This was a retrospective observational study that followed the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines. We used the National Health Insurance Service (NHIS) database of South Korea (NHIS-2022-1-330) [29]. The NHIS is a universal social insurance program that covers 97% of the entire Korean population. The NHIS also manages administrative processes for Medicaid beneficiaries (3% of the population with the lowest income) and reimburses medical professionals for their services. The NHIS database includes information on demographic characteristics, health care use, and diagnostic codes defined by the Korean Classification of Diseases, 7th Revision (KCD-7), and the International Classification of Diseases, 10th revision.

Due to the COVID-19 pandemic, the Korean government temporarily allowed telemedicine in the form of teleconsultations or teleprescriptions beginning on February 24, 2020. Health insurance claims data were extracted for a total of 2 years, 1 year before the policy was implemented (from February 24, 2019, to February 23, 2020) and 1 year after the policy was implemented (from February 24, 2020, to February 23, 2021).

We included all patients who used telemedicine at least once 1 year after the policy was implemented. Telemedicine use was identified with the keywords representing telemedicine in the classification code JX999 (other details) in the database, for which claims are required. However, the code does not specify the modalities of telemedicine. Due to NHIS policy restrictions, we did not receive data for the entire population. Instead, we were provided with a control group at a 5:1 ratio to the telemedicine group, based on age and sex for all telemedicine users. After receiving the data, we defined disease groups using an operational definition tailored to the study’s objectives and extracted the relevant patients accordingly. We excluded individuals who died before the policy was implemented and whose claims data were unavailable to identify outpatient visits or telemedicine through a specification code.

We included 4 chronic disease groups—hypertension, DM, COPD, and common mental disorder groups—for which telemedicine is frequently used [30-33]. We defined patients with chronic disease as those who were prescribed the relevant drug at least 2 times for the primary or secondary diagnosis 1 year before the implementation of the telemedicine policy. All the diagnosis codes for hypertension (I10-I15) and DM (E10-E14) and all the emphysema and COPD codes (J43-J44) were included in the definitions. For common mental disorders, patients with a diagnosis of schizophrenia and psychotic disease and bipolar, depressive, and anxiety disorders were included (F2x, F31-33, and F41). The major classification categories of prescribed drugs were referred to identify proper relevant drugs for each disease (Multimedia Appendix 1).

Sociodemographic variables included the Charlson Comorbidity Index (CCI), residence location, type of disability, and degree of disability [34]. Residence location was categorized into metropolitan cities (cities with more than 1 million people), cities, and rural areas. The detailed definitions of the factors are presented in Multimedia Appendix 2.

We evaluated the health care use rate for each chronic disease group regarding policy implementation. In addition, we evaluated the medication possession ratio (MPR) as a result index to analyze prescription adherence in each disease group [35,36]. The MPR was calculated by dividing the total number of days the medication was prescribed by the number of days in the study period [37]. This study compared the MPR of telemedicine and control groups before and after policy implementation, calculating the MPR for each specific drug based on the corresponding disease. For cases involving multiple drugs, the longest prescription period was used, with duplicate or overlapping days excluded, along with patients who had a history of hospitalization during the analysis period. We evaluated the relevance of each disease and its complications regarding admission to general wards (GWs), emergency departments (EDs), and intensive care units (ICUs) in the telemedicine and control groups after the policy was implemented [38]. Moreover, to overcome the weak relationship between the 1-year admission rate and each type of telemedicine use, we additionally analyzed the rates of admission to GWs, EDs, and ICUs within 1 month after each health care use for both in-person visits and telemedicine visits [39].

All continuous and categorical variables are reported as the mean (SD) and number (percentage), respectively. To test the differences in categorical variables, the chi-square test or Fisher exact test, when appropriate, was used. The effects of the telemedicine policy on the health care use rate, the MPR, and hospital admission rates were analyzed using difference-in-differences (DID) analysis. DID analysis is a method of confirming the difference in average performance between a treatment group and a control group before and after an intervention and was mainly used to confirm the effectiveness of the policy [40]. The DID model specification was as follows:

The use of the NHIS’s customized research database in this study was approved by the NHIS’s Ethics Committee (NHIS-2022-1-330). In addition, the study was approved by Samsung Medical Center’s institutional review board (SMC 2022-01-058). The need for informed consent was waived owing to the observational nature of the study.

A total of 7,891,669 patients, consisting of 1,315,284 patients in the telemedicine group and 6,576,385 patients in the 5-fold matched control group, were initially enrolled by reviewing health insurance claims data. Among these patients, 69,669 patients who died before the telemedicine policy was implemented; 2828 patients who were aged under 18 years; and 193,452 patients who had no claims data after the telemedicine policy was implemented were excluded. According to the operational definitions of the 4 chronic diseases, 1,773,454 patients in the hypertension group; 795,869 patients in the DM group; 37,460 patients in the COPD group; and 167,084 patients in the common mental disorders group were ultimately analyzed (Figure 1).

Among the patients in the telemedicine group, more than half were aged 60 years or older in all 4 disease groups (hypertension: 269,973/401,861, 67.2%; DM: 132,678/201,149, 66%; COPD: 9531/10,291, 92.7%; and common mental disorder group: 29,148/45,483, 64.1%). Telemedicine was used more by female patients in hypertension (220,025/401,861, 54.8%) and common mental disorder (31,136/45,483, 68.5%) groups but less by female patients in the DM (97,650/201,149, 48.5%) and COPD (2557/10,291, 24.8%) groups. Approximately half of the patients in the telemedicine group lived in metropolitan cities (hypertension group: 204,092/401,861, 50.8%; DM group: 101,841/201,149, 50.6%; and common mental disorders group: 22,869/45,483, 50.3%), except patients in the COPD group (4369/10,291, 42.5%), and approximately 10% of patients lived in rural areas (hypertension group: 34,412/401,861, 8.6%; DM group: 17,824/201,149, 8.9%; common mental disorder group: 4402/45,483, 9.7%; and COPD group: 1422/10,291: 13.8%). Approximately half of the patients with DM and COPD had high morbidity severity scores, with CCIs greater than 3 points (DM group: 84,164/201,149, 41.8%; COPD group: 4415/10,291, 42.9%), while 25.2% (11,451/45,483) of patients with the common mental disorder and 18.3% (73,459/401,861) of patients with hypertension had CCIs greater than 3 points. Almost all the patients with hypertension, DM, and common mental disorders in the telemedicine group had no disabilities (hypertension group: 349,811/401,861, 87%; DM group: 171,563/201,149, 85.3%; and common mental disorder group: 37,007/45,483, 81.4%); however, the patients with COPD (1241/10,291, 12.1%) in the telemedicine group had the most severe disabilities among the 4 disease groups (Table 1).

We compared the number of health care uses associated with the diagnosis of each chronic disease. During the post–COVID-19 pandemic period, health care use decreased; however, the decrease in the telemedicine group was less than that in the control group for all patients (mean difference: –0.06 vs –0.28, P<.001 for the hypertension group; –0.06 vs –0.37, P<.001 for the DM group; and –0.56 vs –0.79, P=.001 for the common mental disorder group), except COPD (–0.79 vs –0.80; P=.40). The average number of telemedicine visits in 1 year after the telemedicine policy was implemented was 1.95 (SD 1.86) for the hypertension group, 1.89 (SD 1.76) for the DM group, 1.66 (SD 1.62) for the COPD group, and 2.52 (SD 3.74) for the common mental disorder group (Table 2).

The effect of telemedicine on medication prescriptions, in terms of the MPR and telemedicine policy, resulted in an overall increase in the MPR for both patients with hypertension and DM. For patients with hypertension, the MPR increased in both the telemedicine group (83.8% vs 87%) and the control group (83.9% vs 86.6%); however, the increase was more significant in the telemedicine group, with a 0.5% DID (P<.001). For patients with DM, similar to patients with hypertension, the MPR increased in both the telemedicine group (83.9% vs 87.3%) and the control group (84.2% vs 87.4%), and the difference was greater in the telemedicine group (0.2% DID; P=.009). Moreover, the improvement in the MPR for the COPD and common mental disorder groups was not significant. For patients with COPD, the MPR increased after the policy was implemented in both the telemedicine group (46.4% vs 52.5%) and the control group (43.9% vs 50%); however, the effects of the telemedicine policy were not significant (–0.1% DID; P=.94). Similarly, for patients with common mental disorders, the MPR increased in the telemedicine group (55.3% vs 64.2%) and control group (52.90% vs 61.80%) after the policy was implemented; however, the telemedicine policy effects were not significant (–0.1% DID; P=.78; Table 3 and Multimedia Appendices 3-6). The monthly changes in MPR before and after policy implementation are presented in Multimedia Appendix 7.

The implementation of the telemedicine policy was associated with a decrease in hospitalization rates for patients with hypertension and DM, while its impact on patients with COPD and common mental disorders was not statistically significant (all P>.05). For patients with hypertension, there were decreases in GW admissions (–0.1%; P<.001), ED visits (–0.2%; P<.001), and ICU admissions (–0.1%; P<.001). In patients with DM, there were reductions in GW (–0.1%; P=.02) and ED admissions (–0.1%; P=.047); however, the influence of telemedicine on ICU admissions was not significant (P=.09; Table 4).

Analysis of hospital admission rates within 1 month after health care use in each disease group revealed that patients with DM had greater GW (0.32% vs 0.17%; P<.001), ED (0.16% vs 0.11%; P<.001), and ICU (0.06% vs 0.05%; P=.001) admission rates after in-person visits than after telemedicine visits. For patients with hypertension, in-person visits resulted in higher GW (0.13% vs 0.09%; P<.001) and ICU admission rates (0.11% vs 0.1%; P<.001), but admissions after ED visits (0.05% vs 0.03%; P=.28) were not significantly different from those after telemedicine visits. For patients with COPD, there was no significant difference in admission rates within 1 month between telemedicine and in-person visits (all P>.05). For patients with common mental disorders, telemedicine was associated with lower GW admission rates (0.12% vs 0.21%; P<.001) only, with no significant differences in other admission rates (all P>.05; Table 5).

This study examined the impact of a telemedicine policy on health care use, medication prescription adherence, and admission rates in the early stages among certain chronic disease groups using a DID analysis. In addition, this research highlights the outcomes and early effects of a national telemedicine policy that was implemented during the COVID-19 pandemic; notably, telemedicine consultations were conducted amid the outbreak.

In terms of the patient’s demographics, older patients, particularly those in their 70s, and those living in rural areas generally showed lower telemedicine use. This finding is consistent with previous reports, which attribute the trend to lower digital literacy among older age groups, limited technological infrastructure, or a preference for in-person visits in rural areas [41,42].

We found that patients diagnosed with hypertension or DM who used telemedicine had an increased MPR without increasing rates of admission to GWs, EDs, or ICUs during the first year after the policy was implemented. A major reason for these results could be that telemedicine helped improve adherence to antihypertensive and antidiabetic medications, which are critical factors in managing these disease groups [36,43,44]. We suggest that telemedicine can mitigate the decrease in health care use during lockdown periods and increase prescription adherence, ultimately enhancing the MPR and leading to lower admission rates, thereby improving disease management. When assessing the risk of adverse events associated with telemedicine use, telemedicine visits did not result in increased admission rates compared with in-person visits. Some previous studies support our findings, indicating that telemedicine improves engagement in continuous care and has a positive effect on clinical outcomes, such as lowering hemoglobin A_1c levels and managing BP [45-47]. However, these trends were not observed for patients with COPD or common mental disorders.

Telemedicine for patients with COPD has shown potential in various areas, such as remote education, disease monitoring, cost-effectiveness, and improved access to health care, with studies indicating positive outcomes in rehabilitation enhancement [48-50]. However, despite previous reports, the effects of telemedicine on reducing hospital admissions or exacerbations and improving medication adherence remain uncertain, which is consistent with the findings of our study [51,52]. We infer that the relatively older age of patients with COPD than of patients with other chronic diseases may be linked to these outcomes. Telemedicine adoption often requires familiarization with new platforms, such as mobile apps, which can be challenging for older individuals with limited digital literacy [53,54]. Furthermore, low medication adherence, often due to complex regimens, may account for the unclear impact of telemedicine on medication adherence observed in our study [55].

In our study, the patients in the common mental disorder group, similar to those in the COPD group, displayed a low MPR, and telemedicine did not improve the MPR under these conditions. Telemedicine interventions, such as telephone cognitive behavioral therapy, have been widely adopted in psychiatry. The exclusion of certain types of telemedicine treatment might have limited the observed effectiveness in our study. Nonetheless, during the COVID-19 pandemic, there was a marked preference among patients and health care providers for continuing the use of telemedicine, reflecting its potential benefits in terms of convenience and access to care, although the impact on the MPR and admission rates remained unclear [56-58]. Furthermore, a few studies have indicated that telemedicine, particularly psychotherapy in psychiatry, can broaden treatment possibilities for patients with mood and anxiety disorders, offering significant benefits for elderly patients [59].

In future studies, the long-term effects of telemedicine will need to be evaluated to establish it as a permanent and sustainable national solution. Moreover, assessing the effectiveness of telemedicine requires exploring nonmedication-related outcomes such as clinical laboratory test results, symptom severity, and mortality. In addition, the influence of various telemedicine modalities on disease characteristics should be explored. Identifying the optimal telemedicine approach for the most suitable target populations should be a focus of future research.

First, the data included a 5-fold control group matched to the telemedicine group based on age and sex for all telemedicine users, although not for the entire population, due to NHIS policy restrictions. As a result, variables such as socioeconomic status and comorbidities were not directly matched. While we incorporated fixed effects, such as region and the CCI, in the DID analysis to account for some of these factors, residual confounding may still remain.

Second, although we did not apply survey weights during the analysis, the age and sex distributions in our data were similar to national disease demographics across 4 disease groups [60,61]. Finally, our study focuses on a specific period during the COVID-19 pandemic, when telemedicine adoption surged due to policy changes. Therefore, applying these findings to nonpandemic contexts or different health care settings should be done with caution.

The temporary telemedicine policy effectively increased medication adherence and reduced admission rates for patients with hypertension and DM. However, the effectiveness of the policy was limited for patients with COPD and common mental disorders when analyzed using national claims data. Future studies are needed to demonstrate the long-term effects of telemedicine policies, considering various outcome measures and disease characteristics.