A 43‐year‐old woman presented to an outside hospital with painful plaques and patches on her bilateral lower extremities. Two weeks prior to presentation, she had noticed a single red lesion on her left ankle. Over the next two weeks, the lesion enlarged to involve the lower half of her posterior calf and subsequently turned purple and became exquisitely tender. Similar but smaller purple, tender lesions simultaneously appeared, first over her right shin and then on her bilateral thighs and hips. She also reported fatigue as well as diffuse joint pains in her hands and wrists bilaterally for the past month. She denied any swelling of these joints or functional impairment. She denied fevers, weight loss, headache, sinus symptoms, difficulty breathing, or abdominal pain.

Although we do not yet have a physical exam, the tempo, pattern of spread, and accompanying features allow some early hypotheses to be considered. Distal lower extremity lesions which darkened and spread could be erythema nodosum or erythema induratum. Malignancies rarely have such prominent skin manifestations, although leukemia cutis or an aggressive cutaneous T cell lymphoma might present with disseminated and darkened plaques, and Kaposi's sarcoma is characteristically purple and multifocal. Autoimmune disorders such as sarcoidosis, cutaneous lupus, and psoriasis may similarly present with widespread plaques. Most disseminated infections that start with patches evolve to pustules, ulcers, bullae, or other forms that reflect the invasive nature of the infection; syphilis warrants consideration for any widespread eruption of unknown etiology. Antecedent arthralgias with fatigue suggest an autoimmune condition, although infections such as hepatitis or parvovirus can do the same. Systemic lupus erythematosus (SLE) or rheumatoid arthritis (RA) would be favored initially on account of her demographics and the hand and wrist involvement, and each can be associated with vasculitis.

The significant pain as described is not compatible with most of the aforementioned diagnoses. Its presence, coupled with potential autoimmune symptoms, suggests a vasculitis such as polyarteritis nodosa (which can have prominent diffuse skin involvement), Henoch Schonlein purpura (with its predilection for the lower extremities, including extension to the hips and buttocks), cryoglobulinemia, or SLE‐ or RA‐associated vasculitis. Calciphylaxis is another ischemic vascular disorder that can cause diffuse dark painful lesions, but this only warrants consideration if advanced renal disease is present.

A skin biopsy ofher right hip was taken at the outside hospital. She was discharged on a two‐week course of prednisone for suspected vasculitis while biopsy results were pending. Over the next two weeks, none of the skin lesions improved, despite compliance with this treatment, and the skin over her left posterior calf and right shin lesions began to erode and bleed. In addition, small purple, tender lesions appeared over the pinnae of both ears. Three weeks after her initial evaluation, she presented to another emergency department for ulcerating skin lesions and worsening pain. At that point, the initial skin biopsy result was available and revealed vasculopathy of the small vessels with thrombi but no vasculitis.

The patient had no children,and denied a history of miscarriages. Her past medical history was unremarkable. She did not report any history of thrombotic events. She started a new job as a software engineer one month ago and was feeling stressed about her new responsibilities. She denied any high‐risk sexual behavior and any history of intravenous drug use. She had not traveled recently and did not own any pets. There was no family history of rheumatologic disorders, hypercoagulable states, or thrombotic events.

This picture of occluded but noninflamed vessels shifts the diagnosis away from vasculitis and focuses attention on hypercoagulable states with prominent dermal manifestations, including antiphospholipid antibody syndrome (APLS) and livedoid vasculopathy. In this young woman with arthralgias, consideration of SLE and APLS is warranted. Her recent increase in stress and widespread purpuric and ulcerative lesions could bring to mind a factitious disorder, but the histology results eliminate this possibility.

The patient's temperaturewas 36.5C, her blood pressure was 110/70 mmHg, respiratory rate was 16 breaths per minute, and her heart rate was 65 beats per minute. She was well‐appearing but in moderate pain. She did not have any oral lesions. Her cardiac, respiratory, and abdominal exams were normal. Skin exam revealed a 10‐cm by 4‐cm area of bloody granulation tissue draining serosanguinous fluid, surrounded by stellate palpable netlike purpura on her left posterior calf. There was a similar 4‐cm by 2‐cm ulcerated lesion on her right shin. Both lesions were exquisitely tender to palpation. On her bilateral thighs and hips, there were multiple stellate purpuric patches, all 4 cm in diameter or less, and only minimally tender to palpation. She also had 1‐cm purpuric bullae on the helices of both ears (Figure 1) which were slightly tender to palpation. Splinter hemorrhages were also noted on multiple nail beds bilaterally. Musculoskeletal exam did not reveal any synovitis.

Figure 1

The original purpura on her calf and ear demonstrate a clear demarcation corresponding to cutaneous vascular insufficiency. The development of bullae (ear) and ulceration (calf) are compatible with ischemia. Despite the presence of multiple splinter hemorrhages, the distribution of lesions is very unusual for an embolic phenomenon (eg, endocarditis, cholesterol emboli, or atrial myxoma). The multifocal nature of the skin lesions with progression to well‐demarcated cutaneous necrosis is reminiscent of calciphylaxis or warfarin‐induced skin necrosis, although she lacks the relevant risk factors. A toxin such as cocaine or methamphetamine mediating multifocal vasoconstriction or hypercoagulability should be excluded.

The bilateral ear involvement remains decidedly unusual and makes me wonder if there is something about the ear, such as the nature of its circulation or its potentially lower temperature (as an acral organ) that might render it particularly susceptible, for instance, to cryoglobulinemia or cryofibrinogenemia‐mediated ischemia.

Laboratory studiesdemonstrated: white blood cell count of 1500/mm³ (37.3% neutrophils, 5.1% lymphocytes, 6.7% monocytes, and 1.3% eosinophils); hemoglobin 9.3 g/dl (mean corpuscular volume 91 fL); platelet count 212/mm³; erythrocyte sedimentation rate 62 mm/hr; C‐reactive protein 14.6 mg/L. Serum electrolytes, liver tests, coagulation studies, and urinalysis were normal. Fecal occult blood test was negative.

Her neutropenia and anemia suggest decreased production in the marrow by infection, malignancy, or toxin, or increased destruction, perhaps from an autoimmune process. The associated infections are usually viral, such as human immunodeficiency virus (HIV) and Epstein‐Barr virus (EBV), although their linkage with her cutaneous disease is tenuous. It is possible that malignancy could be present in the marrow with resultant dermal hypercoagulability and ischemia, but this seems unlikely. We do not know about any toxins that she has been exposed to, but these hematologic findings would mandate directed inquiry along those lines. In this young woman with cutaneous ulcers secondary to thrombotic vasculopathy, bicytopenia, antecedent arthralgias without synovitis, and elevated inflammatory markers, I favor an autoimmune process such as SLE, which I would evaluate with an antinuclear antibody (ANA) and antiphospholipid antibody studies.

She was admittedto the hospital and received hydromorphone for pain control. Corticosteroids were not administered. Peripheral blood morphology was normal. Antibodies against HIV1 and 2 were negative, as were antibodies against cytomegalovirus, EBV, parvovirus B19, mycoplasma pneumoniae, and hepatitis C virus. Bilateral lower extremity ultrasound was negative for deep vein thrombosis. Transthoracic echocardiogram was normal. Repeat skin biopsy confirmed small vessel vasculopathy without vasculitis (Figure 2). The results of the following investigations were also negative: ANA, rheumatoid factor, double‐stranded DNA (dsDNA), cyclic citrullinated peptide, ribonucleoprotein (RNP), and anti‐Smith antibodies. C3 and C4 complement levels were normal.

Figure 2

Given how much the histology is driving the clinical reasoning and focusing the differential diagnosis in this case, I agree with the decision to repeat the biopsy. In complex or undiagnosed cases, repeat histology samples allow for confirmation of the original interpretation (often with the perspective of new clinicians and pathologists) and sometimes reveal pathognomonic or additional findings that only appear after the disease has evolved over time. HIV seronegativity helps constrain the differential diagnosis, and parvovirus is another excellent consideration for arthralgias and cytopenias (with the predilection to involve cells lines other than RBCs particularly seen in HIV), although ulcers are not seen with this condition. Herpes simplex virus (HSV) is another viral infection that can cause painful skin ulcerations and cytopenias, although the duration and distribution are highly atypical. The negative ANA and dsDNA and normal complement levels make SLE unlikely. The negative lower extremity ultrasound helps frame the thromboses as a local cutaneous process rather than a systemic hypercoagulable state. Although the peripheral blood smear is normal, a bone marrow biopsy will be necessary to exclude a marrow invasive process, such as leukemia or lymphoma. A bone marrow biopsy would also provide another opportunity to examine tissue for mycobacteria or fungi which can cause ulcerations and cytopenias, although there is little reason to suspect she is susceptible to those pathogens. As this clinical picture fails to fit clearly with an infectious, autoimmune, or neoplastic disorder, I would revisit the possibility of toxinsprescription, complementary, over‐the‐counter, or illegal (eg, cocaine) at this time.

In further discussionwith the patient, she reported using cocaine intranasally for the past three months. Her urine toxicology was positive for cocaine. She was found to have positive perinuclear antineutrophil cytoplasmic antibodies (p‐ANCA), antimyeloperoxidase (MPO) antibodies, anticardiolipin (ACL) antibodies, and lupus anticoagulant (LAC). By hospital day 3, her lesions had significantly improved without any intervention, and her absolute neutrophil count increased to 1080/mm³.

The presence of widespread cutaneous ischemia (with bland thrombosis) and detectable ACL and LAC antibodies is compatible with APLS; the APLS could be deemed primary, because there is no clear associated rheumatologic or other systemic disease. However, neutropenia is not a characteristic of APLS, which has thrombocytopenia as its more frequently associated hematologic abnormality. Livedoid vasculopathy, a related disorder, is also supported by the ACL and LAC results, but also does not feature neutropenia. While the presence of diffuse thrombosis could be attributed to a widespread secondary effect of cocaine vasoconstriction, the appearance of ANCA (which can be drug‐induced, eg, propylthiouracil [PTU]) and the slowly resolving neutropenia during hospitalization without specific treatment is very suggestive of a toxin. The demographic, diffuse skin ulcers, and hematologic and serologic profile is compatible with the recently described toxidrome related to levamisole adulteration of cocaine.

A send‐out studyof a urine sample returned positive for levamisole. Based on purpuric skin lesions with a predilection for the ears, agranulocytosis, and skin biopsy revealing thrombotic vasculopathy, she was diagnosed with levamisole‐adulterated cocaine exposure. One week after discharge, her lower extremity pain and ulcerations were significantly improved. Her absolute neutrophil count increased to 2820/mm.3 Her urine toxicology screen was negative for cocaine.

DISCUSSION

Levamisole was initially developed in 1964 as an antihelminthic agent. Its incidentally discovered immunomodulatory effects led to trials for the treatment of chronic infections, inflammatory bowel disease, rheumatic diseases,1 and nephrotic syndrome in children.2 By 1990, 3 major studies supported levamisole as an adjunctive therapy in melanoma3 and colon cancer.4

Although levamisole appeared to be nontoxic at single or low doses, long‐term use in clinical trials demonstrated that 2.5%‐13% of patients developed life‐threatening agranulocytosis, and up to 10% of those instances resulted in death.5 A distinctive cutaneous pseudovasculitis was noted in children on therapeutic levamisole. They presented with purpura that had a predilection for the ears, cheeks, and thighs,6 and positive serologic markers for ANCA and antiphospholipid antibodies. Skin biopsies of the purpuric lesions revealed leukocytoclastic vasculitis, thrombotic vasculitis, and/or vascular occlusions.

Levamisole was withdrawn from the market in 2000 in the United States due to its side effects,7 but quickly found its way onto the black market. It was first detected in cocaine in 2002, and the percentage of cocaine containing levamisole has steadily been increasing since then. In July 2009, over 70% of cocaine seized by the Drug Enforcement Administration was found to contain levamisole.8 It is unclear exactly why this drug is used as an adulterant in cocaine. Theories include potentiation of the euphoric effects of cocaine, serving as a bulking agent, or functioning as a chemical signature to track distribution.9

The resurgence of levamisole has brought a new face to a problem seen over a decade ago. Current reports of levamisole toxicity describe adults presenting with purpura preferentially involving the ears, neutropenia, positive ANCA, and positive antiphospholipid antibodies.1012 Since 2002, there have been at least 20 confirmed cases of agranulocytosis and two deaths associated with levamisole‐adulterated cocaine.8, 13, 14 In September 2009, the Department of Health and Human Services issued a public health alert warning of an impending increase in levamisole‐related illness.

Levamisole is not detected on routine toxicology screens, but can be tested for using gas chromatography and mass spectrometry. Most laboratories do not offer testing for levamisole and send‐out testing is required. Given its half‐life of 5.6 hours, levamisole can only be detected in the blood within 24 hours, and in the urine within 48‐72 hours of exposure.15, 16 Urine samples are preferred over blood samples, since blood levels decline more rapidly and have lower sensitivity. Cocaine can also be sent out to local or state forensics laboratories to be tested for levamisole. The only definitive treatment for levamisole‐induced cutaneous pseudovasculitis and neutropenia is cessation of toxin exposure.

Although the discussant had familiarity with this toxidrome from local and published cases, he was only able to settle on levamisole toxicity after a series of competing hypotheses were ruled out on the basis of irreconcilable features (vasculitis and histology results; APLS and neutropenia; SLE and negative ANA with no visceral involvement) and by using analogical reasoning (eg, to infer the presence of a toxin on the basis of neutropenia [as seen with chemotherapy and other drugs] and ANCA induction [as seen with PTU]). It was a laborious process of hypothesis testing, but one that ultimately allowed him to crack the case.

A 43‐year‐old woman presented to an outside hospital with painful plaques and patches on her bilateral lower extremities. Two weeks prior to presentation, she had noticed a single red lesion on her left ankle. Over the next two weeks, the lesion enlarged to involve the lower half of her posterior calf and subsequently turned purple and became exquisitely tender. Similar but smaller purple, tender lesions simultaneously appeared, first over her right shin and then on her bilateral thighs and hips. She also reported fatigue as well as diffuse joint pains in her hands and wrists bilaterally for the past month. She denied any swelling of these joints or functional impairment. She denied fevers, weight loss, headache, sinus symptoms, difficulty breathing, or abdominal pain.

Although we do not yet have a physical exam, the tempo, pattern of spread, and accompanying features allow some early hypotheses to be considered. Distal lower extremity lesions which darkened and spread could be erythema nodosum or erythema induratum. Malignancies rarely have such prominent skin manifestations, although leukemia cutis or an aggressive cutaneous T cell lymphoma might present with disseminated and darkened plaques, and Kaposi's sarcoma is characteristically purple and multifocal. Autoimmune disorders such as sarcoidosis, cutaneous lupus, and psoriasis may similarly present with widespread plaques. Most disseminated infections that start with patches evolve to pustules, ulcers, bullae, or other forms that reflect the invasive nature of the infection; syphilis warrants consideration for any widespread eruption of unknown etiology. Antecedent arthralgias with fatigue suggest an autoimmune condition, although infections such as hepatitis or parvovirus can do the same. Systemic lupus erythematosus (SLE) or rheumatoid arthritis (RA) would be favored initially on account of her demographics and the hand and wrist involvement, and each can be associated with vasculitis.

The significant pain as described is not compatible with most of the aforementioned diagnoses. Its presence, coupled with potential autoimmune symptoms, suggests a vasculitis such as polyarteritis nodosa (which can have prominent diffuse skin involvement), Henoch Schonlein purpura (with its predilection for the lower extremities, including extension to the hips and buttocks), cryoglobulinemia, or SLE‐ or RA‐associated vasculitis. Calciphylaxis is another ischemic vascular disorder that can cause diffuse dark painful lesions, but this only warrants consideration if advanced renal disease is present.

A skin biopsy ofher right hip was taken at the outside hospital. She was discharged on a two‐week course of prednisone for suspected vasculitis while biopsy results were pending. Over the next two weeks, none of the skin lesions improved, despite compliance with this treatment, and the skin over her left posterior calf and right shin lesions began to erode and bleed. In addition, small purple, tender lesions appeared over the pinnae of both ears. Three weeks after her initial evaluation, she presented to another emergency department for ulcerating skin lesions and worsening pain. At that point, the initial skin biopsy result was available and revealed vasculopathy of the small vessels with thrombi but no vasculitis.

The patient had no children,and denied a history of miscarriages. Her past medical history was unremarkable. She did not report any history of thrombotic events. She started a new job as a software engineer one month ago and was feeling stressed about her new responsibilities. She denied any high‐risk sexual behavior and any history of intravenous drug use. She had not traveled recently and did not own any pets. There was no family history of rheumatologic disorders, hypercoagulable states, or thrombotic events.

This picture of occluded but noninflamed vessels shifts the diagnosis away from vasculitis and focuses attention on hypercoagulable states with prominent dermal manifestations, including antiphospholipid antibody syndrome (APLS) and livedoid vasculopathy. In this young woman with arthralgias, consideration of SLE and APLS is warranted. Her recent increase in stress and widespread purpuric and ulcerative lesions could bring to mind a factitious disorder, but the histology results eliminate this possibility.

The patient's temperaturewas 36.5C, her blood pressure was 110/70 mmHg, respiratory rate was 16 breaths per minute, and her heart rate was 65 beats per minute. She was well‐appearing but in moderate pain. She did not have any oral lesions. Her cardiac, respiratory, and abdominal exams were normal. Skin exam revealed a 10‐cm by 4‐cm area of bloody granulation tissue draining serosanguinous fluid, surrounded by stellate palpable netlike purpura on her left posterior calf. There was a similar 4‐cm by 2‐cm ulcerated lesion on her right shin. Both lesions were exquisitely tender to palpation. On her bilateral thighs and hips, there were multiple stellate purpuric patches, all 4 cm in diameter or less, and only minimally tender to palpation. She also had 1‐cm purpuric bullae on the helices of both ears (Figure 1) which were slightly tender to palpation. Splinter hemorrhages were also noted on multiple nail beds bilaterally. Musculoskeletal exam did not reveal any synovitis.

Figure 1

The original purpura on her calf and ear demonstrate a clear demarcation corresponding to cutaneous vascular insufficiency. The development of bullae (ear) and ulceration (calf) are compatible with ischemia. Despite the presence of multiple splinter hemorrhages, the distribution of lesions is very unusual for an embolic phenomenon (eg, endocarditis, cholesterol emboli, or atrial myxoma). The multifocal nature of the skin lesions with progression to well‐demarcated cutaneous necrosis is reminiscent of calciphylaxis or warfarin‐induced skin necrosis, although she lacks the relevant risk factors. A toxin such as cocaine or methamphetamine mediating multifocal vasoconstriction or hypercoagulability should be excluded.

The bilateral ear involvement remains decidedly unusual and makes me wonder if there is something about the ear, such as the nature of its circulation or its potentially lower temperature (as an acral organ) that might render it particularly susceptible, for instance, to cryoglobulinemia or cryofibrinogenemia‐mediated ischemia.

Laboratory studiesdemonstrated: white blood cell count of 1500/mm³ (37.3% neutrophils, 5.1% lymphocytes, 6.7% monocytes, and 1.3% eosinophils); hemoglobin 9.3 g/dl (mean corpuscular volume 91 fL); platelet count 212/mm³; erythrocyte sedimentation rate 62 mm/hr; C‐reactive protein 14.6 mg/L. Serum electrolytes, liver tests, coagulation studies, and urinalysis were normal. Fecal occult blood test was negative.

Her neutropenia and anemia suggest decreased production in the marrow by infection, malignancy, or toxin, or increased destruction, perhaps from an autoimmune process. The associated infections are usually viral, such as human immunodeficiency virus (HIV) and Epstein‐Barr virus (EBV), although their linkage with her cutaneous disease is tenuous. It is possible that malignancy could be present in the marrow with resultant dermal hypercoagulability and ischemia, but this seems unlikely. We do not know about any toxins that she has been exposed to, but these hematologic findings would mandate directed inquiry along those lines. In this young woman with cutaneous ulcers secondary to thrombotic vasculopathy, bicytopenia, antecedent arthralgias without synovitis, and elevated inflammatory markers, I favor an autoimmune process such as SLE, which I would evaluate with an antinuclear antibody (ANA) and antiphospholipid antibody studies.

She was admittedto the hospital and received hydromorphone for pain control. Corticosteroids were not administered. Peripheral blood morphology was normal. Antibodies against HIV1 and 2 were negative, as were antibodies against cytomegalovirus, EBV, parvovirus B19, mycoplasma pneumoniae, and hepatitis C virus. Bilateral lower extremity ultrasound was negative for deep vein thrombosis. Transthoracic echocardiogram was normal. Repeat skin biopsy confirmed small vessel vasculopathy without vasculitis (Figure 2). The results of the following investigations were also negative: ANA, rheumatoid factor, double‐stranded DNA (dsDNA), cyclic citrullinated peptide, ribonucleoprotein (RNP), and anti‐Smith antibodies. C3 and C4 complement levels were normal.

Figure 2

Given how much the histology is driving the clinical reasoning and focusing the differential diagnosis in this case, I agree with the decision to repeat the biopsy. In complex or undiagnosed cases, repeat histology samples allow for confirmation of the original interpretation (often with the perspective of new clinicians and pathologists) and sometimes reveal pathognomonic or additional findings that only appear after the disease has evolved over time. HIV seronegativity helps constrain the differential diagnosis, and parvovirus is another excellent consideration for arthralgias and cytopenias (with the predilection to involve cells lines other than RBCs particularly seen in HIV), although ulcers are not seen with this condition. Herpes simplex virus (HSV) is another viral infection that can cause painful skin ulcerations and cytopenias, although the duration and distribution are highly atypical. The negative ANA and dsDNA and normal complement levels make SLE unlikely. The negative lower extremity ultrasound helps frame the thromboses as a local cutaneous process rather than a systemic hypercoagulable state. Although the peripheral blood smear is normal, a bone marrow biopsy will be necessary to exclude a marrow invasive process, such as leukemia or lymphoma. A bone marrow biopsy would also provide another opportunity to examine tissue for mycobacteria or fungi which can cause ulcerations and cytopenias, although there is little reason to suspect she is susceptible to those pathogens. As this clinical picture fails to fit clearly with an infectious, autoimmune, or neoplastic disorder, I would revisit the possibility of toxinsprescription, complementary, over‐the‐counter, or illegal (eg, cocaine) at this time.

In further discussionwith the patient, she reported using cocaine intranasally for the past three months. Her urine toxicology was positive for cocaine. She was found to have positive perinuclear antineutrophil cytoplasmic antibodies (p‐ANCA), antimyeloperoxidase (MPO) antibodies, anticardiolipin (ACL) antibodies, and lupus anticoagulant (LAC). By hospital day 3, her lesions had significantly improved without any intervention, and her absolute neutrophil count increased to 1080/mm³.

The presence of widespread cutaneous ischemia (with bland thrombosis) and detectable ACL and LAC antibodies is compatible with APLS; the APLS could be deemed primary, because there is no clear associated rheumatologic or other systemic disease. However, neutropenia is not a characteristic of APLS, which has thrombocytopenia as its more frequently associated hematologic abnormality. Livedoid vasculopathy, a related disorder, is also supported by the ACL and LAC results, but also does not feature neutropenia. While the presence of diffuse thrombosis could be attributed to a widespread secondary effect of cocaine vasoconstriction, the appearance of ANCA (which can be drug‐induced, eg, propylthiouracil [PTU]) and the slowly resolving neutropenia during hospitalization without specific treatment is very suggestive of a toxin. The demographic, diffuse skin ulcers, and hematologic and serologic profile is compatible with the recently described toxidrome related to levamisole adulteration of cocaine.

A send‐out studyof a urine sample returned positive for levamisole. Based on purpuric skin lesions with a predilection for the ears, agranulocytosis, and skin biopsy revealing thrombotic vasculopathy, she was diagnosed with levamisole‐adulterated cocaine exposure. One week after discharge, her lower extremity pain and ulcerations were significantly improved. Her absolute neutrophil count increased to 2820/mm.3 Her urine toxicology screen was negative for cocaine.

DISCUSSION

Levamisole was initially developed in 1964 as an antihelminthic agent. Its incidentally discovered immunomodulatory effects led to trials for the treatment of chronic infections, inflammatory bowel disease, rheumatic diseases,1 and nephrotic syndrome in children.2 By 1990, 3 major studies supported levamisole as an adjunctive therapy in melanoma3 and colon cancer.4

Although levamisole appeared to be nontoxic at single or low doses, long‐term use in clinical trials demonstrated that 2.5%‐13% of patients developed life‐threatening agranulocytosis, and up to 10% of those instances resulted in death.5 A distinctive cutaneous pseudovasculitis was noted in children on therapeutic levamisole. They presented with purpura that had a predilection for the ears, cheeks, and thighs,6 and positive serologic markers for ANCA and antiphospholipid antibodies. Skin biopsies of the purpuric lesions revealed leukocytoclastic vasculitis, thrombotic vasculitis, and/or vascular occlusions.

Levamisole was withdrawn from the market in 2000 in the United States due to its side effects,7 but quickly found its way onto the black market. It was first detected in cocaine in 2002, and the percentage of cocaine containing levamisole has steadily been increasing since then. In July 2009, over 70% of cocaine seized by the Drug Enforcement Administration was found to contain levamisole.8 It is unclear exactly why this drug is used as an adulterant in cocaine. Theories include potentiation of the euphoric effects of cocaine, serving as a bulking agent, or functioning as a chemical signature to track distribution.9

The resurgence of levamisole has brought a new face to a problem seen over a decade ago. Current reports of levamisole toxicity describe adults presenting with purpura preferentially involving the ears, neutropenia, positive ANCA, and positive antiphospholipid antibodies.1012 Since 2002, there have been at least 20 confirmed cases of agranulocytosis and two deaths associated with levamisole‐adulterated cocaine.8, 13, 14 In September 2009, the Department of Health and Human Services issued a public health alert warning of an impending increase in levamisole‐related illness.

Levamisole is not detected on routine toxicology screens, but can be tested for using gas chromatography and mass spectrometry. Most laboratories do not offer testing for levamisole and send‐out testing is required. Given its half‐life of 5.6 hours, levamisole can only be detected in the blood within 24 hours, and in the urine within 48‐72 hours of exposure.15, 16 Urine samples are preferred over blood samples, since blood levels decline more rapidly and have lower sensitivity. Cocaine can also be sent out to local or state forensics laboratories to be tested for levamisole. The only definitive treatment for levamisole‐induced cutaneous pseudovasculitis and neutropenia is cessation of toxin exposure.

Although the discussant had familiarity with this toxidrome from local and published cases, he was only able to settle on levamisole toxicity after a series of competing hypotheses were ruled out on the basis of irreconcilable features (vasculitis and histology results; APLS and neutropenia; SLE and negative ANA with no visceral involvement) and by using analogical reasoning (eg, to infer the presence of a toxin on the basis of neutropenia [as seen with chemotherapy and other drugs] and ANCA induction [as seen with PTU]). It was a laborious process of hypothesis testing, but one that ultimately allowed him to crack the case.

A 43‐year‐old woman presented to an outside hospital with painful plaques and patches on her bilateral lower extremities. Two weeks prior to presentation, she had noticed a single red lesion on her left ankle. Over the next two weeks, the lesion enlarged to involve the lower half of her posterior calf and subsequently turned purple and became exquisitely tender. Similar but smaller purple, tender lesions simultaneously appeared, first over her right shin and then on her bilateral thighs and hips. She also reported fatigue as well as diffuse joint pains in her hands and wrists bilaterally for the past month. She denied any swelling of these joints or functional impairment. She denied fevers, weight loss, headache, sinus symptoms, difficulty breathing, or abdominal pain.

Although we do not yet have a physical exam, the tempo, pattern of spread, and accompanying features allow some early hypotheses to be considered. Distal lower extremity lesions which darkened and spread could be erythema nodosum or erythema induratum. Malignancies rarely have such prominent skin manifestations, although leukemia cutis or an aggressive cutaneous T cell lymphoma might present with disseminated and darkened plaques, and Kaposi's sarcoma is characteristically purple and multifocal. Autoimmune disorders such as sarcoidosis, cutaneous lupus, and psoriasis may similarly present with widespread plaques. Most disseminated infections that start with patches evolve to pustules, ulcers, bullae, or other forms that reflect the invasive nature of the infection; syphilis warrants consideration for any widespread eruption of unknown etiology. Antecedent arthralgias with fatigue suggest an autoimmune condition, although infections such as hepatitis or parvovirus can do the same. Systemic lupus erythematosus (SLE) or rheumatoid arthritis (RA) would be favored initially on account of her demographics and the hand and wrist involvement, and each can be associated with vasculitis.

The significant pain as described is not compatible with most of the aforementioned diagnoses. Its presence, coupled with potential autoimmune symptoms, suggests a vasculitis such as polyarteritis nodosa (which can have prominent diffuse skin involvement), Henoch Schonlein purpura (with its predilection for the lower extremities, including extension to the hips and buttocks), cryoglobulinemia, or SLE‐ or RA‐associated vasculitis. Calciphylaxis is another ischemic vascular disorder that can cause diffuse dark painful lesions, but this only warrants consideration if advanced renal disease is present.

A skin biopsy ofher right hip was taken at the outside hospital. She was discharged on a two‐week course of prednisone for suspected vasculitis while biopsy results were pending. Over the next two weeks, none of the skin lesions improved, despite compliance with this treatment, and the skin over her left posterior calf and right shin lesions began to erode and bleed. In addition, small purple, tender lesions appeared over the pinnae of both ears. Three weeks after her initial evaluation, she presented to another emergency department for ulcerating skin lesions and worsening pain. At that point, the initial skin biopsy result was available and revealed vasculopathy of the small vessels with thrombi but no vasculitis.

The patient had no children,and denied a history of miscarriages. Her past medical history was unremarkable. She did not report any history of thrombotic events. She started a new job as a software engineer one month ago and was feeling stressed about her new responsibilities. She denied any high‐risk sexual behavior and any history of intravenous drug use. She had not traveled recently and did not own any pets. There was no family history of rheumatologic disorders, hypercoagulable states, or thrombotic events.

This picture of occluded but noninflamed vessels shifts the diagnosis away from vasculitis and focuses attention on hypercoagulable states with prominent dermal manifestations, including antiphospholipid antibody syndrome (APLS) and livedoid vasculopathy. In this young woman with arthralgias, consideration of SLE and APLS is warranted. Her recent increase in stress and widespread purpuric and ulcerative lesions could bring to mind a factitious disorder, but the histology results eliminate this possibility.

The patient's temperaturewas 36.5C, her blood pressure was 110/70 mmHg, respiratory rate was 16 breaths per minute, and her heart rate was 65 beats per minute. She was well‐appearing but in moderate pain. She did not have any oral lesions. Her cardiac, respiratory, and abdominal exams were normal. Skin exam revealed a 10‐cm by 4‐cm area of bloody granulation tissue draining serosanguinous fluid, surrounded by stellate palpable netlike purpura on her left posterior calf. There was a similar 4‐cm by 2‐cm ulcerated lesion on her right shin. Both lesions were exquisitely tender to palpation. On her bilateral thighs and hips, there were multiple stellate purpuric patches, all 4 cm in diameter or less, and only minimally tender to palpation. She also had 1‐cm purpuric bullae on the helices of both ears (Figure 1) which were slightly tender to palpation. Splinter hemorrhages were also noted on multiple nail beds bilaterally. Musculoskeletal exam did not reveal any synovitis.

Figure 1

The original purpura on her calf and ear demonstrate a clear demarcation corresponding to cutaneous vascular insufficiency. The development of bullae (ear) and ulceration (calf) are compatible with ischemia. Despite the presence of multiple splinter hemorrhages, the distribution of lesions is very unusual for an embolic phenomenon (eg, endocarditis, cholesterol emboli, or atrial myxoma). The multifocal nature of the skin lesions with progression to well‐demarcated cutaneous necrosis is reminiscent of calciphylaxis or warfarin‐induced skin necrosis, although she lacks the relevant risk factors. A toxin such as cocaine or methamphetamine mediating multifocal vasoconstriction or hypercoagulability should be excluded.

The bilateral ear involvement remains decidedly unusual and makes me wonder if there is something about the ear, such as the nature of its circulation or its potentially lower temperature (as an acral organ) that might render it particularly susceptible, for instance, to cryoglobulinemia or cryofibrinogenemia‐mediated ischemia.

Laboratory studiesdemonstrated: white blood cell count of 1500/mm³ (37.3% neutrophils, 5.1% lymphocytes, 6.7% monocytes, and 1.3% eosinophils); hemoglobin 9.3 g/dl (mean corpuscular volume 91 fL); platelet count 212/mm³; erythrocyte sedimentation rate 62 mm/hr; C‐reactive protein 14.6 mg/L. Serum electrolytes, liver tests, coagulation studies, and urinalysis were normal. Fecal occult blood test was negative.

Her neutropenia and anemia suggest decreased production in the marrow by infection, malignancy, or toxin, or increased destruction, perhaps from an autoimmune process. The associated infections are usually viral, such as human immunodeficiency virus (HIV) and Epstein‐Barr virus (EBV), although their linkage with her cutaneous disease is tenuous. It is possible that malignancy could be present in the marrow with resultant dermal hypercoagulability and ischemia, but this seems unlikely. We do not know about any toxins that she has been exposed to, but these hematologic findings would mandate directed inquiry along those lines. In this young woman with cutaneous ulcers secondary to thrombotic vasculopathy, bicytopenia, antecedent arthralgias without synovitis, and elevated inflammatory markers, I favor an autoimmune process such as SLE, which I would evaluate with an antinuclear antibody (ANA) and antiphospholipid antibody studies.

She was admittedto the hospital and received hydromorphone for pain control. Corticosteroids were not administered. Peripheral blood morphology was normal. Antibodies against HIV1 and 2 were negative, as were antibodies against cytomegalovirus, EBV, parvovirus B19, mycoplasma pneumoniae, and hepatitis C virus. Bilateral lower extremity ultrasound was negative for deep vein thrombosis. Transthoracic echocardiogram was normal. Repeat skin biopsy confirmed small vessel vasculopathy without vasculitis (Figure 2). The results of the following investigations were also negative: ANA, rheumatoid factor, double‐stranded DNA (dsDNA), cyclic citrullinated peptide, ribonucleoprotein (RNP), and anti‐Smith antibodies. C3 and C4 complement levels were normal.

Figure 2

Given how much the histology is driving the clinical reasoning and focusing the differential diagnosis in this case, I agree with the decision to repeat the biopsy. In complex or undiagnosed cases, repeat histology samples allow for confirmation of the original interpretation (often with the perspective of new clinicians and pathologists) and sometimes reveal pathognomonic or additional findings that only appear after the disease has evolved over time. HIV seronegativity helps constrain the differential diagnosis, and parvovirus is another excellent consideration for arthralgias and cytopenias (with the predilection to involve cells lines other than RBCs particularly seen in HIV), although ulcers are not seen with this condition. Herpes simplex virus (HSV) is another viral infection that can cause painful skin ulcerations and cytopenias, although the duration and distribution are highly atypical. The negative ANA and dsDNA and normal complement levels make SLE unlikely. The negative lower extremity ultrasound helps frame the thromboses as a local cutaneous process rather than a systemic hypercoagulable state. Although the peripheral blood smear is normal, a bone marrow biopsy will be necessary to exclude a marrow invasive process, such as leukemia or lymphoma. A bone marrow biopsy would also provide another opportunity to examine tissue for mycobacteria or fungi which can cause ulcerations and cytopenias, although there is little reason to suspect she is susceptible to those pathogens. As this clinical picture fails to fit clearly with an infectious, autoimmune, or neoplastic disorder, I would revisit the possibility of toxinsprescription, complementary, over‐the‐counter, or illegal (eg, cocaine) at this time.

In further discussionwith the patient, she reported using cocaine intranasally for the past three months. Her urine toxicology was positive for cocaine. She was found to have positive perinuclear antineutrophil cytoplasmic antibodies (p‐ANCA), antimyeloperoxidase (MPO) antibodies, anticardiolipin (ACL) antibodies, and lupus anticoagulant (LAC). By hospital day 3, her lesions had significantly improved without any intervention, and her absolute neutrophil count increased to 1080/mm³.

The presence of widespread cutaneous ischemia (with bland thrombosis) and detectable ACL and LAC antibodies is compatible with APLS; the APLS could be deemed primary, because there is no clear associated rheumatologic or other systemic disease. However, neutropenia is not a characteristic of APLS, which has thrombocytopenia as its more frequently associated hematologic abnormality. Livedoid vasculopathy, a related disorder, is also supported by the ACL and LAC results, but also does not feature neutropenia. While the presence of diffuse thrombosis could be attributed to a widespread secondary effect of cocaine vasoconstriction, the appearance of ANCA (which can be drug‐induced, eg, propylthiouracil [PTU]) and the slowly resolving neutropenia during hospitalization without specific treatment is very suggestive of a toxin. The demographic, diffuse skin ulcers, and hematologic and serologic profile is compatible with the recently described toxidrome related to levamisole adulteration of cocaine.

A send‐out studyof a urine sample returned positive for levamisole. Based on purpuric skin lesions with a predilection for the ears, agranulocytosis, and skin biopsy revealing thrombotic vasculopathy, she was diagnosed with levamisole‐adulterated cocaine exposure. One week after discharge, her lower extremity pain and ulcerations were significantly improved. Her absolute neutrophil count increased to 2820/mm.3 Her urine toxicology screen was negative for cocaine.

DISCUSSION

Levamisole was initially developed in 1964 as an antihelminthic agent. Its incidentally discovered immunomodulatory effects led to trials for the treatment of chronic infections, inflammatory bowel disease, rheumatic diseases,1 and nephrotic syndrome in children.2 By 1990, 3 major studies supported levamisole as an adjunctive therapy in melanoma3 and colon cancer.4

Although levamisole appeared to be nontoxic at single or low doses, long‐term use in clinical trials demonstrated that 2.5%‐13% of patients developed life‐threatening agranulocytosis, and up to 10% of those instances resulted in death.5 A distinctive cutaneous pseudovasculitis was noted in children on therapeutic levamisole. They presented with purpura that had a predilection for the ears, cheeks, and thighs,6 and positive serologic markers for ANCA and antiphospholipid antibodies. Skin biopsies of the purpuric lesions revealed leukocytoclastic vasculitis, thrombotic vasculitis, and/or vascular occlusions.

Levamisole was withdrawn from the market in 2000 in the United States due to its side effects,7 but quickly found its way onto the black market. It was first detected in cocaine in 2002, and the percentage of cocaine containing levamisole has steadily been increasing since then. In July 2009, over 70% of cocaine seized by the Drug Enforcement Administration was found to contain levamisole.8 It is unclear exactly why this drug is used as an adulterant in cocaine. Theories include potentiation of the euphoric effects of cocaine, serving as a bulking agent, or functioning as a chemical signature to track distribution.9

The resurgence of levamisole has brought a new face to a problem seen over a decade ago. Current reports of levamisole toxicity describe adults presenting with purpura preferentially involving the ears, neutropenia, positive ANCA, and positive antiphospholipid antibodies.1012 Since 2002, there have been at least 20 confirmed cases of agranulocytosis and two deaths associated with levamisole‐adulterated cocaine.8, 13, 14 In September 2009, the Department of Health and Human Services issued a public health alert warning of an impending increase in levamisole‐related illness.

Levamisole is not detected on routine toxicology screens, but can be tested for using gas chromatography and mass spectrometry. Most laboratories do not offer testing for levamisole and send‐out testing is required. Given its half‐life of 5.6 hours, levamisole can only be detected in the blood within 24 hours, and in the urine within 48‐72 hours of exposure.15, 16 Urine samples are preferred over blood samples, since blood levels decline more rapidly and have lower sensitivity. Cocaine can also be sent out to local or state forensics laboratories to be tested for levamisole. The only definitive treatment for levamisole‐induced cutaneous pseudovasculitis and neutropenia is cessation of toxin exposure.

Although the discussant had familiarity with this toxidrome from local and published cases, he was only able to settle on levamisole toxicity after a series of competing hypotheses were ruled out on the basis of irreconcilable features (vasculitis and histology results; APLS and neutropenia; SLE and negative ANA with no visceral involvement) and by using analogical reasoning (eg, to infer the presence of a toxin on the basis of neutropenia [as seen with chemotherapy and other drugs] and ANCA induction [as seen with PTU]). It was a laborious process of hypothesis testing, but one that ultimately allowed him to crack the case.

In the United States, unhealthy alcohol use affects medical care on several levels. The prevalence of alcohol problems is 7%20% or higher among outpatients,1 30%40% among emergency room patients, and 50% among patients with trauma.13 In 2006, approximately 430,000 hospital discharges in the United States were for persons with a principal (first‐listed) alcohol‐related diagnosis, and 1.7 million discharges listed at least one alcohol‐related diagnosis, representing 1.3% and 5.3% of all hospital discharges, respectively.4 Alcoholic psychosis (34.5%) and alcohol dependence syndrome (29.5%) together accounted for the majority of principal alcohol‐related diagnoses.4 Additionally, many patients hospitalized for other indications are susceptible to withdrawal symptoms due to physiological habituation to alcohol. Abrupt cessation of alcohol intake causes habituated drinkers to experience symptoms of alcohol withdrawal syndrome (AWS), which significantly increases intensity and cost of care. Trauma patients who develop AWS were found to have increased morbidity, more intensive care and ventilator days, and higher hospital costs than trauma patients without AWS.5 Unfortunately, attempts to develop predictive models to accurately forecast the likelihood of developing severe AWS in an individual case have been modestly successful at best.68

Regimens used to treat AWS have evolved over time, taking advantage of advances in the understanding of addiction neurophysiology. There is no specific ethanol receptor.9 Much of alcohol's acute effects on the central nervous system are mediated by its stimulation of the gamma‐aminobutyric acid (GABA) system, which is neuroinhibitory.10 Chronic alcohol use leads to habituation partly by inducing configuration changes of GABA‐A receptor subunits. This renders the GABA‐A receptor less sensitive to alcohol, barbiturates, and benzodiazepines.11 Although both GABA‐A and GABA‐B receptor activation cause increased GABA neuronal output, the GABA‐A receptor is rendered relatively less sensitive by chronic exposure to alcohol. Baclofen is a pure GABA‐B receptor agonist,12 and its GABA‐B stimulatory effect is maintained even in habituated alcoholics.13, 14 The absence of cross‐tolerance between baclofen and ethanol suggests that low doses of baclofen may be helpful in the management of AWS.

Currently, AWS is usually managed with benzodiazepines, using variable dosing depending on the severity of withdrawal symptoms. Such symptom‐triggered treatment is generally preferred over fixed‐dose regimens,15 in part because when using this method, many cases of AWS can be managed with less medication. Benzodiazepine regimens using high doses have been found to be associated with substantial morbidity and prolonged hospitalizations.16, 17

In a series of small studies, Addolorato's research team has reported decreases in AWS symptoms in association with the use of low doses of baclofen in an outpatient population,18 and has found baclofen to be associated with reduced alcohol craving in the long‐term management of alcohol dependence.11, 19, 20 Addolorato and colleagues' studies of baclofen in relieving AWS symptoms prompted our group to apply the use of baclofen in a larger group of inpatients with AWS.11, 18, 21 We conducted this study to improve understanding of the role of baclofen in the management of acute AWS in an inpatient population of subjects at risk for AWS, drawn from general hospital admissions.

Our primary null hypothesis was that Clinical Institute Withdrawal Assessment of Alcohol Scale, Revised (CIWA‐Ar) scores in acutely withdrawing alcoholic patients are equal in baclofen‐treated and placebo‐treated subject groups. Our secondary null hypothesis was that benzodiazepine doses used to treat acutely withdrawing alcoholic patients are equal in the baclofen‐treated and placebo‐treated groups.

METHODS

The protocol for this study was approved by the Essentia Health Institutional Review Board, Duluth, Minnesota.

This was a randomized, placebo‐controlled, double‐blind trial. Subjects were recruited from among patients who were admitted to 1 of 2 regional general hospitals in Duluth, Minnesota (St. Mary's Medical Center or Miller‐Dwan Medical Center) and who were identified by clinical staff as being at risk for AWS. Potential subjects were not required to have an alcohol‐related condition as their primary reason for admission, but were required to have a history of AWS or of alcohol use suggestive of significant risk for AWS, and to be able to provide informed consent, as described below.

Patients were not eligible for enrollment in the study if they had other active drug dependence in addition to alcohol; were using baclofen at the time of study enrollment; were using benzodiazepines chronically at the time of study enrollment; had a known baclofen or benzodiazepine sensitivity; were unable to take oral medications; were pregnant or breast‐feeding; had a serum creatinine level 2.0; had a history of non‐alcohol withdrawal seizures; required intravenous benzodiazepines to control their AWS; or were unable to complete the consenting procedures.

RESULTS

Seventy‐nine subjects met study inclusion criteria, and provided informed consent for participation in the study. Of these, 44 subjects developed signs of AWS sufficient to meet DSMIV criteria for AWS diagnosis, and were randomized to receive either baclofen or placebo, in addition to benzodiazepine therapy. As summarized in Table 1, subjects who developed signs of AWS were similar to subjects who did not enter withdrawal, differing principally in that those who developed AWS reported more drinks per day, and more significant history of previous AWS.

The 79 subjects who were enrolled were drawn from a population of 237 potential subjects who were screened for the study. The most common reasons that the 158 potential subjects were not enrolled were refusal (29.7%), low risk of withdrawal (19.0%), inability to provide consent (9.5%), and concurrent use of benzodiazepines (8.9%). The 15 patients who were unable to provide consent were those who scored 24 or lower on the MMSE, and did not have a surrogate decision‐maker available.

Of the 44 subjects who were randomized, 31 (18 in the baclofen group, 13 in the placebo group) completed 72 hours of CIWA‐Ar assessments, as summarized in Table 2. These assessments were completed either entirely as inpatients (24 subjects) or with inpatient assessments followed by outpatient assessments after discharge (7 subjects). Discharges prior to 72 hours occurred in 3 of the 18 subjects receiving baclofen, and in 4 of the 13 receiving placebo (odds ratio = 0.45, 95% CI = 0.082.49). Mean CIWA‐Ar scores for the 31 subjects who completed 72 hours of CIWA‐Ar assessments are presented in Figure 1.

Figure 1

Figures 2 and 3 summarize the mean lorazepam doses in each 8‐hour period, and the cumulative lorazepam doses for the subjects in the 2 arms of the study, respectively. The cumulative dose of lorazepam administered to the 31 subjects ranged from 0 to 1035 mg in the 72 hours following randomization, with a range of 1 to 1035 mg in the placebo group and 0 to 39 mg in the baclofen group. The 8 subjects who received the highest doses of lorazepam (20 mg or more) included 1 of the 18 subjects who received baclofen and 7 of the 13 subjects who received placebo (P = 0.004). Only 4 subjects required >50 mg of lorazepam over the 72 hours; all 4 of these were patients in the placebo group (P = 0.023).

Figure 2

Figure 3

Subjects who received baclofen and subjects who received placebo did not differ significantly in regard to the use of sedatives other than benzodiazepines, the use of restraints, the use of intensive care, or other clinical complications. On the basis this analysis, the study's primary null hypothesis was not rejected (the CIWA‐Ar scores in the baclofen and placebo groups were not different), but the secondary null hypothesis was rejected (baclofen was associated with lower likelihood of the use of high doses of benzodiazepines).

DISCUSSION

Benzodiazepines are effective drugs in the treatment of alcohol withdrawal syndrome. They remain the gold standard of treatment. The most effective method of administering benzodiazepines to acutely withdrawing patients has been shown to be variable dosing, based on withdrawal symptoms.

Our small study of acutely withdrawing inpatients confirmed that benzodiazepines, administered at frequencies and doses dependent on AWS symptoms, work well to control CIWA‐Ar scores over a relatively short time span. Our study also demonstrated that the addition of the GABA‐B agonist baclofen orally, at a fixed dose of 30 mg daily, will allow the same level of AWS symptom control, while reducing the risk that high doses of benzodiazepines will be needed to achieve that control. Reducing the use of high‐dose benzodiazepines has the potential to improve patient safety. In addition, since the frequency of nursing assessments parallel CIWA‐AR scores and benzodiazepine dosing frequency, using oral baclofen in this setting has the potential to decrease the nursing time required to control withdrawal symptoms.

A larger study of AWS will be needed to assess the role of baclofen in managing the frequency and severity of the complications of AWS, such as prolonged sedation, intensive care admission, and ventilator days. The current study was not powered to assess differences in the frequency of relatively rare events, and we excluded patients who required intravenous benzodiazepines for AWS symptom management. However, in light of the well‐documented risk of sedation and respiratory depression from high‐dose benzodiazepines, our findings suggest that a larger study of the role of baclofen in AWS management is warranted.

Either baclofen or benzodiazepines may have severe adverse effects in high doses. In this study, we used a low, fixed dose of baclofen (10 mg every 8 hours), a level at which severe side effects such as respiratory depression are uncommon. Our principal finding was that the use of low‐dose baclofen is associated with reduced use of high‐dose benzodiazepines in some AWS patients.

CONCLUSION

These findings suggest that baclofen may have potential as an adjunct in the management of acute alcohol withdrawal.

In the United States, unhealthy alcohol use affects medical care on several levels. The prevalence of alcohol problems is 7%20% or higher among outpatients,1 30%40% among emergency room patients, and 50% among patients with trauma.13 In 2006, approximately 430,000 hospital discharges in the United States were for persons with a principal (first‐listed) alcohol‐related diagnosis, and 1.7 million discharges listed at least one alcohol‐related diagnosis, representing 1.3% and 5.3% of all hospital discharges, respectively.4 Alcoholic psychosis (34.5%) and alcohol dependence syndrome (29.5%) together accounted for the majority of principal alcohol‐related diagnoses.4 Additionally, many patients hospitalized for other indications are susceptible to withdrawal symptoms due to physiological habituation to alcohol. Abrupt cessation of alcohol intake causes habituated drinkers to experience symptoms of alcohol withdrawal syndrome (AWS), which significantly increases intensity and cost of care. Trauma patients who develop AWS were found to have increased morbidity, more intensive care and ventilator days, and higher hospital costs than trauma patients without AWS.5 Unfortunately, attempts to develop predictive models to accurately forecast the likelihood of developing severe AWS in an individual case have been modestly successful at best.68

Regimens used to treat AWS have evolved over time, taking advantage of advances in the understanding of addiction neurophysiology. There is no specific ethanol receptor.9 Much of alcohol's acute effects on the central nervous system are mediated by its stimulation of the gamma‐aminobutyric acid (GABA) system, which is neuroinhibitory.10 Chronic alcohol use leads to habituation partly by inducing configuration changes of GABA‐A receptor subunits. This renders the GABA‐A receptor less sensitive to alcohol, barbiturates, and benzodiazepines.11 Although both GABA‐A and GABA‐B receptor activation cause increased GABA neuronal output, the GABA‐A receptor is rendered relatively less sensitive by chronic exposure to alcohol. Baclofen is a pure GABA‐B receptor agonist,12 and its GABA‐B stimulatory effect is maintained even in habituated alcoholics.13, 14 The absence of cross‐tolerance between baclofen and ethanol suggests that low doses of baclofen may be helpful in the management of AWS.

Currently, AWS is usually managed with benzodiazepines, using variable dosing depending on the severity of withdrawal symptoms. Such symptom‐triggered treatment is generally preferred over fixed‐dose regimens,15 in part because when using this method, many cases of AWS can be managed with less medication. Benzodiazepine regimens using high doses have been found to be associated with substantial morbidity and prolonged hospitalizations.16, 17

In a series of small studies, Addolorato's research team has reported decreases in AWS symptoms in association with the use of low doses of baclofen in an outpatient population,18 and has found baclofen to be associated with reduced alcohol craving in the long‐term management of alcohol dependence.11, 19, 20 Addolorato and colleagues' studies of baclofen in relieving AWS symptoms prompted our group to apply the use of baclofen in a larger group of inpatients with AWS.11, 18, 21 We conducted this study to improve understanding of the role of baclofen in the management of acute AWS in an inpatient population of subjects at risk for AWS, drawn from general hospital admissions.

Our primary null hypothesis was that Clinical Institute Withdrawal Assessment of Alcohol Scale, Revised (CIWA‐Ar) scores in acutely withdrawing alcoholic patients are equal in baclofen‐treated and placebo‐treated subject groups. Our secondary null hypothesis was that benzodiazepine doses used to treat acutely withdrawing alcoholic patients are equal in the baclofen‐treated and placebo‐treated groups.

METHODS

The protocol for this study was approved by the Essentia Health Institutional Review Board, Duluth, Minnesota.

This was a randomized, placebo‐controlled, double‐blind trial. Subjects were recruited from among patients who were admitted to 1 of 2 regional general hospitals in Duluth, Minnesota (St. Mary's Medical Center or Miller‐Dwan Medical Center) and who were identified by clinical staff as being at risk for AWS. Potential subjects were not required to have an alcohol‐related condition as their primary reason for admission, but were required to have a history of AWS or of alcohol use suggestive of significant risk for AWS, and to be able to provide informed consent, as described below.

Patients were not eligible for enrollment in the study if they had other active drug dependence in addition to alcohol; were using baclofen at the time of study enrollment; were using benzodiazepines chronically at the time of study enrollment; had a known baclofen or benzodiazepine sensitivity; were unable to take oral medications; were pregnant or breast‐feeding; had a serum creatinine level 2.0; had a history of non‐alcohol withdrawal seizures; required intravenous benzodiazepines to control their AWS; or were unable to complete the consenting procedures.

RESULTS

Seventy‐nine subjects met study inclusion criteria, and provided informed consent for participation in the study. Of these, 44 subjects developed signs of AWS sufficient to meet DSMIV criteria for AWS diagnosis, and were randomized to receive either baclofen or placebo, in addition to benzodiazepine therapy. As summarized in Table 1, subjects who developed signs of AWS were similar to subjects who did not enter withdrawal, differing principally in that those who developed AWS reported more drinks per day, and more significant history of previous AWS.

The 79 subjects who were enrolled were drawn from a population of 237 potential subjects who were screened for the study. The most common reasons that the 158 potential subjects were not enrolled were refusal (29.7%), low risk of withdrawal (19.0%), inability to provide consent (9.5%), and concurrent use of benzodiazepines (8.9%). The 15 patients who were unable to provide consent were those who scored 24 or lower on the MMSE, and did not have a surrogate decision‐maker available.

Of the 44 subjects who were randomized, 31 (18 in the baclofen group, 13 in the placebo group) completed 72 hours of CIWA‐Ar assessments, as summarized in Table 2. These assessments were completed either entirely as inpatients (24 subjects) or with inpatient assessments followed by outpatient assessments after discharge (7 subjects). Discharges prior to 72 hours occurred in 3 of the 18 subjects receiving baclofen, and in 4 of the 13 receiving placebo (odds ratio = 0.45, 95% CI = 0.082.49). Mean CIWA‐Ar scores for the 31 subjects who completed 72 hours of CIWA‐Ar assessments are presented in Figure 1.

Figure 1

Figures 2 and 3 summarize the mean lorazepam doses in each 8‐hour period, and the cumulative lorazepam doses for the subjects in the 2 arms of the study, respectively. The cumulative dose of lorazepam administered to the 31 subjects ranged from 0 to 1035 mg in the 72 hours following randomization, with a range of 1 to 1035 mg in the placebo group and 0 to 39 mg in the baclofen group. The 8 subjects who received the highest doses of lorazepam (20 mg or more) included 1 of the 18 subjects who received baclofen and 7 of the 13 subjects who received placebo (P = 0.004). Only 4 subjects required >50 mg of lorazepam over the 72 hours; all 4 of these were patients in the placebo group (P = 0.023).

Figure 2

Figure 3

Subjects who received baclofen and subjects who received placebo did not differ significantly in regard to the use of sedatives other than benzodiazepines, the use of restraints, the use of intensive care, or other clinical complications. On the basis this analysis, the study's primary null hypothesis was not rejected (the CIWA‐Ar scores in the baclofen and placebo groups were not different), but the secondary null hypothesis was rejected (baclofen was associated with lower likelihood of the use of high doses of benzodiazepines).

DISCUSSION

Benzodiazepines are effective drugs in the treatment of alcohol withdrawal syndrome. They remain the gold standard of treatment. The most effective method of administering benzodiazepines to acutely withdrawing patients has been shown to be variable dosing, based on withdrawal symptoms.

Our small study of acutely withdrawing inpatients confirmed that benzodiazepines, administered at frequencies and doses dependent on AWS symptoms, work well to control CIWA‐Ar scores over a relatively short time span. Our study also demonstrated that the addition of the GABA‐B agonist baclofen orally, at a fixed dose of 30 mg daily, will allow the same level of AWS symptom control, while reducing the risk that high doses of benzodiazepines will be needed to achieve that control. Reducing the use of high‐dose benzodiazepines has the potential to improve patient safety. In addition, since the frequency of nursing assessments parallel CIWA‐AR scores and benzodiazepine dosing frequency, using oral baclofen in this setting has the potential to decrease the nursing time required to control withdrawal symptoms.

A larger study of AWS will be needed to assess the role of baclofen in managing the frequency and severity of the complications of AWS, such as prolonged sedation, intensive care admission, and ventilator days. The current study was not powered to assess differences in the frequency of relatively rare events, and we excluded patients who required intravenous benzodiazepines for AWS symptom management. However, in light of the well‐documented risk of sedation and respiratory depression from high‐dose benzodiazepines, our findings suggest that a larger study of the role of baclofen in AWS management is warranted.

Either baclofen or benzodiazepines may have severe adverse effects in high doses. In this study, we used a low, fixed dose of baclofen (10 mg every 8 hours), a level at which severe side effects such as respiratory depression are uncommon. Our principal finding was that the use of low‐dose baclofen is associated with reduced use of high‐dose benzodiazepines in some AWS patients.

CONCLUSION

These findings suggest that baclofen may have potential as an adjunct in the management of acute alcohol withdrawal.

In the United States, unhealthy alcohol use affects medical care on several levels. The prevalence of alcohol problems is 7%20% or higher among outpatients,1 30%40% among emergency room patients, and 50% among patients with trauma.13 In 2006, approximately 430,000 hospital discharges in the United States were for persons with a principal (first‐listed) alcohol‐related diagnosis, and 1.7 million discharges listed at least one alcohol‐related diagnosis, representing 1.3% and 5.3% of all hospital discharges, respectively.4 Alcoholic psychosis (34.5%) and alcohol dependence syndrome (29.5%) together accounted for the majority of principal alcohol‐related diagnoses.4 Additionally, many patients hospitalized for other indications are susceptible to withdrawal symptoms due to physiological habituation to alcohol. Abrupt cessation of alcohol intake causes habituated drinkers to experience symptoms of alcohol withdrawal syndrome (AWS), which significantly increases intensity and cost of care. Trauma patients who develop AWS were found to have increased morbidity, more intensive care and ventilator days, and higher hospital costs than trauma patients without AWS.5 Unfortunately, attempts to develop predictive models to accurately forecast the likelihood of developing severe AWS in an individual case have been modestly successful at best.68

Regimens used to treat AWS have evolved over time, taking advantage of advances in the understanding of addiction neurophysiology. There is no specific ethanol receptor.9 Much of alcohol's acute effects on the central nervous system are mediated by its stimulation of the gamma‐aminobutyric acid (GABA) system, which is neuroinhibitory.10 Chronic alcohol use leads to habituation partly by inducing configuration changes of GABA‐A receptor subunits. This renders the GABA‐A receptor less sensitive to alcohol, barbiturates, and benzodiazepines.11 Although both GABA‐A and GABA‐B receptor activation cause increased GABA neuronal output, the GABA‐A receptor is rendered relatively less sensitive by chronic exposure to alcohol. Baclofen is a pure GABA‐B receptor agonist,12 and its GABA‐B stimulatory effect is maintained even in habituated alcoholics.13, 14 The absence of cross‐tolerance between baclofen and ethanol suggests that low doses of baclofen may be helpful in the management of AWS.

Currently, AWS is usually managed with benzodiazepines, using variable dosing depending on the severity of withdrawal symptoms. Such symptom‐triggered treatment is generally preferred over fixed‐dose regimens,15 in part because when using this method, many cases of AWS can be managed with less medication. Benzodiazepine regimens using high doses have been found to be associated with substantial morbidity and prolonged hospitalizations.16, 17

In a series of small studies, Addolorato's research team has reported decreases in AWS symptoms in association with the use of low doses of baclofen in an outpatient population,18 and has found baclofen to be associated with reduced alcohol craving in the long‐term management of alcohol dependence.11, 19, 20 Addolorato and colleagues' studies of baclofen in relieving AWS symptoms prompted our group to apply the use of baclofen in a larger group of inpatients with AWS.11, 18, 21 We conducted this study to improve understanding of the role of baclofen in the management of acute AWS in an inpatient population of subjects at risk for AWS, drawn from general hospital admissions.

Our primary null hypothesis was that Clinical Institute Withdrawal Assessment of Alcohol Scale, Revised (CIWA‐Ar) scores in acutely withdrawing alcoholic patients are equal in baclofen‐treated and placebo‐treated subject groups. Our secondary null hypothesis was that benzodiazepine doses used to treat acutely withdrawing alcoholic patients are equal in the baclofen‐treated and placebo‐treated groups.

METHODS

The protocol for this study was approved by the Essentia Health Institutional Review Board, Duluth, Minnesota.

This was a randomized, placebo‐controlled, double‐blind trial. Subjects were recruited from among patients who were admitted to 1 of 2 regional general hospitals in Duluth, Minnesota (St. Mary's Medical Center or Miller‐Dwan Medical Center) and who were identified by clinical staff as being at risk for AWS. Potential subjects were not required to have an alcohol‐related condition as their primary reason for admission, but were required to have a history of AWS or of alcohol use suggestive of significant risk for AWS, and to be able to provide informed consent, as described below.

Patients were not eligible for enrollment in the study if they had other active drug dependence in addition to alcohol; were using baclofen at the time of study enrollment; were using benzodiazepines chronically at the time of study enrollment; had a known baclofen or benzodiazepine sensitivity; were unable to take oral medications; were pregnant or breast‐feeding; had a serum creatinine level 2.0; had a history of non‐alcohol withdrawal seizures; required intravenous benzodiazepines to control their AWS; or were unable to complete the consenting procedures.

RESULTS

Seventy‐nine subjects met study inclusion criteria, and provided informed consent for participation in the study. Of these, 44 subjects developed signs of AWS sufficient to meet DSMIV criteria for AWS diagnosis, and were randomized to receive either baclofen or placebo, in addition to benzodiazepine therapy. As summarized in Table 1, subjects who developed signs of AWS were similar to subjects who did not enter withdrawal, differing principally in that those who developed AWS reported more drinks per day, and more significant history of previous AWS.

The 79 subjects who were enrolled were drawn from a population of 237 potential subjects who were screened for the study. The most common reasons that the 158 potential subjects were not enrolled were refusal (29.7%), low risk of withdrawal (19.0%), inability to provide consent (9.5%), and concurrent use of benzodiazepines (8.9%). The 15 patients who were unable to provide consent were those who scored 24 or lower on the MMSE, and did not have a surrogate decision‐maker available.

Of the 44 subjects who were randomized, 31 (18 in the baclofen group, 13 in the placebo group) completed 72 hours of CIWA‐Ar assessments, as summarized in Table 2. These assessments were completed either entirely as inpatients (24 subjects) or with inpatient assessments followed by outpatient assessments after discharge (7 subjects). Discharges prior to 72 hours occurred in 3 of the 18 subjects receiving baclofen, and in 4 of the 13 receiving placebo (odds ratio = 0.45, 95% CI = 0.082.49). Mean CIWA‐Ar scores for the 31 subjects who completed 72 hours of CIWA‐Ar assessments are presented in Figure 1.

Figure 1

Figures 2 and 3 summarize the mean lorazepam doses in each 8‐hour period, and the cumulative lorazepam doses for the subjects in the 2 arms of the study, respectively. The cumulative dose of lorazepam administered to the 31 subjects ranged from 0 to 1035 mg in the 72 hours following randomization, with a range of 1 to 1035 mg in the placebo group and 0 to 39 mg in the baclofen group. The 8 subjects who received the highest doses of lorazepam (20 mg or more) included 1 of the 18 subjects who received baclofen and 7 of the 13 subjects who received placebo (P = 0.004). Only 4 subjects required >50 mg of lorazepam over the 72 hours; all 4 of these were patients in the placebo group (P = 0.023).

Figure 2

Figure 3

Subjects who received baclofen and subjects who received placebo did not differ significantly in regard to the use of sedatives other than benzodiazepines, the use of restraints, the use of intensive care, or other clinical complications. On the basis this analysis, the study's primary null hypothesis was not rejected (the CIWA‐Ar scores in the baclofen and placebo groups were not different), but the secondary null hypothesis was rejected (baclofen was associated with lower likelihood of the use of high doses of benzodiazepines).

DISCUSSION

Benzodiazepines are effective drugs in the treatment of alcohol withdrawal syndrome. They remain the gold standard of treatment. The most effective method of administering benzodiazepines to acutely withdrawing patients has been shown to be variable dosing, based on withdrawal symptoms.

Our small study of acutely withdrawing inpatients confirmed that benzodiazepines, administered at frequencies and doses dependent on AWS symptoms, work well to control CIWA‐Ar scores over a relatively short time span. Our study also demonstrated that the addition of the GABA‐B agonist baclofen orally, at a fixed dose of 30 mg daily, will allow the same level of AWS symptom control, while reducing the risk that high doses of benzodiazepines will be needed to achieve that control. Reducing the use of high‐dose benzodiazepines has the potential to improve patient safety. In addition, since the frequency of nursing assessments parallel CIWA‐AR scores and benzodiazepine dosing frequency, using oral baclofen in this setting has the potential to decrease the nursing time required to control withdrawal symptoms.

A larger study of AWS will be needed to assess the role of baclofen in managing the frequency and severity of the complications of AWS, such as prolonged sedation, intensive care admission, and ventilator days. The current study was not powered to assess differences in the frequency of relatively rare events, and we excluded patients who required intravenous benzodiazepines for AWS symptom management. However, in light of the well‐documented risk of sedation and respiratory depression from high‐dose benzodiazepines, our findings suggest that a larger study of the role of baclofen in AWS management is warranted.

Either baclofen or benzodiazepines may have severe adverse effects in high doses. In this study, we used a low, fixed dose of baclofen (10 mg every 8 hours), a level at which severe side effects such as respiratory depression are uncommon. Our principal finding was that the use of low‐dose baclofen is associated with reduced use of high‐dose benzodiazepines in some AWS patients.

CONCLUSION

These findings suggest that baclofen may have potential as an adjunct in the management of acute alcohol withdrawal.

The tension between continuity of care and specialization is not new, but may have reached a tipping point when the hospitalist movement erupted onto the American medical scene in the late 1990s. By definition, when a hospitalist cares for an inpatient, there is some fragmentation of care, which is, at least in theory, avoidableif the primary care provider (PCP) can serve as attending physician in the hospital. Literature has since emerged suggesting that clinical and economic outcomes of care by hospitalists are at least as good as that provided by PCPs, and that patients are not, in general, opposed to hospitalist care.13

However, the degree of discontinuity is not just a feature of whether a hospitalist assumes care of the hospitalized patient. Discontinuity can be exacerbated by changing attendings throughout the hospital stay. And inpatient continuity is a potential issue for both the hospitalist model and traditional model of care (in which the PCP serves as inpatient attending physician). While one might assume that the hospitalist model fosters more inpatient discontinuity because most hospitalistswhether working a 7‐on7‐off schedule or another scheduledo not commit to caring for a patient throughout an entire hospitalization the way a PCP might, this question has not previously been examined. Even if the hospitalist model is a fait accompli in many hospitals, it is worth knowing how inpatient continuity differs between the 2 models.

In this issue of the Journal, Fletcher and colleagues4 used billing data to examine trends in inpatient continuity of care over a 10‐year period ending in 2006, and sought to determine: (1) whether inpatient care has become more fragmented over time (as defined by the number of generalists caring for a patient over the course of an average hospitalization), and (2) whether inpatient care provided by hospitalists tends to be more fragmented than care provided by PCPs. They found that continuity of inpatient care has indeed decreased over time. In 1996, just over 70% of patients received care from 1 generalist; this number declined to just under 60% a decade later, despite a decrease in length‐of‐stay during that period. However, and perhaps surprisingly, patients cared for exclusively by hospitalists saw fewer generalists in the hospital (ie, fewer different hospitalists) than those cared for exclusively by outpatient providers. The authors conclude that the doctorpatient continuity over the course of a hospital stay is not worse in the hospitalist model than in the traditional model. While reassuring, it is important to remember that the patient experience does not begin at admission or end at discharge, and a more patient‐centered analysis might take into account the outpatient providers too (ie, those seeing the patient before admission and after discharge), and would probably show that the hospitalist model indeed leads to more care fragmentation. After all, there are at least 2 providers involved in every patient's care when a hospitalist model is used, whereas a large subset of patients cared for by PCPs would have only 1 provider involved.

While not the primary focus of the analysis, Fletcher and colleagues4 identified additional predictors of inpatient continuity of care. Higher socioeconomic class and white race were associated with lower continuity. This suggests that care fragmentation is not a feature of inferior, or at least cheap, care. In keeping with this observation, there was also enormous geographic variation in inpatient care continuity, marked by greater fragmentation of care in the New England and the mid‐Atlantic regions than in other areas of the country, and more fragmentation in larger hospitals serving heavily populated metropolitan areas. This pattern is strikingly similar to the cost‐of‐care patterns observed by the Dartmouth Atlas researchers.5, 6 Densely populated areas tend to have more specialists per capita and also tend to deliver more expensive carewithout demonstrably higher quality. In parallel, it is easy to see how care fragmentation might increase length‐of‐stay7 and lead to excessive diagnostic testing and consultation. More cooks in the kitchen might make costlier stew.

How hospitalists tackle the issue of inpatient continuity is not only a matter of quality of care, but also a matter of job sustainability. The simple way to maximize continuityworking many consecutive dayscan lead to burnout if taken too far. But there are creative ways to assign admissions that maximize continuity for the average inpatient while allowing providers needed time off. The CICLE initiative (Creating Incentives and Continuity Leading to Efficiency in hospital medicine) at the Johns Hopkins Bayview Medical Center, for instance, assigns physicians to 4‐day cycles of clinical work; the first day of the cycle (a long‐call day) involves admitting a large number of patients during a busy shift, with no new patients admitted on the remaining days of the cycle. Thus, all patients whose length of stay is less than 5 days will have a single attending‐of‐record. Not only does this model increase continuity, it also incentivizes providers to augment throughput: more discharged patients on Tuesday means fewer patients to see on Wednesday, without any expectation to backfill. Other less aggressive but similar approaches are used elsewhere, such as exempting hospitalists from accepting new patients on the last 1 or 2 of the consecutive days they work. We eagerly await data on the impact of these programs on quality of care, patient satisfaction, and provider satisfaction.

The impact of other providers and staff cannot be ignored. While the most important handoff in many cases may indeed be between the PCP and attending hospitalist tasked with coordinating the overall care of the patient, for some patients, there may be a specialist who has known the patient for years who is driving the plan of care. For patients with severe chronic illnesses, such as end‐stage renal disease or asthma, a well‐structured specialty clinic may even serve as a patient‐centered medical home.8 And the current inpatient team includes night coverage physicians (whether moonlighters, house staff, or covering hospitalists), and an ever‐increasing number of non‐physicians who play a critical role in hospital care (non‐physician providers, nurses, social workers, pharmacists, case managers, physical therapists, and others). While it is tempting to focus on the attending physician as the main driver of healthcare quality, continuity, and the inpatient experience, this is an oversimplification.

If there is a take‐home message, it is probably that most hospitalized patients will be cared for by multiple providers and a team of non‐physicians. The Marcus Welby practice model may not be completely dead, but if Dr. Welby were still in practice, it would be a safe bet that he would be slower at computerized order entry than the average intern, that financial pressures would make it hard for him to attend to his hospitalized patients, and that he probably would have turned over much of his inpatient practice to the physicians and non‐physician caregivers who make the hospital their primary workplace.9 Going forward, research should examine ways to optimize care coordination under the hospitalist model,1013 rather than comparing it to the traditional model of inpatient care. The ingredients for success include coordinated care by a committee of caregivers, effective handoffs (throughout hospitalization and at discharge),12, 14 focused and deliberate multidisciplinary communication, and effective patient education,15 regardless of the attending‐du‐jour.

The tension between continuity of care and specialization is not new, but may have reached a tipping point when the hospitalist movement erupted onto the American medical scene in the late 1990s. By definition, when a hospitalist cares for an inpatient, there is some fragmentation of care, which is, at least in theory, avoidableif the primary care provider (PCP) can serve as attending physician in the hospital. Literature has since emerged suggesting that clinical and economic outcomes of care by hospitalists are at least as good as that provided by PCPs, and that patients are not, in general, opposed to hospitalist care.13

However, the degree of discontinuity is not just a feature of whether a hospitalist assumes care of the hospitalized patient. Discontinuity can be exacerbated by changing attendings throughout the hospital stay. And inpatient continuity is a potential issue for both the hospitalist model and traditional model of care (in which the PCP serves as inpatient attending physician). While one might assume that the hospitalist model fosters more inpatient discontinuity because most hospitalistswhether working a 7‐on7‐off schedule or another scheduledo not commit to caring for a patient throughout an entire hospitalization the way a PCP might, this question has not previously been examined. Even if the hospitalist model is a fait accompli in many hospitals, it is worth knowing how inpatient continuity differs between the 2 models.

In this issue of the Journal, Fletcher and colleagues4 used billing data to examine trends in inpatient continuity of care over a 10‐year period ending in 2006, and sought to determine: (1) whether inpatient care has become more fragmented over time (as defined by the number of generalists caring for a patient over the course of an average hospitalization), and (2) whether inpatient care provided by hospitalists tends to be more fragmented than care provided by PCPs. They found that continuity of inpatient care has indeed decreased over time. In 1996, just over 70% of patients received care from 1 generalist; this number declined to just under 60% a decade later, despite a decrease in length‐of‐stay during that period. However, and perhaps surprisingly, patients cared for exclusively by hospitalists saw fewer generalists in the hospital (ie, fewer different hospitalists) than those cared for exclusively by outpatient providers. The authors conclude that the doctorpatient continuity over the course of a hospital stay is not worse in the hospitalist model than in the traditional model. While reassuring, it is important to remember that the patient experience does not begin at admission or end at discharge, and a more patient‐centered analysis might take into account the outpatient providers too (ie, those seeing the patient before admission and after discharge), and would probably show that the hospitalist model indeed leads to more care fragmentation. After all, there are at least 2 providers involved in every patient's care when a hospitalist model is used, whereas a large subset of patients cared for by PCPs would have only 1 provider involved.

While not the primary focus of the analysis, Fletcher and colleagues4 identified additional predictors of inpatient continuity of care. Higher socioeconomic class and white race were associated with lower continuity. This suggests that care fragmentation is not a feature of inferior, or at least cheap, care. In keeping with this observation, there was also enormous geographic variation in inpatient care continuity, marked by greater fragmentation of care in the New England and the mid‐Atlantic regions than in other areas of the country, and more fragmentation in larger hospitals serving heavily populated metropolitan areas. This pattern is strikingly similar to the cost‐of‐care patterns observed by the Dartmouth Atlas researchers.5, 6 Densely populated areas tend to have more specialists per capita and also tend to deliver more expensive carewithout demonstrably higher quality. In parallel, it is easy to see how care fragmentation might increase length‐of‐stay7 and lead to excessive diagnostic testing and consultation. More cooks in the kitchen might make costlier stew.

How hospitalists tackle the issue of inpatient continuity is not only a matter of quality of care, but also a matter of job sustainability. The simple way to maximize continuityworking many consecutive dayscan lead to burnout if taken too far. But there are creative ways to assign admissions that maximize continuity for the average inpatient while allowing providers needed time off. The CICLE initiative (Creating Incentives and Continuity Leading to Efficiency in hospital medicine) at the Johns Hopkins Bayview Medical Center, for instance, assigns physicians to 4‐day cycles of clinical work; the first day of the cycle (a long‐call day) involves admitting a large number of patients during a busy shift, with no new patients admitted on the remaining days of the cycle. Thus, all patients whose length of stay is less than 5 days will have a single attending‐of‐record. Not only does this model increase continuity, it also incentivizes providers to augment throughput: more discharged patients on Tuesday means fewer patients to see on Wednesday, without any expectation to backfill. Other less aggressive but similar approaches are used elsewhere, such as exempting hospitalists from accepting new patients on the last 1 or 2 of the consecutive days they work. We eagerly await data on the impact of these programs on quality of care, patient satisfaction, and provider satisfaction.

The impact of other providers and staff cannot be ignored. While the most important handoff in many cases may indeed be between the PCP and attending hospitalist tasked with coordinating the overall care of the patient, for some patients, there may be a specialist who has known the patient for years who is driving the plan of care. For patients with severe chronic illnesses, such as end‐stage renal disease or asthma, a well‐structured specialty clinic may even serve as a patient‐centered medical home.8 And the current inpatient team includes night coverage physicians (whether moonlighters, house staff, or covering hospitalists), and an ever‐increasing number of non‐physicians who play a critical role in hospital care (non‐physician providers, nurses, social workers, pharmacists, case managers, physical therapists, and others). While it is tempting to focus on the attending physician as the main driver of healthcare quality, continuity, and the inpatient experience, this is an oversimplification.

If there is a take‐home message, it is probably that most hospitalized patients will be cared for by multiple providers and a team of non‐physicians. The Marcus Welby practice model may not be completely dead, but if Dr. Welby were still in practice, it would be a safe bet that he would be slower at computerized order entry than the average intern, that financial pressures would make it hard for him to attend to his hospitalized patients, and that he probably would have turned over much of his inpatient practice to the physicians and non‐physician caregivers who make the hospital their primary workplace.9 Going forward, research should examine ways to optimize care coordination under the hospitalist model,1013 rather than comparing it to the traditional model of inpatient care. The ingredients for success include coordinated care by a committee of caregivers, effective handoffs (throughout hospitalization and at discharge),12, 14 focused and deliberate multidisciplinary communication, and effective patient education,15 regardless of the attending‐du‐jour.

The tension between continuity of care and specialization is not new, but may have reached a tipping point when the hospitalist movement erupted onto the American medical scene in the late 1990s. By definition, when a hospitalist cares for an inpatient, there is some fragmentation of care, which is, at least in theory, avoidableif the primary care provider (PCP) can serve as attending physician in the hospital. Literature has since emerged suggesting that clinical and economic outcomes of care by hospitalists are at least as good as that provided by PCPs, and that patients are not, in general, opposed to hospitalist care.13

However, the degree of discontinuity is not just a feature of whether a hospitalist assumes care of the hospitalized patient. Discontinuity can be exacerbated by changing attendings throughout the hospital stay. And inpatient continuity is a potential issue for both the hospitalist model and traditional model of care (in which the PCP serves as inpatient attending physician). While one might assume that the hospitalist model fosters more inpatient discontinuity because most hospitalistswhether working a 7‐on7‐off schedule or another scheduledo not commit to caring for a patient throughout an entire hospitalization the way a PCP might, this question has not previously been examined. Even if the hospitalist model is a fait accompli in many hospitals, it is worth knowing how inpatient continuity differs between the 2 models.

In this issue of the Journal, Fletcher and colleagues4 used billing data to examine trends in inpatient continuity of care over a 10‐year period ending in 2006, and sought to determine: (1) whether inpatient care has become more fragmented over time (as defined by the number of generalists caring for a patient over the course of an average hospitalization), and (2) whether inpatient care provided by hospitalists tends to be more fragmented than care provided by PCPs. They found that continuity of inpatient care has indeed decreased over time. In 1996, just over 70% of patients received care from 1 generalist; this number declined to just under 60% a decade later, despite a decrease in length‐of‐stay during that period. However, and perhaps surprisingly, patients cared for exclusively by hospitalists saw fewer generalists in the hospital (ie, fewer different hospitalists) than those cared for exclusively by outpatient providers. The authors conclude that the doctorpatient continuity over the course of a hospital stay is not worse in the hospitalist model than in the traditional model. While reassuring, it is important to remember that the patient experience does not begin at admission or end at discharge, and a more patient‐centered analysis might take into account the outpatient providers too (ie, those seeing the patient before admission and after discharge), and would probably show that the hospitalist model indeed leads to more care fragmentation. After all, there are at least 2 providers involved in every patient's care when a hospitalist model is used, whereas a large subset of patients cared for by PCPs would have only 1 provider involved.

While not the primary focus of the analysis, Fletcher and colleagues4 identified additional predictors of inpatient continuity of care. Higher socioeconomic class and white race were associated with lower continuity. This suggests that care fragmentation is not a feature of inferior, or at least cheap, care. In keeping with this observation, there was also enormous geographic variation in inpatient care continuity, marked by greater fragmentation of care in the New England and the mid‐Atlantic regions than in other areas of the country, and more fragmentation in larger hospitals serving heavily populated metropolitan areas. This pattern is strikingly similar to the cost‐of‐care patterns observed by the Dartmouth Atlas researchers.5, 6 Densely populated areas tend to have more specialists per capita and also tend to deliver more expensive carewithout demonstrably higher quality. In parallel, it is easy to see how care fragmentation might increase length‐of‐stay7 and lead to excessive diagnostic testing and consultation. More cooks in the kitchen might make costlier stew.

How hospitalists tackle the issue of inpatient continuity is not only a matter of quality of care, but also a matter of job sustainability. The simple way to maximize continuityworking many consecutive dayscan lead to burnout if taken too far. But there are creative ways to assign admissions that maximize continuity for the average inpatient while allowing providers needed time off. The CICLE initiative (Creating Incentives and Continuity Leading to Efficiency in hospital medicine) at the Johns Hopkins Bayview Medical Center, for instance, assigns physicians to 4‐day cycles of clinical work; the first day of the cycle (a long‐call day) involves admitting a large number of patients during a busy shift, with no new patients admitted on the remaining days of the cycle. Thus, all patients whose length of stay is less than 5 days will have a single attending‐of‐record. Not only does this model increase continuity, it also incentivizes providers to augment throughput: more discharged patients on Tuesday means fewer patients to see on Wednesday, without any expectation to backfill. Other less aggressive but similar approaches are used elsewhere, such as exempting hospitalists from accepting new patients on the last 1 or 2 of the consecutive days they work. We eagerly await data on the impact of these programs on quality of care, patient satisfaction, and provider satisfaction.

The impact of other providers and staff cannot be ignored. While the most important handoff in many cases may indeed be between the PCP and attending hospitalist tasked with coordinating the overall care of the patient, for some patients, there may be a specialist who has known the patient for years who is driving the plan of care. For patients with severe chronic illnesses, such as end‐stage renal disease or asthma, a well‐structured specialty clinic may even serve as a patient‐centered medical home.8 And the current inpatient team includes night coverage physicians (whether moonlighters, house staff, or covering hospitalists), and an ever‐increasing number of non‐physicians who play a critical role in hospital care (non‐physician providers, nurses, social workers, pharmacists, case managers, physical therapists, and others). While it is tempting to focus on the attending physician as the main driver of healthcare quality, continuity, and the inpatient experience, this is an oversimplification.

If there is a take‐home message, it is probably that most hospitalized patients will be cared for by multiple providers and a team of non‐physicians. The Marcus Welby practice model may not be completely dead, but if Dr. Welby were still in practice, it would be a safe bet that he would be slower at computerized order entry than the average intern, that financial pressures would make it hard for him to attend to his hospitalized patients, and that he probably would have turned over much of his inpatient practice to the physicians and non‐physician caregivers who make the hospital their primary workplace.9 Going forward, research should examine ways to optimize care coordination under the hospitalist model,1013 rather than comparing it to the traditional model of inpatient care. The ingredients for success include coordinated care by a committee of caregivers, effective handoffs (throughout hospitalization and at discharge),12, 14 focused and deliberate multidisciplinary communication, and effective patient education,15 regardless of the attending‐du‐jour.

Close supervision of residents potentially leads to fewer errors, lower patient mortality, and improved quality of care.19 An Institute of Medicine (IOM) report3 recommended improving supervision through more frequent consultations between residents and their supervisors. Although current Accreditation Council for Graduate Medical Education (ACGME) guidelines also recommend that attending physicians (attendings) supervise residents, detailed guidance about what constitutes adequate supervision and how it should be implemented is not well defined.10, 11 The ACGME stresses that supervision should promote resident autonomy in clinical care.10 However, when trainees act independently, it might lead to critical communication breakdowns and other patient safety concerns.5, 6, 1214 Although attendings can encourage (or discourage) residents from seeking advice,15, 16 residents also play important roles in asking for help (ie, initiating their own supervision).1719 Additional research is needed on how residents walk the fine line between exercising independence and seeking supervision.

Lack of resident supervision is especially problematic in high‐risk settings such as the medical intensive care unit (ICU), where medical errors are as frequent as 1.7 errors per patient per day,20, 21 and the adverse drug event rate is twice that of non‐ICU settings.22 Because medication errors are one of the most common errors residents make,23, 24 resident interactions with nursing and pharmacy staff may significantly influence medication safety in error‐prone ICUs.2529 Studies of traditional hierarchical supervision tend to overlook how interactions with other professionals influence resident training.12, 18, 30, 31

We define supervision as a process of providing trainees with monitoring, guidance, and feedback^9(p828) as they care for patients.3 Whereas traditionally, supervisors are identified by their positions of formal authority in the medical chain of command; we conceptualize supervision as a process in which professionals engaged in supervisory activities need not have formal authority over their trainees.

To examine how residents seek supervision through both the traditional medical hierarchical chain of command (including attendings, fellows and senior residents) and interprofessional communication channels (including nursing and pharmacy staff), we conducted a qualitative study of residents working in ICUs in three tertiary care hospitals. Using semi‐structured interviews, we asked residents to describe how they experienced supervision as they provided medications to patients. Two broad research questions guided data analysis:

• 1

  How do residents receive supervision from physicians in the traditional medical hierarchy?

• 2

  How do residents receive supervision from other professionals (ie, nurses, staff pharmacists, and clinical pharmacists)?

METHODS

RESULTS

Residents described how they were supervised not only by other physicians within the traditional medical hierarchy, but also by other professionals, including nurses, staff pharmacists, and clinical pharmacists, ie, interprofessional supervision (Figure 1). After presenting these results, we examine how physicians and other professionals used communication strategies during interprofessional supervision. Here we use the term residents to include trainees at all levels, from interns to upper‐level residents, and male pronouns for de‐identification.

Figure 1

DISCUSSION

We analyzed interviews of residents working in medical ICUs to understand their supervision experiences related to medication safety. Although residents espoused beliefs in seeking assistance from supervising physicians and articulated strategies for doing so, many experienced difficulties in initiating supervision through the traditional medical hierarchy. Some residents were embarrassed by their mistaken decisions; others were concerned that their questions would reflect poorly on them.

Residents also received interprofessional supervision from nurses and pharmacists, who proactively monitored, intervened in, and guided residents' decisions. Other professionals evaluated residents' decisions by comparing them to distinctive professional guidelines and routinely used deferential language when conveying their concerns. Residents, in turn, asked other professionals for assistance.

We posit that interprofessional supervision clearly meets an accepted definition of supervision.3, 9 Residents received monitoring, guidance and feedback^9(p828) from other professionals, who engaged in routine monitoring and in situation‐specific double‐checks of residents' clinical decisions, similar to those performed by supervising physicians.30 Moreover, other professionals demonstrated the ability to anticipate a doctor's strengths and weaknesses in particular clinical situations in order to maximize patient safety.^9(p829)

Our study results have implications for graduate medical education (GME) reform. First, trainees experienced supervision as a two‐way interaction.36 Residents balanced the countervailing pressures to act independently or to seek a supervising physician's advice, in part, by developing strategies for deciding when to ask questions. Kennedy et al. identified similar rhetorical strategies.18 By asking questions about their clinical decisions, residents requested that supervising physicians guide their work; thus, they proactively initiated and thereby enacted their own supervision. Fostering the conditions for initiating supervision is essential, especially given the association between lack of effective supervision and adverse outcomes.5, 6, 1214

Second, residents expressed contradictory expectations about seeking advice from supervising physicians. Some residents were wary of approaching attending physicians for fear of appearing incompetent or being ridiculed.12, 16, 18, 31 However, we found that other residents remained reluctant to seek advice despite simultaneously appreciating that attendings encouraged them to ask for assistance. Whereas the perceived approachability of supervising physicians was important,18, 19 our exploratory findings suggest that it may be a necessary, but not a sufficient, condition for creating a learning environment. Creating a supportive learning environmentin which residents feel comfortable in revealing their perceived shortcomings to supervising physicians3begins with cultural changes, such as building medical teams,6 but such changes can be slow to develop.

Third, interprofessional supervision offers a strategy for improving supervision. The ubiquitous involvement of nursing and pharmacy staff in monitoring and intervening in residents' medication‐related decisions could result in overlooking their unique contributions to resident supervision. Mindful that supervising physicians evaluate them, residents selectively sought nonjudgmental advice from professionals outside the medical hierarchy. Therefore, improving supervision could entail offering residents ready access to other professionals who can advise them, especially during late night hours when supervising physicians might not be present.17, 27

The importance of interprofessional supervision has not been adequately recognized and emphasized in GME. Our study findings, if supported by future research, highlight how interpersonal communication techniques could influence both interprofessional supervision and hierarchical supervision among physicians. Medical team training programs3739 emphasize developing skills, such as mutual performance monitoring,^40(p13) by training providers to raise and respond to potentially sensitive questions. Improving supervision by enhancing interpersonal communication skills may be important, not only when relative status differences are clear (ie, physician hierarchy), but also when status differences are ambiguous (ie, residents and other professionals). GME programs could consider incorporating these techniques into their formal curricula, as could programs for nursing and pharmacy staff.

Our study has several limitations. Because of the larger research project objectives, we focused on medication safety in medical ICU settings, where nurses and pharmacists may be especially vigilant and proactive in monitoring residents. Thus, our findings may be specific to medication issues and less relevant outside ICUs. We had a relatively small sample size and do not claim to generalize from it, although we believe it offers meaningful insights. We also did not continue enlarging our sample until reaching redundancy.^35(p202) Nevertheless, the purposeful random sample of residents produced rich information. Indeed, some study results are consistent with previous resident education research,18 adding validity to our findings. Although the interview protocol was not designed specifically to investigate supervision, the resulting interviews yielded abundant data containing residents' detailed descriptions of how they experienced supervision. Whereas we were careful to note whether particular perceptions were unique to one resident, or shared by others, we recognize that the value of residents' observations is assessed by the quality of the insights they provide, not necessarily by the number of residents who described the same experience.

In conclusion, we found that residents experienced difficulties in initiating traditional hierarchical supervision related to medication safety in the ICU. However, they reported ubiquitous interprofessional supervision, albeit limited in scope, which they relied upon for nonjudgmental guidance in their therapeutic decision‐making, especially after‐hours. In our study, interprofessional supervision proved crucial to improving medication safety in the ICU.

Close supervision of residents potentially leads to fewer errors, lower patient mortality, and improved quality of care.19 An Institute of Medicine (IOM) report3 recommended improving supervision through more frequent consultations between residents and their supervisors. Although current Accreditation Council for Graduate Medical Education (ACGME) guidelines also recommend that attending physicians (attendings) supervise residents, detailed guidance about what constitutes adequate supervision and how it should be implemented is not well defined.10, 11 The ACGME stresses that supervision should promote resident autonomy in clinical care.10 However, when trainees act independently, it might lead to critical communication breakdowns and other patient safety concerns.5, 6, 1214 Although attendings can encourage (or discourage) residents from seeking advice,15, 16 residents also play important roles in asking for help (ie, initiating their own supervision).1719 Additional research is needed on how residents walk the fine line between exercising independence and seeking supervision.

Lack of resident supervision is especially problematic in high‐risk settings such as the medical intensive care unit (ICU), where medical errors are as frequent as 1.7 errors per patient per day,20, 21 and the adverse drug event rate is twice that of non‐ICU settings.22 Because medication errors are one of the most common errors residents make,23, 24 resident interactions with nursing and pharmacy staff may significantly influence medication safety in error‐prone ICUs.2529 Studies of traditional hierarchical supervision tend to overlook how interactions with other professionals influence resident training.12, 18, 30, 31

We define supervision as a process of providing trainees with monitoring, guidance, and feedback^9(p828) as they care for patients.3 Whereas traditionally, supervisors are identified by their positions of formal authority in the medical chain of command; we conceptualize supervision as a process in which professionals engaged in supervisory activities need not have formal authority over their trainees.

To examine how residents seek supervision through both the traditional medical hierarchical chain of command (including attendings, fellows and senior residents) and interprofessional communication channels (including nursing and pharmacy staff), we conducted a qualitative study of residents working in ICUs in three tertiary care hospitals. Using semi‐structured interviews, we asked residents to describe how they experienced supervision as they provided medications to patients. Two broad research questions guided data analysis:

• 1

  How do residents receive supervision from physicians in the traditional medical hierarchy?

• 2

  How do residents receive supervision from other professionals (ie, nurses, staff pharmacists, and clinical pharmacists)?

METHODS

RESULTS

Residents described how they were supervised not only by other physicians within the traditional medical hierarchy, but also by other professionals, including nurses, staff pharmacists, and clinical pharmacists, ie, interprofessional supervision (Figure 1). After presenting these results, we examine how physicians and other professionals used communication strategies during interprofessional supervision. Here we use the term residents to include trainees at all levels, from interns to upper‐level residents, and male pronouns for de‐identification.

Figure 1

DISCUSSION

We analyzed interviews of residents working in medical ICUs to understand their supervision experiences related to medication safety. Although residents espoused beliefs in seeking assistance from supervising physicians and articulated strategies for doing so, many experienced difficulties in initiating supervision through the traditional medical hierarchy. Some residents were embarrassed by their mistaken decisions; others were concerned that their questions would reflect poorly on them.

Residents also received interprofessional supervision from nurses and pharmacists, who proactively monitored, intervened in, and guided residents' decisions. Other professionals evaluated residents' decisions by comparing them to distinctive professional guidelines and routinely used deferential language when conveying their concerns. Residents, in turn, asked other professionals for assistance.

We posit that interprofessional supervision clearly meets an accepted definition of supervision.3, 9 Residents received monitoring, guidance and feedback^9(p828) from other professionals, who engaged in routine monitoring and in situation‐specific double‐checks of residents' clinical decisions, similar to those performed by supervising physicians.30 Moreover, other professionals demonstrated the ability to anticipate a doctor's strengths and weaknesses in particular clinical situations in order to maximize patient safety.^9(p829)

Our study results have implications for graduate medical education (GME) reform. First, trainees experienced supervision as a two‐way interaction.36 Residents balanced the countervailing pressures to act independently or to seek a supervising physician's advice, in part, by developing strategies for deciding when to ask questions. Kennedy et al. identified similar rhetorical strategies.18 By asking questions about their clinical decisions, residents requested that supervising physicians guide their work; thus, they proactively initiated and thereby enacted their own supervision. Fostering the conditions for initiating supervision is essential, especially given the association between lack of effective supervision and adverse outcomes.5, 6, 1214

Second, residents expressed contradictory expectations about seeking advice from supervising physicians. Some residents were wary of approaching attending physicians for fear of appearing incompetent or being ridiculed.12, 16, 18, 31 However, we found that other residents remained reluctant to seek advice despite simultaneously appreciating that attendings encouraged them to ask for assistance. Whereas the perceived approachability of supervising physicians was important,18, 19 our exploratory findings suggest that it may be a necessary, but not a sufficient, condition for creating a learning environment. Creating a supportive learning environmentin which residents feel comfortable in revealing their perceived shortcomings to supervising physicians3begins with cultural changes, such as building medical teams,6 but such changes can be slow to develop.

Third, interprofessional supervision offers a strategy for improving supervision. The ubiquitous involvement of nursing and pharmacy staff in monitoring and intervening in residents' medication‐related decisions could result in overlooking their unique contributions to resident supervision. Mindful that supervising physicians evaluate them, residents selectively sought nonjudgmental advice from professionals outside the medical hierarchy. Therefore, improving supervision could entail offering residents ready access to other professionals who can advise them, especially during late night hours when supervising physicians might not be present.17, 27

The importance of interprofessional supervision has not been adequately recognized and emphasized in GME. Our study findings, if supported by future research, highlight how interpersonal communication techniques could influence both interprofessional supervision and hierarchical supervision among physicians. Medical team training programs3739 emphasize developing skills, such as mutual performance monitoring,^40(p13) by training providers to raise and respond to potentially sensitive questions. Improving supervision by enhancing interpersonal communication skills may be important, not only when relative status differences are clear (ie, physician hierarchy), but also when status differences are ambiguous (ie, residents and other professionals). GME programs could consider incorporating these techniques into their formal curricula, as could programs for nursing and pharmacy staff.

Our study has several limitations. Because of the larger research project objectives, we focused on medication safety in medical ICU settings, where nurses and pharmacists may be especially vigilant and proactive in monitoring residents. Thus, our findings may be specific to medication issues and less relevant outside ICUs. We had a relatively small sample size and do not claim to generalize from it, although we believe it offers meaningful insights. We also did not continue enlarging our sample until reaching redundancy.^35(p202) Nevertheless, the purposeful random sample of residents produced rich information. Indeed, some study results are consistent with previous resident education research,18 adding validity to our findings. Although the interview protocol was not designed specifically to investigate supervision, the resulting interviews yielded abundant data containing residents' detailed descriptions of how they experienced supervision. Whereas we were careful to note whether particular perceptions were unique to one resident, or shared by others, we recognize that the value of residents' observations is assessed by the quality of the insights they provide, not necessarily by the number of residents who described the same experience.

In conclusion, we found that residents experienced difficulties in initiating traditional hierarchical supervision related to medication safety in the ICU. However, they reported ubiquitous interprofessional supervision, albeit limited in scope, which they relied upon for nonjudgmental guidance in their therapeutic decision‐making, especially after‐hours. In our study, interprofessional supervision proved crucial to improving medication safety in the ICU.

Close supervision of residents potentially leads to fewer errors, lower patient mortality, and improved quality of care.19 An Institute of Medicine (IOM) report3 recommended improving supervision through more frequent consultations between residents and their supervisors. Although current Accreditation Council for Graduate Medical Education (ACGME) guidelines also recommend that attending physicians (attendings) supervise residents, detailed guidance about what constitutes adequate supervision and how it should be implemented is not well defined.10, 11 The ACGME stresses that supervision should promote resident autonomy in clinical care.10 However, when trainees act independently, it might lead to critical communication breakdowns and other patient safety concerns.5, 6, 1214 Although attendings can encourage (or discourage) residents from seeking advice,15, 16 residents also play important roles in asking for help (ie, initiating their own supervision).1719 Additional research is needed on how residents walk the fine line between exercising independence and seeking supervision.

Lack of resident supervision is especially problematic in high‐risk settings such as the medical intensive care unit (ICU), where medical errors are as frequent as 1.7 errors per patient per day,20, 21 and the adverse drug event rate is twice that of non‐ICU settings.22 Because medication errors are one of the most common errors residents make,23, 24 resident interactions with nursing and pharmacy staff may significantly influence medication safety in error‐prone ICUs.2529 Studies of traditional hierarchical supervision tend to overlook how interactions with other professionals influence resident training.12, 18, 30, 31

We define supervision as a process of providing trainees with monitoring, guidance, and feedback^9(p828) as they care for patients.3 Whereas traditionally, supervisors are identified by their positions of formal authority in the medical chain of command; we conceptualize supervision as a process in which professionals engaged in supervisory activities need not have formal authority over their trainees.

To examine how residents seek supervision through both the traditional medical hierarchical chain of command (including attendings, fellows and senior residents) and interprofessional communication channels (including nursing and pharmacy staff), we conducted a qualitative study of residents working in ICUs in three tertiary care hospitals. Using semi‐structured interviews, we asked residents to describe how they experienced supervision as they provided medications to patients. Two broad research questions guided data analysis:

• 1

  How do residents receive supervision from physicians in the traditional medical hierarchy?

• 2

  How do residents receive supervision from other professionals (ie, nurses, staff pharmacists, and clinical pharmacists)?

METHODS

RESULTS

Residents described how they were supervised not only by other physicians within the traditional medical hierarchy, but also by other professionals, including nurses, staff pharmacists, and clinical pharmacists, ie, interprofessional supervision (Figure 1). After presenting these results, we examine how physicians and other professionals used communication strategies during interprofessional supervision. Here we use the term residents to include trainees at all levels, from interns to upper‐level residents, and male pronouns for de‐identification.

Figure 1

DISCUSSION

We analyzed interviews of residents working in medical ICUs to understand their supervision experiences related to medication safety. Although residents espoused beliefs in seeking assistance from supervising physicians and articulated strategies for doing so, many experienced difficulties in initiating supervision through the traditional medical hierarchy. Some residents were embarrassed by their mistaken decisions; others were concerned that their questions would reflect poorly on them.

Residents also received interprofessional supervision from nurses and pharmacists, who proactively monitored, intervened in, and guided residents' decisions. Other professionals evaluated residents' decisions by comparing them to distinctive professional guidelines and routinely used deferential language when conveying their concerns. Residents, in turn, asked other professionals for assistance.

We posit that interprofessional supervision clearly meets an accepted definition of supervision.3, 9 Residents received monitoring, guidance and feedback^9(p828) from other professionals, who engaged in routine monitoring and in situation‐specific double‐checks of residents' clinical decisions, similar to those performed by supervising physicians.30 Moreover, other professionals demonstrated the ability to anticipate a doctor's strengths and weaknesses in particular clinical situations in order to maximize patient safety.^9(p829)

Our study results have implications for graduate medical education (GME) reform. First, trainees experienced supervision as a two‐way interaction.36 Residents balanced the countervailing pressures to act independently or to seek a supervising physician's advice, in part, by developing strategies for deciding when to ask questions. Kennedy et al. identified similar rhetorical strategies.18 By asking questions about their clinical decisions, residents requested that supervising physicians guide their work; thus, they proactively initiated and thereby enacted their own supervision. Fostering the conditions for initiating supervision is essential, especially given the association between lack of effective supervision and adverse outcomes.5, 6, 1214

Second, residents expressed contradictory expectations about seeking advice from supervising physicians. Some residents were wary of approaching attending physicians for fear of appearing incompetent or being ridiculed.12, 16, 18, 31 However, we found that other residents remained reluctant to seek advice despite simultaneously appreciating that attendings encouraged them to ask for assistance. Whereas the perceived approachability of supervising physicians was important,18, 19 our exploratory findings suggest that it may be a necessary, but not a sufficient, condition for creating a learning environment. Creating a supportive learning environmentin which residents feel comfortable in revealing their perceived shortcomings to supervising physicians3begins with cultural changes, such as building medical teams,6 but such changes can be slow to develop.

Third, interprofessional supervision offers a strategy for improving supervision. The ubiquitous involvement of nursing and pharmacy staff in monitoring and intervening in residents' medication‐related decisions could result in overlooking their unique contributions to resident supervision. Mindful that supervising physicians evaluate them, residents selectively sought nonjudgmental advice from professionals outside the medical hierarchy. Therefore, improving supervision could entail offering residents ready access to other professionals who can advise them, especially during late night hours when supervising physicians might not be present.17, 27

The importance of interprofessional supervision has not been adequately recognized and emphasized in GME. Our study findings, if supported by future research, highlight how interpersonal communication techniques could influence both interprofessional supervision and hierarchical supervision among physicians. Medical team training programs3739 emphasize developing skills, such as mutual performance monitoring,^40(p13) by training providers to raise and respond to potentially sensitive questions. Improving supervision by enhancing interpersonal communication skills may be important, not only when relative status differences are clear (ie, physician hierarchy), but also when status differences are ambiguous (ie, residents and other professionals). GME programs could consider incorporating these techniques into their formal curricula, as could programs for nursing and pharmacy staff.

Our study has several limitations. Because of the larger research project objectives, we focused on medication safety in medical ICU settings, where nurses and pharmacists may be especially vigilant and proactive in monitoring residents. Thus, our findings may be specific to medication issues and less relevant outside ICUs. We had a relatively small sample size and do not claim to generalize from it, although we believe it offers meaningful insights. We also did not continue enlarging our sample until reaching redundancy.^35(p202) Nevertheless, the purposeful random sample of residents produced rich information. Indeed, some study results are consistent with previous resident education research,18 adding validity to our findings. Although the interview protocol was not designed specifically to investigate supervision, the resulting interviews yielded abundant data containing residents' detailed descriptions of how they experienced supervision. Whereas we were careful to note whether particular perceptions were unique to one resident, or shared by others, we recognize that the value of residents' observations is assessed by the quality of the insights they provide, not necessarily by the number of residents who described the same experience.

In conclusion, we found that residents experienced difficulties in initiating traditional hierarchical supervision related to medication safety in the ICU. However, they reported ubiquitous interprofessional supervision, albeit limited in scope, which they relied upon for nonjudgmental guidance in their therapeutic decision‐making, especially after‐hours. In our study, interprofessional supervision proved crucial to improving medication safety in the ICU.

The Joint Commission (TJC) currently accredits approximately 4546 acute care, critical access, and specialty hospitals,1 accounting for approximately 82% of U.S. hospitals (representing 92% of hospital beds). Hospitals seeking to earn and maintain accreditation undergo unannounced on‐site visits by a team of Joint Commission surveyors at least once every 3 years. These surveys address a variety of domains, including the environment of care, infection prevention and control, information management, adherence to a series of national patient safety goals, and leadership.1

The survey process has changed markedly in recent years. Since 2002, accredited hospitals have been required to continuously collect and submit selected performance measure data to The Joint Commission throughout the three‐year accreditation cycle. The tracer methodology, an evaluation method in which surveyors select a patient to follow through the organization in order to assess compliance with selected standards, was instituted in 2004. Soon thereafter, on‐site surveys went from announced to unannounced in 2006.

Despite the 50+ year history of hospital accreditation in the United States, there has been surprisingly little research on the link between accreditation status and measures of hospital quality (both processes and outcomes). It is only recently that a growing number of studies have attempted to examine this relationship. Empirical support for the relationship between accreditation and other quality measures is emerging. Accredited hospitals have been shown to provide better emergency response planning2 and training3 compared to non‐accredited hospitals. Accreditation has been observed to be a key predictor of patient safety system implementation4 and the primary driver of hospitals' patient‐safety initiatives.5 Accredited trauma centers have been associated with significant reductions in patient mortality,6 and accreditation has been linked to better compliance with evidence‐based methadone and substance abuse treatment.7, 8 Accredited hospitals have been shown to perform better on measures of hospital quality in acute myocardial infarction (AMI), heart failure, and pneumonia care.9, 10 Similarly, accreditation has been associated with lower risk‐adjusted in‐hospital mortality rates for congestive heart failure (CHF), stroke, and pneumonia.11, 12 The results of such research, however, have not always been consistent. Several studies have been unable to demonstrate a relationship between accreditation and quality measures. A study of financial and cost‐related outcome measures found no relationship to accreditation,13 and a study comparing medication error rates across different types of organizations found no relationship to accreditation status.14 Similarly, a comparison of accredited versus non‐accredited ambulatory surgical organizations found that patients were less likely to be hospitalized when treated at an accredited facility for colonoscopy procedures, but no such relationship was observed for the other 4 procedures studied.15

While the research to date has been generally supportive of the link between accreditation and other measures of health care quality, the studies were typically limited to only a few measures and/or involved relatively small samples of accredited and non‐accredited organizations. Over the last decade, however, changes in the performance measurement landscape have created previously unavailable opportunities to more robustly examine the relationship between accreditation and other indicators of hospital quality.

At about the same time that The Joint Commission's accreditation process was becoming more vigorous, the Centers for Medicare and Medicaid Services (CMS) began a program of publicly reporting quality data (http://www.hospitalcompare.hhs.gov). The alignment of Joint Commission and CMS quality measures establishes a mechanism through which accredited and non‐accredited hospitals can be compared using the same nationally standardized quality measures. Therefore, we took advantage of this unique circumstancea new and more robust TJC accreditation program and the launching of public quality reportingto examine the relationship between Joint Commission accreditation status and publicly reported hospital quality measures. Moreover, by examining trends in these publicly reported measures over five years and incorporating performance data not found in the Hospital Compare Database, we assessed whether accreditation status was also linked to the pace of performance improvement over time.

By using a population of hospitals and a range of standardized quality measures greater than those used in previous studies, we seek to address the following questions: Is Joint Commission accreditation status truly associated with higher quality care? And does accreditation status help identify hospitals that are more likely to improve their quality and safety over time?

METHODS

RESULTS

For the 16 individual measures used in this study, a total of 4798 hospitals participated in Hospital Compare or reported data to The Joint Commission in 2004 or 2008. Of these, 907 were excluded because the performance data were not available for either 2004 (576 hospitals) or 2008 (331 hospitals) resulting in a missing value for the change in performance score. Therefore, 3891 hospitals (81%) were included in the final analyses. The 907 excluded hospitals were more likely to be rural (50.8% vs 17.5%), be critical access hospitals (53.9% vs 13.9%), have less than 100 beds (77.4% vs 37.6%), be government owned (34.6% vs 22.1%), be for profit (61.4% vs 49.5%), or be unaccredited (79.8% vs 45.8% in 2004; 75.6% vs 12.8% in 2008), compared with the included hospitals (P < 0.001 for all comparisons).