Quality improvement implementation in the nursing home

Health Services Research, Feb, 2003 by Dan R. Berlowitz, Gary J. Young, Elaine C. Hickey, Debra Saliba, Brian S. Mittman, Elaine Czarnowski, Barbara Simon, Jennifer J. Anderson, Arlene S. Ash, Lisa V. Rubenstein, Mark A. Moskowitz

Risk-Adjusted Pressure Ulcer Rates. Our methods for calculating the outcome measure of pressure ulcer development has been described previously (Berlowitz et al. 1996; Berlowitz, Young, Brandeis et al. 2001). Briefly, we determined rates of pressure ulcer development using the Patient Assessment File (PAF), a VA administrative database originally developed for case-mix-based reimbursements in nursing homes according to Resource Utilization Groups version 2 (RUGs II) (Schneider et al. 1988). Information in the PAF is collected on all VA long-term care residents at the time of admission and semi-annually on April 1 and October 1. Thus, long-staying residents will have six months between assessments. By following a resident over time in the database, it is possible to identify changes in health status, such as the development of a pressure ulcer. We defined pressure ulcer development as present when a resident without an ulcer on an "index" assessment had a stage 2 or larger ulcer on a subsequent assessment. Res idents with stage 1 ulcers were considered pressure ulcer-free. Risk-adjustment was performed using a previously derived and validated model that considers 11 resident characteristics. Data used in this study were from three, six-month periods that overlapped our chart abstractions, beginning in October 1997 and ending in April 1999.

Analyses

Descriptive statistics were calculated for each study variable including QI implementation, organizational culture, staff-reported guideline adoption, employee satisfaction, and adherence to guideline recommendations. Analysis of variance was used to test for differences among nursing homes in their extent of QI implementation ([H.sub.1]). Linear regression models were used to test associations among the 35 nursing homes between their scores on these variables. Specifically, we examined each nursing home's organizational culture score in relation to its QI implementation score ([H.sub.2]) and each nursing home's QI implementation score in relation to each of employee satisfaction ([H.sub.3]), staff reported guideline adoption ([H.sub.4]), and adherence with guideline recommendations ([H.sub.4]). Analyses relating QI implementation to pressure ulcer development ([H.sub.5]) were performed at the patient-level using a random effects model. Each resident was assigned the QI implementation score of his or her nurs ing home as the independent variable. The dependent variable (0-E) was the observed rate of pressure ulcer development (1 if a pressure ulcer developed or 0 if not) minus the expected rate (as predicted by the logistic model) for each individual patient.

We further examined these associations by including a limited number of nursing home characteristics as independent variables in the regression models. These included region of the country (East, South, Midwest, and West), nursing home size defined by number of beds, urban versus rural location based on presence in a Metropolitan Standard Statistical Area, and teaching status according to membership on the Council of Teaching Hospitals. As all rural facilities were nonteaching, these last two variables were combined as urban-teaching, urban-nonteaching, and rural. Due to the relatively small number of facilities, each characteristic was used as an independent variable in separate regression models testing each of [H.sub.2], [H.sub.3], and [H.sub.4].