Sociodemographic and clinical characteristics of stroke patients
During the study period from 1st September 2020 to 1st December 2021, a total of 259 stroke patients were treated at DBCS Hospital. From the total patients, this study included 203 stroke patients with complete data on associated variables of interest. Of the 203 stroke patients, 152 (74.88%) were censored or did not live to see the event, and 51 (25.12%) died. Of the total patients included, 99 (48.77%) were male, of whom 17 (33%) died, while 104 were female, of whom 34 (67%) died. Among 124 (61.08%) patients with hypertension, of which 40 patients (78.4%) died and 79 (38.92%) patients without hypertension, of which 11 patients (21.5%) died (Table 1).
Of all patients, 95 (46.80%) were urban patients, of those 25 (49%) died, and 108 (53.20%) were rural patients, of which 26 (51%) died. Of the patients included in the study, 62 (30.54%) had heart disease, 17 (33%) died, and 34 of 141 (69.46%) who did not have heart disease 34 (67%) died. There are 129 (63.55%) ischemic patients, of which 36 (70.5%) have died, 56 (27.59%) are non-heamorrhagic patients, 12 (23.5%) have died and 18 (9.59%) suffered from both types of stroke, of which 3 (5.8 %) died (Table 1).
Table 1 Sociodemographic and clinical characteristics of stroke patients at DBCSH from 1st September 2020 to 1st December 2021 follow-up periods.
Variable
|
categories
|
Frequency (%)
|
Death (%)
|
Censored (%)
|
Sex
|
Male
|
99 (48.77)
|
17 (33)
|
82 (54)
|
Female
|
104 (51.23)
|
34 (67)
|
70 (46)
|
Hypertension
|
No
|
79 (38.92)
|
11 (21.5)
|
68 (44.7)
|
Yes
|
124 (61.08)
|
40 (78.4)
|
84 (55.2)
|
Residence
|
Urban
|
95 (46.80)
|
25 (49)
|
70 (46)
|
Rural
|
108 (53.20)
|
26 (51)
|
82 (54)
|
Cardiac disease
|
No
|
141 (69.46)
|
34 (67)
|
107 (70.4)
|
Yes
|
62 (30.54)
|
17 (33)
|
45 (29.6)
|
Diabetes mellitus
|
No
|
158 (77.83)
|
44 (86.3)
|
114 (75)
|
Yes
|
45 (22.17)
|
7 (13)
|
38 (25)
|
Baseline complication
|
No
|
98 (48.28)
|
18 (35.3)
|
80 (52.6)
|
Yes
|
105 (51.72)
|
33 (64)
|
72 (47.3)
|
Stroke type
|
Ischemic
|
129 (63.55)
|
36 (70.5)
|
93 (61)
|
Haemorrhagic
|
56 (27.59)
|
12 (23.5)
|
44 (29)
|
Both
|
18 (9.59)
|
3 (5.8)
|
15 (10)
|
Age
|
Min.
|
Max.
|
Mean
|
Std. Dev.
|
22
|
96
|
61.54
|
14.32
|
Time
|
N
|
Median (50 %)
|
[95% CI]
|
|
203
|
4.438
|
4.02 4.86
|
Among patients considered in the study, 45 (22.17%) were diabetic, of which 7 (13%) died, 158 (77.83%) were non-diabetic in which 44 (86.3%) died. Of the total patients, 105 (51.72%) had baseline complications, of which 33 (64%) died, 98 (48.28%) had no baseline complications of which 18 (35.3%) dead. The average age of the patients was 61.54 years, the oldest and youngest were 96 and 22 years, respectively, and the standard deviation was 14.32. Finally, the median survival time of patients was 4.438 months (Table 1).
Overall stroke patients’ status
Of the total 259 stroke patients who were treated at DBCS hospital from 1st September 2020 to 1st December 2021 follow-up periods, 25.12% of them were died and the remaining 74.88% were survived during the specified period (Figure 1).
Figure 1 shows stroke patients’ status during the study periods.
Kaplan Meier Estimate of stroke patients
Figure 2 shows the overall survivor estimates from the Kaplan-Meier and hazard functions, and it shows a relatively large number of deaths that occurred early in stroke treatment (Figure 2 (a)). Similarly, patient risk or mortality increased significantly early in the follow-up period (Figure 2 (b)). This seems to be inversely proportional to the survival function, that is, the risk increases as the patient's survival decreases (Figure 2).
Figure 2 (a): Survival Function curve
Figure 2 (b): Hazard Function curve
Figure 2: The K-M plots of survival and hazard function of stroke patients at DBCSH during 1st September 2020 to 1st December 2021 follow-up periods.
Furthermore, plots of Kaplan-Meier survival function estimates for patients based on various categories of factors were also generated. For instance, the graph in Figure 3 (left) shows significant differences between patient sex groups. In other words, relatively high survival rates are observed in males. Similarly, Figure 3 (right) shows that better survival was observed in patients without hypertension or those without hypertension have a longer survival time than those with hypertension (Figure 3).
Figure 3: Plot of Kaplan-Meier curves by sex (left side) and hypertension (right side) of stroke patients at DBCSH, respectively, from 1st September 2020 to 1st December 2021 follow-up periods.
Log-rank test
In log rank test of equality of survival functions, a statistically significant difference in survival time was observed among patient groups of gender, hypertension, cardiac disease, and baseline complications at 5% level of significance. However, there are no statistically significant differences in survival experience between patient categories such as diabetes mellitus, stroke type, and residence (Table 2).
Table 2 Log-rank test in stroke patient characteristics for each categorical variables at DBCSH from 1st September 2020 to 1st December 2021 follow-up periods.
Predictor
|
Chi-square of log rank
|
df.
|
p-value
|
Sex
|
6.97
|
1
|
0.0083*
|
Residence
|
0.05
|
1
|
0.8182
|
Stroke type
|
1.71
|
2
|
0.4243
|
Hypertension
|
9.25
|
1
|
0.0024*
|
Cardiac disease
|
1.36
|
1
|
0.0394*
|
Diabetes mellitus
|
2.58
|
1
|
0.1085
|
Baseline complication
|
4.14
|
1
|
0.0420*
|
* P-value < 0.005 at 5% level of significant
Bivariate Cox PH regression model
Instead of immediate access to the all-inclusive Cox PH multiple regression model presented to determine baseline factors for stroke patient survival, bivariate analyses were performed at the 25% significance level to see the impact of each predictor on stroke patient survival. . Accordingly, predictors such as gender, age, hypertension, diabetes, heart disease and underlying complications were identified as contributing factors and candidates for multivariate analysis (Table 3).
Table 3 Bivariate Cox PH regression estimates of stroke patients at DBCSH from 1st September 2020 to 1st December 2021 follow-up periods.
Predictor
|
Category
|
CHR
|
P-value
|
75% CI
|
Sex
|
Male-reference
|
1
|
|
|
|
Female
|
2.103
|
0.012
|
1.174 3.767
|
Residence
|
Urban-reference
|
1
|
|
|
|
Rural
|
.940
|
0.824
|
.543 1.627
|
Age
|
None
|
1.037
|
0.000
|
1.016 1.058
|
Stroke type
|
Ischemic- reference
|
1
|
|
|
|
Hemorrhagic
|
.716
|
0.316
|
.372 1.376
|
|
Both types
|
.585
|
0.372
|
.180 1.899
|
Hypertension
|
No- reference
|
1
|
|
|
|
Yes
|
2.627
|
0.005
|
1.347 5.124
|
Cardiac disease
|
No-reference
|
1
|
|
|
|
Yes
|
1.188
|
0.026
|
1.066 2.127
|
Diabetes mellitus
|
No- reference
|
1
|
|
|
|
Yes
|
.537
|
0.126
|
.242 1.192
|
Baseline complication
|
No-reference
|
1
|
|
|
|
Yes
|
1.767
|
0.052
|
.995 3.138
|
Multivariable Cox PH regression model
Based on the final fitted Cox PH regression estimation, the hazard ratio for females was 2.894, which implies that females had 2.894 times higher risk of death than males (95% CI: 1.574, 5.320, p = 0.001). Patients' age was also found to be a significant factor for their survival (AHR = 1.041, 95% CI: 1.017, 1.065 and p = 0.002). This suggests that the probability of death in stroke patients increases by 1.041 times as their age increases by one month. Patients with hypertensive stroke were 2.575 times more likely to die than those without hypertension (AHR = 2.575, 95% CI: 1.295, 5.121, p = 0.007). Furthermore, heart disease was also considered as one of the contributing factors to reduce the survival time of stroke patients, with an estimated hazard ratio of 1.483 (95% CI: 1.059, 1.972 and p = 0.042). The estimate indicated that stroke patients with heart disease were 1.483 times more likely to die than patients without heart disease (Table 4).
Table 4 Multivariable Cox PH regression estimates of stroke patients at DBCSH from 1st September 2020 to 1st December 2021 follow-up periods.
Predictor
|
Category
|
AHR
|
P-value
|
95% CI
|
Sex
|
Male-reference
|
1
|
|
|
|
Female
|
2.894
|
0.001
|
1.574 5.320
|
Age
|
Continuous
|
1.041
|
0.002
|
1.017 1.065
|
Diabetes mellitus
|
No- reference
|
1
|
|
|
|
Yes
|
.630
|
0.271
|
.276 1.435
|
Hypertension
|
No- reference
|
1
|
|
|
|
Yes
|
2.575
|
0.007
|
1.295 5.121
|
Cardiac disease
|
No- reference
|
1
|
|
|
|
Yes
|
1.483
|
0.042
|
1.059 1.972
|
Baseline complication
|
No- reference
|
1
|
|
|
|
Yes
|
1.166
|
0.791
|
.596 1.972
|
Test of proportional hazard assumption
The global (overall) schoenfeld residual test was not significant (p = 0.110), indicating that the proportionality hypothesis is satisfied at the 0.05 significance level. In addition, almost each predictor showed no statistically significant p-values, implying that the corresponding hazard hypothesis should not be rejected (see Table 5).
Table 5 Schoenfeld residual estimates used for testing proportional hazards assumption among stroke patients at DBCSH from 1st September 2020 to 1st December 2021 follow-up periods.
Predictor
|
Category
|
rho
|
|
Df.
|
Prob.>
|
Sex
|
Female
|
0.012
|
0.01
|
1
|
0.924
|
Age
|
None
|
-0.040
|
0.09
|
1
|
0.769
|
Residence
|
Rural
|
-0.149
|
1.39
|
1
|
0.239
|
Stroke type
|
Hemorrhagic
|
0.057
|
0.18
|
1
|
0.673
|
Both types
|
-0.232
|
2.77
|
1
|
0.096
|
Hypertension
|
Yes
|
-0.335
|
5.67
|
1
|
0.017
|
Cardiac Disease
|
Yes
|
-0.191
|
2.48
|
1
|
0.115
|
Diabetes Mellitus
|
Yes
|
0.033
|
0.06
|
1
|
0.814
|
Baseline Complication
|
Yes
|
-0.055
|
0.18
|
1
|
0.676
|
Global test
|
|
14.35
|
9
|
0.110
|
Graphical test of proportional assumption
In Figure 4 (a, b and c), -ln (-ln (survival)) cells are often referred to as "log-log" cells for each category of a nominally independent variable as a function of follow-up time that shows testing the proportional hazard hypothesis. Therefore, the curves are parallel to each other, implying that the assumption of proportional risk is not violated. Here, each estimate is obtained by adjusting for the other covariates. From Figure 4 (b), patients without diabetes unfortunately had a shorter survival time than patients with this disease.
Figure 4 (a): Patients' hypertesion status
Figure4 (b): Patients' diabetic mellitus status
Figure 4 (c): Patients' baseline complication
Figure 4: The -ln(-ln S(t)) vs time plot is the graphical test of proportional hazards (PH) assumption of Cox PH model using some stroke patients’ health statuses, such as patients’ hypertesion (a), diabetic mellitus (b), and baseline complication (c) statuses.