Taiwan
NHIRD
Population-based national claims environment used to develop the source model and estimate observed treatment effects.
Home / Accomplishment / Transportability
Case 03 · Methodological extension
Evidence travels from Taiwan to Japan
A proof-of-concept study compares observed estimates in Taiwan NHIRD and Japan DeSC with estimates transported by a model developed in Taiwan.
- Design
- Retrospective target-trial emulation
- Settings
- Taiwan and Japan
- Comparisons
- Three antiplatelet trials
Study aim
A feasibility test for cross-country evidence transport
To evaluate the feasibility of transporting evidence across countries through a proof-of-concept study built on prior IMPRESIVE research, with the goal of informing healthcare decision-making.
Method
Cohort studies using a target-trial emulation framework
Hospitalized patients with acute coronary syndrome are studied separately in each database. The three head-to-head comparisons are prasugrel versus clopidogrel, ticagrelor versus clopidogrel, and ticagrelor versus prasugrel. Permitted aggregate results are then synthesized for cross-database interpretation.
Japan
DeSC
Population-based claims environment used to estimate the observed target-setting effect and evaluate the transported estimate.
| Database | Database start | Database end | Enrolment start | Enrolment end |
|---|---|---|---|---|
| Taiwan NHIRD | 2013-01-01 | 2022-12-31 | 2014-01-01 | 2021-12-31 |
| Japan DeSC | 2016-01-01 | 2023-08-31 | 2017-01-01 | 2022-08-31 |
Target-trial emulation
| Component | Target trial | NHIRD and DeSC emulation |
|---|---|---|
| Eligibility | Age ≥18; hospitalized with a primary diagnosis of acute coronary syndrome, including STEMI, NSTEMI, or unstable angina. Exclude contraindications or prior use of ticagrelor, prasugrel, or clopidogrel. | Same criteria; receipt of a study drug is treated as evidence that the clinical contraindication decision had been made. |
| Treatment strategies | Initiate prasugrel versus clopidogrel; ticagrelor versus clopidogrel; or ticagrelor versus prasugrel. | Same three head-to-head strategies. |
| Assignment | Eligible participants are randomly assigned with the same treatment-assignment probability within each comparison. | Propensity-score inverse-probability weighting is used to balance measured characteristics and approximate comparable treatment-assignment probabilities. |
| Outcomes | MACE; cardiovascular death; stroke; myocardial infarction; all-cause mortality; ACS recurrence; and bleeding events. | Same outcomes, using approved local definitions and code lists. |
| Follow-up | From treatment assignment until outcome, death, or end of study period, whichever occurs first. | Same follow-up rule. |
Transported estimates from NHIRD to DeSC
Switch treatment comparisons to inspect Taiwan estimates, Japan estimates, and estimates transported from the Taiwan model.
Taiwan → Japan
Observed and transported relative-risk estimates
Interpretation boundaries
Differences are the result to explain, not an error to hide
Absolute-risk heterogeneity
Baseline risk can differ substantially across healthcare contexts. Coding intensity, patient selection, and utilization patterns must be examined before transporting absolute estimates.
Relative-effect inconsistency
A relative measure is not automatically transportable. Directional differences may indicate that measured or unmeasured effect modifiers are distributed differently.
Effect-modifier discrepancies
Comorbidities can have different interaction patterns, and rare outcomes can produce wide uncertainty. Calibration must be evaluated outcome by outcome.
Return to the expanding case portfolio
Compare this methodological extension with the established ASCVD and AD/PN cases.