Accounting for changes in coding practices may be important in analyzing trends based on administrative data. Several studies have demonstrated large reductions in mortality over time among pneumonia patients. However, a recent study suggested that this reduction may have been an artifact of case definition because more of the highest-risk patients were being coded under alternative principal diagnoses in recent years.

Using the National Inpatient Sample from 1993 to 2005, we selected hospitalizations with a principal diagnosis of pneumonia or a secondary diagnosis of pneumonia and a principal diagnosis of sepsis or respiratory failure. We performed logistic regression, estimating the likelihood of in-hospital mortality in each year, adjusting for age, sex, and comorbidities.

Over time, there was a substantial increase in the frequency of sepsis and respiratory failure as a principal diagnosis. Length of stay decreased in all 3 principal diagnosis groups. By 2005, the adjusted odds ratio (OR) of death among principal diagnosis pneumonia and respiratory failure hospitalizations decreased to 0.50 (95% confidence interval [CI], 0.49-0.51) and 0.62 (95% CI, 0.58-0.66), respectively, compared with 1993. With all 3 groups combined, there was still a substantial, albeit attenuated, reduction in the risk of mortality (OR₂₀₀₅ 0.70; 95% CI, 0.69-0.72).

Survival of patients with community-acquired pneumonia has improved greatly over time. However, interpretation of such findings based on administrative data must be made with caution and careful attention to case definition and coding trends.