Declining accuracy in disease classification on health insurance claims: should we reconsider classification by principal diagnosis?

The Effects of Diagnostic Definitions in Claims Data on Healthcare Cost Estimates: Evidence from a Large-Scale Panel Data Analysis of Diabetes Care in Japan

BACKGROUND

Inaccurate estimates of diabetes-related healthcare costs can undermine the efficiency of resource allocation for diabetes care. The quantification of these costs using claims data may be affected by the method for defining diagnoses.

OBJECTIVES

The aims were to use panel data analysis to estimate diabetes-related healthcare costs and to comparatively evaluate the effects of diagnostic definitions on cost estimates.

RESEARCH DESIGN

Monthly panel data analysis of Japanese claims data.

SUBJECTS

The study included a maximum of 141,673 patients with type 2 diabetes who received treatment between 2005 and 2013.

MEASURES

Additional healthcare costs associated with diabetes and diabetes-related complications were estimated for various diagnostic definition methods using fixed-effects panel data regression models.

RESULTS

The average follow-up period per patient ranged from 49.4 to 52.3 months. The number of patients identified as having type 2 diabetes varied widely among the diagnostic definition methods, ranging from 14,743 patients to 141,673 patients. The fixed-effects models showed that the additional costs per patient per month associated with diabetes ranged from US$180 [95 % confidence interval (CI) 178-181] to US$223 (95 % CI 221-224). When the diagnostic definition excluded rule-out diagnoses, the diabetes-related complications associated with higher additional healthcare costs were ischemic heart disease with surgery (US$13,595; 95 % CI 13,568-13,622), neuropathy/extremity disease with surgery (US$4594; 95 % CI 3979-5208), and diabetic nephropathy with dialysis (US$3689; 95 % CI 3667-3711).

CONCLUSIONS

Diabetes-related healthcare costs are sensitive to diagnostic definition methods. Determining appropriate diagnostic definitions can further advance healthcare cost research for diabetes and its applications in healthcare policies.

Effects of health guidance on outpatient and pharmacy expenditures: a disease- and drug-specific 3-year observational study using propensity-score matching

BACKGROUND

Evidence is lacking on whether health guidance for metabolic syndrome reduces health care expenditures. The author used propensity-score matching to evaluate the effects of health guidance on health care expenditure.

METHODS

Men who did and did not receive health guidance from a health insurance society (approximately 60 000 covered lives) were matched (n = 397 respectively) using propensity scores. Health insurance claims were compared using cumulative health care expenditures for metabolic syndrome-related outpatient medical care and drug costs for the period from the initial consultation to 3 years later.

RESULTS

No difference was observed between intervention and control groups in cumulative outpatient charges or drug costs related to metabolic syndrome. However, regression analysis using the Tobit model showed that health guidance resulted in a small, nonsignificant reduction in health care expenditure.

CONCLUSIONS

Health guidance for metabolic syndrome did not reduce outpatient charges or drug costs related to metabolic syndrome during the 3-year period after the intervention. Findings from Tobit regression suggest that health guidance might eventually result in savings, but this hypothesis remains untested.

Estimating medical expenditures spent on rule-out diagnoses in Japan

BACKGROUND

According to the regulations concerning reimbursement rules for the uniform coverage scheme in Japan's health insurance system, rule-out diagnoses must be included in a health insurance claim (HIC) to ensure reimbursement for clinical procedures whose results show that a suspected disease is not present. However, estimations of disease-specific medical expenditure by conventional methods have not considered the information on rule-out diagnoses.

OBJECTIVES

To estimate disease-specific medical expenditure for rule-out diagnoses.

METHODS

Data were obtained from 169,622 outpatient HICs in May 2006 from corporate health insurance societies. We used the proportional distribution method to estimate medical expenditure for each of the major disease categories defined by the Classification of Diseases for the use of Social Insurance, which is based on the International Statistical Classification of Diseases and Related Health Problems, 10th Revision.

RESULTS

There were 442,010 diagnoses on the HICs, of which 20,330 (4.60%) were rule-out diagnoses. Rule-out diagnoses accounted for 8.5% of total medical expenditure. The proportion of medical expenditure spent on rule-out diagnoses varied across the major diseases categories, and it was estimated that more than one-third (36.9%) of the medical expenditure on neoplasm is spent on rule-out diagnoses.

CONCLUSIONS

The existence of rule-out diagnoses affects the estimation of disease-specific medical expenditure. Therefore, the estimation of disease-specific medical expenditure and evaluation of prevention and treatment programmes should be improved by utilizing information on rule-out diagnoses.

A statistical analysis of 'rule-out' diagnoses in outpatient health insurance claims in Japan

BACKGROUND

As each clinical procedure must be justified by a corresponding diagnosis in Japanese health insurance claim (HIC), unconfirmed diagnoses marked as rule-out diagnoses will be written on an HIC. However, little is known about the statistical profiles of such rule-out diagnoses.

OBJECTIVES

To illustrate the basic statistical profiles of rule-out diagnoses.

METHODS

We analysed all the diagnoses on 169 622 outpatient HICs in May 2006 from corporate health insurance societies. The proportions of additional diagnoses and rule-out diagnoses in each of the major disease categories defined by the Classification of Diseases for the use of Social Insurance, which is based on the International Statistical Classification of Diseases and Related Health Problems 10th Revision were calculated.

RESULTS

There were 442 010 diagnoses on the HICs, of which 96 253 (21.8%) were principal diagnoses and 345 757 (78.2%) were additional diagnoses. Of the principal diagnoses, 1022 (1.1%) were rule-out diagnoses. Of the additional diagnoses, 19 308 (5.6%) were rule-out diagnoses. The percentage of rule-out diagnoses in the additional diagnosis category was significantly higher than that in the principal diagnosis category. Among the major disease categories, neoplasms showed the highest percentage of rule-out diagnoses for both principal diagnosis and additional diagnosis.

CONCLUSIONS

The existence of rule-out diagnoses affects the results of statistics based on HIC data. Japanese statistics based on HIC data should be improved by utilizing the information on rule-out diagnoses.