Comparative Rankings of Hospital Quality – Does the Data Source Matter? Anne Elixhauser, Ph.D. Bernard Friedman, Ph.D. June 26, 2005 AcademyHealth Research Meeting Background Hospitals are being compared based on readily available data Data on Medicare patients are available from virtually all U.S. hospitals Convenient – but do we know how the Medicare experience reflects hospital quality overall? Purpose of Study Examine the extent to which information on Medicare patients can be extrapolated to the general population when comparing hospital quality Methods – Data Source 2001 Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample (NIS) Sample of 986 hospitals from 33 states – All discharges from each hospital are included No weighting for this study – Used NIS as a convenience sample of hospitals Study population: hospital inpatients from short-term, non-Federal, acute care hospitals Methods – Quality Measures AHRQ Quality Indicators – Based on hospital administrative data – 15 Patient Safety Indicators (PSIs) Risk adjusted using gender, age, comorbidities, and collapsed DRGs – 12 in-hospital mortality measures from the Inpatient Quality Indicators (IQIs) Risk adjusted using APR-DRGs Methods – Rankings For each measure: – Dropped the 10% of hospitals with the fewest Medicare cases – Using Medicare discharges only Ranked hospitals and ordered into deciles – Using all patients (including Medicare) Ranked hospitals and ordered into deciles Comparison of Rankings How many hospitals changed from the highest or lowest rank by at least two deciles? Compared rank using Medicare-only data to rank using all-patient data What % of poorest performing hospitals increased their ranking? What % of best-performing hospitals fell in their ranking? Percent of hospitals in lowest decile that increased rank by at least 2 deciles Patient Safety Indicators 50 45 40 35 30 25 20 15 10 5 0 s l r sth D RG ulce scue ody orax are ip fxha ge nge p fai DVT epsi ence tion e a n ort us re gn b ot h ed c op h orr dera res PE p s is c c era x C m ubit e t o rei um o m ost em op top top osto deh l la r Fo n e w P p h ost os Pos P top nta et -lo De c ailu p u P e h s t d F o x sto P a r o t Po c cid e f P Ia In D A Percent of hospitals in highest decile that fell in rank by at least 2 deciles Patient Safety Indicators 50 45 40 35 30 25 20 15 10 5 0 s l r sth D RG ulce scue ody orax are ip fxha ge nge p fai DVT epsi ence tion e a n ort us re gn b ot h ed c op h orr dera res PE p s is c c era x C m ubit e t o rei um o m ost em op top top osto deh l la r Fo n e w P p h ost os Pos P top nta et -lo De c ailu p u P e h s t d F o x sto P a r o t Po c cid e f P Ia In D A Percent of hospitals in lowest decile that increased rank by at least 2 deciles In-hospital Mortality Indicators 50 45 40 35 30 25 20 15 10 5 0 y nt air BG ep CA iot om eme r A ac an AA Cr repl Hip I AM e F A e a e CH trok hag ct ur oni PTC S orr ip fra neum m P H he GI A CE Percent of hospitals in highest decile that fell in rank by at least 2 deciles In-hospital Mortality Indicators 50 45 40 35 30 25 20 15 10 5 0 y nt air BG ep CA iot om eme r A ac an AA Cr repl Hip I AM e F A e a e CH trok hag ct ur oni PTC S orr ip fra neum m P H he GI A CE Summary of Findings: Patient Safety Indicators Among the top 10% of hospitals (best, or lowest PSI rates): At least 40% of hospitals fell to 3rd decile or lower for: 1/3 of hospitals fell to 3rd decile or lower for: 1/4 of hospitals fell to 3rd decile or lower for: Postop hip fracture Postop physiologic and metabolic derangement Postop respiratory failure Wound dehiscence Anesthesia complications Death in low mortality DRGs Foreign body after procedure Iatrogenic pneumothorax Infection due to medical care Postop hemorrhage Postop PE and DVT Postop sepsis Summary of Findings: In-hospital Mortality Indicators Among top 10% of hospitals (best, or lowest mortality rates): 38% of hospitals fell to 3rd decile or lower for: 12-20% of hospitals fell to 3rd decile or lower for: Craniotomy GI hemorrhage Carotid endarterectomy Abdominal aortic aneurysm repair Among bottom 10% of hospitals (worst, or highest mortality rates): 24% of hospitals rose to 8th decile or higher 21% of hospitals rose to 8th decile or higher 14% of hospitals rose to 8th decile or higher 10% of hospitals rose to 8th decile or higher CABG Craniotomy PTCA Abdominal aortic aneurysm repair Limitations Findings may be unique to these specific indicators Findings may not hold for cutpoints other than deciles Convenience sample of hospitals – not nationally representative – But hospitals are drawn from a sampling frame that comprises 80% of U.S. discharges Conclusions When comparing Medicare and all-payer analyses – Found loose overlap of the top-ranking and poorest-ranking hospitals – Saw largest shifts for PSIs among the topranking hospitals Using Medicare-only data may carry a greater risk of incorrectly labeling a hospital as a top-ranking performer Pre-test ranking approaches