• Health Policy and Management
• /
• v.26 no.1
• /
• pp.71-78
• /
• 2016

The value of using health insurance claim database is continuously rising in healthcare research. In studies where comorbidities act as a confounder, comorbidity adjustment holds importance. Yet researchers are faced with a myriad of options without sufficient information on how to appropriately adjust comorbidity. The purpose of this study is to assist in selecting an appropriate index, look back period, and data range for comorbidity adjustment. No consensus has been formed regarding the appropriate index, look back period and data range in comorbidity adjustment. This study recommends the Charlson comorbidity index be selected when predicting the outcome such as mortality, and the Elixhauser's comorbidity measures be selected when analyzing the relations between various comorbidities and outcomes. A longer look back period and inclusion of all diagnoses of both inpatient and outpatient data led to increased prevalence of comorbidities, but contributed little to model performance. Limited data range, such as the inclusion of primary diagnoses only, may complement limitations of the health insurance claim database, but could miss important comorbidities. This study suggests that all diagnoses of both inpatients and outpatients data, excluding rule-out diagnosis, be observed for at least 1 year look back period prior to the index date. The comorbidity index, look back period, and data range must be considered for comorbidity adjustment. To provide better guidance to researchers, follow-up studies should be conducted using the three factors based on specific diseases and surgeries.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.13 no.6
• /
• pp.2672-2679
• /
• 2012

The study was done to provide basic data of medical quality evaluation after developing the comorbidity disease mortality measurement modeled on the severity-adjustment method of AMI. This study analyzed 699,701 cases of Hospital Discharge Injury Data of 2005 and 2008, provided by the Korea Centers for Disease Control and Prevention. We used logistic regression to compare the risk-adjustment model of the Charlson Comorbidity Index with the predictability and compatibility of our severity score model that is newly developed for calibration. The models severity method included age, sex, hospitalization path, PCI presence, CABG, and 12 variables of the comorbidity disease. Predictability of the newly developed severity models, which has statistical C level of 0.796(95%CI=0.771-0.821) is higher than Charlson Comorbidity Index. This proves that there are differences of mortality, prevalence rate by method of mortality model calibration. In the future, this study outcome should be utilized more to achieve an improvement of medical quality evaluation, and also models will be developed that are considered for clinical significance and statistical compatibility.

• Journal of Digital Convergence
• /
• v.17 no.2
• /
• pp.217-230
• /
• 2019

This study was conducted to develop a customized severity-adjustment method and to evaluate their validity for acute myocardial infarction(AMI) patients to complement the limitations of the existing severity-adjustment method for comorbidities. For this purpose, the subjects of KCD-7 code I20.0 ~ I20.9, which is the main diagnosis of acute myocardial infarction were extracted using the Korean National Hospital Discharge In-depth Injury survey data from 2006 to 2015. Three tools were used for severity-adjustment method of comorbidities : CCI (charlson comorbidity index), ECI (Elixhauser comorbidity index) and the newly proposed CCS (Clinical Classification Software). The results showed that CCS was the best tool for the severity correction, and that support vector machine model was the most predictable. Therefore, we propose the use of the customized method of severity correction and machine learning techniques from this study for the future research on severity adjustment such as assessment of results of medical service.

• Health Policy and Management
• /
• v.21 no.4
• /
• pp.541-564
• /
• 2011

This study aimed to investigate health outcome of acute myocardial infarction (AMI) patients such as mortality and length of stay in hospital and to identify factors associated with the health outcome according to the comorbidity index. Nation-wide representative samples of 3,748 adult inpatients aged between 20-85 years with acute myocardial infarction were derived from the Korea National Hospital Discharge Injury Survey, 2005-2008. Comorbidity index was measured using the Charlson Comorbidity Index (CCI). The data were analyzed using t-test, ANOVA, multiple regression, logistic regression analysis in order to investigate the effect of comorbidity on health outcome. According to the study results, the factors associated with length of hospital stay of acute myocardial infarction patients were gender, insurance type, residential area scale, admission route, PCI perform, CABG perform, and CCI. The factors associated with mortality of acute myocardial infarction patients were age, admission route, PCI perform, and CCI. CCI with a higher length of hospital stay and mortality also increased significantly. This study demonstrated comorbidity risk adjustment for health outcome and presented important data for health care policy. In the future study, more detailed and adequate comorbidity measurement tool should be developed, so patients' severity can be adjusted accurately.

• Journal of Preventive Medicine and Public Health
• /
• v.42 no.4
• /
• pp.267-273
• /
• 2009

Objectives : To compare the performance of three comorbidity measurements (Charlson comorbidity index, Elixhauser s comorbidity and comorbidity selection) with the effect of different comorbidity lookback periods when predicting in-hospital mortality for patients who underwent percutaneous coronary intervention. Methods : This was a retrospective study on patients aged 40 years and older who underwent percutaneous coronary intervention. To distinguish comorbidity from complications, the records of diagnosis were drawn from the National Health Insurance Database excluding diagnosis that admitted to the hospital. C-statistic values were used as measures for in comparing the predictability of comorbidity measures with lookback period, and a bootstrapping procedure with 1,000 replications was done to determine approximate 95% confidence interval. Results : Of the 61,815 patients included in this study, the mean age was 63.3 years (standard deviation: ${\pm}$10.2) and 64.8% of the population was male. Among them, 1,598 2.6%) had died in hospital. While the predictive ability of the Elixhauser's comorbidity and comorbidity selection was better than that of the Charlson comorbidity index, there was no significant difference among the three comorbidity measurements. Although the prevalence of comorbidity increased in 3 years of lookback periods, there was no significant improvement compared to 1 year of a lookback period. Conclusions : In a health outcome study for patients who underwent percutaneous coronary intervention using National Health Insurance Database, the Charlson comorbidity index was easy to apply without significant difference in predictability compared to the other methods. The one year of observation period was adequate to adjust the comorbidity. Further work to select adequate comorbidity measurements and lookback periods on other diseases and procedures are needed.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.11
• /
• pp.126-136
• /
• 2018

The purpose of this study was to develop a severity-adjustment model for predicting mortality in acute stroke patients using machine learning. Using the Korean National Hospital Discharge In-depth Injury Survey from 2006 to 2015, the study population with disease code I60-I63 (KCD 7) were extracted for further analysis. Three tools were used for the severity-adjustment of comorbidity: the Charlson Comorbidity Index (CCI), the Elixhauser comorbidity index (ECI), and the Clinical Classification Software (CCS). The severity-adjustment models for mortality prediction in patients with acute stroke were developed using logistic regression, decision tree, neural network, and support vector machine methods. The most common comorbid disease in stroke patients were hypertension, uncomplicated (43.8%) in the ECI, and essential hypertension (43.9%) in the CCS. Among the CCI, ECI, and CCS, CCS had the highest AUC value. CCS was confirmed as the best severity correction tool. In addition, the AUC values for variables of CCS including main diagnosis, gender, age, hospitalization route, and existence of surgery were 0.808 for the logistic regression analysis, 0.785 for the decision tree, 0.809 for the neural network and 0.830 for the support vector machine. Therefore, the best predictive power was achieved by the support vector machine technique. The results of this study can be used in the establishment of health policy in the future.

• Journal of Periodontal and Implant Science
• /
• v.48 no.2
• /
• pp.92-102
• /
• 2018

Purpose: This study investigated the validity of the Charlson comorbidity index (CCI) as a predictor of periodontal disease (PD) over a 12-year period. Methods: Nationwide representative samples of 149,785 adults aged ${\geq}60$ years with PD (International Classification of Disease, 10th revision [ICD-10], K052-K056) were derived from the National Health Insurance Service-Elderly Cohort during 2002-2013. The degree of comorbidity was measured using the CCI (grade 0-6), including 17 diseases weighted on the basis of their association with mortality, and data were analyzed using multivariate Cox proportional-hazards regression in order to investigate the associations of comorbid diseases (CDs) with PD. Results: The multivariate Cox regression analysis with adjustment for sociodemographic factors (sex, age, household income, insurance status, residence area, and health status) and CDs (acute myocardial infarction, congestive heart failure, peripheral vascular disease, cerebral vascular accident, dementia, pulmonary disease, connective tissue disorders, peptic ulcer, liver disease, diabetes, diabetes complications, paraplegia, renal disease, cancer, metastatic cancer, severe liver disease, and human immunodeficiency virus [HIV]) showed that the CCI in elderly comorbid participants was significantly and positively correlated with the presence of PD (grade 1: hazard ratio [HR], 1.11; P<0.001; grade ${\geq}2$: HR, 1.12, P<0.001). Conclusions: We demonstrated that a higher CCI was a significant predictor of greater risk for PD in the South Korean elderly population.

• Journal of Preventive Medicine and Public Health
• /
• v.39 no.1
• /
• pp.13-20
• /
• 2006

Objectives: To propose a risk-adjustment model with using insurance claims data and to analyze whether or not the outcomes of non-emergent and isolated coronary artery bypass graft surgery (CABG) differed between the low- and high-volume hospitals for the patients who are at different levels of surgical risk. Methods: This is a cross-sectional study that used the 2002 data of the national health insurance claims. The study data set included the patient level data as well as all the ICD-10 diagnosis and procedure codes that were recorded in the claims. The patient's biological, admission and comorbidity information were used in the risk-adjustment model. The risk factors were adjusted with the logistic regression model. The subjects were classified into five groups based on the predicted surgical risk: minimal (<0.5%), low (0.5% to 2%), moderate (2% to 5%), high (5% to 20%), and severe (=20%). The differences between the low- and high-volume hospitals were assessed in each of the five risk groups. Results: The final risk-adjustment model consisted of ten risk factors and these factors were found to have statistically significant effects on patient mortality. The C-statistic (0.83) and Hosmer-Lemeshow test ($x^2=6.92$, p=0.55) showed that the model's performance was good. A total of 30 low-volume hospitals (971 patients) and 4 high-volume hospitals (1,087 patients) were identified. Significant differences for the in-hospital mortality were found between the low- and high-volume hospitals for the high (21.6% vs. 7.2%, p=0.00) and severe (44.4% vs. 11.8%, p=0.00) risk patient groups. Conclusions: Good model performance showed that insurance claims data can be used for comparing hospital mortality after adjusting for the patients' risk. Negative correlation was existed between surgery volume and in-hospital mortality. However, only patients in high and severe risk groups had such a relationship.

• Journal of Preventive Medicine and Public Health
• /
• v.42 no.2
• /
• pp.117-122
• /
• 2009

Objectives : We tried to evaluate the agreement of the Charlson comorbidity index values(CCI) obtained from different sources(medical records and National Health Insurance claims data) for gastric cancer patients. We also attempted to assess the prognostic value of these data for predicting 1-year mortality and length of the hospital stay(length of stay). Methods : Medical records of 284 gastric cancer patients were reviewed, and their National Health Insurance claims data and death certificates were also investigated. To evaluate agreement, the kappa coefficient was tested. Multiple logistic regression analysis and multiple linear regression analysis were performed to evaluate and compare the prognostic power for predicting 1 year mortality and length of stay. Results : The CCI values for each comorbid condition obtained from 2 different data sources appeared to poorly agree(kappa: 0.00-0.59). It was appeared that the CCI values based on both sources were not valid prognostic indicators of 1-year mortality. Only medical record-based CCI was a valid prognostic indicator of length of stay, even after adjustment of covariables($\beta$ = 0.112, 95% CI = [0.017-1.267]). Conclusions : There was a discrepancy between the data sources with regard to the value of CCI both for the prognostic power and its direction. Therefore, assuming that medical records are the gold standard for the source for CCI measurement, claims data is not an appropriate source for determining the CCI, at least for gastric cancer.

• Health Policy and Management
• /
• v.26 no.4
• /
• pp.359-372
• /
• 2016

Background: The purpose of this study was to develop risk-adjustment models for acute stroke mortality that were based on data from Health Insurance Review and Assessment Service (HIRA) dataset and to evaluate the validity of these models for comparing hospital performance. Methods: We identified prognostic factors of acute stroke mortality through literature review. On the basis of the avaliable data, the following factors was included in risk adjustment models: age, sex, stroke subtype, stroke severity, and comorbid conditions. Survey data in 2014 was used for development and 2012 dataset was analysed for validation. Prediction models of acute stroke mortality by stroke type were developed using logistic regression. Model performance was evaluated using C-statistics, $R^2$ values, and Hosmer-Lemeshow goodness-of-fit statistics. Results: We excluded some of the clinical factors such as mental status, vital sign, and lab finding from risk adjustment model because there is no avaliable data. The ischemic stroke model with age, sex, and stroke severity (categorical) showed good performance (C-statistic=0.881, Hosmer-Lemeshow test p=0.371). The hemorrhagic stroke model with age, sex, stroke subtype, and stroke severity (categorical) also showed good performance (C-statistic=0.867, Hosmer-Lemeshow test p=0.850). Conclusion: Among risk adjustment models we recommend the model including age, sex, stroke severity, and stroke subtype for HIRA assessment. However, this model may be inappropriate for comparing hospital performance due to several methodological weaknesses such as lack of clinical information, variations across hospitals in the coding of comorbidities, inability to discriminate between comorbidity and complication, missing of stroke severity, and small case number of hospitals. Therefore, further studies are needed to enhance the validity of the risk adjustment model of acute stroke mortality.