Time to recovery and predictors among under five children with severe pneumonia admitted to public hospitals in West Gojjam zone Northwest Ethiopia Authors: Getie Beza, Mulatu Kassahun, Tsigie Gebrie, Haimanot Degu, Debash Tsegaye, Getnet Gedif Published: 2025 ABSTRACT Pneumonia remains a leading cause of under-five morbidity and mortality worldwide, with the burden particularly high in low- and middle-income countries, including Ethiopia. While most studies emphasize prevalence, limited evidence exists on recovery time and its predictors, despite their critical role in guiding clinical management and reducing hospital burden. Recovery duration reflects illness severity, treatment effectiveness, and resource utilization, and may be influenced by comorbidities, nutritional status, and timely initiation of treatment. Determining time to recovery and its predictors is therefore essential to improve child survival and optimize health system performance. This study aimed to assess the time to recovery from severe pneumonia and its predictors among under-five children admitted to West Gojjam Zone public hospitals in 2023. An institutional-based prospective follow-up study was conducted among 394 under-5 children in West Gojjam-selected public hospitals from April 2023 to July 2023. A systematic random sampling technique was used to select the study participants. Data were collected from mothers/caregivers using structured questionnaires. The data were entered into Epi data version 4.6 and analyzed using Stata version 17. After identifying predictor variables with P-value ≤ 0.25 during the bivariable analysis and by checking model assumption, Weibull regression was performed. A hazard ratio with a 95% confidence interval was used and a P-value < 0.05 was considered a statistically significant. The median time to recovery from severe pneumonia was 4 days with the interquartile range (3–6), and the overall recovery rate from severe pneumonia was 19.75 per 100 person-days (95% CI: 18% − 22%). The recovery from severe pneumonia was found among children with total visits for severe pneumonia ((4–5 times (AHR: 0.47, 95% CI: 0.34–0.66)), Educational level ((diploma and above) (AHR: 1.75, 95% CI: 1.26–2.44)), Maternal body mass index ((malnutrition) (AHR: 0.4, 95% CI: 0.21–0.74)), Cooking place ((Separate Kitchen) (AHR: 1.63, 95% CI: 1.25–2.13)), previous acute upper respiratory tract infection ((No) (AHR: 1.48, 95% CI: 1.12–1.95)), danger sign ((No) (AHR: 1.96, 95% CI: 1.53 2.45 )) and Residence ((urban) (AHR: 1.52, 95% CI: 1.16–2.00)). The findings of this study showed that the median time to recovery was 4 days. The main predictors of time to recovery from severe pneumonia were 4–5 times visiting with severe pneumonia, Educational level, Mother’s body mass index, Cooking place, a danger sign, previous acute upper respiratory tract infection and Residence. Hence, it should be better to give attention to children with identified predictors. Introduction Pneumonia remains one of the leading causes of morbidity and mortality among under-five children worldwide, accounting for nearly 14% of all deaths in this age group1. In 2019, an estimated 740,000 under-five deaths were attributed to pneumonia, with the burden disproportionately concentrated in low- and middle-income countries, particularly in Sub-Saharan Africa and South Asia1,2. Ethiopia is among the 15 high-burden countries, with pneumonia contributing substantially to child mortality3. While pneumonia is a common childhood illness, severe pneumonia is associated with higher risks of complications, prolonged hospitalization, and death4. Severe pneumonia is an infection and inflammation of air space in the lungs5. It is the most prevalent lung infection, affecting millions of children worldwide6. Pneumonia is characterized by cough, fever, chills, fatigue, dyspnea, and rigours and generally presents as an acute infection, with symptoms appearing over the course of hours or days and diagnosed clinically7. Severe pneumonia is a leading infectious disease that requires hospitalization and it is the most common cause of morbidity and mortality in children under the age of five, particularly in low-income communities8. The clinical symptoms of pneumonia in children are not distinct; no single symptom or sign is present9. The causative agent of pneumonia study conducted in the University Hospital of Udine reported that Streptococcus pneumonia was the most prevalent bacterial, fungal, and viral agent in pediatric community-acquired pneumonia (CAP) at any age, while Mycoplasma pneumonia is more common in kids over the age of 5 years10. For fast recovery of children from severe pneumonia among hospitalized clients better use of the different types of diagnosis methods was needed11. Globally, severe Pneumonia is a leading cause of death under 5 more than 800,000 children died due to pneumonia in 201712. Community-acquired Pneumonia is a leading cause of death under 5 more than 671,928 children died in 201913. Severe pneumonia is a leading cause of mortality and morbidity in children, accounting for more than 156 million episodes and 14.9 million hospitalizations each year as an infectious cause of death in children worldwide. It is a leading cause of sickness and mortality among children worldwide, particularly in low-income nations14. Nearly, 2,200 less than 5 years children die daily and over 800,000 children live with pneumonia in 201915. Severe pneumonia, which killed 740 180 children in 2019, accounted for 14% of all mortality in children under the age of five12. In Ethiopia, pneumonia contributes to 16.4% of all deaths under five years of age more than diarrhea, malaria, AIDS and measles combined16. More than 46,302 under 5 children died due to pneumonia in 2019 in Ethiopia13. The mortality rate of those under 5 by SCAP in Ethiopia was 50% in 201917. While much of the literature emphasizes the epidemiology, risk factors, and mortality of pneumonia, less attention has been paid to recovery time and the factors influencing it. Recovery time is an important measure because it reflects not only the severity of the illness and the effectiveness of treatment, but also has implications for hospital stay, healthcare costs, and resource utilization in already constrained health systems5,18. Prolonged recovery may increase the risk of complications, secondary infections, and mortality, whereas timely recovery reduces hospital burden and improves outcomes for both patients and health facilities19. Several studies have reported that recovery time is influenced by a range of clinical, demographic, and health system-related factors2,4,9. Clinical predictors include comorbidities such as malnutrition, anemia, and HIV infection, which are common among children in Sub-Saharan Africa20,21. Delayed presentation to healthcare facilities, poor adherence to treatment, and lack of supportive therapies such as oxygen supplementation have also been shown to prolong recovery22. Conversely, early initiation of appropriate antibiotics and adequate nutritional status are associated with faster recovery23. Despite the importance of recovery time, few studies in Ethiopia have investigated its duration and predictors among under-five children with severe pneumonia. Most available evidence focuses on prevalence and mortality, leaving an important knowledge gap on recovery patterns. Understanding recovery time and its determinants is crucial for improving treatment protocols, reducing length of hospital stay, and guiding policies to allocate resources more efficiently. Therefore, this study aimed to estimate the time to recovery from severe pneumonia and to identify its predictors among under-five children admitted to public hospitals in West Gojjam Zone, Northwest Ethiopia. Materials and methods Study design, area and period Institutional based prospective follow-up study was conducted in west Gojjam zone public hospitals, which are located in the Amhara region, 385 km from Addis Ababa northwest Ethiopia. West Gojjam Zone includes eight public hospitals that provide health care to the community. Merawi Primary Hospital, Finoteselam General Hospital, and Burie Primary Hospital admitted an average of 3980, 5747, and 2835 under five children each year respectively, according to statistics acquired from the selected administrative office of the hospitals’ reports in 2021 and 202224. From April 2023 to July 2023, the study was conducted on children under the age of five who were admitted to selected public hospitals due to severe pneumonia. Source and study population The source population included all children under the age of five who had severe pneumonia and were admitted to public hospitals in the west Gojjam zone. All under five children who have severe pneumonia that were admitted from April 2023 to July 2023 in selected West Gojjam zone public hospitals were our study population. Inclusion criteria All Children aged 1–59 months diagnosed with severe pneumonia and admitted to the selected West Gojam zone public hospitals during the study period were included in the study. Children with congenital heart disease were excluded from the study. Sample size and sampling procedures Sample size determination The sample size was calculated using Stata software Version 17. To have a representative sample of under 5 with severe pneumonia children, the sample size was determined by using the Log-rank test, two-sided significance level (α = 5%), Z a/2 = Z value at 95% confidence interval = 1.96, power 80% and p = cumulative recovery rate. The previous study found that the ratio of the exposure to the non-exposed population was 1:1. Variables such as formula feeding, cow milk feeding, underweight, presence of comorbidities, age, and presence of danger signs at admission were utilized to compute the sample size25. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{The}}\,{\mathrm{total}}\;{\mathrm{sample}}\;{\mathrm{size}}\;{\mathrm{is}}\;{\mathrm{calculated}}\;{\mathrm{as}}\;({\mathrm{n}})=\frac{{{\mathrm{Total}}\;{\mathrm{number}}\;{\mathrm{of}}\;{\mathrm{events}}}}{{{\mathrm{Probability}}\;{\mathrm{of}}\;{\mathrm{an}}\;{\mathrm{event}}}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N=\frac{E}{{P(E)}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$=\frac{{{{\left( {{{\mathrm{Z}}_{\alpha /2}}+{{\mathrm{Z}}_\beta }} \right)}^2}}}{{{\mathrm{P}}\;\left( {\mathrm{E}} \right)\;{{\left( {\ln {\mathrm{HR}}} \right)}^2}{\mathrm{P}}\left( {1 - {\mathrm{P}}} \right)}}$$\end{document} Zα/2: Z value at 95% confidence interval = 1.96. Zβ: the power of the study = 0.8. HR: hazard ratio. E: Total number of events. Thus, the final sample size was chosen for each case from designated hospitals until the desired quota was met (Table 1). Table 1Sample size calculation recovery for TR SP and its predictors among under-five children admitted to West Gojjam public hospitals 20from severe pneumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals 2023.VariablesReferencesSample sizeAHR and 95%CIMode of infancy feeding Formula feeding26,272180.68(0.25–0.58) Cow milk feeding27462.43(2.22–6.62)Presence of comorbidity28560.45Age253541.35 (0.43–0.91)Underweight26562.2 (0.47–0.88)Presence of danger signs at admission281340.61 (0.42–0.98) Table 1Sample size calculation recovery for TR SP and its predictors among under-five children admitted to West Gojjam public hospitals 20from severe pneumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals 2023.VariablesReferencesSample sizeAHR and 95%CIMode of infancy feeding Formula feeding26,272180.68(0.25–0.58) Cow milk feeding27462.43(2.22–6.62)Presence of comorbidity28560.45Age253541.35 (0.43–0.91)Underweight26562.2 (0.47–0.88)Presence of danger signs at admission281340.61 (0.42–0.98) Sample size calculation recovery for TR SP and its predictors among under-five children admitted to West Gojjam public hospitals 20from severe pneumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals 2023. Finally, by adding 10% adjusted loss 354/(1-0.1) = 394, the variable Age was a large sample size given of 394 who were admitted with sever pnumonia in selected West Gojjam Zone public Hospitals 2023. Sampling procedure First, by using simple random sampling (the lottery method), three hospitals were selected out of eight public hospitals found in the west gojjam Zone. The selected hospitals were Merawi Primary Hospital, Finote Selam General Hospital, and Burie Primary Hospital. By calculating the interval from the previous patient flow in each hospital, the total sample size for each was proportionally allocated. study participants were selected at every 2 interval until reaching the total sample size. The first participant was selected by the lottery method of sampling.(Fig. 1). Fig. 1Schematic presentation of sampling techniques for time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Schematic presentation of sampling techniques for time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Variable of the study Dependent variable. Time to recovery from severe pneumonia. Independent variables. Socio-demographic and hygiene characteristics Age, sex, residence, waste disposal system, Cooking Place, animal living with family, Occupation, Family size, Presence of health insurance and maternal educational status. Maternal and child nutritional status Exclusive breastfeeding, maternal BMI, WFH, WFA, HFA. Baseline clinical conditions of the child Chief complaint type, the type of drug prescribed, Presence of co-morbidity, History of previous ARTI, Presence of Lab findings, Deleance to come to the Hospital, vaccination status of the child and presence of danger signs. Operational definitions Recovery defined as discharge from hospital based on clinical improvement, including: normalization of respiratory rate for age, absence of chest indrawing, SpO2 ≥ 92% on room air, ability to feed orally, and physician’s judgment of stability. Event recovery from SP within the follow-up period. Time to recovery The number of days from admission or starting therapy to patient recovery from SP, as determined by calculating the difference from the start of treatment and recovery or censoring. Censored Children who were referred, died, or discharged for any cause without recovering during the study period. Danger sign was defined as loss of consciousness and/or vomiting everything and/or convulsions and/or inability to feed29on admission. Co-morbidities In children with SCAP, having TB, and/or HIV and/or malaria and/or Saver acute malnutrition, and/ or diarrhoea, and/or severe anaemia co-infection on admission. Fully vaccinated full vaccination refers to children, who have received and finished all forms of vaccinations. Partial vaccinated Refers to children who have received at least one dose of PCV. Not vaccinated refers to children who have not received any vaccines. (Children who have not had at least one dose of the PCV vaccine) Data collection tool and procedure Data was collected by face-to-face interviews from mothrs/caregivers using a structured questionnaire, adapted from previous studies5,6,18,26 with some modifications. The questionnaire was prepared in English language and translated to the national language Amharic and then it was also back translated to English to check consistency. These questionnaires contain socio-demographic characteristics, clinical characteristics, nutritional status and socio-economic characteristics. The data was collected by Six BSc nurses data collectors and three supervisors who were supervising the data-gathering procedure. Data quality assurance The questionnaire was translated into the local language to check consistency and a Pre-test was carried out at Injibara General Hospital on (5%) (20 cases) of respondents out of the main study before the actual data collection to ensure its consistency and errors were corrected and the questionnaire was modified before actual data collection. Maintaining the participants’ confidentiality throughout the data collection process and the study was discussed and ascertained during the day’s long training. Six BSc nurses who were data collectors and three supervisors who were supervising the data-gathering procedure were trained for one day. the researcher visited the data collectors as needed to check whether they collected the data appropriately. Data was double-checked for completeness both before and during the cleaning process. Data was checked again for completeness before data entry and during the data cleaning process. Data processing and analysis Data was prepared, edited, coded, and entered into EPI-Data version 4.6 before being exported to STATA version 17 for analysis. Descriptive statistics were used to summarize the data in a frequency distribution table and graphically for categorical data and to clarify in terms of central tendency (median) and dispersion (IQR). Days were used to calculate the time to recovery of the outcome during this study recovery/cure from SCAP). Children who defaulted, died, referred, and not recoverd during the study was regarded as censored. The outcome of every case is dichotomized into censored or recovered. Kaplan-Meier survival estimates and the log-rank test were used to examine the significance of differences in survival outcomes across various categories. Identifying bivariable predictor variables with a cut-off P-value ≤ 0.25 for multivariable analysis. Before running the regression, Cox proportional hazard assumption was cheked by Model goodness-of-fit Schoenfeld residual (global test = 0.0001 and graphically by Cox Snele residual plot (Residual Plots was not a straight line with a slope of 1) so it was considered as not fulfilling the assumptions. multicollinearity was checked by Varianc Infilation Factors. By comparing other survival analysis models (extended Cox regression, stratified Cox regression, Exponential regression, Weibull regression and Gompertz regression) based on considering the Check Model Fitness Assumptions flexibility, interpretability Specific Objective, Base of knowledge and model performance (The lowest AIC and BIC score) and finally, the Weibull regression model was selected. After identifying predictor variables with cut-off point P-value ≤ 0.25 during the bivariable analysis and by checking model assumption, Weibull regression was performed and the adjusted hazard ratio was used as a measure of association. Finally, Multivariable Weibull regression analysis was carried out to identify predictor variables that have a cutoff point of P-value ≤ 0.25 during the bi-variable analysis. A hazard ratio with a 95% confidence interval was used and a P-value < 0.05 was considered statistically significant. Results Socio-demographic and vaccination characteristics A total of 394 children were followed prospectively and the median time of recovery was 4 days with an interquartile range of 3–6 days. Among 394 study participants, 60.41% were male. More than half 71.07% came from rural areas. The median age of the children was 12 months with 58.38% of the children aged 2 to 12 months and 27% aged 13 to 36 months (Fig. 2). Fig. 2Socio-demographic and vaccination Characteristics of time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Socio-demographic and vaccination Characteristics of time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Economic, nutritional and clinical characteristics Among 394 children, 129 (32.74%) wasting (WFH), 28 (7.11%) stunting (HFA), underweight (WFA) 108(27.41%), 174 (44.16%)) co-morbidity, 191 (48.48%) presented with danger signs, 27 (6.45%) maternal malnutrition (Tables 2 and 3). Table 2Baseline clinical condition, nutritional status and economic characteristics of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023.VariablesCategoryFrequency n = 394Percent (%)Chief complaint typeCough And Fast Breathing12130.71Cough, Fast Breathing And Fever25464.47Cough, Fast Breathing And wheezing51.27Cough, Fast Breathing, wheezing and Fever10.25Cough and Fever133.30Antibiotic changedYes4310.91No35189.09Antibiotic prescribedCeftriaxone36693.65Ampicillin And Gentamycin255.58Ceftriaxone and Azithromycin30.76Previous ARTIYes30677.66No8822.34Presence of danger signYes19148.48No20351.52C-morbidityYes17444.16No22055.84Positive laboratory resultYes15539.34No23960.66Positive radiology resultYes6215.74No33284.26Total visit to the hospital with SP1–4 Times36291.885–8 Times328.12Outcome of treatmentCensored369.14Recovery35890.86Delayed to come to the Hospital1–3 day32883.24–5 day5513.96> 6 day112.8 Table 2Baseline clinical condition, nutritional status and economic characteristics of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023.VariablesCategoryFrequency n = 394Percent (%)Chief complaint typeCough And Fast Breathing12130.71Cough, Fast Breathing And Fever25464.47Cough, Fast Breathing And wheezing51.27Cough, Fast Breathing, wheezing and Fever10.25Cough and Fever133.30Antibiotic changedYes4310.91No35189.09Antibiotic prescribedCeftriaxone36693.65Ampicillin And Gentamycin255.58Ceftriaxone and Azithromycin30.76Previous ARTIYes30677.66No8822.34Presence of danger signYes19148.48No20351.52C-morbidityYes17444.16No22055.84Positive laboratory resultYes15539.34No23960.66Positive radiology resultYes6215.74No33284.26Total visit to the hospital with SP1–4 Times36291.885–8 Times328.12Outcome of treatmentCensored369.14Recovery35890.86Delayed to come to the Hospital1–3 day32883.24–5 day5513.96> 6 day112.8 Baseline clinical condition, nutritional status and economic characteristics of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Table 3Nutritional status and economic characteristics of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023.VariablesCategoryFrequency = 394Per cent (%)Occupation typeEmployee369.1Merchant102.2Daily labourer122.03Housewife194.82Farmer31272.No job20.5Waste and fluid disposal systemLatrine disposal only22156Fluid disposal only9323.6Both latrine and fluid30.76Not use both7719.55Health insuranceYes27569.8No11930.20Animal living statusYes6215.74No33284.26Presence of offensiveIn your surrounding odour wastes or fluidsYes30.76No39199.24Family sizeBelow 4 family sizes14637.0565–8 family size10827.4Above 8 family size4010.2Child nutritional statusBreastfeeding 1–7 times per day34286.80Breastfeeding 8 and above times per day5213.20HFA well nourished (z-score ≥-2)36692.89HFA malnourished z -score <-2287.01WFA well nourished (z-score ≥-2)28672.59WFA malnourished z -score <-210827.4WFH well nourished (z-score ≥-224572.59WFH malnourished z -score <-212927.4 Table 3Nutritional status and economic characteristics of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023.VariablesCategoryFrequency = 394Per cent (%)Occupation typeEmployee369.1Merchant102.2Daily labourer122.03Housewife194.82Farmer31272.No job20.5Waste and fluid disposal systemLatrine disposal only22156Fluid disposal only9323.6Both latrine and fluid30.76Not use both7719.55Health insuranceYes27569.8No11930.20Animal living statusYes6215.74No33284.26Presence of offensiveIn your surrounding odour wastes or fluidsYes30.76No39199.24Family sizeBelow 4 family sizes14637.0565–8 family size10827.4Above 8 family size4010.2Child nutritional statusBreastfeeding 1–7 times per day34286.80Breastfeeding 8 and above times per day5213.20HFA well nourished (z-score ≥-2)36692.89HFA malnourished z -score <-2287.01WFA well nourished (z-score ≥-2)28672.59WFA malnourished z -score <-210827.4WFH well nourished (z-score ≥-224572.59WFH malnourished z -score <-212927.4 Nutritional status and economic characteristics of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Treatment outcome and survival function graphs for significant variables Regarding the outcome of the treatment of children with SP, 358 (90.86%) recovered 33 (8.38%) medical referrals, and 3(1%) died during treatment (Fig. 3). Participants’ median time to recover from SP was 4 days with the IQR (3 − 6). The shortest and longest follow-up periods were 1 and 15 days, respectively. The survival rate was higher on the first day after a significant pneumonia diagnosis, but it declined as the follow-up period increased. The total follow-up time was 1793 person-days observation. The overall recovery rate from SP was 19.85 per 100 person-days (95% CI: 0.17 0.22). Fig. 3Treatment outcome for time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Treatment outcome for time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Comparison across groups Children with danger sign experienced significantly longer recovery times compared to those without (log-rank test, p = 0.0001)(Fig. 5). Similarly, children with URTI had delayed recovery (log-rank test, p = 0.0001) (Fig. 6). As we see from Figs. 4, 5, 6 and 7, Kaplan meier survival curve tend to decrease rapidly between the first five to ten days indicating that most children recovered from the disease within this time. Fig. 4The Kaplan–Meier survival estimate compares for time to recovery from sever pnumonia with categories of residence (95% CI-survival function) in West Gojjam zone public hospitals;2023. The Kaplan–Meier survival estimate compares for time to recovery from sever pnumonia with categories of residence (95% CI-survival function) in West Gojjam zone public hospitals;2023. Fig. 5The Kaplan–Meier survival estimate compares time to recovery from sever pnumonia with categories of presence of danger signs (95% CI-survival function) in West Gojjam zone public hospitals; 2023. The Kaplan–Meier survival estimate compares time to recovery from sever pnumonia with categories of presence of danger signs (95% CI-survival function) in West Gojjam zone public hospitals; 2023. Fig. 6The Kaplan–Meier survival estimate compares time to recovery from sever pnumonia with categories of the presence of a history of acute upper respiratory infection (95% CI-survival function) in West Gojjam zone public hospitals; 2023. The Kaplan–Meier survival estimate compares time to recovery from sever pnumonia with categories of the presence of a history of acute upper respiratory infection (95% CI-survival function) in West Gojjam zone public hospitals; 2023. Fig. 7The Kaplan–Meier survival estimate compares time to recovery from sever pnumonia with categories of cooking place (95% CI-survival function) in West Gojjam zone public hospitals; 2023. The Kaplan–Meier survival estimate compares time to recovery from sever pnumonia with categories of cooking place (95% CI-survival function) in West Gojjam zone public hospitals; 2023. Model building of time to recovery from SP Finally, the parametric regression model was employed to fit the data, as the Cox proportional hazard model did not successfully fit the data (it was violated) when testing the assumptions of the Cox proportional hazards regression model using the Schoenfeld residuals (global test) and graphically with the Cox-Snell residuals (Fig. 8). The global test value was 0.001, showing a p-value < 0.05, which indicates that the proportional hazards assumption was violated. Weibull regression offers significant advantages in terms of flexibility, interpretability, and capability. Fig. 8Test of proportional-hazards assumption by Cox Snell residual and Schoenfeld residual test of time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Test of proportional-hazards assumption by Cox Snell residual and Schoenfeld residual test of time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. The model with the lowest AIC score was selected, with a Weibull PH regression model having an AIC score of 594.5243 (Table 4). Table 4Model selection based on lowest information criteria multivariable analysis of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023.Model | N AICBICExponential regression394895.6546959.2762Weibull regression394594.5243662.1223Gompertz PH regression394735.8441783.5603Stratified cox regression3943196.4733252.142Extended cox regrasion3943602.9053646.645 Table 4Model selection based on lowest information criteria multivariable analysis of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023.Model | N AICBICExponential regression394895.6546959.2762Weibull regression394594.5243662.1223Gompertz PH regression394735.8441783.5603Stratified cox regression3943196.4733252.142Extended cox regrasion3943602.9053646.645 Model selection based on lowest information criteria multivariable analysis of time to recovery from sever Pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. As we can see from the Fig. 9 the lines align well or follow the same trend, it would indicate that the Weibull assumption holds, and the model is appropriate for the data. Fig. 9Test of Weibull assumption by Nelson commutative hazard graph test of time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Test of Weibull assumption by Nelson commutative hazard graph test of time to recovery from sever pnumonia and its predictors among under-five children admitted to West Gojjam zone public hospitals; 2023. Predictors of time to recovery from sever pneumonia Predictors that had an association at a p-value of ≤ 0.25 in bivariable Weibull regression were included in multivariable Weibull regression. co-morbidity, Residence, age of the child, AURTI, cooking place, WFH, educational level, maternal BMI, danger sign, and total number of visits were selected through the stepwise backward elimination approach at a p-value of ≤ 0.25. In the final Weibull regression model, Residence, AURTI, cooking place, educational level, maternal BMI, danger sign, and total number of visits became statistically significant predictors of recovery time from SCAP at a p-value of 0.05 level of significance(Table 5). Table 5Bivariable and Multi-variable Weibull regression for time to recovery from sever Pnumoniaand its predictors among 1–59 month children admitted in West Gojam public hospitals, Northwest Ethiopia, 2023.VariablesCategoryOutcome statusCHR 95%CIAHR 95%CI P RecoveryCensoredComorbiditiesYes159(40)15(4)11No199(51)21(5)1.3(1.05–1.61)1.04(0.79–1.20)0.95ResidenceRural252(64)28(7)11Urban106(27)8(2)1.92(1.55–2.54)1.52(1.16-2.00)**0.002AURTIYes278(71)28(7)11No80(20)8(2)1.98(1.55–2.55)1.48(1.12–1.95)**0.005Age of child≤ year208(53)22(6)112-3yr102(26)6(2)1.1(0.87–1.83)0.94(0.71–1.28)0.54 and above48(12)8(2)1.24(0.9–1.70)1.14(0.79–1.65)0.45Cooking placeIn living room93(24)8(2)11In a Separate Kitchen265(67)28(7)1.3(1.07–1.73)1.63(1.25–2.13)*0.034Educational levelIlliterate184(47)24(6)11Grade 1-4 complete57(14)1(0)1.77(1.33–2.42)1.79(1.30–2.45)**0.000Grade 5-12complate52(13)15(4)1.20(0.88–1.32)1.15(0.84–1.59)0.36Diploma and above65(16)21(5)1.56(1.18–2.04)1.75(1.26–2.44)**0.001Maternal BMINormal337(86)28(7)11Undeweight12(3)8(2)0.30(0.16–0.53)0.4(0.21–0.74)**0.004Above9(2)28(7)0.26(0.13–0.51)0.15(0.07–0.30)**0.000WFHSevere38(10)8(2)11Moderate78(20)22(6)1.78(1.2–2.6)1.13(0.76–1.68)0.54Normal242(61)6(2)2.39(1.6–2.39)1.39(0.96–2.01)0.45Danger_signYes171(43)8(2)1No187(47)8(2)2.58(2.08–3.20)1.96(1.53–2.45)**0.000Total number of visits1–3 visit300(76)28(7)0.42(0.29–0.62)14–5 visit54(14)24(6)0.058(0.250.47)0.47(0.34–0.66)**0.016 and Above4(1)1(0)1.65(0.61–4.44)0.85(0.29–2.48)0.77WFH weight forage, HFA height for age, WFH weight for height, 1 = Reference.URATI upper acute respiratory tract infection.motherBMI maternal body mass index.*Statistically significant at 0.05 ** Statistically significant at 0.025. Table 5Bivariable and Multi-variable Weibull regression for time to recovery from sever Pnumoniaand its predictors among 1–59 month children admitted in West Gojam public hospitals, Northwest Ethiopia, 2023.VariablesCategoryOutcome statusCHR 95%CIAHR 95%CI P RecoveryCensoredComorbiditiesYes159(40)15(4)11No199(51)21(5)1.3(1.05–1.61)1.04(0.79–1.20)0.95ResidenceRural252(64)28(7)11Urban106(27)8(2)1.92(1.55–2.54)1.52(1.16-2.00)**0.002AURTIYes278(71)28(7)11No80(20)8(2)1.98(1.55–2.55)1.48(1.12–1.95)**0.005Age of child≤ year208(53)22(6)112-3yr102(26)6(2)1.1(0.87–1.83)0.94(0.71–1.28)0.54 and above48(12)8(2)1.24(0.9–1.70)1.14(0.79–1.65)0.45Cooking placeIn living room93(24)8(2)11In a Separate Kitchen265(67)28(7)1.3(1.07–1.73)1.63(1.25–2.13)*0.034Educational levelIlliterate184(47)24(6)11Grade 1-4 complete57(14)1(0)1.77(1.33–2.42)1.79(1.30–2.45)**0.000Grade 5-12complate52(13)15(4)1.20(0.88–1.32)1.15(0.84–1.59)0.36Diploma and above65(16)21(5)1.56(1.18–2.04)1.75(1.26–2.44)**0.001Maternal BMINormal337(86)28(7)11Undeweight12(3)8(2)0.30(0.16–0.53)0.4(0.21–0.74)**0.004Above9(2)28(7)0.26(0.13–0.51)0.15(0.07–0.30)**0.000WFHSevere38(10)8(2)11Moderate78(20)22(6)1.78(1.2–2.6)1.13(0.76–1.68)0.54Normal242(61)6(2)2.39(1.6–2.39)1.39(0.96–2.01)0.45Danger_signYes171(43)8(2)1No187(47)8(2)2.58(2.08–3.20)1.96(1.53–2.45)**0.000Total number of visits1–3 visit300(76)28(7)0.42(0.29–0.62)14–5 visit54(14)24(6)0.058(0.250.47)0.47(0.34–0.66)**0.016 and Above4(1)1(0)1.65(0.61–4.44)0.85(0.29–2.48)0.77WFH weight forage, HFA height for age, WFH weight for height, 1 = Reference.URATI upper acute respiratory tract infection.motherBMI maternal body mass index.*Statistically significant at 0.05 ** Statistically significant at 0.025. Bivariable and Multi-variable Weibull regression for time to recovery from sever Pnumoniaand its predictors among 1–59 month children admitted in West Gojam public hospitals, Northwest Ethiopia, 2023. WFH weight forage, HFA height for age, WFH weight for height, 1 = Reference. URATI upper acute respiratory tract infection. motherBMI maternal body mass index. *Statistically significant at 0.05 ** Statistically significant at 0.025. Discussion In Ethiopia, pneumonia is a leading cause of morbidity and mortality among children under the age of five. The purpose of this study was to determine the median recovery time from severe pneumonia and its predictors in under-five children admitted in West Gojam Zone public hospitals. In this study, the overall incidence of recovery rate was found to be 19.75 (95%CI, 17, 22) per 100-person-day observation. This result is higher than the study conducted at Debre Markose referral hospital (16.25)27and Gonder referral hospital (13.5)23,30,31. But lower than in Nigist Eleni Mohammed Memorial Comprehensive Specialized Hospital reported were 24.16 per 100 person-days28. The difference may be due to differences in the study area sample size, service quality and presence of comorbidity which as compared to the above studies. In this study, the median time to recover from SP was 4 days, and IQR was 3 days. This was in line with the findings from institution-based research in Pawe General Hospital 4 days32; and Debre Mrkose referral Hospital (4 days)27; and higher than Gondar Comprehensive Specialized Hospital (3 days)23 the finding in Hospitalized Young Nepalese Children (2 days)33 ; and less than studies conducted in Tanzania (10 days)34 United States Centers for Diseases Control and Prevention( 5 days)30. These variations may reflect differences in healthcare quality, treatment protocols, and patient characteristics. Facilities with better staffing, early diagnosis, and reliable access to oxygen therapy are likely to achieve faster recovery compared to those with limited resources. In this study, the recovery among underweight and overweight maternal body mass index was delayed by 60% and 85% respectively as compared to the normal BMI of mothers. This report was in line with the study conducted at Debre Markos Referral Hospital, University of Gondar Comprehensive Specialized Hospital and Asosa General Hospital Gambia by Kuti, and Nepal by Sudha Basnet which showed that malnourished children had prolonged time of recovery23,26,35,36. Maternal malnutrition compromises breastmilk quality and infant immunity, predisposing children to recurrent infections and slower recovery. This aligns with evidence showing that nutritional deficiencies weaken immune function, prolonging hospital stay37. Similarly, children from households with a separate kitchen recovered faster than those exposed to indoor smoke. Indoor air pollution is a well-documented risk factor for respiratory illness severity, and continued exposure during treatment may aggravate symptoms, delaying recovery38. This finding highlights how environmental and household factors interact with medical care in determining outcomes. The presence of comorbidities such as acute upper respiratory infections was also associated with slower recovery. Comorbid conditions increase the physiological burden on children, complicate treatment, and extend the time needed to reach clinical stability. This is consistent with literature from Ethiopia and other low-income countries where malnutrition, anemia, and coinfections like HIV and malaria commonly prolong pneumonia recovery5,32. In this study, the recovery among maternal who have completed a diploma and above educational level was 1.61 times as compared to illiterate mothers. this disparity can be attributed to factors such as better health literacy, improved access to healthcare, higher socioeconomic status, healthier behaviours, and enhanced parenting skills among educated mothers. Children admitted without danger signs recovered more quickly than those with danger signs, which is expected given the greater severity of illness at baseline. Danger signs (e.g., inability to feed, convulsions, hypoxia) often require intensive management, longer antibiotic courses, and close monitoring, explaining the extended recovery period. This report was similar to the study conducted in Debre Markose referral hospital, and Swiss multicenter trial which showed that present in danger sign children had prolonged time of recovery27,39. This suggests Children with severe pneumonia and danger signs exhibit a slower recovery due to the gravity of their condition, requiring intensive medical intervention, including hospitalization and specialized care. Factors such as underlying health issues and potential complications further contribute to the extended recovery process. Residence was another significant predictor: urban children recovered faster than rural counterparts. This may be explained by differences in health-seeking behavior, earlier presentation to facilities, and better continuity of care in urban settings. Rural children may arrive later with more advanced disease, leading to delayed recovery. Overall, these findings underscore that recovery from severe pneumonia is not determined solely by medical treatment but is strongly influenced by healthcare quality, baseline comorbidities, nutritional status, and environmental conditions. Addressing these broader determinants is essential to improving pneumonia outcomes and reducing pediatric hospital burden. Limitations of the study This study has several limitations that should be considered when interpreting the findings. First, since data were collected only from public hospitals in West Gojjam Zone, the results may not be generalizable to private facilities or other regions of Ethiopia where healthcare quality and resources may differ. Second, although a prospective follow-up design was used, some variables such as household factors and caregiver practices were self-reported and may have been subject to recall or social desirability bias. Third, recovery was measured using clinical criteria at discharge, which may not fully capture long-term outcomes or relapses after leaving the hospital. Fourth, unmeasured confounders such as variations in treatment adherence, nutritional supplementation, or environmental exposures could have influenced recovery time but were not accounted for in the analysis. Finally, the relatively small sample size for some subgroups (e.g., children with comorbidities) may have limited the statistical power to detect associations. Despite these limitations, the study provides valuable insights into recovery time and its predictors in a resource-limited setting, contributing evidence that can inform child health policies and clinical management in Ethiopia. Conclusion and recommendation This study found that the median time to recovery from severe pneumonia among under-five children admitted to public hospitals in West Gojjam Zone was 4 days. Several factors, including the number of prior health facility visits, maternal educational status, maternal body mass index, household cooking environment, history of upper respiratory tract infection, and rural residence, were identified as independent predictors of delayed recovery. These findings suggest that both clinical and non-clinical factors influence recovery from severe pneumonia. Interventions to improve maternal education, strengthen nutritional support, and reduce household air pollution could play a critical role in shortening recovery time. Furthermore, strengthening early case detection and ensuring equitable access to quality healthcare for rural children are essential strategies to improve child health outcomes. Overall, this study provides valuable evidence for clinicians and policymakers to design targeted interventions that address the determinants of delayed recovery. Further research should explore long-term outcomes and evaluate the effectiveness of integrated strategies to reduce pneumonia-related morbidity and mortality among under-five children in Ethiopia.