Monday, November 3, 2003

This presentation is part of : Nursing Sensitive Outcomes

Multilevel Analysis of Nursing-Related, Hospital-Level Factors on Patient Outcome in Alberta Acute Hospitals

Carole A. Estabrooks, RN, PhD1, William Kai Midodzi, BSc, MSc, MSc1, Greta G. Cummings, RN, MEd2, Charles Humphrey, MA3, Kathryn L. Ricker, BKin, MSc4, and Phyllis Giovannetti, RN, ScD5. (1) Knowledge Utilization Studies in Practice Unit, Faculty of Nursing, University of Alberta, Edmonton, AB, Canada, (2) Administration, Cross Cancer Institute, Edmonton, AB, Canada, (3) Data Library, University of Alberta, Edmonton, AB, Canada, (4) Department of Educational Psychology, University of Alberta, Edmonton, AB, Canada, (5) Faculty of Nursing, University of Alberta, Edmonton, AB, Canada
Learning Objective #1: N/A
Learning Objective #2: N/A

Objective: To determine the relative effect of nursing derived hospital factors on patient mortality, after accounting for differences in patient and hospital characteristics.

Sample: Patient sample included all 23,873 adult inpatients in Alberta hospitals with a diagnosis of AMI, CHF, COPD, stroke or pneumonia, during the 1997-98 fiscal year.

The hospital sample was derived from the Alberta Nurse Survey (1998). All nurses in acute care hospitals were accessed through the professional association (n=6526, 52.8% response rate). Fifty hospitals with more than 5 respondents and greater than 20 beds were identified and matched to the patient sample.

Variables: Demographic, nurse outcome, inpatient mortality (or the vital status of the patient within 30 days of admission), organizational and practice environment variables were averaged at the hospital level to reflect nursing characteristics of each hospital.

Methods: To control for variation in the patient mix in hospitals, patient level logistic regression was used to develop risk-adjusted scores. These scores adjusted for patients that develop complications based on demographics, pre-existing conditions and co-morbidities. Multilevel logistic regression was used with patients nested in hospitals. This partitioned the variance of adjusted inpatient mortality into patient and hospital levels. The relative effects of nursing derived hospital characteristics were calculated unadjusted and adjusted for specific patient and hospital characteristics.

Findings: Significant variation in patient mortality exists across hospitals (variance =0.083, p<0.0001). Patient age, sex, pre-existing conditions and co-morbidities were strong determinants, (67% variance across hospitals). Hospital factors explained 11%, leaving 22% to be explained by nursing and unknown factors.

Conclusions: Results suggest significant variations in patient mortality largely attributable to differences in patient characteristics and pre-existing conditions. Organizational and nursing related factors (e.g., non-nursing tasks, quality of care) contribute to variation in patient mortality.

Implications: Hospital factors, nurse and support staff levels have a demonstrable impact on patient mortality.

Back to Nursing Sensitive Outcomes
Back to 37th Biennial Convention - Scientific Session
Sigma Theta Tau International