Menu
Menu
Menu
Menu
Objective To compare coverage of key child health policy indicators across provinces and to explore their association with under-five mortality and level of conflict in the Democratic Republic of the Congo. Methods We made a secondary analysis of nationally representative data from 1380 health facilities and 20 792 households in 2017–2018. We analysed provincial-level data on coverage of 23 different indicators for improving common causes of childhood mortality, combined into mean scores for: newborn health, pneumonia, diarrhoea, malaria and safe environment. Using negative binomial regression we compared the scores with provincial-level under-five mortality. With binary logistic regression at the individual level we compared indicators (outcome) with living in a conflict-affected province (exposure). Findings All grouped coverage scores demonstrated large ranges across the 26 provinces: newborn health: 20% to 61%; pneumonia: 26% to 86%; diarrhoea: 25% to 63%; malaria: 22% to 53%; and safe environment: 4% to 53%. The diarrhoea score demonstrated the strongest association with under-five mortality (adjusted coefficient: −0.026; 95% confidence interval: −0.045 to −0.007). Conflict-affected provinces had both the highest as well as the lowest mortality rates and indicator coverages. The odds of coverage were higher in conflict-affected provinces for 13 out of 23 indicators, whereas in provinces unaffected by conflict only one indicator had higher odds of coverage. Conclusion Conflict alone is a poor predictor for child health. Ensuring that children in unaffected provinces are not neglected while addressing the needs of the most vulnerable in conflict settings is important. Prevent, protect and treat strategies for diarrhoeal disease could help improve equity in child survival.
We performed a secondary analysis of data from nationally representative, cross-sectional surveys of health facilities and households in the Democratic Republic of the Congo in 2017–2018. The framework for the study was based on a review of three global action plans to identify key policy indicators for action on common causes of childhood mortality, under the broad themes of prevent, protect and treat. The Democratic Republic of the Congo has an estimated population of 85–100 million14,15 residing across 26 provinces and 516 health zones.16 Health care is offered by public and private operators including faith-based organizations.16 In addition, several NGOs and international organizations operate in the country.17 An estimated 40% of the country’s health-care spending comes from out-of-pocket expenditure, with international donors providing a similar proportion.18 Ethical approval for the study was obtained from the Swedish Ethical Review Authority (Dnr 2020–05190). Data collection and sampling procedures for the data sets have been described elsewhere.5,19,20 We describe here some important details about the data sets; further details are in the supplementary files in the authors’ data repository.21 We obtained data on health indicators and socioeconomic status from two national data sets. The Service and Provision Assessment 2017–201819 used stratified random probability sampling to select 1412 health facilities from a list of all 12 050 operational health facilities, excluding health posts. These facilities were surveyed between October 2017 and April 2018. Of the sampled health facilities, 32 (2.3%) were not surveyed, mainly due to security problems. We extracted data from the inventory section of the data (for example, on medications and equipment), and from the service provider questionnaire (for example on receipt of training in kangaroo mother care). The Multiple Cluster Indicator Survey 2018 household survey5 was designed to provide provincial estimates based on individual-level data using a sample frame based on the 1984 population census. A systematic random sample of 30 households was drawn from each of the 721 clusters giving an overall sample of 21 630 households, of which 20 792 (96.1%) were successfully interviewed between December 2017 and July 2018. Twelve clusters were not visited due to insecurity problems, mainly in Tanganyika and Maniema provinces. We used data from the questionnaires about the household, women and children younger than 5 years. We extracted data on relative socioeconomic status (continuous variable) based on household asset ownership and urban or rural setting. To obtain data on areas of conflict in the Democratic Republic of the Congo we used a third data set. The Uppsala Conflict Data Program Georeferenced Event Data Set contains global temporally and spatially disaggregated data of conflict events.22–25 For an event to be included it must have resulted in at least one death and the actor involved must have been involved in events that together accumulated to at least 25 deaths in one calendar year. We calculated annual levels of conflict for each province between 2013–2018 to match the time frame used to calculate the under-five mortality. We divided provinces into three different conflict categories, adapting the definition from Uppsala University regarding state-based violence: major conflict (if more than 1000 battle-related deaths had occurred in one of the calendar years), minor conflict (more than 25 battle-related deaths) and no conflict (25 deaths or fewer).26 We compiled a list of 47 key policy indicators for action on common causes of childhood mortality from the following documents: (i) Every Newborn action plan;27,28 (ii) Global Action Plan for the Prevention and Control of Pneumonia and Diarrhoea;29 and (iii) Global Technical Strategy for Malaria 2016–2030.30 We reviewed the national health facility and household surveys for available data on coverage of the identified indicators. We used data on 23 different indicators: 10 of the 15 indicators in the Every Newborn action plan,27 11 of the 18 indicators from the Global Action Plan for the Prevention and Control of Pneumonia and Diarrhoea29 and three of the 15 Global Technical Strategy for Malaria 2016–2030 indicators30 (Table 1). We excluded indicators if no data were available, the intervention was not implemented at the time of the survey, the indicator was not focused on the child (maternal indicators, for example) or too few observations were recorded. Details about the excluded indicators are in the supplementary files.32 We set the target coverage at 80% for all indicators, except exclusive breastfeeding (50%) and caesarean section (10%), using the district-level targets set out by the Global Action Plan for the Prevention and Control of Pneumonia and Diarrhoea and the International Vaccine Access Centre.31 DTP: diphtheria–tetanus–pertussis; Hep B: hepatitis B; Hib: Haemophilus influenzae type B. a Authors’ classification. b Service and Provision Assessment Survey 2017–2018 does not include a question on mask size. c Study definition differs from action plan definition. d We only chose zinc, to be consistent with the international vaccine access centre definition.31 Note: Data sources were the Multiple Indicator Cluster Survey, 2017–20185 and Service and Provision Assessment 2017–2018.19 We calculated the indicators according to the definitions on Table 1; some indicators were identical to the source reports whereas other differed in definition and were not reported in the reports. We then combined data for the available indicators into six grouped coverage scores covering common causes of childhood mortality, using the same method as the International Vaccine Access Center:31 (i) newborn health (using indicators from the Every Newborn action plan); (ii) pneumonia; (iii) diarrhoea; (iv) combined pneumonia and diarrhoea (each from the Global Action Plan for the Prevention and Control of Pneumonia and Diarrhoea); (v) malaria (from the Global Technical Strategy for Malaria 2016–2030); and (vi) safe environment. We generated overall grouped scores by adding the coverage for all included indicators and dividing by the number of indicators in each group (Box 1). Numerator: exclusive breastfeeding for 6 months, skilled birth attendance, early postnatal care contact for infants, essential newborn care, newborn resuscitation, kangaroo mother care, treatment of severe neonatal infection, chlorhexidine cord-cleansing, caesarean section, emergency obstetric care Denominator: number of indicators (10) Numerator: exclusive breastfeeding for 6 months, pentavalent vaccine coverage, measles vaccine coverage, pneumococcal vaccine coverage, oral rehydration therapy, zinc for the treatment of diarrhoea Denominator: number of indicators (6) Numerator: exclusive breastfeeding for 6 months, pentavalent vaccine coverage, measles vaccine coverage, pneumococcal vaccine coverage Denominator: number of indicators (4) Numerator: exclusive breastfeeding for 6 months, measles vaccine coverage, oral rehydration therapy, zinc for the treatment of diarrhoea Denominator: number of indicators (4) Numerator: insecticide-treated net, malaria testing, first-line malaria treatment Denominator: number of indicators (3) Numerator: access to improved drinking-water, access to handwashing with soap, access to an improved sanitation facility, access to clean fuel for cooking Denominator: number of indicators (4) a We did not include pneumonia care-seeking, pneumonia treatment and rotavirus vaccine coverage due to lack of data. a We did not include pneumonia care-seeking, pneumonia treatment and rotavirus vaccine coverage due to lack of data. a We did not include pneumonia care-seeking, pneumonia treatment and rotavirus vaccine coverage due to lack of data. Our primary outcome was provincial-level under-five mortality, calculated using the synthetic cohort probability method.33 We collapsed the indicator variables to provincial means and summed these into the six indicator grouped scores (Box 1) as the main exposure variables. We applied sample weights to adjust for sampling method for all data taken from the health facility and household data sets. All numerators and denominators presented here are raw data whereas some percentages are weighted. We performed negative binomial regression (due to overdispersion in the data), to estimate the associations between provincial-level under-five mortality and indicator coverage scores for both grouped and individual indicators. Due to collinearity, we analysed each indicator separately. We adjusted the negative binomial regressions for provincial level of conflict (none, minor or major conflict) and socioeconomic status, reporting the results as an adjusted coefficient. Due to low levels of missing data, we performed a complete case analysis. Differences in mean scores were compared using two-sample t-tests. We performed an individual-level analysis using logistic regression, to explore associations between being covered by an indicator (outcome) and living in a conflict-affected province (exposure), combining major and minor levels of conflict. We adjusted the analysis for household socioeconomic status. The analysis was performed using Stata version 16 (StataCorp, College Station, Texas, United States of America).
N/A
