Factors Associated With 28-Day Hospital Readmission After Stroke in Australia
Background and Purpose—Understanding the factors that contribute to early readmission after discharge following stroke is limited. We aimed to describe the factors associated with 28-day readmission after hospitalization for stroke.
Methods—Factors associated with readmission were classified from the medical record standardized audits of 50 to 100 consecutively admitted patients with stroke from 35 Australian hospitals during multiple time periods (2000–2010). Factors were compared between patients readmitted and not readmitted after stroke hospitalization (n=43) grouped using 5 categories: patient characteristics (n=16; eg, age), clinical processes of care (n=13; eg, admitted into a stroke unit), social circumstances (n=3; eg, living home alone prior), health system (n=6; eg, location of hospital), and health outcome (n=5; eg, length of stay). Multilevel logistic regression modeling was used to examine the association with these independent factors selected if statistical significance P<0.15 or if considered clinically important and readmission status.
Results—Among 3328 patients, 6.5% were readmitted within 28 days (mean age, 75; 48% female; 92% ischemic). After bivariate analyses 14/43 factors from 4/5 categories were associated with readmission after hospitalization for stroke. Two factors from patient and health outcome categories remained independently associated with readmission after multivariable analyses. These were dependent premorbid functional status (adjusted odds ratio, 1.87; 95% confidence interval, 1.25–2.81) and having a severe adverse event during the initial hospitalization for stroke (adjusted odds ratio, 2.81; 95% confidence interval, 1.55–5.12).
Conclusions—This is the first study to comprehensively evaluate factors associated with 28-day readmission after stroke. The factors associated with 28-day readmission are diverse and include potentially modifiable and nonmodifiable factors.
Stroke is an important condition for understanding the likely factors that contribute to avoidable readmissions. This is because stroke is a leading cause of hospitalizations, deaths, and disease burden. In Australia, among older people aged ≥75 years stroke is the leading cause of hospitalization.1 Hospitalizations for patients with stroke are expensive.2
Information on the reasons why hospital readmissions occur within 1 month after discharge following an admission for stroke is limited. Readmission can be an important outcome of the quality of hospital care for patients with stroke.3,4 Readmissions within 1 month of discharge may reflect unresolved problems at discharge,5 the quality of immediate post–hospital care, a more chronically ill population, or because of a combination of these factors.6 A study in the United States provided evidence that patients with acute stroke who were readmitted within 30 days had a worse chance of survival and incurred greater healthcare costs than patients who were not readmitted.7 Preventing avoidable readmissions may reduce costs to the healthcare system.8
Several studies have been undertaken to examine the frequency of readmissions within 1 month of discharge after stroke, mainly from the United States, Australia, Denmark, and Taiwan (Table 1).3–5,8–15 Common data sources used to calculate hospital readmissions rates include Medicare data, registry data, patient interviews, or medical records. Evidence from these studies show that the frequency of hospital readmissions may vary from 6.5% to ≈24.3%.3–5,8–15
The potential factors that may be associated with being readmitted after an initial hospitalization for stroke can be grouped into 5 broad categories: patient characteristics, social circumstances, health system, clinical care or process, and health outcome (Figure 1). A recent systematic review and other subsequent studies investigating factors associated with readmission after stroke have focused on <30 factors, which fall mainly within patient or health system categories (Table 1).3,9–12 Investigators have generally used Medicare data and have assessed <3 clinical care or process variables, such as discharge medication, admitted into a stroke unit, and readmission after stroke.
Since 2004, the New South Wales Stroke Audit Program has been used to monitor patient characteristics, social circumstances, health system, adherence to clinical processes of care, and outcomes before and after stroke unit implementation within hospitals.16,17 The detailed data available from this large study provided a unique opportunity to explore many factors and their potential association with early readmission after an initial hospitalization for stroke.
The aim of this study was to describe the factors associated with readmission after a hospitalization for stroke and determine whether there are any modifiable factors. Our primary hypothesis was that quality of care factors, in addition to patient and system factors, can explain differences between patients with stroke who were readmitted within 28 days and those who were not readmitted.
Data were obtained on patients with acute stroke who were admitted to New South Wales hospitals between 2000 and 2010 at various time periods. Methods of data collection and patient eligibility for this study have been previously published.16,17 In brief, consecutive medical record audits were conducted by trained data abstractors working at the hospitals using a validated data collection tool. Patient eligibility criteria were as follows: confirmed diagnosis of a first-ever or recurrent stroke admission to hospital for acute management and readmission status known (that is cases with unknown or missing information for this question were excluded). On the rare occasion a medical record was not available, then those patients were substituted for the next consecutive patient until the desired sample size was reached for that hospital. All audit data were independently processed and analyzed by externally based research staff.
Readmitted patients were defined as those readmitted to the same hospital within 28 days of discharge for an initial hospitalization for stroke. After careful review of the readmission records, auditors documented in free text, the primary cause for the readmission, referred to here as the readmission diagnosis. The readmission diagnoses for this study were then sorted into consistent disease categories by a medically trained research officer. Separate questions determined whether the readmission was related to the previous admission for stroke, the length of stay, and whether the patient was discharged from hospital for that admission
The variables selected for our analyses were based on findings from our literature review (Table 1) and our earlier research using this data set.16,17 Our literature review (Table 1) involved identifying previous studies investigating early readmission after stroke. Each of the articles was reviewed to estimate the number of factors investigated. Only key findings were highlighted under each of 5 identified broad categories.
Patient characteristic factors: n=16, including variables used to account for stroke severity on the basis of a validated prognostic model developed for predicting stroke outcome.18 These included arm weakness, impaired speech, inability to walk on admission, and incontinence in the first 72 hours. Dependence on admission and before stroke was defined according to the modified Rankin Scale (2–5).
Social circumstances factors: n=3.
Health system factors: n=6.
Clinical processes of care factors: n=13 with brain imaging defined as receiving a computed tomography or MRI scan.
Health outcomes: n=5 the modified Rankin Scale was used to define level of dependence on discharge with being dependent classified as modified Rankin Scale 3 to 519 equivalent to moderate to severe disability. Severe complications during the initial hospitalization were defined as events determined to be incapacitating, life-threatening, and prolonging the hospital admission and patient acuity (eg, falls, urinary tract infection, aspiration pneumonia, other chest infection, decubitus ulcer, deep vein thrombosis, stroke progression, second stroke, and myocardial infarct).
χ-square tests were used for categorical variables and the Wilcoxon Mann–Whitney rank-sum test for continuous variables for bivariate analyses.
Random effects multilevel, logistic regression models were used with level defined as hospital. Readmission status was defined as the dependent variable. A parsimonious approach to model development was used where independent variables with clinical importance or statistical significance (P<0.15) from bivariate analyses were used. Standard techniques were implemented to check for collinearity and the fit of various models were compared using Bayesian Information Criteria. Level of dependence on admission and discharge was collinear with 90% of patients dependent on admission and discharge. Therefore, in the final model we included only the dependent on admission variable. Stata (version 10.1; StataCorp, College Station, TX, 2010) statistical software was used for all analyses and P values of <0.05 were considered significant. Adjusted odds ratio (OR) and 95% confidence intervals (CIs) were calculated.
Of the 4139 patients with acute stroke included in this study, the median age was 74 years (Q1–Q3, 67–83); 50% were women; 92% had an ischemic stroke; 66% had first-ever stroke; and 90% had been discharged from hospital. Readmission status was unknown or missing for 11% of discharged patients. There was no difference in patient characteristics, such as age, or being dependent before admission for those where readmission status was known compared with patients whose readmission status was unknown. Patients with stroke were admitted from 35 Australian hospitals during multiple time periods (2000–2010). Quality of care received improved for patients admitted to stroke units (5%–69%; P<0.001) and receiving computed tomography or MRI scan within 24 hours (84%–96%; P<0.001) during this decade. Outcomes improved during this period with an increase in patients discharged home (54%–69%; P=0.03).
Readmissions to Hospital Within 28 Days of Discharge
Of the 3328 discharged patients with known readmission status 6.5% (95% CI, 3.2–9.7) were admitted to the same hospital within 28 days. The reasons for readmissions are summarized in Figure 2, with the majority being for a stroke or cardiovascular disease. The median length of stay for the readmission was 6 days (Q1–Q3, 2–12). Among the readmitted patients with discharge status recorded, 10% died during the readmission. The frequency of readmissions remained constant during the time period ranging from 4% to 19% (P=0.68).
Findings From the Descriptive Analyses of Factors
Table 2 shows that many of the patient characteristics and social circumstances of readmitted patients were similar to those who were not readmitted. However, readmitted patients were more likely to have a documented history of ischemic heart disease, be incontinent in the first 72 hours of admission and be dependent before admission for the initial hospitalization compared with patients not readmitted.
Differences in the health system, clinical care, and health outcome factors for patients with stroke who were readmitted and not readmitted to hospital are provided in Table 2. Management on a stroke unit did not differ for readmitted patients compared with those not readmitted, even during the poststroke unit implementation period. However, we found that readmitted patients were less likely to receive brain imaging within 24 hours of the initial admission when compared with patients who were not readmitted, but overall a similar proportion had brain imaging during their admission. On further examination, we noted that patients who did not receive brain imaging within 24 hours were more likely to have been admitted to a rural hospital (11%) compared with an urban hospital (6%; P<0.001). However, there was no statistically significant difference among patients in rural or urban hospitals receiving brain imaging during their stroke admission, but more rural patients had their scans done at different locations than the hospital where they were admitted for stroke (50% rural versus 10% urban), and this may have caused time delays. There were no differences among patients who received timely scans and whether they arrived after-hours to the hospital compared with those that did not receive a scan within 24 hours.
Patients who were readmitted were more likely to have been dependent at discharge and experience ≥1 or more severe complications during their initial hospitalization for stroke compared with patients not readmitted (Figure 3). Patients with severe complications had longer lengths of stay (median, 13 days) compared with patients without severe complications (median 8 days, P<0.001). Patients with severe complications were also more likely to be dependent at time of admission (OR, 1.6; 95% CI, 1.2–1.9) or discharge (OR, 9.3; 95% CI, 5.0–17.2).
Findings From Multivariable Analyses
The factors that remained significantly associated with patients being readmitted within 28 days included being dependent before the initial admission and experiencing a severe complication during the admission (Table 3). Our measures of stroke severity were not independent predictors of readmission in our adjusted model.
When we restricted our analyses to patients with no history of prior stroke, being dependent before the initial admission (OR, 2.1; 95% CI, 1.3–3.5) and experiencing a severe complication during the admission (OR, 3.3; 95% CI, 1.6–6.7) retained significance when all the potential factors associated with readmission were included in the model.
This is the first study to have comprehensively assessed a large and broad range of factors that may be associated with readmissions and to determine whether there were any differences in the quality of care received by patients on the basis of readmission status. The results provided evidence that patients who were being readmitted were generally receiving the same quality of care during the initial hospitalization as patients who were not readmitted. Our summary of the literature noted only 1 study by Chuang et al4 that found associations with clinical processes of care, such as need for wound care nursing or other invasive nursing care and adopting a care plan with early readmission after stroke. In our study, we found only 1 potential difference, which had borderline significance in the quality of care factors that differed for readmitted patients. Readmitted patients tended to receive fewer brain scans within 24 hours of admission compared with patients who were not readmitted. This result could not be explained by differences in time of arrival to hospital. However, patients admitted to rural hospitals in our sample were less likely to receive timely imaging because these services are often not collocated. Early brain imaging is required to confirm type of stroke, commence time-dependent therapies effective within the first 24 hours and to exclude stroke mimics.20
In our study, we investigated nonmodifiable factors, such as patients being dependent (having a moderate to severe disability) before their initial stroke admission. We found patients had a nearly 2-fold greater chance of being readmitted if they were dependent. It is interesting that the stroke severity variables we used did not seem to be an independent predictor of readmission. Hospitals treating patients with stroke should flag dependent patients as at risk of readmission. The level of dependence at discharge for stroke was also associated with readmission within 28 days. This indicates that these patients who are frailer or have more severe strokes because of greater comorbidity have a greater chance of readmission. Any other interventions aimed at preventing 28-day readmissions should place more attention on those with a moderate to severe disability. Hospitals should consider whether these patients might benefit from staying in hospital longer or ensure adequate transitions to other healthcare services after discharge from an acute hospital setting to prevent unplanned readmission.21
We also investigated potentially modifiable factors, such as patients with stroke experiencing a severe complication during admission. We found that patients had a ≈3-fold greater chance of being readmitted if they had experienced a severe complication. Our study showed that the likelihood of a severe complication or adverse event increases if you are dependent before stroke or have a severe disability after stroke. Hospitals treating patients with stroke should flag dependent patients before admission as at risk of severe complications and readmissions. Greater vigilance and monitoring may be warranted so that preventable serious adverse events, such as urinary tract infections or falls, are avoided. Targeted intervention programs need to be developed to increase efforts at prevention of complications during admission, especially for patients who were dependent at admission.
Short-term readmissions to hospitals are a worldwide problem. In the United States, Medicare is following readmission data closely. In other disease states, such as chronic heart failure, readmissions have been tied to reduced reimbursements to hospitals.22 One recommendation from our study is that hospitals treating patients with stroke should flag patients who experienced a severe complication during admission. These patients should be targeted for early outpatient follow-up after discharge, as there is recent evidence that these arrangements can potentially reduce readmissions.23
In Australia, data on factors related to readmissions for patients admitted with stroke are limited. The strength of our study is that we were able to use a large data set from a diverse range of hospitals within New South Wales, Australia. Other studies have used administrative data sets and were unable to include a large range of variables covered within the 5 categories we have identified. A limitation of our study was that we were only able to ascertain factors related to readmissions to the same hospital. Therefore, our readmission numbers may be an underestimate if patients within a geographical region went to a different hospital. In our sample, only 5% of patients had been transferred to the participating hospital from a different hospital for their initial stroke admission. It may be possible that some of these people did not return to the participating hospital for a readmission. Furthermore, we did not exclude planned readmissions because this information was not collected. On review of our data, there were potentially 8 readmissions that might be considered a planned readmission (1 for carotid endarterectomy, 1 for permanent pace-maker, and 6 for percutaneous gastrostomy tube insertion). In contrast, the studies using linked administrative Medicare data were able to look at factors associated with readmissions to any hospital and whether the readmissions were planned or unplanned. However, it has been shown that 81% of readmission within 30 days are admitted to the same hospital after a major disease event.24 Future studies are underway in Australia linking Registry data with administrative hospital data to explore further the factors associated with readmissions for patients with stroke to all hospitals.
This is the first time a large number of factors have been explored to assess potential modifiable and nonmodifiable factors associated with readmission. Our study showed that severe complications, which are potentially modifiable, were independently associated with a risk of readmission. Given the growing cost burden of stroke on healthcare systems, this study provides important information for clinicians, health administrators, and policy makers to identify patients at risk of readmission. Use of the categories outlined in this study will facilitate future comparative and comprehensive approaches for investigating this topic.
We acknowledge New South Wales (NSW) Health State-wide services development branch for funding the NSW stroke audit program. We also acknowledge L. Cutler and the NSW Rural Institute of Clinical Services and Training for their support of the rural audit initiative. We also acknowledge the hospital staff, stroke care coordinators, and area directors across Stroke Services NSW Network for supporting this work. The authors thank research assistants for their contribution to data processing from the Florey Institute of Neuroscience and Mental Health and L.C. Quang for her contribution to database management. A/Prof L. Churilov for advice on the multivariable modeling. We thank Dr B. Amatya for recoding all the readmission diagnoses into discrete categories.
- Received December 17, 2012.
- Accepted May 6, 2013.
- © 2013 American Heart Association, Inc.
- 1.↵Australian Institute of Health and Welfare. Australia’s Health 2006. 2006:512.
- Cadilhac DA,
- Carter R,
- Thrift AG,
- Dewey HM
- Lichtman JH,
- Leifheit-Limson EC,
- Jones SB,
- Watanabe E,
- Bernheim SM,
- Phipps MS,
- et al
- Fonarow GC,
- Smith EE,
- Reeves MJ,
- Pan W,
- Olson D,
- Hernandez AF,
- et al
- Langagergaard V,
- Palnum KH,
- Mehnert F,
- Ingeman A,
- Krogh BR,
- Bartels P,
- et al
- Lichtman JH,
- Jones SB,
- Leifheit-Limson EC,
- Wang Y,
- Goldstein LB
- Cadilhac DA,
- Pearce DC,
- Levi CR,
- Donnan GA
- Counsell C,
- Dennis M,
- McDowall M,
- Warlow C
- Sulter G,
- Steen C,
- De Keyser J
- 20.↵National Stroke Foundation. Clinical Guidelines for Stroke Management 2010. Melbourne: National Stroke Foundation; 2010.