Clinical and Medicolegal Evidence Patterns in Granted and Rejected Medical Malpractice Claims in Indonesia: An Explainable Concept-Tagging Study

Abstract

Background. Medical malpractice judgments encode clinical standards, causation, professional evidence, and procedural constraints, but Indonesian decisions have rarely been analysed at the level of independent disputes.

Objective. To identify which judicially accepted or materially discussed clinical–medicolegal evidence patterns most clearly distinguished granted from rejected medical malpractice claims in a purposive corpus of Indonesian court decisions from 2006–2026.

Methods. This exploratory archival study purposively assembled 16 deduplicated medical malpractice disputes decided in Indonesia during 2006–2026. Eight prespecified, auditable concept tags captured judicially accepted or materially discussed clinical–medicolegal evidence. Granted and rejected claims were compared using risk differences, Haldane–Anscombe-corrected odds ratios (ORs), 95% confidence intervals (CIs), and two-sided Fisher exact tests; a sensitivity analysis excluded one low-confidence case.

Results. Six claims were granted and 10 rejected. Articulated breach linked to causation occurred in 6/6 versus 0/10 disputes (corrected OR 273.0, 95% CI 4.80–15515.67); adverse professional findings in 5/6 versus 0/10 (OR 77.0, 95% CI 2.67–2222.91); and organizational failure in 4/6 versus 0/10 (OR 37.8, 95% CI 1.49–956.48). Acceptance of a standard-of-care or medical-risk defence occurred in 0/6 versus 8/10 (OR 0.023, 95% CI 0.001–0.557). Severe harm was nearly universal (6/6 versus 9/10) and did not discriminate outcomes. Directions were unchanged after excluding the low-confidence case.

Conclusion. In this small purposive corpus, litigation outcomes aligned more closely with judicially accepted evidence connecting specific breach, causation, professional findings, and organizational responsibility than with injury severity alone. The findings are exploratory and support larger, independently coded studies rather than case prediction