Feasibility, safety, and impact of the RTS,S/AS01E malaria vaccine when implemented through national immunisation programmes: evaluation of cluster-randomised introduction of the vaccine in Ghana, Kenya, and Malawi

BACKGROUND

The RTS,S/AS01E malaria vaccine (RTS,S) was introduced by national immunisation programmes in Ghana, Kenya, and Malawi in 2019 in large-scale pilot schemes. We aimed to address questions about feasibility and impact, and to assess safety signals that had been observed in the phase 3 trial that included an excess of meningitis and cerebral malaria cases in RTS,S recipients, and the possibility of an excess of deaths among girls who received RTS,S than in controls, to inform decisions about wider use.

METHODS

In this prospective evaluation, 158 geographical clusters (66 districts in Ghana; 46 sub-counties in Kenya; and 46 groups of immunisation clinic catchment areas in Malawi) were randomly assigned to early or delayed introduction of RTS,S, with three doses to be administered between the ages of 5 months and 9 months and a fourth dose at the age of approximately 2 years. Primary outcomes of the evaluation, planned over 4 years, were mortality from all causes except injury (impact), hospital admission with severe malaria (impact), hospital admission with meningitis or cerebral malaria (safety), deaths in girls compared with boys (safety), and vaccination coverage (feasibility). Mortality was monitored in children aged 1-59 months throughout the pilot areas. Surveillance for meningitis and severe malaria was established in eight sentinel hospitals in Ghana, six in Kenya, and four in Malawi. Vaccine uptake was measured in surveys of children aged 12-23 months about 18 months after vaccine introduction. We estimated that sufficient data would have accrued after 24 months to evaluate each of the safety signals and the impact on severe malaria in a pooled analysis of the data from the three countries. We estimated incidence rate ratios (IRRs) by comparing the ratio of the number of events in children age-eligible to have received at least one dose of the vaccine (for safety outcomes), or age-eligible to have received three doses (for impact outcomes), to that in non-eligible age groups in implementation areas with the equivalent ratio in comparison areas. To establish whether there was evidence of a difference between girls and boys in the vaccine's impact on mortality, the female-to-male mortality ratio in age groups eligible to receive the vaccine (relative to the ratio in non-eligible children) was compared between implementation and comparison areas. Preliminary findings contributed to WHO's recommendation in 2021 for widespread use of RTS,S in areas of moderate-to-high malaria transmission.

FINDINGS

By April 30, 2021, 652 673 children had received at least one dose of RTS,S and 494 745 children had received three doses. Coverage of the first dose was 76% in Ghana, 79% in Kenya, and 73% in Malawi, and coverage of the third dose was 66% in Ghana, 62% in Kenya, and 62% in Malawi. 26 285 children aged 1-59 months were admitted to sentinel hospitals and 13 198 deaths were reported through mortality surveillance. Among children eligible to have received at least one dose of RTS,S, there was no evidence of an excess of meningitis or cerebral malaria cases in implementation areas compared with comparison areas (hospital admission with meningitis: IRR 0·63 [95% CI 0·22-1·79]; hospital admission with cerebral malaria: IRR 1·03 [95% CI 0·61-1·74]). The impact of RTS,S introduction on mortality was similar for girls and boys (relative mortality ratio 1·03 [95% CI 0·88-1·21]). Among children eligible for three vaccine doses, RTS,S introduction was associated with a 32% reduction (95% CI 5-51%) in hospital admission with severe malaria, and a 9% reduction (95% CI 0-18%) in all-cause mortality (excluding injury).

INTERPRETATION

In the first 2 years of implementation of RTS,S, the three primary doses were effectively deployed through national immunisation programmes. There was no evidence of the safety signals that had been observed in the phase 3 trial, and introduction of the vaccine was associated with substantial reductions in hospital admission with severe malaria. Evaluation continues to assess the impact of four doses of RTS,S.

FUNDING

Gavi, the Vaccine Alliance; the Global Fund to Fight AIDS, Tuberculosis and Malaria; and Unitaid.

Development and Validation of the Phoenix Criteria for Pediatric Sepsis and Septic Shock

Importance

The Society of Critical Care Medicine Pediatric Sepsis Definition Task Force sought to develop and validate new clinical criteria for pediatric sepsis and septic shock using measures of organ dysfunction through a data-driven approach.

Objective

To derive and validate novel criteria for pediatric sepsis and septic shock across differently resourced settings.

Design, Setting, and Participants

Multicenter, international, retrospective cohort study in 10 health systems in the US, Colombia, Bangladesh, China, and Kenya, 3 of which were used as external validation sites. Data were collected from emergency and inpatient encounters for children (aged <18 years) from 2010 to 2019: 3 049 699 in the development (including derivation and internal validation) set and 581 317 in the external validation set.

Exposure

Stacked regression models to predict mortality in children with suspected infection were derived and validated using the best-performing organ dysfunction subscores from 8 existing scores. The final model was then translated into an integer-based score used to establish binary criteria for sepsis and septic shock.

Main Outcomes and Measures

The primary outcome for all analyses was in-hospital mortality. Model- and integer-based score performance measures included the area under the precision recall curve (AUPRC; primary) and area under the receiver operating characteristic curve (AUROC; secondary). For binary criteria, primary performance measures were positive predictive value and sensitivity.

Results

Among the 172 984 children with suspected infection in the first 24 hours (development set; 1.2% mortality), a 4-organ-system model performed best. The integer version of that model, the Phoenix Sepsis Score, had AUPRCs of 0.23 to 0.38 (95% CI range, 0.20-0.39) and AUROCs of 0.71 to 0.92 (95% CI range, 0.70-0.92) to predict mortality in the validation sets. Using a Phoenix Sepsis Score of 2 points or higher in children with suspected infection as criteria for sepsis and sepsis plus 1 or more cardiovascular point as criteria for septic shock resulted in a higher positive predictive value and higher or similar sensitivity compared with the 2005 International Pediatric Sepsis Consensus Conference (IPSCC) criteria across differently resourced settings.

Conclusions and Relevance

The novel Phoenix sepsis criteria, which were derived and validated using data from higher- and lower-resource settings, had improved performance for the diagnosis of pediatric sepsis and septic shock compared with the existing IPSCC criteria.

International Consensus Criteria for Pediatric Sepsis and Septic Shock

Importance

Sepsis is a leading cause of death among children worldwide. Current pediatric-specific criteria for sepsis were published in 2005 based on expert opinion. In 2016, the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) defined sepsis as life-threatening organ dysfunction caused by a dysregulated host response to infection, but it excluded children.

Objective

To update and evaluate criteria for sepsis and septic shock in children.

Evidence Review

The Society of Critical Care Medicine (SCCM) convened a task force of 35 pediatric experts in critical care, emergency medicine, infectious diseases, general pediatrics, nursing, public health, and neonatology from 6 continents. Using evidence from an international survey, systematic review and meta-analysis, and a new organ dysfunction score developed based on more than 3 million electronic health record encounters from 10 sites on 4 continents, a modified Delphi consensus process was employed to develop criteria.

Findings

Based on survey data, most pediatric clinicians used sepsis to refer to infection with life-threatening organ dysfunction, which differed from prior pediatric sepsis criteria that used systemic inflammatory response syndrome (SIRS) criteria, which have poor predictive properties, and included the redundant term, severe sepsis. The SCCM task force recommends that sepsis in children be identified by a Phoenix Sepsis Score of at least 2 points in children with suspected infection, which indicates potentially life-threatening dysfunction of the respiratory, cardiovascular, coagulation, and/or neurological systems. Children with a Phoenix Sepsis Score of at least 2 points had in-hospital mortality of 7.1% in higher-resource settings and 28.5% in lower-resource settings, more than 8 times that of children with suspected infection not meeting these criteria. Mortality was higher in children who had organ dysfunction in at least 1 of 4-respiratory, cardiovascular, coagulation, and/or neurological-organ systems that was not the primary site of infection. Septic shock was defined as children with sepsis who had cardiovascular dysfunction, indicated by at least 1 cardiovascular point in the Phoenix Sepsis Score, which included severe hypotension for age, blood lactate exceeding 5 mmol/L, or need for vasoactive medication. Children with septic shock had an in-hospital mortality rate of 10.8% and 33.5% in higher- and lower-resource settings, respectively.

Conclusions and Relevance

The Phoenix sepsis criteria for sepsis and septic shock in children were derived and validated by the international SCCM Pediatric Sepsis Definition Task Force using a large international database and survey, systematic review and meta-analysis, and modified Delphi consensus approach. A Phoenix Sepsis Score of at least 2 identified potentially life-threatening organ dysfunction in children younger than 18 years with infection, and its use has the potential to improve clinical care, epidemiological assessment, and research in pediatric sepsis and septic shock around the world.

Verbal Autopsy to Assess Postdischarge Mortality in Children With Suspected Sepsis in Uganda

BACKGROUND

Reducing child mortality in low-income countries is constrained by a lack of vital statistics. In the absence of such data, verbal autopsies provide an acceptable method to determining attributable causes of death. The objective was to assess potential causes of pediatric postdischarge mortality in children younger than age 5 years (under-5) originally admitted for suspected sepsis using verbal autopsies.

METHODS

Secondary analysis of verbal autopsy data from children admitted to 6 hospitals across Uganda from July 2017 to March 2020. Structured verbal autopsy interviews were conducted for all deaths within 6 months after discharge. Two physicians independently classified a primary cause of death, up to 4 alternative causes, and up to 5 contributing conditions using the Start-Up Mortality List, with discordance resolved by consensus.

RESULTS

Verbal autopsies were completed for 361 (98.6%) of the 366 (5.9%) children who died among 6191 discharges (median admission age: 5.4 months [interquartile range, 1.8-16.7]; median time to mortality: 28 days [interquartile range, 9-74]). Most deaths (62.3%) occurred in the community. Leading primary causes of death, assigned in 356 (98.6%) of cases, were pneumonia (26.2%), sepsis (22.1%), malaria (8.5%), and diarrhea (7.9%). Common contributors to death were malnutrition (50.5%) and anemia (25.7%). Reviewers were less confident in their causes of death for neonates than older children (P < .05).

CONCLUSIONS

Postdischarge mortality frequently occurred in the community in children admitted for suspected sepsis in Uganda. Analyses of the probable causes for these deaths using verbal autopsies suggest potential areas for interventions, focused on early detection of infections, as well as prevention and treatment of underlying contributors such as malnutrition and anemia.

Repeatability of Pulse Oximetry Measurements in Children During Triage in 2 Ugandan Hospitals

BACKGROUND

In low- and middle-income countries, health workers use pulse oximeters for intermittent spot measurements of oxygen saturation (SpO2). However, the accuracy and reliability of pulse oximeters for spot measurements have not been determined. We evaluated the repeatability of spot measurements and the ideal observation time to guide recommendations during spot check measurements.

METHODS

Two 1-minute measurements were taken for the 3,903 subjects enrolled in the study conducted April 2020-January 2022 in Uganda, collecting 1 Hz SpO2 and signal quality index (SQI) data. The repeatability between the 2 measurements was assessed using an intraclass correlation coefficient (ICC), calculated using a median of all seconds of non-zero SpO2 values for each recording (any quality, Q1) and again with a quality filter only using seconds with SQI 90% or higher (good quality, Q2). The ICC was also recalculated for both conditions of Q1 and Q2 using the initial 5 seconds, then the initial 10 seconds, and continuing with 5-second increments up to the full 60 seconds. Lastly, the whole minute ICC was calculated with good quality (Q2), including only records where both measurements had a mean SQI of more than 70% (Q3).

RESULTS

The repeatability ICC with condition Q1 was 0.591 (95% confidence interval [CI]=0.570, 0.611). Using only the first 5 seconds of each measurement reduced the repeatability to 0.200 (95% CI=0.169, 0.230). Filtering with Q2, the whole-minute ICC was 0.855 (95% CI=0.847, 0.864). The ICC did not improve beyond the first 35 seconds. For Q3, the repeatability rose to 0.908 (95% CI=0.901, 0.914).

CONCLUSIONS

Training guidelines must emphasize the importance of signal quality and duration of measurement, targeting a minimum of 35 seconds of adequate-quality, stable data. In addition, the design of new devices should incorporate user prompts and force quality checks to encourage more accurate pulse oximetry measurements.

ACORN (A Clinically-Oriented Antimicrobial Resistance Surveillance Network) II: protocol for case based antimicrobial resistance surveillance

Background: Antimicrobial resistance surveillance is essential for empiric antibiotic prescribing, infection prevention and control policies and to drive novel antibiotic discovery. However, most existing surveillance systems are isolate-based without supporting patient-based clinical data, and not widely implemented especially in low- and middle-income countries (LMICs). Methods: A Clinically-Oriented Antimicrobial Resistance Surveillance Network (ACORN) II is a large-scale multicentre protocol which builds on the WHO Global Antimicrobial Resistance and Use Surveillance System to estimate syndromic and pathogen outcomes along with associated health economic costs. ACORN-healthcare associated infection (ACORN-HAI) is an extension study which focuses on healthcare-associated bloodstream infections and ventilator-associated pneumonia. Our main aim is to implement an efficient clinically-oriented antimicrobial resistance surveillance system, which can be incorporated as part of routine workflow in hospitals in LMICs. These surveillance systems include hospitalised patients of any age with clinically compatible acute community-acquired or healthcare-associated bacterial infection syndromes, and who were prescribed parenteral antibiotics. Diagnostic stewardship activities will be implemented to optimise microbiology culture specimen collection practices. Basic patient characteristics, clinician diagnosis, empiric treatment, infection severity and risk factors for HAI are recorded on enrolment and during 28-day follow-up. An R Shiny application can be used offline and online for merging clinical and microbiology data, and generating collated reports to inform local antibiotic stewardship and infection control policies. Discussion: ACORN II is a comprehensive antimicrobial resistance surveillance activity which advocates pragmatic implementation and prioritises improving local diagnostic and antibiotic prescribing practices through patient-centred data collection. These data can be rapidly communicated to local physicians and infection prevention and control teams. Relative ease of data collection promotes sustainability and maximises participation and scalability. With ACORN-HAI as an example, ACORN II has the capacity to accommodate extensions to investigate further specific questions of interest.

Pediatric prognostic models predicting inhospital child mortality in resource-limited settings: An external validation study

Background and Aims

Prognostic models provide evidence-based predictions and estimates of future outcomes, facilitating decision-making, patient care, and research. A few of these models have been externally validated, leading to uncertain reliability and generalizability. This study aims to externally validate four models to assess their transferability and usefulness in clinical practice. The models include the respiratory index of severity in children (RISC)-Malawi model and three other models by Lowlavaar et al.

Methods

The study used data from the Clinical Information Network (CIN) to validate the four models where the primary outcome was in-hospital mortality. 163,329 patients met eligibility criteria. Missing data were imputed, and the logistic function was used to compute predicted risk of in-hospital mortality. Models' discriminatory ability and calibration were determined using area under the curve (AUC), calibration slope, and intercept.

Results

The RISC-Malawi model had 50,669 pneumonia patients who met the eligibility criteria, of which the case-fatality ratio was 4406 (8.7%). Its AUC was 0.77 (95% CI: 0.77-0.78), whereas the calibration slope was 1.04 (95% CI: 1.00 -1.06), and calibration intercept was 0.81 (95% CI: 0.77-0.84). Regarding the external validation of Lowlavaar et al. models, 10,782 eligible patients  were included, with an in-hospital mortality rate of 5.3%. The primary model's AUC was 0.75 (95% CI: 0.72-0.77), the calibration slope was 0.78 (95% CI: 0.71-0.84), and the calibration intercept was 0.37 (95% CI: 0.28-0.46). All models markedly underestimated the risk of mortality.

Conclusion

All externally validated models exhibited either underestimation or overestimation of the risk as judged from calibration statistics. Hence, applying these models with confidence in settings other than their original development context may not be advisable. Our findings strongly suggest the need for recalibrating these model to enhance their generalizability.

Recalibrating prognostic models to improve predictions of in-hospital child mortality in resource-limited settings

BACKGROUND

In an external validation study, model recalibration is suggested once there is evidence of poor model calibration but with acceptable discriminatory abilities. We identified four models, namely RISC-Malawi (Respiratory Index of Severity in Children) developed in Malawi, and three other predictive models developed in Uganda by Lowlaavar et al. (2016). These prognostic models exhibited poor calibration performance in the recent external validation study, hence the need for recalibration.

OBJECTIVE

In this study, we aim to recalibrate these models using regression coefficients updating strategy and determine how much their performances improve.

METHODS

We used data collected by the Clinical Information Network from paediatric wards of 20 public county referral hospitals. Missing data were multiply imputed using chained equations. Model updating entailed adjustment of the model's calibration performance while the discriminatory ability remained unaltered. We used two strategies to adjust the model: intercept-only and the logistic recalibration method.

RESULTS

Eligibility criteria for the RISC-Malawi model were met in 50,669 patients, split into two sets: a model-recalibrating set (n = 30,343) and a test set (n = 20,326). For the Lowlaavar models, 10,782 patients met the eligibility criteria, of whom 6175 were used to recalibrate the models and 4607 were used to test the performance of the adjusted model. The intercept of the recalibrated RISC-Malawi model was 0.12 (95% CI 0.07, 0.17), while the slope of the same model was 1.08 (95% CI 1.03, 1.13). The performance of the recalibrated models on the test set suggested that no model met the threshold of a perfectly calibrated model, which includes a calibration slope of 1 and a calibration-in-the-large/intercept of 0.

CONCLUSIONS

Even after model adjustment, the calibration performances of the 4 models did not meet the recommended threshold for perfect calibration. This finding is suggestive of models over/underestimating the predicted risk of in-hospital mortality, potentially harmful clinically. Therefore, researchers may consider other alternatives, such as ensemble techniques to combine these models into a meta-model to improve out-of-sample predictive performance.

Using self-supervised feature learning to improve the use of pulse oximeter signals to predict paediatric hospitalization

Background: The success of many machine learning applications depends on knowledge about the relationship between the input data and the task of interest (output), hindering the application of machine learning to novel tasks. End-to-end deep learning, which does not require intermediate feature engineering, has been recommended to overcome this challenge but end-to-end deep learning models require large labelled training data sets often unavailable in many medical applications. In this study, we trained self-supervised learning (SSL) models for automatic feature extraction from raw photoplethysmography (PPG) obtained using a pulse oximeter, with the aim of predicting paediatric hospitalization.  Methods: We compared logistic regression models fitted using features extracted using SSL with models trained using both clinical and SSL features. In addition, we compared end-to-end deep learning models initialized randomly or using weights from the SSL models. We also compared the performance of SSL models trained on labelled data alone (n=1,031) with SSL trained using both labelled and unlabelled signals (n=7,578). Results: Logistic regression models were more predictive of hospitalization when trained on features extracted using labelled PPG signals only compared to SSL models trained on both labelled and unlabelled signals (AUC 0.83 vs 0.80). However, features extracted using SSL model trained on both labelled and unlabelled PPG signals were more predictive of hospitalization when concatenated with clinical features (AUC 0.89 vs 0.87). The end-to-end deep learning model had an AUC of 0.80 when initialized using the SSL model trained on all PPG signals, 0.77 when initialized using SSL trained on labelled data only, and 0.73 when initialized randomly. Conclusions: This study shows that SSL can extract features from PPG signals that are predictive of hospitalization or initialize end-to-end deep learning models. Furthermore, SSL can leverage larger unlabelled data sets to improve performance of models fitted using small labelled data sets.

A pragmatic randomized controlled trial of standard care versus steroids plus standard care for treatment of pneumonia in adults admitted to Kenyan hospitals (SONIA)

Background: It is unclear if adjunctive steroid therapy reduces mortality in community-acquired pneumonia, as very few studies have had mortality as a primary outcome. This question has become even more relevant following demonstration of a mortality benefit of dexamethasone when used in patients with COVID-19 who had severe disease. This has led to increased prescription of steroids in adults with community acquired pneumonia in low-resource settings even when their COVID-19 diagnosis is uncertain due to low testing rates. This pragmatic parallel randomised-controlled open-label trial will determine if adjunctive low-dose steroids for treatment of adults admitted to hospital with community acquired pneumonia whose SARS-CoV-2 status is either unknown or negative reduces mortality. Methods: We will enroll and randomize 2180 patients admitted with a clinical diagnosis of community acquired pneumonia into two arms; in Stratum A-participants will receive standard care for the treatment of community acquired pneumonia. In Stratum B-participants will receive a 10-day course of low-dose oral corticosteroids in addition to standard care. All participants will be followed up to 30 days post randomization and their final status recorded (alive or dead). An immunology sub study will be conducted on a subset of the trial participants (50 per arm) to determine the correlation of pre-existing and treatment induced immune and metabolic changes with study outcomes. Discussion: Mortality among adults admitted to hospital with community acquired pneumonia in resource-limited settings is high. Steroids are readily available in these settings. If the addition of steroids to standard care for community acquired pneumonia is found to be beneficial, this easily scalable intervention would significantly reduce the currently high mortality associated with the illness.

The CINAMR (Clinical Information Network-Antimicrobial Resistance) Project: A pilot microbial surveillance using hospitals linked to regional laboratories in Kenya: Study Protocol

Background: Antimicrobial resistance (AMR) is a global threat and is thought to be acute in low-and middle-income country (LMIC) settings, including in Kenya, but there is limited unbiased surveillance that can provide reliable estimates of its burden. Current efforts to build capacity for microbiology testing in Kenya are unlikely to result in systematic routine microbiological testing in the near term. Therefore, there is little prospect for microbiological support to inform clinical diagnoses nor for indicating the burden of AMR and for guiding empirical choice of antibiotics. Objective: We aim to build on an existing collaboration, the Clinical Information Network (CIN), to pilot microbiological surveillance using a 'hub-and-spoke' model where selected hospitals are linked to high quality microbiology research laboratories. Methods: Children admitted to paediatric wards of 12 participating hospitals will have a sample taken for blood culture at admission before antibiotics are started. Indication for blood culture will be a clinician's prescription of antibiotics. Samples will then be transported daily to the research laboratories for culture and antibiotic susceptibility testing and results relayed back to clinicians for patient management. The surveillance will take place for 6 months in each hospital. Separately, we shall conduct semi-structured interviews with frontline health workers to explore the feasibility and utility of this approach. We will also seek to understand how the availability of microbiology results might inform antibiotic stewardship, and as an interim step to the development of better national or regional laboratories linked to routine surveillance. Conclusions: If feasible, this approach is less costly and periodic 'hub-and-spoke' surveillance can be used to track AMR trends and to broadly guide empirical antibiotic guidance meaning it is likely to be more sustainable than establishing functional microbiological facilities in each hospital in a LMIC setting.

Using Machine Learning Methods Incorporating Individual Reader Annotations to Classify Paediatric Chest Radiographs in Epidemiological Studies

Introduction: Epidemiological studies that involve interpretation of chest radiographs (CXRs) suffer from inter-reader and intra-reader variability. Inter-reader and intra-reader variability hinder comparison of results from different studies or centres, which negatively affects efforts to track the burden of chest diseases or evaluate the efficacy of interventions such as vaccines. This study explores machine learning models that could standardize interpretation of CXR across studies and the utility of incorporating individual reader annotations when training models using CXR data sets annotated by multiple readers.

Methods: Convolutional neural networks were used to classify CXRs from seven low to middle-income countries into five categories according to the World Health Organization's standardized methodology for interpreting paediatric CXRs. We compared models trained to predict the final/aggregate classification with models trained to predict how each reader would classify an image and then aggregate predictions for all readers using unweighted mean.

Results: Incorporating individual reader's annotations during model training improved classification accuracy by 3.4% (multi-class accuracy 61% vs 59%). Model accuracy was higher for children above 12 months of age (68% vs 58%). The accuracy of the models in different countries ranged between 45% and 71%.

Conclusions: Machine learning models can annotate CXRs in epidemiological studies reducing inter-reader and intra-reader variability. In addition, incorporating individual reader annotations can improve the performance of machine learning models trained using CXRs annotated by multiple readers.

A proposed de-identification framework for a cohort of children presenting at a health facility in Uganda

Data sharing has enormous potential to accelerate and improve the accuracy of research, strengthen collaborations, and restore trust in the clinical research enterprise. Nevertheless, there remains reluctancy to openly share raw data sets, in part due to concerns regarding research participant confidentiality and privacy. Statistical data de-identification is an approach that can be used to preserve privacy and facilitate open data sharing. We have proposed a standardized framework for the de-identification of data generated from cohort studies in children in a low-and-middle income country. We applied a standardized de-identification framework to a data sets comprised of 241 health related variables collected from a cohort of 1750 children with acute infections from Jinja Regional Referral Hospital in Eastern Uganda. Variables were labeled as direct and quasi-identifiers based on conditions of replicability, distinguishability, and knowability with consensus from two independent evaluators. Direct identifiers were removed from the data sets, while a statistical risk-based de-identification approach using the k-anonymity model was applied to quasi-identifiers. Qualitative assessment of the level of privacy invasion associated with data set disclosure was used to determine an acceptable re-identification risk threshold, and corresponding k-anonymity requirement. A de-identification model using generalization, followed by suppression was applied using a logical stepwise approach to achieve k-anonymity. The utility of the de-identified data was demonstrated using a typical clinical regression example. The de-identified data sets was published on the Pediatric Sepsis Data CoLaboratory Dataverse which provides moderated data access. Researchers are faced with many challenges when providing access to clinical data. We provide a standardized de-identification framework that can be adapted and refined based on specific context and risks. This process will be combined with moderated access to foster coordination and collaboration in the clinical research community.

Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis

BACKGROUND

Antimicrobial resistance (AMR) poses a major threat to human health around the world. Previous publications have estimated the effect of AMR on incidence, deaths, hospital length of stay, and health-care costs for specific pathogen-drug combinations in select locations. To our knowledge, this study presents the most comprehensive estimates of AMR burden to date.

METHODS

We estimated deaths and disability-adjusted life-years (DALYs) attributable to and associated with bacterial AMR for 23 pathogens and 88 pathogen-drug combinations in 204 countries and territories in 2019. We obtained data from systematic literature reviews, hospital systems, surveillance systems, and other sources, covering 471 million individual records or isolates and 7585 study-location-years. We used predictive statistical modelling to produce estimates of AMR burden for all locations, including for locations with no data. Our approach can be divided into five broad components: number of deaths where infection played a role, proportion of infectious deaths attributable to a given infectious syndrome, proportion of infectious syndrome deaths attributable to a given pathogen, the percentage of a given pathogen resistant to an antibiotic of interest, and the excess risk of death or duration of an infection associated with this resistance. Using these components, we estimated disease burden based on two counterfactuals: deaths attributable to AMR (based on an alternative scenario in which all drug-resistant infections were replaced by drug-susceptible infections), and deaths associated with AMR (based on an alternative scenario in which all drug-resistant infections were replaced by no infection). We generated 95% uncertainty intervals (UIs) for final estimates as the 25th and 975th ordered values across 1000 posterior draws, and models were cross-validated for out-of-sample predictive validity. We present final estimates aggregated to the global and regional level.

FINDINGS

On the basis of our predictive statistical models, there were an estimated 4·95 million (3·62-6·57) deaths associated with bacterial AMR in 2019, including 1·27 million (95% UI 0·911-1·71) deaths attributable to bacterial AMR. At the regional level, we estimated the all-age death rate attributable to resistance to be highest in western sub-Saharan Africa, at 27·3 deaths per 100 000 (20·9-35·3), and lowest in Australasia, at 6·5 deaths (4·3-9·4) per 100 000. Lower respiratory infections accounted for more than 1·5 million deaths associated with resistance in 2019, making it the most burdensome infectious syndrome. The six leading pathogens for deaths associated with resistance (Escherichia coli, followed by Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa) were responsible for 929 000 (660 000-1 270 000) deaths attributable to AMR and 3·57 million (2·62-4·78) deaths associated with AMR in 2019. One pathogen-drug combination, meticillin-resistant S aureus, caused more than 100 000 deaths attributable to AMR in 2019, while six more each caused 50 000-100 000 deaths: multidrug-resistant excluding extensively drug-resistant tuberculosis, third-generation cephalosporin-resistant E coli, carbapenem-resistant A baumannii, fluoroquinolone-resistant E coli, carbapenem-resistant K pneumoniae, and third-generation cephalosporin-resistant K pneumoniae.

INTERPRETATION

To our knowledge, this study provides the first comprehensive assessment of the global burden of AMR, as well as an evaluation of the availability of data. AMR is a leading cause of death around the world, with the highest burdens in low-resource settings. Understanding the burden of AMR and the leading pathogen-drug combinations contributing to it is crucial to making informed and location-specific policy decisions, particularly about infection prevention and control programmes, access to essential antibiotics, and research and development of new vaccines and antibiotics. There are serious data gaps in many low-income settings, emphasising the need to expand microbiology laboratory capacity and data collection systems to improve our understanding of this important human health threat.

FUNDING

Bill & Melinda Gates Foundation, Wellcome Trust, and Department of Health and Social Care using UK aid funding managed by the Fleming Fund.

Malaria hospitalisation in East Africa: age, phenotype and transmission intensity

BACKGROUND

Understanding the age patterns of disease is necessary to target interventions to maximise cost-effective impact. New malaria chemoprevention and vaccine initiatives target young children attending routine immunisation services. Here we explore the relationships between age and severity of malaria hospitalisation versus malaria transmission intensity.

METHODS

Clinical data from 21 surveillance hospitals in East Africa were reviewed. Malaria admissions aged 1 month to 14 years from discrete administrative areas since 2006 were identified. Each site-time period was matched to a model estimated community-based age-corrected parasite prevalence to provide predictions of prevalence in childhood (PfPR2-10). Admission with all-cause malaria, severe malaria anaemia (SMA), respiratory distress (RD) and cerebral malaria (CM) were analysed as means and predicted probabilities from Bayesian generalised mixed models.

RESULTS

52,684 malaria admissions aged 1 month to 14 years were described at 21 hospitals from 49 site-time locations where PfPR2-10 varied from < 1 to 48.7%. Twelve site-time periods were described as low transmission (PfPR2-10 < 5%), five low-moderate transmission (PfPR2-10 5-9%), 20 moderate transmission (PfPR2-10 10-29%) and 12 high transmission (PfPR2-10 ≥ 30%). The majority of malaria admissions were below 5 years of age (69-85%) and rare among children aged 10-14 years (0.7-5.4%) across all transmission settings. The mean age of all-cause malaria hospitalisation was 49.5 months (95% CI 45.1, 55.4) under low transmission compared with 34.1 months (95% CI 30.4, 38.3) at high transmission, with similar trends for each severe malaria phenotype. CM presented among older children at a mean of 48.7 months compared with 39.0 months and 33.7 months for SMA and RD, respectively. In moderate and high transmission settings, 34% and 42% of the children were aged between 2 and 23 months and so within the age range targeted by chemoprevention or vaccines.

CONCLUSIONS

Targeting chemoprevention or vaccination programmes to areas where community-based parasite prevalence is ≥10% is likely to match the age ranges covered by interventions (e.g. intermittent presumptive treatment in infancy to children aged 2-23 months and current vaccine age eligibility and duration of efficacy) and the age ranges of highest disease burden.

The effect of introduction of routine immunization for rotavirus vaccine on paediatric admissions with diarrhoea and dehydration to Kenyan Hospitals: an interrupted time series study

Background: Dehydration secondary to diarrhoea is a major cause of hospitalization and mortality in children aged less than five years. Most diarrhoea cases in childhood are caused by rotavirus, and routine introduction of rotavirus vaccine is expected to reduce the incidence and severity of dehydration secondary to diarrhoea in vaccinated infants. Previously, studies have examined changes in admissions with stools positive for rotavirus but this study reports on all admissions with dehydration secondary to diarrhoea regardless of stool rotavirus results. We aimed to assess the changes in all-cause severe diarrhoea and dehydration (DAD) admissions following the vaccine’s introduction. Methods: We examined changes in admissions of all clinical cases of DAD before and after introduction of routine vaccination with rotavirus vaccine in July 2014 in Kenya. We use data from 13 public hospitals currently involved in a clinical network, the Clinical Information Network (CIN). Routinely collected data for children aged 2-36 months were examined. We used a segmented mixed effects model to assess changes in the burden of diarrhoea and dehydration after introduction of rotavirus vaccine. For sensitivity analysis, we examined trends for non-febrile admissions (surgical or burns). Results: There were 17,708 patients classified as having both diarrhoea and dehydration. Average monthly admissions due to DAD for each hospital before vaccine introduction (July 2014) was 35 (standard deviation: ±22) and 17 (standard deviation: ±12) after vaccine introduction.  Segmented mixed effects regression model showed there was a 33% (95% CI, 30% to 38%) decrease in DAD admissions immediately after the vaccine was introduced to the Kenya immunization program in July 2014.  There was no change in admissions due to non-febrile admissions pre-and post-vaccine introduction.  Conclusion: The rotavirus vaccine, after introduction into the Kenya routine immunization program resulted in reduction of all-cause admissions of diarrhoea and dehydration in children to public hospitals.

Criteria for Pediatric Sepsis-A Systematic Review and Meta-Analysis by the Pediatric Sepsis Definition Taskforce

OBJECTIVE

To determine the associations of demographic, clinical, laboratory, organ dysfunction, and illness severity variable values with: 1) sepsis, severe sepsis, or septic shock in children with infection and 2) multiple organ dysfunction or death in children with sepsis, severe sepsis, or septic shock.

DATA SOURCES

MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials were searched from January 1, 2004, and November 16, 2020.

STUDY SELECTION

Case-control studies, cohort studies, and randomized controlled trials in children greater than or equal to 37-week-old postconception to 18 years with suspected or confirmed infection, which included the terms "sepsis," "septicemia," or "septic shock" in the title or abstract.

DATA EXTRACTION

Study characteristics, patient demographics, clinical signs or interventions, laboratory values, organ dysfunction measures, and illness severity scores were extracted from eligible articles. Random-effects meta-analysis was performed.

DATA SYNTHESIS

One hundred and six studies met eligibility criteria of which 81 were included in the meta-analysis. Sixteen studies (9,629 patients) provided data for the sepsis, severe sepsis, or septic shock outcome and 71 studies (154,674 patients) for the mortality outcome. In children with infection, decreased level of consciousness and higher Pediatric Risk of Mortality scores were associated with sepsis/severe sepsis. In children with sepsis/severe sepsis/septic shock, chronic conditions, oncologic diagnosis, use of vasoactive/inotropic agents, mechanical ventilation, serum lactate, platelet count, fibrinogen, procalcitonin, multi-organ dysfunction syndrome, Pediatric Logistic Organ Dysfunction score, Pediatric Index of Mortality-3, and Pediatric Risk of Mortality score each demonstrated significant and consistent associations with mortality. Pooled mortality rates varied among high-, upper middle-, and lower middle-income countries for patients with sepsis, severe sepsis, and septic shock (p < 0.0001).

CONCLUSIONS

Strong associations of several markers of organ dysfunction with the outcomes of interest among infected and septic children support their inclusion in the data validation phase of the Pediatric Sepsis Definition Taskforce.

Using Machine Learning Methods Incorporating Individual Reader Annotations to Classify Paediatric Chest Radiographs in Epidemiological Studies

Introduction: Epidemiological studies that involve interpretation of chest radiographs (CXRs) suffer from inter-reader and intra-reader variability. Inter-reader and intra-reader variability hinder comparison of results from different studies or centres, which negatively affects efforts to track the burden of chest diseases or evaluate the efficacy of interventions such as vaccines. This study explores machine learning models that could standardize interpretation of CXR across studies and the utility of incorporating individual reader annotations when training models using CXR data sets annotated by multiple readers. Methods: Convolutional neural networks were used to classify CXRs from seven low to middle-income countries into five categories according to the World Health Organization's standardized methodology for interpreting paediatric CXRs. We compared models trained to predict the final/aggregate classification with models trained to predict how each reader would classify an image and then aggregate predictions for all readers using unweighted mean. Results: Incorporating individual reader's annotations during model training improved classification accuracy by 3.4% (multi-class accuracy 61% vs 59%). Model accuracy was higher for children above 12 months of age (68% vs 58%). The accuracy of the models in different countries ranged between 45% and 71%. Conclusions: Machine learning models can annotate CXRs in epidemiological studies reducing inter-reader and intra-reader variability. In addition, incorporating individual reader annotations can improve the performance of machine learning models trained using CXRs annotated by multiple readers.

Surveillance strategies using routine microbiology for antimicrobial resistance in low- and middle-income countries

BACKGROUND

Routine microbiology results are a valuable source of antimicrobial resistance (AMR) surveillance data in low- and middle-income countries (LMICs) as well as in high-income countries. Different approaches and strategies are used to generate AMR surveillance data.

OBJECTIVES

We aimed to review strategies for AMR surveillance using routine microbiology results in LMICs and to highlight areas that need support to generate high-quality AMR data.

SOURCES

We searched PubMed for papers that used routine microbiology to describe the epidemiology of AMR and drug-resistant infections in LMICs. We also included papers that, from our perspective, were critical in highlighting the biases and challenges or employed specific strategies to overcome these in reporting AMR surveillance in LMICs.

CONTENT

Topics covered included strategies of identifying AMR cases (including case-finding based on isolates from routine diagnostic specimens and case-based surveillance of clinical syndromes), of collecting data (including cohort, point-prevalence survey, and case-control), of sampling AMR cases (including lot quality assurance surveys), and of processing and analysing data for AMR surveillance in LMICs.

IMPLICATIONS

The various AMR surveillance strategies warrant a thorough understanding of their limitations and potential biases to ensure maximum utilization and interpretation of local routine microbiology data across time and space. For instance, surveillance using case-finding based on results from clinical diagnostic specimens is relatively easy to implement and sustain in LMIC settings, but the estimates of incidence and proportion of AMR is at risk of biases due to underuse of microbiology. Case-based surveillance of clinical syndromes generates informative statistics that can be translated to clinical practices but needs financial and technical support as well as locally tailored trainings to sustain. Innovative AMR surveillance strategies that can easily be implemented and sustained with minimal costs will be useful for improving AMR data availability and quality in LMICs.

Using self-supervised feature learning to improve the use of pulse oximeter signals to predict paediatric hospitalization

Background: The success of many machine learning applications depends on knowledge about the relationship between the input data and the task of interest (output), hindering the application of machine learning to novel tasks. End-to-end deep learning, which does not require intermediate feature engineering, has been recommended to overcome this challenge but end-to-end deep learning models require large labelled training data sets often unavailable in many medical applications. In this study, we trained machine learning models to predict paediatric hospitalization given raw photoplethysmography (PPG) signals obtained from a pulse oximeter. We trained self-supervised learning (SSL) for automatic feature extraction from PPG signals and assessed the utility of SSL in initializing end-to-end deep learning models trained on a small labelled data set with the aim of predicting paediatric hospitalization.Methods: We compared logistic regression models fitted using features extracted using SSL with end-to-end deep learning models initialized either randomly or using weights from the SSL model. We also compared the performance of SSL models trained on labelled data alone (n=1,031) with SSL trained using both labelled and unlabelled signals (n=7,578). Results: The SSL model trained on both labelled and unlabelled PPG signals produced features that were more predictive of hospitalization compared to the SSL model trained on labelled PPG only (AUC of logistic regression model: 0.78 vs 0.74). The end-to-end deep learning model had an AUC of 0.80 when initialized using the SSL model trained on all PPG signals, 0.77 when initialized using SSL trained on labelled data only, and 0.73 when initialized randomly. Conclusions: This study shows that SSL can improve the classification of PPG signals by either extracting features required by logistic regression models or initializing end-to-end deep learning models. Furthermore, SSL can leverage larger unlabelled data sets to improve performance of models fitted using small labelled data sets.

Pulse oximetry adoption and oxygen orders at paediatric admission over 7 years in Kenya: a multihospital retrospective cohort study

OBJECTIVES

To characterise adoption and explore specific clinical and patient factors that might influence pulse oximetry and oxygen use in low-income and middle-income countries (LMICs) over time; to highlight useful considerations for entities working on programmes to improve access to pulse oximetry and oxygen.

DESIGN

A multihospital retrospective cohort study.

SETTINGS

All admissions (n=132 737) to paediatric wards of 18 purposely selected public hospitals in Kenya that joined a Clinical Information Network (CIN) between March 2014 and December 2020.

OUTCOMES

Pulse oximetry use and oxygen prescription on admission; we performed growth-curve modelling to investigate the association of patient factors with study outcomes over time while adjusting for hospital factors.

RESULTS

Overall, pulse oximetry was used in 48.8% (64 722/132 737) of all admission cases. Use rose on average with each month of participation in the CIN (OR: 1.11, 95% CI 1.05 to 1.18) but patterns of adoption were highly variable across hospitals suggesting important factors at hospital level influence use of pulse oximetry. Of those with pulse oximetry measurement, 7% (4510/64 722) had hypoxaemia (SpO2 <90%). Across the same period, 8.6% (11 428/132 737) had oxygen prescribed but in 87%, pulse oximetry was either not done or the hypoxaemia threshold (SpO2 <90%) was not met. Lower chest-wall indrawing and other respiratory symptoms were associated with pulse oximetry use at admission and were also associated with oxygen prescription in the absence of pulse oximetry or hypoxaemia.

CONCLUSION

The adoption of pulse oximetry recommended in international guidelines for assessing children with severe illness has been slow and erratic, reflecting system and organisational weaknesses. Most oxygen orders at admission seem driven by clinical and situational factors other than the presence of hypoxaemia. Programmes aiming to implement pulse oximetry and oxygen systems will likely need a long-term vision to promote adoption, guideline development and adherence and continuously examine impact.

Malaria infection and severe disease risks in Africa

The relationship between community prevalence of Plasmodium falciparum and the burden of severe, life-threatening disease remains poorly defined. To examine the three most common severe malaria phenotypes from catchment populations across East Africa, we assembled a dataset of 6506 hospital admissions for malaria in children aged 3 months to 9 years from 2006 to 2020. Admissions were paired with data from community parasite infection surveys. A Bayesian procedure was used to calibrate uncertainties in exposure (parasite prevalence) and outcomes (severe malaria phenotypes). Each 25% increase in prevalence conferred a doubling of severe malaria admission rates. Severe malaria remains a burden predominantly among young children (3 to 59 months) across a wide range of community prevalence typical of East Africa. This study offers a quantitative framework for linking malaria parasite prevalence and severe disease outcomes in children.

Prevalence and fluid management of dehydration in children without diarrhoea admitted to Kenyan hospitals: a multisite observational study

OBJECTIVES

To examine the prevalence of dehydration without diarrhoea among admitted children aged 1-59 months and to describe fluid management practices in such cases.

DESIGN

A multisite observational study that used routine in-patient data collected prospectively between October 2013 and December 2018.

SETTINGS

Study conducted in 13 county referral hospitals in Kenya.

PARTICIPANTS

Children aged 1-59 months with admission or discharge diagnosis of dehydration but had no diarrhoea as a symptom or diagnosis. Children aged <28 days and those with severe acute malnutrition were excluded.

RESULTS

The prevalence of dehydration in children without diarrhoea was 3.0% (2019/68 204) and comprised 15.9% (2019/12 702) of all dehydration cases. Only 55.8% (1127/2019) of affected children received either oral or intravenous fluid therapy. Where fluid treatment was given, the volumes, type of fluid, duration of fluid therapy and route of administration were similar to those used in the treatment of dehydration secondary to diarrhoea. Pneumonia (1021/2019, 50.6%) and malaria (715/2019, 35.4%) were the two most common comorbid diagnoses. Overall case fatality in the study population was 12.9% (260/2019).

CONCLUSION

Sixteen per cent of children hospitalised with dehydration do not have diarrhoea but other common illnesses. Two-fifths do not receive fluid therapy; a regimen similar to that used in diarrhoeal cases is used in cases where fluid is administered. Efforts to promote compliance with guidance in routine clinical settings should recognise special circumstances where guidelines do not apply, and further studies on appropriate management for dehydration in the absence of diarrhoea are required.

A bibliometric analysis of COVID-19 research in Africa

BACKGROUND

The COVID-19 pandemic has led to an unprecedented global research effort to build a body of knowledge that can inform mitigation strategies. We carried out a bibliometric analysis to describe the COVID-19 research output in Africa in terms of setting, study design, research themes and author affiliation.

METHODS

We searched for articles published between 1 December 2019 and 3 January 2021 from various databases including PubMed, African Journals Online, medRxiv, Collabovid, the WHO global research database and Google. All article types and study design were included.

RESULTS

A total of 1296 articles were retrieved. 46.6% were primary research articles, 48.6% were editorial-type articles while 4.6% were secondary research articles. 20.3% articles used the entire continent of Africa as their study setting while South Africa (15.4%) was the most common country-focused setting. The most common research topics include 'country preparedness and response' (24.9%) and 'the direct and indirect health impacts of the pandemic' (21.6%). However, only 1.0% of articles focus on therapeutics and vaccines. 90.3% of the articles had at least one African researcher as author, 78.5% had an African researcher as first author, while 63.5% had an African researcher as last author. The University of Cape Town leads with the greatest number of first and last authors. 13% of the articles were published in medRxiv and of the studies that declared funding, the Wellcome Trust was the top funding body.

CONCLUSIONS

This study highlights Africa's COVID-19 research and the continent's existing capacity to carry out research that addresses local problems. However, more studies focused on vaccines and therapeutics are needed to inform local development. In addition, the uneven distribution of research productivity among African countries emphasises the need for increased investment where needed.

Neonatal mortality in Kenyan hospitals: a multisite, retrospective, cohort study

BACKGROUND

Most of the deaths among neonates in low-income and middle-income countries (LMICs) can be prevented through universal access to basic high-quality health services including essential facility-based inpatient care. However, poor routine data undermines data-informed efforts to monitor and promote improvements in the quality of newborn care across hospitals.

METHODS

Continuously collected routine patients' data from structured paper record forms for all admissions to newborn units (NBUs) from 16 purposively selected Kenyan public hospitals that are part of a clinical information network were analysed together with data from all paediatric admissions ages 0-13 years from 14 of these hospitals. Data are used to show the proportion of all admissions and deaths in the neonatal age group and examine morbidity and mortality patterns, stratified by birth weight, and their variation across hospitals.

FINDINGS

During the 354 hospital months study period, 90 222 patients were admitted to the 14 hospitals contributing NBU and general paediatric ward data. 46% of all the admissions were neonates (aged 0-28 days), but they accounted for 66% of the deaths in the age group 0-13 years. 41 657 inborn neonates were admitted in the NBUs across the 16 hospitals during the study period. 4266/41 657 died giving a crude mortality rate of 10.2% (95% CI 9.97% to 10.55%), with 60% of these deaths occurring on the first-day of admission. Intrapartum-related complications was the single most common diagnosis among the neonates with birth weight of 2000 g or more who died. A threefold variation in mortality across hospitals was observed for birth weight categories 1000-1499 g and 1500-1999 g.

INTERPRETATION

The high proportion of neonatal deaths in hospitals may reflect changing patterns of childhood mortality. Majority of newborns died of preventable causes (>95%). Despite availability of high-impact low-cost interventions, hospitals have high and very variable mortality proportions after stratification by birth weight.

Examining unit costs for COVID-19 case management in Kenya

Introduction

We estimated unit costs for COVID-19 case management for patients with asymptomatic, mild-to-moderate, severe and critical COVID-19 disease in Kenya.

Methods

We estimated per-day unit costs of COVID-19 case management for patients. We used a bottom-up approach to estimate full economic costs and adopted a health system perspective and patient episode of care as our time horizon. We obtained data on inputs and their quantities from data provided by three public COVID-19 treatment hospitals in Kenya and augmented this with guidelines. We obtained input prices from a recent costing survey of 20 hospitals in Kenya and from market prices for Kenya.

Results

Per-day, per-patient unit costs for asymptomatic patients and patients with mild-to-moderate COVID-19 disease under home-based care are 1993.01 Kenyan shilling (KES) (US$18.89) and 1995.17 KES (US$18.991), respectively. When these patients are managed in an isolation centre or hospital, the same unit costs for asymptomatic patients and patients with mild-to-moderate disease are 6717.74 KES (US$63.68) and 6719.90 KES (US$63.70), respectively. Per-day unit costs for patients with severe COVID-19 disease managed in general hospital wards and those with critical COVID-19 disease admitted in intensive care units are 13 137.07 KES (US$124.53) and 63 243.11 KES (US$599.51).

Conclusion

COVID-19 case management costs are substantial, ranging between two and four times the average claims value reported by Kenya’s public health insurer. Kenya will need to mobilise substantial resources and explore service delivery adaptations that will reduce unit costs.

Derivation and internal validation of a data-driven prediction model to guide frontline health workers in triaging children under-five in Nairobi, Kenya

Background: Many hospitalized children in developing countries die from infectious diseases. Early recognition of those who are critically ill coupled with timely treatment can prevent many deaths. A data-driven, electronic triage system to assist frontline health workers in categorizing illness severity is lacking. This study aimed to develop a data-driven parsimonious triage algorithm for children under five years of age. Methods: This was a prospective observational study of children under-five years of age presenting to the outpatient department of Mbagathi Hospital in Nairobi, Kenya between January and June 2018. A study nurse examined participants and recorded history and clinical signs and symptoms using a mobile device with an attached low-cost pulse oximeter sensor. The need for hospital admission was determined independently by the facility clinician and used as the primary outcome in a logistic predictive model. We focused on the selection of variables that could be quickly and easily assessed by low skilled health workers. Results: The admission rate (for more than 24 hours) was 12% (N=138/1,132). We identified an eight-predictor logistic regression model including continuous variables of weight, mid-upper arm circumference, temperature, pulse rate, and transformed oxygen saturation, combined with dichotomous signs of difficulty breathing, lethargy, and inability to drink or breastfeed. This model predicts overnight hospital admission with an area under the receiver operating characteristic curve of 0.88 (95% CI 0.82 to 0.94). Low- and high-risk thresholds of 5% and 25%, respectively were selected to categorize participants into three triage groups for implementation.  Conclusion: A logistic regression model comprised of eight easily understood variables may be useful for triage of children under the age of five based on the probability of need for admission. This model could be used by frontline workers with limited skills in assessing children. External validation is needed before adoption in clinical practice.

Employing learning health system principles to advance research on severe neonatal and paediatric illness in Kenya

We have worked to develop a Clinical Information Network (CIN) in Kenya as an early form of learning health systems (LHS) focused on paediatric and neonatal care that now spans 22 hospitals. CIN's aim was to examine important outcomes of hospitalisation at scale, identify and ultimately solve practical problems of service delivery, drive improvements in quality and test interventions. By including multiple routine settings in research, we aimed to promote generalisability of findings and demonstrate potential efficiencies derived from LHS. We illustrate the nature and range of research CIN has supported over the past 7 years as a form of LHS. Clinically, this has largely focused on common, serious paediatric illnesses such as pneumonia, malaria and diarrhoea with dehydration with recent extensions to neonatal illnesses. CIN also enables examination of the quality of care, for example that provided to children with severe malnutrition and the challenges encountered in routine settings in adopting simple technologies (pulse oximetry) and more advanced diagnostics (eg, Xpert MTB/RIF). Although regular feedback to hospitals has been associated with some improvements in quality data continue to highlight system challenges that undermine provision of basic, quality care (eg, poor access to blood glucose testing and routine microbiology). These challenges include those associated with increased mortality risk (eg, delays in blood transfusion). Using the same data the CIN platform has enabled conduct of randomised trials and supports malaria vaccine and most recently COVID-19 surveillance. Employing LHS principles has meant engaging front-line workers, clinical managers and national stakeholders throughout. Our experience suggests LHS can be developed in low and middle-income countries that efficiently enable contextually appropriate research and contribute to strengthening of health services and research systems.

Derivation and internal validation of a data-driven prediction model to guide frontline health workers in triaging children under-five in Nairobi, Kenya

Background: Many hospitalized children in developing countries die from infectious diseases. Early recognition of those who are critically ill coupled with timely treatment can prevent many deaths. A data-driven, electronic triage system to assist frontline health workers in categorizing illness severity is lacking. This study aimed to develop a data-driven parsimonious triage algorithm for children under five years of age. Methods: This was a prospective observational study of children under-five years of age presenting to the outpatient department of Mbagathi Hospital in Nairobi, Kenya between January and June 2018. A study nurse examined participants and recorded history and clinical signs and symptoms using a mobile device with an attached low-cost pulse oximeter sensor. The need for hospital admission was determined independently by the facility clinician and used as the primary outcome in a logistic predictive model. We focused on the selection of variables that could be quickly and easily assessed by low skilled health workers. Results: The admission rate (for more than 24 hours) was 12% (N=138/1,132). We identified an eight-predictor logistic regression model including continuous variables of weight, mid-upper arm circumference, temperature, pulse rate, and transformed oxygen saturation, combined with dichotomous signs of difficulty breathing, lethargy, and inability to drink or breastfeed. This model predicts overnight hospital admission with an area under the receiver operating characteristic curve of 0.88 (95% CI 0.82 to 0.94). Low- and high-risk thresholds of 5% and 25%, respectively were selected to categorize participants into three triage groups for implementation.  Conclusion: A logistic regression model comprised of eight easily understood variables may be useful for triage of children under the age of five based on the probability of need for admission. This model could be used by frontline workers with limited skills in assessing children. External validation is needed before adoption in clinical practice.

Prognostic models for predicting in-hospital paediatric mortality in resource-limited countries: a systematic review

OBJECTIVES

To identify and appraise the methodological rigour of multivariable prognostic models predicting in-hospital paediatric mortality in low-income and middle-income countries (LMICs).

DESIGN

Systematic review of peer-reviewed journals.

DATA SOURCES

MEDLINE, CINAHL, Google Scholar and Web of Science electronic databases since inception to August 2019.

ELIGIBILITY CRITERIA

We included model development studies predicting in-hospital paediatric mortality in LMIC.

DATA EXTRACTION AND SYNTHESIS

This systematic review followed the Checklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies framework. The risk of bias assessment was conducted using Prediction model Risk of Bias Assessment Tool (PROBAST). No quantitative summary was conducted due to substantial heterogeneity that was observed after assessing the studies included.

RESULTS

Our search strategy identified a total of 4054 unique articles. Among these, 3545 articles were excluded after review of titles and abstracts as they covered non-relevant topics. Full texts of 509 articles were screened for eligibility, of which 15 studies reporting 21 models met the eligibility criteria. Based on the PROBAST tool, risk of bias was assessed in four domains; participant, predictors, outcome and analyses. The domain of statistical analyses was the main area of concern where none of the included models was judged to be of low risk of bias.

CONCLUSION

This review identified 21 models predicting in-hospital paediatric mortality in LMIC. However, most reports characterising these models are of poor quality when judged against recent reporting standards due to a high risk of bias. Future studies should adhere to standardised methodological criteria and progress from identifying new risk scores to validating or adapting existing scores.

PROSPERO REGISTRATION NUMBER

CRD42018088599.

The Clinical Profile of Severe Pediatric Malaria in an Area Targeted for Routine RTS,S/AS01 Malaria Vaccination in Western Kenya

BACKGROUND

The malaria prevalence has declined in western Kenya, resulting in the risk of neurological phenotypes in older children. This study investigates the clinical profile of pediatric malaria admissions ahead of the introduction of the RTS,S/AS01 vaccine.

METHODS

Malaria admissions in children aged 1 month to 15 years were identified from routine, standardized, inpatient clinical surveillance data collected between 2015 and 2018 from 4 hospitals in western Kenya. Malaria phenotypes were defined based on available data.

RESULTS

There were 5766 malaria admissions documented. The median age was 36 months (interquartile range, 18-60): 15% were aged between 1-11 months of age, 33% were aged 1-23 months of age, and 70% were aged 1 month to 5 years. At admission, 2340 (40.6%) children had severe malaria: 421/2208 (19.1%) had impaired consciousness, 665/2240 (29.7%) had an inability to drink or breastfeed, 317/2340 (13.6%) had experienced 2 or more convulsions, 1057/2340 (45.2%) had severe anemia, and 441/2239 (19.7%) had severe respiratory distress. Overall, 211 (3.7%) children admitted with malaria died; 163/211 (77% deaths, case fatality rate 7.0%) and 48/211 (23% deaths, case fatality rate 1.4%) met the criteria for severe malaria and nonsevere malaria at admission, respectively. The median age for fatal cases was 33 months (interquartile range, 12-72) and the case fatality rate was highest in those unconscious (44.4%).

CONCLUSIONS

Severe malaria in western Kenya is still predominantly seen among the younger pediatric age group and current interventions targeted for those <5 years are appropriate. However, there are increasing numbers of children older than 5 years admitted with malaria, and ongoing hospital surveillance would identify when interventions should target older children.

Examining which clinicians provide admission hospital care in a high mortality setting and their adherence to guidelines: an observational study in 13 hospitals

BACKGROUND

We explored who actually provides most admission care in hospitals offering supervised experiential training to graduating clinicians in a high mortality setting where practices deviate from guideline recommendations.

METHODS

We used a large observational data set from 13 Kenyan county hospitals from November 2015 through November 2018 where patients were linked to admitting clinicians. We explored guideline adherence after creating a cumulative correctness of Paediatric Admission Quality of Care (cPAQC) score on a 5-point scale (0-4) in which points represent correct, sequential progress in providing care perfectly adherent to guidelines comprising admission assessment, diagnosis and treatment. At the point where guideline adherence declined the most we dichotomised the cPAQC score and used multilevel logistic regression models to explore whether clinician and patient-level factors influence adherence.

RESULTS

There were 1489 clinicians who could be linked to 53 003 patients over a period of 3 years. Patients were rarely admitted by fully qualified clinicians and predominantly by preregistration medical officer interns (MOI, 46%) and diploma level clinical officer interns (COI, 41%) with a median of 28 MOI (range 11-68) and 52 COI (range 5-160) offering care per study hospital. The cPAQC scores suggest that perfect guideline adherence is found in ≤12% of children with malaria, pneumonia or diarrhoea with dehydration. MOIs were more adherent to guidelines than COI (adjusted OR 1.19 (95% CI 1.07 to 1.34)) but multimorbidity was significantly associated with lower guideline adherence.

CONCLUSION

Over 85% of admissions to hospitals in high mortality settings that offer experiential training in Kenya are conducted by preregistration clinicians. Clinical assessment is good but classifying severity of illness in accordance with guideline recommendations is a challenge. Adherence by MOI with 6 years' training is better than COI with 3 years' training, performance does not seem to improve during their 3 months of paediatric rotations.

Smart triage: triage and management of sepsis in children using the point-of-care Pediatric Rapid Sepsis Trigger (PRST) tool

BACKGROUND

Sepsis is the leading cause of death and disability in children. Every hour of delay in treatment is associated with an escalating risk of morbidity and mortality. The burden of sepsis is greatest in low- and middle-income countries where timely treatment may not occur due to delays in diagnosis and prioritization of critically ill children. To circumvent these challenges, we propose the development and clinical evaluation of a digital triage tool that will identify high risk children and reduce time to treatment. We will also implement and clinically validate a Radio-Frequency Identification system to automate tracking of patients. The mobile platform (mobile device and dashboard) and automated patient tracking system will create a low cost, highly scalable solution for critically ill children, including those with sepsis.

METHODS

This is pre-post intervention study consisting of three phases. Phase I will be a baseline period where data is collected on key predictors and outcomes before implementation of the digital triage tool. In Phase I, there will be no changes to healthcare delivery processes in place at the study hospitals. Phase II will involve model derivation, technology development, and usability testing. Phase III will be the intervention period where data is collected on key predictors and outcomes after implementation of the digital triage tool. The primary outcome, time to treatment initiation, will be compared to assess effectiveness of the digital health intervention.

DISCUSSION

Smart technology has the potential to overcome the barrier of limited clinical expertise in the identification of the child at risk. This mobile health platform, with sensors and data-driven applications, will provide real-time individualized risk prediction to rapidly triage patients and facilitate timely access to life-saving treatments for children in low- and middle-income countries, where specialists are not regularly available and deaths from sepsis are common.

TRIAL REGISTRATION

Clinical Trials.gov Identifier: NCT04304235, Registered 11 March 2020.

Methodological rigor of prognostic models for predicting in-hospital paediatric mortality in low- and middle-income countries: a systematic review protocol

Introduction: In low- and middle-income countries (LMICs) where healthcare resources are often limited, making decisions on appropriate treatment choices is critical in ensuring reduction of paediatric deaths as well as instilling proper utilisation of the already constrained healthcare resources. Well-developed and validated prognostic models can aid in early recognition of potential risks thus contributing to the reduction of mortality rates. The aim of the planned systematic review is to identify and appraise the methodological rigor of multivariable prognostic models predicting in-hospital paediatric mortality in LMIC in order to identify statistical and methodological shortcomings deserving special attention and to identify models for external validation. Methods and analysis: This protocol has followed the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Protocols. A search of articles will be conducted in MEDLINE, Google Scholar, and CINAHL (via EbscoHost) from inception to 2019 without any language restriction. We will also perform a search in Web of Science to identify additional reports that cite the identified studies. Data will be extracted from relevant articles in accordance with the Cochrane Prognosis Methods' guidance; the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies. Methodological quality assessment will be performed based on prespecified domains of the Prediction study Risk of Bias Assessment Tool. Ethics and dissemination: Ethical permission will not be required as this study will use published data. Findings from this review will be shared through publication in peer-reviewed scientific journals and, presented at conferences. It is our hope that this study will contribute to the development of robust multivariable prognostic models predicting in-hospital paediatric mortality in low- and middle-income countries. Registration: PROSPERO ID CRD42018088599; registered on 13 February 2018.

Examining the effectiveness of zinc treatment in children admitted with diarrhoea in Kenya's public hospitals: an observational comparative effectiveness study

Background

Kenyan paediatric treatment protocols recommend the use of zinc supplement for all children with diarrhoea. However, there is limited evidence of benefit for young children aged 1-5 months and those who are well-nourished. We examine effectiveness of zinc supplementation for children admitted with diarrhoea to Kenya’s public hospitals with different nutritional and age categories. This is to determine whether the current policy where zinc is prescribed for all children with diarrhoea is appropriate.

Methods

We explore the effect of zinc treatment on time to discharge for children aged 1-5 and 6-59 months and amongst those classified as either severely – moderately under-nourished or well-nourished. To overcome the challenges associated with non-random allocation of treatments and missing data in these observational data, we use propensity score methods and multiple imputation to minimize bias.

Results

The analysis included 1645 (1-5 months) and 11 546 (6-59 months) children respectively. The estimated sub-distribution hazard ratios for being discharged in the zinc group vs the non-zinc group were 1.25 (95% confidence interval (CI) = 1.07, 1.46) and 1.17 (95% CI = 1.10, 1.24) in these respective age categories. Zinc treatment was associated with shorter time to discharge in both well and under-nourished children.

Conclusion

Zinc treatment, in general, was associated with shorter time to discharge. In the absence of significant adverse effects, these data support the continued use of zinc for admissions with diarrhoea including those aged 1-5 months and in those who are well-nourished.

Risk factors for death among children aged 5-14 years hospitalised with pneumonia: a retrospective cohort study in Kenya

INTRODUCTION

There were almost 1 million deaths in children aged between 5 and 14 years in 2017, and pneumonia accounted for 11%. However, there are no validated guidelines for pneumonia management in older children and data to support their development are limited. We sought to understand risk factors for mortality among children aged 5-14 years hospitalised with pneumonia in district-level health facilities in Kenya.

METHODS

We did a retrospective cohort study using data collected from an established clinical information network of 13 hospitals. We reviewed records for children aged 5-14 years admitted with pneumonia between 1 March 2014 and 28 February 2018. Individual clinical signs were examined for association with inpatient mortality using logistic regression. We used existing WHO criteria (intended for under 5s) to define levels of severity and examined their performance in identifying those at increased risk of death.

RESULTS

1832 children were diagnosed with pneumonia and 145 (7.9%) died. Severe pallor was strongly associated with mortality (adjusted OR (aOR) 8.06, 95% CI 4.72 to 13.75) as were reduced consciousness, mild/moderate pallor, central cyanosis and older age (>9 years) (aOR >2). Comorbidities HIV and severe acute malnutrition were also associated with death (aOR 2.31, 95% CI 1.39 to 3.84 and aOR 1.89, 95% CI 1.12 to 3.21, respectively). The presence of clinical characteristics used by WHO to define severe pneumonia was associated with death in univariate analysis (OR 2.69). However, this combination of clinical characteristics was poor in discriminating those at risk of death (sensitivity: 0.56, specificity: 0.68, and area under the curve: 0.62).

CONCLUSION

Children >5 years have high inpatient pneumonia mortality. These findings also suggest that the WHO criteria for classification of severity for children under 5 years do not appear to be a valid tool for risk assessment in this older age group, indicating the urgent need for evidence-based clinical guidelines for this neglected population.

Derivation and internal validation of a data-driven prediction model to guide frontline health workers in triaging children under-five in Nairobi, Kenya

Background: Many hospitalized children in developing countries die from infectious diseases. Early recognition of those who are critically ill coupled with timely treatment can prevent many deaths. A data-driven, electronic triage system to assist frontline health workers in categorizing illness severity is lacking. This study aimed to develop a data-driven parsimonious triage algorithm for children under five years of age. Methods: This was a prospective observational study of children under-five years of age presenting to the outpatient department of Mbagathi Hospital in Nairobi, Kenya between January and June 2018. A study nurse examined participants and recorded history and clinical signs and symptoms using a mobile device with an attached low-cost pulse oximeter sensor. The need for hospital admission was determined independently by the facility clinician and used as the primary outcome in a logistic predictive model. We focused on the selection of variables that could be quickly and easily assessed by low skilled health workers. Results: The admission rate (for more than 24 hours) was 12% (N=138/1,132). We identified an eight-predictor logistic regression model including continuous variables of weight, mid-upper arm circumference, temperature, pulse rate, and transformed oxygen saturation, combined with dichotomous signs of difficulty breathing, lethargy, and inability to drink or breastfeed. This model predicts overnight hospital admission with an area under the receiver operating characteristic curve of 0.88 (95% CI 0.82 to 0.94). Low- and high-risk thresholds of 5% and 25%, respectively were selected to categorize participants into three triage groups for implementation.  Conclusion: A logistic regression model comprised of eight easily understood variables may be useful for triage of children under the age of five based on the probability of need for admission. This model could be used by frontline workers with limited skills in assessing children. External validation is needed before adoption in clinical practice.

Effect of enhancing audit and feedback on uptake of childhood pneumonia treatment policy in hospitals that are part of a clinical network: a cluster randomized trial

BACKGROUND

The World Health Organization (WHO) revised its clinical guidelines for management of childhood pneumonia in 2013. Significant delays have occurred during previous introductions of new guidelines into routine clinical practice in low- and middle-income countries (LMIC). We therefore examined whether providing enhanced audit and feedback as opposed to routine standard feedback might accelerate adoption of the new pneumonia guidelines by clinical teams within hospitals in a low-income setting.

METHODS

In this parallel group cluster randomized controlled trial, 12 hospitals were assigned to either enhanced feedback (n = 6 hospitals) or standard feedback (n = 6 hospitals) using restricted randomization. The standard (network) intervention delivered in both trial arms included support to improve collection and quality of patient data, provision of mentorship and team management training for pediatricians, peer-to-peer networking (meetings and social media), and multimodal (print, electronic) bimonthly hospital specific feedback reports on multiple indicators of evidence guideline adherence. In addition to this network intervention, the enhanced feedback group received a monthly hospital-specific feedback sheet targeting pneumonia indicators presented in multiple formats (graphical and text) linked to explicit performance goals and action plans and specific email follow up from a network coordinator. At the start of the trial, all hospitals received a standardized training on the new guidelines and printed booklets containing pneumonia treatment protocols. The primary outcome was the proportion of children admitted with indrawing and/or fast-breathing pneumonia who were correctly classified using new guidelines and received correct antibiotic treatment (oral amoxicillin) in the first 24 h. The secondary outcome was the proportion of correctly classified and treated children for whom clinicians changed treatment from oral amoxicillin to injectable antibiotics.

RESULTS

The trial included 2299 childhood pneumonia admissions, 1087 within the hospitals randomized to enhanced feedback intervention, and 1212 to standard feedback. The proportion of children who were correctly classified and treated in the first 24 h during the entire 9-month period was 38.2% (393 out of 1030) and 38.4% (410 out of 1068) in the enhanced feedback and standard feedback groups, respectively (odds ratio 1.11; 95% confidence interval [CI] 0.37-3.34; P = 0.855). However, in exploratory analyses, there was evidence of an interaction between type of feedback and duration (in months) since commencement of intervention, suggesting a difference in adoption of pneumonia policy over time in the enhanced compared to standard feedback arm (OR = 1.25, 95% CI 1.14 to 1.36, P < 0.001).

CONCLUSIONS

Enhanced feedback comprising increased frequency, clear messaging aligned with goal setting, and outreach from a coordinator did not lead to a significant overall effect on correct pneumonia classification and treatment during the 9-month trial. There appeared to be a significant effect of time (representing cumulative effect of feedback cycles) on adoption of the new policy in the enhanced feedback compared to standard feedback group. Future studies should plan for longer follow-up periods to confirm these findings.

TRIAL REGISTRATION

US National Institutes of Health-ClinicalTrials.gov identifier (NCT number) NCT02817971 . Registered September 28, 2016-retrospectively registered.

Assessment of Myocardial Function in Kenyan Children With Severe, Acute Malnutrition: The Cardiac Physiology in Malnutrition (CAPMAL) Study

Importance

Mortality among African children hospitalized with severe malnutrition remains high, with sudden, unexpected deaths leading to speculation about potential cardiac causes. Malnutrition is considered high risk for cardiac failure, but evidence is limited.

Objective

To investigate the role of cardiovascular dysfunction in African children with severe, acute malnutrition (SAM).

Design, Setting, and Participants

A prospective, matched case-control study, the Cardiac Physiology in Malnutrition (CAPMAL) study, of 88 children with SAM (exposed) vs 22 severity-matched patients without SAM (unexposed) was conducted between March 7, 2011, and February 20, 2012; data analysis was performed from October 1, 2012, to March 1, 2016.

Exposures

Echocardiographic and electrocardiographic (ECG) recordings (including 7-day Holter monitoring) at admission, day 7, and day 28.

Main Outcomes and Measures

Findings in children with (cases) and without (controls) SAM and in marasmus and kwashiorkor phenotypes were compared.

Results

Eighty-eight children (52 with marasmus and 36 with kwashiorkor) of the 418 admitted with SAM and 22 severity-matched controls were studied. A total of 63 children (57%) were boys; median age at admission was 19 months (range, 12-39 months). On admission, abnormalities more common in cases vs controls included severe hypokalemia (potassium <2.5 mEq/L) (18 of 81 [22%] vs 0%), hypoalbuminemia (albumin level <3.4 g/dL) (66 of 88 [75%] vs 4 of 22 [18%]), and hypothyroidism (free thyroxine level <0.70 ng/dL or thyrotropin level >4.2 mU/L) (18 of 74 [24%] vs 1 of 21 [5%]) and were associated with typical electrocardiographic changes (T-wave inversion: odds ratio, 7.3; 95% CI, 1.9-28.0; P = .001), which corrected as potassium levels improved. Fourteen children with SAM (16%) but no controls died. Myocardial mass was lower in cases on admission but not by day 7. Results of the Tei Index, a measure of global cardiac function, were within the reference range and similar in cases (median, 0.37; interquartile range [IQR], 0.26-0.45) and controls (median, 0.36; IQR, 0.28-0.42). Echocardiography detected no evidence of cardiac failure among children with SAM, including those receiving intravenous fluids to correct hypovolemia. Cardiac dysfunction was generally associated with comorbidity and typical of hypovolemia, with low cardiac index (median, 4.9 L/min/m2; IQR, 3.9-6.1 L/min/m2), high systemic vascular resistance index (median, 1333 dyne seconds/cm5/m2; IQR, 1133-1752 dyne seconds/cm5/m2), and with few differences between the marasmus and kwashiorkor manifestations of malnutrition. Seven-day continuous ECG Holter monitoring during the high-risk initial refeeding period demonstrated self-limiting significant ventricular arrhythmias in 33 of 55 cases (60%) and 6 of 18 controls (33%) (P = .049); none were temporally related to adverse events, including fatalities.

Conclusions and Relevance

There is little evidence that African children with SAM are at greater risk of cardiac dysfunction or clinically significant arrhythmias than those without SAM or that marasmus and kwashiorkor differed in cardiovascular profile. These findings should prompt a review of current guidelines.

Determining predictors of sepsis at triage among children under 5 years of age in resource-limited settings: A modified Delphi process

Sepsis is a life-threatening dysfunction of the immune system leading to multiorgan failure that is precipitated by infectious diseases and is a leading cause of death in children under 5 years of age. It is necessary to be able to identify a sick child at risk of developing sepsis at the earliest point of presentation to a healthcare facility so that appropriate care can be provided as soon as possible. Our study objective was to generate a list of consensus-driven predictor variables for the derivation of a prediction model that will be incorporated into a mobile device and operated by low-skilled healthcare workers at triage. By conducting a systematic literature review and examination of global guideline documents, a list of 72 initial candidate predictor variables was generated. A two-round modified Delphi process involving 26 experts from both resource-rich and resource-limited settings, who were also encouraged to suggest new variables, yielded a final list of 45 predictor variables after evaluating each variable based on three domains: predictive potential, measurement reliability, and level of training and resources required. The final list of predictor variables will be used to collect data and contribute to the derivation of a prediction model.

Magnitude and pattern of improvement in processes of care for hospitalised children with diarrhoea and dehydration in Kenyan hospitals participating in a clinical network

OBJECTIVE

WHO recommends optimisation of available interventions to reduce deaths of under-five children with diarrhoea and dehydration (DD). Clinical networks may help improve practice across many hospitals but experience with such networks is scarce. We describe magnitude and patterns of changes in processes of care for children with DD over the first 3 years of a clinical network.

METHODS

Observational study involving children aged 2-59 months with DD admitted to 13 hospitals participating in the clinical network. Processes of individual patient care including agreement of assessment, diagnosis and treatment according to WHO guidelines were combined using the composite Paediatric Admission Quality of Care (PAQC) score (range 0-6).

RESULTS

Data from 7657 children were analysed and improvements in PAQC scores were observed. Predicted mean PAQC score for all the hospitals at enrolment was 59.8% (95% CI: 54.7, 64.9) but showed a wide variation (variance 10.7%, 95% CI: 5.8, 19.6). Overall mean PAQC score increased by 13.8% (95% CI: 8.7-18.9, SD between hospitals: ±8.2) in the first 12 months, with an average 0.9% (95% CI: 0.3-1.5, SD ± 1.0) increase per month and plateaued thereafter, and changes were similar in two groups of hospitals joining the network at different times.

CONCLUSION

Adherence to guidelines for children admitted with DD can be improved through participation in a clinical network but improvement is limited, not uniform for all aspects of care and contexts and occurs early. Future research should address these issues.

Risk factors for mortality and effect of correct fluid prescription in children with diarrhoea and dehydration without severe acute malnutrition admitted to Kenyan hospitals: an observational, association study

BACKGROUND

Diarrhoea causes many deaths in children younger than 5 years and identification of risk factors for death is considered a global priority. The effectiveness of currently recommended fluid management for dehydration in routine settings has also not been examined.

METHODS

For this observational, association study, we analysed prospective clinical data on admission, immediate treatment, and discharge of children age 1-59 months with diarrhoea and dehydration, which were routinely collected from 13 Kenyan hospitals. We analysed participants with full datasets using multivariable mixed-effects logistic regression to assess risk factors for in-hospital death and effect of correct rehydration on early mortality (within 2 days).

FINDINGS

Between Oct 1, 2013, and Dec 1, 2016, 8562 children with diarrhoea and dehydration were admitted to hospital and eligible for inclusion in this analysis. Overall mortality was 9% (759 of 8562 participants) and case fatality was directly correlated with severity. Most children (7184 [84%] of 8562) with diarrhoea and dehydration had at least one additional diagnosis (comorbidity). Age of 12 months or younger (adjusted odds ratio [AOR] 1·71, 95% CI 1·42-2·06), female sex (1·41, 1·19-1·66), diarrhoea duration of more than 14 days (2·10, 1·42-3·12), abnormal respiratory signs (3·62, 2·95-4·44), abnormal circulatory signs (2·29, 1·89-2·77), pallor (2·15, 1·76-2·62), use of intravenous fluid (proxy for severity; 1·68, 1·41-2·00), and abnormal neurological signs (3·07, 2·54-3·70) were independently associated with in-hospital mortality across hospitals. Signs of dehydration alone were not associated with in-hospital deaths (AOR 1·08, 0·87-1·35). Correct fluid prescription significantly reduced the risk of early mortality (within 2 days) in all subgroups: abnormal respiratory signs (AOR 1·23, 0·68-2·24), abnormal circulatory signs (0·95, 0·53-1·73), pallor (1·70, 0·95-3·02), dehydration signs only (1·50, 0·79-2·88), and abnormal neurological signs (0·86, 0·51-1·48).

INTERPRETATION

Children at risk of in-hospital death are those with complex presentations rather than uncomplicated dehydration, and the prescription of recommended rehydration guidelines reduces risk of death. Strategies to optimise the delivery of recommended guidance should be accompanied by studies on the management of dehydration in children with comorbidities, the vulnerability of young girls, and the delivery of immediate care.

FUNDING

The Wellcome Trust.

The Prevalence and Management of Dehydration amongst Neonatal Admissions to General Paediatric Wards in Kenya-A Clinical Audit

An audit of randomly selected case records of 810 patients admitted to 13 hospitals between December 2015 and November 2016 was done. Prevalence of dehydration was 19.7% (2293 of 11 636) [95% CI: 17.1-22.6%], range across hospitals was 9.4% to 27.0%. Most cases with dehydration were clinically diagnosed (82 of 153; 53.6%), followed by excessive weight loss (54 of 153; 35.3%) and abnormal urea/electrolytes/creatinine (23 of 153; 15.0%). Documentation of fluids prescribed was poor but, where data were available, Ringers lactate (30 of 153; 19.6%) and 10% dextrose (18 of 153; 11.8%) were mostly used. Only 17 of 153 (11.1%) children had bolus fluid prescription, and Ringer's lactate was most commonly used for bolus at a median volume per kilogram body weight of 20 ml/kg (interquartile range, 12-30 ml/kg). Neonatal dehydration is common, but current documentation may underestimate the burden. Heterogeneity in practice likely reflects the absence of guidelines that in turn reflects a lack of research informing practical treatment guidelines.

High Frequency of Blackwater Fever Among Children Presenting to Hospital With Severe Febrile Illnesses in Eastern Uganda

Background

In the Fluid Expansion as a Supportive Treatment (FEAST) trial, an unexpectedly high proportion of participants from eastern Uganda presented with blackwater fever (BWF).

Methods

We describe the prevalence and outcome of BWF among trial participants and compare the prevalence of 3 malaria-protective red blood cell polymorphisms in BWF cases vs both trial (non-BWF) and population controls.

Results

Of 3170 trial participants, 394 (12.4%) had BWF. The majority (318 [81.0%]) presented in eastern Uganda and were the subjects of further analysis. BWF cases typically presented with both clinical jaundice (254/318 [80%]) and severe anemia (hemoglobin level <5 g/dL) (238/310 [77%]). Plasmodium falciparum parasitemia was less frequent than in non-BWF controls, but a higher proportion were positive for P. falciparum histidine rich protein 2 (192/246 [78.0%]) vs 811/1154 [70.3%]; P = .014), suggesting recent antimalarial treatment. Overall, 282 of 318 (88.7%) received transfusions, with 94 of 282 (33.3%) and 9 of 282 (3.4%) receiving 2 or 3 transfusions, respectively. By day 28, 39 of 318 (12.3%) BWF cases and 154 of 1554 (9.9%) non-BWF controls had died (P = .21), and 7 of 255 (3.0%) vs 13/1212 (1%), respectively, had severe anemia (P = .036). We found no association with G6PD deficiency. The prevalence of both the sickle cell trait (10/218 [4.6%]) and homozygous α+thalassemia (8/216 [3.7%]) were significantly lower among cases than among population controls (334/2123 [15.7%] and 141/2114 [6.6%], respectively), providing further support for the role of malaria.

Conclusions

We report the emergence of BWF in eastern Uganda, a condition that, according to local investigators, was rare until the last 7 years. We speculate that this might relate to the introduction of artemisinin-based combination therapies. Further studies investigating this possibility are urgently required.

Does audit and feedback improve the adoption of recommended practices? Evidence from a longitudinal observational study of an emerging clinical network in Kenya

Background

Audit and feedback (A&F) is widely used in healthcare but there are few examples of how to deploy it at scale in low-income countries. Establishing the Clinical Information Network (CIN) in Kenya provided an opportunity to examine the effect of A&F delivered as part of a wider set of activities to promote paediatric guideline adherence.

Methods

We analysed data collected from medical records on discharge for children aged 2–59 months from 14 Kenyan hospitals in the CIN. Hospitals joined CIN in phases and for each we analysed their initial 25 months of participation that occurred between December 2013 and March 2016. A total of 34 indicators of adherence to recommendations were selected for evaluation each classified by form of feedback (passive, active and none) and type of task (simple or difficult documentation and those requiring cognitive work). Performance change was explored graphically and using generalised linear mixed models with attention given to the effects of time and use of a standardised paediatric admission record (PAR) form.

Results

Data from 60 214 admissions were eligible for analysis. Adherence to recommendations across hospitals significantly improved for 24/34 indicators. Improvements were not obviously related to nature of feedback, may be related to task type and were related to PAR use in the case of documentation indicators. There was, however, marked variability in adoption and adherence to recommended practices across sites and indicators. Hospital-specific factors, low baseline performance and specific contextual changes appeared to influence the magnitude of change in specific cases.

Conclusion

Our observational data suggest some change in multiple indicators of adherence to recommendations (aspects of quality of care) can be achieved in low-resource hospitals using A&F and simple job aides in the context of a wider network approach.