Let's think about what we need to do ourselves. We have to acknowledge that orders we write drive up health care costs.1 AMA President, Nancy H. Nielsen, MD, PhD

As the most prominent providers of inpatient care, hospitalists should be aware that, of the total annual expenditures on US healthcare ($2.3 trillion in 2007),2 approximately one‐third goes to hospital‐based medical care, over one‐half of which (57%) is covered by public funds through Medicare and Medicaid3; this high cost of healthcare is increasingly being blamed for unnecessarily burdening our economy and preventing our industries from being globally competitive. I believe that the high proportion of spending on inpatient care places hospitalists firmly in the center of the debate on how to reduce healthcare costs. It is well known that the United States spends about twice as much per capita as other industrialized countries on healthcare,4 without evidence of superior health outcomes.5 However, it is also known that remarkable local and regional variations in healthcare spending also exist within the US, again, without evidence of superior health outcomes in the higher‐spending regions.6 Both of these observations suggest that we are spending many healthcare dollars on things that evidently do not improve the health of our patients. How much of this waste is administrative, operational, or clinical is debatable and remains the focus of growing national healthcare reform efforts.711 However, from the hospitalist perspective, we should be especially wary of providing so‐called flat‐of‐the‐curve medicine, that is, a level of intensity of care that provides no incremental health benefit.12 The purpose of this editorial is to challenge hospitalists to collectively examine how much of our inpatient spending is potentially unnecessary, and how we, as specialists in inpatient medicine, can assume a critical role in controlling healthcare costs.

To illustrate the issue, consider the following clinical scenario, managed in different ways by different hospitalists, with approximate costs itemized in Table 1. The patient is an elderly woman who presents to the emergency room with syncope occurring at church. The first hospitalist takes time to gather history from the patient, family, eyewitnesses, and the primary care physician, and requests a medication list and outside medical records, which reveal several recent and relevant cardiac and imaging studies. He performs a careful examination, discovers orthostatic hypotension, and his final diagnosis is syncope related to volume depletion from a recently added diuretic as well as a mild gastroenteritis. The patient is rehydrated and discharged home from the emergency room in the care of her family, and asked to hold her diuretic until seen by her family physician in 1 or 2 days. The second hospitalist receives the call from the emergency room and tells the staff to get the patient a telemetry bed. He sees the patient 2 hours later when she gets to the floor. The family has gone home and the mildly demented patient does not recall much of the event or her past medical history. The busy hospitalist constructs a broad differential diagnosis and writes some quick orders to evaluate the patient for possible stroke, seizure, pulmonary embolism, and cardiac ischemia or arrhythmia. He also asks cardiology and neurology to give an opinion. The testing is normal, and the patient is discharged with a cardiac event monitor and an outpatient tilt‐table test scheduled.

Although the above scenarios purposely demonstrate 2 extremes of care, I suspect most readers would agree that each hospitalist has his or her own style of practice, and that these differences in style inevitably result in significant differences in the total cost of healthcare delivered. This variation in spending among individual physicians is perhaps more easily understood than the striking variations in healthcare spending seen when different states, regions, and hospitals are compared. For example, annual Medicare spending per beneficiary has varied widely from state to state, from $5436 in Iowa to $7995 in New York (in 2004), a 47% difference.13 Specific analysis of inpatient spending variations is presented in the Dartmouth Atlas of Health Care 2008, which reports healthcare spending in the last 2 years of life for patients with at least 1 chronic illness.14 While the average Medicare inpatient spending per capita for these patients was about $25,000, the state‐specific spending varied widely from $37,040 in New Jersey to $17,135 in Idaho. There was also significant variation in spending within individual states (ie, New York: Binghamton, $18,339; Manhattan, $57,000) and between similar types of hospitals (UCLA Medical Center, $63,900; Massachusetts General Hospital, $43,058). Yet there is no evidence that higher‐spending regions produce better health outcomes.6 Interestingly, the observed differences in spending within the US were primarily due to the volume and intensity of care, not the price of care, as has been seen in some comparisons of the US with other industrialized countries.8, 15 In overall Medicare expenditures, higher‐spending locations tended to have a more inpatient‐based and specialist‐oriented pattern of practice, with higher utilization of inpatient consultations, diagnostic testing, and minor procedures.6

Although the wide variation in spending observed is a bit baffling, the encouraging aspect of this data is that some places are apparently doing it right; that is, providing their patients with a much higher value per healthcare dollar. Ultimately, if the higher‐spending locations modeled the lower‐spending locations, we would have the potential to reduce overall healthcare costs by as much as 30% without harming health.9

What are the possible reasons that we are providing unnecessary care? There are both environment‐dependent and physician‐dependent reasons, which I will outline here. The first 3 reasons represent areas that would seem to require system‐wide change, whereas the remaining 7 reasons are perhaps more amenable to local and/or national hospitalist‐directed efforts.

• 1

  Working in a litigious environment promotes unnecessary testing and consultations with the intent of reducing our exposure to malpractice liability, so‐called defensive medicine.16

• 2

  A reimbursement system that is primarily fee‐for‐service encourages physicians to provide more care and involve more physicians in the care of each patient, with little or no incentive to spend less, a core problem that was recently highlighted in a public Society of Hospital Management (SHM) statement.17

• 3

  The lack of integrated medical record systems promotes waste by leading to duplicate testing, simply because we cannot easily obtain old records to confirm whether tests were previously done. Interestingly, data from the Commonwealth Fund conclude that US physicians order duplicate diagnostic tests (a test repeated within 2 years) at more than twice the rate of Canada and the United Kingdom, while the nation with the lowest rate of duplicate testing, The Netherlands, has the highest rate of electronic medical record use (98%).18

• 4

  Working with patients (or families) with high expectations who insist upon aggressive testing, treatment, and referral to specialists inflates spending, especially if associated with futile and expensive end‐of‐life care.

• 5

  The involvement of one or more specialists may subsequently lead to even more aggressive care ordered by each specialist.

• 6

  The availability and promotion of new technology (diagnostic testing, medical devices, etc.) may prompt us to make use of it simply because it is there, with or without evidence of a health benefit. Our natural curiosity or fascination with information, or our desire to do an overly complete evaluation, works against cost containment.

• 7

  Local trends or traditions within our specific work environment, as suggested by the variability data, may have a strong influence on our individual practice. In such a setting, inadequate knowledge of the cost‐effectiveness of various tests and treatment options likely leads to unnecessary health care spending.

• 8

  A hospitalist work environment in which a high patient load is carried will inevitably result in less time to gather a detailed history and obtain old records or other information that could help narrow a differential diagnosis and minimize unnecessary or duplicate testing.

• 9

  Preventable readmissions resulting from inadequate coordination of care add cost,19 a phenomenon highly dependent on efficient information systems and proper physician‐physician communication.20

• 10

  An overestimation of the need for inpatient evaluation and treatment (vs. outpatient) leads to unnecessary admissions and a longer average length‐of‐stay, each of which add dramatically to total healthcare costs. This is not only dependent on our individual threshold for admitting and discharging patients, but also on our efficiency in diagnosing and treating acute conditions. The fact that the average length‐of‐stay for congestive heart failure admissions, for example, ranges in different regions from 4.9 to 6.1 days (with costs of $9143 and $12,528, respectively)21 is enough to show that there is room for progress.

What joint efforts could be made to minimize unnecessary inpatient spending? The following are my personal opinions and suggestions (Table 2). Most importantly, I believe every physician deserves prompt and accurate feedback regarding their spending patterns, accompanied by valid comparisons to national and local standards, to demonstrate where they stand on the spectrum of healthcare spending. We are currently far behind other industries in our ability, as physicians, to evaluate what we are spending money on, how much, and why. If I knew, for example, that my spending was in the 95th percentile of all hospitalists in community hospitals similar to mine, I would be prompted to investigate where the differences were and why. In an informal survey of hospitalist colleagues, I found that the majority do not receive any data on the costs associated with their care, and are largely unaware of the actual cost of the inpatient tests they commonly order. Developing a secure, user‐friendly database of individual physician spending patterns relative to national and local standards could be a preliminary step, and would likely require a unified effort between government agencies, professional societies, hospitals, and the insurance industry. However, once available, the increased transparency and clarity of spending variations would hopefully prompt introspection and change. In the absence of hard data, however, individual self‐assessment on spending patterns could also be offered through the development of an online simulated case‐based examination in which a physician could gain a general idea of how his evaluation and treatment of a case scenario compares to his hospitalist colleagues, and to what degree each of his clinical decisions affects the overall cost of care. There are many excellent quality improvement tools offered through SHM but none that specifically address the cost of care.

Second, hospitalists need quick access to current evidence‐based guidelines regarding the true clinical value, or cost‐effectiveness, of testing and treatment for common inpatient conditions, including specific admission criteria. A single source or clearinghouse of guidelines, sponsored by SHM, may be particularly helpful, especially if it focuses on clarifying areas of highest variability in inpatient spending. In addition, I believe that, given the critically important interface between emergency medicine and hospital medicine, joint guidelines between the 2 groups would potentially be very helpful in controlling costs by limiting unnecessary admissions. Advocacy for comparative effectiveness research to establish validity in these guidelines will be fundamental22, 23; however, I suspect the common sense question: Will this added cost improve my patient's outcome? also needs to be applied more generously, since many individual clinical scenarios will not likely lend themselves to formal study. For discussion, some sample case scenarios are presented (Table 3).

Third, hospitalists could potentially benefit from the development of patient education materials, available through SHM, that address the cost‐effectiveness of common inpatient tests and treatments with the goal of decreasing patient demand for unnecessary testing. Education regarding advanced directives and end‐of‐life care decision‐making could be particularly valuable in minimizing futile care, as it is well‐documented that transitioning to palliative care as soon as it is appropriate reduces healthcare spending greatly during the end‐of‐life period.2427 At the same time, we need to be careful to reassure our patients that we are not trying to ration care, but are instead minimizing the risks and costs for them associated with unnecessary care. In my experience, most patients, if given appropriate time, attention, and education, are willing to accept the final recommendation of their physician.

Fourth, intensified federal and state advocacy in several areas could help reduce spending. For example, advocacy for medical liability reform may reduce the atmosphere of defensive medicine, although I suspect that because old habits die hard, it may take a full generation of decreased liability risk to actually change practice patterns. Advocacy for the development of a national, or at least more uniform, electronic medical record, may decrease duplicate testing and improve efficiency. Advocacy for value‐based reimbursement models may help dampen costs resulting from a predominantly fee‐for‐service environment.28

Fifth, and perhaps most fundamental to the future of our specialty, encouraging the broad professional development of hospitalists as a true specialists in inpatient medicine (based on the SHM Core Competencies,)29 could help minimize the unnecessary costs associated with specialist‐oriented care.6 With the desire to create, in the near future, a formal board‐certification in hospital medicine comes an obligation to develop broad knowledge and broad skill sets that are truly unique to our profession, whereas deferring to a specialist‐oriented pattern of care actually shrinks us down to something less than a traditional internist, rather than a unique entity.30 With our 24/7 focus on inpatient care, we should easily be able to demonstrate our superiority in safety, quality, and efficiency, all of which are closely linked to increased value per healthcare dollar. If, however, our focus is blurred by an overly productivity‐based practice, in which patient volume and procedures take precedence, we will not be able to claim any special value to the system.

Last, supporting efforts to improve coordination of care and transitions of care could reduce costs associated with unnecessary readmissions or posthospital complications. A recent policy statement from several professional societies, including SHM, highlights the importance of these transitions,20, 31 and within the past year, SHM has launched the successful Project BOOST (Better Outcomes for Older adults through Safe Transitions) to help in this effort.32

Unfortunately, there is an inherent problem with all of the above proposals: the assumption that physicians actually want to reduce healthcare spending. Since everyone who works in the medical industry benefits financially in some way from the current high levels of spending on healthcare, reducing spending is counterintuitive for many, and the incentives to spend more will likely persist until some form of spending targets or limits are set.33 Moreover, since physicians traditionally do not like to be told how to practice medicine, history would predict that, without attractive incentives, nothing will change. This is the fundamental and unfortunate dilemma that has apparently pushed us to the eleventh hour of a healthcare crisis.

Another concern with an extreme atmosphere of cost cutting is the risk of swinging too far in the opposite direction, focusing so intently on cost that we begin to compromise quality or access to care in order to achieve spending targets. Reassuringly, however, the data suggest that there is plenty of room for us to cut costs without harming health outcomes.

Despite these obstacles, during this historic time in US healthcare, I believe hospitalists have a unique and perhaps transient opportunity to demonstrate their singular commitment to rational healthcare spending and by doing so to gain significant influence in shaping the impending healthcare reforms. If we speak and act with one voice, with transparency, and with the proper data, we could be the first and only professional society to not only demonstrate our current pattern of spending, but also our potential for reducing spending and our plan on how to get there.

Let's think about what we need to do ourselves. We have to acknowledge that orders we write drive up health care costs.1 AMA President, Nancy H. Nielsen, MD, PhD

As the most prominent providers of inpatient care, hospitalists should be aware that, of the total annual expenditures on US healthcare ($2.3 trillion in 2007),2 approximately one‐third goes to hospital‐based medical care, over one‐half of which (57%) is covered by public funds through Medicare and Medicaid3; this high cost of healthcare is increasingly being blamed for unnecessarily burdening our economy and preventing our industries from being globally competitive. I believe that the high proportion of spending on inpatient care places hospitalists firmly in the center of the debate on how to reduce healthcare costs. It is well known that the United States spends about twice as much per capita as other industrialized countries on healthcare,4 without evidence of superior health outcomes.5 However, it is also known that remarkable local and regional variations in healthcare spending also exist within the US, again, without evidence of superior health outcomes in the higher‐spending regions.6 Both of these observations suggest that we are spending many healthcare dollars on things that evidently do not improve the health of our patients. How much of this waste is administrative, operational, or clinical is debatable and remains the focus of growing national healthcare reform efforts.711 However, from the hospitalist perspective, we should be especially wary of providing so‐called flat‐of‐the‐curve medicine, that is, a level of intensity of care that provides no incremental health benefit.12 The purpose of this editorial is to challenge hospitalists to collectively examine how much of our inpatient spending is potentially unnecessary, and how we, as specialists in inpatient medicine, can assume a critical role in controlling healthcare costs.

To illustrate the issue, consider the following clinical scenario, managed in different ways by different hospitalists, with approximate costs itemized in Table 1. The patient is an elderly woman who presents to the emergency room with syncope occurring at church. The first hospitalist takes time to gather history from the patient, family, eyewitnesses, and the primary care physician, and requests a medication list and outside medical records, which reveal several recent and relevant cardiac and imaging studies. He performs a careful examination, discovers orthostatic hypotension, and his final diagnosis is syncope related to volume depletion from a recently added diuretic as well as a mild gastroenteritis. The patient is rehydrated and discharged home from the emergency room in the care of her family, and asked to hold her diuretic until seen by her family physician in 1 or 2 days. The second hospitalist receives the call from the emergency room and tells the staff to get the patient a telemetry bed. He sees the patient 2 hours later when she gets to the floor. The family has gone home and the mildly demented patient does not recall much of the event or her past medical history. The busy hospitalist constructs a broad differential diagnosis and writes some quick orders to evaluate the patient for possible stroke, seizure, pulmonary embolism, and cardiac ischemia or arrhythmia. He also asks cardiology and neurology to give an opinion. The testing is normal, and the patient is discharged with a cardiac event monitor and an outpatient tilt‐table test scheduled.

Although the above scenarios purposely demonstrate 2 extremes of care, I suspect most readers would agree that each hospitalist has his or her own style of practice, and that these differences in style inevitably result in significant differences in the total cost of healthcare delivered. This variation in spending among individual physicians is perhaps more easily understood than the striking variations in healthcare spending seen when different states, regions, and hospitals are compared. For example, annual Medicare spending per beneficiary has varied widely from state to state, from $5436 in Iowa to $7995 in New York (in 2004), a 47% difference.13 Specific analysis of inpatient spending variations is presented in the Dartmouth Atlas of Health Care 2008, which reports healthcare spending in the last 2 years of life for patients with at least 1 chronic illness.14 While the average Medicare inpatient spending per capita for these patients was about $25,000, the state‐specific spending varied widely from $37,040 in New Jersey to $17,135 in Idaho. There was also significant variation in spending within individual states (ie, New York: Binghamton, $18,339; Manhattan, $57,000) and between similar types of hospitals (UCLA Medical Center, $63,900; Massachusetts General Hospital, $43,058). Yet there is no evidence that higher‐spending regions produce better health outcomes.6 Interestingly, the observed differences in spending within the US were primarily due to the volume and intensity of care, not the price of care, as has been seen in some comparisons of the US with other industrialized countries.8, 15 In overall Medicare expenditures, higher‐spending locations tended to have a more inpatient‐based and specialist‐oriented pattern of practice, with higher utilization of inpatient consultations, diagnostic testing, and minor procedures.6

Although the wide variation in spending observed is a bit baffling, the encouraging aspect of this data is that some places are apparently doing it right; that is, providing their patients with a much higher value per healthcare dollar. Ultimately, if the higher‐spending locations modeled the lower‐spending locations, we would have the potential to reduce overall healthcare costs by as much as 30% without harming health.9

What are the possible reasons that we are providing unnecessary care? There are both environment‐dependent and physician‐dependent reasons, which I will outline here. The first 3 reasons represent areas that would seem to require system‐wide change, whereas the remaining 7 reasons are perhaps more amenable to local and/or national hospitalist‐directed efforts.

• 1

  Working in a litigious environment promotes unnecessary testing and consultations with the intent of reducing our exposure to malpractice liability, so‐called defensive medicine.16

• 2

  A reimbursement system that is primarily fee‐for‐service encourages physicians to provide more care and involve more physicians in the care of each patient, with little or no incentive to spend less, a core problem that was recently highlighted in a public Society of Hospital Management (SHM) statement.17

• 3

  The lack of integrated medical record systems promotes waste by leading to duplicate testing, simply because we cannot easily obtain old records to confirm whether tests were previously done. Interestingly, data from the Commonwealth Fund conclude that US physicians order duplicate diagnostic tests (a test repeated within 2 years) at more than twice the rate of Canada and the United Kingdom, while the nation with the lowest rate of duplicate testing, The Netherlands, has the highest rate of electronic medical record use (98%).18

• 4

  Working with patients (or families) with high expectations who insist upon aggressive testing, treatment, and referral to specialists inflates spending, especially if associated with futile and expensive end‐of‐life care.

• 5

  The involvement of one or more specialists may subsequently lead to even more aggressive care ordered by each specialist.

• 6

  The availability and promotion of new technology (diagnostic testing, medical devices, etc.) may prompt us to make use of it simply because it is there, with or without evidence of a health benefit. Our natural curiosity or fascination with information, or our desire to do an overly complete evaluation, works against cost containment.

• 7

  Local trends or traditions within our specific work environment, as suggested by the variability data, may have a strong influence on our individual practice. In such a setting, inadequate knowledge of the cost‐effectiveness of various tests and treatment options likely leads to unnecessary health care spending.

• 8

  A hospitalist work environment in which a high patient load is carried will inevitably result in less time to gather a detailed history and obtain old records or other information that could help narrow a differential diagnosis and minimize unnecessary or duplicate testing.

• 9

  Preventable readmissions resulting from inadequate coordination of care add cost,19 a phenomenon highly dependent on efficient information systems and proper physician‐physician communication.20

• 10

  An overestimation of the need for inpatient evaluation and treatment (vs. outpatient) leads to unnecessary admissions and a longer average length‐of‐stay, each of which add dramatically to total healthcare costs. This is not only dependent on our individual threshold for admitting and discharging patients, but also on our efficiency in diagnosing and treating acute conditions. The fact that the average length‐of‐stay for congestive heart failure admissions, for example, ranges in different regions from 4.9 to 6.1 days (with costs of $9143 and $12,528, respectively)21 is enough to show that there is room for progress.

What joint efforts could be made to minimize unnecessary inpatient spending? The following are my personal opinions and suggestions (Table 2). Most importantly, I believe every physician deserves prompt and accurate feedback regarding their spending patterns, accompanied by valid comparisons to national and local standards, to demonstrate where they stand on the spectrum of healthcare spending. We are currently far behind other industries in our ability, as physicians, to evaluate what we are spending money on, how much, and why. If I knew, for example, that my spending was in the 95th percentile of all hospitalists in community hospitals similar to mine, I would be prompted to investigate where the differences were and why. In an informal survey of hospitalist colleagues, I found that the majority do not receive any data on the costs associated with their care, and are largely unaware of the actual cost of the inpatient tests they commonly order. Developing a secure, user‐friendly database of individual physician spending patterns relative to national and local standards could be a preliminary step, and would likely require a unified effort between government agencies, professional societies, hospitals, and the insurance industry. However, once available, the increased transparency and clarity of spending variations would hopefully prompt introspection and change. In the absence of hard data, however, individual self‐assessment on spending patterns could also be offered through the development of an online simulated case‐based examination in which a physician could gain a general idea of how his evaluation and treatment of a case scenario compares to his hospitalist colleagues, and to what degree each of his clinical decisions affects the overall cost of care. There are many excellent quality improvement tools offered through SHM but none that specifically address the cost of care.

Second, hospitalists need quick access to current evidence‐based guidelines regarding the true clinical value, or cost‐effectiveness, of testing and treatment for common inpatient conditions, including specific admission criteria. A single source or clearinghouse of guidelines, sponsored by SHM, may be particularly helpful, especially if it focuses on clarifying areas of highest variability in inpatient spending. In addition, I believe that, given the critically important interface between emergency medicine and hospital medicine, joint guidelines between the 2 groups would potentially be very helpful in controlling costs by limiting unnecessary admissions. Advocacy for comparative effectiveness research to establish validity in these guidelines will be fundamental22, 23; however, I suspect the common sense question: Will this added cost improve my patient's outcome? also needs to be applied more generously, since many individual clinical scenarios will not likely lend themselves to formal study. For discussion, some sample case scenarios are presented (Table 3).

Third, hospitalists could potentially benefit from the development of patient education materials, available through SHM, that address the cost‐effectiveness of common inpatient tests and treatments with the goal of decreasing patient demand for unnecessary testing. Education regarding advanced directives and end‐of‐life care decision‐making could be particularly valuable in minimizing futile care, as it is well‐documented that transitioning to palliative care as soon as it is appropriate reduces healthcare spending greatly during the end‐of‐life period.2427 At the same time, we need to be careful to reassure our patients that we are not trying to ration care, but are instead minimizing the risks and costs for them associated with unnecessary care. In my experience, most patients, if given appropriate time, attention, and education, are willing to accept the final recommendation of their physician.

Fourth, intensified federal and state advocacy in several areas could help reduce spending. For example, advocacy for medical liability reform may reduce the atmosphere of defensive medicine, although I suspect that because old habits die hard, it may take a full generation of decreased liability risk to actually change practice patterns. Advocacy for the development of a national, or at least more uniform, electronic medical record, may decrease duplicate testing and improve efficiency. Advocacy for value‐based reimbursement models may help dampen costs resulting from a predominantly fee‐for‐service environment.28

Fifth, and perhaps most fundamental to the future of our specialty, encouraging the broad professional development of hospitalists as a true specialists in inpatient medicine (based on the SHM Core Competencies,)29 could help minimize the unnecessary costs associated with specialist‐oriented care.6 With the desire to create, in the near future, a formal board‐certification in hospital medicine comes an obligation to develop broad knowledge and broad skill sets that are truly unique to our profession, whereas deferring to a specialist‐oriented pattern of care actually shrinks us down to something less than a traditional internist, rather than a unique entity.30 With our 24/7 focus on inpatient care, we should easily be able to demonstrate our superiority in safety, quality, and efficiency, all of which are closely linked to increased value per healthcare dollar. If, however, our focus is blurred by an overly productivity‐based practice, in which patient volume and procedures take precedence, we will not be able to claim any special value to the system.

Last, supporting efforts to improve coordination of care and transitions of care could reduce costs associated with unnecessary readmissions or posthospital complications. A recent policy statement from several professional societies, including SHM, highlights the importance of these transitions,20, 31 and within the past year, SHM has launched the successful Project BOOST (Better Outcomes for Older adults through Safe Transitions) to help in this effort.32

Unfortunately, there is an inherent problem with all of the above proposals: the assumption that physicians actually want to reduce healthcare spending. Since everyone who works in the medical industry benefits financially in some way from the current high levels of spending on healthcare, reducing spending is counterintuitive for many, and the incentives to spend more will likely persist until some form of spending targets or limits are set.33 Moreover, since physicians traditionally do not like to be told how to practice medicine, history would predict that, without attractive incentives, nothing will change. This is the fundamental and unfortunate dilemma that has apparently pushed us to the eleventh hour of a healthcare crisis.

Another concern with an extreme atmosphere of cost cutting is the risk of swinging too far in the opposite direction, focusing so intently on cost that we begin to compromise quality or access to care in order to achieve spending targets. Reassuringly, however, the data suggest that there is plenty of room for us to cut costs without harming health outcomes.

Despite these obstacles, during this historic time in US healthcare, I believe hospitalists have a unique and perhaps transient opportunity to demonstrate their singular commitment to rational healthcare spending and by doing so to gain significant influence in shaping the impending healthcare reforms. If we speak and act with one voice, with transparency, and with the proper data, we could be the first and only professional society to not only demonstrate our current pattern of spending, but also our potential for reducing spending and our plan on how to get there.

Let's think about what we need to do ourselves. We have to acknowledge that orders we write drive up health care costs.1 AMA President, Nancy H. Nielsen, MD, PhD

As the most prominent providers of inpatient care, hospitalists should be aware that, of the total annual expenditures on US healthcare ($2.3 trillion in 2007),2 approximately one‐third goes to hospital‐based medical care, over one‐half of which (57%) is covered by public funds through Medicare and Medicaid3; this high cost of healthcare is increasingly being blamed for unnecessarily burdening our economy and preventing our industries from being globally competitive. I believe that the high proportion of spending on inpatient care places hospitalists firmly in the center of the debate on how to reduce healthcare costs. It is well known that the United States spends about twice as much per capita as other industrialized countries on healthcare,4 without evidence of superior health outcomes.5 However, it is also known that remarkable local and regional variations in healthcare spending also exist within the US, again, without evidence of superior health outcomes in the higher‐spending regions.6 Both of these observations suggest that we are spending many healthcare dollars on things that evidently do not improve the health of our patients. How much of this waste is administrative, operational, or clinical is debatable and remains the focus of growing national healthcare reform efforts.711 However, from the hospitalist perspective, we should be especially wary of providing so‐called flat‐of‐the‐curve medicine, that is, a level of intensity of care that provides no incremental health benefit.12 The purpose of this editorial is to challenge hospitalists to collectively examine how much of our inpatient spending is potentially unnecessary, and how we, as specialists in inpatient medicine, can assume a critical role in controlling healthcare costs.

To illustrate the issue, consider the following clinical scenario, managed in different ways by different hospitalists, with approximate costs itemized in Table 1. The patient is an elderly woman who presents to the emergency room with syncope occurring at church. The first hospitalist takes time to gather history from the patient, family, eyewitnesses, and the primary care physician, and requests a medication list and outside medical records, which reveal several recent and relevant cardiac and imaging studies. He performs a careful examination, discovers orthostatic hypotension, and his final diagnosis is syncope related to volume depletion from a recently added diuretic as well as a mild gastroenteritis. The patient is rehydrated and discharged home from the emergency room in the care of her family, and asked to hold her diuretic until seen by her family physician in 1 or 2 days. The second hospitalist receives the call from the emergency room and tells the staff to get the patient a telemetry bed. He sees the patient 2 hours later when she gets to the floor. The family has gone home and the mildly demented patient does not recall much of the event or her past medical history. The busy hospitalist constructs a broad differential diagnosis and writes some quick orders to evaluate the patient for possible stroke, seizure, pulmonary embolism, and cardiac ischemia or arrhythmia. He also asks cardiology and neurology to give an opinion. The testing is normal, and the patient is discharged with a cardiac event monitor and an outpatient tilt‐table test scheduled.

Although the above scenarios purposely demonstrate 2 extremes of care, I suspect most readers would agree that each hospitalist has his or her own style of practice, and that these differences in style inevitably result in significant differences in the total cost of healthcare delivered. This variation in spending among individual physicians is perhaps more easily understood than the striking variations in healthcare spending seen when different states, regions, and hospitals are compared. For example, annual Medicare spending per beneficiary has varied widely from state to state, from $5436 in Iowa to $7995 in New York (in 2004), a 47% difference.13 Specific analysis of inpatient spending variations is presented in the Dartmouth Atlas of Health Care 2008, which reports healthcare spending in the last 2 years of life for patients with at least 1 chronic illness.14 While the average Medicare inpatient spending per capita for these patients was about $25,000, the state‐specific spending varied widely from $37,040 in New Jersey to $17,135 in Idaho. There was also significant variation in spending within individual states (ie, New York: Binghamton, $18,339; Manhattan, $57,000) and between similar types of hospitals (UCLA Medical Center, $63,900; Massachusetts General Hospital, $43,058). Yet there is no evidence that higher‐spending regions produce better health outcomes.6 Interestingly, the observed differences in spending within the US were primarily due to the volume and intensity of care, not the price of care, as has been seen in some comparisons of the US with other industrialized countries.8, 15 In overall Medicare expenditures, higher‐spending locations tended to have a more inpatient‐based and specialist‐oriented pattern of practice, with higher utilization of inpatient consultations, diagnostic testing, and minor procedures.6

Although the wide variation in spending observed is a bit baffling, the encouraging aspect of this data is that some places are apparently doing it right; that is, providing their patients with a much higher value per healthcare dollar. Ultimately, if the higher‐spending locations modeled the lower‐spending locations, we would have the potential to reduce overall healthcare costs by as much as 30% without harming health.9

What are the possible reasons that we are providing unnecessary care? There are both environment‐dependent and physician‐dependent reasons, which I will outline here. The first 3 reasons represent areas that would seem to require system‐wide change, whereas the remaining 7 reasons are perhaps more amenable to local and/or national hospitalist‐directed efforts.

• 1

  Working in a litigious environment promotes unnecessary testing and consultations with the intent of reducing our exposure to malpractice liability, so‐called defensive medicine.16

• 2

  A reimbursement system that is primarily fee‐for‐service encourages physicians to provide more care and involve more physicians in the care of each patient, with little or no incentive to spend less, a core problem that was recently highlighted in a public Society of Hospital Management (SHM) statement.17

• 3

  The lack of integrated medical record systems promotes waste by leading to duplicate testing, simply because we cannot easily obtain old records to confirm whether tests were previously done. Interestingly, data from the Commonwealth Fund conclude that US physicians order duplicate diagnostic tests (a test repeated within 2 years) at more than twice the rate of Canada and the United Kingdom, while the nation with the lowest rate of duplicate testing, The Netherlands, has the highest rate of electronic medical record use (98%).18

• 4

  Working with patients (or families) with high expectations who insist upon aggressive testing, treatment, and referral to specialists inflates spending, especially if associated with futile and expensive end‐of‐life care.

• 5

  The involvement of one or more specialists may subsequently lead to even more aggressive care ordered by each specialist.

• 6

  The availability and promotion of new technology (diagnostic testing, medical devices, etc.) may prompt us to make use of it simply because it is there, with or without evidence of a health benefit. Our natural curiosity or fascination with information, or our desire to do an overly complete evaluation, works against cost containment.

• 7

  Local trends or traditions within our specific work environment, as suggested by the variability data, may have a strong influence on our individual practice. In such a setting, inadequate knowledge of the cost‐effectiveness of various tests and treatment options likely leads to unnecessary health care spending.

• 8

  A hospitalist work environment in which a high patient load is carried will inevitably result in less time to gather a detailed history and obtain old records or other information that could help narrow a differential diagnosis and minimize unnecessary or duplicate testing.

• 9

  Preventable readmissions resulting from inadequate coordination of care add cost,19 a phenomenon highly dependent on efficient information systems and proper physician‐physician communication.20

• 10

  An overestimation of the need for inpatient evaluation and treatment (vs. outpatient) leads to unnecessary admissions and a longer average length‐of‐stay, each of which add dramatically to total healthcare costs. This is not only dependent on our individual threshold for admitting and discharging patients, but also on our efficiency in diagnosing and treating acute conditions. The fact that the average length‐of‐stay for congestive heart failure admissions, for example, ranges in different regions from 4.9 to 6.1 days (with costs of $9143 and $12,528, respectively)21 is enough to show that there is room for progress.

What joint efforts could be made to minimize unnecessary inpatient spending? The following are my personal opinions and suggestions (Table 2). Most importantly, I believe every physician deserves prompt and accurate feedback regarding their spending patterns, accompanied by valid comparisons to national and local standards, to demonstrate where they stand on the spectrum of healthcare spending. We are currently far behind other industries in our ability, as physicians, to evaluate what we are spending money on, how much, and why. If I knew, for example, that my spending was in the 95th percentile of all hospitalists in community hospitals similar to mine, I would be prompted to investigate where the differences were and why. In an informal survey of hospitalist colleagues, I found that the majority do not receive any data on the costs associated with their care, and are largely unaware of the actual cost of the inpatient tests they commonly order. Developing a secure, user‐friendly database of individual physician spending patterns relative to national and local standards could be a preliminary step, and would likely require a unified effort between government agencies, professional societies, hospitals, and the insurance industry. However, once available, the increased transparency and clarity of spending variations would hopefully prompt introspection and change. In the absence of hard data, however, individual self‐assessment on spending patterns could also be offered through the development of an online simulated case‐based examination in which a physician could gain a general idea of how his evaluation and treatment of a case scenario compares to his hospitalist colleagues, and to what degree each of his clinical decisions affects the overall cost of care. There are many excellent quality improvement tools offered through SHM but none that specifically address the cost of care.

Second, hospitalists need quick access to current evidence‐based guidelines regarding the true clinical value, or cost‐effectiveness, of testing and treatment for common inpatient conditions, including specific admission criteria. A single source or clearinghouse of guidelines, sponsored by SHM, may be particularly helpful, especially if it focuses on clarifying areas of highest variability in inpatient spending. In addition, I believe that, given the critically important interface between emergency medicine and hospital medicine, joint guidelines between the 2 groups would potentially be very helpful in controlling costs by limiting unnecessary admissions. Advocacy for comparative effectiveness research to establish validity in these guidelines will be fundamental22, 23; however, I suspect the common sense question: Will this added cost improve my patient's outcome? also needs to be applied more generously, since many individual clinical scenarios will not likely lend themselves to formal study. For discussion, some sample case scenarios are presented (Table 3).

Third, hospitalists could potentially benefit from the development of patient education materials, available through SHM, that address the cost‐effectiveness of common inpatient tests and treatments with the goal of decreasing patient demand for unnecessary testing. Education regarding advanced directives and end‐of‐life care decision‐making could be particularly valuable in minimizing futile care, as it is well‐documented that transitioning to palliative care as soon as it is appropriate reduces healthcare spending greatly during the end‐of‐life period.2427 At the same time, we need to be careful to reassure our patients that we are not trying to ration care, but are instead minimizing the risks and costs for them associated with unnecessary care. In my experience, most patients, if given appropriate time, attention, and education, are willing to accept the final recommendation of their physician.

Fourth, intensified federal and state advocacy in several areas could help reduce spending. For example, advocacy for medical liability reform may reduce the atmosphere of defensive medicine, although I suspect that because old habits die hard, it may take a full generation of decreased liability risk to actually change practice patterns. Advocacy for the development of a national, or at least more uniform, electronic medical record, may decrease duplicate testing and improve efficiency. Advocacy for value‐based reimbursement models may help dampen costs resulting from a predominantly fee‐for‐service environment.28

Fifth, and perhaps most fundamental to the future of our specialty, encouraging the broad professional development of hospitalists as a true specialists in inpatient medicine (based on the SHM Core Competencies,)29 could help minimize the unnecessary costs associated with specialist‐oriented care.6 With the desire to create, in the near future, a formal board‐certification in hospital medicine comes an obligation to develop broad knowledge and broad skill sets that are truly unique to our profession, whereas deferring to a specialist‐oriented pattern of care actually shrinks us down to something less than a traditional internist, rather than a unique entity.30 With our 24/7 focus on inpatient care, we should easily be able to demonstrate our superiority in safety, quality, and efficiency, all of which are closely linked to increased value per healthcare dollar. If, however, our focus is blurred by an overly productivity‐based practice, in which patient volume and procedures take precedence, we will not be able to claim any special value to the system.

Last, supporting efforts to improve coordination of care and transitions of care could reduce costs associated with unnecessary readmissions or posthospital complications. A recent policy statement from several professional societies, including SHM, highlights the importance of these transitions,20, 31 and within the past year, SHM has launched the successful Project BOOST (Better Outcomes for Older adults through Safe Transitions) to help in this effort.32

Unfortunately, there is an inherent problem with all of the above proposals: the assumption that physicians actually want to reduce healthcare spending. Since everyone who works in the medical industry benefits financially in some way from the current high levels of spending on healthcare, reducing spending is counterintuitive for many, and the incentives to spend more will likely persist until some form of spending targets or limits are set.33 Moreover, since physicians traditionally do not like to be told how to practice medicine, history would predict that, without attractive incentives, nothing will change. This is the fundamental and unfortunate dilemma that has apparently pushed us to the eleventh hour of a healthcare crisis.

Another concern with an extreme atmosphere of cost cutting is the risk of swinging too far in the opposite direction, focusing so intently on cost that we begin to compromise quality or access to care in order to achieve spending targets. Reassuringly, however, the data suggest that there is plenty of room for us to cut costs without harming health outcomes.

Despite these obstacles, during this historic time in US healthcare, I believe hospitalists have a unique and perhaps transient opportunity to demonstrate their singular commitment to rational healthcare spending and by doing so to gain significant influence in shaping the impending healthcare reforms. If we speak and act with one voice, with transparency, and with the proper data, we could be the first and only professional society to not only demonstrate our current pattern of spending, but also our potential for reducing spending and our plan on how to get there.

Congestive heart failure (CHF) is a common disease with high mortality and morbidity.1, 2 Better physiological understanding has led to significant advances in therapy in recent years, with synthesis of this evidence into widely available treatment guidelines.3, 4 However, patients who have had an acute hospitalization with heart failure continue to have a high rate of symptomatic relapse, with up to 25% readmitted within 3 months.2 One of the major challenges in heart failure therapy is to avert these relapses to prevent hospital readmission.

Angiotensin‐converting enzyme (ACE) inhibitors, beta‐blockers, and spironolactone have promised a reduction in hospitalization rates as well as mortality; however, suboptimal prescribing5 and adherence to therapy6, 7 may limit their anticipated benefits. This has led to interest in improved systems of care to reduce hospital utilization. Such approaches have included improved systems for optimizing medications,68 comprehensive discharge planning and postdischarge support,914 and self‐management and case management strategies1517 to enhance patient participation in care.

Combinations of these strategies are known as disease management programs (DMPs), and trials of such combination strategies to improve patient outcomes have been promising.1823 Recognized features4 include skilled multidisciplinary team care; individualized guideline‐based treatment plans that may include dietary and exercise programs as well as optimal pharmacological therapy; patient education and self‐management strategies; improved integration between hospital and community care providers; vigilant follow‐up including prompt review after hospitalization; ready access to expert assessment in the event of deterioration; and regular monitoring with expert titration of therapy, through clinics, home visits, or telemonitoring. Several randomized controlled trials have suggested that DMPs may reduce heart failure‐related9, 1517 and all‐cause9, 10 readmissions. Meta‐analyses12, 1823 have demonstrated reduction in risk of all‐cause readmission of 12% to 25% as well as a reduction in mortality of 14% to 25%.

Trials of DMPs have generally involved careful participant selection, and differences in methods and outcome reporting have led some reviewers to be circumspect in their interpretation of the impact of these programs on readmission rates.23 A large, real‐world quality improvement program conducted as part of the Royal Australasian College of Physicians Clinical Support Systems Project provided an opportunity to measure whether a multifaceted program targeting a representative group of patients with CHF and their healthcare providers could reduce readmission rates. As previously published, this program delivered measurable improvements in processes of care including evidence‐based prescribing, adherence, multidisciplinary involvement, and discharge communication, associated with a reduction in 12‐month mortality.24

Objective

The Brisbane Cardiac Consortium sought to improve processes of care for patients with CHF by using evidence‐based strategies targeting patients and their healthcare providers to optimize uptake of management guidelines, improve discharge processes between hospital and primary care, and increase patient participation in care. We hypothesized that the program would reduce hospital readmissions in the intervention patients in the first 12 months following discharge.

Methods

Results

There were 220 patients identified with a clinical diagnosis of CHF during the baseline period, and 235 during the intervention period. Figure 1 shows ascertainment, in‐hospital mortality, and eligibility rates for the 2 cohorts. Eighty‐nine (45%) of baseline patients and 76 (35%) of intervention patients received intensive posthospital follow‐up as described above. Information on readmission was available for 197 baseline patients and 219 intervention patients discharged alive; this is the sample used for all analyses in this report. Table 1 shows the demographic and clinical characteristics of these patients. Table 2 summarizes the previously reported improvements in processes of care.

Figure 1

Duing the 12‐month follow‐up, 107 (49%) of intervention patients were readmitted to the hospital compared to 71 (36%) of control patients, representing a 1.7‐fold increase in the adjusted probability of readmission in the intervention group (odds ratio [OR] = 1.71, 95% confidence interval [CI] = 1.14‐2.56; P = 0.009). As shown in Table 3, this was partly balanced by a trend toward reduced post‐hospital mortality, such that no significant difference was seen in readmission‐free survival.

Time‐to‐event analysis (Figures 2 and 3) demonstrated similar findings, with a significant reduction in time to first readmission in the intervention group (adjusted hazard ratio [HR] = 1.43; 95% CI = 1.04‐1.97; P = 0.046) but no difference in time to death or first readmission (adjusted HR = 1.14; 95% CI = 0.86‐1.46; P = 0.36).

Figure 2

Figure 3

There was a trend to increased readmissions attributed to heart failure: 47 (21.5%) of intervention patients compared to 33 (16.7%) in the baseline group (OR = 1.30; 95% CI = 0.87‐1.93; P = 0.20). No significant difference was demonstrated in the frequency of readmissions (average 0.75 admission per participant per year in baseline, compared to 0.93 intervention; P = 0.32) nor the mean number of days in hospital in 12 months subsequent to the index admission (5.9 in the baseline group compared to 6.5 in the intervention group; P = 0.1).

Subgroup analysis by intervention intensity showed similar results, with 42 of 76 (55.3%) intensive group participants in the intervention group and 36 of 89 (40.4%) in the baseline group requiring hospital readmission within 12 months. The HR for death or readmission was estimated to be 1.27 (95% CI = 0.85‐1.9).

Discussion

In this study, heart failure patients who received a multidisciplinary intervention (including inpatient education, self‐management support, improved timely medical follow‐up, and better integration between hospital and primary care) showed a trend to improved 1‐year post‐hospital survival, but this appeared to be at the cost of increased readmissions among survivors. This occurred despite our previously reported improved optimization of pharmacological therapy both in‐hospital and posthospital with this program.18

There are a number of potential explanations for this finding, which have important implications for adoption of disease management programs. First, the intervention may not have been of sufficient intensity. Programs primarily aimed at educating providers and patients in evidence‐based guidelines, without structured postdischarge support, have not always improved clinical outcomes.26 In our study, general practitioners were supported to provide improved postdischarge care to their CHF patients, but direct postdischarge patient support was only provided to consenting patients and was limited in scope. There is still some debate about which elements of successful DMPs are most important for efficacy. Most authorities support the central importance of medication optimization, intensive education, and self‐care support. Taylor et al.23 found stronger evidence for programs using individual case management or outreach rather than clinic‐based interventions. Yu et al.27 concluded that outpatient drug titration and ready access to specialist review were factors contributing to success. In our program, even the more intensive intervention did not include regular clinical review by specialist nurses, a system for rapid review in the event of deterioration or supervised drug titration protocols. Furthermore, strategies which prompted more frequent primary care review and improved patient, carer, and general practitioner recognition of disease deterioration may have provided more opportunities to initiate readmission, especially in the absence of an alternative care pathway such as rapid‐access clinics or outreach services.28

Second, this study may reflect the reality of generalizing randomized controlled trial data to an unselected population. Many trials enrolled patients with high anticipated event rates but excluded patients with complex comorbidities, poor life expectancy, and cognitive impairment. Such studies enrolled a high‐risk population (10%‐48% of screened patients randomized) who had a relatively high readmission rate (50%‐60% at 6‐12 months) compared to our unselected population. These studies may overstate the benefits of applying heart failure DMPs in an unselected population. Galbreath et al.29 enrolled a self‐selected community sample of heart failure patients into a disease management program incorporating education, self‐management, telephone support, and advice to primary care providers and home health providers. Like our model, they demonstrated a survival benefit in the intervention group but no reduction in hospital or other healthcare utilization.

Third, only about one‐half of the readmissions were due to heart failure, again reflecting the complexity of this real‐world patient group. Interventions that focus on a single disease in patients with complex comorbidities might be expected to have only limited impact on their subsequent healthcare needs.

Fourth, findings may reflect differences in patient characteristics between the 2 cohorts. While statistical adjustment for measured differences did not have any significant impact on results, unmeasured patient characteristics may have introduced bias. The beforeafter nature of the study also raises the possibility that temporal trends in care practices influenced patient outcomes, such as changing patterns of drug and device therapies. There is conflicting evidence in the literature regarding trends in CHF readmission rates,3032 but it is possible that health system factors external to the study contributed to a higher readmission rate in the later cohort.

Finally, there was a trend toward reduction in mortality within the intervention cohort. These additional survivors might be expected to have more advanced heart failure or other comorbid disease, and therefore may have been more susceptible to deterioration and the need for inpatient care.

Conclusions

We acknowledge the weaknesses inherent in this nonrandomized study design, including convenience sampling, measured and unmeasured confounders and temporal trends in processes and systems of care. Nonetheless, this real world study suggests a note of caution in the widespread enthusiasm for chronic disease management programs. A complex bundle of interventions that resulted in measurable improvements in adherence to evidence‐based guidelines, discharge processes, integration between care providers, and patient education appeared to prolong life expectancy but increase hospital utilization. Mortality reduction in an incurable chronic disease such as heart failure will increase the burden of disease (and therefore treatment costs) unless treatments concurrently reduce disability and the frequency of symptomatic relapse.33 Whether this balance is achieved will depend on patient selection and the intensity and/or components of the intervention. These factors have not been fully defined in the literature to date.

Our study suggests that a widely applied, discharge‐focused intervention which primarily augmented the CHF management knowledge of care providers and patients, and enhanced attendance within the existing care model of primary care and internal medicine/cardiology outpatient services, improved the quality of care and may have reduced mortality at the cost of higher hospital utilization. It raises questions about whether a disease management service can achieve the uncertain promise of reduced readmissions in a cost‐effective manner outside of a high‐risk experimental population.

Congestive heart failure (CHF) is a common disease with high mortality and morbidity.1, 2 Better physiological understanding has led to significant advances in therapy in recent years, with synthesis of this evidence into widely available treatment guidelines.3, 4 However, patients who have had an acute hospitalization with heart failure continue to have a high rate of symptomatic relapse, with up to 25% readmitted within 3 months.2 One of the major challenges in heart failure therapy is to avert these relapses to prevent hospital readmission.

Angiotensin‐converting enzyme (ACE) inhibitors, beta‐blockers, and spironolactone have promised a reduction in hospitalization rates as well as mortality; however, suboptimal prescribing5 and adherence to therapy6, 7 may limit their anticipated benefits. This has led to interest in improved systems of care to reduce hospital utilization. Such approaches have included improved systems for optimizing medications,68 comprehensive discharge planning and postdischarge support,914 and self‐management and case management strategies1517 to enhance patient participation in care.

Combinations of these strategies are known as disease management programs (DMPs), and trials of such combination strategies to improve patient outcomes have been promising.1823 Recognized features4 include skilled multidisciplinary team care; individualized guideline‐based treatment plans that may include dietary and exercise programs as well as optimal pharmacological therapy; patient education and self‐management strategies; improved integration between hospital and community care providers; vigilant follow‐up including prompt review after hospitalization; ready access to expert assessment in the event of deterioration; and regular monitoring with expert titration of therapy, through clinics, home visits, or telemonitoring. Several randomized controlled trials have suggested that DMPs may reduce heart failure‐related9, 1517 and all‐cause9, 10 readmissions. Meta‐analyses12, 1823 have demonstrated reduction in risk of all‐cause readmission of 12% to 25% as well as a reduction in mortality of 14% to 25%.

Trials of DMPs have generally involved careful participant selection, and differences in methods and outcome reporting have led some reviewers to be circumspect in their interpretation of the impact of these programs on readmission rates.23 A large, real‐world quality improvement program conducted as part of the Royal Australasian College of Physicians Clinical Support Systems Project provided an opportunity to measure whether a multifaceted program targeting a representative group of patients with CHF and their healthcare providers could reduce readmission rates. As previously published, this program delivered measurable improvements in processes of care including evidence‐based prescribing, adherence, multidisciplinary involvement, and discharge communication, associated with a reduction in 12‐month mortality.24

Objective

The Brisbane Cardiac Consortium sought to improve processes of care for patients with CHF by using evidence‐based strategies targeting patients and their healthcare providers to optimize uptake of management guidelines, improve discharge processes between hospital and primary care, and increase patient participation in care. We hypothesized that the program would reduce hospital readmissions in the intervention patients in the first 12 months following discharge.

Methods

Results

There were 220 patients identified with a clinical diagnosis of CHF during the baseline period, and 235 during the intervention period. Figure 1 shows ascertainment, in‐hospital mortality, and eligibility rates for the 2 cohorts. Eighty‐nine (45%) of baseline patients and 76 (35%) of intervention patients received intensive posthospital follow‐up as described above. Information on readmission was available for 197 baseline patients and 219 intervention patients discharged alive; this is the sample used for all analyses in this report. Table 1 shows the demographic and clinical characteristics of these patients. Table 2 summarizes the previously reported improvements in processes of care.

Figure 1

Duing the 12‐month follow‐up, 107 (49%) of intervention patients were readmitted to the hospital compared to 71 (36%) of control patients, representing a 1.7‐fold increase in the adjusted probability of readmission in the intervention group (odds ratio [OR] = 1.71, 95% confidence interval [CI] = 1.14‐2.56; P = 0.009). As shown in Table 3, this was partly balanced by a trend toward reduced post‐hospital mortality, such that no significant difference was seen in readmission‐free survival.

Time‐to‐event analysis (Figures 2 and 3) demonstrated similar findings, with a significant reduction in time to first readmission in the intervention group (adjusted hazard ratio [HR] = 1.43; 95% CI = 1.04‐1.97; P = 0.046) but no difference in time to death or first readmission (adjusted HR = 1.14; 95% CI = 0.86‐1.46; P = 0.36).

Figure 2

Figure 3

There was a trend to increased readmissions attributed to heart failure: 47 (21.5%) of intervention patients compared to 33 (16.7%) in the baseline group (OR = 1.30; 95% CI = 0.87‐1.93; P = 0.20). No significant difference was demonstrated in the frequency of readmissions (average 0.75 admission per participant per year in baseline, compared to 0.93 intervention; P = 0.32) nor the mean number of days in hospital in 12 months subsequent to the index admission (5.9 in the baseline group compared to 6.5 in the intervention group; P = 0.1).

Subgroup analysis by intervention intensity showed similar results, with 42 of 76 (55.3%) intensive group participants in the intervention group and 36 of 89 (40.4%) in the baseline group requiring hospital readmission within 12 months. The HR for death or readmission was estimated to be 1.27 (95% CI = 0.85‐1.9).

Discussion

In this study, heart failure patients who received a multidisciplinary intervention (including inpatient education, self‐management support, improved timely medical follow‐up, and better integration between hospital and primary care) showed a trend to improved 1‐year post‐hospital survival, but this appeared to be at the cost of increased readmissions among survivors. This occurred despite our previously reported improved optimization of pharmacological therapy both in‐hospital and posthospital with this program.18

There are a number of potential explanations for this finding, which have important implications for adoption of disease management programs. First, the intervention may not have been of sufficient intensity. Programs primarily aimed at educating providers and patients in evidence‐based guidelines, without structured postdischarge support, have not always improved clinical outcomes.26 In our study, general practitioners were supported to provide improved postdischarge care to their CHF patients, but direct postdischarge patient support was only provided to consenting patients and was limited in scope. There is still some debate about which elements of successful DMPs are most important for efficacy. Most authorities support the central importance of medication optimization, intensive education, and self‐care support. Taylor et al.23 found stronger evidence for programs using individual case management or outreach rather than clinic‐based interventions. Yu et al.27 concluded that outpatient drug titration and ready access to specialist review were factors contributing to success. In our program, even the more intensive intervention did not include regular clinical review by specialist nurses, a system for rapid review in the event of deterioration or supervised drug titration protocols. Furthermore, strategies which prompted more frequent primary care review and improved patient, carer, and general practitioner recognition of disease deterioration may have provided more opportunities to initiate readmission, especially in the absence of an alternative care pathway such as rapid‐access clinics or outreach services.28

Second, this study may reflect the reality of generalizing randomized controlled trial data to an unselected population. Many trials enrolled patients with high anticipated event rates but excluded patients with complex comorbidities, poor life expectancy, and cognitive impairment. Such studies enrolled a high‐risk population (10%‐48% of screened patients randomized) who had a relatively high readmission rate (50%‐60% at 6‐12 months) compared to our unselected population. These studies may overstate the benefits of applying heart failure DMPs in an unselected population. Galbreath et al.29 enrolled a self‐selected community sample of heart failure patients into a disease management program incorporating education, self‐management, telephone support, and advice to primary care providers and home health providers. Like our model, they demonstrated a survival benefit in the intervention group but no reduction in hospital or other healthcare utilization.

Third, only about one‐half of the readmissions were due to heart failure, again reflecting the complexity of this real‐world patient group. Interventions that focus on a single disease in patients with complex comorbidities might be expected to have only limited impact on their subsequent healthcare needs.

Fourth, findings may reflect differences in patient characteristics between the 2 cohorts. While statistical adjustment for measured differences did not have any significant impact on results, unmeasured patient characteristics may have introduced bias. The beforeafter nature of the study also raises the possibility that temporal trends in care practices influenced patient outcomes, such as changing patterns of drug and device therapies. There is conflicting evidence in the literature regarding trends in CHF readmission rates,3032 but it is possible that health system factors external to the study contributed to a higher readmission rate in the later cohort.

Finally, there was a trend toward reduction in mortality within the intervention cohort. These additional survivors might be expected to have more advanced heart failure or other comorbid disease, and therefore may have been more susceptible to deterioration and the need for inpatient care.

Conclusions

We acknowledge the weaknesses inherent in this nonrandomized study design, including convenience sampling, measured and unmeasured confounders and temporal trends in processes and systems of care. Nonetheless, this real world study suggests a note of caution in the widespread enthusiasm for chronic disease management programs. A complex bundle of interventions that resulted in measurable improvements in adherence to evidence‐based guidelines, discharge processes, integration between care providers, and patient education appeared to prolong life expectancy but increase hospital utilization. Mortality reduction in an incurable chronic disease such as heart failure will increase the burden of disease (and therefore treatment costs) unless treatments concurrently reduce disability and the frequency of symptomatic relapse.33 Whether this balance is achieved will depend on patient selection and the intensity and/or components of the intervention. These factors have not been fully defined in the literature to date.

Our study suggests that a widely applied, discharge‐focused intervention which primarily augmented the CHF management knowledge of care providers and patients, and enhanced attendance within the existing care model of primary care and internal medicine/cardiology outpatient services, improved the quality of care and may have reduced mortality at the cost of higher hospital utilization. It raises questions about whether a disease management service can achieve the uncertain promise of reduced readmissions in a cost‐effective manner outside of a high‐risk experimental population.

Congestive heart failure (CHF) is a common disease with high mortality and morbidity.1, 2 Better physiological understanding has led to significant advances in therapy in recent years, with synthesis of this evidence into widely available treatment guidelines.3, 4 However, patients who have had an acute hospitalization with heart failure continue to have a high rate of symptomatic relapse, with up to 25% readmitted within 3 months.2 One of the major challenges in heart failure therapy is to avert these relapses to prevent hospital readmission.

Angiotensin‐converting enzyme (ACE) inhibitors, beta‐blockers, and spironolactone have promised a reduction in hospitalization rates as well as mortality; however, suboptimal prescribing5 and adherence to therapy6, 7 may limit their anticipated benefits. This has led to interest in improved systems of care to reduce hospital utilization. Such approaches have included improved systems for optimizing medications,68 comprehensive discharge planning and postdischarge support,914 and self‐management and case management strategies1517 to enhance patient participation in care.

Combinations of these strategies are known as disease management programs (DMPs), and trials of such combination strategies to improve patient outcomes have been promising.1823 Recognized features4 include skilled multidisciplinary team care; individualized guideline‐based treatment plans that may include dietary and exercise programs as well as optimal pharmacological therapy; patient education and self‐management strategies; improved integration between hospital and community care providers; vigilant follow‐up including prompt review after hospitalization; ready access to expert assessment in the event of deterioration; and regular monitoring with expert titration of therapy, through clinics, home visits, or telemonitoring. Several randomized controlled trials have suggested that DMPs may reduce heart failure‐related9, 1517 and all‐cause9, 10 readmissions. Meta‐analyses12, 1823 have demonstrated reduction in risk of all‐cause readmission of 12% to 25% as well as a reduction in mortality of 14% to 25%.

Trials of DMPs have generally involved careful participant selection, and differences in methods and outcome reporting have led some reviewers to be circumspect in their interpretation of the impact of these programs on readmission rates.23 A large, real‐world quality improvement program conducted as part of the Royal Australasian College of Physicians Clinical Support Systems Project provided an opportunity to measure whether a multifaceted program targeting a representative group of patients with CHF and their healthcare providers could reduce readmission rates. As previously published, this program delivered measurable improvements in processes of care including evidence‐based prescribing, adherence, multidisciplinary involvement, and discharge communication, associated with a reduction in 12‐month mortality.24

Objective

The Brisbane Cardiac Consortium sought to improve processes of care for patients with CHF by using evidence‐based strategies targeting patients and their healthcare providers to optimize uptake of management guidelines, improve discharge processes between hospital and primary care, and increase patient participation in care. We hypothesized that the program would reduce hospital readmissions in the intervention patients in the first 12 months following discharge.

Methods

Results

There were 220 patients identified with a clinical diagnosis of CHF during the baseline period, and 235 during the intervention period. Figure 1 shows ascertainment, in‐hospital mortality, and eligibility rates for the 2 cohorts. Eighty‐nine (45%) of baseline patients and 76 (35%) of intervention patients received intensive posthospital follow‐up as described above. Information on readmission was available for 197 baseline patients and 219 intervention patients discharged alive; this is the sample used for all analyses in this report. Table 1 shows the demographic and clinical characteristics of these patients. Table 2 summarizes the previously reported improvements in processes of care.

Figure 1

Duing the 12‐month follow‐up, 107 (49%) of intervention patients were readmitted to the hospital compared to 71 (36%) of control patients, representing a 1.7‐fold increase in the adjusted probability of readmission in the intervention group (odds ratio [OR] = 1.71, 95% confidence interval [CI] = 1.14‐2.56; P = 0.009). As shown in Table 3, this was partly balanced by a trend toward reduced post‐hospital mortality, such that no significant difference was seen in readmission‐free survival.

Time‐to‐event analysis (Figures 2 and 3) demonstrated similar findings, with a significant reduction in time to first readmission in the intervention group (adjusted hazard ratio [HR] = 1.43; 95% CI = 1.04‐1.97; P = 0.046) but no difference in time to death or first readmission (adjusted HR = 1.14; 95% CI = 0.86‐1.46; P = 0.36).

Figure 2

Figure 3

There was a trend to increased readmissions attributed to heart failure: 47 (21.5%) of intervention patients compared to 33 (16.7%) in the baseline group (OR = 1.30; 95% CI = 0.87‐1.93; P = 0.20). No significant difference was demonstrated in the frequency of readmissions (average 0.75 admission per participant per year in baseline, compared to 0.93 intervention; P = 0.32) nor the mean number of days in hospital in 12 months subsequent to the index admission (5.9 in the baseline group compared to 6.5 in the intervention group; P = 0.1).

Subgroup analysis by intervention intensity showed similar results, with 42 of 76 (55.3%) intensive group participants in the intervention group and 36 of 89 (40.4%) in the baseline group requiring hospital readmission within 12 months. The HR for death or readmission was estimated to be 1.27 (95% CI = 0.85‐1.9).

Discussion

In this study, heart failure patients who received a multidisciplinary intervention (including inpatient education, self‐management support, improved timely medical follow‐up, and better integration between hospital and primary care) showed a trend to improved 1‐year post‐hospital survival, but this appeared to be at the cost of increased readmissions among survivors. This occurred despite our previously reported improved optimization of pharmacological therapy both in‐hospital and posthospital with this program.18

There are a number of potential explanations for this finding, which have important implications for adoption of disease management programs. First, the intervention may not have been of sufficient intensity. Programs primarily aimed at educating providers and patients in evidence‐based guidelines, without structured postdischarge support, have not always improved clinical outcomes.26 In our study, general practitioners were supported to provide improved postdischarge care to their CHF patients, but direct postdischarge patient support was only provided to consenting patients and was limited in scope. There is still some debate about which elements of successful DMPs are most important for efficacy. Most authorities support the central importance of medication optimization, intensive education, and self‐care support. Taylor et al.23 found stronger evidence for programs using individual case management or outreach rather than clinic‐based interventions. Yu et al.27 concluded that outpatient drug titration and ready access to specialist review were factors contributing to success. In our program, even the more intensive intervention did not include regular clinical review by specialist nurses, a system for rapid review in the event of deterioration or supervised drug titration protocols. Furthermore, strategies which prompted more frequent primary care review and improved patient, carer, and general practitioner recognition of disease deterioration may have provided more opportunities to initiate readmission, especially in the absence of an alternative care pathway such as rapid‐access clinics or outreach services.28

Second, this study may reflect the reality of generalizing randomized controlled trial data to an unselected population. Many trials enrolled patients with high anticipated event rates but excluded patients with complex comorbidities, poor life expectancy, and cognitive impairment. Such studies enrolled a high‐risk population (10%‐48% of screened patients randomized) who had a relatively high readmission rate (50%‐60% at 6‐12 months) compared to our unselected population. These studies may overstate the benefits of applying heart failure DMPs in an unselected population. Galbreath et al.29 enrolled a self‐selected community sample of heart failure patients into a disease management program incorporating education, self‐management, telephone support, and advice to primary care providers and home health providers. Like our model, they demonstrated a survival benefit in the intervention group but no reduction in hospital or other healthcare utilization.

Third, only about one‐half of the readmissions were due to heart failure, again reflecting the complexity of this real‐world patient group. Interventions that focus on a single disease in patients with complex comorbidities might be expected to have only limited impact on their subsequent healthcare needs.

Fourth, findings may reflect differences in patient characteristics between the 2 cohorts. While statistical adjustment for measured differences did not have any significant impact on results, unmeasured patient characteristics may have introduced bias. The beforeafter nature of the study also raises the possibility that temporal trends in care practices influenced patient outcomes, such as changing patterns of drug and device therapies. There is conflicting evidence in the literature regarding trends in CHF readmission rates,3032 but it is possible that health system factors external to the study contributed to a higher readmission rate in the later cohort.

Finally, there was a trend toward reduction in mortality within the intervention cohort. These additional survivors might be expected to have more advanced heart failure or other comorbid disease, and therefore may have been more susceptible to deterioration and the need for inpatient care.

Conclusions

We acknowledge the weaknesses inherent in this nonrandomized study design, including convenience sampling, measured and unmeasured confounders and temporal trends in processes and systems of care. Nonetheless, this real world study suggests a note of caution in the widespread enthusiasm for chronic disease management programs. A complex bundle of interventions that resulted in measurable improvements in adherence to evidence‐based guidelines, discharge processes, integration between care providers, and patient education appeared to prolong life expectancy but increase hospital utilization. Mortality reduction in an incurable chronic disease such as heart failure will increase the burden of disease (and therefore treatment costs) unless treatments concurrently reduce disability and the frequency of symptomatic relapse.33 Whether this balance is achieved will depend on patient selection and the intensity and/or components of the intervention. These factors have not been fully defined in the literature to date.

Our study suggests that a widely applied, discharge‐focused intervention which primarily augmented the CHF management knowledge of care providers and patients, and enhanced attendance within the existing care model of primary care and internal medicine/cardiology outpatient services, improved the quality of care and may have reduced mortality at the cost of higher hospital utilization. It raises questions about whether a disease management service can achieve the uncertain promise of reduced readmissions in a cost‐effective manner outside of a high‐risk experimental population.

Gastrointestinal bleeding (GIB) is a frequent reason for acute hospitalization, with estimated rates of hospitalization at 375 per 100,000 per year in the United States.1 GIB is not a specific disease but rather a diverse set of conditions that lead to the clinical manifestations associated with bleeding into the gastrointestinal tract. One of the most commonly used organizing frameworks in gastrointestinal bleeding is the differentiation between upper gastrointestinal bleeding (UGIB) and lower gastrointestinal bleeding (LGIB). There are important differences in the etiologies between the 2 sources. For example, acid‐related disease is a common etiology in UGIB but does not occur in LGIB. While some aspects of the acute management are shared between UGIB and LGIB, important differences exist in the management, including initial endoscopy and medication choice. There have been few direct comparisons of rates, resource use, and clinical outcomes between UGIB and LGIB.

Historically, rates of UGIB have been reported to exceed those of LGIB by 2‐fold to 8‐fold.25 Protocols, clinical practice guidelines, and policy decisions reflect this emphasis on UGIB.68 Among 9 guidelines hosted by National Guideline Clearinghouse addressing GIB, 6 were focused on UGIB, 2 on both UGIB and LGIB, and only 1 on LGIB.9 There are several reasons to believe that these relative incidence rates may not be accurate. First, recent advances in therapy and prevention of UGIB, such as the treatment of Helicobacter pylori infection; proton pump inhibitors (PPIs); and selective cyclooxygenase‐2 (COX‐2) inhibitors, may have affected the epidemiology of gastrointestinal bleeding.1016 Among these therapies, only COX‐2 inhibitors may also reduce the incidence of LGIB.14, 1618 Therefore, these advances may result in a disproportionate drop in UGIB relative to LGIB. In addition, known risk factors for both LGIB and UGIB, including advancing age and renal failure, are increasing in the general population.5, 19, 20 Finally, given the recent increased recommendations for aspirin therapy and systemic anticoagulation, exposure to aspirin and warfarin have increased, both risk factors for LGIB and UGIB.2124 Indeed, recent studies in the epidemiology of UGIB do suggest a changing pattern of etiologies of UGIB reflecting these advances.25 One study examining rates of both UGIB and LGIB demonstrate a decrease in hospitalizations overall for GIB driven by a reduction in UGIB while at the same time reporting an increase in the incidence of hospitalization for LGIB.1

In addition to a changing epidemiology, a second reason for a potential underestimation of LGIB incidence is one of methodology. There are well‐recognized limitations with using purely administrative data due to difficulties in accurately identifying patients with LGIB.26

Studies using large administrative databases may not accurately identify LGIB because of the poor sensitivity and specificity of International Classification of Diseases, Ninth revision, Clinical Modification (ICD‐9) codes for LGIB.5 While there are standard methods of identifying patients with UGIB using ICD‐9 codes,19 there is not an accepted standard for LGIB. Thus, estimates using only ICD‐9 codes may overidentify or underidentify patients with LGIB. Prior studies that have most accurately identified patients with LGIB used a 2‐step method to address this issue. The initial ICD‐9 identification included a high sensitivity/low specificity approach. These identified patient charts undergo chart review to confirm the presence of an LGIB.5 This method is labor intensive and cannot be done using administrative databases. No direct comparison of UGIB to LGIB among hospitalized patients using this 2‐step method has been done recently.

The current emphasis on UGIB as seen in the published guidelines could also be supported if patients with UGIB had greater resource utilization or worse clinical outcomes. Limited direct comparisons for these outcomes are available. However, 1 administrative database study reported similar mortality rates for UGIB (2.7%) and LGIB (2.9%) in 2006.1 No direct comparisons of other clinical outcomes or resource use outcomes are available. Therefore, the emphasis on UGIB in publications and guidelines is best supported by the incidence rates that are, as has already been discussed, problematic.

We conducted a retrospective cohort study to examine the incidences of UGIB and LGIB among patients admitted to an academic medical center over 2 years using methods designed to optimally identify patients with either UGIB or LGIB. Our study also examined differences in clinical outcomes and resource utilization between subjects with UGIB and LGIB to examine the relative severity of these 2 clinical entities. These results may be useful in determining the need to reconsider clinical approaches as well as protocols and guidelines among patients with gastrointestinal bleeding.

Patients and Methods

Results

During the 2 years of observation, a total of 7741 subjects were admitted to the internal medicine service and enrolled in the hospitalist study. Of these, 1014 had a primary or secondary ICD‐9 code that may be consistent with UGIB or LGIB and underwent chart review to determine if they had an acute GIB. Out of 1014 subjects, 647 were determined not to have an acute GI hemorrhage and were excluded from the remaining analyses; 367 of the 1104 subjects identified by ICD‐9 codes were found to have a clinical presentation consistent with GIB and were included in this study. A total of 180 of these 367 had UGIB and 187 had LGIB. The mean age was 62.4 years, 56.7% were female, 82.6% were African American, 12.7% were Caucasian, and the mean Charlson index was 1.5. (Table 1) Among baseline characteristics, both gender and age were statistically associated with a difference in rates of upper vs. lower source bleeding, with LGIB patients more likely to be female (P = 0.01) and older (P < 0.001). Etiologies of UGIB include erosive disease, peptic ulcer disease, variceal bleeding, arteriovenous malformation, and malignancy. Etiologies of LGIB include: diverticulosis, colitis, arteriovenous malformation, cancer, ischemic colitis, polyp, hemorrhoidal bleed, ulcer, inflammatory bowel disease, other, and not determined (Table 2).

Baseline use of medications known to be associated with either increased or decreased risk of GIB was common. Approximately one‐third of subjects with both LGIB and UGIB used aspirin and 10% used warfarin. LGIB subjects were less likely to use an nsNSAID (P < 0.001), but more likely to use a proton pump inhibitor (PPI) (P = 0.06) (Table 3).

Key initial clinical presentation findings included vital sign abnormalities and admission hemoglobin levels. While hypotension was not common (4.7%), resting tachycardia (37%) and orthostasis (16%) were seen frequently. Subjects with LGIB were significantly less likely than those with UGIB to present with orthostasis (8.8% vs. 21.0%, respectively; P = 0.006) and resting tachycardia (32.3% vs. 42.5%, respectively; P = 0.04). Subjects with LGIB had a higher admission hemoglobin than those with UGIB (10.7 vs. 9.7, respectively; P < 0.001) (Table 4).

We also examined several clinical outcomes. When comparing LGIB to UGIB patients for these clinical outcomes using bivariate and multivariate statistics, there was no difference for in‐hospital mortality (1.1% vs. 1.1%), transfer to ICU (16.0% vs. 13.9%), 30‐day readmission (5.9% vs.7.8%), number of red blood cell (RBC) transfusions (2.7 vs. 2.4), or need for GI surgery (1.1% vs. 0.0%). The mean drop in hemoglobin was greater among subjects with LGIB compared to UGIB (1.9 g/dL vs. 1.5 g/dL, respectively) by both bivariate (P = 0.01) and multivariate (P = 0.003) analyses (Table 5).

Mean costs were $11,892 for LGIB and $14,301 for UGIB and median costs were $7,890 for LGIB and $9,548 for UGIB, but were not statistically different. LOS was also similar between subjects with LGIB (5.1 days) and UGIB (5.7 days). In bivariate and multivariate analyses, UGIB subjects had a similar mean number of endoscopic procedures (1.3) compared to LGIB subjects (1.2). Thirteen percent of subjects with UGIB required a second EGD while only 8% of subjects with LGIB required 2 colonoscopies. In addition, 29% of subjects with LGIB received an EGD while only 16% of subjects with an UGIB received a colonoscopy (P = 0.001) (Table 6).

Conclusions

This study represents one of the largest direct comparisons of LGIB to UGIB not based on administrative databases. The most striking finding was the nearly equal rates of LGIB and UGIB. There are 2 likely explanations for this surprising result. First, there may be methodological reasons that we identified a greater proportion of true LGIBs; our study used a highly sensitive search strategy of ICD‐9 coding with confirmatory chart abstraction to ensure that as many LGIB and UGIB cases would be identified as possible while also excluding cases not meeting accepted criteria for GIB. The second possibility is that there is an actual change in epidemiology of GIB. Known risk factors for LGIB are increasing such as advancing age, increased use of chronic aspirin therapy, and renal disease. At the same time, significant advances in the treatment and prevention of UGIB have been made. Recent studies have demonstrated similar trends in admissions for upper and lower GI complications, suggesting that there may be a changing epidemiology due primarily to reductions in upper GI complications.1, 16

Either explanation would have implications for the care of patients with GIB. Clinical decision‐making based on prior literature would support that in ambiguous clinical situations and initial evaluation for an UGIB is appropriate. Most risk stratification literature and clinical guidelines focus on UGIB. If rates of LGIB and UGIB are similar, then existing clinical decision protocols may need to be reevaluated to incorporate the higher likelihood of LGIB. This reevaluation would be less important if the clinical outcomes or resource utilization of UGIB was significantly greater than that for LGIB, but we did not find this was the case. Similarly, if the ability to distinguish between LGIB and UGIB were robust on clinical signs and symptoms, then a reevaluation would be less important. However, we found fairly similar numbers of patients initially receiving evaluation for UGIB then being evaluated for LGIB as we found patients initially receiving evaluation for LGIB then being evaluated for UGIB. This suggests the potential benefit of clinical decision protocols that could better distinguish between UGIB and LGIB and account for the potentially higher incidence of LGIB than previously thought.

In addition to affecting the attention paid to LGIB for acute management, a changed understanding of incidence could also affect the attention paid to prevention of LGIB. Of the recent nonendoscopic advances in the treatment and prevention of GIB, only the use of COX‐2s (when used in place of traditional nsNSAIDs) reduces the risk of both LGIB and UGIB;14, 1618 H .pylori treatment and PPIs only prevent UGIB. Therefore, if the clinical and financial burdens of LGIB are similar to those seen in UGIB, more attention may need to be focused on preventing LGIB.

Baseline medication use was notable primarily for the similarities between UGIB and LGIB. Agents known to affect the rates of GIB were common in both groups. Over one‐third of the population was using aspirin and 10% were taking warfarin. Over 20% of subjects were taking an nsNSAID or a COX‐2 inhibitor. Almost one‐quarter of subjects were taking a PPI, agents known to decrease rates of UGIB and potentially increase LGIB through the risk of C. difficile colitis. Notably, the only statistically significant difference in baseline medication use between subjects with UGIB and LGIB was the more than 3‐fold higher use of nsNSAIDs in patients with UGIB as compared to LGIB. While current guidelines are not clear and consistent about which populations of at‐risk patients should receive GI prophylaxis,2830 these results suggest that patients admitted with GIB are very likely to be taking medications which impact the risk of GIB.

In terms of disease severity, the clinical presentation at admission suggests a greater degree of hemodynamic instability among subjects with UGIB. Rates of orthostatic hypotension and resting tachycardia are higher in UGIB subjects, as well as having a lower mean hemoglobin levels at presentation. However, despite the more severe clinical presentation, clinical outcomes did not differ significantly between the 2 bleeding sources. Thus, the most relevant clinical outcomes suggest that the severity of both LGIB and UGIB are similar. This similarity again suggest that the clinical burden of LGIB is not significantly different than UGIB.

Our results concerning resource utilization demonstrate a similar pattern. While the point estimates for costs and LOS suggest that UGIB may be associated with higher resource utilization, these differences were not significant in either bivariate or multivariate analyses. Those subjects with UGIB did receive more total endoscopic procedures than subjects with LGIB. More interesting though was that 24% of all subjects received an endoscopy of the opposite site (LGIB with EGD and UGIB with colonoscopy). These results suggest that the site of bleeding is not clear in a significant proportion of patients who present with GIB. These additional endoscopies are associated with increased risk, costs, LOS, and discomfort to patients. Improving our ability to accurately predict the source (upper vs. lower) of bleeding would allow us to reduce the number of these excess endoscopies. Additionally, it is interesting that despite the almost universal use of endoscopies, 20% of LGIB and 14% of UGIB subjects could not have a specific etiology identified during endoscopy or subsequent workup.

There are some important limitations to this study. While the sample size is among the largest of its type involving chart abstraction, it may be underpowered to detect some differences. Additionally, our results are from a single urban academic medical center with a patient population that is predominantly African American, which may limit generalizability. This study required consent and therefore only examines a subset of patients admitted to the medical center with GIB, which could potentially introduce bias into the sample. However, it is not clear why there would be systematic differences in subjects who choose to consent vs. those who decide not to consent that would affect the results of this study in substantive ways.

Despite significant efforts at identifying all subjects with GIB admitted during this time period, there were potential methodological reasons that may have resulted in some cases being missed. Only subjects admitted to a medicine service were approached for consent. All subjects in this medical center with GIB are admitted to a medicine service. We captured all subjects who were initially admitted to a medicine service as well as those admitted initially to an ICU and then transferred to the floor at any point prior to discharge. It is possible, though, that a subject would be admitted to an ICU for GIB and die prior to being transferred to the floor. While it is the impression of the director of the ICU that this would be a very unusual event, as most of the patients would be discharged to the floor prior to death (personal communication), given the very low mortality rate seen in this study, small numbers of missed events could have a significant impact on the interpretation of in‐hospital mortality results. It is also important to note that this medical center did not have the ability to perform endoscopy prior to admission for patients with GIB at the time of the study; all patients who presented with GIB would have been admitted and identified for this study. Finally, we were unable to routinely identify the rationale for obtaining an endoscopic exam. We assumed that all endoscopic exams were done for the purpose of evaluating and/or treating the GIB for which the subject was admitted. It is possible that some subjects had additional endoscopies for other reasons, which would have led to our overestimating the rates of additional endoscopies for GIB.

This study highlights the similarities between LGIB and UGIB rather than the differences. There were few significant differences between the 2 bleeding sources in terms of incidence, clinical outcomes, and resource utilization. In fact, the study also suggests that determining the source of bleeding may not be clear, given the high rates of opposite site endoscopies. While this study did reveal several similarities between UGIB and LGIB, it also highlights the need to identify improved strategies to improve the sensitivity and specificity of identification of LGIB compared to UGIB, both for clinical purposes and for research. The value of such improved clinical algorithms have the potential to improve both the cost and outcomes of care, while better algorithms for separating UGIB and LGIB using administrative data might help produce more precise estimates of costs and clinical outcomes, and aid in the development of risk stratification models.

Gastrointestinal bleeding (GIB) is a frequent reason for acute hospitalization, with estimated rates of hospitalization at 375 per 100,000 per year in the United States.1 GIB is not a specific disease but rather a diverse set of conditions that lead to the clinical manifestations associated with bleeding into the gastrointestinal tract. One of the most commonly used organizing frameworks in gastrointestinal bleeding is the differentiation between upper gastrointestinal bleeding (UGIB) and lower gastrointestinal bleeding (LGIB). There are important differences in the etiologies between the 2 sources. For example, acid‐related disease is a common etiology in UGIB but does not occur in LGIB. While some aspects of the acute management are shared between UGIB and LGIB, important differences exist in the management, including initial endoscopy and medication choice. There have been few direct comparisons of rates, resource use, and clinical outcomes between UGIB and LGIB.

Historically, rates of UGIB have been reported to exceed those of LGIB by 2‐fold to 8‐fold.25 Protocols, clinical practice guidelines, and policy decisions reflect this emphasis on UGIB.68 Among 9 guidelines hosted by National Guideline Clearinghouse addressing GIB, 6 were focused on UGIB, 2 on both UGIB and LGIB, and only 1 on LGIB.9 There are several reasons to believe that these relative incidence rates may not be accurate. First, recent advances in therapy and prevention of UGIB, such as the treatment of Helicobacter pylori infection; proton pump inhibitors (PPIs); and selective cyclooxygenase‐2 (COX‐2) inhibitors, may have affected the epidemiology of gastrointestinal bleeding.1016 Among these therapies, only COX‐2 inhibitors may also reduce the incidence of LGIB.14, 1618 Therefore, these advances may result in a disproportionate drop in UGIB relative to LGIB. In addition, known risk factors for both LGIB and UGIB, including advancing age and renal failure, are increasing in the general population.5, 19, 20 Finally, given the recent increased recommendations for aspirin therapy and systemic anticoagulation, exposure to aspirin and warfarin have increased, both risk factors for LGIB and UGIB.2124 Indeed, recent studies in the epidemiology of UGIB do suggest a changing pattern of etiologies of UGIB reflecting these advances.25 One study examining rates of both UGIB and LGIB demonstrate a decrease in hospitalizations overall for GIB driven by a reduction in UGIB while at the same time reporting an increase in the incidence of hospitalization for LGIB.1

In addition to a changing epidemiology, a second reason for a potential underestimation of LGIB incidence is one of methodology. There are well‐recognized limitations with using purely administrative data due to difficulties in accurately identifying patients with LGIB.26

Studies using large administrative databases may not accurately identify LGIB because of the poor sensitivity and specificity of International Classification of Diseases, Ninth revision, Clinical Modification (ICD‐9) codes for LGIB.5 While there are standard methods of identifying patients with UGIB using ICD‐9 codes,19 there is not an accepted standard for LGIB. Thus, estimates using only ICD‐9 codes may overidentify or underidentify patients with LGIB. Prior studies that have most accurately identified patients with LGIB used a 2‐step method to address this issue. The initial ICD‐9 identification included a high sensitivity/low specificity approach. These identified patient charts undergo chart review to confirm the presence of an LGIB.5 This method is labor intensive and cannot be done using administrative databases. No direct comparison of UGIB to LGIB among hospitalized patients using this 2‐step method has been done recently.

The current emphasis on UGIB as seen in the published guidelines could also be supported if patients with UGIB had greater resource utilization or worse clinical outcomes. Limited direct comparisons for these outcomes are available. However, 1 administrative database study reported similar mortality rates for UGIB (2.7%) and LGIB (2.9%) in 2006.1 No direct comparisons of other clinical outcomes or resource use outcomes are available. Therefore, the emphasis on UGIB in publications and guidelines is best supported by the incidence rates that are, as has already been discussed, problematic.

We conducted a retrospective cohort study to examine the incidences of UGIB and LGIB among patients admitted to an academic medical center over 2 years using methods designed to optimally identify patients with either UGIB or LGIB. Our study also examined differences in clinical outcomes and resource utilization between subjects with UGIB and LGIB to examine the relative severity of these 2 clinical entities. These results may be useful in determining the need to reconsider clinical approaches as well as protocols and guidelines among patients with gastrointestinal bleeding.

Patients and Methods

Results

During the 2 years of observation, a total of 7741 subjects were admitted to the internal medicine service and enrolled in the hospitalist study. Of these, 1014 had a primary or secondary ICD‐9 code that may be consistent with UGIB or LGIB and underwent chart review to determine if they had an acute GIB. Out of 1014 subjects, 647 were determined not to have an acute GI hemorrhage and were excluded from the remaining analyses; 367 of the 1104 subjects identified by ICD‐9 codes were found to have a clinical presentation consistent with GIB and were included in this study. A total of 180 of these 367 had UGIB and 187 had LGIB. The mean age was 62.4 years, 56.7% were female, 82.6% were African American, 12.7% were Caucasian, and the mean Charlson index was 1.5. (Table 1) Among baseline characteristics, both gender and age were statistically associated with a difference in rates of upper vs. lower source bleeding, with LGIB patients more likely to be female (P = 0.01) and older (P < 0.001). Etiologies of UGIB include erosive disease, peptic ulcer disease, variceal bleeding, arteriovenous malformation, and malignancy. Etiologies of LGIB include: diverticulosis, colitis, arteriovenous malformation, cancer, ischemic colitis, polyp, hemorrhoidal bleed, ulcer, inflammatory bowel disease, other, and not determined (Table 2).

Baseline use of medications known to be associated with either increased or decreased risk of GIB was common. Approximately one‐third of subjects with both LGIB and UGIB used aspirin and 10% used warfarin. LGIB subjects were less likely to use an nsNSAID (P < 0.001), but more likely to use a proton pump inhibitor (PPI) (P = 0.06) (Table 3).

Key initial clinical presentation findings included vital sign abnormalities and admission hemoglobin levels. While hypotension was not common (4.7%), resting tachycardia (37%) and orthostasis (16%) were seen frequently. Subjects with LGIB were significantly less likely than those with UGIB to present with orthostasis (8.8% vs. 21.0%, respectively; P = 0.006) and resting tachycardia (32.3% vs. 42.5%, respectively; P = 0.04). Subjects with LGIB had a higher admission hemoglobin than those with UGIB (10.7 vs. 9.7, respectively; P < 0.001) (Table 4).

We also examined several clinical outcomes. When comparing LGIB to UGIB patients for these clinical outcomes using bivariate and multivariate statistics, there was no difference for in‐hospital mortality (1.1% vs. 1.1%), transfer to ICU (16.0% vs. 13.9%), 30‐day readmission (5.9% vs.7.8%), number of red blood cell (RBC) transfusions (2.7 vs. 2.4), or need for GI surgery (1.1% vs. 0.0%). The mean drop in hemoglobin was greater among subjects with LGIB compared to UGIB (1.9 g/dL vs. 1.5 g/dL, respectively) by both bivariate (P = 0.01) and multivariate (P = 0.003) analyses (Table 5).

Mean costs were $11,892 for LGIB and $14,301 for UGIB and median costs were $7,890 for LGIB and $9,548 for UGIB, but were not statistically different. LOS was also similar between subjects with LGIB (5.1 days) and UGIB (5.7 days). In bivariate and multivariate analyses, UGIB subjects had a similar mean number of endoscopic procedures (1.3) compared to LGIB subjects (1.2). Thirteen percent of subjects with UGIB required a second EGD while only 8% of subjects with LGIB required 2 colonoscopies. In addition, 29% of subjects with LGIB received an EGD while only 16% of subjects with an UGIB received a colonoscopy (P = 0.001) (Table 6).

Conclusions

This study represents one of the largest direct comparisons of LGIB to UGIB not based on administrative databases. The most striking finding was the nearly equal rates of LGIB and UGIB. There are 2 likely explanations for this surprising result. First, there may be methodological reasons that we identified a greater proportion of true LGIBs; our study used a highly sensitive search strategy of ICD‐9 coding with confirmatory chart abstraction to ensure that as many LGIB and UGIB cases would be identified as possible while also excluding cases not meeting accepted criteria for GIB. The second possibility is that there is an actual change in epidemiology of GIB. Known risk factors for LGIB are increasing such as advancing age, increased use of chronic aspirin therapy, and renal disease. At the same time, significant advances in the treatment and prevention of UGIB have been made. Recent studies have demonstrated similar trends in admissions for upper and lower GI complications, suggesting that there may be a changing epidemiology due primarily to reductions in upper GI complications.1, 16

Either explanation would have implications for the care of patients with GIB. Clinical decision‐making based on prior literature would support that in ambiguous clinical situations and initial evaluation for an UGIB is appropriate. Most risk stratification literature and clinical guidelines focus on UGIB. If rates of LGIB and UGIB are similar, then existing clinical decision protocols may need to be reevaluated to incorporate the higher likelihood of LGIB. This reevaluation would be less important if the clinical outcomes or resource utilization of UGIB was significantly greater than that for LGIB, but we did not find this was the case. Similarly, if the ability to distinguish between LGIB and UGIB were robust on clinical signs and symptoms, then a reevaluation would be less important. However, we found fairly similar numbers of patients initially receiving evaluation for UGIB then being evaluated for LGIB as we found patients initially receiving evaluation for LGIB then being evaluated for UGIB. This suggests the potential benefit of clinical decision protocols that could better distinguish between UGIB and LGIB and account for the potentially higher incidence of LGIB than previously thought.

In addition to affecting the attention paid to LGIB for acute management, a changed understanding of incidence could also affect the attention paid to prevention of LGIB. Of the recent nonendoscopic advances in the treatment and prevention of GIB, only the use of COX‐2s (when used in place of traditional nsNSAIDs) reduces the risk of both LGIB and UGIB;14, 1618 H .pylori treatment and PPIs only prevent UGIB. Therefore, if the clinical and financial burdens of LGIB are similar to those seen in UGIB, more attention may need to be focused on preventing LGIB.

Baseline medication use was notable primarily for the similarities between UGIB and LGIB. Agents known to affect the rates of GIB were common in both groups. Over one‐third of the population was using aspirin and 10% were taking warfarin. Over 20% of subjects were taking an nsNSAID or a COX‐2 inhibitor. Almost one‐quarter of subjects were taking a PPI, agents known to decrease rates of UGIB and potentially increase LGIB through the risk of C. difficile colitis. Notably, the only statistically significant difference in baseline medication use between subjects with UGIB and LGIB was the more than 3‐fold higher use of nsNSAIDs in patients with UGIB as compared to LGIB. While current guidelines are not clear and consistent about which populations of at‐risk patients should receive GI prophylaxis,2830 these results suggest that patients admitted with GIB are very likely to be taking medications which impact the risk of GIB.

In terms of disease severity, the clinical presentation at admission suggests a greater degree of hemodynamic instability among subjects with UGIB. Rates of orthostatic hypotension and resting tachycardia are higher in UGIB subjects, as well as having a lower mean hemoglobin levels at presentation. However, despite the more severe clinical presentation, clinical outcomes did not differ significantly between the 2 bleeding sources. Thus, the most relevant clinical outcomes suggest that the severity of both LGIB and UGIB are similar. This similarity again suggest that the clinical burden of LGIB is not significantly different than UGIB.

Our results concerning resource utilization demonstrate a similar pattern. While the point estimates for costs and LOS suggest that UGIB may be associated with higher resource utilization, these differences were not significant in either bivariate or multivariate analyses. Those subjects with UGIB did receive more total endoscopic procedures than subjects with LGIB. More interesting though was that 24% of all subjects received an endoscopy of the opposite site (LGIB with EGD and UGIB with colonoscopy). These results suggest that the site of bleeding is not clear in a significant proportion of patients who present with GIB. These additional endoscopies are associated with increased risk, costs, LOS, and discomfort to patients. Improving our ability to accurately predict the source (upper vs. lower) of bleeding would allow us to reduce the number of these excess endoscopies. Additionally, it is interesting that despite the almost universal use of endoscopies, 20% of LGIB and 14% of UGIB subjects could not have a specific etiology identified during endoscopy or subsequent workup.

There are some important limitations to this study. While the sample size is among the largest of its type involving chart abstraction, it may be underpowered to detect some differences. Additionally, our results are from a single urban academic medical center with a patient population that is predominantly African American, which may limit generalizability. This study required consent and therefore only examines a subset of patients admitted to the medical center with GIB, which could potentially introduce bias into the sample. However, it is not clear why there would be systematic differences in subjects who choose to consent vs. those who decide not to consent that would affect the results of this study in substantive ways.

Despite significant efforts at identifying all subjects with GIB admitted during this time period, there were potential methodological reasons that may have resulted in some cases being missed. Only subjects admitted to a medicine service were approached for consent. All subjects in this medical center with GIB are admitted to a medicine service. We captured all subjects who were initially admitted to a medicine service as well as those admitted initially to an ICU and then transferred to the floor at any point prior to discharge. It is possible, though, that a subject would be admitted to an ICU for GIB and die prior to being transferred to the floor. While it is the impression of the director of the ICU that this would be a very unusual event, as most of the patients would be discharged to the floor prior to death (personal communication), given the very low mortality rate seen in this study, small numbers of missed events could have a significant impact on the interpretation of in‐hospital mortality results. It is also important to note that this medical center did not have the ability to perform endoscopy prior to admission for patients with GIB at the time of the study; all patients who presented with GIB would have been admitted and identified for this study. Finally, we were unable to routinely identify the rationale for obtaining an endoscopic exam. We assumed that all endoscopic exams were done for the purpose of evaluating and/or treating the GIB for which the subject was admitted. It is possible that some subjects had additional endoscopies for other reasons, which would have led to our overestimating the rates of additional endoscopies for GIB.

This study highlights the similarities between LGIB and UGIB rather than the differences. There were few significant differences between the 2 bleeding sources in terms of incidence, clinical outcomes, and resource utilization. In fact, the study also suggests that determining the source of bleeding may not be clear, given the high rates of opposite site endoscopies. While this study did reveal several similarities between UGIB and LGIB, it also highlights the need to identify improved strategies to improve the sensitivity and specificity of identification of LGIB compared to UGIB, both for clinical purposes and for research. The value of such improved clinical algorithms have the potential to improve both the cost and outcomes of care, while better algorithms for separating UGIB and LGIB using administrative data might help produce more precise estimates of costs and clinical outcomes, and aid in the development of risk stratification models.

Gastrointestinal bleeding (GIB) is a frequent reason for acute hospitalization, with estimated rates of hospitalization at 375 per 100,000 per year in the United States.1 GIB is not a specific disease but rather a diverse set of conditions that lead to the clinical manifestations associated with bleeding into the gastrointestinal tract. One of the most commonly used organizing frameworks in gastrointestinal bleeding is the differentiation between upper gastrointestinal bleeding (UGIB) and lower gastrointestinal bleeding (LGIB). There are important differences in the etiologies between the 2 sources. For example, acid‐related disease is a common etiology in UGIB but does not occur in LGIB. While some aspects of the acute management are shared between UGIB and LGIB, important differences exist in the management, including initial endoscopy and medication choice. There have been few direct comparisons of rates, resource use, and clinical outcomes between UGIB and LGIB.

Historically, rates of UGIB have been reported to exceed those of LGIB by 2‐fold to 8‐fold.25 Protocols, clinical practice guidelines, and policy decisions reflect this emphasis on UGIB.68 Among 9 guidelines hosted by National Guideline Clearinghouse addressing GIB, 6 were focused on UGIB, 2 on both UGIB and LGIB, and only 1 on LGIB.9 There are several reasons to believe that these relative incidence rates may not be accurate. First, recent advances in therapy and prevention of UGIB, such as the treatment of Helicobacter pylori infection; proton pump inhibitors (PPIs); and selective cyclooxygenase‐2 (COX‐2) inhibitors, may have affected the epidemiology of gastrointestinal bleeding.1016 Among these therapies, only COX‐2 inhibitors may also reduce the incidence of LGIB.14, 1618 Therefore, these advances may result in a disproportionate drop in UGIB relative to LGIB. In addition, known risk factors for both LGIB and UGIB, including advancing age and renal failure, are increasing in the general population.5, 19, 20 Finally, given the recent increased recommendations for aspirin therapy and systemic anticoagulation, exposure to aspirin and warfarin have increased, both risk factors for LGIB and UGIB.2124 Indeed, recent studies in the epidemiology of UGIB do suggest a changing pattern of etiologies of UGIB reflecting these advances.25 One study examining rates of both UGIB and LGIB demonstrate a decrease in hospitalizations overall for GIB driven by a reduction in UGIB while at the same time reporting an increase in the incidence of hospitalization for LGIB.1

In addition to a changing epidemiology, a second reason for a potential underestimation of LGIB incidence is one of methodology. There are well‐recognized limitations with using purely administrative data due to difficulties in accurately identifying patients with LGIB.26

Studies using large administrative databases may not accurately identify LGIB because of the poor sensitivity and specificity of International Classification of Diseases, Ninth revision, Clinical Modification (ICD‐9) codes for LGIB.5 While there are standard methods of identifying patients with UGIB using ICD‐9 codes,19 there is not an accepted standard for LGIB. Thus, estimates using only ICD‐9 codes may overidentify or underidentify patients with LGIB. Prior studies that have most accurately identified patients with LGIB used a 2‐step method to address this issue. The initial ICD‐9 identification included a high sensitivity/low specificity approach. These identified patient charts undergo chart review to confirm the presence of an LGIB.5 This method is labor intensive and cannot be done using administrative databases. No direct comparison of UGIB to LGIB among hospitalized patients using this 2‐step method has been done recently.

The current emphasis on UGIB as seen in the published guidelines could also be supported if patients with UGIB had greater resource utilization or worse clinical outcomes. Limited direct comparisons for these outcomes are available. However, 1 administrative database study reported similar mortality rates for UGIB (2.7%) and LGIB (2.9%) in 2006.1 No direct comparisons of other clinical outcomes or resource use outcomes are available. Therefore, the emphasis on UGIB in publications and guidelines is best supported by the incidence rates that are, as has already been discussed, problematic.

We conducted a retrospective cohort study to examine the incidences of UGIB and LGIB among patients admitted to an academic medical center over 2 years using methods designed to optimally identify patients with either UGIB or LGIB. Our study also examined differences in clinical outcomes and resource utilization between subjects with UGIB and LGIB to examine the relative severity of these 2 clinical entities. These results may be useful in determining the need to reconsider clinical approaches as well as protocols and guidelines among patients with gastrointestinal bleeding.

Patients and Methods

Results

During the 2 years of observation, a total of 7741 subjects were admitted to the internal medicine service and enrolled in the hospitalist study. Of these, 1014 had a primary or secondary ICD‐9 code that may be consistent with UGIB or LGIB and underwent chart review to determine if they had an acute GIB. Out of 1014 subjects, 647 were determined not to have an acute GI hemorrhage and were excluded from the remaining analyses; 367 of the 1104 subjects identified by ICD‐9 codes were found to have a clinical presentation consistent with GIB and were included in this study. A total of 180 of these 367 had UGIB and 187 had LGIB. The mean age was 62.4 years, 56.7% were female, 82.6% were African American, 12.7% were Caucasian, and the mean Charlson index was 1.5. (Table 1) Among baseline characteristics, both gender and age were statistically associated with a difference in rates of upper vs. lower source bleeding, with LGIB patients more likely to be female (P = 0.01) and older (P < 0.001). Etiologies of UGIB include erosive disease, peptic ulcer disease, variceal bleeding, arteriovenous malformation, and malignancy. Etiologies of LGIB include: diverticulosis, colitis, arteriovenous malformation, cancer, ischemic colitis, polyp, hemorrhoidal bleed, ulcer, inflammatory bowel disease, other, and not determined (Table 2).

Baseline use of medications known to be associated with either increased or decreased risk of GIB was common. Approximately one‐third of subjects with both LGIB and UGIB used aspirin and 10% used warfarin. LGIB subjects were less likely to use an nsNSAID (P < 0.001), but more likely to use a proton pump inhibitor (PPI) (P = 0.06) (Table 3).

Key initial clinical presentation findings included vital sign abnormalities and admission hemoglobin levels. While hypotension was not common (4.7%), resting tachycardia (37%) and orthostasis (16%) were seen frequently. Subjects with LGIB were significantly less likely than those with UGIB to present with orthostasis (8.8% vs. 21.0%, respectively; P = 0.006) and resting tachycardia (32.3% vs. 42.5%, respectively; P = 0.04). Subjects with LGIB had a higher admission hemoglobin than those with UGIB (10.7 vs. 9.7, respectively; P < 0.001) (Table 4).

We also examined several clinical outcomes. When comparing LGIB to UGIB patients for these clinical outcomes using bivariate and multivariate statistics, there was no difference for in‐hospital mortality (1.1% vs. 1.1%), transfer to ICU (16.0% vs. 13.9%), 30‐day readmission (5.9% vs.7.8%), number of red blood cell (RBC) transfusions (2.7 vs. 2.4), or need for GI surgery (1.1% vs. 0.0%). The mean drop in hemoglobin was greater among subjects with LGIB compared to UGIB (1.9 g/dL vs. 1.5 g/dL, respectively) by both bivariate (P = 0.01) and multivariate (P = 0.003) analyses (Table 5).

Mean costs were $11,892 for LGIB and $14,301 for UGIB and median costs were $7,890 for LGIB and $9,548 for UGIB, but were not statistically different. LOS was also similar between subjects with LGIB (5.1 days) and UGIB (5.7 days). In bivariate and multivariate analyses, UGIB subjects had a similar mean number of endoscopic procedures (1.3) compared to LGIB subjects (1.2). Thirteen percent of subjects with UGIB required a second EGD while only 8% of subjects with LGIB required 2 colonoscopies. In addition, 29% of subjects with LGIB received an EGD while only 16% of subjects with an UGIB received a colonoscopy (P = 0.001) (Table 6).

Conclusions

This study represents one of the largest direct comparisons of LGIB to UGIB not based on administrative databases. The most striking finding was the nearly equal rates of LGIB and UGIB. There are 2 likely explanations for this surprising result. First, there may be methodological reasons that we identified a greater proportion of true LGIBs; our study used a highly sensitive search strategy of ICD‐9 coding with confirmatory chart abstraction to ensure that as many LGIB and UGIB cases would be identified as possible while also excluding cases not meeting accepted criteria for GIB. The second possibility is that there is an actual change in epidemiology of GIB. Known risk factors for LGIB are increasing such as advancing age, increased use of chronic aspirin therapy, and renal disease. At the same time, significant advances in the treatment and prevention of UGIB have been made. Recent studies have demonstrated similar trends in admissions for upper and lower GI complications, suggesting that there may be a changing epidemiology due primarily to reductions in upper GI complications.1, 16

Either explanation would have implications for the care of patients with GIB. Clinical decision‐making based on prior literature would support that in ambiguous clinical situations and initial evaluation for an UGIB is appropriate. Most risk stratification literature and clinical guidelines focus on UGIB. If rates of LGIB and UGIB are similar, then existing clinical decision protocols may need to be reevaluated to incorporate the higher likelihood of LGIB. This reevaluation would be less important if the clinical outcomes or resource utilization of UGIB was significantly greater than that for LGIB, but we did not find this was the case. Similarly, if the ability to distinguish between LGIB and UGIB were robust on clinical signs and symptoms, then a reevaluation would be less important. However, we found fairly similar numbers of patients initially receiving evaluation for UGIB then being evaluated for LGIB as we found patients initially receiving evaluation for LGIB then being evaluated for UGIB. This suggests the potential benefit of clinical decision protocols that could better distinguish between UGIB and LGIB and account for the potentially higher incidence of LGIB than previously thought.

In addition to affecting the attention paid to LGIB for acute management, a changed understanding of incidence could also affect the attention paid to prevention of LGIB. Of the recent nonendoscopic advances in the treatment and prevention of GIB, only the use of COX‐2s (when used in place of traditional nsNSAIDs) reduces the risk of both LGIB and UGIB;14, 1618 H .pylori treatment and PPIs only prevent UGIB. Therefore, if the clinical and financial burdens of LGIB are similar to those seen in UGIB, more attention may need to be focused on preventing LGIB.

Baseline medication use was notable primarily for the similarities between UGIB and LGIB. Agents known to affect the rates of GIB were common in both groups. Over one‐third of the population was using aspirin and 10% were taking warfarin. Over 20% of subjects were taking an nsNSAID or a COX‐2 inhibitor. Almost one‐quarter of subjects were taking a PPI, agents known to decrease rates of UGIB and potentially increase LGIB through the risk of C. difficile colitis. Notably, the only statistically significant difference in baseline medication use between subjects with UGIB and LGIB was the more than 3‐fold higher use of nsNSAIDs in patients with UGIB as compared to LGIB. While current guidelines are not clear and consistent about which populations of at‐risk patients should receive GI prophylaxis,2830 these results suggest that patients admitted with GIB are very likely to be taking medications which impact the risk of GIB.

In terms of disease severity, the clinical presentation at admission suggests a greater degree of hemodynamic instability among subjects with UGIB. Rates of orthostatic hypotension and resting tachycardia are higher in UGIB subjects, as well as having a lower mean hemoglobin levels at presentation. However, despite the more severe clinical presentation, clinical outcomes did not differ significantly between the 2 bleeding sources. Thus, the most relevant clinical outcomes suggest that the severity of both LGIB and UGIB are similar. This similarity again suggest that the clinical burden of LGIB is not significantly different than UGIB.

Our results concerning resource utilization demonstrate a similar pattern. While the point estimates for costs and LOS suggest that UGIB may be associated with higher resource utilization, these differences were not significant in either bivariate or multivariate analyses. Those subjects with UGIB did receive more total endoscopic procedures than subjects with LGIB. More interesting though was that 24% of all subjects received an endoscopy of the opposite site (LGIB with EGD and UGIB with colonoscopy). These results suggest that the site of bleeding is not clear in a significant proportion of patients who present with GIB. These additional endoscopies are associated with increased risk, costs, LOS, and discomfort to patients. Improving our ability to accurately predict the source (upper vs. lower) of bleeding would allow us to reduce the number of these excess endoscopies. Additionally, it is interesting that despite the almost universal use of endoscopies, 20% of LGIB and 14% of UGIB subjects could not have a specific etiology identified during endoscopy or subsequent workup.

There are some important limitations to this study. While the sample size is among the largest of its type involving chart abstraction, it may be underpowered to detect some differences. Additionally, our results are from a single urban academic medical center with a patient population that is predominantly African American, which may limit generalizability. This study required consent and therefore only examines a subset of patients admitted to the medical center with GIB, which could potentially introduce bias into the sample. However, it is not clear why there would be systematic differences in subjects who choose to consent vs. those who decide not to consent that would affect the results of this study in substantive ways.

Despite significant efforts at identifying all subjects with GIB admitted during this time period, there were potential methodological reasons that may have resulted in some cases being missed. Only subjects admitted to a medicine service were approached for consent. All subjects in this medical center with GIB are admitted to a medicine service. We captured all subjects who were initially admitted to a medicine service as well as those admitted initially to an ICU and then transferred to the floor at any point prior to discharge. It is possible, though, that a subject would be admitted to an ICU for GIB and die prior to being transferred to the floor. While it is the impression of the director of the ICU that this would be a very unusual event, as most of the patients would be discharged to the floor prior to death (personal communication), given the very low mortality rate seen in this study, small numbers of missed events could have a significant impact on the interpretation of in‐hospital mortality results. It is also important to note that this medical center did not have the ability to perform endoscopy prior to admission for patients with GIB at the time of the study; all patients who presented with GIB would have been admitted and identified for this study. Finally, we were unable to routinely identify the rationale for obtaining an endoscopic exam. We assumed that all endoscopic exams were done for the purpose of evaluating and/or treating the GIB for which the subject was admitted. It is possible that some subjects had additional endoscopies for other reasons, which would have led to our overestimating the rates of additional endoscopies for GIB.

This study highlights the similarities between LGIB and UGIB rather than the differences. There were few significant differences between the 2 bleeding sources in terms of incidence, clinical outcomes, and resource utilization. In fact, the study also suggests that determining the source of bleeding may not be clear, given the high rates of opposite site endoscopies. While this study did reveal several similarities between UGIB and LGIB, it also highlights the need to identify improved strategies to improve the sensitivity and specificity of identification of LGIB compared to UGIB, both for clinical purposes and for research. The value of such improved clinical algorithms have the potential to improve both the cost and outcomes of care, while better algorithms for separating UGIB and LGIB using administrative data might help produce more precise estimates of costs and clinical outcomes, and aid in the development of risk stratification models.

Acute upper gastrointestinal hemorrhage (UGIH) is one of the most common hospital admissions for acute care. Estimates indicate that 300,000 patients (100‐150 cases per 100,000 adults) are admitted annually with an associated economic impact of $2.5 billion.15 The current standard management of UGIH requires hospital admission and esophagogastroduodenoscopy (EGD) by a gastroenterologist for diagnosis and/or treatment. This management strategy results in a high consumption of hospital resources and costs.

Simultaneously, hospitalists have dramatically changed the delivery of inpatient care in the United States and are recognized as a location‐driven subspecialty for the care of acute hospitalized patients, similar to emergency medicine. Currently there are 20,000 hospitalists, and more than one‐third of general medicine inpatients are cared for by hospitalists.6, 7

Previous studies have shown that hospitalist care offers better or comparable outcomes, with lower overall length of stay (LOS) and costs compared to traditional providers.810 However, most of these studies were performed in single institutions, had weak designs or little‐to‐no adjustment for severity of illness, or were limited to 7 specific diseases (pneumonia, congestive heart failure [CHF], chest pain, ischemic stroke, urinary tract infection, chronic obstructive lung disease [COPD], and acute myocardial infarction [AMI]).8

Furthermore, less is known about the effect of hospitalists on conditions that may be dependent upon specialist consultation for procedures and/or treatment plans. In this study, gastroenterologists performed diagnostic and/or therapeutic endoscopy work as consultants to the attending physicians in the management of acute inpatient UGIH.

To explore the effects of hospitalists on care of patients with acute UGIH, we examined data from the Multicenter Hospitalist (MCH) trial. The objectives of our study were to compare clinical outcomesin‐hospital mortality and complications (ie, recurrent bleeding, intensive care unit [ICU] transfer, decompensation, transfusion, reendoscopy, 30‐day readmission)and efficiency (LOS and costs) in hospitalized acute UGIH patients cared for by hospitalists and nonhospitalists in 6 academic centers in the United States during a 2‐year period.

Patients and Methods

Results

Discussion

This is the first study that has looked at the effect of hospitalists on clinical outcomes and efficiency in patients admitted for acute UGIH, a condition highly dependent upon another specialty for procedures and management. This is also one of only a few studies on UGIH that adjusted for severity of illness (Rockall score), comorbidity, performance of early endoscopypatient‐level confounders usually unaccounted for in prior research.

We show that hospitalists and nonhospitalists caring for acute UGIH patients had overall similar unadjusted outcomes; except for blood transfusion and 30‐day readmission rates. Unfortunately, due to the small number of events for readmissions, we were unable to perform adjusted analysis for readmission. Differences between hospitalists and nonhospitalists on blood transfusion rates were not substantiated on multivariable adjustments.

As for efficiency, univariable and multivariable analyses revealed that LOS was similar between provider types while costs were greater in UGIH patients attended by hospitalists.

Reductions in resource use, particularly costs, may be achieved by increasing throughput (eg, reducing LOS) or by decreasing service intensity (eg, using fewer ancillary services and specialty consultations).26 Specifically in acute UGIH, LOS is significantly affected by performance of early EGD.27, 28 In these studies, gastroenterologist‐led teams, compared to internists and surgeons, have easier access to endoscopy, thus reducing LOS and overall costs.27, 28

Similarly, prior studies have shown that the mechanism by which hospitalists lower costs is by decreasing LOS.810, 29 There are several hypotheses on how hospitalists affect LOS. Hospitalists, by being available all day, are thought to respond quickly to acute symptoms or new test results, are more efficient in navigating the complex hospital environment, or develop greater expertise as a result of added inpatient experience.8 On the downside, although the hospitalist model reduces overall LOS and costs, they also provide higher intensity of care as reflected by greater costs when broken down per hospital day.29 Thus, the cost differential we found may represent higher intensity of care by hospitalists in their management of acute UGIH, as higher intensity care without decreasing LOS can translate to higher costs.

In addition, patients with acute UGIH are unique in several respects. In contrast to diseases like heart failure, COPD, and pneumonia, in which the admitting provider has the option to request a subspecialist consultation, all patients with acute UGIH need a gastroenterologist to perform endoscopy as part of the management. These patients are usually admitted to general medicine wards, aggressively resuscitated with intravenous fluids, with a nonurgent gastroenterology consult or EGD performed on the next available schedule.

Aside from LOS being greatly affected by performance of early EGD and/or delay in consulting gastroenterology, sicker patients require longer hospitalization and drive LOS and healthcare costs up. It was therefore crucial that we accounted for severity of illness, comorbidity, and performance of early EGD in our regression models for LOS and costs. This approach allows us to acquire a more accurate estimate on the effects of hospitalist on LOS and costs in patients admitted with acute UGIH.

Our findings suggest that the academic hospitalist model of care may not have as great of an impact on hospital efficiency in certain patient groups that require nonurgent subspecialty consultations. Future studies should focus on elucidating these relationships.

Limitations

This study has several limitations. First, clinical data were abstracted at 6 sites by different abstractors so it is possible there were variations in how data were collected. To reduce variation, a standardized abstraction form with instructions was developed and the primary investigator (PI) was available for specific questions during the abstraction process. Second, only 5 out of the 6 sites used TSI accounting systems. Although similar, interhospital costs captured by TSI may vary among sites in terms of classifying direct and indirect costs, potentially resulting in misclassification bias in our cost estimates.17 We addressed these issues by including the hospital site variable in our regression models, regardless of its significance. Third, consent rates across sites vary from 70% to 85%. It is possible that patients who refused enrollment in the MCH trial are systematically different and may introduce bias in our analysis.

Furthermore, the study was designed as a natural experiment based on a rotational call cycle between hospitalist‐led and nonhospitalist‐led teams. It is possible that the order of patient assignment might not be completely naturally random as we intended. However, the study period was for 2 years and we expect the effect of order would have averaged out in time.

There are many hospitalist models of care. In terms of generalizability, the study pertains only to academic hospitalists and may not be applicable to hospitalists practicing in community hospitals. For example, the nonhospitalist comparison group is likely different in the community and academic settings. Community nonhospitalists (traditional practitioners) are usually internists covering both inpatient and outpatient responsibilities at the same time. In contrast, academic nonhospitalists are internists or subspecialists serving as ward attendings for a limited period (usually 1 month) with considerable variation in their nonattending responsibilities (eg, research, clinic, administration). Furthermore, academic nonhospitalist providers might be a self‐selected group by their willingness to serve as a ward attending, making them more hospitalist‐like. Changes and variability of inpatient attendings may also affect our findings when compared to prior work. Finally, it is also possible that having residents at academic medical centers may attenuate the effect of hospitalists more than in community‐based models.

Conclusions/Implications

Compared to nonhospitalists, academic hospitalist care of acute UGIH patients had similar overall clinical outcomes. However, our finding of similar LOS yet higher costs for patients cared for by hospitalists support 1 proposed mechanism in which hospitalists decrease healthcare costs: providing higher intensity of care per day of hospitalization. However, in academic hospitalist models, this higher intensity hypothesis should be revisited, especially in certain patient groups in which timing and involvement of subspecialists may influence discharge decisions, affecting LOS and overall costs.

Due to inherent limitations in this observational study, future studies should focus on verifying and elucidating these relationships further. Lastly, understanding which patient groups receive the greatest potential benefit from this model will help guide both organizational efforts and quality improvement strategies.