Outcome of nursing home-acquired pneumonia: Derivation and application of a practical model to predict 30 day mortality

OBJECTIVES: To derive a prediction model of 30 day mortality for nursing ho me-acquired pneumonia (NHAP) based on factors that can be readily identifie d by nursing home staff at the time of diagnosis and to apply the model to management issues related to NHAP including clarifying the importance of pr epneumonia functional status as a predictor of outcome of NHAP. DESIGN: This was a retrospective chart review of 37& episodes of NHAP treat ed in the nursing home or hospital during two periods: November 1997 to Apr il 1998 and November 1998 to April 1999. SETTING: Eleven nursing homes in the greater Buffalo, NY region. PARTICIPANTS: Nursing home residents with radiographically proven pneumonia who had at least one of the following signs/symptoms: cough, fever, purule nt sputum, respiratory rate greater than or equal to 25 breaths/minute, loc alized auscultatory findings, or pleuritic pain. MEASUREMENTS: Status (alive or dead) of each resident at 30 days (30 day mo rtality) after diagnosis of NHAP was the dependent variable. Factors predic ting 30 day mortality were identified by logistic regression analysis. A sc oring system was developed based on the results of the logistic model. Each episode of NHAP in the derivation cohort was scored using the model and th e cohort was stratified by the model score into six categories or risk for mortality (0-5). The predictability of the model in the derivation cohort w as measured using receiver operator characteristics curve analysis. RESULTS: Of 378 episodes of NHAP, 74% were treated initially in the nursing home and 26% were hospitalized initially for treatment. The overall 30 day mortality was 21.4%; however, the mortality rate was significantly higher for those treated initially in the hospital (29.6% vs 16.6%; P = .012). Log istic regression analysis identified four predictors of 30 day mortality: ( 1) respiratory rate > 30 breaths/ minute (2 points), (2) pulse >125 beats/m inute (1 point), (3) altered mental status (1 point), and (4) a history of dementia (1 point). Applying the scoring system to each episode in the deri vation cohort demonstrated increasing mortality with increasing score. The c statistic for the model in the derivation cohort was .74. Based on the se verity of NHAT, model episodes treated initially in the hospital were more acutely ill than those who were treated initially in the nursing home, and episodes treated with a parenteral antibiotic in the nursing home were more acutely ill than those who were treated with an oral agent. Functional sta tus was not a predictor of 30 day mortality although there was a trend of h igher mortality in the most dependent group (P = .065). The severity of NHA P model was able to define low and high risk mortality groups within a func tional status category. CONCLUSIONS: A severity of NHAP model was derived from a large cohort of ep isodes in multiple facilities. The model had reasonable discriminatory powe r in the derivation cohort. The model may aid clinicians in making treatmen t decisions in the nursing home setting and in making hospitalization decis ions. Although prepneumonia functional status provides a reasonable estimat e of NHAP severity and prognosis, the severity of NHAP model permitted furt her refinement of these estimates. The severity of NHAP model requires vali dation before it can be recommended for general use.