Cost-effectiveness of patient navigation programs for stroke patients–A systematic review

Objective Stroke remains a leading cause of premature death, impairment and reduced quality of life. Its aftercare is performed by numerous different health care service providers, resulting in a high need for coordination. Personally delivered patient navigation (PN) is a promising approach for managing pathways through health care systems and for improving patient outcomes. Although PN in stroke care is evolving, no summarized information on its cost-effectiveness in stroke survivors is available. Hence, the aim of this systematic review is to analyze the level of evidence on the cost-effectiveness of PN for stroke survivors. Methods A systematic literature search without time limitations was carried out in PubMed, EMBASE, CENTRAL, CINAHL as well as PsycINFO and supplemented by a manual search. Randomized controlled trials published prior to April 2020 in English or German were considered eligible if any results regarding the cost-effectiveness of PN for stroke survivors were reported. The review was conducted according to PRISMA guidelines. Quality of included studies was assessed with the RoB2 tool. Main study characteristics and cost-effectiveness results were summarized and discussed. Results The search identified 1442 records, and two studies met the inclusion criteria. Quality of included studies was rated moderate and high. Programs, settings and cost-effectiveness results were heterogeneous, with one study showing a 90% probability of being cost-effective at a willingness to pay of $25600 per QALY (health/social care perspective) and the other showing similar QALYs and higher costs. Conclusions Since only two studies were eligible, this review reveals a large gap in knowledge regarding the cost-effectiveness of PN for stroke survivors. Furthermore, no conclusive statement about the cost-effectiveness can be made. Future attempts to evaluate PN for stroke survivors are necessary and should also involve cost-effectiveness issues.


Introduction
Stroke is the second most common cause of death (5.5 million cases annually). Its burden on health care systems remains high, with 13.7 million annual incident cases worldwide [1]. Strokes are mostly associated with psychological, cognitive and physical impairments, which become more onerous with increasing severity [2]. Due to these impairments, the risk of hospital readmission post stroke is high, and the costs of post stroke care can be very high, depending on severity and the length of hospital stay [3,4]. A stroke affects not only the lives of patients but also the lives of their relatives and caregivers, highlighting the individual burden and the public health relevance of stroke [5,6].
Thus, stroke survivors need long-term care and support in various areas of life. Care in general and the transition from hospital to outpatient care and rehabilitation in particular present many challenges. These challenges include sufficient information or assistance in accessing health care after hospital discharge (e.g., locating services, scheduling appointments) as well as appropriate education and training regarding the transition to the home environment [7].
One option for enhancing the transition between different care environments and thus improving patient-oriented outcomes such as quality of life is the concept of patient navigation (PN). Initially introduced in the context of barriers to timely breast cancer care for African-American women in Harlem, New York, current PN programs mostly focus on the abolition of one or multiple barriers to accessing appropriate health care services over a wide range of diseases [8][9][10]. Such programs can be very heterogeneous and be conducted by a variety of usually specifically trained health care professionals or lay health workers [11,12]. The components of PN programs vary with targeted barriers and can include for example, help in care coordination, emotional support, transportation services, educational modules and financial assistance components [13].
Within the context of stroke, PN is an evolving and promising approach for addressing barriers to appropriate health care and improving patient-oriented outcomes of stroke survivors [14][15][16][17][18]. Among others, PN for stroke survivors is associated with improvements in quality of life, treatment and prevention compliance as well as reduced hospital readmissions.
However, the general cost-effectiveness of such programs has not yet been investigated in a systematic and quantitative way. Therefore, the aim of this systematic review is to quantitatively analyze the scientific literature to assess the level of evidence for randomized controlled trials (RCTs) on the cost-effectiveness of PN programs for stroke survivors compared to a control group.

Search strategy
The PubMed, EMBASE, CENTRAL, CINAHL and PsycINFO databases were systematically searched for publications prior to April 2020 using predefined keywords and Medical Subject Headings (MeSH) to identify RCTs regarding the cost-effectiveness of PN programs for patients after experiencing a stroke. Broad search term combinations (keywords and MeSH) related to stroke, PN and cost-effectiveness analyses were used. The PubMed search strategy (see S1 Appendix) was adapted to the different databases. The systematic search included the gray literature especially conferences, books and dissertations and was supplemented by a manual screening of the reference lists of key studies, reviews, and study registers to identify additional eligible studies. Furthermore, the authors of study protocols were contacted to retrieve potentially unpublished results. Article inclusion was restricted to 1. RCTs 2. reporting cost-effectiveness results of PN programs for patients with a confirmed stroke, 3. that were written in German or English.
Based on a modification of the cancer-related PN definition by Wells et al. 2008, PN was defined as an intervention that focuses on the detection and elimination of individual barriers to appropriate health care and that is carried out personally (e.g., by care managers, nurses, lay health workers) for a defined episode of stroke-related care [11]. Since this review focuses on stroke survivors, articles were a priori excluded if they solely addressed caregivers or the results were not distinguishable between caregivers and patients. No further restrictions regarding stroke type, the number of previous strokes of participants, reported effectiveness and cost outcomes, and the evaluation perspective were applied. Similarly, no restrictions regarding follow-up lengths, publication date, or participant characteristics (such as age and sex) were used. Articles were excluded if they did not meet the criteria above.

Study selection and quality assessment
Records from all retrieved search results (systematic and manual) were downloaded and merged with Endnote version X9.2 for Windows (Clarivate Analytics, Boston, MA, USA). In the first step, the titles and abstracts of all retrieved records were screened. In the next step, the full texts were reviewed for potential study inclusion. Data from the included studies were extracted with a standardized spreadsheet (including relevant information on the PN intervention, study design and setting, participants, outcomes, and results). The revised Cochrane tool for assessing risk of bias in randomized trials (RoB2) was used to assess the quality of the included articles [20]. The RoB2 tool assesses the risk of bias of randomized trials in five distinct domains: 1) the randomization process, 2) deviations from the intended interventions, 3) missing outcome data, 4) measurement of the outcome, and 5) selection of the reported results. The tool specification for cluster-randomized trials contains an additional sixth domain, "bias arising from the timing of identification and recruitment of individual participants in relation to timing of randomization", which was applied if applicable. Each domain contains several signaling questions. The five or six domain judgments lead to an overall riskof-bias judgment. All domains and the overall judgment can result in "low risk of bias", "some concerns", and "high risk of bias". The study protocols and trial registry records of the included studies were incorporated into the assessment. Uncertainties regarding specific domain questions were tried to resolve through personal contact with the corresponding authors. The RoB2 results were used for descriptive purposes.
The entire screening, data extraction and quality assessment process was independently performed by two authors (B.K. and C.D.). These authors worked closely together with an experienced senior author (T.R.) throughout the review process. Discrepancies regarding article feasibility and risk of bias were discussed between the authors, and consensus was achieved.

Study selection
The initial database and manual search identified 1442 records (PubMed n = 960, EMBASE n = 147, CENTRAL n = 208, CINAHL n = 103, PsycINFO n = 20, manual n = 4). After removal of duplicates, 1340 records were screened, and if applicable, abstracts and full texts were reviewed for inclusion. A total of 1338 records did not meet the inclusion criteria. Among these, a focus on something other than stroke (n = 353) and interventions that did not meet the PN definition (n = 657) were the predominant reasons for exclusion. Full text reviews were carried out for 12 records. The PN definition was not met in eight studies [21][22][23][24][25][26][27][28][29], and two studies did not present cost-effectiveness results [14,30]. Hence, two studies were included in this review [31,32]. Fig 1 displays the PRISMA flowchart to illustrate the article identification and selection process. If more than one exclusion criterion was met, the first criterion observed was stated. The interrater reliability between the two reviewers was 100% during the full-text review. Table 1 provides the main characteristics of the two included studies [31,32]. Both studies were performed in the United Kingdom between 2009 and 2015. The cluster RCT carried out by Forster et al. 2015 collected and analyzed data from 32 stroke care coordinators (SCCs) and 800 participants (intervention group n = 401). The RCT by Rodgers et al. 2019 included 573 participants (intervention group n = 285) newly diagnosed with stroke. Both studies included slightly more men than women, and the participant ages were comparable (see Table 1). Both studies used PN as the primary intervention and incorporated cost-effectiveness analyses. However, the PN programs and settings were different. The study by Forster et al. delivered PN with an evidence-based care system through SCCs compared to usual national health care recommended stroke care coordination in a pragmatic cluster RCT setting with a 12-month follow-up. The intervention focused on the longer-term needs and problems of patients after stroke and of their carers [31]. The PN of the study by Rodgers et al. was nested in an early discharge setting after stroke and compared participants in the intervention group to usual care alone over a 24-month follow-up period. PN was delivered by senior members of an early discharge unit (e.g., physiotherapist, occupational therapist, nurse). Over an 18-month intervention period, regular care was accompanied by continuous interviews to identify patient needs in the context of rehabilitation and goal setting. Action plans were agreed upon and reviewed to maximize participants' recovery and ability to manage their day-to-day activities.

Study characteristics
In both studies, costs were determined by means of self-reported resource use questionnaires. Resource consumption was monetarily valued by both research groups by attaching standardized unit costs afterwards. The resulting total costs were calculated from the health and social care perspective over the specific follow-up lengths in both studies, while  WTP, the authors found a 90% probability of cost-effectiveness for the intervention. More details about the results of the studies are presented in Table 2.

Quality assessment
Risk of bias was rated overall "low" for the study by Forster et al., with all domains of the extended RoB2 tool for cluster RCTs rating "low". For the study by Rodgers and colleagues, two of the five domains of the RoB2 tool were judged as "some concerns", resulting in an overall risk of bias judgment of "some concerns". The main concerns arose from the study protocol deviation regarding the cost-effectiveness results [33]. In contrast to the planned analyses, only QALYs and not the scores of the Nottingham Extended Activities of Daily Living (NEADL) scale at 24 months were linked to total costs. The adjusted mean NEADL scale difference of 1.8 (95% CI: -0.7 to 4.2) in favor of the intervention group was smaller than the minimum clinically important difference on the NEADL scale of 6. Hence, the intervention did not have a favorable effect on patients' functional status as assessed with the NEADL scale. Only when QALYs were linked to total costs over 24 months did the intervention seem to be cost-effective. The corresponding author was contacted to comment on the topic and referred to the later published full report of the study (see also the discussion) [34]. More information about the risk of bias assessment is displayed in Table 3.

Author/Country Effectiveness Results
Cost Results

Discussion
This review aimed to assess and summarize the level of evidence on the cost-effectiveness of PN programs for stroke survivors. To the best of our knowledge, this study is the first systematic review of the cost-effectiveness of PN in the context of stroke. We identified only two studies meeting the inclusion criteria, revealing a large gap in knowledge on the topic. Just like the underlying cost-effectiveness results, the programs were heterogeneous, with one study showing a 90% probability of being cost-effective at a WTP of $25600 per QALY (health/social care perspective) and the other showing similar QALYs and higher costs (both health/social care and societal perspectives). These results, in combination with the small number of included studies, do not allow for a conclusive statement about the overall cost-effectiveness of PN for stroke survivors, and they demonstrate the need for further research in the field. Our quality assessment of the included studies ranged from moderate (overall RoB2: "some concerns") to high quality ("low risk of bias"). After our final search was completed, the full report related to the study by Rogers et al. 2019 was published [34]. This report was brought to our attention by the corresponding author. The report included further, hitherto unpublished and less favorable cost-effectiveness results related to the NEADL scale, which was the primary cost-effectiveness outcome according to the study protocol (see the results section). The publication of these results resolved some of our risk of bias concerns. However, we did not change our overall risk of bias rating. In our opinion, these results should have been mentioned in the main publication, and we believe that we correctly scored this publication in line with the RoB2 scoring guide.
Compared to PN for patients with different types of cancer or chronic diseases, the results of the studies included here are not as promising. PN in these fields has been shown to be more beneficial in terms of improving patient-oriented outcomes (e.g., quality of life) and having equal or lower health care costs [9,[11][12][13]. One explanation could be the heterogeneity in the aftercare needs of stroke survivors and their generally high unmet needs, which vary tremendously on the individual patient level. Inevitably, stroke survivors with severe hemiparesis living in a nursing home have needs that are different from those of stroke survivors with less severe or no spasticity issues living at home. This heterogeneity might lead to difficulties with respect to the construction of PN programs [35,36]. However, the heterogeneity of PN programs, e.g., in terms of navigation goals and the personal provision of services, and the different courses of disease across the continuum of diseases contradict the comparison across the care continuum.
Following the high standards for systematic reviews and making additional attempts to gather unpublished results were strengths of the present review. However, this review has several limitations that should be noted. The main concern arises from the small number of included studies, which calls into question the generalizability of the results. One reason for the low number of studies meeting our inclusion criteria may be our restriction to RCTs. Since modeling studies, which are often used in health economics research, are susceptible to uncertainty mostly due to the model assumptions, the restriction to RCTs was a priori defined to focus only on the highest level of evidence. Another reason can be found in the language restrictions of the author group, which allowed only studies published in English and German to be included. Nevertheless, we used a broad search strategy which was not restricted to RCTs and included the gray literature. Yet, we were unable to identify further relevant or ongoing studies, even when considering gray literature and additional languages at the level of title screening. Studies that implemented PN but that did not report cost-effectiveness outcomes were excluded. These studies could still contribute to the evaluation of PN; however, this review aimed to assess the cost-effectiveness of PN programs for stroke survivors and not the general effectiveness of PN. Moreover, even though studies meeting our broad definition of PN and our other inclusion criteria could be identified with our search strategy, it is noteworthy that no study labeled its intervention as PN. This finding suggests that the terminology remains uncommon in the context of stroke. Even though the term stroke nurse navigator was introduced more than 10 years ago in the United States and the underlying study showed promising results with respect to increased drug compliance, physician follow-up and potentially decreased rehospitalization rates [17]. Considering the scarcity of the term PN in the context of stroke, highlights the importance of broad search terms, which we applied, to identify the available body of evidence in the field.
Furthermore, both included studies were conducted only in the UK. Since health care resource consumption and costs are influenced by local health care structures (e.g., different payment systems, health provider incentives), the transferability of these results to other countries is limited [37]. In addition, different national health care systems face different degrees of fragmentation [38,39]. These differences in the degree of fragmentation may also have an influence on PN effectiveness, which might further limit the generalizability of the results. As more providers are involved in patient care, the need for coordination and cooperation is increasing and could be managed by PN.

Conclusions
The cost-effectiveness of PN for stroke survivors is an underrepresented research field. Only two studies meeting the inclusion criteria could be identified. Thus, our study reveals a large gap in knowledge, and no conclusive statement about the cost-effectiveness of PN programs for stroke survivors can be made. Since one of two studies suggested cost-effectiveness of PN for stroke survivors while the other presented comparable quality of life and cost results in both groups, the research into PN in stroke survivor care should be pursued further to close the revealed gap in knowledge. This could help stakeholders make more informed decisions with respect to the promotion and implementation of PN programs for stroke survivors. Although the implementation of new health care models such as PN depends in part on their cost-effectiveness, future PN studies in regard to stroke survivors are needed and should include comprehensive health economic analysis strategies.
Supporting information S1