Emily Koech
Research Subjects
Emily Koech has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Kristen Alyce Stafford
Center for International Health, Education and Biosecurity, Institute of Human Virology, School of Medicine, University of Maryland, 725 W Lombard St, Baltimore, MD, 21201, United States, 1 410-706-8614
2 shared publications - Abraham Katana
Division of Global HIV & TB, U.S. Centers for Disease Control and Prevention, Nairobi, Kenya
1 shared publication - Alash’le Abimiku
International Research Center of Excellence, Institute of Human Virology Nigeria, Abuja, Nigeria
1 shared publication - Anthony K. Ngugi
Department of Population Health, Aga Khan University Nairobi, Nairobi, Kenya
1 shared publication Caroline Ngunu
Nairobi County Government, Nairobi, Kenya
1 shared publication- Christina Mwachari
Department of Laboratory, Maryland Global Initiatives Corporation (MGIC), Nairobi
1 shared publication
Latest Publications
- Open Access
- 4 February 2026
- Published by JMIR Publications Inc. in JMIR Medical Informatics
- Vol. 14 (1) , e78964
Prediction of First and Multiple Antiretroviral Therapy Interruptions in People Living With HIV: Comparative Survival Analysis Using Cox and Explainable Machine Learning Models
Background: The Cox proportional hazards (CPH) model is a common choice for analyzing time-to-treatment interruptions in patients on antiretroviral therapy (ART), valued for its straightforward interpretability and flexibility in handling time-dependent covariates. Machine learning (ML) models have increasingly been adapted for handling temporal data, with added advantages of handling complex, nonlinear relationships and large datasets, and providing clear practical interpretations. Objective: This study aims to compare the predictive performance of the traditional CPH model and ML models in predicting treatment interruptions among patients on ART, while also providing both global and individual-level explanations to support personalized, data-driven interventions for improving treatment retention. Methods: Using data from 621,115 patients who started ART between 2017 and 2023, in Kenya, we compared the performance of the CPH with the following ML models—gradient boosting machine, extreme gradient boosting, regularized generalized linear models (Ridge, Lasso, and Elastic-Net), and recursive partitioning—in predicting first and multiple treatment interruptions. Explainable surrogate technique (model-agnostic) was applied to interpret the best performing model’s predictions globally, using variable importance and partial dependence profiles, and at individual level, using breakdown additive, Shapley Additive Explanations, and ceteris paribus. Results: The recursive partitioning model achieved the best performance with a predictive concordance index score of 0.81 for first treatment interruptions and 0.89 for multiple interruptions, outperforming the CPH model, which scored 0.78 and 0.87 for the same scenarios, respectively. Recursive partitioning’s performance can be attributed to its ability to model nonlinear relationships and automatically detect complex interactions. The global model-agnostic explanations aligned closely with the interpretations offered by hazard ratios in the CPH model, while offering additional insights into the impact of specific features on the model’s predictions. The breakdown additive and Shapley Additive Explanations explainers demonstrated how different variables contribute to the predicted risk at the individual patient level. The ceteris paribus profiles further explored the time-varying model to illustrate how changes in a patient’s covariates over time could impact their predicted risk of treatment interruption. Conclusions: Our results highlight the superior predictive performance of ML models and their ability to provide patient-specific risk predictions and insights that can support targeted interventions to reduce treatment interruptions in ART care.
- 4 February 2026
- Published by JMIR Publications Inc. in JMIR Medical Informatics
- Vol. 14 (1) , e78964
Open AccessPrediction of First and Multiple Antiretroviral Therapy Interruptions in People Living With HIV: Comparative Survival Analysis Using Cox and Explainable Machine Learning Models
Background: The Cox proportional hazards (CPH) model is a common choice for analyzing time-to-treatment interruptions in patients on antiretroviral therapy (ART), valued for its straightforward interpretability and flexibility in handling time-dependent covariates. Machine learning (ML) models have increasingly been adapted for handling temporal data, with added advantages of handling complex, nonlinear relationships and large datasets, and providing clear practical interpretations. Objective: This study aims to compare the predictive performance of the traditional CPH model and ML models in predicting treatment interruptions among patients on ART, while also providing both global and individual-level explanations to support personalized, data-driven interventions for improving treatment retention. Methods: Using data from 621,115 patients who started ART between 2017 and 2023, in Kenya, we compared the performance of the CPH with the following ML models—gradient boosting machine, extreme gradient boosting, regularized generalized linear models (Ridge, Lasso, and Elastic-Net), and recursive partitioning—in predicting first and multiple treatment interruptions. Explainable surrogate technique (model-agnostic) was applied to interpret the best performing model’s predictions globally, using variable importance and partial dependence profiles, and at individual level, using breakdown additive, Shapley Additive Explanations, and ceteris paribus. Results: The recursive partitioning model achieved the best performance with a predictive concordance index score of 0.81 for first treatment interruptions and 0.89 for multiple interruptions, outperforming the CPH model, which scored 0.78 and 0.87 for the same scenarios, respectively. Recursive partitioning’s performance can be attributed to its ability to model nonlinear relationships and automatically detect complex interactions. The global model-agnostic explanations aligned closely with the interpretations offered by hazard ratios in the CPH model, while offering additional insights into the impact of specific features on the model’s predictions. The breakdown additive and Shapley Additive Explanations explainers demonstrated how different variables contribute to the predicted risk at the individual patient level. The ceteris paribus profiles further explored the time-varying model to illustrate how changes in a patient’s covariates over time could impact their predicted risk of treatment interruption. Conclusions: Our results highlight the superior predictive performance of ML models and their ability to provide patient-specific risk predictions and insights that can support targeted interventions to reduce treatment interruptions in ART care.
- Open Access
- 30 June 2022
- Published by AOSIS in African Journal of Laboratory Medicine
- Vol. 11 (1) , 6
Implementing an intensified mentorship approach towards accelerated medical laboratory accreditation in 10 counties in Kenya
Background: Despite Kenya's roll-out of the Strengthening Laboratory Management Towards Accreditation programme in 2010, most laboratories had not made significant or tangible improvements towards accreditation by 2016. In April 2016, the University of Maryland, Baltimore enrolled 27 facilities in the standard Strengthening Laboratory Management Towards Accreditation programme. Objective: This study aimed to describe and evaluate the implementation of an intensified mentorship strategy on laboratory accreditation. Methods: In October 2017, the University of Maryland, Baltimore implemented intensive mentorship in 27 hospital laboratories in Nairobi, Kiambu, Meru, Embu, Muranga, Nyeri, Laikipia, Nyandarua, Tharaka-Nithi, and Kirinyaga counties in Kenya. Laboratories were paired with competent mentors whose skills were matched to facility gaps. Baseline and follow-up assessments were done between April 2016 and March 2019 using the World Health Organization's Stepwise Laboratory Quality Improvement Process Towards Accreditation Checklist and overall scores of the 12 Quality System Essentials and star ratings (from zero to five, based on scores) used to evaluate the effectiveness of the intensified mentorship. Results: In September 2017, 14 laboratories scored zero stars, three scored one star, eight scored two stars, one scored three stars, and one laboratory was accredited. By March 2019, eight laboratories were accredited, five scored four stars, 10 scored three stars, three scored two stars, and only one scored one star. The average score change with the intensified approach was 81.5 versus 53.9 for the standard approach. Conclusion: The intensified mentorship strategy resulted in fast-tracked progress towards laboratory accreditation and can be adopted in similar resource-limited settings.
- 30 June 2022
- Published by AOSIS in African Journal of Laboratory Medicine
- Vol. 11 (1) , 6
Open AccessImplementing an intensified mentorship approach towards accelerated medical laboratory accreditation in 10 counties in Kenya
Background: Despite Kenya's roll-out of the Strengthening Laboratory Management Towards Accreditation programme in 2010, most laboratories had not made significant or tangible improvements towards accreditation by 2016. In April 2016, the University of Maryland, Baltimore enrolled 27 facilities in the standard Strengthening Laboratory Management Towards Accreditation programme. Objective: This study aimed to describe and evaluate the implementation of an intensified mentorship strategy on laboratory accreditation. Methods: In October 2017, the University of Maryland, Baltimore implemented intensive mentorship in 27 hospital laboratories in Nairobi, Kiambu, Meru, Embu, Muranga, Nyeri, Laikipia, Nyandarua, Tharaka-Nithi, and Kirinyaga counties in Kenya. Laboratories were paired with competent mentors whose skills were matched to facility gaps. Baseline and follow-up assessments were done between April 2016 and March 2019 using the World Health Organization's Stepwise Laboratory Quality Improvement Process Towards Accreditation Checklist and overall scores of the 12 Quality System Essentials and star ratings (from zero to five, based on scores) used to evaluate the effectiveness of the intensified mentorship. Results: In September 2017, 14 laboratories scored zero stars, three scored one star, eight scored two stars, one scored three stars, and one laboratory was accredited. By March 2019, eight laboratories were accredited, five scored four stars, 10 scored three stars, three scored two stars, and only one scored one star. The average score change with the intensified approach was 81.5 versus 53.9 for the standard approach. Conclusion: The intensified mentorship strategy resulted in fast-tracked progress towards laboratory accreditation and can be adopted in similar resource-limited settings.
- 1 September 2021
- Published by SAGE Publications in AIDS Research and Human Retroviruses
- Vol. 37 (9) , 642-646
Factors Associated with Loss to Follow-Up Among Patients Receiving HIV Treatment in Nairobi, Kenya
Objective: We investigated factors associated with loss to follow-up in 24 urban health facilities in Nairobi, Kenya. Materials and Methods: We conducted a retrospective analysis of routinely collected data to assess factors associated with lost to follow-up (LTFU) in the period October 1, 2016 to June 30, 2017. LTFU was defined as no antiretroviral therapy (ART) refill for ≥90 days and no documentation of transfer, death, or treatment cessation in the patient chart, and if no lapse of ≥90 days between ART refills, patients were considered retained in care. Multivariable logistic regression modeling was used to compute odds ratios and 95% CI for LTFU. Results: Our analysis included 633 individuals who were LTFU and 13,098 individuals retained in care. Most participants (69.6%) were women, and median age was 33.0 years (interquartile range, 27.2–38.3 years). Median ART duration was shorter among those LTFU (0.4 years) compared to retained patients (2.5 years, p<0.0001). Being male (adjusted odds ratio (aOR) 1.30; 95% confidence interval (CI) 1.04, 1.63, P=0.02), transferring into facilities while already receiving ART (aOR 11.58; 95% CI 8.23, 16.29, P<0.0001), and having a shorter ART duration (<6 months) were associated with increased odds of LTFU. Patients who transferred into a facility while already receiving ART had the highest adjusted odds of being LTFU compared to those retained in care. Conclusions: In this urban and highly mobile population, transferring into facilities while already receiving ART was strongly associated with LTFU. Focusing programming efforts on patients transferring between urban clinics to identify reasons for transfer and potential barriers to treatment adherence could help improve patient outcomes. Supplementary case management and support may be needed to promote a seamless transition and ensure uninterrupted engagement in HIV care and treatment.
- 1 September 2021
- Published by SAGE Publications in AIDS Research and Human Retroviruses
- Vol. 37 (9) , 642-646
Factors Associated with Loss to Follow-Up Among Patients Receiving HIV Treatment in Nairobi, Kenya
Objective: We investigated factors associated with loss to follow-up in 24 urban health facilities in Nairobi, Kenya. Materials and Methods: We conducted a retrospective analysis of routinely collected data to assess factors associated with lost to follow-up (LTFU) in the period October 1, 2016 to June 30, 2017. LTFU was defined as no antiretroviral therapy (ART) refill for ≥90 days and no documentation of transfer, death, or treatment cessation in the patient chart, and if no lapse of ≥90 days between ART refills, patients were considered retained in care. Multivariable logistic regression modeling was used to compute odds ratios and 95% CI for LTFU. Results: Our analysis included 633 individuals who were LTFU and 13,098 individuals retained in care. Most participants (69.6%) were women, and median age was 33.0 years (interquartile range, 27.2–38.3 years). Median ART duration was shorter among those LTFU (0.4 years) compared to retained patients (2.5 years, p<0.0001). Being male (adjusted odds ratio (aOR) 1.30; 95% confidence interval (CI) 1.04, 1.63, P=0.02), transferring into facilities while already receiving ART (aOR 11.58; 95% CI 8.23, 16.29, P<0.0001), and having a shorter ART duration (<6 months) were associated with increased odds of LTFU. Patients who transferred into a facility while already receiving ART had the highest adjusted odds of being LTFU compared to those retained in care. Conclusions: In this urban and highly mobile population, transferring into facilities while already receiving ART was strongly associated with LTFU. Focusing programming efforts on patients transferring between urban clinics to identify reasons for transfer and potential barriers to treatment adherence could help improve patient outcomes. Supplementary case management and support may be needed to promote a seamless transition and ensure uninterrupted engagement in HIV care and treatment.