Predictive modelling of linear growth faltering among pediatric patients with Diarrhea in Rural Western Kenya: an explainable machine learning approach DOI Creative Commons
Billy Ogwel,

Vincent H. Mzazi,

Alex O Awuor

et al.

BMC Medical Informatics and Decision Making, Journal Year: 2024, Volume and Issue: 24(1)

Published: Dec. 2, 2024

Abstract Introduction Stunting affects one-fifth of children globally with diarrhea accounting for an estimated 13.5% stunting. Identifying risk factors its precursor, linear growth faltering (LGF), is critical to designing interventions. Moreover, developing new predictive models LGF using more recent data offers opportunity enhance model accuracy, interpretability and capture insights. We employed machine learning (ML) derive validate a among enrolled in the Vaccine Impact on Diarrhea Africa (VIDA) study Enterics Global Heath (EFGH) ― Shigella rural western Kenya. Methods used 7 diverse ML algorithms retrospectively build prognostic prediction (≥ 0.5 decrease height/length age z-score [HAZ]) 6–35 months. de-identified from VIDA ( n = 1,106) combined synthetic 8,894) development, which entailed split-sampling K-fold cross-validation over-sampling technique, EFGH-Shigella 655) temporal validation. Potential predictors 65) included demographic, household-level characteristics, illness history, anthropometric clinical were identified boruta feature selection explanatory analysis interpretability. Results The prevalence development validation cohorts was 187 (16.9%) 147 (22.4%), respectively. Feature following 6 variables ranked by importance: (16.6%), temperature (6.0%), respiratory rate (4.1%), SAM (3.4%), rotavirus vaccination (3.3%), skin turgor (2.1%). While all showed good capability, gradient boosting achieved best performance (area under curve % [95% Confidence Interval]: 83.5 [81.6–85.4] 65.6 [60.8–70.4]) datasets, Conclusion Our findings accentuate enduring relevance established whilst demonstrating practical utility rapid identification at-risk children.

Language: Английский

Shigella Detection and Molecular Serotyping With a Customized TaqMan Array Card in the Enterics for Global Health (EFGH): Shigella Surveillance Study DOI Creative Commons
Jie Liu, Paul F. Garcia Bardales, Kamrul Islam

et al.

Open Forum Infectious Diseases, Journal Year: 2024, Volume and Issue: 11(Supplement_1), P. S34 - S40

Published: March 1, 2024

Quantitative polymerase chain reaction (qPCR) targeting ipaH has been proven to be highly efficient in detecting Shigella clinical samples compared culture-based methods, which underestimate burden by 2- 3-fold. qPCR assays have also developed for speciation and serotyping, is critical both vaccine development evaluation. The Enterics Global Health (EFGH) surveillance study will utilize a customized real-time PCR-based TaqMan Array Card (TAC) interrogating 82 targets, the detection differentiation of spp, sonnei, flexneri serotypes, other diarrhea-associated enteropathogens, antimicrobial resistance (AMR) genes. Total nucleic acid extracted from rectal swabs or stool samples, assayed on TAC. analysis performed determine likely attribution particular etiologies diarrhea using quantification cycle cutoffs derived previous studies. results conventional culture, phenotypic susceptibility approaches EFGH. TAC enables simultaneous diarrheal etiologies, principal pathogen subtypes, AMR high sensitivity assay more accurate estimation Shigella-attributed disease burden, informing policy design future trials.

Language: Английский

Citations

8
Derivation and validation of a clinical predictive model for longer duration diarrhea among pediatric patients in Kenya using machine learning algorithms DOI Creative Commons
Billy Ogwel,

Vincent H. Mzazi,

Alex O Awuor

et al.

BMC Medical Informatics and Decision Making, Journal Year: 2025, Volume and Issue: 25(1)

Published: Jan. 15, 2025

Abstract Background Despite the adverse health outcomes associated with longer duration diarrhea (LDD), there are currently no clinical decision tools for timely identification and better management of children increased risk. This study utilizes machine learning (ML) to derive validate a predictive model LDD among presenting facilities. Methods was defined as episode lasting ≥ 7 days. We used ML algorithms build prognostic models prediction < 5 years using de-identified data from Vaccine Impact on Diarrhea in Africa ( N = 1,482) development Enterics Global Health Shigella 682) temporal validation champion model. Features included demographic, medical history examination collected at enrolment both studies. conducted split-sampling employed K-fold cross-validation over-sampling technique development. Moreover, critical predictors their impact were obtained an explainable agnostic approach. The determined based area under curve (AUC) metric. Model calibrations assessed Brier, Spiegelhalter’s z -test its accompanying p -value. Results There significant difference prevalence between cohorts (478 [32.3%] vs 69 [10.1%]; 0.001). following variables decreasing order: pre-enrolment days (55.1%), modified Vesikari score(18.2%), age group (10.7%), vomit (8.8%), respiratory rate (6.5%), vomiting (6.4%), frequency (6.2%), rotavirus vaccination (6.1%), skin pinch (2.4%) stool (2.4%). While all showed good capability, random forest achieved best performance (AUC [95% Confidence Interval]: 83.0 [78.6–87.5] 71.0 [62.5–79.4]) datasets, respectively. slight deviations perfect calibration, these not statistically (Brier score 0.17, Spiegelhalter -value 0.219). Conclusions Our suggests derived could be rapidly identify risk LDD. Integrating into decision-making may allow clinicians target closer observation enhanced management.

Language: Английский

Citations

0
Optimizing Vaccine Trials for Enteric Diseases: The Enterics for Global Health (EFGH) Shigella Surveillance Study DOI Creative Commons
Kirsten Vannice, Calman A. MacLennan, Jessica E. Long

et al.

Open Forum Infectious Diseases, Journal Year: 2024, Volume and Issue: 11(Supplement_1), P. S1 - S5

Published: March 1, 2024

In this introductory article, we describe the rationale for Enterics Global Health (EFGH) Shigella surveillance study, which is largely to optimize design and implementation of pivotal vaccine trials in target population infants young children living low- middle-income countries. Such optimization will ideally lead a shorter time availability population. We also provide brief description articles included supplement.

Language: Английский

Citations

3
Predictive modelling of linear growth faltering among pediatric patients with Diarrhea in Rural Western Kenya: an explainable machine learning approach DOI Creative Commons
Billy Ogwel,

Vincent H. Mzazi,

Alex O Awuor

et al.

BMC Medical Informatics and Decision Making, Journal Year: 2024, Volume and Issue: 24(1)

Published: Dec. 2, 2024

Abstract Introduction Stunting affects one-fifth of children globally with diarrhea accounting for an estimated 13.5% stunting. Identifying risk factors its precursor, linear growth faltering (LGF), is critical to designing interventions. Moreover, developing new predictive models LGF using more recent data offers opportunity enhance model accuracy, interpretability and capture insights. We employed machine learning (ML) derive validate a among enrolled in the Vaccine Impact on Diarrhea Africa (VIDA) study Enterics Global Heath (EFGH) ― Shigella rural western Kenya. Methods used 7 diverse ML algorithms retrospectively build prognostic prediction (≥ 0.5 decrease height/length age z-score [HAZ]) 6–35 months. de-identified from VIDA ( n = 1,106) combined synthetic 8,894) development, which entailed split-sampling K-fold cross-validation over-sampling technique, EFGH-Shigella 655) temporal validation. Potential predictors 65) included demographic, household-level characteristics, illness history, anthropometric clinical were identified boruta feature selection explanatory analysis interpretability. Results The prevalence development validation cohorts was 187 (16.9%) 147 (22.4%), respectively. Feature following 6 variables ranked by importance: (16.6%), temperature (6.0%), respiratory rate (4.1%), SAM (3.4%), rotavirus vaccination (3.3%), skin turgor (2.1%). While all showed good capability, gradient boosting achieved best performance (area under curve % [95% Confidence Interval]: 83.5 [81.6–85.4] 65.6 [60.8–70.4]) datasets, Conclusion Our findings accentuate enduring relevance established whilst demonstrating practical utility rapid identification at-risk children.

Language: Английский

Citations

0