Towards panel data specifications of efficiency measures for English acute hospitals

This paper reports work undertaken for the Department of Health to explore different approaches of measuring hospital efficiency. The emphasis throughout is on developing adjusted cost-efficiency measures in line with NHS Trusts performance objectives. Previous work described the derivation of three residual-based cost indices (CCI, 2CCI and 3CCI), each with increasing adjustment in terms of case mix, factor prices and environmental factors for a single year’s data (1995/6) (Söderlund & van der Merwe, 1999). This study explores further options based on the previous work by: (1) supplementing hospital level with specialty level data; (2) studying a 4-year panel from 1994/5 to 1997/8; (3) estimating models with non-symmetric error terms and including Trust-specific effects when measuring inefficiency. Although the paper argues that panel data models may have certain advantages over cross-sectional ones, the results suggest that data pooling across years provide robust parameter estimates. Longitudinal fixed effect models may however be useful to construct efficiency indices while stochastic frontier models have the advantage of taking account of random noise. Specialty level models proved inferior to whole hospital estimations. The paper argues that the degree of variation between hospitals in terms of efficiency is not that great and scope for efficiency enhancement is primarily attainable by optimising capacity and activity levels in the long run. Increased activity levels may however have adverse consequences such as increased hospital infection rates, poorer quality of care and a lack of capacity to deal with emergency demand. The paper argues that the Department of Health might consider a shift from the adjusted cost index approach used in this normative benchmarking framework to the more conventional efficiency analysis approach using a total cost function, and more flexible functional forms, allowing for a more defensible interpretation of the residuals as inefficiency.