Garvin JH (Center for Health Equity Research and Promotion, NewCourtland Center for Transitions and Health), Benson P
(Center for Health Equity Research and Promotion), Field S
(Center for Health Equity Research and Promotion), Roche D
(Center for Health Equity Research and Promotion), Groeneveld P
(Center for Health Equity Research and Promotion, University of Pennsylvania), Leecaster M
(IDEAS Center, VA Salt Lake City Health Care System and University of Utah), Weiner M
(Center for Health Equity Research and Promotion, University of Pennsylvania)
Comorbidities coded with the International Classification of Diseases 9th Revision Clinical Modification (ICD-9-CM) are commonly used to risk adjust non-random patient groups. This presentation will compare the frequencies of comorbidities associated with the Elixhauser Comorbidity Measure (ECM) found in a non-validated and a validated administrative dataset.
The data were compiled from patients who were discharged from one VA medical center in 2003 with a principal diagnosis of Congestive Heart Failure (CHF). A coding expert, blinded to the original codes assigned, undertook manual chart review of the 182 inpatient records to determine which comorbid conditions from the ECM should be coded. The findings of the coding expert were validated by the Chief of Coding and Compliance of the medical center. The resulting validated data were compared to the original unvalidated codes assigned at the time of patient discharge. The frequencies of ECM comorbidities in both datasets were determined. The Kappa statistic and the bootstrap confidence intervals were calculated as were descriptive and summary statistics.
The Kappa statistic was 0.375 with a 95% confidence interval of 0.333 to 0.417. There were a total of 374 comorbid conditions found in the original data set with an average of 2 comorbidities per patient compared to 1926 comorbidities in the validated dataset with an average of 6.5 per patient. Common diagnoses with a physiological relationship to CHF, such as hypertension, pulmonary hypertension, and valve disease were frequently missed as were other important diagnoses like depression and diabetes. Renal failure was not included in the calculations due to a change in coding rules and definitions between 2003 and 2008.
Comorbidites in the ECM were significantly underrepresented in the original patient data. In short, the lack of accurate code assignment for comorbidities not only affects payment, but may also affect the accuracy of risk adjustment processes.
The accuracy of risk adjustment processes may be compromised if comorbidity codes are not correctly assigned. It is important to highlight the need to evaluate code assignment of this type because risk adjustment is frequently used for health services research as well as for resource allocation and performance measurement within the VA.