In this retrospective cohort study, we analyzed data from the administrative database of health insurance claims in Mie Prefecture, Japan. Health insurance is mandatory for residents in Japan. The database stores information related to the National Health Insurance and Advanced Elderly Medical Service System; those covered are self-employed individuals as well as retirees and their dependents aged 75 years or above. These insured under that scheme amounted to 37.8% of the total Japanese population in 2020 [14]. The claims data include diagnoses, use of medical care (face-to-face or telehealth), prescriptions, and expenditures related to medical procedures.

The diagnosis information is recorded using the International Classification of Diseases 10th Revision (ICD-10) codes and Japanese diagnosis codes, with flags indicating the diagnosis. With links to insurance information, the database includes patients’ sex, date of birth, and place of residence.

We conducted this study in Mie Prefecture. The prefecture is located toward the center of Japan; it had a population of 1.77 million (22nd highest among the 47 prefectures) and an aging rate of 29.4% (18th highest) [15]. Figure 1 shows the geographic and demographic details for Mie Prefecture in 2020; Table 1 presents information about the medical facilities.

The term “telehealth” is polysemous and defined differently in different contexts. In this study, we categorized telehealth services as follows: online (using videoconferencing systems or devices), telephone, and monitoring (using a monitoring device such as a cardiac pacemaker and continuous positive airway pressure). We identified individuals who used any telehealth service at least once from January 2017 to September 2021. We determined the role of telehealth for each patient, such as for prescriptions and home care for COVID-19. Furthermore, we obtained data about sex, age when the patient used the service, and diagnoses. To calculate the monthly proportion of people using telehealth services, we identified the number of patients who received face-to-face consultations (visits to clinics or hospitals). The diagnoses were categorized by diseases according to ICD-10. We followed previous studies regarding the selection of diseases of interest for telehealth, mapping the targeted diseases, and ICD-10 classification [16,17]. We summarized the following: sex, age, type and frequency of telehealth services used, medical expenses recorded with the same claims as used with telehealth, and disease or condition when patients used telehealth. We determined the characteristics of medical institutions that provided telehealth at least once with respect to clinics or hospitals, number of beds, and location (within or outside Mie Prefecture) using a unique number code applied to medical institutions.

We compared the characteristics of telehealth users for a year before and after April 2020 as the reference time point, which was the first declaration of the state of emergency in Japan [18]. We also identified patients who used telehealth for initial consultations following the Japanese government announcement on April 10, 2020, that patients could do so.

The data set included in the study was provided by the regional public health care insurers in Mie prefecture with the approval of the personal information protection commission in charge of each of the local government areas and by contract with the local governments and the insurers.

The outcome was a change in the monthly proportion of telehealth users to beneficiaries. We evaluated the outcomes by stratifying patients into 2 age groups: 65 years old and older and younger than 65 years. To evaluate the impact of each disease, we applied stratification based on diseases and conditions. The secondary outcome was the change in the monthly proportion of telehealth users to beneficiaries.

We compared continuous variables for the 2 groups using the Wilcoxon rank sum test. Categorical data (such as sex) were compared using the chi-square test with Yates correction.

To determine the outcomes, we undertook an interrupted time-series analysis (ITSA) [19,20] to evaluate the association between the reference time point and the outcomes from January 2017 to April 2021. Also referred to as segmented time-series regression, ITSA is applied to statistically measure changes in various levels and trends in a postintervention period compared with a pre-intervention period [21]. Owing to the frequency of telehealth use, we used ITSA in the present study with a log-linear model as the link function and assumed a Poisson error structure with adjusted seasonality, which could also be called a segmented time-series Poisson regression:

Y_t represents the monthly number of telehealth users to beneficiaries at time point t; beneficiaries_t signifies the number of monthly beneficiaries. We set it to calculate the proportion of telehealth users to beneficiaries as a log-offset term. T_t represents the time since January 2017; T₀ is a dummy variable signifying t at the reference time point, the 40th month in this study. X_t is a dummy variable indicating the time after the reference time point: After April 1, 2020, it was 1; before, it was 0. β₀ signifies the level of Y_t at t=0. β₁ represents the pretrend; β₂ is the outcome change immediately after the outbreak of COVID-19 in Japan and the policy responses to it; β₃ is the change in the trend compared with before the reference time point. harmonic_t was set as the Fourier term to adjust for seasonality [19].

To remove the effect of an immediate increase at the reference point, we excluded 1 month’s observations before and after that point [22]. For subgroup analysis, the same Poisson regression was applied to the groups under 65 years old and 65 years old and older. For the sensitivity analysis, we changed the bandwidth for the removal for the overall telehealth user population from 0 months to 2 months; we did so to assess robustness and goodness of fit for each model based on the Akaike information criterion (AIC).

We performed all statistical analyses using R version 4.1.0 (R Foundation for Statistical Computing, Vienna, Austria). All statistical tests were 2-sided; we set the significance level at 5%, and we used 95% CIs for interval estimations of the outcomes.

Table 1 presents the basic information about patients who used telehealth. During the observation period, 13,618 patients (33.71 per 1000 beneficiaries, 18.62 per 1000 outpatients) used telehealth before April 2020, and 28,853 (71.42 per 1000 beneficiaries, 39.46 per 1000 outpatients) did so after that time. An immediate increase after the reference time point was also significantly evident with ITSA (RR 2.97; 95% CI 2.14-2.31). However, no change in the trend was observed (change in trend 1.00; 95% CI 1.00-1.01). Among the patients who used telehealth, there was a higher proportion of women than men. After the reference point, there were increases in the frequency of telehealth use per patient, the number of telehealth users for initial consultations, and medical expenses. Regarding diseases and conditions associated with telehealth use, we observed an increase in hyperlipidemia, cardiovascular disease, hyperuricemia, and mental disease. Most telehealth services were provided by clinics, but we identified an increase in the number of hospitals using telehealth. We observed an increase in the proportion of telehealth utilization in medical institutions outside Mie Prefecture compared with inside the prefecture.

Recent studies have reported an increase in telehealth use with COVID-19 [17,23]. However, no reports have examined how telehealth utilization has changed in a country such as Japan, where patients have free access to medical care and facilities are abundant but where the government has imposed restrictions on telehealth. With ITSA, this study confirmed the immediate increase in telehealth use after April 2020. However, no following change in the trend was evident with either model. This is not consistent with results from a previous study in the United Kingdom, where universal health coverage is provided through the National Health Service, similar to Japan [24]. When focusing on the Japanese policy strategy for telehealth from the point of view of providers, until at least 2022, telehealth has not spread widely in Japan because face-to-face consultations are more profitable for medical institutions than telehealth. Therefore, to expand the use of telehealth, not only deregulation but also a review of the incentive system are considered necessary. In addition, since deregulation of telehealth was a temporary measure until April 2022, medical institutions’ use of telemedicine may have been inhibited. This should also be researched, by extending the observation period.

We should state that our database did not include all patients; employee-insured patients, for example, were excluded. Our database mainly covered the population who were older than 65 years, which allowed us to capture the impact of telehealth utilization on this group. Nevertheless, we obtained consistent results when the ITSA was conducted with a subgroup analysis—with patients divided into those aged under 65 years and those aged 65 years and older. Whether similar results can be obtained for the entire population younger than 65 years of age is a matter for additional analysis.

We found that telehealth use was mostly by telephone and for prescriptions. Such utilization has the potential to serve as an alternative to conventional consultations, but additional verification of patient safety is required. We observed that online consultation was rarely used by public insurance services. In the Japanese health care system, telehealth is mainly limited to follow-up purposes, and it is likely that telephone consultation was chosen because many cases had little need for face-to-face consultation. Since online consultation is a relatively new system in Japan, having been issued in 2018, it is necessary to continue to investigate its prevalence. In addition, the Japanese government’s deregulation of telehealth in April 2020, which was announced as a temporary measure, may have slowed the increase in the number of medical institutions intending to adopt telehealth or caused them to defer decisions about it.

With respect to telehealth use, we observed changes in patient characteristics and in disease composition as well as increased medical costs after the time reference point. Our results suggest that the Japanese government should not only examine the spread of telehealth in terms of the increased number of cases but also focus on changes in the characteristics of patients who use telehealth. If drug treatment becomes the main treatment strategy, medical resource allocation should be considered in conjunction with pharmacy allocation. The complementary role of telehealth should also be evaluated: It is difficult to provide only telehealth when physical treatment, such as nutritional guidance and rehabilitation, is involved.

Japanese patients are not generally treated by primary care physicians or general practitioners; not all patients have specific physicians they consult on a regular basis. Given that more patients are using telehealth for initial consultations, it is necessary to promote a communication infrastructure that offers the following: vetting of eligible patients, clarification of criteria for the type of treatment (changing to face-to-face consultations), and collection of the required information. With regard to the type of medical institutions providing telehealth, the number of hospitals has increased; thus, the government needs to design a system that takes into account changes in providers. The increasing use of medical institutions outside the patient’s prefecture of residence suggests that access to telehealth is affected by factors other than geography.

One strength of the present study was identifying the use of telehealth and how it changed after COVID-19 in part of Japan, which has free access to medical care and abundant medical facilities. However, several limitations of this study deserve mention. First, this was an observational study limited to Mie Prefecture. Therefore, caution should be exercised regarding the applicability of our results to the general population; our results should be verified for Japan as a whole. Second, this study did not cover all insured individuals, which may have led to selection bias. In particular, it should be noted that our results strongly reflected the trends among older adults; there was a lack of information on telehealth use among people enrolled in employee insurance (ie, the population of working age). Third, several studies have reported better outcomes in improving treatment adherence and preventing serious events with telehealth use [25-27]. Longer observation periods than in the present study would be needed to determine the incidence of events that could serve as medical outcomes; thus, this is a matter for future study. Finally, we observed changes in patient and health care provider characteristics and trends only as a temporary state. However, it is possible that circumstances after COVID-19 will continue to change dynamically. Therefore, further data collection and analysis over a longer period are needed.

The observations of this study from January 2017 to September 2021 indicate that telehealth use temporarily increased after COVID-19; however, the diffusion trend remained unchanged. The attributes of patients who used telehealth and the medical institutions that provided it underwent change: Decision makers for related policies need to design their systems in light of this environment.