Estimating Prevalence and Healthcare Utilization for Treatment-Resistant Depression in Japan: A Retrospective Claims Database Study

Health insurance claims data from non-governmental employees and their family members between July 2009 and March 2015 sourced from multiple health insurance associations were retrieved from the Japan Medical Data Center (JMDC). During this period, a total of 2,958,220 patients were enrolled in the JMDC database. The database provides comprehensive patient and clinical information, including patient demographics, diagnostic codes, dates and types of procedures, dispensed prescription drugs, medical services provided to inpatients and outpatients, and expenditures. The JMDC database has been used to investigate a wide range of conditions in Japan such as schizophrenia [23] or cardiovascular disease [24]. All personally identifiable information was de-identified to protect patient privacy. Therefore, no informed consent was necessary.

A retrospective cohort design was used. JMDC database members joined the study cohort on 1 April 2012 if they were within 18–60 years of age (i.e., were born between 1952 and 1994), had no prior depression diagnosis (ICD 10; F32.x, F33.x, F34.1, F41.2, F43.2, F53.0), no previous diagnoses of other mental diseases (ICD 10; F0;F1;F2;F3F20; F30;F31), and no prior prescription of any antidepressant medication before 31 March 2012. Antidepressant medication was defined as a medication that is approved for the treatment of depression in Japan. The list of antidepressant medications can be found in Supplementary Table 1. Patients were censored if they develop other psychiatric conditions as mentioned above after they were included in the study cohort.

Based on these criteria, 98,552 patients who were continuous JMDC database members from 1 April 2012 through 31 March 2013 were included in this analysis (Table 1). Maximal follow-up time was until 31 March 2015.

A pharmaceutically treated depression (PTD) treatment interval began when a study subject received a depression diagnosis and simultaneously or subsequently (within less than 30 days) a dispensation of an antidepressant medication.

The start of a treatment interval was the date of depression diagnosis or the date of antidepressant medication dispensation, whichever was earlier. A treatment interval was terminated when the subject received no depression diagnosis and was dispensed no antidepressant medication for 120 days. The duration of a PTD interval was defined as the number of months from PTD index month to the month a treatment interval ends. It is important to recall that what we call a treatment interval of PTD is a sequence of dispensing of antidepressant medications and visits with a depression diagnosis, and that this represents a treatment interval of care for PTD rather than a clinical episode of clinical depression. In particular, medication may include some prophylaxis.

A TRD treatment interval is a PTD treatment interval in which two treatment regimens have failed. A treatment regimen fails when, at least 15 days after it began, an antidepressant medication is added or substituted for another medication. We assumed 15 days as adequate threshold, because evidence suggest that nonresponse is predicted by a lack of symptom improvement during the first 14 days of therapy [25]. Moreover, this is a common definition used in database analysis and choosing this threshold value allows for a better comparability across database studies [26].

As we are not able to observe symptom improvements in claims databases, we acknowledge that there is some uncertainty around this definition.

Hospital claims were characterized by resource utilizations as: hospitalization in emergency department, hospitalization in psychiatric emergency department, psychiatric hospitalization, all hospitalizations, psychiatric office visit (out-patient), and office visit (out-patient). Total cost per patient and total cost per patient-year were calculated. Cost per person-time and cost per person capture of different information and the former measure adjusts for time while the latter does not. Total cost refers to the sum of costs paid by insurance and by the patients via co-insurance schemes and that are paid out of pocket. The co-insurance rate is 30% for people under 70 years of age and 20% for those above 70 years of age. Maximum co-payments are capped depending on household income ranging from 35,400 JPY (US$307) per month for the population earning less than 1 million (US$8700) a year to 252,600 JPY per month (US$2200) for those with an annual income above 11.6 million JPY (US$100,900). Once a person turns 70 years old, the upper limit of the co-payments is reduced drastically ranging from 8000 (US$70 USD) to 80,000 JPY (US$700) per month depending on income and whether the cost is related to in-patient or out-patient services. At the age of 75 years, everyone in Japan switches to an insurance scheme for the elderly that was established 2008. The common co-payment rate for this plan is 10% [27]

This study employed a cohort design because TRD is defined by a temporal sequence of events (unsuccessful treatment regimens) that may occur in a subject with PTD. A cohort perspective also facilitates estimation of TRD incidence and duration of treatment. This was a descriptive study so no effect measure was calculated. The incidence of PTD and TRD were estimated by proportion per year. Group comparisons were conducted using the Kruskal–Wallis test with post hoc Scheffe’s rank sum multiple contrast tests. All statistical analyses were performed using R version 3.2.1 (The R Foundation for Statistical Computing, Vienna, Austria).

In total, 1143 (1.2%) subjects experienced 1154 PTD treatment intervals for the 98,552 patients contained in the selected sample (Table 1). Of the patients who met the inclusion criteria for PTD, 11 patients had more than one treatment interval during the study period (Table 2). If a patient had at least one TRD treatment interval, the patient was classified as a TRD patient. Reasons for censoring among subjects with PTD but not TRD and among subjects with TRD are shown in Table 2.

One hundred and thirty-seven patients developed TRD within 1 year. TRD and non-TRD/PTD patients (i.e., patients with PTD who did not develop TRD) were of a similar mean age (43.4 and 42.8 years, respectively) (Table 3). Among patients with PTD (whether or not they developed TRD), 556/1143 (48.6%) were male. In contrast, among those with TRD, 84/137 (61.3%) were male. TRD treatment intervals had longer durations (mean 22.0 months, median 25.0 months) than did PTD intervals without TRD (mean 11.0 months, median 7.0 months).

Our study estimated the 12-month incidence of PTD as 1.2% for the privately insured population of Japan. Recent epidemiologic studies of community residents reporting the prevalence of major depression according to DSM-IV criteria was 1–2% for 12 months and 3–7% for lifetime in Japan [28]. The estimated differences between Japan and the USA or other Western countries may be partially due a greater reluctance to report depression among people in the Japanese population than in Western countries. Compared to Western countries, Japanese patients with depression were reported to have decreased behaviors of seeking any medical treatment or of consulting a psychiatrist (27% sought any; 14% consulted a psychiatrist), which is lower than that reported from many Western countries, and approximately half of that in the USA [28]. Accordingly, previous epidemiological studies have shown a lower prevalence of depression in East-Asian countries including Korea, Japan, and China compared with the West [29]. Findings from the World Mental Health Japan Survey 2002–2003 [21] for instance suggest that the majority of people with a recent psychiatric disorder did not utilize mental healthcare or other support systems, despite service improvement over time. Such cross-cultural differences would equate to a higher diagnostic threshold for depression in East-Asian countries [30, 31]. Presumably, these cultural differences might also influence the gender effects on depression. Considering Japan is in the East-Asian sociocultural realm, similar estimates after accounting for sociocultural differences may be anticipated for this study. Similar to the global trend, women exhibited a greater risk than men for depressive disorders in Japan, although age and social class distributions were different from other countries [32]. Twelve percent of PTD patients developed TRD within a year, which is in the range of estimates from other countries [4, 20, 33]. For Taiwan, which is culturally close to Japan, this proportion was estimated to be 21% [26]. As we utilized a database whose members include employees and their families, patients with a very severe history of depression might not be able to stay in employment and are therefore under-represented in this sample. This would imply a downward bias of the TRD incidence rate. The majority of TRD patients were male in our analysis, which is in contrast to findings from Taiwan [26] or the USA [7]. Again, the composition of our database might be a potential explanation for this finding.

Prescription patterns in the study population showed that SSRIs were the most frequently used drugs, similar to the usage reported for six other East-Asian countries (at 40 sites) [32]. In contrast, a similar study in the USA reported that SNRIs were most frequently used [7].

Another finding is the wide usage of sulpiride, which confirms results of previous studies [34]. A Japanese chart review for instance found that sulpiride was the most frequently prescribed antidepressant with a share of 40.3% in 367 outpatients with a major depressive disorder [35]. The wide use of sulpiride is somewhat unexpected given the limited evidence of its efficacy in depression. A review of the UK HTA agency NICE in 2011 identified only one clinical trial that found a statistically significant greater mean change in the 21-item Hamilton Depression Rating Scale with 150–300 mg sulpiride compared with placebo. NICE concluded that due to the small number of studies and lack of study quality assessment, the effectiveness of second-generation antipsychotics in the treatment of major depressive disorder is unclear [36]. Table 4 indicates that many subjects were receiving more than one medication for depression at a time, and sulpiride may have been used to augment an antidepressant. In a Japanese clinical trial, this has been recommended as a successful strategy for accelerating antidepressant response [37].

Our results also suggest that more than one antidepressant medication was prescribed on average, especially in the later treatment lines, although the link between polypharmacy and efficacy has not been established [38] and polypharmacy is not recommended in international and Japanese guidelines. Instead, clinical guidelines recommend monotherapy with a second-generation antidepressant for acute-phase treatment [39]. Despite this, only 26% of Japanese psychiatrists treat patients with monotherapy according to a survey by Ueshima et al. [40]. A Japanese claims database analysis of 7,338 Japanese patients with depression also observed various patterns of polypharmacy [41].

Some psychiatric service utilization including psychiatric hospital days, overall office visits, and psychiatric office visits for TRD patients was higher compared to PTD without TRD and compared to patients without PTD. This is partly due to the fact that when a patient had a prescription renewed, that event was considered to be an outpatient visit. The finding of increased service utilization of the TRD population holds for both the per-patient-year and per-patient perspective, and echoes recent findings from Brazil [42] or the USA [4]. No similar difference was seen for all types of hospital days, where PTD patients who were not TRD had higher resource utilization.

Total treatment costs per person among TRD patients of 1.01 million JPY were higher than for patients who only exhibited PTD (0.643 million JPY) or compared with the non-PTD population (0.327 million JPY per patient). Those are only direct medical costs that accrue to the health insurer and a recent health economic study found that direct medical costs constitute only 14.2% of the total cost associated with depression in Japan [43]. The biggest fraction of the economic burden resulted from indirect costs such as productivity losses [44]. Although the magnitude of healthcare expenditure/costs for TRD was smaller than that reported in the USA [7], this study shows the same directionality. That treatment costs are higher in the USA is a well-established result in the literature and the cited Japanese health economic study [44] found direct treatment costs to be only 50% of those in the USA [45]. That TRD patients do incur higher costs to the healthcare system has been reported for a number of other countries such as Brazil [42] or the USA [14, 46‐48].

This study had some limitations. The study’s definition could not take dosage into account when defining antidepressant medication regimens due to substantial amounts of missing data, which is likely to affect the estimated TRD incidence. However, by not using dosage as a factor, this allows more regimens to be counted and thus would capture more cases compared with a study that imposed a dose requirement. Moreover, this study only investigated pharmacological interventions among patients with depression. The role of non-pharmacological treatments, such as psychotherapy, was not investigated here. For that reason, the estimates from this study are likely to be a lower bound for the actual disease incidence. On the other hand, psychotherapy is only rarely prescribed in Japan compared to its wide use in the USA [49]. Another potential limitation is that claims were used as the basis of defining events for analysis of study. In clinical practice, prophylactic measures of continued antidepressant prescriptions are taken because those that have had episodes of depression are at increased risk for sequential episodes [45, 50]. Prophylactic treatment behavior would be indistinguishable from ongoing clinical depression, so some of the changed regimens may have been initially effective. This may have affected the estimates of the durations of episodes of PTD and TRD and caused some PTD cases that were not TRD to be misclassified as TRD cases. This misclassification would result in an overestimation of the incidence of TRD. However, this limitation is common in retrospective database studies of TRD and is difficult to avoid. Measurement of patient-reported outcomes or physician assessment, such as the antidepressant treatment response questionnaire (ATRQ) or Mini International Neuropsychiatric Interview (MINI) for treatment response, would not have been available in claims data. Future studies should validate the findings using alternative definitions of TRD in order to address the high degree of uncertainty there is surrounding our results. Moreover, a common problem in any database analysis is the coding quality [51]. More often than not, hospitals or physicians designate medical codes that provide the highest reimbursement rates, but disease codes do not always reflect clinical reality.