Influence of age and sex on length of stay among COVID-19 patients at Kenyatta National Hospital Infectious Disease Unit, Kenya

¹Department of Community Health, School of Public Health, Amref International University, Nairobi, Kenya, ²Department of Health Systems Management and Development, School of Public Health, Amref International University, Nairobi, Kenya, ³Department of Nursing, University of Nairobi, Nairobi, Kenya

^&Corresponding author
Vivian Manyeki, Department of Community Health, School of Public Health, Amref International University, Nairobi, Kenya

Introduction    

The length of hospital stay (LOS) is an important health service measure because it reflects both the clinical course of disease and the operational efficiency of inpatient care. During COVID-19 surges, LOS directly affected bed availability, oxygen demand, intensive care unit (ICU) capacity, staffing, and treatment costs [1,2]. In low- and middle-income countries, where critical care capacity is constrained, accurate understanding of factors associated with LOS supports more realistic surge planning and resource allocation [3].

Global evidence has consistently linked older age and male sex with severe COVID-19 outcomes, particularly mortality and need for intensive care [4-7]. However, whether these demographic characteristics independently predict LOS is less consistent. LOS may be shortened by early death, prolonged by delayed recovery, or influenced by admission and discharge policies that changed rapidly during the early pandemic period [1]. Consequently, mortality and LOS should be interpreted together rather than treating LOS as a direct proxy for severity. Evidence from African settings remains limited but increasingly shows that COVID-19 hospitalization outcomes are influenced by local demographic profiles, comorbidity patterns, referral systems, oxygen availability and evolving public health protocols [8-10]. In Kenya, multicentre evidence showed that admitted COVID-19 patients were relatively young, and that age and male sex were more clearly associated with mortality than with duration of hospitalization [9]. More recent African evidence further indicates that COVID-19 outcomes varied by epidemic wave, vaccination period and underlying comorbidities, highlighting the importance of situating early pandemic data within its policy and clinical context [11,12].

This study examined the influence of age and sex on LOS among patients admitted with COVID-19 at the Kenyatta National Hospital Infectious Disease Unit (KNH-IDU) in Kenya, during the peak COVID-19 period before vaccine availability and while admission and discharge protocols were still evolving. The study focused on age and sex because they are routinely available demographic variables used in triage, surveillance, and service planning, but their utility for predicting LOS in this setting required local empirical assessment.

Methods     

Study design: this was a single-site retrospective cohort study based on a review of routinely collected electronic and paper medical records. The cohort design was appropriate because exposure information, clinical course, and outcomes were drawn from records of patients admitted during a defined period. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) items for cohort studies were considered [13].

Study setting: the study was conducted at KNH-IDU in Nairobi County, Kenya. KNH is the largest public tertiary teaching and referral hospital in Kenya and was designated to manage COVID-19 cases during the early pandemic response. In March 2020, KNH established a dedicated COVID-19 treatment facility with 102 beds, including 6 ICU beds. The unit included general ward beds for mild to moderate disease, high-dependency care for patients requiring close monitoring and oxygen support, and ICU capacity for critically ill patients requiring advanced respiratory support. These surge arrangements made KNH-IDU a relevant setting for examining LOS during a high demand and resource constrained phase of the pandemic.

Study population and eligibility: the study population comprised all patients admitted to KNH-IDU with laboratory-confirmed COVID-19 between 1^st June and 30^th November 2020. COVID-19 confirmation was based on a positive reverse transcriptase polymerase chain reaction or rapid antigen-based test recorded in the medical file. The study period was selected because it captured the early peak COVID-19 period in Kenya, including the July 2020 peak and the larger October 2020 peak, before national vaccine rollout and while clinical protocols were evolving. Patients were eligible if they had a confirmed diagnosis, an admission date, and a discharge or death outcome recorded. Records with unresolved missing admission or discharge dates or without laboratory confirmation were excluded.

Sampling approach: a census sampling approach was used. All eligible records from the defined period were enumerated from KNH-IDU admission logs and health information records, then cross-checked against electronic and physical patient files. This approach yielded 558 eligible patients, comprising 385 survivors and 173 non-survivors. Since the entire eligible cohort was included, no sample size calculation was required, and no probabilistic sampling interval was applied.

Variables and measurement: the main outcome was LOS, defined as the number of days from hospital admission to discharge or death. LOS was analyzed as a continuous variable using medians and ranges because of skewed distribution, and as a binary variable for regression analysis. Prolonged LOS was defined as greater than 8 days, based on the observed median LOS in the study cohort. The primary exposures were age and sex. Age was categorized as less than 20 years, 20-40 years, and more than 40 years, showing the distribution of the cohort and reporting categories used during the COVID-19 response. Sex was recorded as female or male. Survival status was used for stratification because death is a terminal event that may shorten LOS and complicate interpretation.

Potential confounders included disease severity at admission, comorbidity status, and oxygen therapy. Disease severity was classified using Kenya Ministry of Health COVID-19 clinical management categories. Mild disease included symptoms without evidence of pneumonia, moderate disease included clinical pneumonia without severe respiratory distress or hypoxia, and severe disease included respiratory distress, oxygen saturation below 90% on room air, ICU admission or need for advanced respiratory support.

Data collection: data were abstracted using a structured REDCap-based data abstraction tool designed by the principal investigator and guided by the study objectives, hospital records structure, and relevant COVID-19 literature. The tool captured demographic characteristics, admission and discharge dates, referral status, clinical severity, comorbidities, oxygen therapy, ICU admission, laboratory findings and outcomes. It was pretested on 10% of the intended sample using COVID-19 records from Mbagathi County Referral Hospital before 1^st June 2020. The pretest assessed clarity, completeness, and agreement with source records. However, no substantive changes were required after pretesting.

Three trained health records professionals retrieved and abstracted data. Two health records supervisors oversaw data collection. Before data abstraction, the team completed a two-day training on the study objectives, record retrieval, REDCap navigation, variable definitions, and confidentiality procedures. Data quality checks included double entry verification, duplicate detection using patient identifiers, and source document verification for outliers, particularly LOS values exceeding 60 days.

Statistical analysis: data were analyzed using Stata version 15. Frequencies and percentages summarized categorical variables. Medians and ranges summarized LOS because of non-normal distribution. Comparisons of categorical variables used Chi-square or Fisher exact tests, as appropriate. The Wilcoxon rank-sum test was used for non-parametric continuous comparisons where relevant. Logistic regression estimated crude and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for prolonged LOS. Models for age and sex were stratified by survival status and adjusted for clinically relevant covariates, including disease severity, comorbidity status and oxygen therapy. Missingness was below 5% for analyzed variables. Multiple imputation by chained equations and complete case sensitivity analyses produced comparable results. Therefore, the main findings are presented using the imputed dataset.

Ethical considerations: ethical approval was obtained from the Kenyatta National Hospital - University of Nairobi Ethics and Research Committee (approval number P354/04/2022) and the Amref Health Africa Ethics and Scientific Review Committee (approval number P1074-2021). A research permit was also obtained from the National Commission for Science, Technology and Innovation (NACOSTI) (approval number P/24/36842). Because the study used retrospective secondary patient records involving a large number of previously admitted patients, the ethics approvals included a waiver of individual informed consent. Patient identifiers were removed from the dataset, access was restricted to the study team, and encrypted files were used for data storage.

Results     

The final cohort included 558 patients admitted with confirmed COVID-19, of whom 385 (69.0%) survived, and 173 (31.0%) died. Overall median LOS was 7.5 days (range 0-183 days). Median LOS was 8 days among survivors and 6 days among non-survivors, indicating that mortality status influenced interpretation of LOS because some non-survivors died early during admission.

The age distribution differed between survivors and non-survivors. Among survivors, 3 (0.8%) were aged less than 20 years, 133 (34.5%) were aged 20-40 years, and 249 (64.7%) were aged more than 40 years. Among non-survivors, 2 (1.2%) were aged less than 20 years, 35 (20.2%) were aged 20-40 years, and 136 (78.8%) were aged more than 40 years. Despite this survival-status pattern, age was not independently associated with prolonged LOS in either stratum. Among survivors, adjusted ORs for prolonged LOS were 0.67 (95% CI 0.05-8.60) for patients aged 20-40 years and 0.68 (95% CI 0.05-8.52) for patients aged more than 40 years, compared with those aged less than 20 years. Among non-survivors, adjusted ORs were 0.52 (95% CI 0.03-9.60) and 0.64 (95% CI 0.04-11.28), respectively. Wide confidence intervals indicated the very small number of patients aged less than 20 years (Table 1).

Male patients constituted a larger proportion of both survivors and non-survivors. Among survivors, males had a median LOS of 9 days compared with 7 days for females. Male sex showed a non-significant tendency toward prolonged LOS among survivors (adjusted OR 1.56, 95% CI 0.99-2.43). Among non-survivors, males had a median LOS of 7 days compared with 5 days for females, with a similarly non-significant tendency toward prolonged LOS (adjusted OR 1.85, 95% CI 0.96-3.57). Thus, sex-related differences in median LOS were observed descriptively, but the adjusted associations did not meet statistical significance (Table 2).

Discussion     

This study examined whether age and sex influenced LOS among patients admitted with COVID-19 at KNH-IDU during the early peak COVID-19 period in Kenya. A key finding was that neither age nor sex independently predicted prolonged LOS after stratification by survival status and adjustment for clinical factors. Age was more clearly related to survival status, with non-survivors concentrated among patients older than 40 years, but this did not translate into a statistically significant association with prolonged LOS. Male sex showed a consistent but non-significant tendency toward longer LOS in both survivors and non-survivors.

The average age of study participants was approximately 49 years, with a median age of 48 years. This profile is younger than many cohorts reported from Europe and North America [14] and is broadly consistent with the younger demographic structure reported in African COVID-19 cohorts [9,10]. The lack of a significant association between age and prolonged LOS may therefore depict the local demographic profile, admission protocols during the early pandemic period, and the influence of death on LOS. Other studies have shown that older age strongly predicts severe outcomes and mortality [4,13], but the relationship with LOS is less direct because early death can shorten hospitalization while recovery from severe disease can prolong it [1,15-17].

The male-to-female ratio was approximately 1.35:1, indicating a predominance of male admissions. This is in line with global and African evidence showing higher risks of severe disease, ICU admission, or mortality among males [6,7,9]. In the current study, males had longer median LOS than females in both survival strata, but the adjusted estimates crossed the null. This indicates that sex may be clinically relevant for risk profiling, but in the present study it was not a sufficiently strong independent predictor of prolonged LOS for planning bed occupancy.

The findings should be interpreted within the early pandemic policy environment. During the study period, vaccination had not yet been introduced in Kenya, and hospital admission and discharge practices were strongly influenced by national isolation, testing and clinical recovery protocols. More recent African evidence shows that hospitalization, mortality and utilization patterns changed across later waves and vaccination periods [11,12]. Therefore, this study contributes historical but still relevant evidence for pandemic preparedness by showing that routinely available demographic variables alone were insufficient for predicting LOS in a national referral setting.

The study has several strengths. It used a census of all eligible KNH-IDU admissions over a defined early pandemic period, reducing sampling bias within the facility. It also stratified analyses by survival status, which is important because death is a terminal outcome that changes interpretation of LOS. Data abstraction was conducted using a pretested REDCap-based tool with trained records personnel and multiple quality checks. However, the study also had limitations. It was based on retrospective records from a single national referral facility, and findings may not generalize to primary care settings, home-based care, private hospitals, or later pandemic waves. Some clinically relevant variables, including treatment changes over time and post-discharge outcomes, were not available consistently. The small number of patients younger than 20 years resulted in wide confidence intervals and limited precision for age group comparisons.

Conclusion     

Age and sex were not independent predictors of prolonged LOS among COVID-19 patients admitted to KNH-IDU during the early peak pandemic period. Although older age was more common among non-survivors and males had slightly longer median stays, adjusted estimates were not statistically significant. Hospital surge planning in similar settings should prioritize clinical severity, oxygen and ICU needs, survival status, and evolving admission and discharge protocols rather than relying on age and sex alone.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

Vivian Manyeki conceptualized the study, coordinated data collection, led the analysis, and drafted the manuscript; Margaret Nyongesa supervised the study design, data interpretation, and manuscript revision; Dorcas Maina contributed to methodological design, data quality assurance, interpretation of findings, and critical manuscript revision. All the authors read and approved the final version of this manuscript.

Acknowledgments     

The authors acknowledge Kenyatta National Hospital Infectious Disease Unit and the health records team for facilitating access to records used in the study. The authors also acknowledge the supervisory support received from Amref International University and the University of Nairobi during the thesis from which this manuscript was developed.

Tables     

Table 1: age group and length of stay among confirmed COVID-19 patients at Kenyatta National Hospital Infectious Disease Unit, Kenya, June-November 2020, by survival status (N = 558)

Table 2: sex and length of stay among confirmed COVID-19 patients at Kenyatta National Hospital Infectious Disease Unit, Kenya, June-November 2020, by survival status (N = 558)

References