Is the socioeconomic inequality in stroke prognosis changing over time and does quality of care play a role?

Abstract

Introduction:

In a publicly financed healthcare system we aimed to study the development in socioeconomic disparity in ischemic stroke outcomes over time. In addition, we study whether the healthcare system affects these outcomes through the quality of early stroke care when adjustments are made for various patient characteristics incl. comorbidity and stroke severity.

Patients and methods:

Using nationwide, detailed individual-level register-data we analysed how income-related and education-related inequality in 30-day mortality and 30-day readmission risk developed between 2003 and 2018. In addition, focusing on income-related inequality, we applied mediation analyses to estimate the mediating role of quality of acute stroke care on 30-day mortality and 30-day readmission.

Results:

A total of 97,779 individual ischemic stroke patients were registered in Denmark with a first ever stroke in the study period. Three-point-seven percent died within 30 days of their index-admission and 11.5% were readmitted within 30 days of discharge. The income-related inequality in mortality remained virtually unchanged over time from an RR of 0.53 (95% CI: 0.38; 0.74) in 2003–06 to RR 0.69 (95% CI: 0.53; 0.89)) in 2015–18 when high income was compared to low income (Family income-time interaction: RR 1.00 (95% CI: 0.98–1.03)). A similar but less uniform trend was found for the education-related inequality in mortality (Education-time interaction: RR 1.00 (95% CI: 0.97–1.04)). The income-related disparity in 30-day readmission was smaller than in 30-day mortality and it diminished over time from 0.70 (95% CI: 0.58; 0.83) to 0.97 (95% CI: 0.87; 1.10). The mediation analysis showed no systematic mediating effect of quality of care on neither mortality nor readmission. However, it cannot be ruled out that residual confounding may have washed out some mediating effects.

Discussion and Conclusion:

The socioeconomic inequality in stroke mortality and re-admission risk has yet to be eliminated. Additional studies from different settings are warranted in order to clarify the impact of socioeconomic inequality of quality of acute stroke care.

Keywords

Healthcare disparities quality of health care inequality mediation analysis

Introduction

Socioeconomic position is a well-documented risk factor for stroke both via an association with stroke risk factors such as diabetes, high blood pressure and smoking, and as an apparent independent risk factor.^1,2 Socioeconomic inequality is also present in the clinical outcomes following stroke, including mortality, recurrence and disability.^2–6 Over the past decades, the risk of dying following a stroke has declined,⁷ but very little is known about how inequality in stroke outcomes has developed over time.

In addition to the inequalities reported with respect to risk factors, incidence and prognosis, socioeconomic inequalities have also been reported in access to and quality of stroke care.^5,8–10 This entails a risk that healthcare systems may play a role in stroke outcome inequality. The socioeconomic inequality in quality of care has recently been found to persist despite systematic and prolonged efforts to minimise it.¹¹ However, it remains unclear to which extent inequity in quality of acute stroke care contributes to socioeconomic disparities in clinical outcomes among patients with stroke.

In a setting in which the health-policy explicitly includes ensuring equality in care, we aimed to examine the development in socioeconomic disparity in stroke outcomes over time and to study the potential impact of disparities in quality of acute stroke care. Specifically, we used detailed individual-level data to analyse the development in income-related and education-related disparity in 30-day mortality and 30-day readmission risk. In addition, focusing on income-related inequality, we aimed to estimate how much of the disparity in prognosis was associated with the differences in quality of care.

Methods

Study design, population and setting

A nationwide study of consecutive stroke patients was conducted based on public, nationwide, population-based registers. The study was set in Denmark where acute stroke care is exclusively provided at public hospitals. All hospital care is tax financed and free of charge for all citizens. More than 90% of all acute stroke patients were treated at specialized multidisciplinary stroke units in the study period. Care became more centralized during the period as the number of stroke units was reduced from >50 to 25, including eight centres providing i.v. thrombolysis and three comprehensive stroke centres providing both i.v. thrombolysis and thrombectomy.

All adult patients who were residents in Denmark and hospitalised between 1 January 2003 and 31 December 2018 with an ischemic stroke were eligible for inclusion. An ischemic stroke was defined using the International Classification of Diseases, 10th revision (ICD-10): I63 cerebral infarction. We included only patients’ first ever stroke event.

We identified the stroke patients in the Danish Stroke Registry (DSR), a national clinical quality database. It is mandatory for all Danish hospital departments providing early stroke care to report data on all stroke events to the DSR. The estimated sensitivity of the DSR is 97%.¹² The DSR was established in 2003 as part of a national initiative to monitor and improve quality of care in Denmark and the registry holds detailed data on the stroke events and the provided early stroke care.^12,13

Using the patients’ personal identification numbers, we linked the data from the DSR to other national registers.¹²

Exposure

The exposure studied was socioeconomic situation operationalised as highest achieved education and disposable family income. Education data were drawn from the Danish register of the citizens’ education – highest completed education at the time of the stroke.¹³ Using the 2011 International Standard Classification of Education (ISCED), we categorised education into low (0–2), medium (3–4) and high (5–8).¹⁴ Income data were drawn from the Danish National Family Income Register under Statistics Denmark, which holds official records from the Salary Information Register and the Central Taxpayers’ Register.¹⁵ The disposable income reflects a family’s spending opportunity and includes salaries, public pension, private pension and other earnings after taxes and interest expenses. We used the income from the year prior to each stroke to avoid any effect of the stroke on the patients’ income. Mimicking the income distribution in the elderly population where most are retired and simultaneously emphasising the focus on the disparities, we categorised family income as low (up to the 25th percentile), middle (the 25th–75th percentile) and high (from the 75th percentile). The categorisation was made for each calendar year separately to remove any inflation. Family income has previously been proposed as a socioeconomic exposure when studying an elderly population.^16–18

Outcomes

We studied two outcomes: all-cause mortality within 30 days of the index admission and all-cause readmission within 30 days of discharge. Information on patients’ mortality was drawn from The Danish Civil Registration System, which holds public records of all Danish citizens’ status including their home address, emigration date and date of death (12). Data regarding readmission were extracted from The Danish National Patient Register, which holds records of all Danish patients. Upon discharge from the hospital or after an outpatient visit, patients are registered with one primary diagnosis (and additional diagnoses when relevant) according to the ICD-10.^19,20

Quality of early stroke care

The quality of early stroke care is monitored in the DSR using a set of process performance measures. The measures are defined by a national multidisciplinary panel clinical expert panel and reflect key recommendations from the national clinical guidelines for early stroke care.^21,22 These comply with the guidelines of the European Stroke Organisation. The performance measures cover early diagnostics, treatment and rehabilitation and reflect areas of care which can be considered to be under the control of the health care system, that is, not dependent on actions from patients or relatives. Healthcare professionals’ complete registrations based on written instructions and individual evaluation of the patient.²² In case contraindications are identified by the clinical staff providing care for the patient, for example, severe dementia precluding use of oral anticoagulant therapy or projected early death making active rehabilitation unethical, the patient is considered ineligible for the care intervention and thus the given performance measure. The number and definitions of the performance measures have changed over the years; however, a core set of measures have remained unchanged and are relevant for the majority of the patient population enabling temporal analyses (Table 1). The selected measures have in previous studies been associated with clinical outcomes, including 30-day mortality and length of stay.^23,24 Based on this set of measures, we operationalised quality of care as the number of relevant measures met for each patient (ranging from 0 to 5).

Table 1.

Core performance measures.

• Ischaemic stroke patient without atrial fibrillation received platelet-inhibitor therapy within 2 days of admission / Ischaemic stroke patient with atrial fibrillation received oral anticoagulation therapy within 14 days after admission

• A stroke patient was assessed by a physiotherapist about the need for rehabilitation (including type and extent) within the second day of admission

• A stroke patient was assessed by an occupational therapist about the need for rehabilitation (including type and extent) within the second day of admission

• A stroke patient received a nutritional risk assessment within the second day of admission

• A stroke patient was admitted to a stroke unit within the second day of admission

Covariates

From the DSR we obtained data on the following sociodemographic characteristics: sex, age at the time of admission and cohabitant status (whether they live alone or not). We also obtained data regarding the patients’ health status from the DSR: stroke severity (very severe, severe, moderate or mild according to the Scandinavian Stroke Scale Score at the time of the stroke admission²⁵) and their stroke-related comorbidity prior to or diagnosed during the stroke admission. The stroke-related comorbidity included atrial fibrillation, hypertension, diabetes, and acute myocardial infarction. Lastly, as a proxy for health-related behaviour, we retrieved smoking status (yes or no) from the DSR.

Statistical analyses

All effects were represented by risk ratios (RR) and estimated using Poisson regression with a robust variance estimator to account for clustering of patients at hospital level – that is. the observations were not independent. Assuming that data were missing at random, we applied multiple imputation (MI) using chained equations for missing data in confounding variables and generated 10 datasets for each of the analyses. Outcome, exposure, covariates and the composite quality of care measure were included as predictors in the MI models. For each set of exposure, outcome and time-period, we applied three models; an unadjusted, an adjusted by sex and age, and a fully adjusted model using all covariates described above. The time-trend analyses of the association between the exposures (education and family income, modelled separately) and each of the clinical outcomes (30-day mortality and 30-day readmission) were conducted across the entire stroke population included. We performed interaction analyses to test for interaction with time.

For the mediation analysis, we focused on the association between family income and the clinical outcomes, and separated the total effects into a natural direct and natural indirect effect with quality of care as an intermediate variable following the mediation approach suggested by Lange et al.²⁶ The direct effect reflects the part of the association between SES and outcome, which is not affected by the quality of care, while the indirect effect reflects the part of the association which can be entirely explained by the quality of care. Contrary to conventional mediation analysis, natural direct effect is not found by conditioning on the mediation variable, but rather by enforcing similar distribution of mediation among the exposure groups using one of the groups as reference. The direct and natural direct effects are similar when the mediation variable does not modify the direct effect. The natural indirect effect is found by subtracting the natural direct effect from the total effect. The assumed role of the variables on the causal pathway is depicted in a naïve path diagram in Figure 1. These associations were stratified by the early (2003–10) and late period (2011–2018). Model adjustment was conducted using inverse probability of exposure weighting. Specific sample size calculation for mediation analysis was not conducted.²⁷ The mediation analyses were restricted to patients without contraindications to any of the examined core performance measures in order to ensure homogeneity and that differences in care did not merely reflect differences in clinical need. We expected that the assumption of no unmeasured confounding could be violated by unregistered prognostic factors potentially affecting especially the mediator-outcome relationship. Thus, in addition to adjustment for all covariates, a sensitivity analysis with adjustment for all covariates except Scandinavian Stroke Scale Score was also done as stroke severity may have a particular important role as mediator of socioeconomic disparities in clinical outcomes.²⁸

Figure 1.

Simplified path diagram relating exposure to mediator and the outcomes in the presence of the included confounders.

Supplementary analyses were conducted to investigate potential associations between patients’ income level and whether they had contraindications to each of the examined performance measures. We applied univariable logistic regression for these analyses, thus, associations were represented by odds ratios (OR).

Effects were presented with 95% confidence interval (CI). Data management and time trend analyses were conducted in Stata version 16²⁹ and mediation analyses were conducted in R.³⁰

Research ethics

This study did not involve patient contact and informed patient consent was therefore not required under Danish law. The project was approved by the Danish Data Protection Agency. Data were handled in accordance with the Person Data Act and followed internal university policies on good research practice.

Results

A total of 97,779 unique ischemic stroke patients were registered in Denmark with a first-time stroke during the study period. The distribution of contraindications for the core performance measures is provided in the Supplemental Table 1. Patients who were not eligible for all the performance measures listed in Table 1 were then excluded and the analyses focused on the 59,066 remaining patients.

All patient characteristics varied significantly between the three income groups (see Table 2). Compared with the low-income group, patients in the high-income group were younger, had less comorbidity, had milder strokes, more often lived with a partner and were more often males.

Table 2.

Patient characteristics. Number (%) unless otherwise stated.

Family Income		Low	Medium	High	p-Value
Family Income		N = 14,719	N = 29,930	N = 14,417	p-Value
Age, median (interquartile range)		79 (70–85)	73 (64–81)	65 (57–74)	<0.001
Sex	Female	7800 (53.0%)	14,095 (47.1%)	5198 (36.1%)	<0.001
Sex	Male	6919 (47.0%)	15,835 (52.9%)	9219 (63.9%)	<0.001
Acute myocardial infarction	No	13,246 (90.9%)	27,277 (91.9%)	13,477 (94.1%)	<0.001
Acute myocardial infarction	Yes	1325 (9.1%)	2411 (8.1%)	844 (5.9%)	<0.001
Hypertension	No	6600 (45.5%)	13,843 (46.9%)	7440 (52.2%)	<0.001
Hypertension	Yes	7921 (54.5%)	15,675 (53.1%)	6826 (47.8%)	<0.001
Diabetes	No	12,338 (84.2%)	25,391 (85.2%)	12,835 (89.5%)	<0.001
Diabetes	Yes	2309 (15.8%)	4420 (14.8%)	1508 (10.5%)	<0.001
Atrial fibrillation	No	7958 (94.3%)	15,623 (94.7%)	7599 (96.6%)	<0.001
Atrial fibrillation	Yes	484 (5.7%)	867 (5.3%)	271 (3.4%)	<0.001
Stroke severity scale	Very severe	983 (7.0%)	1326 (4.7%)	496 (3.6%)	<0.001
	Severe	1538 (11.0%)	2463 (8.6%)	939 (6.8%)
	Moderate	3278 (23.5%)	5786 (20.3%)	2038 (14.8%)
	Mild	8177 (58.5%)	18,930 (66.4%)	10,254 (74.7%)
Civil status	Cohabitant	6010 (42.9%)	15,680 (54.3%)	10,754 (76.2%)	<0.001
Civil status	Living alone	8002 (57.1%)	13,185 (45.7%)	3364 (23.8%)	<0.001
Smoking	No	7998 (62.8%)	16,320 (60.7%)	8527 (64.5%)	<0.001
	Yes	4744 (37.2%)	10,555 (39.3%)	4692 (35.5%)

Multiple imputation was used to handle missing data in the confounding variables. The following degrees of missingness were present: cohabitation status 3.5%, atrial fibrillation 44.5%, hypertension 1.3%, diabetes 0.5%, acute myocardial infarction 0.8%, stroke severity 4.8% and smoking status 10.6%.

Interaction analyses showed that the inequality gap in family income increased over the studied period. When comparing the low-income group to the high-income group the mean difference increased with 7631 (95% CI: 6658; 8604) DKK per year.

In total 2169 (3.7%) of the included patients died within 30 days of their index-admission and 6771 (11.5%) were readmitted within 30 days of discharge. The mortality rate varied across family income with higher mortality rate among low-income patients compared to high-income patients. Over time the difference narrowed. The differences in readmission rates were less systematic (see Table 3).

Table 3.

Mortality and readmission over time and income group, n (%) unless otherwise stated.

Family income	30-day mortality				30-day readmission
Family income	n	Low	Medium	High	Low	Medium	High
2003–06	11,340	196 (7.3)	166 (2.9)	47 (1.7)	303 (11.2)	543 (9.4)	225 (7.9)
2007–10	14,132	251 (7.6)	240 (3.3)	81 (2.3)	422 (12.7)	792 (11.0)	332 (9.3)
2011–14	15,660	226 (5.9)	262 (3.28)	82 (2.1)	504 (13.2)	942 (11.8)	459 (12.0)
2015–18	17,934	261 (5.4)	289 (3.2)	68 (1.6)	637 (13.1)	1145 (12.9)	468 (11.3)

The time-trend analyses, illustrated in Table 4, showed that income-related disparity in mortality may have decreased somewhat over the study period, that is, the unadjusted RR was 0.23 (95% CI: 0.17; 0.31) in 2003-06 when comparing high income with ow income, whereas the corresponding RR in 2015-18 was 0.31 (95% CI: 0.23; 0.40). The corresponding fully adjusted estimates were 0.53 (95% CI: 0.38; 0.74) and 0.69 (95% CI: 0.53; 0.89), respectively. The education-related inequality in mortality may also have decreased over time but under a less straight forward pattern.

Table 4.

Temporal development in 30-day mortality and all-cause readmission risk according to family income and education among patients with acute stroke in Denmark 2003–2018.

	30-day mortality				30-day readmission
	Medium vs low		High vs low		Medium vs low		High vs low
	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI
Family income
Unadjusted
2003–06	0.40	(0.32; 0.48)	0.23	(0.17; 0.31)	0.84	(0.73; 0.96)	0.70	(0.60; 0.83)
2007–10	0.44	(0.38; 0.52)	0.30	(0.23; 0.38)	0.86	(0.77; 0.96)	0.73	(0.64; 0.84)
2011–14	0.56	(0.47; 0.66)	0.36	(0.28; 0.47)	0.90	(0.81; 0.99)	0.91	(0.81; 1.03)
2015–18	0.61	(0.51; 0.71)	0.31	(0.23; 0.40)	0.98	(0.90; 1.08)	0.86	(0.77; 0.96)
Adjusted for sex and age
2003–06	0.52	(0.42; 0.64)	0.43	(0.31; 0.61)	0.83	(0.72; 0.95)	0.68	(0.57; 0.81)
2007–10	0.61	(0.51; 0.72)	0.62	(0.49; 0.80)	0.86	(0.77; 0.96)	0.72	(0.63; 0.84)
2011–14	0.73	(0.61; 0.87)	0.70	(0.54; 0.90)	0.91	(0.82; 1.00)	0.94	(0.83; 1.06)
2015–18	0.80	(0.68; 0.95)	0.58	(0.44; 0.76)	1.01	(0.92; 1.10)	0.91	(0.82; 1.03)
Full adjustment^a
2003–06	0.62	(0.51; 0.76)	0.53	(0.38; 0.74)	0.84	(0.73; 0.97)	0.70	(0.58; 0.83)
2007–10	0.72	(0.61; 0.85)	0.82	(0.64; 1.05)	0.86	(0.77; 0.97)	0.74	(0.64; 0.86)
2011–14	0.79	(0.67; 0.94)	0.86	(0.67; 1.10)	0.92	(0.83; 1.02)	0.98	(0.86; 1.11)
2015–18	0.88	(0.75; 1.03)	0.69	(0.53; 0.89)	1.03	(0.94; 1.13)	0.97	(0.87; 1.10)
Education
Unadjusted
2003–06	0.79	(0.60; 1.04)	0.48	(0.29; 0.80)	1.07	(0.93; 1.22)	0.81	(0.65; 1.02)
2007–10	0.65	(0.52; 0.81)	0.62	(0.44; 0.86)	0.97	(0.87; 1.08)	0.96	(0.82; 1.12)
2011–14	0.59	(0.48; 0.73)	0.82	(0.64; 1.06)	0.92	(0.84; 1.01)	0.90	(0.79; 1.02)
2015–18	0.71	(0.59; 0.84)	0.54	(0.42; 0.70)	0.95	(0.87; 1.03)	0.95	(0.85; 1.06)
Adjusted for sex and age
2003–06	0.97	(0.73; 1.29)	0.57	(0.35; 0.95)	1.04	(0.90; 1.20)	0.79	(0.63; 1.00)
2007–10	0.80	(0.64; 0.99)	0.74	(0.53; 1.03)	0.99	(0.89; 1.11)	0.98	(0.83; 1.14)
2011–14	0.78	(0.64; 0.96)	1.04	(0.81; 1.33)	0.94	(0.85; 1.03)	0.91	(0.80; 1.04)
2015–18	1.00	(0.84; 1.19)	0.74	(0.56; 0.93)	0.99	(0.90; 1.08)	0.99	(0.88; 1.10)
Full adjustment^a
2003–06	1.00	(0.76; 1.32)	0.66	(0.40; 1.09)	1.05	(0.91; 1.21)	0.82	(0.65; 1.03)
2007–10	0.85	(0.69; 1.05)	0.81	(0.59; 1.12)	1.00	(0.90; 1.12)	1.01	(0.86; 1.18)
2011–14	0.82	(0.67; 1.01)	1.14	(0.90; 1.45)	0.95	(0.86; 1.05)	0.93	(0.82; 1.06)
2015–18	1.07	(0.91; 1.27)	0.85	(0.66; 1.09)	1.01	(0.92; 1.11)	1.04	(0.93; 1.17)

RR: relative risk, CI: confidence interval; vs: versus.

Adjusted for age, sex, atrial fibrillation, hypertension, diabetes, and acute myocardial infarction, stroke severity, cohabitation status and smoking.

The income-related disparity in 30-day readmission was smaller and turned insignificant over time from 0.70 (95% CI: 0.58; 0.83) to 0.97 (95% CI: 0.87; 1.10) when adjusting for the included covariates. There were no significant education-related differences. Adjustment for covariates reduced the strengths of the associations, but the patterns remained (Table 4).

We found no evidence of an overall time trend in the interaction analysis for neither of the SES variables on neither of 30-day mortality nor readmission (see Table 5).

Table 5.

Time interaction with family income and education in 30-day mortality and all-cause readmission risk among patients with acute stroke in Denmark 2003–2018.

	30-day mortality				30-day readmission
	Medium vs low		High vs low		Medium vs low		High vs low
	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI
Family income-time interaction	1.02	(1.00; 1.04)	1.00	(0.98; 1.03)	1.01	(1.00; 1.02)	1.02	(1.00; 1.03)
Education-time interaction	1.00	(0.98; 1.03)	1.00	(0.97; 1.04)	0.99	(0.98; 1.00)	1.01	(0.99; 1.02)

RR: relative risk; CI: confidence interval; vs: versus.

All estimates are adjusted for age, sex, atrial fibrillation, hypertension, diabetes, and acute myocardial infarction, stroke severity, cohabitation status and smoking.

Table 6 outlines the results of the mediation analyses. A clear negative association was found between income and mortality as well as readmission in the selected population only including patients, who were eligible for all examined care performance measures. However, the association was strongest for mortality. The analyses revealed consistent – though naturally diminishing – patterns across model advancements in terms of adjustments for potential confounding/mediation by patient-related characteristics. The stratification of the analyses into two periods revealed that the identified income-related disparities appeared to diminish over time both in relation to mortality and readmission. Specifically, when comparing high-income with low-income patients total unadjusted RR was 0.27 (95% CI: 0.22; 0.32) and the fully adjusted RR 0.69 (95% CI: 0.55; 0.86) in 2003-10 whereas the corresponding estimates were 0.34 (95% CI: 0.28; 0.40) and 0.77 (95% CI: 0.65; 0.92) in 2011–2018.

Table 6.

Mediation analyses. Direct, indirect, and total effects in relative risks (95% confidence interval).

	Unadjusted						Age and sex adjustment						Full adjustment^a
	2003–2010			2011–2018			2003–2010			2011–2018			2003–2010			2011–2018
RR of 30-day mortality (95% CI)
Medium vs low income
Natural direct	0.43	(0.38;	0.49)	0.58	(0.52;	0.66)	0.57	(0.50;	0.66)	0.77	(0.68;	0.87)	0.70	(0.61;	0.80)	0.84	(0.75;	0.94)
Natural indirect	0.98	(0.97;	0.99)	1.00	(0.99;	1.00)	0.99	(0.98;	1.00)	1.00	(0.99;	1.00)	0.99	(0.97;	1.01)	1.00	(0.99;	1.00)
Total	0.42	(0.37;	0.48)	0.58	(0.52;	0.66)	0.57	(0.50;	0.65)	0.77	(0.68;	0.87)	0.69	(0.60;	0.79)	0.84	(0.75;	0.94)
High vs low income
Natural direct	0.27	(0.23;	0.33)	0.34	(0.28;	0.41)	0.54	(0.44;	0.66)	0.65	(0.54;	0.78)	0.69	(0.55;	0.86)	0.78	(0.65;	0.93)
Natural indirect	0.98	(0.97;	0.99)	0.98	(0.98;	0.99)	1.00	(0.99;	1.01)	0.99	(0.98;	1.00)	0.99	(0.98;	1.01)	0.99	(0.99;	1.00)
Total	0.27	(0.22;	0.32)	0.34	(0.28;	0.40)	0.54	(0.44;	0.66)	0.64	(0.53;	0.77)	0.69	(0.55;	0.86)	0.77	(0.65;	0.92)
RR of 30-day readmission (95% CI)
Medium vs low income
Natural direct	0.85	(0.78;	0.93)	0.94	(0.88;	1.01)	0.85	(0.78;	0.92)	0.96	(0.90;	1.03)	0.86	(0.78;	0.94)	0.98	(0.91;	1.05)
Natural indirect	1.00	(0.99;	1.00)	1.00	(1.00;	1.00)	1.00	(0.99;	1.00)	1.00	(1.00;	1.00)	1.00	(0.99;	1.00)	1.00	(1.00;	1.00)
Total	0.85	(0.78;	0.93)	0.94	(0.88;	1.01)	0.85	(0.78;	0.92)	0.96	(0.90;	1.03)	0.86	(0.78;	0.94)	0.98	(0.91;	1.05)
High vs low income
Natural direct	0.72	(0.65;	0.80)	0.89	(0.82;	0.96)	0.70	(0.63;	0.79)	0.93	(0.86;	1.01)	0.73	(0.64;	0.83)	0.98	(0.90;	1.07)
Natural indirect	1.00	(0.99;	1.00)	1.00	(0.99;	1.00)	1.00	(1.00;	1.00)	1.00	(1.00;	1.00)	1.00	(1.00;	1.00)	1.00	(1.00;	1.00)
Total	0.72	(0.65;	0.80)	0.89	(0.82;	0.96)	0.70	(0.63;	0.79)	0.93	(0.86;	1.01)	0.73	(0.64;	0.83)	0.98	(0.90;	1.07)

RR: relative risk; CI: confidence interval; vs: versus.

Adjusted for age, sex, atrial fibrillation, hypertension, diabetes, and acute myocardial infarction, stroke severity, cohabitation status and smoking.

The indirect relative risks express the extent to which the disparity in quality of care may contribute to the disparity in clinical outcomes. Apart from a few exceptions the estimates showed no indirect effects. The only exceptions were the unadjusted mortality risks where the natural indirect risk ratios for low versus high family income were 0.98 (95% CI: 0.97; 0.99) in 2003–10 and 0.98 (95% CI: 0.98; 0.99) in 2011–18. In a sensitivity analysis with adjustment for all covariates except Scandinavian Stroke Scale score, the maximum indirect relative risk was 0.98 (95% CI: 0.97; 0.99).

The supplementary analysis of the association between income level and contraindications showed no consistent inequality across the performance measures but high-income patients had significantly higher odds of being considered as ineligible for early assessment by a physiotherapist and an occupational therapist compared to low-income patients (see Supplemental Table 2).

Discussion

In this nationwide study covering patients admitted with acute ischemic stroke at Danish hospitals in the 2003–2018 period, substantial income-related inequality in 30-day mortality was observed; however, the inequality in this risk appeared to decline somewhat during the study period, despite the increasing income gap we observed. No uniform temporal trends were established regarding the education-related inequality in 30-day mortality. The inequalities in readmission risk and changes herein were generally more limited.

Generally, only sparse data exist on temporal changes in socioeconomic disparities in post-stroke clinical outcomes. It has been demonstrated that post stroke mortality has decreased and post stroke return to work has improved over the past decades,^31,32 but the development in disparities herein is uncertain. A previous German study of inequality trends in stroke outcomes reported income-related inequalities in stroke-free and stroke-affected life years was constant between 2006 and 2016.⁷

Our mediation analyses indicated that the quality of care as reflected by five selected performance measures reflecting basic elements of early in-hospital stroke care had no or only very limited role in the disparities in mortality and readmission risk. A number of factors may have contributed to this puzzling finding. First, only a limited number of care performance measures were included in the analyses. Although these measures have in previous studies been linked with better clinical outcomes, the extent of the disparity in quality of care may have been too small to play a measurable role on mortality and readmission risk, which are quite blunt and typically insensitive clinical outcomes. These outcomes were chosen primarily for a pragmatic reason as valid information was available for the entire study population. Functional level, for example, Modified Rankin score, would likely have been a more on sensitive and clinically relevant information, however, this information is unfortunately not available on the general stroke population in the DSR.

Previous studies of socioeconomic inequality in stroke outcomes point in different directions. Studies have found that the socioeconomic inequality in stroke mortality starts to manifest already within the first 30-days,³³ that the short-term mortality inequality was small or even insignificant,³⁴ or that it exists in both the short term and long term perspective.³⁵ Since a focus of the current study was the possible role of early hospital-based stroke care rather than long-term care provided by rehabilitation institutions, municipalities, and general practitioners, we chose to restrict the observation time for the clinical outcomes to 30 days. Further studies are needed to cover broader spectrum of socioeconomic inequality in long-term post-stroke outcomes, for example in relation to functional capacity (e.g. in the Modified Rankin score) or return to work. Though mortality has the advantage of being robust against misclassification, more sensitive outcome measures could potentially be of even greater relevance to clinicians, politicians and patients.

Another consideration when studying socioeconomic inequality in relation to stroke is the operationalisation of socioeconomic position, especially since our time trend analyses showed opposite trends for education-related and income-related inequalities. In the design phase, the income focus was prioritised in the mediation analyses as previous studies of the Danish stroke population had found no significant association between education and quality of care.^11,36 Economic indicators of socioeconomic position have previously been identified as more sensible indicators than education and occupation in relation to mortality.³⁷ Focusing specifically on late-life health, income has been found to be a preferable indicator of socioeconomic situation,¹⁷ however, household wealth may serve as an alternative measure.³⁷ Geyer et al. recommended that the different socioeconomic indicators should not be used interchangeably as they relate to differing causal mechanisms.³⁸ Whereas income may determine patients’ opportunities for healthy living, education may affect their ability to understand health advice and manage their disease accordingly.³⁸ These measures therefore reflect fundamentally different perspectives on socioeconomic disparity and may therefore provide complementary insights.

A key strength of this study is the high-quality data sources that the study was based upon. We analysed individual-level data covering all first-time ischemic stroke patients admitted in Denmark. Follow-up was complete and detailed data on important prognostic factors were available, including key comorbidity factors, stroke severity and smoking. This allowed us to adjust for the potential confounding effects these covariates could have on health status. However, the handling of these covariates in the data analyses also presents a challenge as they may both have roles as confounding and mediating factors in the examined associations. Stroke severity is an example of this, as the variable has previously been reported to mediate as much as half of the income-related inequality in 3-month case fatality.³⁸ In addition, stroke severity may in itself also be associated with the quality of care and may therefore have represent mediator-outcome-confounding, which is particular troublesome in mediation analyses.³⁹ However, even though uncertainties on the exact nature of the role of the covariates exist (and consequently how best to handle the covariates in the data analyses), the consistency of our findings across the different statistical models was noteworthy in the mediation analyses.

Finally, the selected mediator and the way we conceptualised the quality of care introduces both important strengths and limitations to this study. A unique strength lies in the fact that each of the included performance measures involves a professional consideration of the individual patient’s potential contraindications. This implies that only patients with similar clinical needs were included in the analyses. However, inherent to this evaluation of each patient’s contraindications lies also risk of misclassification. The supplementary analysis shows some income-related differences in contraindications verging on a skewness towards high-income patients having more contraindications than low-income patients. It is not possible to conclude from this, that the assessment of clinical needs was systematically different according to socioeconomic position, however, some caution is relevant when interpreting the data as some systematic misclassification cannot entirely be excluded despite the regular audits and detailed written instructions on how to register patients in the DSR. Misclassification of the mediating variable has previously been found to cause more substantial underestimations of the indirect effects than similar misclassification of the exposure variable in mediation analyses.⁴⁰ In addition to this, previous studies have shown that socioeconomic inequality exists for different domains of stroke care in addition to the five performance measures selected for the mediation analyses in the current study, including the use of revascularization therapy and patient-related pre-hospital time delay.^11,41 However, these elements of care are critical dependent on other factors than the performance of the healthcare system, in particular public stroke awareness and pre-hospital patient-related time delay,⁴² and were therefore not considered in the current study.

As demonstrated in the current study, socioeconomic disparities in stroke outcomes continue to represent an important clinical and public health challenge despite extensive efforts to standardise and improve the quality of early stroke care. Much of the disparity may be ascribed to upstream factors outside the control or influence of healthcare systems, for example, labour market access, housing or air pollution, however, this should not preclude efforts to further clarify how health care systems can most effectively improve the chances of good clinical outcomes also among socioeconomic disadvantaged patients be it in stroke prevention efforts, prehospital or in hospital care, or downstream during rehabilitation.

Supplemental Material

sj-docx-1-eso-10.1177_23969873221146591 – Supplemental material for Is the socioeconomic inequality in stroke prognosis changing over time and does quality of care play a role?

Supplemental material, sj-docx-1-eso-10.1177_23969873221146591 for Is the socioeconomic inequality in stroke prognosis changing over time and does quality of care play a role? by Vibe Bolvig Hyldgård, Rikke Søgaard, Jan Brink Valentin, Theis Lange, Dorte Damgaard and Søren Paaske Johnsen in European Stroke Journal

Footnotes

Acknowledgements

None

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

Ethical approval

The project was approved by the Danish Data Protection Agency. Data were handled in accordance with the Person Data Act and followed internal university policies on good research practice.

Informed consent

Not applicable. This study did not involve patient contact and informed patient consent was therefore not required under Danish law.

Guarantor

VBH

Contributorship

VBH, RS, SPJ, JBV and TL collaborated on designing the study. SPJ retrieved the data. VBH handled the data management and descriptive analyses. VBH, JBV and SPJ performed the statistical analyses. VBH, RS, SPJ, JBV, TL and DD interpreted and discussed the results. VBH and SPJ drafted the manuscript and revised it in accordance with comments and inputs from RS, JBV, TL and DD.

ORCID iD

Vibe Bolvig Hyldgård

Supplemental material

Supplemental material for this article is available online.

References

Aguiar de Sousa

von Martial

Abilleira

, et al Access to and delivery of acute ischaemic stroke treatments: A survey of national scientific societies and stroke experts in 44 European countries. Eur Stroke J 2019; 4: 13–28.

Bray

Paley

Hoffman

, et al Socioeconomic disparities in first stroke incidence, quality of care, and survival: a nationwide registry-based cohort study of 44 million adults in England. Lancet Public Health 2018; 3: e185–e193.

Addo

Ayerbe

Mohan

, et al Socioeconomic status and stroke an updated review. Stroke 2012; 43: 1186–1191.

Bettger

Zhao

Bushnell

, et al The association between socioeconomic status and disability after stroke: findings from the adherence evaluation after ischemic stroke longitudinal (AVAIL) registry. BMC Public Health 2014; 14: 281.

Chen

McKevitt

Crichton

, et al Socioeconomic deprivation and provision of acute and long-term care after stroke: the south London stroke register cohort study. J Neurol Neurosurg Psychiatry 2014; 85: 1294–1300.

Pennlert

Asplund

Glader

, et al Socioeconomic status and the risk of stroke recurrence - persisting gaps observed in a nationwide Swedish study 2001 to 2012. Stroke 2017; 48: 1518–1523.

Tetzlaff

Geyer

Tetzlaff

, et al Income inequalities in stroke incidence and mortality: trends in stroke-free and stroke-affected life years based on German health insurance data. PLoS One 2020; 15: e0227541–e.

Langagergaard

Palnum

Mehnert

, et al Socioeconomic differences in quality of care and clinical utcome after stroke A nationwide population-based study. Stroke 2011; 42: 2896–2902.

Kimball

Neal

Waters

, et al Race and income disparity in ischemic stroke care: nationwide inpatient sample database, 2002 to 2008. J Stroke Cerebrovasc Dis 2014; 23: 17–24.

10.

Pan

Chen

, et al Socioeconomic status and the quality of acute stroke care - the China national stroke registry. Stroke 2016; 47: 2836–2842.

11.

Hyldgård

Johnsen

Støvring

, et al

Socioeconomic status and acute stroke care: has the inequality gap been closed?

Clin Epidemiol 2019; 11: 933–941.

12.

Schmidt

Pedersen

Sørensen

. The Danish civil registration system as a tool in epidemiology. Eur J Epidemiol 2014; 29: 541–549.

13.

Jensen

Rasmussen

. Danish education registers. Scand J Public Health 2011; 39: 91–94.

14.

Publishing O, UNESCO Institute for S, Eurostat. ISCED 2011 operational manual: guidelines for classifying national education programmes and related qualifications. Paris: OECD, 2015.

15.

Baadsgaard

Quitzau

. Danish registers on personal income and transfer payments. Scand J Public Health 2011; 39: 103–105.

16.

Galobardes

Shaw

Lawlor

, et al Indicators of socioeconomic position (part 1). J Epidemiol Community Health 2006; 60: 7–12.

17.

Darin-Mattsson

Fors

Kåreholt

. Different indicators of socioeconomic status and their relative importance as determinants of health in old age. Int J Equity Health 2017; 16: 173–211.

18.

Avlund

Holstein

Osler

, et al Social position and health in old age: the relevance of different indicators of social position. Scand J Public Health 2003; 31: 126–136.

19.

Lynge

Sandegaard

Rebolj

. The Danish national patient register. Scand J Public Health 2011; 39: 30–33.

20.

Schmidt

Sandegaard

, et al The Danish national patient registry: a review of content, data quality, and research potential. Clin Epidemiol 2015; 7: 449–490.

21.

Wildenschild

Mehnert

Thomsen

, et al Registration of acute stroke: validity in the Danish stroke registry and the Danish national registry of patients. Clin Epidemiol 2013; 6: 27–36.

22.

Johnsen

Ingeman

Hundborg

, et al The Danish stroke registry. Clin Epidemiol 2016; 8: 697–702.

23.

Ingeman

Pedersen

Hundborg

, et al Quality of care and mortality among patients with stroke: a nationwide follow-up study. Med Care 2008; 46: 63–69.

24.

Svendsen

Ehlers

Andersen

, et al Quality of care and length of hospital stay among patients with stroke. Med Care 2009; 47: 575–582.

25.

Lindenstrøm

Boysen

Waage Christiansen

, et al Reliability of Scandinavian neurological stroke scale. Cerebrovasc Dis 1991; 1: 103–107.

26.

Lange

Hansen

Sørensen

, et al Applied mediation analyses: a review and tutorial. Epidemiol Health 2017; 39: e2017035.

27.

Lee

Cashin

Lamb

, et al A guideline for reporting mediation analyses of randomized trials and observational studies: the AGReMA statement. J Am Med Assoc 2021; 326: 1045–1056.

28.

Lindmark

Norrving

Eriksson

. Socioeconomic status and survival after stroke - using mediation and sensitivity analyses to assess the effect of stroke severity and unmeasured confounding. BMC Public Health 2020; 20: 554.

29.

StataCorp. Stata statistical software: release 16. College Station, TX: StataCorp LLC, 2019.

30.

R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2021, https://www.R-project.org/ (accessed 2 June 2022).

31.

Lee

Shafe

Cowie

. UK stroke incidence, mortality and cardiovascular risk management 1999-2008: time-trend analysis from the general practice research database. BMJ Open 2011; 1: e000269.

32.

Hannerz

Mortensen

Poulsen

, et al Time trend analysis of return to work after stroke in Denmark 1996-2006. Int J Occup Med Environ Health 2012; 25: 200–204.

33.

Lindmark

Glader

Asplund

, et al, Riks-Stroke Collaboration. Socioeconomic disparities in stroke case fatality–observations from risks-stroke, the Swedish stroke register. Int J Stroke 2014; 9: 429–436.

34.

Andersen

Olsen

. Social inequality by income in short- and long-term cause-specific mortality after stroke. J Stroke Cerebrovasc Dis 2019; 28: 1529–1536.

35.

Belleudi

Sciattella

Agabiti

, et al Socioeconomic differences in one-year survival after ischemic stroke: the effect of acute and post-acute care-pathways in a cohort study. BMC Public Health 2016; 16: 408.

36.

Hyldgård

Johnsen

Søgaard

. Index-based inequality in quality of care: an empirical comparison of apples and pears. Clin Epidemiol 2021; 13: 791–800.

37.

Daly

Duncan

McDonough

, et al Optimal indicators of socioeconomic status for health research. Am J Public Health 2002; 92: 1151–1157.

38.

Geyer

Hemström

Peter

, et al Education, income, and occupational class cannot be used interchangeably in social epidemiology. Empirical evidence against a common practice. J Epidemiol Community Health 2006; 60: 804–810.

39.

Ding

Vanderweele

. Sharp sensitivity bounds for mediation under unmeasured mediator-outcome confounding. Biometrika 2016; 103: 483–490.

40.

Blakely

McKenzie

Carter

. Misclassification of the mediator matters when estimating indirect effects. J Epidemiol Community Health 2013; 67: 458–466.

41.

Buus

SMØ

Schmitz

Cordsen

, et al Socioeconomic inequalities in reperfusion therapy for acute ischemic stroke. Stroke 2022; 53: 2307–2316.

42.

Mainz

Andersen

Valentin

, et al Treatment delays and chance of reperfusion therapy in patients with acute stroke: a Danish nationwide study. Cerebrovasc Dis 2022; 31: 1–8.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.05 MB