Logo Lanfrica

afiadata/ppb-kenya-recalls

Domaine:

healthcare

Type de record:

dataset
Créateur:
afi
Hôte:
Structured data on medicine and medical device recalls and rapid alerts issued by the Pharmacy and Poisons Board (PPB) of Kenya. # PPB Kenya Recalls Dataset Structured data on medicine and medical device recalls and rapid alerts issued by the Pharmacy and Poisons Board (PPB) of Kenya, covering 2016 to 2025. ## Source All records originate from the PPB website (pharmacyboardkenya.org). Rapid alerts include both PPB-issued notices and WHO alerts circulated by PPB. The data was compiled and structured from publicly available recall notices and PDF documents. ## Files ### Structured data | File | Description | |------|-------------| | `recalls_combined.csv` | All recall records from 2016 to 2025, stacked from two source files | | `rapid_alerts.csv` | Rapid alerts and public notices from 2018 to 2025 | | `manifest_recalls.csv` | Maps each recall PDF to its corresponding record in `recalls_combined.csv` | | `manifest_rapid_alerts.csv` | Maps each rapid alert PDF to its corresponding record in `rapid_alerts.csv` | | `curation_log.csv` | Documents manual data corrections and records flagged for review | ### Full-text corpus | File | Description | |------|-------------| | `recalls_full_text.jsonl` | Extracted text from recall PDFs, one record per line | | `rapid_alerts_full_text.jsonl` | Extracted text from rapid alert PDFs, one record per line | Text was extracted using pdfplumber. Scanned PDFs were processed with Tesseract OCR at 300 DPI. Each record includes a `status` field: `ok` (digital text), `ocr_ok` (OCR), or `ocr_empty` (extraction failed). ### Source PDFs | Folder | Description | |--------|-------------| | `recall_pdfs/` | Official recall notice PDFs from PPB | | `rapid_alerts_pdfs/` | Rapid alert and public notice PDFs | PDF filenames follow the pattern `{year}_{seq}.{product_name}.pdf` where `seq` reflects the numbering used on the PPB website at time of download, which does not always correspond to the official `recall_ref`. ## Coverage - **Recalls**: 183 records, 2016–2025. Records before 2022 do not have an official `recall_ref` in the source data. - **Rapid alerts**: 62 …