emergentai/PRESCRIP-TEC-uganda-cervical-cancer-screening

PRESCRIP-TEC Uganda Cervical Cancer Screening Dataset

Dataset Summary

This dataset provides anonymized records from the PRESCRIP-TEC initiative in Uganda. It focuses on cervical cancer screening, primarily through HPV self-testing, VIA (Visual Inspection with Acetic Acid), and related follow-ups. Data was collected via Village Health Teams (VHTs) among other clinical personal. The dataset includes demographic, socioeconomic, reproductive health, and screening outcome variables for 7,181 women. It has been preprocessed to remove personally identifiable information (PII) and columns with excessive missing values (>97%) to ensure privacy and usability.

PII dropped

-'fcm_name'

-'registration_date'

-'date_of_birth'

-'fcm_id'

-'beneficiary_name'

-'mobile_number_of_the_contact_person'

-'name_of_contact_person_of_mobile_number'

-'auto_generated_id'

-'outreach_number'

-'name_of_contact_person_of_mobile_number'

-'location_reference'

-'household_number'

Key themes:

• Household and demographic characteristics.

• Reproductive and health history (e.g., parity, contraception use, HIV status).

• Awareness and prior screening for cervical cancer.

• HPV swab collection and test results (including genotypes like HPV 16/18/45).

• Referrals and follow-up dates.

• Limitations: High missingness in some columns (e.g., referral-related fields >80% missing, as not all participants reached those phases). Data is from Uganda only; may not generalize.

The dataset Descriptions are based on the PRESCRIP-TEC variables list (UG-Updated-PRESCRIP variables.doc)

Data Splits

• Single split: train (full dataset, 7,181 rows). No validation/test splits provided; users can create their own.

• Preprocessing

• Removed PII: e.g., names, DOB, registration dates, mobile numbers, IDs.

• Dropped columns with 100% missing values (e.g., reasons for pelvic operations).

• Dropped columns with >97% missing values (e.g., AI assessment, complications).

• Result: 50 columns retained.

• No imputation; missing values preserved (use df.isna().mean() for details).

Citation Information

@dataset{prescrip_tec_uganda_2024,

  author = {PRESCRIP-TEC Project},

  title = {PRESCRIP-TEC Uganda Cervical Cancer Screening Dataset},

  year = {2024},

  url = {https://huggingface.co/datasets/Beijuka/PRESCRIP-TEC-uganda-cervical-cancer-screening}

}