Accuracy of estimation of disease-specific costs in national health care expenditure: An example of renal failure and dialysis

Presenter: Etsuji Okamoto, National Institute of Public Health

Abstract

Rationale: Estimation of disease-specific health care is important for effective health policy development. While it is rather easy to estimate the entire national health care expenditure (NHCE), it is often difficult to estimate disease-specific portion of NHCE. Japan's disease-specific NHCE relies on a nation-wide sampling survey on health insurance claims called Social Insurance Claims Survey (SICS) conducted in every June but the accuracy of such estimation is difficult to verify because of lack of standards against which the estimates are compared. Recently the government published NHCE in 2006 and showed that the expenditure for renal failure had declined by 31% since previous year, an unbelievable figure given the constant increase of Japan's dialysis patients posing a serious doubt on the accuracy and reliability of the SICS.

Objectives: To verify the accuracy of estimated number of outpatient health insurance claims whose primary diagnoses are renal failure in SICS by comparing the estimates against the population survey conducted by Japan Society of Dialysis Therapy (JSDT).

Methodology: Estimated number of outpatient health insurance claims with primary diagnoses of renal failure by SICS for 22 years (1986-2007) and the number of patients on dialysis surveyed by JSDT for the same period were obtained. Since the survey data by JSDT has been stable showing a consistent upward trend, their data were used as a standard against which the estimated data of SICS were evaluated. A linear trend line was drawn from JSDT data and the estimated data of SICS were evaluated using a ratio of SICS estimates to the JSDT data in each year.

Result: The linear trend line was expressed as Y = 10051*X + 55537, where Y is the expected number of dialysis patients and X is the number of years since 1986 meaning that the number of dialysis patients is increasing by 10,051 every year. The coefficient of determination (R square) was 0.9966 and was considered to be reliable enough. The cumulative ratio of estimates by SICS to the N of dialysis patients by JSDT was 0.958, which means that SICS slightly underestimated the number of patients. The ratio fluctuated considerably particularly after 1998. The largest aberration was observed in 1999, when the SICS overestimated the number of patients by as much as 77% and a large year-to-year fluctuation was observed for the last three years.

Conclusions: SICS has increasingly become inaccurate in estimating the number of patients of renal failure and dialysis particularly for a last few years. The inaccuracy was such that the estiamtes are not reliable enough for policy making. The reason of the deterioration of sampling error remains unanswered and needs further investigation.