This study is an observational multi-country cohort study that aims to build algorithms that can identify children between 5 and 16 years of age admitted for proven or suspected sepsis who are at risk of mortality after they are discharged in East Africa.

In low- and middle-income countries, about 5% of children discharged after hospitalization for sepsis will die in the weeks after returning home. Doctors and parents are often unaware of this period of vulnerability and are poorly equipped to identify or handle this critical situation. This project builds on past work that developed and evaluated models and the Smart Discharges program to predict, during hospitalization, an individual child's risk of recurrent illness and mortality, as well as to provide additional post-discharge support to at-risk children.

Participants will be enrolled from facilities once they are admitted, collecting clinical and social variables. They will then be followed until 6 months post-discharge to understand what happens to them after they return home. This data will be evaluated to identify which variables collected at facilities can be predictive of mortality and recurrent illness after discharge.

PURPOSE The purpose of our research is to develop and expand the Smart Discharges approach to improve outcomes among all pediatric populations, and ultimately to build a scalable solution for improving post-discharge outcomes.

HYPOTHESIS Post-discharge mortality can be predicted, using admission variables, for children under 16 years of age who are admitted with suspected sepsis. Furthermore, a high proportion of these children will experience long-term reductions in health-related quality of life over the post-discharge period.

JUSTIFICATION With an improved understanding of risk, and an ability to determine risk at the bedside among older children, the existing Smart Discharge program can be expanded to be more inclusive, applicable to all pediatric patients with suspected sepsis, further simplifying implementation initiatives. This also further increases optimization of resource allocation within the health system. Such programs in precision public health can not only save lives and resources, but are much more likely to be scalable in economically strained environments

OBJECTIVES

The study aims to:

1. Develop and internally validate a clinical risk prediction model for identifying children 5-<16 years of age admitted for proven or suspected sepsis who are at high-risk of post-discharge death
2. Quantitatively evaluate health-related quality of life (HRQL) and functional status up to 12 months post-discharge in children 5-<16 years of age admitted for proven or suspected sepsis.
3. Qualitatively assess post-discharge experiences of children 5-<16 years of age admitted for proven or suspected sepsis, and the experiences of their families and health workers providing care.

RESEARCH DESIGN This is a mixed-methods study at a regionally representative sample of public hospitals providing pediatric care in East Africa. There will be three phases to this study, each corresponding to a specific objective.

Phase I - Model Derivation and internal validation: The investigators will conduct an observational cohort study in 4000 children 5-16 years of age informed through direct observation of the child prior to facility discharge and through telephone follow-up visits at 2-, 4-, and 6-months post-discharge. The objective of this study is to i) build a model to predict post-discharge mortality within 6 months of discharge among children 5-16 years of age.

Phase II - HRQL & Functional Status Evaluation: The investigators will conduct an observational cohort study in a random subset of 500 children aged 5 - <16 years enrolled in phase I. Using survey-based methodology, the investigators will conduct face-to-face caregiver interviews at admission, and telephone interviews will be conducted 2-,4-, 6-, and 12-months post-discharge. The objective of this study is to determine the HRQL and Functional Status Score (FSS) trajectory of sepsis during the first 6 months following discharge.

Phase IIIa: The investigators will conduct a qualitative study using focus group discussions with caregivers and community health workers. The objective of this study is to better understand the experiences of patients and caregivers during the post-discharge period, as well as the barriers and facilitators to effective care seeking during the post-discharge period.

Phase IIIb: The investigators will conduct a qualitative study using key informant interviews with health workers. The objective of this study is to better understand the perspective of health workers' discharge decision-making as well as caregiver interactions with the health system during the discharge and post-discharge period.

STATISTICAL ANALYSIS PLAN Phase I Our initial analysis will examine age- and sex-based interactions to identify an optimal set of key variables for modelling. This approach allows us to account for multiple factors that may contribute to age or sex based differences in vulnerability. With our sample size the study is powered to include age/sex as individual and interactive predictors. The study is not powered to build models within segregated age/sex groups, as this is not our intended output. Derivation of prediction models will be based on optimization of the Area Under the Receiver Operating Characteristic (AUROC) and specificity across a variety of modeling and variable selection approaches (e.g. logistic regression, elastic net, support vector machines). Model performance will be based on appropriate resampling techniques for internal validation (e.g., cross-validation, bootstrapping). The study will focus on developing parsimonious predictive models (e.g., 5-10 predictor variables) with high sensitivity (>80%) to reduce false negatives and maximize model use in resource-limited settings. AUROC, sensitivity, and specificity will be reported for each model, along with positive and negative predictive values. The investigators will compare country-specific metrics to ensure consistency across settings, and consider country-specific recalibration if model performance is lower than expected (>5% drop). The investigators will also summarize all risk factors for children who do and do not experience poor outcomes and estimate univariate associations. Outcomes and risk factors will be reported by sex and age category.

Phase II The investigators will use one-way Analysis if Variance (ANOVA) to test for significant difference in mean scores across groups of interest (e.g., sex, nutritional status, study site), and post-hoc Tukey HSD (Honest Significance Difference) test to perform pair-comparison in mean scores between different groups. They will also conduct descriptive statistical analysis with survival curves stratified by groups of interest for outcome event, and a mixed effects model, which can be used to model the longitudinal change in the outcome scores. The investigators will use univariable and multivariable logistic regression modeling to perform exploratory analysis examining associations between markers of critical illness (e.g., vital signs and clinical signs and symptoms collected in Aim 1) and our primary outcome for this Aim (persistent deterioration of HRQL). Odds ratios with 95% confidence intervals will be reported for all risk factors. Descriptive themes include barriers to care and post-discharge health-seeking behavior, while interpretive themes focus on caregiver and health worker perceptions of child death and disability, quality of life post-discharge, and the role of the health system in maintaining health. The investigators will employ member checking with a sample of participants to ensure validity of our results.

Phase III A team of researchers and clinicians from each of our study sites will lead data analysis. They will analyze focus group discussions using a framework method, which allows themes to be developed inductively from participants and deductively from existing literature. Through an iterative process, transcripts will be coded and analyzed for descriptive and interpretive themes using NVivo 12. Codes will be grouped and organized to summarize common themes from participants' lived experiences. Descriptive themes include barriers to care and post-discharge health-seeking behavior, while interpretive themes focus on caregiver and health worker perceptions of child death and disability, quality of life post-discharge, and the role of the health system in maintaining health. The team of researchers will employ member checking with a sample of participants to ensure validity of our results.

Inclusion Criteria:

• Age between 0 and 16 years admitted to pediatric ward of hospital for proven or suspected infection
• Provide informed consent
• Ability to provide contact information (phone number or address) for phone or in-person follow-up

Exclusion Criteria:

• Child lives outside of the catchment area of a study site
• Admitted for elective procedures, trauma, or short-term observation
• Language barriers

The study population is children admitted to hospital facilities with suspected or proven sepsis between 5-16 years of age in the catchment area of the 8 study sites: Mbarara Regional Referral Hospital (Uganda), Jinja Regional Referral Hospital (Uganda), Uganda Martyrs Ibanda Hospital (Uganda), Holy Innocents Children's Hospital (Uganda), Sekou-Toure Regional Referral Hospital (Tanzania), Bugando Medical Centre (Tanzania), Masaka Hospital (Rwanda), University Teaching Hospital of Kigali (Rwanda), Ruhengeri Referral Hospital (Rwanda).

After the study period, a de-identified copy of the data will be prepared for deposition in a repository with open access with proper governance mechanisms. The investigators will make every effort to prevent re-identification of subjects by coding data that has the potential of being identifiable. For example, they will convert all dates into meaningful decimal numbers (date of birth into days since birth and date of recruitment will be reduced to month of recruitment) and all locations will be coded into data that is useful but not specific (such as address converted to distance and direction from facility). The investigators will ensure that data elements with small numbers of subjects (less than 10) will be coded or lumped to avoid identification. The de-identified study data will be made available using a data hosting service (e.g., Dataverse, Vivli, etc.)

Data will be deposited to an open access repository with moderated access within 2 years of study completion

Moderated access on a case-by-case basis.