Introduction
The term stroke refers to a clinical syndrome, of presumed vascular origin, typified by rapidly developing signs of focal or global disturbance of cerebral function, lasting more than 24 h (World Health Organization, 1978). Stroke is the second most common cause of death in adults worldwide (Lopez and Mathers, Reference Lopez and Mathers2006) and, with heart disease, is predicted to be the biggest global cause of loss of disability-adjusted life years by 2025 (Murray and Lopez, Reference Murray and Lopez1997).
Two-thirds of patients will survive their first stroke but many are left with long-term disability (Hare et al., Reference Hare, Rogers, Lester, McManus and Mant2006), and will have ongoing needs for rehabilitation, and healthcare and social-care services. Our earlier study described the information needs of patients and caregivers after stroke (Allison et al., Reference Allison, Evans, Kilbride and Campbell2008). A number of other studies have attempted to establish the longer-term needs of patients and caregivers. Murray et al. (Reference Murray, Ashworth, Forster and Young2003) identified that patients reported long-term difficulties in a number of areas including the physical, social, and emotional consequences of stroke. Hare et al. (Reference Hare, Rogers, Lester, McManus and Mant2006) identified that patients and caregivers indicated the importance of using primary care as the first point of contact. The National Stroke Strategy in the United Kingdom (Department of Health, 2007) recommends that people and their caregivers should have access to a review from primary care providers of their healthcare and social-care status and secondary prevention needs at six weeks and six months after discharge, and then annually.
Who should undertake these reviews and within what model of care is uncertain as there is little guidance on the feasibility or effectiveness of different models of stroke follow-up in primary care. Mitchell et al. (Reference Mitchell, Brown, Erikssen and Tieman2008) reported a systematic review of care planning involving primary care in the management of people with stroke but the majority of studies included involved coordinating the process of hospital discharge, rather than longer-term management of such people. Given this present focus of evidence on hospital-based processes, we undertook this review to investigate the evidence for the effectiveness of different models of primary care-based follow-up after stroke.
Methods
We conducted a systematic review of published research articles of interventions in primary care follow-up after stroke. We defined these as interventions involving both a healthcare and social-care element, which would be suitable to meet the requirements of the post-discharge and annual review of the stroke strategy. We interpreted this as interventions that included a review of medications and an assessment of longer-term disability and caregiver needs, as well as the provision of information and signposting to other services. We did not include studies that considered secondary prevention in isolation or that evaluated ongoing community-based rehabilitation. The review excluded studies in which follow-up interventions were provided by secondary care alone, but included interventions involving reaching out into primary care. The review primarily focused on how and by whom the models of follow-up care were coordinated. Articles were included if they were written in English, had targeted a population of people over the age of 18 years in primary care with a diagnosis of stroke, and where any of the following outcomes were evaluated: patient satisfaction, patient health outcome, levels of depression, patient behaviour, levels of function, quality of life, caregiver burden scales, and readmissions to hospital.
Search strategy
Multiple search strategies including electronic online database searches and hand searching of individual journals were used. The following electronic databases were searched: MEDLINE (1950–May 2008), EMBASE (1980–May 2008), CINAHL (1982–May 2008), AMED (1985–May 2008), and the Cochrane Library (1980–May 2008). The following key words were used in combinations: stroke, cerebrovascular, long-term care, chronic disease, primary health care, family practice, and community health services. Relevant papers were also identified by citation tracking, using reference lists from journals.
One investigator (RA) conducted the search. All titles or abstracts retrieved were then screened by at least two of the other investigators (RA, TA, LS, or RD) to decide whether the study was suitable for inclusion. Disagreements about eligibility were resolved by consensus or adjudication by a third investigator.
Data extraction and analysis
Key data were extracted from the identified papers using a structured form, which included the following components: general study information (title, author, and country of study), details of the intervention, study characteristics (population data, intervention, comparator, and outcomes), and findings including length of follow-up. The theoretical basis of the intervention was also extracted (if reported).
All relevant studies were independently rated by at least two of the investigators (RA, TA, LS, or RD) using the Quality Assessment Tool for quantitative studies developed by the Effective Public Health Practice Project at McMaster's University in Canada (Effective Public Health Practice Project, 2008). This tool rates a number of components for risk of bias including sample selection, study design, identification of confounding factors, blinding of assessors, reliability and validity of data collection methods, recording of withdrawals, and intervention integrity and analysis. The tool produces an overall rating of strong, moderate, or weak. It has been shown to have good content and construct validity, in addition to good reliability, and has been used in at least 20 systematic reviews (Thomas et al., Reference Thomas, Ciliska, Dobbins and Micucci2004). In this review, κ-statistics were used to analyse inter-observer reliability of the overall quality rating. Any discrepancy in data extraction or quality rating was resolved through discussion.
Results
Over 20 000 references were initially identified, although the majority of these did not evaluate models of care. A total of 128 abstracts were retrieved, but 119 were excluded because they focused only on secondary prevention of stroke or on ongoing community-based rehabilitation. In total, nine studies were suitable for quality assessment. All of the studies referred to were randomised controlled trials. The articles by Claiborne (Reference Claiborne2006a; Reference Claiborne2006b) and Mant et al. (Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005) presented data from the same study, but because different outcome measures were reported in the former and different time scales in the latter, they were included separately in this systematic review.
Interventions
The studies that were of sufficient quality for inclusion evaluated interventions provided by stroke support workers, care coordinators, case managers, or a care management model linked with systems of recall and clinical guidelines in primary care. All the studies identified patients before hospital discharge but did not report how long this was after the stroke. They all evaluated interventions that were provided for periods of three months up to nine months, with final follow-up usually at the end of the intervention period. Five studies were from the United Kingdom and four from North America.
In seven of the studies (Dennis et al., Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997; Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005; Lincoln et al., Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003; Tilling et al., Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005; Claiborne Reference Claiborne2006a; Reference Claiborne2006b), there did not appear to be a formal established relationship between the intervention worker and other primary care services. In the remaining two studies, there was evidence of a relationship linking the intervention with the patient's general practitioner (GP), although in one of the studies the nature of this link was only loosely described (Mayo et al., Reference Mayo, Nadeau, Ahmed, White, Grad, Huang, Yaffe and Wood-Dauphinee2008), and in the other (Allen et al., Reference Allen, Hazelett, Jarjoura, Wickstrom, Hua, Weinhardt and Wright2002), although the protocol detailed joint working, the detail of that relationship was not recorded.
This corresponds to the findings of Mitchell et al. (Reference Mitchell, Brown, Erikssen and Tieman2008) who found that few studies looking at multidisciplinary care planning described the actual role provided by GPs.
Theoretical basis
The theoretical basis of the organisational model was difficult to establish in the majority of the reports because most studies presented only an introduction and a review of the methods. Four of the studies (Dennis et al., Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997; Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005; Lincoln et al., Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003) evaluated existing components of care. The study of stroke support workers conducted by Tilling et al. (Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005) was preceded by two postal surveys, one of stroke support workers and one of patients, both suggesting what the intervention should include. The remaining studies used theoretical models of case management (Mayo et al., Reference Mayo, Nadeau, Ahmed, White, Grad, Huang, Yaffe and Wood-Dauphinee2008), chronic disease management (Allen et al., Reference Allen, Hazelett, Jarjoura, Wickstrom, Hua, Weinhardt and Wright2002), and care coordination (Claiborne Reference Claiborne2006a; Reference Claiborne2006b), but did not formally examine the a priori evidence for these models.
Methodological quality
The methodological details reported in the papers were of variable quality. All of the studies described selection criteria and gave a full account of patients who withdrew from the study. Using the McMaster Tool, three studies were found to be of high quality, five were of moderate quality, and two were rated weak. The quality of studies is presented in Table 1. The most common shortcomings related to lack of sample size calculations and inadequate assessor blinding. The inter-rater agreement for judging the quality of the studies was moderate (κ-coefficient 0.59).
Review of included studies
BP = blood pressure; SA-SIP30 = Stroke-Adapted 30-item Sickness Impact Profile version; GDS = Geriatric depression scale; FAI = Frenchay Activities Index; GHQ = General Health Questionnaire; HADS = Hospital Anxiety and Depression Scale; FSO = Family Support Organiser; Not EADL = Nottingham extended activities of daily living; RMI = Rivermead Mobility Index; LHS = London Handicap scale; TUG = Timed Up and Go.
Participants
The participants in the studies are shown in Table 1. A total of 1425 patients and 267 caregivers participated in the studies included in the review. Broadly, the studies included a representative sample of people after stroke, although most studies excluded people who had previously lived in nursing homes, and two of the studies only included people with caregivers (Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005).
Outcomes
Most studies did not identify a single primary outcome measure but reported instead on a range of outcomes including physical functioning, mood, quality of life, hospital readmissions, and health service utilisation. In one of the studies (Allen et al., Reference Allen, Hazelett, Jarjoura, Wickstrom, Hua, Weinhardt and Wright2002), traditional measures were subsumed within composite measures, which meant that identifying the effect on individual outcomes was impossible. The heterogeneity of the interventions meant that it was not possible to meta-analyse the data. A summary of the findings is given in Table 1.
Physical functioning
Six of the nine studies included physical functioning as an outcome measure. The well-established Barthel Index was used as a measure in five of the studies. In addition, one study (Mayo et al., Reference Mayo, Nadeau, Ahmed, White, Grad, Huang, Yaffe and Wood-Dauphinee2008) evaluated the impact of the intervention on gait speed and Timed up and Go test, two studies reported the use of the Rivermead Mobility Index (Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005), three studies examined the use of the Frenchay Activities Index (Dennis et al., Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997; Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005), and one used the Nottingham Extended Activities of Daily Living Scale (Lincoln et al., Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003). None of these studies showed any significant differences between the intervention and control groups in physical functioning.
Mood
Eight of the studies included an outcome measure of mood. Four studies used the Hospital Anxiety and Depression Scale (HADS; Dennis et al., Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997; Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005; Tilling et al., Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005), and the remainder used the Centre for Epidemiological Studies Depression Scale (Allen et al., Reference Allen, Hazelett, Jarjoura, Wickstrom, Hua, Weinhardt and Wright2002), the General Health Questionnaire (Lincoln et al., Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003), and the Geriatric Depression Scale (Claiborne, Reference Claiborne2006b; Mayo et al., Reference Mayo, Nadeau, Ahmed, White, Grad, Huang, Yaffe and Wood-Dauphinee2008). Claiborne (Reference Claiborne2006b) showed that patients receiving the care coordinator intervention were significantly less depressed than those in the control group. This study was the smallest with only 28 participants in total and was rated ‘weak’ with the McMaster Tool. In all of the other studies, no difference was demonstrated between the intervention and control groups.
Quality of life
Six of the studies measured quality of life, using SF-36 (Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005; Mayo et al., Reference Mayo, Nadeau, Ahmed, White, Grad, Huang, Yaffe and Wood-Dauphinee2008), Dartmouth Co-op charts (Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005), Euroquol EQ-5D (Mayo et al., Reference Mayo, Nadeau, Ahmed, White, Grad, Huang, Yaffe and Wood-Dauphinee2008), Re-integration to Normal Living Index (Tilling et al., Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005), and Stroke-Adapted Sickness Impact Profile (Allen et al., Reference Allen, Hazelett, Jarjoura, Wickstrom, Hua, Weinhardt and Wright2002), respectively. One study (Claiborne, Reference Claiborne2006b) indicated a significant improvement in the mental health component of SF-36, but this was a small study of overall weak quality. No other studies showed any significant difference on measures of quality of life.
Health service utilisation
Two of the studies examined health service utilisation. Mayo et al. (Reference Mayo, Nadeau, Ahmed, White, Grad, Huang, Yaffe and Wood-Dauphinee2008) showed that there were no differences in health service utilisation between groups receiving care from a stroke coordinator and those receiving standard care. Claiborne (Reference Claiborne2006a) showed that patients receiving case management from a social worker attended significantly more outpatient appointments but had less emergency attendances than those in a control group. However, the sample size was small (n = 28) and there was large variation in each individual's use of services.
Patient behaviour
Only one study (Claiborne, Reference Claiborne2006b) reported the effects of the intervention on patient behaviour. Claiborne measured behaviour through an index of self-care that included concordance with medication regimes and attendance at hospital appointments. Although the study showed a significant improvement in adherence to self-care in favour of intervention, it is unclear how this was measured on the index or how the validity of the measure was established.
Satisfaction
Five studies reported patient and caregiver satisfaction with services. Four of these studies (Dennis et al., Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997; Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005; Tilling et al., Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005) used a formally developed satisfaction questionnaire, and the other (Lincoln et al., Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003) used a questionnaire designed by the study team.
With regard to patient satisfaction, none of the studies showed a significant difference in overall satisfaction, although some of the studies identified differences in response to individual aspects of satisfaction. For example, in Dennis et al. (Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997), patients in the intervention group felt someone had listened to their needs. Tilling et al. (Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005) showed small non-significant effects with higher satisfaction with information about stroke prevention but lower satisfaction with community services, and Lincoln et al. (Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003) found that patients in the intervention group were more satisfied with information about community services and about emotional support.
For caregivers, two studies (Mant et al., Reference Mant, Carter, Wade and Winner2000; Tilling et al., Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005) showed no differences in caregiver satisfaction. Lincoln et al. (Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003) showed a statistically higher level of satisfaction for caregivers in the questions concerning provision of information on emotional support and practical help, but this did not translate into a difference in overall satisfaction. Dennis et al. (Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997) found that carergivers receiving follow-up were significantly more likely to feel that someone had listened and were less likely to feel neglected.
Perceived knowledge
Three of the studies examined the perceived knowledge of both patients and caregivers (Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005; Lincoln et al., Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003), and one looked at patients alone (Allen et al., Reference Allen, Hazelett, Jarjoura, Wickstrom, Hua, Weinhardt and Wright2002). All of these studies used questionnaires designed by each study team, usually to assess knowledge of whom to call in the event of another stroke, how to access information, and secondary prevention. None of the studies described how these questionnaires had been validated and the findings were inconsistent. Lincoln et al. (Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003) showed that both patients and caregivers in the intervention group had a higher perceived knowledge of stroke, risk factors, and resources available. Allen et al. (Reference Allen, Hazelett, Jarjoura, Wickstrom, Hua, Weinhardt and Wright2002) also showed a positive benefit for patients. However, Mant et al. (Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005) established no differences for either patients or caregivers in respect of knowledge regarding stroke.
Caregiver strain
Five of the studies measured caregiver burden: four used the Carer Strain Index (Mant et al., Reference Mant, Carter, Wade and Winner2000; Reference Mant, Winner, Roche and Wade2005; Lincoln et al., Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003; Tilling et al., Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005) and one used the Care Giving Hassles Scale (Dennis et al., Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997). Although the trend in one study was for less strain in caregivers in the intervention group (Dennis et al., Reference Dennis, O'Rourke, Slattery, Staniforth and Warlow1997), there were no significant differences observed in favour of the intervention group in any of the studies. Four of the studies also looked at care-giver mood and socialisation. Although two studies showed no impact on caregiver mood using HADS (Tilling et al., Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005) or General Health Questionnaire (Lincoln et al., Reference Lincoln, Francis, Lilley, Sharma and Summerfield2003), Mant et al. (Reference Mant, Carter, Wade and Winner2000) showed that caregivers in the intervention group reported less depression on the HADS and were involved in significantly more social activities. These positive effects were continued at follow-up at one year.
Adverse events
Adverse events or side effects of the interventions were not reported in any of the studies.
In summary, patients and caregivers receiving formal primary care follow-up delivered using a range of models did not show any gains in physical function, mood, or quality of life. Some patients and caregivers receiving follow-up were more satisfied with some aspects of communication, and had a higher perceived knowledge of stroke, but these were not consistently shown. In two of the studies, patients receiving follow-up had fewer emergency attendances when compared with patients not receiving follow-up.
Discussion
The aim of this systematic review was to evaluate the effectiveness of models of primary care-based follow-up after stroke. The studies identified compared groups of patients and caregivers who received contact with stroke support workers, care coordinators, and case managers with those who received standard care. From the evidence reviewed, there is insufficient evidence to support or refute the value of these models of follow-up after stroke.
Most of the studies included in this review provided an inadequate description of their methods to allow a full assessment of their methodological quality. Where detailed methods were described, most studies were found to be of only moderate quality. Given the relatively small sample size of most of these trials, a lack of statistically significant differences between groups may simply reflect a lack of statistical power rather than the absence of a real effect. The heterogeneity of the interventions meant that it was not possible to meta-analyse the data.
The choice of outcomes may also have affected the ability to demonstrate an effect. First, it is unclear whether the outcomes selected would be sensitive to change. For example, models of follow-up based on general support and information transfer are unlikely to directly affect physical function, and since the models did not include a formal intervention to manage depression, such as a psychological therapy, they may be unlikely to affect mood. Only two of the studies identified primary outcome measures (Tilling et al., Reference Tilling, Coshall, McKevitt, Daneski and Wolfe2005; Mayo et al., Reference Mayo, Nadeau, Ahmed, White, Grad, Huang, Yaffe and Wood-Dauphinee2008). Most of the studies measured a wide range of outcomes and it was often unclear exactly which outcome the intervention was targeting. It is also possible that the choice of outcomes was not ideal, as some may have shown ceiling effects. The Barthel Index, for example, focuses on basic physical function, and in three of the studies, most participants had achieved the maximum scores at the time of the final measure.
The majority of studies did not report a sound theoretical basis for the development of the intervention given. It is essential that a ‘complex intervention’, such as a model of healthcare and social-care follow-up, has a theoretical basis (Medical Research Council, 2008). Theoretical models of chronic disease management (Wagner, Reference Wagner1998) have been used to improve approaches to care in other conditions, such as diabetes (Wagner et al., Reference Wagner, Grothaus, Sandhu, Galvin, McGregor, Artz and Coleman2001) and chronic pulmonary disease (Adams et al., Reference Adams, Smith, Allan, Anzueto, Pugh and Cornell2007), and consideration should be given to the value of these models for stroke.
Finally, although the interventions provided took place in the community, there appeared to be little attention paid to the extent and quality of communication between intervention staff and the patient's GP. There is significant evidence to show how poor communication between primary care and secondary care clinicians, or other community services may have a negative impact on patient care (White and Marriot, Reference White and Marriot2004; Farquhar et al., Reference Farquhar, Barclay, Earl, Grande, Emery and Crawford2005). We would, therefore, suggest that future studies of follow-up give consideration to the mechanisms of communication and interface with primary care staff.
The National Stroke Strategy (Department of Health, 2007) recommends that stroke survivors and their caregivers are provided with a healthcare and social-care review by primary care professionals at key points following their discharge from hospital, and that patients and caregivers have identified the importance of primary care as a first point of contact (Hare et al., Reference Hare, Rogers, Lester, McManus and Mant2006). Currently, there is very little evidence on which to base models for providing these reviews. Further research is underway to evaluate a comprehensive assessment approach to post-acute stroke care (Murray et al., Reference Murray, Young, Forster, Herbert and Ashworth2006), but attention may still need to be given on how assessment roles link with GPs. Further research is needed to inform future practice, and examples of chronic disease management, such as those increasingly used in the United States, may be useful in developing models of follow-up after stroke.
Conclusions
Overall, the findings do not support the use of stroke support workers, care coordinators, or case managers working in the ways described in these studies to deliver the primary care-based healthcare and social-care review after stroke. The lack of a theoretical model on which to base the worker's intervention, and the lack of a defined relationship or route of communication with the stroke survivor's GP may be potential reasons for the lack of positive effect.
The lack of a sound theoretical basis for the development of primary care-based follow-up after stroke highlights the need for urgent work to develop this. The Complex Interventions Framework (Medical Research Council, 2008) provides some guidance on how this could be developed. The limited quality of the studies considered in this review highlights the need for high-quality trials of primary care-based interventions aimed at following up patients who have sustained a stroke.
Appendix 1 : Search Strategy
MEDLINE, CINAHL, EMBASE, AMED, BNI
1. Stroke
2. Cerebrovascular
3. CVA
4. CVE
5. Vascular accident
6. Long-term care
7. Chronic disease
8. Primary health care
9. Family practice
10. Community health services
11. Follow-up
12. 1, 2, 3, 4, or 5
13. 6, 7, 8, 9, 10, or 11
14. 12 and 13
Cochrane Stroke Group Trials Register, Cochrane Central Register of Controlled Trials (CENTRAL), Cochrane Effective Practice, and Organisation of Care Group (EPOC)
1. Stroke
2. Long-term care
3. Chronic disease
4. Primary health care
5. Family practice
6. Community health services
7. Follow-up
8. 2 or 3 or 4 or 5 or 6 or 7
9. 1 and 8
Department of Health data, Kings Fund
