Socioeconomic Differences in Quality of Care and Clinical Outcome After Stroke
A Nationwide Population-Based Study
Background and Purpose—The association among socioeconomic status, quality of care, and clinical outcome after stroke remains poorly understood. In a Danish nationwide follow-up study, we examined whether socioeconomic-related differences in acute stroke care occur and, if so, whether they explain socioeconomic differences in case-fatality and readmission risk.
Methods—Using population-based public registries, we identified and followed all patients aged ≤65 years admitted with stroke from 2003 to 2007 (n=14 545). We compared the proportion of patients receiving 7 specific processes of care according to income, educational attainment, and employment status. Furthermore, we computed 30-day and 1-year hazard ratios for death and readmission adjusted for patient characteristics and received processes of acute stroke care.
Results—For low-income patients and disability pensioners, the relative risk of receiving all of the relevant processes of care was 0.82 (95% CI, 0.78 to 0.86) and 0.83 (95% CI, 0.79 to 0.87), respectively, compared with high-income patients and employed patients. Adjusted 30-day and 1-year hazard ratios for death for unemployed patients were 1.57 (95% CI, 1.25 to 1.97) and 1.58 (1.32 to 1.88), respectively, compared with employed patients. Unemployed patients also had a higher risk of readmission. The differences in mortality and readmission risk remained after controlling for received processes of acute stroke care.
Conclusions—Low socioeconomic status was associated with a lower chance of receiving optimal acute stroke care. However, the differences in acute care did not appear to explain socioeconomic differences in mortality and readmission risk.
There is strong evidence for the existence of a socioeconomic gradient in stroke incidence and overall mortality with higher rates of stroke in low socioeconomic groups being a consistent finding.1,2 Differential distribution of clinical and lifestyle factors are possible explanations for these socioeconomic differences.3,4 However, existing literature also suggests that socioeconomic-related differences may exist in care with patients with low socioeconomic status (SES) receiving fewer relevant diagnostic examinations and less care than patients with high SES.5–7 Still, the association among SES, quality of care, and clinical outcome remains poorly understood. No consistent pattern of inequity in stroke care has been found.4,8 The fact that the majority of studies on this topic did not have detailed individual-level data on SES and quality of care (in particular, the timing of specific processes of care) may have contributed to this uncertainty.
The evidence of an association between SES and stroke outcome is also controversial, because some studies have found a strong relationship between SES and poststroke mortality,6,7,9–12 whereas others have reported either no association or a weak association.3,5,13,14 Furthermore, there are sparse data on the possible implications of differences in care on SES-related differences in mortality.7
Therefore, we conducted a nationwide follow-up study of patients with stroke in Denmark, a country with tax-supported health care to all residents, including free access to hospitals, which, in theory, should guarantee equal access to treatment independent of SES. We examined whether SES-related differences in acute stroke care occur and, if so, whether they affect clinical outcome including 30-day and 1-year case-fatality and readmission risk.
The Danish National Indicator Project
The Danish National Indicator Project (DNIP) is a nationwide initiative to monitor and improve the quality of care for specific diseases, including stroke.15 The project does this by developing evidence-based quality criteria related to the structure, process, and outcome of health care and, subsequently, by monitoring the fulfillment of these criteria. Participation in the project is mandatory for all hospitals in Denmark treating patients with acute stroke.
All patients (≥18 years) admitted to Danish hospitals with acute stroke (including both intracerebral hemorrhage, ischemic and no-specified stroke) according to the World Health Organization criteria (ie, rapidly developed clinical signs of focal or global disturbance of cerebral function, lasting >24 hours or until death, with no apparent nonvascular cause16) are eligible for inclusion in DNIP. Patients with subdural hematoma, epidural or subarachnoidal hemorrhage, retinal infarct, and infarct caused by trauma, infection, surgery, or an intracerebral malignant process are not included. We identified all admissions of patients aged 18 to 65 years with stroke registered in DNIP from January 1, 2003, to December 31, 2007. Patients >65 years of age were excluded, because it is difficult to use employment status and income as a reflection of SES when the patients are above retirement age. Although some patients had >1 stroke during the study period, we only included the first stroke event registered in the study period (n=14 983) to ensure independence between the events. Furthermore, only patients residing in Denmark with available data on SES were included. Data from 14 545 patients (97.1%) were available for further analysis.
The Integrated Database for Labor Market Research (IDA) contains socioeconomic information at the individual level of Danish citizens.17 From the IDA database we obtained information on employment status the year before hospital admission for each patient (employed, unemployed, disability pensioner). “Unemployed” comprised patients who were unemployed or received early retirement benefits and other economically inactive persons. Early retirement is voluntary and is only possible for persons who are still able to work. We also retrieved information on all personal income, including net annual value for each patient and cohabiting partner. This broad definition of income was used in an attempt to reflect the wealth of each patient, because it has been suggested that wealth is a more sensitive indicator of SES than income.18 To account for yearly variation in income, we calculated the average income in the 5 years before admission for the patient and cohabiting partner. All patients were divided into tertiles of increasing income.
Information on the highest completed educational level registered the year before admission was obtained from the Student Registry of Statistics Denmark.19 Patients were divided into 3 groups according to highest attained education: (1) short-, medium-, and long-term higher education; (2) vocational education and upper secondary school; and (3) primary and lower secondary school.
Processes of Acute Stroke Care
A national expert panel identified 7 processes of acute stroke care covering the acute phase of stroke15: early admission to a specialized stroke unit, early administration of antiplatelet or anticoagulant therapy, early CT/MRI scan, and early assessment by a physiotherapist, an occupational therapist, and of nutritional risk. A timeframe was defined for each process to capture the timeliness of the process. The timeframe was the second day of hospitalization for all processes, except initiation of oral anticoagulant therapy in which it was the 14th day of hospitalization. From 2005 onward, the timeframe for early examination with CT/MRI was changed to the day of admission.
A specialized stroke unit was defined as a hospital department/unit that exclusively or primarily is dedicated to patients with stroke and which is characterized by multidisciplinary teams, a staff with a specific interest in stroke, involvement of relatives, and continuous education of the staff. Administration of antiplatelet and oral anticoagulant therapy was defined as continuous use of the drugs and not merely a single dose. Assessment by a physiotherapist and occupational therapist was defined as a formal bedside assessment of the patient's need for rehabilitation, whereas assessment of nutritional risk was defined as an assessment following the recommendations of the European Society for Parental and Enteral Nutrition.
On hospital admission, data on care and patient characteristics were prospectively collected for each patient using a standardized form by the staff caring for the patient. Patients were classified as eligible or noneligible for the specific processes of care depending on whether the staff treating the patient identified contraindications, for example, atrial fibrillation precluding oral anticoagulant therapy, or rapid spontaneous recovery of motor symptoms making early assessment by a physiotherapist and occupational therapist irrelevant. Detailed written instructions were available to the staff, which specified criteria for deeming a patient ineligible for the care processes.
Data on patient characteristics included: age, sex, civil status (cohabitant, living alone, other), housing (own home, nursing home, other), Scandinavian Stroke Scale score (a measure of stroke severity20), previous stroke and myocardial infarction, previous/current atrial fibrillation, hypertension, diabetes mellitus or intermittent claudication, smoking habits (smoker, exsmoker, never), and alcohol intake (≤14/21, >14/21 drinks per week for women and men, respectively). We computed the Charlson comorbidity index score for each patient based on all discharge diagnoses recorded before the stroke hospitalization. The index has previously been adapted for use with hospital discharge registry data and has been reported to be useful also among patients with stroke.21,22
Data on previous hospitalizations were obtained from the National Patient Registry, which contains data on all discharges from all nonpsychiatric hospitals in Denmark since 1977.23 We defined 3 levels of comorbidity: 0 (“none”), 1 to 2 (“low”), and >2 (“high”). We excluded diabetes, myocardial infarction, and former stroke from the index and adjusted for these conditions separately due to their strong prognostic role.
Mortality and Readmission
We computed 30-day and 1-year risks of death and readmission with any diagnosis after stroke. Mortality was ascertained from the Civil Registration System, which keeps electronic records on dates of birth, death, and emigration for all Danish citizens.24 Information on readmission was ascertained from the National Patient Registry.
We first calculated, for each stratum of income, education, and employment, the proportion of patients receiving all relevant processes of care and then assessed whether the patients had received the individual processes. We computed relative risks (RRs) for each stratum of income, education, and employment, using “high income,” “long education,” and “employed” as references. To analyze the combined effect of income, education, and employment on care, we also constructed 3 patient groups in which all 3 indicators of SES were high, medium, or low. Second, we used Cox proportional hazards regression to compute hazard ratios for the time to death and readmission within 30 days or 1 year after stroke according to the SES indicators. Follow-up began on the date of admission with stroke (in analyses on risk of readmission, follow-up began on date of discharge) and ended on the date of death, emigration, new admission (only in analyses on risk of readmission), or after 30 days/1 year, whichever came first. First, we adjusted the crude hazard ratios for patient characteristics and hospital department. Second, we additionally adjusted for the proportion of received processes of care that each patient was deemed eligible for. To examine the interrelations among the 3 different indicators of SES, we finally mutually adjusted for the socioeconomic factors (eg, models examining the effects of income on mortality were adjusted for education and employment). The analyses were repeated in strata of men and women to examine if sex modified the associations. We used multiple imputation to impute missing values for covariates. We generated 5 imputed data sets, and the hazard ratios were then averaged across the 5 imputations. Besides all measured covariates, we included the event indicator and the Nelson-Aalen estimator of the cumulative hazard to the survival time in the imputation model.25 However, for comparison, a data set consisting of only complete records was also examined. We analyzed data using STATA Version 10.0 (StataCorp). The study was approved by the Danish Data Protection Agency (record no. 2008-41-2274).
Patients with a low income, a short education, and disability pensioners had more comorbidity; were more likely to smoke and live alone; and less likely to live in their own home compared with patients with a high income, a long education, or patients who were employed (Supplemental Table I; http://stroke.ahajournals.org).
Processes of Acute Stroke Care
Table 1 presents, according to income, education, and employment, the proportion of patients who received all relevant processes of care. We found substantial differences between patients with high and low SES. The largest differences were seen for patients with a low income and disability pensioners where the RR of receiving all relevant processes of care was 0.82 (95% CI: 0.78 to 0.86) and 0.83 (95% CI: 0.79 to 0.87), respectively, compared to high-income patients and employed patients.
When combining employment status, income, and education, we found that the RR of receiving all relevant processes of care for disability pensioners with a low income and a short education was 0.72 (95% CI, 0.66 to 0.79) compared with employed patients with a high income and a long education. For unemployed patients with a medium income and medium education, the corresponding RR was 0.91 (95% CI, 0.81 to 1.02).
There were only minor differences in the proportions of patients deemed ineligible or with missing data for the specific processes of care among the subgroups of income, education, and employment.
The differences in the proportions of eligible patients who received the individual processes were moderate to small when comparing low-income patients to high-income patients (ie, RRs ranging from 0.89 [95% CI, 0.85 to 0.92] to 0.98 [95% CI, 0.86 to 1.12]) and when comparing patients with a short education to those with a long education (ie, RRs ranging from 0.96 [95% CI, 0.94 to 0.99] to 0.99 [95% CI, 0.97 to 1.01]). Similarly, there were slightly fewer disability pensioners who received each process of care compared with employed patients (ie, RRs ranging from 0.87 [95% CI, 0.84 to 0.91] to 0.97 [95% CI, 0.95 to 0.99]), except for oral anticoagulant therapy (RR, 1.14; 95% CI, 1.00 to 1.30; (Supplemental Table II).
Overall, 697 patients (4.8%) died within 30 days of follow-up and 1202 patients (8.3%) died within 1 year. The crude hazard ratios showed a higher 30-day and 1-year case-fatality among patients with a medium/low income and among unemployed/disability pensioners, whereas there was no substantially increased case-fatality among patients with shorter education (Table 2). As expected, adjusting for patient characteristics attenuated the effect of income and employment. Further adjustment for differences in the proportion of received relevant processes of care had only a marginal impact on the adjusted hazard ratios. The fully adjusted analyses showed a profoundly increased risk of death within 30 days and 1 year for unemployed patients and within 1 year for disability pensioners compared with employed patients. Unemployed patients with medium income and education had an increased 30-day and 1-year mortality with adjusted hazard ratios of 1.94 (95% CI, 1.21 to 3.11) and 2.14 (95% CI, 1.49 to 3.07), respectively, when compared with patients in the highest SES group. Adjusting for care did not affect the estimates. In contrast, there was no significantly increased mortality for disability pensioners with a low income and short education (data not shown). No substantial differences were found when stratifying the analyses according to sex (data not shown). Furthermore, excluding patients receiving early retirement benefit from the unemployed group (1620 of 3360 patients) had virtually no impact on the findings.
Within 30 days of follow-up, 1318 (9.1%) of the patients were readmitted to the hospital. After 1 year, 4498 (31%) patients had been readmitted at least once. The crude hazard ratios showed a higher 30-day and 1-year readmission risk among patients with a medium/low income and among unemployed/disability pensioners, whereas patients with a short education had an increased 1-year readmission risk (Table 3). The fully adjusted analyses showed a marginally increased risk of readmission within 1 year for patients with a medium/low income and a more substantially increased risk of readmission within 30 days and 1 year for unemployed/disability pensioners. We found no major difference in 30-day readmission risk among the 3 combined exposure groups, but within 1 year, the hazard ratio for readmission was 1.22 (95% CI, 1.00 to 1.49) for the medium SES group and 1.64 (95% CI, 1.39 to 1.94) for the lowest SES group when compared with patients in the highest SES group and after adjusting for patient characteristics. Again, adjusting for acute care and stratifying by sex did not affect the estimates.
All findings were virtually unchanged when the analyses were redone on a data set consisting of only patients with complete information on all covariates (n=9033).
Our finding of SES-related differences in quality of acute stroke care corroborates 3 previous studies.5–7 A study from Finland reported that among 6903 first stroke events, patients from high-income groups were more likely to be treated at a university hospital, be examined by a neurologist, and have CT or MRI.6 A large population-based Canadian study concluded that despite a universal health insurance program, patients in the lowest income group waited longer time for carotid surgery and were less likely to be treated by neurologists and receive in-hospital rehabilitation.7 Similarly, a study of 2709 patients with stroke from Scotland suggested reduced access to key items of acute care among low-SES patients.5 By contrast, a study from London found no consistent pattern of inequality in care according to SES (measured by occupation) using 22 indicators of evidence-based care.8 Also, a small Swedish study reported no difference in stroke unit care or medication according to educational level.4 The existing studies have used different SES indicators and only 3 studies had individual-level measures.4,6,8 Our use of more detailed SES data suggested that the quality of care was more associated with income and employment status than with educational level. Furthermore, for some of the processes of care we evaluated (ie, initiation of antiplatelet therapy, examination with CT/MRI, assessment by a physiotherapist/occupational therapist), nearly all patients in our study population (>95%) received the intervention in question before discharge, although not necessarily before the defined timeframe.26 Our results therefore indicate later initiation of therapy among low-SES patients.
The increased case-fatality among unemployed and disability pensioners in our study could not be explained by SES-related differences in acute hospital care. This confirms the findings of Kapral et al, who reported that although they adjusted for differences in care, SES (measured as median neighborhood income) remained a predictor of mortality.7 Several studies have reported an association between SES and mortality after stroke based on different SES indicators but without controlling for differences in care.6,9,10,12 In contrast, other studies reported no association between SES and stroke mortality when using SES indicators based on occupation or area-based deprivation indices.3,5,13,14 We found no association between educational level and mortality, which confirms previous observations.4,9,10
A few studies have reported sex specific socioeconomic-related differences in mortality. Jakovljevic et al found an association between low education and 28-day and 1-year case-fatality, which was limited to women aged 25 to 59 years.6 The authors also found a relationship between income and 28-day case-fatality, but only among men. Likewise, a Swedish study reported a higher 28-day and 1-year mortality in low-income groups limited to men.12 In contrast, we found no systematic sex-related differences in the association between SES and clinical outcome.
The higher readmission risk among patients with low SES in our study probably reflects a higher vulnerability in this patient group, even after adjustment for differences in prognostic profile. In accordance with our finding, 2 previous studies reported an association between low SES and risk of readmission for cardiovascular events after stroke.3,13 A small Swedish study found that income was inversely associated with recurrent stroke, however only in women.12 In contrast, we found no sex-related difference in readmission risk according to SES.
The strengths of our study include the prospective population-based design with detailed data on patient characteristics, processes of acute stroke care and complete follow-up on mortality and readmission risk. SES can be indicated by a number of (sub-)concepts such as employment status, occupational status, educational attainment, income, and wealth and consequently usually understood as a multidimensional concept.27 Measures of SES may be derived directly from individuals or indirectly through the characteristics of the area in which they live (area-based measures). Area-based measures are widely used and have the advantage that they may also capture information about the cultural and physical environment in which the patients live. However, use of area-based measures to estimate an individual's SES results in considerable misclassification and individual-level measures are therefore often preferred if data are available.28
The use of data collected in a nonstandardized setting during routine clinical work is a limitation that potentially affects the accuracy of our data. However, participation in DNIP is mandatory for all departments treating patients with acute stroke in Denmark, and great effort is made to ensure the validity of the DNIP.15 In particular, a yearly structured audit is conducted nationally, regionally, and locally, which includes validation of the completeness of patient registration against hospital discharge registries. Furthermore, any misclassification is unlikely to depend on SES. The fact that the eligibility for the specific processes of care was determined by the staff might be a cause for concern as health professionals could prioritize differently. However, we found only minor differences in the proportions of patients considered eligible for care according to income, education, and employment.
In conclusion, our nationwide study showed that despite a tax-financed universal healthcare system, lower SES was associated with poorer quality of acute hospital care among younger patients with stroke. Furthermore, we found a substantially increased mortality and risk of readmission among unemployed and disability pensioners. Differences in acute stroke care could not explain these findings. Continuous efforts are warranted to ensure optimal stroke care to all patient groups and to explore possible future interventions aimed at reducing the social gradient in stroke outcome.
Sources of Funding
This work was supported by grants from The Danish Heart Association and Trygfonden.
The online-only Data Supplement is available at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.611871/-/DC1.
- Received December 21, 2010.
- Revision received April 10, 2011.
- Accepted May 5, 2011.
- © 2011 American Heart Association, Inc.
- Cesaroni G,
- Agabiti N,
- Forastiere F,
- Perucci CA
- Weir NU,
- Gunkel A,
- McDowall M,
- Dennis MS
- Jakovljevic D,
- Sarti C,
- Sivenius J,
- Torppa J,
- Mahonen M,
- Immonen-Raiha P,
- et al
- Kapral MK,
- Wang H,
- Mamdani M,
- Tu JV
- McKevitt C,
- Coshall C,
- Tilling K,
- Wolfe C
- Arrich J,
- Lalouschek W,
- Mullner M
- Saposnik G,
- Jeerakathil T,
- Selchen D,
- Baibergenova A,
- Hachinski V,
- Kapral MK
- Li C,
- Hedblad B,
- Rosvall M,
- Buchwald F,
- Khan FA,
- Engstrom G
- Aslanyan S,
- Weir CJ,
- Lees KR,
- Reid JL,
- McInnes GT
- Peltonen M,
- Rosen M,
- Lundberg V,
- Asplund K
- Mainz J,
- Krog BR,
- Bjornshave B,
- Bartels P
Declaration of Content: Integrated Database for Labour Marked Research (IDA); 2007. Available at: www.dst.dk/HomeUK/Guide/documentation/Varedeklarationer/emnegruppe/emne.aspx?sysrid=1013. Accessed September 1, 2010.
Declaration of Content: Student Register; 2009. Available at: www.dst.dk/HomeUK/Guide/documentation/Varedeklarationer/emnegruppe/emne.aspx?sysrid=113371. Accessed September 1, 2010.
Multicenter trial of hemodilution in ischemic stroke—background and study protocol. Scandinavian Stroke Study Group. Stroke. 1985;16:885–890.
- Goldstein LB,
- Samsa GP,
- Matchar DB,
- Horner RD
- Frank L
The Danish National Indicator Project; 2010. Available at: www.nip.dk. Accessed September 1, 2010.
- Ainley J,
- Graetz B,
- Long M,
- Batten M
- Graetz B
- Ainley J,
- Graetz B,
- Long M,
- Batten M
- Ainley J,
- Long M