Digital clinical support tools to improve child health: development, implementation, and evaluation of ePOCT+ to support healthcare providers in the management of sick children at primary care health facilities in Tanzania

Bacterial antimicrobial resistance due to inappropriate antibiotic use, and poor quality of care are major contributors to the unacceptably high childhood mortality in Tanzania. Electronic Clinical Decision Support Algorithms (CDSAs) are evidence based digital health tools based on clinical guidelines that guide health providers through a consultation to ultimately propose the diagnosis and treatment based on the inputs entered. Such tools have been found to reduce antibiotic prescription and improve quality of care. Nonetheless, there is a lack of pragmatic studies evaluating CDSAs in Tanzania, and there are many remaining challenges with previously developed CDSAs.
The aim of this project was to improve quality of care and reduce antibiotic prescription at primary care level health facilities in Tanzania. This was done by first developing the ePOCT+ clinical decision support algorithm, addressing challenges identified in other CDSAs, and secondly evaluating the effect of ePOCT+ on antibiotic prescription in a pragmatic cluster randomized controlled trial (DYNAMIC Tanzania study).
To improve uptake, adherence, safety, and potential for antibiotic stewardship, ePOCT+ expanded the clinical scope of the clinical algorithms, expanded the age range of patients it could manage, and assured comprehensive input from clinical and digital experts, as well as health provider end-users. Numerous meetings were conducted, Delphi processes were utilized, and comprehensive piloting was performed to develop ePOCT+. In order to assure safety, a systematic review was conducted to identify the best performing predictors of severe disease to integrate within the algorithm.
The DYNAMIC Tanzania study was a pragmatic, open-label, parallel-group, cluster randomized trial in 40 primary health facilities in the Mbeya and Morogoro regions of Tanzania. Randomization of health facilities were stratified by region, council, level of health facility, and attendance rate. The intervention consisted of the ePOCT+ CDSA with supporting IT infrastructure, C-reactive protein rapid test, hemoglobin rapid test, pulse oximeter, training and supportive mentorship. Co-primary outcomes were 1) antibiotic prescription at the time of the initial consultation (superiority analysis), and 2) clinical failure at day 7 defined as “not cured” and “not improved”, or unscheduled hospitalization (non-inferiority analysis). Secondary safety outcomes include unscheduled re-attendance visits, non-referred secondary hospitalization and death by day 7. Analyses were performed using a random effects logistic regression model using the cluster and patient as random effects, with further adjustment using fixed effect terms for randomization stratification factors, and baseline characteristics.
The systematic review on predictors of severe disease in febrile children presenting from the community identified 18 studies evaluating 200 prognostic factors and 25 clinical prediction models in 24 530 children. There were few outpatient and primary care studies identified. The most common and best preforming predictors of severe disease were malnutrition, altered consciousness, markers of acidosis, and poor peripheral perfusion.
In expanding the age scope of ePOCT+ to manage children 1 day to 15 years, and based on feedback from previous studies and CDSAs, additional illnesses were integrated in the ePOCT+ clinical algorithm. These include trauma, urinary tract infection, and abdominal pain, selected based on 1) incidence, 2) morbidity/mortality, and 3) feasibility at primary care. A Delphi survey among 30 Tanzanian health providers evaluated feasibility, acceptability and reliability of integrating specific predictors within ePOCT+, notably predictors identified within the systematic review. Feasibility tests in over 200 patients in 20 health facilities, and pilots in over 2000 consultations, lead to modifications to ePOCT+ based on user-experience feedback and observations, notably providing option to not measure some clinical signs when not feasible, allow health providers to accept or refuse a proposed diagnosis or treatment, provide alternative medicines in case of stock-outs, and highlighting clinical elements that would result in referral.
The DYNAMIC Tanzania cluster randomized trial took place between December 2021 to October 2022. Over 40,000 children under 15 years of age were enrolled in 20 health facilities (clusters) where health providers could use ePOCT+, and 20 health facilities where health providers provided care as usual. The co-primary outcomes found that the use of ePOCT+ resulted in a 3-fold reduction in the likelihood of a sick child receiving an antibiotic prescription compared to children in usual care health facilities. Despite substantially fewer antibiotics prescriptions, the co-primary outcome of clinical failure was non-inferior. There were also no differences in the secondary outcomes of death, secondary hospitalization, and additional medications after the initial consultations between study arms by day 7.
In conclusion, the ePOCT+ electronic clinical decision support algorithm if implemented to scale could help address the urgent problem of antimicrobial resistance by safely reducing antibiotic prescribing. Transfer of ownership to the ministry of health of Tanzania and integration within the Tanzanian digital health landscape will be essential in order to achieve wide scale implementation.