Childhood mortality during and after acute illness in sub-Saharan Africa and South Asia - The CHAIN cohort study

Objectives Mortality during acute illness among children in low- and middle-income settings remain unacceptably high and there is increasing recognition of the importance of post-discharge mortality. A comprehensive understanding of pathways underlying mortality among acutely ill children is needed to develop interventions and improve guidelines. We aimed to determine the incidence, timing and contributions of proximal and underlying exposures for mortality among acutely ill young children from admission to hospital until 6 months after discharge in sub-Saharan Africa and South Asia in the context of guideline-based care. Design A prospective stratified cohort study recruiting acutely ill children at admission to hospital with follow up until 180 days after discharge from hospital (November 2016-July 2019). Setting Nine urban and rural hospitals in sub-Saharan Africa and South Asia across a range of facility levels, and local prevalences of HIV and malaria. Participants Inclusion criteria were age 2-23 months, admission to hospital with acute, non-traumatic medical illness and stratified into three groups by anthropometry. Children were excluded if currently receiving pulmonary resuscitation, had a known condition requiring surgery within 6 months or known terminal illness with death expected within 6 months. Main outcome measures Acute mortality occurring within 30-days from admission; post-discharge mortality within 180-days from discharge; characteristics with direct and indirect associations with mortality within a multi-level a priori framework including demographic, clinical, anthropometric characteristics at admission and discharge from hospital, and pre-existing child-, caregiver- and household-level characteristics. Results Of 3101 participants (median age 11 months), 1218 were severely wasted/kwashiorkor, 763 moderately wasted and 1120 were not wasted. Of 350 deaths, 182 (52%) occurred during index admission, 234 (67%) within 30-days of admission and 168 (48%) within 180-days post-discharge. Ninety (54%) post-discharge deaths occurred at home. The ratio of inpatient to post-discharge mortality was consistent across anthropometric strata and sites. Large high and low risk groups could be disaggregated for both early and post-discharge mortality. Structural equation models identified direct pathways to mortality and multiple socioeconomic, clinical and nutritional domains acting indirectly through anthropometric status. Conclusions Among diverse sites in Africa and South Asia, almost half of mortality occurs post-discharge. Despite being highly predictable, these deaths are not addressed in current guidelines. A fundamental shift to a risk-based approach to inpatient and post-discharge management is needed to further reduce childhood mortality and clinical trials of these approaches with outcomes of mortality, readmission and cost are warranted. Trial Registration ClinicalTrials.gov: NCT03208725


Introduction
Most seriously ill children in low-and middle-income countries (LMICs) access the healthcare system at some point 1 and many are admitted to hospitals. Reported paediatric inpatient case fatality ratios vary widely between hospitals from 2% to >20%. [2][3][4][5] Postdischarge mortality is a widely recognized phenomenon with reported case fatality rates from 1-2% to >20% from paediatric studies in LMICs with widely varying inclusion criteria, follow up duration and loss to follow-up. 4 6-10 Variation in case fatality observed between facilities likely reflects differences in care seeking, 11 local prevalence of malnutrition, HIV and other comorbidities, and facility-level differences in staffing and resources. 3 12 However, within children admitted to each hospital there is also a broad range of illness types and severity, and underlying risks. While several risk scores based on clinical signs and anthropometry have been devised, 13-21 few include social or underlying risks, and none are in widespread use or have impacted case fatality at scale.
Syndrome-based protocols defined in World Health Organization (WHO) 22 guidelines have been widely adopted, although often not fully implemented. They typically include recommendations by severity within some common clinical syndromes, but do not fully address children meeting criteria for multiple syndromes or having underlying comorbidities, 23 and are often based on limited or low-quality evidence. 6 24 25 They provide little advice on post-discharge care or for children with challenging social circumstances. We aimed to describe the epidemiology of deaths associated with hospitalization and identify pathways associated with acute mortality within 30-days of an admission to hospital and mortality during 180-days post-discharge in the context of guideline-based care in a range of facilities in sub-Saharan Africa and South Asia.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021

Methods
The Childhood Acute Illness & Nutrition (CHAIN) cohort was a prospective stratified cohort. 12 Acutely ill children aged 2−23 months were systematically enrolled at admission to hospital between November 20, 2016 and January 31, 2019 (eMethods 1 & 2). Nine participating hospitals were selected to reflect a range of rural and urban environments, and malaria and Pakistan, Civil Hospital-Karachi; and Uganda, Mulago Hospital-Kampala, as previously described. 26 Children were screened for eligibility at admission for non-surgical, nontraumatic conditions. Inclusion and exclusion criteria are listed in eMethods 2. Primary outcomes were 30-day mortality and 180-day post-discharge mortality, which are in-line with reporting best practices 27 and, in contrast to examining inpatient mortality, are unaffected by differential discharge decisions.
A pre-study audit was conducted to identify areas of divergence from guidelines 5 and ongoing care audits were undertaken using the treatment database. Internal quality control and external study monitoring (Westat ® , MD, USA) were conducted. Ethical approval was obtained from the Oxford University Tropical Research Ethics Committee and ethics committees at all participating institutions.

Enrolment
Anthropometry is consistently a strong predictor of mortality. 9 13 15 To weight towards higherrisk children, enrolment was therefore stratified by mid-upper arm circumference (MUAC), a marker of mortality risk that is more easily measured in sick children and is less affected by dehydration than weight/length-based anthropometry. Three strata, not wasted (NW), moderately wasted (MW) and severely wasted or kwashiorkor (edematous malnutrition) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 (SWK) defined in eMethods 2 were enrolled in a 2:1:2 ratio, deliberately over-representing MW and SWK. Standardized demographic, medical history, underlying medical conditions and clinical examination data were collected (eMethods 3-6).

Follow-up
At discharge, clinical examination and anthropometry were performed. SWK children were discharged when completing feeds and were all linked to an outpatient therapeutic feeding program. MW children did not routinely receive supplementary feeds post-discharge reflecting current policy at the study sites, whilst SWK children received supplementary feeds as part of outpatient nutritional rehabilitation. Study follow-up visits were scheduled at 45, 90 and 180 days after discharge when vital status, anthropometry and health status were assessed. Non-attendees were traced through home visits, or vital status determined by telephone.

Patient and public involvement
Patients were not directly involved in the design of the study. However, the CHAIN cohort was conceived and designed in response to concerns raised by parents and carers involved in our prior clinical trials regarding their concerns that the prescribed medical and nutritional treatments were not addressing the 'real causes' of their child's problems, and difficulties providing adequate care and accessing health advice after discharge from hospital. In addition, health care workers expressed concerns that acutely ill children were dying despite doing everything possible to follow WHO and national guidelines. Further, they felt inadequately prepared to plan and undertake treatment for children with multiple morbidities, that many children's problems arose from profound social disadvantage.

Study size
Pearson's χ 2 test suggested 2600 children were required post-discharge for 80% power to detect a difference in post-discharge case fatality between NW and MW children of 2.4% vs.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 5.1% based on prior Kenyan data, 9 assuming an α =0.05, allowing 10% lost to follow-up (LTFU) and a two-sided hypothesis.

Statistical analysis
All analyses were conducted according to an a priori statistical analysis plan accounting for the stratified design, specifying handling of missing data and imputation methods following STROBE guidelines (Statistical Analysis Plan in the Supplementary Appendix). We calculated proportions and absolute event rates for deaths during the index admission, from admission to 30 days later and from discharge to180 days later.
For compatibility with prior studies and to examine the potential for risk stratification, exposures were investigated using standard survival models using a Weibull distribution with a random intercept for site and weighted for the stratified selection and loss to follow up (LTFU) (eTable 2, 3 & eFigure 2). To generate a more comprehensive understanding of the relationships between a priori domains of exposures (Tables 1, 3 & eTable 4) and their direct and indirect effects on mortality, we developed an a priori conceptual framework (eFigure 1) based on the UNICEF conceptual framework 28 linking immediate and underlying exposures through undernutrition and illness to mortality. Anthropometry and signs of illness severity domain scores at discharge were calculated independently of those at admission.
We built survival structural equation models (SEM) according to our framework with a Weibull distribution, a random intercept for site and weighted as per the survival regression analyses. Outcomes were adjusted for age and sex. In reporting, we distinguish direct pathways (from a domain to an outcome) and indirect pathways (from a domain to an outcome via another domain). Findings from SEM models were considered the primary outputs on effects of exposures.
Multivariable model performance was assessed by bootstrapped area under receiver operating characteristic curve (AUC) with 1,000 replications. Statistical analyses were . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 conducted using STATA v15.1 (StataCorp, College Station, TX, USA) and R v3.6.2 (R Foundation for Statistical Computing). Detailed methods, including approaches to missing data, are given in Statistical Analysis Plan & eMethods 6-9 in the Supplementary Appendix.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021

All deaths
Overall, 350 (11%) children died between admission and 180 days following hospital discharge (Figure 2A). Causes of death and deaths by clinical syndromes and sites are given in eTables 20 & 21.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 Figure 2B).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021

Survival regression models
The distributions of children within exposure domain scores are shown in eTable 24.

Sensitivity analyses
Sensitivity analyses across a range of plausible sampling weights in LMIC settings and using different anthropometric indices, provided results similar to the primary models (eTables 25−33 & eFigure 6). The relationship between MUAC on a continuous scale and mortality . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 stratified by admission illness severity tertiles is shown in eFigure 7. Unlike later postdischarge deaths, deaths during the first month post-discharge were associated with age and leaving against medical advice (eTable 34).

Risk stratification
At admission, the lowest and highest risk quintiles predicted by the regression models had a 1.3% and 24% risk of 30-day mortality respectively (eTable 35). At discharge, the lowest two risk quintiles (40% of discharged children) had a 1.0% risk, and the highest risk quintile (20%

180-day post-discharge mortality
At discharge, signs of illness severity (high: aHR 3.68 (95%CI 1.56−8.67)), improved anthropometry (aHR 0.70 (95%CI 0.56−0.89)) and discharge against medical advice (aHR 2.05 (95%CI 1.25−3.35)) were directly associated with mortality. Duration of admission was . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ; https://doi.org/10.1101/2021.11.24.21266806 doi: medRxiv preprint not directly associated with post-discharge mortality and acted as a 'sink' since many domains impacted duration of admission.
Several important indirect effects were observed. Household-level exposures influenced discharge against medical advice, caregiver characteristics and underlying medical conditions. Reduced access to healthcare, underlying medical conditions and child-level nutritional exposures affected admission nutritional status, with child-level nutritional exposure also being associated with underlying medical conditions. Underlying medical conditions also affected the change in nutritional status from admission to discharge.

30-day and 180-day post-discharge mortality models
In the confirmatory factor analysis, employment, mental health and education were the main measures of the caregiver characteristics domain (eTable 4). In both 30-day and postdischarge mortality models, lower anthropometric strata were associated with less severe signs of illness at admission. Model performance is shown in Table 3 and eTables 38 & 39.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ; https://doi.org/10.1101/2021.11.24.21266806 doi: medRxiv preprint

Discussion
Our findings suggest that acutely ill children would benefit from greater risk stratification to target interventions beyond short-term therapeutic and supplemental feeding, including prior to hospitalization. General risk stratification among paediatric admissions may permit low risk children to be identified and discharged earlier, freeing-up staff and resources for children at higher risk and saving costs to families.
Across different geographical settings almost half of all deaths occurred after discharge from hospital and most post-discharge deaths occurred at home. The ratio of in-patient to postdischarge deaths was preserved across geographic settings and anthropometric strata and matched the highest estimates previously reported. 4 Although mortality risk falls during the initial month following discharge, significant risk persists for the entire six-month postdischarge period.
Post-discharge mortality was as predictable as inpatient mortality using information at admission, discharge and on the child's social circumstances. Clinical features at index discharge have not been previously assessed in prognostic studies 13 15 and, not surprisingly, these were more strongly associated with post-discharge death than admission features.
Importantly, multiple pathways converged on length of hospital stay which was not associated with post-discharge mortality, suggesting that extending inpatient treatment is unlikely to reduce mortality for children judged ready for discharge. As expected, leaving against medical advice increased mortality, suggesting a need to identify these children early in order to address caregiver concerns.
In the context of severe acute illness, our analysis indicates that low MUAC represents a high-risk state (eFigure 7) associated with mechanisms beyond short-term reduced nutrient intake (Figure 3). Importantly, children with low MUAC admitted with low illness severity had . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ; https://doi.org/10.1101/2021.11.24.21266806 doi: medRxiv preprint far lower mortality risk, as is seen in outpatient programs for uncomplicated severe malnutrition.
Caregiver characteristics, dominated by maternal employment and mental health, directly influenced post-discharge mortality and appeared more important than household-level exposures or access to healthcare. These findings support previous work, 29 suggesting that interventions specifically targeting maternal wellbeing, agency and financial independence are needed. These findings may also partially explain the lack of efficacy of individual interventions against post-discharge mortality in recent clinical trials. 4 30-33 Strengths of the CHAIN cohort include its size, geographic diversity, standardized care and comprehensive data collection, with generalizability from being across a broad range of geographic and epidemiologic settings. Limitations included difficulty in differentiating situational stressors and experiences from longer term psychosocial strain, lack of biochemical determination of nutrient status, challenges achieving full guideline adherence, and difficulties in obtaining some datapoints for children who died early. Causes of death are based on all available information but may be unreliable. We chose to retain sepsis as a cause of death as a primary focus of infection was not always evident and/or children met criteria for multiple syndromes.
Clinical trials and research to support a fundamental shift towards a generic, structured, riskbased approach to inpatient care and assessment at discharge to inform continued care with outcomes of mortality, readmission, costs and implementation are needed. Identification at admission and structured review may allow earlier discharge for very low-risk children, freeing-up resources and staff for higher-risk children. Whilst this is currently done in some larger hospitals through a Paediatric Observation Ward (POW), neither admission to the POW or evaluation after 24 hours are undertaken using formal criteria, and these decisions . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ; https://doi.org/10. 1101/2021 are often made by junior or intern staff with limited paediatrics training or experience. The approach can be extended to county-and district-level hospitals.
For higher-risk children, given that just over half of post-discharge deaths occur at home, 4 34 'down-referral' to the responsibility of a named community health worker or nurse, like the role of health visitors in high-income settings should be evaluated. Training for mothers and family members in recognizing danger signs and continued structured contact with a health professional by phone and short messaging services to deliver health advice and facilitate recognition of children in need of medical evaluation. Facilitated access to the emergency room for recently admitted children, avoiding queueing and waiving hospital costs, including caregiver bed charges, may help to overcome hesitancy in re-presenting to hospital. 25 Guideline revision to provide specific care for readmitted children is also needed. Some of these interventions may be better suited to before/after or step-wedge designs than individually randomized trials.  l  a  c  e  o  f  d  e  a  t  h  ,  c  a  r  e  -s  e  e  k  i  n  g  a  n  d  c  a  r  e  p  a  t  h  w  a  y  p  r  o  g  r  e  s  s  i  o  n  i  n  t  h  e  f  i  n  a  l  i  l  l  n  e  s  s  e  s  o  f  c  h  i  l  d  r  e  n  u  n  d  e  r  f  i  v  e  y  e  a  r  s  o  f  a  g  e  i  n  s  u  b  -S  a  h  a  r  a  n  A  f  r  i  c  a  :  a  s  y  s  t  e  m  a  t  i  c  r  e I  l  l  n  e  s  s  a  n  d  N  u  t  r  i  t  i  o  n  (  C  H  A  I  N  )  N  e  t  w  o  r  k  :  a  p  r  o  t  o  c  o  l  f  o  r  a  m  u  l  t  i  -s  i  t  e  p  r  o  s  p  e  c  t  i  v  e  c  o  h  o  r  t  s  t  u  d  y  t  o  i  d  e  n  t  i  f  y  m  o  d  i  f  i  a  b  l  e  r  i  s  k  f  a  c  t  o  r  s  f  o  r  m  o  r  t  a  l  i  t  y  a  m  o  n  g  a  c  u  t  e  l  y  i  l  l  c  h  i  l  d  r  e  n  i  n  A  f  r  i  c  a  a  n   . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review) preprint
The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ; https://doi.org/10. 1101/2021   . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ; https://doi.org/10.1101/2021.11.24.21266806 doi: medRxiv preprint    Removing all pre-existing characteristics from the models resulted in bootstrapped AUCs of a) 0.77 (95%CI: 0.74 to 0.79) and b) 0.78 (95%CI 0.75 to 0.82).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2021. ; https://doi.org/10.1101/2021.11.24.21266806 doi: medRxiv preprint