PS4-37 THE "WORKING" DIAGNOSIS: CHANGES IN THE PNEUMONIA DIAGNOSIS AMONG HOSPITALIZED VETERANS

Wednesday, October 21, 2015
Grand Ballroom EH (Hyatt Regency St. Louis at the Arch)
Poster Board # PS4-37

Barbara Jones, MD1, Makoto Jones, MD2, Zhou Xi3, Brian Sauer4, Brett South5, Qing Zeng-Treitler5, Wendy Chapman5 and Matthew Samore, MD6, (1)VHA Salt Lake City, University of Utah, and Intermountain Health Care, Salt Lake City, UT, (2)VHA Salt Lake CIty and University of Utah, Salt Lake City, UT, (3)SLC VA Health System & University of Utah, Salt Lake City, UT, (4)University of Uah, Salt Lake City, UT, (5)University of Utah, Salt Lake City, UT, (6)University of Utah, Salt Lake City, UT. VA Salt LAke City Health Care System, SLC, UT, Salt Lake City, UT

Purpose: To examine the alignment between the diagnosis for pneumonia made in the emergency department (ED) versus the final diagnosis at hospital discharged among hospitalized Veterans

Method: Using the VA national data repository, we identified all ED visits across the VA system with chest imaging between Jan 01 2007-Dec 31 2012. Among those patients who were hospitalized, we extracted International Classification of Disease-9th Edition (ICD-9) codes that were assigned at completion of the ED visit ("ED-assigned ICD-9") and ICD-9 codes assigned at the time of discharge ("Discharge ICD-9").  We defined a diagnosis of pneumonia as a primary ICD-9 for pneumonia, or pneumonia as a secondary diagnosis with respiratory failure or sepsis as a primary diagnosis. We compared the alignment between ED-assigned and discharge ICD-9 codes using a 2x2 contingency table. Of those hospitalizations with a different principal diagnosis at discharge, we identified the most common alternative diagnoses. We then examined variation in the alignment between ED-assigned and discharge ICD-9s, defined as the proportion of true positives and true negatives, at the provider and facility levels.

Result: 1,675,984  hospitalizations were identified during the study period among 94 ED's and 54,136 providers.  Of the 84,631 hospitalizations with an initial ED-assigned ICD-9 code for pneumonia, 43% (36,576) were discharged with a different diagnosis, the five most common being COPD (18%) , CHF (6%), Urinary tract infection (4%), non-ST-elevation MI (1%), and acute renal failure (1%).  Of the 1,591,353 hospitalizations with an initial ED-assigned diagnosis not consistent with pneumonia, 64,324 (4%) were discharged with a diagnosis of pneumonia. The most common initial diagnoses for this group were shortness of breath (7%), COPD  (6%), CHF (3%), chest pain (3%), and atrial fibrillation (1%).  For those hospitalizations with an initial diagnosis for pneumonia, we found substantial variation in the proportion of alignment with the discharge ICD-9 (positive predictive value) at both the provider and the facility levels (Figure).

Conclusion: We found poor alignment between initial and final diagnosis for pneumonia among a large dataset of hospitalized Veterans, suggesting substantial uncertainty in the diagnosis of pneumonia. We also observed variation across providers and facilities. This variation provides an opportunity to identify outliers in decision-making.