Characteristics of children readmitted with severe pneumonia in Kenyan hospitals.

BACKGROUND: Pneumonia is a leading cause of childhood morbidity and mortality. Hospital re-admission may signify missed opportunities for care or undiagnosed comorbidities. METHODS: We conducted a retrospective cohort study including children aged ≥ 2 months-14 years hospitalised with severe pneumonia between 2013 and 2021 in a network of 20 primary referral hospitals in Kenya. Severe pneumonia was defined using the 2013 World Health Organization criteria, and re-admission was based on clinical documentation from individual patient case notes. We estimated the prevalence of re-admission, described clinical management practices, and modelled risk factors for re-admission and inpatient mortality. RESULTS: Among 20,603 children diagnosed with severe pneumonia, 2,274 (11.0%, 95% CI 10.6-11.5) were readmitted. Re-admission was independently associated with age (12-59 months vs. 2-11 months: adjusted odds ratio (aOR) 1.70, 1.54-1.87; >5 years vs. 2-11 months: aOR 1.85, 1.55-2.22), malnutrition (weight-for-age-z-score (WAZ) <-3SD vs. WAZ> -2SD: aOR 2.05, 1.84-2.29); WAZ - 2 to -3 SD vs. WAZ> -2SD: aOR 1.37, 1.20-1.57), wheeze (aOR 1.17, 1.03-1.33) and presence of a concurrent neurological disorder (aOR 4.42, 1.70-11.48). Chest radiography was ordered more frequently among those readmitted (540/2,274 [23.7%] vs. 3,102/18,329 [16.9%], p < 0.001). Readmitted patients more frequently received second-line antibiotics (808/2,256 [35.8%] vs. 5,538/18,173 [30.5%], p < 0.001), TB medication (69/2,256 [3.1%] vs. 298/18,173 [1.6%], p < 0.001), salbutamol (530/2,256 [23.5%] vs. 3,707/18,173 [20.4%], p = 0.003), and prednisolone (157/2,256 [7.0%] vs. 764/18,173 [4.2%], p < 0.001). Inpatient mortality was 2,354/18,329 (12.8%) among children admitted with a first episode of severe pneumonia and 269/2,274 (11.8%) among those who were readmitted (adjusted hazard ratio (aHR) 0.93, 95% CI 0.82-1.07). Age (12-59 months vs. 2-11 months: aHR 0.62, 0.57-0.67), male sex (aHR 0.81, 0.75-0.88), malnutrition (WAZ <-3SD vs. WAZ >-2SD: aHR 1.87, 1.71-2.05); WAZ - 2 to -3 SD vs. WAZ >-2SD: aHR 1.46, 1.31-1.63), complete vaccination (aHR 0.74, 0.60-0.91), wheeze (aHR 0.87, 0.78-0.98) and anaemia (aHR 2.14, 1.89-2.43) were independently associated with mortality. CONCLUSIONS: Children readmitted with severe pneumonia account for a substantial proportion of pneumonia hospitalisations and deaths. Further research is required to develop evidence-based approaches to screening, case management, and follow-up of children with severe pneumonia, prioritising those with underlying risk factors for readmission and mortality.

Hospital Mortality - a neglected but rich source of information supporting the transition to higher quality health systems in low and middle income countries.

BACKGROUND: There is increasing focus on the strength of primary health care systems in low and middle-income countries (LMIC). There are important roles for higher quality district hospital care within these systems. These hospitals are also sources of information of considerable importance to health systems, but this role, as with the wider roles of district hospitals, has been neglected. KEY MESSAGES: As we make efforts to develop higher quality health systems in LMIC we highlight the critical importance of district hospitals focusing here on how data on hospital mortality offers value: i) in understanding disease burden; ii) as part of surveillance and impact monitoring; iii) as an entry point to exploring system failures; and iv) as a lens to examine variability in health system performance and possibly as a measure of health system quality in its own right. However, attention needs paying to improving data quality by addressing reporting gaps and cause of death reporting. Ideally enabling the collection of basic, standardised patient level data might support at least simple case-mix and case-severity adjustment helping us understand variation. Better mortality data could support impact evaluation, benchmarking, exploration of links between health system inputs and outcomes and critical scrutiny of geographic variation in quality and outcomes of care. Improved hospital information is a neglected but broadly valuable public good. CONCLUSION: Accurate, complete and timely hospital mortality reporting is a key attribute of a functioning health system. It can support countries' efforts to transition to higher quality health systems in LMIC enabling national and local advocacy, accountability and action.

An exploration of mortality risk factors in non-severe pneumonia in children using clinical data from Kenya.

BACKGROUND: Childhood pneumonia is the leading infectious cause of mortality in children younger than 5 years old. Recent updates to World Health Organization pneumonia guidelines recommend outpatient care for a population of children previously classified as high risk. This revision has been challenged by policymakers in Africa, where mortality related to pneumonia is higher than in other regions and often complicated by comorbidities. This study aimed to identify factors that best discriminate inpatient mortality risk in non-severe pneumonia and explore whether these factors offer any added benefit over the current criteria used to identify children with pneumonia requiring inpatient care. METHODS: We undertook a retrospective cohort study of children aged 2-59 months admitted with a clinical diagnosis of pneumonia at 14 public hospitals in Kenya between February 2014 and February 2016. Using machine learning techniques, we analysed whether clinical characteristics and common comorbidities increased the risk of inpatient mortality for non-severe pneumonia. The topmost risk factors were subjected to decision curve analysis to explore if using them as admission criteria had any net benefit above the current criteria. RESULTS: Out of 16,162 children admitted with pneumonia during the study period, 10,687 were eligible for subsequent analysis. Inpatient mortality within this non-severe group was 252/10,687 (2.36%). Models demonstrated moderately good performance; the partial least squares discriminant analysis model had higher sensitivity for predicting mortality in comparison to logistic regression. Elevated respiratory rate (≥70 bpm), age 2-11 months and weight-for-age Z-score (WAZ) < -3SD were highly discriminative of mortality. These factors ranked consistently across the different models. For a risk threshold probability of 7-14%, there is a net benefit to admitting the patient sub-populations with these features as additional criteria alongside those currently used to classify severe pneumonia. Of the population studied, 70.54% met at least one of these criteria. Sensitivity analyses indicated that the overall results were not significantly affected by variations in pneumonia severity classification criteria. CONCLUSIONS: Children with non-severe pneumonia aged 2-11 months or with respiratory rate ≥ 70 bpm or very low WAZ experience risks of inpatient mortality comparable to severe pneumonia. Inpatient care is warranted in these high-risk groups of children.

Correction to: An exploration of mortality risk factors in non-severe pneumonia in children using clinical data from Kenya.

CORRECTION: The original article contains an omission in the Acknowledgements sub-section of the Declarations.

Inappropriate prescription of cough remedies among children hospitalised with respiratory illness over the period 2002-2015 in Kenya.

OBJECTIVE: To examine trends in prescription of cough medicines over the period 2002-2015 in children aged 1 month to 12 years admitted to Kenyan hospitals with cough, difficulty breathing or diagnosed with a respiratory tract infection. METHODS: We reviewed hospitalisation records of children included in four studies providing cross-sectional prevalence estimates from government hospitals for six time periods between 2002 and 2015. Children with an atopic illness were excluded. Amongst eligible children, we determined the proportion prescribed any adjuvant medication for cough. Active ingredients in these medicines were often multiple and were classified into five categories: antihistamines, antitussives, mucolytics/expectorants, decongestants and bronchodilators. From late 2006, guidelines discouraging cough medicine use have been widely disseminated and in 2009 national directives to decrease cough medicine use were issued. RESULTS: Across the studies, 17 963 children were eligible. Their median age and length of hospital stay were comparable. The proportion of children who received cough medicines shrank across the surveys: approximately 6% [95% CI: 5.4, 6.6] of children had a prescription in 2015 vs. 40% [95% CI: 35.5, 45.6] in 2002. The most common active ingredients were antihistamines and bronchodilators. The relative proportion that included antihistamines has increased over time. CONCLUSIONS: There has been an overall decline in the use of cough medicines among hospitalised children over time. This decline has been associated with educational, policy and mass media interventions.

Timing of Most Recent Health Care Visit by Older People Living With and Without HIV.

The objective of this article is to document factors associated with the recency of health-care service utilization by people aged 50 years and over living with and without HIV in Uganda. A survey was conducted with 510 Ugandans aged 50 and older, living with and without HIV. The survey included information on sociodemographic characteristics, health state, self-reported chronic conditions, and timing of most recent visit to a health-care facility (time since last visit [TSLV]). We use ordinal logistic regression to identify independent factors associated TSLV. Independent factors associated with TSLV (>6 months) include age, OR = 2.40 [95% CI 1.08-5.37] for those aged 80 years and above, urban respondents, OR = 0.6 [95%CI 0.38-0.94], HIV-positive respondents, OR = 0.33 [95%CI 0.18-0.59], and better health. To understand the meaning of these finding, further investigation should examine (a) how best to define and measure older persons' health-care service needs and (b) older persons' decision-making processes around the timing of their access to health-care facilities.

Quality of surgical care in hospitals providing internship training in Kenya: a cross sectional survey.

OBJECTIVE: To evaluate services in hospitals providing internship training to graduate doctors in Kenya. METHODS: A survey of 22 internship training hospitals was conducted. Availability of key resources spanning infrastructure, personnel, equipment and drugs was assessed by observation. Outcomes and process of care for pre-specified priority conditions (head injury, chest injury, fractures, burns and acute abdomen) were evaluated by auditing case records. RESULTS: Each hospital had at least one consultant surgeon. Scheduled surgical outpatient clinics, major ward rounds and elective (half day) theatre lists were provided once per week in 91%, 55% and 9%, respectively. In all other hospitals, these were conducted twice weekly. Basic drugs were not always available (e.g. gentamicin, morphine and pethidine in 50%, injectable antistaphylococcal penicillins in 5% hospitals). Fewer than half of hospitals had all resources needed to provide oxygen. One hundred and forty-five of 956 cases evaluated underwent operations under general or spinal anaesthesia. We found operation notes for 99% and anaesthetic records for 72%. Pre-operatively measured vital signs were recorded in 80% of cases, and evidence of consent to operation was found in 78%. Blood loss was documented in only one case and sponge and instrument counts in 7%. CONCLUSIONS: Evaluation of surgical services would be improved by development and dissemination of clear standards of care. This survey suggests that internship hospitals may be poorly equipped and documented care suggests inadequacies in quality and training.

Assessment of paediatric inpatient care during a multifaceted quality improvement intervention in Kenyan district hospitals--use of prospectively collected case record data.

BACKGROUND: In assessing quality of care in developing countries, retrospectively collected data are usually used given their availability. Retrospective data however suffer from such biases as recall bias and non-response bias. Comparing results obtained using prospectively and retrospectively collected data will help validate the use of the easily available retrospective data in assessing quality of care in past and future studies. METHODS: Prospective and retrospective datasets were obtained from a cluster randomized trial of a multifaceted intervention aimed at improving paediatric inpatient care conducted in eight rural Kenyan district hospitals by improving management of children admitted with pneumonia, malaria and diarrhea and/or dehydration. Four hospitals received a full intervention and four a partial intervention. Data were collected through 3 two weeks surveys conducted at baseline, after 6 and 18 months. Retrospective data was sampled from paediatric medical records of patients discharged in the preceding six months of the survey while prospective data was collected from patients discharged during the two week period of each survey. Risk Differences during post-intervention period of 16 quality of care indicators were analyzed separately for prospective and retrospective datasets and later plotted side by side for comparison. RESULTS: For the prospective data there was strong evidence of an intervention effect for 8 of the indicators and weaker evidence of an effect for one indicator, with magnitude of effect sizes varying from 23% to 60% difference. For the retrospective data, 10 process (these include the 8 indicators found to be statistically significant in prospective data analysis) indicators had statistically significant differences with magnitude of effects varying from 10% to 42%. The bar-graph comparing results from the prospective and retrospective datasets showed similarity in terms of magnitude of effects and statistical significance for all except two indicators. CONCLUSION: Multifaceted interventions can help improve adoption of clinical guidelines and hence improve the quality of care. The similar inference reached after analyses based on prospective assessment of case management is a useful finding as it supports the utility of work based on examination of retrospectively assembled case records allowing longer time periods to be studied while constraining costs. TRIAL REGISTRATION: Current Controlled Trials ISRCTN42996612. Trial registration date: 20/11/2008.

Hypothermia amongst neonatal admissions in Kenya: a retrospective cohort study assessing prevalence, trends, associated factors, and its relationship with all-cause neonatal mortality.

Background: Reports on hypothermia from high-burden countries like Kenya amongst sick newborns often include few centers or relatively small sample sizes. Objectives: This study endeavored to describe: (i) the burden of hypothermia on admission across 21 newborn units in Kenya, (ii) any trend in prevalence of hypothermia over time, (iii) factors associated with hypothermia at admission, and (iv) hypothermia's association with inpatient neonatal mortality. Methods: A retrospective cohort study was conducted from January 2020 to March 2023, focusing on small and sick newborns admitted in 21 NBUs. The primary and secondary outcome measures were the prevalence of hypothermia at admission and mortality during the index admission, respectively. An ordinal logistic regression model was used to estimate the relationship between selected factors and the outcomes cold stress (36.0°C-36.4°C) and hypothermia (<36.0°C). Factors associated with neonatal mortality, including hypothermia defined as body temperature below 36.0°C, were also explored using logistic regression. Results: A total of 58,804 newborns from newborn units in 21 study hospitals were included in the analysis. Out of these, 47,999 (82%) had their admission temperature recorded and 8,391 (17.5%) had hypothermia. Hypothermia prevalence decreased over the study period while admission temperature documentation increased. Significant associations were found between low birthweight and very low (0-3) APGAR scores with hypothermia at admission. Odds of hypothermia reduced as ambient temperature and month of participation in the Clinical Information Network (a collaborative learning health platform for healthcare improvement) increased. Hypothermia at admission was associated with 35% (OR 1.35, 95% CI 1.22, 1.50) increase in odds of neonatal inpatient death. Conclusions: A substantial proportion of newborns are admitted with hypothermia, indicating a breakdown in warm chain protocols after birth and intra-hospital transport that increases odds of mortality. Urgent implementation of rigorous warm chain protocols, particularly for low-birth-weight babies, is crucial to protect these vulnerable newborns from the detrimental effects of hypothermia.

External validation of a paediatric Smart triage model for use in resource limited facilities.

Models for digital triage of sick children at emergency departments of hospitals in resource poor settings have been developed. However, prior to their adoption, external validation should be performed to ensure their generalizability. We externally validated a previously published nine-predictor paediatric triage model (Smart Triage) developed in Uganda using data from two hospitals in Kenya. Both discrimination and calibration were assessed, and recalibration was performed by optimizing the intercept for classifying patients into emergency, priority, or non-urgent categories based on low-risk and high-risk thresholds. A total of 2539 patients were eligible at Hospital 1 and 2464 at Hospital 2, and 5003 for both hospitals combined; admission rates were 8.9%, 4.5%, and 6.8%, respectively. The model showed good discrimination, with area under the receiver-operator curve (AUC) of 0.826, 0.784 and 0.821, respectively. The pre-calibrated model at a low-risk threshold of 8% achieved a sensitivity of 93% (95% confidence interval, (CI):89%-96%), 81% (CI:74%-88%), and 89% (CI:85%-92%), respectively, and at a high-risk threshold of 40%, the model achieved a specificity of 86% (CI:84%-87%), 96% (CI:95%-97%), and 91% (CI:90%-92%), respectively. Recalibration improved the graphical fit, but new risk thresholds were required to optimize sensitivity and specificity.The Smart Triage model showed good discrimination on external validation but required recalibration to improve the graphical fit of the calibration plot. There was no change in the order of prioritization of patients following recalibration in the respective triage categories. Recalibration required new site-specific risk thresholds that may not be needed if prioritization based on rank is all that is required. The Smart Triage model shows promise for wider application for use in triage for sick children in different settings.

Comparison between the Smart Triage model and the Emergency Triage Assessment and Treatment guidelines in triaging children presenting to the emergency departments of two public hospitals in Kenya.

Several triage systems have been developed, but little is known about their performance in low-resource settings. Evaluating and comparing novel triage systems to existing triage scales provides essential information about their added value, reliability, safety, and effectiveness before adoption. This study included children aged < 15 years who presented to the emergency departments of two public hospitals in Kenya between February and December 2021. We compared the performance of Emergency Triage Assessment and Treatment (ETAT) guidelines and Smart Triage (ST) models (ST model with independent triggers, and recalibrated ST model with independent triggers) in categorizing children into emergency, priority, and non-urgent triage categories. Sankey diagrams were used to visualize the distribution of children into similar or different triage categories by ETAT and ST models. Sensitivity, specificity, negative and positive predictive values for mortality and admission were calculated. 5618 children were enrolled, and the majority (3113, 55.4%) were aged between one and five years of age. Overall admission and mortality rates were 7% and 0.9%, respectively. ETAT classified 513 (9.2%) children into the emergency category compared to 1163 (20.8%) and 1161 (20.7%) by the ST model with independent triggers and recalibrated model with independent triggers, respectively. ETAT categorized 3089 (55.1%) children as non-urgent compared to 2097 (37.4%) and 2617 (46.7%) for the respective ST models. ETAT classified 191/395 (48.4%) admitted patients as emergencies compared to more than half by all the ST models. ETAT and ST models classified 25/49 (51%) and 39/49 (79.6%) deceased children as emergencies. Sensitivity for admission and mortality was 48.4% and 51% for ETAT and 74.9% and 79.6% for the ST models, respectively. Smart Triage shows potential for identifying critically ill children in low-resource settings, particularly when combined with independent triggers and performs comparably to ETAT. Evaluation of Smart Triage in other contexts and comparison to other triage systems is required.

Evaluating the documentation of vital signs following implementation of a new comprehensive newborn monitoring chart in 19 hospitals in Kenya: A time series analysis.

Multi-professional teams care for sick newborns, but nurses are the primary caregivers, making nursing care documentation essential for delivering high-quality care, fostering teamwork, and improving patient outcomes. We report on an evaluation of vital signs documentation following implementation of the comprehensive newborn monitoring chart using interrupted time series analysis and a review of filled charts. We collected post-admission vital signs (Temperature (T), Pulse (P), Respiratory Rate (R) and Oxygen Saturation (S)) documentation frequencies of 43,719 newborns with a length of stay > 48 hours from 19 public hospitals in Kenya between September 2019 and October 2021. The primary outcome was an ordinal categorical variable (no monitoring, monitoring 1 to 3 times, 4 to 7 times and 8 or more times) based on the number of complete sets of TPRS. Descriptive analyses explored documentation of at least one T, P, R and S. The percentage of patients in the no-monitoring category decreased from 68.5% to 43.5% in the post-intervention period for TPRS monitoring. The intervention increased the odds of being in a higher TPRS monitoring category by 4.8 times (p<0.001) and increased the odds of higher monitoring frequency for each vital sign, with S recording the highest odds. Sicker babies were likely to have vital signs documented in a higher monitoring category and being in the NEST360 program increased the odds of frequent vital signs documentation. However, by the end of the intervention period, nearly half of the newborns did not have a single full set of TPRS documented and there was heterogenous hospital performance. A review of 84 charts showed variable documentation, with only one chart being completed as designed. Vital signs documentation fell below standards despite increased documentation odds. More sustained interventions are required to realise the benefits of the chart and hospital-specific performance data may help customise interventions.

Cross-sectional study to predict subnational levels of health workers' knowledge about severe malaria treatment in Kenya.

OBJECTIVES: This study applied a Bayesian hierarchical ecological spatial model beyond predictor analysis to test for the best fitting spatial effects model to predict subnational levels of health workers' knowledge of severe malaria treatment policy, artesunate dosing, and preparation. SETTING: County referral government and major faith-based hospitals across 47 counties in Kenya in 2019. DESIGN AND PARTICIPANTS: A secondary analysis of cross-sectional survey data from 345 health workers across 89 hospitals with inpatient departments who were randomly selected and interviewed. OUTCOME MEASURES: Three ordinal outcome variables for severe malaria treatment policy, artesunate dose and preparation were considered, while 12 individual and contextual predictors were included in the spatial models. RESULTS: A third of the health workers had high knowledge levels on artesunate treatment policy; almost three-quarters had high knowledge levels on artesunate dosing and preparation. The likelihood of having high knowledge on severe malaria treatment policy was lower among nurses relative to clinicians (adjusted OR (aOR)=0.48, 95% CI 0.25 to 0.87), health workers older than 30 years were 61% less likely to have high knowledge about dosing compared with younger health workers (aOR=0.39, 95% CI 0.22 to 0.67), while health workers exposed to artesunate posters had 2.4-fold higher odds of higher knowledge about dosing compared with non-exposed health workers (aOR=2.38, 95% CI 1.22 to 4.74). The best model fitted with spatially structured random effects and spatial variations of the knowledge level across the 47 counties exhibited neighbourhood influence. CONCLUSIONS: Knowledge of severe malaria treatment policies is not adequately and optimally available among health workers across Kenya. The factors associated with the health workers' level of knowledge were cadre, age and exposure to artesunate posters. The spatial maps provided subnational estimates of knowledge levels for focused interventions.

Handling missing data in a composite outcome with partially observed components: simulation study based on clustered paediatric routine data.

Composite scores are useful in providing insights and trends about complex and multidimensional quality of care processes. However, missing data in subcomponents may hinder the overall reliability of a composite measure. In this study, strategies for handling missing data in Paediatric Admission Quality of Care (PAQC) score, an ordinal composite outcome, were explored through a simulation study. Specifically, the implications of the conventional method employed in addressing missing PAQC score subcomponents, consisting of scoring missing PAQC score components with a zero, and a multiple imputation (MI)-based strategy, were assessed. The latent normal joint modelling MI approach was used for the latter. Across simulation scenarios, MI of missing PAQC score elements at item level produced minimally biased estimates compared to the conventional method. Moreover, regression coefficients were more prone to bias compared to standards errors. Magnitude of bias was dependent on the proportion of missingness and the missing data generating mechanism. Therefore, incomplete composite outcome subcomponents should be handled carefully to alleviate potential for biased estimates and misleading inferences. Further research on other strategies of imputing at the component and composite outcome level and imputing compatibly with the substantive model in this setting, is needed.

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.

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.

Antibiotic use in Kenyan public hospitals: Prevalence, appropriateness and link to guideline availability.

OBJECTIVE: To examine prescription patterns and explore to what extent guidelines are available and how they might influence treatment appropriateness among hospitalised patients in Kenyan hospitals. METHODS: Data on antimicrobial usage were collected from hospitalised patients across 14 Kenyan public hospitals. For each prescription, appropriateness of treatment was defined using available local and international treatment guidelines and through consensus with local medical specialists. Association between appropriate treatment, guideline availability and other possible explanatory factors was explored using univariate and multiple regression analysis. RESULTS: There were 1675 (46.7%) of the 3590 hospitalised patients on antimicrobials with 3145(94%) of the 3363 antimicrobial prescriptions being antibiotics. Two patients (0.1%), had treatment based on available antibiotic susceptibility tests. Appropriate treatment was assessed in 1502 patients who had a single diagnosis. Of these, 805 (53.6%) received appropriate treatment. Physical availability of treatment guidelines increased the odds of receiving appropriate treatment Odds Ratio 6.44[95% CI 4.81-8.64]. CONCLUSION: Appropriate antibiotic prescription remains a challenge in Kenyan public hospitals. This may be improved by the availability of context-specific, up-to-date, and readily accessible treatment guidelines across all the departments, and by providing better diagnostic support.

Evaluating the foundations that help avert antimicrobial resistance: Performance of essential water sanitation and hygiene functions in hospitals and requirements for action in Kenya.

BACKGROUND: Water Sanitation and Hygiene (WASH) in healthcare facilities is critical in the provision of safe and quality care. Poor WASH increases hospital-associated infections and contributes to the rise of antimicrobial resistance (AMR). It is therefore essential for governments and hospital managers to know the state of WASH in these facilities to set priorities and allocate resources. METHODS: Using a recently developed survey tool and scoring approach, we assessed WASH across four domains in 14 public hospitals in Kenya (65 indicators) with specific assessments of individual wards (34 indicators). Aggregate scores were generated for whole facilities and individual wards and used to illustrate performance variation and link findings to specific levels of health system accountability. To help interpret and contextualise these scores, we used data from key informant interviews with hospital managers and health workers. RESULTS: Aggregate hospital performance ranged between 47 and 71% with five of the 14 hospitals scoring below 60%. A total of 116 wards were assessed within these facilities. Linked to specific domains, ward scores varied within and across hospitals and ranged between 20% and 80%. At ward level, some critical indicators, which affect AMR like proper waste segregation and hand hygiene compliance activities had pooled aggregate scores of 45 and 35% respectively. From 31 interviews conducted, the main themes that explained this heterogenous performance across facilities and wards included differences in the built environment, resource availability, leadership and the degree to which local managers used innovative approaches to cope with shortages. CONCLUSION: Significant differences and challenges exist in the state of WASH within and across hospitals. Whereas the senior hospital management can make some improvements, input and support from the national and regional governments are essential to improve WASH as a basic foundation for averting nosocomial infections and the spread of AMR as part of safe, quality hospital care in Kenya.

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.

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.