2012 National Meeting

3073 — Two Latent-Trait Model for Outpatient and Inpatient “Quality of Care” in a Multi-Center Cluster Randomized Trial

Zhao X, Center for Health Equity Research and Promotion (CHERP), VA Pittsburgh Healthcare System; Roslyn SA, University of Pittsburgh School of Public Health; Ye F, University of Pittsburgh School of Education; Fine MJ, CHERP;

Quality of care is an abstract and multidimensional construct that cannot be measured directly. We implemented a two latent-trait model to quantify intervention effects on two dimensions of quality of care with regard to multiple binary performance metrics for outpatients and inpatients presenting to the ED. We illustrate use of this model in a 32-site cluster randomized clinical trial, the Emergency Department Community-Acquired Pneumonia (EDCAP) study, which compared the effectiveness of three site-level guideline implementation strategies on quality of ED care for outpatients and inpatients with CAP.

The correlation structure between multiple indictors is incorporated into two latent constructs that are simultaneously regressed on the interventions. The EDCAP outcomes were binary indicators of four recommended processes of outpatient care (oxygen assessment, first dose of antibiotics in ED, treatment with compliant antibiotics, and compliant antibiotic therapy upon discharge) and four recommended processes of inpatient care (oxygen assessment, blood cultures before antibiotics, antibiotics within 4 hours, and compliant antibiotic therapy in ED). These outcomes were collected at the patient-level and clustered at the provider and site levels. Bayesian estimation is implemented in WinBUGS.

High intensity intervention sites had significantly higher mean quality of outpatient and inpatient care than sites with low or moderate intensity interventions; low and moderate intensity sites were similar. Two processes of outpatient care (first dose of antibiotics in ED and compliant antibiotic therapy upon discharge) and two processes of inpatient care (blood cultures before antibiotics and compliant antibiotic therapy in ED) best discriminated between sites. A few “outlier” sites were identified using the estimated latent score. The estimated intervention effect is stronger for outpatient than for inpatient processes of care.

This two latent-trait model provides a summary measure of the “quality of care” for each site with regard to multiple indicators, allows assessment of overall intervention effects, quantifies relationships between outcomes, and identifies those outcomes that best discriminate between sites.

This two latent-trait model provides a comprehensive way to quantify “quality of care” with regard to multiple separate indicators for outpatients and inpatients.