Development and Validation of the N-DROP Score: A Penalized Logistic Regression Model for Predicting 30-Day in-Hospital Mortality in Patients with Nursing- and Healthcare-Associated Pneumonia.

BACKGROUND: Nursing- and healthcare-associated pneumonia (NHCAP) refers to pneumonia occurring in nursing-home residents or individuals with frequent healthcare exposure. Although the Age, Dehydration, Respiratory Status, Orientation Disturbance, Low Blood Pressure (A-DROP) score is widely used to assess community-acquired pneumonia severity in Japan, it does not account for frailty or functional status, potentially limiting its predictive accuracy for NHCAP. This study aimed to develop an improved mortality prediction model for NHCAP. METHODS: We developed and validated a penalized logistic regression model to predict 30-day in-hospital mortality in patients with NHCAP using a nationwide Japanese inpatient database. Patients hospitalised for NHCAP between April 2018 and March 2020 were enrolled. Candidate predictors were selected based on clinical relevance. Logistic least absolute shrinkage and selection operator (LASSO) regression was employed to identify influential variables, from which the most important predictors were used to construct a risk score model (N-DROP) based on the A-DROP framework. RESULTS: The 30-day in-hospital mortality in 116,185 eligible patients was 11.7%. LASSO regression identified five variables, and three key predictors-A-DROP score, impaired oral intake, and diminished activities of daily living (ADL)-were retained for N-DROP. N-DROP assigned one point per A-DROP score increment, one point for impaired oral intake, and one or two points based on ADL dependence. N-DROP demonstrated improved discriminatory performance compared with A-DROP [C-statistic: 0.746 (95% CI 0.734-0.757) vs. 0.708 (95% CI 0.695-0.720); P<0.001]. CONCLUSION: The novel N-DROP model improves prediction of in-hospital mortality compared with A-DROP in patients with NHCAP.