Development of MDS-Based Predication Model for COPD Severity in Nursing Home Residents.

Abstract

Assessing chronic obstructive pulmonary disease (COPD) severity is challenging in nursing home (NH) residents due to incomplete symptom assessments and exacerbation history. The objective of this study was to predict COPD severity in NH residents using the Minimum Data Set (MDS), a clinical assessment of functional capabilities and health needs. A cohort analysis of prospectively collected longitudinal data was conducted. Residents from geographically varied Medicare-certified NHs with age ≥60 years, COPD diagnosis, and ≥6 months NH residence at enrollment were included. Residents with severe cognitive impairment were excluded. Demographic characteristics, medical history, and MDS variables were extracted from medical records. The care provider-completed COPD Assessment Test (CAT) and COPD exacerbation history were used to categorize residents by Global Initiative for Chronic Lung Disease (GOLD) A to D groups. Multivariate multinomial logit models mapped the MDS to GOLD A to D groups with stepwise selection of variables. Nursing home residents (N = 175) were 64% women and had a mean age of 77.9 years. Among residents, GOLD B was most common (A = 13.1%; B = 44.0%; C = 5.7%; D = 37.1%). Any long-acting bronchodilator (LABD) use and any dyspnea were significant predictors of GOLD A to D groups. The predicted MDS-GOLD group (A = 6.9%; B = 52.6%; C = 4.6%; D = 36.0%) showed good model fit (correctly predicted = 60.6%). Nursing home residents may underuse group-recommended LABD treatment (no LABD: B = 53.2%; C = 80.0%; D = 40.0%). The MDS, completed routinely for US NH residents, could potentially be used to estimate COPD severity. Predicted COPD severity with additional validation could provide a map to evidence-based treatment guidelines and may help to individualize treatment pathways for NH residents.