Coordination of patient care between primary care and oncology care providers is vital to care quality and outcomes across the cancer continuum, yet it is known to be challenging. We conducted a systematic review to evaluate current or new models of care and/or interventions aimed at improving coordination between primary care and oncology care providers for patients with adult breast and/or colorectal cancer. MEDLINE, EMBASE, CINAHL, Cochrane Library Database of Systematic Reviews, and the Centre for Reviews and Dissemination were searched for existing English language studies published between January 2000 and 15 May 2015. Systematic reviews, meta-analyses, randomised controlled trials (RCTs) and non-randomised studies were included if they evaluated a specific model/intervention that was designed to improve care coordination between primary care and oncology care providers, for any stage of the cancer continuum, for patients with adult breast and/or colorectal cancer. Two reviewers extracted data and assessed risk of bias. Twenty-two studies (5 systematic reviews, 6 RCTs and 11 non-randomised studies) were included and varied with respect to the targeted phase of the cancer continuum, type of model or intervention tested, and outcome measures. The majority of studies showed no statistically significant changes in any patient, provider or system outcomes. Owing to conceptual and methodological limitations in this field, the review is unable to provide specific conclusions about the most effective or preferred model/intervention to improve care coordination. Imprecise results that lack generalisability and definitiveness provide limited evidence to base the development of future interventions and policies.
Trial registration number CRD42015025006.
Cancer is the leading cause of death in Canada1 and primary care is the first and most frequent point of contact for patients with cancer within the healthcare system. Although cancer care forms a modest (∼10%) workload component for primary care providers,2 the intricacy and urgency of patients’ care needs are unique and can be complex. These needs include unintended chronic and late-occurring complications of cancer and its treatment, possible oversight of post-treatment surveillance regimens and other multiple concurrent chronic conditions, including those that place patients at higher risk of the adverse effects from cancer treatments.3
The coordination of patient care between primary care and oncology care providers is vital to improve the quality and outcomes of care across the cancer continuum;4 however, this coordination of care has been very challenging.2 ,5 ,6 For example, the poor adherence to recommended cancer-specific surveillance regimens and preventative services may be influenced by the ambiguity regarding providers’ roles and diffusion of responsibility, suboptimal collaboration between care providers and increased volume of patients with cancer.7–9 Initiatives to improve care coordination, and evaluations of shared models of care, have been studied; however, a synthesis of this knowledge is required to understand the tactics that can best optimise this goal. To this end, the purpose of this systematic review is to evaluate models of care and interventions designed to improve coordination of care between primary care and oncology care providers for adult patients with breast and/or colorectal cancer. Choice of patient group was restricted to these two diagnoses because of the high prevalence of these diseases,10 and to align with the mandate of the Canadian Team to Improve Community-Based Cancer Care along the Continuum (CanIMPACT) programme of research.
This project is part of a larger research programme called CanIMPACT, which aims to improve coordination of care along the cancer care continuum between primary care provider and cancer specialist communities.11
Protocol and registration
The systematic review has been conducted and reported using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines.12 Details of the protocol for this systematic review were registered on PROSPERO.13
MEDLINE, EMBASE, CINAHL, Cochrane Library Database of Systematic Reviews and the Centre for Reviews and Dissemination were searched for existing English language studies published between January 2000 and November 2014. The Cochrane Central Register of Controlled Trials was searched for ongoing studies. The full original search strategy can be found in the online supplementary material. Reference lists of relevant systematic reviews were manually searched to identify eligible studies. The search was updated in May 2015 to include articles published between November 2014 and 15 May 2015.
Eligibility criteria were defined a priori and included the following elements:
Included studies evaluated a specific model of care or intervention (1) designed to improve care coordination between primary care and oncology care providers, AND (2) for any stage of the cancer continuum, AND (3) for adult breast and/or colorectal cancer patients. A model of care/intervention was conceptualised as a programme or project that is intended to improve some aspect of cancer care coordination, delivery, organisation or patient care. Examples of ‘interventions’ are the implementation of case management and primary care-led follow-up.
Systematic reviews, meta-analyses, randomised controlled trials (RCTs) and non-randomised studies, evaluating the effectiveness of a model of care or intervention on patient, provider or system outcomes, were included. Retrospective and descriptive studies, unpublished data, abstracts and conference proceedings were excluded.
The model of care or intervention being evaluated had to include both primary care providers (eg, family physicians, nurse practitioners, family practice nurses, community pharmacists and physiotherapists) and oncology care providers (eg, medical/radiation oncologists, general/family practitioners in oncology, oncology specialists, generalist surgeons and advanced practice or specialist nurses). The participants involved in the evaluation could include members of the healthcare provider group and/or patients who were treated under the model of care being evaluated.
For RCTs and non-randomised studies, the comparator group could receive another model of care intervention, standard/usual care or no intervention.
Types of outcome measures
Study eligibility was not dependent on reported outcomes; patient, provider and system outcome measures at all time points were included. Patient outcomes, such as survival, quality of life and chronic adverse effects, were prioritised.
Literature search results were uploaded to, and deduplicated in, EndNote X7 reference management software. Titles and abstracts, followed by full texts of records meeting the initial screening criteria, were retrieved and examined independently by two reviewers (research assistants). Disagreements related to screening were resolved through discussion to reach consensus.
One research assistant extracted data from the included studies using a pilot-tested form, and a second research assistant verified the extracted data through a formal audit process to reduce errors and bias. Disagreements were resolved by consensus. The following data were extracted from each article: (1) study design; (2) setting; (3) risk of bias assessment (more information below); (4) characteristics of participating providers; (5) characteristics of participating patients; (6) model of care or intervention characteristics (including purpose, description and implementation); (7) outcome measures; and (8) results.
Risk of bias assessment
The two reviewers independently assessed the risk of bias for each included study. Disagreements were resolved by consensus. The Assessing Methodological Quality of Systematic Reviews (AMSTAR) Checklist14 was used to assess systematic reviews and Cochrane Collaboration tools15 ,16 were used to assess randomised and non-randomised studies.
The results were summarised in tables according to study design and a systematic narrative synthesis was conducted for each type of study. A meta-analysis was not possible due to heterogeneity in the outcome measures used in the studies.
Results of the search
The original search generated 5064 references. An additional 703 were identified through the search update, resulting in a total of 5767 articles; 130 full-text articles were assessed for eligibility. Twenty-two articles met the eligibility criteria and were included in this review. A PRISMA flow chart outlining the study selection process and reasons for exclusions is found in figure 1.
Details of included studies
Five systematic reviews,17–21 six RCTs22–27 and 11 non-randomised studies28–38 comprised the review. The six RCTs and 11 non-randomised studies were not already included in the five systematic reviews. The findings from each type of study are discussed below.
The number of studies included in the five systematic reviews17–21 ranged from 19 to 51. The types of study designs, the number of studies included, the model(s) of care evaluated and the outcomes measured varied across the systematic reviews, making comparisons difficult. None of the reviews specifically focused on breast cancer or colorectal cancer, but did include these patient groups. Three of the five systematic reviews were considered to be of moderate quality (ie, AMSTAR score of 5-8/11) and two were considered to be of low quality (ie, AMSTAR score of 0-4/11). table 1 provides AMSTAR results and table 2 provides study details and outcomes.
Overall, the reviews did not support any one model or intervention over another to improve continuity and/or coordination of care. This was due to heterogeneity of outcomes, lack of clinical-centred or person-centred outcomes, and the overall low quality of the studies. The most definitive conclusion that emerged from the existing reviews was from Howell et al,18 who found that primary care and nurse-led models of care are equivalent in post-surgical colorectal cancer populations and following adjuvant treatment for breast cancer. These results are similar to earlier conclusions by Lewis et al,20 who demonstrated no statistically significant difference in survivorship, recurrence of cancer or psychological morbidity between physician-led and nurse-led follow-up care. All systematic reviews concluded that better quality investigations are warranted.
Randomised controlled studies
Six RCTs were included in the review, one each from Canada,23 Australia,24 the UK25 and the USA,26 and two from Denmark.22 ,27 Studies were designed for all cancers,22 haematological, breast, ovarian and colorectal cancers,24 breast, colorectal and lung cancers,25 ,26 breast cancer only23 and colorectal cancer only.27 Two RCTs examined interventions that targeted transition from diagnosis through to treatment,26 ,27 one study targeted only treatment,24 two studies targeted transition from treatment through to survivorship22 ,25 and one study targeted survivorship specifically.23 The types of model designs or interventions included case management,22 ,24 ,25 ,27 nurse navigation26 and the dissemination of survivorship care plans.23 Outcomes assessed in the studies also varied. All six RCTs examined patient outcomes, including satisfaction with care,22 ,23 ,27 self-advocacy,25 dropout,25 quality of life,23 ,26 patient experience,23 ,26 cancer-related distress,23 anxiety/depression and empowerment24 and adverse effects of treatment.24 Four studies examined provider outcomes, such as satisfaction with care provided,22 ,24 ,27 and system outcomes, such as number of patient visits with a general practitioner and/or oncologist23 ,27 and costs after diagnosis.26 The key issues regarding study bias were lack of blinding of participants and personnel, and selective reporting of outcomes. table 3 provides a Cochrane risk of bias table and table 4 describes study details and outcomes. Overall, there were no significant changes reported in any of the measured outcomes.
Eleven non-randomised studies examining models of care/interventions to improve care coordination between primary care and oncology care providers were eligible for inclusion in the review.28–38 The countries of origin of these studies included three each from the USA30 ,31 ,37 and the UK,28 ,32 ,35 two each from Canada27 ,36 and Australia,29 ,38 and one from New Zealand.36 The types of cancer targeted in the studies included breast only (n=4),29 ,30 ,34 ,38 colorectal only (n=4),28 ,33 ,35 ,36 either breast or colorectal (n=2)31 ,37 and breast, colorectal and other (n=1).32 The majority of studies piloted interventions for patients in the transition from treatment to survivorship (n=8),31–38 with only one intervention focused on diagnosis28 and two focused on the transition from diagnosis through to treatment.29 ,30 Intervention strategies included the use of nurse navigators (n=3),29 ,35 ,36 treatment or survivorship care plans (n=3),30 ,31 ,37 referral letters/forms (n=1),28 transfer into survivorship assessments (n=1)38 and shared care/creation of multidisciplinary teams (n=1).32 Two studies reported interventions that comprised a number of strategies from those previously listed.33 ,34 The outcomes assessed in the studies ranged from patient outcomes (eg, perceptions of quality of care,30 satisfaction with care,30–33 ,35 ,37 quality of life,35 psychological morbidity34), to provider outcomes (eg, satisfaction with care32 ,35) to system outcomes (eg, wait times,29 number of tests requested or appointments,36 patient referrals28 ,38 and cost savings35). Eight studies were judged to be of serious risk of bias and three were judged to be of moderate risk of bias. table 5 provides A Cochrane Risk of Bias Assessment Tool: for Non-Randomized Studies of Interventions (ACROBAT-NRSI) results and table 6 describes the study details and outcomes. All of the non-randomised studies presented inconclusive results.
This review evaluated models or intervention aimed at improving coordination of care between primary care and oncology care providers, for patients with breast and/or colorectal cancer. The 22 included studies varied with respect to type of model (most incorporating nurse navigation25–27 ,29 ,33 ,35 ,36) or intervention (most incorporating survivorship care plans23 ,31 ,33 ,34 ,37) evaluated. Many different outcome measures were used and assessed the impact at provider, patient and system levels. The conceptual and methodological limitations with the studies make it challenging to provide specific conclusions about the model or intervention tactic that would lead to changes in patient, provider and system outcomes.
Two conceptual issues with the research until now became evident when conducting the review. First, the majority of the reviewed studies provide little rationale for the selection of the model or intervention being tested, with tactics being chosen by investigator preference rather than by a systematic process of building from previous research.39 ,40 Second, the evaluation of the model or intervention is often a secondary consideration, not the primary objective, of the investigation. As a result, the implementation of the model or intervention was infrequently monitored, making it difficult to decipher whether the results were attributable to the model or intervention. For example, in two studies, the results of patient needs interviews/assessments were provided to primary care providers who were encouraged to participate in patient care, but their involvement was neither enforced nor documented.22 ,25 Consequently, there is a haphazard progression in inquiry such that the field is not advancing in logical sequence as might have been seen more often in clinical investigations (eg, evaluation of new chemotherapy agents). Researchers and practitioners in this field are encouraged to pursue inquiries based on theoretical41 ,42 and evidence-based rationale to build on previously published work.43 Planning for monitoring of implementation would further our knowledge of the effectiveness and feasibility of selected models or interventions in practice.
This synthesis of research also pointed to a number of methodological concerns in this field. With only six RCTs meeting eligibility criteria,22–27 it is evident that non-experimental pilot and feasibility studies are dominating the field. However, the included studies are of low to moderate quality; common methodological flaws include small sample sizes and inadequate statistical power,24 ,25 a lack of baseline data collected to examine changes over time,22 ,25 ,26 and a lack of clarity about the statistical significance of the results and their importance from a clinical and health system perspective.25 Moreover, very few studies investigated any one model or intervention type, making it challenging for a body of knowledge to be accumulated. Compounding this issue is when studies do investigate the same model design or intervention type, yet study authors use different nomenclature. For example, the concept of ‘nurse navigation’ in Wagner et al26 is referred to as ‘case management’ in two other RCTs,25 ,27 although the model appears to be essentially identical. Even when several studies investigate similar intervention types, an array of provider, patient and system outcomes are assessed, and the quality of the measurement of these varied outcomes is often questionable. By way of illustration, patients’ needs/satisfaction with care is best defined by how it is experienced by patients;23–25 however, studies often examine practitioners’ evaluation of care and extrapolate the outcome as patient satisfaction.27 Similar methodological flaws and the heterogeneity in outcome measures were noted in the included systematic reviews.17–21
The primary potential limitation of this systematic review was that all relevant literature may not have been captured and/or reported. For example, a number of terms (eg, ‘care coordination’, ‘continuity of care’ and ‘care integration’) exist for similar concepts in this field, and are often interchangeable. To mitigate this limitation, we used a rigorous, systematic methodology12 and our search strategy included a number of search terms encompassing the variety in nomenclature (see online supplementary material).
Overall, researchers, clinicians and administrators are left with imprecise results that lack both generalisability as well as definitiveness, providing limited data to build better interventions. Therefore, the development of provincial or national policies based on a strong evidence base remains unlikely. The most robust conclusion that can be made from this systematic review is that there has been little progress in this field. The ageing population, combined with an overall greater life expectancy for those living with cancer, will lead to an increased burden on the healthcare system. The Canadian healthcare system can neither afford nor has the human resource capacity to continue with business as usual. Ongoing demands will be placed on primary care and oncology care providers, without adequate evidence to direct the most suitable model designs (for the most appropriate patients and contexts), and without enough support to optimise the collaboration between these healthcare providers. High quality and adequately powered prospective experimental designs in this field are required to optimise patient experience, provider satisfaction and system performance.
The authors would like to thank: Julie Makarski and Tricia Waldron for their support with study planning and design; Kate Kerkvliet, Keneisha Gardiner and Heather Kilgour for assistance with acquisition of data, and data entry and analysis; and Sharon Matthias, Margaret Tompson and Richard Wassersug for providing the patient perspective to the project.
Contributors JRT, MCB and MV conceived the design of the study, led the implementation, as well as the drafting and modification of this paper. All authors contributed to the analysis and interpretation of the data, reviewed drafts and provided important intellectual contributions during the revisions, and approved of this final draft. All authors can attest to the accuracy and integrity of the work.
Funding The CanIMPACT project is funded by the Canadian Institutes of Health Research (reference number: 128272).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/