| To Register SMDM Homepage | |
|
|
|
||||
Purpose: Few studies have investigated the volume-outcome relationship for non-surgical care, and none have examined the implications of seemingly justified volume-based referral policies in a local market. We estimate the effect of hospital volume on risk-adjusted 30-day mortality and in-hospital costs for patients with acute stroke in northeast Ohio (NEO).
Methods: This is a retrospective cohort of 13,096 patients admitted for acute stroke to one of 29 hospitals in NEO between 1991-97. Hospital volume was measured in 3 ways: 1) average annual number of stroke cases; 2) number of stroke cases in the 365 days preceding a given patient’s admission; and 3) average daily census, including non-stroke patients. Risk adjustment included stroke severity (c-statistic 0.865), hospital teaching status, and patient socioeconomic markers. In-hospital costs were estimated from patient charges and hospital-specific cost-to-charge ratios; we also examined discharges to extended care and readmissions within 30 days.
Results: Median (interquartile range) annual stroke volume and costs were 148 (89-206) and $5214 ($3638-8010, 1997 dollars), respectively; 30-day mortality was 15.0%. Increasing volume was associated with decreasing risk-adjusted 30-day mortality ( per 100-patient increase in volume, adjusted OR 0.897; 95% CI: 0.827,0.974); however, the two smallest volume hospitals had lower than predicted mortality, and the only plausible volume threshold would have excluded 26 of 29 hospitals. Interestingly, average daily census had at least as strong an effect size and pseudo–R2 as condition-specific volume measures. OLS regressions on log costs showed no significant effect of volume (p=0.338), nor were there significant volume effects on discharge disposition (p=0.509) or 30-day readmission rates (p=0.388).
Discussion: Despite a significant inverse relationship of volume to stroke mortality and a favorable overall cost-benefit relationship, the volume-outcome relationship is imprecise and there is no threshold that clearly discriminates hospital performance. Further, the equivalence of the alternative volume measures in predicting outcome imply quite different quality improvement strategies. In this regional analysis, volume is an inadequate proxy for quality and a poor guide to selective referral.
See more of Oral Concurrent Session B - Health Services Research
See more of The 26th Annual Meeting of the Society for Medical Decision Making (October 17-20, 2004)