KEY DETERMINANTS OF LONG-TERM POST-STROKE MORTALITY IN GHANA

Abstract

Background:

Stroke affects a predominantly young to middle-aged population in Africa and is associated with poor outcomes. There are limited data on patient-level determinants of long-term stroke survival on the continent.

Purpose:

To assess factors associated with long-term, all-cause mortality among stroke survivors in a Ghanaian medical system.

Methods:

We analyzed the dataset of clinical trial involving hypertensive stroke survivors (n=60) who enrolled in a 9-month study primarily assessing the effect of an m-health intervention on blood pressure control. This was a single tertiary center study conducted in a Ghanaian medical center. Participants or relatives were contacted by a phone call to assess vital status 4.5 years after stroke onset. Demographic, psycho-social and vascular risk factors data were collected during the study. Cox proportional hazards regression modeling was used to assess the factors associated with death.

Results:

Among the study participants, long term information was obtainable on 58 of 60 (97%). After a median follow-up of 52 months [IQR: 48 to 53 months], 16 participants had died, resulting in a mortality rate of 27.6% (95% CI of 16.6% - 40.9%). Upon adjustment for confounders, the two factors independently associated with long-term mortality were resistant hypertension (Hazard Ratio 3.99; 95% CI: 1.29 – 12.37] and depression (Hazard Ratio 1.18; 95% CI: 1.05 – 1.31; per unit rise on the Hamilton Depression Scale).

Conclusion:

In this convenience sample of recent stroke patients in Ghana, over a quarter had died within 5 years of index stroke onset. Resistant hypertension and depression may be modifiable therapeutic targets to improve outcomes in these patients.

INTRODUCTION

Suffering a stroke in sub-Saharan Africa portends a grim short-to-long term prognosis due to a combination of adverse patient- and system-related factors. Nearly 30% of stroke cases die within 30-days of stroke onset and 84% succumb within three years.^1–5 Data on the determinants of long-term mortality of stroke survivors in SSA are lacking. Although socio-demographic and stroke severity indicators may be important contributors to these poor outcomes, there could potentially be modifiable factors associated with post-stroke mortality yet to be identified on the sub-continent.

None of the modifiable risk factors for stroke occurrence, recurrence and its subsequent outcomes are as strongly correlated as systemic arterial hypertension. For instance, the population attributable risk (PAR) of hypertension for stroke occurrence among indigenous Africans is 91% (95% CI:88 – 94%), the highest in reported literature.⁶ Given how ubiquitous hypertension is among stroke survivors in Africa, a study focused on the determinants of long-term outcomes in this population seems warranted. While the treatment and control of blood pressure remains an important goal of secondary prevention after surviving a stroke, this goal is often elusive. Apart from medication non-adherence⁷ and therapeutic inertia⁸, resistant hypertension is probably an important, albeit overlooked, potential contributor to poor control of hypertension among stroke survivors.^9–11

Our objective for this study is to assess the factors which were longitudinally associated with mortality up to 5 years of surviving a stroke in Ghana. We approached this by deploying a telephonic follow up strategy among stroke survivors who were involved in a feasibility clinical trial titled Phone-based Intervention under nurse guidance after stroke (PINGS).¹² PINGS sought to test whether an m-health technology-enabled, nurse-led, multi-level integrated approach would be effective in improving blood pressure control among stroke survivors compared with usual care.

METHODS

Study Design & settings:

This is a prospective study involving stroke survivors who were recruited into the PINGS trial. The protocol for the PINGS trial has been previously published.¹² In brief, PINGS was a two-arm cluster randomized controlled pilot trial involving 60 stroke survivors within 1 month of stroke onset. Participants were randomly allocated into four clusters of 15 patients each per physician: two clusters in the intervention arm and two in the usual care arm. Participants in the intervention arm received a Blue-toothed UA-767Plus BT BP device and a smartphone for monitoring and reporting BP measurements and medication intake under nurse guidance for 3 months. Tailored motivational messages were delivered based upon levels of adherence to the medication intake. Both groups were then followed for 6 more months to compare BP control at months 3 (interim) and 9 (final) as primary outcomes. The first and last trial participant were enrolled on 7^th January 2017 and 16^th June, 2017 respectively while the last trial follow-up was completed on 16^th September, 2018.

In the present study, we contacted all study participants and their relatives using telephone calls. These calls were performed between July 1 to 15, 2021. In brief, patients or their relatives were called using phone numbers provided. We chose this approach due to its convenience, its low cost, the wide geographic distribution of stroke survivors and its ability to capture the study outcome of interest principally vital status of participant as dead or alive. Call duration was approximately 5 to 10 minutes. Verbal informed consent was sought based on ethical approval by the Committee of Human Research Publication and Ethics of the Kwame Nkrumah University of Science & Technology, Kumasi, Ghana.

Data collection:

Variables collected included age at stroke onset, gender, location of residence and educational status as socio-demographic variables. In addition, information of stroke type, stroke severity, functional status using the Modified Rankin scale, and Barthel’s index at enrollment into the trial were also collected. Stroke types were determined based on cranial computerized tomography (CT) scans performed within 10 days after stroke.

• Hypertension was defined using a blood pressure cutoff of higher than or equal to 140/90 mmHg for up to 72 hours after stroke, a history of hypertension, or use of antihypertensive before or after stroke onset.

• Resistant hypertension is defined as having a blood pressure that remains above the recommended target goal of 140/90 mm Hg despite treatment with three classes of antihypertensive medications or treatment with four or more classes of antihypertensive medications with any level of BP.¹³

• Diabetes mellitus was define based on history of diabetes mellitus, use of medications for diabetes mellitus, an HBA1c >6.5% or a fasting blood glucose level greater than 7.0 mmol/L measured after the post-acute phase of stroke.¹⁴

• For obesity, we assessed waist circumference using the WHO guidelines. ¹⁵

• Depression: Post-stroke depression was assessed at baseline visit using the 21-item Hamilton Depression Rating Scale¹⁶ (HDRS), one of the most widely used instruments in stroke studies with a sensitivity of 0.84 (95% CI: 0.75–0.90) and specificity of 0.83 (95%CI: 0.72–0.90).¹⁷ HDRS is used to rate the severity of depression by assessing mood, feelings of guilt, suicide ideation, insomnia, agitation or retardation, anxiety, weight loss and somatic symptoms. Each item on the questionnaire is scored on a 3 or 5 point scale depending on the item on a Likert scale. Participants with a score of 0 to 7 points were considered normal, 8–13 as mild depression, 14–18 as moderate depression, 19–22 as severe depression and >23 as very severe depression. The HDRS questionnaire was first translated into Twi (a local dialect widely spoken in the Ashanti region, where the study was conducted), pre-tested and back-translated into English language to establish semantic equivalence. HDRS assessments were performed by two experienced Research Officers who received four weeks of training on the study instruments until proficiency was attained with inter-rater agreement of >95% among hospital-based volunteers.

Follow-Up:

The present analysis involves 60 stroke survivors who enrolled into the PINGS trial and data was closed for the analysis in July 2021 to allow at least 48 months follow-up for the last subject who enrolled into the study. We used the following definitions for the vital status of study participants:

• Loss to follow-up: Any subject who could not be reached by phone to assess vital status.

• Death was ascertained by verbal autopsy from a relative of the stroke survivor. A verbal autopsy is a method of gathering health information about a deceased individual to determine his or her cause of death. Descriptions of events before death were acquired through interviews with a relative familiar with the deceased and analyzed by FSS to assign probable cause of death. For those who had died, we requested relatives to please provide date of death to help calculate duration of follow-up.

• Alive: For those who were alive, we assessed their functional status using the Modified Rankin Score.

Participants or their relatives were asked for clinical evidence of recurrent strokes defined as new onset or worsening of neurological deficits of sudden onset. Those alive were also asked whether they were still on secondary prevention medications or not.

Statistical Analysis:

Means and medians were compared using the Student’s t-test or Mann-Whitney’s U-test for paired comparisons. Proportions were compared using the Chi-squared or the Fisher’s exact test. Kaplan-Meier survival plots were constructed to assess proportions of subjects under follow-up with observations censored at either date of death, or time of phone contact to confirm being alive. Collinearity of variables was assessed by visual inspection of survival curves. Cox Proportional Hazards regression model was employed to assess the determinants of deaths during follow-up. In this model, variables such as age, gender, stroke type (ischemic, hemorrhagic, not-typed), resistant hypertension, depression, vascular risk factors and functional status of stroke survivors were selected as independent variables to be tested in the model based on their known associations with poor stroke outcomes. In bivariate analysis, factors associated with the dependent variable at a p-value level of 0.10 were included in the multivariable model. In our primary outcome analysis, subjects who were lost to follow up were excluded from our analysis. In all analysis, two-tailed p-values <0.05 were considered statistically significant with no adjustments for multiple comparisons. Statistical analysis was performed using SPSS version 19.

RESULTS

Disposition of study participants on follow-up:

Among 60 participants who enrolled into the PINGS trial, two (2) participants could not be reached by means of telephone calls and were excluded. Out of the 58 who could be reached, 16 had died and 42 were alive. The proportion who had died constituted 27.6% (95%CI of 16.6% - 40.9%) of those whose vital status could be assessed. The median duration of follow-up was 52 months [IQR: 48 to 53 months] with a mean (SD) duration of 46 (14) months. Using the Modified Rankin Score, 18 (31.0%) had a score of 1, 18 (31.0) had a score of 2, 4 (6.9%) had a score of 3, 1 (1.7%) each had a score of 4 and 5 respectively and 16 (27.6%) had a score of 6 (dead).

Two participants out of 29 (6.9%) who received the PINGS had recurrent strokes while 4 out of 29 (13.8%) on usual care had recurrent strokes, p=0.39. Among those who were alive (n=42), 1 out of 21 (4.8%) who were assigned to the PINGS intervention had stopped all blood pressure medications compared with 6 out of 21 (28.6%) for those who were in the usual care arm, p=0.04.

Comparison of participants according to vital status:

The median [IQR] age of study participants was 56.5 [44.0 – 64.3] years with a preponderance of male comprising 63.8% of the sample. As shown in Table 1, the median age between those who died compared with those who survived were non-significantly different at 60 years versus 51.5 years respectively. However, there were proportionally more female deaths at 42.9% (n=21) than male deaths at 18.9% (n=37), p=0.05. There were no differences in proportion of death by location of residence and educational attainment. Also there were no significant differences in survival status by stroke type, stroke severity and baseline functional status nor exposure to the PINGS intervention. Resistant hypertension was highly prevalent among those who died at 62.4% compared with 26.2% among those who survived, p=0.01. In addition, baseline median [IQR] Hamilton depression scale scores were significantly higher among those who died than those who were alive, 14 [12.5 – 17.5] versus 9 [6.5 – 12.0], p = 0.0001. (Table 1)

Table 1.

Characteristics of Study Participants According to survival status

Characteristic	Dead (n=16)	Alive (n=42)	Total (n=58)	P-value
Age, median (IQR)	60.0 (45.0 – 66.8)	51.5 (43.8 – 62.0)	56.5 (44.0 – 64.3)	0.32
Sex				0.05
Female	9 (56.3)	12 (28.6)	21 (36.2)
Male	7 (43.7)	30 (71.4)	37 (63.8)
Residence				0.91
Urban	10 (62.6)	24 (57.2)	34 (58.6)
Semi-urban	3 (18.7)	8 (19.0)	11 (19.0)
Rural	3 (18.7)	10 (23.8)	13 (22.4)
Educational attainment				0.36
No formal education	0 (0.0)	3 (7.1)	3 (5.2)
Primary education	10 (62.4)	17 (40.5)	27 (46.6)
Secondary education	5 (31.3)	12 (28.6)	17 (29.3)
Tertiary education	1 (6.3)	8 (19.0)	9 (15.5)
Postgraduate	0 (0.0)	2 (4.8)	2 (3.4)
Stroke Type				0.74
Ischemic stroke	10 (62.4)	27 (64.3)	37 (63.8)
Intracerebral hemorrhage	5 (31.3)	10 (23.8)	15 (25.9)
Untyped	1 (6.3)	5 (11.9)	6 (10.3)
Intervention allocation				1.00
Usual care	8 (50.0)	21 (50.0)	29 (50.0)
PINGS	8 (50.0)	21 (50.0)	29 (50.0)
Waist circumference	89.9 ± 11.2	92.3 ± 12.1	91.6 ± 11.8	0.50
NIHSS baseline	5.6 ± 3.8	4.3 ± 3.1	4.7 ± 3.3	0.17
Modified Rankin Scale at baseline	2.5 [2.0 – 3.0]	2.0 [2.0 – 3.0]	2.0 [2.0 – 3.0]	0.28
Barthel’s Index	77.5 [50.0 – 100.0]	85.0 [60.0 – 100.0]	85.0 [55.0 – 100.0]	0.61
HDS	14 [12.5 – 17.5]	9 [6.5 – 12.0]	10 [8 – 14]	0.0001
Resistant Hypertension	10 (62.4)	11 (26.2)	21 (36.2)	0.01
Blood glucose	7.9 ± 3.2	7.1 ± 1.8	7.3 ± 2.3	0.19
Diabetes	5 (31.3)	9 (21.4)	14 (24.1)	0.43
Statin	11 (68.8)	33 (78.6)	44 (75.9)	0.42
Antiplatelets	8 (50.0)	26 (61.9)	34 (58.6)	0.41
Number of Antihypertensives	3 [2 – 3]	2 [2 −3]	2 [ 2–3]	0.48
MMAS-9	11.5 [9.3–13.0]	12.0 [8.0–14.0]	12.0 [9.0 – 13.3]	0.88

NIHSS= National Institute of Health Stroke Scale; MRS = Modified Rankin Score; HDS = Hamilton’s Depression Scale, Modified Morisky Medication adherence scale.

Predictors of long-term stroke mortality:

By Kaplan-Meier curves plots the estimated probability of death at 12 months was 6.9%, at 24 months was 15.5%, at 36 months was 17.3%, at 48 months was 25.9% and 54 months was 27.6%. In unadjusted analysis, having resistant hypertension was associated with death at a hazard ratio of 3.48 [ 1.26 – 9.59] shown in Figure 1 and increasing severity of depression i.e. none, mild-to-moderate vs severe was incrementally associated with higher hazards of mortality as shown in Figure 2. Upon adjustment for confounders, the two factors independently associated with long-term mortality were resistant hypertension aHR of 3.99 [95% CI: 1.29 – 12.37] while each unit rise on the Hamilton Depression Scale was associated with aHR of 1.18 [1.05– 1.31], shown in Table 2. There was a trend towards higher mortality hazards among females than males, aHR of 2.32 [0.72–7.46], p=0.15. A sensitivity analysis showed that females had higher baseline HDRS scores than males 13.1 ± 5.6 compared with 9.6 ± 5.4, p = 0.03.

Figure 1:

Hazards of death according to Resistant Hypertension status

Figure 2:

Hazards of death in relation to Depression assessed using Hamilton Depression scale

Table 2.

Predictors of 4.5-year mortality Among Stroke Survivors in the PINGS Trial

Predictor	Unadjusted HR (95% CI)	P-value	Adjusted HR (95% CI)	P-value
Age, each 10 years older	1.15 [0.79–1.67]	0.47	--	--
Sex		0.06		0.15
Female	2.56 [0.95–6.88]		2.32 [0.72–7.46]
Male	1.00		1.00
Resistant Hypertension	3.48 [1.26–9.59]	0.02	3.99 [1.29–12.37]	0.016
National Institute of Health Stroke Scale, each unit higher	1.11 [0.97–1.27]	0.14	0.96 [0.80–1.14]	0.63
Hamilton’s depression scale, each unit higher	1.14 [1.06–1.23]	0.0008	1.18 [1.05–1.31]	0.0036

DISCUSSION

This is a post-hoc follow-up of a cohort of 60 Ghanaians with a recent stroke enrolled into an RCT to test the feasibility of a mHealth intervention for blood pressure control. Vital status, our primary study outcome was available for 58 (97%) participants who contributed a median duration of 52 months ( ≈ 4.5 years). Approximately 28% [95% CI: 17% - 41%] of participants died within the period of follow-up with females more likely than males to succumb. A recent meta-analysis of 91 studies from sub-Saharan Africa showed that 1-, 3- and 5-years poststroke case fatality were 33.2%, 40.1% and 39.4% respectively.² A previous community-based Tanzanian study reported a much higher case fatality of 84% within three years of stroke onset.⁵ An important reason for the lower overall mortality observed in our cohort is that the participants selected to participate in our m-health RCT mostly had mild-to-moderate stroke severity and functional status. On average the median NIHSS score was 5 and were all hypertensive stroke survivors.

We have previously reported on the interim outcomes at month 3 and final outcomes at month 9 for the PINGS trial wherein the proportion of participants achieving BP <140/90 mmHg among the intervention arm was 50% vs 44% among those on usual care at month 9.^18,19 In the present analysis, we show that the risk of post-trial mortality was not independently associated with exposure to the PINGS intervention. While exposure to the PINGS intervention did not translate to mortality reduction benefit, we observed that there were proportionally more recurrent strokes among those assigned to usual care (13.8%) compared with those who were assigned to PINGS (6.9%). Additionally, among those who were alive, ≈ 95% of those assigned to PINGS were still on antihypertensive medications prescribed by their physicians compared with ≈ 70% usual care, p=0.04. These observations suggest that the PINGS intervention-which entailed patient education on blood pressure risk management after stroke- may provide enduring benefits in terms of compliance with secondary prevention therapies to mitigate CVD risk. In our final adjusted models, two factors were independently associated with long-term stroke fatality, namely having resistant hypertension and depression.

Resistant hypertension was associated with an almost 4-fold higher adjusted hazard of mortality compared with those without treatment resistant hypertension. The confidence interval for this estimate is admittedly wide due to the low sample size. Although the literature is replete with data on the adverse outcomes of resistant hypertension, ^20–23 no previous study has provided specific estimates of the long-term mortality outcomes of resistant hypertension among stroke survivors. For instance, using a nationally representative sample from the U.S., Kaczmarski et al. demonstrated among the NHANES cohort that treatment resistant hypertension increased risk of cardiovascular mortality by 47% [95% CI: 10–96%] and all-cause mortality by 33% (95% CI: 13–57%).²⁴ Our data shows that resistant hypertension profoundly accentuates the risk of mortality among stroke survivors perhaps more than it does in other patient sub-populations. We have previously demonstrated among Ghanaians that that prevalence of resistant hypertension is substantially higher among stroke survivors at 45.3% (95% CI: 42.4–48.2) vs. stroke-free hypertensives at 19.9% (95%CI: 18.4–21.4), p<0.0001.²⁵ Furthermore, refractory hypertension, a more severe phenotype of RH was seven-fold commoner among Ghanaian hypertensive stroke survivors than stroke-free hypertensive individuals.²⁶ An estimated 17.7% of US adults taking antihypertensive medication had apparent treatment resistant hypertension based on BP cut-off value of <140/90 mm Hg with a pronounced association among blacks.^27–29 Perhaps, the poor outcomes of stroke survivors with resistant hypertension requires a closer scrutiny for more targeted management to reduce overall adverse outcomes.

We also interrogated our data to assess for associations between depression risk assessed using the Hamilton depression scale deployed at enrollment into the PINGS trial and long-term survival after the trial. For each unit rise on the Hamilton depression scale, adjusted hazard of mortality increased by 18% (95%CI: 5 – 31%). This finding corroborates the established association between post-stroke depression and risk of mortality. For instance, stroke survivors in the South London Stroke Register with post-stroke depression were at a 41% higher hazard of death [95% CI: 13%−77%].³⁰ A systematic review and meta-analysis conducted in 2012 reported a pooled hazard ratio of 1.52 [95%CI: 1.02 – 2.26] for mortality from post-stroke depression.³¹ An updated meta-analysis in 2018, re-emphasized the adverse association between post-stroke depression and mortality with a hazard ratio of 1.50 [95%CI: 1.28–1.75].³² Interestingly, post-stroke depression was found to portend a higher risk for short-term mortality with a relative risk of 1.70 than for long-term mortality at 1.35 in this updated meta-analysis.³² It is also noteworthy that the associations between post-stroke depression and risk of mortality appear more pronounced among those younger than 65 years³³, the age group which is largely reflected in the current study. We have shown in previous studies that post-stroke depression is quite rife in Ghana, a low-middle income country saddled with endemic poverty and limited social support systems for stroke survivors.^34,35 Thus the contribution of post-stroke depression to the poor outcomes of stroke in sub-Saharan Africa could be quite substantial but remains largely understudied. We also observed a trend towards a higher mortality hazard among women although not statistically significant in our final model. A sensitivity analysis comparing males to females showed that females were more likely to have hugher baseline depression score than male, depression being independently associated with mortality. A previous study by the SIREN group showed that women were more likely to have severer stroke in Ghana and Nigeria compared with men perhaps increasing their proclivity towards higher mortality.³⁶ However, in the present study baseline modified Rankin and NIHSS scores between males and females were not significantly different.

The main limitations of this study are that it is a post-hoc analysis of a pilot trial with a modest sample size of 60 participants of which 58 were evaluable because they could be reached by phone. Many of the conclusions drawn are applicable to a selected cohort of hypertensive stroke survivors and not generalizable to stroke survivors without hypertension. In Africa however, the population attributable risk of hypertension to stroke occurrence is over 90% and is perhaps among the highest on the globe.⁶ Thus a focus on hypertension among stroke survivors is probably well intentioned. An important strength of the study is its prospective design with nearly 4.5 years of follow-up outcomes. Very few studies in Africa have reported on long-term outcomes of stroke, hence this is an added contribution to the literature. A much larger second phase of the PINGS trial across 10 study sites in Ghana is currently ongoing with results expected in 2025. ³⁷

Conclusion:

This long-term post-trial vital status assessment of a sample of Ghanaian stroke survivors with hypertension revealed that almost 30% had died within 5 years of stroke onset. Independent predictors of mortality were resistant hypertension phenotype and early post-stroke depression, two factors which are amenable to targeted treatment.

Highlights.

The determinants of long-term mortality among stroke survivors in Africa is unknown

We found 28% of stroke survivors had died after 4.5 years of follow-up

Resistant hypertension relatively increased risk of mortality by 4-fold

Depression increased risk of death by 18% per unit rise on HDS