Projects

  1. Children Investment Fund Foundation, Kenya

    Project: Impact Evaluation of Faya Sexual Reproductive Health Education Program in Kenya.

    Role: Research Assistant

    Date: March 2021 – December 2022.

    About the project

    Faya was a two-year Children’s Investment Fund Foundation (CIFF) investment targeting adolescents between the ages of 15 to 19 years old in Homa Bay, Kilifi, Mombasa, and Siaya counties. The program, implemented by Amref Health Africa (Amref) sought to increase access to quality SRH education among adolescents in Kenya and to link adolescents to health services through the delivery of comprehensive sexual education via a Life Skills Education (LSE) Toolkit. I was responsible for mapping the households, collecting data using ODK collect and conducting interviews for randomized control trial (RCT) and quasi-experimental design to measure the changes in key outcomes of interest and conducting the process evaluation.

  2. Kemri Welcome Trust

    Project Name: Efficacy of Pneumonia treatment guidelines in reducing mortality using comparative Machine Learning Models

    Period: 2019 - 2022

    Role: Research Analyst

    About the Role

    In this study, we evaluate the extent to which clinicians follow pneumonia treatment guidelines among children aged 2- 59 months without severe malnutrition and not ex posed to HIV in level one referral hospitals. The study employed a retrospective cross sectional study design, where data is collected from hospital records. It is based on data from a cluster randomized trial that was conducted in 12 Kenyan hospitals. The clini cal information network (CIN) data set was used. Random forest, logistic regression, and decision trees machine learning models were employed. The study found that 94% of children who were correctly diagnosed with severe pneumonia were more likely to die as comparedtothosechildrennotcorrectlydiagnosedwithseverepneumonia, holdingother factors constant. Also, 14.9% of children who were correctly diagnosed with non-severe pneumonia were morelikely to die as compared to those children not correctly diagnosed with non-severe pneumonia, holding other factors constant. On the other hand, correct administration of treatment for non-severe pneumonia reduced the chances of mortal ity by 32.4%, holding other factors constant. Machine learning techniques were used in validating these findings. The conclusion was that the results could be repetitive with approximately 70% accuracy.