This was a Hospital-based retrospective cohort study, which involved the analysis of all COVID-19 patients admitted at KCMC from 10 ^th March 2020 to 26 ^th January 2022 which comprises of all four waves, to determine the mortality rate and predictors of mortality among COVID-19 patients at KCMC Hospital in Northern-Tanzania.

Data available on the excel isolation database was used, whereby the patient’s information on sociodemographic characteristics, clinical features and comorbidities was obtained by retrieving them from the electronic medical record files. The electronic files contain such information from the patients was filled in by the clinicians. A total of 1117 COVID-19 patients were retrieved from the electronic files, but only 547 confirmed patients records were complete and therefore available for analysis in this study.

All COVID-19 patients admitted at KCMC Hospital in Northern Tanzania from 10 ^th March 2020 to 26 ^th Jan 2022 whose information’s were available in the dataset. The study excluded participants whose information’s were incomplete such as unknown discharge status, date of admission and date of discharge. Data used in this study were collected between 10 ^th March 2020 to 26 ^th Jan 2022 by recollecting information from all admitted COVID-19 cases. A total of 1117 COVID-19 patients were retrieved from the electronic files, but only 547 confirmed patients records were complete and therefore available for analysis in this study again duplicate patients record were identified and removed before analysis ( Figure 1 ).

The dependent variable (event) was death due to COVID-19, and was defined based on the certified cause of death due to COVID-19 documented in the patient electronic files. Coded as 1 if the patient died and 0 If the patient survived to the discharge date.

Independent variables included patients socio-demographic characteristics, Clinical features at admission and comorbidities. The study’s selection and categorization of independent variables were based on previous literature. ²⁰

Patients background characteristics include age in years (<60 and ≥60 years), Sex (Male and Female) and Region (Arusha, Kilimanjaro and others) and were measured based on relative or patient self-report and presentation.

Clinical features were measured based on defined WHO cutoff point, Disease severity Oxygen saturation ≥94% on RA + No clinical or radiological evidence of pneumonia (mild), 90–94% on RA+ clinical or radiological evidence of non-severe pneumonia (moderate), 80–89% on RA+ clinical or radiological evidence of severe pneumonia (severe),whereby Fever, headache, nausea, Cough, Difficulty in breathing, Chest pain, generalized body weakness, sore throat, runny nose, smell loss, loss of taste, Abdominal pain and diarrhea were recorded and measured based on patient/relative report and clinician assessment (“Yes” if had any & “No” if had not).

Includes Hypertension, Diabetes, Liver disease, Asthma, Renal disease, Neurological disease, Immunodeficiency, Malignancy were measured if patient had history of comorbidities or was taking any anti-comorbidities drugs i.e. Antihypertensive Antidiabetic, Antiviral, Anticancer drug, and management he/she was receiving (“Yes” if had any & “No” if had not) while Abnormal chest x-ray and abnormal lung auscultation were defined based on radiological finding (“Yes” if abnormal & “No” if normal).

Data were cleaned, checked for consistency, quality, summarized, and analyzed using Stata version 15 (Stata corp. 2017. Stata statistical software: Release 15. College Station, TX: StataCorp LLC). In Descriptive statistics, numerical variables (age and Oxygen saturation at the time of attending hospital) were summarized by median (IQR) while categorical variables were summarized by frequency and percentages. Pearson Chi-square or Fisher’s exact (for <5 expected count in a cell) was used to test the association between mortality and socio-demographic, clinical and comorbidities features, A significance level of 0.05 was used. Kaplan-Meier survival estimate showing overall survival experience: and Nelson-Aalen cumulative hazard plots with Log-rank test was used to compare the mortality experience. Predictors of mortality were assessed by using survival models including, Cox, and Weibull regression. a Akaike Information Criterion and graphical test of Weibull hazard assumptions a straight line was obtained and hence prove that the Weibull model is adequate for this analysis, ²¹ Weibull regression was found to be best fit for this analysis. Weibull regression model which take into account a time variable was used to determine the predictors of mortality.

Crude Weibull regression was carried out to identify the association between mortality and sociodemographic, clinical and comorbidities variables. The Crude hazard ratio (CHR) and corresponding 95% CIs were reported whereby some variables of clinical significance based on literature such as hypertension, diabetes, age, sex, and those with a p-value of ≤0.1 were considered in the adjusted analysis. The Akaike information Criterion was used for model selection where the model with the smallest Akaike was considered as the best fit. Multivariable analysis of exposures had a significant association with the outcome and those with p-value of ≤0.1 were entered in the development of the final model. The variables with the p-value of ≤0.05 in the final model were considered as statistically significant. Adjusted hazard ratio (AHR) and corresponding 95% CIs were reported.

Ethical consideration

Permission letter to use COVID-19 data with reference No. KCMC/P.I/Vol.IX was granted from KCMC Hospital executive director office. Patient ID were used to maintain confidentiality.

Ethical approval to carry out the current study was obtained from the Kilimanjaro Christian Medical College Research Ethics and Review Committee (KCMU-CREC) with clearance number PG 15/2021.

Consent form

For secondary data analysis of existing COVID-19 patient records used retrospective, de-identified data, therefore, individual informed consent was not required, as the study qualified for a waiver of consent approved by the ethics committee in accordance with institutional and national ethical guidelines for minimal risk record-based research.

Among 547 patients who tested positive to COVID-19 and met inclusion criteria, the majority of them 324 (60%) were aged 60 years and above with a median age of 63 (IQR: 50–75). Of whom 547, More than half of them were males 310 (56.7%), ( Table 1 ). Most of them were present with the following clinical features fever 333 (60.8%), Severe form of disease 243 (44.4%), Difficult in Breathing 424 (73.3%), Chest pain 252 (46.1%), Generalized Body Weakness 390 (71.3%), more than half of the participants 375 (68.6%) were exposed during wave 2 & 3, their median oxygen saturation at the time of attending the hospital was 70% (IQR: 53–82), ( Table 2 ) majority of the participants 491 (90%) were from the Kilimanjaro region with abnormal chest x-ray 458 (83.7%), abnormal lung auscultation 274 (50.1%), hypertension 255 (46.6%) and diabetes 159 (29.1%) being the most common comorbidities ( Table 3 ).

Of 547 COVID-19 patients, 189 (34.6%) died and 358 (65.5%) were discharged alive. Their median survival time was 7 days (IQR: 3–12 Days). The overall mortality rate was 32.33 per 1000 person-day. Mortality occurred in 141 (43.5%) of the elderly patients (P < 0.001), 113 (36.5%) in male patients but was not statistically significant (P = 0.285) and 172 (35.0%) from Kilimanjaro region (P = 0.742), ( Tables 2 & 3 ).

Clinical features, Death was statistically higher in 140 (57.6%) of patients with severe form of disease (P < 0.001), 154(38.4%) among patients who present with difficulty in breathing (P = 0.002), 133 (35.5%) and 31 (43.7%) in patients exposed during wave 2 & 3 and 4 respectively (P = 0.03), 19 (25.7%) in patients who reported to loss taste but was marginally significant higher (P-value =0.084) and in 10 (22.7%) which was also marginally significantly higher among who reported to have diarrhea episodes (P-value 0.085) ( Table 2 ).

Comorbidities, Mortality also occurred in 66 (41.5%) patients with diabetes (P = 0.028), 172 (37.6%) in patients with abnormal chest x-ray (P < 0.001), 110 (40.2%) in patients with abnormal lung auscultation (P = 0.006), and 13 (54.2%) in patients with renal disease (P = 0.003) ( Table 3 ).

The Kaplan-Meier curve depicts overall survival among COVID-19 patients admitted to KCMC Hospital over the course of the study. Nelson-Aalen cumulative hazard plots and the log rank test reveal differences in mortality experience. The mortality rate per 1000 person-day was (MR = 46.7, 95%CI; 39.4–55.4) in patients over 60, and (MR = 37.1, 95%CI; 30.8–44.5) in male patients. Furthermore, patients with moderate and severe forms of the disease had mortality rates of (MR = 19.2, 95%CI; 13.3–27.4) and (MR = 55.0, 95%CI; 46.4–65.2), respectively. The mortality rate among patients presenting with breathing difficulties was (MR = 35.4, 95%CI; 30.1–41.6), and in diabetic patients was (MR = 39.4, 95%CI; 30.7–50.6). The mortality rate in hypertensive patients was (MR = 36.9, 95%CI; 30.0–45.2). Patients with abnormal chest x-rays and lung auscultations had a mortality rate of (MR = 34.5, 95%CI; 29.6–40.2) and (MR = 37.8, 95%CI; 31.2–45.9), respectively, and those with renal disease had a mortality rate of (MR = 37.8, 95%CI; 31.2–45.9) ( Figure 2 ).

The predictors associated with COVID-19 mortality are displayed in ( Tables 4, 5, and 6 ). In crude analysis, the hazard of mortality among patients aged ≥60 years was 2.62 times significant higher, compared to those aged <60 years (CHR = 2.62; 95%CI 1.88–3.66; P < 0.001). Males had 1.35 times significant higher the hazard of mortality compared to females (CHR = 1.35; 95%CI 1.00–1.83; P = 0.049). Compared to the mild form of the disease, patients with moderate form had a 1.91 times higher hazard of mortality (CHR = 1.91; 95%CI 1.03–3.53; P < 0.041). Compared to the mild form of the disease, patients with the severe form had a 5.54 times higher hazard of mortality (CHR = 5.54; 95%CI 3.25–9.44; P < 0.001). Patients who present with Difficult breathing had 1.56 times the significant higher hazard of mortality compared to those without (CHR = 1.56; 95%CI 1.06–2.28; P = 0.023), compared to patients living without diabetes, diabetes patients had 1.33 times the significant higher hazard of mortality (CHR = 1.33; 95%CI 0.98–1.82; P = 0.066). Patients found with abnormal chest X-rays had 1.75 times the significant higher hazard of mortality compared to patients without abnormal chest x-ray, (CHR = 1.75; 95%CI 1.06–2.88; P = 0.028). Patients with abnormal lung auscultation had 1.41 times the significant higher hazard of mortality compared to a patient with normal lung auscultations (CHR = 1.41; 95%CI 1.05–1.89; P = 0.024). Lastly, compared to patients without renal disease, patients with renal disease had a 1.89 times higher hazard of mortality (CHR = 1.89; 95%CI 1.08–3.33; P = 0.027).

In adjusted analysis, Higher significant hazard of mortality were among elderly aged ≥60 years (AHR = 2.01; 95%CI 1.41–2.87; P < 0.001), Male patients (AHR = 1.27; 95%CI 0.93–1.73; P = 0.128), Patients diagnosed with moderate (AHR = 1.67; 95%CI 0.89–3.13; P = 0.112) and severe form of disease (AHR = 4.44; 95%CI 2.56–7.73; P < 0.001), those exposure during wave 2&3 (AHR = 1.58; 95%CI 1.03–2.41; P = 0.036) and in diabetes patients (AHR = 1.25; 95%CI 0.88–1.79; P = 0.212) ( Table 7 ).

More attention to older men patients as the findings of this study shows being older male increase the risk of mortality.

Ensure that standard and supportive care are also available at the primary health facilities to allow early intervention to take place among patients diagnosed with the severe form of the disease because some patients were transferred late to referral hospitals which turnout in poor clinical outcome.

A further large multicenter quantitative study should be carried out by using both signs, symptoms, pre-existing comorbidities, laboratory parameters, and radiological and pathological findings to identify the potential predictors of mortality among COVID-19 patients, thus can allow external validation. Furthermore, a clinical trial study is needed to explore the sex differences in susceptibility, severity and outcomes of coronavirus diseases in 2019.