F1000Research F1000Research 2046-1402 F1000 Research Limited London, UK 10.12688/f1000research.133981.2 Research Article Articles Association between no referral versus health centres referrals preceding hospital admission and in-hospital child mortality in rural Malawi

[version 2; peer review: 1 approved with reservations, 4 not approved]

 Selstø Annlaug Data Curation Formal Analysis Investigation Methodology Writing – Original Draft Preparation Writing – Review & Editing https://orcid.org/0009-0002-3704-8502 a 1 2 King Carina Conceptualization Formal Analysis Funding Acquisition Investigation Methodology Writing – Original Draft Preparation Writing – Review & Editing https://orcid.org/0000-0002-6885-6716 1 3 Hildenwall Helena Conceptualization Formal Analysis Funding Acquisition Investigation Methodology Writing – Original Draft Preparation Writing – Review & Editing 1 4 5 Zadutsa Beatiwel Conceptualization Data Curation Writing – Review & Editing 6 Banda Lumbani Data Curation Writing – Review & Editing 6 Phiri Everlisto Data Curation Writing – Review & Editing 6 Makwenda Charles Conceptualization Data Curation Writing – Review & Editing 6 Department of Global Public Health, Karolinska Institutet, Stockholm, 171 77 Solna, Sweden Division of Mental and Physical Health, Norwegian Institute of Public Health, Oslo, Postboks 222 Skøyen, 0213, Norway Institute for Global Health, University College London, London, WC1E 6BT, UK Department of Clinical Science, Intervention and Technology, Karolinska Institutet, Stockholm, 141 52 Huddinge, Sweden Astrid Lindgren Children's Hospital, Karolinska University Hospital, Stockholm, Huddinge, 141 86, Sweden Parent and Child Health Initiative, Lilongwe, Malawi annlaug.selsto@fhi.no

No competing interests were disclosed.

 7 1 2025 2023 12 1053 2 1 2025 Copyright: © 2025 Selstø A et al. 2025 This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Background

The under-five mortality rate counted 71 deaths per 1000 live births for children under five in Sub Saharan Africa in 2022, far above the SDG under five-mortality goal of 25 deaths per live 1000 birth. Referral issues are major challenges to child health and a hypothesis is that care at lower-level facilities delays quality emergency care. This study investigates the extent of no-referral versus health centres-referrals prior to hospital admission for sick children, and with inpatient mortality associations in Malawi.

 Methods

We conducted a cohort study of children aged 0-12 years admitted to hospitals in Mchinji district. Data-collection was done from September 2019 to April 2020 at one district-hospital and three community-hospitals. Information was collected from caregivers, patient files and ward admission registers. The primary analysis assesses the association between referral from a health centre and in-hospital outcomes using logistic regression. Inpatient mortality was the primary outcome.

 Result

The 4926 included children demonstrated that 86.6% had gone straight to hospital without referral. The majority (67.9%) were admitted due to malaria, sepsis (21.3%) and pneumonia (13.3%). Referred children had a case fatality rate of 5.3%, while the non-referred had a case fatality rate of 2.5%. Referred children had higher odds of dying (AOR: 2.0, CI 95%: 1.3-3.0), compared to not referred children. Children with anaemia (AOR: 4.1, CI 95%: 2.7-6.3) and malnutrition (AOR: 6.3, CI 95%: 2.7-14.6) had significantly higher odds of dying, than those without these conditions.

 Conclusion

Most children admitted to hospital were taken without a referral, and these children had better survival than those who were referred. While being referred could be a proxy for being more severely sick the mortality difference emphasises a need to better understand care-seeking pathways, including referral challenges, to direct interventions to improve timely provision of care for sick children.

 In-hospital child mortality; Care-seeking Referral; Paediatric; Malawi; low-income country The Swedish Researc Concil (2017-05579) The Laerdal Foundation (40348) Pediatric Health Initiative ASwasfundedwithatravelgrant Einhorn Family Foundation This work was supported by The Swedish Research Council [2017-05579]; The Laerdal Foundation [40348]; Einhorn Family Foundation and Pediatric Health Initiative (travel grant awarded to AS) The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Amendments from Version 1

Adjustments have been made according to reviewers. Among these the title has been changed to better address what this article is presenting. Strengths and limitations have been elaborated. We also found a comma mistake on our result om AOR. This has been corrected in the new version.

 Introduction

Many countries, mostly in Sub-Saharan Africa (SSA), continue to suffer from a high child mortality rate. 1 In 2022 the under-5 mortality was 71 deaths per 1,000 live births in SSA, and 40 per 1,000 for Malawi. 2 This is far from the Sustainable Development Goal (SDG) 3,2 to decrease the under-five mortality to less than 25 deaths per 1,000 live births by 2030. 3 The Lancet Global Health Commission on Quality Healthcare argued in 2018 that poor quality of healthcare is a now a bigger barrier than access to care in reducing child mortality and accounts for as much as 60% of avoidable deaths in low and middle-income countries (LMIC). 4 The Integrated Management of Childhood Illness (IMCI) and integrated Community Case Management (iCCM) programs, were developed to improve the quality of care for sick children in low-income settings. Children with signs of severe illness in the community or in the health centre should be referred to higher level facilities for further management. 5 IMCI does not include any components of emergency case management, and primary healthcare facilities in low-income countries commonly lack the resources to treat severely ill children. 6 Furthermore, challenges in the identification of severely ill children have been reported under IMCI, potentially leading to missed referrals. 7

Delays in care seeking for young children in LMIC have been reported due to guardians’ inability to recognise illness, household challenges in mobilising resources and/or deciding to seek care, especially in poor and less educated households. 8 – 10 This, along with referral challenges within the health system including waiting times, poor triage, and lack of referral transport, could lead to increased risks for children who are brought to primary care with an acute illness, as adequate care at a higher-level facility may be further delayed.

This study was performed to assess the care-seeking patterns, and more specifically on referrals, prior to being admitted to hospital, to establish the association between pre-admission referrals and inpatient mortality for children in Mchinji District, Malawi. We also investigated which characteristics who was associated with higher mortality.

 Methods Study setting

Malawi was one of ten countries in SSA that achieved the Millennium Development Goal 4 to reduce child mortality. 11 Despite the improvements in child mortality in Malawi, the mortality rate still falls short of the SDGs. The Malawian healthcare system, who heavily depends on external financing, 11 , 12 is organized in public (free of charge), private for profit and private not for profit sector. 12 The Ministry of Health supports services in the private not for profit sector in essential care, such as maternal and child health. This sector is dominated by non-governmental organisations and religious institutions, such as The Christian Health Association of Malawi (CHAM). 13 Malawi adopted and implemented IMCI in 1999 and iCCM was introduced in Malawi in 2007. The World Health Organisations (WHO) Emergency Triage Assessment and Treatment (ETAT) 14 has been implemented in Malawian tertiary and secondary hospitals.

Mchinji is a district in the Central Region of Malawi with a mainly rural population and with an under-five population of approximately 90,000 in 2018. In 2018 574.000 of the population in Mchinji lived in rural areas while only 28.000 lived in urban. 15 The country economy heavily relies on agriculture, with 80% of the population working in this field. 16 Data was collected at the paediatric department of the main district hospital, Mchinji District Hospital (MDH) and the three CHAM community hospitals within the district. The primary care health centres do not offer any inpatient care, and children in need of admission are referred to hospital for further care.

The recruitment and data collection were planned for one year from the 9 th of September 2019 but was stopped in April 2020 due to the Covid-19 pandemic.

 Study design

This was an observational cohort study involving all children admitted to hospitals within Mchinji district during the study period.

 Participants

Children aged 0-12 years old who were admitted to a paediatric ward at any of the four hospitals within Mchinji District during the study period were eligible for enrolment to the study. The only exclusion criteria were neonates (age <28 days) who were born within the included hospitals, as they were unable to present with a referral history.

 Data sources

Data collection involved interviews with guardians and patient file review which was performed by study data collectors. Two of these were based at MDH and one each in the three community hospitals. The data collectors had a minimum education of completed secondary school and could read and write in English and Chichewa. All data collectors attended one week of training on the study protocol and ethics in research. They were supervised on a weekly basis by field supervisors and a clinical officer. Data collectors approached caretakers of admitted children in the hospitals and asked for their consent to participate in the study. Following consent, data was collected on whether the caretaker had brought the child straight to the hospital, or whether another health facility had been visited before coming to the hospital. Patient demographics, including patient age, sex, admission diagnosis, the time and day of the hospital admission and outcomes were recorded from patient files and ward admission books. Data collectors follow up on patient outcomes. Data was entered into password protected tablets using Open Data Kit (ODK) for electronic data collection. A data collection supervisor and a data manager checked the data for completeness on a weekly basis. In our case the data was then exported to Stata for analysis. Data can also be analysed in other software such as R. As data was collected at the hospital during admission, and records were kept over all patients, the risk of loss to follow up bias was minimal. All admitted children in the relevant age groups were invited to participate, eliminating the risk of selection bias.

 Sample size

The initial plan was to include all admitted children for one year using routine hospital admission data from the Mchinji district. The assumption was that data from Mchinji would sufficiently be representative to comfortable make generalisations of the whole population of interest. While data collection was stopped due to the Covid-19 pandemic the inclusion of close to 5000 children was still deemed enough to yield statistically significant results.

 Data analysis

The primary outcome variable for the analysis was inpatient mortality, and primary exposure was if the child had been referred from elsewhere prior to hospital admission. Age and sex of the child, day and time of admission and admission diagnosis were considered potential confounders. Sex was determined from medical records and any difference was explored as there are differences in mortality patterns between sexes in children under-five. All variables were converted into categorical variables. Admission days were categorized as weekdays (Monday to Friday) or weekends (Saturday and Sunday), and time as “day” (8 am to 8 pm) or “night” (8 pm to 8 am). This gave an indication if hospital admissions outside of daytime working hours, when there are less staff on the wards, were associated with higher mortality. Children could be assigned multiple admission diagnoses, we created binary variables for diagnosis category, meaning a total of more than 100% is presented in the diagnosis variable.

Proportions were stratified by facility type (district hospital versus community hospitals) and compared using chi 2 ( Table 1). Logistic regression analysis was done to determine the associations between the main exposure and outcome ( Table 2). All statistical analysis was preformed using Stata/IC 16.1. Stata was chosen as the researchers had license and experience using this software. Alternatively, the free statistical software R could have been used.

 Ethical considerations

Guardians were informed about the study and provided verbal consent for their minors to participate in the study prior to any data collection. Due to literacy levels, study information was given verbally in Chichewa, and consent given verbally –the informed consent was subsequently recorded in the electronic data collection form. Refusal to participate had no impact on the care provided to the patients and study procedure including consenting was approved by the Malawi College of Medicine Research and Ethics Committee (reference: P11/18/25389).

 Results

A total of 4926 children, 2322 (47.1%) female and 2604 (52.9) males, were admitted to hospitals in Mchinji district during the study period. Of these, 172/4926 (3.5%) where less than one month old, 852 (17.3%) where 1 month to 12 months old, 2758 (56.0%) were between 1 and 5 years, and 1138 (23.1%) were aged 5 to 12 years. 4265 (86.6%) of the total admitted children had been brought straight to hospital, and 661 (13.4%) children had been referred from another health care provider ( Table 1). The proportion of admitted children that had been referred was higher at the district hospital compared to the community hospitals (29.0%, vs 3.9%, p<0.001). Out of the 4926 children who were admitted, 141 died with an overall in-hospital mortality of 2.9%. The in-hospital mortality was 5.3% for referred children compared to 2.5% for the non-referred cases (p<0.001). The in-hospital mortality was 3.3% at the district hospital compared to 2.1% at community hospitals for non-referrals (p=0.021). For referred cases, the in-hospital mortality was 5.3% at the district hospital and 5.2% at the community hospitals (p=0.998). The overall in-hospital mortality was higher at the district hospital at 3.9%, compared to 2.2% at the community hospitals (p=0.01). Of the 545 children referred to the district hospitals, 33 had been referred from any of the included community hospitals and 5 of these died (CFR 15.1%).

Distribution of patient characteristics and mortality between hospitals.
Total Mchinji district hospital Community hospitals
Deaths (n/N) Case fatality rate (%) Deaths (n/N) Case fatality rate (%) Deaths (n/N) Case fatality rate (%)
Referral
Referred 35/661 5.3 29/545 5.3 6/116 5.2
Non-referred 106/4265 2.5 44/1332 3.3 62/2933 2.1
Age
<29 days 6/172 3.6 2/41 4.9 4/131 3.1
1–12 months 33/852 3.9 18/283 6.4 15/569 2.6
>1–5 years 79/2758 2.9 37/1042 3.6 42/1716 2.5
>5–12 years 23/1138 2.0 16/505 3.2 7/633 1.1
Sex
Female 71/2322 3.1 38/870 4.4 33/1452 2.3
Male 70/2604 2.7 35/1007 3.5 35/1587 2.2
Admission diagnosis
Pneumonia 17/656 2.6 12/354 3.4 5/302 1.3
Malaria 97/3345 2.9 44/1025 4.3 53/2320 2.3
Diarrhoea 8/343 2.3 2/74 2.7 6/269 2.2
Sepsis 22/1047 2.1 6/248 2.4 16/799 2.0
Anaemia 37/441 8.4 21/180 11.7 16/261 6.1
Malnutrition 8/58 13.8 3/34 8.8 5/24 20.8
Trauma 3/198 1.5 2/121 1.7 1/77 1.3
Other 10/439 2.3 4/152 2.6 6/287 2.1
Admission time
Weekday 113/3603 3.1 59/1424 4.1 54/2179 2.5
Weekend 28/1323 2.1 14/453 3.1 14/870 1.6
Day 120/3978 3.0 65/1683 3.9 55/2295 2.4
Night * 21/948 2.2 8/194 4.1 13/754 1.7

Admission any time from 8 pm to 8 am the following day.

 Referral history and in-hospital mortality risk

The adjusted odds ratio of in-hospital mortality was 1.9 (95% CI:1.3-2.9) for children who were admitted to hospital who had been referred compared to those who was not referred. Table 2 presents the unadjusted (UOR) and adjusted Odds Ratios (AOR) for the variables included in the logistic regression analysis.

Association of patient characteristics and presence of referral with in-hospital mortality.
Variable Unadjusted odds ratio p-value 95% CI Adjusted odds ratio p-value 95% CI
Referral
Not referred 1.0
Referred 2.2 <0.001 1.5-3.2 1.9 0.001 1.3-2.9
Age
<29 days 1.0
1–12 months 1.1 0.810 0.5-2.7 0.7 0.545 0.3-2.0
>1–5 years 0.8 0.637 0.4-1.9 0.5 0.106 0.2-1.2
>5–12 years 0.6 0.239 0.2-1.4 0.3 0.028 0.1-0.9
Sex
Male 1.0
Female 1.2 0.438 0.8-1.6 1.1 0.488 0.8-1.6
Admission diagnosis *
Pneumonia 0.9 0.655 0.5-1.5 1.0 0.904 0.6-1.9
Malaria 1.0 0.818 0.7-1.5 1.4 0.195 0.6-2.2
Sepsis 0.7 0.098 0.4– 1.1 0.9 0.520 0.5-1.4
Diarrhoea 0.8 0.542 0.4-1.6 1.0 0.950 0.5-2.2
Anaemia 3.9 <0.001 2.6-5.7 4.1 <0.001 2.6-6.3
Malnutrition 5.7 <0.001 2.6-12.3 6.2 <0.001 2.7-14.4
Trauma 0.5 0.255 0.2-1.6 1.0 0.971 0.3-3.4
Other 0.7 0.443 0.4-1.5 0.9 0.719 0.4-1.8
Admission time
Weekday 1.0 0.059 0.4-1.0 0.8 0.201 0.5-1.2
Weekend 0.7
Night 1.0
Day 1.4 0.185 0.9-2.2 1.2 0.483 0.7-1.9

More than one diagnosis per child possible.

For the different age-groups, neonates (age <29 days) represented 3.5% of all admissions (172/4926), with a CFR of 3.6%. Infants, aged 1 to 12 months, represented 17.3% (852/4926) of the admitted children and had the highest CFR at 3.9%. The biggest group of admitted children (56.0%) were between 1 and 5 years of age (2758/4926) while 1138 children (23.1%) were between 5 to 12 years of age. The children aged 1-5 year olds had a CFR of 2.9% and the CFR for children aged 5-12 years old was 2.0%. Only a small difference was seen depending on sex, with 52.9% of the admitted children being males and 47.1% females, and with a CFRs of 2.7% and 3.1%, respectively (p-value: 0.437).

Most of the other included factors did not show any statistical significance. The exceptions were children in the age-group from 5 to 12 years, whose AOR for in-hospital mortality was 0.3 compared to neonates (95% CI: 0.1-0.9).

The diagnosis with the highest CFR was malnutrition at 13.8%, followed by anaemia at 8.5% with an AOR of 6.2 (95% CI 2.7-14.4) and AOR 4.1 (95% CI: 2.6-6.3) respectively. A bigger proportion of children died from malnutrition at the community hospitals than at the district hospital, however the result was not significant (20.8%, vs 8.8%, p<0.191). The proportion that died of anaemia was higher in the district hospital (11.7%) than at the community hospitals (6.2%, p=0.039).

 Discussion

We observed a higher in-hospital mortality among children who had been referred from a lower-level facilities compared to children who were brought straight to hospital in a rural Malawian setting. Most children who are admitted to hospital had been brought there without seeking any previous care at lower-level facilities. In terms of diagnosis, the evidence suggests children with malnutrition and anaemia carry an increased mortality risk.

The in-hospital mortality was twice as high for children who had been referred to hospital from a lower-level facility compared to children who were brought straight to the hospital. The CFR was highest for children who had been referred from any of the community hospitals to the district hospital. While the study did not establish the actual causes of the mortality, and can be assumed that referred children were likely to be sicker than the non-referred children, and thus at higher risk of dying. However additional explanations to the mortality difference could be that adequate care was delayed during the time it took to seek and receive care elsewhere, followed by the time it took to complete referrals. Lower-level facilities have limited abilities to offer stabilisation of severely ill children with transport issues for referrals reported as a main constraint. 17 , 18 Delays may also occur within facilities including the decision to refer a child. 19 While IMCI was rolled out in 1999 and include guidelines on when to refer a sick child, challenges in the sustainability have been reported including limitations in equipment, training, and adherence to guidelines. 18 , 20 To ensure that all in need are referred early, current IMCI guidelines may need more objective definitions for timely stabilisation. 21 , 22

The majority of children admitted to the included hospitals had come straight to the hospital. This may suggest a good awareness among guardians on when a child’s illness requires hospital care. Alternatively, there is a general preference for hospital care which could be due to lack of trust in lower level facilities. 23 A study conducted in Southern Malawi showed that patient satisfaction with the primary healthcare system demonstrated considerable variation 24 and better facility quality is associated with a higher utilisation of sick child healthcare services. 25

The highest mortality was seen among children diagnosed with malnutrition and anaemia. This confirms findings from previous studies 26 – 28 and children with these diagnoses also suffer an increased risk of post-discharge mortality. 29 , 30 It is possible the ETAT guidelines need to pay more attention to malnourished children with multiple diagnoses for priority and stabilisation at admission, and IMCI and iCCM should focus on earlier referral for children with malnutrition and anaemia. For age differences, young children had the highest mortality. While neonates generally suffer the highest mortality rates, in this study the mortality was highest among infants. Part of this may be explained by the study exclusion of babies born within the facilities, a study on place of death and care-seeking prior to death that poor illness recognition is a major driver in neonatal deaths, whereas death despite care-seeking among older children indicates inadequate quality care and referral. 31

The strengths of this study include a district wide approach with a big sample of children included. However, there are some major limitations that preclude any conclusion on the overall mortality-risks in the study population. Firstly, it is possible that referred children as a group were suffering from more severe conditions than the ones who were not-referred. The study did not collect any information on the symptoms at the time the decision to seek care was made or whether a progression occurred during the time spent on care seeking prior to reaching the hospital. Indeed, the high CFR among children referred from community hospitals to the district hospital may reflect on illness severity and a higher need for more advanced care among these children, rather than failures within the referral system. Secondly, we did not collect any data on reasons for referrals and whether the selection of children to be referred adhered to guidelines. Thirdly, there is also the possibility that caretakers who were recommended referral to hospital could not make their way there, causing an under-estimation of post-referral mortality. Similarly, we did not assess post-discharge mortality which is reportedly high in many SSA settings. 29

To conclude, the higher in-hospital mortality demonstrated between children who were referred from a health centre compared to not-referred cases may be caused by more severe illness and greater need for more advanced care. However, given the overall high CFRs among referred cases it is also plausible that it is linked to poor quality care at primary healthcare facilities, inadequate guidelines for when to refer a sick child and/or or difficulties to completing referrals. Future studies to better understand the demonstrated mortality pattern should include data on illness severity and caretaker’s reasons for care-seeking choices. The vulnerability of children with malnutrition and/or anaemia should be especially considered when assessing sick children at all levels of the health care system.

 Data availability

Repository: Harvard Dataverse: Mchinji ward admission data. https://doi.org/10.7910/DVN/E9ZJNH. 32

This project contains the following underlying data:

230522 Ward Admission Data.xls (variables used in analysis)

Data dictionary Mchinji hospital data.xlsx (data dictionary)

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).

 Acknowledgements

We would like to thank all the children and their parents/caretakers who participated in this study, and the healthcare workers who supported our data collector teams in their work. We are also grateful to the data collectors for their hard work, and the Mchinji District Health Management Team for their input and support.

 References Hug L Sharrow D You D : Levels & Trends in Child Mortality, Report 2019. 2019 [cited 2019 29th December]. Reference Source UN and D.o.E.a.S. Affairs : Goals 3. Ensure healthy lives and promote well-being for all at all ages. Targets and Indicators. 2024. Reference Source Group, W.B : Mortality rate, under-5 (per 1,000 live births) - Sub-Saharan Africa, Malawi. cited2024. Kruk ME : High-quality health systems in the Sustainable Development Goals era: time for a revolution. Lancet Glob. Health. 2018;6(11):e1196–e1252. 30196093 10.1016/S2214-109X(18)30386-3 PMC7734391 WHO: Integrated Management of childhood Illness Chart Booklet. 2014 [cited 2020 18th April]. Reference Source Johansson EW : Accessibility of basic paediatric emergency care in Malawi: analysis of a national facility census. BMC Public Health. 2020;20(1):992. 32580762 10.1186/s12889-020-09043-3 PMC7315502 Krüger C Heinzel-Gutenbrunner M Ali M : Adherence to the integrated management of childhood illness guidelines in Namibia, Kenya, Tanzania and Uganda: evidence from the national service provision assessment surveys. BMC Health Serv. Res. 2017;17(1):822. 29237494 10.1186/s12913-017-2781-3 PMC5729502 Ahmed T : Healthcare utilization and maternal and child mortality during the COVID-19 pandemic in 18 low- and middle-income countries: An interrupted time-series analysis with mathematical modeling of administrative data. PLoS Med. 2022;19(8):e1004070. 36040910 10.1371/journal.pmed.1004070 PMC9426906 Källander K : Barriers on the pathway to survival for children dying from treatable illnesses in Inhambane province, Mozambique. J. Glob. Health. 2019;9(1):010809. 31275569 10.7189/jogh.09.010809 PMC6596358 Noordam AC : Association between caregivers’ knowledge and care seeking behaviour for children with symptoms of pneumonia in six sub-Saharan African Countries. BMC Health Serv. Res. 2017;17(1):107. 28153011 10.1186/s12913-017-2060-3 PMC5290628 Molyneux M Molyneux E : Reaching Millennium Development Goal 4. Lancet Glob. Health. 2016;4(3):e146–e147. 10.1016/S2214-109X(16)00009-7 Malawi, G.o.t.R.o : Malawi Health Sector Strategic Plan III (HSSP III) 2023-2030. Reforming for Universal Health Coverage;2023. Makwero MT : Delivery of primary health care in Malawi. Afr. J. Prim. Health Care Fam. Med. 2018;10(1):e1–e3. 29943590 10.4102/phcfm.v10i1.1799 WHO: Emergency Triage Assessment and Treatment (ETAT) course. 2005 [cited 2020 18th April]. Reference Source Chiunda C M K : 2018 Malawi Population and Housing. Census. Government of Malawi;2019. Group, W.B . The World Bank in Malawi. 2024. King C : Paediatric Emergency Triage, Assessment and Treatment (ETAT) - preparedness for implementation at primary care facilities in Malawi. Glob. Health Action. 2021;14(1):1989807. 34779363 10.1080/16549716.2021.1989807 PMC8592602 Robertson SK Manson K Fioratou E : IMCI and ETAT integration at a primary healthcare facility in Malawi: a human factors approach. BMC Health Serv. Res. 2018;18(1):1014. 30594185 10.1186/s12913-018-3803-5 PMC6310991 King C : Care-seeking patterns amongst suspected paediatric pneumonia deaths in rural Malawi. Gates Open Res. 2020;4:178. 10.12688/gatesopenres.13208.1 Kilov K : Integrated Management of Childhood Illnesses (IMCI): a mixed-methods study on implementation, knowledge and resource availability in Malawi. BMJ Paediatr. Open. 2021;5(1):e001044. 34013071 10.1136/bmjpo-2021-001044 PMC8098945 King C : Prospective cohort study of referred Malawian children and their survival by hypoxaemia and hypoglycaemia status. Bull. World Health Organ. 2022;100(5):302–314B. 35521039 10.2471/BLT.21.287265 PMC9047421 Thunberg A : Hypoxemia, hypoglycemia and IMCI danger signs in pediatric outpatients in Malawi. Plos Global Public Health. 2022;2(4):e0000284. 36962312 10.1371/journal.pgph.0000284 PMC10021275 Kahabuka C : Why caretakers bypass Primary Health Care facilities for child care - a case from rural Tanzania. BMC Health Serv. Res. 2011;11(1):315. 22094076 10.1186/1472-6963-11-315 PMC3234197 Dullie L : Performance of primary care in different healthcare facilities: a cross-sectional study of patients’ experiences in Southern Malawi. BMJ Open. 2019;9(7):e029579. 31324683 10.1136/bmjopen-2019-029579 PMC6661549 Liu L : Exploring the association between sick child healthcare utilisation and health facility quality in Malawi: a cross-sectional study. BMJ Open. 2019;9(7):e029631. 31352421 10.1136/bmjopen-2019-029631 PMC6661667 Berkley JA : Prognostic indicators of early and late death in children admitted to district hospital in Kenya: cohort study. BMJ. 2003;326(7385):361. 10.1136/bmj.326.7385.361 Ngwalangwa F : Risk Factors for Mortality in Severely Ill Children Admitted to a Tertiary Referral Hospital in Malawi. Am. J. Trop. Med. Hyg. 2019;101(3):670–675. 31287044 10.4269/ajtmh.19-0127 PMC6726928 Obonyo CO : In-hospital morbidity and mortality due to severe malarial anemia in western Kenya. Am. J. Trop. Med. Hyg. 2007;77(6 Suppl):23–28. 18165471 10.4269/ajtmh.77.6.suppl.23 Childhood Acute I Nutrition N : Childhood mortality during and after acute illness in Africa and south Asia: a prospective cohort study. Lancet Glob. Health. 2022;10(5):e673–e684. Kwambai TK : Post-discharge morbidity and mortality in children admitted with severe anaemia and other health conditions in malaria-endemic settings in Africa: a systematic review and meta-analysis. Lancet Child Adolesc. Health. 2022;6(7):474–483. 35605629 10.1016/S2352-4642(22)00074-8 PMC10196725 Price J : Place of death, care-seeking and care pathway progression in the final illnesses of children under five years of age in sub-Saharan Africa: a systematic review. J. Glob. Health. 2019;9(2):020422. 31673338 10.7189/jogh.09.020422 PMC6815655 Helena H : Mchinji ward admission data. Harvard Dataverse .2023; V3. 10.7910/DVN/E9ZJNH 10.5256/f1000research.176145.r440239 Reviewer response for version 2 Baltzell Kimberly 1 Referee University of California San Francisco, San Francisco, USA

Competing interests: No competing interests were disclosed.

 31 12 2025 Copyright: © 2025 Baltzell K 2025 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation reject

Summary: This article summarizes a study from central Malawi, designed to compare mortality rates between sick children from 0-12 years old presenting directly to the hospital with those presenting to the hospital from a health centre. Three hospitals were included in the study, one at the district level and three at the community level. Approximately 5000 children were included in the study, with the majority presenting directly at the hospital. Investigators used patient records and ward admission books to obtain information on general demographics and time and diagnosis at admission. Data collectors then followed up on patient outcomes. The investigators hypothesized that children first seen at a primary health centre were likely to have worse outcomes due to a delay in seeking needed care. Findings supported the hypothesis as children referred for care were significantly more likely to die than children who reported straight to the hospital for care.

Background: The authors clearly outline the setting and context for the study, including the majority of the study population live in rural areas. Current literature supports much of the background information.

Methods: While the authors hypothesize poor quality of care at health centres may be responsible for differences in children mortality. However, they did not collect information on health centre admitting diagnosis, or time/distance from health centre to the hospitals for those who were referred.

Results: It is not clear if “referred” means those coming from community hospitals as well as primary health centres, although the differences in mortality rates are noted. It is also not clear if the cases seen at the community hospitals and then referred to the district hospital (33) were excluded or categorized with health centre referrals.

Discussion: The authors note that children sent directly to the hospital may have guardians who recognized illness sooner, it could be a question of simple proximity. It is possible that those sent directly to the hospital lived closer and it was the closest point of care.

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

No

Are all the source data underlying the results available to ensure full reproducibility?

Partly

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

No

Are sufficient details of methods and analysis provided to allow replication by others?

Yes

Reviewer Expertise:

I am a nurse scientist with a history of work in Malawi and also maternal and neonatal health.

I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

 10.5256/f1000research.176145.r442888 Reviewer response for version 2 Blair Alden 1 Referee University of California San Francisco, San Francisco, California, USA

Competing interests: No competing interests were disclosed.

 30 12 2025 Copyright: © 2025 Blair A 2025 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation reject

This study proposes a quite interesting query, as to whether under-5 mortality in low-resource settings is partially driven by delays in higher care that occur in primary care settings. The authors hypothesize this to be the case and utilize a large dataset of patient-level data from multiple hospitals in Mchinji District, Malawi. This sees them with a great deal of data from which to assess their exploration as to whether there is a statistically significant difference in mortality between those who self-referred directly to a hospital and those who were referred from a peripheral health center.

It is unsurprising to see that those patients who were referred from peripheral primary care health centers had higher mortality than those who were direct-entry to hospitals. While the authors propose that this backs up their hypothesis, the study design does not allow them to directly answer this question. 

The main issues are that of the representativeness between the two groups compared; those self-referring and those who were referred from the primary health centers. One can assume that those who are referred are by nature sicker patients if only due to the fact that their underlying condition could not be cared for at the primary health center. For this reason, their baseline health status is likely worse. Similarly, patients who are able to direct-refer to a hospital likely live closer to said facility as it was their first stop, suggesting that those referred have temporal and distance-related aspects to their profile. However, the authors were unable to account for these known confounders in their study, which would be important. 

To truly answer the research question, it would be important for the authors to be able to collect and analyze data from the peripheral health centers to see what their overall mortality rate was, not just the mortality rate for the patients who they refer to hospitals. This would show if patient who first seek care at the health centers have higher mortality than those who direct refer to hospitals. As the current comparison stands, we are only comparing a small subset of that population to the larger population of everyone seeking direct care from a hospital, making the two samples problematic for comparison.

Despite the issues with the framing of the paper, it is well written and thoughtful. The authors have a great deal of data and present it clearly. A reworking of the exploration to perhaps focus on the drivers of mortality would be beneficial as the dataset itself cannot answer the research question proposed.

Is the work clearly and accurately presented and does it cite the current literature?

Yes

If applicable, is the statistical analysis and its interpretation appropriate?

No

Are all the source data underlying the results available to ensure full reproducibility?

Partly

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

No

Are sufficient details of methods and analysis provided to allow replication by others?

Yes

Reviewer Expertise:

I am a mixed-methods epidemiologist whose work focuses on maternal and neonatal health in low- and middle-income countries, with a specific focus on rural and remote areas.

I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

 10.5256/f1000research.147003.r319092 Reviewer response for version 1 Swaminathan Sumathi 1 Referee St John’s National Academy of Health Sciences, St John's Research Institute, Bengaluru, Karnataka, India

Competing interests: No competing interests were disclosed.

 21 10 2024 Copyright: © 2024 Swaminathan S 2024 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation reject

Your Report

Please provide a full report, expanding on your answers to the questions above. In particular, if you answered “no” or “partly” to any of the questions, please give constructive and specific details as to how the authors can address any criticisms. Please indicate clearly which points must be addressed to make the article scientifically sound. ( 50 words min.)

The manuscript describes the type of admissions at one district hospital and 3 rural or community hospitals in the Mchingi district at Malawi. The aim of the study as per the authors, is to “assess the care-seeking patterns prior to being admitted to hospital, and to establish the association between pre-admission referrals and inpatient mortality for children in Mchinji District, Malawi”.

In the hierarchical health care system, the highest level of health care in a district must be the district hospital with more advanced facilities compared to the rural or community hospitals followed by the health centers available. While this needs to be described, it should be noted that the issue of inpatient mortality due to “self-referrals and referrals” does not arise here. Obviously lower-level hospitals with much less facilities will need to refer those children requiring advanced facilities for treatment based on the severity of their illness, to the community or district hospital depending on the level of treatment required. Self-referrals to district hospitals are likely only among those caregivers who live close to the district hospital rather than those who live in far off rural communities. Mortality is dependent on the severity of illness and the most severe cases are referred from the lower-level hospitals to the district hospital or community hospitals. Self -referrals may not necessarily be for children requiring intensive and advanced medical care as the degree of severity in condition may not be high. So, it is not surprising that mortality among referred cases is higher. And self-referrals lower.    

For the study exposure is defined as the referral and outcome as mortality. Do the authors mean referral to be type of referral? Based on the argument above, the type of referral as an exposure will not be a good choice for this study. Instead, the factors associated with mortality among those referred to higher facilities may be a better question to be examined.

From my readings of a government and World Bank documents, healthcare in Malawi and particularly Mchinji district operates through both government and private hospitals run by the Christian Health Association of Malawi. One document indicates that the Central Malawi region had 8 district hospitals and 16 Rural hospitals. It is not clear why the specific district was chosen and how the hospitals were chosen for the study. Authors state that “While all district in Malawi were not included it is assumed that data from Mchinji is sufficiently representative to comfortable make generalizations to the whole population of interest.” The sampling needs to be clearly explained.

The authors state that this is an observational retrospective study design. This needs to be explained as the rest of the methods states that consent was taken from caregivers as the children got admitted. From my understanding the hospital case records as and when children were admitted must have been referred to. Therefore, this must have been a prospective study.

In Table 2, there are 2 types of referrals mentioned- not referred and referred. There is no reference to self-referral. From this table it is obvious that the factors associated with mortality in referred cases is important to be studied.  The duration of hospital stays before death also needs to be reported.

The data uploaded does not give specific details of diagnosis, duration and severity of condition. It should be noted that children are referred only if they cannot be managed by lower-level healthcare providers. From the present minimal data shown, most children have died within a couple of days of admission, whether self-referred or referred by a healthcare provider. There is not sufficient data to indicate why the mortality occurred. Further several children have had co-morbidities apart from the primary condition. This has not been reported in the analysis. There are several with “other” diagnosis done as per the raw data. What this refers to is not clear.

With minimal data presented, the discussion can obviously not be clearly written. Overall, the whole manuscript should be based on more robust data collected.

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

Partly

Are all the source data underlying the results available to ensure full reproducibility?

Partly

Is the study design appropriate and is the work technically sound?

No

Are the conclusions drawn adequately supported by the results?

No

Are sufficient details of methods and analysis provided to allow replication by others?

No

Reviewer Expertise:

Nutrition, Public health, Epidemiology

I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

Selstø Annlaug Mental and Physical Health, Norwegian Institute of Public Health, Oslo, Norway

Competing interests: No competing interests to declare.

 23 12 2024

Dear Sumathi Swaminathan

We appreciate the comments and have revised the manuscript accordingly. Please see our responses  listed under, and our new version of our article.

Thank you for your reflections and comments on the health care sytem and the referral mechanism. We agree with you, and will still argue that our findings bear some interest in the difference observed in case fatality rates between those referred from another health centre and those who come straight to the hospital. We have discussed the limitations implied by the lack of severity assessments in the limitations section at the end of the discussion. We have also provided more information on the healthcare systemin Malawi under study setting. Additionally, we have added information on numbers referred from community hospitals to the district hospital with added sentences in the discussion to reflect on how severity of illness may relate to referral status.

We are not investigating “type of referral” but rather “referral or not referred”. In this sense “self-referral” might sound confusing. We have therefore renamed the exposure to referred or not-referred.

On the matter of chosen hospital in Mchinji we have included the three community hospitals whereof all are CHAM ( Christian Health Association of Malawi, the major health provider among NGOs) facilities. These are the hospitals operating within Mchinji district. The study is limited to Mchinji district only. 

On the matter of representativeness of the sample: We had data collectors present at all facilities ranked as “hospitals” within Mchinji district during the study period. These collected information from all admitted children. The initially planned sample was for one full year of data collection which we expected would give enough numbers to ensure statistically significant results. Unfortunately, data collection was stopped early due to the covid pandemic which made us halt data collection at close to 5000 included children. However, this number turned out to show statistically significant results in terms of mortality difference.

Study design: We have changed this and removed the word “retrospective

Types of referral and factors associated with mortality: We agree that our findings merit further investigations, and we are currently working on studies to provide more in-depth data to understand the situation better. We are aware that the current study contains several limitations as stated at the end of the discussion.

Our primary outcome is hospital mortality, and the duration of admission was not included as part of data collection in this study.

We agree with you that there are several additional factors that are highly relevant to understand the reasons for fatal outcomes among admitted children and that the lack of these factors highly impacts the conclusion made from our study. Still, we believe that the finding of a higher case fatality among those referred compared to those who were not referred from another facility provides an interesting information also in that it emphasizes the need to better understand background characteristics and referral practices between different groups. We do by no means think this manuscript can provide a full picture and the various limitations are stated and discussed at the end of the manuscript. We have also added information both in results and discussion on the presumed impact of illness severity on referral decisions and outcomes.

 Kind regards,

on behalf of the research team

Annlaug Selstø

 10.5256/f1000research.147003.r319089 Reviewer response for version 1 Vonasek Bryan 1 Referee https://orcid.org/0000-0002-5657-0980 Michigan State University College of Osteopathic Medicine, East Lansing, Michigan, USA

Competing interests: No competing interests were disclosed.

 18 9 2024 Copyright: © 2024 Vonasek B 2024 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation reject

This manuscript focuses on a very important topic and includes a large dataset to explore associations with in-hospital mortality. However, there are several major limitations in the study design and interpretation of results that need to be addressed. 

MAJOR CONSIDERATIONS

1. The title and elsewhere seem to indicate "care-seeking" as what is being explored as an exposure variable. But the main exposure variable being explored is clearly whether participants were referred or not, which is completely different than "care-seeking." When I think of "care-seeking," it is the decision making (or lack thereof) from sick individual or his/her caregiver or community. This isn't directly assessed at all in this manuscript. All sections of the manuscript need to be revised to address this issue.

2. The study design is stated as an "observational retrospective cohort study"--but then enrollment with informed consent is described. Please clarify about the study design.

3. The methods for calculation of unadjusted and adjusted odds ratios are not described at all.

4. Given that you seem to capture data from all four government hospitals in the district, is it possible to evaluate those referred to MDH from the community hospitals, in comparison to those admitted to the community hospital that weren't referred to MDH? At least report how many of the referrals to MDH were referred from the three community hospitals in the study.  

MINOR CONSIDERATIONS

1. Methods > Study Setting: The goal of <25 deaths per 1000 is stated. It would be informative to include what this number is currently (maybe at the national level).

2. Methods > Participants: No need to also mention the exposure and outcome variables here. 

3. Ethical considerations: No need to mention COMREC twice.

4. Why are the diagnoses of meningitis and sepsis combined?

5. Table 1 - add CFR units

6. Is it necessary to present both chi-square results and odds ratios? Probably don't need chi-square results if the ORs are there for the same comparisons.

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

Partly

Are all the source data underlying the results available to ensure full reproducibility?

Yes

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

Partly

Are sufficient details of methods and analysis provided to allow replication by others?

No

Reviewer Expertise:

Under 5 mortality in Malawi

I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

Selstø Annlaug Mental and Physical Health, Norwegian Institute of Public Health, Oslo, Norway

Competing interests: No competing interest to declare.

 23 12 2024

Dear Bryan Vonasek

Thank you for your thorough reading and review of our article. We appreciate your comments and have revised the manuscript accordingly. Please see our responses listed below. Pleas also see our lates updated version of our manuscript.

For your major considerations:

1. Thank you for this valuable comment on our title. We agree with the feedback that "care-seeking" is not correctly reflecting what has been studied in this work. Referral is just one factor out of several in the process of reaching out for professional medical care. We have revised the title and the manuscript to better describe what has been done.

2. On the matter of study design we agree with your comment and has removed the wording retrospective throughout. Children and their guardians were approached at the hospital where they were admitted and asked to consent to the study whereby we received approval to collect outcome information data.

3. Conserning UOR and AOR we describe that a logistic regression was done under “data analysis”. This was a univariate analysis followed by multiple logistic regression to achieve adjusted odds ratios.

4. Thank you for the useful comment on referrals to MDH versus community hospitals. We have added information on the numbers of referred from community hospitals to the district hospital. We have also added some sentences in the discussion to reflect on this. Please note that the community hospitals are CHAM facilities while the district hospital is under the government.

MINOR CONSIDERATIONS

1. Thank you for the comment requestibg information on child mortality in Malawi under Study setting.. The information requested is written in the beginning of the introduction.

2. Under methods/participation we have removed information on the exposure and outcome variables.

3. Duplicate on ethical considerations has been removed.

4. Results on meningitis and sepsis were combined as the diagnosis of meningitis commonly is made without any lumbar puncture there may be a big overlap between meningitis and sepsis why we choose to combined these. But there are very few with "meningitis" in our dataset so we chose to separate these after your comments. 

5. In table 1 CFR units (%) has been added.  

6. We have changed the result section to report primarily on ORs, thus removing results from chi square analysis.

Kind regards,

on behalf of the research team

Annlaug Selstø

 10.5256/f1000research.147003.r248631 Reviewer response for version 1 Ahmed Faisal Elgasim 1 Referee https://orcid.org/0000-0001-7188-2715 Urology Research Center, Al-Thora General Hospital, Department of Urology, School of Medicine, Ibb University of Medical Sciences, Ibb, Yemen

Competing interests: No competing interests were disclosed.

 21 3 2024 Copyright: © 2024 Ahmed FE 2024 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation approve-with-reservations

The authors present a useful study of the Association between care-seeking at health centers preceding hospital admission and in-hospital child mortality. Instead of an interesting study, it still needs extensive revision.

My comments/queries:

Abstract:

Background: Please enrich the background with a few extra details.

Method: The main outcome of your method needs to be clearly defined.

Result: The data does not clearly show what percentage of the children had risk factors associated with death. Additionally, some factors not mentioned, such as prematurity, failure to thrive, etc, may associated with death.

Introduction:

Briefly mention the difference in health care facilities offered in different other health facilities in your country.

Several analogous studies exist in the current literature. It is essential to underscore the significance of integrating this study into the existing body of research.

 Method:

Please mention the exclusion criteria.

The sample size should be revised. For larger populations, such as a population of 10,000, a comparatively small minimum ratio of 10 percent (1,000) of individuals is required to ensure the representativeness of the sample.

Please define the outcome and variables included in your study such as mortality, risk factors included, etc....

What are the reasons for being referred by another healthcare provider? It should be mentioned.

The authors need to mention which protocols were followed for the collected data of mortality to validate the results.

How was the checklist pre-validated? Authors can mention that all collected that reviewed, and checked by the Supervisor and the principal investigator for completeness, and consistency, and if any missing info was found, mitigation was done.

Authors need to mention in more detail the drawbacks of a retrospective study, especially since they have had no control over many variables.

The authors also need to emphasize that no causation was advocated for. Furthermore, the data was collected from a secondary source.

I think a table of the components and indicators of the Pathway to Mortality would have been good to include so it could be compared to the components chosen for this analysis.

Please mention the statistical analysis section including how you present the qualitative data, quantitative data, statistical tests used, etc…

 Result:

Give characteristics of study participants (eg demographic, clinical, social). Please mention the patients' age, and gender, then go to specific details.

The low number of people who sought care from community health workers is interesting but the reason for this is not explained.

Table 1: The data does not clearly show what percentage of variables. Please include the percentage for each variable.

Please mention the case fatality rate on your method.

This section is hard to follow, and it should be improved. Please start with a general description of the study participants. Then, mention the specific analysis comparison between your groups. the tables also should be revised similarly.

Before mentioning the adjusted OR table, you should mention the table of univariate analysis.

The disease severity is not mentioned. Why?

 The duration of hospital admission should be mentioned.

Where is the reference group for each variable in Table 2?

The OR should be in the reported range of 95%CI, however, some adjusted OR are out of the reported period. Please recheck your analysis.

">1-5 years, >6-12 years"; please revise it through the tables.

 Discussion:

- The first paragraph should summarize your result. Not a selected part of the study. Please revise this section.

- Please mention additional Strengths and limitations of your study.

- The factors discussed are very few. Please discuss the important findings of the obtained study and compare the findings with previous reports with accurate justifications.

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

Partly

Are all the source data underlying the results available to ensure full reproducibility?

Partly

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

Partly

Are sufficient details of methods and analysis provided to allow replication by others?

Partly

Reviewer Expertise:

I am a urologist. I have worked for 5 years in the field with governmental hospitals in Yemen mainly in the field of surgery and trauma.

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

Selstø Annlaug Mental and Physical Health, Norwegian Institute of Public Health, Oslo, Norway

Competing interests: No competing interests to declare.

 20 12 2024

Dear Editors

Thank you for the opportunity to review our manuscript. We appreciate the comments form the reviewers and have revised the manuscript accordingly. Please see our responses to the reviewers listed under each comment below.

Abstract:

Background: please see information added in new version.

Method: The main outcome has been clarified under abstract methods.

Results: Thank you for this comment. While we agree there are multiple potential mortality risks, it was beyond the scope of this study to address all potential risk factors as the main focus is on comparing hospital outcomes and previous referral-history.

 Introduction

Information on the Malawian health system has been added under “study setting” as we think it fits better there than in the introduction.

Information has been added to the introduction to highlight the difference in child mortality and progress to SDG 3 in Malawi compared to SSA as a whole.

 Method: 

The exclusion criteria have been clarified.

As all children admitted to the four included hospitals in the whole of Mchinji district were included we expected to have enough numbers to identify any significant difference. Despite the need to stop the study earlier than expected due to Covid, we still included enough children to demonstrate a statistically significant difference.

The exposure and outcome variables as well as potential confounders are listed under “data analysis”.

Malawian health workers are supposed to adhere to national guidelines on referral decisions. We did not collect data on specific reasons for referrals in this study.

A sentence has been added to “data sources” to clarify that data collectors were present at all hospitals to follow up on outcomes, verified from patient files and hospital records.

Information on completeness checks has been added under “data sources”.

A presentation of the different limitations in this study, including the unavailability of information on reason for referrals, is available at the end of the discussion section.

The limitations of not having all possible risk factors available in this report are mentioned at the end of the discussion section.

We agree with the reviewer that multiple additional components could have been included to get a more thorough understanding of all factors leading to fatal outcomes. However, this is a small study with the main aim to compare differences between children brought straight to hospital with those being referred. The limitations of being a small study are discussed under study limitations.

This was, as stated under study design, a cohort study. An explanation of statistical methods used is available under “data analysis”. The study did not include any qualitative data. We have added reference to the two included tables under “data analysis”.

 Results:

We did not collect data on social and demographic factors. Background characteristics are presented in Table 1 and text has been added to the first part of the result section.

We did not collect any specific information on referral numbers from community health workers, thus we do not fully understand this question. The limitations caused by the unavailability of pre-referral data are discussed at the end of the discussion. 

Thank you for the feedback. The percentage symbol has been added in the heading of the table.

The case fatality rates are presented in Table 1 as well as in the end of the first section of “results”.

We have changed the text to present more on general description of study participants before moving in the comparisons between groups.

Table 2 presents the unadjusted/univariate as well as the adjusted ORs.

We did not collect any specific data on illness severity on condition. The limitations of this approach are discussed at the end of the discussion section.

Our primary outcome is hospital mortality, and the duration of admission is thus not included in this study.

The reference group is the first group mentioned for each variable, indicated with an UOR of 1.0 as in our experience is common practice in scientific reports.

Thank you for pointing out concern on 95%-CI – we have revised all the numbers reported to ensure they are correctly reflecting the analysis done.

These are the age groups used in our analysis; <29 days ,1-12 months, >1-5 years and >5-12 years The > indicates that a child of 5 years and one day will be in the  >5-12 years group while a child of exactly 5 years will be in the >1-5 years group

 Discussion: 

We present our main findings at the start of the discussion. This aligns with the objective of the study.

Strengths and limitations are discussed at the end of the discussion section.

As this is not a big study, we think we can only discuss what we have found, highlight the limitations of our study and relate our findings to previous work done.