Efficiency measurement of 89 public provincial hospitals in Thailand: Parametric and non-parametric estimation methods

This study examined whether or not the behavior of Thailand's provincial hospitals were consistent with the cost minimization behavioral objective. Cost minimization implies a non-existence of cost inefficiency. The study was based on frontier methodology using both the parametric and non-parametric estimation methods to examine sensitivity of results to estimation choices. The resulting analyses also revealed the sources of cost inefficiency, nature of technological change, and the growth of total factor productivity in the provincial hospital sector.; Strong evidences from all estimation methods suggested that significant cost inefficiency exists invalidating the hypothesized cost minimization objective. Cost inefficiency substantially raised hospital costs above the minimum costs by approximately 10--30 percent depending on the estimation methods. Two-thirds of the cost inefficiency was attributable to technical inefficiency, and the allocative component contributed to the remaining variations. The total potential cost saving from eliminating the technical and allocative inefficiency was estimated to be 1--3 billion baht or approximately US{dollar}40--120 million per year. Increasing competition and devising compensation policies that will enhance efficiency and minimize selectivity bias and moral hazard can be crucial keys to realize potential cost saving.; There was a negligible productivity gain over the 6-year period. Changes in technology dominated changes in inefficiency in determining changes in productivity. Evidence supported the hypothesis that rising health care cost in Thailand was primarily a result of cost-increasing technical change rather than increasing inefficiency.; Both the parametric and non-parametric estimation methods provided strong evidence supporting the notion that hospitals regularly maintained stand-by capacity to meet the desired probability of turning away patients. The common practice in the past literature of implicitly assumed separable cost function was strongly rejected by the data. Both findings clearly suggested that the past practice of ignoring the influence of demand uncertainty on hospital costs, specifying "behavioral cost function", and using highly aggregated output were seriously flawed. The resulting specification error produced a serious bias of the estimates of the cost structure and cost efficiency. Efficiency estimates were found to be sensitive to estimation choices. The conventional approach of regression analysis remains a preferred method.