Factors associated with late hospital arrival in acute stroke patients of Bangladesh

Abstract

Objectives

Underutilisation of thrombolysis is a major problem in patients with stroke in Bangladesh as patients do not arrive within the therapeutic window due to delays in their way to emergency department. This study aims to assess the time delay from patients’ symptom onset to arrival in the hospital and the factors that are associated with it.

Methods

This cross-sectional survey of a prospective cohort of stroke patients was conducted between January and March 2023. 448 stroke patients meeting the inclusion criteria were enrolled in the study from five tertiary-level hospitals in Bangladesh. After obtaining informed consent, trained data collectors conducted face-to-face interviews of the patient/patients’ guardians via a pretested structured questionnaire. Stata (V.16) was used to analyse data. Median and IQRs were used to summarise quantitative variables, and qualitative variables were summarised using frequency and relative frequency. Pearson’s χ² test and Mann-Whitney U test were used to explore the bivariate relationship between predictor and outcome variables. Finally, a binary logistic regression model was fit to explore the factors associated with delayed arrival (>4.5 hours) at the hospital.

Results

The median age of the patients was 61 years (54–70) and 63% were men. The majority hailed from rural (59.6%) areas and had primary (25.89%) education. The patients had an overall median prehospital delay of 14 (8–28) hours, 3 (1–6) hours of decision delay, 1 (0–2) hours of medical contact delay, and 14 (6.5–25.75) hours of referral delay. Patients with master’s education (adjusted OR (AOR): 0.04, p=0.023) and private transport (AOR: 0.26, p=0.029) had a lower chance of late arrival. However, patients having unknown onset, self-medicating, having a previous history of stroke, and being admitted to a private hospital had a significantly higher chance of late arrival.

Conclusion

Nearly 90% of the patients were late to arrive (>4.5 hours) at hospital and referral delay comprises the majority of the prehospital delay. Therefore, fast symptom recognition and the urgency of seeking healthcare as soon as symptoms appear should be the focus of public awareness efforts.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Bangladesh has a high burden of stroke-related mortality and morbidity.

• Existing limited evidence has highlighted delays in hospital arrival for stroke patients in many countries, including Bangladesh, often exceeding the critical window for effective treatment.

WHAT THIS STUDY ADDS

• This cross-sectional study reveals that 87.72% of the stroke patients in Bangladesh arrive to hospital more than 4.5 hours after stroke onset. Decision delay, medical contact delay and referral delay all contributed. Predictors of delay were education lower than a master’s degree, unknown onset/found, previous history of stroke, self-medicating after onset of symptoms, ambulance transport, involvement of both hemispheres and treatment in a private hospital.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The findings emphasise the necessity for improved public awareness campaigns on stroke symptom recognition and response. Furthermore, it suggests the need for a centralised referral system to facilitate timely access to transport and thrombolysis, potentially in a dedicated stroke centre.

Introduction

Bangladesh has an age-standardised and gender-standardised stroke mortality of 54.8/100 000 individuals per year.¹ However, the incidence of stroke has not been adequately explored in epidemiological studies. The stroke prevalence in Bangladesh was reported to be 0.2%, 0.3%, 0.2%, 1% and 1% among people aged 40–49, 50–59, 60–69, 70–79 and 80+ years, respectively.² Also, stroke prevalence is nearly identical in men and women in urban and rural locations.³ It rises with age to the point where the proportion of cerebral infarction to intracerebral haemorrhage is 2.91 in the population.⁴ According to the most recent WHO data published in 2020, there were 134 166 stroke-related fatalities in Bangladesh, accounting for 18.74% of all deaths.²

The timeliness of care for patients enduring acute ischaemic strokes (AIS) is significantly impacted by the variability in hospitals’ ability to perform thrombolysis in Bangladesh. Although previous research has investigated the causes of delays in emergency department presentations and the administration of thrombolysis, it has generally neglected the critical delays that result from referrals from hospitals that lack the requisite facilities.^{5 6} In Bangladesh, the absence of a centralised referral mechanism and the variability of specialised care services further exacerbate these delays, resulting in protracted prehospital delays (PHDs) that have a substantial impact on treatment outcomes in AIS cases.^{5 7} Failure to recognise stroke symptoms by patients, caregivers and physicians, lack of awareness of thrombolysis as a treatment, absence of a witness at the time of onset and the improvement of symptoms after the onset are all contributing factors to PHD. The reasons for the non-use of thrombolysis include the absence of infrastructure, organisational and economic barriers, and inadequate policy and governance in the context of stroke care.⁵ To address this lacuna, this study will offer insights into the ways in which referral delays contribute to the overall PHD in stroke management. Developing strategies to improve timely access to thrombolysis and ultimately improve patient outcomes in a healthcare system characterised by varied service capabilities is contingent on an understanding of these dynamics.⁵

Numerous multicentre randomised controlled trials have conclusively demonstrated that intravenous thrombolysis is an efficient therapy for AIS.⁸ Thrombolysis is typically administered within 4.5 hours of symptoms beginning in AIS⁹; hence, its effectiveness may be restricted by this short therapeutic window. Underutilisation of thrombolysis is a widespread problem in many countries, as patients present late for assessment of this treatment, resulting in PHD. In a developing nation like ours, reducing PHD can make a significant impact in stroke management, considering the burden of stroke-associated complications. However, there is a dearth of rigorous data addressing this topic in Bangladesh.¹⁰ Therefore, this multi-centre study was carried out to investigate the factors contributing to individuals with acute stroke arriving late and identify strategies for reducing the delay.

Methods

Study design, setting and participants

This cross-sectional survey of a prospective cohort of stroke patients was conducted at five tertiary-level hospitals in different cities of Bangladesh. The details of the healthcare system of Bangladesh are found in online supplemental file 1. Between January 2023 and March 2023, all the patients admitted to the neurology department of these hospitals either directly through the emergency department or by referral from other institutions were approached for enrolment into this study on meeting the inclusion criteria. The following criteria were considered before including a patient in the study: (1) patients with a stroke diagnosis (International Classification of Diseases codes I61: non-traumatic intracerebral haemorrhage; I63: ischaemic stroke (cerebral infarction); or G45: transient ischaemic attack, TIA) based on clinical examination by a neurologist and neuroimaging (CT and MRI scans), (2) at least 18 years of age and (3) patient/patient’s guardian provided written informed consent. Patients were excluded if they had (1) any main diagnosis other than stroke, (2) an in-hospital stroke, (3) inability to respond to the structured questionnaire and absence of a witness to the prehospital phase who might respond to the structured questionnaire or (4) patients who had previously received thrombolysis treatment or had been referred from a medical facility with thrombolysis capacity before being admitted to one of the study institutions. In the case of recurrent stroke, only the first event was included. Finally, a total of 448 stroke patients were enrolled in the study (figure 1).

Figure 1. Patient enrollment.

Data collection

One trained physician in each institution (a total of five) interviewed all the recruited patients or their guardians in person using a structured questionnaire (online supplemental file 2) after obtaining informed consent. All interviews were conducted at the bedside and completed within 3 days of admission. The questionnaire had two parts: (1) sociodemographics and data regarding the prehospital period and (2) clinical data. Sociodemographic information about the patient and data regarding the prehospital period was recorded from the interview, while clinical data of the patient were obtained from the medical records.

Variables

PHD was the main outcome variable of this study. PHD was defined as the time between the onset of the symptoms of stroke and the time of admission into the study hospitals. When a patient or family member first recognised stroke symptoms, which was considered to be the time of symptom onset. If the symptoms happened while the patient was sleeping, the time they woke up was recorded as the time of onset because only then they could seek medical help.¹¹ For patients whose onset of symptoms was unknown, the last known time without symptoms was regarded as the time of symptom onset.¹² If the exact time of stroke onset could not be specified, the time was estimated to be morning (8:00), noon (12:00), afternoon (16:00), evening (20:00) and night (0:00).¹³ PHD is broken down into decision delay (time interval between symptom onset and decision to seek healthcare), medical contact delay (time interval between decision to seek healthcare and first medical contact) and referral delay (time interval between first medical contact and admission into the study hospitals) (figure 2).¹⁴ PHD was dichotomised into two groups: (1) early arrival (≤4.5 hours) and (2) late arrival (>4.5 hours). The cut-off time of 4.5 hours was chosen according to the expert’s opinion, and it corresponds to the time window in which Intravenous thrombolysis can be administered.¹²

Figure 2. Definition of total prehospital delay and different delays. ICU, intensive care unit.

Sociodemographic variables and variables regarding the prehospitalisation period included age, gender, religion, educational level, monthly family income, marital status, region of residence (urban, rural), type of family (nuclear, extended), whether living alone before the stroke, the distance of primary care centre, shortest distance on the road between the geographic location of stroke onset and study hospital, mode of transport (ambulance, private transport, public transport), mode of admission (direct admission, referred from primary centre, referred from a government hospital, referred from a private hospital, referred from a physician) and type of study hospital (private, public). Clinical variables included risk factors, symptoms, mode of symptom onset, diagnosis, location of the lesion, Glasgow coma scale score (GCS), modified National Institutes of Health stroke scale (mNIHSS) score and modified Rankin score (mRS). A quantitative and qualitative categorisation of the variables are provided in online supplemental file 3.

Statistical analysis

Statistical analysis was done using Stata (V.16; StataCorp, College Station, Texas). A histogram, a normal Q-Q plot, and the Shapiro-Wilks test were performed to check for the normality of continuous data distribution. Since the distribution of PHD times was skewed, the median was used as a measure of central tendency, and the IQR was used as a measure of dispersion for continuous variables. Categorical variables were summarised using frequency and relative frequency. Mann-Whitney U test, Pearson’s χ² test and Fisher’s exact test were used to explore the bivariate relationship between outcome and predictor variables. In order to identify the predictors of late hospital arrival of stroke patients, we fit a multiple logistic regression model to estimate the OR with a 95% CI after adjusting for covariates. In the multivariate analysis, we included variables based on our review of existing literature pertinent to our study objectives. We also included variables with a p value of ≤0.2 in the univariate analysis. All significance tests were two tailed, and a p value of less than 0.05 was considered statistically significant.

Patient and public involvement

A stroke survivor and his wife (who was present at the onset of his stroke and the entire treatment duration) were involved in the review of the study design and research tools. The patients were not involved in the recruitment and conduct of the study. We are unable to directly disseminate the findings to the study participants.

Results

Baseline characteristics of the patients

The median age of the patients was 61 years (54–70) and 63% were men. Nearly half of them (45.09%) were from low-income households, and the majority (59.60%) lived in rural areas. More than half of the participants (50.89%) had a primary or below level of education. The median distance of the primary care centre was 5 km (IQR, 2–8), and the median of the shortest route on the road between the geographic location at stroke onset and the study hospital was 30 km (IQR, 10–100). Most of the patients were from government study hospitals (88.17%). After the onset of the symptoms, 77.46% consulted qualified physicians, while others chose to see non-qualified practitioners or self-medicate, etc. Approximately 67% arrived at the hospital by ambulance, while the remainder took public and private transportation. Only one-third of the participants were admitted directly to the hospital, while the others were referred by other government or private hospitals. Among the participants, 261 (58.26%) were diagnosed with cerebral infarction, 158 (35.27%) with non-traumatic intracerebral haemorrhage and 29 (6.47%) with TIA. In terms of stroke symptoms, 63.62% of participants lost their balance, 75.89% showed arm or leg weakness and 63.62% of patients had speech difficulties. In 60 (13.39%) patients, the exact moment of symptom start could be ascertained, and 120 (26.79%) patients awoke with symptoms, whereas the onset of symptoms was unknown in the remaining 268 (59.82%) patients. Many patients had a history of smoking (32.81%), diabetes mellitus (46.21%), hypertension (78.57%) and previous (26.56%) or family history (24.56%) of stroke. Overall, 181 (40.40%) patients/patients’ attendant suspected stroke after symptoms, whereas 158 patients/patients’ attendant (35.27%) did not realise the symptoms of a stroke. On admission, the median mNIHSS score was 12 points (IQR, 7–18), the median GCS score was 10 points (IQR, 8–13), and the median mRS score was 4 (IQR, 3–5) (table 1).

Table 1. Baseline characteristics of the patients (n=448).

Characteristics	Total (n=448)
Age (in years), median (IQR)	61 (54–70)
Age category (in years), n (%)
<55	115 (25.67%)
55–69	209 (46.65%)
≥70	124 (27.68%)
Gender, n (%)
Female	166 (37.05%)
Male	282 (62.95%)
Region of residence, n (%)
Rural	267 (59.60%)
Urban	181 (40.40%)
Distance of primary care centre (in km), median (IQR)	5 (2–8)
Shortest distance on road between geographic location of stroke onset and study hospital (in km), median (IQR)	30(10–100)
Religion, n (%)
Islam	410 (91.52%)
Sanatan	36 (8.04%)
Buddhism	2 (0.45%)
Highest education, n (%)
No formal education	90 (20.09)
Informal education	22 (4.91%)
Primary	116 (25.89%)
Secondary	87 (19.42)
Higher secondary	79 (17.63%)
Bachelor	44 (9.82%)
Masters	10 (2.23%)
Monthly family income (in BDT), n (%)
<14K	202 (45.09%)
15–29K	144 (32.14%)
30–60K	89 (19.87%)
>60K	13 (2.90%)
Marital status, n (%)
Unmarried	5 (1.12%)
Married	407 (90.85%)
Widowed	36 (8.04%)
Type of family, n (%)
Nuclear	226 (50.45%)
Extended	222 (49.55%)
Living home alone before the stroke, n (%)
No	392 (87.50%)
Yes	56 (12.50%)
Mode of transport, n (%)
Ambulance	299 (66.74%)
Private	81 (18.08%)
Public	68 (15.18%)
Mode of admission, n (%)
Direct admission	141 (31.47%)
Referred from primary centre and general practitioner	117 (26.12%)
Referred from government hospital	135 (30.13%)
Referred from private hospital	55 (12.28%)
Type of study hospital, n (%)
Private	53 (11.83%)
Government	395 (88.17%)
Diagnosis, n (%)
I61: nontraumatic intracerebral haemorrhage	158 (35.27%)
I63: cerebral infarction	261 (58.26%)
G45: transient ischaemic attack	29 (6.47%)
Location of lesion, n (%)
Right hemisphere	182 (40.63%)
Left hemisphere	156 (34.82%)
Bilateral	72 (16.07%)
Cerebellum	26 (5.80%)
Brainstem	12 (2.68%)
Symptoms, n (%)
Loss of balance	285 (63.62%)
Blurry vision	116 (25.89%)
Facial dropping	132 (29.46%)
Arm or leg weakness	340 (75.89%)
Speech difficulty	285 (63.62%)
Symptom onset, n (%)
Uncertain/found with stroke symptoms	268 (59.82%)
Onset during sleep and woke up with stroke symptoms	120 (26.79%)
Spontaneous onset	60 (13.39%)
Behaviour after the onset of symptom, n (%)
Did not consider the symptoms to be serious	41 (9.15%)
Waited to see symptoms would resolve	44 (9.82%)
Didn't realise the symptoms of a stroke	158 (35.27%)
Suspected as stroke	181 (40.40%)
Call family members	24 (5.36%)
First medical action after the onset of symptoms, n (%)
Visiting qualified doctor	347 (77.46%)
Visiting non-qualified practitioner	39 (8.71%)
Self-medication	36 (8.04%)
Tele consultation	25 (5.58%)
Direct Admission	1 (0.22%)
Medical history/risk factors, n (%)
Smoking	147 (32.81%)
Diabetes mellitus	207 (46.21%)
Sedentary lifestyles	91 (20.31%)
Hypertension	352 (78.57%)
Previous history of stroke	119 (26.56%)
Family history of stroke	111 (24.78%)
Hypercholesterolemia	95 (21.21%)
Atrial fibrillation	13 (2.90%)
Total GCS score, median(IQR)	10 (8–13)
Total mNIHSS score, median (IQR)	12 (7–18)
Total mRS score, median (IQR)	4 (3–5)

BDTBangladeshi taka (1$ = 120 BDT)GCSGlasgow coma scalemNIHSSmodified National Institutes of Health Stroke ScalemRSmodified Rankin Score

Prehospital delay

Table 2 shows a right-skewed distribution of PHDs with a wide IQR. The median of all the delays was 14 (IQR, 8–28) hours, where the median of decision delays was three (IQR, 1–6) hours, and the median medical contact delay was one and a half (IQR, 1–3) hours. Overall, referral delays were 14 (IQR, 6.5–25.75) hours, and this was much longer for government hospitals, 19.75 (IQR, 9–45), and for private hospitals, 15 (IQR, 7–28) than the primary care centre and registered physicians, 8.5 (IQR, 4–21).

Table 2. Total prehospital delay, decision delay, medical contact delay and referral delay for patients with stroke (n=448).

Characteristics	N (%)		Time in hours, median (IQR)
Total prehospital delay	448	(100%)	14 (8–28)
Decision delay	448	(100%)	3 (1-6)
Medical contact delay*
Overall	448	(100%)	1.5 (1-3)
Visit to primary care centre and general practitioner	117	(15.85%)	2 (1-4)
Admission to government hospital	135	(30.13%)	1 (1-2)
Admission to private hospital	55	(12.28%)	1 (1-3)
Admission to study hospital	141	(31.47%)	2 (1-4)
Referral delay*
Overall	307	(100%)	14 (6.5–25.75)
From primary care centre	71	(23.13%)	10 (4–18)
From government hospital	135	(43.97%)	19.75 (9-45)
From private hospital	55	(17.92%)	15 (7–28)
From registered physician	46	(14.98%)	8.5 (4-21)

^*Significant difference between the groups (pp<0.001).

Factors associated with PHD of stroke patients

Of the 448 study participants, 55 (12.28%) came early to the hospital, whereas 393 (87.72%) arrived late. A bivariate comparison of demographic and presentation characteristics between early and late arrival shows that female gender, rural residence, longer distance to the primary care centre, higher distance in shortest route on the road between the geographic location of stroke onset and study hospital, lower education status, transportation through ambulance, government hospital, loss of balance at onset, visiting qualified doctor after the onset of symptoms, higher total NIHSS score and higher total modified ranking score were significantly associated with late arrival (table 3).

Table 3. Baseline characteristics of patients stratified by pre-hospital delay (n=448).

	Characteristics	Median prehospital delay (IQR) (in hours)	Early arrival (≤4.5 hours) (N=55, 12.28%)	Late arrival (>4.5 hours) (N=393, 87.72%)	P value
Age (years), median (IQR)		–	62 (55–70)	61 (54–70)	0.89
Age category (years), n (%)	<55	13 (7–48)	13 (23.64%)	102 (25.95%)	0.927
	55–69	15 (8–26)	26 (47.27%)	183 (46.56%)
	≥70	14 (10–24)	16 (29.09%)	108 (27.48%)
Gender, n (%)	Female	24 (10–36)	12 (21.82%)	154 (39.19%)	0.012
	Male	12 (6–24)	43 (78.18%)	239 (60.81%)
Region of residence, n (%)	Rural	20 (10–36)	16 (29.09%)	251 (63.87%)	<0.001
	Urban	10 (5–24)	39 (70.91%)	142 (36.13%)
Distance of primary care centre (km), median (IQR)		–	2 (1–3)	5 (3–8)	<0.001
Shortest distance on road between geographic location of stroke onset and study hospital (km), median (IQR)		–	3 (1.5–6)	40(13-106)	<0.001
Religion, n (%)	Islam	14 (8–28)	50 (90.91%)	360 (91.60%)	0.260
	Sanatan	13.5 (10–24)	4 (7.27%)	32 (8.14%)
	Buddhism	8 (4–12)	1 (1.82%)	1 (0.25%)
Highest education, n (%)	No formal education	18 (10–30)	3 (5.45%)	87 (22.14%)	<0.001
	Informal education	24 (12–26)	–	22 (5.60%)
	Primary	20 (10–30)	13 (23.64%)	103 (26.21%)
	Secondary	24 (12–36)	3 (5.45%)	84 (21.37%)
	Higher secondary	8 (3–16)	25 (45.45%)	54 (13.74%)
	Bachelor	10 (6–24)	5 (9.09%)	39 (9.92%)
	Masters	3 (1–6)	6 (10.91%)	4 (1.02%)
Monthly family income (in BDT), n (%)	<14K	14 (10–26)	21 (38.18%)	181 (46.06%)	0.275
	15–29K	12 (6–24)	24 (43.64%)	120 (30.53%)
	30–60K	24 (10–48)	9 (16.36%)	80 (20.36%)
	>60K	10 (6–24)	1 (1.82%)	12 (3.05%)
Marital status, n (%)	Unmarried	7 (6–12)	1 (1.82%)	4 (1.02%)	0.174
	Married	12 (7–28)	53 (96.36%)	354 (90.08%)
	Widowed	24 (17–48)	1 (1.82%)	35 (8.91%)
Type of family, n (%)	Nuclear	12 (6–24)	32 (58.18%)	194 (49.36%)	0.221
	Extended	14.5 (10–30)	23 (41.82%)	199 (50.64%)
Living home alone before the stroke, n (%)	No	15 (8–30)	48 (87.27%)	344 (87.53%)	0.957
	Yes	12 (6–24)	7 (12.73%)	49 (12.47%)
Mode of transport, n (%)	Ambulance	24 (10–36)	11 (20.00%)	288 (73.28%)	<0.001
	Private	6 (2–10)	32 (58.18%)	49 (12.47%)
	Public	12 (6–48)	12 (21.82%)	56 (14.25%)
Mode of admission, n (%)	Direct admission	8 (4–12)	40 (72.73%)	101 (25.70%)	<0.001
	Referred from primary centre and general practitioner	20 (5–30)	10 (8.55%)	107 (91.45%)
	Referred from government hospital	24 (12–48)	3 (5.45%)	132 (33.59%)
	Referred from private hospital	24 (10–48)	2 (3.64%)	53 (13.49%)
Type of study hospital, n (%)	Private	12 (5–20)	11 (20.00%)	42 (10.69%)	0.045
	Government	16 (8–28)	44 (80.00%)	351 (89.31%)
Diagnosis, n (%)	I61:nontraumatic intracerebral haemorrhage	20 (10–30)	14 (25.45%)	144 (36.64%)	0.215
	I63:cerebral infarction	12 (6–26)	38 (69.09%)	223 (56.74%)
	G45: transient ischaemic attack	12 (7–24)	3 (5.45%)	26 (6.62%)
Location of lesion, n (%)	Right hemisphere	12 (7–24)	27 (49.09%)	155 (39.44%)	0.236
	Left hemisphere	12 (8–30)	18 (32.73%)	138 (35.11%)
	Bilateral	24 (12–48)	4 (7.27%)	68 (17.30%)
	Cerebellum	16 (6–24)	5 (9.09%)	21 (5.34%)
	Brainstem	24 (12–30)	1 (1.82%)	11 (2.80%)
Symptoms, n (%)	Loss of balance	14 (7–28)	42 (76.36%)	243 (61.83%)	0.036
	Blurry vision	12 (6–24)	18 (32.73%)	98 (24.94%)	0.217
	Facial dropping	14.5 (8–25)	21 (38.18%)	111 (28.24%)	0.130
	Arm or leg weakness	12 (8–26)	42 (76.36%)	298 (75.83%)	0.931
	Speech difficulty	12 (7–24)	36 (65.45%)	249 (63.36%)	0.762
Symptom onset, n (%)	Uncertain/found with stroke symptoms	14 (8–28)	29 (52.73%)	239 (60.81%)	<0.001
	Onset during sleep and woke up with stroke symptoms	20 (10–30)	6 (10.91%)	114 (29.01%)
	Spontaneous onset	6 (4–14)	20 (36.36%)	40 (10.18%)
Behaviour after the onset of symptom, n (%)	Did not consider the symptoms to be serious	12 (8–28)	4 (7.27%)	37 (9.41%)	0.249
	Waited to see symptoms would going	24 (8–48)	3 (5.45%)	41 (10.43%)
	Didn't realise the symptoms of a stroke	17 (9–30)	18 (32.73%)	140 (35.62%)
	Suspected as stroke	12 (6–24)	29 (52.73%)	152 (38.68%)
	Call family members	24 (7–96)	1 (1.82%)	23 (5.85%)
First medical action after the onset of symptoms, n (%)	Visiting qualified doctor	16 (8–30)	38 (69.09%)	309 (78.63%)	0.013
	Visiting non-qualified practitioner	19 (6–36)	7 (12.73%)	32 (8.14%)
	Self-medication	12 (6–15)	3 (5.45%)	33 (8.40%)
	Tele consultation	12 (5–22)	6 (10.91%)	19 (4.83%)
	Direct admission	2 (2–2)	1 (1.82%)	–
Medical history/risk factors, n (%)	Smoking	12 (6–24)	17 (30.91%)	130 (33.08%)	0.748
	Diabetes mellitus	12 (6–24)	31 (56.36%)	176 (44.78%)	0.107
	Sedentary lifestyles	15 (8–25)	16 (29.09%)	75 (19.08%)	0.084
	Hypertension	12 (7–24)	47 (85.45%)	305 (77.61%)	0.184
	Previous history of stroke	12 (7–24)	109 (27.74%)	109 (27.74%)	0.133
	Family history of stroke	12 (6–24)	98 (24.94%)	98 (24.94%)	0.834
	Hypercholesterolemia	12 (6–24)	79 (20.10%)	79 (20.10%)	0.127
	Atrial fibrillation	20 (12–30)	13 (3.31%)	13 (3.31%)	0.171
Total GCS score, median(IQR)		–	10 (8–13)	10 (8–13)	0.984
Total mNIHSS score, median (IQR)		–	10 (6–14)	12 (7–18)	0.035
Total mRS score, median (IQR)		–	3 (2–5)	4 (3–5)	<0.001

Significant p- values (< 0.05) are in bold.

BDTBangladeshi taka (1$ = 120 BDT)GCSGlasgow coma scalemNIHSSmodified National Institutes of Health Stroke ScalemRSmodified Rankin Score

The logistic regression model demonstrated that PHD was shorter by 97% among patients with master’s degrees (adjusted OR (AOR)=0.03, 95% CI 0.00 to 0.46; p=0.012) compared with those with no formal education. Patients who woke up with stroke symptoms had 4.15 times higher odds (AOR=4.15, 95% CI 1.21 to 14.23; p=0.024), and patients who had known onset of stroke symptoms had a 76% lower chance (AOR=0.24, 95% CI 0.07 to 0.86; p=0.028) of arriving late to the hospital than those who with uncertain symptom onset.

Those who took self-medication after having a stroke were 6.38 times more likely to be late (AOR=6.38, 95% CI 1.08 to 37.76; p=0.041) to the hospital compared with those who visited a qualified physician. Stroke patients who were transported to the hospital by private transportation were 74% less likely to be late (AOR=0.26, 95% CI 0.08 to 0.82; p=0.021), while those who were transported by public transportation were 86% less likely to be late (AOR=0.14, 95% CI 0.04 to 0.51; p=0.003) at the hospital than those who were transported by ambulance. Stroke patients referred from the primary centre (did not have thrombolysis facility) and general practitioner had 3.98 times more chances of being late for hospital arrival (AOR=3.98, 95% CI 1.19 to 13.29; p=0.025) than those who got direct admission.

A patient with a previous history of stroke was 9.64 times more likely to arrive late at the hospital than those (AOR=9.64, 95% CI 2.23 to 41.65; p=0.002) who had not. Stroke patients who went to a private hospital were 4.21 times more (AOR=4.21, 95% CI 1.13 to 15.71; p=0.032) late than those who went to a public hospital (table 4). Gender despite being a significant variable in bivariate analysis, it did not hold up its significance in the multivariate analysis.

Table 4. Multivariate logistic regression investigating the factors associated with the late arrival of patients with stroke.

Predictors	AOR	95% CI	P value
Education
No formal education	Reference
Primary	0.82	0.13 to 5.08	0.831
Secondary	1.38	0.16 to 12.09	0.772
Higher secondary	0.6	0.10 to 3.65	0.58
Bachelor	1.67	0.18 to 15.06	0.649
Masters	0.03	0.00 to 0.46	0.012
Type of family
Extended	Reference
Nuclear	2.9	0.94 to 8.94	0.064
Symptom onset
Uncertain/found	Reference
Wake-up	4.15	1.21 to 14.23	0.024
Known	0.24	0.07 to 0.86	0.028
First medical action after symptom onset
Visiting qualified physician	Reference
Visiting non-qualified practitioner	0.64	0.12 to 3.38	0.599
Self-medication	6.38	1.08 to 37.76	0.041
Tele consultation	0.26	0.04 to 1.58	0.143
Mode of transport
Ambulance	Reference
Private	0.26	0.08 to 0.82	0.021
Public	0.14	0.04 to 0.51	0.003
Mode of admission
Direct admission	Reference
Referred from primary centre and general practitioner	3.98	1.19 to 13.29	0.025
Referred from government hospital	2.9	0.46 to 18.52	0.26
Referred from private hospital	2.99	0.40 to 22.40	0.286
Previous history of stroke
No	Reference
Yes	9.64	2.23 to 41.65	0.002
Shortest distance on the road between the geographic location of stroke onset and study hospital (in km)	1.12	1.06 to 1.18	<0.001
Type of study hospital
Public	Reference
Private	4.21	1.13 to 15.71	0.032
Stroke location
Right hemisphere	Reference
Left hemisphere	1.02	0.36 to 2.85	0.971
Bilateral	12.51	1.64 to 95.17	0.015
Cerebellum	0.69	0.11 to 4.38	0.693
Brainstem	1.46	0.11 to 19.56	0.775

Significant p- values (< 0.05) are in bold.

Discussion

In 26 nations, PHD has not reduced since 2006, with most patients arriving after 3 hours.¹⁵ A similar pattern was observed here in this cross-sectional survey of stroke patients in Bangladesh. Here, 87.72% reported PHDs of >4.5 hours, which was found to be linked to education, help-seeking behaviour, transportation and self-medication. Stroke history and symptom onset were unmodifiable risk factors. Besides, the median PHD was 14 hours, which is substantially greater than the 3–6 hours recorded by most developed nations.¹⁶ This is a valid outcome that can be explained by the limited resources, facilities and primary care services of a developing country.

Like the study conducted by Nowacki et al,¹⁷ our study indicated that the higher educational status of patients decreased PHD, although earlier studies found no association.^{18 19} Hence, the role of education in early arrival is controversial. Here, bystander educational status and previous experience of stroke might have influenced the outcome. On the other hand, people living in nuclear families were more likely to have delayed presentation, which validates previous results²⁰ that living alone is a factor in PHD. This factor had a 4.52 OR compared with extended families, which emphasises the fact that a well-working social support and network is important for early admission. The social network is, however, not readily changeable.

The use of emergency medical services has been demonstrated to reduce PHD, according to several studies.^{21 22} However, the lack of easily accessible emergency ambulance services in Bangladesh forces patients to rely on alternative forms of transportation to reach the hospital. Hence, PHD was not decreased by the use of ambulances in this study. Instead, stroke patients who were taken to the hospital by private or public transportation were 74% and 83% less likely to be late, respectively. This result warrants re-evaluating the emergency medical response and transportation policies. Adaptation of emergency medical response service via a telephone-aided emergency first responders-led preclinical care can help lower the fatalities and delays.²³

Referral pattern is a significant driver of delay. If patients arrived through referral, the PHD was, naturally, significantly increased. In this study, a large number of patients (68.53%) were referred by primary care, hospitals and physicians, which caused significant delays. Besides, the majority of study participants could have been assessed for thrombolytic therapy if the first medical facilities had access to acute stroke care. The need for better healthcare facilities in Bangladesh is not going to be resolved soon; however, creating awareness of the alarming symptoms and benefits of early management can play a vital role as an immediate step.

In this study, the median decision delay was 3 hours. Unfortunately, the reasons for the significant delay in making a decision were obscure. Still, it might be presumed that they were unaware of the need to get medical attention. Here, stroke patients were observed to be taking medication on their own without consulting a physician. This practice caused significant delays in hospitalisation, which supports the findings of another study by Rabin et al.²⁴ In case of younger individuals’ stroke is not readily considered despite having the stroke symptoms, mainly due to the perception that stroke affects the elderly.²⁵ Besides, a study by Barber et al showed that some individuals may decide to wait to see if symptoms improve, which can lead to delays in seeking medical assistance.²⁶ Organisational challenges, uncertainty regarding the onset of symptoms, waiting for symptoms to resolve, perceived severity of symptoms, and clinical improvement are all factors that contribute to this delay, according to the study. These factors underscore the necessity of immediate medical intervention and the lack of patient education and awareness regarding stroke symptoms. The results emphasise the necessity of targeted public health initiatives to enhance awareness and education regarding stroke symptoms and the significance of a prompt medical response. In order to enhance patient outcomes and ensure that a greater number of individuals benefit from available treatments such as intravenous tissue plasminogen activator (IV TPA), it is imperative to address these issues. It is also important to understand how people perceive their symptoms and what action should be taken.

In the current study, mNIHSS score, GCS and mRS were used to understand the clinical condition of the patients. According to the study, the mNIHSS and mRS score were found to be significantly associated with the hospital arrival of the patients. In patients who arrived late, the median mNIHSS score was 12 (vs 10 in early arrival group) and the median mRS score was 4 (vs 3 in early arrival group), which indicates poorer clinical condition^{27 28} in late hospital arrival group. While the late arrival could be responsible for this poorer clinical outcome, we still lack substantial evidence in this regard. Future studies can help understand this relationship.

Our study explored parameters linked with PHD in Bangladeshi stroke patients. Patients’ ignorance of stroke symptoms and poor perception of stroke consequences would delay prehospital care. Prior studies also highlighted lack of understanding and awareness of typical stroke symptoms can lead to misinformation and improper treatment seeking.²⁹ Suspected stroke patients should be encouraged to attend facilities with stroke treatment. Reducing referral times should be considered. Besides, stroke screening and thrombolysis in primary care may prevent severity and complications.

Since this was a cross-sectional survey, causality cannot be inferred for the associations presented in this study. However, by presenting the adjusted ORs from the multiple logistic regression model, we attempted to account for the effect of the confounders. Although the participants were enrolled within 3 days of the hospital admission to ensure sufficient time for the mental adjustment of the patient/guardian after the incident to get quality data during the interview. To reduce recall bias, we attempted to recruit patients as early as possible after their admission. Although we tried to include all the admitted stroke patients during the study period, we could not enrol patients discharged before enrolment and patients with severe symptoms (admitted into intensive care units, dead), which may have introduced selection bias.

Conclusion

This is the first multicentre study to assess the PHD of acute stroke patients in Bangladesh and to investigate the factors associated with delayed hospital arrival. An alarmingly 87.72% of stroke patients arrived late (>4.5 hours after symptom onset) in the hospital, and a median of 3 hours was required to make a decision by the relatives/bystanders of the patients to seek medical help. Therefore, mass public information campaigns on stroke symptom recognition, patient-friendly transport infrastructures and establishing more dedicated stroke centres are a must to ensure that patients receive quality care within the optimum time.

Data availability statement

Data are available upon reasonable request.