JMIR Publications

ABSTRACT

Background:

The completeness of adverse event (AE) reports is crucial for assessing putative causal relationships. Little information is known about the underlying factors contributing to reports’ completeness. Malaysian reports surpass the global average vigiGrade completeness score (~0.44), approaching the well-documented benchmark (0.80) with a recent five-year average score of 0.79 as of 2019.

Objective:

This study aimed to employ a data-driven approach to explore the main drivers of the completeness of Malaysian AE reports in VigiBase over a recent 15-year period using vigiGrade. A secondary objective was to understand the actions taken by the Malaysian authorities that preceded the relatively high report completeness across different timeframes.

Methods:

We studied 132,738 Malaysian reports (2005-2019) recorded in VigiBase as of February 2021, split into INTDIS (63,943 in 2005-2016) and E2B (68,795 in 2015-2019) subsets. For machine learning (ML) analyses, we performed a two-stage feature selection followed by a random forest (RF) classifier algorithm to identify the top important features predicting well-documented Malaysian reports. We then examined the magnitude, prevalence, and direction of feature effects using a Shapley additive explanations (SHAP) algorithm for tree-based models (TreeExplainer). We also conducted time series analyses to study and summarise chronological trends and potential influences of key interventions on reporting quality.

Results:

Of all 132,738 reports studied, 42.8% (56,877) were well-documented, with 65.4% (53,929/82,497) well-documented since 2015. Over two-thirds (67.1%, 46,186) of the Malaysian E2B reports were well-documented, compared to 16.7% (10,691) of the historical INTDIS reports. For INTDIS reports, higher pharmacovigilance centre staffing was the primary feature positively associated with well-documentation. In recent E2B reports, the top important positive features included reaction abated upon drug dechallenge, reaction onset/drug use duration within 1 week, dosing interval within 1 day, reports from public specialist hospitals, reports by pharmacists, and reaction duration between 1-6 days. Contrarily, reports from product registration holders and other health care professionals, and reactions involving product substitution issues negatively affected the quality of E2B reports. Multifaceted strategies and interventions, comprising policy changes, continuity of education, and human resource development, laid the groundwork for AE reporting in Malaysia, while advancements in technological infrastructure, the pharmacovigilance database, and reporting tools concurred with increases in both the quantity and quality of AE reports.

Conclusions:

Our data-driven approach using interpretable machine learning methods has identified specific features that positively or negatively contribute to the completeness of Malaysian AE reports. Time series analyses further depicted how Malaysia has built up and strengthened its pharmacovigilance capacity over time through the adoption of multifaceted strategies and interventions to enhance AE reporting. These findings provide a foundation for future research and strategies to enhance pharmacovigilance practices, and, ultimately, public health outcomes.