Why and how to use mixed methods in primary health care research

Introduction

Mixed methods (MM) are increasingly popular in primary care research (1). It consists of using both qualitative (QUAL) and quantitative (QUAN) methods and integrating them to study complex phenomena. In MM, integration of QUAL and QUAN is done at some levels and stages of the research process (research questions, methodological approaches, designs, procedures, results and interpretation) (2,3). However, planning, and conducting MM studies, as well as training graduate students in family medicine is challenging for clinicians and academics. Several MM designs and integration strategies have been proposed. However, there is a need for practical guidance on MM in primary health care (MM-PHC) research in two areas: (i) why and how to design MM-PHC research? (ii) How to integrate QUAL and QUAN methods? Addressing this need will contribute to upskill primary care clinicians and researchers to conduct MM research.

Based on a literature review (2,4) and our experience in teaching and performing MM-PHC research, we first explain the rationale for using MM-PHC and then propose a twofold practical guide on (i) how to design an MM-PHC study and (ii) integrate QUAL and QUAN methods. In addition, this article provides links and references to freely accessible resources and illustrates three common MM designs with real field studies that are conducted by graduate students.

Why to use mixed methods in primary care research?

Primary care research is complex and multifaceted (5,6), and MM research has been advocated to provide new insights into this complexity (7). Integrating both QUAN and QUAL methods can optimize the breadth and depth of a study and help taking into account the socio-cultural context and the real-world environment (8). In MM, the data collected are more comprehensive and provide a more complete understanding of the problem and potential solutions. In a review of 232 MM studies (9), 16 reasons for conducting MM have been identified including (i) enhance or build upon QUAL findings with QUAN findings, and vice versa; (ii) provide a comprehensive understanding of a phenomenon (e.g. variables and viewpoints); (iii) triangulate results; (iv) combine diverse viewpoints; (v) facilitate the sampling (e.g. using a survey to select interview participants); and (vi) develop and test instruments (e.g. develop a questionnaire using focus groups).

What are the types of mixed method designs?

Several MM designs have been developed to guide researchers on the integration of the QUAL and QUAN methods. The three common types are convergent, sequential exploratory and sequential explanatory (Fig. 1).

Figure 1.

Common mixed method designs.

Convergent MM design

This design combines the QUAL and QUAN methods during data collection and analysis steps. The QUAL and QUAN methods are usually (but not necessarily) concomitant. The results of the QUAL and QUAN are compared or combined.

Sequential MM design

This design involves two phases where one method (either QUAL or QUAN) is used first, and its results inform the other. The two subtypes of sequential designs are as follows:

• (i) Sequential exploratory: QUAN method is informed by results obtained using a QUAL method. For example, QUAL results are first obtained, and then the QUAN methods are used to statistically generalize the QUAL results.

• (ii) Sequential explanatory: QUAL method is informed by results obtained using a QUAN method. For example, QUAN results are first obtained and then QUAL methods, and results are used to interpret the QUAN results.

How to choose an appropriate study design?

Choosing an appropriate study design requires logical and purposeful planning. Two key decisions can help for planning an MM design (10):

• (i) Timing of the QUAL and QUAN methods

• Determine when QUAN and QUAL data will be collected. For example, in a sequential study design, researchers inevitably choose to start by collecting and analysing QUAN or QUAL data first.

• (ii) Level of integration of the QUAL and QUAN methods

• Determine whether the mixing will occur during data collection, data analysis or during interpretation. Integration—the explicit interrelation of the QUAN and QUAL methods (11)—is crucial in MM. Three main strategies of integration can be used: connection of phases, comparison of results and assimilation of data (Table 1) (4).

Table 1.

Common types and strategies of integration used in mixed method research

Types of integration	Integration strategies
1.Connection of phases	1.1.Connecting the results of the QUAL phase to data collection of the QUAN phase
	1.2.Connecting the results of the QUAN phase to data collection of the QUAL phase
	1.3.Following a thread
2.Comparison of results	2.1.Comparing QUAL and QUAN results obtained from separate data collection and analysis
	2.2.Comparing QUAL and QUAN results obtained from interdependent data collection and analysis
	2.3.Comparing divergences of QUAL and QUAN results
3.Assimilation of data	3.1.Transforming QUAL data into QUAN data (quantitizing)
	3.2.Transforming QUAN data into QUAL data (qualitizing)
	3.3.Merging QUAL and QUAN data

Adapted from Pluye et al. (4).

Three practical examples: the pan-Canadian research program on dementia care

Major neurocognitive disorders (dementia) are characterized by progressive cognitive decline, leading to increased morbidity and mortality (12). It is the seventh leading cause of death globally (13). As part of the Canadian Consortium on Neurodegeneration in Aging (14), we developed a research program combining several MM studies.

We will present three practical examples: (i) a convergent design study evaluating innovative collaborative care models (collCMs) for persons living with dementia (PWD); (ii) a sequential exploratory study of inappropriate acute hospital use in PWD; and (iii) a sequential explanatory study aiming to understand the various reasons why case management does not address PWDs’ needs.

Convergent mixed method design

CollCMs are complex interventions providing patient-centred, comprehensive, continuous and interprofessional care to PWDs. They are difficult to implement. Evaluation of these complex interventions necessitate going beyond evaluation of the impact on outcomes. It requires the assessment of characteristics of the models, as well as their implementation dynamics (15–17). Our objectives are:

1. To identify the facilitators and barriers for the successful implementation of collCMs in Ontario and Quebec;

2. To determine the association between key factors (organizational, individual and clinical characteristics) and quality of follow-up care; and

3. To understand why some key factors are associated with better quality of follow-up care based on the implementation strategy put in place.

A convergent MM design involving 22 family medicine groups (FMGs) is used. QUAN and QUAL methods are conducted in parallel and then integrated.

QUAL methods (objective 1)

We are using a multiple case study approach. We selected cases (FMGs) using a purposeful maximum variation sampling method based on type of collCM and rural/urban location. We now analyse data from documents, interviews and focus groups with managers, clinicians and patients-caregivers.

QUAN methods (objective 2)

We are performing a cross-sectional observational study using chart reviews and surveys on the same FMGs (35 charts per FMG). The primary outcome is a quality of dementia follow-up score.

Integration

Comparison of data (strategy 2.1, Table 1): We will merge QUAL and QUAN data to jointly review both data types using a matrix (18), with columns representing sites and rows representing findings, both QUAN data (outcome) and QUAL data (case summaries of facilitators and barriers). It will allow drawing conclusions on the link between implementation strategies, models of care used and quality of dementia follow-up. In this study, using both QUAN and QUAL data will provide a better understanding of the implementation strategies put in place (as identified in the QUAL data) and the factors associated with better quality of follow-up care (as measured using QUAN methods).

Sequential exploratory design

Evaluation of primary health care performance requires accurate indicators. To date, accurate indicators are lacking to measure potentially inappropriate acute hospital use in PWDs. We will use an MM sequential exploratory design to develop accurate indicators and measure the trends of inappropriate acute hospital use over the last decade in PWDs using administrative databases in Quebec and Ontario.

Phase 1: QUAL methods

To better understand the phenomenon of inappropriate acute hospital use, we will conduct a qualitative descriptive study using interviews, focus groups and thematic analyses among four groups of stakeholders: decision-makers, clinicians, PWDs-caregivers and researchers. It will allow us to identify different themes on inappropriate acute hospital use and their determinants.

Phase 2: QUAN methods

To measure trends over time of inappropriate acute hospital use, the themes derived from Phase 1 will be converted into indicators and operationalized, into the provincial administrative databases. Repeated, yearly cohorts of adults, aged 65 years and older, in Ontario and Quebec with a new diagnosis of dementia between 2002 and 2014 will be used. Administrative data on inappropriate hospital acute use for PWDs will be collected on each subject for 1 year following their index date of diagnosis of dementia.

Integration

Connection of phases (strategy 1.1, Table 1): We will analyse QUAL data and select significant QUAL findings (Phase 1) that will be used in the QUAN data collection (Phase 2). In this study, because of the paucity of the literature on this topic, the first QUAL phase is essential to inform the selection of indicators to measure potentially inappropriate acute hospital use in PWDs that will be used for the second QUAN phase.

Sequential explanatory design

The Quebec Alzheimer Plan introduced case managers in FMGs. Often nurses and social workers help to empower PWDs and their informal caregivers in their own decision-making, in addition to coordinating their care. Though case management may be an effective approach to dementia care (19), little is known about how this model of care handles the cultural diversity of patients. Given the rich ethnic landscape of Quebec, Canada, we want to assess whether cultural tensions and conflicts exist in the complex and challenging environment of dementia care.

To face this gap, our research questions are as follows: (i) What is the number of met and unmet needs across different ethnic groups of PWDs and their informal caregivers under case management? (ii) What are the PWDs and caregivers’ accounts of their experiences with the illness through case management?

We will use a sequential explanatory MM design. We will first perform a quantitative phase to measure the number of met and unmet needs of PWDs and caregivers. Afterwards, we will perform a qualitative phase to explain the quantitative results.

Phase 1: QUAN methods

We will conduct a cross-sectional clinical study in four following steps: First, we will recruit 180 pairs of PWDs and their main informal caregiver across six FMGs in Quebec where case management is implemented. Second, we will collect socio-demographic data and pay special attention to ethnicity, language, immigration history and spiritual beliefs. Third, we will use two validated questionnaires to assess the met and unmet needs of PWDs and their caregivers: (i) the Camberwell Assessment of Need for Elderly for the needs of PWDs and (ii) Carers’ Needs Assessment for Dementia for the needs of caregivers. Fourth, we will compare questionnaire scores across different ethnic groups using multiple linear regression.

Phase 2: QUAL methods

We will then conduct a qualitative descriptive study to understand PWDs’ and caregivers’ accounts of their experiences (who, what and where) and to explain the quantitative results. We will purposively sample 20 pairs of PWDs-caregivers who took part in Phase 1 based on ethnic groups and the level of their met and unmet needs. We will conduct individual semi-structured interviews and apply inductive thematic analysis (20).

Integration

This will occur twice, between QUAN and QUAL phases and at the end of the study. The QUAN results will inform the sampling of the QUAL study (connection of phases, strategy 1.2, Table 1). Then, the QUAL results will serve to enrich, justify or challenge QUAN results (comparison of results, strategy 2.2, Table 1). They will also explain the results on the number of met and unmet needs and their variation across ethnic groups as measured in the first QUAN phase.

Conclusion

The use of MM is growingly popular in primary care research. This article has described three main designs and integration strategies of QUAL and QUAN methods by using practical examples. While we have completed an extensive number of MM research (e.g. (19,21–24)), the examples provided in this article are from our current research program on dementia and mainly combine QUAL studies with QUAN cross-sectional studies and administrative databases. Several other combinations of QUAL and QUAN studies can be found in MM (4) such as merging a QUAL study with a randomized controlled trial to better understand how and why a program works or does not work (25). It can serve as a basis for those engaged in PHC research to gain better understanding of why and how to use MM.

If you are interested in learning more on MM research, here are some freely accessible resources:

http://www.mixedmethods.org/resources.html

https://study.sagepub.com/creswell3e

https://obssr.od.nih.gov/training/mixed-methods-research/

http://mmira.wildapricot.org/

http://methodesmixtesfrancophonie.pbworks.com/ (French)

Declaration

Funding: Canadian Institutes of Health Research; the Canadian Consortium on Neurodegeneration in Aging (CCNA); or the Fonds de recherche du Québec–Santé (FRQS).

Ethical approval: ethical boards of the West-Central Montreal Health Center Research Review Office; the Institute for Clinical Evaluative Sciences (ICES); or the Institut national de santé publique du Québec (INSPQ).

Conflict of interest: none.