Project Details
Title:
NORA (Healthcare)
Details:
Adverse drug reactions (ADRs): harmful, unintended responses to medication are a leading cause of preventable hospitalization worldwide.
In Ghana, the national pharmacovigilance system relies on the Yellow Card scheme, which depends entirely on healthcare workers manually identifying and reporting suspected reactions. This process is slow and inconsistent and captures only a fraction of true ADR signals, leaving dangerous drug safety gaps undetected across the country.
The problem is compounded by a technology gap. Existing biomedical NLP tools are trained exclusively on North American and European clinical text. They fail on Ghanaian clinical writing, which blends formal medical register with Ghanaian English idioms, pidgin constructions, and regulatory language specific to Ghana FDA publications. No NLP system has ever been built specifically for Ghanaian pharmacovigilance data.
The consequence is a blind spot: ADR signals buried in ward notes, clinic reports, and patient interviews go undetected not because the data does not exist but because no tool can read it. For a country with a growing pharmaceutical market and limited pharmacovigilance capacity, this represents a direct and measurable risk to patient safety
NORA has four core goals: Automate ADR detection: Classify Ghanaian clinical text as containing an adverse drug reaction or not, reducing reliance on manual review by pharmacovigilance officers. Extract structured drug safety entities: Identify and label drugs, reactions, severity indicators, and patient demographics from unstructured clinical text to support downstream reporting. Handle Ghanaian-specific language: Reliably process the blend of formal medical register, Ghanaian English idioms, and Pidgin constructions that cause generic biomedical NLP tools to fail. Reduce documentation burden: Pre-populate Yellow Card-style structured reports from NER output, so pharmacovigilance officers review and confirm rather than write from scratch. Together, these goals move Ghana's ADR reporting from a slow, manual, under-resourced process toward an automated signal detection pipeline built specifically for Ghanaian clinical realities.