Taiwan's National Health Insurance Research Database: A Silent Force Behind Global Changes in Clinical Practice and Guidelines

The Taiwan National Health Insurance Research Database (NHIRD) stands as one of the world's most comprehensive healthcare information systems. Emerging from Taiwan's universal National Health Insurance program in 1995, the NHIRD systematically captures claims data from beneficiaries who represent over 99.5% of the population. Since 2000, this repository has documented an extensive range of healthcare interactions, encompassing patient demographics, outpatient and inpatient claims, prescription records, and detailed medical facility information. Following its transition in 2016 to management by Taiwan's Ministry of Health and Welfare Data Science Centre, the NHIRD now supports advanced cross-database linkages, incorporating key clinical variables such as laboratory results, cancer staging, disability assessments, and socioeconomic indicators [1], representing a key strength when compared to other national health insurance databases (Table 1).

The NHIRD's strength lies in its well-validated data across a spectrum of health conditions, from common ailments to severe diseases including ischemic stroke, atrial fibrillation (AF), hypertension, diabetes, hyperlipidemia, and cancer. With demonstrated modest to high sensitivity and positive predictive values, this database serves as an invaluable resource for understanding disease burden, optimizing treatment strategies, and evaluating outcomes [1, 2].

To demonstrate the NHIRD's influence on clinical practice, we looked into its contributions to the 2024 European Society of Cardiology (ESC) Guidelines for Atrial Fibrillation (AF) Management. Of the 1248 references screened from the guideline, approximately 1% (12 studies) originated from NHIRD analyses (Table 2), each providing significant insights into AF management strategies [15]. Several studies identified high-risk populations at elevated risk for thromboembolic events and stroke among Taiwanese patients [3, 4, 7, 8, 11]. A particularly influential study by Hsu et al. found that patients with concurrent hypertrophic cardiomyopathy (HCM) and AF experienced a high incidence of stroke, regardless of their CHA2DS2-VASc score. This finding directly influenced current practice, leading to a Class IB recommendation for non-vitamin K antagonist oral anticoagulants (NOACs) in HCM patients with AF [4, 15].

Further research enhanced our understanding of risk assessment strategies. Two studies demonstrated that dynamic monitoring of CHA2DS2-VASc and HAS-BLED scores provided superior risk prediction compared to static baseline measurements [5, 6]. This evidence supported the current Class IB recommendation for periodic reassessment of both thromboembolic and bleeding risks in AF management.

The database also yielded important insights into therapeutic approaches. Three studies investigating antidiabetic agents revealed potential protective effects against AF onset and related hospitalizations [12-14]. These findings influenced the guidelines' Class IIa recommendation to consider metformin or SGLT2 inhibitors for AF prevention in patients with type 2 diabetes mellitus [12, 13, 15]. Additional investigations expanded our knowledge of NOAC therapy, examining mortality outcomes in specific populations such as cancer patients [9], and evaluating their safety and efficacy in very elderly patients aged 90 years or older [10].

The NHIRD has proven valuable for pharmacoepidemiologic research and pharmacovigilance efforts. A landmark investigation by Lee et al. in 2015 examined the association between fluoroquinolone therapy and the risk of aortic complications through a nested case–control analysis of 1477 cases and 147,700 matched controls [16]. The study revealed increased risks of aortic aneurysm or dissection with both current use (rate ratio [RR], 2.43; 95% confidence interval [CI], 1.83–3.22) and past use (RR, 1.48; 95% CI, 1.18–1.86) of fluoroquinolones. These findings gained international validation when Pasternak et al. published corroborating evidence from Swedish national registers in 2018, demonstrating elevated risks of aortic complications with fluoroquinolone use compared to amoxicillin [17]. The cumulative evidence led the U.S. Food and Drug Administration to issue a safety warning that same year, advising against fluoroquinolone use in patients with or at risk for aortic aneurysm and recommending immediate discontinuation if suspicious symptoms develop [18].

Beyond drug safety, The NHIRD has also contributed significantly to our understanding of therapeutic benefits across various pharmacological interventions. Researchers have extensively studied medications including statins, metformin, and newer antidiabetic agents—such as thiazolidinediones, dipeptidyl peptidase-4 inhibitors, SGLT2 inhibitors, and GLP-1 receptor agonists—to evaluate their impact on diverse health outcomes [19]. A frequently cited investigation by Lee et al. examined metformin's potential protective effects against gastrointestinal cancers in a large Taiwanese cohort [20]. Their analysis revealed that diabetic patients not using anti-hyperglycemic medications showed at least double the cancer incidence compared to non-diabetic individuals, while metformin users demonstrated cancer rates similar to those without diabetes.

In addition, target trial emulation (TTE) has been increasingly embraced as a methodological framework for enhancing causal inference from observational data. A recent study by Yen et al. [21] applied sequential TTE in alignment with the two-step framework proposed by Hernán and colleagues [22]. Using NHIRD data, they compared the risks of major adverse kidney and cardiovascular events between patients with stage 5 chronic kidney disease, SGLT2 inhibitor users and nonusers. Importantly, the study followed two critical principles: synchronization of eligibility determination and treatment assignment at time zero, which avoided immortal time bias and maintained the comparability of treatment groups; and adequate confounding adjustment, which ensured that baseline differences between users and nonusers were addressed [21, 22]. By using these practices, the study shows how the target trial framework offers a methodical and disciplined procedure for operationalizing good practices in study design and data analysis, thereby enabling more credible and actionable causal inferences to be drawn from observational data [23].

Population-based studies provide high-quality evidence with important clinical implications, particularly for rare autoimmune diseases. Wei et al. found that patients with immune thrombocytopenic purpura (ITP) had a 26.8-fold increased risk of developing systemic lupus erythematosus (SLE) over 16 years [24]. Their findings suggest that ITP may represent an early manifestation of SLE, underscoring the need for timely rheumatology referral to facilitate early diagnosis before organ involvement occurs. Early recognition of evolving SLE can also inform treatment decisions, such as delaying splenectomy or initiating disease-modifying therapy.

Population-based research plays a critical role in elucidating disease progression, guiding earlier interventions, and ultimately improving long-term patient outcomes. As analytical techniques evolve and cross-database linkages expand, the NHIRD is poised to make further impactful contributions to global healthcare.

Writing – original draft preparation: Ting-Chun Tseng. Writing – review and editing: Ting-Chun Tseng and Renin Chang. Supervision: Jin-Shuen Chen and James Cheng-Chung Wei. All authors have read and agreed to the published version of the manuscript.

Renin Chang is currently the associate editor of International Journal of Rheumatic Diseases and had no role in the peer review process of or decision to publish this article. The other authors declared no conflicts of interest in writing this paper.