Time to recovery from severe pneumonia and its predictors among pediatric patients admitted in Mizan-Tepi University Teaching Hospital, South West Ethiopia, 2022

The study was conducted at the MTUTH. MTUTH is located 584 km from Addis Ababa. It serves the Bench Sheko Zone, neighboring woredas of Kaffa, Sheka zones, and partial Gambella regions (Majang zone) as a referral center, resulting in a catchment population of more than two million people. MTUTH provides various services in medical, surgical, ophthalmological, psychiatric, pediatric, and neonatal intensive care, under 5 out pertinent department (OPD), adult OPDs, and other related services. The study was conducted from 15 January 2022 to 30 January 2022.

An institutional-based retrospective cohort study was employed.

All children with severe pneumonia admitted to the pediatric ward of MTUTH were the source participants, and all children admitted to the pediatric ward with severe pneumonia during the study period were the study population.

All children aged from 1 month to 15 years with severe pneumonia admitted to the hospital during the study period were eligible for inclusion. Medical charts with incomplete or missing key information were also excluded.

The required sample size for this study was determined by using the STATA statistical package (Cox-model), version 14, based on the following assumptions: HR of 0.68 which is the selected covariate (formula feeding) taken from a study done in Gondar Hospital, a variability standard deviation (SD) of 0.5, 5% margin of error, 95% confidence interval (CI) and 80% power, and probability of recovery (event) 0.3636.¹⁶ The minimum sample size required was 581. After adding a 5% non-response rate, the final sample size for this study was 610.

The 3-year data of all children with pneumonia admitted to the pediatric ward between 01 January 2019 and 30 December 2021 were enrolled using the admission registration book by recording their medical record numbers sequentially. A total of 2124 medical registrations were obtained from the hospital over the past 3 years. The required number of medical registration cards was selected using a simple random sampling technique with a computer random number generation system in Excel. The selected medical cards were obtained from the medical records office.

The dependent variable was the time to recovery from severe pneumonia in children dichotomized as recovered or censored. The independent variables of the study included the following: socio-demographic variables (sex, age, and residency), clinical presentation and therapeutic-related variables (fever, antibiotic administration, danger signs during admission sign [inability to feed/drink, vomiting everything, head nodding, grunting]), drug regimen, oxygen saturation at admission, preterm birth history, low birth weight history, comorbidity, and nutritional and feeding characteristics (nutritional status [stunting, wasting, underweight], mode of infancy feeding [cow milk feeding, formula feeding, breastfeeding], and exclusive breastfeeding history).

The checklist was adapted from a similar study conducted at the Debre Markos Referral Hospital¹⁷ and modified after reviewing the related literature. ^15–17

The tool provides information on socio-demographic characteristics, clinical presentation and therapeutic characteristics, nutrition, and feeding characteristics. Data were collected through chart review, following the outcome of children starting from the day of admission in the pediatric ward up to the occurrence of an outcome (event or censored) in the hospital.

After checking for data completeness and consistency, the collected data were coded and entered into Epi-data version 4.4.2.1 statistical software packages, and the data were then transported to STATA version 14 for cleaning and analysis. The data were then described using relative frequencies and percentages. Kaplan–Meier survival estimates were used to estimate the median survival time and cumulative probability of survival at a given point in time. A life table was used to estimate the probability of survival at different time intervals during the follow-up period. The Log-rank test was used to compare the time to recovery differences between the different groups. The assumptions of the Cox proportional hazards (CPH) regression model were tested using the Schoenfeld residual test (global test). Additionally, the multicollinearity of each independent variable was checked using the variance inflation factor (VIF). Finally, the goodness of fit of model was checked using the Cox Snell residual graph. The CPH model was used to identify the independent predictors of mortality. Based on bivariate analysis, variables with a P-value <0.25 in the bivariate analysis were transferred to the multivariable analysis. Finally in multivariable analysis, those variables a P-value <0.05 at 95% CI level were considered as an independent predictor for time to recovery of children who had pneumonia. The final measure of the association between independent and dependent variables was expressed as an adjusted hazard ratio (AHR). Finally, the findings are presented in tables, graphs, and text.

To assure the quality, the study pretest was conducted on 5% (31) of the sample size, 2 weeks before the actual data collection period, which was not included in the actual study. Training was provided to data collectors and supervisors on how and what information should be collected from medical records for 1 day. The supervisor checked the data to ensure completeness and clarity. Completeness was checked before the data was entered.

Out of the 591 participants, males accounted for 359 (60.74%). Three hundred eight (52.12%) of them were in the age group of <1 year. Concerning to residency 380 (64.30%) of children came from rural areas (Table 1).

Two hundred forty-eight (41.96%) of the participants had hypoxia (oxygen saturation <90) during admission. Three hundred eighty-two (64.64%) children had fever during admission. Approximately 404 (68.36%) of participants have respiratory distress at the time of admission. The most common antibiotic first given to respondents 338 (57.19%) was ceftriaxone. For 324 (54.82%) children, antibiotics were changed after admission. Concerning the birth weight of children, 530 (89.68%) respondents had birth weights of 2500–4000g. Only 84 (14.21%) of participants were preterm (gestational age <37 weeks) (Table 2).

Of the total study participants, 397 (67.17%) were on exclusive breastfeeding. Related to baseline nutritional status, 82 (13.87%), 94 (15.91%), and 94 (15.91%) of the participants were characterized as underweight, wasting, and stunting, respectively. Of all study participants, 379 (64.13%) of them were on breast milk feeding (Table 3).

A total of 591 Children were followed for 1–11 days. Of 591 children, 541 (91.54%; 95% CI: 89.00–93.53) recovered from severe pneumonia during the follow-up period/hospitalization period and 50 (8.46%; 95% CI: 6.46–10.99) were censored. Children were followed for 2245 person-day observations, with an incidence rate of 24.10 (95% CI: 22.15–26.21) deaths per 1000 person-day observations (hospitalization period). The overall median recovery time of children in this study was 4 days (95% CI: 2–6). The Kaplan–Meier survival graph also confirmed the median recovery time of children (Figure 1).

In this study, the incidence rates for different age groups of children were 22.47 (95% CI: 19.96–25.28), 26.10 (95% CI: 22.32–30.40), and 26. 05 (95% CI: 21.52–31.54) in the <1 year, 1–5 years, and in >5 years, respectively (Table 4).

The Kaplan–Meier graphs show the overall cumulative survival and failure probability among children (Figure 2).

Comparisons of the survival time differences between different groups of categorical covariates were done through the Kaplan–Meier survival graph and statistical log-rank test at a 5% level of significance.

In this study, there was a recovery time difference between the categories of mode of infant feeding (P = 0.0148), breastfeeding status (P = 0.0331), respiratory distress (P = 0.0072), vomiting everything (P = 0.0121), altered level of consciousness (P = 0.0031), and inability to suck/feed (P = 0.0002) compared to their counterparts.

The assumptions of the CPH model were assessed using the Schoenfeld residual/global test, which became non-significant (0.4575) indicating that the proportional hazard assumption of CPH regression was met. The multicollinearity of each independent variable was checked using the VIF and the mean-VIF for these variables was 1.26.

In bivariate Cox regression analysis, predictors such as age, respiratory distress, vomiting, exclusive breast-feeding, antibiotic change, mode of infant feeding, altered level of consciousness, and inability to suck/ feed were associated with time to recovery. In the final CPH model, exclusively breastfeeding and inability to suck/feed were found to be independent predictors of recovery from severe pneumonia in children at 95% CI level. The time to recovery of children with severe pneumonia who did not have exclusively breastfeeding was observed to increase 1.3 times (AHR = 1.3; 95% CI: 1.03–1.66) than their counterparts. This indicates that children who were exclusively breastfed had a shorter time to recovery than those who were not exclusively breastfed. In addition to this, children who had a history of inability to suck/feed were observed to have 19% lower hazard of recovery from severe pneumonia (AHR = 0.81; 95% CI: 0.65–0.99) than their counterparts (Table 5).