Developing and implementing a model of small area variation of hospital use.

Abstract

The objectives of this dissertation were to (1) develop a methodology to assess if variation in rates are significantly different among small areas; (2) to develop a conceptual model for assessing why rates differ; and, (3) to apply the methodology to purposefully selected counties which are significantly different for hospital utilization rates; to assess the model and find significant predictors, and to interpret the findings in light of the restrictive sampling and the conceptual model. The probability model was based on a Beta-binomial distribution to estimate the likelihood of a statistically significant difference in rates, given a number of the sample characteristics (population, resources, rate, variation). Probability tables were produced to reflect the most common combinations of these characteristics. The binomial model was also used to sample the counties with significant differences in use rates. Five hypotheses were formulated and tested. The effect of need, demand, availability of resources, urban status, and accessibility of resources were estimated through regression analysis. Significant associations between need, demand, and accessibility variables and differences in use rates, were found. Specifically, the need variables explained a significant proportion of the difference in hospital use rates among the eighteen counties. The positive association supported the hypothesis that the greater the population's need for care, the higher will be the use rate in the county. The demand variables failed to achieve statistical significance. As for the availability of resources, there was support for the hypothesis that the more the in-county beds the higher the likelihood of using them. Accessibility of care and the urban or rural status of a county were not significant variables in the explication of inter-county differences in hospital discharge rates. The implications of these findings to health policy were discussed.