Melissa L. P. Mattison, MD

Older adults with high levels of medical complexity occupy an increasing fraction of beds in acute‐care hospitals in the United States.[1, 2] By 2007, patients age 65 years and older accounted for nearly half of adult inpatient days of care.[1] These patients are commonly cared for by hospitalists who number more than 40,000.[3] Although hospitalists are most often trained in internal medicine, they have typically received limited formal geriatrics training. Increasingly, access to experts in geriatric medicine is limited.[4] Further, hospitalists and others who practice in acute care are limited by the lack of research to address the needs of the older adult population, specifically in the diagnosis and management of conditions encountered during acute illness.

To better support hospitalists in providing acute inpatient geriatric care, the Society of Hospital Medicine (SHM) partnered with the Association of Specialty Professors to develop a research agenda to bridge this gap. Using methodology from the James Lind Alliance (JLA) and the Patient Centered Outcomes Research Institute (PCORI), the SHM joined with older adult advocacy groups, professional societies of providers, and funders to create a geriatric‐focused acute‐care research agenda, highlighting 10 key research questions.[5, 6, 7] The goal of this approach was to produce and promote high integrity, evidence‐based information that comes from research guided by patients, caregivers, and the broader healthcare community.[8] In this article, we describe the methodology and results of this agenda‐setting process, referred to as the Acute Care of Older Patients (ACOP) Priority Setting Partnership.

METHODS

RESULTS

Table 1 lists the organizations who engaged in 1 or more parts of the topic generation process. Eighteen stakeholder organizations agreed to participate in the convening process. Ten organizations did not respond to our solicitation and 1 declined to participate.

Seventeen stakeholder organizations obtained survey responses from a total of 580 individuals (range, 3150 per organization), who were asked to identify important unanswered questions in the acute care of older persons. Survey respondents were typically female (77%), white (85%), aged 45 to 65 years (65%), and identified themselves as health professionals (90%). Twenty‐six percent of respondents also identified as patients or family caregivers. Their surveys included 1299 individual questions.

Figure 1 summarizes our collation and prioritization process and reports the numbers of questions resulting at each stage. Nine hundred nineteen questions were removed during the first review conducted by steering committee members, and 31 question categories were identified. An additional 305 questions were removed in the second review, with 75 questions remaining. As the final step of the collating process, literature review identified 39 relevant questions not already suggested or moved forward through our consultation and collation process. These questions were added to the list of unanswered questions.

Figure 1

In the first round of prioritization, this list of 114 questions was emailed to each stakeholder organization (Table 1). After the stakeholder voting process was completed, 29 unanswered questions remained (see Supporting Information, Appendix C, in the online version of this article). These questions were refined and prioritized in the in‐person meeting to create the final list of 10 questions. The stakeholders present in the meeting represented 13 organizations (Table 1). Using the NGT with several rounds of small group breakouts and large group deliberation, 9 of the top 10 questions were selected from the list of 29. One additional highly relevant question that had been removed earlier in the collation process regarding workforce was added back by the stakeholder group.

This prioritized research agenda appears in Table 2 and below, organized alphabetically by topic.

1. Advanced care planning: What approaches for determining and communicating goals of care across and within healthcare settings are most effective in promoting goal‐concordant care for hospitalized older patients?
2. Care transitions: What is the comparative effectiveness of transitional care models on patient‐centered outcomes for hospitalized older adults?
3. Delirium: What practices are most effective for consistent recognition, prevention, and treatment of delirium subtypes among hospitalized older adults?
4. Dementia: Does universal assessment of hospitalized older adults for cognitive impairment (eg, at presentation and/or discharge) lead to more appropriate application of geriatric care principles and improve patient‐centered outcomes?
5. Depression: Does identifying depressive symptoms during a hospital stay and initiating a therapeutic plan prior to discharge improve patient‐centered and/or disease‐specific outcomes?
6. Medications: What systems interventions improve medication management for older adults (ie, appropriateness of medication choices and dosing, compliance, cost) in the hospital and postacute care?
7. Models of care: For which populations of hospitalized older adults does systematic implementation of geriatric care principles/processes improve patient‐centered outcomes?
8. Physical function: What is the comparative effectiveness of interventions that promote in‐hospital mobility, improve and preserve physical function, and reduce falls among older hospitalized patients?
9. Surgery: What perioperative strategies can be used to optimize care processes and improve outcomes in older surgical patients?
10. Training: What is the most effective approach to training hospital‐based providers in geriatric and palliative care competencies?

Table 2 also contains a capsule summary of the scope of the problem addressed by each research priority, a capsule summary of related work in the content area (what is known) not intended as a systematic review, and proposed dimensions or subquestions suggested by the stakeholders at the final prioritization meeting

DISCUSSION

Older hospitalized patients account for an increasing number and proportion of hospitalized patients,[1, 2] and hospitalists increasingly are responsible for inpatient care for this population.[3] The knowledge required for hospitalists to deliver optimal care and improve outcomes has not kept pace with the rapid growth of either hospitalists or hospitalized elders. Through a rigorous prioritization process, we identified 10 areas that deserve the highest priority in directing future research efforts to improve care for the older hospitalized patient. Assessment, prevention, and treatment of geriatric syndromes in the hospital account for almost half of the priority areas. Additional research is needed to improve advanced care planning, develop new care models, and develop training models for future hospitalists competent in geriatric and palliative care competencies.

A decade ago, the American Geriatric Society and the John A. Hartford Foundation embarked upon a research agenda aimed at improving the care of hospitalized elders cared for by specialists (ie, New Frontiers in Geriatrics Research: An Agenda for Surgical and Related Medical Specialties).[9] This effort differed in many important ways from the current priortization process. First, the New Frontiers agenda focused upon specific diseases, whereas the ACOP agenda addresses geriatric syndromes that cut across multiple diseases. Second, the New Frontiers agenda was made by researchers and based upon published literature, whereas the ACOP agenda involved the input of multiple stakeholders. Finally, the New Frontiers prioritized a research agenda across a number of surgical specialties, emergency medicine, and geriatric rehabilitation. Hospital medicine, however, was still early in its development and was not considered a unique specialty. Since that time, hospital medicine has matured into a unique specialty, with increased numbers of hospitalists,[3] increased research in hospital medicine,[17] and a separate recertification pathway for internal medicine licensure.[18] To date, there has not been a similar effort performed to direct geriatric research efforts for hospital medicine.

For researchers working in the field of hospital medicine, this list of topics has several implications. First, as hospitalists are commonly generalists, hospitalist researchers may be particularly well‐suited to study syndromes that cut across specialties. However, this does raise concerns about funding sources, as most National Institutes of Health institutes are disease‐focused. Funders that are not disease‐focused such as PCORI, National Institute on Aging, National Institute of Nursing Research, and Agency for Healthcare Research and Quality, and private foundations (Hartford, Robert Wood Johnson, and Commonwealth) may be more fruitful sources of funding for this work, but funding may be challenging. Nonetheless, the increased focus on patient‐centered work may increase funders' interest in such work. Second, the topics on this list would suggest that interventions will not be pharmacologic, but will focus on nonpharmacologic, behavioral, and social interventions. Similarly, outcomes of interest must expand beyond utilization metrics such as length of stay and mortality, to include functional status and symptom management, and goal‐concordant care. Therefore, research in geriatric acute care will necessarily be multidisciplinary.

Although these 10 high‐priority areas have been selected, this prioritized list is inherently limited by our methodology. First, our survey question was not focused on a disease state, and this wording may have resulted in the list favoring geriatric syndromes rather than common disease processes. Additionally, the resulting questions encompass large research areas and not specific questions about discrete interventions. Our results may also have been skewed by the types of engaged respondents who participated in the consultation, collating, and prioritization phases. In particular, we had a large response from geriatric medicine nurses, whereas some stakeholder groups provided no survey responses. Thus, these respondents were not representative of all possible stakeholders, nor were the survey respondents necessarily representative of each of their organizations. Nonetheless, the participants self‐identified as representative of diverse viewpoints that included patients, caregivers, and advocacy groups, with the majority of stakeholder organizations remaining engaged through the completion of the process. Thus, the general nature of this agenda helps us focus upon larger areas of importance, leaving researchers the flexibility to choose to narrow the focus on a specific research question that may include potential interventions and unique outcomes. Finally, our methodology may have inadvertently limited the number of patient and family caregiver voices in the process given our approach to large advocacy groups, our desire to be inclusive of healthcare professional organizations, and our survey methodology. Other methodologies may have reached more patients and caregivers, yet many healthcare professionals have served as family caregivers to frail elders requiring hospitalization and may have been in an ideal position to answer the survey.

In conclusion, several forces are shaping the future of acute inpatient care. These include the changing demographics of the hospitalized patient population, a rapid increase in the proportion of multimorbid hospitalized older adults, an inpatient workforce (hospitalists, generalists, and subspecialists) with potentially limited geriatrics training, and gaps in evidence‐based guidance to inform diagnostic and therapeutic decision making for acutely ill older patients. Training programs in hospital medicine should be aware of and could benefit from the resulting list of unanswered questions. Our findings also have implications for training to enrich education in geriatrics. Moreover, there is growing recognition that patients and other stakeholders deserve a greater voice in determining the direction of research. In addition to efforts to improve patient‐centeredness of research, these areas have been uniquely identified by stakeholders as important, and therefore are in line with newer priorities of PCORI. This project followed a road map resulting in a patient‐centered research agenda at the intersection of hospital medicine and geriatric medicine.[7] In creating this agenda, we relied heavily on the framework proposed by PCORI. We propose to pursue a dissemination and evaluation strategy for this research agenda as well as additional prioritization steps. We believe the adoption of this methodology will create a knowledge base that is rigorously derived and most relevant to the care of hospitalized older adults and their families. Its application will ultimately result in improved outcomes for hospitalized older adults.

Older adults with high levels of medical complexity occupy an increasing fraction of beds in acute‐care hospitals in the United States.[1, 2] By 2007, patients age 65 years and older accounted for nearly half of adult inpatient days of care.[1] These patients are commonly cared for by hospitalists who number more than 40,000.[3] Although hospitalists are most often trained in internal medicine, they have typically received limited formal geriatrics training. Increasingly, access to experts in geriatric medicine is limited.[4] Further, hospitalists and others who practice in acute care are limited by the lack of research to address the needs of the older adult population, specifically in the diagnosis and management of conditions encountered during acute illness.

To better support hospitalists in providing acute inpatient geriatric care, the Society of Hospital Medicine (SHM) partnered with the Association of Specialty Professors to develop a research agenda to bridge this gap. Using methodology from the James Lind Alliance (JLA) and the Patient Centered Outcomes Research Institute (PCORI), the SHM joined with older adult advocacy groups, professional societies of providers, and funders to create a geriatric‐focused acute‐care research agenda, highlighting 10 key research questions.[5, 6, 7] The goal of this approach was to produce and promote high integrity, evidence‐based information that comes from research guided by patients, caregivers, and the broader healthcare community.[8] In this article, we describe the methodology and results of this agenda‐setting process, referred to as the Acute Care of Older Patients (ACOP) Priority Setting Partnership.

METHODS

RESULTS

Table 1 lists the organizations who engaged in 1 or more parts of the topic generation process. Eighteen stakeholder organizations agreed to participate in the convening process. Ten organizations did not respond to our solicitation and 1 declined to participate.

Seventeen stakeholder organizations obtained survey responses from a total of 580 individuals (range, 3150 per organization), who were asked to identify important unanswered questions in the acute care of older persons. Survey respondents were typically female (77%), white (85%), aged 45 to 65 years (65%), and identified themselves as health professionals (90%). Twenty‐six percent of respondents also identified as patients or family caregivers. Their surveys included 1299 individual questions.

Figure 1 summarizes our collation and prioritization process and reports the numbers of questions resulting at each stage. Nine hundred nineteen questions were removed during the first review conducted by steering committee members, and 31 question categories were identified. An additional 305 questions were removed in the second review, with 75 questions remaining. As the final step of the collating process, literature review identified 39 relevant questions not already suggested or moved forward through our consultation and collation process. These questions were added to the list of unanswered questions.

Figure 1

In the first round of prioritization, this list of 114 questions was emailed to each stakeholder organization (Table 1). After the stakeholder voting process was completed, 29 unanswered questions remained (see Supporting Information, Appendix C, in the online version of this article). These questions were refined and prioritized in the in‐person meeting to create the final list of 10 questions. The stakeholders present in the meeting represented 13 organizations (Table 1). Using the NGT with several rounds of small group breakouts and large group deliberation, 9 of the top 10 questions were selected from the list of 29. One additional highly relevant question that had been removed earlier in the collation process regarding workforce was added back by the stakeholder group.

This prioritized research agenda appears in Table 2 and below, organized alphabetically by topic.

1. Advanced care planning: What approaches for determining and communicating goals of care across and within healthcare settings are most effective in promoting goal‐concordant care for hospitalized older patients?
2. Care transitions: What is the comparative effectiveness of transitional care models on patient‐centered outcomes for hospitalized older adults?
3. Delirium: What practices are most effective for consistent recognition, prevention, and treatment of delirium subtypes among hospitalized older adults?
4. Dementia: Does universal assessment of hospitalized older adults for cognitive impairment (eg, at presentation and/or discharge) lead to more appropriate application of geriatric care principles and improve patient‐centered outcomes?
5. Depression: Does identifying depressive symptoms during a hospital stay and initiating a therapeutic plan prior to discharge improve patient‐centered and/or disease‐specific outcomes?
6. Medications: What systems interventions improve medication management for older adults (ie, appropriateness of medication choices and dosing, compliance, cost) in the hospital and postacute care?
7. Models of care: For which populations of hospitalized older adults does systematic implementation of geriatric care principles/processes improve patient‐centered outcomes?
8. Physical function: What is the comparative effectiveness of interventions that promote in‐hospital mobility, improve and preserve physical function, and reduce falls among older hospitalized patients?
9. Surgery: What perioperative strategies can be used to optimize care processes and improve outcomes in older surgical patients?
10. Training: What is the most effective approach to training hospital‐based providers in geriatric and palliative care competencies?

Table 2 also contains a capsule summary of the scope of the problem addressed by each research priority, a capsule summary of related work in the content area (what is known) not intended as a systematic review, and proposed dimensions or subquestions suggested by the stakeholders at the final prioritization meeting

DISCUSSION

Older hospitalized patients account for an increasing number and proportion of hospitalized patients,[1, 2] and hospitalists increasingly are responsible for inpatient care for this population.[3] The knowledge required for hospitalists to deliver optimal care and improve outcomes has not kept pace with the rapid growth of either hospitalists or hospitalized elders. Through a rigorous prioritization process, we identified 10 areas that deserve the highest priority in directing future research efforts to improve care for the older hospitalized patient. Assessment, prevention, and treatment of geriatric syndromes in the hospital account for almost half of the priority areas. Additional research is needed to improve advanced care planning, develop new care models, and develop training models for future hospitalists competent in geriatric and palliative care competencies.

A decade ago, the American Geriatric Society and the John A. Hartford Foundation embarked upon a research agenda aimed at improving the care of hospitalized elders cared for by specialists (ie, New Frontiers in Geriatrics Research: An Agenda for Surgical and Related Medical Specialties).[9] This effort differed in many important ways from the current priortization process. First, the New Frontiers agenda focused upon specific diseases, whereas the ACOP agenda addresses geriatric syndromes that cut across multiple diseases. Second, the New Frontiers agenda was made by researchers and based upon published literature, whereas the ACOP agenda involved the input of multiple stakeholders. Finally, the New Frontiers prioritized a research agenda across a number of surgical specialties, emergency medicine, and geriatric rehabilitation. Hospital medicine, however, was still early in its development and was not considered a unique specialty. Since that time, hospital medicine has matured into a unique specialty, with increased numbers of hospitalists,[3] increased research in hospital medicine,[17] and a separate recertification pathway for internal medicine licensure.[18] To date, there has not been a similar effort performed to direct geriatric research efforts for hospital medicine.

For researchers working in the field of hospital medicine, this list of topics has several implications. First, as hospitalists are commonly generalists, hospitalist researchers may be particularly well‐suited to study syndromes that cut across specialties. However, this does raise concerns about funding sources, as most National Institutes of Health institutes are disease‐focused. Funders that are not disease‐focused such as PCORI, National Institute on Aging, National Institute of Nursing Research, and Agency for Healthcare Research and Quality, and private foundations (Hartford, Robert Wood Johnson, and Commonwealth) may be more fruitful sources of funding for this work, but funding may be challenging. Nonetheless, the increased focus on patient‐centered work may increase funders' interest in such work. Second, the topics on this list would suggest that interventions will not be pharmacologic, but will focus on nonpharmacologic, behavioral, and social interventions. Similarly, outcomes of interest must expand beyond utilization metrics such as length of stay and mortality, to include functional status and symptom management, and goal‐concordant care. Therefore, research in geriatric acute care will necessarily be multidisciplinary.

Although these 10 high‐priority areas have been selected, this prioritized list is inherently limited by our methodology. First, our survey question was not focused on a disease state, and this wording may have resulted in the list favoring geriatric syndromes rather than common disease processes. Additionally, the resulting questions encompass large research areas and not specific questions about discrete interventions. Our results may also have been skewed by the types of engaged respondents who participated in the consultation, collating, and prioritization phases. In particular, we had a large response from geriatric medicine nurses, whereas some stakeholder groups provided no survey responses. Thus, these respondents were not representative of all possible stakeholders, nor were the survey respondents necessarily representative of each of their organizations. Nonetheless, the participants self‐identified as representative of diverse viewpoints that included patients, caregivers, and advocacy groups, with the majority of stakeholder organizations remaining engaged through the completion of the process. Thus, the general nature of this agenda helps us focus upon larger areas of importance, leaving researchers the flexibility to choose to narrow the focus on a specific research question that may include potential interventions and unique outcomes. Finally, our methodology may have inadvertently limited the number of patient and family caregiver voices in the process given our approach to large advocacy groups, our desire to be inclusive of healthcare professional organizations, and our survey methodology. Other methodologies may have reached more patients and caregivers, yet many healthcare professionals have served as family caregivers to frail elders requiring hospitalization and may have been in an ideal position to answer the survey.

In conclusion, several forces are shaping the future of acute inpatient care. These include the changing demographics of the hospitalized patient population, a rapid increase in the proportion of multimorbid hospitalized older adults, an inpatient workforce (hospitalists, generalists, and subspecialists) with potentially limited geriatrics training, and gaps in evidence‐based guidance to inform diagnostic and therapeutic decision making for acutely ill older patients. Training programs in hospital medicine should be aware of and could benefit from the resulting list of unanswered questions. Our findings also have implications for training to enrich education in geriatrics. Moreover, there is growing recognition that patients and other stakeholders deserve a greater voice in determining the direction of research. In addition to efforts to improve patient‐centeredness of research, these areas have been uniquely identified by stakeholders as important, and therefore are in line with newer priorities of PCORI. This project followed a road map resulting in a patient‐centered research agenda at the intersection of hospital medicine and geriatric medicine.[7] In creating this agenda, we relied heavily on the framework proposed by PCORI. We propose to pursue a dissemination and evaluation strategy for this research agenda as well as additional prioritization steps. We believe the adoption of this methodology will create a knowledge base that is rigorously derived and most relevant to the care of hospitalized older adults and their families. Its application will ultimately result in improved outcomes for hospitalized older adults.

Older adults with high levels of medical complexity occupy an increasing fraction of beds in acute‐care hospitals in the United States.[1, 2] By 2007, patients age 65 years and older accounted for nearly half of adult inpatient days of care.[1] These patients are commonly cared for by hospitalists who number more than 40,000.[3] Although hospitalists are most often trained in internal medicine, they have typically received limited formal geriatrics training. Increasingly, access to experts in geriatric medicine is limited.[4] Further, hospitalists and others who practice in acute care are limited by the lack of research to address the needs of the older adult population, specifically in the diagnosis and management of conditions encountered during acute illness.

To better support hospitalists in providing acute inpatient geriatric care, the Society of Hospital Medicine (SHM) partnered with the Association of Specialty Professors to develop a research agenda to bridge this gap. Using methodology from the James Lind Alliance (JLA) and the Patient Centered Outcomes Research Institute (PCORI), the SHM joined with older adult advocacy groups, professional societies of providers, and funders to create a geriatric‐focused acute‐care research agenda, highlighting 10 key research questions.[5, 6, 7] The goal of this approach was to produce and promote high integrity, evidence‐based information that comes from research guided by patients, caregivers, and the broader healthcare community.[8] In this article, we describe the methodology and results of this agenda‐setting process, referred to as the Acute Care of Older Patients (ACOP) Priority Setting Partnership.

METHODS

RESULTS

Table 1 lists the organizations who engaged in 1 or more parts of the topic generation process. Eighteen stakeholder organizations agreed to participate in the convening process. Ten organizations did not respond to our solicitation and 1 declined to participate.

Seventeen stakeholder organizations obtained survey responses from a total of 580 individuals (range, 3150 per organization), who were asked to identify important unanswered questions in the acute care of older persons. Survey respondents were typically female (77%), white (85%), aged 45 to 65 years (65%), and identified themselves as health professionals (90%). Twenty‐six percent of respondents also identified as patients or family caregivers. Their surveys included 1299 individual questions.

Figure 1 summarizes our collation and prioritization process and reports the numbers of questions resulting at each stage. Nine hundred nineteen questions were removed during the first review conducted by steering committee members, and 31 question categories were identified. An additional 305 questions were removed in the second review, with 75 questions remaining. As the final step of the collating process, literature review identified 39 relevant questions not already suggested or moved forward through our consultation and collation process. These questions were added to the list of unanswered questions.

Figure 1

In the first round of prioritization, this list of 114 questions was emailed to each stakeholder organization (Table 1). After the stakeholder voting process was completed, 29 unanswered questions remained (see Supporting Information, Appendix C, in the online version of this article). These questions were refined and prioritized in the in‐person meeting to create the final list of 10 questions. The stakeholders present in the meeting represented 13 organizations (Table 1). Using the NGT with several rounds of small group breakouts and large group deliberation, 9 of the top 10 questions were selected from the list of 29. One additional highly relevant question that had been removed earlier in the collation process regarding workforce was added back by the stakeholder group.

This prioritized research agenda appears in Table 2 and below, organized alphabetically by topic.

1. Advanced care planning: What approaches for determining and communicating goals of care across and within healthcare settings are most effective in promoting goal‐concordant care for hospitalized older patients?
2. Care transitions: What is the comparative effectiveness of transitional care models on patient‐centered outcomes for hospitalized older adults?
3. Delirium: What practices are most effective for consistent recognition, prevention, and treatment of delirium subtypes among hospitalized older adults?
4. Dementia: Does universal assessment of hospitalized older adults for cognitive impairment (eg, at presentation and/or discharge) lead to more appropriate application of geriatric care principles and improve patient‐centered outcomes?
5. Depression: Does identifying depressive symptoms during a hospital stay and initiating a therapeutic plan prior to discharge improve patient‐centered and/or disease‐specific outcomes?
6. Medications: What systems interventions improve medication management for older adults (ie, appropriateness of medication choices and dosing, compliance, cost) in the hospital and postacute care?
7. Models of care: For which populations of hospitalized older adults does systematic implementation of geriatric care principles/processes improve patient‐centered outcomes?
8. Physical function: What is the comparative effectiveness of interventions that promote in‐hospital mobility, improve and preserve physical function, and reduce falls among older hospitalized patients?
9. Surgery: What perioperative strategies can be used to optimize care processes and improve outcomes in older surgical patients?
10. Training: What is the most effective approach to training hospital‐based providers in geriatric and palliative care competencies?

Table 2 also contains a capsule summary of the scope of the problem addressed by each research priority, a capsule summary of related work in the content area (what is known) not intended as a systematic review, and proposed dimensions or subquestions suggested by the stakeholders at the final prioritization meeting

DISCUSSION

Older hospitalized patients account for an increasing number and proportion of hospitalized patients,[1, 2] and hospitalists increasingly are responsible for inpatient care for this population.[3] The knowledge required for hospitalists to deliver optimal care and improve outcomes has not kept pace with the rapid growth of either hospitalists or hospitalized elders. Through a rigorous prioritization process, we identified 10 areas that deserve the highest priority in directing future research efforts to improve care for the older hospitalized patient. Assessment, prevention, and treatment of geriatric syndromes in the hospital account for almost half of the priority areas. Additional research is needed to improve advanced care planning, develop new care models, and develop training models for future hospitalists competent in geriatric and palliative care competencies.

A decade ago, the American Geriatric Society and the John A. Hartford Foundation embarked upon a research agenda aimed at improving the care of hospitalized elders cared for by specialists (ie, New Frontiers in Geriatrics Research: An Agenda for Surgical and Related Medical Specialties).[9] This effort differed in many important ways from the current priortization process. First, the New Frontiers agenda focused upon specific diseases, whereas the ACOP agenda addresses geriatric syndromes that cut across multiple diseases. Second, the New Frontiers agenda was made by researchers and based upon published literature, whereas the ACOP agenda involved the input of multiple stakeholders. Finally, the New Frontiers prioritized a research agenda across a number of surgical specialties, emergency medicine, and geriatric rehabilitation. Hospital medicine, however, was still early in its development and was not considered a unique specialty. Since that time, hospital medicine has matured into a unique specialty, with increased numbers of hospitalists,[3] increased research in hospital medicine,[17] and a separate recertification pathway for internal medicine licensure.[18] To date, there has not been a similar effort performed to direct geriatric research efforts for hospital medicine.

For researchers working in the field of hospital medicine, this list of topics has several implications. First, as hospitalists are commonly generalists, hospitalist researchers may be particularly well‐suited to study syndromes that cut across specialties. However, this does raise concerns about funding sources, as most National Institutes of Health institutes are disease‐focused. Funders that are not disease‐focused such as PCORI, National Institute on Aging, National Institute of Nursing Research, and Agency for Healthcare Research and Quality, and private foundations (Hartford, Robert Wood Johnson, and Commonwealth) may be more fruitful sources of funding for this work, but funding may be challenging. Nonetheless, the increased focus on patient‐centered work may increase funders' interest in such work. Second, the topics on this list would suggest that interventions will not be pharmacologic, but will focus on nonpharmacologic, behavioral, and social interventions. Similarly, outcomes of interest must expand beyond utilization metrics such as length of stay and mortality, to include functional status and symptom management, and goal‐concordant care. Therefore, research in geriatric acute care will necessarily be multidisciplinary.

Although these 10 high‐priority areas have been selected, this prioritized list is inherently limited by our methodology. First, our survey question was not focused on a disease state, and this wording may have resulted in the list favoring geriatric syndromes rather than common disease processes. Additionally, the resulting questions encompass large research areas and not specific questions about discrete interventions. Our results may also have been skewed by the types of engaged respondents who participated in the consultation, collating, and prioritization phases. In particular, we had a large response from geriatric medicine nurses, whereas some stakeholder groups provided no survey responses. Thus, these respondents were not representative of all possible stakeholders, nor were the survey respondents necessarily representative of each of their organizations. Nonetheless, the participants self‐identified as representative of diverse viewpoints that included patients, caregivers, and advocacy groups, with the majority of stakeholder organizations remaining engaged through the completion of the process. Thus, the general nature of this agenda helps us focus upon larger areas of importance, leaving researchers the flexibility to choose to narrow the focus on a specific research question that may include potential interventions and unique outcomes. Finally, our methodology may have inadvertently limited the number of patient and family caregiver voices in the process given our approach to large advocacy groups, our desire to be inclusive of healthcare professional organizations, and our survey methodology. Other methodologies may have reached more patients and caregivers, yet many healthcare professionals have served as family caregivers to frail elders requiring hospitalization and may have been in an ideal position to answer the survey.

In conclusion, several forces are shaping the future of acute inpatient care. These include the changing demographics of the hospitalized patient population, a rapid increase in the proportion of multimorbid hospitalized older adults, an inpatient workforce (hospitalists, generalists, and subspecialists) with potentially limited geriatrics training, and gaps in evidence‐based guidance to inform diagnostic and therapeutic decision making for acutely ill older patients. Training programs in hospital medicine should be aware of and could benefit from the resulting list of unanswered questions. Our findings also have implications for training to enrich education in geriatrics. Moreover, there is growing recognition that patients and other stakeholders deserve a greater voice in determining the direction of research. In addition to efforts to improve patient‐centeredness of research, these areas have been uniquely identified by stakeholders as important, and therefore are in line with newer priorities of PCORI. This project followed a road map resulting in a patient‐centered research agenda at the intersection of hospital medicine and geriatric medicine.[7] In creating this agenda, we relied heavily on the framework proposed by PCORI. We propose to pursue a dissemination and evaluation strategy for this research agenda as well as additional prioritization steps. We believe the adoption of this methodology will create a knowledge base that is rigorously derived and most relevant to the care of hospitalized older adults and their families. Its application will ultimately result in improved outcomes for hospitalized older adults.

Delirium is a common and costly problem in hospitalized medical patients. It is present on admission in 10% to 31% of cases and develops in up to 56% of patients during hospitalization.[1, 2] Prompt identification and treatment of the cause of delirium is important, because delirium is associated with increased morbidity and mortality, long‐term cognitive impairment, higher cost of care, increased length of stay, and more frequent discharge to an extended care facility.[3, 4, 5, 6]

Delirium can be caused or worsened by a variety of factors including adverse drug events, metabolic abnormalities, infections, immobilization, the use of tethers (eg, physical restraints, bladder catheters, telemetry), and disruption of sleepwake cycle.[7] An appropriate history, medication review, physical examination, and tailored laboratory evaluation is sufficient workup in the majority of cases.[8] However, neurologic processes, such as intracranial mass, intracranial hemorrhage, or stroke, can also present as delirium and require head imaging for diagnosis.

Because head imaging is a costly limited resource, a number of studies have aimed to determine which patients with delirium require this evaluation. The majority of research has focused on head computed tomography (CT) in patients presenting for evaluation to the emergency department (ED). In ED patients presenting with delirium, acute confusion, or altered mental status, head imaging identifies acute intracranial pathologic findings in 14% to 39% of cases.[9, 10, 11, 12, 13, 14] Only 2 studies have evaluated patients with delirium who have already been admitted to the hospital. One study involved patients admitted to a neurology unit with acute confusion and found that the yield of head imaging (head CT and magnetic resonance imaging) was 14% for acute intracranial pathology.[15] Another study reviewed patients admitted to an acute delirium unit and found a similar rate of positive findings on head CT (14.5%).[16] Neither study specified whether the head imaging occurred during initial presentation or later in the hospitalization.

Factors that increase the likelihood that delirium is caused by acute intracranial pathology include acute neurologic deficit, recent history of fall or head trauma, and significantly impaired consciousness.[9, 10, 11, 12, 13, 14, 15, 16, 17] Based on these findings, current guidelines and expert clinical statements recommend head imaging for patients with acute neurologic deficit, recent head trauma, or recent fall.[18, 19, 20]

Expert clinical statements also recommend imaging in cases where the cause is unidentified after appropriate medical testing or where delirium continues despite treatment.[8, 21] Yet the utility of head CT performed for nonresolving delirium or delirium that develops during hospitalization in the absence of recent fall, head trauma, or new neurologic deficits is not known. Our study aimed to determine the diagnostic yield of performing a head CT in this patient population. We hypothesized that the diagnostic yield of head CT in this population would be low.

METHODS

RESULTS

Of 1714 head CT studies performed on hospitalized medical patients from January 1, 2010 to November 30, 2012, 398 studies were performed for an indication of delirium, altered mental status, confusion, encephalopathy, somnolence, or unresponsiveness in patients who were admitted for >24 hours. One hundred seventy‐eight studies were excluded (137 for admitting diagnosis of intracranial process, recent fall, or head trauma, and 41 for new neurologic deficit). There were 220 scans included in the study performed on 210 patients.

Table 1 displays characteristics of the 210 patients who underwent CT head imaging. Of the 42 patients on anticoagulation, 15 were potentially supratherapeutic; 10 were on warfarin (INR range, 3.37.7) and 5 were on intravenous heparin infusion (PTT range, 101>150 seconds). None of these individuals had positive or equivocal findings on head CT.

The main outcomes of the 220 included head CT scans and a separate analysis of the 60 head CT scans performed for indications of somnolence or unresponsiveness are shown in Table 2. The 6 (2.7%) positive and 4 (1.8%) equivocal head CT findings are listed in Table 3. Of the 3 positive results in patients on anticoagulation, 2 were on warfarin with an INR of 2.1 and 2.4, respectively, and another was on warfarin and therapeutic enoxaparin (dosed 1 mg/kg twice daily) with an INR of 1.6. The median time from admission to positive head CT was 8 days, with a range of 2 to 28 days. All of the positive head CT studies resulted in change of management. All equivocal head CT studies resulted in repeat imaging. None of these repeat head imaging studies diagnosed acute intracranial pathology. Chronic findings identified included 111 (50.5%) involution or atrophy, 95 (43.2%) small vessel ischemic disease, 31 (14.1%) prior stroke, and 18 (8.2%) other chronic abnormalities (eg, cyst or meningioma).

DISCUSSION

In this retrospective review, we determined that there is a low diagnostic yield of head CT imaging for identifying the cause of nonresolving or new‐onset delirium in hospitalized medical patients. Only 2.7% of head CT scans resulted in identifying an acute intracranial process. Because of the low number of positive results, no risk factor associations could be made from our study.

The low diagnostic yield of head imaging in hospitalized patients with delirium is particularly important for clinicians who care for hospitalized medical patients. Prior to this study, the yield of head CT scans in hospitalized medical patients with nonresolving or new‐onset delirium was unknown. In cases with known risk factors, such as recent fall, head trauma, or acute neurologic deficit, the guidelines recommend head CT imaging.[18, 19, 20] However, in the absence of these findings, the guidelines do not make any recommendation regarding when and in whom to perform head imaging. Expert statements recommend considering head CT imaging when the cause is not identified after appropriate testing or delirium continues despite treatment.[8, 21] Given these recommendations and lack of data, there is no clear standard of care for ordering head CT imaging when hospitalized patients experience delirium in the absence of known risk factors. The low diagnostic yield in this study suggests that head CT imaging is unlikely to diagnose the cause of delirium in hospitalized patients with nonresolving or new‐onset delirium.

The diagnostic yield of head CT for diagnosis of acute intracranial process in delirium was lower in our study than prior studies, which found between 14.0 and 39.1%.[9, 10, 11, 12, 13, 14, 15, 16] This was expected, as our study excluded patients with new neurologic deficits, recent fall or trauma, or an admitting diagnosis of an intracranial process. Even with these exclusions, we still allowed for a number of findings that prior studies considered to be high risk for intracranial pathology, such as age over 73 years, use of anticoagulation, and deterioration in consciousness level or Glasgow coma score under 14.[10, 11, 16] The inclusion and exclusion criteria were designed to create a generalizable study population without a clear standard of care based on current guidelines and expert statements.

Though the rate of positive findings found in our study is low, it likely overestimates the overall yield of head CT in hospitalized patients with delirium. This is because most hospitalized patients with delirium never receive head imaging. Presumably, ordering clinicians have deemed these patients to be at higher risk for intracranial processes than the average hospitalized patient with delirium who does not receive a head CT. Thus, the true rate of positive findings in head CT imaging in delirious hospitalized medical patients is likely lower than what we identified.

Although head CT had a low diagnostic yield, the positive and equivocal studies had a high impact on clinical care. All of the positive and equivocal head CT results produced a change in management. The equivocal findings led to repeat head imaging; however, none of the repeat images identified the cause of delirium. The positive results produced a more significant change in management, ranging from a higher platelet transfusion target, reversal of anticoagulation, repeat advanced head imaging, neurosurgery consultation, and a change in goals of care to a focus on comfort. No patients in our study underwent neurosurgical intervention.

The challenge for inpatient clinicians is to weigh the low diagnostic yield of head CT with the consequences of a missed or delayed diagnosis of an acute intracranial process. The low diagnostic yield leads to unnecessary cost, resource utilization, radiation exposure, and downstream evaluation of insignificant or indeterminate results when head CT is performed. Alternatively, a missed or delayed diagnosis can lead to potentially reversible morbidity and mortality. Given this, we feel that the routine use of head CT in the evaluation of delirium in hospitalized patients is unnecessary. However, there may be a subset of patients with delirium with an increased risk of acute intracranial processes that would benefit from head imaging. Further research is needed to identify this high‐risk population.

There are a number of limitations to our study. It is a retrospective chart review, which introduces a possibility of bias and relies on proper and thorough documentation. In addition, the diagnosis of delirium was made by individual clinicians without the use of a standardized delirium assessment tool. Furthermore, it is possible there may have been CT scans that were not identified due to mischaracterization of indication, or studies may have been included in individuals with new neurologic deficit or recent fall or trauma that were not documented or clinically appreciated. Finally, the study was conducted on medicine and medical subspecialty patients at a single academic tertiary care institution, potentially limiting the generalizability to patients in other settings.

In conclusion, our study suggests that the diagnostic yield of head CT to evaluate delirium in hospitalized patients in the absence of recent fall, head trauma, or new neurologic deficit is low. The routine use of head CT in evaluation of these individuals is unnecessary. However, there may be a subset of high‐risk individuals in which head CT imaging would be indicated. Further research is needed to identify these high‐risk individuals.

Delirium is a common and costly problem in hospitalized medical patients. It is present on admission in 10% to 31% of cases and develops in up to 56% of patients during hospitalization.[1, 2] Prompt identification and treatment of the cause of delirium is important, because delirium is associated with increased morbidity and mortality, long‐term cognitive impairment, higher cost of care, increased length of stay, and more frequent discharge to an extended care facility.[3, 4, 5, 6]

Delirium can be caused or worsened by a variety of factors including adverse drug events, metabolic abnormalities, infections, immobilization, the use of tethers (eg, physical restraints, bladder catheters, telemetry), and disruption of sleepwake cycle.[7] An appropriate history, medication review, physical examination, and tailored laboratory evaluation is sufficient workup in the majority of cases.[8] However, neurologic processes, such as intracranial mass, intracranial hemorrhage, or stroke, can also present as delirium and require head imaging for diagnosis.

Because head imaging is a costly limited resource, a number of studies have aimed to determine which patients with delirium require this evaluation. The majority of research has focused on head computed tomography (CT) in patients presenting for evaluation to the emergency department (ED). In ED patients presenting with delirium, acute confusion, or altered mental status, head imaging identifies acute intracranial pathologic findings in 14% to 39% of cases.[9, 10, 11, 12, 13, 14] Only 2 studies have evaluated patients with delirium who have already been admitted to the hospital. One study involved patients admitted to a neurology unit with acute confusion and found that the yield of head imaging (head CT and magnetic resonance imaging) was 14% for acute intracranial pathology.[15] Another study reviewed patients admitted to an acute delirium unit and found a similar rate of positive findings on head CT (14.5%).[16] Neither study specified whether the head imaging occurred during initial presentation or later in the hospitalization.

Factors that increase the likelihood that delirium is caused by acute intracranial pathology include acute neurologic deficit, recent history of fall or head trauma, and significantly impaired consciousness.[9, 10, 11, 12, 13, 14, 15, 16, 17] Based on these findings, current guidelines and expert clinical statements recommend head imaging for patients with acute neurologic deficit, recent head trauma, or recent fall.[18, 19, 20]

Expert clinical statements also recommend imaging in cases where the cause is unidentified after appropriate medical testing or where delirium continues despite treatment.[8, 21] Yet the utility of head CT performed for nonresolving delirium or delirium that develops during hospitalization in the absence of recent fall, head trauma, or new neurologic deficits is not known. Our study aimed to determine the diagnostic yield of performing a head CT in this patient population. We hypothesized that the diagnostic yield of head CT in this population would be low.

METHODS

RESULTS

Of 1714 head CT studies performed on hospitalized medical patients from January 1, 2010 to November 30, 2012, 398 studies were performed for an indication of delirium, altered mental status, confusion, encephalopathy, somnolence, or unresponsiveness in patients who were admitted for >24 hours. One hundred seventy‐eight studies were excluded (137 for admitting diagnosis of intracranial process, recent fall, or head trauma, and 41 for new neurologic deficit). There were 220 scans included in the study performed on 210 patients.

Table 1 displays characteristics of the 210 patients who underwent CT head imaging. Of the 42 patients on anticoagulation, 15 were potentially supratherapeutic; 10 were on warfarin (INR range, 3.37.7) and 5 were on intravenous heparin infusion (PTT range, 101>150 seconds). None of these individuals had positive or equivocal findings on head CT.

The main outcomes of the 220 included head CT scans and a separate analysis of the 60 head CT scans performed for indications of somnolence or unresponsiveness are shown in Table 2. The 6 (2.7%) positive and 4 (1.8%) equivocal head CT findings are listed in Table 3. Of the 3 positive results in patients on anticoagulation, 2 were on warfarin with an INR of 2.1 and 2.4, respectively, and another was on warfarin and therapeutic enoxaparin (dosed 1 mg/kg twice daily) with an INR of 1.6. The median time from admission to positive head CT was 8 days, with a range of 2 to 28 days. All of the positive head CT studies resulted in change of management. All equivocal head CT studies resulted in repeat imaging. None of these repeat head imaging studies diagnosed acute intracranial pathology. Chronic findings identified included 111 (50.5%) involution or atrophy, 95 (43.2%) small vessel ischemic disease, 31 (14.1%) prior stroke, and 18 (8.2%) other chronic abnormalities (eg, cyst or meningioma).

DISCUSSION

In this retrospective review, we determined that there is a low diagnostic yield of head CT imaging for identifying the cause of nonresolving or new‐onset delirium in hospitalized medical patients. Only 2.7% of head CT scans resulted in identifying an acute intracranial process. Because of the low number of positive results, no risk factor associations could be made from our study.

The low diagnostic yield of head imaging in hospitalized patients with delirium is particularly important for clinicians who care for hospitalized medical patients. Prior to this study, the yield of head CT scans in hospitalized medical patients with nonresolving or new‐onset delirium was unknown. In cases with known risk factors, such as recent fall, head trauma, or acute neurologic deficit, the guidelines recommend head CT imaging.[18, 19, 20] However, in the absence of these findings, the guidelines do not make any recommendation regarding when and in whom to perform head imaging. Expert statements recommend considering head CT imaging when the cause is not identified after appropriate testing or delirium continues despite treatment.[8, 21] Given these recommendations and lack of data, there is no clear standard of care for ordering head CT imaging when hospitalized patients experience delirium in the absence of known risk factors. The low diagnostic yield in this study suggests that head CT imaging is unlikely to diagnose the cause of delirium in hospitalized patients with nonresolving or new‐onset delirium.

The diagnostic yield of head CT for diagnosis of acute intracranial process in delirium was lower in our study than prior studies, which found between 14.0 and 39.1%.[9, 10, 11, 12, 13, 14, 15, 16] This was expected, as our study excluded patients with new neurologic deficits, recent fall or trauma, or an admitting diagnosis of an intracranial process. Even with these exclusions, we still allowed for a number of findings that prior studies considered to be high risk for intracranial pathology, such as age over 73 years, use of anticoagulation, and deterioration in consciousness level or Glasgow coma score under 14.[10, 11, 16] The inclusion and exclusion criteria were designed to create a generalizable study population without a clear standard of care based on current guidelines and expert statements.

Though the rate of positive findings found in our study is low, it likely overestimates the overall yield of head CT in hospitalized patients with delirium. This is because most hospitalized patients with delirium never receive head imaging. Presumably, ordering clinicians have deemed these patients to be at higher risk for intracranial processes than the average hospitalized patient with delirium who does not receive a head CT. Thus, the true rate of positive findings in head CT imaging in delirious hospitalized medical patients is likely lower than what we identified.

Although head CT had a low diagnostic yield, the positive and equivocal studies had a high impact on clinical care. All of the positive and equivocal head CT results produced a change in management. The equivocal findings led to repeat head imaging; however, none of the repeat images identified the cause of delirium. The positive results produced a more significant change in management, ranging from a higher platelet transfusion target, reversal of anticoagulation, repeat advanced head imaging, neurosurgery consultation, and a change in goals of care to a focus on comfort. No patients in our study underwent neurosurgical intervention.

The challenge for inpatient clinicians is to weigh the low diagnostic yield of head CT with the consequences of a missed or delayed diagnosis of an acute intracranial process. The low diagnostic yield leads to unnecessary cost, resource utilization, radiation exposure, and downstream evaluation of insignificant or indeterminate results when head CT is performed. Alternatively, a missed or delayed diagnosis can lead to potentially reversible morbidity and mortality. Given this, we feel that the routine use of head CT in the evaluation of delirium in hospitalized patients is unnecessary. However, there may be a subset of patients with delirium with an increased risk of acute intracranial processes that would benefit from head imaging. Further research is needed to identify this high‐risk population.

There are a number of limitations to our study. It is a retrospective chart review, which introduces a possibility of bias and relies on proper and thorough documentation. In addition, the diagnosis of delirium was made by individual clinicians without the use of a standardized delirium assessment tool. Furthermore, it is possible there may have been CT scans that were not identified due to mischaracterization of indication, or studies may have been included in individuals with new neurologic deficit or recent fall or trauma that were not documented or clinically appreciated. Finally, the study was conducted on medicine and medical subspecialty patients at a single academic tertiary care institution, potentially limiting the generalizability to patients in other settings.

In conclusion, our study suggests that the diagnostic yield of head CT to evaluate delirium in hospitalized patients in the absence of recent fall, head trauma, or new neurologic deficit is low. The routine use of head CT in evaluation of these individuals is unnecessary. However, there may be a subset of high‐risk individuals in which head CT imaging would be indicated. Further research is needed to identify these high‐risk individuals.

Delirium is a common and costly problem in hospitalized medical patients. It is present on admission in 10% to 31% of cases and develops in up to 56% of patients during hospitalization.[1, 2] Prompt identification and treatment of the cause of delirium is important, because delirium is associated with increased morbidity and mortality, long‐term cognitive impairment, higher cost of care, increased length of stay, and more frequent discharge to an extended care facility.[3, 4, 5, 6]

Delirium can be caused or worsened by a variety of factors including adverse drug events, metabolic abnormalities, infections, immobilization, the use of tethers (eg, physical restraints, bladder catheters, telemetry), and disruption of sleepwake cycle.[7] An appropriate history, medication review, physical examination, and tailored laboratory evaluation is sufficient workup in the majority of cases.[8] However, neurologic processes, such as intracranial mass, intracranial hemorrhage, or stroke, can also present as delirium and require head imaging for diagnosis.

Because head imaging is a costly limited resource, a number of studies have aimed to determine which patients with delirium require this evaluation. The majority of research has focused on head computed tomography (CT) in patients presenting for evaluation to the emergency department (ED). In ED patients presenting with delirium, acute confusion, or altered mental status, head imaging identifies acute intracranial pathologic findings in 14% to 39% of cases.[9, 10, 11, 12, 13, 14] Only 2 studies have evaluated patients with delirium who have already been admitted to the hospital. One study involved patients admitted to a neurology unit with acute confusion and found that the yield of head imaging (head CT and magnetic resonance imaging) was 14% for acute intracranial pathology.[15] Another study reviewed patients admitted to an acute delirium unit and found a similar rate of positive findings on head CT (14.5%).[16] Neither study specified whether the head imaging occurred during initial presentation or later in the hospitalization.

Factors that increase the likelihood that delirium is caused by acute intracranial pathology include acute neurologic deficit, recent history of fall or head trauma, and significantly impaired consciousness.[9, 10, 11, 12, 13, 14, 15, 16, 17] Based on these findings, current guidelines and expert clinical statements recommend head imaging for patients with acute neurologic deficit, recent head trauma, or recent fall.[18, 19, 20]

Expert clinical statements also recommend imaging in cases where the cause is unidentified after appropriate medical testing or where delirium continues despite treatment.[8, 21] Yet the utility of head CT performed for nonresolving delirium or delirium that develops during hospitalization in the absence of recent fall, head trauma, or new neurologic deficits is not known. Our study aimed to determine the diagnostic yield of performing a head CT in this patient population. We hypothesized that the diagnostic yield of head CT in this population would be low.

METHODS

RESULTS

Of 1714 head CT studies performed on hospitalized medical patients from January 1, 2010 to November 30, 2012, 398 studies were performed for an indication of delirium, altered mental status, confusion, encephalopathy, somnolence, or unresponsiveness in patients who were admitted for >24 hours. One hundred seventy‐eight studies were excluded (137 for admitting diagnosis of intracranial process, recent fall, or head trauma, and 41 for new neurologic deficit). There were 220 scans included in the study performed on 210 patients.

Table 1 displays characteristics of the 210 patients who underwent CT head imaging. Of the 42 patients on anticoagulation, 15 were potentially supratherapeutic; 10 were on warfarin (INR range, 3.37.7) and 5 were on intravenous heparin infusion (PTT range, 101>150 seconds). None of these individuals had positive or equivocal findings on head CT.

The main outcomes of the 220 included head CT scans and a separate analysis of the 60 head CT scans performed for indications of somnolence or unresponsiveness are shown in Table 2. The 6 (2.7%) positive and 4 (1.8%) equivocal head CT findings are listed in Table 3. Of the 3 positive results in patients on anticoagulation, 2 were on warfarin with an INR of 2.1 and 2.4, respectively, and another was on warfarin and therapeutic enoxaparin (dosed 1 mg/kg twice daily) with an INR of 1.6. The median time from admission to positive head CT was 8 days, with a range of 2 to 28 days. All of the positive head CT studies resulted in change of management. All equivocal head CT studies resulted in repeat imaging. None of these repeat head imaging studies diagnosed acute intracranial pathology. Chronic findings identified included 111 (50.5%) involution or atrophy, 95 (43.2%) small vessel ischemic disease, 31 (14.1%) prior stroke, and 18 (8.2%) other chronic abnormalities (eg, cyst or meningioma).

DISCUSSION

In this retrospective review, we determined that there is a low diagnostic yield of head CT imaging for identifying the cause of nonresolving or new‐onset delirium in hospitalized medical patients. Only 2.7% of head CT scans resulted in identifying an acute intracranial process. Because of the low number of positive results, no risk factor associations could be made from our study.

The low diagnostic yield of head imaging in hospitalized patients with delirium is particularly important for clinicians who care for hospitalized medical patients. Prior to this study, the yield of head CT scans in hospitalized medical patients with nonresolving or new‐onset delirium was unknown. In cases with known risk factors, such as recent fall, head trauma, or acute neurologic deficit, the guidelines recommend head CT imaging.[18, 19, 20] However, in the absence of these findings, the guidelines do not make any recommendation regarding when and in whom to perform head imaging. Expert statements recommend considering head CT imaging when the cause is not identified after appropriate testing or delirium continues despite treatment.[8, 21] Given these recommendations and lack of data, there is no clear standard of care for ordering head CT imaging when hospitalized patients experience delirium in the absence of known risk factors. The low diagnostic yield in this study suggests that head CT imaging is unlikely to diagnose the cause of delirium in hospitalized patients with nonresolving or new‐onset delirium.

The diagnostic yield of head CT for diagnosis of acute intracranial process in delirium was lower in our study than prior studies, which found between 14.0 and 39.1%.[9, 10, 11, 12, 13, 14, 15, 16] This was expected, as our study excluded patients with new neurologic deficits, recent fall or trauma, or an admitting diagnosis of an intracranial process. Even with these exclusions, we still allowed for a number of findings that prior studies considered to be high risk for intracranial pathology, such as age over 73 years, use of anticoagulation, and deterioration in consciousness level or Glasgow coma score under 14.[10, 11, 16] The inclusion and exclusion criteria were designed to create a generalizable study population without a clear standard of care based on current guidelines and expert statements.

Though the rate of positive findings found in our study is low, it likely overestimates the overall yield of head CT in hospitalized patients with delirium. This is because most hospitalized patients with delirium never receive head imaging. Presumably, ordering clinicians have deemed these patients to be at higher risk for intracranial processes than the average hospitalized patient with delirium who does not receive a head CT. Thus, the true rate of positive findings in head CT imaging in delirious hospitalized medical patients is likely lower than what we identified.

Although head CT had a low diagnostic yield, the positive and equivocal studies had a high impact on clinical care. All of the positive and equivocal head CT results produced a change in management. The equivocal findings led to repeat head imaging; however, none of the repeat images identified the cause of delirium. The positive results produced a more significant change in management, ranging from a higher platelet transfusion target, reversal of anticoagulation, repeat advanced head imaging, neurosurgery consultation, and a change in goals of care to a focus on comfort. No patients in our study underwent neurosurgical intervention.

The challenge for inpatient clinicians is to weigh the low diagnostic yield of head CT with the consequences of a missed or delayed diagnosis of an acute intracranial process. The low diagnostic yield leads to unnecessary cost, resource utilization, radiation exposure, and downstream evaluation of insignificant or indeterminate results when head CT is performed. Alternatively, a missed or delayed diagnosis can lead to potentially reversible morbidity and mortality. Given this, we feel that the routine use of head CT in the evaluation of delirium in hospitalized patients is unnecessary. However, there may be a subset of patients with delirium with an increased risk of acute intracranial processes that would benefit from head imaging. Further research is needed to identify this high‐risk population.

There are a number of limitations to our study. It is a retrospective chart review, which introduces a possibility of bias and relies on proper and thorough documentation. In addition, the diagnosis of delirium was made by individual clinicians without the use of a standardized delirium assessment tool. Furthermore, it is possible there may have been CT scans that were not identified due to mischaracterization of indication, or studies may have been included in individuals with new neurologic deficit or recent fall or trauma that were not documented or clinically appreciated. Finally, the study was conducted on medicine and medical subspecialty patients at a single academic tertiary care institution, potentially limiting the generalizability to patients in other settings.

In conclusion, our study suggests that the diagnostic yield of head CT to evaluate delirium in hospitalized patients in the absence of recent fall, head trauma, or new neurologic deficit is low. The routine use of head CT in evaluation of these individuals is unnecessary. However, there may be a subset of high‐risk individuals in which head CT imaging would be indicated. Further research is needed to identify these high‐risk individuals.