Mapping Evidence and Constructing an Analytic Framework: An Umbrella Review Summarizing Primary Care Family Physicians’ Experiences With Clinical Integration

Abstract

PURPOSE

Strengthening primary care’s integration function is a systematic approach to promote integrated care. Understanding the factors influencing the process of integrating care for every patient is crucial for effective intervention planning. The objective of this study was to generate an analytic framework and evidence map of the barriers and facilitators perceived by family physicians (FPs) in clinical integration, a process to coordinate health care services across time, place, discipline, diseases, and patient demographics.

METHODS

Using the Joanna Briggs Institute umbrella review methodology, we searched the MEDLINE, Embase, and CINAHL databases, identifying 90 reviews (2010-2022) on primary care FPs and clinical integration. We adopted a best-fit framework approach to group the factors into a customized clinical integration framework, reflecting how a health care system functions. Two evidence maps were created to visualize the reviews’ distribution. We validated the framework with another 21 reviews (2022-2024).

RESULTS

The analytic framework consisted of 9 themes and 21 subthemes based on 2,891 factors derived from external and internal sources within primary care practices. Several subthemes were common across themes related to individuals (FPs, physicians other than family physicians and allied health providers, patients) and operating units (systems, organizations, practices), highlighting shared elements. The professional theme was the most significant, appearing in 86% of the reviews and including subthemes related to diseases, clinical guidelines, and teamwork. In contrast, themes related to systems, organizations, and practices were reported less frequently (48%, 22%, and 23%).

CONCLUSION

The complex interactions among factors, subthemes, and themes elucidate challenges in finding a universal strategy or implementing initiatives. The generated evidence maps indicated knowledge gaps to guide future research work.

INTRODUCTION

Care integration involves coordinating comprehensive, ongoing health care services across providers and settings to meet patient care needs.^1,2 Achieving integrated care is essential to realizing patient- or person-centered care^3,4 and can potentially advance the Quintuple Aims to promote care quality, improve care access, increase patient and provider satisfaction, and reduce health care costs.^5-7 Globally, governments have aimed to deliver integrated care for the past 2 decades.^8-10 However, interventions with a narrow disease focus can unintentionally increase the system’s complexity.¹¹ Thus, systematic approaches are favored for reducing system fragmentation.

Strengthening the integration of primary care is a systematic approach endorsed by international agencies, including the World Health Organization,¹ the National Health Service (NHS) Federation,¹² the Scottish government,¹³ and the Institute of Medicine (IOM).¹⁴ This approach leverages the commonality shared by integrated care and primary care: delivering patient-centered care that is continuous, coordinated, and comprehensive.^1,2,15,16 Continuous care refers to longitudinal care over time, coordination involves synchronizing services from multiple providers deliberately, and comprehensive care addresses each patient’s biomedical, social, and psychological needs.

The first step to developing health care intervention(s) is understanding the problem and its context.^17,18 To strengthen primary care’s integration, it is salient to understand the extent of, and interactions among, factors influencing the process of integrating care for each patient across providers and settings (aka clinical/service integration). Such a high- and system-level picture remains lacking, while many pieces of information have been reported in reviews with narrow focuses. During the study planning stage, we customized a theoretical framework to synthesize interrelated factors into themes that represent potential intervention targets, such as provider groups. Our overarching goal is to generate a comprehensive map to guide future intervention development.

Aim

This qualitative systematic review aimed to create comprehensive evidence maps and an analytic framework of family physician (FP)–perceived factors that influence clinical integration to guide future studies.

METHODS

This umbrella review included systematic review studies with qualitative data describing FPs’ experiences with clinical care integration. We selected reviews rather than single primary studies, as we anticipated a high volume of studies under this broad topic. The team developed and published the systematic review protocol¹⁹ with the guidance of the published Joanna Briggs Institute (JBI) systematic review methodology framework.²⁰ Two reviewers independently selected reviews, appraised the quality of the included ones, and extracted and synthesized the data. Any disagreements throughout these stages were resolved through discussions between reviewers or by involving a third reviewer.

Eligibility Criteria

We included reviews that met the following criteria using the Population-Concept-Context (PCC) framework²¹ (Supplemental Table 1):

• Participants are primary care FPs in community or office settings. Primary care FPs refer to physicians who have completed training in the family medicine specialty¹⁵ and who provide holistic care based on long-term patient-physician relationships to address patient health care needs in the community.²² We refined this criterion because some identified articles did not specify participants and/or settings. Articles were considered eligible if most participants were FPs (≥50%), if the setting was primary care, or both.

• Concept refers to experiences of coordinating health care services across time, place, and discipline for patients. The process involves service determination (aka disease management, care planning, treatment decisions), referrals connecting patients to the required services, and communication among all involved services.

• Context is qualitative or mixed reviews published in English in 2010-2022. The review can be systematic, scoping, or narrative to ensure we capture a broader range of factors.

Review Search

An information specialist built the search strategies, tested them against 13 review articles from other sources, and revised them with feedback from the first author (O.L.T.), who randomly screened the titles. The absence of a universal definition of clinical integration posed a challenge in identifying relevant articles.²³ To strengthen the search strategy, team members iteratively discussed keywords according to the inclusion criteria and 8 taxonomies describing integrated primary care.²⁴ The final search strategies included MeSH terms and keywords related to 6 concepts: (Physician) AND (Primary Care) AND (Attitudes) AND (Qualitative) AND (Reviews) AND (limits (2010-2021) and (English)) (Supplemental Table 2). The information specialist executed the search in March 2021, which was rerun in April 2022. The databases included MEDLINE, CINHAL, and Embase. We excluded the Cochrane Central Register of Controlled Trials (CENTRAL) database because evaluating intervention effectiveness was not within the scope of this review. Reviewers manually searched the references of all included reviews. The first author (O.L.T.) imported the identified articles into the Covidence platform (Veritas Health Innovation Ltd) and removed duplicates.

Review Selection

Reviewers screened 1,861 unduplicated articles in 2 stages: title and abstract, and full text. We documented the reasons for exclusion at the full-text stage. A total of 90 reviews met the inclusion criteria. The selection process was summarized in a Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) diagram (Supplemental Figure 1).

Critical Appraisal

We appraised the quality of the included reviews using the JBI critical appraisal tool for systematic reviews.^20,25 Reviewers independently assessed each review against 11 questions, examining methodological strengths and potential bias in its design, conduct, and analysis (Supplemental Table 3). Responses to each question were categorized as meeting (yes), partially meeting (unclear), not meeting (no), or not applicable to the evaluation criteria. Reviewers assigned 1 point for each “yes” response and zero points for other responses. Points were summed, classifying reviews as high (9-11 points), medium (7-8 points), or low (≤6 points) quality. We conducted the quality assessment to describe review quality and did not establish a threshold for inclusion criteria.

Data Extraction

We extracted 2 sets of data from the included reviews: review characteristics and qualitative data. Review characteristics, including study identification (ID), title, author, year of publication, number of studies, setting, and participant types, were extracted into a Covidence data extraction template. The qualitative data was extracted into an Excel (Microsoft Corp) worksheet due to their heterogeneity in terms of data formats (text, tables, and figures). The qualitative data included FP-perceived factors and their associated (original) themes and subthemes. A third reviewer (O.L.T.) compared the extracted data from the reviewers to minimize data entry errors and generated the final data sheet for data mapping and analysis.

Data Synthesis

We adopted a “best-fit framework approach” due to its high level of structures and capacity to manage large quantities of contextual data about participants’ experiences,^26,27 as well as its ability to guide the development of an analytic framework.²⁸ Reviewers synthesized the data using the approach outlined in the following 3 stages:

Stage 1. Framework Development

During protocol development,¹⁹ the team customized a “clinical integration framework” consisting of 8 a priori themes to capture factors outside and within a primary care practice.

Stage 2. Data Mapping

Reviewers mapped the extracted FP-perceived factors (codes) onto the “clinical integration framework.” Considering the large size and heterogeneity of the data, 3 reviewers (O.L.T., W.Y.H., G.K.), medical professionals differing in specialty and career stage, experimentally and individually coded 5% to 10% of the data from select reviews focusing on various topics. Disagreements were resolved through discussion among the reviewer team, with persistent disagreements presented to the entire team via monthly video conferencing from January to June 2022. After each monthly meeting, previous codes were adjusted based on feedback, and another 5% to 10% of the data was coded. Subsequently, 3 reviewers (O.L.T., G.K., Celine Young) coded the data by repeating a 5-step cycle.

1. Independently reviewed the extracted codes with their associated, original themes/subthemes

2. Reassigned each code to 1 of the 8 a priori themes that was the best fit and as close as possible to their original ones

3. Inductively developed descriptive subthemes to group similar codes within each theme

4. Compared their coding and reached consensus by discussion

5. Re-examined their earlier work and adjusted the coding as needed

In the end, we assigned each code (extracted factor, level 3) to a subtheme (level 2) and a theme (level 1)

Stage 3. Data Analysis

We initially created 8-20 subthemes within each theme. To reduce the complexity of the data analysis, we reduced the number of subthemes by merging those that were overlapping or similar. For example, we merged several types of resources, captured by 6 subthemes, into a new subtheme, “resource availability and distribution.”

After completing the data coding, we assembled themes, subthemes and factors into an analytic framework (Figures 1A and 1B). We calculated the number and percentage of reviews for each subtheme and theme, stratified by 3 variables: focus, quality, and outcome type of the included reviews. The outcome types include barriers, facilitators, and factors. We then generated 2 evidence maps to visually present the percentage distribution by theme (Figure 2) and subtheme (Figure 3).

Figure 1A.

Analytic Framework of Themes, Subthemes, and Example Codes (Factors)

^aCodes extracted or generated from the updated search conducted in October 2024.

^bOther health care providers include physicians other than family physicians and allied health care providers.

Figure 1B.

Analytic Framework of Themes, Subthemes, and Example Codes (Factors)

^aCodes extracted or generated from the updated search conducted in October 2024.

^bOther health care providers include physicians other than family physicians and allied health care providers.

Figure 2.

Evidence Map by Theme

FP = family physician.

Note: Each bubble represents the percentage of the 90 included reviews that reported factors related to each theme, stratified by 3 variables: focus, quality, and outcome type of the included reviews. The outcome types include barriers, facilitators, and factors. Reviews were counted more than once if they reported a combination of factors, barriers, and/or facilitators.

Figure 3.

Evidence Map by Subtheme

FP = family physician.

Note: Each bubble represents the percentage of the 90 included reviews that reported factors related to each theme, stratified by 3 variables: focus, quality, and outcome type of the included reviews. The outcome types include barriers, facilitators, and factors. Reviews were counted more than once if they reported a combination of factors, barriers, and/or facilitators.

Framework Validation

We conducted an updated search on October 16, 2024, which yielded 638 unduplicated articles after removing 59 duplicates. The first author (O.L.T.) screened the articles, and a total of 21 reviews met the inclusion criteria. We used this new data to validate the analysis framework (Figure 3).

RESULTS

Characteristics of the Included Reviews

We included 90 eligible reviews. Of them, 85 reviews consisted of 1,246 primary qualitative studies (Table 1, Supplemental Table 4). Seventy-six reviews (84%) reported findings from multiple countries. Fifty-five reviews (61%) reported both qualitative and quantitative data. Sixty-seven reviews (74%) focused on primary care and general practice settings. Most reviews focused on disease management and care planning (54 reviews, 60%), followed by care integration across sectors or providers (20 reviews, 22%), such as primary-secondary or FP-oncologist care. The critical appraisal score ranged from 6 to 10, with the maximum score being 11. All reviews lost 1 point in publication bias, which was expected. The quality was high for 50 reviews (56%), medium for 37 reviews (41%) and low for the remaining 3 reviews (3%) (Supplemental Table 4).

Table 1.

Characteristics of the Included Studies

	Reviews for analysis		Reviews for validationa
	No. (%)	%	Number	%
Year of publication
2010-2014	16 (18)		0 (0)	0
2015-2019	43 (48)	48	0 (0)	0
2020-2022	31 (34)	34	8 (38)	38
2023-2024	0 (0)	0	13 (62)	62
Countries
Multiple	76 (84)	84	15 (71)	71
United Kingdom	5 (6)	6	2 (10)	10
United States	2 (2)	2	2 (10)	10
Australia	1 (1)	1	1 (5)	5
Germany	1 (1)	1	0 (0)	0
New Zealand	0 (0)	0	1 (5)	5
Unknown	5	6	0 (0)	0
Type of data
Mixed	55 (61)	61	14 (67)	67
Qualitative only	35 (39)	39	7 (33)	33
Settings
Primary care and general practice	67 (74)	74	19 (90)	90
Primary health care	8 (9)	9	0 (0)	0
Primary—secondary care	5 (6)	6	1 (5)	5
Primary care—public health	1 (1)	1	0 (0)	0
Unknown or mixed	9 (10)	10	1 (5)	5
Focus of the review
Disease management and care planning	54 (60)	60	12 (57)	57
Care integration across sectors or providers	20 (22)	22	3 (14)	14
Disease diagnosis and testing	9 (10)	10	1 (5)	5
Disease prevention and screening	2 (2)	2	3 (14)	14
Follow-up care	2 (2)	2	2 (9)	9
Othersb	3 (3)	3	0 (0)	0

^aReviews were identified from the updated search conducted in November 2024.

^bFitness to work, return to work.

Among the 21 reviews identified from the updated search, most reviews focused on disease management and care planning (12 reviews, 57%), followed by care integration across sectors or providers (3 reviews, 14%) (Table 1).

Themes and Subthemes

The team categorized a total of 2,891 codes into 9 themes and 21 subthemes. We began with 8 a priori themes, which were expanded to 9 themes during the data synthesis stage. This expansion primarily occurred due to various factors concerning FPs and non-FP providers. To distinguish FPs’ roles within practices from those of non-FP providers within the system, we renamed the provider theme as “FP” within the People-Provide-Practice (3-P) model, mirroring primary care practices, and added a provider theme to the Rainbow Model of Integrated Care (RMIC) model, representing the system. Definitions of the 9 themes were summarized in Supplemental Table 5. The type of data reported by each review was documented in Supplemental Table 6.

We inductively developed a total of 21 inter-related subthemes among the 9 themes. Of those subthemes, 3 were shared among system, organization, and practice themes, as they were all operating units despite size differences. Similarly, another 3 subthemes were shared among patient, FP, and provider themes as they are all individuals in specific roles. Definition of subthemes and example codes (factors) were summarized in Supplemental Table 7.

Analytic Framework and Validation

We organized themes, subthemes, and example codes into our customized clinical integration framework (Figures 1A and 1B). Family physicians perceived their daily work as influenced by various factors originating both externally and internally within their primary care practices. The 9 themes featured in the operation units (system, organization, and practice); individuals (providers, FPs, and patients); medicine and professional work (professional); beliefs and attitudes; stigma and culture (normative); and administrative management, finance, and communication (functional). Normative and functional themes have the potential to impact 1 or more of the remaining themes, which are hierarchically layered within a health care system. We used the following example factors to highlight several subthemes that directly influence care integration.

One “teamwork and collaboration” factor (professional theme) was the model of care. A shared care model with patient at the center facilitated care integration,^29-32 in contrast to acting as a barrier when providers were simply placed together without mutual respect and knowledge.²⁹

One financial factor was the availability of funding (functional theme). The lack of payments could hinder collaboration among providers²⁹ or the integration of pharmacists into general practices.³³ Funding became a facilitator when it was sufficient to support collaboration across sectors of primary care and public health,³³ promote shared care among providers,²⁹ or enable the inclusion of general practice pharmacists.³⁴

Communication technology factors included interoperability and standardization (functional theme), both of which impacted care integration through patient data sharing among providers or across sectors. Standardized data served as a facilitator for data sharing,³³ while the lack of standardization acted as a barrier.^35,36 Interoperability also posed a barrier to data sharing when data was stored in separate or unlinked databases.^35-37

We used the data from the 21 reviews identified from the updated search to validate the framework. Most factors were still related to FPs, providers, and patients. There were no changes to the subthemes and themes, except for the addition of 2 level-3 factors under the functional theme of the analytic framework (Figure 3): payment type (finance subtheme) and data management (communication, technology subtheme).^38,39

Evidence Maps

At the theme level, the professional theme was the most significant, appearing in 86% of the reviews and including subthemes related to diseases, roles and identities, clinical guidelines, and teamwork (Figure 2). System, organization, and practice-related themes were less frequently reported (48%, 22%, and 23%) to describe their characteristics, resources, and service planning.

Four subthemes stood out for their popularity: FPs’ characteristics, knowledge, and experience (78%); patients’ characteristics, knowledge, and experience (70%); beliefs and attitudes (62%); and roles and identities (60%).

DISCUSSION

To the best of our knowledge, this review is the first to comprehensively summarize FP-perceived factors influencing care integration from various sources and levels within a health care system. The framework, built upon almost 3,000 factors from 90 reviews consisting of over 1,200 primary qualitative studies, illustrated the complex interactions within a health care system. Evidence gaps existed at both theme and subtheme levels. Organizations and practices remained underevaluated for their influence on care integration. Within both themes, the subthemes of “service planning and integration” and “resource availability and distribution” were overlooked by most included reviews.

Our report serves as an umbrella review, a systemic review of systematic reviews on a broad topic. It is distinguished by its generalizability and capacity to inform evidence-based intervention planning. We cast a wide net to capture factors regardless of patient populations, target diseases, and services, thus ensuring the relevance of our findings to most primary care settings. Additionally, our “best-fit framework” approach involved a mixed deductive-inductive process, leveraging both thematic and framework analyses. This approach allowed us to organize factors into a framework comprising common themes that serve as potential intervention targets. Lastly, our report underscored the significance of common factors across multiple themes or domains. For instance, service planning is an example that can occur at the practice, organization, or system level. By targeting these common factors, primary care’s care integration functions can potentially be systematically strengthened.

As expected, we noted a lower number of studies reporting on practice- or organization-level factors for 2 potential reasons. The primary reason is our data synthesis rules. We grouped certain functional factors at the practice, organization, and system levels into a specific functional theme (ie, staff and IT infrastructure), which resulted in fewer factors remaining in the practice or organization themes. For example, staff factors were captured by 4 reviews that did not report any factors related to practice, organization, or system. The other reason is the possible limitation of our search strategy in identifying all relevant articles due to the lack of a universal definition of care integration.

Surprisingly, FPs often found themselves dealing with barriers, facilitators, and factors associated with 3 distinct subthemes: beliefs and attitudes, roles and identities, and disease. Firstly, the notable frequency of the “beliefs and attitudes” subtheme may result from 2 causes: the promotion of shared decision making and the increasing trend of seeking health information online. The former adds complexity to medical decision making, as FPs must integrate patient perspectives.⁴⁰ Meanwhile, the latter has the potential to shape FPs’ and patients’ beliefs, subsequently influencing decision-making processes. Second, the high popularity of the “roles and identities” subtheme likely coincided with the advent of team-based care, which commenced in the early 2000s in Canada. Close collaboration among FPs and other providers within a team necessitates a clear delineation of responsibilities for each member. Unclear definitions can lead to role conflicts, impeding the effectiveness of team-based care.^41-43 Third, the “disease” subtheme reflects the ongoing challenges in health care. Despite advances in medicine, research evidence is limited for diseases that do not have clear definitions,^44-47 cannot be objectively diagnosed with blood tests,^44,48 or lack effective treatment options.^49-53 Family physicians often encounter challenges when trying to diagnose or manage these illnesses promptly. To summarize, the 3 unexpected subthemes could be FPs’ pain points in their daily practice, which warrants further research studies.

Our original intention was to report factors, including both barriers and facilitators. The reviews included in this study, however, did not consistently or clearly distinguish between the 2. This issue affected 51% of the extracted codes, which were categorized as “factors” in this study. During the synthesis stage, we observed that most factors are context dependent and can function as either barriers or facilitators. For example, factors related to roles and identities, as discussed in the previous paragraph, illustrate this point. As a result, the analytic framework was built around these factors.

Similar to this study, most of the reviews included in the validation process focused on disease management and care planning (12 reviews, 57%), followed by care integration across sectors or providers (3 reviews, 14%) (Table 1). Unsurprisingly, most factors were still related to FPs, providers, and patients. We used these new data to validate the analysis framework (Figure 3). There were no changes to the subthemes and themes, except for the addition of 2 level-3 factors under the functional theme of the analytic framework (Figure 3): payment type (finance subtheme) and data management (communication, technology subtheme). This can be attributed to the increasing complexity of provider payments and the rapid growth of communication technology.

Limitation

Future Studies

Governments and jurisdictions have embarked on primary care reforms to promote integrated care. The reform in Canada has focused on physician payments and team-based primary care since 1998.⁵⁴ Team-based care reduced emergency department use in Quebec and Alberta, but the pay-for-performance model in Ontario had no effect on preventive care.⁵⁵ Despite 2 decades of reform initiatives, Ontario FPs still expressed the need in 2018 to fundamentally change primary care delivery.⁵⁶ Family physicians’ burnout rates continued to rise from 2017 to 2021 due to increased workload,⁵⁷ which further worsened during the subsequent pandemic era.⁵⁸ The Canadian experience explained the challenges inherent in promoting care integration, echoing the research evidence: integrated care initiatives rarely lead to clearly positive effects.⁵ Local context matters, as strategies effective in one jurisdiction may not be readily translated to another. Building upon our findings, the team intends to survey local health care providers to identify potential intervention targets. Moreover, additional research and strategies are needed to address 3 unexpected common subthemes: beliefs and attitudes, roles and identities, and diseases.

CONCLUSION

Despite years of dedicated effort, health care systems remain complex, siloed, and fragmented. Our evidence map and framework illustrate the complex interactions between factors, subthemes, and overarching themes, explaining the inherent challenges in finding a universally applicable strategy or implementing initiatives at any level. Clinical leaders and decision makers must bear these factors in mind when experimenting with interventions. Targeting common elements shared across themes can systematically strengthen care integration. Additional research and strategies are needed to effectively address 3 unexpectedly prevalent subthemes: beliefs and attitudes, roles and identities, and diseases.