The study’s objective was to determine malaria morbidity and mortality at Kinshasa Referral General Hospital (KRGH) or to investigate the status of RBM impact indicators, and factors associated with them.
Study design and area
This was a cross-sectional study of patients admitted at KRGH between January 2017 and December 2018, a period of two years. The facility was selected because it is the first and biggest referral hospital in the country. The hospital receives clients from all parts of Kinshasa and DRC. It has a capacity of 1,101 beds and an average of 8,468 admissions per year.
KRGH is located in Kinshasa, the capital city of DRC. Kinshasa’s planned urban structures were completed in the late seventies when its population was less than 1,200,000. Today, it has an estimated 14,340,000 inhabitants in an area of 9,965 km2 with only 600 km2 of that urbanized. The upper socioeconomic class of Kinshasa lives in this urban and planned area. Dwellings are modern, roads are paved and access to safe water, electricity and sanitation is relatively better. The outskirts of the city of Kinshasa, mainly inhabited by the lower socioeconomic class, are not urbanized but not planned. Dwellings are substandard and often similar to slum communities [16]. Roads and sanitary systems are in a disrepair state or non-existent, vehicle wrecks and puddles are scattered all over the area, a situation conducive for mosquito breeding and malaria outbreaks.
Sample size, sampling approach and data collection
The sample size was calculated using a sampling error of 0.05 and a beta level of 0.20 [17]. The proportion of baseline malaria morbidity/mortality among inpatients was assumed to be 20% [18]. The expected magnitude of association between malaria morbidity/mortality and exposure factors of interest (SES and residential area) was set at 1.68 odds ratio. This led to an estimated sample size of 808, which was increased by 20%. This brought the sample size to 969 participants to allow for any exclusion.
Client hospital registration numbers were used to create the sampling frame. The records included patient files, referral/discharge summaries and mortuary records. Using the computer table of random numbers, 969 patients were randomly selected. Information on the following variables was collected from the patients: age, gender, date of admission, date of discharge, mode of discharge (alive, dead, referred, or escaped), SES and residential area. In addition, information on the top causes of admission at the hospital was also collected. After excluding 99 patients for missing data and discrepancies between different data sources, the study sample size was finally brought to 870.
Data analysis
IBM SPSS version 21 (Chicago, IL) was used for analysis of the data. The distribution of inpatients by illness was computed to obtain the profile of major causes of admission at KRGH in 2017 and 2018. Malaria-specific morbidity rate was estimated by dividing the number of inpatients admitted for malaria over the total number of admissions in the same year, times 1000. Malaria-specific mortality rate was calculated by dividing the number of deaths due to malaria over the total number of deaths that occurred in the same year in the hospital, times 1000. Case fatality rates were calculated by dividing the number of malaria-related deaths over the total number malaria admissions in the same year, times 100. Comparisons between groups/subgroups were made using Chi-square test for trend, Chi-square test with correction of Yates or Fisher exact test. Using malaria morbidity/mortality as the dependent variable and year of admission, age, gender, socioeconomic status, and residential area as independent variables a multivariate logistic regression model was developed to identify predictors of malaria-related morbidity/mortality and to estimate odds ratios and their 95% confidence intervals. Only variables that achieved a p <0.09 significance level were further investigated in the main model. All likely confounders available in the patient records were investigated. Variance inflation factor (VIF) diagnostic was performed for redundancy and multicollinearity of covariates. How well the model fits the data was estimated using the Hosmer-Lemeshow test of goodness of fit.