A retrospective medical-record study at one Thai tertiary teaching hospital identified 730 ICD-10-flagged adverse-drug-reaction admissions among 29,927 admissions and purposively reviewed 200. Within that sample, 66.0% were judged preventable and 67.1% had probable or definite drug causality; warfarin was the most frequently suspected drug. The single-centre retrospective design and code-based case ascertainment make these associative estimates, not national incidence or causal evidence.
Key findings
- Among 29,927 admissions, 730 were code-flagged and 200 were reviewed in detail; the screening count must not be conflated with the analytic sample.
- Within the reviewed sample, 67.1% were rated probable or definite for drug causality and 66.0% preventable; this does not mean that 66% of all hospital admissions were preventable ADR admissions.
- Cardiovascular disorders represented 50.5% and nervous-system disorders 17.0% of reviewed reactions. Warfarin was the leading suspected drug at 38.0%, followed by paracetamol at 5.0%.
- Suspected medicines were discontinued in 82.5% of cases, whereas discharge counselling was documented in 56.0%, indicating a process gap that merits prospective evaluation.
- Ages 36–65 and over 65 were associated with higher odds of documented ADR admission than the reference group, but wide confidence intervals preclude causal interpretation.
Why this matters globally
ADR-related admission is a global medication-safety problem. The study highlights potentially actionable processes—dose optimisation, warfarin monitoring and discharge counselling—but one hospital's records do not show that these measures reduce admissions or are cost-effective elsewhere. Multicentre surveillance and controlled implementation studies are needed.
Thai researcher contribution
Researchers from Khon Kaen University and Praboromarajchanok Institute brought Thai hospital records into the evidence base on ADR causality, preventability and management, adding locally relevant patient, medicine and service-system context to a field often dominated by high-income-country data.
Limitations to consider
This was a retrospective cross-sectional review from one tertiary hospital. ICD-10 ascertainment can miss uncoded events or include misclassified cases; purposively selecting 200 of 730 flags limits overall prevalence estimates, and causality and preventability ratings require reviewer judgement. The dosage-related odds ratio depends on model coding not explained in the abstract, so its direction should not guide policy alone.