Predictors of Mortality in Children Admitted With Pneumonia to a Tertiary Hospital in Ghana: A Retrospective Cohort Study Authors: Sandra Kwarteng Owusu, Richard Kwaku Kwarteng Owusu, Birgit Agyeiwah Baah, Justice Sylverken, Nicholas Karikari Mensah, Naomi Dianne Adjetey, Serwah Bonsu Asafo‐Agyei, Larko Dormeyo Owusu, John Adabie Appiah, Samuel Blay Nguah, Emmanuel Ameyaw, Naana Wireko‐Brobby, Joseph Marfo, Augustine Kwame Afful, Lilliam Ambroggio, Daniel Ansong, Joyce Ashong, Lord Nkansah Published: 2026 ABSTRACT ABSTRACTBackground and AimPneumonia contributes substantially to the childhood mortality burden in low and middle‐income countries (LMICs). Reported predictors of mortality in childhood pneumonia in LMIC are varied. This study aimed to describe predictors of mortality among children with pneumonia in a tertiary hospital in Ghana.MethodsA retrospective cohort study that reviewed demographics, clinical features, admission outcomes, and predictors of in‐hospital mortality among Ghanaian children admitted with pneumonia. Outcome measures of interest were survival and death.ResultsA total of 630 children were eligible for the study; 394 (62.5%) had medical records that were at least 85% complete. 56% (n = 220) were males. Children who died had a median age of 8 months (IQR: 4–17) and 12 months (IQR: 6–24) for the survivors. Among children with pneumonia, 74% (n = 290) were younger than 24 months, and 73.3% (n = 239) were fully immunized. Mortality rate was 19% (n = 75/394), 34 children died within 24 h of admission, and a cumulative 42 children died within 48 h of admission. Hypoxemia at presentation was more frequent, 38.6% (n = 27), among those who died compared to the survivors, 17% (n = 47). Lower chest wall indrawing (LCI) occurred in both groups (80% vs. 60%, respectively). Predictors of mortality included; presence of diarrhea (aHR = 2.19, 95% CI:1.02–4.71, p = 0.05), hypoxemia (aHR = 2.08, 95% CI:1.09–3.96, p = 0.03).ConclusionWe observed a high burden of pneumonia in very young children. The predictors of mortality were hypoxemia and diarrhea. Background Pneumonia causes substantial hospitalizations and mortality in children, particularly in low‐ and middle‐income countries (LMICs) [1]. The significant number of pneumonia hospitalizations in young children places an undue burden on populations with under‐resourced health systems and weak economies [2, 3]. In addition to mortality, children who survive childhood pneumonia may have long term complications. Worldwide, the introduction of pneumococcal conjugate vaccines (PCVs) and improved access to antiretroviral therapy in countries with a high HIV burden have remarkably reduced pneumonia mortality in young children [1, 3], and childhood pneumonia deaths have fallen from 700,000 in 2015 [4] to 502,000 in 2021 [1]. Despite these improvements, pneumonia remains the leading cause of death in children under 5 years of age worldwide, with the largest burden in LMICs [5]. In Ghana, a similar trend was observed, with pneumonia‐related mortality among children under five decreasing from 8.39% to 6.96% between 2015 and 2020 [6]. Despite these improvements, data from UNICEF shows that in 2021, an estimated 3500 deaths occurred among children under 5 years from pneumonia in Ghana [7] and highlights the need for additional efforts to improve outcomes and reduce pneumonia mortality. Young infants, lack of immunization, retroviral infection, severe acute malnutrition, delayed referral, and presence of co‐morbidities have been reported as predictors of death from childhood pneumonia [8, 9, 10]. Additionally, clinical factors such as hypoxemia at presentation, chest wall indrawing, seizures, and diarrhea have been variably reported to be associated with mortality in children admitted with pneumonia [10]. It is noteworthy that hypoxemia has consistently been reported not only to be associated with poorer outcomes but also as an independent risk factor for in‐hospital death among children with pneumonia [11, 12, 13, 14]. In Ghana, however, there is limited data on factors predicting pneumonia mortality in children, which impedes evidence‐based clinical decision‐making and the development of additional public health interventions. The main objective of this study was to review the demographics, immunization status, clinical features, admission outcomes, and to identify predictors of in‐hospital mortality among Ghanaian children admitted to a tertiary hospital with pneumonia. Methods Study Design We conducted a retrospective cohort study of children aged 2–59 months admitted to the Komfo Anokye Teaching Hospital's (KATH) Paediatric Emergency Unit in Kumasi, Ghana, with WHO‐defined pneumonia, from January 1, 2016, to December 31, 2020. We defined the time origin as the admission date and the follow‐up period as the duration of hospital stay, hence permitting the use of Kaplan–Meier methods and Cox regression. The time scale represents days from admission to the outcome (i.e., survival or death). Children who were discharged alive were censored as they were no longer at risk for in‐hospital death. Study Site KATH is a tertiary hospital in the Kumasi metropolis, the capital city of the Ashanti Region of Ghana. The city of Kumasi and the surrounding areas are in the Middle Belt of Ghana, with a population of about 4 million [15]. It is a major referral center for many regions in Ghana. The Pediatric Emergency Unit (PEU) of KATH runs a 24‐h service. All children with medical emergencies are referred to this unit for assessments and initial emergency management. Annually, an estimated 500 children are seen with pneumonia at the Komfo Anokye Teaching Hospital, with 250 admitted to the PEU. Data Collection The Department of Child Health at KATH currently uses the WHO criteria for diagnosing pneumonia and severe pneumonia in children [16]. Each case of pneumonia in children included in this study was based on a diagnosis made by the attending pediatrician during the period of admission. The study used two sources of data for pneumonia admissions in children. Hard copies (hospital folders) of clinical records and electronic data from the Lightwave Health Information Management System (LHIMS) at the KATH‐PEU. Research assistants and medical doctors were trained to extract the data from clinical records. Data on demographic variables, immunization status, clinical presentation, clinical outcomes (mortality or survival), complications (such as pleural effusion), and duration of hospital stay were obtained from medical records. Extracted clinical data included the following: sex, age in completed months, duration of exclusive breastfeeding, maturity at birth (preterm or term), birth weight, and previous episodes of respiratory illness. For the current pneumonia illness, data obtained included the following variables: fever, cough, fast breathing, difficulty in breathing, vomiting, poor feeding, episodes of diarrhea, and episodes of seizures. Physical examination findings documented included: temperature, respiratory rate, oxygen saturation, and signs of respiratory distress (nasal flaring, head bobbing, lower chest wall in‐drawing, and grunting). Chest examination findings, nature of breath sounds, additional sounds (wheezes or crepitations), and weight were captured from the medical records. The data obtained was checked for errors and cleaned to ensure accuracy. Only data with a minimum of 85% completeness were included in the data analysis. A case of pneumonia was included in the data if it satisfied the following: a child under 5 years who presented with cough or difficulty breathing and age‐appropriate fast breathing. We excluded children whose medical record suggested chronic lung disease, such as bronchopulmonary dysplasia, suppurative lung disease, and bronchiectasis. Outcome The primary outcome for this study was in‐hospital death. Secondary outcomes included time to death and length of hospital stay, defined as the period from admission until death or discharge among survivors. Analysis of Data Summary statistics: median (interquartile range) was used for numerical variables such as age, duration of hospital stay, and anthropometric parameters. Categorical variables are presented as counts and percentages and compared between those who died and survivors. Cox regression was used to estimate hazard ratios for mortality. Unadjusted and adjusted hazard ratios with 95% confidence intervals are presented. Kaplan–Meier analysis was used to estimate survival probabilities. All tests were two‐sided. Statistical significance was determined a priori as a p value of less than or equal to 0.05. Stata SE Version 10 software (StataCorp, Texas, USA; www.stata.com) was used for the analysis. Ethical Considerations Ethical approval was obtained from the Institutional Review Board of the Komfo Anokye Teaching Hospital with approval number KATH/IRB/AP/081/21. Results Population Description A total of 630 children were eligible for the study; 394 (62.5%) had complete medical records (Figure 1), and 56% (n = 220) were males. Children who died had a median age of 8 months (IQR: 4–17) and 12 months (IQR: 6–24) for the survivors. Among children with pneumonia, 74% (n = 290) were younger than 24 months. In total, 73.3% (n = 239) of children were up to date on immunizations. (Table 1) The median hospital stay was 2 days (IQR, 1–5) for those who died and 4 days (IQR, 3–8) for those who survived (Table 3). Flow diagram illustrating study cohort selection and clinical outcomes. Table 1Demographic characteristics of children admitted with pneumonia in a tertiary hospital in Ghana.CharacteristicDied, n = 75Survived, n = 319AgeLess than 6 months27 (36%)75 (23.5%)6–11 months20 (26.7%)74 (23.2%)12–23 months14 (18.7%)80 (25.1%)24–59 months14 (18.7%)90 (28.2%)Median age (IQR)8.00 (4.00, 17.00)12.00 (6.00, 24.00)SexFemale37 (49.3%)137 (42.9%)Male38 (50.7%)182 (57.1%)ImmunizationComplete36 (65.5%)203 (74.9%)Incomplete19 (34.5%)68 (25.1%)Maturity at birthPreterm4 (5.6%)27 (8.7%)Term68 (94.4%)285 (91.3%)Birth weight< 2.56 (22.2%)33 (20.5%)2.5–3.513 (48.1%)78 (48.4%)> 3.58 (29.6%)50 (31.1%)WeightMedian (IQR)6.00 (4.70, 7.55)8.00 (6.00, 10.70)Caregiver educational statusBasic16 (24.6%)63 (21.3%)JHS22 (33.8%)103 (34.8%)None11(16.9%)38 (12.8%)SHS/MSLC4 (6.2%)24 (8.1%)Tertiary11 (16.9%)62 (20.9%)Vocational/technical1 (1.5%)6 (2.0%) Z scoreWeight for ageMedian (IQR)−2.07 (−3.74,−0.63)−0.99 (−2.36, −0.01)Height for ageMedian (IQR)−0.99 (−1.97, 1.00)−0.10 (−1.55, 1.01)Weight for heightMedian (IQR)−2.83 (−4.33, −1.86)−1.45 (−2.83, −0.13)John Wiley & Sons, Ltd. Demographic characteristics of children admitted with pneumonia in a tertiary hospital in Ghana. History of Presentation Poor feeding was slightly more common among those who died, 36.1% (n = 26), compared to survivors, 32.9% (n = 100). Seizures were more frequently reported among those who died, 15.5% (n = 11), than survivors, 7.7% (n = 23). Diarrhea and weight loss were higher among those who died, at 26.4% and 25.7%, respectively, compared to 11.9% and 12.9% among survivors (Table 2). Table 2Reported symptoms at presentation in children hospitalised with pneumonia.CharacteristicDied n = 75Survived, n = 319Presence of coughYes53 (72.6%)262 (84.8%)Presence of vomitingYes21 (28.8%)84 (27.4%)Presence of poor feedingYes26 (36.1%)100 (32.9%)Fast breathingYes27 (38%)120 (40.1%)SeizuresYes11 (15.5%)23 (7.7%)DiarrheaYes19 (26.4%)36 (11.9%)Reported weight lossYes18 (25.7%)39 (12.9%)John Wiley & Sons, Ltd. Reported symptoms at presentation in children hospitalised with pneumonia. Clinical Presentation and Predictors of In‐Hospital Mortality In total, the mortality rate was 19% (n = 75/394). Hypoxemia at presentation was more frequent, 38.6% (n = 27) among those who died, compared to those who survived, 17.2% (n = 47). Lower chest wall indrawing was prevalent in both groups (80% vs. 57%) (Table 3). The anthropometric parameters were generally lower in children who died compared to children who survived. Median z scores for weight for age and weight for height were −2.07 and −2.83, respectively, among those who died, compared with −0.99 and −1.45 among survivors (Table 1). The Kaplan–Meier estimates for mortality and the independent variables are shown in Figure 2. Unadjusted hazard ratios showed that diarrhea (HR 2.27, 95% CI:1.34–3.85, p= 0.002), oxygen saturation (HR 2.17, 95% CI:1.33–3.52, p= 0.002) and LCI (HR 2.0, 95% CI: 1.09–3.68, p= 0.03) were statistically significantly associated with in‐hospital mortality. The adjusted hazard ratio, however, showed that only diarrhea (aHR = 2.19, 95% CI: 1.02–4.71, p= 0.05) and hypoxemia (aHR = 2.08, 95% CI: 1.09–3.96, p= 0.03) were statistically significant predictors of mortality in pneumonia (Table 4). Table 3Clinical features and signs of respiratory distress among children hospitalised with pneumonia.Clinical features on arrivalDied, N = 75Survived, N = 319Temperature< = 38.017 (34%)97 (47.3%)> 38.033 (66%)108 (52.7%)Respiratory rate> = 5046 (61.3%)162 (52.9%)< 5029 (38.7%)144 (47.1%)Oxygen saturation> 90.043 (61.4%)226 (82.8%)< = 90.027 (38.6%)47 (17.2%)Head bobbingPresent8 (12.5%)12 (4.2%)Chest wall indrawingPresent52 (80%)176 (59.7%)GruntingPresent8 (12.9%)32 (11.1%)Central cyanosisPresent3 (4.8%)16 (5.6%)CrepitationsYes30 (47.6%)134 (46.5%)WheezingYes4 (6.5%)67 (23.4%)Breath soundsBroncho vesicular2 (3.3%)4 (1.4%)Bronchial40 (65.6%)184 (63.0%)Harsh0 (0.0%)4 (1.4%)Vesicular19 (31.1%)100 (34.2%)Hospital stay (days)Median duration (IQR)2.00 (1.0–5.0)4.00 (3.0–8.0)John Wiley & Sons, Ltd. Clinical features and signs of respiratory distress among children hospitalised with pneumonia. Composite graph of the Kaplan–Meier survival estimates for mortality and independent predictors. Table 4Cox regression model for predictors of mortality among children hospitalised with pneumonia.CharacteristicHR (95% CI) p valueaHR (95% CI) p valueAge (month)Less than 6 months1.04 (0.58, 1.87)0.901.01 (0.38, 2.66)0.996–11 months0.64 (0.33, 1.22)0.170.87 (0.32, 2.37)0.7812–23 months————24–59 months0.72 (0.37, 1.38)0.720.67 (0.22, 2.08)0.49Sex (male)0.91 (0.57, 1.43)0.680.99 (0.52, 1.92)0.98Seizures1.52 (0.79, 2.90)0.211.10 (0.44, 2.75)0.83Diarrhea2.27 (1.34, 3.85)0.0022.19 (1.02, 4.71)0.05SpO2 > 90————≤ 902.17 (1.33, 3.52)0.0022.08 (1.09, 3.96)0.03LCI2.00 (1.09, 3.68)0.031.74 (0.81, 3.76)0.16Resp. rate≥ 501.16 (0.73, 1.86)0.521.00 (0.52, 1.94)> 0.99< 50————Received 3 doses of PCV0.80 (0.44, 1.45)0.471.19 (0.49, 2.90)0.70Abbreviations: LCI, lower chest wall in drawing; PCV, pneumococcal conjugate vaccine; SpO2, oxygen saturation.John Wiley & Sons, Ltd. Cox regression model for predictors of mortality among children hospitalised with pneumonia. Abbreviations: LCI, lower chest wall in drawing; PCV, pneumococcal conjugate vaccine; SpO2, oxygen saturation. Discussion In this study, we found hypoxemia and diarrhea to be predictors of mortality. Hypoxemia as a predictor of mortality in this study is consistent with other studies on childhood pneumonia that reported predictors of mortality [14, 17, 18]. It is important to note that in our study, one out of five children managed for pneumonia presented with hypoxemia, highlighting the need to incorporate pulse oximetry into triaging systems performed at various levels of health care. Although this study did not collect data on the type of respiratory support provided pre‐referral or during transit to our center, exploring these factors in future studies to identify gaps and develop strategies will ensure that children receive adequate respiratory support before and during transfer. Increased insensible loss from rapid respiration [19], and suboptimal oral fluid intake [20] among children with pneumonia, predispose them to dehydration, which is worsened when they have diarrhea. Dehydration resulting from diarrhea disrupts the metabolic process and exacerbates respiratory dysfunction, which worsens their clinical outcome. Wilkes et al. [10], in a recently published systematic review of childhood pneumonia also found that diarrhea was significantly associated with mortality. The review also reported that other factors, such as cyanosis, lower chest indrawing (LCI), tachypnoea, and malnutrition predict mortality. However, our study did not identify a statistically significant association between malnutrition, cyanosis, LCI, tachypnoea, and mortality in pneumonia. This finding may be attributable to the retrospective study design and incomplete data availability, which influenced data completeness and the final sample size used in the analysis. Immunization has been identified as a key protective strategy for preventing childhood pneumonia [1]. Lack of immunization increases the risk of severe pneumonia among children [21], and it is also an independent predictor of mortality in studies by Wilkes et al. and Chakraborty et al. [10, 17]. In this study, although comparative percentage mortality was higher among children with incomplete than those with complete immunization (22% vs. 15%), it was not identified as a predictor of mortality. Mortality in children admitted with pneumonia was 19%, which was high when compared to 8.6% and 9.3% reported by Bokade et al. [8] and Cao et al. [9], respectively. The high mortality observed could be explained by the limited respiratory support services, such as continuous positive airway pressure devices and intensive care units with mechanical ventilators in the hospital [22]. In addition, the absence of full medical records for some of the children admitted with pneumonia, whose information was excluded during the period reviewed, could have accounted for the observed mortality rate. It is worthy of note that children less than 24 months have consistently been identified as a high‐risk group for severe pneumonia in published literature [10, 23]. Similarly, in our study, children younger than 24 months accounted for 73% of the admissions and 81% of deaths. Limitations The generalization of our findings should be done with caution, as there were limitations in our study. Being a retrospective study, medical records were not available for some potential patients, and this could account for the observed mortality rate. We appreciate that there may have been other shortfalls in documenting all clinical information accurately. Despite these limitations, findings from this study provide a rich body of primary data on childhood pneumonia in Ghana, which will serve as a guide for more robust prospective studies in future. Conclusion In conclusion, the burden of pneumonia is high in children under 2 years old. Mortality from childhood pneumonia remains high. Hypoxemia and diarrhea are significant predictors of mortality in children admitted with pneumonia. The first few days following admission are crucial, and continuous monitoring with optimal medical care and interventions should be a priority. Author Contributions Sandra Kwarteng Owusu: conceptualization, writing – original draft, writing – review and editing, supervision. Richard Kwaku Kwarteng Owusu: data curation, writing – review and editing. Agyeiwa Birgit Baah: data curation, writing – review and editing. Justice Sylverken: data curation, writing – review and editing. Nicholas Karikari Mensah: writing – review and editing, data curation. Naomi Dianne Adjetey: writing – review and editing, data curation. Serwah Bonsu Asafo‐Agyei: data curation, writing – review and editing. Larko Dormeyo Owusu: writing – review and editing, data curation. John Adabie Appiah: writing – review and editing, data curation. Samuel Blay Nguah: writing – review and editing, data curation. Emmanuel Ameyaw: data curation, writing – review and editing. Naana Wireko Brobby: writing – review and editing, data curation. Joseph Marfo: data curation, writing – review and editing. Augustine Kwame Afful: data curation, writing – review and editing. Lilliam Ambroggio: writing – review and editing. Daniel Ansong: data curation, writing – review and editing, formal analysis. Joyce Ashong: writing – review and editing. Lord Nkansah: writing – review and editing, formal analysis. Funding The authors have nothing to report. Conflicts of Interest The authors declare no conflicts of interest. Transparency Statement The lead author, Kwarteng Owusu Sandra, affirms that this manuscript is an honest, accurate, and transparent account of the study being reported, that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.