Trends in mortality and medical spending in patients hospitalized for community-acquired pneumonia: 1993-2005

Abstract

Background: Community-acquired pneumonia (CAP) is the most common infectious cause of death in the United States. To understand the effect of efforts to improve quality and efficiency of care in CAP, we examined the trends in mortality and costs among hospitalized CAP patients.

Methods: Using the National Inpatient Sample between 1993 and 2005, we studied 569,524 CAP admissions. The primary outcome was mortality at discharge. We used logistic regression to evaluate the mortality trend, adjusting for age, gender, and comorbidities. To account for the effect of early discharge practices, we also compared daily mortality rates and performed a Cox proportional hazards model. We used a generalized linear model to analyze trends in hospitalization costs, which were derived using cost-to-charge ratios.

Results: Over time, length of stay declined, while more patients were discharged to other facilities. The frequency of many comorbidities increased. Age/gender-adjusted mortality decreased from 8.9% to 4.1% (P < 0.001). In multivariable analysis, the mortality risk declined through 2005 (odds ratio, 0.50; 95% confidence interval, 0.48-0.53), compared with the reference year 1993. The daily mortality rates demonstrated that most of the mortality reduction occurred early during hospitalization. After adjusting for early discharge practices, the risk of mortality still declined through 2005 (hazard ratio, 0.74; 95% confidence interval, 0.70-0.78). Median hospitalization costs exhibited a moderate reduction over time, mostly because of reduced length of stay.

Conclusions: Mortality among patients hospitalized for CAP has declined. Lower in-hospital mortality at a reduced cost suggests that pneumonia is a case of improved productivity in health care.

Figure 1

Selected comorbidities of patients hospitalized for community-acquired pneumonia. Error bars indicate 95% confidence intervals. CHF denotes congestive heart failure; DM denotes diabetes mellitus.

Figure 2

Mortality trends adjusted for age and sex (A) and multivariable-adjusted odds ratio of death relative to the reference year 1993 (B). The multivariate model is adjusted for age, sex, and comorbid conditions. Error bars indicate 95% confidence intervals. LTC denotes long term care.

Figure 3

Daily mortality rates for 1993/1994 and 2004/2005 among patients still hospitalized on each hospital day. Error bars indicate 95% confidence intervals.

Figure 4

Trends in median hospitalization costs, total and per hospital day. The trend in predicted mean costs is based on a generalized linear model that specified cost as the dependent variable and age, sex, and comorbid conditions as covariates. One model also included length of stay (LOS).