In this paper, we aimed to (1) introduce LQR to family medicine researchers and clinicians; (2) highlight important steps, considerations, challenges and opportunities related to LQR; and (3) signpost to other literature that may be helpful for those new to LQR. We discuss key steps related to conducting LQR with particular emphasis on applications in family medicine and primary healthcare research. The steps presented are structured around key stages of any qualitative study, including planning and design, recruitment and data collection, and data analysis. We also discuss ethical considerations as cross-cutting these stages. As experienced qualitative researchers in primary care, we draw on wider applied healthcare research when discussing these steps as well as our experience of conducting qualitative research (including longitudinal design) while highlighting papers that have used longitudinal qualitative design in primary care in order to showcase the different approaches and the value of LQR in this particular setting. Table 1 provides more information on selected examples of longitudinal qualitative studies in family medicine and community health research. This is not an exhaustive list; rather, these were chosen by the authors to illustrate the diversity of LQR in primary care, including methodological approaches, aims and samples.
It is important to highlight that the steps and considerations presented in this paper are applicable to ‘experiential’ methodological approaches14 (such as interpretative phenomenological analysis (IPA), narrative analysis or thematic analysis) which focus on understanding people’s views and experiences rather than discursive approaches (such as discourse analysis and conversation analysis) which are concerned with how language is used to construct a particular version of a reality. In relation to methods of data collection, we focus here on steps relevant to interviews and focus groups as these are one of the most common methods in LQR15 and in primary care research as they allow exploration of people’s views and experiences.
Step 1: deciding on the length of data collection and timing and number of interviews
When employing LQR, researchers need to decide on three key inter-related aspects: (1) the length of the time needed for data collection, (2) timing and spacing of the interviews, and (3) number of interviews. Neale1 suggested thinking of the overall length of data collection as the time frame of a qualitative longitudinal study, and of the number, timing and spacing of the interviews as the study tempo. Together, time frame and tempo can be seen as a framework for designing and conducting LQR.
Length of data collection
One of the key considerations when designing LQR is to decide on the overall length of the data collection period. The time frames of published longitudinal studies in primary care vary, as they are (rightly) guided by the study focus and research question. For example, a recent study on experiences of primary care HCPs on implementing remote consultations during the COVID-19 pandemic focused on the first wave of the pandemic and thus collected data over a period of 4 months (April–July 2020)20 (see also table 1). However, some topics may not lend themselves to such clear cut-off points. For example, when studying patients’ recovery from a particular treatment or adjustment to a new diagnosis, for example, asthma, it may be difficult for researchers to decide for how long they should follow the participants. Being guided by clinical information (eg, the ‘usual’ recovery time) might be useful for researchers while being mindful that individual trajectories may differ. Also, stopping data collection sooner may not capture some of the aspects of recovery. Finally, researchers may want to consider other factors such as availability of resources and staff.
Timing and spacing of interviews
The timing of the interviews is also crucial, and researchers may want to consider three approaches.
Approach 1: data collection around researcher-led events
One approach to deciding the timing of the interviews may involve researchers trying to define ‘events’ which may act as important time points for data collection. These time points need to be decided in relation to the population and topic under study. When exploring patient experiences of the illness trajectory, the timing of the interviews may be based on the events linked to their journey, which, for example, may involve talking to patients shortly after being diagnosed or after they complete their treatment.16 17 One may also conduct interviews before and after the event, with the aim of understanding and comparing one’s expectations and experiences.18 19 For example, Gordon et al studied the process of transition from trainee to trained doctor and conducted interviews with participants before they graduated from their degree (thus studying their expectations of what it means to become a doctor) and after they obtained their degree.20 Similarly, Lester et al interviewed patients within 6 months of inception into the early intervention service and after being discharged to primary care.19 Neale1 suggested that having such clear events can be very helpful for establishing a clear baseline as well as a closure point for a qualitative longitudinal study and urged researchers to think carefully how the beginning and the end of the study will be defined.
Approach 2: data collection around participant-led events
Defining such events may not always be beneficial, and some highlighted the benefits of flexibility in deciding the timing of data collection1 7 and allowing for conducting interviews around, for example, unexpected events. This may mean that researchers would be guided by the participants, who would advise when they experience any significant events in relation to the phenomena of interest. This approach is known as the mirroring process,1 where data collection mirrors the events in participants’ lives. Consequently, researchers may be collecting data around events defined in the same way but which may not occur at the same intervals for all participants.21 However, others also highlighted the drawbacks of this approach, noting that in larger samples, it may be difficult to keep track of all the participants and conduct interviews around the key events.4
Approach 3: data collection based on pre-established, regular intervals
Deciding on the events may not always be possible. For example, Murray et al13 highlighted that defining key events in relation to patient experience requires a researcher to have an understanding of a ‘typical’ illness trajectory of a given condition. For certain conditions, these may be less well defined. In such situations, it may be useful to conduct interviews at regular intervals over a period of time to facilitate an in-depth understanding of issues during a particular period rather than around key events. Nissim et al, who studied the experiences of patients with advanced cancer with the focus on the desire for hastened death, largely adopted this approach by interviewing their participants at 2–4 month intervals.22 However, they also shortened these intervals in a number of scenarios including when participants started new treatments; self-reported measures indicated a change in their physical or psychosocial distress or patients’ condition began deteriorating. The study thus highlights the benefits of a flexible approach, combining data collection at regular intervals with participant-led events.
Number of interviews
The third aspect, the number of the interviews, will be partially framed by the two aspects discussed previously: the length of data collection and timing of the interviews. For example, if a researcher decides to collect data over a 1-year period, the significant events participants experience in their lives will somewhat indicate the number of interviews as well. However, it still leaves researchers scope to decide on the number of interviews. Neale et al1 23 suggested that one way of viewing LQR may be to see it on a spectrum from intensive to extensive. The most intensive approach may mean that the numerous data collection points can lead to almost blurred boundaries between time points, which have been referred to as a ‘description through time’.1 23
Regardless of the approach, it is crucial to plan the length of data collection, timing and frequency of the interviews to facilitate the primary aim of the LQR, which is to study change and continuity.11 Also, these three aspects will need to be guided not only by the research question but also by resources and (existing) expertise within the team. Having multidisciplinary teams can be beneficial as advice from both clinicians and patients24 on, for example, a typical clinical pathway or illness trajectory can be crucial in deciding on the timing and number of interviews and, ultimately, the success of the study.
Step 2: planning recruitment: attrition versus oversampling
Recruitment and sampling are important aspects of all qualitative research but can be particularly challenging in LQR. One of the key aspects is to decide on sample size, which may have numerous implications.
First, researchers need to strike a fine balance between sampling a sufficient number of participants and oversampling. While it is often recommended that researchers allow for sample attrition by recruiting more participants for the initial interviews than needed, in the studies where retention is high, this may lead to an excessively large sample and dataset. In this case, researchers may have to decide whether they want to follow up all the participants or a subset, taking the implications of that into account. For example, Calman et al described how initial oversampling of their participants (caregivers of cancer patients) led to a decision not to interview some participants at subsequent time points, which created tensions between researchers and participants.4 Equally, high levels of attrition may lead to bias in the sample and the subsequent study results.25 For example, Lester et al highlighted that they faced problems in accessing contact details of participants for the follow-up interview 3 years later and thus had relatively high attrition rate (33% of participants took part in the follow-up interview).19
Second, as in all qualitative research, sample size will influence the depth of the analysis. Smaller samples may allow a more in-depth understanding of individual experiences and lend themselves well to methodologies that value an idiographic approach (eg, phenomenology). For example, Smith conducted interviews with four women as case studies to develop an initial theory of transitions to motherhood. In contrast, larger sample sizes may allow, through their breadth, the identification of patterns and the influence of external factors shaping individuals’ experiences to be recognised.1 Neale1 also pointed out that a small number of participants do not necessarily mean a small dataset, given the number of times participants might be interviewed.
Step 3: approaching data collection: asking the same or different questions
One of the key considerations when conducting subsequent interviews in LQR is deciding on what questions will be asked at each time point. Holland et al3 suggested two approaches: one involves researchers asking the same set of questions at each time point, thus facilitating close mapping of the data at all time points. The second involves anchoring data collection on specific topics of interest which in turn may mean asking the same and/or different questions related to these topics. The reason for this is that some questions will only be relevant at particular time points. This will be especially relevant for researchers conducting interviews around key events, as described earlier. For example, in interviews with patients with limiting illness over 18 months, Worth et al covered the majority of the same topics at all time points (eg, patients’ needs) while also asking some questions only at particular time points (eg, about illness history at interview one only).26 In contrast, Lester et al used two different topic guides and asked different questions at each time point.19 Researchers need to consider whether and how participants may be encouraged to discuss change in their lives. Here, we outline three approaches for doing this. One approach may involve providing a participant with a summary of a previous interview at the beginning of each interview.27 This may be useful in providing a starting point for the subsequent interview and can be an example of member checking, an approach used to enhance trustworthiness of qualitative data.28 However, care must be taken when preparing such a summary as researchers could inadvertently impose their interpretations of the previous interview, thus distorting the participant’s story and affecting how they approach the subsequent interview. An approach facilitating a dialogue where a researcher summarises the previous interview(s) while inviting participants’ views on it might be more constructive. Such an approach can provide an opportunity for participant reflection and enhance analysis (see step 4). A second approach may involve focusing the summary on more factual events, thus acting as a reminder of when the last interview took place. A third approach may be to encourage participants to reflect on any changes they have witnessed in their lives. Researchers may want to ask participants directly whether they have experienced any changes in their lives, as well as what remained the same, and more importantly, how they feel about it. For example, Lawton et al used this approach when studying the experiences of patients with diabetes transitioning from specialist care to primary care. They encouraged participants to reflect on any changes related to their contact with diabetes services and HCPs since their last interview, as well as their understandings of why their service contact had changed over time.29
Step 4: planning and conducting the analysis, and writing-up findings
When planning and conducting analysis in LQR, the researchers should consider (1) aims and questions that guide the analysis, (2) which approach to the analysis and writing up is appropriate for the research questions and objectives, and (3) practicalities involved in the analysis.
Aims and questions guiding LQR analysis
As the main aim of LQR is to identify change and continuity over time in phenomena of interest, the analysis needs to focus on exploring and identifying how and why change occurs or not over the study period. The analysis may also aim to identify different types of change. Researchers may want to explore the types of change proposed by Lewis5: individual, service, policy and structural; narrative (ie, ‘unfolding of individual stories’); participant’s reinterpretations (ie, ‘rethinking or retelling of experiences described earlier’); and researcher’s reinterpretations (eg, of what the participant described earlier). It is also important to identify when and why change is absent, and what remains stable and consistent and why.
Asking different types of questions of the data can help guide the analysis. Researchers need to relate to the research aims and questions, explore the different types of change and strive for a comprehensive understanding of the dataset, including how the analytical categories and codes relate to each other. For example, Saldaña7 suggests using three types of questions to facilitate the analytical process: framing questions to capture the context and influences (eg, what contextual and intervening conditions appear to influence and affect participant changes through time?); descriptive questions to capture information to help answer questions (eg, what happens, increases or emerges through time?); and analytical and interpretive questions to integrate the descriptive and framing information (eg, which changes interrelate through time?).
Approaches to analysis
After deciding on research question/aims and the data collection methods, researchers need to select an appropriate analytical approach. They may want to consider two common approaches (or a combination of both) commonly described: (1) recurrent cross-sectional (ie, comparing multiple time points) and (2) trajectory or longitudinal (ie, identifying development or narratives over time).4 9 12 Using a theoretical approach/framework may help decide on the approach to analysis and presentation of findings4; for example, Murphy et al used the normalisation process theory constructs to structure the coding framework in their longitudinal study of the implementation of remote consultations in primary care during the COVID-19 pandemic.
Recurrent cross-sectional analysis focuses on changes and themes at different time points at the level of the whole sample.4 9 12 In this approach, each round of data analysis explores a particular moment in time, but it should also aim to capture the temporal aspect and change between time points. Helpful approaches include thematic analysis (especially when the codes used capture aspects of change/stability) and framework-based analysis (eg, charting themes per time points). The strength of the recurrent cross-sectional approach is in that it enables a comparison and identification of patterns across the whole sample and that it often remains grounded in the data. The drawback may be difficulty in capturing individual narratives over time,9 which in turn may result in a description of each time point rather than an understanding of change.4
Trajectory or longitudinal analysis focuses on change of individuals or groups to identify trajectories of change over time.12 It needs to include the same participants at different time points and can be facilitated by developing ‘case’ summaries or narratives that capture the changes and key themes across time for each participant. IPA might be particularly suitable to this approach, and a framework analysis may also be useful to identify the types of trajectories for subgroups of participants. The trajectory approach enables capturing and presenting the temporality of data, but it can be more difficult to capture and present patterns across the sample. With complex LQR datasets, combining cross-sectional and trajectory approaches and multiple types of analysis may be needed to capture the various aspects of the data.9
Writing up the findings
As writing up the findings can form a part of, or refine, the analysis in qualitative research, it may be helpful to consider the analytical approach together with an approach to presenting the findings. Farr and Nizza15 identified two common approaches to presenting the findings in longitudinal IPA papers, which may be relevant to other methodologies as well.
In the ‘themes tied to time points’ approach, each theme captures a time point or stage and includes a description of all aspects of the participants’ experiences relevant to that time point. In other words, each theme illustrates the different experiences apparent at each time point that contribute to the overall process of change/transition over time (eg, pre-event and postevent). Therefore, different themes could be identified at different time points. This approach may help present findings from the recurrent cross-sectional analysis. For example, Smith described women’s transitions to motherhood at different time points during pregnancy and after giving birth, with themes such as ‘Early pregnancy: adjustment and uncertainty’.18
In the ‘themes spanning time’ approach, the findings are presented in one set of themes with each theme describing change over time. It may be that the focus is on a subset of themes to allow for including a more nuanced and in-depth account of change and commonalities and differences between participants. This approach may help present findings from the trajectory/longitudinal analysis and include descriptions of ‘cases’ or groups/types of trajectories. Most studies in primary care presented in table 1 used this approach; for example, Lawton et al captured the changes over time in patients’ perceptions and experiences of transitions from secondary to primary diabetes care within themes, such as ‘Practice-based care: a mixed blessing’.29
A combined approach is also possible, such as with one theme that is divided into time points and other themes which span time. For example, Murphy et al30 first summarised the changes over time during the transition to remote consulting and then used theoretical concepts to describe different types of changes. In all approaches it can help to present a single case and then highlight similarities and differences with other participants, label the quotes to indicate the participant and time point, and use paired quotes from the same participant that show change/progression (eg, before and after)15 (eg, see Lester et al19).
Practical considerations
With LQR analysis, researchers need to carefully think through and plan how to best manage the practical aspects of the analysis. First, this involves data management—LQR often generates large amounts of data over a prolonged period, so thoughtful, consistent and secure data management is critical. Researchers should plan consistent record-keeping (eg, labelling the interviews, researchers involved, recording summaries and reflections) and data anonymisation processes (eg, when and how it should occur) to allow engagement with participants over longer time while protecting their identities. It may help to record other relevant details to inform the analysis and interpretation, especially at later stages and in longer studies. These may include contextual details, such as records of events, changing policies, media stories, etc, which are relevant to the research topic. Agreeing a consistent approach to note-taking and researcher reflexivity can help capture, access and use the researchers’ reflections and reinterpretations over time.
Second, it is important to plan when the analysis takes place (eg, after each interview, sets of interviews, time points, at the end), what tools and strategies are used to manage the process and the amount of data, and who is involved. For example, Lewis5 describes using different types of analysis at different points: summarising each interview after the interview within a framework organised by key themes (cross-sectional analysis) and developing ‘whole case’ summaries (a narrative analysis) after a number of interviews. Similarly, Thompson and Holland9 describe a provisional analysis after each interview focused on processual features (eg, structure and absences), substantive content and researcher’s reflections; then after a few interviews writing a ‘summary narrative’ for each location, identifying local themes and ‘case profiles’, and tracing changes and continuities in the individual narratives over time. Summarising and conducting a provisional analysis after each interview can help inform future data collection and make incremental progress with the analysis. Analysing the whole dataset at a later or the end stage of the study may be overwhelming when faced with a large dataset but may also enable more global, holistic meaning-making through an immersive and intense analysis of the whole dataset (rather than small ‘chunks’ of the data). Working with bigger datasets or longer studies often requires a team-based approach with different researchers contributing to different parts of the study and analytical process, which highlights the need for planning and consistency of the aspects discussed previously.
Finally, researchers need to consider the implications of the sampling on analysis and data management. Overly heterogeneous samples add complexity as comparisons could be made not only between participants and time points but also between groups of participants. For example, Calman et al4 describe the challenges of analysing data from patients with different types of cancer and trying to identify common trajectories. Analysing data from multiple types of health professionals or patients can make it more difficult to identify common trajectories than when focusing the data collection and analysis on a more homogeneous group from the outset. However, the differences between subgroups of participants may only become apparent during data collection and analysis. In this case, the use of frameworks can help with analysing and comparing data between groups of participants.
