Higher Quality of Care for Hospitalized Older Adults is Associated with Improved Survival One Year After Discharge Vineet Arora MD, MA, Melissa Fish BA, Anirban Basu PhD, Jared Olson MD, Colleen Plein BA, Kalpana Suresh MS, Greg Sachs MD, David O Meltzer MD, PhD Academy Health Annual Research Meeting June 29, 2009 Quality of care for hospitalized older adults Increasing use of quality measures for hospital care Public reporting and pay-forperformance Ideally, adherence to quality measures is associated with improved patient outcomes Need to test and develop meaningful quality measures for hospitalized older patients Increasing numbers of hospitalized older patients Greater severity of illness and more comorbidities Different goals and preferences of health care “Hazards of hospitalization” (i.e. delirium, pressure ulcers) ACOVE: Assessing Care of Vulnerable Elders Series of if/then statements specifically developed for vulnerable elders (Wenger, 2001) Associated with decreased mortality for community dwelling elders (Higashi, 2005) Adapted for hospitalized older patients (Arora, 2006) Association with patient outcomes still unknown Aim of study To assess the relationship between quality of care for hospitalized vulnerable elders (using ACOVE quality measures) and post-discharge mortality Methods: patient selection Patients age 65 and older recruited from University of Chicago Hospitalist Project from May 2004 through June 2007 Excluded Length of stay <1 day (Meltzer, 2002) Identified as vulnerable elders VES-13: 13-item Vulnerable Elders Survey (Saliba, 2001) Score of 3 or higher = “vulnerable” (at greater risk for functional decline or death) Died in hospital or discharged to hospice or comfort care age, self-rated health, functional ability Hard to consider the impact of quality of care during a short stay Death may not have been an unexpected outcome Transferred from an ICU Greater severity of disease Methods: Quality of Care Chart abstraction based on select ACOVE quality measures (Arora, 2006) 6 measures for all patients Effort to improve mobility, discharge planning, assessment of cognitive status, functional ability, nutritional status, and pain 10 other measures assessed only when relevant If a patient has dementia, pressure ulcers, etc. Methods: Mortality data Social Security Death Index Publicly available website Used matching algorithm based on first and last name, date of birth, zip code, and last four digits of Social Security number (Calle, 1993) Recorded dates of death Able to calculate time to death from discharge date http://ssdi.rootsweb.ancestry.com/ Methods: data analysis Composite quality score (Higashi, 2006) quality indicators met quality score = × 100% quality indicators triggered Adherence to universally applicable individual QIs Effort to improve mobility, discharge planning, assessment of cognitive status, functional ability, nutritional status, and pain Methods: data analysis Cox proportional hazards models to test association between quality score and mortality adjusting for: Demographics (age, sex, race, marital status) Markers of patient illness Likely to have an independent effect on mortality VES-13 score DNR/DNI status Charlson comorbidity score # of quality measures triggered length of stay # baseline ADL limitations clustered by attending physician Repeated models to test association between adherence to individual QIs and mortality Results: patient population Between May 2004 and June 2007 3633 (69%) older inpatients consented 2040 (56%) identified as vulnerable 1861 (91%) eligible for chart review did not die in hospital, not discharged to hospice, not transferred from ICU 1856 (99.7%) charts reviewed Study population (n = 1856) Characteristic Age in years Mean* 79.8 ± 8.3 Female, n (%) 1281 (69%) African-American, n (%) 1355 (73%) Single, n (%) 1311 (71%) Impaired cognitive status, n (%) 106 (6%) VES-13 score 5.8 ± 2.1 Length of hospital stay in days 5.9 ± 9.0 # of ADL limitations at admission 2.7 ± 2.7 Charlson comorbidity score 1.9 ± 1.6 DNR/DNI order present, n (%) 177 (10%) Quality of care score 59.3 ± 19.2 Died within 30 days, n (%) 114 (6.1%) Died within 1 year, n (%) 495 (26.7%) *Except where noted Study population (n = 1856) Characteristic Age in years Mean* 79.8 ± 8.3 Female, n (%) 1281 (69%) African-American, n (%) 1355 (73%) Single, n (%) 1311 (71%) Impaired cognitive status, n (%) 106 (6%) VES-13 score 5.8 ± 2.1 Length of hospital stay in days 5.9 ± 9.0 # of ADL limitations at admission 2.7 ± 2.7 Charlson comorbidity score 1.9 ± 1.6 DNR/DNI order present, n (%) 177 (10%) Quality of care score 59.3 ± 19.2 Died within 30 days, n (%) 114 (6.1%) Died within 1 year, n (%) 495 (26.7%) *Except where noted Cox proportional hazards regression Hazards ratio = 0.82 (95% CI = 0.68-1.00) p value < 0.05 Relationship between disease severity covariates and mortality Covariate Cox hazards ratio (95% CI) p value VES-13 score 1.09 (1.04-1.15) <0.001 DNR/DNI status 1.89 (1.41-2.54) <0.001 Number of QIs triggered 1.06 (1.00-1.12) 0.036 Charlson comorbidity score 1.10 (1.04-1.17) 0.002 Length of stay 1.40 (1.25-1.58) <0.001 Number of ADL limitations 1.07 (1.03-1.12) 0.001 As predicted, sicker patients are more likely to die Individual measures and mortality Quality indicator Hazards ratio (95% CI) p value Cognitive status screening 0.66 (0.39-1.10) 0.11 Functional ability screening 0.90 (0.70-1.16) 0.53 Mobility improvement plan 0.81 (0.65-1.01) 0.07 Discharge planning 0.89 (0.65-1.20) 0.43 Nutritional status screening 0.61 (0.40-0.93) 0.02 Pain assessment 0.93 (0.78-1.11) 0.42 *Derived from Cox proportional hazards regression model testing the relationship between predictor variables & post-discharge patient mortality, adjusted for race, gender, age, marital status, VES-13 score, DNR/DNI status, Charlson comorbidity score, number of quality indicators triggered, length of hospital stay, and number of baseline ADL limitations Limitations Cannot assume causality Could be unmeasured confounders driving mortality One institution Mortality may not be the most relevant outcome for elderly patients Completing research examining functional decline Conclusions Hospitalized older persons who receive higher quality of care (as measured by ACOVE quality indicators) are less likely to die one year after discharge Consistent with prior research on communitydwelling elders Association especially strong for the assessment of nutritional status Mechanism is unclear Implications Possible mechanisms for these findings: Better quality of care improves post-discharge survival Inpatients that will live longer receive higher quality of care Adherence to certain process of care measures is a proxy for an unmeasured variable influencing survival Future research needed to confirm possible mechanisms for these findings In particular, examining the effect of nutrition interventions for hospitalized VE’s Questions? varora@uchicago.edu Hospitalist Study Support Staff Korry Schwanz Hui Tang Ben Vekhter Ainoa Mayo Meryl Prochaska Supported by: Juned Siddique DrPH, Northwestern University Hartford Foundation Health Outcomes Research Scholars Award 1T35AG028785-01A1 NIA Short Term Aging-related Research Program 1U18HS016967-01 AHRQ CERT 1 RO1 GM075292-01 NIGMS Effectiveness of TEACH Research Donald W Reynolds Foundation: CHAMP EXTRA SLIDES Quality of Care Delivered & Post-Discharge Mortality Number of Time period post- patients who died (%) discharge *Quality Odds ratio (95% confidence interval)† p value 30 days 114 (6.1) 0.88 (0.76-1.02) 0.08 60 days 191 (10.3) 0.90 (0.81-1.00) 0.05 90 days 246 (13.3) 0.94 (0.86-1.04) 0.22 1 year 495 (26.7) 0.93 (0.87-1.00) 0.04 of care calculated as percent of quality indicators triggered that were actually met, here divided by 10 †Derived from multivariate logistic regression analysis, adjusting for race, gender, age, marital status, VES-13 score, DNR/DNI status, Charlson comorbidity score, number of quality indicators triggered, length of hospital stay, and number of baseline ADL limitations Methods: disease severity Measures of disease severity Age VES-13 score Number of ADL limitations } Assessed in initial study recruitment interview DNR/DNI status Length of hospital stay Number of quality measures for which a patient is eligible Charlson comorbidity score } Assessed in chart abstraction