Risk Factors and Outcomes of Hospitalized Stroke Patients in Lusaka, Zambia

Abstract

Background:

Limited data exists about stroke risk factors and outcomes in sub-Saharan African countries, including Zambia. We aim to fill this gap by describing features of hospitalized stroke patients at University Teaching Hospital (UTH), the national referral hospital in Lusaka, Zambia.

Methods:

We conducted a retrospective study of consecutive adults with stroke admitted to UTH’s inpatient neurology service from October 2018 to March 2019. Strokes were classified as ischemic or hemorrhagic based on CT scan results and unknown if CT scan was not obtained. Chi-square analyses and t-tests were used to compare characteristics between cohorts with differing stroke subtypes.

Results:

Adults with stroke constituted 43% (n=324) of all neurological admissions, had an average age of 60±18 years, and 62% of the cohort was female. Stroke subtypes were 58% ischemic, 28% hemorrhagic, and 14% unknown. Hypertension was present in 80% of all strokes and was significantly associated with hemorrhagic stroke (p=0.03). HIV was present in 18% of all strokes and did not significantly differ by stroke subtype. Diabetes (16%), heart disease (34%), atrial fibrillation (9%), and past medical history of stroke (22%) were all significantly more common in patients with ischemic stroke (p<0.05). In-hospital mortality was 24% overall and highest among individuals with hemorrhagic strokes (33%, p=0.005).

Conclusions:

This Zambian stroke cohort is notable for its young age, significant HIV burden, high in-hospital mortality, and high rates of uncontrolled hypertension. Our results demonstrate Zambia’s substantial stroke burden, significant contribution of HIV to stroke, and the need to improve primary stroke prevention.

Introduction

Stroke is one of the leading causes of disability and mortality worldwide. It accounts for 42% of disability-adjusted life years attributed to neurological disorders¹ and causes long-standing disability in nearly 50% of survivors.² While stroke incidence is decreasing in high-income countries,^3,4 it is significantly increasing in low- and middle-income countries (LMICs).⁵ This includes most countries in sub-Saharan Africa (SSA), such that the African continent is posited to have the highest incidence of stroke worldwide.⁶ Stroke prevalence is only expected to continue to increase in SSA as the African population ages and the burden of traditional stroke risk factors such as hypertension and diabetes becomes more prevalent.^7,8 While SSA shoulders a disproportionate burden of stroke, it also continues to have the fewest neurologists of any world region to counter this substantial and rising stroke burden.^{9, 10} The paucity of neurologists and neurological research capacity in SSA often precludes systematic collection of neuroepidemiological data, including stroke, and the development of locally relevant interventions to improve stroke care.

Zambia is an LMIC in southern Africa with a high burden of stroke such that it is the eighth leading cause of death nationally, surpassed primarily by infectious causes of death including HIV/AIDS and Tuberculosis, but also ischemic heart disease.¹¹ These data illustrate Zambia’s double burden of disease with rising rates of non-communicable diseases in the midst of ongoing high rates of infectious diseases.

Zambia’s first-ever post-graduate neurology training program and dedicated neurology inpatient service were begun in 2018. This newly developing system of specialty neurological care allows for a better understanding of local stroke epidemiology and stroke systems of care. We aimed to understand the characteristics, including stroke subtypes, risk factors, and outcomes, of hospitalized stroke patients at the University Teaching Hospital (UTH), the national referral hospital in the capital city of Lusaka, during the initial development of specialized neurology care and training in Zambia.

Methods

Study Population and Setting.

We completed a retrospective cohort study of all individuals ≥18 years old admitted to the inpatient neurology service at UTH with a suspected or confirmed clinical diagnosis of stroke made by post-graduate neurology trainees or consultant neurologists between October 2018 and March 2019. UTH is a 1,600 bed hospital that serves as Zambia’s national referral hospital to which patients are referred from across the country for specialized inpatient and outpatient care. It is a government hospital, so it is part of Zambia’s subsidized healthcare system. Patients admitted to UTH do not pay a daily bed fee or consultant fees. Basic laboratory investigations are provided free of charge, but patients must pay for more advanced investigations, including imaging studies, before the study can be completed. Medications stocked in the hospital pharmacy are also free of charge. Zambia also recently instituted a national health insurance scheme, but this study was conducted prior to its inception.

At the time of this study, standard of care for patients admitted to UTH with suspected stroke included, finances permitting, non-contrast CT brain, electrocardiogram, echocardiogram, rapid HIV test, lipid panel, fasting blood glucose or hemoglobin A1c, and rapid plasmin reagin (RPR). No acute interventions for stroke (i.e. intravenous tissue plasminogen activator, endovascular therapies) were available, and limited neurosurgical interventions, including hematoma evacuation and external ventricular drain, could be considered in limited cases. Vascular imaging was not routinely obtained as vascular interventions such as carotid endarterectomy and carotid stenting are not available in Zambia. Routine blood pressure monitoring was obtained with thrice daily vital signs, but no cardiac monitors were available for non-ICU patients. Intravenous blood pressure medications were commonly unavailable, so hypertension was generally treated with oral agents such as nifedipine or lisinopril. Aspirin was the only routinely used antiplatelet and was available through the hospital pharmacy. Statin medications and anticoagulants generally had to be purchased through private pharmacies.

Study Procedures.

Retrospective chart review was conducted for all patients meeting the criteria above. Extracted data included demographics, length of stay, clinical presentation, medical co-morbidities, family history, social history, pertinent findings from neurological examinations performed by neurology trainees or neurologists, results of laboratory and imaging investigations, final stroke diagnoses, and vital status at discharge. Complications such as fever, aspiration pneumonia, and intensive care unit (ICU) intervention were also recorded. Stroke types were defined as ischemic or hemorrhagic based on CT imaging. Strokes were classified as unknown if there was no CT scan obtained during admission due to CT scanner malfunctioning, patients’ inability to afford a CT scan, or patients’ death prior to imaging. Etiologies specific for hemorrhagic strokes and ischemic strokes were separately classified based on the treating neurologist’s clinical impression after results of routine clinical investigations and imaging were obtained.

Standard Registrations and Ethical Approvals.

This study was approved by the Johns Hopkins Institutional Review Board, the University of Zambia Biomedical Research Ethics Council, and the Zambia National Health Research Authority. Because no identifiable patient information was extracted, a waiver of consent was obtained.

Risk factor definitions.

Hypertension was defined as a self-reported history of high blood pressure, anti-hypertension medication use, or blood pressures requiring anti-hypertensive medications to maintain a level ≤ 140/90 mmHg more than 72 hours after stroke.¹² Diabetes was defined by self-report, diabetes medication use, or standard laboratory cutoffs including HbA1c ≥6.5% or fasting blood glucose level ≥7mmol/L (126 mg/dL), or random blood glucose level ≥11mmol/L (200 mg/dL).^13,14 Hyperlipidemia was defined by self-report, statin use, or standard laboratory cutoffs including low density lipoprotein (LDL) >130mg/dl, high density lipoprotein (HDL) <50mg/dL, total cholesterol >200 mg/dL, or triglycerides >150mg/dL.¹⁵ Heart disease was defined by self-report or electrocardiogram (ECG) and/or echocardiography (Echo) imaging showing evidence of cardiac abnormalities including heart failure, valvular disease, etc. Atrial fibrillation was defined by self-reported history or ECG or Echo confirming atrial fibrillation.

Hospitalization details definitions.

Altered mental status was defined as confusion or disorientation, somnolence or reduced level of consciousness during an examination performed by neurology trainees or consultant neurologists and was noted as present at stroke onset if reported by patient or caregiver. Fever was defined as temperatures ≥38°C. Aspiration pneumonia was defined as either a witnessed aspiration event recorded in the patient’s file by nursing or medical staff or antibiotic medications prescribed for aspiration pneumonia during the patient’s hospitalization. Antibiotics were sometimes prescribed after definitive evidence of aspiration pneumonia were obtained (e.g. chest x-ray) but were also often prescribed empirically based on clinical suspicion alone due to resource limitations in obtaining chest x-rays and other laboratory investigations to confirm the diagnosis.

Stroke subtype categories.

Ischemic strokes were categorized based on non-contrast CT scan appearance. Strokes were classified as large artery strokes when at least half of a large vessel territory demonstrated ischemic changes on CT scan. Ischemic strokes were classified as cardioembolic strokes when infarcts were present in multiple vascular territories or when large vessel strokes were present in the setting of confirmed atrial fibrillation. Ischemic strokes were classified as lacunar strokes when small areas of ischemia did not correspond to a large vessel vascular distribution and were demonstrated in the subcortical or brainstem regions. Hemorrhagic strokes included intracerebral hemorrhages categorized based on location of bleeding (i.e. deep structures such as the basal ganglia, thalamus, and internal capsule; lobar, brainstem, and cerebellum) and subarachnoid hemorrhages. Hemorrhagic transformation of ischemic strokes were classified as ischemic strokes.

Statistical Analyses.

Study data were collected and managed using REDCap electronic data capture tools hosted at Johns Hopkins University School of Medicine.^{16, 17} Descriptive statistics are presented as means and standard deviations for parametric continuous variables, medians and interquartile ranges for non-parametric continuous variables, and frequencies for categorical variables. Analysis of variance, Kruskal-Wallis tests and chi-square or Fisher’s exact tests were used to compare each type of variable, respectively, between stroke types. Stata 14 (College Station, Texas)¹⁸ was used for all analyses, and the threshold for statistical significance was set at p<0.05.

Results

Admissions for stroke accounted for 43% (n=324) of all inpatient neurology admissions during the study period. Average age was 60±18 years, and the majority were women (n=200, 62%) (Table 1). Ischemic strokes were most common (n=188; 58%) followed by hemorrhagic (n=91, 28%) and unknown (n=45, 14%) strokes. Participants with hemorrhagic stroke were significantly younger (53 ± 14 years) than those with ischemic (62 ± 18 years) and unknown (65 ± 15 years) strokes (p<0.001). Participants with hemorrhagic stroke were less likely to be female (49%, n=45) than those with ischemic (63%; n=118) and unknown (82% female; n=32) strokes (p<0.001).

TABLE 1.

Demographic characteristics and cerebrovascular risk factors are presented for the overall cohort and then compared between stroke subtypes.

	Overall (n=324)		Ischemic Stroke (n=188)	Hemorrhagic Stroke (n=91)	Unknown Stroke (n=45)	p
Age (years) [mean (SD)]	60 (18)		62 (18)	53 (14)	65 (15)	<0.001
Female [n (%)]	200 (62%)		118 (63%)	45 (49%)	32 (82%)	<0.001
Risk Factors [n (%)]
Hypertension	259 (80%)		142 (76%)	81 (89%)	36 (80%)	0.03
New diagnosis	53 (16%)		25 (13%)	25 (27%)	3 (7%)	0.003
Taking BP medications (% of known diagnosis, n=206)	93 (45%)		59 (50%)	22 (39%)	12 (36%)	0.21
Diabetes	51 (16%)		36 (19%)	6 (7%)	9 (20%)	0.02
New diagnosis	22 (7%)		14 (7%)	5 (6%)	3 (7%)	0.91
Taking diabetes medications (% of known diagnosis, n=29)	19 (66%)		15 (68%)	0(0%)	4 (67%)	0.44
Hyperlipidemia	44 (14%)		28 (15%)	12 (13%)	4 (9%)	0.57
New diagnosis	40 (12%)		24 (13%)	12 (13%)	4 (9%)	0.81
Heart Disease	111 (34%)		77 (41%)	25 (27%)	9 (20%)	0.008
New diagnosis (n=114 with ECHO or ECG)	105 (32%)		72 (38%)	8 (9%)	9 (20%)	<0.001
Atrial Fibrillation	30 (9%)		26 (14%)	1 (1%)	3 (7%)	0.001
New diagnosis	14 (4%)		11 (6%)	0 (0%)	3 (7%)	0.02
Prior stroke	71 (22%)		49 (26%)	11 (12%)	11 (24%)	0.03
HIV Infection	58 (18%)		35 (19%)	18 (20%)	5 (11%)	0.6
New diagnosis (n=56)	7 (12%)		3 (9%)	2 (11%)	2 (40%)	0.21
Taking antiretroviral therapy (% of those with HIV, n=58)	48 (83%)		31 (98%)	15 (83%)	2 (40%)	0.046
Current alcohol use	47 (16%)		26 (15%)	17 (20%)	4 (10%)	0.59
Current tobacco use	26 (8%)		14 (8%)	8 (9%)	4 (10%)	0.89
Family history of stroke	48 (15%)		31 (16%)	10 (11%)	7 (16%)	0.5
Family history of hypertension	135 (42%)		80 (42%)	40 (44%)	15 (33%)	0.46
Family history of diabetes	49 (15%)		31 (16%)	12 (13%)	6 (13%)	0.8

Hypertension was the most prevalent risk factor among all stroke types (80%) and was significantly more common among participants with hemorrhagic strokes than ischemic or unknown strokes (Table 1). A new diagnosis of hypertension at time of hospitalization was also significantly associated with hemorrhagic stroke (p=0.003). One-fifth of patients with hypertension (and 16% of the overall cohort) were newly diagnosed at the time of their stroke, and only 45% of patients with known hypertension were on medication at the time of their stroke.

HIV infection was present in 18% of participants, making it the fourth most common risk factor identified, but its prevalence did not vary between stroke subtypes (Table 1). Among people with HIV (PWH), 12% were newly diagnosed with HIV at the time of their stroke presentation, and 83% were taking antiretroviral therapy (ART) at the time of their stroke. ART use was significantly higher among individuals with ischemic stroke compared to those with hemorrhagic and unknown strokes (p=0.046). In the overall cohort, diabetes was present in 16% of all participants but was significantly more prevalent amongst individuals with ischemic strokes (p<0.02), as was heart disease (p=0.008), atrial fibrillation (p<0.001), and a prior history of stroke (p=0.03).

Clinical presentation also varied by stroke type. Individuals with hemorrhagic stroke were significantly more likely to present with altered mentation at stroke onset (47% hemorrhagic vs 27% ischemic vs 15% unknown stroke, p<0.001) and at 24 hours after stroke onset (20% hemorrhagic vs 4% ischemic vs 11% unknown strokes, p<0.001). Individuals with hemorrhagic stroke and unknown stroke also had significantly higher systolic blood pressures at the time of presentation (176±34 mmHg and 171±32 mmHg respectively) compared to adults with ischemic strokes (157±36 mmHg, p<0.001) (Table 2).

TABLE 2.

Clinical presentation and initial examination findings for the overall cohort and by stroke subtype

Stroke characteristics	Overall (n=324)		Ischemic Stroke (n=188)	Hemorrhagic Stroke (n=91)	Unknown Stroke (n=45)	p
Clinical presentation
Altered mental status at onset per history	84 (26%)		12 (27%)	43 (47%)	29 (15%)	<0.001
Symptom progression after onset	26 (8%)		16 (8%)	9 (10%)	1 (2%)	0.28
Headache 2 hours after onset	50 (15%)		24 (14%)	23 (25%)	3 (7%)	0.007
Examination
Altered mental status on examination [n (%)] (n=244)	146 (60%)		76 (53%)	46 (70%)	24 (69%)	0.04
Initial Systolic BP [mean (SD)] (n=290)	165 (36)		157 (36)	176 (34)	171 (32)	<0.001
Initial heart rate [mean (SD)] (n=264)	90 (20)		90 (20)	87 (25)	91 (21)	0.5

Ischemic strokes were primarily large artery strokes (40%, n=76) followed by lacunar strokes (36%, n=68), and, lastly, strokes suspected to be from embolic sources (23%, n=44) (Figure 1). Hemorrhagic strokes were primarily from hypertension (86%, n=78), with fewer due to aneurysmal hemorrhage (4%; n=4), cerebral amyloid angiopathy (4%; n=4), and other factors (5%, n=5).

FIGURE 1.

Further classification by stroke subtype showing ischemic stroke by type of vessel involved (1A) and hemorrhagic stroke by hemorrhage location (1B).

Post-stroke complications in our stroke cohort were common, including aspiration pneumonia (16%) and fever (16%) (Table 3). However, complication rates were not significantly different across stroke subtypes. In-hospital mortality was 24% (n=77) for all strokes. Mortality was significantly higher for individuals with hemorrhagic strokes (n=29, 33%) and unknown strokes (n=15, 33%) compared to those with ischemic strokes (n=33, 18%, p<0.001).

TABLE 3.

In-hospital complications and in-hospital mortality for overall stroke cohort and by subtype.

	Overall (n=324)		Ischemic Stroke (n=188)	Hemorrhagic Stroke (n=91)	Unknown Stroke (n=45)	p
In-Hospital Mortality
In-hospital mortality	77 (24%)		33 (18%)	29 (33%)	15 (34%)	0.005
In-Hospital Post-Stroke Complications
Cared for in ICU [n (%)]	4 (1%)		2 (1%)	2 (2%)	0 (0%)	0.78
Fever	50 (15%)		27 (14%)	16 (18%)	7 (16%)	0.78
Aspiration pneumonia	51 (16%)		31 (16%)	13 (14%)	7 (16%)	0.93
NG tube	65 (20%)		38 (20%)	17 (19%)	10 (22%)	0.89

Discussion

The average age in our stroke cohort (60 ± 18 years) was much younger than stroke cohorts in high-income countries, such as European countries that report average stroke ages between 71 and 75 years of age¹⁹, and the United States, where three-quarters of strokes are in people over the age of 65 (compared to 40% of participants in our Zambian cohort).²⁰ However, this finding was similar to other cohorts from tertiary hospitals in the SSA region, including cohorts from Zimbabwe²¹, Malawi²², Ghana²³ and Uganda²⁴ in which average age ranged from 60 to 64 years. The discrepancy in age of stroke onset among SSA cohorts compared to those from high-income settings can likely be attributed to two primary factors: the high HIV prevalence in SSA countries and higher rates of uncontrolled hypertension that result in a greater proportion of hemorrhagic strokes which often present in younger people. However, novel and non-traditional risk factors such as diet²⁵, indoor air pollution, and stress²⁶ may also be contributing factors and should be further systematically evaluated in these settings.

HIV was the fourth most common risk factor for stroke in this cohort, and its prevalence did not vary between stroke types. HIV has been established as an independent risk factor for both ischemic and hemorrhagic strokes and is thought to double the risk of each.^{27, 28} The exact mechanism by which HIV increases stroke risk remains unclear. Postulated mechanisms include vasculopathy, accelerated atherosclerosis, and central nervous system opportunistic infections.²⁹ This is further complicated by the possible metabolic consequences of ART such as endothelial dysfunction and hyperlipidemia. HIV prevalence in our cohort was 1.5 times the national HIV prevalence, again demonstrating that stroke is more prevalent among PWH than the general population and its mechanisms deserve further study.

Several limitations complicate understanding how HIV increases risk for stroke. Many studies investigating potential mechanisms have been conducted in high-income countries where HIV prevalence in the general population and in stroke cohorts is much lower than in SSA.³⁰ PWH in high-income settings often have greater access to antiretroviral therapy which is often begun at higher nadir CD4 T cell counts and often have higher rates of comorbidities such as diabetes, obesity and injection drug use that further contribute to their increased risk of stroke.³¹ As such, it becomes difficult to isolate the effect of HIV on stroke in high-income settings. Furthermore, many prior studies of HIV-associated stroke conducted in low-income settings have occurred in the pre-ART era³², thus increasing the odds of stroke due to central nervous system opportunistic infections rather than the direct effects of HIV.³³ As such, cohorts such as ours with high rates of HIV infection, higher rates of ART use, and relatively lower rates of comorbidities such as diabetes, hyperlipidemia and prior drug abuse represent ideal populations in which to further investigate mechanisms of HIV-associated stroke in the ART era.

Hypertension is the most prevalent risk factor for stroke in both high- and low- and middle-income countries globally, regardless of country income level.^{34, 35} In high-income countries, hypertension is prevalent in the range of 60–80% of patients with stroke.³⁶ In our cohort, hypertension was present in 80% of patients which was similar to the high prevalence reported in other SSA countries, such as Nigeria and Ghana (95%)¹³ and Malawi (74%)³⁸ but substantially higher than previously reported in Zimbabwe (58%)²³.

In both high-income countries and LMICs, uncontrolled hypertension is significantly associated with hemorrhagic strokes.³⁹ The high prevalence of lacunar strokes in our cohort also suggests cerebral small vessel disease is a major contributor to ischemic strokes in Zambia. However, hemorrhagic strokes are higher in frequency in LMICs than high-income countries largely reflecting the disparity in achieving adequate blood-pressure control.⁴⁰ In high-income settings, a significantly higher proportion of individuals with hypertension have well-controlled blood pressures on anti-hypertensive medications.⁴¹ In SSA, undiagnosed and untreated or inadequately treated hypertension remains a substantial challenge as evidenced by a systematic review showing that nearly 70% of individuals with hypertension were unaware of their status and only 7% of those on anti-hypertensive medications had achieved adequate control of their blood pressures.⁴² As 16% of our patients with hypertension were newly diagnosed and only 45% of patients with known hypertension were on medications, the need for improvement in care for hypertension, including early diagnosis, appropriate disease control, and enhanced patient awareness about medication compliance in SSA, cannot be understated.

Individuals with hemorrhagic stroke are also known to have higher rates of mortality than those with ischemic strokes.³⁹ Thus, the high frequency of hemorrhagic stroke in our cohort also likely substantially contributes to the high rates of in-hospital mortality (24%) seen in this cohort. This has also been seen in other SSA stroke cohorts where in-hospital mortality has been reported to range from 20% to 45%.^{13, 21, 38} Increased mortality and earlier age of stroke onset due to hemorrhagic stroke is evidenced particularly in stroke cohorts in Northwestern Nigeria⁴³ that had higher hemorrhagic stroke frequencies than our cohort (35% vs. 28%), as well as younger average age of stroke patients (55 vs. 60 years) and increased mortality (37% vs. 24%). The higher risk of mortality from hemorrhagic strokes demonstrated in high-income settings is further compounded in low-resourced settings by limited neurosurgical interventions and, often, delayed blood pressure control due to limited availability of many classes of blood pressure medications, especially intravenous formulations.

Other contributing factors to the high mortality seen in our cohort included high rates of in-hospital post-stroke complications, such as aspiration pneumonia, and scarce critical care capacity. There are many well-established strategies for reducing post-stroke aspiration pneumonia, including swallow screens for dysphagia, nasogastric tubes for nutrition in the event of swallow dysfunction common to stroke, and adjustable beds for aspiration prevention. However, these seemingly low-cost interventions are often unavailable in many LMICs, including, to a large extent, in our setting. As a result, the lack of specialist personnel (speech and language pathologists), capital resources, and dedicated stroke units in SSA are significant challenges to substantially reduce post-stroke complications in LMICs.

Of note, our cohort also demonstrated a significant female predominance among people with stroke. This female predominance is likely largely a reflection of Zambia’s population demographics that show the male:female ratio to be 0.87 between ages 55–64 and 0.78 for those 65 and above.⁴⁴ Thus, Zambia has a female predominant population in the age ranges most affected by stroke. Our study also shows predominance of women (82%, n=32) among patients with unknown strokes (n=45), suggesting that fewer women obtain CT scans after their strokes. Further studies are necessary to understand this discrepancy, including sex differences in early stroke mortality and disparities women might face in accessing pre- and post-stroke care.

The population of patients with unknown strokes is unique from higher-income settings but important as many healthcare facilities in SSA lack adequate access to neuroimaging, with many not having any access at all. Therefore, most hospitals in the region still take care of this specific group of patients. While the demographic characteristics of this population were largely more similar to patients with confirmed ischemic strokes than those with confirmed hemorrhagic strokes, the patients with unknown stroke had the highest mortality while those with ischemic stroke had the lowest mortality. We suspect this is largely because many patients with unknown stroke died within the first days of their admission before neuroimaging could be obtained. Therefore, they were likely patients with more severe strokes or those with late presentations to the hospital after complications such as aspiration pneumonia had developed at home. This population also had the highest rate of previous strokes which may suggest challenges in adhering to secondary stroke prevention medications⁴⁵ or reticence on the part of healthcare workers to prescribe appropriate secondary prevention therapies (i.e. antiplatelet therapy) if neuroimaging was also not obtained at the time of their prior stroke.⁴⁶

Our findings are largely similar to a prior study of stroke epidemiology conducted at our hospital in 2010.⁴⁷ Both studies found high frequency of hemorrhagic strokes, hypertension as the most prevalent risk factor of stroke, a higher proportion of strokes among women, and a high frequency of HIV infection among individuals with stroke. However, compared to the 2010 study, our study found a lower proportion of hemorrhagic strokes (35% vs. 28%, p=0.07), increased prevalence of hypertension among the overall cohort (71% vs. 80%, p=0.02), and decreased HIV prevalence (25% vs. 18%, p=0.04). Additionally, our findings revealed a higher average age (55 vs. 60 years, p<0.001) and reduced in-hospital mortality (40% vs. 24%, p<0.001). The changes in stroke epidemiology such as increased hypertension prevalence, decreased proportion of hemorrhagic strokes, and higher average age of stroke onset likely reflects increased diagnosis, treatment, and control of hypertension and other modifiable risk factors of stroke. Decreased in-hospital mortality is likely at least partially due to the emerging systems of specialized neurological care and training in Zambia.

Limitations of our study include lack of stroke severity scales in our patients (eg. National Institutes of Health Stroke Scale, modified Rankin scale, Glasgow Coma Scale) as these scales were not routinely conducted during the timeframe from which these retrospective data were retrieved. In addition, 14% of strokes in this study remained uncategorized by subtype, which reflects the sporadic unavailability of CT scans due to scanner malfunction, individuals’ inability to afford imaging studies, and delays in obtaining CT imaging such that patient death sometimes occurred prior to imaging. We were also unable to characterize ischemic strokes as per the TOAST or STRIVE criteria due to incomplete workups. Although carotid ultrasound and CT-angiogram are available, we do not currently have acute stroke interventions, such as intravenous thrombolytics or endovascular therapies, nor do we have vascular surgeons or interventional radiologists who can intervene on such pathology if it is identified. Given that patients are paying for these expensive tests out-of-pocket, they are not routinely obtained since they will not change management in our setting. As a result, our categorization of large vessel strokes indicates only that a large vessel arterial distribution was observed and could not distinguish large vessel disease from cardioembolic disease. Similarly, lack of vessel imaging prevented us from identifying strokes in multiple vascular territories that may have been due to vasculitis or hypercoagulable disorders which would have been classified as cardioembolic strokes in this study. Of note, while subarachnoid hemorrhage and intracerebral hemorrhage are driven by different pathophysiological mechanisms, we included patients with both conditions in the same group (i.e. hemorrhagic stroke) because the small number of individuals with subarachnoid hemorrhage (n=4) made it difficult to consider them separately. However, sensitivity analyses excluding these individuals from the hemorrhagic stroke group did not result in any substantial changes to any of the reported analyses (data not shown). Finally, given the retrospective nature of this study, missing data for several variables limited the sample size of multivariable models. As such, multivariable models to determine predictors of mortality were not completed.

We were further unable to assess all stroke risk factors in all patients due to intermittent unavailability of laboratory reagents, dependence on robust patient charts maintained on paper, and circumstances in which patients were unable to afford requested investigations. For example, only 19% (n=61) of participants had a lipid profile completed, and only 31% (n=100) underwent both electrocardiogram and echocardiograph. As such, it is likely that our analysis of risk factors and in-hospital complications are underestimated. Similarly, routine screening for deep vein thrombosis (DVT) was not routinely undertaken at the time of this study, so we did not report this data as we strongly suspected the detected cases of DVT underestimated the true occurrence of this post-stroke complication. Lastly, as stroke patients referred to UTH are patients for whom stroke care required escalation of care from local clinics and primary health centers to the national referral hospital, our cohort likely represents the most severe stroke cases. As such, it is possible that our results, including rates of in-hospital complications and mortality, are overestimates of these poor outcomes compared to national averages. Alternatively, access to specialist neurological evaluations and a wider range of investigations, including neuroimaging, may have resulted in improved outcomes compared to what is available at first-level facilities.

However, in this study, risk factors and outcome data are reported from the first six months of Zambia’s first post-graduate neurology training program.⁴⁸ As such, the reported data were collected by neurology residents and verified by neurology faculty during patient encounters, and, as such, reflect a greater degree of precision in the clinical characterization of this stroke cohort than data collected from charts of patients evaluated by non-neurology healthcare workers. Secondly, including consecutive adults admitted with stroke to UTH over six consecutive months provides an important understanding of the current status of stroke epidemiology in Zambia and highlights the real-world challenges of caring for stroke patients in SSA settings with significant resource limitations such as the not infrequent unavailability of CT scans.

In summary, this cohort of Zambian adults with stroke is notable for high prevalence of hypertension, younger age with a higher frequency of hemorrhagic strokes, and higher rates of in-hospital mortality than stroke cohorts in high-income settings and was further notable for its high HIV prevalence. Our study demonstrates the need for a better understanding of both the pre- and post-stroke factors that account for high stroke-related mortality, the mechanisms by which HIV is an independent risk factor for stroke, the role of non-traditional risk factors of stroke in Zambia, and patients’ post-discharge morbidity and mortality due to stroke.

Highlights.

1. Average age of this Zambian stroke cohort was younger than most stroke cohorts in high-income countries.

2. Uncontrolled hypertension contributed to a high proportion of ischemic and hemorrhagic strokes but was most common in individuals with hemorrhagic strokes.

3. HIV was the fourth most common risk factor in this Zambian stroke cohort.

4. Post-stroke aspiration pneumonia substantially contributed to high in-hospital mortality.