Edward Ratner

Decision aids for advance care planning: An overview of the state of the science

Advance care planning is a way to inform care choices for a patient who cannot express a preference and a planning tool that helps patients begin to prioritize their treatment goals. The preferences of seriously ill patients for life-sustaining interventions depend on their care goals. Some prioritize living longer to achieve life goals, whereas others may not wish to be kept alive when meaningful recovery or a particular quality of life is no longer possible (13). Religious and spiritual values and beliefs also affect goals of care (4, 5). Advance care planning helps to honor patient preferences and goals if incapacitating illness or injury prevents adequate communication (6). Decision aids help patients consider health care options. Such aids for advance care planning support the 3 key components of the process: learning about anticipated conditions and options for care; considering these options; and communicating preferences for future care, either orally or in writing. The most important information a decision aid can provide to a decision maker depends on the patients current health status and the predictability of illness trajectories (Figure). A healthy person may benefit most from general decision aids focused on choice of health care proxies and goals of care for hypothetical catastrophic situations, such as after loss of function or cognition or terminal illness. For patients with a life-threatening illness, appropriate aids focus on decisions to accept, withhold, or terminate specific treatments. Advance care planning with decision aids takes place in various settings; it is often done outside clinical settings, particularly among healthy older adults. Nonclinical partners in shared decision making may include family members, caregivers, or attorneys or other professionals. Figure. Continuum of health states during which advance care planning may be considered. Opportunity exists for expansion and improvement of advance care planning. A 2003 Agency for Healthcare Research and Quality (AHRQ) literature summary (7) found that fewer than 50% of the severely or terminally ill patients who were studied had an advance directive (a common outcome of the advance care planning process) in their medical records (811). Furthermore, only 12% of patients with an advance directive had received input from their physician in its development (9), and physicians were accurate only about 65% of the time when predicting patient preferences; they tended to assume that patients would want less life-prolonging treatment than they actually desired, even after reviewing the patients advance directive (12). Decision aids may improve participation in advance care planning and the effectiveness of communication by facilitating clear documentation across platforms and providers and by offering insights into why patients make the decisions they do. This review, commissioned as a technical brief by the AHRQ Effective Health Care Program, provides an overview of advance care planning decision aids for adults. It describes available tools, identifies a framework for future research, and summarizes published studies that used a decision aid as an intervention. Methods Key Informants In November 2013, we conducted semistructured telephone interviews (Appendix Table 1) with 7 key informants, including practicing clinicians and attorneys involved in advance care planning, experts in medical law and medical ethics, consumer advocates, and decision aid researchers and developers. We identified these informants via frequently listed and cited authors of relevant literature, Internet searches for persons with potentially relevant viewpoints, and nominations by other key informants. They contributed information about decision aids, the context in which they are used, and important issues to consider. Appendix Table 1. Interview Probes for Key Informants Literature Search We searched MEDLINE (via Ovid), the Cochrane Library, PsycINFO, and CINAHL from January 1990 to May 2014 using a search strategy based on relevant Medical Subject Headings terms and text words (Appendix Table 2). We also conducted a gray literature search of federal and state government Web sites, the Ottawa Hospital Research Institutes Decision Aid Library Inventory, Web sites of professional organizations, and leads from key informants for decision aids available to the public and in use. Appendix Table 2. MEDLINE Search Strategy We screened abstracts and full-text articles to identify English-language studies of any sample size and design that assessed the effect of a decision aid on outcomes relevant to advance care planning. We excluded studies that involved children or advance planning for psychiatric care. We also excluded studies of decision aids for current (not future or hypothetical) end-of-life decisions; studies of forms for completing advance directives, living wills, or provider orders for life-sustaining treatment that did not include an educational component, help clarify values, or prompt action; and studies that focused on implementation science questions. The reviewers read the full text of selected articles and used a standardized data extraction form to collect reported information about study populations, decision aids, and outcomes. One reviewer abstracted data by using standardized abstraction tables, and a second reviewer provided a quality check. We used the criteria developed by the International Patient Decision Aids Standards (IPDAS) Collaboration to provide a structure for describing and comparing decision aids. These criteria have been used formally to judge quality and effectiveness in existing systematic literature reviews (13, 14). Because we followed technical brief methods, we did not synthesize outcomes, rate risk of bias, or grade the strength of evidence of the literature. Role of the Funding Source The Minnesota Evidence-based Practice Center (EPC) prepared this technical brief with funding from AHRQ. The EPC collaborated with AHRQ to develop the research protocol. Staff at AHRQ helped formulate questions and reviewed the draft report but were not involved in the study selection, data extraction, or drafting of the manuscript for publication. The full report is available at www.effectivehealthcare.ahrq.gov. Results Existing Advance Care Planning Decision Aids and Context for Use In shared clinical decision making, patients and clinicians use evidence-based knowledge, weigh options against treatment goals, and consensually arrive at a clinically prudent decision concordant with patient preferences (15, 16). Although advance care planning lies within the bounds of clinical decision making, it differs from many well-studied decision processes for medical procedures (such as surgical or nonsurgical options for cancer) because patients can make decisions with no health care provider involvement by using readily available, do-it-yourself decision aids. These aids tend to target persons with only general risks for life-threatening conditions, for whom advance care planning may involve considering a wide range of possible future scenarios, eliciting preferred goals of care, or choosing a health care proxy. Although not exhaustive, Table 1 summarizes advance care planning decision aids that target a general, predominantly healthy, older adult audience. These aids, identified through the gray literature search and by key informants, are relatively easy to find online by using common search engines. The most popular issues they address include designation of a health care proxy, clarification of values and desire for comfort care at the end of life, information on living wills or other advance directives, conversation prompts for talking to loved ones or physicians about wishes, and general preferences for various life-sustaining treatments. These aids vary in the degree to which they include the 3 components of our working definition of advance care planning decision aids, which is based on the IPDAS criteria (13, 14): an education component, a structured approach to thinking about the choices a patient faces, and a way for those choices to be communicated. Table 1. Examples of General Advance Care Planning Decision Aids Publicly Available on the Internet General decision aids for advance care planning are often used in conjunction with tools to document the decisions. Health care preferences can be documented in an advance directive and stored at a Web site, such as MyDirectives (www.MyDirectives.com). One or more proxies and their powers can be documented in a durable power of attorney for health care or as part of a more comprehensive advance directive. Health care providers can record advance care planning results (from oral discussions or an advance directive) in health care records; a specific order (such as a do-not-resuscitate order); or a template most commonly called a Physician Orders for Life-Sustaining Treatment form (found at www.polst.org), which has the advantage of serving as standing orders. Most patients gain clarity about what information can best support specific advance care planning treatment decisions as they move from hypothetical to actual clinical decisions and their familiarity with health states increases or when the health state for which a decision is needed becomes more certain. For patients with predictable progressive disease (such as amyotrophic lateral sclerosis), chronic critical illness, or frailty, a structured approach to decisions in advance care planning often requires information on prognosis. Table 2, which is not exhaustive, summarizes decision aids for advance care planning that target patients with a life-limiting illness, for which the decision trajectory is often more clearly defined. These tools are distinct from the general population tools in Table 1 because they are more likely to focus on a single advance care planning topic. They also are more likely to be designed by decision-making

Dying on the Streets: Homeless Persons’ Concerns and Desires about End of Life Care

BackgroundThere is little understanding about the experiences and preferences at the end of life (EOL) for people from unique cultural and socioeconomic backgrounds. Homeless individuals are extreme examples of these overlooked populations; they have the greatest risk of death, encounter barriers to health care, and lack the resources and relationships assumed necessary for appropriate EOL care. Exploring their desires and concerns will provide insight for the care of this vulnerable and disenfranchised population, as well as others who are underserved.ObjectiveExplore the concerns and desires for EOL care among homeless persons.DesignQualitative study utilizing focus groups.ParticipantsFifty-three homeless persons recruited from agencies providing homeless services.MeasurementsIn-depth interviews, which were audiotaped and transcribed.ResultsWe present 3 domains encompassing 11 themes arising from our investigation, some of which are previously unreported. Homeless persons worried about dying and EOL care; had frequent encounters with death; voiced many unique fears, such as dying anonymously and undiscovered; favored EOL documentation, such as advance directives; and demonstrated ambivalence towards contacting family. They also spoke of barriers to EOL care and shared interventions to improve dying among the very poor and estranged.ConclusionsHomeless persons have significant personal experience and feelings about death, dying, and EOL care, much of which is different from those previously described in the EOL literature about other populations. These findings have implications not only for homeless persons, but for others who are poor and disenfranchised.

Decision Aids for Advance Care Planning

Advance care planning is a way to inform care choices for a patient who cannot express a preference and a planning tool that helps patients begin to prioritize their treatment goals. The preferences of seriously ill patients for life-sustaining interventions depend on their care goals. Some prioritize living longer to achieve life goals, whereas others may not wish to be kept alive when meaningful recovery or a particular quality of life is no longer possible (13). Religious and spiritual values and beliefs also affect goals of care (4, 5). Advance care planning helps to honor patient preferences and goals if incapacitating illness or injury prevents adequate communication (6). Decision aids help patients consider health care options. Such aids for advance care planning support the 3 key components of the process: learning about anticipated conditions and options for care; considering these options; and communicating preferences for future care, either orally or in writing. The most important information a decision aid can provide to a decision maker depends on the patients current health status and the predictability of illness trajectories (Figure). A healthy person may benefit most from general decision aids focused on choice of health care proxies and goals of care for hypothetical catastrophic situations, such as after loss of function or cognition or terminal illness. For patients with a life-threatening illness, appropriate aids focus on decisions to accept, withhold, or terminate specific treatments. Advance care planning with decision aids takes place in various settings; it is often done outside clinical settings, particularly among healthy older adults. Nonclinical partners in shared decision making may include family members, caregivers, or attorneys or other professionals. Figure. Continuum of health states during which advance care planning may be considered. Opportunity exists for expansion and improvement of advance care planning. A 2003 Agency for Healthcare Research and Quality (AHRQ) literature summary (7) found that fewer than 50% of the severely or terminally ill patients who were studied had an advance directive (a common outcome of the advance care planning process) in their medical records (811). Furthermore, only 12% of patients with an advance directive had received input from their physician in its development (9), and physicians were accurate only about 65% of the time when predicting patient preferences; they tended to assume that patients would want less life-prolonging treatment than they actually desired, even after reviewing the patients advance directive (12). Decision aids may improve participation in advance care planning and the effectiveness of communication by facilitating clear documentation across platforms and providers and by offering insights into why patients make the decisions they do. This review, commissioned as a technical brief by the AHRQ Effective Health Care Program, provides an overview of advance care planning decision aids for adults. It describes available tools, identifies a framework for future research, and summarizes published studies that used a decision aid as an intervention. Methods Key Informants In November 2013, we conducted semistructured telephone interviews (Appendix Table 1) with 7 key informants, including practicing clinicians and attorneys involved in advance care planning, experts in medical law and medical ethics, consumer advocates, and decision aid researchers and developers. We identified these informants via frequently listed and cited authors of relevant literature, Internet searches for persons with potentially relevant viewpoints, and nominations by other key informants. They contributed information about decision aids, the context in which they are used, and important issues to consider. Appendix Table 1. Interview Probes for Key Informants Literature Search We searched MEDLINE (via Ovid), the Cochrane Library, PsycINFO, and CINAHL from January 1990 to May 2014 using a search strategy based on relevant Medical Subject Headings terms and text words (Appendix Table 2). We also conducted a gray literature search of federal and state government Web sites, the Ottawa Hospital Research Institutes Decision Aid Library Inventory, Web sites of professional organizations, and leads from key informants for decision aids available to the public and in use. Appendix Table 2. MEDLINE Search Strategy We screened abstracts and full-text articles to identify English-language studies of any sample size and design that assessed the effect of a decision aid on outcomes relevant to advance care planning. We excluded studies that involved children or advance planning for psychiatric care. We also excluded studies of decision aids for current (not future or hypothetical) end-of-life decisions; studies of forms for completing advance directives, living wills, or provider orders for life-sustaining treatment that did not include an educational component, help clarify values, or prompt action; and studies that focused on implementation science questions. The reviewers read the full text of selected articles and used a standardized data extraction form to collect reported information about study populations, decision aids, and outcomes. One reviewer abstracted data by using standardized abstraction tables, and a second reviewer provided a quality check. We used the criteria developed by the International Patient Decision Aids Standards (IPDAS) Collaboration to provide a structure for describing and comparing decision aids. These criteria have been used formally to judge quality and effectiveness in existing systematic literature reviews (13, 14). Because we followed technical brief methods, we did not synthesize outcomes, rate risk of bias, or grade the strength of evidence of the literature. Role of the Funding Source The Minnesota Evidence-based Practice Center (EPC) prepared this technical brief with funding from AHRQ. The EPC collaborated with AHRQ to develop the research protocol. Staff at AHRQ helped formulate questions and reviewed the draft report but were not involved in the study selection, data extraction, or drafting of the manuscript for publication. The full report is available at www.effectivehealthcare.ahrq.gov. Results Existing Advance Care Planning Decision Aids and Context for Use In shared clinical decision making, patients and clinicians use evidence-based knowledge, weigh options against treatment goals, and consensually arrive at a clinically prudent decision concordant with patient preferences (15, 16). Although advance care planning lies within the bounds of clinical decision making, it differs from many well-studied decision processes for medical procedures (such as surgical or nonsurgical options for cancer) because patients can make decisions with no health care provider involvement by using readily available, do-it-yourself decision aids. These aids tend to target persons with only general risks for life-threatening conditions, for whom advance care planning may involve considering a wide range of possible future scenarios, eliciting preferred goals of care, or choosing a health care proxy. Although not exhaustive, Table 1 summarizes advance care planning decision aids that target a general, predominantly healthy, older adult audience. These aids, identified through the gray literature search and by key informants, are relatively easy to find online by using common search engines. The most popular issues they address include designation of a health care proxy, clarification of values and desire for comfort care at the end of life, information on living wills or other advance directives, conversation prompts for talking to loved ones or physicians about wishes, and general preferences for various life-sustaining treatments. These aids vary in the degree to which they include the 3 components of our working definition of advance care planning decision aids, which is based on the IPDAS criteria (13, 14): an education component, a structured approach to thinking about the choices a patient faces, and a way for those choices to be communicated. Table 1. Examples of General Advance Care Planning Decision Aids Publicly Available on the Internet General decision aids for advance care planning are often used in conjunction with tools to document the decisions. Health care preferences can be documented in an advance directive and stored at a Web site, such as MyDirectives (www.MyDirectives.com). One or more proxies and their powers can be documented in a durable power of attorney for health care or as part of a more comprehensive advance directive. Health care providers can record advance care planning results (from oral discussions or an advance directive) in health care records; a specific order (such as a do-not-resuscitate order); or a template most commonly called a Physician Orders for Life-Sustaining Treatment form (found at www.polst.org), which has the advantage of serving as standing orders. Most patients gain clarity about what information can best support specific advance care planning treatment decisions as they move from hypothetical to actual clinical decisions and their familiarity with health states increases or when the health state for which a decision is needed becomes more certain. For patients with predictable progressive disease (such as amyotrophic lateral sclerosis), chronic critical illness, or frailty, a structured approach to decisions in advance care planning often requires information on prognosis. Table 2, which is not exhaustive, summarizes decision aids for advance care planning that target patients with a life-limiting illness, for which the decision trajectory is often more clearly defined. These tools are distinct from the general population tools in Table 1 because they are more likely to focus on a single advance care planning topic. They also are more likely to be designed by decision-making

Experiences With and Attitudes Toward Death and Dying Among Homeless Persons

BackgroundHomeless persons face many barriers to health care, have few resources, and experience high death rates. They live lives of disenfranchisement and neglect. Few studies have explored their experiences and attitudes toward death and dying. Unfortunately, studies done in other populations may not apply to homeless persons. Exploring these experiences and attitudes may provide insight into life, health care, and end-of-life (EOL) concerns of this population.ObjectiveTo explore the experiences and attitudes toward death and dying among homeless persons.DesignQualitative study utilizing focus groups.ParticipantsFifty-three homeless persons recruited from homeless service agencies.MeasurementsIn-depth interviews, which were audiotaped and transcribed.ResultsWe present seven themes, some of which are previously unreported. Homeless persons described many significant experiences with death and dying, and many participants suffered losses while very young. These encounters influenced participants’ attitudes toward risks and risky behavior: e.g., for some, these experiences provided justification for high-risk behaviors and influenced their behaviors while living on the streets. For others, they may be associated with their homelessness. Finally, these experiences informed their attitudes toward death and dying as well as EOL care; homeless persons believe that care will be poor at the EOL.ConclusionsFindings from this study have implications for addressing social services, health promotion, prevention, and EOL care for homeless persons, as well as for others who are poor and disenfranchised.

Effect of an End-of-Life Planning Intervention on the Completion of Advance Directives in Homeless Persons: A Randomized Trial

BACKGROUND Few interventions have focused on improving end-of-life care for underserved populations, such as homeless persons. OBJECTIVE To determine whether homeless persons will complete a counseling session on advance care planning and fill out a legal advance directive designed to assess care preferences and preserve the dignity of marginalized persons. DESIGN Prospective, single-blind, randomized trial comparing self-guided completion of an advance directive with professionally assisted advance care planning. (ClinicalTrials.gov registration number: NCT00546884) SETTING 8 sites serving homeless persons in Minneapolis, Minnesota. PARTICIPANTS 262 homeless persons recruited between November 2007 and August 2008. INTERVENTION Minimal, self-guided intervention consisting of advance directive forms and written educational information versus a one-on-one advance planning intervention consisting of counseling and completing an advance directive with a social worker. MEASUREMENTS Rate of advance directive completion, assessed by inspection of completed documents. RESULTS The overall completion rate for advance directives was 26.7% (95% CI, 21.5% to 32.5%), with a higher rate in the counselor-guided group (37.9%) than in the self-guided group (12.8%) (CI of adjusted difference, 15.3 to 34.3 percentage points). This difference persisted across all sites and most subgroups. The advance directives 4 clinical scenarios found a preference for surrogate decision making in 29% to 34% of written responses. LIMITATIONS Sampling was limited to a more stable subset of the homeless population in Minneapolis and may have been subject to selection bias. Modest compensation to complete the preintervention survey could have influenced participants to complete advance directives. CONCLUSION Both a simple and complex intervention successfully engaged a diverse sample of homeless persons in advance care planning. One-on-one assistance significantly increased the completion rate. Homeless persons can respond to an intervention to plan for end-of-life care and can express specific preferences for care or a surrogate decision maker, but additional studies are needed to assess the effect of these directives on subsequent care. PRIMARY FUNDING SOURCE National Institute for Nursing Research and National Center on Minority Health and Health Disparities.

Engaging Homeless Persons in End of Life Preparations

BackgroundThere are no prospective studies that have investigated the effects of an intervention to improve end of life (EOL) care in an underserved population.ObjectiveTo determine whether homeless persons will complete an advance directive (AD).DesignRandomized trial comparing two modes of providing an opportunity for homeless persons to complete an AD. Half of the subjects were randomized to a self-guided group (SG) who were given an AD and written instructions; the other half were given the same material but, in addition, were offered the opportunity to receive guidance to complete the AD (CG).ParticipantsFifty-nine homeless persons recruited from a drop-in center.MeasuresRate of AD completion and baseline and 3-month follow-up EOL-related knowledge, attitudes, and behaviors.ResultsThe overall AD completion rate was 44%, with a statistically significant higher completion rate of 59% in the CG group compared to 30% in the self-guided only group. Frequency of worry about death decreased among those who filled out an AD from 50% to 12.5%, and also among those who did not (25% to 12.5%) (p < .05). Among those who filled out an AD, there were increases in plans to write down EOL wishes (56% to 100%; p < .05) and plans to talk about these wishes with someone (63% to 94%; p < .05).ConclusionThis study demonstrates that people living in dire economic and social situations will complete an AD when offered the opportunity. While offering guidance resulted in higher rates of completion; even a simple self-guided AD process can achieve completion of ADs in this population.

Perception, satisfaction and utilization of the VALUE home telehealth service.

We conducted a randomized controlled trial to evaluate the perception, satisfaction and utilization of a home telehealth service for frail elderly people living independently in their home communities. Control group subjects continued with their usual care and intervention group subjects were able to supplement their usual care with the use of a web portal. The web portal allowed videoconferencing and electronic messaging between home care nurses and clients, ordering health-related and home care services, access to health-related information and general access to the Internet. A total of 99 eligible people (59 female, 40 male) from one urban and one rural study site agreed to participate in the study. Eighty-four subjects were active participants for nine months. There were no significant differences in perception of technology between the two groups at baseline. At 60-day follow-up, the intervention group was significantly more positive towards technology compared to their own baseline (P < 0.001) and compared to the 60-day scores for the controls (P < 0.001). The intervention group indicated that overall the telehealth service met their expectations (mean score 9 out of 10) and they would recommend it to others (mean score 9.5 out of 10). All subjects were able to use the portal without difficulty after brief instructions from the nurses. A total of 1054 videoconferences were conducted with the intervention subjects. Fifty-six of these (5%) had to be discontinued after successful connection because of technical problems. Intervention subjects made fewer emergency department visits than control subjects, more visits to the eye doctor, fewer visits in all categories of home care utilization, and lower use of transportation services. Frail elderly people are able to adopt home telehealth technologies which may enable them to maintain independent living arrangements.

Resolution of the multiple inert gas method for estimating V̇a/Q̇ maldistribution

During steady-state infusion of a mixture of dissolved inert gases, their elimination by the lung depends on the distribution of ventilation/perfusion (VA/Q) ratios. Thus, certain features of the VA/Q distribution can be inferred from inert gas measurement. Because of the: (1) complexity of the lung, and (2) experimental errors, the ability of such a technique to describe the shape and position of the VA/Q distribution accurately is limited. In this report we present an analysis of the resolution of the method for 9 representative sets of inert gas data, taking account of both of the above factors. These 9 sets span the range of commonly observed data, both in health and in diseases such as asthma, interstitial fibrosis, chronic obstructive lung disease and respiratory distress syndromes. Both error-free and error-containing data are studied and by linear programming methods, bounds are placed on maximum and minimum possible perfusion in several regions of the VA/Q spectrum. Modality is also studied by linear programming. The results show that the resolving power of the method depends greatly on the specific case under study. When groups of units are separated in VA/Q by a decade, this can be determined with considerable confidence. Shunt and low VA/Q areas can generally be well resolved, but when distributions are very broad, resolution is limited.

Workforce Development in Geriatric Home Care

With the rapidly aging population, it is anticipated that within two decades several million more individuals in the United States with functional impairment and serious ill health will need home health care. This article discusses workforce development, which is a critical issue for future planning, as recently highlighted by the Institute of Medicine (IOM). Key aspects of recruitment, training, and retention of home care workers are discussed, including those who provide basic support for activities of daily living as well as a variety of skilled professionals: therapists, nurses, pharmacists, and physicians. Although the geriatric workforce shortage affects all care settings, it is especially critical in home health care, in part because we are starting with far too few clinicians to meet the medical needs of homebound elderly. A combination of actions is needed, including educational programs, such as those developed by the American Academy of Home Care Physicians (AAHCP), changes in financial incentives, and changes in the culture and practice of health care, to make the home the primary focus of care for these vulnerable, underserved individuals rather than an afterthought.

Physical Activity Interventions in Preventing Cognitive Decline and Alzheimer-Type Dementia: A Systematic Review

Forty-seven million people worldwide live with dementia (1), and this number is expected to triple by 2050 (2). Despite evidence that the overall incidence of dementia has declined in the United States (3, 4), the number of U.S. adults older than 70 years with dementia or mild cognitive impairment (MCI) increases as our population ages (5, 6). Dementia severely erodes functioning and quality of life, creates burden and stress on families, and leads to institutionalization. Dementia-related costs exceed those of heart disease and cancer and often are paid directly by families (7). Therefore, preventing dementia is an urgent public health priority. Many believe that an active lifestyle may prevent cognitive decline and dementia. Findings of several reviews, primarily those looking at cohort studies, suggest that physical activity may reduce or delay the development of potential modifiable risk factors for cognitive decline, such as obesity, diabetes, and hypertension (813). However, the relationships among physical activity, other risk factors, and cognitive decline are complex and interrelated. Findings of associations from cohort studies alone cannot clarify whether physical activity affects cognitive decline directly, indirectly through the reduction of medical risk factors, or both. Previous systematic reviews of randomized controlled trials report some cognitive benefits of physical activity interventions, although the certainty and clinical importance of these findings have not always been clear (14, 15). This systematic review reports a synthesis of the evidence assessing the effectiveness of physical activity interventions in slowing cognitive decline and delaying the onset of cognitive impairment and dementia in adults without diagnosed cognitive impairments. Methods We developed and followed a standard protocol (16). Our full technical report (17) contains details on methods and findings, an analysis of studies addressing secondary prevention in adults with MCI, and an evaluation of comparative effectiveness. Data Sources and Searches We searched bibliographic databases, including MEDLINE, EMBASE, and PsycINFO via Ovid, as well as the Cochrane Library, to identify controlled trials published in any language from January 2009 through July 2017. (See Part A of the Supplement, for search strategies.) We identified studies published before 2009 by citation searching relevant systematic reviews. Supplement. Data Supplement Study Selection Two investigators independently reviewed titles and abstracts of search results and screened the full text of potentially eligible references. Disagreements about eligibility were resolved by consensus. We included randomized controlled trials of physical activity interventions with any sample size and large (n > 500) prospective quasi-experimental cohort studies with comparator groups if they enrolled adults without diagnosed cognitive impairments, had follow-up of at least 6 months, were published in English, and reported 1 of our preselected primary or intermediate outcomes. We excluded trials enrolling pure subgroups of patients with major medical conditions or conditions that may explain changes in cognitive function (namely stroke, Parkinson disease, cancer, and traumatic brain injury). Our main outcomes of interest were MCI or dementia. Intermediate outcomes included measures of cognitive function assessed by instruments that tested cognition across several domains or those that specifically tested executive function, attention, and processing speed, or memory. Intermediate outcomes were categorized as follows: broad measures intended to capture several cognitive domains that were either brief cognitive tests (category 1) or more comprehensive multidomain neuropsychological tests (category 2) and domain-specific neuropsychological tests or subscales of broader instruments that assessed executive function, attention, and processing speed (category 3) or memory (category 4). Part B of the Supplement shows a list of the intermediate outcomes reported from the studies and our categorization of those outcomes. Data Extraction and Quality Assessment One reviewer extracted the study population, treatment characteristics, and funding source from all eligible studies. Risk of bias was assessed independently by 2 investigators using an instrument developed with guidance from the Agency for Healthcare Research and Quality (AHRQ) (18). Risk of bias for each reported outcome was rated as low, medium, or high on the basis of adequacy of randomization and allocation concealment, masking, attrition, use of intention-to-treat analyses, selectiveness of outcome reporting, and confidence that results were believable given limitations. Outcomes and adverse effects were extracted from eligible trials with low or moderate risk of bias, and a second investigator checked the extraction. Data Synthesis and Analysis We grouped studies by type of physical activity intervention and analyzed results by direction of effect and statistical significance. We found it impossible to assess the clinical significance of findings of the intermediate outcomes across all studies, because many different instruments were used and we did not always find information on the degree of change in specific instrument scores or subscores that would indicate clinical importance. (Part B of the Supplement shows the information we did find about clinically important changes in specific instrument scores.) In addition, results were measured, analyzed, and reported in many different ways. When sufficient data were available (from more than 1 study or 1 study with 500 participants), 1 investigator assessed the strength of evidence for unique comparisons. These assessments were confirmed through consensus. We assessed strength of evidence by using 5 required domains: study limitations (risk of bias of eligible studies for a given comparison), directness (single, direct link between intervention and outcome), consistency (similarity of effect direction and size), precision (degree of certainty around an estimate that includes attention to small sample sizes with power to detect only large differences), and reporting bias (19). On the basis of these factors, the overall strength of evidence for each outcome from a given intervention was rated as high, moderate, low, or insufficient. Role of the Funding Source This review was funded by the National Institute on Aging and AHRQ. These agencies and members of the National Academies Committee on Preventing Dementia and Cognitive Impairment helped refine the scope and reviewed a draft report of findings. The authors are solely responsible for the content preparation, writing of the manuscript, and decision to submit the manuscript for publication. Results Of 32 eligible studies that compared interventions using physical activity components with an inactive control in adults without a cognitive impairment diagnosis (2051), 16all of which were randomized trialswere considered to have low to medium risk of bias (20, 23, 25, 29, 30, 33, 3537, 39, 40, 43, 45, 46, 48, 50). Inactive controls in the trials with low to medium risk of bias included waitlist, usual care, no-intervention, and attention (that is, education and information) groups. Most trials were government funded. Most studies enrolled older adults; some limited enrollment to men or women. Total sample sizes ranged from 42 to 1635 participants. Trials rarely reported adverse effects; those that did showed no differences between groups, with 1 exception. Intervention components, frequency, and duration varied. (Part C of the Supplement contains the literature flow diagram; part D contains evidence tables.) The Table shows overall conclusions and strength-of-evidence ratings. Details of studies considered to have low to medium risk of bias are described later. For any cognitive outcome, evidence was insufficient to draw conclusions about most interventions (aerobic training, resistance training, tai chi, physical activity with diet, and physical activity with a cognitive component). Low-strength evidence showed that multicomponent physical activity interventions of 1 to 2 years did not improve multidomain neurologic performance; executive function, attention, and processing speed; or memory compared with an attention control. Low-strength evidence showed that an intervention combining physical activity, diet, and cognitive training benefited multidomain neuropsychological test performance and executive function, attention, and processing speed compared with an attention control; however, evidence was insufficient to draw conclusions about the efficacy of this intervention on memory. Moderate-strength evidence showed that more participants in the intervention than the control groups had musculoskeletal pain. Table. Conclusions: Physical Activity Versus Inactive Comparisons in Adults Physical Activity Interventions Multicomponent Physical Activity Four trials (n= 1885) with low to medium risk of bias examined multicomponent physical activity interventions. Components included flexibility, strength, balance, endurance, and aerobic training (36, 45, 46, 50). Enrollment criteria varied by trial. Sink and colleagues (45) and Williamson and colleagues (50) enrolled sedentary adults older than 70 years, most of whom were white women. Mean Modified Mini-Mental State Examination (MMSE) scores were higher than 90 points (on a scale of 0 to 100 points). Taylor-Piliae and colleagues (46) enrolled adults, mostly white college-educated women, older than 60 years. Napoli and colleagues (36) enrolled frail, obese older adults, most of whom were white women; mean Modified MMSE score was 96 points. Interventions during the trials lasted from 6 months to 2 years. Sink and colleagues (45) (n= 1635) reported diagnostic outcomes that showed no difference in the incidence of MCI (odds ratio, 1.14 [95% CI, 0.79 to 1.62]) or dementia (odds ratio, 0.9