Table 3.

Primary Care Practice Characteristics Associated with Using More NPs/10,000 PC Patients in 1999, Controlling for NPs/10,000 PC Patients in 1996^*

Independent Variable	Parameter Estimate	Standard Error	P-Value	95% CI
Intecept	.80	.19	<.0001	(0.42, 1.18)
NPs/10,000 PC patients in 1996	.31	.07	<.0001	(0.17, 0.44)
Managed-care-oriented provider  education programs	.09	.04	.02	(0.02, 0.16)
NP training program	.26	.13	.05	(0.005, 0.52)
Specialist physicians/10,000 PC patients	.13	.08	.09	(−0.02, 0.29)

^*Note: We performed bivariate analyses to screen predictors for their potential use in a multivariate linear regression model at the level of alpha=.20. Twelve predictors remained after the screening process: geographic region, urban/rural location, state NP supply, state HMO penetration, academic affiliation, specialist physicians/10,000 PC patients, facility complexity, use of managed-care-oriented provider education, practice guidelines implementation, gate-keeping, primary care firm systems, and whether the practice provided NP training. We entered these 12 variables into multivariate model selection and cross-validation. The variables shown in Table 3 appeared most frequently during the 30 cross-validation runs. (Adjusted R-square=.21)