Cost of illness of ischemic heart disease in Japan: a time trend and future projections

Ischemic heart disease (IHD/ICD10: I20-I25) is the second leading cause of death in Japan and accounts for 40% of deaths due to heart diseases. Recently, the number of deaths (NDy) by IHD has been almost stable; however, the average age of death has increased [1]. In 2014, 73,862 individuals (men 41,902; women 31,960) died because of IHD in Japan.

The influence of the social burden in IHD is expected to change with aging of the Japanese population, because IHD has a high mortality rate in the older age group.

The objectives of this study were to estimate the social burden of IHD using the cost of illness (COI) method and to predict future changes. The COI method has been widely used since the 1960s after being introduced by Rice et al. as a method that can economically evaluate disease burden [2‐12]. As the disease was already studied by another group, the results of which were similar to those evaluated above, the calculation method used in this study is similar to that used in our previous studies [13‐19]. We estimated the economic burden of major forms of cancer from 1996 and projected the future economic burden [13‐17]. These analyses showed that social aging and increase in the average age of death had an impact on the decreased economic burden by devaluing human capital. However, there were no previous studies concerning the economic burden of IHD, despite it being a significant cause of deaths. This study estimated the economic burden of IHD on time trends and determined the effects of change and the aging of the burden that IHD poses on society by surveying the change predicted in the future.

This COI study used a top–down method to estimate resource consumption using governmental aggregated data. COI was estimated using the incidence-based approach. The calculation method used in this study was the same as that in our previous studies investigating COI [13‐19]. COI was estimated as the sum of direct cost (DC) and indirect cost (IC), with IC divided into morbidity cost (MbC) and mortality cost (MtC) as follows:

DC is a medical cost directly related to the disease and includes costs associated with treatment, hospitalization, laboratory investigations, and drugs. Furthermore, DC comprises hospitalization cost (HC) and outpatient cost (OC). DC was calculated using the following equation:

HC was determined by multiplying the inpatient cost per day (iCd) with the total person-days of hospitalization (THD). OC was determined by multiplying the OC per day (oCd) with the total person-days of outpatient visits (TOVy).

Here, we calculated the annual medical costs from the total medical expenses using the “Survey of National Medical Care Insurance Services” [20].

MbC is the opportunity cost lost resulting from hospitalization and visit to hospitals. We calculated MbC using the following equation:

LVd is the 1-day labor value per person. We calculated TOVy and THD in 5-year age groups based on the “Patient Survey” [21] conducted every 3 years by the Japanese government. We determined the labor values in the 5-year age groups based on data from “Basic Survey on Wage Structure” [22], “Labor Force Survey” [23], “Estimates of Monetary Valuation of Unpaid Work” and “Evaluations of Domestic Labor” [24]. We determined MbC by assuming a 1-day labor value loss per day at the hospital and a half-day labor value loss per outpatient visit. We calculated LVd and THD as follows:where Iy is the annual income per person, ULVy is the annual monetary valuation of unpaid work per person, HPy is the annual number of hospitalized patients, and ALOS is the average length of hospital stay. MtC is measured as the loss of human capital (human capital method), which was calculated using the following equation:where NDy is the number of deaths and LVl is the lifetime labor value per person. We calculated NDy of IHD for each 5-year age group based on “Vital Statistics of the Ministry of Health, Labor and Welfare” [1]. We calculated LVl by summing the present value of the potential future income should the patient survives. By estimating and projecting the COI values, the rate of contribution of DC, MbC, and MtC to the overall COI variation was calculated as follows:

\( {\mathrm{Cost}}_t^i \): DC, MbC, and MtC costs at year t

\( {\mathrm{Cost}}_0^i \): DC, MbC, and MtC costs at the baseline year

COI _t : COI at year t

COI₀: COI at the baseline year

With regard to the potential future labor value, we conducted a sensitivity analysis for the discount rate. Thus, the base case discount rate was 3%, and our analyses included a discount rate of 0–5%.

We used data from “Population Estimates” [25] published by the Ministry of Internal Affairs and Communications for 1996–2014. The data for 2017–2029 were drawn from the “Population Statistics of Japan” [26] published by the National Institute of Population and Social Security Research. We used the projection based on medium fertility and medium mortality assumptions. The year 2014 was selected as the benchmark for the 1-day labor value by sex and 5-year age groups. Two methods were used for the future projection of MbC and MtC: the “fixed” and “variable” methods. We estimated the rate of change of four health-related indicators over the past 18 years, namely, the mortality rate, per capita outpatient visits, per capita outpatient hospitalizations, and ALOS. The “fixed” method fixed the health-related indicators of each age group at the 2014 level and changed only the future population and age structure. The “variable” method accounted for changes in health-related indicators in addition to population and age structure. Regarding variable model estimation, we estimated linear (linear model) and logarithmic regression for upward trends or exponential regression for downward trends (exponential/logarithmic model) by changing health-related indicators. We then assessed the mixed model estimation by adopting a higher approximation of the decision coefficient every 5 years. Theoretically, the mixed model was the most valid, and the fixed model could be considered the reference. With regard to ALOS, we adopted the value of 2.7 days from Norway, which was obtained from OECD Health Data 2014 (Statistics and Indicators), because it was the lowest value among all 34 OECD countries (2014) for IHD. This “2.7 days” was applied for age groups that were calculated to be < 2.7 days from the estimated value of 2017–2029.

The study protocol was approved by the Ethics Committee of the Toho University School of Medicine (reference number: A16019).

The past estimation showed that a slight increase was found in the COI of IHD. DC had an unevenly increasing trend from 1996 to 2014. MbC showed decrease after peaking in 1999. MtC also had a tendency to increase until 2011, but decreased in 2014.

DC was influenced by THD, TOVy, iCd, and oCd. THD and TOVy decreased from 1996 to 2014. However, iCd and oCd remarkably increased, and the effect of this increase exceeded the effect of THD and TOVy decrease. The development of new therapies for IHD, such as drug and intra-vascular interventions, has been speculated to influence the increase in the unit cost.

MbC was calculated by multiplying TOVy/THD with LVd, and the main factor of the decrease of MbC was the decrease of ALOS. Recently, ALOS of all age groups have shown a remarkable decrease. Health policies encouraging shorter ALOS were introduced, including DPC/PDPS (Diagnosis Procedure Combination/Per-Diem Payment System), which is a reimbursement system introduced in 2003, where hospitals can receive more money with shorter length of stay, and admission criteria based on severity and needs for acute care hospitals in 2008 [27, 28].

MtC showed a tendency to increase. MtC was calculated by multiplying NDy with LVl (mean human capital value). In case of IHD, both NDy and LVl showed increasing tendencies until 2014. MtC was considered to be influenced by change in mortality rate, social aging (i.e., increasing number of elderly people), and change in LVl. At first, the reduction in mortality rate decreases NDy. Previously, a decrease in the mortality rate was observed in the elderly group (Table 1). If population structure, relative population of each age group, was stable, the NDy would decrease. In contrast, social aging has the effect of increasing overall NDy by increasing the population of elderly people with high mortality rate. Therefore, the increase in NDy in the past estimation has been speculated to occur because social aging effect exceeded the effect of the decrease in mortality rate. The increase in the average age of death by social aging also has the effect of reducing LVl at the time of death because the LVl of elderly people is low. In contrast, the increase in LVl of each sex and age group has the effect of increasing MtC. LVl showed remarkable increase in the elderly in particular from 1996 to 2014. Particularly, an increase of > 36.0% was found in the 70–74-year age group of both sexes, which was considered to be the result of the increase in labor force participation rate in the elderly and the increase in unpaid work, such as care for the spouse [21, 29]. The average age of death increased from 1996 to 2014, but the effect of reducing the LVl was speculated to be limited because the average age of death was already high in 1996, and the rate of increase was small. In other words, we can infer that the MtC increased because the effect of the increase in LVl in the elderly exceeded the effect of increase in the average age of death. Under the influence of these results, COI has been speculated to increase from 1996 to 2014.

Future projection of COI using the fixed model indicated that the COI of IHD would increase. However, if the trend of health-related indicators (mortality rate, per capita outpatient visits/hospitalizations, and ALOS) continues, COI will decrease in both linear and exponential/logarithmic models. The mixed model, which we consider to be the most credible model, also predicted that COI will show a decreasing tendency.

In the future, the proportion of the population in the ≥ 65 year age group will increase further, and NDy due to IHD will increase, but LVl at the time of death (value of human capital) will decrease because of the higher average age of death.

In the fixed model, NDy, DC, MbC, and MtC were predicted to increase. The effect of the decrease in per capita MtC was considered to offset the effect of the increase in NDy. In contrast, the mixed model showed that COI would decrease. As for the unit cost, the increase was estimated by the past trends; however, a remarkable decrease was predicted in THD and TOVy from 2017 to 2029. Therefore, it was estimated that DC will decrease. NDy and MtC decreased, the average age of death increased, and the per capita MtC decreased.

Previous studies have suggested that COI of major cancers decreased because of devaluation of human capital caused by the increase in the average age of death [13‐17]. However, the average age of death by IHD was already high in 2014, and it is predicted to increase gradually. Therefore, COI of IHD was projected to be influenced mainly by the decrease in MtC due to the decrease in NDy. The effect of the decrease in the human capital value was expected to be small.

This study has several limitations. First, the COI method used here did not take the quality of the medical treatment provided or patients’ quality of life into consideration. Therefore, it does not examine the cost effectiveness of individual medical management. However, it is still useful because it enables future estimations and to regard the impacts of the aging population. Second, our model could not catch the effect of LVl increase on future MtC because our model fixed LVl in 2014. Our past estimation showed that LVl of the elderly was increasing, and it increased the MtC, which might underestimate our future projection.

However, projecting precise LVl per person classified by age groups is difficult. We cannot assume a permanent increase of labor force participation rate and unpaid work of the elderly, and these effects on future LVl are considered to be restrictive. In addition, the study period was relatively short, and dramatic changes occurred in the health care system during this period. However, the change among the different methods for determining projections was small, and therefore, the projections are likely to be accurate for the near future.

The COI of IHD was estimated using Japanese government statistics. The past estimation showed that the COI of IHD increased. DC increased because of the increasing unit costs of developing new IHD therapies, MbC decreased because of the decrease in ALOS, and MtC increased because of social aging. In the mixed model, which we believe has the highest degree of relevance in future projection, the COI of IHD was predicted to decrease from 2017 to 2029 if the trend of health-related indicators (mortality rate, per capita outpatient visits/hospitalizations, and ALOS) continues. Possible factors contributing the change in the COI of IHD include aging of the population, DC, MtC, and NDy, and the impact of the aging of the population was considered the most significant.