A Spatial Analysis of Regional Variation in In-Patient Care Using Examples from Five Selected

A spatial analysis of regional variation in in-patient care using examples from five selected procedures

Introduction:

Since the early 1970s practice variation has been discussed in many papers triggered by the article of Wennberg/Gittelsohn (1973). Different explanations have been presented including socioeconomic factors and health resource characteristics which have some explanatory power but a large share of variation remains unexplained. Communication between physicians and experts who cluster regionally (university hospitals) may contribute to the understanding of variations. Therefore, we use a regression model that takes into account the values of the dependent variable in neighboring districts to analyze if the diffusion of medical knowledge follows regional pattern.

Methodology:

Using a regression model that controls for spatial dependence (spatial regression model) we analyze the relation between variation in medical procedures and existence of a university hospital while controlling for health resource characteristics and socioeconomic factors. The dependent variable is the age-standardized number of operations per 100.000 at district level for appendectomy, tonsillectomy, hysterectomy and cholecystectomy, and the share of caesareans on all births based on the Operation and Procedure Code (OPS). In a second step, we estimate the spatial multiplier which allows us to estimate the effect of a change in the surgery rate in the considered district on neighboring districts. We can thus estimate the transregional impact of university hospitals (access to current medical knowledge). The study is based on the DRG (diagnosis related groups) statistics for the years 2006 to 2011.

Results:

We found a negative and significant relation between the existence of a university hospital and the number of surgeries for appendectomy, cholecystectomy, hysterectomy and tonsillectomy. The coefficients lie between -10.8 and -33.50. Additionally, we are able to show that health resource characteristics as the size of the hospital department, the supply of attending and assistant doctors, and competition between the hospitals are significant associated with the number of surgeries. The spatial lag coefficient is significant and positive for all procedures. The spatial multiplier for university hospital indicates that the existence of a university hospital in one district is negatively associated with the number of surgeries in neighboring districts.

Discussion:

We are able to show that the existence of a university hospital is negatively associated with the number of surgeries not just in their own district but also in neighboring districts what confirms the hypothesis of knowledge spillover triggered by communication.