To examine the association between having a nurse practitioner/physician assistant (NP/PA) on staff, other nursing home (NH) characteristics, and the rate of potentially preventable/avoidable hospitalizations of long-stay residents, as defined using a list of ambulatory care-sensitive (ACS) diagnoses. Cross-sectional prospective study using Minimum Data Set (MDS) assessments, Centers for Medicare and Medicaid Services inpatient claims and eligibility records, On-line Survey Certification Automated Records, (OSCAR) and Area Resource File (ARP). Freestanding urban NHs in Maine, Kansas, New York, and South Dakota. Residents of 663 facilities with a quarterly or annual MDS assessment in the 2nd quarter of 1997, who had a prior MDS assessment at least 160 days before, and who were not health maintenance organization members throughout 1997 (N=54,631). A 180-day multinomial outcome was defined as having any hospitalization with primary ACS diagnosis, otherwise having been hospitalized, otherwise died, and otherwise remained in the facility. Multilevel models show that facilities with NP/PAs were associated with lower hospitalization rates for ACS conditions (adjusted odds ratio (AOR)=0.83), but not with other hospitalizations. Facilities with more physicians were associated with higher ACS hospitalizations (ACS, AOR=1.14, and non-ACS, AOR=1.10). Facilities providing intravenous therapy, and those that operate a nurses' aide training program were associated with fewer hospitalizations of both types. Employment of NP/PAs in NHs, the provision of intravenous therapy, and the operation of certified nurse assistant training programs appear to reduce ACS hospitalizations, and may be feasible cost-saving policy interventions.