Predictors of 30-day and 90-day mortality among hemorrhagic and ischemic stroke patients in urban Uganda: a prospective hospital-based cohort study

We conducted a prospective hospital-based cohort study between December 2016 and March 2019 to investigate predictors of 30-day and 90-day mortality and functional disability among HS and IS patients at Nsambya Hospital, Kampala, Uganda. The hospital is a large urban tertiary referral hospital, which lies approximately 5 km southeast of Kampala, Uganda’s capital city. It receives patients from all parts of the country and offers medical care in all disciplines. Most of the patients pay personal private fees for treatment, while a minority has medical insurance covering the costs. Nsambya Hospital has around 19, 000 admissions per year and receives about 300 outpatients per day.

The participants were patients who presented at Nsambya hospital- Kampala within 7 days of the onset of stroke symptoms, during the study period and, who met the WHO stroke definition: rapidly developing clinical signs of focal or global disturbance of cerebral function, lasting for more than 24 h or until death, with no apparent non-vascular cause [19] with neuro imaging confirmed stroke results.

The eligibility criteria for participation in the study included: 1) being an adult aged ≥18 years old; 2) confirmed cases of IS and HS on brain neuro-imaging; and 3) first-ever or recurrent stroke. We excluded participants who were: 1) unable to consent or for whom consent could not be obtained from a caregiver; (2) unable to communicate and without a caregiver respondent; (3) those who died within 24 h on the ward before neuroimaging was done; and 4) those who presented more than 7 days after onset of symptoms.

All consecutive patients aged ≥18 years old with a confirmed diagnosis of HS or IS stroke were enrolled prospectively over a 2-year period. Both first-ever stroke and recurrent stroke patients were included. The study was based on the standardized WHO stepwise approach to stroke surveillance (STEPS) [20]. All information was collected by a trained medical team according to the STEPS Stroke Manual Instructions [20]. Within 24 h of admission, we obtained information about events that occurred within 7 days after the stroke event. Data on social demographic characteristics, social history, past medical history, stroke severity, pre-stroke functional ability, clinical characteristics and family history was collected. If patients were unable to respond to questions, family members served as proxy respondents. All participants underwent physical examination and assessment including recording blood pressure, physical function using a mRS [21] and neurological examination using the NIHSS [21] and Glasgow Coma Scale (GCS). Computed tomography (CT) scan was performed in all patients within the first 48 h after admission. Stroke types were differentiated based on CT scan interpretations into IS and HS.

All study participants were followed up prospectively at 30 days and 90 days by the study team. Participants were given an appointment to attend the neurology outpatient clinic post-stroke. Patients or their family members were contacted to remind them about the appointment date. Patients who failed to attend the follow-up visits were contacted and interviewed by telephone and, if not possible, their relatives were contacted, or a home visit was carried out. Information about functional status at 30 and 90 days using the mRS, stroke recurrence, mortality, mortality date and cause were recorded. Patients who had missed their day 30 or day 90 follow-up visits and did not present at the clinic within 30 days, and could not be contacted by phone, were considered as ‘lost-to-follow up’.

The primary outcome of the study was mortality at 30 and 90 days, and mortality was defined as the number of deaths among stroke patients which occurred within these time limits. The primary outcome was assessed using the mRS. For each patient, social demographic data (age, sex, level of education and marital status), information about presence of hypertension, DM, smoking habits, alcohol consumption, HIV infection, past medical history and a family history of Hypertension, DM and stroke were obtained. Hypertension was defined as either current use of antihypertensive medication or history of being diagnosed with hypertension prior to stroke or documented blood pressure of greater than or equal to 140 mmHg systolic or 90 mmHg diastolic. Diabetes mellitus was defined as current use of antidiabetic drugs, diagnose of type I or type II diabetes before stroke or a documented non-fasting blood glucose of greater than 11.1 mmol/L or fasting blood glucose of greater than 7.0 mmol/L outside of the acute phase of stroke (to exclude acute transient elevation of glucose as a stress response after stroke). Premorbid functional status was assessed using the mRS: (i) mRS 0–2 was considered good, (ii) mRS 3 was fair, and (iii) 4–5 indicated poor outcome. The length of hospital stay (LOS) for a single stroke hospitalization was defined as the time spent in hospital from admission until death or discharge to home. The severity of the neurological deficit at admission was assessed using the NIHSS. Higher figures correspond to higher stroke severity: (i) NIHSS 0–6 was considered mild; (ii) 7–12 was moderate; (iii) 13–20 was severe; (iv) ≥21 was very severe stroke. Level of consciousness was assessed using the GCS: (i) Good GCS (13–15); mild brain injury (alert), (ii) moderate GCS (9–12); moderate brain injury (drowsy), (iii) poor GCS (< 9); severe brain injury (unconscious).

A sample size of 140 participants was needed to assess the primary outcome (30-day and 90-day mortality) of this study. The sample size was based on an anticipated 35% mortality for each of the two mortality measures [13]. An 80% power, 5% level of significance, 95% confidence interval were assumed. The sample was adjusted for a 10% rate of non-response. All eligible participants were consecutively enrolled during the study period.

Data was double entered in OpenClinica (OC), cleaned, and exported to Stata15.0 (StataCorp, College Station, TX, USA) for analysis. We resolved discrepancies by checking the source documents for clarification on all possible confounders. Categorical demographic and clinical characteristics were summarized by counts and percentages. Continuous variables were summarized by means and standard deviations or medians and interquartile ranges. We used the Kaplan-Meier technique to estimate time to mortality after admission and presented the estimated mortality rates per 1000 person-days for the study variables. To compare the mortality distributions between the groups for the different variables of interest, a log rank test was used. Mortality rate was assessed within 30 days and at 90 days of follow-up. Cox proportional hazards regression was used to determine independent predictors of mortality among participants and these were expressed as estimated hazard ratios (HRs) with their corresponding 95% confidence intervals (CIs). In the bivariate proportional hazards analysis, variables significant at p ≤ 0.15 [22] were subsequently considered for the adjusted model. We considered age and gender as priori confounders; these were included in the adjusted model regardless of their unadjusted p values. The P- values represented the differences in the variable sub categories for 30- and 90-day mortality.

During the study period, 153 stroke patients attending Nsambya hospital were screened for eligibility. Of these, 12 (7%) were excluded: those who died within 24 h before CT scan was done (n = 4); those who had unspecified stroke (n = 2); those who had transient ischemic attacks (n = 4) and those who presented more than 7 days after onset of symptoms (n = 2). Thus a total of 141 patients with confirmed IS or HS that consented for enrolment were included, of these five were lost to follow up before outcome assessment at 30 days, 77 completed the 90 days of follow-up, a total of 44 patients died within 90 days post-stroke period (31 died between 0 and 30 days, while 13 died between > 30–90 days); 15 were lost to follow up between 30 and 90 days (Fig. 1).

Of the 141 participants included in the analysis, 48 (34%) were male, mean age was 63.2 (+ 15.4) years old, 90 (64%) had IS, while 51 (36%) had HS; 65 (46%) attained secondary education and above; 81 (57%) were ≥ 60 years. The mean length of hospital stay was 9.4 (SD ± 10.7) days. Median time between the onset of symptoms and medical attention for IS patients was higher (2 days (Interquartile range [IQR]: 1–4 days) compared to HS (1 day [IQR]: 1–5 days) (Table 1). Overall mortality was 44 (31%); 31 (23%) patients died within the first 30 days post-stroke and, an additional 13 (14%) died within 90 days post-stroke. The mean follow-up duration was 61 (SD ± 34) days, with a total of 8274 person days of follow up at 90 days. The overall mortality for HS was 19 (37.3%) and 25 (27.8%) for IS (Table 2). Figure 2 shows cumulative mortalities during the 3 months follow-up period according to stroke type. We observed a statistically significant higher mortality rates among females compared to males at 30 days (p = 0.039). Thirty-day and 90-day mortality increased with increasing age at diagnosis (p = 0.036). In unadjusted analyses, individuals aged ≥60 years old, females, patients without formal education, patients who were divorced or separated, severe stroke NIHSS > 21, longer hospital stay > 14–29 days, low GCS < 9 and HS type presented higher mortality rates across 90-day follow up period compared to other subgroups (Table 1).

The majority 132(93%) of patients were fully independent (mRS 0–2) before stroke. On admission, mean NIHSS was 15 (SD ±10.1) and 42(30%) of the patients had severe stroke (NIHSS ≥21). The mean GCS was 12 (SD ±2.8), and only 21 (15%) had GCS < 9 in all stroke, majority with a GCS < 9 died within 1 month; 58 (41%) had severe disability (mRS 4–5) at 30 days post-stroke and 38 (36%) at 90 days (Table 1).

Univariate analysis showed that age ≥ 60 years (uHR = 2.3, 95% CI: 1.03–5.16), smoking (uHR = 4.64, 95% CI: 1.11–19.4), stroke severity (NIHSS > 21) (uHR = 5.9, 95% CI: 2.33–14.9), prolonged hospital stay of ≥14 days (uHR = 2.64, 95% CI: 1.17–6.00), lower GCS < 9 (uHR = 4.71, 95% CI: 2.24–9.87), and poor functional status at enrolment (mRS 4–5) (uHR = 4.89, 95% CI: 1.17–20.5), significantly increased the risk of mortality at 30 days. Age ≥ 60 years old (uHR = 1.95 (95% CI: 1.02–3.73), longer hospital stay of 14–29 days (uHR = 3.2 (95% CI: 1.53–6.78), lower GCS < 9 (uHR = 4.33, 95% CI: 2.24–8.36), stroke severity (NIHSS ≥21) (uHR = 4.66 (95% CI: 2.26–9.58) and poor functional status at enrolment (mRS 4–5) (uHR = 3.58, 95% CI: 1.28–10.0) increased the risk of mortality at 90 days (Table 3).

After adjustment for age and sex in a multivariable analysis, lower GCS < 9 (aHR =3.49 (95% CI: 1.39–8.75) was a significant predictor of 30-day mortality, whereas lower GCS of < 9 (aHR =4.34 (95% CI: 1.85–10.2), stroke severity (NIHSS ≥21) (aHR) = 2.63 (95% CI: (1.68–10.5) and haemorrhagic stroke type (aHR = 2.30, 95% CI: 1.13–4.66) were significant predictors of 90-day mortality. Shorter hospital stay (< 14 days) (aHR = 0.31 (95% CI: 0.11–0.93) and being married (aHR = 0.22 (95% CI: 0.06–0.84) were associated with reduced 30- and 90-day mortality respectively (Table 3).

Measures for secondary prevention of stroke, particularly proper treatment, referral, and intensive follow-up including drug compliance are desperately needed in Africa. We hope that taking into account the predictors for mortality for each stroke type in the present study, individual patient management can be better optimized and outcomes improved in Uganda. Further studies to examine the relationship between the type of stroke and immediate cause of death are warranted. The knowledge of the immediate cause of death and other contributory factors will greatly enhance effective management of acute stroke.