The increasing availability of medical treatments to decrease

Transcription

1 Higher Vulnerable Elders Survey Scores Predict Death and Functional Decline in Vulnerable Older People Lillian C. Min, MD, Marc N. Elliott, PhD, z Neil S. Wenger, MD, wz and Debra Saliba, MD, MPH z OBJECTIVES: To examine whether the Vulnerable Elders Survey (VES-13) score predicts risk of death and functional decline in vulnerable older adults. DESIGN: Longitudinal evaluation with mean follow-up of 11 months (range 8 14 months). SETTING: Two managed care organizations in the United States. PARTICIPANTS: Four hundred twenty community-dwelling older people identified as having moderate to high risk of death and functional decline based on a VES-13 score of 3 or higher. These older people were enrolled in the Assessing Care of Vulnerable Elders observational study. MEASUREMENTS: Baseline: VES-13 score, sex, income, cognitive score, and number of medical diagnoses. Outcome measures: functional decline and death. RESULTS: VES-13 scores strongly predicted death and functional decline (Po.001, area under the receiver operating curve ). The estimated combined risk of death and decline rose with VES-13 score, increasing from 23% for older people with a VES-13 score of 3 to 60% for those with a score of 10. Other measures (sex, comorbidity) were not significant predictors of death or decline over this period after controlling for VES-13 score. CONCLUSION: The VES-13 score is useful as a screening tool to detect risk of health deterioration in already vulnerable older populations, and higher scores reflect greater risk over a short follow-up period. J Am Geriatr Soc 54: , Key words: mortality; functional decline; screening tool From the Divisions of Geriatrics and w General Internal Medicine and Health Services Research, University of California, Los Angeles; z RAND, Santa Monica, California; and Greater Los Angeles Veterans Affairs Healthcare System, Los Angeles, California. Dr. Min is supported by a Geriatrics Health Outcomes Research Scholars Award from the Hartford/Foundation for Health in Aging and by the Bureau of Health Professions Geriatric Faculty Training Program ( ). Dr. Saliba was supported by a Career Development Award from the Veterans Affairs Health Services Research and Development Service. Address correspondence to Lillian C. Min, MD, Department of Medicine, Division of Geriatrics, University of California, Los Angeles, Le Conte Avenue, Suite 2339, Box , Los Angeles, CA lmin@mednet.ucla.edu DOI: /j x The increasing availability of medical treatments to decrease mortality and preserve function in older patients creates a growing need for an easily administered tool to screen for vulnerability to death and functional decline. A brief screening instrument could aid providers assessments of older adults with high levels of comorbidity, improve prioritization of care for patients, and allow for better targeting of evaluations and healthcare services. 1,2 The Vulnerable Elders Survey (VES-13) 3 is a simple 13- item screening tool that asks older people to report their age, their ability to perform six physical and five functional activities, and their self-rated health. It aims to identify a group of community-dwelling older people at risk for death or decline and who might therefore benefit from improved detection and care of prevalent medical and geriatric conditions known to result in functional decline and mortality. Nonclinical personnel can administer the VES-13 during a brief (o5-minute) telephone interview and derive a score ranging from 0 (lowest risk) to 10 (highest risk). Using Medicare Current Beneficiary Survey (MCBS) data, one study found that VES-13 scores of 3 or higher identified a group of older people at a 4.2 times greater risk of functional decline or death 2 years later than those with scores less than 3 (49.5% vs 11.8%). The VES-13 was nearly as predictive as a much-lengthier tool that required identification of medical conditions. 3 The prognostic value of the VES-13 has not been verified prospectively in an independent community sample. It is also not known whether it predicts risk within follow-up intervals of less than 2 years. In addition, the differential properties of VES-13 scores of 5 and higher have not been tested, because the 2-year prognosis in these individuals in the MCBS sample was uniformly poor. An interpretation of these higher scores might be desirable in the clinic setting (e.g., alerting clinicians to the most vulnerable patients who might need specialized geriatric care), in population-based targeting of older people who are predominantly vulnerable at baseline, and in allowing healthcare organizations to better understand the extent of vulnerability within their at-risk group. Finally, although self-reported medical conditions did not contribute to the VES-13 in the MCBS cohort over 2 years, it would be useful to reconsider the role of JAGS 54: , 2006 r 2006, Copyright the Authors Journal compilation r 2006, The American Geriatrics Society /06/$15.00

2 508 MIN ET AL. MARCH 2006 VOL. 54, NO. 3 JAGS comorbidity in predicting risk, because comorbidity might have a more important role over shorter follow-up times in a more-vulnerable group. METHODS Participants The validation sample was drawn from a population of 3,207 community-dwelling older people ( 65) enrolled in two large managed care organizations. The Assessing Care of Vulnerable Elders (ACOVE) Study 4 successfully contacted 88% (n 5 2,810) by telephone, and 90% of these (n 5 2,521) agreed to be screened (cumulative response rate 5 79%). Ten percent (n 5 243) did not meet ACOVE Study inclusion criteria: not health plan members (n 5 54), subject or proxy unable to participate in screening due to poor health (n 5 18) or non-english speaking (n 5 122), receiving cancer therapy (n 5 49). Nonclinical interviewers conducted screening telephone interviews with the remaining 2,278 eligible older people to determine functional status and VES-13 score (average VES-13 interview time 5 4 minutes). Twenty-one percent (n 5 475) of these older people were identified as vulnerable (VES-13 score 3), of whom 88% (n 5 420) consented to participate in a comprehensive assessment of the quality of their medical care. The 55 who refused at this stage did not significantly differ from the 420 participants in terms of age, sex, or baseline VES-13 score. Data This study included a 13-month medical record review and a subsequent in-depth interview that queried functional status items identical to the screening interview. Medical records were the source of information for chronic illness and were reviewed for the 89% of patients (n 5 372) with adequate records and at least some medical care beyond influenza vaccination. The National Death Index (NDI) provided death information for the entire sample. Because of variation in respondent availability, the follow-up interview was conducted 8.4 months to 13.8 months (mean standard deviation months) after the baseline interview. Outcomes Patients health outcomes were represented in the same manner as in the derivation of the VES-13 survey, as a multinomial outcome with three categories: death before the end of the field period (10/20/00), functional decline among those surviving until the end of the data collection period (defined as change from no functional disability to any functional disability, an increase of two or more in the total disability count, or admission to a nursing home), or no death or decline. In additional logistic regression models, outcome was defined as death or decline versus no death or decline, combining the first two categories. Twelve functional disabilities were considered for the functional decline outcome: requiring assistance with bathing, feeding, toileting, transfer, walking, dressing, shopping, managing finances, light housework, using the telephone, managing medications, and preparing meals. The outcome of one individual who completed a follow-up interview and later died before October 20, 2000, was classified as a death. The VES-13 and Other Candidate Predictors The main predictor of interest, baseline VES-13 score, was computed from age ( point, points), selfrated health (fair or poor 5 1 point), difficulty with one or more physical activities (stooping, crouching, or kneeling; walking one quarter of a mile; lifting 10 pounds; heavy housework; reaching above shoulder level; writing or grasping small objects 5 1 point for each activity, maximum of 2 points), and requiring assistance with any of five activities (shopping, light housework, finances, walking across room, or bathing 5 4 points). VES-13 scores can range from a minimum of 0 (no risk factors for decline) to a maximum of 10 (greatest number of risk factors for decline). Because those with scores of 0 to 2 were excluded from the study, the scoring range for this analysis was 3 to 10. VES-13 was tested in the model as a linear variable after preliminary exploration, because a set of indicator variables suggested a monotonic relationship with outcome. Comorbidity was calculated as a count of 15 chronic medical conditions associated with mortality or functional decline in older adults: dementia, 5,6 depression, 7 10 diabetes mellitus, 8,11 hearing loss, 12,13 heart failure, 10,14 hypertension, 8,15,16 ischemic heart disease, 8,10 osteoarthritis, 10,17 osteoporosis, 8,18,19 stroke, 8,10 atrial fibrillation, 20,21 urinary incontinence, 9,22 chronic renal failure, 23 chronic obstructive lung disease, 10,24 and pressure ulcers. 25 A missing indicator was used for 48 patients with inadequate medical records. The effect of comorbidity was evaluated using a joint test of the count and missing conditions indicator. Finally, age, measured continuously in years, was tested for its predictive ability beyond the three-category scoring employed by the VES-13. To determine whether bias resulted from varying time to re-interview (i.e., patients with longer follow-up intervals may have had greater opportunity to decline or were more difficult to interview due to deterioration in health), time exposure was also tested as a covariate in the first multinomial logistic regression. Loss-to Follow-Up Weights This analysis also applied weights to adjust for possible bias if selective loss to follow-up had occurred. These weights reflected the inverse probability of obtaining follow-up information and were derived using the full sample of 420 older people. Those who died before October 20, 2000, received a full weight of one because all deaths were obtained from the NDI. For the remaining 387 living participants, income, sex, cognitive status, 26 and Mental Health Index 27 score were used to estimate probability of re-interview. Analysis First, a multinomial logistic regression of the three-level health outcome (death, functional decline, or no decline) was performed to determine whether VES-13 score predicted death and functional decline outcomes independently in the patients for whom follow-up information (interview and death data) was available, weighting for

3 JAGS MARCH 2006 VOL. 54, NO. 3 VES-13 SCORES PREDICT DEATH AND FUNCTIONAL DECLINE 509 loss to follow-up. Bootstrapped 95% confidence intervals were calculated for predicted probabilities of each of the three outcomes for VES-13 scores from 3 to 10. The effect of re-interview time exposure on functional decline was tested within this multinomial model. Because it was not statistically significant (b-coefficient , P 5.95), it was excluded from subsequent models that considered functional decline as any part of the outcome. Second, the overall predictive value of the VES-13 was evaluated in a logistic regression of the combined outcome of death or decline (vs no decline), because these two outcomes are not necessarily distinguishable (i.e., functional decline can precede death or re-interview), again weighting for loss to follow-up. A receiver operating characteristic curve was calculated for this model. Third, to evaluate mortality effects in the full sample of 420 older people, a Cox proportional hazards survival analysis was performed to determine the effect of the VES- 13 score on death at 13 months. Full information on this outcome was available through the NDI, so this model did not require weighting. Kaplan-Meier survival curves were plotted to examine whether high scorers dying early in the follow-up interval (i.e., older people in the midst of a rapid decline) could explain an observed effect of the VES-13 score. Fourth, the potential for improvements in VES-13 scoring was assessed by adding continuous age, sex, and a count of comorbid conditions to the multinomial and survival analyses as predictors. RESULTS The mean baseline VES-13 score for the group identified as vulnerable was 5.3 (range 3 10). The mean number of conditions was 2.2 (range 0 7). The mean age of patients at baseline was 80.8, and two-thirds were female. Using the Blessed Orientation-Memory Concentration test 26 to assess baseline cognitive status, 62% were rated good (score 17) and 15% as poor (score17), and 23% were not tested because of representation by proxy respondent. At baseline, nearly all with VES-13 scores of 3 to 5 were free from activity of daily living (ADL) or instrumental activity of daily living (IADL) impairments (97% and 90%, respectively). For those with VES-13 scores of 6 or higher, half were independent in all ADLs, and 11% were independent in all IADLs. Three hundred eight vulnerable older people were considered for the multinomial longitudinal outcome analysis; 33 (11%) had died by the end of the follow-up period, and 275 (71%) completed a follow-up interview. Seventy-two (26%) of the re-interviewed participants had functional decline, half of whom were free of any ADL or IADL impairments at baseline. Fifty-four interviews (20%) were conducted with a proxy respondent. In the multivariable logistic model of loss to follow-up (using the full sample of 420 to determine loss-to follow-up weights), low or missing income and male sex were associated with lower probability of follow-up (Po.05 for all). Higher VES-13 scores significantly predicted greater odds of decline and death in the multinomial logistic model. The odds of functional decline versus no decline were multiplied by 1.18 for each 1-point increase in VES-13 score Figure 1. Predicted probability of no decline, death, or decline at 8 to 14 months according to Vulnerable Elders Survey (VES-13) score (N 5 308). Greater VES-13 score (x-axis) in the range of 3 to 10 (normal value 5 0) is associated with greater predicted probability (y-axis) of death and decline in vulnerable older patients. Y-axis values corresponding to the curve drawn above the light gray middle zone represent the sum of the probability of decline and death because the zones have been stacked (i.e., at each level of the VES-13, the sum of the probabilities of all three outcomes equals 1). For example, an older person with a VES-13 score of 7 has a 14.0% predicted probability of death plus an additional 27.2% predicted probability of decline. The combined probability of death or decline is 14.0%127.2% % (vs 58.8% for no decline). Bottom (dark gray) zone 5 predicted probability of death. Middle (light gray) zone 5 predicted probability of decline. Top (white) zone 5 predicted probability of no decline. Solid lines 5 smoothed curves drawn through predicted probabilities. Dotted lines 5 bootstrapped 95% confidence limits around predicted probabilities. (P 5.008), and the odds of death versus no decline were multiplied by 1.50 per 1-point increase in VES-13 score (Po.001). The effect of the VES-13 score on predicted probability of death and functional decline over mean follow-up time of Sensitivity Specificity Figure 2. Vulnerable Elders Survey (VES-13) predicts death or decline versus no decline at 8 to 14 months (N 5 308). VES-13 scores discriminate a dichotomous outcome (death or decline vs no decline) with an area under the receiver operating characteristic curve of VES-13 scores range from 4 to 10. 4

4 510 MIN ET AL. MARCH 2006 VOL. 54, NO. 3 JAGS Proportion of sample (n=420) remaining By VES-13 score of 7 or higher versus score of 6 or lower Days after enrollment VES-13 score: 7-10 VES-13 score: 3-6 Figure 3. Kaplan-Meier survival estimates of time to death according to Vulnerable Elders Survey (VES-13) scores of 7 or higher versus 6 or lower. Older people with VES-13 scores of 7 to 10 experience shorter time to death than those with scores of 3 to 6. Deaths (24 of the higher scorers and 9 of the lower scorers) did not occur disproportionately early in the follow-up interval. 11 months is displayed graphically in Figure 1. For a VES- 13 score of 3, the predicted probability of death was 0.04, and the predicted probability of functional decline was 0.19 (combined probability of 0.23); at the highest VES-13 score of 10, the predicted probability of death increased to 0.30, and the predicted probability of functional decline increased to 0.30 (combined probability of 0.60). When death or decline was considered as a joint outcome, the odds of the combined outcome were multiplied by 1.25 for each 1-point increase in VES-13 score (Po.001), and the area under the receiver operating characteristic curve was 0.66 (Figure 2). The Cox proportional hazards model predicting time to death from continuous VES-13 scores resulted in a hazard ratio of 1.39 (Po.001) (i.e., the relative risk of death at any point in time was multiplied by 1.39 for every additional unit of VES-13 score). For example, risk of death for an older person with a score of 8 would be 5.19 times higher than the risk for an older person with a score of 3, because 1.39 (8 3) Figure 3 illustrates the Kaplan-Meier survival curves for two groups: those with VES-13 scores of 3 to 6 and those with scores of 7 to 10. Deaths accumulated throughout the follow-up period in both groups. Additional predictors were then introduced into the models. In the expanded multinomial logistic regression, the odds ratios (ORs) associated with the VES-13 remained significant (OR for decline, P 5.008; OR for death, Po.001). Other covariates were not significant for decline (age OR , P 5.76; male sex OR , P 5.87; number of conditions OR , P 5.11) or for death (age OR , P 5.52; male sex OR , P 5.59; number of conditions OR , P 5.14). Similarly, the VES-13 was the only significant predictor (hazard ratio , Po.001) of death at 13 months in the expanded multivariable Cox proportional hazards analysis of 420 patients. DISCUSSION In this prospective validation study, higher VES-13 scores predicted greater risk for death and decline over a mean follow-up time of 11 months in a population of vulnerable community-dwelling older adults. The relationship between VES-13 score and the odds of death or decline increased linearly from scores of 3 through 10, and each additional point substantially increased the risk of health deterioration. This study found that VES-13, an easily administered tool (o5 minutes by telephone by a nonclinician), might further differentiate those at greater short-term risk of death and decline in an already at-risk population. For the purposes of screening a more-vulnerable population for deterioration, higher VES-13 scores effectively differentiate higher-risk from moderate-risk patients. The VES-13 can serve as a screening tool for identifying vulnerability that is not already apparent to clinicians. It relies on information readily provided by patients: self-rated health, age, and ability to perform five ADLs and six physical tasks. Half of the patients who eventually underwent functional decline were free of any ADL or IADL impairments at baseline. In addition, functional status items such as those in the VES-13 are not routinely obtained or assessed during generalist office visits. 28 This study extends current understanding of VES-13 scores. The original derivation study found that a VES-13 score of 3 or more identified older people who were vulnerable to death or decline at 2 years. The current study expands the potential use of this tool, showing that differentiating scores higher than 3 (e.g., scores from 5 to 10) can provide meaningful additional prediction of risk of functional decline and death. The differential risks associated with higher scores are observable over a shorter follow-up interval (mean of 11 months). In addition, the current study confirms an earlier finding that collecting additional medical condition information would not significantly improve the VES-13, thus avoiding added survey burden. A recent analysis of quality of care also supports the clinical utility of identifying short-term risk of death or decline using the VES-13. In a population identified as vulnerable (VES-13 score 3), those who received better-thanaverage quality of medical and geriatric care experienced significantly lower mortality than those who received worse care. This difference was seen as soon as 1.5 years after the care was delivered. 29 Thus, the VES-13 seems to identify older people likely to benefit over a short time from higher quality of care. The current study has some limitations. First, this evaluation of the VES-13 score was performed on a sample frailer than the average community population, because those with VES-13 scores of 0 to 2 were excluded. Despite this restricted sample range, 1-point changes in this simple scoring system predict substantial differences in deterioration. Therefore, the current study findings are limited to persons who score 3 or higher on the VES-13 and serve primarily to differentiate among at-risk patients. A second possible limitation is that some of the older people may have been experiencing an acute or rapid trajectory of decline that was readily apparent to clinicians at baseline, thus obviating the need for screening. This study did not test providers recognition of patients functional limitations, although other studies have shown this to be a problem. 28 In addition, the survival analyses showed that the predictive ability of the VES-13 remains robust over the entire observation period.

5 JAGS MARCH 2006 VOL. 54, NO. 3 VES-13 SCORES PREDICT DEATH AND FUNCTIONAL DECLINE 511 A third limitation was that possible bias might have been introduced by loss to follow-up. To address this limitation, this study employed loss-to follow-up weights and additional analyses using the death outcome only because this information did not rely on the follow-up interview. Despite these limitations, this study provides useful insight into the potential use of the VES-13 in older people at greater baseline risk (i.e., those who score 3). For this population, increasing scores can discriminate risk for death or decline within approximately the next year. This ability to differentiate risk may be particularly useful for clinical practices or research samples that find a large number of vulnerable older people (i.e., VES-13 score 3) when baseline screening is conducted. Future studies might compare the predictive ability, acceptability, and feasibility of the VES-13 with those of other prediction tools. The predictive ability of the VES-13 in different ethnic and language groups should also be evaluated. In addition, intervention trials could measure whether care quality improves if clinicians are aware of vulnerability or whether knowledge of VES-13 score improves the targeting of specific interventions such as comprehensive geriatric assessment. In conclusion, this study demonstrated that the VES-13 predicted death and functional decline over a mean followup of 11 months in a group of older people who were independent of the original derivation (MCBS) study. This practical survey instrument can be used to identify which vulnerable older people are at the highest risk of decline in clinic and population-based settings. ACKNOWLEDGMENTS We recognize the technical assistance of Patricia Smith. Ms. Caren Kamberg and Dr. Takahiro Higashi contributed to the acquisition of NDI data. We thank Robin Hertz, PhD, senior director of outcomes research/population studies at Pfizer Inc. for her support. Financial Disclosure: This project was supported by a contract from Pfizer Inc. to RAND. Lillian Min received research support from the Geriatrics Health Outcomes Research Scholars Award from the Hartford/Foundation for Health in Aging ( ) and the Bureau of Health Professions Geriatric Faculty Training Program ( ). Debra Saliba received research support from the Veterans Affairs Health Services Research and Development Service. Author Contributions: Lillian Min and Marc Elliott: concept and design, analysis and interpretation of data, preparation of manuscript. Neil Wenger and Debra Saliba: concept and design, acquisition of subjects and data, analysis and interpretation of data, preparation of manuscript. Sponsor s Role: The funding source had no role in the design, analysis, or interpretation of the study or in the decision to submit the results for publication. REFERENCES 1. Gill TM, Baker DI, Gottschalk M et al. A program to prevent functional decline in physically frail, elderly persons who live at home. N Engl J Med 2002;347: Winograd CH, Gerety MB, Brown E et al. Targeting the hospitalized elderly for geriatric consultation. J Am Geriatr Soc 1988;36: Saliba D, Elliott M, Rubenstein LZ et al. The Vulnerable Elders Survey: A tool for identifying vulnerable older people in the community. J Am Geriatr Soc 2001;49: Wenger NS, Solomon DH, Roth CP et al. The quality of medical care provided to vulnerable community-dwelling older patients. Ann Intern Med 2003;139: Aguero-Torres H, Fratiglioni L, Guo Z et al. Mortality from dementia in advanced age: A 5-year follow-up study of incident dementia cases. J Clin Epidemiol 1999;52: Aguero-Torres H, Fratiglioni L, Guo Z et al. Dementia is the major cause of functional dependence in the elderly: 3-year follow-up data from a populationbased study. Am J Public Health 1998;88: Yaffe K, Edwards ER, Covinsky KE et al. Depressive symptoms and risk of mortality in frail, community-living elderly persons. Am J Geriatr Psychiatry 2003;11: Wang L, van Belle G, Kukull WB et al. Predictors of functional change: A longitudinal study of nondemented people aged 65 and older. J Am Geriatr Soc 2002;50: Cho CY, Alessi CA, Cho M et al. The association between chronic illness and functional change among participants in a comprehensive geriatric assessment program. J Am Geriatr Soc 1998;46: Guccione AA, Felson DT, Anderson JJ et al. The effects of specific medical conditions on the functional limitations of elders in the Framingham Study. Am J Public Health 1994;84: Wu JH, Haan MN, Liang J et al. Diabetes as a predictor of change in functional status among older Mexican Americans: A population-based cohort study. Diabetes Care 2003;26: Reuben DB, Mui S, Damesyn M et al. The prognostic value of sensory impairment in older persons. J Am Geriatr Soc 1999;47: Keller BK, Morton JL, Thomas VS et al. The effect of visual and hearing impairments on functional status. J Am Geriatr Soc 1999;47: Pulignano G, Del Sindaco D, Tavazzi L et al. Clinical features and outcomes of elderly outpatients with heart failure followed up in hospital cardiology units: Data from a large nationwide cardiology database (IN-CHF Registry). Am Heart J 2002;143: Staessen JA, Gasowski J, Wang JG et al. Risks of untreated and treated isolated systolic hypertension in the elderly: Meta-analysis of outcome trials. Lancet 2000;355: Glynn RJ, Field TS, Rosner B et al. Evidence for a positive linear relation between blood pressure and mortality in elderly people. Lancet 1995;345: Ling SM, Fried LP, Garrett ES et al. Knee osteoarthritis compromises early mobility function: The Women s Health and Aging Study II. J Rheumatol 2003;30: Johnell O, Kanis JA, Oden A et al. Mortality after osteoporotic fractures. Osteoporos Int 2004;15: Chrischilles EA, Butler CD, Davis CS et al. A model of lifetime osteoporosis impact. Arch Intern Med 1991;151: Molander U, Dey DK, Sundh V et al. ECG abnormalities in the elderly: Prevalence, time and generation trends and association with mortality. Aging Clin Exp Res 2003;15: Kannel WB, Wolf PA, Benjamin EJ et al. Prevalence, incidence, prognosis, and predisposing conditions for atrial fibrillation: Population-based estimates. Am JCardiol1998;82:2N 9N. 22. Nuotio M, Tammela TL, Luukkaala T et al. Urgency and urge incontinence in an older population: Ten-year changes and their association with mortality. Aging Clin Exp Res 2002;14: Fried LF, Shlipak MG, Crump C et al. Renal insufficiency as a predictor of cardiovascular outcomes and mortality in elderly individuals. J Am Coll Cardiol 2003;41: Fried TR, Pollack DM, Tinetti ME. Factors associated with six-month mortality in recipients of community-based long-term care. J Am Geriatr Soc 1998;46: Brown G. Long-term outcomes of full-thickness pressure ulcers: Healing and mortality. Ostomy Wound Manage 2003;49: Katzman R, Brown T, Fuld P et al. Validation of a short Orientation-Memory- Concentration Test of cognitive impairment. Am J Psychiatry 1983;140: Veit CT, Ware JE Jr. The structure of psychological distress and well-being in general populations. J Consult Clin Psychol 1983;51: Calkins DR, Rubenstein LV, Cleary PD et al. Failure of physicians to recognize functional disability in ambulatory patients. Ann Intern Med 1991;114: Higashi T, Shekelle PG, Adams J et al. Quality of care is associated with survival in vulnerable older patients. Ann Intern Med 2005;143:

6