Clinical predictors of COVID-19 in-hospital mortality in Anjouan, Comoros: implications for triage strategies in small Island developing States

¹Laboratory of Biology and Health, Faculty of Science, Ibn Tofail University, Kenitra, Morocco, ²Higher Institute of Nursing Professions and Health Techniques, Rabat, Morocco, ³University Hassan First of Settat, Higher Institute of Health Sciences, Health Sciences and Technologies Laboratory, Settat, Morocco, ⁴Medical Service, Regional Hospital Centre of Anjouan, Mutsamudu, Comoros

^&Corresponding author
Salim Irnat, Laboratory of Biology and Health, Faculty of Science, Ibn Tofail University, Kenitra, Morocco

Introduction    

Public health crises place extreme pressure on health systems, revealing how early clinical assessment, care coordination, and patient triage can decisively shape outcomes. Such crises consistently expose critical vulnerabilities, even in systems with advanced infrastructure [1,2]. In Africa, successive public health threats, from COVID-19 and Ebola to cholera and yellow fever, accentuate systemic challenges such as scarce resources, fragile infrastructure, and the urgent need for rapid, coordinated responses [3,4]. Timely case detection, effective care delivery, and efficient resource mobilisation are therefore crucial to mitigating morbidity and mortality during outbreaks [5,6].

The COVID-19 pandemic has intensified these strains across the continent. A critical paradox has emerged: while reported prevalence remains comparatively low in Africa [7], mortality rates among confirmed cases are disproportionately high [8,9]. This discrepancy reflects inequitable care access, diagnostic delays, and inconsistent quality of clinical management. In such constrained environments, rapid identification of high-risk patients and adaptive implementation of evidence-based interventions become essential strategies for reducing preventable deaths [10,11].

Health systems in low-resource settings face converging challenges, including insufficient hospital capacity, shortages of qualified personnel, limited access to oxygen and essential medicines, and logistical barriers to patient monitoring [12-14]. During epidemic surges, the emergence of more transmissible variants further exacerbated these pressures, making initial triage decisions (hospital admission versus home-based monitoring) a critical determinant of outcomes [15,16].

The island of Anjouan in the Union of the Comoros illustrates these challenges. During the second COVID-19 wave, constrained infrastructure and a limited healthcare workforce necessitated difficult triage decisions, with some patients monitored at home and others admitted directly according to clinical severity and available resources. However, evidence on COVID-19 outcomes and triage practices from small island developing states remains limited, hindering the development of context-specific preparedness strategies. This context provides a valuable setting for examining factors associated with mortality among hospitalised patients and for deriving practical lessons applicable to similar settings. Within this framework, this study aims to: (1) identify factors independently associated with mortality among COVID-19 patients hospitalised in Anjouan during the peak of the second epidemic wave; and (2) assess whether the initial management pathway (home-based monitoring versus direct hospitalisation) was associated with clinical outcomes, to inform triage strategies and epidemic response in resource-limited island settings.

Methods     

Study design and population: this retrospective observational study was conducted at the regional referral hospital of Anjouan, Union of the Comoros, during the peak of the second COVID-19 pandemic wave. The study included all symptomatic, RT-PCR-confirmed COVID-19 patients who were hospitalised between 8 January and 9 February 2021, regardless of whether they were initially managed at home or directly admitted to the hospital. This exhaustive inclusion was chosen to provide a comprehensive overview of patient care pathways, clinical practices, and resource allocation under conditions of intense epidemic pressure.

Data collection and variables: data were collected using a standardised clinical form to ensure systematic and uniform documentation for all included patients. The form captured variables across the following domains: i) sociodemographic characteristics: age, sex, occupation, place of residence, and mode of transportation to the healthcare facility; ii) clinical presentation: symptoms at admission, including fever, chills, fatigue, cough, dyspnea, sputum production, headache, and myalgia; iii) medical history: pre-existing comorbidities including hypertension, diabetes, asthma, obesity, and cardiovascular disease; iv) clinical course: dates of admission and discharge, initial management setting (home-based versus direct hospitalisation), and final clinical outcome (discharged alive or deceased); v) physical examination: vital signs measured at admission, including heart rate, respiratory rate, systolic and diastolic blood pressure, and peripheral oxygen saturation (SpO₂).

Case definition and severity classification: a confirmed COVID-19 case was defined as any individual with a positive RT-PCR test for SARS-CoV-2. Patients were categorised into four clinical severity levels at admission, based on respiratory parameters, vital signs, and comorbidity status: i) mild: symptoms without evidence of pneumonia, heart rate ≤90 bpm; respiratory rate ≤20 breaths/min; oxygen saturation (SpO₂) ≥94%; absence of significant comorbidities; ii) moderate: pneumonia (respiratory rate 20-29 breaths/min and/or SpO₂ 90-93%); or presence of comorbidities (e.g., hypertension, diabetes, obesity) or age ≥60 years, without signs of severe respiratory distress; iii) severe: severe pneumonia (respiratory rate ≥30 breaths/min and/or SpO₂ <90%), dyspnea at rest; or requirement for supplemental oxygen therapy; iv) critical: acute respiratory distress syndrome (ARDS), septic shock, or multi-organ failure requiring life-saving interventions such as mechanical ventilation or vasopressor support.

Vital signs classification: to ensure consistency in clinical assessment, vital signs were categorised as follows: i) blood pressure: hypotension (systolic blood pressure <90 mmHg or diastolic blood pressure <60 mmHg); normal (systolic 90-139 mmHg and diastolic 60-89 mmHg), hypertension (systolic ≥140 mmHg or diastolic ≥90 mmHg); ii) heart rate: normal (≤90 bpm), tachycardia (>90 bpm); iii) respiratory rate: normal (≤20 breaths/min), tachypnea (>20 breaths/min); iv) oxygen saturation: normal (SpO₂ ≥95%), hypoxemia (SpO₂ <95%).

Statistical analysis: data were analysed using SPSS version 21.0. Continuous variables were expressed as mean ± standard deviation, and categorical variables as frequencies and percentages. Comparisons between groups were performed using appropriate statistical tests according to variable type and distribution. For analytical purposes, clinical severity was dichotomised into mild/moderate versus severe/critical, and blood pressure was grouped as normal versus abnormal (hypertension or hypotension). Multivariate binary logistic regression was performed to identify factors independently associated with in-hospital mortality. Variables with p ≤ 0.10 in bivariate analysis or with established clinical relevance were considered for inclusion. Given the limited number of events (n=14), a maximum of three variables were retained in the final model to avoid overfitting. Results were reported as odds ratios (OR) with 95% confidence intervals (CI). Statistical significance was set at p < 0.05.

Ethical considerations: this retrospective study was authorised by the health authorities of Anjouan. All data were extracted from routine clinical records and fully anonymised before analysis. Given the retrospective, non-interventional nature of the study and the use of de-identified data, the requirement for individual informed consent was waived.

Results     

Patient characteristics and initial management: during the study period, a total of 66 COVID-19 patients were hospitalised in Anjouan (Table 1). The cohort had a mean age of 49.0 ± 16.9 years, with a slight male predominance (53.0%). Most patients resided in urban areas (74.2%). The majority were transported to the hospital by fire brigade (57.6%) or ambulance (16.7%), while private vehicles (22.7%) and public transportation (3.0%) were less frequently used. More than half of the patients (n = 39; 59.1%) were initially managed at home before being hospitalised, while 27 patients (40.9%) were directly admitted. Comparison between these two groups showed broadly comparable sociodemographic and clinical profiles, with the notable exception of age. Patients directly admitted were significantly younger than those initially managed at home (42.4 ± 15.6 years vs 53.6 ± 16.4 years; p=0.007).

Comorbidities and clinical presentation: comorbidities were common, including hypertension (34.8%), diabetes (33.3%), and obesity (19.7%). Their distribution was similar between the two management groups. The most frequent symptoms at admission were fatigue (72.7%), dyspnea (66.7%), chills (66.7%), and cough (66.7%), followed by fever (42.4%), headache (37.9%), myalgia (27.3%), and sputum production (19.7%). Fever showed a trend toward higher frequency in the home-based care group compared to directly hospitalised patients (51.3% vs 29.6%; p = 0.08), while sputum production was more common among immediately hospitalised patients (29.6% vs 12.8%; p = 0.09), although neither difference reached statistical significance.

Vital signs and clinical outcomes: vital parameters at admission were similar between the two groups. Mean respiratory rate was 27.6 ± 7.3 breaths/min, mean heart rate was 93.5 ± 15.9 bpm, and mean oxygen saturation was 90.7 ± 11.7%. Mean systolic blood pressure was 131.7 ± 21.5 mmHg, and mean diastolic blood pressure was 76.1 ± 13.6 mmHg. The overall mortality rate was 21.0%. Mortality was higher among patients initially managed at home compared to those directly hospitalised (25.6% vs 14.8%), although this difference did not reach statistical significance (p = 0.29).

Factors associated with in-hospital mortality: in univariate analysis (Table 2), several clinical and physiological factors were associated with increased odds of in-hospital mortality. Statistically significant associations were observed for hypoxemia (SpO₂ <95%; OR = 12.35; p = 0.002), severe or critical disease (OR = 9.60; p = 0.006), tachypnea (respiratory rate >20/min; OR = 8.36; p = 0.049), tachycardia (heart rate >90 bpm; OR = 6.22; p = 0.005), obesity (OR = 4.82; p = 0.020), diabetes (OR = 3.62; p = 0.039), and abnormal blood pressure (OR = 3.62; p = 0.039). The presence of any comorbidity (OR = 3.96; p = 0.052) and older age ≥60 years (OR = 3.00; p = 0.078) showed trends toward association but did not reach statistical significance.

In multivariate analysis, three variables remained independently associated with in-hospital mortality. Disease severity was the strongest predictor: patients presenting with severe or critical forms had a 14-fold higher risk of death compared to those with mild or moderate forms (aOR = 14.29; 95% CI: 2.11-96.93; p = 0.006). Tachycardia was independently associated with a six-fold higher risk of death (aOR = 6.13; 95% CI: 1.16-32.29; p = 0.032). Notably, male sex was associated with a lower risk of death (aOR = 0.12; 95% CI: 0.02-0.74; p = 0.022). The overall model demonstrated acceptable explanatory power (Nagelkerke R² = 0.42). The initial management pathway was not significantly associated with mortality (OR = 1.98; p = 0.295).

Discussion     

This study identified disease severity at admission and tachycardia as independent predictors of in-hospital mortality among COVID-19 patients hospitalised in Anjouan during the peak of the second epidemic wave. Patients presenting with severe or critical forms had a 14-fold higher risk of death, while those with tachycardia had a six-fold increased mortality risk. Notably, the type of initial management (home-based monitoring versus direct hospitalisation) was not independently associated with mortality, suggesting that clinical outcomes depend more on timely recognition of deterioration and prompt escalation of care than on the initial location of management. An unexpected finding was the protective association of male sex, which contrasts with global epidemiological trends and warrants further investigation.

The mean age of patients (49 years) was lower than that typically reported in high-income countries, where mean ages generally range from 55 to 65 years [17,18]. This difference reflects the younger demographic structure of low-income countries but also highlights that severe COVID-19 can occur at earlier ages. These findings suggest that clinical vigilance should not be limited to older age groups in similar settings. A slight male predominance was observed in the cohort (53%), consistent with international literature identifying male sex as a risk factor for severe COVID-19 [19]. However, multivariate analysis revealed an unexpected protective association of male sex (aOR = 0.12; 95% CI: 0.02-0.74), contrasting with the well-documented higher mortality among men reported globally. Several hypotheses may explain this finding. First, the limited number of events (14 deaths) may have produced unstable estimates, as reflected by the wide confidence interval. Second, context-specific factors may be involved: women in Anjouan may have experienced delays in seeking care or in referral, leading to more severe disease at hospital presentation. Third, the distribution of comorbidities or unmeasured confounders (such as pregnancy status, socioeconomic factors, or household roles affecting exposure) may differ between sexes in this population. Finally, differences in health-seeking behaviors between men and women in this cultural context could influence the timing and severity of presentation. These hypotheses warrant further investigation in larger, multicenter cohorts to determine whether this finding reflects a true local epidemiological pattern or a statistical artefact.

Analysis of transportation modalities showed that most patients were conveyed through emergency services, notably by fire brigade (57.6%) or ambulance (16.7%), while the use of private vehicles or public transportation remained limited. This observation highlights the central role of logistics and access to emergency transport in facilitating timely hospital presentation. Although no direct association with mortality was demonstrated in this study, previous research has shown that transportation barriers are associated with delays in access to care and poorer outcomes in resource-limited settings. Comparable findings have been reported in Ethiopia [20], where limited prehospital emergency services and inequities in ambulance availability were documented during the COVID-19 pandemic. Similarly, a systematic review emphasised transport logistics as a critical component of health system resilience in low- and middle-income countries [21]. Evidence from other settings has also shown that transportation barriers during the COVID-19 pandemic contributed to delayed care and poorer health outcomes [22]. Taken together, these findings support the integration of transport planning and early referral systems into pandemic preparedness strategies.

Approximately one-third of patients presented with at least one comorbidity, most commonly hypertension (34.8%), diabetes (33.3%), and obesity (19.7%). Although these conditions were not consistently associated with mortality in multivariate analysis (likely due to limited statistical power), their role in severe COVID-19 and adverse outcomes has been widely documented [23,24]. Given the high prevalence of chronic non-communicable diseases in the Comorian population, these findings underscore the importance of integrating chronic disease management into pandemic preparedness strategies. Previous studies conducted in low- and middle-income countries have shown that weak primary healthcare systems and poor control of chronic conditions increase vulnerability during infectious disease outbreaks [25]. Early identification and close monitoring of patients with comorbidities should therefore remain a priority, in both hospital and community settings. Strengthening primary healthcare services, ensuring early screening and appropriate management of conditions such as hypertension and diabetes, and training healthcare workers to recognise high-risk patients have been identified as effective approaches to mitigate severe outcomes during epidemic crises [26,27].

Initial management pathways were central to this analysis. More than half of patients (59.1%) were initially monitored at home before hospitalisation, including some presenting signs of severity such as tachypnea. While delayed admission has been associated with increased mortality in larger cohorts [28,29], the type of initial management was not independently associated with mortality in our study (OR = 1.98; p = 0.295). This finding suggests that outcomes may depend less on the initial location of care than on the timely recognition of clinical deterioration and prompt escalation to hospital-based management. In resource-limited settings where hospital capacity is constrained, home-based monitoring may represent a viable initial strategy provided that clear triage protocols are in place and simple monitoring tools, such as pulse oximeters, are available to detect early signs of deterioration. These results support the development of structured home monitoring programs with defined escalation criteria, rather than defaulting to hospitalisation for all confirmed cases.

Disease severity at admission emerged as the strongest independent predictor of mortality, with patients presenting severe or critical forms having a 14-fold higher risk of death compared to those with mild or moderate disease (aOR = 14.29; 95% CI: 2.11-96.93). This finding aligns with evidence from both African and international cohorts [30,31] and underscores the critical importance of early severity assessment. Tachycardia (heart rate >90 bpm) was also independently associated with mortality (aOR = 6.13; 95% CI: 1.16-32.29), consistent with its recognised role as a marker of physiological stress, systemic inflammation, and cardiovascular compromise in COVID-19 patients [28]. In settings where access to advanced diagnostic tools is limited, simple clinical parameters (respiratory rate, oxygen saturation, and heart rate) can effectively stratify risk and identify high-risk patients at the point of first contact.

In-hospital mortality in this cohort (21%) was comparable to rates reported in other African cohorts [30,32], but substantially higher than global averages [33]. Deaths often occurred early in the hospital course, among patients presenting with hypoxemia, tachycardia, or tachypnea, highlighting that delays in recognising severe cases may contribute to rapid clinical deterioration and death [34]. Multinational analyses have shown higher mortality among hospitalised patients in African facilities compared with non-African hospitals, reflecting systemic constraints such as limited access to intensive care, mechanical ventilation, and advanced therapeutics [35]. Conversely, in high-income countries, better-developed intensive care infrastructure, systematic early monitoring, and rapid escalation of care are associated with lower COVID-19-related mortality [36]. These comparisons indicate that the clinical trajectory of COVID-19 patients depends substantially on the health system's capacity to detect severity early and intervene promptly. They underscore the importance of investing in critical care infrastructure, ensuring the availability of oxygen and essential medicines, and developing context-adapted clinical protocols in low-resource settings.

Taken together, these findings highlight that health system preparedness for epidemic crises extends beyond resource availability. Effective crisis management also requires organising information flows, prioritising patients efficiently through structured triage, and ensuring early response capacity at the community level. The experience of Anjouan during the second COVID-19 wave illustrates both the challenges and the adaptive strategies that emerge in constrained environments. Beyond COVID-19, these findings provide transferable methodological and operational lessons for strengthening health systems in low-resource island settings facing recurrent public health threats, including cholera, dengue, and future respiratory pathogen outbreaks.

Strengths and limitations of the study: this study presents several important strengths. The exhaustive inclusion of all confirmed COVID-19 cases hospitalised during the study period, including those initially managed at home before admission, provides a comprehensive overview of local care pathways under epidemic pressure. The systematic comparison between initial management strategies and clinical characteristics offers valuable insights into triage practices in a resource-limited island context. The use of multivariate logistic regression allowed identification of factors independently associated with mortality while controlling for potential confounders. Finally, the operational relevance of the findings supports practical implications for improving early assessment, referral protocols, and clinical prioritisation in similar low-resource settings facing future epidemic surges. Several limitations should also be acknowledged. First, the relatively small number of patients (66 patients, including 14 deaths) limited statistical power to detect associations for certain well-established risk factors and contributed to wide confidence intervals. Second, the single-centre design limits generalizability to other settings. However, the study was conducted in the main referral hospital managing COVID-19 cases in Anjouan during the peak of the second wave, making the cohort representative of hospitalised patients in this context. Third, missing data for some variables may have introduced minor information bias, though overall data completeness was acceptable. Finally, the lack of post-hospitalisation follow-up prevented evaluation of medium- and long-term outcomes. Despite these limitations, the study provides a detailed snapshot of acute clinical trajectories and management patterns, offering valuable lessons for clinical decision-making and health system planning in similar settings.

Perspectives and recommendations: these findings highlight several priorities for strengthening epidemic response in similar settings. Consolidating local research capacity is essential; conducting multicenter studies across the Comoros archipelago or in collaboration with other East African island nations would allow validation of these observations in larger cohorts and improve understanding of regional specificities, thereby informing evidence-based public health policies. Improving frontline triage tools should also be a priority; developing and validating simple clinical checklists incorporating respiratory rate, heart rate, and oxygen saturation could facilitate earlier identification of high-risk patients and support clinical decision-making, even in settings with limited diagnostic capacity. Community-based monitoring programs deserve further evaluation; home follow-up protocols combining pulse oximetry with structured clinical reassessment or telemonitoring could help reduce delays in hospital referral and mitigate mortality, particularly during periods of hospital overcrowding. These lessons should be integrated into proactive pandemic preparedness plans, including frontline staff training on danger sign recognition, standardised triage protocols with clear escalation criteria, and strengthened management of chronic comorbidities. Finally, given the high prevalence of comorbidities observed in this cohort, strengthening chronic disease management may help reduce population vulnerability during future epidemics.

Conclusion     

This study found that in-hospital mortality among COVID-19 patients hospitalised in Anjouan during the second epidemic wave was 21%. Disease severity at admission and tachycardia were independently associated with death, while home-based monitoring before hospitalisation was not. An unexpected protective association of male sex was also observed, contrasting with global trends and warranting further investigation in larger cohorts. These findings suggest that clinical outcomes depend more on timely recognition of deterioration than on the initial location of care. Early severity assessment using simple clinical parameters (respiratory rate, heart rate, and oxygen saturation) can help identify high-risk patients and facilitate prompt escalation of care. Strengthening frontline triage protocols, expanding access to pulse oximetry, and reinforcing healthcare worker training on danger sign recognition may improve survival in similar resource-limited settings. The Anjouan experience offers valuable lessons for building more resilient health systems capable of responding effectively to future epidemic threats.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

Salim Irnat: conception and study design, data analysis and interpretation, manuscript drafting. Mohamed Anssoufouddine and Mouhamadi Abdalli Mari: data collection, data cleaning and database harmonisation, access to patient medical records. Ouassima Erefai: support in statistical analysis, interpretation of results, and manuscript revision. Fatine Hadrya, Abdelmajid Soulaymani, Abdelrhani Mokhtari, and Hinde Hami: methodological guidance, critical revision of the manuscript for important intellectual content, and contribution to the final interpretation of findings. All the authors read and approved the final version of this manuscript.

Tables     

Table 1: comparison of patient characteristics by initial management type

Table 2: univariate and multivariate logistic regression analysis of factors associated with in-hospital mortality

References