Nigeria Outbreak Analysis Model
AutoScientist Challenge 2026 | Healthcare Category
Author: Hussein Adeiza (mabera)
Role: Licensed Environmental Health Officer, Abuja Nigeria
Base Model: GPT-OSS 20B
Fine-tuned with: AutoScientist by Adaption Labs
Model Description
This is a LoRA adapter fine-tuned to interpret raw weekly epidemiological surveillance indicators and produce structured analytical reasoning, in the style of a surveillance epidemiologist reviewing a situation report. Unlike question-and-answer formats, this model takes raw statistics as input and produces trend analysis, operational flags and historical context as output.
Training Data
- Source: Official NCDC Lassa Fever Situation Reports (2023, 2025, 2026), with corroborating citations from WHO, NETEC, Guardian Nigeria and Vanguard Nigeria
- Dataset: 5 prompt-completion pairs expanded via Adaptive Data
- Languages: English, Hausa, Yoruba
- Quality improvement: 34.3% (Grade B → A)
- Kaggle: https://www.kaggle.com/datasets/yunusahusseinadeiza/nigeria-lassa-fever-situation-report-interpreter
Training Metrics
- Win rate: 54% adapted vs 46% base model
- Base model: gpt-oss-20b
- Method: LoRA — House Special + Reasoning Traces + Hallucination mitigation
- Dataset quality: 7.0 → 9.4 (+34.3% improvement, Grade A)
Key Finding From Source Data
Nigeria's 2026 Lassa fever case fatality rate has been consistently 5 to 9 percentage points higher than the same reporting week in 2025, across every comparable period available in official NCDC reporting, indicating a genuine year-on-year deterioration in outcomes rather than reporting variance.
Why This Matters
Public health surveillance teams need to interpret weekly indicators quickly and accurately, not just retrieve facts. This model is trained to do that interpretive work directly, flagging operational concerns like healthcare worker infections and contextualizing fatality rates against historical baselines, while staying strictly within what the cited source data supports.
Important Note
This dataset describes a real, ongoing outbreak with real fatalities. All training data and model outputs are constrained to stay within what official NCDC figures and corroborating public health reporting actually support, with no speculative medical claims.
Credits
Powered by Adaptive Data — Adaption Labs
AutoScientist Challenge 2026 | Healthcare Category
Data: NCDC, WHO Disease Outbreak News, NETEC, Guardian Nigeria, Vanguard Nigeria