key: cord-0750268-75df6h0q
authors: King, Carina; Zadutsa, Beatiwel; Banda, Lumbani; Phiri, Everlisto; McCollum, Eric D; Langton, Josephine; Desmond, Nicola; Qazi, Shamim Ahmad; Nisar, Yasir Bin; Makwenda, Charles; Hildenwall, Helena
title: Prospective cohort study of referred Malawian children and their survival by hypoxaemia and hypoglycaemia status
date: 2022-05-01
journal: Bull World Health Organ
DOI: 10.2471/blt.21.287265
sha: ad9c6391ac0e841c1d747e0c224e8f02cc980a21
doc_id: 750268
cord_uid: 75df6h0q

OBJECTIVE: To investigate survival in children referred from primary care in Malawi, with a focus on hypoglycaemia and hypoxaemia progression. METHODS: The study involved a prospective cohort of children aged 12 years or under referred from primary health-care facilities in Mchinji district, Malawi in 2019 and 2020. Peripheral blood oxygen saturation (SpO(2)) and blood glucose were measured at recruitment and on arrival at a subsequent health-care facility (i.e. four hospitals and 14 primary health-care facilities). Children were followed up 2 weeks after discharge or their last clinical visit. The primary study outcome was the case fatality ratio at 2 weeks. Associations between SpO(2) and blood glucose levels and death were evaluated using Cox proportional hazards models and the treatment effect of hospitalization was assessed using propensity score matching. FINDINGS: Of 826 children recruited, 784 (94.9%) completed follow-up. At presentation, hypoxaemia was moderate (SpO(2): 90–93%) in 13.1% (108/826) and severe (SpO(2): < 90%) in 8.6% (71/826) and hypoglycaemia was moderate (blood glucose: 2.5–4.0 mmol/L) in 9.0% (74/826) and severe (blood glucose: < 2.5 mmol/L) in 2.3% (19/826). The case fatality ratio was 3.7% (29/784) overall but 26.3% (5/19) in severely hypoglycaemic children and 12.7% (9/71) in severely hypoxaemic children. Neither moderate hypoglycaemia nor moderate hypoxaemia was associated with mortality. CONCLUSION: Presumptive pre-referral glucose treatment and better management of hypoglycaemia could reduce the high case fatality ratio observed in children with severe hypoglycaemia. The morbidity and mortality burden of severe hypoxaemia was high; ways of improving hypoxaemia identification and management are needed.

Global initiatives to reduce child mortality have generally focused on improving early access to basic treatment for common illnesses using tools such as the World Health Organization's (WHO) Integrated Management of Childhood Illness strategy and integrated community case management. [1] [2] [3] In the absence of gold-standard diagnostic techniques for conditions such as pneumonia, these approaches rely primarily on subjective clinical assessment for syndromic case management. Children with signs of severe illness are referred to hospital for supportive care. However, emergency care is weak in under-resourced health systems, referrals can be difficult for caregivers, there may be delays due to a lack of transportation, and financial barriers are common. [4] [5] [6] Children may, therefore, arrive at referral hospitals when their illness is at a late stage and treatment may be less effective. In Malawi, 33% (39/118) of deaths reported among children at a tertiary referral hospital in 2017 occurred in the first 24 hours after admission. 7 Hypoxaemia and hypoglycaemia are objective measures associated with paediatric mortality. [7] [8] [9] [10] Indeed, Integrated Management of Childhood Illness protocols include hypoxaemia, defined as peripheral blood oxygen saturation (SpO 2 ) below 90%, as a referral criterion and the threshold for initiating oxygen treatment. General danger signs (Box 1) are inadequate for identifying hypoxaemia and children may not receive the oxygen they need. 12, 13 An SpO 2 level below 93% has been associated with deaths in children with clinical pneumonia in sub-Saharan Africa but evidence is lacking on the optimal referral threshold. 10, 14, 15 Similarly, Integrated Management of Childhood Illness protocols recognize the risk posed by hypoglycaemia to children and recommend presumptive pre-referral treatment for those with danger signs. Currently, WHO defines hypoglycaemia in a well-nourished child as a blood glucose concentration below 2.5 mmol/L 1 though increased mortality has been reported in children admitted with higher concentrations. 8, 9, 16, 17 A recent trial in Malawi found that hypoglycaemia treatment in hospitalized children with a blood glucose concentration between 2.5 and 5.0 mmol/L was not associated with survival. 18 More evidence is needed on how best to detect, monitor and treat hypoglycaemia in children.

Early detection of moderate hypoxaemia (SpO 2 between 90 and 93%) and moderate hypoglycaemia (blood glucose concentration between 2.5 and 4.0 mmol/L) in primary care may help reduce child mortality. The aim of our study was to investigate the survival of children referred from primary health-care facilities in Malawi, with a focus on clinical progression in those who presented with moderate hypoglycaemia or moderate hypoxaemia.

In this preplanned secondary analysis of a prospective cohort study, we assessed the survival of children referred from primary health-care facilities to hospitals in Mchinji district, Malawi. Mchinji had a population of approximately 600 000 in 2015 to 2016 and a mortality rate in children younger than 5 years of 123 per 1000 live births. 19 In particular, we followed children with severe or moderate hypoxaemia or hypoglycaemia (Box 1) from recruitment to presentation at another facility ( Fig. 1 ). Children were recruited at all 14 functional, government, primary health-care facilities in Mchinji district that provided outpatient paediatric care: two dispensaries, 11 health centres and one rural hospital with no inpatient care. Three rural hospitals and a district hospital acted as referral facilities. The nearest tertiary referral hospital was in the neighbouring district of Lilongwe (no data were collected from this facility). To be eligible for inclusion in the study, referred children had to be aged between 0 months and 12 years and be resident in Mchinji district. Recruitment started on 1 July 2019 and was intended to last 12 months. However, enrolment was terminated early, on 6 April 2020, because of the coronavirus disease 2019 (COVID- 19) pandemic. Follow-ups were completed on 13 June 2020.

We employed 20 non-clinical data collectors resident in Mchinji district. They underwent one week's residential training in study procedures and blood glucose and SpO 2 measurement. At the end of the training, data collectors were individually assessed on their interpretation of different clinical scenarios to verify their understanding of good-quality SpO 2 and blood glucose measurements. Clinical staff at all study facilities attended a 2-day refresher training course provided by the district health management team that covered Integrated Management of Childhood Illness protocols. Attendees' knowledge before and after training was not formally assessed.

Study participants were recruited during standard operating hours (i.e. 08:00 to 15:00, Monday to Friday) from the primary health-care facilities; emergency cases seen outside these times may have been missed. Children were assessed routinely by facility staff who alerted data collectors when a child was referred to another facility (Fig. 1) . After obtaining informed consent from caregivers, data collectors measured SpO 2 with a Lifebox pulse oximeter (Lifebox Foundation, London, England) using the big toe (or finger in children older than 2 years). Then the blood glucose concentration was measured in a capillary sample using Accu-Chek Aviva (Roche, Basel, Switzerland). If severe hypoglycaemia or hypoxaemia was detected, the health-care provider was alerted. Other clinical data were extracted from the child's health passport or the caregiver's report and contact details, sociodemographic information and details of care-seeking and treatment were obtained using a brief questionnaire. Enrolment was kept brief and study staff were instructed not to interfere with caregiver decisionmaking. A unique study barcode sticker was placed inside each recruited child's health passport.

Outpatients and inpatients at the four referral hospitals and the 14 primary health-care facilities were screened daily by data collectors to document the onward care of recruited children. Both SpO 2 and blood glucose were measured again by study staff at these locations when a recruited child was identified (Fig. 1 ). Children were followed up by phone or household visit 2 weeks after hospital discharge or their last confirmed outpatient visit to confirm survival and obtain details of any additional formal or informal care-seeking. b We did not extract information on movement as a danger sign and severe chest indrawing was not disaggregated from chest indrawing. c As 5-12 year olds are not included in Integrated Management of Childhood Illness protocols, we used the same danger signs as 2-59-month-olds.

Survival of child referrals, Malawi Carina King et al.

A new illness episode was registered if a child presented after the 2-week follow-up period had been completed and, therefore, it was possible for an individual child to be recruited more than once. Verbal autopsies were conducted for children who died using WHO's 2016 verbal autopsy instrument. 20 However, because of COVID-19, verbal autopsies were completed for only eight of the 29 children who died. Data were entered and uploaded daily onto tablet computers using CommCare software (Dimagi Inc., Cambridge, United States of America) and the full list of currently recruited children was visible to all data collectors. Data collectors were supervised by the project manager (a clinical officer), the data manager, and study monitoring and evaluation staff. Problems with implementation were dealt with during frequent supervision visits and monthly group meetings. Ethical approval was obtained from the Research and Ethics Committee at the University of Malawi's College of Medicine (P.11/18/2538). Caregivers provided informed verbal consent at recruitment and each subsequent interaction.

We recorded hypoxaemia and hypoglycaemia severity at recruitment and compared changes between recruitment and subsequent clinical visits using paired t-tests for means and Wilcoxon signed-rank tests for medians. The primary study outcome was the case fatality ratio and the primary exposures of interest were the SpO 2 level and the blood glucose concentration at recruitment (Box 1). The case fatality ratio was calculated as the number of deaths occurring between recruitment (day 0) and 14 days after hospital discharge or the last confirmed clinical visit divided by the number of children who completed follow-up. Associations with the case fatality ratio were estimated using multivariable Cox proportional hazards models, adjusted for recruitment facility clusters. The survival time was censored at death or 2 weeks after hospital discharge or the last confirmed clinical visit. For children who died on the day of recruitment, the survival time was taken to be 0.5 days. Missing SpO 2 values at recruitment were included as a distinct category because of previous evidence of an association with mortality. 12 We adopted the same approach for missing glucose values. Models were adjusted for the presence of general danger signs (Box 1), severe underweight, age, sex and hospital admission.

We were unable to adjust for oxygen or dextrose treatment using multivariable adjustment, interaction terms, stratification or propensity score matching because of confounding by indication (i.e. the most severely ill children were more likely to receive oxygen but were less likely to survive). 21, 22 However, we conducted exploratory analyses using hospital admission as a proxy for treatment: we performed a stratified analysis by hospital admission and estimated the treatment effect of admission using propensity score matching. 23 All analyses were performed using Stata v. 14 (StataCorp LLC, College Station, USA).

In total, 834 episodes of child illness were screened and 826 children were recruited, of whom 784 (94.9%) completed follow-up ( Fig. 2 Routine assessment by health-care provider. Health-care provider decides about referral. SpO 2 and blood glucose measured by study sta . SpO 2 and blood glucose measured by study sta . Routine assessment by health-care provider.

Oxygen and dextrose treatment of admitted children checked daily by study sta until discharge or death. Study sta con rm survival 14 days after hospital discharge or last clinic visit. follow-ups involved household visits (71.9%; 564/784) and caregivers were contacted on average 1.7 times (range: 1-12). The median follow-up time was 14 days (range: 0-77). The children's median age was 36 months (interquartile range, IQR: 16-73) and more boys were recruited than girls: 52.9% (437/826) versus 47.1% (389/826), respectively.

We recorded 29 deaths within 2 weeks of hospital discharge or the last confirmed clinical visit, which gave a case fatality ratio of 3.7% (29/784). We recruited 13 children more than once, of whom two died (15.4%). The case fatality ratio was highest in infants younger than 2 months (15.8%; 6/38) and lowest in children aged 5 to 12 years (2.3%; 6/258; P-value: < 0.001). No significant difference in case fatality ratio was observed by sex (P-value: 0.850). The median time from recruitment to death was 1 day (IQR: 0-4) and 44.8% (13/29) of deaths occurred within 24 hours. Seven of the 13 children who died on the day of recruitment were not admitted (available in data repository). 24 Following recruitment, 37.1% (306/826) of children were admitted to hospital and 41.7% (344/826) attended another facility.

Overall, 8.6% (71/826) of children were severely hypoxaemic at recruitment and 13.1% (108/826) were moderately hypoxaemic (Table 1) . Severe hypoxaemia was significantly more frequent in children younger than 2 months (28.9%; 11/38) than in those aged 5 to 12 years (4.1%; 11/270; P-value: < 0.001). The case fatality ratio among children who completed follow-up was 13.9% (9/65) in those with severe hypoxaemia, 3.9% (4/104) in those with moderate hypoxaemia and 2.3% (14/605) in those with normoxaemia. Among all severely hypoxaemic children ( (Table 3 ; available from: https:// www .who .int/ publications/ journals/ bulletin/ ). The median SpO 2 increased after recruitment in both those with severe hypoxaemia (from 84% to 92%; P-value: < 0.001) and moderate hypoxaemia (from 92% to 95%; P-value: 0.006).

Of the 292 children who were admitted to hospital and completed follow-up, 34 (11.6%) were severely hypoxaemic on arrival (Fig. 3) ; 28 of the 34 (82.4%) received oxygen. Of the 49 children with moderate hypoxaemia who were admitted, 12 (24.5%) had progressed to severe hypoxaemia and one of the 12 died (case fatality ratio: 8.3%) -this child did not receive oxygen. Of the 42 children with severe hypoxaemia at recruitment who were admitted, 15 (35.7%) remained severely hypoxaemic on arrival at hospital and five of the 15 died (case fatality ratio: 33.3%). The case fatality ratio for children with an SpO 2 below 90% at hospital admission was 20.6% (7/34), which was similar to the ratio for severely hypoxaemic children who were not admitted (17.4%; 4/23).

Overall, 2.3% (19/826) of children were severely hypoglycaemic at recruitment and 9.0% (74/826) were moderately hypoglycaemic ( Table 1 ). The case fatality ratio among children who completed follow-up was 27.8% (5/18) in those with severe hypoglycaemia, 5.6% (4/71) in those with moderate hypoglycaemia and 2.9% (20/687) in those with normoglycaemia. Of the 19 severely hypoglycaemic children overall (Table 4) , 15 (79.0%) presented with a danger sign and the most frequent diagnoses were malaria (52.6%; 10/19), malnutrition (31.6%; 6/19) and sepsis or meningitis (21.1%; 4/19). Severe hypoglycaemia was more frequent in girls than boys: 3.1% (12/389) versus 1.6% (7/437), respectively (P-value: 0.160).

Presentation at another facility after recruitment ( Although there was no difference in the mean blood glucose concentration between recruitment and arrival at another facility overall, the mean was significantly higher on subsequent measure-ment for both severely and moderately hypoglycaemic children (Table 5) . No severely or moderately hypoglycaemic child was given pre-referral glucose treatment at recruitment.

Of the 292 children admitted to hospital, six (2.1%) had severe hypoglycaemia at admission; four of the six (66.7%) received dextrose treatment (Fig. 4) . Of the 26 children with moderate hypoglycaemia at recruitment who were subsequently admitted, two (7.7%) had severe hypoglycaemia at admission and 13 (50.0%) had a normal glucose level. The case fatality ratio was similar among children who had moderate or severe hypoglycaemia at admission: 17.7% (3/17) versus 16.7% (1/6), respectively. However, the ratio was 42.9% (3/7) among severely hypoglycaemic children who were not admitted.

The results of the adjusted Cox proportional hazards model for survival are presented in Table 6 . Both severe hypoxaemia (adjusted hazard ratio, aHR, compared with normoxaemia: 4.05; 95% confidence interval, CI: 1.65 to 9.94) and severe hypoglycaemia (aHR compared with normoglycaemia: 7.60; 95% CI: 2.07 to 27.92) at recruitment were independently associated with death. There was no significant association with either moderate hypoxaemia or moderate hypoglycaemia.

In the analysis in which children were stratified by hospital admission, admission appeared to decrease the hazard of death for both those with severe hypoxaemia and those with severe hypoglycaemia. Among children with severe hypoxaemia, the aHR for death compared with normoxaemia was 9.14 in those who were not admitted versus 2.34 in those who were. Among children with severe hypoglycaemia, the aHR for death compared with normoglycaemia was 15.74 in those who were not admitted versus 4.12 in those who were. However, the CIs for these hazard ratios were wide (available in the data repository). 24 Overall, the treatment effect of hospital admission was estimated to be a 1.39% (95% CI: −6.81 to 4.02) reduction in the case fatality ratio among those admitted (data repository). 24 The estimated effect was larger for children with moderate or severe hypoxaemia but was not significant. 

We found that severe hypoxaemia and severe hypoglycaemia were significantly associated with death among children referred from primary health-care facilities to hospitals in Malawi. Although neither moderate hypoxaemia nor moderate hypoglycaemia was significantly associated with increased mortality, our exploratory analyses suggested that hospital admission may decrease the risk. Only 37% (292/784) of children in the study were admitted to hospital and, overall, more than 95% of those with moderate hypoxaemia or hypoglycaemia who were not admitted survived, irrespective of the presence of danger signs. However, over a quarter of referred children with severe hypoglycaemia died and these deaths predominantly occurred within 24 hours, which suggests that the severity of the disease may have been recognized late and care-seeking may have been delayed, as has been observed in previous studies in Malawi. 5, 26 The Integrated Management of Childhood Illness protocol recommends presumptive hypoglycaemia treatment before referral. 1 In 2019, the majority of facilities in Mchinji had a glucometer and a stock of dextrose. 27 Although it would be unreasonable to expect that blood glucose measurements would be carried out routinely at our study facilities as they are not included in the protocol, the fact that no hypoglycaemic child reportedly received presumptive glucose treatment points to a gap in the protocol's implementation. Nevertheless, we observed an increase in the mean glucose concentration after recruitment in both moderately and severely hypoglycaemic children, which suggests that health-care providers may have given caregivers advice on feeding. Alternatively, many of the most acutely hypoglycaemic children may have died before hospital admission. Our findings support the use of presumptive glucose treatment. However, greater efforts must be made to ensure this happens, along with subsequent glucose monitoring and management. 28 In the Integrated Management of Childhood Illness algorithm for respiratory infections, an SpO 2 below 90% is an indication for referral. 1 In agreement with previous reports, 29 we found that hypoxaemia was relatively common, even in the absence of pneumonia. Although functional oximeters were reportedly available in 29.8% (14/47) of sampled facilities in Malawi, 30 healthcare workers often made referral decisions without using pulse oximetry. We found that 28.9% of infants younger than 2 months were severely hypoxaemic, similar to the 22.6% (53/235) reported in a previous study from Malawi. 31 The quality of oximetry measurements in these infants can be poor due to badly fitting probes, non-cooperation or perfusion issues. However, reported diagnoses were consistent with conditions where hypoxaemia was expected (e.g. congenital heart disease, asphyxia, apnoea, pneumonia and sepsis). Given the role of pulse oximetry in detecting congenital heart disease in neonates, 32 which is often asymptomatic, neonatal SpO 2 measurements must be feasible and reliable. More broadly, we observed that respiratory rates were rarely documented, which corresponds with previous findings that respiratory examinations are often poorly conducted in Malawi. [33] [34] [35] [36] There is, therefore, a need to improve pneumonia diagnosis and management.

We found that both moderate hypoxaemia and moderate hypoglycaemia at recruitment were associated with a non-significant increase in the hazard of death among children, which contradicts previous hospital-based studies. 7, 8, 14, 31 Moreover, our exploratory analysis, though it lacked statistical power, suggested that hospitalization may have reduced mortality in these patient groups. Strikingly, 11 of the 12 children with moderate hypoxaemia who progressed to severe hypoxaemia by hospital admission survived -the child who died did not receive oxygen. In contrast, a third of children admitted with persistently severe hypoxaemia died, even though most received oxygen. These findings suggest that earlier identification and prompt care-seeking could reduce mortality. 37 However, given confounding by indication (i.e. the most severely ill children are more likely to receive oxygen but also to die) and potential survivorship bias (i.e. children have to survive long enough to reach hospital), well-designed trials are needed to provide evidence for guideline reviews.

The influence of dextrose treatment on survival was less clear. Although moderate hypoglycaemia at recruitment was not a significant risk factor for death, the case fatality ratio in children with moderate hypoglycaemia at hospital admission was 17.7%, higher than for any other admission hypoglycaemia category. The recent SugarFACT trial in Malawi failed to show that treatment improved survival in children with hypoglycaemia, 18 which reinforces the need for better understanding of the management of these patients. Our observation that blood glucose and SpO 2 categories changed between recruitment and hospitalization in most children raises the important question of whether serial measurements are preferable to one-off spot checks for case management and for identifying the need for urgent care and outpatient monitoring. 38 Although few infants younger than 2 months were recruited, they had the highest case fatality ratio of all age groups. We were surprised to find that 32.7% (270/826) of children recruited were aged 5 to 12 years and that their case fatality ratio was comparable to that of children aged 12 to 59 months: 2.3% (6/258) versus 3.1% (12/390), respectively. This older age group is overlooked, being neither explicitly included in an Integrated Management of Childhood Illness chart booklet nor targeted by sustainable development goals. 39 Moreover, measurement of SpO 2 and blood glucose levels do not appear to be informative for this age group and more research is warranted.

Our study had three key limitations. First, because of the COVID-19 pandemic, we stopped recruitment before the planned closure date and verbal autopsies were not completed for all deaths. During follow-ups and verbal autopsies, we asked about care-seeking to validate data collection at facilities. Given that the response rate varied by survival status, we chose not to use these data and it is possible, therefore, that we were not able to confirm all instances of onward care. To minimize the possibility that children admitted out of hours were missed, hospital-based data collectors reviewed patient charts each morning. Nevertheless, children who presented to primary-care facilities out of hours would have been missed, resulting in lower recruitment and the under-ascertainment of onward care. Second, we used non-clinical data collectors and it is plausible that some of the variation in hypoxaemia category between recruitment and subsequent care resulted from measurement quality issues as oximetry in young infants requires skill. Finally, we relied on routine clinical assessment and decision-making by health-care workers for deciding on study eligibility and it is possible that some hypoxaemic children who should have been referred were missed. We were unable to validate key clinical variables and problems with routine data quality were appar-ent (e.g. the absence of respiratory rate data) despite Integrated Management of Childhood Illness refresher training.

Mortality among children with severe hypoxaemia or hypoglycaemia who were referred from primary care in Malawi was high. For hypoglycaemia, our findings support current recommendations for presumptive glucose treatment but further research is needed to determine the optimal threshold for treatment and the best management for this group. For hypoxaemia, timely care-seeking, routine pulse oximetry, and earlier identification and referral of severely hypoxaemic children could reduce the risk of death. However, given that most referred children in our study were not subsequently admitted to hospital but survived, greater understanding of how best to manage moderately hypoxaemic children is needed. Optimal management must take into account the burden placed by referral on the health system and on patients as well as the clinical benefits of treatment. ■ 

Проспективное когортное исследование малавийских детей, направленных на лечение, и их выживаемость при гипоксемии и гипогликемии Цель Изучить выживаемость детей, направленных из учреждений первичной медико-санитарной помощи в Малави, с упором на гипогликемию и прогрессирование гипоксемии. Методы В исследование была включена проспективная когорта детей в возрасте до 12 лет включительно, направленных из учреждений первичной медико-санитарной помощи в районе Мчинджи, Малави, в 2019 и 2020 годах. Насыщение кислородом периферической крови (SpO 2 ) и уровень глюкозы в крови измерялись во время набора в исследование и по прибытии в соответствующее медицинское учреждение (четыре больницы и 14 учреждений первичной медико-санитарной помощи). Последующее наблюдение проводили через 2 недели после выписки или последнего визита ребенка в клинику. Первичным результатом исследования был показатель летальности через 2 недели. Взаимосвязь между SpO 2 , уровнями глюкозы в крови и смертностью оценивалась с использованием модели пропорциональных рисков Кокса. Терапевтический эффект госпитализации оценивался с использованием псевдорандомизации.

Результаты Из 826 детей, включенных в исследование, 784 (94,9%) прошли последующее наблюдение. На момент первичного осмотра гипоксемия была умеренной степени (SpO 2 : 90-93%) у 13,1% детей (108/826) и тяжелой степени (SpO 2 : < 90%) у 8,6% детей (71/826). Гипогликемия была умеренной степени (глюкоза в крови: 2,5-4,0 ммоль/л) у 9,0% детей (74/826) и тяжелой степени (глюкоза в крови: < 2,5 ммоль/л) у 2,3% детей (19/826). Показатель летальности составил 3,7% (29/784) в целом, но 26,3% (5/19) у детей с тяжелой гипогликемией и 12,7% (9/71) у детей с тяжелой гипоксемией. Ни умеренная гипогликемия, ни умеренная гипоксемия не были связаны со смертностью. Вывод Пробное догоспитальное лечение глюкозой и более эффективное лечение гипогликемии могут снизить высокий показатель летальности, наблюдаемый у детей с тяжелой гипогликемией. Бремя заболеваемости и смертности от тяжелой гипоксемии было высоким, поэтому необходимы способы совершенствования механизмов выявления и лечения гипоксемии.

Objetivo Estudiar la supervivencia de los niños remitidos desde la atención primaria en Malawi, con un enfoque en la evolución de la hipoglucemia y la hipoxemia. Métodos El estudio incluyó una cohorte prospectiva de niños de 12 años o menos remitidos desde centros de atención primaria de salud ubicados en el distrito de Mchinji, Malawi, en 2019 y 2020. La saturación de oxígeno en sangre periférica (SpO 2 ) y la glucemia se midieron en el momento de selección y a la llegada a un centro sanitario posterior (es decir, cuatro hospitales y 14 centros de atención primaria). Se realizó un seguimiento de los niños dos semanas después del alta o de su última visita médica. El resultado principal del estudio fue la tasa de mortalidad a las dos semanas. Las asociaciones entre los niveles de SpO 2 y de glucemia y la muerte se evaluaron mediante modelos de riesgo proporcional de Cox, y el efecto del tratamiento de la hospitalización se evaluó mediante el emparejamiento del índice de propensión.

Resultados De los 826 niños seleccionados, 784 (94,9 %) completaron el seguimiento. Cuando se presentaron, la hipoxemia era moderada (SpO 2 : 90-93 %) en el 13,1 % (108/826) y grave (SpO 2 : <90 %) en el 8,6 % (71/826), mientras que la hipoglucemia era moderada (glucemia: 2,5-4,0 mmol/l) en el 9,0 % (74/826) y grave (glucemia: <2,5 mmol/l) en el 2,3 % (19/826). La tasa de letalidad fue del 3,7 % (29/784) en general, pero del 26,3 % (5/19) en los niños con hipoglucemia grave y del 12,7 % (9/71) en los niños con hipoxemia grave. Ni la hipoglucemia moderada ni la hipoxemia moderada se asociaron a la mortalidad. Conclusión El tratamiento presunto de la glucosa antes de la remisión y el mejor tratamiento de la hipoglucemia podrían reducir la elevada tasa de letalidad que se observa en los niños con hipoglucemia grave. La carga de morbilidad y mortalidad de la hipoxemia grave fue elevada; en consecuencia, es preciso mejorar la identificación y el tratamiento de la hipoxemia. (6); (ii) three children unconscious and receiving care; and (iii) a biologically plausible value could not be obtained for one child. d Reasons for missing data were: (i) no test strips available (one child); (ii) no lancet available (two children); (iii) glucometer not working (four children); and (iv) transport for referral was found before the test could be completed (one child). e Diagnosis made by the health-care provider at recruitment. f Children could receive more than one diagnosis.

teams/ maternal -newborn -child -adolescent -health -and -ageing/ child -health/ integrated -management -of -childhood -illness

Geneva & New York: World Health Organization & United Nations Children's Fund

Integrated Management of Childhood Illness: management of the sick young infant aged up to 2 months: IMCI chart booklet. Geneva: World Health Organization

Quality of hospital care for seriously ill children in less-developed countries

Care-seeking patterns amongst suspected paediatric pneumonia deaths in rural Malawi. Gates Open Res

Coping with paediatric referral -Ugandan parents' experience. Lancet

Risk factors for mortality in severely ill children admitted to a tertiary referral hospital in Malawi

Blood glucose as a predictor of mortality in children admitted to the hospital with febrile illness in Tanzania

Mini-review: management of hypoglycaemia in children aged 0-59 months

Hypoxaemia as a mortality risk factor in acute lower respiratory infections in children in low and middle-income countries: systematic review and meta-analysis

WHO child growth standards: length/height-for-age, weight-for-age, weightfor-length, weight-for-height and body mass index-for-age: methods and development. Geneva: World Health Organization

Predictive value of pulse oximetry for mortality in infants and children presenting to primary care with clinical pneumonia in rural Malawi: a data linkage study

Pulse oximetry for children with pneumonia treated as outpatients in rural Malawi. Bull World Health Organ

Predicting hospitalised paediatric pneumonia mortality risk: an external validation of RISC and mRISC, and local tool development

The impact of pulse oximetry on diagnosis, management and outcomes of acute febrile illness in low-income and middle-income countries: a systematic review

Abnormal blood glucose concentrations on admission to a rural Kenyan district hospital: prevalence and outcome

Relationship between glycemic levels and treatment outcome among critically ill children admitted into emergency room in Enugu

Effect on mortality of increasing the cutoff blood glucose concentration for initiating hypoglycaemia treatment in severely sick children aged 1 month to 5 years in Malawi (SugarFACT): a pragmatic, randomised controlled trial

Malawi Demographic and Health Survey 2015-16. Zomba & Rockville: National Statistical Office of Malawi & ICF

Geneva: World Health Organization

When do confounding by indication and inadequate risk adjustment bias critical care studies? A simulation study. Crit Care

Demystifying propensity scores

Estimation of average treatment effects based on propensity scores

Supplementary Appendix file for the manuscript Progression of hypoxaemia and hypoglycaemia in children referred from primary care facilities in Malawia prospective cohort study

Determinants of healthcare seeking for childhood illnesses among caregivers of under-five children in urban slums in Malawi: a population-based cross-sectional study

Paediatric Emergency Triage, Assessment and Treatment (ETAT) -preparedness for implementation at primary care facilities in Malawi. Glob Health Action

Reducing mortality from severe malaria in Sierra Leonean children by applying the World Health Organization's standard malarial protocol with additional sublingual glucose: a continuous quality improvement report

Diagnosis of pneumonia and malaria in Nigerian hospitals: a prospective cohort study

Integrated Management of Childhood Illnesses (IMCI): a mixed-methods study on implementation, knowledge and resource availability in Malawi

The epidemiology of hypoxemic pneumonia among young infants in Malawi

Pulse oximetry screening for critical congenital heart defects in asymptomatic newborn babies: a systematic review and meta-analysis

Determining the quality of IMCI pneumonia care in Malawian children

Paediatr Int Child Health

Distribution and determinants of pneumonia diagnosis using Integrated Management of Childhood Illness guidelines: a nationally representative study in Malawi

Quality of case management for pneumonia and diarrhea among children seen at health facilities in southern Malawi

Determinants of Integrated Management of Childhood Illness (IMCI) non-severe pneumonia classification and care in Malawi health facilities: analysis of a national facility census

Recognising and treatment seeking for acute bacterial meningitis in adults and children in resource-poor settings: a qualitative study

Anticipating the future: prognostic tools as a complementary strategy to improve care for patients with febrile illnesses in resource-limited settings

United Nations

We thank all participating children, caregivers and health-care workers, the EREMISS study team (i.e. data collectors, monitoring and evaluation officers and administrative staff at the Parent and Child Health Initiative), the district health officer and district health management team in Mchinji district, and Henrike Habel

The study was funded by grants from the Swedish Research Council (2017-05579), the Laerdal Foundation (40348) and the Einhorn Family Foundation.Competing interests: CK and EDM are independent advisors on pulse oximetry to the Lifebox Foundation. Other authors declare no conflict of interest. The hazard of death between study recruitment and 14 days after hospital discharge or the last clinical visit. c The proportional hazards assumption was tested using Schoenfeld residuals and was found not to be violated (P-value: 0.201). d Definitions of normoxaemia, hypoxaemia, normoglycaemia and hypoglycaemia are given in Box 1. e As all eight children with missing data survived, they were dropped from the model because of perfect prediction. f Danger signs are described in Box 1.