Cost-effectiveness of a New Opportunistic Screening Strategy for Walk-in Fingertip HbA_1c Testing at Community Pharmacies in Japan

Type 2 diabetes is often asymptomatic and can remain undiagnosed and untreated for many years, until complications appear (1,2). Detection of prediabetes or early-stage diabetes would provide an opportunity to intervene and delay the development of type 2 diabetes; however, there are disputes about the direct evidence of screening itself (3–6).

There are several different approaches to screening for type 2 diabetes, namely, universal, targeted, community-based, and opportunistic screening. These approaches have been used alone and in combination (6). Nationwide universal screening (Specific Health Checkup [SHC]) includes type 2 diabetes screening in Japan (7) and the National Health Service Health Check program in the U.K., a nationwide targeted screening, which includes a diabetes check as well (8). Opportunistic screening is also common in many countries (6). This involves testing for type 2 diabetes that is conducted at the time of scheduled clinical visits for the management of other diseases (8). Also, opportunistic screening at community pharmacies has been in place in some countries, such as the U.S., Switzerland, the U.K., and Australia (8,9).

In Japan, undiagnosed cases of type 2 diabetes have been estimated to account for ∼35% of the total 9.4 million cases among adults in 2015 (10). Currently, the SHC represents one of three available opportunities for type 2 diabetes screening, in addition to conventional opportunistic screening in clinics and walk-in fingertip HbA_1c testing at a Specimen Measurement Office (SMO). The Japanese government launched the SHC and Specific Counseling Guidance in 2008 as a nationwide mass screening program to detect metabolic syndrome among adults 40–74 years of age. Insurers are required to allow insured persons 40–74 years of age to participate in the SHC program. During an SHC visit, HbA_1c or fasting blood glucose level should be screened to detect type 2 diabetes (11). However, although it is a nationwide program, the SHC coverage was only ∼48.6% in 2014 (12).

A new opportunistic community-based strategy was launched in Japan to detect lifestyle-related diseases including diabetes, by creating the SMOs, which offer HbA_1c testing, in April 2014. This legislation was prompted by the success of a trial diabetes screening project, Revolution of Access to Diabetes Diagnosis (RADD), which was implemented at 10 pharmacies in Adachi Ward, Tokyo, as a joint program of the nonprofit Adachi Diabetes Mellitus Society and the Adachi Ward Pharmacist Association (13). On 28 February 2018, there were 1,586 SMOs in Japan, mostly established on the premises of community pharmacies or drug stores (14). In Adachi Ward, the RADD evolved in 2015 into an official subsidized program for HbA_1c testing at SMOs, to increase participation in diabetes screening. In this locally subsidized program, only 5 U.S. dollars (USD) (equivalent to 500 Japanese yen [JPY]) per test is charged to residents of Adachi Ward, compared with the regular fee of USD 10 (JPY 1,000) at the 10 RADD pharmacies (15).

HbA_1c testing at the SMOs is mainly targeted to people who miss or fail to present at SHC visits. The screening complements conventional health services using an innovative approach. If opportunistic screening at SMOs becomes widespread in Japan, it could offer an important resource for people who are not tested by the SHC program. It is suggested that it could also provide a new mode of care in community settings for diverse health care professionals seeking better management of people with diabetes. This article aimed to assess the value-for-money of HbA_1c testing services at SMOs from a societal perspective by conducting a cost-effectiveness analysis.

We conducted a cost-effectiveness analysis to evaluate the introduction of HbA_1c testing services at SMOs in Japan by adopting a societal perspective. We compared two scenarios: 1) status quo, defined as HbA_1c testing that is available only during SHC visits and conventional opportunistic screening in clinics, and 2) HbA_1c testing that is available at SMOs as a complement to the status quo scenario.

In the model, the data on SMOs were taken from the RADD Project (16). Data were used from 1,296 people who were 40–74 years of age from a total of 2,024 people who provided HbA_1c data to SMOs between October 2010 and May 2014. The total study population (n = 2,024) had a mean age of 52.1 years (age range, 17–92 years), and 48% were male. We also carried out a literature search to find relevant available evidence on type 2 diabetes and its complications in Japan. The PubMed database; Igaku Chuo Zasshi (Japan Medical Abstracts Society), a Japanese medical literature database; annual government statistical reports; Ministry of Health, Labour and Welfare (MHLW) grant system; and local government reports and tariffs were accessed using various combinations of relevant terms, such as “type 2 diabetes.” We then used those sources to estimate the annual transition probabilities of complications as well as costs.

The process modeled for HbA_1c testing in pharmacy settings at SMOs is as follows. A customer at a pharmacy asks the pharmacist to have their HbA_1c tested. The pharmacist confirms that the customer is not being treated for diabetes, to exclude patients who have received a diagnosis of type 2 diabetes. After receiving an explanation about the HbA_1c test from the pharmacist, the customer undergoes testing. The pharmacist gives the customer their results together with information about how to interpret the results. The customer is advised to visit their family doctor if their HbA_1c level is not <6.0% (42 mmol/mol) (16).

In this model, we assumed the participant cohort to be individuals 40–74 years of age who sought walk-in fingertip HbA_1c testing. In addition, we assumed a hypothetical scenario of only one HbA_1c test at an SMO for the age-groups 40–44, 45–49, 50–54, 55–59, 60–64, 65–69, and 70–74 years. At the initial stage of the model, every person is assumed to have no diagnosis of or have not been treated for type 2 diabetes. The model consisted of a screening module with a decision tree (Fig. 1A) and a disease progression module with a Markov model (Fig. 1B).

Two scenarios were modeled as follows: 1) status quo and 2) complementary HbA_1c testing at SMOs. The model was a combination of a decision tree and a Markov model. The decision tree included two main nodes, as follows: 1) the decision node, which represented the decision made by pharmacy managers about whether to establish an SMO that provided HbA_1c testing on the pharmacy premises, and 2) chance nodes, which represented the possible events that could occur (17). In Japan, 6% (42 mmol/mol) has been adopted by the National Health and Nutrition Survey as the cutoff value for defining the risk for the development of type 2 diabetes. Therefore, in the current study, a participant with an HbA_1c test result above this cutoff value was defined as a person at risk for diabetes. Additional details are shown in the Supplementary Appendix, and other probabilities assigned to the chance nodes are shown in Supplementary Table 1.

The following two kinds of Markov models followed the decision tree: a Markov positive model (Markov [+] model), which suggests the prognosis for a person with type 2 diabetes (Fig. 1B), and a Markov negative model (Markov [−] model, which suggests the future for a person without type 2 diabetes (Fig. 1C).

The Markov (+) model comprised a series of five health states that a patient can have at a given point in time (17): 1) type 2 diabetes without complications, 2) end-stage renal disease (ESRD), 3) cardiovascular disease (CVD), 4) amputation, 5) blindness, and 6) death. For type 2 diabetes, each person was assumed to have a single complication rather than multiple complications. Each transition probability defines the rate at which each complication is reached with or without treatment as well as the probability of death from no diabetes or diabetes without complications as well as diabetes with a complication. We assumed that the age-specific probability of death for those without diabetes and those with diabetes without complications or with blindness only were the same (18).

The Markov (−) model comprised the following two states: 1) no type 2 diabetes at a given point in time and 2) death. Annual transition probabilities (death with no type 2 diabetes) were determined using vital statistics (19). The effects of early intervention for type 2 diabetes were included in the Markov (+) model. One Markov cycle was set at 1 year. Individuals who did not undergo HbA_1c testing at SMOs were omitted from this model because they would have produced no effects in terms of cost and effectiveness in either scenario.

Utility weights used to calculate quality-adjusted life-years (QALYs) were 1.0 for individuals without type 2 diabetes, 0.910 for individuals with early-stage type 2 diabetes without complications (20), 0.884 for individuals with advanced type 2 diabetes without complications (20), 0.746 for individuals with CVD (21), 0.620 for individuals with ESRD (22), 0.559 for individuals with amputation (22), 0.536 for individuals with blindness (22), and 0 for the individual who died. The deterioration of utility is assumed to occur at years 6 and 10 for untreated case patients and treated case patients, respectively (23,24). The utility weight for each undetected case of type 2 diabetes was assumed to be the same as the value for each detected case of type 2 diabetes without complications (25).

In this study, costs were considered for HbA_1c testing and treatment for type 2 diabetes and complications, including the costs of confirmatory HbA_1c testing by a doctor. These cost data were mainly estimated using the published literature and expert opinion. Costs are expressed in USD as well as JPY. Cost was not adjusted for the consumer price index in this article because consumer prices have barely changed. These costs were used for modeling in the decision tree and Markov model (Table 1). The cost for HbA_1c testing at SMOs is shared by pharmacies and individuals. The costs for confirmatory testing and treatment for type 2 diabetes and complications are shared by social insurance and patients. The direct costs for treatment for type 2 diabetes were mainly estimated using medical fee tariffs; costs for sectors other than health and productivity losses were uncounted for in this study.

The cost of each HbA_1c test at SMOs, USD 12 (JPY 1,229), was estimated based on data from the RADD Project. The costs consist of fixed/variable capital costs and labor cost. The fixed capital cost, USD 1 (JPY 122), is calculated from the annualized capital cost of the testing equipment and an expected 732 tests, according to the RADD Project, in a year. The variable capital cost includes the price of a test kit: USD 6 (JPY 560) per participant. The pharmacist’s labor cost for a 15-min test, USD 5 (JPY 547), is calculated from an average hourly wage of USD 22 (JPY 2,189) for a pharmacist (26). The cost of an HbA_1c test at the SHC was estimated to be USD 9 (JPY 918) (27).

The costs of confirmatory testing and treatment for type 2 diabetes and complications were estimated as follows. For people with positive HbA_1c test results at SMOs, a cost of USD 78 (JPY 7,800) was set for additional testing (22). The annual costs for outpatient treatment for type 2 diabetes was set at USD 3,700 (JPY 370,000) (28). The annual cost for ESRD treatment was set at USD 48,230 (JPY 4,823,000) (29). The cost in the first year for CVD treatment was set at USD 19,990 (JPY 1,999,000), and USD 1,790 (JPY 179,000) in the second year (21). The cost of amputation in the first year was set at USD 22,910 (JPY 2,291,000) (30). The annual cost of blindness was set at USD 760 (JPY 76,000) (30). Most data were estimated from the medical care fee schedule in Japan.

We took a life-long time horizon, and the Markov cycle was repeated as the age stratum reached 100 years old.

Costs and outcomes were discounted at a rate of 3% (17).

In this study, we performed one-way sensitivity analyses for our model assumptions. In the analyses, lower and upper limits were applied as plausible values from references in case there was an absence of information on values, such as 95% CI (Supplementary Table 1).

We also conducted a probabilistic sensitivity analysis using a set of 1,000 Monte Carlo simulations. The distributions for variables of probabilities, utilities, and the discount rate were assumed to have triangular distribution, and those for variables of costs were assumed to have a γ-distribution. The lower and upper limit ranges in the triangular distribution were applied in the same way as in the one-way sensitivity analysis.

In this study, we compared two scenarios to be chosen by pharmacy managers: 1) status quo, defined as HbA_1c testing only at SHC visits and conventional opportunistic screening in clinics, and 2) HbA_1c testing at SMOs as a complement to the status quo. We calculated the respective costs and effects of these scenarios, and the difference between costs and effectiveness was compared. That is, we calculated incremental cost-effectiveness ratios (ICERs; cost per QALYs) over the lifetime analytic horizon as the primary end point of the cost-effectiveness analysis. ICERs were defined as follows:

Model outcomes were analyzed using TreeAge Pro 2016 (TreeAge Software, Inc., Williamstown, MA).

The total cost per individual for those individuals 40–74 years of age was estimated to be USD 4,951 (JPY 495,053) for HbA_1c testing at SMOs, compared with USD 5,478 (JPY 547,775) for the status quo scenario. QALYs were estimated to be 17.996 for HbA_1c testing at SMOs compared with 17.975 for the status quo. Therefore, the total gain in QALYs was greater, and the total cost was lower for HbA_1c testing at SMOs compared with the status quo. Thus, this cost-effectiveness analysis showed that compared with the status quo, introducing HbA_1c testing in SMOs is more effective and less costly, which means cost saving with improved QALYs for the population 40–74 years of age (Table 2). We conducted a subgroup analysis by each age-group and assumed the participant cohort to be 40–74 years of age (in 5-year increments). The largest outcome gain in QALYs was estimated to be among individuals 60–64 years of age. The introduction of HbA_1c testing at SMOs remained dominant in all age-groups from 40 to 74 years (Table 2). We also conducted a subgroup analysis focusing on the participant cohort as specific age-groups in those participants 40–74 years of age (in 5-year increments) (Supplementary Table 2).

Sensitivity analysis is commonly used for handling uncertainty. Figure 2A shows the results of one-way sensitivity analysis for the top 10 variables that effect ICER, depicted as tornado diagrams. The parameters that have a potential effect on ICER could differ depending on the variable. Each transition probability for progression to ESRD among individuals with type 2 diabetes with and without treatment for type 2 diabetes affected the ICERs. ICER was sensitive to the utility of ESRD and to each probability for death because of ESRD among both male and female participants.

Figure 2B shows the results of probabilistic sensitivity analyses by depicting 1,000 ICERs produced by Monte Carlo simulations. HbA_1c testing in SMOs was estimated to cost less and gain more in 63.8% of simulations. This was also estimated to be cost-effective in 90.6% of simulations, including dominant cases, which were under the threshold of USD 50,000 (JPY 5,000,000) per QALY gained; 97.7% of simulations, including dominant cases, were under the threshold of USD 100,000 (JPY 10,000,000) per QALY. Figure 2C shows scatter plots of the cost-effectiveness plane; 0% of simulations showed negative QALY gained.

This study assessed the cost-effectiveness of a new walk-in fingertip HbA_1c testing strategy at SMOs in Japan, such as at community pharmacies. The results showed that HbA_1c testing services available at SMOs offer cost savings compared with the status quo of HbA_1c testing only at SHC visits and conventional opportunistic screening in clinics.

We focused on walk-in fingertip HbA_1c testing for the early detection of diabetes through HbA_1c screening in a pharmacy setting, a unique point of this study. If given the opportunity, some individuals who test positive in this setting would seek confirmatory testing and receive a diagnosis by a doctor, which could lead to earlier intervention. This study is different from other cost-effectiveness studies that have focused on type 2 diabetes screening strategies such as universal, community-based, targeted, and opportunistic screening conducted together with management for other diseases like hypertension (6,31). Complementary HbA_1c testing in SMOs could benefit the population who live far from the SHC or checkup sites or conventional opportunistic screening.

Based on the RADD Project background, we modeled walk-in fingertip HbA_1c testing in an SMO setting. In this model, we took the following points into account. First, the percentage of individuals who have never undergone routine SHC checkups was 21.2% on average (for individuals 40–74 years of age). Second, the proportion of individuals with HbA_1c level ≥6.0 mmol/mol among all participants (age range, 40–74 years) who had never had a health checkup was 32.8%. Therefore, ∼7% (21.2% × 32.8%) of participants newly learned that they possibly had type 2 diabetes in the pharmacy settings. There is an undiagnosed population with type 2 diabetes of ∼3.3 million in Japan (10). This model is generalized for the population who are unaware of walk-in fingertip HbA_1c testing besides a lack of access to SHC checkups.

We consider our results to be stable and robust since most simulations are under the threshold of USD 50,000 (JPY 5,000,000) per QALY gained. In one-way sensitivity analysis, this model was most sensitive to probabilities of transition from type 2 diabetes to ESRD with and without treatment for type 2 diabetes. The speed of the progress to ESRD by early detection of type 2 diabetes and intervention for effective management to delay to ESRD could affect the ICER results (32). The probability for death because of ESRD was also a sensitive parameter to ICER. However, the crude death rate has remained stable at between 9.2% and 10.2% for the last 20 years (33).

Our results will be useful to managers of pharmacies or other health institutions and/or to policy makers in local government. Local policies, such as the subsidy program in Adachi Ward, Tokyo, for HbA_1c testing services provided in 10 pharmacies as SMOs since April 2015, are worth consideration. The similar subsidy programs are currently implemented in Mihara-city, Hiroshima, and Gyoda-city, Saitama, as well. Our results also have implications for policy makers in foreign countries that use community pharmacies for diabetes risk assessment programs (9).

This study has several limitations that must be considered when interpreting the results. First, the model in this study is a simplified economic model. We input four major complications of type 2 diabetes and considered each as a single item in the model. However, the consequences of type 2 diabetes vary in the real world, including the complications used in the model. Additionally, we simplified the definition of CVD to include coronary heart disease and stroke in the model. The cost was also determined using the same assumption (21). Second, although the effect of treatment on type 2 diabetes would depend upon treatment uptake, types, and adherence, we assumed a population with 100% adherence to treatment in the model. The ICER may be biased as either lower or higher when participants stop medication. Whereas the effectiveness is overestimated, costs might be either lower or higher. Third, it is quite difficult to identify the transition probability of each complication for the population of individuals who did not undergo testing. Therefore, those transition probabilities were estimated using sequential/serial data, such as checkup data. Additionally, the possibility of multiple complications may not have been excluded from the probabilities. Fourth, the utility weight was estimated as higher, at 9 years for treated case patients and 5 years for untreated case patients (23,24). This estimation of a 4-year gap might be an underestimated outcome; however, this gap affected the outcomes slightly. Fifth, testing HbA_1c does not always provide a robust method for screening for type 2 diabetes. However, we considered that it was a suitable method in the pharmacy setting because of convenience. Finally, we should consider the different progression of type 2 diabetes at a timing of SMO testing as heterogeneity. The effect of treatment for type 2 diabetes, that is, the transition probability of complications, could differ depending on the baseline. However, there are no data identifying the exact duration of type 2 diabetes, especially for patients who have never undergone testing or received medication. The base case was assumed using results from national surveys, such as the National Health and Nutrition Survey, and/or those from retrospective/prospective research surveys in Japan.

In summary, this study suggests that the introduction of HbA_1c testing in SMOs can be justified as an efficient use of finite health resources. Therefore, introducing opportunistic HbA_1c testing at SMOs that are mostly in community pharmacy settings is worth consideration by pharmacy managers and programs subsidized by local governments.

Acknowledgments. The authors thank the Adachi-ku Pharmaceutical Association, the Adachi Diabetes Mellitus Society, The Adachi-ku Medical Association, and The Cooperation Council for Specimen Measurement Offices. The authors also thank Akiko Nagai of the Ayase Community Pharmacy.

Funding. This study was funded by The Research Foundation for Pharmaceutical Sciences. This research was supported by the Mitsukoshi Health and Welfare Foundation research fund (2010).

The funding source did not play a role in any aspect of the study or in the decision to submit the manuscript for publication.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.S. wrote the manuscript and researched the data. M.K. contributed to the discussion and edited the manuscript. S.-l.H. contributed to the research and discussion. R.O. contributed to the research. N.Y. contributed to the discussion. A.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.