Predictors of Adoption of at least a “Basic” Electronic Medical Record System.^*

*In this analysis, we used adoption of a “basic” electronic medical record system as the outcome. This definition was taken from Jha and DesRoches, and required a system to include electronic management of demographic information, computerized provider order entry, and lab and imaging results [3], [4]. We began our analysis by conducing bivariate analyses, to determine which of a series of candidate predictors appeared to have a relationship with the outcome variable. We used the following candidate predictors: patient age, gender, race/ethnicity, and source of payment, and, at the hospital level, region, metropolitan vs. non-metropolitan (i.e. urban vs. rural), ownership, and teaching status. Candidate predictors were eliminated from further consideration if bivariate chi-squared testing resulted in a p-value≥0.20. Remaining candidate predictors were fitted to a multivariate logistic regression model, constructed via stepwise backward elimination until all remaining independent covariates had p<0.05 in their type 3 analyses of effects. Accordingly, the following predictors were eliminated from the model: patient-level variables, age, gender, race/ethnicity, and source of payment (insurance type); and hospital-level variables, ownership, teaching status. Only region and urban/rural status were significant predictors of adoption of at least a “basic” system.