Societal cost-of-illness analysis of self-injection of biologics for patients with rheumatoid arthritis in Japan using claims, surveys, and national statistics Authors: Kazuhiko Takahata, Eiichi Tanaka, Ryoko Sakai, Manabu Akazawa Published: 2025 ABSTRACT BackgroundThe objective of this study was to evaluate the societal cost-of-illness (COI) associated with self-injection (SI) versus non-SI of biological disease-modifying antirheumatic drugs (bDMARDs) in Japanese patients with rheumatoid arthritis (RA) using real-world claims data, patient surveys, and official statistics.MethodsWe conducted a descriptive cross-sectional COI analysis from a societal perspective using a microcosting approach, integrating (1) health insurance claims for direct medical costs, (2) a web-based patient survey for direct non-medical and indirect costs, and (3) government statistics for wage/time estimates. Costs were calculated as annual per-patient estimates (1 USD = 150 JPY). Productivity loss was estimated using a working-age patient model. A one-way sensitivity analysis and 10-year cost projections were performed.ResultsThe SI group incurred higher direct medical costs than the non-SI group ($9,280 vs. $7,298), mainly attributed to increased bDMARD prescriptions. Compared with the non-SI group, the SI group was associated with fewer outpatient visits, lower travel costs ($64 vs. $86), and less time lost ($136 vs. $315). The SI group experienced slightly higher productivity loss due to activity impairment ($5,896 vs. $5,123). The total annual COI per patient was $15,376 and $12,822 for the SI and non-SI groups, respectively. Sensitivity analysis identified activity impairment as a key driver of costs, with cost differences between groups persisting over the 10-year projection period.ConclusionAlthough SI increases bDMARD costs, it may reduce the burden of medical visits and improve treatment convenience, underscoring the need to explore its broader societal impact, particularly productivity loss and access to care. Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13561-025-00710-y. Introduction Rheumatoid arthritis (RA) is a chronic systemic autoimmune inflammatory disease that primarily affects the joints and periarticular soft tissues [1]. In Japan, approximately 826,000 patients (76.3% females) have been diagnosed with RA, with an estimated prevalence of 0.65% [2]. In recent years, the launch of biological disease-modifying antirheumatic drugs (bDMARDs) has substantially advanced the treatment of RA, making clinical remission a realistic goal [3]. Nevertheless, the associated treatment costs remain high [4], with reports indicating that annual medical expenses per patient approach approximately 1,950 USD [5]. Moreover, RA imposes a considerable socioeconomic burden by decreasing work productivity and increasing caregiving needs in the long term, thereby elevating indirect costs [6]. As of April 2025, 12 bDMARDs, including 9 originator biologics and 3 biosimilars, have been approved for the treatment of RA in Japan, with 10 of these available for self-injection (SI). Providing a portfolio of SI devices tailored to meet patient needs may enhance patient experience, thereby improving adherence and long-term treatment outcomes [7]. SI allows patients or their caregivers to administer medications at home or in other non-clinical settings. This approach can reduce various medical-related expenses, including (i) direct medical costs such as medical and drug costs [8], (ii) direct non-medical costs such as travel-related costs [9], and (iii) indirect costs such as the loss of patients’ or caregivers’ work productivity. In Europe and North America, multiple real-world studies have reported that SI improves patient convenience and treatment adherence while reducing dependency on healthcare institutions, contributing to a more efficient use of healthcare resources [7, 10]. Notably, comprehensive studies evaluating the cost-of-illness (COI) of SI in RA therapy, including both direct and indirect costs, remain limited globally, and evidence from real-world clinical practice in Japan is particularly scarce. To address this gap in the literature, we previously conducted two studies in Japan. First, in our web-based survey of Japanese patients with RA, more than 80% of bDMARD users reported selecting SI, citing convenience and reducing the frequency of hospital visits as primary reasons. Although the SI group tended to incur higher out-of-pocket expenses per visit, their total annual out-of-pocket burden was lower than that of the non-SI group, suggesting that reduced visit frequency may help alleviate patient costs [11] Second, our claims-based analysis revealed that the SI group had markedly higher monthly bDMARD prescription volumes, leading to greater direct medical costs, compared with the non-SI group. However, the SI group was also associated with substantially fewer hospital visits, potentially resulting in reduced time loss and indirect costs. These findings suggest that, although SI may appear more expensive from a payer’s perspective, it may offer better cost-effectiveness from a broader societal perspective [12]. Japan represents a unique and critical case study owing to its rapidly aging population, comprehensive universal healthcare coverage, and proactive promotion of SI for biologics [13]. Thus, evaluating the socioeconomic impact of these strategies can provide valuable insights that can be applied to other nations facing similar demographic and healthcare delivery challenges. The objective of the current study was to compare the annual COI, including both direct and indirect costs, between SI and non-SI groups among patients with RA in Japan. Furthermore, we conducted a 10-year scenario-based cost simulation to assess the long-term economic implications, considering potential changes in healthcare delivery systems and patient demographics. Methods Study design and perspective Japan’s universal healthcare system permits the use of SI under specific conditions, including instructions from a physician with sufficient clinical experience, appropriate coordination for emergency situations, and guidance on safe disposal of injection-related waste [14]. As SI is administered by patients or caregivers who have received proper medical instructions, no clinically meaningful differences in efficacy or safety are expected between patients with and without SI. This assumption is supported by findings reported in our previous study, which found no notable differences in safety outcomes associated with SI use [15]. We conducted a cross-sectional COI analysis following the methodological principles outlined by Larg and Moss [16], incorporating a microcosting approach [17]. The analysis was performed from a societal perspective, and three cost categories were comprehensively evaluated: (i)Direct medical costs (e.g., medical and drug costs).(ii)Direct non-medical costs (e.g., travel costs), and.(iii)Indirect costs (e.g., productivity losses due to illness or time spent on healthcare-related activities). (i)Direct medical costs (e.g., medical and drug costs). Direct medical costs (e.g., medical and drug costs). (ii)Direct non-medical costs (e.g., travel costs), and. Direct non-medical costs (e.g., travel costs), and. (iii)Indirect costs (e.g., productivity losses due to illness or time spent on healthcare-related activities). Indirect costs (e.g., productivity losses due to illness or time spent on healthcare-related activities). All costs were estimated as annual per-patient costs in USD, using an exchange rate of 1 USD = 150 JPY. In line with best practices for COI studies, sensitivity analyses were performed to evaluate the impact of uncertainty on key cost parameters. Data sources and study population This study estimated the components of COI using three primary data sources (Table 1). Table 1List of parametersCategory (mean, 95% CI)UnitSI groupNon-SI groupp-valueData sourceDirect medical costDrug costsUSD/month701 (689‒712)516 (503‒528)*<0.05.1)bDMARDsUSD/month694 (684‒705)499 (489‒508)*<0.05.1)Number of bDMARD prescriptionsUnits/month2.9 (2.9‒3.0)1.9 (1.9‒1.9)*<0.05.1)Other DMARDsUSD/month7 (5‒8)17 (14‒19)*<0.05.1)Medical service feesUSD/month73 (69–77)93 (86‒99)*<0.05.1)HospitalizationUSD/month16 (15‒17)28 (25‒30)*<0.05.1)ExaminationUSD/month22 (22‒23)23 (23‒23)*<0.05.1)Imaging examinationUSD/month13 (13‒14)17 (16‒18)*<0.05.1)RehabilitationUSD/month5 (5‒6)8 (7‒9)*<0.05.1)SurgeryUSD/month5 (4‒6)6 (4‒7)*<0.05.1)DispensingUSD/month11 (11‒11)11 (11‒11)*<0.05.1)Direct non-medical costTraveling costUSD/visit7 (4‒9)3 (2‒5)*<0.05.2)Indirect costAverage monthly wage (ages 30–60)USD/month1,708 (1,448‒1,967)N/A3)Average hourly wage (ages 30–60)USD/h10 (8‒11)N/A3)Productivity loss associated with outpatient visitsUSD/visit14 (13‒15)13 (12‒14)*<0.05.2) + 3)Round-trip travel time for outpatient visitsmin/visit38.1 (29.6‒46.6)27.9 (19.0‒36.8)*<0.05.2)Waiting time before consultationmin/visit39.1 (1.6‒76.5)N/A3)Duration of medical consultationmin/visit9.7 (0‒19.9)N/A3)Total patient time per outpatient visith/visit1.4 (0.5‒2.4)1.3 (0.3‒2.2)*<0.05.2) + 3)Monthly productivity loss due to activity impairmentUSD/month491 (399‒579)427 (195‒626)> 0.052) + 3)Activity Impairment%/month28.8 (23.3‒33.9)25.0 (11.4‒36.7)> 0.052)OtherAnnual frequency of hospital visitsvisit/year9.5 (9.4‒9.6)25.0 (24.5‒25.6)*<0.05.1)J-HAQ-0.61 (0.48‒0.74)0.63 (0.31‒0.92)> 0.052)Data sources:1): Health Insurance Claims Data2): Patient Survey3): Official Statistical Data*A p-value <0.05 indicates a statistically significant difference between the SI and non-SI groupsAbbreviations: bDMARDs biological disease-modifying antirheumatic drugs, SI self-injection, CI confidence interval, N/A not applicable, J-HAQ Japanese Health Assessment Questionnaire Table 1List of parametersCategory (mean, 95% CI)UnitSI groupNon-SI groupp-valueData sourceDirect medical costDrug costsUSD/month701 (689‒712)516 (503‒528)*<0.05.1)bDMARDsUSD/month694 (684‒705)499 (489‒508)*<0.05.1)Number of bDMARD prescriptionsUnits/month2.9 (2.9‒3.0)1.9 (1.9‒1.9)*<0.05.1)Other DMARDsUSD/month7 (5‒8)17 (14‒19)*<0.05.1)Medical service feesUSD/month73 (69–77)93 (86‒99)*<0.05.1)HospitalizationUSD/month16 (15‒17)28 (25‒30)*<0.05.1)ExaminationUSD/month22 (22‒23)23 (23‒23)*<0.05.1)Imaging examinationUSD/month13 (13‒14)17 (16‒18)*<0.05.1)RehabilitationUSD/month5 (5‒6)8 (7‒9)*<0.05.1)SurgeryUSD/month5 (4‒6)6 (4‒7)*<0.05.1)DispensingUSD/month11 (11‒11)11 (11‒11)*<0.05.1)Direct non-medical costTraveling costUSD/visit7 (4‒9)3 (2‒5)*<0.05.2)Indirect costAverage monthly wage (ages 30–60)USD/month1,708 (1,448‒1,967)N/A3)Average hourly wage (ages 30–60)USD/h10 (8‒11)N/A3)Productivity loss associated with outpatient visitsUSD/visit14 (13‒15)13 (12‒14)*<0.05.2) + 3)Round-trip travel time for outpatient visitsmin/visit38.1 (29.6‒46.6)27.9 (19.0‒36.8)*<0.05.2)Waiting time before consultationmin/visit39.1 (1.6‒76.5)N/A3)Duration of medical consultationmin/visit9.7 (0‒19.9)N/A3)Total patient time per outpatient visith/visit1.4 (0.5‒2.4)1.3 (0.3‒2.2)*<0.05.2) + 3)Monthly productivity loss due to activity impairmentUSD/month491 (399‒579)427 (195‒626)> 0.052) + 3)Activity Impairment%/month28.8 (23.3‒33.9)25.0 (11.4‒36.7)> 0.052)OtherAnnual frequency of hospital visitsvisit/year9.5 (9.4‒9.6)25.0 (24.5‒25.6)*<0.05.1)J-HAQ-0.61 (0.48‒0.74)0.63 (0.31‒0.92)> 0.052)Data sources:1): Health Insurance Claims Data2): Patient Survey3): Official Statistical Data*A p-value <0.05 indicates a statistically significant difference between the SI and non-SI groupsAbbreviations: bDMARDs biological disease-modifying antirheumatic drugs, SI self-injection, CI confidence interval, N/A not applicable, J-HAQ Japanese Health Assessment Questionnaire List of parameters Data sources: 1): Health Insurance Claims Data 2): Patient Survey 3): Official Statistical Data *A p-value <0.05 indicates a statistically significant difference between the SI and non-SI groups Abbreviations: bDMARDs biological disease-modifying antirheumatic drugs, SI self-injection, CI confidence interval, N/A not applicable, J-HAQ Japanese Health Assessment Questionnaire (1) Health insurance claims data. (2) Patient reports. (3) Official statistics published by the Ministry of Health, Labour and Welfare (MHLW). Each dataset reflected different cost components—claims data for direct medical costs, survey data for direct non-medical and indirect costs, and national statistics for wage and price indices. These data sources were not directly merged but analyzed separately for each cost category. Integrated COI estimates were derived through a representative patient model that reflects the demographic and socioeconomic characteristics of Japanese patients with RA. Health insurance claims data Direct medical costs were estimated using administrative claims data recorded between April 2014 and August 2023, provided by DeSC Healthcare, Inc. (Tokyo, Japan) [18], and previously analyzed by the authors [12]. Propensity score matching was employed to mitigate potential selection bias inherent in the non-randomized, observational nature of our study. This approach ensured comparability between the SI and non-SI patient groups by balancing baseline demographic and clinical characteristics. Propensity scores were estimated using logistic regression, incorporating 25 covariates, including demographic, clinical, and administrative factors. Specifically, the model included sex, age, the interval between definitive RA diagnosis and initiation of the first bDMARD (representing treatment duration), type of first bDMARD used, type of insurance, dependent category, and comorbidities. The comorbidities comprised hypertension, osteoporosis, dyslipidemia, diabetes, and depression (defined using Anatomical Therapeutic Chemical codes); and heart failure, chronic pulmonary disease, cerebrovascular disease, renal disease, and dementia (defined using the International Classification of Diseases, 10th revision (ICD-10) codes based on the Charlson comorbidity criteria). These covariates were selected to capture patient-level differences in clinical background, treatment history, and socioeconomic characteristics that could influence treatment selection and cost structure. To prevent multiple entries for the same individual, all records were consolidated at the patient ID level to ensure one record per patient. One-to-one nearest-neighbor matching was applied with a caliper width of 0.1. Post-matching balance was evaluated using standardized mean differences (SMDs), with all covariates achieving SMD < 0.1, indicating satisfactory balance. A complete list of covariates and SMD results is presented in Supplementary Table S1. A total of 6,784 patients with RA were included, with 3,392 patients assigned to each group. Regarding the demographic characteristics, patients (mean age: 73.5 ± 11.7 years; female: 73.2%) were slightly older than the national RA population with prescriptions of disease-modifying antirheumatic drugs (DMARDs) [2, 19]. Patients were categorized into the SI and non-SI groups based on the presence of claims indicating the receipt of SI instructions. Costs included those for DMARDs (penicillamine, sodium aurothiomalate, auranofin, sulfasalazine, bucillamine, mizoribine, methotrexate, actarit, lobenzarit sodium, tacrolimus hydrate, leflunomide, iguratimod, bDMARDs [etanercept, etanercept biosimilar, tocilizumab, adalimumab, adalimumab biosimilar, abatacept, golimumab, certolizumab pegol, sarilumab, and ozoralizumab], tofacitinib citrate, baricitinib, and upadacitinib hemihydrate), hospitalization, diagnostic examinations, imaging, rehabilitation, surgical procedures (synovectomy, total hip arthroplasty, arthrodesis, and arthroplasty), and dispensing. The annual frequency of hospital visits and average monthly direct medical costs were calculated for each group, with unit costs (drug prices and medical service fees) adjusted using the 2024 revision of the Japanese National Health Insurance fee schedule [20]. Patient-reported survey To estimate direct non-medical and indirect costs, a web-based survey was conducted among Japanese patients with RA using bDMARDs, as previously reported [11]. Participants were recruited from a patient panel maintained by Medilead, Inc. (Tokyo, Japan) [21–23]. A total of 79 valid responses were analyzed. The survey collected data on travel costs, round-trip travel time to outpatient visits (Supplementary 1 in Online Resource 1), activity impairment, and the Japanese Health Assessment Questionnaire (J-HAQ) [24]. Patients were categorized into two groups based on their self-reported injection method: the SI group and the non-SI group. The demographic characteristics of respondents (mean age: 60.2 ± 12.2 years; female: 64.6%) were consistent with those of the national RA population [2, 19]. Official statistical data Official statistics published by the MHLW were used to support the cost estimation. Wage data were obtained from the 2023 Basic Survey on Wage Structure [25] and the Monthly Labor Survey. Outpatient visit times (including waiting and consultation times) were estimated using the 2020 Patient Survey data [26] (Supplementary 2 and 3 in Online Resource 1). According to the MHLW definition, the Basic Survey on Wage Structure covers employees working in major industries and excludes self-employed individuals and unpaid family workers. Therefore, the estimation of indirect costs in this study reflects productivity loss among employed workers only. This definition was considered appropriate because the indirect cost component was intended to represent labor-related economic impacts within the general employed population. Representative patient model and wage estimation A representative patient model was constructed to reflect demographic and socioeconomic characteristics of working-age patients with RA in Japan. In this study, the patient questionnaire data were not directly linked to the individual-level claims data; rather, both were treated as independent data sources. The estimated values of direct non-medical costs and indirect costs derived from the questionnaire were incorporated into the representative patient model to ensure consistency with the demographic and socioeconomic characteristics of Japanese patients with RA. Specifically, wage- and age-related parameters were weighted according to the age distribution (30–60 years) of working-age patients with RA reported in a national epidemiological study [2, 19]. This approach enabled integrated cost estimation representing the target population, while avoiding any individual-level data linkage or potential double counting. Patients in their 70 s, comprising the largest age group (28.6%), were excluded from the productivity loss calculations owing to retirement. The model focused on patients in their 30 s to 60 s, and wage-related parameters were estimated using weighted averages based on age distributions [2, 19]. Accordingly, the weighted average monthly wage was estimated at 1,708 USD (95% confidence interval [CI]: 1,448–1,967), and the wage growth rate was 1.74% (95% CI: 1.30%–2.18%). The mean weighted age was 57.1 years. These values were used to estimate hourly wages for productivity loss and conduct sensitivity analyses (Supplementary 4 in Online Resource 1). Annual COI calculation formula The annual COI was estimated using the following formula: Annual COI = (Direct Medical Costs [USD/month] × 12) + (Traveling Costs × Number of Visits[visit/year]) + (Time per Visit × Hourly Wage × Number of Visits [visit/year]) + (Activity Impairment Loss [USD/month] × 12). Indirect costs were applied only to employed individuals. The analysis was conducted across four subgroups based on SI and employment status. SI group × employed (SI-employed).SI group × unemployed (SI-unemployed).Non-SI group × employed (non-SI-employed).Non-SI group × unemployed (non-SI-unemployed). SI group × employed (SI-employed). SI group × employed (SI-employed). SI group × unemployed (SI-unemployed). SI group × unemployed (SI-unemployed). Non-SI group × employed (non-SI-employed). Non-SI group × employed (non-SI-employed). Non-SI group × unemployed (non-SI-unemployed). Non-SI group × unemployed (non-SI-unemployed). Scenario analysis To estimate long-term trends in medical and societal costs, we conducted a 10-year cost projection using the following Japan-specific macroeconomic assumptions: - Medical fee revision rate: +0.61% (every 2 years).- Drug price revision rate: −1.06% (every 2 years). - Medical fee revision rate: +0.61% (every 2 years). - Medical fee revision rate: +0.61% (every 2 years). - Drug price revision rate: −1.06% (every 2 years). - Drug price revision rate: −1.06% (every 2 years). Consumer price index (CPI) growth rate: +1.63% per year. - Wage growth rate: +1.74% per year. In this projection, disease progression, treatment regimen, and employment status were assumed to remain constant over time, as the model aimed to isolate the effects of macroeconomic variables on long-term cost trends rather than simulate clinical or behavioral changes. Using these parameters, we simulated the 10-year trends in direct medical costs and total COI for both the SI and non-SI groups (Supplementary 5‒8 in Online Resource 1). The 10-year time horizon was considered appropriate based on the weighted average age of 57.1 years in the representative patient with RA model used in this study (Supplementary 4 in Online Resource 1), reflecting a reasonable remaining work-life expectancy. Based on the conceptual framework proposed by Getzen (2000, 2013) [27, 28] and the Society of Actuaries (2021) [29], this study referred to a multilevel modeling approach to long-term healthcare cost growth. However, unlike traditional applications that incorporate general inflation (CPI) across all categories, this study did not apply CPI adjustments to medical service fees, drug prices, or wages. This decision was based on Japan’s official fee schedule and drug reimbursement prices, which are revised through centrally controlled processes that reflect macroeconomic conditions, including inflation expectations [30, 31]. Likewise, wage growth was modeled using the observed nominal wage trends from national labor statistics without CPI inflation overlays [25]. In contrast, CPI adjustments were applied only to direct non-medical costs, specifically travel costs, to reflect real-world inflation in transportation-related costs. Although the original model proposed by Getzen included a gross domestic product (GDP)-based saturation mechanism to constrain long-term healthcare cost growth, this element was excluded owing to the disease-specific focus of the study on treatment-level cost trajectories (i.e., SI of bDMARDs in RA) [32]. This approach aligns with the practical structure of Japan’s regulated health system and ensures internal consistency in the component-level COI simulation (Supplementary 9 in Online Resource 1). Sensitivity analysis To establish the robustness of the findings, one-way sensitivity analyses were conducted by varying the key cost parameters within 95% CIs. The parameters were as follows: drug costs, medical service fees, travel costs, productivity loss associated with outpatient visits, and productivity loss due to impaired activity. drug costs, medical service fees, travel costs, productivity loss associated with outpatient visits, and productivity loss due to impaired activity. drug costs, medical service fees, travel costs, productivity loss associated with outpatient visits, and productivity loss due to impaired activity. The impact of each parameter on total COI was visualized using a tornado diagram to identify the most influential cost drivers. Scenario sensitivity analysis To assess the robustness of our base-case projections, a scenario-based sensitivity analysis was performed using both pessimistic and optimistic macroeconomic assumptions (Supplementary 9 in Online Resource 1). The pessimistic scenario assumed no inflation (CPI: 0%), no scheduled increases in medical service fees (0%), a continued decline in drug prices (− 2%, every 2 years), and stagnant wages (0%).The optimistic scenario assumed accelerated inflation (CPI: +3%), scheduled increases in medical service fees (+ 3% every two years), stable drug prices (0%), and a wage growth of + 3%. The pessimistic scenario assumed no inflation (CPI: 0%), no scheduled increases in medical service fees (0%), a continued decline in drug prices (− 2%, every 2 years), and stagnant wages (0%). The pessimistic scenario assumed no inflation (CPI: 0%), no scheduled increases in medical service fees (0%), a continued decline in drug prices (− 2%, every 2 years), and stagnant wages (0%). The optimistic scenario assumed accelerated inflation (CPI: +3%), scheduled increases in medical service fees (+ 3% every two years), stable drug prices (0%), and a wage growth of + 3%. The optimistic scenario assumed accelerated inflation (CPI: +3%), scheduled increases in medical service fees (+ 3% every two years), stable drug prices (0%), and a wage growth of + 3%. These alternative scenarios were applied uniformly over a 10-year projection period to simulate the range of potential cost outcomes under varying macroeconomic conditions. Statistical analysis All cost-related outcomes are reported as mean values with 95% CIs. Depending on the data distribution, CIs were calculated using non-parametric bootstrap procedures with 10,000 replicates. Between-group comparisons were performed using both the 95% CI overlap and p-values < 0.05, which were considered statistically significant. Data analyses were performed using the R software version 3.2.2 (R Foundation for Statistical Computing, Vienna, Austria), SAS version 9.4 (SAS Institute Inc., Cary, NC, USA), and Microsoft SQL Server 2022 (Microsoft Corporation, Redmond, WA, USA). Results Comparative costs and outcomes between the SI and Non-SI groups (Table 1) Direct medical costs The total drug cost was significantly higher in the SI group (701 USD/month [95% CI: 689–712]) than in the non-SI group (516 USD/month [95% CI: 503–528]) (p < 0.05). This difference was largely attributable to the cost of bDMARDs, which amounted to 694 USD/month (95% CI: 684–705) in the SI group and 499 USD/month (95% CI: 489–508) in the non-SI group (p < 0.05). The average number of bDMARD prescriptions per month was also higher in the SI group (2.9 vs. 1.9, p < 0.05), whereas the cost of other DMARDs was significantly lower in the SI group (7 USD vs. 17 USD, p < 0.05). Medical service fees were lower in the SI group (73 USD/month [95% CI: 69–77]) than in the non-SI group (93 USD/month [95% CI: 86–99]) (p < 0.05). Specifically, the SI group had significantly lower costs for hospitalization (16 USD vs. 28 USD), imaging examinations (13 USD vs. 17 USD), and rehabilitation (5 USD vs. 8 USD) (p < 0.05). Direct non-medical costs (Traveling costs) Traveling costs per outpatient visit were significantly higher in the SI group (7 USD [95% CI: 4–9]) than in the non-SI group (3 USD [95% CI: 2–5]) (p < 0.05). Indirect costs The productivity loss per outpatient visit was slightly higher in the SI group (14 USD [95% CI: 13–15]) than in the non-SI group (13 USD [95% CI: 12–14]) (p < 0.05). Furthermore, the average round-trip travel time was significantly longer in the SI group (38.1 min [95% CI: 29.6–46.6]) than in the non-SI group (27.9 min [95% CI: 19.0–36.8]) (p < 0.05). Total patient time per outpatient visit, including travel, waiting, and consultation, was 1.4 h (95% CI: 0.5–2.4) in the SI group and 1.3 h (95% CI: 0.3–2.2) in the non-SI group (p < 0.05). The monthly productivity loss due to activity impairment did not differ significantly between the groups (491 USD [95% CI: 399–579 USD] in the SI group vs. 427 USD [95% CI: 195–626 USD] in the non-SI group; p > 0.05). Likewise, activity impairment was comparable between groups (28.8% [95% CI: 23.3–33.9] vs. 25.0% [95% CI: 11.4–36.7]; p > 0.05). Other measures The annual frequency of hospital visits was significantly lower in the SI group (9.5 visits/year [95% CI: 9.4–9.6]) than in the non-SI group (25.0 visits/year [95% CI: 24.5–25.6]) (p < 0.05). The mean J-HAQ score, which reflects functional disability, was comparable between groups (0.61 [95% CI: 0.48–0.74] in the SI group vs. 0.63 [95% CI: 0.31–0.92] in the non-SI group; p > 0.05). COI comparison between SI and non-SI groups (base year analysis) (Table 2) Direct medical costs As shown in Table 2, the annual direct medical costs were higher in the SI group (9,280 USD [95% CI: 9,096–9,442]) than in the non-SI group (7,298 USD [95% CI: 7,072–7,448]).Table 2Annual medical costs for patients with rheumatoid arthritis (mean and 95% confidence interval)Category (mean, 95% CI)UnitSI groupNon-SI groupDirect medical costAnnual direct medical costsUSD/year9,280(9,096‒9,442)7,298(7,072‒7,448)Drug costsUSD/year8,410(8,273‒8,545)6,187(6,039‒6,332)Medical service feesUSD/year870(823‒927)1,111(1,033‒1,186)Direct non-medical costsTraveling costUSD/year64(37‒87)86(51‒117)Indirect costsAnnual productivity loss associated with outpatient visitsUSD/year 136(124–147)315(284–645)Annual productivity loss due to activity impairmentUSD/year 5,896(4,782–6,948)5,123(2,342–7,514)Values are presented as mean (95% confidence interval)Abbreviations: SI self-injection, CI confidence interval Table 2Annual medical costs for patients with rheumatoid arthritis (mean and 95% confidence interval)Category (mean, 95% CI)UnitSI groupNon-SI groupDirect medical costAnnual direct medical costsUSD/year9,280(9,096‒9,442)7,298(7,072‒7,448)Drug costsUSD/year8,410(8,273‒8,545)6,187(6,039‒6,332)Medical service feesUSD/year870(823‒927)1,111(1,033‒1,186)Direct non-medical costsTraveling costUSD/year64(37‒87)86(51‒117)Indirect costsAnnual productivity loss associated with outpatient visitsUSD/year 136(124–147)315(284–645)Annual productivity loss due to activity impairmentUSD/year 5,896(4,782–6,948)5,123(2,342–7,514)Values are presented as mean (95% confidence interval)Abbreviations: SI self-injection, CI confidence interval Annual medical costs for patients with rheumatoid arthritis (mean and 95% confidence interval) Values are presented as mean (95% confidence interval) Abbreviations: SI self-injection, CI confidence interval This difference was primarily driven by higher bDMARD costs in the SI group than in the non-SI group (8,410 USD vs. 6,187 USD, respectively). Conversely, the non-SI group had higher consultation fees than the SI group (1,111 USD vs. 870 USD, respectively), primarily due to more frequent outpatient visits. Direct non-medical costs (Traveling costs) Traveling costs were lower in the SI group (64 USD [95% CI: 37–87]) than in the non-SI group (86 USD [95% CI: 51–117]), reflecting fewer outpatient visits (9.5 vs. 25.0 visits/year, respectively). Indirect costs (Time loss and activity impairment) The annual cost of productivity loss due to outpatient visit time was lower in the SI group (136 USD [95% CI: 124–147]) than in the non-SI group (315 USD [95% CI: 284–645]). In contrast, productivity loss due to activity impairment was slightly higher in the SI group (5,896 USD [95% CI: 4,782–6,948]) than in the non-SI group (5,123 USD [95% CI: 2,342–7,514]), although the difference was statistically non-significant. Comparison of total annual COI (Fig. 1) Summing all cost categories, the annual total COI per employed patient was higher in the SI group (15,376 USD) than in the non-SI group (12,822 USD), with a difference of 2,554 USD. Fig. 1Breakdown of annual cost-of-illness Note: The figure illustrates the annual cost-of-illness (COI) per patient divided into drug costs, medical service fees, productivity losses (due to activity impairment and outpatient visits), and travel costs. Each bar represents the mean total cost for each subgroup Breakdown of annual cost-of-illness Note: The figure illustrates the annual cost-of-illness (COI) per patient divided into drug costs, medical service fees, productivity losses (due to activity impairment and outpatient visits), and travel costs. Each bar represents the mean total cost for each subgroup Among unemployed patients (assuming zero indirect costs), the total COI was 9,344 USD in the SI group and 7,383 USD in the non-SI group, with a difference of 1,961 USD. Results of sensitivity analysis (Fig. 2) As illustrated in the tornado diagram (Fig. 2), productivity loss due to activity impairment was identified as the most influential factor in the non-SI group, potentially leading to an annual COI variation of up to 2,586 USD within the 95% CI range.The second most influential factor in the SI group was activity impairment (variation: 1,083 USD).Fig. 2Tornado diagram Note: Tornado diagram showing the contribution of each cost component to the annual cost difference between self-injection (SI) and non-SI groups Tornado diagram Note: Tornado diagram showing the contribution of each cost component to the annual cost difference between self-injection (SI) and non-SI groups Other factors, such as outpatient-related productivity loss, drug costs, consultation fees, and travel costs, had minimal impact (variation: <150 USD), supporting the robustness of the findings. Ten-year cost projections Base scenario (Fig. 3) Among employed individuals, COI increased over the 10-year period in both the SI and non-SI groups. In particular, the SI-employed group had the highest annual COI, increasing from approximately 15,000 USD in year 1 to over 16,000 USD by year 10, primarily due to the increased productivity loss resulting from activity impairment. Although COI also increased in the non-SI-employed group, it remained consistently lower than that in the SI group throughout the observation period.Fig. 3Comparison of annual cost-of-illness (initial year vs. year 10; base scenario) Note: Comparison of the annual cost-of-illness (COI) per patient in self-injection (SI) and non-SI groups by employment status at baseline and in year 10 (base scenario). The base scenario assumes average macroeconomic conditions, including medical service fee revision (+ 0.61% every 2 years), drug price revision (− 1.06% every 2 years), consumer price index (CPI) growth (+ 1.63% annually), and wage growth (+ 1.74% annually) Comparison of annual cost-of-illness (initial year vs. year 10; base scenario) Note: Comparison of the annual cost-of-illness (COI) per patient in self-injection (SI) and non-SI groups by employment status at baseline and in year 10 (base scenario). The base scenario assumes average macroeconomic conditions, including medical service fee revision (+ 0.61% every 2 years), drug price revision (− 1.06% every 2 years), consumer price index (CPI) growth (+ 1.63% annually), and wage growth (+ 1.74% annually) In contrast, unemployed individuals did not incur productivity loss, resulting in a lower COI compared with employed individuals. Their COI primarily comprised drug and medical service fees. In both the SI and non-SI-unemployed groups, COI declined slightly over the 10-year period, with the non-SI-unemployed group maintaining the lowest annual COI among all groups. Scenario sensitivity analysis (Fig. 4) In the pessimistic scenario, a declining trend was observed across all groups, primarily owing to the effects of drug price revisions. In contrast, under the optimistic scenario, the annual COI increased across all subgroups over the 10-year period. In particular, among employed individuals in the SI group, COI increased from approximately 15,000 USD in year 1 to approximately 18,000 USD in year 10, representing the largest increase among all groups. This increase was primarily driven by rising productivity losses due to activity impairment and higher medical service fees. In the unemployed groups, where productivity loss was not incurred, the overall COI levels remained substantially lower than in employed individuals. Although a slight upward trend was observed, the values remained almost constant across the study period. Among all the subgroups, the non-SI-unemployed group consistently showed the lowest COI over the 10-year period.Fig. 4Comparison of annual cost-of-illness (year 1 vs. year 10; pessimistic and optimistic scenarios) Note: Comparison of the annual cost-of-illness (COI) per patient in self-injection (SI) and non-SI groups by employment status at baseline and year 10 under the pessimistic and optimistic scenarios The pessimistic scenario assumes average macroeconomic conditions, including medical service fee revision (0%), drug price revision (− 2% every 2 years), consumer price index (CPI) growth (0%), and wage growth (0%) The optimistic scenario assumes average macroeconomic conditions, including medical service fee revisions (+ 3% every 2 years), drug price revisions (0%), consumer price index (CPI) growth (+ 3% annually), and wage growth (+ 3% annually) Comparison of annual cost-of-illness (year 1 vs. year 10; pessimistic and optimistic scenarios) Note: Comparison of the annual cost-of-illness (COI) per patient in self-injection (SI) and non-SI groups by employment status at baseline and year 10 under the pessimistic and optimistic scenarios The pessimistic scenario assumes average macroeconomic conditions, including medical service fee revision (0%), drug price revision (− 2% every 2 years), consumer price index (CPI) growth (0%), and wage growth (0%) The optimistic scenario assumes average macroeconomic conditions, including medical service fee revisions (+ 3% every 2 years), drug price revisions (0%), consumer price index (CPI) growth (+ 3% annually), and wage growth (+ 3% annually) Discussion The current study is one of the few comprehensive evaluations of societal COI associated with SI versus non-SI bDMARD use in Japanese patients with RA. By integrating health insurance claims data, patient-reported surveys, and official statistical data, we examined the cost structure differences between these treatment delivery methods from a societal perspective. Our findings revealed that SI was associated with higher direct medical costs, primarily because of the increased cost of bDMARDs. Although fewer hospital visits helped lower some non-medical and indirect costs, these savings did not fully offset the higher medication expenses, leading to a slightly greater COI in the SI group. Importantly, our study did not assess clinical outcomes such as adherence, treatment satisfaction, or quality of life, which have been previously identified as potential benefits of SI [33–36]. As such, this analysis remains descriptive and does not address the overall value or cost-effectiveness of SI. Further research integrating clinical outcomes with economic data is essential to comprehensively assess the comparative value of SI and non-SIs. Based on the results of the sensitivity analysis, a key finding of this study is that productivity loss due to activity impairment accounted for the largest proportion of the total societal costs in both the SI and non-SI groups. This finding aligns with previous research demonstrating that indirect costs may be equivalent to or exceed direct medical costs in the management of RA [37]. Therefore, regardless of SI, these findings support the notion that maintaining adequate disease control in patients with RA is the most crucial factor in preventing decline in work productivity. In Japan, extending the working life of older adults has emerged as an important societal and policy objective [38, 39]. In our study, productivity loss accounted for a large proportion of the total societal costs, suggesting that work-related outcomes are particularly relevant in chronic diseases such as RA. Self-injectable biologics may contribute to improved disease control and physical functioning [7, 40], and these improvements may, in turn, enable patients to remain employed for prolonged periods [41]. As access to and acceptance of such treatment options continue to increase, they may help reduce productivity loss and support labor force participation among individuals with RA. Therefore, treatment strategies that promote disease control and self-management may positively impact clinical outcomes while reducing the broader economic burden. However, additional studies are required to confirm these benefits. The 10-year cost projections further indicated that SI may reduce reliance on medical institutions, but could result in higher COI, especially for employed individuals. Variations in wage growth substantially impact indirect costs, suggesting that macroeconomic assumptions play a critical role in estimating the long-term societal burden. While policy interventions, such as workplace support, may help mitigate productivity loss, their effectiveness was not assessed in this study and warrants further investigation. Although the current analysis focuses on Japan, the findings may be relevant to other Asian countries encountering similar demographic and healthcare challenges. Nevertheless, the generalizability of our findings should be confirmed through context-specific studies. This study has several limitations that should be considered when interpreting the findings. First, productivity loss was estimated using self-reported activity impairment scores as proxies for presenteeism. Although commonly applied in health economics research, this method may not accurately reflect actual productivity loss or absenteeism and is subject to reporting bias. Second, the cross-sectional design and relatively small sample size of the non-SI group may have limited the precision of some cost estimates, resulting in wide CIs. Third, although the claims-based cohort (mean age: 73.5 years) was older than the survey respondents (mean age: 60.2 years), both datasets included patients receiving treatment with bDMARDs. According to the 2024 Clinical Practice Guideline for Rheumatoid Arthritis issued by the Japan College of Rheumatology [42], bDMARDs are generally recommended for patients who persistently exhibit disease activity despite adequate treatment with conventional synthetic DMARDs. Furthermore, large-scale Japanese database studies have shown that the prescription rate of bDMARDs among older patients is nearly equivalent to that among younger patients [2]. Based on these findings, it is unlikely that substantial differences in disease activity or functional status exist between the two samples. Consequently, the impact of differences in age distribution on the estimation of direct and indirect costs was considered limited. Fourth, although propensity score matching was performed to adjust for observable differences in the claims analysis, residual confounding may persist owing to the voluntary nature of SI selection (e.g., disease severity, self-management ability, and treatment preferences). Fifth, certain cost components, such as traveling and time costs, were approximated using average values from surveys and official data, which may fail to capture real-world heterogeneity. Sixth, indirect costs were estimated using wage data from the Basic Survey on Wage Structure, which covers salaried employees but excludes self-employed individuals and unpaid family workers. Therefore, productivity losses related to unpaid or non-salaried activities, such as caregiving and household work among retirees or homemakers, may have been underestimated. Seventh, clinical outcomes such as disease activity, functional status, or quality-adjusted life years were not assessed, limiting the ability to evaluate cost-effectiveness. Eighth, the assumptions used in the 10-year projection model, including future changes in fees, drug prices, wages, and inflation, are based on past trends and may not reflect future conditions. Ninth, regional differences in access to healthcare facilities may influence the feasibility and selection of SI, particularly in rural and underserved areas. However, this study did not assess the impact of geographical factors or residence-based accessibility on treatment selection, which may limit the generalizability of our findings across diverse settings. Finally, although our model-based approach provided a useful framework for evaluating societal costs from a population-level perspective, it did not fully capture the complexities experienced by older patients with RA. In real-world settings, advanced age is frequently accompanied by multiple comorbidities, cognitive impairment, and reduced self-management ability, which can limit the feasibility of using bDMARD or SI. Future research incorporating age-stratified analyses and patient-centered factors is warranted to better reflect the heterogeneity of real-world treatment experiences, particularly among older adults. These limitations suggest that the results should be interpreted with caution and underscore the need for further research to evaluate the long-term economic and clinical impacts of SI. Conclusion Collectively, this study offers descriptive evidence of the societal COI differences between SI and non-SI of bDMARDs in Japanese patients with RA. Although SI was associated with higher bDMARD-related costs, reductions in outpatient-related burdens partially offset the direct non-medical and indirect costs. Across both groups, productivity loss due to activity impairment was the largest contributor to the total COI, highlighting the importance of considering labor-related outcomes in economic evaluations of RA management. These findings may inform healthcare planning and contribute to future discussions on optimal treatment options for RA using biologics. Nevertheless, additional studies incorporating clinical outcomes, treatment adherence, and cost-effectiveness indicators are needed to support more comprehensive evidence-based decisions regarding the value of SI. Supplementary Information Supplementary Material 1. Supplementary Material 1.