Community-acquired pneumonia: can it be defined with claims data?

Area of Science:

• Health Services Research
• Clinical Informatics
• Epidemiology

Background:

• Administrative data, such as International Classification of Diseases, 9th revision, Clinical Modification (ICD9-CM) codes, are valuable for health research.
• However, these codes lack specificity for community-acquired pneumonia (CAP), a critical clinical differentiator.
• This limitation hinders accurate analysis of CAP using administrative datasets.

Purpose of the Study:

• To evaluate the accuracy of administrative data-based methods for identifying community-acquired pneumonia (CAP) patients.
• To compare the agreement of different administrative data classification systems with clinician chart review.
• To assess the impact of different identification methods on length of stay (LOS) and mortality in CAP populations.

Main Methods:

• Classified 212 patients with pneumonia diagnosis codes into CAP and non-CAP groups using three administrative data methods: Diagnosis Related Groups (DRGs) alone, principal diagnosis alone, and a complex algorithm.
• Compared the administrative classifications against a gold standard of clinician chart review.
• Analyzed length of stay (LOS) and mortality rates for CAP populations identified by each method and by clinical review.

Main Results:

• Agreement between administrative data methods and clinician chart review ranged from 80% to 86%, with DRG classification performing the poorest.
• Populations identified using administrative data showed similar mortality rates compared to the clinically defined group.
• Mean LOS was shorter in populations identified by administrative data (7.91–9.70 days) compared to the clinically defined population (10.85 days).

Conclusions:

• Administrative data can be utilized to identify CAP patients, but accuracy varies by method.
• Classification using DRGs is less reliable for identifying CAP.
• Studies relying on administrative claims data to define CAP populations may underestimate the actual length of stay (LOS).