A comparison of the Charlson comorbidity index derived from medical records and claims data from patients undergoing lung cancer surgery in Korea: a population-based investigation
1 Department of Nursing, College of Medicine, Chosun University, Gwangju, Korea
2 Department of Preventive Medicine, Korea University, Seoul, Korea
3 Department of Preventive Medicine, University of Ulsan College of Medicine, Seoul, Korea
4 Department of Preventive Medicine, Konkuk University School of Medicine, Seoul, Korea
5 National Cancer Control Research Institute and Hospital, National Cancer Center, Seoul, Korea
6 Graduate School of Korea University, Department of Public Health, Seoul, Korea
BMC Health Services Research 2010, 10:236 doi:10.1186/1472-6963-10-236Published: 13 August 2010
Calculating the Charlson comorbidity index (CCI) from medical records is a time-consuming and expensive process. The objectives of this study are to 1) measure agreement between medical record and claims data for CCI in lung cancer patients and 2) predict health outcomes of lung cancer patients based on CCIs from both data sources.
We studied 392 patients who underwent surgery for pathologic stages I-III of lung cancer. The kappa value was used to measure the agreement between the 17 comorbidities of the CCI prevalence obtained from medical records and claims data. Multiple linear regression analyses were used to evaluate the relationships between CCI and length of stay and reimbursement cost.
Out of 17 comorbidities identified in the Charlson comorbidity index, ten had a higher prevalence, four had a lower prevalence and three had a similar prevalence in claims data to those of medical records. The kappa values calculated from the two databases ranged from 0.093 to 0.473 for nine comorbidities. In predicting length of stay and reimbursement cost after surgical resection for lung cancer patients, the CCI scores derived from both the medical records and claims data were not statistically significant.
Poor agreement between medical record data and claims data may result from different motivations for collecting data. Further studies are needed to determine an appropriate method for predicting health outcomes based on these data sources.