Care-seeking patterns amongst suspected paediatric pneumonia deaths in rural Malawi

Setting

At the time of data collection Mchinji district had an approximate population of 450,000, with 85% living as rural subsistence farmers. The under-five mortality rate was 63/1000 livebirths in the 2015–2016 Malawi Demographic and Health Survey. Healthcare was provided for free by 354 community healthcare workers (known locally as Health Surveillance Assistants), 11 government primary healthcare centres, and one referral district hospital, and for a small fee in four rural hospitals.

Data collection

Full details of the community surveillance system have been published previously¹⁴. Briefly, deaths were recorded and reported monthly by 1059 village-level key informants. Deaths were also identified at household visits conducted for all children born in the district at four months and one year of age. Data were submitted to the office for cleaning. A list of all community reported deaths amongst 0–59 months olds in Mchinji district was generated monthly for VA interviews. VAs were conducted by nine senior fieldworkers, all with previous experience of conducting VAs. They received one week’s training, including: translating the data collection tool, conducting mock interviews and using smart phones for data collection.

VAs interviews were randomised to conduct an open narrative or not, as part of a separate research question on the role of narratives in data quality, rapport and interview procedures¹⁵. This was done at the point of interview using a random number generator within the Open Data Kit (ODK) Collect application, version 1.4, used to collect data¹⁶. The open narrative was unstructured, conducted at the start of the VA and was either audio recorded then transcribed or captured using paper notes. Data collectors could record the information in the format they preferred, in English or Chichewa. Open narratives in Chichewa were later translated by the data collectors and data entry clerks. We used the standardized World Health Organization (WHO) 2012 VA questionnaire¹⁷.

Pneumonia death definition

Suspected pneumonia deaths were identified in three ways: 1. the caregiver responded yes to both the child having a cough and fast breathing in the two weeks prior to death, according to the WHO 2012 VA questionnaire; 2. the caregiver explicitly stated that the child had died of pneumonia during the VA; or 3. the cause of death was classified as ‘acute respiratory infection’ by InterVA-4. We excluded deaths occurring in the neonatal period (aged 0–28 days).

Quantitative analysis

We extracted data from the open narratives using a custom-designed tool in Excel, based on the ‘Pathways to Survival’ framework (Table 1)^13,18. We did not consider indicators relating to wellness inside the home (e.g. breastfeeding, hygiene) or outside the home (e.g. imminisation, water and sanitation), or indicators of quality of care inside the home. The tool was developed by TC and CK using three randomly selected narratives to define fields and categories. Data were then double coded from a random set of ten narratives. Disagreements between coding were discussed and the data extraction tool updated; a further 10 narratives were double coded to check for consistency. The remaining narratives were coded by CK. Data were described using proportions and means, and compared with chi² and t-tests. We stratified patterns of care-seeking by sex and age groups.

Qualitative analysis

Narratives were analysed using a framework approach¹⁹. Themes were pre-defined, based on the Pathway to Survival framework¹⁸, as: wellness, illness recognition, care-seeking decisions, referral procedures, and quality of care. Concepts were inductively coded within these themes by CK, and a random sub-set of ten narratives were double coded by TC. The interpretation of the data was discussed with two of the data collectors who conducted VAs to check for context-appropriate interpretation. Disagreements in interpretation were discussed until consensus was reached.

Ethics

The data used were collected as part of the VacSurv Study, which was approved by the National Health Sciences Research Ethics Committee in Malawi (reference: 837), London School of Hygiene and Tropical Medicine (reference: 6047) and Centers for Disease Control and Prevention (reference: 6268). Verbal informed consent was taken from all respondents, and confirmed in the electronic data form; written consent was not sought due to literacy rates.

Patient and Public Involvement

Before starting any data collection, the study protocol was presented to the District Executive Committee and District Health Management teams in Mchinji for input and approval. Extensive community engagement was conducted through village development committees and traditional leaders. Community consent was sought from traditional authorities during study introduction, who also assisted in informing communities of verbal autopsies.

Cohort description

During the cohort period, 4,855 death events were reported (Figure 1). Of the 1673 confirmed post-neonatal under-five deaths, 395 (24%) were classified as suspected pneumonia, representing a post-neonatal pneumonia mortality rate of 8.2 per 1000 livebirths. A total of 178 (45%) were randomised to open narrative, with balance between the two arms in terms of key demographics and care-seeking proxies (Table 2). Overall 171 suspected pneumonia deaths with complete data were included in analysis. Of these 37% (n=64/171) were completed in Chichewa and translated to English. The majority of deaths were classified as suspected pneumonia by InterVA (153/171, 89%).

Figure 1. Participant inclusion flow-chart.

Quantitative analysis

Overall, 86% (n=146/171) of children were taken to a public or private healthcare facility during their final illness episode. 44% (n=76/171) of caregivers sought care more than once in the final illness episode, with the most reported attendances being six over a 35-day period, visiting a health centre twice and hospital four times (Table 3). A sole visit to hospital occurred in (n = 50/171) 29%, and a single visit to a health centre in (n=17/171) 10%. Care-seeking at pharmacy shops was not explicitly reported. Of those where information on time to care-seeking could be extracted, 44% (n=36/81) of caregivers sought care within 24 hours of illness recognition. The median time from illness recognition to death was three days (IQR: 1 - 5). In 48 narratives (28%), the caregiver stated the child has a pre-existing condition or risk factor, such a birth defect or being born prematurely.

A total of 119/171 (68%) children ever visited a hospital, among whom 70% (n=83/119) were admitted. Amongst those admitted, caregivers specifically reported the child was given oxygen in 25% (n=21/83) of cases. Oxygen was more frequently reported amongst infants aged 1–11 months than those aged 12–59 months (32% vs 11%, p-value = 0.058), with no difference observed by sex (26% male vs 25% female). Pneumonia symptoms were recognised by caregivers less frequently in older children, but they were taken to hospital more frequently. Female children were taken to hospital more often than male children (74% vs 64%), but admitted less often than male children (68% vs 75%) – neither difference was statistically significant (Table 4).

Half of the children died within a healthcare setting (43% hospital, 5% health centre and 2% private clinics). Of the 64 (37%) children who died at home, 42% (n=27/64) had visited a hospital and 17% (n=11/64) had been admitted; we were unable to extract time from discharge to death. The remaining 13% of children died in transit, either from their home to seek care, or during referral from primary to secondary care (Figure 2). The majority of children aged 12–59 months died within a healthcare facility, while infants more often died at home (p-value = 0.028, Table 4).

Figure 2. Summary of care-seeking according to the Pathways to Survival framework, for first, second and third location.

Qualitative analysis

Wellness. Wellness includes prevent and protect indicators within households, with the sub-themes of: breastfeeding as a barometer of health; the role of a normal delivery; and sickly children.

Breastfeeding was frequently used to express the overall health of the child. A child who was able to breastfeed, and a caregiver who was able to breastfeed their child from birth, were presented as being healthy. While the converse was presented both as a cause of illness in the child or a sign of severe illness.

"the baby was ok, she was able to suck breastmilk" (2 months, female)

“He was not breastfeeding properly, because [I was] not producing milk. Then the child died” (13 months, male)

Like the mother’s ability to breastfeed, the health of the child was linked to the mother having a ‘normal’ pregnancy and delivery. Narratives often started with the woman’s pregnancy, highlighting whether they had attended antenatal care or had an uncomplicated delivery. In several narratives, the respondents indicated that the child had been born prematurely.

In a sub-group of narratives, caregivers described children who were repeatedly sick from recurrent and concurrent episodes of pneumonia, malaria, diarrhoea, and anaemia. In these cases, caregivers reported multiple episodes of care-seeking and described other comorbidities as explanations for their recurrent infections (e.g. prematurity, asthma, and congenital malformations). Misconceptions of asthma were apparent.

“This baby has been ill on and off since she was born. We were not surprised because the baby was born with asthma, and another problem was that the baby was born with a large chest.” (10 months, female.)

Conversely, there were narratives where respondents expressed surprise that their child was ill, referring to them having been vaccinated and growing strongly.

Illness recognition. Illness recognition includes: pneumonia-specific signs and symptoms; overlap between pneumonia and asthma, and signs of severe illness.

Caregivers reported a range of clinical presentations for pneumonia, including; cough, difficulty breathing, noisy breathing, chest indrawing and fast breathing - in line with the IMCI chart booklet for diagnosing pneumonia. Caregivers differentiated between the severity of these different signs, with a cough often reported as the first sign of illness, which wasn’t necessarily treated, followed by the more serious sign of difficulty breathing. Fever was frequently reported, and was almost exclusively associated with malaria.

"after some hours had passed the coughing started, and after a day or two passed the baby started having difficulties in breathing, so we went to the hospital." (5 months, female.)

Several caregivers stated that their children were ‘born with asthma’ and that asthma triggered them to seek care, or their child died of a sudden asthma attack. However, it was often unclear from the narratives whether the child had asthma, pneumonia, or both. It should be noted that in the Chichewa transcripts, distinct terms are used for pneumonia (zibayo) and asthma (mphumo), and field staff reflected that they were not used interchangeably.

"the child developed the problem of suffering from asthma - he had been admitted several times because of these attacks and during those times he was given medications like amoxicillin, Bactrim [cotrimoxazole], prednisone, Panadol and salbutamol." (24 months, male.)

Danger signs that triggered the caregiver to consider the illness severe included fainting and convulsions. A commonly used term was ‘weakness’ of the child (e.g. “we just saw our child was weak”, 12 months, female), and this was often provided as the trigger to seek care. This, and “dizzy eyes” may reflect the WHO general danger sign of lethargy. A child’s failure to breastfeed was also understood as being a sign of them being severely ill, and led to several caregivers stating that their children had died of starvation.

“I believe that the child died of hunger because she was not feeding and coughing.” (10 months, female.)

Care-seeking decisions. Sub-themes within care-seeking decisions included: delaying seeking care until the illness was severe; and the plurality of care. In several cases the respondent stated that they recognised the child was unwell, but did not seek care immediately. Rather they waited until a more severe sign of illness was seen (e.g. weakness, failing to feed or difficulty breathing).

In Malawi, traditional and western medicine exist in parallel. This plurality in healthcare systems was apparent in several narratives with caregivers reporting attending both types of care for the same episode of illness.

"the baby kept on crying and could hardly be stopped. We thought it could be evil spirits… therefore traditional medicine was given but there was no change, so we decided to visit the health centre." (4 months, male.)

The sequential access of different types of care, was attributed to the failure of the care sought either to cure the child or a perception that the care was not sufficient, or on the advice of neighbours and relatives:

"the child was given some panadols […] we therefore resorted to traditional drugs." (6 months, female.)

Referral procedures. Referral procedures, covering both healthcare worker’s recommendation to seek further care and the process of getting to further care, included the sub-themes of: community advice; transportation barriers; and advice to wait and see. Advice to initially seek care and further care, was received from different community sources, namely relatives, traditional birth attendants and community healthcare workers. It should be noted respondents reported that community members recommended further care be sought both through formal and informal providers.

"when we reached home we were advised to proceed to [the hospital] and indeed we started off for the hospital." (10 months, female.)

Following community and healthcare provider referrals, challenges in getting to the facility due to transportation issues were reported. These barriers included lack of ambulance access, issues in finding the funds to arrange private transport and in the example given, the inappropriate use of an ambulance.

"the ambulance that carried us also took some nurses who had different errands, and we stopped at the market. By the time we reached [the hospital] the child was pronounced dead." (10 months, male.)

There were several stories of children taken to health centres, given treatment and sent home with the advice to return if the child did not improve –in line with WHO IMCI protocols for children without severe illness.

"the doctor said that if we see no signs of any change, we should go to bigger hospital… there was no change and indeed we went to [the hospital].” (16 months, female.)

Quality of care. Several quality of care challenges were reported during the open narratives, relating both to pneumonia specific issues, and wider health system challenges. Sub-themes include challenges in oxygen provision, barriers to blood provision, inappropriate discharge, missed opportunities to treat, and delays in care provision.

Oxygen was reported by many caregivers as being given at the time of the child’s death, with many children dying shortly after being given oxygen. There were three specific narratives where caregivers reported the power going out while receiving oxygen, resulting in the child’s death.

"He was put on oxygen for 50 minutes… shortly after the electricity at the hospital went out and the doctors removed the child from the oxygen machine. After 20 minutes the electricity was on again and the doctors came to take the child and put them on the oxygen machine. […] unfortunately they found that the child was already dead." (7 months, male.)

Many caregivers also reported that their child was prescribed a blood transfusion, but issues inserting intravenous lines and the lack of available blood meant they didn’t receive it. A major gap in quality of care reported from caregivers was that their child was discharged from hospital while they were still unwell:

"Although she was discharged the child was not okay…very shortly after, it’s when the child died here at home, the same day she was discharged from hospital." (8 months, female.)

Caregivers reported several situations of missed opportunities for diagnosis and treatment, including lacking medications and failing to diagnose the child with any illness during their examination. In these cases, children were sent home with no diagnosis, treatment or referral, leading to delays in care provision.

"we took the baby to [the clinic] and there was no medication at the facility. We went back home and the coughing continued… come the next day, the baby did not survive." (2 months, male.)

Within facilities, some reported delays in the child being seen by a healthcare provider, and in one such case the caregiver attributed her child’s death to this delay:

"I blame the negligence of medical personnel at the hospital for delaying admission, and the time they took to start assisting the baby." (8 months, male.)

Underlying data

Harvard Dataverse: Replication Data for: "Care-seeking patterns amongst suspected paediatric pneumonia deaths in rural Malawi: a mixed method study"

https://doi.org/10.7910/DVN/66YIME³⁴

This project contains the following underlying data:

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

The open narratives cannot be sufficiently anonymised, as they contain detailed accounts of individual death events with locations, dates, demographic and clinical details. These are therefore not available in an open repository. For researchers wishing to duplicate or re-analyse these data for data for research purposes only, contact Dr Carina King (carina.king@ki.se) or Prof. Neil French (N.French@liverpool.ac.uk), to discuss the signing of a data sharing agreement, following review by the National Health Sciences Research Ethics Committee in Malawi (contact:infor@ncst.mw).