Quality of early stroke care and long-term mortality in patients with acute stroke: A nationwide follow-up study

Abstract

Introduction:

High quality of early stroke care is essential for optimizing the chance of a good patient outcome. The quality of care may be monitored by process performance measures (PPMs) and previous studies have found an association between fulfilment of PPMs and short-term mortality. However, the association with long-term mortality remains to be determined. We aimed to evaluate the association between fulfilment of PPMs and long-term mortality for patients with acute stroke in Denmark.

Patients and methods:

We used data from Danish health care registers between 2008 and 2020 to identify all patients admitted with incident stroke (haemorrhagic (ICH) or ischaemic stroke). The quality of early stroke care was assessed using 10 PPMs. Mortality was compared using Cox proportional hazard ratios, risk ratios computed using Poisson regression, and standardized relative survival.

Results:

We included 102,742 patients; 9804 cases of ICH, 88,591 cases of ischaemic stroke, and 4347 cases of unspecified strokes. The cumulative 10-year mortality risk was 56.8%. Fulfilment of the individual PPMs was associated with adjusted hazard rate ratios of death between 0.76 and 0.96. Patients with 100% fulfilment of all PPMs had a lower 10-year post-stroke mortality (adjusted risk ratio 0.90) compared to the patients with 0%–49% fulfilment and a standardized relative survival of 81.3%, compared to the general population.

Conclusion:

High quality of early stroke care was associated with lower long-term mortality following both ICH and ischaemic stroke, which emphasizes the importance of continued attention on the ability of stroke care providers to deliver high quality of early care.

Graphical abstract.

Introduction

In stroke care, early recognition, fast admission to stroke centres, and a multidisciplinary team-based treatment approach including both pharmacological and non-pharmacological patient specific care, are significant components of high treatment quality.^1–3 Fulfilment of evidence-based stroke process performance measures (PPMs) based on international guidelines on early stroke care has been reported to be associated with lower all-cause mortality and improved functional outcomes albeit results were ambiguous.^4–14 To ensure high standards of stroke care, continuous monitoring of these evidence-based PPMs is of the essence, and there is an ongoing effort to optimize and implement this practice worldwide.^4,15 PPMs should ideally reflect multiple domains of care, for example, admission to a stroke unit, early brain imaging, early mobilization, screening for dysphagia, and antiplatelet therapy. Fulfilment of selected PPMs has been associated with lower short-term (30–90-days) mortality, but it remains to be determined if the effects are sustainable with regard to long-term mortality in the same registry population. ¹⁶ We hypothesized that the level of fulfilment of PPMs in stroke care is associated with lower all-cause long-term mortality. To this end, we evaluated both the combined, as well as the individual role of each evidence-based PPM of early stroke care, and relative survival long-term among Danish patients with incident acute stroke.

Patients and methods

Study context, design, and data sources

The Danish public healthcare system provides free and universal health care coverage. All Danish inhabitants are assigned a unique 10-digit personal identifier number, which allows linkage of nationwide administrative registries at an individual level. ¹⁷ Patients with suspected stroke are all exclusively treated at public hospitals. The Danish Stroke Registry (DSR) is a nationwide clinical registry, including all adult patients with acute stroke or transient cerebral ischaemic attack (TIA) in Denmark since 2003. ¹⁵ It is mandatory for Danish hospitals to report all acute stroke admissions to the registry. The DSR contains prospectively collected data on evidence-based PPMs of early stroke care as well as sociodemographic, medical and lifestyle prognostic factors. ¹⁵ The registration of stroke events in the DSR have been reported to have a high sensitivity and positive predictive value (>90%). ¹⁵ Furthermore, the completeness of the registration of variables is also above 90% for all variables. ¹⁸ We linked data from the DSR with the Danish Civil Registration System (DCRS), and the Danish National Patient Register (DNPR).^15,17,19 DCRS contains information on birth date, sex, vital status, residency, and emigration. ¹⁷ Vital status is updated daily, where the date of death is available. DNPR includes information on all hospital contacts, including hospital department, date and time of admission and discharge, diagnoses and type of treatment. ¹⁹ Data on the general population (lifetables) was obtained from Statistics Denmark. ²⁰

Study population

The DSR was used to identify all stroke admissions from January 1st, 2008 to December 31st, 2020. The population was subsequently, restricted to patients with incident (first-time) strokes. ¹⁵ We considered a stroke episode to be incident if the patient did not have a previous stroke event recorded in DSR from 2003, DSR did not otherwise indicate a history of stroke, and the patient did not have a stroke registered in DNPR at any previous time. All patients (⩾18 years) were included if one of the following ICD-10 codes were applied; ICH (I61), cerebral infraction (I63), and unspecified stroke (not specified as haemorrhage or infarction) (I64). Patients were excluded if the verified diagnosis was TIA (G45), information on vital status was missing, or if the patient was not eligible for a minimum of three PPMs.

Patient characteristics

From the DSR we retrieved information on age, sex, marital status, housing, stroke subtype, stroke severity (Scandinavian Stroke Scale (SSS)) on admission, comorbidities (acute myocardial infarction, atrial fibrillation, hypertension, diabetes mellitus), and information on treatment with intravenous thrombolysis or/and thrombectomy (Table 1 and Supplemental Table 1).^15,21 Birth dates were obtained from DCRS ¹⁷ and linked with admission dates from the DSR to verify age at admission. ¹⁵

Table 1.

Characteristics of cohort including 102,742 patients with incident stroke. Patient characteristics were identified during admission for incident stroke and were reported in Danish Stroke Registry, 2008–2020. Co-morbidities were determined based on medical history at the time of admission, or during workups in stroke unit. Revascularisation includes both thrombolysis and thrombectomy.

Patient characteristics	N (%)
Age
18–65	33,401 (32.5)
>65–80	42,438 (41.3)
>80	26,903 (26.2)
Age
Median, IQR	72 (62–81)
Sex
Male	55,450 (54.0)
Female	47,292 (46.0)
Marital status
Living with a partner	57,213 (55.7)
Living alone	41,025 (39.9)
Housing
Undisclosed	4504 (4.4)
Own home	93,335 (90.8)
Nursing home or institution	5306 (5.2)
Undisclosed	4101 (4.0)
Type of stroke
Intracerebral haemorrhage I61	9804 (9.5)
Cerebral infarction I63	88,591 (86.2)
Unspecified I64	4347 (4.2)
Scandinavian Stroke Scale
Severe 0–25	12,775 (13.0)
Moderate > 26–42	17,180 (17.5)
Mild > 43–58	68,317 (69.5)
Undisclosed	4470
Acute myocardial infarction
Yes	7386 (7.2)
No	93,254 (90.8)
Undisclosed	2102 (2.0)
Atrial fibrillation
Yes	17,178 (16.7)
No	84,093 (81.8)
Undisclosed	1471 (1.4)
Hypertension
Yes	54,467 (53.0)
No	46,335 (45.1)
Undisclosed	1940 (1.9)
Diabetes mellitus
Yes	14,020 (13.6)
No	87,390 (85.1)
Undisclosed	1332 (1.3)
Treatment with thrombolysis after 2008
Yes	14,523 (14.1)
Treatment with thrombectomy after 2008
Yes	3168 (3.1)
Revascularization after 2008
Yes	15,419 (15.3)

Process performance measures

The quality of early stroke care was assessed using 10 predefined PPMs for early stroke care (Supplemental Table 2). ¹⁵ The PPMs were defined by a multidisciplinary national committee of clinical stroke experts who have been responsible for selection based on a review of scientific evidence, clinical relevance, and pragmatic consideration of the feasibility of obtaining valid nationwide data in in routine clinical practice.

The number of patients included in the analysis for a specific PPM reflected a varying proportion of patients eligible for individual PPMs, for example, 65,737 out of 102,698 patients were considered eligible for mobilization by the staff treating the patient, and out of them 54,801 received the treatment within the first day of admission which corresponded to 83.4% fulfilment. The registration in DSR could either be yes, no, undisclosed, or not eligible. Not eligible were patients for example, who were moribund on arrival or patients with very mild symptoms, where some processes of care like physical and occupational therapy assessment were considered clearly irrelevant. A timeframe was defined for each individual PPM to capture the timeliness of the care intervention (Supplemental Table 2). Quality of early stroke care was assessed both as fulfilment of individual PPMs and as an opportunity-based score (composite measure). ²² Fulfilment was divided into four categories based on percentage of relevant PPMs fulfilled: 0%–49% (very low quality of care; reference group), 50%–74% (low quality of care), 75%–99% (moderate quality of care), and 100% (high quality of care). To support the categories, we performed a sensitivity analysis and displayed risk ratios for each relevant level of fulfilment with 100% fulfilment as reference. We also displayed results from related models where the level of fulfilment was included as a continuous, splined predictor (Supplemental Table 3 and Figures 1–6).

Outcome

Mortality was defined by vital status in DCRS during follow-up. ¹⁷ The study outcome was all-cause mortality.

Statistics

Continuous variables were described with medians and interquartile ranges. Binary and categorical variables were described with counts and percentages. All estimates were presented with 95% confidence intervals (CI).

The cumulative risk of death in the main cohort was estimated utilizing the Kaplan Meier estimator.

We estimated the hazard rate ratios (HRR) of mortality for fulfilling each of the 10 individual PPMs with Cox proportional hazard regression. Violations of the proportional hazard’s assumption was assessed using visual inspection of Andersen plots. The dose-response relationship between PPM-fulfilment and mortality was analysed in two ways. First, using Poisson regression, we estimated Risk-Ratios (RR) between ‘very low fulfilment’ and each of the three other fulfilment categories. Both the HRRs in the individual PPM analyses, and the RRs in the dose-response analyses were adjusted for sex, age, marital status, housing, stroke type, SSS, prior stroke, myocardial infract, atrial fibrillation, hypertension, diabetes, and used cluster robust variance to account for within hospital correlation. Secondly, we estimated the age, sex, and calendar year standardized relative survival (RelS), using the external STATA-package Strel2. ²³ The RelS-estimates were standardized to the entire Danish population.

All patients were followed from day five after hospitals admission until date of death or end of follow-up (31 December 2021), which ever came first. Analyses on the individual PPMs were restricted to patients considered eligible for each of the PPMs and follow-up in these analyses started 5 days after stroke admission which allowed for completion of the individual PPM. Patients were excluded if they died within the time allowed to complete individual PPMs, thus not permitting fulfilment. In the dose-response analyses based on the opportunity-based composite score of overall PPM fulfilment, follow-up started 5 days after the date of stroke admission. ¹⁵ All missing values in covariates were imputed using multiple imputation with chained equations. Imputations were made using a 10 nearest neighbour model and 10 imputed data sets were generated. Additional sex- and age-stratified analyses was performed to identify potential effect modification. To investigate whether the increased use of revascularisation therapy influenced the association between fulfilment of the PPMs and long-term mortality, we conducted three sensitivity analyses. In the first analysis, the population was restricted to patients who had received revascularisation therapy. In the second analysis, the study period was stratified to cover 2008–2014 and 2015–2020, respectively. In the third sensitivity analysis of the association with cardiovascular mortality was performed. We performed all data management and statistical analyses in Stata 16.1 (StataCorp, 2017, College Station, TX). ²⁴

Ethics

The study was register-based and as per Danish regulations individual consent of participation was not required. Due to confidentiality these data cannot be shared. However, data can be accessed through the Danish Health Data Authority and Statistics Denmark by researchers employed by authorized institutions and with the required approval from the Danish Data Protection Agency.

Results

In total, 102,742 patients were included in the main cohort, of which 9804 patients had ICH, 88,591 with ischaemic stroke, and 4347 patients with unspecified stroke (Figure 1). Stroke severity at the time of stroke unit admission was classified as mild in 70% of the patients. ²¹ The median age at incident stroke was 72 years [IQR 62–81] and 46% were women. A total of 14,523 (16.06%) patients received treatment with intravenous thrombolysis, whereas 3168 (3.6%) patients were treated with thrombectomy out of 88,591 (17.4% revascularized) patients with ischaemic stroke. The cumulative number of deaths upon 10 years follow-up from admission were 43,120 cases. Median follow-up time was 1566 days (705:2270) corresponding to 4.11 years (1.69:7.53) (Table 1, Figure 1, and Supplemental Table 4). A total of 2314 patients died within 4 days. Compared to patients with 100% fulfilment of PPMs, the patients who died early and did not survive until start of follow-up were older (median 81 years (72–87) vs 71 years (61–80)), had more severe stroke (median SSS 11 (4–26) points vs 52 (42–56) points), female sex 44.2% versus 56.1%, and living in a nursing home 13.5% versus 4.2% (Supplemental Table 1).

Figure 1.

Flowchart of study population selection.

The fully adjusted HHRs ranged from 0.76 (0.70:0.81) for treatment with antiplatelet to 0.96 (0.90:1.02) for carotid vessel imaging (Table 2). Poisson regression analysis indicated a dose-response relationship pattern between the percentage of relevant PPMs fulfilled (opportunity-based score) and 10-years mortality. For patients who had 100% fulfilment of PPMs mortality was lower (adjusted RR = 0.90 (0.88:0.93)) compared to 0%–49% fulfilment of PPMs after 10-years (Table 3 and Figure 2). The supplementary sex- and age-stratified analyses for 10-years mortality showed that the association between fulfilment of PPMs and 10-years mortality was independent of sex, but diminish with higher age, that is, the adjusted RRs among women were 0.56 (0.48:0.66) in age group 18–65 years, 0.85 (0.79:0.91) in age group > 65–80 years and 1.00 (0.98:1.02) among women above 80 years (Supplemental Tables 5–7). The corresponding adjusted RRs for men were 0.76 (0.68:0.86) in age group 18–65, 0.91 (0.86:0.96) in age group > 65–80 years and 0.97 (0.94:1.00) above 80 years (Supplemental Tables 8–10).

Table 2.

Association between individual process performance measures and 10 years all-cause mortality.

Hazard rate ratio for mortality
Process performance measures	Unadjusted	Adjusted for sex and age	Fully adjusted
Admission to a stroke unit	0.74 (0.69:0.79)	0.81 (0.76:0.86)	0.83 (0.79:0.88)
Examination with CT/MRI	0.88 (0.84:0.92)	0.96 (0.93:1.00)	0.88 (0.85:0.92)
Mobilization	0.58 (0.55:0.62)	0.67 (0.64:0.71)	0.85 (0.81:0.88)
Indirect dysphagia screening	0.78 (0.75:0.81)	0.85 (0.82:0.89)	0.90 (0.87:0.94)
Assessment by physiotherapist	0.86 (0.82:0.91)	0.85 (0.82:0.88)	0.89 (0.85:0.92)
Assessment by occupational therapist	0.86 (0.81:0.91)	0.85 (0.82:0.89)	0.89 (0.84:0.93)
Assessment for nutritional risk	0.67 (0.63:0.71)	0.76 (0.72:0.80)	0.87 (0.84:0.89)
Treatment with antiplatelet	0.56 (0.52:0.60)	0.65 (0.61:0.69)	0.76 (0.70:0.81)
Carotid vessel imaging with ultrasound	0.92 (0.87:0.99)	0.92 (0.87:0.98)	0.96 (0.90:1.02)
Treatment with anticoagulant	0.67 (0.60:0.75)	0.73 (0.66:0.80)	0.83 (0.73:0.96)

Fully adjusted: Sex, age, marital status, housing, stroke type, Scandinavian stroke scale, prior stroke, myocardial infarct, atrial fibrillation, hypertension, diabetes, and department/hospital.

Table 3.

Mortality according to proportion of fulfilled process performance measures (opportunity-based composite measure) among patients with incident stroke.

Adjusted risk ratios
Time from admission	Fulfilment 0%–49%	Fulfilment 50%–74%	Fulfilment 75%–99%	Fulfilment 100%
30-days	1	0.82 (0.75:0.90)	0.73 (0.67:0.80)	0.71 (0.65:0.78)
90-days	1	0.81 (0.77:0.85)	0.71 (0.67:0.75)	0.68 (0.64:0.72)
365-days	1	0.84 (0.81:0.87)	0.75 (0.71:0.79)	0.68 (0.65:0.72)
3-years	1	0.87 (0.84:0.89)	0.80 (0.78:0.83)	0.74 (0.72:0.77)
5-years	1	0.91 (0.88:0.93)	0.87 (0.84:0.89)	0.81 (0.78:0.83)
6-years	1	0.93 (0.90:0.95)	0.88 (0.86:0.91)	0.83 (0.80:0.85)
7-years	1	0.93 (0.91:0.95)	0.90 (0.87:0.93)	0.85 (0.82:0.87)
8-years	1	0.93 (0.91:0.95)	0.91 (0.88:0.94)	0.86 (0.84:0.88)
9-years	1	0.94 (0.92:0.97)	0.92 (0.90:0.95)	0.89 (0.87:0.91)
10-years	1	0.94 (0.92:0.97)	0.92 (0.89:0.95)	0.90 (0.88:0.93)
11-years	1	0.96 (0.94:0.99)	0.93 (0.90:0.95)	0.92 (0.89:0.94)
12-years	1	0.97 (0.93:1.00)	0.92 (0.89:0.96)	0.93 (0.89:0.96)
13-years	1	0.99 (0.95:1.03)	0.94 (0.91:0.97)	0.94 (0.90:0.98)

Fully adjusted: Sex, age, marital status, housing, stroke type, Scandinavian stroke scale, prior stroke, myocardial infarct, atrial fibrillation, hypertension, diabetes, and department/hospital.

Figure 2.

Mortality after admission for incident stroke by fulfilment level among patients who survived 4 days, were relevant for minimum 3 process performance measures, and according to percentage fulfilled.

Standardized relative survival analysis showed that compared to the general population, patients with 100% fulfilment of PPMs had a 19% (RelS = 81.3 (80.3:82.2)) higher 10-years mortality, whereas patients with 0%–49% of PPMs fulfilled had a 40% (RelS = 60.2 (58.0:62.2)) higher mortality (Figure 3 and Supplemental Table 11).

Figure 3.

Mortality after admission for incident stroke by fulfilment level among patients who survived 4 days, were relevant for minimum 3 process performance measures, and according to percentage fulfilled compared to the general population (consisted of the entire Danish population including our main cohort).

When restricting the population to patients who did not receive revascularization 10-years mortality RR 0.83 (0.80:0.86), and RR 0.96 (0.85:1.09) for those who received revascularisation therapy (Supplemental Tables 12–13). When stratified according to study period to 2008–2014 (Supplemental Tables 14–18). Versus 2015–2020 (Supplemental Tables 19–23), we found that there was a difference in the association between the PPMs and mortality (Supplemental Tables 15 + 20). Additionally, when restricting the outcome to cardiovascular the associations were slightly stronger as compared with the associations with all-cause mortality (Supplemental Table 24–28).

Discussion

In this nationwide, population-based study, we found fulfilment of relevant PPMs reflecting the quality of early stroke care to be associated with lower 10-years all-cause mortality after acute stroke. Some indication of a dose-response pattern was observed when analysing the association between an opportunity-based composite score of quality of early stroke care and mortality, although this pattern appeared to be weakened with duration of follow-up. When restricting the population, the time period 2015–2020 is more representative of contemporary stroke care. Regarding thrombectomy the treatment was first generally used from 2015. Also, the associations between quality of early stroke care and long-term mortality appeared diminished with increasing age, whereas no sex-related differences were observed. Long-term mortality, especially among the oldest patients, will be influenced by many factors, for example, co-morbidities that develop during long-term follow-up. It is thus likely more challenging to demonstrate long-term benefits of acute stroke treatment in this population. This should, however, not lead to a nihilistic approach to treatment in the elderly as early benefits are also demonstrated in this patient group.

The strengths of this study included the use of a nationwide stroke registry with prospectively collected detailed individual patient data on fulfilment of PPMs for early stroke care from all stroke units in Denmark. Nonetheless, the accuracy of the recorded data is supported by the high completeness of the variables, the detailed manual of data definitions that is used as a reference for all reporting departments and the annual audits of the reports from the registry, where the data and performance of all Danish stroke units is systematically evaluated to ensure a uniform registration practice and high quality of care. Also, we accounted for any clustering effect at the individual hospital level to consider that patients admitted to the same hospital may potentially be exposed to common factors related to local differences in care practice. Further, the analyses were restricted to incident strokes without contraindications for the specific PPMs. The major limitation is the observational study design. The design precludes the direct assessment of a cause-effect relationship due to the risk of confounding. However, we used different approaches to reduce the possible risk of confounding, including multivariable adjustment for potential confounding variables and restriction of the study population to patients, who were considered eligible for the individual PPMs being investigated. Another aspect was all-cause mortality as the chosen outcome of interest. Many patients may not die from causes related to stroke, which could affect the relative risk estimates in a conservative direction compared if the analysis was limited to causes directly related to death by stroke. However, it is complicated to accurately define if the cause of death is directly or indirectly related to a stroke event, that may have preceded death by several years. Further to this, patients may have had additional strokes after their incident episode. The quality of treatment in the subsequent hospitalizations would also affect the risk of death. Despite its importance, mortality is not the only relevant endpoint for patient with incident stroke. Examination of other endpoints for example, functional level after discharge would clearly also be of major relevance. However, unfortunately such data was not available for this study.

The current data support the opportunity-based composite score of quality of early stroke care, and mortality indicate a dose-response pattern. As such the panel is by no way exhaustive of all potentially relevant acute stroke care components. ²⁵ The associations between individual evidence-based PPMs for example, dysphagia screen^8–10 or measures compiled into bundles (temperature measurement, blood sugar measurement, and swallow management^6,11–13) and patient outcomes have previously been investigated, but with inconsistent results. Such inconsistent findings may arise from small sample sizes, uncertain selection of study populations, variable study designs (randomized, non-randomized, observational, registry-based) length of follow-up, varying composition of specific evidence-based PPMs, and variable choice of outcome for example, mortality or pneumonia.^6,8–14,16 It remains to be specified which PPM measures have the most significant impact in the stroke unit effect or if the stroke unit benefits are based on a composite impact. Further, it may be speculated that hospitals which fulfil a high proportion of PPMs also perform well on other aspects of treatment not included in the PPMs. The current study showed that early stroke care, exception for carotid artery imaging, was associated with lower all-cause long-term mortality. Furthermore, examining the potential mediating role of the quality of acute stroke care on inequities in 30-days mortality and 30-days readmission from an earlier study using the DSR data no, or only a minor, mediating role of lower quality of care was observed, which seems to be in contrast with the findings of our current study. ¹⁵ However, several factors may have contributed to this difference, including residual confounding in the previous study, the narrower set of PPMs assessed in the analysis and differences in the studied outcomes (30-days mortality/readmission vs long-term mortality). ²⁶ Our results supported the relevance of PPM monitoring. Such systematic monitoring and reporting are currently recommended in the Action Plan for Stroke. ³ It is evident, from previous and current data, that PPMs should include multi-disciplinary interventions reflecting different domains of early stroke care. For post- stroke survivors, future research needs to investigate the composite effect of a high level of stroke care on morbidity and functional level including the ability to return to work.

Conclusion

We found that high quality of early stroke care compared with very low quality of care was associated with lower long-term mortality and followed a dose-response relationship. These findings applied to both sexes under the age of 80 years. The results from our registry-based study indicated a link between a range of specific early stroke care elements and lower long-term all-cause mortality. This highlights the relevance of an early multidisciplinary approach and the importance of stroke centres dedicated to continued quality improvement in early stroke care.