| Show to measure the rurality of the small area mean household included population income percent below poverty percent of high school graduates ethnicity and percent of the population who were rural residents Mean household income was standardized ie this variable was transformed to havemean ofand standard error of0 The Census defined rural residents as those who live outside urbanized areas in places with less than 2500 residents Gesler Ricketts 1992 The 1996 Utah Health Status Survey Bureau of Surveillance and Analysis 1997 provided the percentage of population with insurance The 1994 bed supply was obtained from the Department of Health Bureau of Primary Care Rural and Ethnic Health for each zip code Bed supply was calculated for each small area by counting each small areaown bed supply plus the bed supply of adjacent small areas except whensmall area was itself an entire county orgroup of counties Provider data family officebased as such are practitioners practitioners general available from licensure data However the address at whichphysician receives licensure forms can behome address rather thanwork address and spot checks of these data determined that they did not match actual provider supply in the small areas thus this variable was eliminated from consideration Technical Details Analysis was done by least-squares regression weighted by the inverse standard error of the ACSH rate This effectively adjusts for the variation in population size among small areas Diagnostic statistics were used to examine outlier observations and collinearity of variables Bamett Lewis 1978 Belsley Kuh Welsch 1980 Kleinbaum Kupper Muller 1988 This was an ecological analysis Rothman Greenland 1998 in that all of the explanatory variables were at the small area level not the individual level Thus they operated as contextual factors In fact race ethnicity data were not effectively available at the individual level for hospitalizations in Utah since these data were not mandated to be collected and many hospitals do not obtain them In addition local area race ethnicity data were not available for two small areas which were created after 1990 These two areas were dropped from analysis Racial ethnic compositions of small areas were highly correlated with SES Regression models became unstable when both were included Therefore SES was selected for inclusion in the regression models To examine Wennbergclaim one model examined the remaining effects of other variables adjusting for bed supply Since others have argued that local socioeconomic status is part of the chain of logic leading to hospitalization second model examined the effects of other variables adjusting for SES The third model was fited by computerized forward backward stepwise regression Final models were fitted using5 signiiicance level as an inclusion criterion Results When adjusted for bed supply variables in addition to bed supply were found to be statistically significant predictors of ACSH for each disease process see Model Table Thus Wennbergfinding was not bome out in the smaller areas examined in Utah After controliing for bed supply asthma hospitalization rates were predicted by proportionately fewer rural residents and fewer high school graduates pneumonia hospitalizations by lower percent of high school graduates and diabetes hospitalizations by proportionately fewer rural residents and fewer high school graduates When socioeconomic status was adjusted similar results were obtained except that bed supply was no longerpneumoniapredictor significant hospitalizations adjusting for local SES see Model Table When using computerized stepwise selection to fit the model asthma hospitalizations insmall area were significantly predicted by lower SES especially lower high school graduation rates fewer rural residents and larger bed supply see Model Table Pneumonia hospitalizations were significantly predicted by lower SES especially lower high school graduation rates and lower local bed supply The effect of hospital bed supply on pneumonia hospitalization rate was not in the direction 18 Digital image 2005 Marriott Library University of Utah Al rights reserved |
