclics and Janssen Biotech, Inc.
Whole plant treats malaria better
artemisinin is derived
Credit: Jorge Ferreira
Preclinical research suggests that using the whole plant Artemesia annua, from which the drug artemisinin is extracted, may treat malaria more effectively than artemisinin itself.
Whole-plant treatment withstood the evolution of resistance and remained effective for up to 3 times longer than pure artemisinin.
Whole-plant therapy was also more effective in killing rodent parasites that have previously evolved resistance to pure artemisinin.
Stephen Rich, PhD, of the University of Massachusetts Amherst, and his colleagues reported these findings in PNAS.
The team previously showed that the whole-plant approach is more effective at killing rodent malaria than purified artemisinin.
In the present study, the investigators conducted a series of experiments to determine the rates at which parasites become resistant to whole-plant treatment compared to the rate with pure artemisinin, and if the whole-plant treatment can overcome resistance to pharmaceutical artemisinin.
The team chose 2 rodent malaria species for particular characteristics. They chose Plasmodium yoelii because an artemisinin-resistant strain exists and could be used to test whether the whole plant can overcome that resistance.
And they chose Plasmodium chabaudi because, among several species of rodent malaria, it most closely biologically resembles the deadliest of the 5 human malaria parasites, Plasmodium falciparum.
“Conducting these experiments in different rodent malaria species also provides a robust test of the therapy,” Dr Rich noted.
To determine the respective evolutionary rates of resistance to whole-plant therapy and artemisinin, Dr Rich and his colleagues conducted artificial evolution experiments. The goal was to compare the rates at which resistance to these two treatments arises in serial passage among wild-type parasite lines.
In this technique, parasite proliferation rates determine resistance. Resistant parasites are expected to reach a certain target level at the same time, whether treatment is present or absent. Sensitive parasite strains will grow more slowly in the presence of treatment and reach the target later than untreated strains.
The investigators found that artemisinin-treated parasites achieved stable resistance to low-dose (100 mg/kg) therapy on passage 16. Those parasites were then treated with a doubled artemisinin dose, and they became resistant to this after an additional 24 passages.
By comparison, parasites did not become resistant to even the low dose of whole-plant therapy (100 mg/kg) after 49 passages.
From this, the investigators concluded that the whole-plant therapy lasts at least 3 times longer than its artemisinin counterpart, and at least twice as long as the doubled dose of pure artemisinin.
“This is especially important given the recent reports of resistance to artemisinin in malaria-endemic regions of the world,” Dr Rich said.
He and his colleagues also tested whether dried, whole-plant therapy can overcome existing resistance to pharmaceutical artemisinin.
They fed groups of mice infected with artemisinin-resistant malaria either the whole-plant therapy or artemisinin mixed with water. Single treatments were given in low (40 mg) and high (200 mg) doses. Control groups received a mouse chow placebo.
The investigators then measured the parasite levels in the rodents’ bloodstream at 9 points after treatment began.
Mice given either the low or high dose of whole-plant therapy showed a significantly greater reduction in parasitemia than those in their respective artemisinin groups. As expected for these resistant parasites, parasitemia in mice in the low-dose artemisinin group did not differ from controls.
The investigators said consuming the whole plant may be more effective than the single purified drug because the whole plant “may constitute a naturally occurring combination therapy that augments artemisinin delivery and synergizes the drug’s activity.”
Dr Rich did note that the exact mechanisms of whole-plant therapy’s effectiveness still need to be identified. But he also said the antimalarial activity of whole-plant therapy against artemisinin-resistant parasites provides “compelling reasons to further explore the role of non-pharmaceutical forms of artemisinin to treat human malaria.”
artemisinin is derived
Credit: Jorge Ferreira
Preclinical research suggests that using the whole plant Artemesia annua, from which the drug artemisinin is extracted, may treat malaria more effectively than artemisinin itself.
Whole-plant treatment withstood the evolution of resistance and remained effective for up to 3 times longer than pure artemisinin.
Whole-plant therapy was also more effective in killing rodent parasites that have previously evolved resistance to pure artemisinin.
Stephen Rich, PhD, of the University of Massachusetts Amherst, and his colleagues reported these findings in PNAS.
The team previously showed that the whole-plant approach is more effective at killing rodent malaria than purified artemisinin.
In the present study, the investigators conducted a series of experiments to determine the rates at which parasites become resistant to whole-plant treatment compared to the rate with pure artemisinin, and if the whole-plant treatment can overcome resistance to pharmaceutical artemisinin.
The team chose 2 rodent malaria species for particular characteristics. They chose Plasmodium yoelii because an artemisinin-resistant strain exists and could be used to test whether the whole plant can overcome that resistance.
And they chose Plasmodium chabaudi because, among several species of rodent malaria, it most closely biologically resembles the deadliest of the 5 human malaria parasites, Plasmodium falciparum.
“Conducting these experiments in different rodent malaria species also provides a robust test of the therapy,” Dr Rich noted.
To determine the respective evolutionary rates of resistance to whole-plant therapy and artemisinin, Dr Rich and his colleagues conducted artificial evolution experiments. The goal was to compare the rates at which resistance to these two treatments arises in serial passage among wild-type parasite lines.
In this technique, parasite proliferation rates determine resistance. Resistant parasites are expected to reach a certain target level at the same time, whether treatment is present or absent. Sensitive parasite strains will grow more slowly in the presence of treatment and reach the target later than untreated strains.
The investigators found that artemisinin-treated parasites achieved stable resistance to low-dose (100 mg/kg) therapy on passage 16. Those parasites were then treated with a doubled artemisinin dose, and they became resistant to this after an additional 24 passages.
By comparison, parasites did not become resistant to even the low dose of whole-plant therapy (100 mg/kg) after 49 passages.
From this, the investigators concluded that the whole-plant therapy lasts at least 3 times longer than its artemisinin counterpart, and at least twice as long as the doubled dose of pure artemisinin.
“This is especially important given the recent reports of resistance to artemisinin in malaria-endemic regions of the world,” Dr Rich said.
He and his colleagues also tested whether dried, whole-plant therapy can overcome existing resistance to pharmaceutical artemisinin.
They fed groups of mice infected with artemisinin-resistant malaria either the whole-plant therapy or artemisinin mixed with water. Single treatments were given in low (40 mg) and high (200 mg) doses. Control groups received a mouse chow placebo.
The investigators then measured the parasite levels in the rodents’ bloodstream at 9 points after treatment began.
Mice given either the low or high dose of whole-plant therapy showed a significantly greater reduction in parasitemia than those in their respective artemisinin groups. As expected for these resistant parasites, parasitemia in mice in the low-dose artemisinin group did not differ from controls.
The investigators said consuming the whole plant may be more effective than the single purified drug because the whole plant “may constitute a naturally occurring combination therapy that augments artemisinin delivery and synergizes the drug’s activity.”
Dr Rich did note that the exact mechanisms of whole-plant therapy’s effectiveness still need to be identified. But he also said the antimalarial activity of whole-plant therapy against artemisinin-resistant parasites provides “compelling reasons to further explore the role of non-pharmaceutical forms of artemisinin to treat human malaria.”
artemisinin is derived
Credit: Jorge Ferreira
Preclinical research suggests that using the whole plant Artemesia annua, from which the drug artemisinin is extracted, may treat malaria more effectively than artemisinin itself.
Whole-plant treatment withstood the evolution of resistance and remained effective for up to 3 times longer than pure artemisinin.
Whole-plant therapy was also more effective in killing rodent parasites that have previously evolved resistance to pure artemisinin.
Stephen Rich, PhD, of the University of Massachusetts Amherst, and his colleagues reported these findings in PNAS.
The team previously showed that the whole-plant approach is more effective at killing rodent malaria than purified artemisinin.
In the present study, the investigators conducted a series of experiments to determine the rates at which parasites become resistant to whole-plant treatment compared to the rate with pure artemisinin, and if the whole-plant treatment can overcome resistance to pharmaceutical artemisinin.
The team chose 2 rodent malaria species for particular characteristics. They chose Plasmodium yoelii because an artemisinin-resistant strain exists and could be used to test whether the whole plant can overcome that resistance.
And they chose Plasmodium chabaudi because, among several species of rodent malaria, it most closely biologically resembles the deadliest of the 5 human malaria parasites, Plasmodium falciparum.
“Conducting these experiments in different rodent malaria species also provides a robust test of the therapy,” Dr Rich noted.
To determine the respective evolutionary rates of resistance to whole-plant therapy and artemisinin, Dr Rich and his colleagues conducted artificial evolution experiments. The goal was to compare the rates at which resistance to these two treatments arises in serial passage among wild-type parasite lines.
In this technique, parasite proliferation rates determine resistance. Resistant parasites are expected to reach a certain target level at the same time, whether treatment is present or absent. Sensitive parasite strains will grow more slowly in the presence of treatment and reach the target later than untreated strains.
The investigators found that artemisinin-treated parasites achieved stable resistance to low-dose (100 mg/kg) therapy on passage 16. Those parasites were then treated with a doubled artemisinin dose, and they became resistant to this after an additional 24 passages.
By comparison, parasites did not become resistant to even the low dose of whole-plant therapy (100 mg/kg) after 49 passages.
From this, the investigators concluded that the whole-plant therapy lasts at least 3 times longer than its artemisinin counterpart, and at least twice as long as the doubled dose of pure artemisinin.
“This is especially important given the recent reports of resistance to artemisinin in malaria-endemic regions of the world,” Dr Rich said.
He and his colleagues also tested whether dried, whole-plant therapy can overcome existing resistance to pharmaceutical artemisinin.
They fed groups of mice infected with artemisinin-resistant malaria either the whole-plant therapy or artemisinin mixed with water. Single treatments were given in low (40 mg) and high (200 mg) doses. Control groups received a mouse chow placebo.
The investigators then measured the parasite levels in the rodents’ bloodstream at 9 points after treatment began.
Mice given either the low or high dose of whole-plant therapy showed a significantly greater reduction in parasitemia than those in their respective artemisinin groups. As expected for these resistant parasites, parasitemia in mice in the low-dose artemisinin group did not differ from controls.
The investigators said consuming the whole plant may be more effective than the single purified drug because the whole plant “may constitute a naturally occurring combination therapy that augments artemisinin delivery and synergizes the drug’s activity.”
Dr Rich did note that the exact mechanisms of whole-plant therapy’s effectiveness still need to be identified. But he also said the antimalarial activity of whole-plant therapy against artemisinin-resistant parasites provides “compelling reasons to further explore the role of non-pharmaceutical forms of artemisinin to treat human malaria.”
Medication Warnings for Adults
Many computerized provider order entry (CPOE) systems suffer from having too much of a good thing. Few would question the beneficial effect of CPOE on medication order clarity, completeness, and transmission.[1, 2] When mechanisms for basic decision support have been added, however, such as allergy, interaction, and duplicate warnings, reductions in medication errors and adverse events have not been consistently achieved.[3, 4, 5, 6, 7] This is likely due in part to the fact that ordering providers override medication warnings at staggeringly high rates.[8, 9] Clinicians acknowledge that they are ignoring potentially valuable warnings,[10, 11] but suffer from alert fatigue due to the sheer number of messages, many of them judged by clinicians to be of low‐value.[11, 12]
Redesign of medication alert systems to increase their signal‐to‐noise ratio is badly needed,[13, 14, 15, 16] and will need to consider the clinical significance of alerts, their presentation, and context‐specific factors that potentially contribute to warning effectiveness.[17, 18, 19] Relatively few studies, however, have objectively looked at context factors such as the characteristics of providers, patients, medications, and warnings that are associated with provider responses to warnings,[9, 20, 21, 22, 23, 24, 25] and only 2 have studied how warning acceptance is associated with medication risk.[18, 26] We wished to explore these factors further. Warning acceptance has been shown to be higher, at least in the outpatient setting, when orders are entered by low‐volume prescribers for infrequently encountered warnings,[24] and there is some evidence that patients receive higher‐quality care during the day.[27] Significant attention has been placed in recent years on inappropriate prescribing in older patients,[28] and on creating a culture of safety in healthcare.[29] We therefore hypothesized that our providers would be more cautious, and medication warning acceptance rates would be higher, when orders were entered for patients who were older or with more complex medical problems, when they were entered during the day by caregivers who entered few orders, when the medications ordered were potentially associated with greater risk, and when the warnings themselves were infrequently encountered.
METHODS
Setting and Caregivers
Johns Hopkins Bayview Medical Center (JHBMC) is a 400‐bed academic medical center serving southeastern Baltimore, Maryland. Prescribing caregivers include residents and fellows who rotate to both JHBMC and Johns Hopkins Hospital, internal medicine hospitalists, other attending physicians (including teaching attendings for all departments, and hospitalists and clinical associates for departments other than internal medicine), and nurse practitioners and physician assistants from most JHBMC departments. Nearly 100% of patients on the surgery, obstetrics/gynecology, neurology, psychiatry, and chemical dependence services are hospitalized on units dedicated to their respective specialty, and the same is true for approximately 95% of medicine patients.
Order Entry
JHBMC began using a client‐server order entry system by MEDITECH (Westwood, MA) in July 2003. Provider order entry was phased in beginning in October 2003 and completed by the end of 2004. MEDITECH version 5.64 was being used during the study period. Medications may generate duplicate, interaction, allergy, adverse reaction, and dose warnings during a patient ordering session each time they are ordered. Duplicate warnings are generated when the same medication (no matter what route) is ordered that is either on their active medication list, was on the list in the preceding 24 hours, or that is being ordered simultaneously. A drug‐interaction database licensed from First DataBank (South San Francisco, CA) is utilized, and updated monthly, which classifies potential drug‐drug interactions as contraindicated, severe, intermediate, and mild. Those classified as contraindicated by First DataBank are included in the severe category in MEDITECH 5.64. During the study period, JHBMC's version of MEDITECH was configured so that providers were warned of potential severe and intermediate drug‐drug interactions, but not mild. No other customizations had been made. Patients' histories of allergies and other adverse responses to medications can be entered by any credentialed staff member. They are maintained together in an allergies section of the electronic medical record, but are identified as either allergy or adverse reactions at the time they are entered, and each generates its own warnings.
When more than 1 duplicate, interaction, allergy, or adverse reaction warning is generated for a particular medication, all appear listed on a single screen in identical fonts. No visual distinction is made between severe and intermediate drug‐drug interactions; for these, the category of medication ordered is followed by the category of the medication for which there is a potential interaction. A details button can be selected to learn specifically which medications are involved and the severity and nature of the potential interactions identified. In response to the warnings, providers can choose to either override them, erase the order, or replace the order by clicking 1 of 3 buttons at the bottom of the screen. Warnings are not repeated unless the medication is reordered for that patient. Dose warnings appear on a subsequent screen and are not addressed in this article.
Nurses are discouraged from entering verbal orders but do have the capacity to do so, at which time they encounter and must respond to the standard medication warnings, if any. Medical students are able to enter orders, at which time they also encounter and must respond to the standard medication warnings; their orders must then be cosigned by a licensed provider before they can be processed. Warnings encountered by nurses and medical students are not repeated at the time of cosignature by a licensed provider.
Data Collection
We collected data regarding all medication orders placed in our CPOE system from October 1, 2009 to April 20, 2010 for all adult patients. Intensive care unit (ICU) patients were excluded, in anticipation of a separate analysis. Hospitalizations under observation were also excluded. We then ran a report showing all medications that generated any number of warnings of any type (duplicate, interaction, allergy, or adverse reaction) for the same population. Warnings generated during readmissions that occurred at any point during the study period (ranging from 1 to 21 times) were excluded, because these patients likely had many, if not all, of the same medications ordered during their readmissions as during their initial hospitalization, which would unduly influence the analysis if retained.
There was wide variation in the number of warnings generated per medication and in the number of each warning type per medication that generated multiple warnings. Therefore, for ease of analysis and to ensure that we could accurately determine varying response to each individual warning type, we thereafter focused on the medications that generated single warnings during the study period. For each single warning we obtained patient name, account number, event date and time, hospital unit at the time of the event, ordered medication, ordering staff member, warning type, and staff member response to the warning (eg, override warning or erase order [accept the warning]). The response replace was used very infrequently, and therefore warnings that resulted in this response were excluded. Medications available in more than 1 form included the route of administration in their name, and from this they were categorized as parenteral or nonparenteral. All nonparenteral or parenteral forms of a given medication were grouped together as 1 medication (eg, morphine sustained release and morphine elixir were classified as a single‐medication, nonparenteral morphine). Medications were further categorized according to whether or not they were on the Institute for Safe Medication Practice (ISMP) List of High‐Alert Medications.[30]
The study was approved by the Johns Hopkins Institutional Review Board.
Analysis
We collected descriptive data about patients and providers. Age and length of stay (LOS) at the time of the event were determined based on the patients' admit date and date of birth, and grouped into quartiles. Hospital units were grouped according to which service or services they primarily served. Medications were grouped into quartiles according to the total number of warnings they generated during the study period. Warnings were dichotomously categorized according to whether they were overridden or accepted. Unpaired t tests were used to compare continuous variables for the 2 groups, and [2] tests were used to compare categorical variables. A multivariate logistic regression was then performed, using variables with a P value of <0.10 in the univariate analysis, to control for confounders and identify independent predictors of medication warning acceptance. All analyses were performed using Intercooled Stata 12 (StataCorp, College Station, TX).
RESULTS
A total of 259,656 medication orders were placed for adult non‐ICU patients during the 7‐month study period. Of those orders, 45,835 generated some number of medication warnings.[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20] The median number of warnings per patient was 4 (interquartile range [IQR]=28; mean=5.9, standard deviation [SD]=6.2), with a range from 1 to 84. The median number of warnings generated per provider during the study period was 36 (IQR=6106, mean=87.4, SD=133.7), with a range of 1 to 1096.
There were 40,391 orders placed for 454 medications for adult non‐ICU patients, which generated a single‐medication warning (excluding those with the response replace, which was used 20 times) during the 7‐month study period. Data regarding the patients and providers associated with the orders generating single warnings are shown in Table 1. Most patients were on medicine units, and most orders were entered by residents. Patients' LOS at the time the orders were placed ranged from 0 to 118 days (median=1, IQR=04; mean=4.0, SD=7.2). The median number of single warnings per patient was 4 (IQR=28; mean=6.1, SD=6.5), with a range from 1 to 84. The median number of single warnings generated per provider during the study period was 15 (IQR=373; mean=61.7, SD=109.6), with a range of 1 to 1057.
| No. (%) | |
|---|---|
| |
| Patients (N=6,646) | |
| Age | |
| 1545 years | 2,048 (31%) |
| 4657 years | 1,610 (24%) |
| 5872 years | 1,520 (23%) |
| 73104 years | 1,468 (22%) |
| Gender | |
| Male | 2,934 (44%) |
| Hospital unita | |
| Medicine | 2,992 (45%) |
| Surgery | 1,836 (28%) |
| Neuro/psych/chem dep | 1,337 (20%) |
| OB/GYN | 481 (7%) |
| Caregivers (N=655) | |
| Resident | 248 (38%)b |
| Nurse | 154 (24%) |
| Attending or other | 97 (15%) |
| NP/PA | 69 (11%) |
| IM hospitalist | 31 (5%) |
| Fellow | 27 (4%) |
| Medical student | 23 (4%) |
| Pharmacist | 6 (1%) |
Patient and caregiver characteristics for the medication orders that generated single warnings are shown in Table 2. The majority of medications were nonparenteral and not on the ISMP list (Table 3). Most warnings generated were either duplicate (47%) or interaction warnings (47%). Warnings of a particular type were repeated 14.5% of the time for a particular medication and patient (from 2 to 24 times, median=2, IQR=22, mean=2.7, SD=1.4), and 9.8% of the time for a particular caregiver, medication, and patient (from 2 to 18 times, median=2, IQR=22, mean=2.4, SD=1.1).
| Variable | No. of Warnings (%)a | No. of Warnings Accepted (%)a | P |
|---|---|---|---|
| |||
| Patient age | |||
| 1545 years | 10,881 (27) | 602 (5.5%) | <0.001 |
| 4657 years | 9,733 (24) | 382 (3.9%) | |
| 5872 years | 10,000 (25) | 308 (3.1%) | |
| 73104 years | 9,777 (24) | 262 (2.7%) | |
| Patient gender | |||
| Female | 23,395 (58) | 866 (3.7%) | 0.074 |
| Male | 16,996 (42) | 688 (4.1%) | |
| Patient length of stay | |||
| <1 day | 10,721 (27) | 660 (6.2%) | <0.001 |
| 1 day | 10,854 (27) | 385 (3.5%) | |
| 24 days | 10,424 (26) | 277 (2.7%) | |
| 5118 days | 8,392 (21) | 232 (2.8%) | |
| Patient hospital unit | |||
| Medicine | 20,057 (50) | 519 (2.6%) | <0.001 |
| Surgery | 10,274 (25) | 477 (4.6%) | |
| Neuro/psych/chem dep | 8,279 (21) | 417 (5.0%) | |
| OB/GYN | 1,781 (4) | 141 (7.9%) | |
| Ordering caregiver | |||
| Resident | 22,523 (56) | 700 (3.1%) | <0.001 |
| NP/PA | 7,534 (19) | 369 (4.9%) | |
| IM hospitalist | 5,048 (13) | 155 (3.1%) | |
| Attending | 3225 (8) | 219 (6.8%) | |
| Fellow | 910 (2) | 34 (3.7%) | |
| Nurse | 865 (2) | 58 (6.7%) | |
| Medical student | 265 (<1) | 17 (6.4%) | |
| Pharmacist | 21 (<1) | 2 (9.5%) | |
| Day ordered | |||
| Weekday | 31,499 (78%) | 1276 (4.1%) | <0.001 |
| Weekend | 8,892 (22%) | 278 (3.1%) | |
| Time ordered | |||
| 00000559 | 4,231 (11%) | 117 (2.8%) | <0.001 |
| 06001159 | 11,696 (29%) | 348 (3.0%) | |
| 12001759 | 15,879 (39%) | 722 (4.6%) | |
| 18002359 | 8,585 (21%) | 367 (4.3%) | |
| Administration route (no. of meds) | |||
| Nonparenteral (339) | 27,086 (67%) | 956 (3.5%) | <0.001 |
| Parenteral (115) | 13,305 (33%) | 598 (4.5%) | |
| ISMP List of High‐Alert Medications status (no. of meds)[30] | |||
| Not on ISMP list (394) | 27,503 (68%) | 1251 (4.5%) | <0.001 |
| On ISMP list (60) | 12,888 (32%) | 303 (2.4%) | |
| No. of warnings per med (no. of meds) | |||
| 11062133 (7) | 9,869 (24%) | 191 (1.9%) | <0.001 |
| 4681034 (13) | 10,014 (25%) | 331 (3.3%) | |
| 170444 (40) | 10,182 (25%) | 314 (3.1%) | |
| 1169 (394) | 10,326 (26%) | 718 (7.0%) | |
| Warning type (no. of meds)b | |||
| Duplicate (369) | 19,083 (47%) | 1041 (5.5%) | <0.001 |
| Interaction (315) | 18,894 (47%) | 254 (1.3%) | |
| Allergy (138) | 2,371 (6%) | 243 (10.0%) | |
| Adverse reaction (14) | 43 (0.1%) | 16 (37%) | |
| Variable | Adjusted OR | 95% CI |
|---|---|---|
| ||
| Patient age | ||
| 1545 years | 1.00 | Reference |
| 4657 years | 0.89 | 0.771.02 |
| 5872 years | 0.85 | 0.730.99 |
| 73104 years | 0.91 | 0.771.08 |
| Patient gender | ||
| Female | 1.00 | Reference |
| Male | 1.26 | 1.131.41 |
| Patient length of stay | ||
| <1 day | 1.00 | Reference |
| 1 day | 0.65 | 0.550.76 |
| 24 days | 0.49 | 0.420.58 |
| 5118 days | 0.49 | 0.410.58 |
| Patient hospital unit | ||
| Medicine | 1.00 | Reference |
| Surgery | 1.45 | 1.251.68 |
| Neuro/psych/chem dep | 1.35 | 1.151.58 |
| OB/GYN | 2.43 | 1.923.08 |
| Ordering caregiver | ||
| Resident | 1.00 | Reference |
| NP/PA | 1.63 | 1.421.88 |
| IM hospitalist | 1.24 | 1.021.50 |
| Attending | 1.83 | 1.542.18 |
| Fellow | 1.41 | 0.982.03 |
| Nurse | 1.92 | 1.442.57 |
| Medical student | 1.17 | 0.701.95 |
| Pharmacist | 3.08 | 0.6714.03 |
| Medication factors | ||
| Nonparenteral | 1.00 | Reference |
| Parenteral | 1.79 | 1.592.03 |
| HighAlert Medication status (no. of meds)[30] | ||
| Not on ISMP list | 1.00 | Reference |
| On ISMP list | 0.37 | 0.320.43 |
| No. of warnings per medication | ||
| 11062133 | 1.00 | Reference |
| 4681034 | 2.30 | 1.902.79 |
| 170444 | 2.25 | 1.852.73 |
| 1169 | 4.10 | 3.424.92 |
| Warning type | ||
| Duplicate | 1.00 | Reference |
| Interaction | 0.24 | 0.210.28 |
| Allergy | 2.28 | 1.942.68 |
| Adverse reaction | 9.24 | 4.5218.90 |
One thousand five hundred fifty‐four warnings were erased (ie, accepted by clinicians [4%]). In univariate analysis, only patient gender was not associated with warning acceptance. Patient age, LOS, hospital unit at the time of order entry, ordering caregiver type, day and time the medication was ordered, administration route, presence on the ISMP list, warning frequency, and warning type were all significantly associated with warning acceptance (Table 2).
Older patient age, longer LOS, presence of the medication on the ISMP list, and interaction warning type were all negatively associated with warning acceptance in multivariable analysis. Warning acceptance was positively associated with male patient gender, being on a service other than medicine, being a caregiver other than a resident, parenteral medications, lower warning frequency, and allergy or adverse reaction warning types (Table 3).
The 20 medications that generated the most single warnings are shown in Table 4. Medications on the ISMP list accounted for 8 of these top 20 medications. For most of them, duplicate and interaction warnings accounted for most of the warnings generated, except for parenteral hydromorphone, oral oxycodone, parenteral morphine, and oral hydromorphone, which each had more allergy than interaction warnings.
| Medication | ISMP Listb | No. of Warnings | Duplicate, No. (%)c | Interaction, No. (%)c | Allergy, No. (%)c | Adverse Reaction, No. (%)c |
|---|---|---|---|---|---|---|
| ||||||
| Hydromorphone injectable | Yes | 2,133 | 1,584 (74.3) | 127 (6.0) | 422 (19.8) | |
| Metoprolol | 1,432 | 550 (38.4) | 870 (60.8) | 12 (0.8) | ||
| Aspirin | 1,375 | 212 (15.4) | 1,096 (79.7) | 67 (4.9) | ||
| Oxycodone | Yes | 1,360 | 987 (72.6) | 364 (26.8) | 9 (0.7) | |
| Potassium chloride | 1,296 | 379 (29.2) | 917 (70.8) | |||
| Ondansetron injectable | 1,167 | 1,013 (86.8) | 153 (13.1) | 1 (0.1) | ||
| Aspart insulin injectable | Yes | 1,106 | 643 (58.1) | 463 (41.9) | ||
| Warfarin | Yes | 1,034 | 298 (28.8) | 736 (71.2) | ||
| Heparin injectable | Yes | 1,030 | 205 (19.9) | 816 (79.2) | 9 (0.3) | |
| Furosemide injectable | 980 | 438 (45.0) | 542 (55.3) | |||
| Lisinopril | 926 | 225 (24.3) | 698 (75.4) | 3 (0.3) | ||
| Acetaminophen | 860 | 686 (79.8) | 118 (13.7) | 54 (6.3) | 2 (0.2) | |
| Morphine injectable | Yes | 804 | 467 (58.1) | 100 (12.4) | 233 (29.0) | 4 (0.5) |
| Diazepam | 786 | 731 (93.0) | 41 (5.2) | 14 (1.8) | ||
| Glargine insulin injectable | Yes | 746 | 268 (35.9) | 478 (64.1) | ||
| Ibuprofen | 713 | 125 (17.5) | 529 (74.2) | 54 (7.6) | 5 (0.7) | |
| Hydromorphone | Yes | 594 | 372 (62.6) | 31 (5.2) | 187 (31.5) | 4 (0.7) |
| Furosemide | 586 | 273 (46.6) | 312 (53.2) | 1 (0.2) | ||
| Ketorolac injectable | 487 | 39 (8.0) | 423 (86.9) | 23 (4.7) | 2 (0.4) | |
| Prednisone | 468 | 166 (35.5) | 297 (63.5) | 5 (1.1) | ||
DISCUSSION
Medication warnings in our study were frequently overridden, particularly when encountered by residents, for patients with a long LOS and on the internal medicine service, and for medications generating the most warnings and on the ISMP list. Disturbingly, this means that potentially important warnings for medications with the highest potential for causing harm, for possibly the sickest and most complex patients, were those that were most often ignored by young physicians in training who should have had the most to gain from them. Of course, this is not entirely surprising. Despite our hope that a culture of safety would influence young physicians' actions when caring for these patients and prescribing these medications, these patients and medications are those for whom the most warnings are generated, and these physicians are the ones entering the most orders. Only 13% of the medications studied were on the ISMP list, but they generated 32% of the warnings. We controlled for number of warnings and ISMP list status, but not for warning validity. Most likely, high‐risk medications have been set up with more warnings, many of them of lower quality, in an errant but well‐intentioned effort to make them safer. If developers of CPOE systems want to gain serious traction in using decision support to promote prescribing safe medications, they must take substantial action to increase attention to important warnings and decrease the number of clinically insignificant, low‐value warnings encountered by active caregivers on a daily basis.
Only 2 prior studies, both by Seidling et al., have specifically looked at provider response to warnings for high risk medications. Interaction warnings were rarely accepted in 1,[18] as in our study; however, in contrast to our findings, warning acceptance in both studies was higher for drugs with dose‐dependent toxicity.[18, 26] The effect of physician experience on warning acceptance has been addressed in 2 prior studies. In Weingart et al., residents were more likely than staff physicians to erase medication orders when presented with allergy and interaction warnings in a primary care setting.[20] Long et al. found that physicians younger than 40 years were less likely than older physicians to accept duplicate warnings, but those who had been at the study hospital for a longer period of time were more likely to accept them.[23] The influence of patient LOS and service on warning acceptance has not previously been described. Further study is needed looking at each of these factors.
Individual hospitals tend to avoid making modifications to order entry warning systems, because monitoring and maintaining these changes is labor intensive. Some institutions may make the decision to turn off certain categories of alerts, such as intermediate interaction warnings, to minimize the noise their providers encounter. There are even tools for disabling individual alerts or groups of alerts, such as that available for purchase from our interaction database vendor.[31] However, institutions may fear litigation should an adverse event be attributed to a disabled warning.[15, 16] Clearly, a comprehensive, health system‐wide approach is warranted.[13, 15] To date, published efforts describing ways to improve the effectiveness of medication warning systems have focused on either heightening the clinical significance of alerts[14, 21, 22, 32, 33, 34, 35, 36] or altering their presentation and how providers experience them.[21, 36, 37, 38, 39, 40, 41, 42, 43] The single medication warnings our providers receive are all presented in an identical font, and presumably response to each would be different if they were better distinguished from each other. We also found that a small but significant number of warnings were repeated for a given patient and even a given provider. If the providers knew they would only be presented with warnings the first time they occurred for a given patient and medication, they might be more attuned to the remaining warnings. Previous studies describe context‐specific decision support for medication ordering[44, 45, 46]; however, only 1 has described the use of patient context factors to modify when or how warnings are presented to providers.[47] None have described tailoring allergy, duplicate, and interaction warnings according to medication or provider types. If further study confirms our findings, modulating basic warning systems according to severity of illness, provider experience, and medication risk could powerfully increase their effectiveness. Of course, this would be extremely challenging to achieve, and is likely outside the capabilities of most, if not all, CPOE systems, at least for now.
Our study has some limitations. First, it was limited to medications that generated a single warning. We did this for ease of analysis and so that we could ensure understanding of provider response to each warning type without bias from simultaneously occurring warnings; however, caregiver response to multiple warnings appearing simultaneously for a particular medication order might be quite different. Second, we did not include any assessment of the number of medications ordered by each provider type or for each patient, either of which could significantly affect provider response to warnings. Third, as previously noted, we did not include any assessment of the validity of the warnings, beyond the 4 main categories described, which could also significantly affect provider response. However, it should be noted that although the validity of interaction warnings varies significantly from 1 medication to another, the validity of duplicate, allergy, and adverse reaction warnings in the described system are essentially the same for all medications. Fourth, it is possible that providers did modify or even erase their orders even after selecting override in response to the warning; it is also possible that providers reentered the same order after choosing erase. Unfortunately auditing for actions such as these would be extremely laborious. Finally, the study was conducted at a single medical center using a single order‐entry system. The system in use at our medical center is in use at one‐third of the 6000 hospitals in the United States, though certainly not all are using our version. Even if a hospital was using the same CPOE version and interaction database as our institution, variations in patient population and local decisions modifying how the database interacts with the warning presentation system might affect reproducibility at that institution.
Commonly encountered medication warnings are overridden at extremely high rates, and in our study this was particularly so for medications on the ISMP list, when ordered by physicians in training. Warnings of little clinical significance must be identified and eliminated, the most important warnings need to be visually distinct to increase user attention, and further research should be done into the patient, provider, setting, and medication factors that affect user responses to warnings, so that they may be customized accordingly and their significance increased. Doing so will enable us to reap the maximum possible potential from our CPOE systems, and increase the CPOE's power to protect our most vulnerable patients from our most dangerous medications, particularly when cared for by our most inexperienced physicians.
Acknowledgements
The authors thank, in particular, Scott Carey, Research Informatics Manager, for assistance with data collection. Additional thanks go to Olga Sherman and Kathleen Ancinich for assistance with data collection and management.
Disclosures: This research was supported in part by the Johns Hopkins Institute for Clinical and Translational Research. All listed authors contributed substantially to the study conception and design, analysis and interpretation of data, drafting the article or revising it critically for important intellectual content, and final approval of the version to be published. No one who fulfills these criteria has been excluded from authorship. This research received no specific grant from any funding agency in the public, commercial, or not‐for‐profit sectors. The authors have no competing interests to declare.
- , , , et al., Effect of computerized physician order entry and a team intervention on prevention of serious medication errors. JAMA. 1998;280:1311–1316.
- , , , , , . Effects of computerized provider order entry on prescribing practices. Arch Intern Med. 2000;160:2741–2747.
- , , , et al. Effects of computerized clinician decision support systems on practitioner performance and patient outcomes: a systematic review. JAMA. 2005;293:1223–1238.
- , , , et al. The effect of computerized physician order entry with clinical decision support on the rates of adverse drug events: a systematic review. J Gen Intern Med. 2008;23:451–458.
- , , . The impact of computerized physician medication order entry in hospitalized patients—a systematic review. Int J Med Inform. 2008;77:365–376.
- , , , et al. What evidence supports the use of computerized alerts and prompts to improve clinicians' prescribing behavior? J Am Med Inform Assoc. 2009;16:531–538.
- , , , , . Does computerized provider order entry reduce prescribing errors for hospital inpatients? A systematic review. J Am Med Inform Assoc. 2009;16:613–623.
- , , , . Overriding of drug safety alerts in computerized physician order entry. J Am Med Inform Assoc. 2006;13:138–147.
- , , , , , . Evaluating clinical decision support systems: monitoring CPOE order check override rates in the Department of Veterans Affairs' Computerized Patient Record System. J Am Med Inform Assoc. 2008;15:620–626.
- , , . GPs' views on computerized drug interaction alerts: questionnaire survey. J Clin Pharm Ther. 2002;27:377–382.
- , , , et al. Clinicians' assessments of electronic medication safety alerts in ambulatory care. Arch Intern Med. 2009;169:1627–1632.
- , , , , . A mixed method study of the merits of e‐prescribing drug alerts in primary care. J Gen Intern Med. 2008;23:442–446.
- . CPOE and clinical decision support in hospitals: getting the benefits: comment on “Unintended effects of a computerized physician order entry nearly hard‐stop alert to prevent a drug interaction.” Arch Intern Med. 2010;170:1583–1584.
- , , . Critical drug‐drug interactions for use in electronic health records systems with computerized physician order entry: review of leading approaches. J Patient Saf. 2011;7:61–65.
- , , , , . Clinical decision support systems could be modified to reduce 'alert fatigue' while still minimizing the risk of litigation. Health Aff (Millwood). 2011;30:2310–2317.
- , , , . Critical issues associated with drug‐drug interactions: highlights of a multistakeholder conference. Am J Health Syst Pharm. 2011;68:941–946.
- , , , , , . Development of a context model to prioritize drug safety alerts in CPOE systems. BMC Med Inform Decis Mak. 2011;11:35.
- , , , et al. Factors influencing alert acceptance: a novel approach for predicting the success of clinical decision support. J Am Med Inform Assoc. 2011;18:479–484.
- , , , . How to improve the delivery of medication alerts within computerized physician order entry systems: an international Delphi study. J Am Med Inform Assoc. 2011;18:760–766.
- , , , , , . Physicians' decisions to override computerized drug alerts in primary care. Arch Intern Med. 2003;163:2625–2631.
- , , , et al. Improving acceptance of computerized prescribing alerts in ambulatory care. J Am Med Inform Assoc. 2006;13:5–11.
- , , , . Optimizing the acceptance of medication‐based alerts by physicians during CPOE implementation in a community hospital environment. AMIA Annu Symp Proc. 2007:701–705.
- , , , . The use of a CPOE log for the analysis of physicians' behavior when responding to drug‐duplication reminders. Int J Med Inform. 2008;77:499–506.
- , , , et al. Overrides of medication alerts in ambulatory care. Arch Intern Med. 2009;169:305–311.
- , , , , , . Drug safety alert generation and overriding in a large Dutch university medical centre. Pharmacoepidemiol Drug Saf. 2009;18:941–947.
- , , , et al. Patient‐specific electronic decision support reduces prescription of excessive doses. Qual Saf Health Care. 2010;19:e15.
- , , , et al. Survival from in‐hospital cardiac arrest during nights and weekends. JAMA. 2008;299:785–792.
- , . Managing medications in clinically complex elders: “There's got to be a happy medium.” JAMA. 2010;304:1592–1601.
- Agency for Healthcare Research and Quality. Safety culture. Available at: http://psnet.ahrq.gov/primer.aspx?primerID=5. Accessed October 29, 2013.
- Institute for Safe Medication Practice. List of High‐Alert Medications. Available at: http://www.ismp.org/Tools/highalertmedications.pdf. Accessed June 18, 2013.
- First Databank. FDB AlertSpace. Available at: http://www.fdbhealth.com/solutions/fdb‐alertspace. Accessed July 3, 2014.
- , , , et al. Improving allergy alerting in a computerized physician order entry system. Proc AMIA Symp. 2000:2–6.
- , , , , , . A clinical data warehouse‐based process for refining medication orders alerts. J Am Med Inform Assoc. 2012;19:782–785.
- , , , et al. Drug‐drug interactions that should be non‐interruptive in order to reduce alert fatigue in electronic health records. J Am Med Inform Assoc. 2013;20:489–493.
- , , , et al. High‐priority drug‐drug interactions for use in electronic health records. J Am Med Inform Assoc. 2012;19:735–743.
- , , , , . Design of decision support interventions for medication prescribing. Int J Med Inform. 2013;82:492–503.
- , , , et al. A randomized trial of the effectiveness of on‐demand versus computer‐triggered drug decision support in primary care. J Am Med Inform Assoc. 2008;15:430–438.
- , , , et al. Tiering drug‐drug interaction alerts by severity increases compliance rates. J Am Med Inform Assoc. 2009;16:40–46.
- , , , et al. A review of human factors principles for the design and implementation of medication safety alerts in clinical information systems. J Am Med Inform Assoc. 2010;17:493–501.
- , , , et al. Unintended effects of a computerized physician order entry nearly hard‐stop alert to prevent a drug interaction: a randomized controlled trial. Arch Intern Med. 2010;170:1578–1583.
- , , , , , . Randomized clinical trial of a customized electronic alert requiring an affirmative response compared to a control group receiving a commercial passive CPOE alert: NSAID—warfarin co‐prescribing as a test case. J Am Med Inform Assoc. 2010;17:411–415.
- , , , , . Making electronic prescribing alerts more effective: scenario‐based experimental study in junior doctors. J Am Med Inform Assoc. 2011;18:789–798.
- , , , et al. Development and preliminary evidence for the validity of an instrument assessing implementation of human‐factors principles in medication‐related decision‐support systems—I‐MeDeSA. J Am Med Inform Assoc. 2011;18(suppl 1):i62–i72.
- , , , et al. Medication‐related clinical decision support in computerized provider order entry systems: a review. J Am Med Inform Assoc. 2007;14:29–40.
- , , , et al. Physicians' perceptions on the usefulness of contextual information for prioritizing and presenting alerts in Computerized Physician Order Entry systems. BMC Med Inform Decis Mak. 2012;12:111.
- , , , et al. Computerized clinical decision support systems for drug prescribing and management: a decision‐maker‐researcher partnership systematic review. Implement Sci. 2011;6:89.
- , . A successful model and visual design for creating context‐aware drug‐drug interaction alerts. AMIA Annu Symp Proc. 2011;2011:339–348.
Many computerized provider order entry (CPOE) systems suffer from having too much of a good thing. Few would question the beneficial effect of CPOE on medication order clarity, completeness, and transmission.[1, 2] When mechanisms for basic decision support have been added, however, such as allergy, interaction, and duplicate warnings, reductions in medication errors and adverse events have not been consistently achieved.[3, 4, 5, 6, 7] This is likely due in part to the fact that ordering providers override medication warnings at staggeringly high rates.[8, 9] Clinicians acknowledge that they are ignoring potentially valuable warnings,[10, 11] but suffer from alert fatigue due to the sheer number of messages, many of them judged by clinicians to be of low‐value.[11, 12]
Redesign of medication alert systems to increase their signal‐to‐noise ratio is badly needed,[13, 14, 15, 16] and will need to consider the clinical significance of alerts, their presentation, and context‐specific factors that potentially contribute to warning effectiveness.[17, 18, 19] Relatively few studies, however, have objectively looked at context factors such as the characteristics of providers, patients, medications, and warnings that are associated with provider responses to warnings,[9, 20, 21, 22, 23, 24, 25] and only 2 have studied how warning acceptance is associated with medication risk.[18, 26] We wished to explore these factors further. Warning acceptance has been shown to be higher, at least in the outpatient setting, when orders are entered by low‐volume prescribers for infrequently encountered warnings,[24] and there is some evidence that patients receive higher‐quality care during the day.[27] Significant attention has been placed in recent years on inappropriate prescribing in older patients,[28] and on creating a culture of safety in healthcare.[29] We therefore hypothesized that our providers would be more cautious, and medication warning acceptance rates would be higher, when orders were entered for patients who were older or with more complex medical problems, when they were entered during the day by caregivers who entered few orders, when the medications ordered were potentially associated with greater risk, and when the warnings themselves were infrequently encountered.
METHODS
Setting and Caregivers
Johns Hopkins Bayview Medical Center (JHBMC) is a 400‐bed academic medical center serving southeastern Baltimore, Maryland. Prescribing caregivers include residents and fellows who rotate to both JHBMC and Johns Hopkins Hospital, internal medicine hospitalists, other attending physicians (including teaching attendings for all departments, and hospitalists and clinical associates for departments other than internal medicine), and nurse practitioners and physician assistants from most JHBMC departments. Nearly 100% of patients on the surgery, obstetrics/gynecology, neurology, psychiatry, and chemical dependence services are hospitalized on units dedicated to their respective specialty, and the same is true for approximately 95% of medicine patients.
Order Entry
JHBMC began using a client‐server order entry system by MEDITECH (Westwood, MA) in July 2003. Provider order entry was phased in beginning in October 2003 and completed by the end of 2004. MEDITECH version 5.64 was being used during the study period. Medications may generate duplicate, interaction, allergy, adverse reaction, and dose warnings during a patient ordering session each time they are ordered. Duplicate warnings are generated when the same medication (no matter what route) is ordered that is either on their active medication list, was on the list in the preceding 24 hours, or that is being ordered simultaneously. A drug‐interaction database licensed from First DataBank (South San Francisco, CA) is utilized, and updated monthly, which classifies potential drug‐drug interactions as contraindicated, severe, intermediate, and mild. Those classified as contraindicated by First DataBank are included in the severe category in MEDITECH 5.64. During the study period, JHBMC's version of MEDITECH was configured so that providers were warned of potential severe and intermediate drug‐drug interactions, but not mild. No other customizations had been made. Patients' histories of allergies and other adverse responses to medications can be entered by any credentialed staff member. They are maintained together in an allergies section of the electronic medical record, but are identified as either allergy or adverse reactions at the time they are entered, and each generates its own warnings.
When more than 1 duplicate, interaction, allergy, or adverse reaction warning is generated for a particular medication, all appear listed on a single screen in identical fonts. No visual distinction is made between severe and intermediate drug‐drug interactions; for these, the category of medication ordered is followed by the category of the medication for which there is a potential interaction. A details button can be selected to learn specifically which medications are involved and the severity and nature of the potential interactions identified. In response to the warnings, providers can choose to either override them, erase the order, or replace the order by clicking 1 of 3 buttons at the bottom of the screen. Warnings are not repeated unless the medication is reordered for that patient. Dose warnings appear on a subsequent screen and are not addressed in this article.
Nurses are discouraged from entering verbal orders but do have the capacity to do so, at which time they encounter and must respond to the standard medication warnings, if any. Medical students are able to enter orders, at which time they also encounter and must respond to the standard medication warnings; their orders must then be cosigned by a licensed provider before they can be processed. Warnings encountered by nurses and medical students are not repeated at the time of cosignature by a licensed provider.
Data Collection
We collected data regarding all medication orders placed in our CPOE system from October 1, 2009 to April 20, 2010 for all adult patients. Intensive care unit (ICU) patients were excluded, in anticipation of a separate analysis. Hospitalizations under observation were also excluded. We then ran a report showing all medications that generated any number of warnings of any type (duplicate, interaction, allergy, or adverse reaction) for the same population. Warnings generated during readmissions that occurred at any point during the study period (ranging from 1 to 21 times) were excluded, because these patients likely had many, if not all, of the same medications ordered during their readmissions as during their initial hospitalization, which would unduly influence the analysis if retained.
There was wide variation in the number of warnings generated per medication and in the number of each warning type per medication that generated multiple warnings. Therefore, for ease of analysis and to ensure that we could accurately determine varying response to each individual warning type, we thereafter focused on the medications that generated single warnings during the study period. For each single warning we obtained patient name, account number, event date and time, hospital unit at the time of the event, ordered medication, ordering staff member, warning type, and staff member response to the warning (eg, override warning or erase order [accept the warning]). The response replace was used very infrequently, and therefore warnings that resulted in this response were excluded. Medications available in more than 1 form included the route of administration in their name, and from this they were categorized as parenteral or nonparenteral. All nonparenteral or parenteral forms of a given medication were grouped together as 1 medication (eg, morphine sustained release and morphine elixir were classified as a single‐medication, nonparenteral morphine). Medications were further categorized according to whether or not they were on the Institute for Safe Medication Practice (ISMP) List of High‐Alert Medications.[30]
The study was approved by the Johns Hopkins Institutional Review Board.
Analysis
We collected descriptive data about patients and providers. Age and length of stay (LOS) at the time of the event were determined based on the patients' admit date and date of birth, and grouped into quartiles. Hospital units were grouped according to which service or services they primarily served. Medications were grouped into quartiles according to the total number of warnings they generated during the study period. Warnings were dichotomously categorized according to whether they were overridden or accepted. Unpaired t tests were used to compare continuous variables for the 2 groups, and [2] tests were used to compare categorical variables. A multivariate logistic regression was then performed, using variables with a P value of <0.10 in the univariate analysis, to control for confounders and identify independent predictors of medication warning acceptance. All analyses were performed using Intercooled Stata 12 (StataCorp, College Station, TX).
RESULTS
A total of 259,656 medication orders were placed for adult non‐ICU patients during the 7‐month study period. Of those orders, 45,835 generated some number of medication warnings.[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20] The median number of warnings per patient was 4 (interquartile range [IQR]=28; mean=5.9, standard deviation [SD]=6.2), with a range from 1 to 84. The median number of warnings generated per provider during the study period was 36 (IQR=6106, mean=87.4, SD=133.7), with a range of 1 to 1096.
There were 40,391 orders placed for 454 medications for adult non‐ICU patients, which generated a single‐medication warning (excluding those with the response replace, which was used 20 times) during the 7‐month study period. Data regarding the patients and providers associated with the orders generating single warnings are shown in Table 1. Most patients were on medicine units, and most orders were entered by residents. Patients' LOS at the time the orders were placed ranged from 0 to 118 days (median=1, IQR=04; mean=4.0, SD=7.2). The median number of single warnings per patient was 4 (IQR=28; mean=6.1, SD=6.5), with a range from 1 to 84. The median number of single warnings generated per provider during the study period was 15 (IQR=373; mean=61.7, SD=109.6), with a range of 1 to 1057.
| No. (%) | |
|---|---|
| |
| Patients (N=6,646) | |
| Age | |
| 1545 years | 2,048 (31%) |
| 4657 years | 1,610 (24%) |
| 5872 years | 1,520 (23%) |
| 73104 years | 1,468 (22%) |
| Gender | |
| Male | 2,934 (44%) |
| Hospital unita | |
| Medicine | 2,992 (45%) |
| Surgery | 1,836 (28%) |
| Neuro/psych/chem dep | 1,337 (20%) |
| OB/GYN | 481 (7%) |
| Caregivers (N=655) | |
| Resident | 248 (38%)b |
| Nurse | 154 (24%) |
| Attending or other | 97 (15%) |
| NP/PA | 69 (11%) |
| IM hospitalist | 31 (5%) |
| Fellow | 27 (4%) |
| Medical student | 23 (4%) |
| Pharmacist | 6 (1%) |
Patient and caregiver characteristics for the medication orders that generated single warnings are shown in Table 2. The majority of medications were nonparenteral and not on the ISMP list (Table 3). Most warnings generated were either duplicate (47%) or interaction warnings (47%). Warnings of a particular type were repeated 14.5% of the time for a particular medication and patient (from 2 to 24 times, median=2, IQR=22, mean=2.7, SD=1.4), and 9.8% of the time for a particular caregiver, medication, and patient (from 2 to 18 times, median=2, IQR=22, mean=2.4, SD=1.1).
| Variable | No. of Warnings (%)a | No. of Warnings Accepted (%)a | P |
|---|---|---|---|
| |||
| Patient age | |||
| 1545 years | 10,881 (27) | 602 (5.5%) | <0.001 |
| 4657 years | 9,733 (24) | 382 (3.9%) | |
| 5872 years | 10,000 (25) | 308 (3.1%) | |
| 73104 years | 9,777 (24) | 262 (2.7%) | |
| Patient gender | |||
| Female | 23,395 (58) | 866 (3.7%) | 0.074 |
| Male | 16,996 (42) | 688 (4.1%) | |
| Patient length of stay | |||
| <1 day | 10,721 (27) | 660 (6.2%) | <0.001 |
| 1 day | 10,854 (27) | 385 (3.5%) | |
| 24 days | 10,424 (26) | 277 (2.7%) | |
| 5118 days | 8,392 (21) | 232 (2.8%) | |
| Patient hospital unit | |||
| Medicine | 20,057 (50) | 519 (2.6%) | <0.001 |
| Surgery | 10,274 (25) | 477 (4.6%) | |
| Neuro/psych/chem dep | 8,279 (21) | 417 (5.0%) | |
| OB/GYN | 1,781 (4) | 141 (7.9%) | |
| Ordering caregiver | |||
| Resident | 22,523 (56) | 700 (3.1%) | <0.001 |
| NP/PA | 7,534 (19) | 369 (4.9%) | |
| IM hospitalist | 5,048 (13) | 155 (3.1%) | |
| Attending | 3225 (8) | 219 (6.8%) | |
| Fellow | 910 (2) | 34 (3.7%) | |
| Nurse | 865 (2) | 58 (6.7%) | |
| Medical student | 265 (<1) | 17 (6.4%) | |
| Pharmacist | 21 (<1) | 2 (9.5%) | |
| Day ordered | |||
| Weekday | 31,499 (78%) | 1276 (4.1%) | <0.001 |
| Weekend | 8,892 (22%) | 278 (3.1%) | |
| Time ordered | |||
| 00000559 | 4,231 (11%) | 117 (2.8%) | <0.001 |
| 06001159 | 11,696 (29%) | 348 (3.0%) | |
| 12001759 | 15,879 (39%) | 722 (4.6%) | |
| 18002359 | 8,585 (21%) | 367 (4.3%) | |
| Administration route (no. of meds) | |||
| Nonparenteral (339) | 27,086 (67%) | 956 (3.5%) | <0.001 |
| Parenteral (115) | 13,305 (33%) | 598 (4.5%) | |
| ISMP List of High‐Alert Medications status (no. of meds)[30] | |||
| Not on ISMP list (394) | 27,503 (68%) | 1251 (4.5%) | <0.001 |
| On ISMP list (60) | 12,888 (32%) | 303 (2.4%) | |
| No. of warnings per med (no. of meds) | |||
| 11062133 (7) | 9,869 (24%) | 191 (1.9%) | <0.001 |
| 4681034 (13) | 10,014 (25%) | 331 (3.3%) | |
| 170444 (40) | 10,182 (25%) | 314 (3.1%) | |
| 1169 (394) | 10,326 (26%) | 718 (7.0%) | |
| Warning type (no. of meds)b | |||
| Duplicate (369) | 19,083 (47%) | 1041 (5.5%) | <0.001 |
| Interaction (315) | 18,894 (47%) | 254 (1.3%) | |
| Allergy (138) | 2,371 (6%) | 243 (10.0%) | |
| Adverse reaction (14) | 43 (0.1%) | 16 (37%) | |
| Variable | Adjusted OR | 95% CI |
|---|---|---|
| ||
| Patient age | ||
| 1545 years | 1.00 | Reference |
| 4657 years | 0.89 | 0.771.02 |
| 5872 years | 0.85 | 0.730.99 |
| 73104 years | 0.91 | 0.771.08 |
| Patient gender | ||
| Female | 1.00 | Reference |
| Male | 1.26 | 1.131.41 |
| Patient length of stay | ||
| <1 day | 1.00 | Reference |
| 1 day | 0.65 | 0.550.76 |
| 24 days | 0.49 | 0.420.58 |
| 5118 days | 0.49 | 0.410.58 |
| Patient hospital unit | ||
| Medicine | 1.00 | Reference |
| Surgery | 1.45 | 1.251.68 |
| Neuro/psych/chem dep | 1.35 | 1.151.58 |
| OB/GYN | 2.43 | 1.923.08 |
| Ordering caregiver | ||
| Resident | 1.00 | Reference |
| NP/PA | 1.63 | 1.421.88 |
| IM hospitalist | 1.24 | 1.021.50 |
| Attending | 1.83 | 1.542.18 |
| Fellow | 1.41 | 0.982.03 |
| Nurse | 1.92 | 1.442.57 |
| Medical student | 1.17 | 0.701.95 |
| Pharmacist | 3.08 | 0.6714.03 |
| Medication factors | ||
| Nonparenteral | 1.00 | Reference |
| Parenteral | 1.79 | 1.592.03 |
| HighAlert Medication status (no. of meds)[30] | ||
| Not on ISMP list | 1.00 | Reference |
| On ISMP list | 0.37 | 0.320.43 |
| No. of warnings per medication | ||
| 11062133 | 1.00 | Reference |
| 4681034 | 2.30 | 1.902.79 |
| 170444 | 2.25 | 1.852.73 |
| 1169 | 4.10 | 3.424.92 |
| Warning type | ||
| Duplicate | 1.00 | Reference |
| Interaction | 0.24 | 0.210.28 |
| Allergy | 2.28 | 1.942.68 |
| Adverse reaction | 9.24 | 4.5218.90 |
One thousand five hundred fifty‐four warnings were erased (ie, accepted by clinicians [4%]). In univariate analysis, only patient gender was not associated with warning acceptance. Patient age, LOS, hospital unit at the time of order entry, ordering caregiver type, day and time the medication was ordered, administration route, presence on the ISMP list, warning frequency, and warning type were all significantly associated with warning acceptance (Table 2).
Older patient age, longer LOS, presence of the medication on the ISMP list, and interaction warning type were all negatively associated with warning acceptance in multivariable analysis. Warning acceptance was positively associated with male patient gender, being on a service other than medicine, being a caregiver other than a resident, parenteral medications, lower warning frequency, and allergy or adverse reaction warning types (Table 3).
The 20 medications that generated the most single warnings are shown in Table 4. Medications on the ISMP list accounted for 8 of these top 20 medications. For most of them, duplicate and interaction warnings accounted for most of the warnings generated, except for parenteral hydromorphone, oral oxycodone, parenteral morphine, and oral hydromorphone, which each had more allergy than interaction warnings.
| Medication | ISMP Listb | No. of Warnings | Duplicate, No. (%)c | Interaction, No. (%)c | Allergy, No. (%)c | Adverse Reaction, No. (%)c |
|---|---|---|---|---|---|---|
| ||||||
| Hydromorphone injectable | Yes | 2,133 | 1,584 (74.3) | 127 (6.0) | 422 (19.8) | |
| Metoprolol | 1,432 | 550 (38.4) | 870 (60.8) | 12 (0.8) | ||
| Aspirin | 1,375 | 212 (15.4) | 1,096 (79.7) | 67 (4.9) | ||
| Oxycodone | Yes | 1,360 | 987 (72.6) | 364 (26.8) | 9 (0.7) | |
| Potassium chloride | 1,296 | 379 (29.2) | 917 (70.8) | |||
| Ondansetron injectable | 1,167 | 1,013 (86.8) | 153 (13.1) | 1 (0.1) | ||
| Aspart insulin injectable | Yes | 1,106 | 643 (58.1) | 463 (41.9) | ||
| Warfarin | Yes | 1,034 | 298 (28.8) | 736 (71.2) | ||
| Heparin injectable | Yes | 1,030 | 205 (19.9) | 816 (79.2) | 9 (0.3) | |
| Furosemide injectable | 980 | 438 (45.0) | 542 (55.3) | |||
| Lisinopril | 926 | 225 (24.3) | 698 (75.4) | 3 (0.3) | ||
| Acetaminophen | 860 | 686 (79.8) | 118 (13.7) | 54 (6.3) | 2 (0.2) | |
| Morphine injectable | Yes | 804 | 467 (58.1) | 100 (12.4) | 233 (29.0) | 4 (0.5) |
| Diazepam | 786 | 731 (93.0) | 41 (5.2) | 14 (1.8) | ||
| Glargine insulin injectable | Yes | 746 | 268 (35.9) | 478 (64.1) | ||
| Ibuprofen | 713 | 125 (17.5) | 529 (74.2) | 54 (7.6) | 5 (0.7) | |
| Hydromorphone | Yes | 594 | 372 (62.6) | 31 (5.2) | 187 (31.5) | 4 (0.7) |
| Furosemide | 586 | 273 (46.6) | 312 (53.2) | 1 (0.2) | ||
| Ketorolac injectable | 487 | 39 (8.0) | 423 (86.9) | 23 (4.7) | 2 (0.4) | |
| Prednisone | 468 | 166 (35.5) | 297 (63.5) | 5 (1.1) | ||
DISCUSSION
Medication warnings in our study were frequently overridden, particularly when encountered by residents, for patients with a long LOS and on the internal medicine service, and for medications generating the most warnings and on the ISMP list. Disturbingly, this means that potentially important warnings for medications with the highest potential for causing harm, for possibly the sickest and most complex patients, were those that were most often ignored by young physicians in training who should have had the most to gain from them. Of course, this is not entirely surprising. Despite our hope that a culture of safety would influence young physicians' actions when caring for these patients and prescribing these medications, these patients and medications are those for whom the most warnings are generated, and these physicians are the ones entering the most orders. Only 13% of the medications studied were on the ISMP list, but they generated 32% of the warnings. We controlled for number of warnings and ISMP list status, but not for warning validity. Most likely, high‐risk medications have been set up with more warnings, many of them of lower quality, in an errant but well‐intentioned effort to make them safer. If developers of CPOE systems want to gain serious traction in using decision support to promote prescribing safe medications, they must take substantial action to increase attention to important warnings and decrease the number of clinically insignificant, low‐value warnings encountered by active caregivers on a daily basis.
Only 2 prior studies, both by Seidling et al., have specifically looked at provider response to warnings for high risk medications. Interaction warnings were rarely accepted in 1,[18] as in our study; however, in contrast to our findings, warning acceptance in both studies was higher for drugs with dose‐dependent toxicity.[18, 26] The effect of physician experience on warning acceptance has been addressed in 2 prior studies. In Weingart et al., residents were more likely than staff physicians to erase medication orders when presented with allergy and interaction warnings in a primary care setting.[20] Long et al. found that physicians younger than 40 years were less likely than older physicians to accept duplicate warnings, but those who had been at the study hospital for a longer period of time were more likely to accept them.[23] The influence of patient LOS and service on warning acceptance has not previously been described. Further study is needed looking at each of these factors.
Individual hospitals tend to avoid making modifications to order entry warning systems, because monitoring and maintaining these changes is labor intensive. Some institutions may make the decision to turn off certain categories of alerts, such as intermediate interaction warnings, to minimize the noise their providers encounter. There are even tools for disabling individual alerts or groups of alerts, such as that available for purchase from our interaction database vendor.[31] However, institutions may fear litigation should an adverse event be attributed to a disabled warning.[15, 16] Clearly, a comprehensive, health system‐wide approach is warranted.[13, 15] To date, published efforts describing ways to improve the effectiveness of medication warning systems have focused on either heightening the clinical significance of alerts[14, 21, 22, 32, 33, 34, 35, 36] or altering their presentation and how providers experience them.[21, 36, 37, 38, 39, 40, 41, 42, 43] The single medication warnings our providers receive are all presented in an identical font, and presumably response to each would be different if they were better distinguished from each other. We also found that a small but significant number of warnings were repeated for a given patient and even a given provider. If the providers knew they would only be presented with warnings the first time they occurred for a given patient and medication, they might be more attuned to the remaining warnings. Previous studies describe context‐specific decision support for medication ordering[44, 45, 46]; however, only 1 has described the use of patient context factors to modify when or how warnings are presented to providers.[47] None have described tailoring allergy, duplicate, and interaction warnings according to medication or provider types. If further study confirms our findings, modulating basic warning systems according to severity of illness, provider experience, and medication risk could powerfully increase their effectiveness. Of course, this would be extremely challenging to achieve, and is likely outside the capabilities of most, if not all, CPOE systems, at least for now.
Our study has some limitations. First, it was limited to medications that generated a single warning. We did this for ease of analysis and so that we could ensure understanding of provider response to each warning type without bias from simultaneously occurring warnings; however, caregiver response to multiple warnings appearing simultaneously for a particular medication order might be quite different. Second, we did not include any assessment of the number of medications ordered by each provider type or for each patient, either of which could significantly affect provider response to warnings. Third, as previously noted, we did not include any assessment of the validity of the warnings, beyond the 4 main categories described, which could also significantly affect provider response. However, it should be noted that although the validity of interaction warnings varies significantly from 1 medication to another, the validity of duplicate, allergy, and adverse reaction warnings in the described system are essentially the same for all medications. Fourth, it is possible that providers did modify or even erase their orders even after selecting override in response to the warning; it is also possible that providers reentered the same order after choosing erase. Unfortunately auditing for actions such as these would be extremely laborious. Finally, the study was conducted at a single medical center using a single order‐entry system. The system in use at our medical center is in use at one‐third of the 6000 hospitals in the United States, though certainly not all are using our version. Even if a hospital was using the same CPOE version and interaction database as our institution, variations in patient population and local decisions modifying how the database interacts with the warning presentation system might affect reproducibility at that institution.
Commonly encountered medication warnings are overridden at extremely high rates, and in our study this was particularly so for medications on the ISMP list, when ordered by physicians in training. Warnings of little clinical significance must be identified and eliminated, the most important warnings need to be visually distinct to increase user attention, and further research should be done into the patient, provider, setting, and medication factors that affect user responses to warnings, so that they may be customized accordingly and their significance increased. Doing so will enable us to reap the maximum possible potential from our CPOE systems, and increase the CPOE's power to protect our most vulnerable patients from our most dangerous medications, particularly when cared for by our most inexperienced physicians.
Acknowledgements
The authors thank, in particular, Scott Carey, Research Informatics Manager, for assistance with data collection. Additional thanks go to Olga Sherman and Kathleen Ancinich for assistance with data collection and management.
Disclosures: This research was supported in part by the Johns Hopkins Institute for Clinical and Translational Research. All listed authors contributed substantially to the study conception and design, analysis and interpretation of data, drafting the article or revising it critically for important intellectual content, and final approval of the version to be published. No one who fulfills these criteria has been excluded from authorship. This research received no specific grant from any funding agency in the public, commercial, or not‐for‐profit sectors. The authors have no competing interests to declare.
Many computerized provider order entry (CPOE) systems suffer from having too much of a good thing. Few would question the beneficial effect of CPOE on medication order clarity, completeness, and transmission.[1, 2] When mechanisms for basic decision support have been added, however, such as allergy, interaction, and duplicate warnings, reductions in medication errors and adverse events have not been consistently achieved.[3, 4, 5, 6, 7] This is likely due in part to the fact that ordering providers override medication warnings at staggeringly high rates.[8, 9] Clinicians acknowledge that they are ignoring potentially valuable warnings,[10, 11] but suffer from alert fatigue due to the sheer number of messages, many of them judged by clinicians to be of low‐value.[11, 12]
Redesign of medication alert systems to increase their signal‐to‐noise ratio is badly needed,[13, 14, 15, 16] and will need to consider the clinical significance of alerts, their presentation, and context‐specific factors that potentially contribute to warning effectiveness.[17, 18, 19] Relatively few studies, however, have objectively looked at context factors such as the characteristics of providers, patients, medications, and warnings that are associated with provider responses to warnings,[9, 20, 21, 22, 23, 24, 25] and only 2 have studied how warning acceptance is associated with medication risk.[18, 26] We wished to explore these factors further. Warning acceptance has been shown to be higher, at least in the outpatient setting, when orders are entered by low‐volume prescribers for infrequently encountered warnings,[24] and there is some evidence that patients receive higher‐quality care during the day.[27] Significant attention has been placed in recent years on inappropriate prescribing in older patients,[28] and on creating a culture of safety in healthcare.[29] We therefore hypothesized that our providers would be more cautious, and medication warning acceptance rates would be higher, when orders were entered for patients who were older or with more complex medical problems, when they were entered during the day by caregivers who entered few orders, when the medications ordered were potentially associated with greater risk, and when the warnings themselves were infrequently encountered.
METHODS
Setting and Caregivers
Johns Hopkins Bayview Medical Center (JHBMC) is a 400‐bed academic medical center serving southeastern Baltimore, Maryland. Prescribing caregivers include residents and fellows who rotate to both JHBMC and Johns Hopkins Hospital, internal medicine hospitalists, other attending physicians (including teaching attendings for all departments, and hospitalists and clinical associates for departments other than internal medicine), and nurse practitioners and physician assistants from most JHBMC departments. Nearly 100% of patients on the surgery, obstetrics/gynecology, neurology, psychiatry, and chemical dependence services are hospitalized on units dedicated to their respective specialty, and the same is true for approximately 95% of medicine patients.
Order Entry
JHBMC began using a client‐server order entry system by MEDITECH (Westwood, MA) in July 2003. Provider order entry was phased in beginning in October 2003 and completed by the end of 2004. MEDITECH version 5.64 was being used during the study period. Medications may generate duplicate, interaction, allergy, adverse reaction, and dose warnings during a patient ordering session each time they are ordered. Duplicate warnings are generated when the same medication (no matter what route) is ordered that is either on their active medication list, was on the list in the preceding 24 hours, or that is being ordered simultaneously. A drug‐interaction database licensed from First DataBank (South San Francisco, CA) is utilized, and updated monthly, which classifies potential drug‐drug interactions as contraindicated, severe, intermediate, and mild. Those classified as contraindicated by First DataBank are included in the severe category in MEDITECH 5.64. During the study period, JHBMC's version of MEDITECH was configured so that providers were warned of potential severe and intermediate drug‐drug interactions, but not mild. No other customizations had been made. Patients' histories of allergies and other adverse responses to medications can be entered by any credentialed staff member. They are maintained together in an allergies section of the electronic medical record, but are identified as either allergy or adverse reactions at the time they are entered, and each generates its own warnings.
When more than 1 duplicate, interaction, allergy, or adverse reaction warning is generated for a particular medication, all appear listed on a single screen in identical fonts. No visual distinction is made between severe and intermediate drug‐drug interactions; for these, the category of medication ordered is followed by the category of the medication for which there is a potential interaction. A details button can be selected to learn specifically which medications are involved and the severity and nature of the potential interactions identified. In response to the warnings, providers can choose to either override them, erase the order, or replace the order by clicking 1 of 3 buttons at the bottom of the screen. Warnings are not repeated unless the medication is reordered for that patient. Dose warnings appear on a subsequent screen and are not addressed in this article.
Nurses are discouraged from entering verbal orders but do have the capacity to do so, at which time they encounter and must respond to the standard medication warnings, if any. Medical students are able to enter orders, at which time they also encounter and must respond to the standard medication warnings; their orders must then be cosigned by a licensed provider before they can be processed. Warnings encountered by nurses and medical students are not repeated at the time of cosignature by a licensed provider.
Data Collection
We collected data regarding all medication orders placed in our CPOE system from October 1, 2009 to April 20, 2010 for all adult patients. Intensive care unit (ICU) patients were excluded, in anticipation of a separate analysis. Hospitalizations under observation were also excluded. We then ran a report showing all medications that generated any number of warnings of any type (duplicate, interaction, allergy, or adverse reaction) for the same population. Warnings generated during readmissions that occurred at any point during the study period (ranging from 1 to 21 times) were excluded, because these patients likely had many, if not all, of the same medications ordered during their readmissions as during their initial hospitalization, which would unduly influence the analysis if retained.
There was wide variation in the number of warnings generated per medication and in the number of each warning type per medication that generated multiple warnings. Therefore, for ease of analysis and to ensure that we could accurately determine varying response to each individual warning type, we thereafter focused on the medications that generated single warnings during the study period. For each single warning we obtained patient name, account number, event date and time, hospital unit at the time of the event, ordered medication, ordering staff member, warning type, and staff member response to the warning (eg, override warning or erase order [accept the warning]). The response replace was used very infrequently, and therefore warnings that resulted in this response were excluded. Medications available in more than 1 form included the route of administration in their name, and from this they were categorized as parenteral or nonparenteral. All nonparenteral or parenteral forms of a given medication were grouped together as 1 medication (eg, morphine sustained release and morphine elixir were classified as a single‐medication, nonparenteral morphine). Medications were further categorized according to whether or not they were on the Institute for Safe Medication Practice (ISMP) List of High‐Alert Medications.[30]
The study was approved by the Johns Hopkins Institutional Review Board.
Analysis
We collected descriptive data about patients and providers. Age and length of stay (LOS) at the time of the event were determined based on the patients' admit date and date of birth, and grouped into quartiles. Hospital units were grouped according to which service or services they primarily served. Medications were grouped into quartiles according to the total number of warnings they generated during the study period. Warnings were dichotomously categorized according to whether they were overridden or accepted. Unpaired t tests were used to compare continuous variables for the 2 groups, and [2] tests were used to compare categorical variables. A multivariate logistic regression was then performed, using variables with a P value of <0.10 in the univariate analysis, to control for confounders and identify independent predictors of medication warning acceptance. All analyses were performed using Intercooled Stata 12 (StataCorp, College Station, TX).
RESULTS
A total of 259,656 medication orders were placed for adult non‐ICU patients during the 7‐month study period. Of those orders, 45,835 generated some number of medication warnings.[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20] The median number of warnings per patient was 4 (interquartile range [IQR]=28; mean=5.9, standard deviation [SD]=6.2), with a range from 1 to 84. The median number of warnings generated per provider during the study period was 36 (IQR=6106, mean=87.4, SD=133.7), with a range of 1 to 1096.
There were 40,391 orders placed for 454 medications for adult non‐ICU patients, which generated a single‐medication warning (excluding those with the response replace, which was used 20 times) during the 7‐month study period. Data regarding the patients and providers associated with the orders generating single warnings are shown in Table 1. Most patients were on medicine units, and most orders were entered by residents. Patients' LOS at the time the orders were placed ranged from 0 to 118 days (median=1, IQR=04; mean=4.0, SD=7.2). The median number of single warnings per patient was 4 (IQR=28; mean=6.1, SD=6.5), with a range from 1 to 84. The median number of single warnings generated per provider during the study period was 15 (IQR=373; mean=61.7, SD=109.6), with a range of 1 to 1057.
| No. (%) | |
|---|---|
| |
| Patients (N=6,646) | |
| Age | |
| 1545 years | 2,048 (31%) |
| 4657 years | 1,610 (24%) |
| 5872 years | 1,520 (23%) |
| 73104 years | 1,468 (22%) |
| Gender | |
| Male | 2,934 (44%) |
| Hospital unita | |
| Medicine | 2,992 (45%) |
| Surgery | 1,836 (28%) |
| Neuro/psych/chem dep | 1,337 (20%) |
| OB/GYN | 481 (7%) |
| Caregivers (N=655) | |
| Resident | 248 (38%)b |
| Nurse | 154 (24%) |
| Attending or other | 97 (15%) |
| NP/PA | 69 (11%) |
| IM hospitalist | 31 (5%) |
| Fellow | 27 (4%) |
| Medical student | 23 (4%) |
| Pharmacist | 6 (1%) |
Patient and caregiver characteristics for the medication orders that generated single warnings are shown in Table 2. The majority of medications were nonparenteral and not on the ISMP list (Table 3). Most warnings generated were either duplicate (47%) or interaction warnings (47%). Warnings of a particular type were repeated 14.5% of the time for a particular medication and patient (from 2 to 24 times, median=2, IQR=22, mean=2.7, SD=1.4), and 9.8% of the time for a particular caregiver, medication, and patient (from 2 to 18 times, median=2, IQR=22, mean=2.4, SD=1.1).
| Variable | No. of Warnings (%)a | No. of Warnings Accepted (%)a | P |
|---|---|---|---|
| |||
| Patient age | |||
| 1545 years | 10,881 (27) | 602 (5.5%) | <0.001 |
| 4657 years | 9,733 (24) | 382 (3.9%) | |
| 5872 years | 10,000 (25) | 308 (3.1%) | |
| 73104 years | 9,777 (24) | 262 (2.7%) | |
| Patient gender | |||
| Female | 23,395 (58) | 866 (3.7%) | 0.074 |
| Male | 16,996 (42) | 688 (4.1%) | |
| Patient length of stay | |||
| <1 day | 10,721 (27) | 660 (6.2%) | <0.001 |
| 1 day | 10,854 (27) | 385 (3.5%) | |
| 24 days | 10,424 (26) | 277 (2.7%) | |
| 5118 days | 8,392 (21) | 232 (2.8%) | |
| Patient hospital unit | |||
| Medicine | 20,057 (50) | 519 (2.6%) | <0.001 |
| Surgery | 10,274 (25) | 477 (4.6%) | |
| Neuro/psych/chem dep | 8,279 (21) | 417 (5.0%) | |
| OB/GYN | 1,781 (4) | 141 (7.9%) | |
| Ordering caregiver | |||
| Resident | 22,523 (56) | 700 (3.1%) | <0.001 |
| NP/PA | 7,534 (19) | 369 (4.9%) | |
| IM hospitalist | 5,048 (13) | 155 (3.1%) | |
| Attending | 3225 (8) | 219 (6.8%) | |
| Fellow | 910 (2) | 34 (3.7%) | |
| Nurse | 865 (2) | 58 (6.7%) | |
| Medical student | 265 (<1) | 17 (6.4%) | |
| Pharmacist | 21 (<1) | 2 (9.5%) | |
| Day ordered | |||
| Weekday | 31,499 (78%) | 1276 (4.1%) | <0.001 |
| Weekend | 8,892 (22%) | 278 (3.1%) | |
| Time ordered | |||
| 00000559 | 4,231 (11%) | 117 (2.8%) | <0.001 |
| 06001159 | 11,696 (29%) | 348 (3.0%) | |
| 12001759 | 15,879 (39%) | 722 (4.6%) | |
| 18002359 | 8,585 (21%) | 367 (4.3%) | |
| Administration route (no. of meds) | |||
| Nonparenteral (339) | 27,086 (67%) | 956 (3.5%) | <0.001 |
| Parenteral (115) | 13,305 (33%) | 598 (4.5%) | |
| ISMP List of High‐Alert Medications status (no. of meds)[30] | |||
| Not on ISMP list (394) | 27,503 (68%) | 1251 (4.5%) | <0.001 |
| On ISMP list (60) | 12,888 (32%) | 303 (2.4%) | |
| No. of warnings per med (no. of meds) | |||
| 11062133 (7) | 9,869 (24%) | 191 (1.9%) | <0.001 |
| 4681034 (13) | 10,014 (25%) | 331 (3.3%) | |
| 170444 (40) | 10,182 (25%) | 314 (3.1%) | |
| 1169 (394) | 10,326 (26%) | 718 (7.0%) | |
| Warning type (no. of meds)b | |||
| Duplicate (369) | 19,083 (47%) | 1041 (5.5%) | <0.001 |
| Interaction (315) | 18,894 (47%) | 254 (1.3%) | |
| Allergy (138) | 2,371 (6%) | 243 (10.0%) | |
| Adverse reaction (14) | 43 (0.1%) | 16 (37%) | |
| Variable | Adjusted OR | 95% CI |
|---|---|---|
| ||
| Patient age | ||
| 1545 years | 1.00 | Reference |
| 4657 years | 0.89 | 0.771.02 |
| 5872 years | 0.85 | 0.730.99 |
| 73104 years | 0.91 | 0.771.08 |
| Patient gender | ||
| Female | 1.00 | Reference |
| Male | 1.26 | 1.131.41 |
| Patient length of stay | ||
| <1 day | 1.00 | Reference |
| 1 day | 0.65 | 0.550.76 |
| 24 days | 0.49 | 0.420.58 |
| 5118 days | 0.49 | 0.410.58 |
| Patient hospital unit | ||
| Medicine | 1.00 | Reference |
| Surgery | 1.45 | 1.251.68 |
| Neuro/psych/chem dep | 1.35 | 1.151.58 |
| OB/GYN | 2.43 | 1.923.08 |
| Ordering caregiver | ||
| Resident | 1.00 | Reference |
| NP/PA | 1.63 | 1.421.88 |
| IM hospitalist | 1.24 | 1.021.50 |
| Attending | 1.83 | 1.542.18 |
| Fellow | 1.41 | 0.982.03 |
| Nurse | 1.92 | 1.442.57 |
| Medical student | 1.17 | 0.701.95 |
| Pharmacist | 3.08 | 0.6714.03 |
| Medication factors | ||
| Nonparenteral | 1.00 | Reference |
| Parenteral | 1.79 | 1.592.03 |
| HighAlert Medication status (no. of meds)[30] | ||
| Not on ISMP list | 1.00 | Reference |
| On ISMP list | 0.37 | 0.320.43 |
| No. of warnings per medication | ||
| 11062133 | 1.00 | Reference |
| 4681034 | 2.30 | 1.902.79 |
| 170444 | 2.25 | 1.852.73 |
| 1169 | 4.10 | 3.424.92 |
| Warning type | ||
| Duplicate | 1.00 | Reference |
| Interaction | 0.24 | 0.210.28 |
| Allergy | 2.28 | 1.942.68 |
| Adverse reaction | 9.24 | 4.5218.90 |
One thousand five hundred fifty‐four warnings were erased (ie, accepted by clinicians [4%]). In univariate analysis, only patient gender was not associated with warning acceptance. Patient age, LOS, hospital unit at the time of order entry, ordering caregiver type, day and time the medication was ordered, administration route, presence on the ISMP list, warning frequency, and warning type were all significantly associated with warning acceptance (Table 2).
Older patient age, longer LOS, presence of the medication on the ISMP list, and interaction warning type were all negatively associated with warning acceptance in multivariable analysis. Warning acceptance was positively associated with male patient gender, being on a service other than medicine, being a caregiver other than a resident, parenteral medications, lower warning frequency, and allergy or adverse reaction warning types (Table 3).
The 20 medications that generated the most single warnings are shown in Table 4. Medications on the ISMP list accounted for 8 of these top 20 medications. For most of them, duplicate and interaction warnings accounted for most of the warnings generated, except for parenteral hydromorphone, oral oxycodone, parenteral morphine, and oral hydromorphone, which each had more allergy than interaction warnings.
| Medication | ISMP Listb | No. of Warnings | Duplicate, No. (%)c | Interaction, No. (%)c | Allergy, No. (%)c | Adverse Reaction, No. (%)c |
|---|---|---|---|---|---|---|
| ||||||
| Hydromorphone injectable | Yes | 2,133 | 1,584 (74.3) | 127 (6.0) | 422 (19.8) | |
| Metoprolol | 1,432 | 550 (38.4) | 870 (60.8) | 12 (0.8) | ||
| Aspirin | 1,375 | 212 (15.4) | 1,096 (79.7) | 67 (4.9) | ||
| Oxycodone | Yes | 1,360 | 987 (72.6) | 364 (26.8) | 9 (0.7) | |
| Potassium chloride | 1,296 | 379 (29.2) | 917 (70.8) | |||
| Ondansetron injectable | 1,167 | 1,013 (86.8) | 153 (13.1) | 1 (0.1) | ||
| Aspart insulin injectable | Yes | 1,106 | 643 (58.1) | 463 (41.9) | ||
| Warfarin | Yes | 1,034 | 298 (28.8) | 736 (71.2) | ||
| Heparin injectable | Yes | 1,030 | 205 (19.9) | 816 (79.2) | 9 (0.3) | |
| Furosemide injectable | 980 | 438 (45.0) | 542 (55.3) | |||
| Lisinopril | 926 | 225 (24.3) | 698 (75.4) | 3 (0.3) | ||
| Acetaminophen | 860 | 686 (79.8) | 118 (13.7) | 54 (6.3) | 2 (0.2) | |
| Morphine injectable | Yes | 804 | 467 (58.1) | 100 (12.4) | 233 (29.0) | 4 (0.5) |
| Diazepam | 786 | 731 (93.0) | 41 (5.2) | 14 (1.8) | ||
| Glargine insulin injectable | Yes | 746 | 268 (35.9) | 478 (64.1) | ||
| Ibuprofen | 713 | 125 (17.5) | 529 (74.2) | 54 (7.6) | 5 (0.7) | |
| Hydromorphone | Yes | 594 | 372 (62.6) | 31 (5.2) | 187 (31.5) | 4 (0.7) |
| Furosemide | 586 | 273 (46.6) | 312 (53.2) | 1 (0.2) | ||
| Ketorolac injectable | 487 | 39 (8.0) | 423 (86.9) | 23 (4.7) | 2 (0.4) | |
| Prednisone | 468 | 166 (35.5) | 297 (63.5) | 5 (1.1) | ||
DISCUSSION
Medication warnings in our study were frequently overridden, particularly when encountered by residents, for patients with a long LOS and on the internal medicine service, and for medications generating the most warnings and on the ISMP list. Disturbingly, this means that potentially important warnings for medications with the highest potential for causing harm, for possibly the sickest and most complex patients, were those that were most often ignored by young physicians in training who should have had the most to gain from them. Of course, this is not entirely surprising. Despite our hope that a culture of safety would influence young physicians' actions when caring for these patients and prescribing these medications, these patients and medications are those for whom the most warnings are generated, and these physicians are the ones entering the most orders. Only 13% of the medications studied were on the ISMP list, but they generated 32% of the warnings. We controlled for number of warnings and ISMP list status, but not for warning validity. Most likely, high‐risk medications have been set up with more warnings, many of them of lower quality, in an errant but well‐intentioned effort to make them safer. If developers of CPOE systems want to gain serious traction in using decision support to promote prescribing safe medications, they must take substantial action to increase attention to important warnings and decrease the number of clinically insignificant, low‐value warnings encountered by active caregivers on a daily basis.
Only 2 prior studies, both by Seidling et al., have specifically looked at provider response to warnings for high risk medications. Interaction warnings were rarely accepted in 1,[18] as in our study; however, in contrast to our findings, warning acceptance in both studies was higher for drugs with dose‐dependent toxicity.[18, 26] The effect of physician experience on warning acceptance has been addressed in 2 prior studies. In Weingart et al., residents were more likely than staff physicians to erase medication orders when presented with allergy and interaction warnings in a primary care setting.[20] Long et al. found that physicians younger than 40 years were less likely than older physicians to accept duplicate warnings, but those who had been at the study hospital for a longer period of time were more likely to accept them.[23] The influence of patient LOS and service on warning acceptance has not previously been described. Further study is needed looking at each of these factors.
Individual hospitals tend to avoid making modifications to order entry warning systems, because monitoring and maintaining these changes is labor intensive. Some institutions may make the decision to turn off certain categories of alerts, such as intermediate interaction warnings, to minimize the noise their providers encounter. There are even tools for disabling individual alerts or groups of alerts, such as that available for purchase from our interaction database vendor.[31] However, institutions may fear litigation should an adverse event be attributed to a disabled warning.[15, 16] Clearly, a comprehensive, health system‐wide approach is warranted.[13, 15] To date, published efforts describing ways to improve the effectiveness of medication warning systems have focused on either heightening the clinical significance of alerts[14, 21, 22, 32, 33, 34, 35, 36] or altering their presentation and how providers experience them.[21, 36, 37, 38, 39, 40, 41, 42, 43] The single medication warnings our providers receive are all presented in an identical font, and presumably response to each would be different if they were better distinguished from each other. We also found that a small but significant number of warnings were repeated for a given patient and even a given provider. If the providers knew they would only be presented with warnings the first time they occurred for a given patient and medication, they might be more attuned to the remaining warnings. Previous studies describe context‐specific decision support for medication ordering[44, 45, 46]; however, only 1 has described the use of patient context factors to modify when or how warnings are presented to providers.[47] None have described tailoring allergy, duplicate, and interaction warnings according to medication or provider types. If further study confirms our findings, modulating basic warning systems according to severity of illness, provider experience, and medication risk could powerfully increase their effectiveness. Of course, this would be extremely challenging to achieve, and is likely outside the capabilities of most, if not all, CPOE systems, at least for now.
Our study has some limitations. First, it was limited to medications that generated a single warning. We did this for ease of analysis and so that we could ensure understanding of provider response to each warning type without bias from simultaneously occurring warnings; however, caregiver response to multiple warnings appearing simultaneously for a particular medication order might be quite different. Second, we did not include any assessment of the number of medications ordered by each provider type or for each patient, either of which could significantly affect provider response to warnings. Third, as previously noted, we did not include any assessment of the validity of the warnings, beyond the 4 main categories described, which could also significantly affect provider response. However, it should be noted that although the validity of interaction warnings varies significantly from 1 medication to another, the validity of duplicate, allergy, and adverse reaction warnings in the described system are essentially the same for all medications. Fourth, it is possible that providers did modify or even erase their orders even after selecting override in response to the warning; it is also possible that providers reentered the same order after choosing erase. Unfortunately auditing for actions such as these would be extremely laborious. Finally, the study was conducted at a single medical center using a single order‐entry system. The system in use at our medical center is in use at one‐third of the 6000 hospitals in the United States, though certainly not all are using our version. Even if a hospital was using the same CPOE version and interaction database as our institution, variations in patient population and local decisions modifying how the database interacts with the warning presentation system might affect reproducibility at that institution.
Commonly encountered medication warnings are overridden at extremely high rates, and in our study this was particularly so for medications on the ISMP list, when ordered by physicians in training. Warnings of little clinical significance must be identified and eliminated, the most important warnings need to be visually distinct to increase user attention, and further research should be done into the patient, provider, setting, and medication factors that affect user responses to warnings, so that they may be customized accordingly and their significance increased. Doing so will enable us to reap the maximum possible potential from our CPOE systems, and increase the CPOE's power to protect our most vulnerable patients from our most dangerous medications, particularly when cared for by our most inexperienced physicians.
Acknowledgements
The authors thank, in particular, Scott Carey, Research Informatics Manager, for assistance with data collection. Additional thanks go to Olga Sherman and Kathleen Ancinich for assistance with data collection and management.
Disclosures: This research was supported in part by the Johns Hopkins Institute for Clinical and Translational Research. All listed authors contributed substantially to the study conception and design, analysis and interpretation of data, drafting the article or revising it critically for important intellectual content, and final approval of the version to be published. No one who fulfills these criteria has been excluded from authorship. This research received no specific grant from any funding agency in the public, commercial, or not‐for‐profit sectors. The authors have no competing interests to declare.
- , , , et al., Effect of computerized physician order entry and a team intervention on prevention of serious medication errors. JAMA. 1998;280:1311–1316.
- , , , , , . Effects of computerized provider order entry on prescribing practices. Arch Intern Med. 2000;160:2741–2747.
- , , , et al. Effects of computerized clinician decision support systems on practitioner performance and patient outcomes: a systematic review. JAMA. 2005;293:1223–1238.
- , , , et al. The effect of computerized physician order entry with clinical decision support on the rates of adverse drug events: a systematic review. J Gen Intern Med. 2008;23:451–458.
- , , . The impact of computerized physician medication order entry in hospitalized patients—a systematic review. Int J Med Inform. 2008;77:365–376.
- , , , et al. What evidence supports the use of computerized alerts and prompts to improve clinicians' prescribing behavior? J Am Med Inform Assoc. 2009;16:531–538.
- , , , , . Does computerized provider order entry reduce prescribing errors for hospital inpatients? A systematic review. J Am Med Inform Assoc. 2009;16:613–623.
- , , , . Overriding of drug safety alerts in computerized physician order entry. J Am Med Inform Assoc. 2006;13:138–147.
- , , , , , . Evaluating clinical decision support systems: monitoring CPOE order check override rates in the Department of Veterans Affairs' Computerized Patient Record System. J Am Med Inform Assoc. 2008;15:620–626.
- , , . GPs' views on computerized drug interaction alerts: questionnaire survey. J Clin Pharm Ther. 2002;27:377–382.
- , , , et al. Clinicians' assessments of electronic medication safety alerts in ambulatory care. Arch Intern Med. 2009;169:1627–1632.
- , , , , . A mixed method study of the merits of e‐prescribing drug alerts in primary care. J Gen Intern Med. 2008;23:442–446.
- . CPOE and clinical decision support in hospitals: getting the benefits: comment on “Unintended effects of a computerized physician order entry nearly hard‐stop alert to prevent a drug interaction.” Arch Intern Med. 2010;170:1583–1584.
- , , . Critical drug‐drug interactions for use in electronic health records systems with computerized physician order entry: review of leading approaches. J Patient Saf. 2011;7:61–65.
- , , , , . Clinical decision support systems could be modified to reduce 'alert fatigue' while still minimizing the risk of litigation. Health Aff (Millwood). 2011;30:2310–2317.
- , , , . Critical issues associated with drug‐drug interactions: highlights of a multistakeholder conference. Am J Health Syst Pharm. 2011;68:941–946.
- , , , , , . Development of a context model to prioritize drug safety alerts in CPOE systems. BMC Med Inform Decis Mak. 2011;11:35.
- , , , et al. Factors influencing alert acceptance: a novel approach for predicting the success of clinical decision support. J Am Med Inform Assoc. 2011;18:479–484.
- , , , . How to improve the delivery of medication alerts within computerized physician order entry systems: an international Delphi study. J Am Med Inform Assoc. 2011;18:760–766.
- , , , , , . Physicians' decisions to override computerized drug alerts in primary care. Arch Intern Med. 2003;163:2625–2631.
- , , , et al. Improving acceptance of computerized prescribing alerts in ambulatory care. J Am Med Inform Assoc. 2006;13:5–11.
- , , , . Optimizing the acceptance of medication‐based alerts by physicians during CPOE implementation in a community hospital environment. AMIA Annu Symp Proc. 2007:701–705.
- , , , . The use of a CPOE log for the analysis of physicians' behavior when responding to drug‐duplication reminders. Int J Med Inform. 2008;77:499–506.
- , , , et al. Overrides of medication alerts in ambulatory care. Arch Intern Med. 2009;169:305–311.
- , , , , , . Drug safety alert generation and overriding in a large Dutch university medical centre. Pharmacoepidemiol Drug Saf. 2009;18:941–947.
- , , , et al. Patient‐specific electronic decision support reduces prescription of excessive doses. Qual Saf Health Care. 2010;19:e15.
- , , , et al. Survival from in‐hospital cardiac arrest during nights and weekends. JAMA. 2008;299:785–792.
- , . Managing medications in clinically complex elders: “There's got to be a happy medium.” JAMA. 2010;304:1592–1601.
- Agency for Healthcare Research and Quality. Safety culture. Available at: http://psnet.ahrq.gov/primer.aspx?primerID=5. Accessed October 29, 2013.
- Institute for Safe Medication Practice. List of High‐Alert Medications. Available at: http://www.ismp.org/Tools/highalertmedications.pdf. Accessed June 18, 2013.
- First Databank. FDB AlertSpace. Available at: http://www.fdbhealth.com/solutions/fdb‐alertspace. Accessed July 3, 2014.
- , , , et al. Improving allergy alerting in a computerized physician order entry system. Proc AMIA Symp. 2000:2–6.
- , , , , , . A clinical data warehouse‐based process for refining medication orders alerts. J Am Med Inform Assoc. 2012;19:782–785.
- , , , et al. Drug‐drug interactions that should be non‐interruptive in order to reduce alert fatigue in electronic health records. J Am Med Inform Assoc. 2013;20:489–493.
- , , , et al. High‐priority drug‐drug interactions for use in electronic health records. J Am Med Inform Assoc. 2012;19:735–743.
- , , , , . Design of decision support interventions for medication prescribing. Int J Med Inform. 2013;82:492–503.
- , , , et al. A randomized trial of the effectiveness of on‐demand versus computer‐triggered drug decision support in primary care. J Am Med Inform Assoc. 2008;15:430–438.
- , , , et al. Tiering drug‐drug interaction alerts by severity increases compliance rates. J Am Med Inform Assoc. 2009;16:40–46.
- , , , et al. A review of human factors principles for the design and implementation of medication safety alerts in clinical information systems. J Am Med Inform Assoc. 2010;17:493–501.
- , , , et al. Unintended effects of a computerized physician order entry nearly hard‐stop alert to prevent a drug interaction: a randomized controlled trial. Arch Intern Med. 2010;170:1578–1583.
- , , , , , . Randomized clinical trial of a customized electronic alert requiring an affirmative response compared to a control group receiving a commercial passive CPOE alert: NSAID—warfarin co‐prescribing as a test case. J Am Med Inform Assoc. 2010;17:411–415.
- , , , , . Making electronic prescribing alerts more effective: scenario‐based experimental study in junior doctors. J Am Med Inform Assoc. 2011;18:789–798.
- , , , et al. Development and preliminary evidence for the validity of an instrument assessing implementation of human‐factors principles in medication‐related decision‐support systems—I‐MeDeSA. J Am Med Inform Assoc. 2011;18(suppl 1):i62–i72.
- , , , et al. Medication‐related clinical decision support in computerized provider order entry systems: a review. J Am Med Inform Assoc. 2007;14:29–40.
- , , , et al. Physicians' perceptions on the usefulness of contextual information for prioritizing and presenting alerts in Computerized Physician Order Entry systems. BMC Med Inform Decis Mak. 2012;12:111.
- , , , et al. Computerized clinical decision support systems for drug prescribing and management: a decision‐maker‐researcher partnership systematic review. Implement Sci. 2011;6:89.
- , . A successful model and visual design for creating context‐aware drug‐drug interaction alerts. AMIA Annu Symp Proc. 2011;2011:339–348.
- , , , et al., Effect of computerized physician order entry and a team intervention on prevention of serious medication errors. JAMA. 1998;280:1311–1316.
- , , , , , . Effects of computerized provider order entry on prescribing practices. Arch Intern Med. 2000;160:2741–2747.
- , , , et al. Effects of computerized clinician decision support systems on practitioner performance and patient outcomes: a systematic review. JAMA. 2005;293:1223–1238.
- , , , et al. The effect of computerized physician order entry with clinical decision support on the rates of adverse drug events: a systematic review. J Gen Intern Med. 2008;23:451–458.
- , , . The impact of computerized physician medication order entry in hospitalized patients—a systematic review. Int J Med Inform. 2008;77:365–376.
- , , , et al. What evidence supports the use of computerized alerts and prompts to improve clinicians' prescribing behavior? J Am Med Inform Assoc. 2009;16:531–538.
- , , , , . Does computerized provider order entry reduce prescribing errors for hospital inpatients? A systematic review. J Am Med Inform Assoc. 2009;16:613–623.
- , , , . Overriding of drug safety alerts in computerized physician order entry. J Am Med Inform Assoc. 2006;13:138–147.
- , , , , , . Evaluating clinical decision support systems: monitoring CPOE order check override rates in the Department of Veterans Affairs' Computerized Patient Record System. J Am Med Inform Assoc. 2008;15:620–626.
- , , . GPs' views on computerized drug interaction alerts: questionnaire survey. J Clin Pharm Ther. 2002;27:377–382.
- , , , et al. Clinicians' assessments of electronic medication safety alerts in ambulatory care. Arch Intern Med. 2009;169:1627–1632.
- , , , , . A mixed method study of the merits of e‐prescribing drug alerts in primary care. J Gen Intern Med. 2008;23:442–446.
- . CPOE and clinical decision support in hospitals: getting the benefits: comment on “Unintended effects of a computerized physician order entry nearly hard‐stop alert to prevent a drug interaction.” Arch Intern Med. 2010;170:1583–1584.
- , , . Critical drug‐drug interactions for use in electronic health records systems with computerized physician order entry: review of leading approaches. J Patient Saf. 2011;7:61–65.
- , , , , . Clinical decision support systems could be modified to reduce 'alert fatigue' while still minimizing the risk of litigation. Health Aff (Millwood). 2011;30:2310–2317.
- , , , . Critical issues associated with drug‐drug interactions: highlights of a multistakeholder conference. Am J Health Syst Pharm. 2011;68:941–946.
- , , , , , . Development of a context model to prioritize drug safety alerts in CPOE systems. BMC Med Inform Decis Mak. 2011;11:35.
- , , , et al. Factors influencing alert acceptance: a novel approach for predicting the success of clinical decision support. J Am Med Inform Assoc. 2011;18:479–484.
- , , , . How to improve the delivery of medication alerts within computerized physician order entry systems: an international Delphi study. J Am Med Inform Assoc. 2011;18:760–766.
- , , , , , . Physicians' decisions to override computerized drug alerts in primary care. Arch Intern Med. 2003;163:2625–2631.
- , , , et al. Improving acceptance of computerized prescribing alerts in ambulatory care. J Am Med Inform Assoc. 2006;13:5–11.
- , , , . Optimizing the acceptance of medication‐based alerts by physicians during CPOE implementation in a community hospital environment. AMIA Annu Symp Proc. 2007:701–705.
- , , , . The use of a CPOE log for the analysis of physicians' behavior when responding to drug‐duplication reminders. Int J Med Inform. 2008;77:499–506.
- , , , et al. Overrides of medication alerts in ambulatory care. Arch Intern Med. 2009;169:305–311.
- , , , , , . Drug safety alert generation and overriding in a large Dutch university medical centre. Pharmacoepidemiol Drug Saf. 2009;18:941–947.
- , , , et al. Patient‐specific electronic decision support reduces prescription of excessive doses. Qual Saf Health Care. 2010;19:e15.
- , , , et al. Survival from in‐hospital cardiac arrest during nights and weekends. JAMA. 2008;299:785–792.
- , . Managing medications in clinically complex elders: “There's got to be a happy medium.” JAMA. 2010;304:1592–1601.
- Agency for Healthcare Research and Quality. Safety culture. Available at: http://psnet.ahrq.gov/primer.aspx?primerID=5. Accessed October 29, 2013.
- Institute for Safe Medication Practice. List of High‐Alert Medications. Available at: http://www.ismp.org/Tools/highalertmedications.pdf. Accessed June 18, 2013.
- First Databank. FDB AlertSpace. Available at: http://www.fdbhealth.com/solutions/fdb‐alertspace. Accessed July 3, 2014.
- , , , et al. Improving allergy alerting in a computerized physician order entry system. Proc AMIA Symp. 2000:2–6.
- , , , , , . A clinical data warehouse‐based process for refining medication orders alerts. J Am Med Inform Assoc. 2012;19:782–785.
- , , , et al. Drug‐drug interactions that should be non‐interruptive in order to reduce alert fatigue in electronic health records. J Am Med Inform Assoc. 2013;20:489–493.
- , , , et al. High‐priority drug‐drug interactions for use in electronic health records. J Am Med Inform Assoc. 2012;19:735–743.
- , , , , . Design of decision support interventions for medication prescribing. Int J Med Inform. 2013;82:492–503.
- , , , et al. A randomized trial of the effectiveness of on‐demand versus computer‐triggered drug decision support in primary care. J Am Med Inform Assoc. 2008;15:430–438.
- , , , et al. Tiering drug‐drug interaction alerts by severity increases compliance rates. J Am Med Inform Assoc. 2009;16:40–46.
- , , , et al. A review of human factors principles for the design and implementation of medication safety alerts in clinical information systems. J Am Med Inform Assoc. 2010;17:493–501.
- , , , et al. Unintended effects of a computerized physician order entry nearly hard‐stop alert to prevent a drug interaction: a randomized controlled trial. Arch Intern Med. 2010;170:1578–1583.
- , , , , , . Randomized clinical trial of a customized electronic alert requiring an affirmative response compared to a control group receiving a commercial passive CPOE alert: NSAID—warfarin co‐prescribing as a test case. J Am Med Inform Assoc. 2010;17:411–415.
- , , , , . Making electronic prescribing alerts more effective: scenario‐based experimental study in junior doctors. J Am Med Inform Assoc. 2011;18:789–798.
- , , , et al. Development and preliminary evidence for the validity of an instrument assessing implementation of human‐factors principles in medication‐related decision‐support systems—I‐MeDeSA. J Am Med Inform Assoc. 2011;18(suppl 1):i62–i72.
- , , , et al. Medication‐related clinical decision support in computerized provider order entry systems: a review. J Am Med Inform Assoc. 2007;14:29–40.
- , , , et al. Physicians' perceptions on the usefulness of contextual information for prioritizing and presenting alerts in Computerized Physician Order Entry systems. BMC Med Inform Decis Mak. 2012;12:111.
- , , , et al. Computerized clinical decision support systems for drug prescribing and management: a decision‐maker‐researcher partnership systematic review. Implement Sci. 2011;6:89.
- , . A successful model and visual design for creating context‐aware drug‐drug interaction alerts. AMIA Annu Symp Proc. 2011;2011:339–348.
© 2015 Society of Hospital Medicine
Reducing Inappropriate Acid Suppressives
Prior studies have found that up to 70% of acid‐suppressive medication (ASM) use in the hospital is not indicated, most commonly for stress ulcer prophylaxis in patients outside of the intensive care unit (ICU).[1, 2, 3, 4, 5, 6, 7] Accordingly, reducing inappropriate use of ASM for stress ulcer prophylaxis in hospitalized patients is 1 of the 5 opportunities for improved healthcare value identified by the Society of Hospital Medicine as part of the American Board of Internal Medicine's Choosing Wisely campaign.[8]
We designed and tested a computerized clinical decision support (CDS) intervention with the goal of reducing use of ASM for stress ulcer prophylaxis in hospitalized patients outside the ICU at an academic medical center.
METHODS
Study Design
We conducted a quasiexperimental study using an interrupted time series to analyze data collected prospectively during clinical care before and after implementation of our intervention. The study was deemed a quality improvement initiative by the Beth Israel Deaconess Medical Center Committee on Clinical Investigations/Institutional Review Board.
Patients and Setting
All admissions >18 years of age to a 649‐bed academic medical center in Boston, Massachusetts from September 12, 2011 through July 3, 2012 were included. The medical center consists of an East and West Campus, located across the street from each other. Care for both critically ill and noncritically ill medical and surgical patients occurs on both campuses. Differences include greater proportions of patients with gastrointestinal and oncologic conditions on the East Campus, and renal and cardiac conditions on the West Campus. Additionally, labor and delivery occurs exclusively on the East Campus, and the density of ICU beds is greater on the West Campus. Both campuses utilize a computer‐based provider order entry (POE) system.
Intervention
Our study was implemented in 2 phases (Figure 1).
Baseline Phase
The purpose of the first phase was to obtain baseline data on ASM use prior to implementing our CDS tool designed to influence prescribing. During this baseline phase, a computerized prompt was activated through our POE system whenever a clinician initiated an order for ASM (histamine 2 receptor antagonists or proton pump inhibitors), asking the clinician to select the reason/reasons for the order based on the following predefined response options: (1) active/recent upper gastrointestinal bleed, (2) continuing preadmission medication, (3) Helicobacter pylori treatment, (4) prophylaxis in patient on medications that increase bleeding risk, (5) stress ulcer prophylaxis, (6) suspected/known peptic ulcer disease, gastritis, esophagitis, gastroesophageal reflux disease, and (7) other, with a free‐text box to input the indication. This indications prompt was rolled out to the entire medical center on September 12, 2011 and remained active for the duration of the study period.
Intervention Phase
In the second phase of the study, if a clinician selected stress ulcer prophylaxis as the only indication for ordering ASM, a CDS prompt alerted the clinician that Stress ulcer prophylaxis is not recommended for patients outside of the intensive care unit (ASHP Therapeutic Guidelines on Stress Ulcer Prophylaxis. Am J Health‐Syst Pharm. 1999, 56:347‐79). The clinician could then select either, For use in ICUOrder Medication, Choose Other Indication, or Cancel Order. This CDS prompt was rolled out in a staggered manner to the East Campus on January 3, 2012, followed by the West Campus on April 3, 2012.
Outcomes
The primary outcome was the rate of ASM use with stress ulcer prophylaxis selected as the only indication in a patient located outside of the ICU. We confirmed patient location in the 24 hours after the order was placed. Secondary outcomes were rates of overall ASM use, defined via pharmacy charges, and rates of use on discharge.
Statistical Analysis
To assure stable measurement of trends, we studied at least 3 months before and after the intervention on each campus. We used the Fisher exact test to compare the rates of our primary and secondary outcomes before and after the intervention, stratified by campus. For our primary outcomeat least 1 ASM order with stress ulcer prophylaxis selected as the only indication during hospitalizationwe developed a logistic regression model with a generalized estimating equation and exchangeable working correlation structure to control for admission characteristics (Table 1) and repeated admissions. Using a term for the interaction between time and the intervention, this model allowed us to assess changes in level and trend for the odds of a patient receiving at least 1 ASM order with stress ulcer prophylaxis as the only indication before, compared to after the intervention, stratified by campus. We used a 2‐sided type I error of <0.05 to indicate statistical significance.
| Study Phase | Campus | |||
|---|---|---|---|---|
| East | West | |||
| Baseline, n=3,747 | Intervention, n=6,191 | Baseline, n=11,177 | Intervention, n=5,285 | |
| ||||
| Age, y, mean (SD) | 48.1 (18.5) | 47.7 (18.2) | 61.0 (18.0) | 60.3 (18.1) |
| Gender, no. (%) | ||||
| Female | 2744 (73.2%) | 4542 (73.4%) | 5551 (49.7%) | 2653 (50.2%) |
| Male | 1003 (26.8%) | 1649 (26.6%) | 5626 (50.3%) | 2632 (49.8%) |
| Race, no. (%) | ||||
| Asian | 281 (7.5%) | 516 (8.3%) | 302 (2.7%) | 156 (3%) |
| Black | 424 (11.3%) | 667 (10.8%) | 1426 (12.8%) | 685 (13%) |
| Hispanic | 224 (6%) | 380 (6.1%) | 619 (5.5%) | 282 (5.3%) |
| Other | 378 (10.1%) | 738 (11.9%) | 776 (6.9%) | 396 (7.5%) |
| White | 2440 (65.1%) | 3890 (62.8%) | 8054 (72%) | 3766 (71.3%) |
| Charlson score, mean (SD) | 0.8 (1.1) | 0.7 (1.1) | 1.5 (1.4) | 1.4 (1.4) |
| Gastrointestinal bleeding, no. (%)* | 49 (1.3%) | 99 (1.6%) | 385 (3.4%) | 149 (2.8%) |
| Other medication exposures, no. (%) | ||||
| Therapeutic anticoagulant | 218 (5.8%) | 409 (6.6%) | 2242 (20.1%) | 1022 (19.3%) |
| Prophylactic anticoagulant | 1081 (28.8%) | 1682 (27.2%) | 5999 (53.7%) | 2892 (54.7%) |
| NSAID | 1899 (50.7%) | 3141 (50.7%) | 1248 (11.2%) | 575 (10.9%) |
| Antiplatelet | 313 (8.4%) | 585 (9.4%) | 4543 (40.6%) | 2071 (39.2%) |
| Admitting department, no. (%) | ||||
| Surgery | 2507 (66.9%) | 4146 (67%) | 3255 (29.1%) | 1578 (29.9%) |
| Nonsurgery | 1240 (33.1%) | 2045 (33%) | 7922 (70.9%) | 3707 (70.1%) |
| Any ICU Stay, no. (%) | 217 (5.8%) | 383 (6.2%) | 2786 (24.9%) | 1252 (23.7%) |
RESULTS
There were 26,400 adult admissions during the study period, and 22,330 discrete orders for ASM. Overall, 12,056 (46%) admissions had at least 1 charge for ASM. Admission characteristics were similar before and after the intervention on each campus (Table 1).
Table 2 shows the indications chosen each time ASM was ordered, stratified by campus and study phase. Although selection of stress ulcer prophylaxis decreased on both campuses during the intervention phase, selection of continuing preadmission medication increased.
| Study Phase | Campus | |||
|---|---|---|---|---|
| East | West | |||
| Baseline, n=2,062 | Intervention, n=3,243 | Baseline, n=12,038 | Intervention, n=4,987 | |
| ||||
| Indication* | ||||
| Continuing preadmission medication | 910 (44.1%) | 1695 (52.3%) | 5597 (46.5%) | 2802 (56.2%) |
| PUD, gastritis, esophagitis, GERD | 440 (21.3%) | 797 (24.6%) | 1303 (10.8%) | 582 (11.7%) |
| Stress ulcer prophylaxis | 298 (14.4%) | 100 (3.1%) | 2659 (22.1%) | 681 (13.7%) |
| Prophylaxis in patient on medications that increase bleeding risk | 226 (11.0%) | 259 (8.0%) | 965 (8.0%) | 411 (8.2%) |
| Active/recent gastrointestinal bleed | 154 (7.5%) | 321 (9.9%) | 1450 (12.0%) | 515 (10.3) |
| Helicobacter pylori treatment | 6 (0.2%) | 2 (0.1%) | 43 (0.4%) | 21 (0.4%) |
| Other | 111 (5.4%) | 156 (4.8%) | 384 (3.2%) | 186 (3.7%) |
Table 3 shows the unadjusted comparison of outcomes between baseline and intervention phases on each campus. Use of ASM with stress ulcer prophylaxis as the only indication decreased during the intervention phase on both campuses. There was a nonsignificant reduction in overall rates of use on both campuses, and use on discharge was unchanged. Figure 2 demonstrates the unadjusted and modeled monthly rates of admissions with at least 1 ASM order with stress ulcer prophylaxis selected as the only indication, stratified by campus. After adjusting for the admission characteristics in Table 1, during the intervention phase on both campuses there was a significant immediate reduction in the odds of receiving an ASM with stress ulcer prophylaxis selected as the only indication (East Campus odds ratio [OR]: 0.36, 95% confidence interval [CI]: 0.180.71; West Campus OR: 0.41, 95% CI: 0.280.60), and a significant change in trend compared to the baseline phase (East Campus 1.5% daily decrease in odds of receiving ASM solely for stress ulcer prophylaxis, P=0.002; West Campus 0.9% daily decrease in odds of receiving ASM solely for stress ulcer prophylaxis, P=0.02).
| Study Phase | Campus | |||||
|---|---|---|---|---|---|---|
| East | West | |||||
| Baseline, n=3,747 | Intervention, n=6,191 | P Value* | Baseline, n=11,177 | Intervention, n=5,285 | P Value* | |
| ||||||
| Outcome | ||||||
| Any inappropriate acid‐suppressive exposure | 4.0% | 0.6% | <0.001 | 7.7% | 2.2% | <0.001 |
| Any acid‐suppressive exposure | 33.1% | 31.8% | 0.16 | 54.5% | 52.9% | 0.05 |
| Discharged on acid‐suppressive medication | 18.9% | 19.6% | 0.40 | 34.7% | 34.7% | 0.95 |
DISCUSSION
In this single‐center study, we found that a computerized CDS intervention resulted in a significant reduction in use of ASM for the sole purpose of stress ulcer prophylaxis in patients outside the ICU, a nonsignificant reduction in overall use, and no change in use on discharge. We found low rates of use for the isolated purpose of stress ulcer prophylaxis even before the intervention, and continuing preadmission medication was the most commonly selected indication throughout the study.
Although overall rates of ASM use declined after the intervention, the change was not statistically significant, and was not of the same magnitude as the decline in rates of use for the purpose of stress ulcer prophylaxis. This suggests that our intervention, in part, led to substitution of 1 indication for another. The indication that increased the most after rollout on both campuses was continuing preadmission medication. There are at least 2 possibilities for this finding: (1) the intervention prompted physicians to more accurately record the indication, or (2) physicians falsified the indication in order to execute the order. To explore these possibilities, we reviewed the charts of a random sample of 100 admissions during each of the baseline and intervention phases where continuing preadmission medication was selected as an indication for an ASM order. We found that 6/100 orders in the baseline phase and 7/100 orders in the intervention phase incorrectly indicated that the patient was on ASM prior to admission (P=0.77). This suggests that scenario 1 above is the more likely explanation for the increased use of this indication, and that the intervention, in part, simply unmasked the true rate of use at our medical center for the isolated purpose of stress ulcer prophylaxis.
These findings have implications for others attempting to use computerized CDS to better understand physician prescribing. They suggest that information collected through computer‐based interaction with clinicians at the point of care may not always be accurate or complete. As institutions increasingly use similar interventions to drive behavior, information obtained from such interaction should be validated, and when possible, patient outcomes should be measured.
Our findings suggest that rates of ASM use for the purpose of stress ulcer prophylaxis in the hospital may have declined over the last decade. Studies demonstrating that up to 70% of inpatient use of ASM was inappropriate were conducted 5 to 10 years ago.[1, 2, 3, 4, 5] Since then, studies have demonstrated risk of nosocomial infections in patients on ASM.[9, 10, 11] It is possible that the low rate of use for stress ulcer prophylaxis in our study is attributable to awareness of the risks of these medications, and limited our ability to detect differences in overall use. It is also possible, however, that a portion of the admissions with continuation of preadmission medication as the indication were started on these medications during a prior hospitalization. Thus, some portion of preadmission use is likely to represent failed medication reconciliation during a prior discharge. In this context, hospitalization may serve as an opportunity to evaluate the indication for ASM use even when these medications show up as preadmission medications.
There are additional limitations. First, the single‐center nature limits generalizability. Second, the first phase of our study, designed to obtain baseline data on ASM use, may have led to changes in prescribing prior to implementation of our CDS tool. Additionally, we did not validate the accuracy of each of the chosen indications, or the site of initial prescription in the case of preadmission exposure. Last, our study was not powered to investigate changes in rates of nosocomial gastrointestinal bleeding or nosocomial pneumonia owing to the infrequent nature of these complications.
In conclusion, we designed a simple computerized CDS intervention that was associated with a reduction in ASM use for stress ulcer prophylaxis in patients outside the ICU, a nonsignificant reduction in overall use, and no change in use on discharge. The majority of inpatient use represented continuation of preadmission medication, suggesting that interventions to improve the appropriateness of ASM prescribing should span the continuum of care. Future studies should investigate whether it is worthwhile and appropriate to reevaluate continued use of preadmission ASM during an inpatient stay.
Acknowledgements
The authors acknowledge Joshua Guthermann, MBA, and Jane Hui Chen Lim, MBA, for their assistance in the early phases of data analysis, and Long H. Ngo, PhD, for his statistical consultation.
Disclosures: Dr. Herzig was funded by a Young Clinician Research Award from the Center for Integration of Medicine and Innovative Technology, a nonprofit consortium of Boston teaching hospitals and universities, and grant number K23AG042459 from the National Institute on Aging. Dr. Marcantonio was funded by grant number K24AG035075 from the National Institute on Aging. The funding organizations had no involvement in any aspect of the study, including design, conduct, and reporting of the study. Dr. Herzig had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs. Herzig and Marcantonio were responsible for the study concept and design. Drs. Herzig, Feinbloom, Howell, and Ms. Adra and Mr. Afonso were responsible for the acquisition of data. Drs. Herzig, Howell, Marcantonio, and Mr. Guess were responsible for the analysis and interpretation of the data. Dr. Herzig drafted the manuscript. All of the authors participated in the critical revision of the manuscript for important intellectual content. Drs. Herzig and Marcantonio were responsible for study supervision. The authors report no conflicts of interest.
- , . Stress ulcer prophylaxis in hospitalized patients not in intensive care units. Am J Health Syst Pharm. 2007;64(13):1396–1400.
- , . Magnitude and economic impact of inappropriate use of stress ulcer prophylaxis in non‐ICU hospitalized patients. Am J Gastroenterol. 2006;101(10):2200–2205.
- , . Stress‐ulcer prophylaxis for general medical patients: a review of the evidence. J Hosp Med. 2007;2(2):86–92.
- , , , et al. Hospital use of acid‐suppressive medications and its fall‐out on prescribing in general practice: a 1‐month survey. Aliment Pharmacol Ther. 2003;17(12):1503–1506.
- , , , , , . Inadequate use of acid‐suppressive therapy in hospitalized patients and its implications for general practice. Dig Dis Sci. 2005;50(12):2307–2311.
- , . Brief report: reducing inappropriate usage of stress ulcer prophylaxis among internal medicine residents. A practice‐based educational intervention. J Gen Intern Med. 2006;21(5):498–500.
- , , . Inappropriate continuation of stress ulcer prophylactic therapy after discharge. Ann Pharmacother. 2007;41(10):1611–1616.
- , , , et al. Choosing wisely in adult hospital medicine: five opportunities for improved healthcare value. J Hosp Med. 2013;8(9):486–492.
- , , , , . Risk of Clostridium difficile diarrhea among hospital inpatients prescribed proton pump inhibitors: cohort and case‐control studies. CMAJ. 2004;171(1):33–38.
- , , , et al. Iatrogenic gastric acid suppression and the risk of nosocomial Clostridium difficile infection. Arch Intern Med. 2010;170(9):784–790.
- , , , . Acid‐suppressive medication use and the risk for hospital‐acquired pneumonia. JAMA. 2009;301(20):2120–2128.
- Healthcare Cost and Utilization Project. Clinical classifications software (CCS) for ICD‐9‐CM. December 2009. Agency for Healthcare Research and Quality, Rockville, MD. Available at: www.hcup‐us.ahrq.gov/toolssoftware/ccs/ccs.jsp. Accessed June 18, 2014.
Prior studies have found that up to 70% of acid‐suppressive medication (ASM) use in the hospital is not indicated, most commonly for stress ulcer prophylaxis in patients outside of the intensive care unit (ICU).[1, 2, 3, 4, 5, 6, 7] Accordingly, reducing inappropriate use of ASM for stress ulcer prophylaxis in hospitalized patients is 1 of the 5 opportunities for improved healthcare value identified by the Society of Hospital Medicine as part of the American Board of Internal Medicine's Choosing Wisely campaign.[8]
We designed and tested a computerized clinical decision support (CDS) intervention with the goal of reducing use of ASM for stress ulcer prophylaxis in hospitalized patients outside the ICU at an academic medical center.
METHODS
Study Design
We conducted a quasiexperimental study using an interrupted time series to analyze data collected prospectively during clinical care before and after implementation of our intervention. The study was deemed a quality improvement initiative by the Beth Israel Deaconess Medical Center Committee on Clinical Investigations/Institutional Review Board.
Patients and Setting
All admissions >18 years of age to a 649‐bed academic medical center in Boston, Massachusetts from September 12, 2011 through July 3, 2012 were included. The medical center consists of an East and West Campus, located across the street from each other. Care for both critically ill and noncritically ill medical and surgical patients occurs on both campuses. Differences include greater proportions of patients with gastrointestinal and oncologic conditions on the East Campus, and renal and cardiac conditions on the West Campus. Additionally, labor and delivery occurs exclusively on the East Campus, and the density of ICU beds is greater on the West Campus. Both campuses utilize a computer‐based provider order entry (POE) system.
Intervention
Our study was implemented in 2 phases (Figure 1).
Baseline Phase
The purpose of the first phase was to obtain baseline data on ASM use prior to implementing our CDS tool designed to influence prescribing. During this baseline phase, a computerized prompt was activated through our POE system whenever a clinician initiated an order for ASM (histamine 2 receptor antagonists or proton pump inhibitors), asking the clinician to select the reason/reasons for the order based on the following predefined response options: (1) active/recent upper gastrointestinal bleed, (2) continuing preadmission medication, (3) Helicobacter pylori treatment, (4) prophylaxis in patient on medications that increase bleeding risk, (5) stress ulcer prophylaxis, (6) suspected/known peptic ulcer disease, gastritis, esophagitis, gastroesophageal reflux disease, and (7) other, with a free‐text box to input the indication. This indications prompt was rolled out to the entire medical center on September 12, 2011 and remained active for the duration of the study period.
Intervention Phase
In the second phase of the study, if a clinician selected stress ulcer prophylaxis as the only indication for ordering ASM, a CDS prompt alerted the clinician that Stress ulcer prophylaxis is not recommended for patients outside of the intensive care unit (ASHP Therapeutic Guidelines on Stress Ulcer Prophylaxis. Am J Health‐Syst Pharm. 1999, 56:347‐79). The clinician could then select either, For use in ICUOrder Medication, Choose Other Indication, or Cancel Order. This CDS prompt was rolled out in a staggered manner to the East Campus on January 3, 2012, followed by the West Campus on April 3, 2012.
Outcomes
The primary outcome was the rate of ASM use with stress ulcer prophylaxis selected as the only indication in a patient located outside of the ICU. We confirmed patient location in the 24 hours after the order was placed. Secondary outcomes were rates of overall ASM use, defined via pharmacy charges, and rates of use on discharge.
Statistical Analysis
To assure stable measurement of trends, we studied at least 3 months before and after the intervention on each campus. We used the Fisher exact test to compare the rates of our primary and secondary outcomes before and after the intervention, stratified by campus. For our primary outcomeat least 1 ASM order with stress ulcer prophylaxis selected as the only indication during hospitalizationwe developed a logistic regression model with a generalized estimating equation and exchangeable working correlation structure to control for admission characteristics (Table 1) and repeated admissions. Using a term for the interaction between time and the intervention, this model allowed us to assess changes in level and trend for the odds of a patient receiving at least 1 ASM order with stress ulcer prophylaxis as the only indication before, compared to after the intervention, stratified by campus. We used a 2‐sided type I error of <0.05 to indicate statistical significance.
| Study Phase | Campus | |||
|---|---|---|---|---|
| East | West | |||
| Baseline, n=3,747 | Intervention, n=6,191 | Baseline, n=11,177 | Intervention, n=5,285 | |
| ||||
| Age, y, mean (SD) | 48.1 (18.5) | 47.7 (18.2) | 61.0 (18.0) | 60.3 (18.1) |
| Gender, no. (%) | ||||
| Female | 2744 (73.2%) | 4542 (73.4%) | 5551 (49.7%) | 2653 (50.2%) |
| Male | 1003 (26.8%) | 1649 (26.6%) | 5626 (50.3%) | 2632 (49.8%) |
| Race, no. (%) | ||||
| Asian | 281 (7.5%) | 516 (8.3%) | 302 (2.7%) | 156 (3%) |
| Black | 424 (11.3%) | 667 (10.8%) | 1426 (12.8%) | 685 (13%) |
| Hispanic | 224 (6%) | 380 (6.1%) | 619 (5.5%) | 282 (5.3%) |
| Other | 378 (10.1%) | 738 (11.9%) | 776 (6.9%) | 396 (7.5%) |
| White | 2440 (65.1%) | 3890 (62.8%) | 8054 (72%) | 3766 (71.3%) |
| Charlson score, mean (SD) | 0.8 (1.1) | 0.7 (1.1) | 1.5 (1.4) | 1.4 (1.4) |
| Gastrointestinal bleeding, no. (%)* | 49 (1.3%) | 99 (1.6%) | 385 (3.4%) | 149 (2.8%) |
| Other medication exposures, no. (%) | ||||
| Therapeutic anticoagulant | 218 (5.8%) | 409 (6.6%) | 2242 (20.1%) | 1022 (19.3%) |
| Prophylactic anticoagulant | 1081 (28.8%) | 1682 (27.2%) | 5999 (53.7%) | 2892 (54.7%) |
| NSAID | 1899 (50.7%) | 3141 (50.7%) | 1248 (11.2%) | 575 (10.9%) |
| Antiplatelet | 313 (8.4%) | 585 (9.4%) | 4543 (40.6%) | 2071 (39.2%) |
| Admitting department, no. (%) | ||||
| Surgery | 2507 (66.9%) | 4146 (67%) | 3255 (29.1%) | 1578 (29.9%) |
| Nonsurgery | 1240 (33.1%) | 2045 (33%) | 7922 (70.9%) | 3707 (70.1%) |
| Any ICU Stay, no. (%) | 217 (5.8%) | 383 (6.2%) | 2786 (24.9%) | 1252 (23.7%) |
RESULTS
There were 26,400 adult admissions during the study period, and 22,330 discrete orders for ASM. Overall, 12,056 (46%) admissions had at least 1 charge for ASM. Admission characteristics were similar before and after the intervention on each campus (Table 1).
Table 2 shows the indications chosen each time ASM was ordered, stratified by campus and study phase. Although selection of stress ulcer prophylaxis decreased on both campuses during the intervention phase, selection of continuing preadmission medication increased.
| Study Phase | Campus | |||
|---|---|---|---|---|
| East | West | |||
| Baseline, n=2,062 | Intervention, n=3,243 | Baseline, n=12,038 | Intervention, n=4,987 | |
| ||||
| Indication* | ||||
| Continuing preadmission medication | 910 (44.1%) | 1695 (52.3%) | 5597 (46.5%) | 2802 (56.2%) |
| PUD, gastritis, esophagitis, GERD | 440 (21.3%) | 797 (24.6%) | 1303 (10.8%) | 582 (11.7%) |
| Stress ulcer prophylaxis | 298 (14.4%) | 100 (3.1%) | 2659 (22.1%) | 681 (13.7%) |
| Prophylaxis in patient on medications that increase bleeding risk | 226 (11.0%) | 259 (8.0%) | 965 (8.0%) | 411 (8.2%) |
| Active/recent gastrointestinal bleed | 154 (7.5%) | 321 (9.9%) | 1450 (12.0%) | 515 (10.3) |
| Helicobacter pylori treatment | 6 (0.2%) | 2 (0.1%) | 43 (0.4%) | 21 (0.4%) |
| Other | 111 (5.4%) | 156 (4.8%) | 384 (3.2%) | 186 (3.7%) |
Table 3 shows the unadjusted comparison of outcomes between baseline and intervention phases on each campus. Use of ASM with stress ulcer prophylaxis as the only indication decreased during the intervention phase on both campuses. There was a nonsignificant reduction in overall rates of use on both campuses, and use on discharge was unchanged. Figure 2 demonstrates the unadjusted and modeled monthly rates of admissions with at least 1 ASM order with stress ulcer prophylaxis selected as the only indication, stratified by campus. After adjusting for the admission characteristics in Table 1, during the intervention phase on both campuses there was a significant immediate reduction in the odds of receiving an ASM with stress ulcer prophylaxis selected as the only indication (East Campus odds ratio [OR]: 0.36, 95% confidence interval [CI]: 0.180.71; West Campus OR: 0.41, 95% CI: 0.280.60), and a significant change in trend compared to the baseline phase (East Campus 1.5% daily decrease in odds of receiving ASM solely for stress ulcer prophylaxis, P=0.002; West Campus 0.9% daily decrease in odds of receiving ASM solely for stress ulcer prophylaxis, P=0.02).
| Study Phase | Campus | |||||
|---|---|---|---|---|---|---|
| East | West | |||||
| Baseline, n=3,747 | Intervention, n=6,191 | P Value* | Baseline, n=11,177 | Intervention, n=5,285 | P Value* | |
| ||||||
| Outcome | ||||||
| Any inappropriate acid‐suppressive exposure | 4.0% | 0.6% | <0.001 | 7.7% | 2.2% | <0.001 |
| Any acid‐suppressive exposure | 33.1% | 31.8% | 0.16 | 54.5% | 52.9% | 0.05 |
| Discharged on acid‐suppressive medication | 18.9% | 19.6% | 0.40 | 34.7% | 34.7% | 0.95 |
DISCUSSION
In this single‐center study, we found that a computerized CDS intervention resulted in a significant reduction in use of ASM for the sole purpose of stress ulcer prophylaxis in patients outside the ICU, a nonsignificant reduction in overall use, and no change in use on discharge. We found low rates of use for the isolated purpose of stress ulcer prophylaxis even before the intervention, and continuing preadmission medication was the most commonly selected indication throughout the study.
Although overall rates of ASM use declined after the intervention, the change was not statistically significant, and was not of the same magnitude as the decline in rates of use for the purpose of stress ulcer prophylaxis. This suggests that our intervention, in part, led to substitution of 1 indication for another. The indication that increased the most after rollout on both campuses was continuing preadmission medication. There are at least 2 possibilities for this finding: (1) the intervention prompted physicians to more accurately record the indication, or (2) physicians falsified the indication in order to execute the order. To explore these possibilities, we reviewed the charts of a random sample of 100 admissions during each of the baseline and intervention phases where continuing preadmission medication was selected as an indication for an ASM order. We found that 6/100 orders in the baseline phase and 7/100 orders in the intervention phase incorrectly indicated that the patient was on ASM prior to admission (P=0.77). This suggests that scenario 1 above is the more likely explanation for the increased use of this indication, and that the intervention, in part, simply unmasked the true rate of use at our medical center for the isolated purpose of stress ulcer prophylaxis.
These findings have implications for others attempting to use computerized CDS to better understand physician prescribing. They suggest that information collected through computer‐based interaction with clinicians at the point of care may not always be accurate or complete. As institutions increasingly use similar interventions to drive behavior, information obtained from such interaction should be validated, and when possible, patient outcomes should be measured.
Our findings suggest that rates of ASM use for the purpose of stress ulcer prophylaxis in the hospital may have declined over the last decade. Studies demonstrating that up to 70% of inpatient use of ASM was inappropriate were conducted 5 to 10 years ago.[1, 2, 3, 4, 5] Since then, studies have demonstrated risk of nosocomial infections in patients on ASM.[9, 10, 11] It is possible that the low rate of use for stress ulcer prophylaxis in our study is attributable to awareness of the risks of these medications, and limited our ability to detect differences in overall use. It is also possible, however, that a portion of the admissions with continuation of preadmission medication as the indication were started on these medications during a prior hospitalization. Thus, some portion of preadmission use is likely to represent failed medication reconciliation during a prior discharge. In this context, hospitalization may serve as an opportunity to evaluate the indication for ASM use even when these medications show up as preadmission medications.
There are additional limitations. First, the single‐center nature limits generalizability. Second, the first phase of our study, designed to obtain baseline data on ASM use, may have led to changes in prescribing prior to implementation of our CDS tool. Additionally, we did not validate the accuracy of each of the chosen indications, or the site of initial prescription in the case of preadmission exposure. Last, our study was not powered to investigate changes in rates of nosocomial gastrointestinal bleeding or nosocomial pneumonia owing to the infrequent nature of these complications.
In conclusion, we designed a simple computerized CDS intervention that was associated with a reduction in ASM use for stress ulcer prophylaxis in patients outside the ICU, a nonsignificant reduction in overall use, and no change in use on discharge. The majority of inpatient use represented continuation of preadmission medication, suggesting that interventions to improve the appropriateness of ASM prescribing should span the continuum of care. Future studies should investigate whether it is worthwhile and appropriate to reevaluate continued use of preadmission ASM during an inpatient stay.
Acknowledgements
The authors acknowledge Joshua Guthermann, MBA, and Jane Hui Chen Lim, MBA, for their assistance in the early phases of data analysis, and Long H. Ngo, PhD, for his statistical consultation.
Disclosures: Dr. Herzig was funded by a Young Clinician Research Award from the Center for Integration of Medicine and Innovative Technology, a nonprofit consortium of Boston teaching hospitals and universities, and grant number K23AG042459 from the National Institute on Aging. Dr. Marcantonio was funded by grant number K24AG035075 from the National Institute on Aging. The funding organizations had no involvement in any aspect of the study, including design, conduct, and reporting of the study. Dr. Herzig had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs. Herzig and Marcantonio were responsible for the study concept and design. Drs. Herzig, Feinbloom, Howell, and Ms. Adra and Mr. Afonso were responsible for the acquisition of data. Drs. Herzig, Howell, Marcantonio, and Mr. Guess were responsible for the analysis and interpretation of the data. Dr. Herzig drafted the manuscript. All of the authors participated in the critical revision of the manuscript for important intellectual content. Drs. Herzig and Marcantonio were responsible for study supervision. The authors report no conflicts of interest.
Prior studies have found that up to 70% of acid‐suppressive medication (ASM) use in the hospital is not indicated, most commonly for stress ulcer prophylaxis in patients outside of the intensive care unit (ICU).[1, 2, 3, 4, 5, 6, 7] Accordingly, reducing inappropriate use of ASM for stress ulcer prophylaxis in hospitalized patients is 1 of the 5 opportunities for improved healthcare value identified by the Society of Hospital Medicine as part of the American Board of Internal Medicine's Choosing Wisely campaign.[8]
We designed and tested a computerized clinical decision support (CDS) intervention with the goal of reducing use of ASM for stress ulcer prophylaxis in hospitalized patients outside the ICU at an academic medical center.
METHODS
Study Design
We conducted a quasiexperimental study using an interrupted time series to analyze data collected prospectively during clinical care before and after implementation of our intervention. The study was deemed a quality improvement initiative by the Beth Israel Deaconess Medical Center Committee on Clinical Investigations/Institutional Review Board.
Patients and Setting
All admissions >18 years of age to a 649‐bed academic medical center in Boston, Massachusetts from September 12, 2011 through July 3, 2012 were included. The medical center consists of an East and West Campus, located across the street from each other. Care for both critically ill and noncritically ill medical and surgical patients occurs on both campuses. Differences include greater proportions of patients with gastrointestinal and oncologic conditions on the East Campus, and renal and cardiac conditions on the West Campus. Additionally, labor and delivery occurs exclusively on the East Campus, and the density of ICU beds is greater on the West Campus. Both campuses utilize a computer‐based provider order entry (POE) system.
Intervention
Our study was implemented in 2 phases (Figure 1).
Baseline Phase
The purpose of the first phase was to obtain baseline data on ASM use prior to implementing our CDS tool designed to influence prescribing. During this baseline phase, a computerized prompt was activated through our POE system whenever a clinician initiated an order for ASM (histamine 2 receptor antagonists or proton pump inhibitors), asking the clinician to select the reason/reasons for the order based on the following predefined response options: (1) active/recent upper gastrointestinal bleed, (2) continuing preadmission medication, (3) Helicobacter pylori treatment, (4) prophylaxis in patient on medications that increase bleeding risk, (5) stress ulcer prophylaxis, (6) suspected/known peptic ulcer disease, gastritis, esophagitis, gastroesophageal reflux disease, and (7) other, with a free‐text box to input the indication. This indications prompt was rolled out to the entire medical center on September 12, 2011 and remained active for the duration of the study period.
Intervention Phase
In the second phase of the study, if a clinician selected stress ulcer prophylaxis as the only indication for ordering ASM, a CDS prompt alerted the clinician that Stress ulcer prophylaxis is not recommended for patients outside of the intensive care unit (ASHP Therapeutic Guidelines on Stress Ulcer Prophylaxis. Am J Health‐Syst Pharm. 1999, 56:347‐79). The clinician could then select either, For use in ICUOrder Medication, Choose Other Indication, or Cancel Order. This CDS prompt was rolled out in a staggered manner to the East Campus on January 3, 2012, followed by the West Campus on April 3, 2012.
Outcomes
The primary outcome was the rate of ASM use with stress ulcer prophylaxis selected as the only indication in a patient located outside of the ICU. We confirmed patient location in the 24 hours after the order was placed. Secondary outcomes were rates of overall ASM use, defined via pharmacy charges, and rates of use on discharge.
Statistical Analysis
To assure stable measurement of trends, we studied at least 3 months before and after the intervention on each campus. We used the Fisher exact test to compare the rates of our primary and secondary outcomes before and after the intervention, stratified by campus. For our primary outcomeat least 1 ASM order with stress ulcer prophylaxis selected as the only indication during hospitalizationwe developed a logistic regression model with a generalized estimating equation and exchangeable working correlation structure to control for admission characteristics (Table 1) and repeated admissions. Using a term for the interaction between time and the intervention, this model allowed us to assess changes in level and trend for the odds of a patient receiving at least 1 ASM order with stress ulcer prophylaxis as the only indication before, compared to after the intervention, stratified by campus. We used a 2‐sided type I error of <0.05 to indicate statistical significance.
| Study Phase | Campus | |||
|---|---|---|---|---|
| East | West | |||
| Baseline, n=3,747 | Intervention, n=6,191 | Baseline, n=11,177 | Intervention, n=5,285 | |
| ||||
| Age, y, mean (SD) | 48.1 (18.5) | 47.7 (18.2) | 61.0 (18.0) | 60.3 (18.1) |
| Gender, no. (%) | ||||
| Female | 2744 (73.2%) | 4542 (73.4%) | 5551 (49.7%) | 2653 (50.2%) |
| Male | 1003 (26.8%) | 1649 (26.6%) | 5626 (50.3%) | 2632 (49.8%) |
| Race, no. (%) | ||||
| Asian | 281 (7.5%) | 516 (8.3%) | 302 (2.7%) | 156 (3%) |
| Black | 424 (11.3%) | 667 (10.8%) | 1426 (12.8%) | 685 (13%) |
| Hispanic | 224 (6%) | 380 (6.1%) | 619 (5.5%) | 282 (5.3%) |
| Other | 378 (10.1%) | 738 (11.9%) | 776 (6.9%) | 396 (7.5%) |
| White | 2440 (65.1%) | 3890 (62.8%) | 8054 (72%) | 3766 (71.3%) |
| Charlson score, mean (SD) | 0.8 (1.1) | 0.7 (1.1) | 1.5 (1.4) | 1.4 (1.4) |
| Gastrointestinal bleeding, no. (%)* | 49 (1.3%) | 99 (1.6%) | 385 (3.4%) | 149 (2.8%) |
| Other medication exposures, no. (%) | ||||
| Therapeutic anticoagulant | 218 (5.8%) | 409 (6.6%) | 2242 (20.1%) | 1022 (19.3%) |
| Prophylactic anticoagulant | 1081 (28.8%) | 1682 (27.2%) | 5999 (53.7%) | 2892 (54.7%) |
| NSAID | 1899 (50.7%) | 3141 (50.7%) | 1248 (11.2%) | 575 (10.9%) |
| Antiplatelet | 313 (8.4%) | 585 (9.4%) | 4543 (40.6%) | 2071 (39.2%) |
| Admitting department, no. (%) | ||||
| Surgery | 2507 (66.9%) | 4146 (67%) | 3255 (29.1%) | 1578 (29.9%) |
| Nonsurgery | 1240 (33.1%) | 2045 (33%) | 7922 (70.9%) | 3707 (70.1%) |
| Any ICU Stay, no. (%) | 217 (5.8%) | 383 (6.2%) | 2786 (24.9%) | 1252 (23.7%) |
RESULTS
There were 26,400 adult admissions during the study period, and 22,330 discrete orders for ASM. Overall, 12,056 (46%) admissions had at least 1 charge for ASM. Admission characteristics were similar before and after the intervention on each campus (Table 1).
Table 2 shows the indications chosen each time ASM was ordered, stratified by campus and study phase. Although selection of stress ulcer prophylaxis decreased on both campuses during the intervention phase, selection of continuing preadmission medication increased.
| Study Phase | Campus | |||
|---|---|---|---|---|
| East | West | |||
| Baseline, n=2,062 | Intervention, n=3,243 | Baseline, n=12,038 | Intervention, n=4,987 | |
| ||||
| Indication* | ||||
| Continuing preadmission medication | 910 (44.1%) | 1695 (52.3%) | 5597 (46.5%) | 2802 (56.2%) |
| PUD, gastritis, esophagitis, GERD | 440 (21.3%) | 797 (24.6%) | 1303 (10.8%) | 582 (11.7%) |
| Stress ulcer prophylaxis | 298 (14.4%) | 100 (3.1%) | 2659 (22.1%) | 681 (13.7%) |
| Prophylaxis in patient on medications that increase bleeding risk | 226 (11.0%) | 259 (8.0%) | 965 (8.0%) | 411 (8.2%) |
| Active/recent gastrointestinal bleed | 154 (7.5%) | 321 (9.9%) | 1450 (12.0%) | 515 (10.3) |
| Helicobacter pylori treatment | 6 (0.2%) | 2 (0.1%) | 43 (0.4%) | 21 (0.4%) |
| Other | 111 (5.4%) | 156 (4.8%) | 384 (3.2%) | 186 (3.7%) |
Table 3 shows the unadjusted comparison of outcomes between baseline and intervention phases on each campus. Use of ASM with stress ulcer prophylaxis as the only indication decreased during the intervention phase on both campuses. There was a nonsignificant reduction in overall rates of use on both campuses, and use on discharge was unchanged. Figure 2 demonstrates the unadjusted and modeled monthly rates of admissions with at least 1 ASM order with stress ulcer prophylaxis selected as the only indication, stratified by campus. After adjusting for the admission characteristics in Table 1, during the intervention phase on both campuses there was a significant immediate reduction in the odds of receiving an ASM with stress ulcer prophylaxis selected as the only indication (East Campus odds ratio [OR]: 0.36, 95% confidence interval [CI]: 0.180.71; West Campus OR: 0.41, 95% CI: 0.280.60), and a significant change in trend compared to the baseline phase (East Campus 1.5% daily decrease in odds of receiving ASM solely for stress ulcer prophylaxis, P=0.002; West Campus 0.9% daily decrease in odds of receiving ASM solely for stress ulcer prophylaxis, P=0.02).
| Study Phase | Campus | |||||
|---|---|---|---|---|---|---|
| East | West | |||||
| Baseline, n=3,747 | Intervention, n=6,191 | P Value* | Baseline, n=11,177 | Intervention, n=5,285 | P Value* | |
| ||||||
| Outcome | ||||||
| Any inappropriate acid‐suppressive exposure | 4.0% | 0.6% | <0.001 | 7.7% | 2.2% | <0.001 |
| Any acid‐suppressive exposure | 33.1% | 31.8% | 0.16 | 54.5% | 52.9% | 0.05 |
| Discharged on acid‐suppressive medication | 18.9% | 19.6% | 0.40 | 34.7% | 34.7% | 0.95 |
DISCUSSION
In this single‐center study, we found that a computerized CDS intervention resulted in a significant reduction in use of ASM for the sole purpose of stress ulcer prophylaxis in patients outside the ICU, a nonsignificant reduction in overall use, and no change in use on discharge. We found low rates of use for the isolated purpose of stress ulcer prophylaxis even before the intervention, and continuing preadmission medication was the most commonly selected indication throughout the study.
Although overall rates of ASM use declined after the intervention, the change was not statistically significant, and was not of the same magnitude as the decline in rates of use for the purpose of stress ulcer prophylaxis. This suggests that our intervention, in part, led to substitution of 1 indication for another. The indication that increased the most after rollout on both campuses was continuing preadmission medication. There are at least 2 possibilities for this finding: (1) the intervention prompted physicians to more accurately record the indication, or (2) physicians falsified the indication in order to execute the order. To explore these possibilities, we reviewed the charts of a random sample of 100 admissions during each of the baseline and intervention phases where continuing preadmission medication was selected as an indication for an ASM order. We found that 6/100 orders in the baseline phase and 7/100 orders in the intervention phase incorrectly indicated that the patient was on ASM prior to admission (P=0.77). This suggests that scenario 1 above is the more likely explanation for the increased use of this indication, and that the intervention, in part, simply unmasked the true rate of use at our medical center for the isolated purpose of stress ulcer prophylaxis.
These findings have implications for others attempting to use computerized CDS to better understand physician prescribing. They suggest that information collected through computer‐based interaction with clinicians at the point of care may not always be accurate or complete. As institutions increasingly use similar interventions to drive behavior, information obtained from such interaction should be validated, and when possible, patient outcomes should be measured.
Our findings suggest that rates of ASM use for the purpose of stress ulcer prophylaxis in the hospital may have declined over the last decade. Studies demonstrating that up to 70% of inpatient use of ASM was inappropriate were conducted 5 to 10 years ago.[1, 2, 3, 4, 5] Since then, studies have demonstrated risk of nosocomial infections in patients on ASM.[9, 10, 11] It is possible that the low rate of use for stress ulcer prophylaxis in our study is attributable to awareness of the risks of these medications, and limited our ability to detect differences in overall use. It is also possible, however, that a portion of the admissions with continuation of preadmission medication as the indication were started on these medications during a prior hospitalization. Thus, some portion of preadmission use is likely to represent failed medication reconciliation during a prior discharge. In this context, hospitalization may serve as an opportunity to evaluate the indication for ASM use even when these medications show up as preadmission medications.
There are additional limitations. First, the single‐center nature limits generalizability. Second, the first phase of our study, designed to obtain baseline data on ASM use, may have led to changes in prescribing prior to implementation of our CDS tool. Additionally, we did not validate the accuracy of each of the chosen indications, or the site of initial prescription in the case of preadmission exposure. Last, our study was not powered to investigate changes in rates of nosocomial gastrointestinal bleeding or nosocomial pneumonia owing to the infrequent nature of these complications.
In conclusion, we designed a simple computerized CDS intervention that was associated with a reduction in ASM use for stress ulcer prophylaxis in patients outside the ICU, a nonsignificant reduction in overall use, and no change in use on discharge. The majority of inpatient use represented continuation of preadmission medication, suggesting that interventions to improve the appropriateness of ASM prescribing should span the continuum of care. Future studies should investigate whether it is worthwhile and appropriate to reevaluate continued use of preadmission ASM during an inpatient stay.
Acknowledgements
The authors acknowledge Joshua Guthermann, MBA, and Jane Hui Chen Lim, MBA, for their assistance in the early phases of data analysis, and Long H. Ngo, PhD, for his statistical consultation.
Disclosures: Dr. Herzig was funded by a Young Clinician Research Award from the Center for Integration of Medicine and Innovative Technology, a nonprofit consortium of Boston teaching hospitals and universities, and grant number K23AG042459 from the National Institute on Aging. Dr. Marcantonio was funded by grant number K24AG035075 from the National Institute on Aging. The funding organizations had no involvement in any aspect of the study, including design, conduct, and reporting of the study. Dr. Herzig had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs. Herzig and Marcantonio were responsible for the study concept and design. Drs. Herzig, Feinbloom, Howell, and Ms. Adra and Mr. Afonso were responsible for the acquisition of data. Drs. Herzig, Howell, Marcantonio, and Mr. Guess were responsible for the analysis and interpretation of the data. Dr. Herzig drafted the manuscript. All of the authors participated in the critical revision of the manuscript for important intellectual content. Drs. Herzig and Marcantonio were responsible for study supervision. The authors report no conflicts of interest.
- , . Stress ulcer prophylaxis in hospitalized patients not in intensive care units. Am J Health Syst Pharm. 2007;64(13):1396–1400.
- , . Magnitude and economic impact of inappropriate use of stress ulcer prophylaxis in non‐ICU hospitalized patients. Am J Gastroenterol. 2006;101(10):2200–2205.
- , . Stress‐ulcer prophylaxis for general medical patients: a review of the evidence. J Hosp Med. 2007;2(2):86–92.
- , , , et al. Hospital use of acid‐suppressive medications and its fall‐out on prescribing in general practice: a 1‐month survey. Aliment Pharmacol Ther. 2003;17(12):1503–1506.
- , , , , , . Inadequate use of acid‐suppressive therapy in hospitalized patients and its implications for general practice. Dig Dis Sci. 2005;50(12):2307–2311.
- , . Brief report: reducing inappropriate usage of stress ulcer prophylaxis among internal medicine residents. A practice‐based educational intervention. J Gen Intern Med. 2006;21(5):498–500.
- , , . Inappropriate continuation of stress ulcer prophylactic therapy after discharge. Ann Pharmacother. 2007;41(10):1611–1616.
- , , , et al. Choosing wisely in adult hospital medicine: five opportunities for improved healthcare value. J Hosp Med. 2013;8(9):486–492.
- , , , , . Risk of Clostridium difficile diarrhea among hospital inpatients prescribed proton pump inhibitors: cohort and case‐control studies. CMAJ. 2004;171(1):33–38.
- , , , et al. Iatrogenic gastric acid suppression and the risk of nosocomial Clostridium difficile infection. Arch Intern Med. 2010;170(9):784–790.
- , , , . Acid‐suppressive medication use and the risk for hospital‐acquired pneumonia. JAMA. 2009;301(20):2120–2128.
- Healthcare Cost and Utilization Project. Clinical classifications software (CCS) for ICD‐9‐CM. December 2009. Agency for Healthcare Research and Quality, Rockville, MD. Available at: www.hcup‐us.ahrq.gov/toolssoftware/ccs/ccs.jsp. Accessed June 18, 2014.
- , . Stress ulcer prophylaxis in hospitalized patients not in intensive care units. Am J Health Syst Pharm. 2007;64(13):1396–1400.
- , . Magnitude and economic impact of inappropriate use of stress ulcer prophylaxis in non‐ICU hospitalized patients. Am J Gastroenterol. 2006;101(10):2200–2205.
- , . Stress‐ulcer prophylaxis for general medical patients: a review of the evidence. J Hosp Med. 2007;2(2):86–92.
- , , , et al. Hospital use of acid‐suppressive medications and its fall‐out on prescribing in general practice: a 1‐month survey. Aliment Pharmacol Ther. 2003;17(12):1503–1506.
- , , , , , . Inadequate use of acid‐suppressive therapy in hospitalized patients and its implications for general practice. Dig Dis Sci. 2005;50(12):2307–2311.
- , . Brief report: reducing inappropriate usage of stress ulcer prophylaxis among internal medicine residents. A practice‐based educational intervention. J Gen Intern Med. 2006;21(5):498–500.
- , , . Inappropriate continuation of stress ulcer prophylactic therapy after discharge. Ann Pharmacother. 2007;41(10):1611–1616.
- , , , et al. Choosing wisely in adult hospital medicine: five opportunities for improved healthcare value. J Hosp Med. 2013;8(9):486–492.
- , , , , . Risk of Clostridium difficile diarrhea among hospital inpatients prescribed proton pump inhibitors: cohort and case‐control studies. CMAJ. 2004;171(1):33–38.
- , , , et al. Iatrogenic gastric acid suppression and the risk of nosocomial Clostridium difficile infection. Arch Intern Med. 2010;170(9):784–790.
- , , , . Acid‐suppressive medication use and the risk for hospital‐acquired pneumonia. JAMA. 2009;301(20):2120–2128.
- Healthcare Cost and Utilization Project. Clinical classifications software (CCS) for ICD‐9‐CM. December 2009. Agency for Healthcare Research and Quality, Rockville, MD. Available at: www.hcup‐us.ahrq.gov/toolssoftware/ccs/ccs.jsp. Accessed June 18, 2014.
