“Lives and times”: The case for qualitative longitudinal research in anatomical sciences education

Abstract

Qualitative longitudinal research (QLR) focuses on changes in perceptions, interpretations, or practices through time. Despite longstanding traditions in social science, QLR has only recently appeared in anatomical sciences education (ASE). While some existing methodology papers guide researchers, they take a narrow view of QLR and lack specificity for ASE. This discursive article aims to (1) describe what QLR is and its benefits, its philosophies, methodologies and methods, considerations, and quality indicators, and (2) critically discuss examples of QLR in ASE. Underpinned by relativist ontology and subjectivist epistemology, time can be understood as fluid/subjective or fixed/objective. QLR is a flexible, creative, and exploratory methodology, often associated with other methodologies. Sampling is typically purposive, with repeated and recursive data collection methods, and complex three‐strand analyses (themes, cases, and time), enabling cross‐sectional and longitudinal analyses. QLR involves ethical, relationship, analytical, dissemination, and funding considerations. Key quality indicators relate to qualitative research as well as temporal aspects. Most of the nine ASE papers reviewed explored changes in anatomy learners, but few labeled their methodology as QLR. Just under half described their sampling as purposive, most employed pre‐planned and standardized repeated interviews, analyzed their data cross‐sectionally, and utilized qualitative data analysis software. Most cited the confirmability and transferability of their studies, but few cited credibility and dependability elements. Study timeframes and tempos were generally clear, but details of longitudinal retention/attrition were often lacking, and longitudinal data analysis was not often conducted. We therefore provide recommendations for the conduct of QLR in ASE.

INTRODUCTION

Qualitative longitudinal research (QLR, or alternatively longitudinal qualitative research: LQR) can be described as qualitative research with “special emphasis on time, the elapse of time and the changes or stability of practices, perceptions and interpretations when following actors through time.” ^{1 , p. 8} While “researching lives through time,” ^{2 , p. 2} has longstanding traditions in the social sciences since the temporal turn in the 1980s, ² QLR has gained popularity in health professions education research (HPER) in recent years. ³ , ⁴ , ⁵ , ⁷ Although QLR is still in its infancy in anatomical sciences education (ASE), several QLR studies have been published recently in ASE. These include examples of studies exploring changes in learners (e.g., anatomy learners' study habits ⁸ ), changes in educators (e.g., the development of gross anatomy near‐peer tutors' personal and professional competencies ⁹ ), or changes in anatomy learning artifacts (i.e., anatomy textbooks ¹⁰ ). These QLR ASE studies have diverse findings, for example, that gross anatomy tutoring improves near‐peer tutors' knowledge and skills, ⁹ or that donor dissection changes medical students' ethical perceptions. ¹¹ However, for readers interested in these study findings, we recommend reading the original articles.

QLR matters in ASE, like the broader field of health professions education (HPE), because education inevitably involves change, development, growth, transformation, journeys, and transition. Indeed, “knowing about the journey, with its stops and starts, detours, transitions, and reversals, enriches understanding of events and accomplishments along the way.” ^{12 , p. 1254} QLR is therefore an optimal methodology to explore transitions (most typically participants' transitions, but these could be transitions in settings over time), including how those transitions are negotiated and experienced emotionally, how people make sense of their past, present, and future (often in a non‐linear manner), and the factors influencing their transitions. ² , ¹³ , ¹⁴

QLR is thought to be a complex qualitative methodology, especially in relation to data analysis, ¹⁵ and few methodology papers exist in HPER to guide researchers. ¹² , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ These papers have begun to map the QLR terrain in HPER, but they can (a) take a narrow view of QLR methodology (e.g., repeated interviews), ¹⁵ (b) focus on select rather than comprehensive elements of QLR (e.g., conceptualizations of time; ¹² reflexivity; ¹⁷ theory in QLR), ¹⁸ (c) provide limited examples of QLR, ¹² , ¹⁵ , ¹⁶ or limited critical analyses of examples, ¹² and (d) only one includes QLR examples relevant to an ASE context. ¹² Therefore, this discursive article for ASE researchers aims to move beyond these HPER methodology papers to (1) describe what QLR is and its potential benefits for ASE, its philosophical underpinnings, diverse methodologies, and methods (sampling, data collection, and analysis), as well as its considerations and quality indicators, and (2) critically discuss nine diverse illustrative cases of QLR in ASE, to provide ASE researcher recommendations.

DESCRIPTION

What is QLR and its benefits?

Spanning the methodological traditions of qualitative and longitudinal research, QLR includes “qualitative studies, with repeated data collection, that focus on the temporality (e.g., time and change) of a phenomenon.” ^{19 , p. 1} As described by Neale, ^{13 , p. 1} QLR possesses a “dual identity” as both qualitative and longitudinal and, as such, “explores dynamic processes through an in‐depth qualitative lens… [providing] insights into how people narrate, understand and shape their unfolding lives and the evolving world of which they are a part.”

While definitions vary, ¹⁴ researchers have characterized QLR as qualitative research including multiple rounds of data collection with the same participants and/or in the same setting, and recursive data collection (e.g., interviews requiring participants to reconsider earlier narratives and imagine future possibilities). ¹² , ¹⁸ , ¹⁹ The number of data collection points is project, phenomenon, and methodology dependent, but many researchers suggest a minimum of two data points for QLR, to enable time‐related comparisons. ² , ¹⁴ , ²⁰ , ²¹ , ²² , ²³ Researchers have mostly not suggested a minimum study duration, arguing that a sufficient time‐period is one that allows for an adequate exploration of meaningful change. ¹⁹ , ²⁰ , ²³ , ²⁴ So, researchers should consider how liable their phenomenon of inquiry is to change, alongside the expected rapidity of that change, to determine an optimal study duration and the number/frequency of data collection points. ¹ , ¹⁴ , ²³ Neale ^{2 , p. 4} referred to this as “temporal logic,” helping researchers to determine when studies should begin and end.

QLR enables researchers to deeply explore critical time‐periods, turning points, changes, and continuities through time, ¹ , ² , ²² , ²³ , ²⁵ as well as build trusting longitudinal relationships and promoting participant disclosure. ²² , ²⁶ QLR can capture and beautifully illustrate the changes in participants' attitudes, perceptions, experiences, practices, and identities, as well as how individuals reflect on their past, present, and future in dynamic ways. ¹ , ¹⁴ , ¹⁶ , ²³ Moreover, specific QLR methods (e.g., diaries) can offer significant benefits for collecting unobtrusive and close‐to‐experience data, with rich and evolving data stimulating participants' ongoing reflective practices. ²⁷ , ²⁸

The philosophical underpinnings of QLR and temporal theory

Interpretivist philosophies underpinning qualitative research are central to QLR, ¹³ , ²⁹ , ³⁰ including a relativist ontology, subjectivist epistemology, and axiology privileging language, social interaction, and context, as well as the temporality of human experience. ¹⁹ , ²² , ²⁹ , ³⁰ , ³¹ See Box 1 for a glossary of QLR terms employed in this paper. Although QLR is thought to be flexible, innovative, and diverse, time is its keystone. ¹ , ² , ¹³ , ¹⁹ , ²¹ , ²³ , ³¹ , ³² Therefore, QLR further values thinking dynamically, ² , ¹³ as well as the trusting relationship between researcher and participant. ¹⁸

BOX 1. Glossary of qualitative longitudinal research (QLR) terms (in alphabetical order).

Axiology: The values underpinning a research approach and its philosophical foundations. ³⁰ Qualitative research is typically shaped by the values of language, social interaction, and context. ³⁰

Confirmability: The extent to which qualitative findings and qualitative researcher interpretations of data align with those qualitative data and could be confirmed by other researchers. ⁴⁹ , ⁵⁰ This might include researchers keeping reflexive journals or applying triangulation methods such as data or investigator triangulation. ⁴⁹

Credibility: The extent to which reported qualitative findings adequately represent the multiple realities of participants, so are credible from their perspectives. ⁴⁹ , ⁵⁰ This might include researchers establishing their expertise and prolonged participant engagement. ⁴⁹

Dependability: The extent to which qualitative findings are repeatable if the qualitative study was conducted with the same analysts, context, and participants. ⁴⁹ This might include researchers conducting audit trails and/or providing rich descriptions of qualitative methods. ⁴⁹

Diachronic analysis: Focuses on analyzing the unfolding data through time, so is conducted after all data are collected. ² , ¹⁴

Epistemology: Described as the nature of knowledge or knowing. ⁵⁰

Interpretivist: An approach to research common in qualitative research, which sees the researcher's job as identifying participants' subjective meanings through a process of interpretation. ²⁹

Ontology: Described as the nature of reality or being. ⁵⁰

Over time: A conceptualization of time as fixed, linear, and objective. ¹⁸

Purposive: A type of sampling common in qualitative approaches that samples participants aligned with the research purpose, so can include different types such as maximum‐variation sampling for diversity, or theoretical sampling for specific participants enabling the development of certain constructs. ²⁹

Reflexivity: The extent to which qualitative researchers (either individually or collectively, as in team‐reflexivity) reflect on how their assumptions and their relationships with the topic of inquiry and participants influence their study design, as well as their collection and interpretation of data. ⁵⁰

Synchronic analysis: Focuses on analyzing one snapshot in time, so is conducted during or after each data collection wave, so is synchronized with the data collection. ² , ¹⁴ , ²³

Tempo: Intertwined with the QLR timeframe, tempo refers to the pace of the study, reflecting the number, spacing, and continuity of data collection waves. ²

Three‐strand analysis: This involves integrating the analysis of themes, cases, and time (specifically through time) in QLR, which Balmer et al. described as: “a braid that fits into a conceptual framework to tell a cohesive story.” ^{47 , p. 1256}

Through time: A conceptualization of time as dynamic, unbounded, and subjective. ¹⁸

Timeframe: Intertwined with the QLR tempo, timeframe refers to the study duration, reflecting the start and end of qualitative data collection. ²

Transferability: The extent to which qualitative findings are applicable to other settings or groups. ⁵⁰ This might include researchers including diverse samples with sufficient size to enable transferability ⁵¹ and/or from multiple settings. ⁵⁰

Triangulation: The extent to which the qualitative study design employs multiple researchers, multiple data sources, or multiple methods (data collection/analysis) to establish convergence. ⁵⁰

Trustworthiness: A broad marker of quality in qualitative research incorporating other quality elements outlined by Lincoln and Guba ⁴⁰ such as credibility, dependability, confirmability, and transferability. ⁴⁹ , ⁵⁰

Within QLR, time is often seen as a “vehicle rather than a topic,” ^{2 , p. 3} but has been conceptualized in diverse ways. Firstly, time can be understood as fluid, subjective, and complex (thinking “through” time) and/or fixed, objective, and linear (thinking “over” time). ² , ¹² , ¹³ , ¹⁸ , ¹⁹ Interestingly, while researchers argue that time can be fluid and fixed, ¹² several researchers ¹³ , ¹⁸ prefer conceptualizations “through” time rather than “over” time because “through” time emphasizes the dynamism and fluidity of time, the collaborative and recursive nature of longitudinal research, and helps to distinguish QLR from cross‐sectional and two timepoint, follow‐up studies. ¹² Within this fluid/fixed conceptualization, time can be further considered as synchronic (snapshot in time) and/or diachronic (unfolding through time). ² , ¹³ , ²⁶ For example, researchers can provide a synchronic take on data presenting themes by cases at one point in time, as well as a diachronic (i.e., temporal) exploration of patterns through time of themes by cases. ² , ¹⁴ , ²³

Secondly, time can be considered across three planes: (1) prospective–retrospective: orientating to past, present, and future; (2) micro–macro: orientating to individuals/small collectives or larger samples; and (3) intensive–extensive: orientating to different timeframes (e.g., short‐term or long‐term), tempos (e.g., the number of, and spacing between, data collection stages), and temporality. ¹ , ² , ¹² , ¹³ , ¹⁶ , ¹⁹ , ²¹ , ³¹ Indeed, timing relates to the ideal time to explore participants' experiences or perceptions, whereas tempo describes the intensity of time, and temporality processes such as lifecycles or phases. ¹⁸

Thirdly, Audulv et al. ¹⁹ identified how longitudinal perspectives were conceptualized in the 299 health QLR papers they reviewed; in decreasing order of frequency, longitudinal terms were employed as follows: “change,” “over time,” “process,” “transition,” “implementation,” “development,” and “longitudinal.” Interestingly, in the papers reviewed, they found that time was conceptualized in three fundamental ways: (a) as the phenomenon of interest (e.g., exploring change, trajectories, or processes); (b) as the study outcomes (e.g., exploring mechanisms underpinning intervention outcomes); and/or (c) as the study context (e.g., exploring experiences over time without change being a primary interest). They sometimes found that papers conceptualized time in multiple ways, or time was not described at all.

Finally, Ayalon et al. ²⁰ provided a conceptual framework of change and stability in QLR including considerations such as who defines change/stability? (e.g., participants or researchers?), what has changed over time? (e.g., new topics at different times or enduring topics discussed in different ways?), and in what timeframes? (e.g., within the same or different data collection stage?). Ultimately, these theoretical understandings of time, as well as the philosophical assumptions of QLR, serve to guide the design of chosen methodologies and methods to ensure internal coherence (and thus quality) in QLR. ³³

Common methodologies and methods for QLR

QLR has been described as a flexible, creative, and exploratory methodology, ¹ , ² , ¹³ capturing dynamic research processes, often with recurring cycles of sampling, recruitment, data collection, and data analysis. ² Interestingly, qualitative longitudinal researchers can be seen as “bricoleurs,” rather than methodological purists, integrating numerous techniques and tools to design bespoke studies. ² Methodologies often associated with QLR include ethnography, biography, narrative, phenomenology, case study, and grounded theory, as well as qualitative longitudinal arms in mixed methods studies or randomized controlled trials. ¹ , ¹³ , ¹⁹ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ These are not always labeled as QLR, however. ¹⁹ , ³¹ Methods used in QLR are diverse in terms of sampling (e.g., purposive, theoretical), data collection (e.g., ethnographic immersion, repeated interviews, participatory approaches such as diary methods, and document analysis), ² , ¹⁶ , ¹⁹ , ²⁸ and data analysis (e.g., synchronic and/or diachronic analyses, cross‐sectional and/or longitudinal analyses, and individual and/or group‐level analyses). ² , ¹⁴ , ¹⁶ , ²¹ , ²³ , ²⁴ , ²⁵ , ³⁴

As indicated above, sampling in QLR is often purposive, based on participants' shared experiences of the topic of inquiry. ²² , ²³ Noteworthy is that qualitative longitudinal researchers need to consider not only the sampling of cases, but also their sampling throughout time, and especially considering the tempo of data collection. ² Indeed, qualitative longitudinal researchers can employ a time‐based or event‐based approach to determining the timing of data collection stages. ²⁸ While sample sizes for qualitative research are variable depending on studies' intended purposes, underpinning philosophies, and methodologies, ³⁵ , ³⁶ they are typically small (e.g., 9–17 interviews or 4–8 focus groups), ³⁷ but can be smaller still for in‐depth QLR, given that individuals or collectives can be tracked relatively intensively over modest time‐periods or less intensively over longer time‐periods. ² , ¹³ Therefore, it is not unusual to see single‐digit sample sizes for QLR such as Balmer et al.'s ³⁸ study exploring six medical trainees' 9‐year journeys across medical school and different residency programs.

QLR data collection methods are diverse, flexible, and iterative, employing less standardized methods of data collection and analysis. ¹ The most common data collection method in nursing and health QLR studies is interviews (employed in 74% and 55.9% of studies reviewed, respectively). ¹⁹ , ²² A further 20% of nursing and 32.8% of health QLR studies reviewed used a combination of data collection methods, such as interviews alongside observations, focus group discussions, surveys, questionnaires, or diaries. ¹⁹ , ²² Repeated qualitative interviews with reasonably small sample sizes are also standard in HPER. ¹⁵ Interestingly, data collection methods often correspond with the entities (either individuals, cases or dyads, groups, or settings) followed longitudinally, with, for example, individual interviews typically corresponding with following individuals longitudinally and focus group discussions for following groups longitudinally. ¹⁹ Studies also employ diverse timeframes and tempos, with follow‐up of 6 months‐6 years, and from 2 to 20 follow‐up waves. ²² Researchers might adjust the timeframe and tempo of a study as it unfolds, based on the data collected (i.e., participant‐adapted data collection) and unexpected changes in participants' lives, ² , ¹⁹ as well as other features relating to data collection (e.g., interview schedules, the data collectors and/or analysts). ²⁰ Indeed, Audulv et al. ¹⁹ suggested that timeframe and tempo are often inter‐connected, with shorter durations corresponding to faster tempos of data collection and vice versa. However, some level of standardization is often required to enable comparisons across time and participants in terms of stability and change, so researchers often pre‐plan data collection. ¹ , ² , ¹⁴ , ¹⁹

Although some argue that “there is no correct or best way to analyse any qualitative data,” ^{25 , p. S212} qualitative longitudinal analysis typically employs thematic and case analysis that is situated within an omnipresent temporal structure. ² Qualitative longitudinal researchers suggest that analyses should attend to the integration between themes, cases, and time (specifically “through” time), in what has been called a “three‐strand” data analysis approach. ² , ^{12 , p. 1256} Indeed, data are typically analyzed both cross‐sectionally and longitudinally. ¹⁵ , ²⁵ Essentially, what changes from one data collection stage to the next for the whole sample, versus how does a participant's story unfold through time. ¹⁵ Researchers differentiate between in‐depth longitudinal case analysis (as in pen portraits, case profiles, case histories, case matrices, analytical summaries, and/or framework grids mapping cases against time for each theme) and cross‐case thematic analysis that are influenced temporally (including framework grids and matrices, illustrating changing themes across cases). ² , ¹⁴ , ²³ , ²⁴ , ²⁵ , ²⁷ , ³⁴ Furthermore, some recommend visual representations of longitudinal patterns such as a theme by time matrix. ²⁷ While data analysis may start before the completion of all data collection stages, researchers must collect data from all timepoints before conducting a complete longitudinal analysis. ²¹

Considerations for QLR

QLR has numerous ethical, relationship, analytical, dissemination, and funding considerations, ² , ¹⁵ , ¹⁶ , ¹⁷ , ²¹ , ²⁶ , ³¹ , ³² which might partly account for the relative dearth of QLR studies in ASE. Indeed, these considerations can seem somewhat overwhelming, particularly for novice qualitative researchers. ¹⁵

From an ethical perspective, QLR has been described by some as constituting “intrusion into people's lives,” ^{21 , p. 122} Consequently, researchers must manage difficult decisions about the extent to which they actively encourage repeated participation over time, versus honoring participants' rights to withdraw from longitudinal research at any point and without giving reasons, thereby recognizing consent as a continuous process. ² , ¹⁶ , ²² , ²⁶ Furthermore, unintentional disclosure is arguably more likely with repeated than one‐off interviews. ²² Through prospective in‐the‐moment QLR (such as with diary methods), researchers might hear ethically concerning information from participants as the study unfolds, which they feel compelled to act on and to protect the participant or others (e.g., students, patients). ¹² , ¹⁶ Therefore, qualitative longitudinal researchers need to balance proactive strategies (i.e., a priori/planned strategies such as informed consent), with reactive ethical strategies (i.e., in‐the‐moment unplanned strategies such as agreements to breach confidentiality to protect vulnerable actors at risk) throughout the study duration. ² , ³² Other ethical issues can relate to the co‐productive nature of QLR, raising issues over privacy, anonymity and confidentiality, ownership, reputation, exploitation, and maintaining professional researcher–participant boundaries. ² , ²¹ , ²² , ²³ , ³¹

From a relationship perspective, having sufficient access to participants during the study duration is key, along with oversampling to account for attrition, and minimizing attrition through respect, relational engagement, and incentives. ¹ , ² , ¹⁴ , ¹⁵ , ¹⁶ , ²¹ , ²³ , ²⁷ , ³² Attrition, for example, can negatively impact the explanatory power of QLR. ¹ Qualitative samples may be more restrictive because fewer people are able/willing to commit to studies longer‐term. ¹ Several researchers ¹² , ¹⁵ , ¹⁸ , ³² have argued that trusting relationships are central to QLR and that such trust can influence the candor, depth, and richness (and hence quality) of data collected through time, but can also render participants and researchers vulnerable. ¹² , ¹⁸ , ²¹ , ²³ , ²⁵ Over longer periods of time, there may be changes in the research team affecting continuity of data collection and analysis, as well as trusting relationships. ¹⁴ , ²¹ , ³² Balmer and Richards ^{18 , p. 281} described managing this process through non‐finalizability (“no one meaning is final and no one's meaning is final”), second person address (speaking with participants rather than about them, and “walking alongside” participants), ² , ¹³ with no claim of privilege (recognizing that participants are experts in telling their own stories, and as such, researchers do not assume interpretive authority).

From an analytical perspective, QLR requires more complex and labor‐intensive analyses to account for dynamic temporal changes in often large amounts of qualitative data. ² , ¹² , ¹⁴ , ¹⁵ , ²¹ , ²³ , ²⁴ , ²⁷ , ³⁴ For example, diachronic analysis typically occurs after all data have been collected and includes data at all timepoints, whereas synchronic analysis involves analyzing data after each wave to inform subsequent waves of data collection. ¹² , ¹⁴ Such voluminous data can quickly become unmanageable, ² so qualitative longitudinal researchers can employ computer‐assisted qualitative data analysis software (CAQDAS) to help manage voluminous data. ²²

In terms of disseminating QLR analyses, challenges relate to the articulation of methods (complex sampling, data collection, and data analysis) in a parsimonious fashion, as well as the volume of data, some of which can lack an analytical endpoint. ¹ , ¹⁵ SmithBattle et al., ²² for example, found that many nursing QLR studies lacked sufficient detail on sampling, data collection and analysis, retention strategies, and attrition, resulting in challenges establishing the transferability and credibility of study findings. The prohibitive word count restrictions of journals may mean that researchers struggle to articulate the complexity and richness of QLR findings too. ¹⁴ , ²¹ Researchers can also struggle to decide what to present (e.g., individuals and/or cohort, cross‐sectional and/or longitudinal data), as well as when to publish (i.e., whether to publish earlier rounds of data collection or wait until data are fully collected). ²³

Finally, from a funding perspective, given the dearth of funding for, and underfunding of, HPER, ³⁹ it can be challenging to find sufficient funds for QLR. ¹² , ²¹ , ²³ , ²⁴ , ²⁷ As a result, QLR studies are often designed pragmatically based on resources, rather than temporal logic. ² Researchers might decide on shorter timeframes and less intensive tempos when they have limited funding.

Quality indicators of QLR

Quality in QLR focuses on aspects common to other qualitative approaches such as credibility (e.g., believability of the findings), transferability (e.g., whether the findings resonate with other contexts), confirmability (e.g., whether other researchers corroborate the interpretations), and dependability (e.g., whether the study is repeatable). ²⁹ , ³⁰ , ⁴⁰ Moreover, quality in QLR relates to its temporality. ¹⁷ , ²² , ²³ , ²⁷ , ⁴¹ Serving credibility in QLR, Balmer et al. ¹² suggested that data should be collected recursively, that is, encouraging participants to reflect on previous data (e.g., “last time we spoke, you said X, where are you at with this now?”), as well as imagining forward (e.g., “how do you think this might change into the future?”). Indeed, several researchers argue that optimal QLR should incorporate retrospective aspects into a prospective design to blend a temporal exploration of past, present, and future. ² , ²⁰ , ³² Scholars also suggest that analysis should attend to themes, cases, and time (specifically “through” time) in a “three‐strand” data analysis. ^{12 , p. 1256}

Serving transferability in QLR, numerous retention strategies have been articulated in the broader health literature, including researchers having multiple contact details for participants (emails, addresses, and phone numbers), persistent “sleuthing” to keep in touch with participants, debriefing participants after data collection stages (also known as participant check‐ins), ²⁵ and providing participation incentives to improve retention and thus sample size through the study duration. ^{22 , p. 4}

Serving confirmability in QLR, qualitative longitudinal researchers should consider reflexively how time influences, and perhaps changes, their own thinking about the research and its unfolding methods, findings, and interpretations. ¹² , ¹⁷ , ²⁶ Furthermore, relational reflexivity is enabled by the longitudinal relationship developed between researchers and participants through multiple research encounters over time: “in which both researcher and participant can consider and reconsider their own understandings of experiences and build more nuanced insights.” ^{17 , p. 1224}

Serving dependability in QLR, Nevedal et al. ¹⁴ suggested that QLR studies should consider temporal research questions focusing on change, with comparisons between data collected from two or more time‐periods. They argue that researchers should provide clarity regarding what is compared (e.g., interview questions, themes, and cases), when data are analyzed (synchronic, diachronic, or both), and how analyses change over time (if at all). While Neale ² suggested that timeframes and tempos should be flexible, they should not be so flexible that problems are created in the management and analysis of data. Key criticisms of QLR in nursing/health include the merging of qualitative data across time‐periods, thereby ignoring the temporality of participants' experiences. ¹⁹ , ²² Furthermore, QLR studies are not always clearly identified/labeled as such. ²² Although formal reporting standards for QLR do not yet exist, ²² we argue that reporting standards for observational longitudinal research are helpful. ⁴¹ Alongside recommendations for QLR criteria thought to influence study dependability, ²² the following temporal details should be made explicit in QLR studies: study start and end dates; participant numbers at the start of the study and at each subsequent data collection stage; the number of data collection stages, and the timing in‐between (alongside the consistency of data collection methods); the procedures to retain participants across each stage (including any attempts at over‐sampling); the attrition at each stage of data collection and the reasons (where ethics permit); the longitudinal analysis methods (e.g., analysis within and across data collection stages; synchronic, diachronic, or both); and whether attrition is taken into account in the analysis. ¹⁴ , ²² , ⁴¹

DISCUSSION

In this section, we critically discuss nine illustrative cases of QLR in ASE (see Table 1). Using Web of Science, we ran various non‐systematic searches to identify examples of QLR in ASE over the last ten years, employing search terms such as “qualitative longitudinal research” and “anatomy education.” We also searched key HPER journals (e.g., Academic Medicine, Advances in Health Sciences Education, Medical Education, Medical Teacher, Nurse Education Today), as well as key anatomical journals (e.g., Anatomical Sciences Education, Annals of Anatomy). Like Balmer et al., ¹² we did not limit our search to studies labeled specifically as QLR (or LQR). We identified nine articles from seven studies over the last ten years focusing on (or including) anatomy education and employing QLR, even though researchers did not always label their studies as such. We summarized the key method‐related features of each of these articles (see Table 1). First, we documented the study aims, including identifying how time was articulated. Second, we identified the study methodology espoused by the authors. Third, we identified the sampling approach and samples, including which entities were followed through time, what (if any) participant retention strategies were used, and the longitudinal attrition. Fourth, we identified the data collection methods, including the study timeframe and tempo for data collection and whether data collection was pre‐planned or adapted and flexible or standardized. Fifth, we documented the data analysis methods, and what (if any) software was used. Finally, we identified quality issues with each paper, including strategies to establish credibility, transferability, confirmability, and dependability. ²⁹ , ³⁰ , ⁴⁰ We also identified whether time issues were clear (such as study timeframe and tempo) and the extent to which data were analyzed longitudinally. Those wishing to better understand the findings of these studies are encouraged to read the articles.

TABLE 1.

Overview of qualitative longitudinal research in anatomical sciences education. ^a

Author (date)	Study aim	Methodology b	Sampling and samples	Data collection methods	Data analysis methods	Qualitative quality issues	Time quality issues
Alvarez and Schultz 9	To explore the development of professional and personal competencies through gross anatomy peer tutoring; time is seen as improved knowledge, skills, and abilities	Longitudinal mixed‐methods design combining qualitative and quantitative methods; Quantitative elements followed through time (improvements in knowledge, skills, and abilities)	Sampling unclear; 24 peer tutors of gross anatomy (year 2 medical students); Retention strategies included financial compensation; Participation rates and attrition across study not explicit (although we assume 0% attrition given 72 interviews in total over three timepoints)	Three questionnaires and semi‐structured interviews over 18‐month period collecting pre‐planned and standardized quantitative and qualitative data over time (T1 = beginning of gross anatomy course; T2 = 4–6 months later; T3 = 4–6 months after the second interview; 72 questionnaires and interviews in total)	Structured content analysis employed using deductive and inductive approaches; SPSS used to evaluate questionnaire data longitudinally (e.g., repeated measures ANOVA); MAXQDA employed for qualitative data and categorized by the roles of the doctor espoused in CanMEDS and NKLM frameworks, but this was not explored longitudinally	Two analysts coded data independently, discussing any disagreements and amending coding (credibility); Authors acknowledge limited generalizability and self‐selection bias	Timeframe and tempo clear; QLR analysis not conducted
Johnson and Gallagher 8	To explore the (deep and surface) cognitive processes employed by undergraduate students while learning anatomy and physiology; Time is seen as changed study habits	Qualitative, comparative case study involving the 3P model of teaching and learning processes (presage, process, and product)	Winnowed sample of 11 biology or health students with preferences for surface (n = 6) or deep (n = 5) approaches to learning; Retention strategies included multiple incentives (e.g., coupons, gift cards, food); Participation rate at each timepoint and attrition clear (n = 11 at T1; n = 10 at T2; n = 8 at T3)	Three semi‐structured interviews over 8‐month period across two semesters collecting pre‐planned and standardized qualitative data using interview protocol (but sometimes tailored based on participants' weekly diaries; 29 interviews in total); Weekly diaries pre‐planned and standardized (10 pairs of diary prompts)	Two analysts coded qualitative data through 7 steps: (a) bracketing; (b) attribute and descriptive coding; (c) deidentifying data; (d) block coding using a priori themes; (e) first cycle open coding; (f) code mapping grouping process codes; and (g) second cycle coding, but this was not explored longitudinally; CAQDAS use for qualitative analysis unclear	Developing results were presented to other education researchers for critical review (credibility); Researchers established an audit trail including memos (dependability); study triangulated data from several sources (confirmability); authors argue for study “validity and reliability” (p. 490)	Timeframe and tempo clear; QLR analysis not conducted
Pyörälä et al. 42	To explore medical and dental students' self‐reported study uses of mobile note taking (including their annotated anatomy notes); time is seen as changed note taking practices	Action research project involving online questionnaires, plus focus groups	Sampling not stated; 124 medical and 52 dental students with iPads from their first to fifth year of study; questionnaire response rates clear for each year (73–95%); focus group sizes roughly clear (range provided for some); Retention strategies included participatory approach: “constantly collaborated with… participants in the study” (p. 8); participation rate at each timepoint and attrition clear	Data collection over a 5‐year period employing pre‐planned and standardized annual online questionnaires with closed and open‐ended questions (five data points); three focus groups in the first year of the study, and two focus groups in the third year of the study	Inductive, qualitative content analysis was employed by two analysts to independently code open‐ended questions; authors compare six themes quantitatively (%) between the years for the cohort; CAQDAS use for qualitative analysis unclear	Authors make claims for the trustworthiness of their analysis through researcher/data triangulation (confirmability), member checking (dependability), as well as discussing transferability	Timeframe and tempo provided but imprecise (e.g., seasons); QLR analysis conducted, but presented quantitatively (%)
Silva et al. 10 , c	To analyze the anatomical illustrations in two textbooks (published 1755 and 1764) and two anatomical atlases (published 1822 and 1823) for Chilean medical students' human anatomy learning and dissection practice; time is seen as evolution of anatomical illustration quality	Historical analysis of textbooks and anatomical atlases	Sampling not stated; two textbooks and two anatomical atlases published in different years (1755, 1764, 1822, and 1823) across a 68‐year period; attrition not applicable	Directed search of academic repositories to identify relevant textbooks and anatomical atlases displaying anatomical illustrations that were used by medical students in Chile from 1758 to 1833 (four data points over 75‐year search period)	The authors conducted bibliographic (including author, title, publication place, and language), quantitative (including image labeling and image classifications as per others' work), and descriptive–qualitative analyses (focusing on design characteristics and image quality); relevant data extracted into a table; CAQDAS use for qualitative analysis unclear	Quality elements unclear	Timeframe and tempo clear; qualitative data analyzed cross‐sectionally (although one claim is made for the evolution of book printing techniques in conclusions)
Stephens et al. 11 , d	To understand how anatomy learning through donor dissection influences medical students' ethics perceptions; time is seen as changed ethics perceptions	Inductive, exploratory cross‐sectional and longitudinal qualitative study	Purposive sampling; 207 medical students completing online discussion forums during three semesters, and 24 students participating in at least 1/11 interviews at the end of each semester; two cohorts over 18‐month period; no retention strategies mentioned; participation rate at each timepoint clear for discussion forums (Cohort A: n = 103 at T1, n = 20 at T2, n = 25 at T3; Cohort B: n = 84 at T1, n = 9 at T2, n = 5 at T3); Authors mention last‐minute interview participant withdrawals and dropout across the study but attrition unclear	Data collection of multiple sources occurred over 18‐month period (6 waves) including pre‐planned and standardized online discussion forums during three semesters and semi‐structured interviews (one per semester; 11 interviews in total); five interview participants took part in more than one interview at different timepoints	5‐step framework analysis involving NVivo: (a) familiarization; (b) developing coding framework; (c) indexing; (d) charting (including exploring themes and timepoints); (e) mapping and interpretation; one longitudinal case study (student participating in all three timepoints) was identified, analyzed and presented	Authors clearly articulate team reflexivity (confirmability), as well as discuss transferability of their findings	Timeframe and tempo clear; longitudinal analysis presents temporal thematic changes across the whole sample, as well as an illustrative longitudinal case
Stephens et al. 43 , d	To identify the positive and negative impacts of anatomy education on preclinical medical students' tolerance of uncertainty; time is seen as changed stimuli, moderators, and appraisals/responses	Qualitative longitudinal research study drawing on Hillen and colleagues' 47 tolerance of uncertainty model	Purposive sampling; two cohorts of first‐year medical students (207 students participating in online discussion forums, and 24 students participating in interviews across three interview time‐periods of 18 months); no retention strategies mentioned; authors mention participation in discussion forums declining over time but attrition unclear	Pre‐planned and standardized qualitative data collected in 6 waves including online discussion forums during three semesters (18 months) in years 1 and 2 of the medical program, and semi‐structured interviews at the end of each semester (T1‐T3; 11 interviews in total); five interview participants took part in two interviews at two different timepoints	Abductive approach to framework analysis using NVivo including inductive analysis to build theory on tolerance of ambiguity, and deductive analysis drawing on the Hillen et al. 47 model; authors also explored temporal changes across the entire dataset; one illustrative longitudinal case provided	Authors clearly articulate team reflexivity (confirmability), as well as discuss conceptual generalizability (transferability)	Timeframe and tempo clear; longitudinal analysis presents temporal thematic changes across the whole sample, as well as an illustrative longitudinal case
Stephens et al. 44 , e	To explore perceived moderators of clinical medical students' perceptions of, and responses to, uncertainty; time is not explicitly conceptualized	Longitudinal qualitative study employing in‐semester reflective diaries and end‐of‐semester group or individual semi‐structured interviews, drawing on Hillen et al. 47 as initial conceptual framework	Purposive sampling; two clinical year cohorts (23 from year 3 and 18 from year 5); of 41 medical students recruited, 35 completed all eight stages of data collection through two semesters; retention strategies included giving participants choice of diary method (audio, typed or handwritten) and interview (individual or group); participation and attrition rates across the eight stages of data collection unclear	Participants completed eight data collection stages, including minimum of six pre‐planned and standardized in‐semester reflective diary entries (three per semester approx. six weeks apart; 230 entries in total), plus two pre‐planned and standardized individual or group semi‐structured interviews (one at the end of each semester; 40 interviews in total); total data collection duration of 10 months/two semesters	Abductive approach to team‐based framework analysis using NVivo and drawing on the Hillen et al. 47 model; authors identified four broad moderator themes but did not analyze any temporal aspects for their presentation of findings	Authors clearly articulate team reflexivity (confirmability), discuss transferability, as well as outline study rigor including information power, crystallization, and use of existing theory (confirmability)	Timeframe and tempo clear; QLR analysis not conducted
Stephens et al. 45 , e	To explore how medical students describe their responses to uncertainty, and how these change (if at all) over time; time is seen as changed responses to uncertainty	Longitudinal qualitative study employing in‐semester reflective diaries and end‐of‐semester group or individual semi‐structured interviews, and drawing on Hillen et al. 47 as initial conceptual framework	Purposive sampling; two clinical year cohorts (23 from year 3 and 18 from year 5); of 41 medical students recruited, 35 completed all eight stages of data collection through two semesters; retention strategies included giving participants choice of interview method (individual or group); participation rates across time for diaries are provided (n = 41, 40, 39, 38, 37, 35) but attrition unclear for interviews	Participants completed eight data collection stages, including minimum six pre‐planned and standardized in‐semester reflective diary entries (three per semester; 230 entries in total), plus two pre‐planned and standardized individual or group semi‐structured interviews (one at the end of each semester; 40 interviews in total); total data collection duration of two semesters	Abductive approach to framework analysis using NVivo and drawing on the Hillen et al. 47 model; authors explored patterns across the response themes from the diary data only over time	Authors clearly articulate team reflexivity (confirmability)	Timeframe clear but tempo not wholly clear; QLR analysis conducted
Walser et al. 46	To explore the teaching activities of near‐peers during the dissection course, and whether these change over time; time is seen as changed teaching activities	Methodology unclear; described as quantitative and qualitative analysis of semi‐structured reports (i.e., teaching logs of tutor activity and individual reflection)	Sampling unclear; 23 student tutors participating in a train‐the‐tutor program; high retention due to daily teaching logs being mandatory part of training, but response rate for reports was 565/575 = 98.3%, so small attrition (10 reports); participation and attrition rates at each time point unclear	565 pre‐planned and standardized semi‐structured reports (i.e., teaching logs of tutor activity and individual reflection) documented daily across the 25‐day dissection course	Quantitative analysis using SPSS (including pie charts categorizing activities and average relative time expenditure for each category); quantitative longitudinal analysis including paired t‐tests; qualitative analysis of two open‐ended questions from teaching logs employing grounded theory and frequencies, but this was not explored longitudinally; CAQDAS use for qualitative analysis unclear	Three analysts independently coded qualitative data before group discussion and agreement (credibility); authors discuss bias and transferability	Timeframe and tempo clear; QLR analysis not conducted

^a Web of Science searches to identify qualitative longitudinal research (QLR) in key health professions education research journals (e.g., Academic Medicine, Advances in Health Sciences Education, Medical Education, Medical Teacher, Nurse Education Today), and QLR relating to anatomical sciences education (e.g., Anatomical Sciences Education, Annals of Anatomy).

^b We employ the methodology terminology used by the authors wherever possible, although we have also constructed the methodology of these papers as QLR.

^c We gleaned data from this Spanish‐language paper employing Google translate.

^d Stephens et al. ¹¹ , ⁴³ are from the same study.

^e Stephens et al. ⁴⁴ , ⁴⁵ are from the same study.

As seen in Table 1, these studies explored changes in the context of anatomy education, with changes related mostly to anatomy learners (n = 6), but also to near‐peer anatomy educators (n = 2), and anatomy learning artifacts (i.e., textbooks: n = 1). Across the nine papers, time was mostly constructed as change (n = 6) but was also conceptualized as improvement (n = 1), evolution (n = 1), or it was not explicitly conceptualized (n = 1), consistent with previous health research. ¹⁹

Only some papers labeled their methodology as QLR (or LQR; n = 4); other methodologies explicitly cited included mixed‐methods (n = 1), case study (n = 1), action research (n = 1), and historical analysis (n = 1), consistent with previous health research. ¹⁹ One study did not cite their methodology (n = 1). Sampling approaches across the papers were mostly purposive (n = 4), but also included “winnowed” sampling (n = 1), or sampling was not stated/unclear (n = 4). Sample sizes included n = 4 for documents, but ranged from n = 11–207 for participants (average of n = 91 participants), so were arguably large. ³⁵ Indeed, the average participant sample was somewhat larger in ASE than has been found in previous health, nursing and HPE QLR (average of n = 20, 22, and 18 participants, respectively). ¹² , ¹⁹ , ²² Note that large sample sizes in qualitative research are not necessarily perceived as better, because they can prohibit detailed case‐oriented analysis (and micro‐analysis) of data that is expected in some qualitative methodologies such as interpretive phenomenological analysis. ² , ³⁵ , ³⁶ , ⁴⁸ Sample retention strategies were diverse including (in decreasing order of frequency) financial incentives (n = 2); flexible data collection methods (n = 2), participatory approaches (n = 1), food (n = 1), or mandatory participation (n = 1). Sample retention strategies were not mentioned in some papers (n = 2) or were not applicable (n = 1). Sample attrition across studies was mostly unclear (n = 5), sometimes clear (n = 3), or not relevant (n = 1).

Diverse data collection methods were employed across the papers including (in decreasing order of frequency) individual or group interviews (n = 7), diaries (n = 3), documents (n = 2), questionnaires (n = 2), and discussion forums (n = 2). This is consistent with previous health and nursing research, ¹⁹ , ²² but is arguably more diverse than in HPER more broadly, possibly because of the more stringent definition of QLR applied by Balmer et al. ¹² Interestingly, data collection methods were typically pre‐planned and standardized (n = 8), consistent with previous health and HPE QLR. ¹² , ¹⁹ The data collection duration for study participants was also diverse, ranging from 25 days to 5 years (average of 18 months), except for the historical analysis of texts (68‐year time‐period), consistent with previous health, nursing and HPE QLR, which also had diverse study durations. ¹² , ¹⁹ , ²² The number of data collection waves for studies with individual or group interviews ranged from two to three (average of 2.6 per study). The number of data collection waves for other participant‐centered data collection methods (with more frequent tempos) were higher, such as daily teaching logs (25 data points), or weekly or monthly diaries (6–10 data points, average of 7.3). Consistent with Audulv et al., ¹⁹ the ASE papers suggested that timeframe and tempo were inter‐connected, with shorter study durations (e.g., 25 days) corresponding to faster tempos of data collection (e.g., daily). In terms of the qualitative data analyzed in the papers, five conducted cross‐sectional analysis and four conducted longitudinal analysis. The types of analysis mentioned were mostly thematic or framework analysis (n = 5), but also included content analysis (n = 2), descriptive–qualitative analysis (n = 1), or grounded theory (n = 1). Five of the papers described using qualitative data analysis software (e.g., MAXQDA or Nvivo) to support the analytical process, consistent with previous nursing QLR. ²²

In terms of qualitative quality issues, the papers cited diverse elements of qualitative trustworthiness; papers mostly cited the confirmability of their studies (e.g., reflexivity, triangulation: n = 6), and transferability (n = 5), but fewer studies cited credibility (e.g., multiple analysts: n = 3) and dependability elements (e.g., member checking: n = 2). One study purported to meet the quality indicators most often associated with positivist research (i.e., reliability and validity), which are not appropriate markers for qualitative research. ³⁰ In terms of quality elements relevant to longitudinal research, the timeframe and tempo of data collection were mostly clear (n = 7), but sometimes timeframe was clear, but tempo was not (n = 1), or both lacked precision (n = 1). As above, despite conducting QLR, oftentimes data were not analyzed and/or reported longitudinally within the papers (n = 5). This is relatively consistent with a previous review of nursing QLR, which found that papers often lacked adequate information about sampling, retention strategies, attrition, data collection, and longitudinal analysis, making it challenging for readers to understand the trustworthiness of study findings. ²²

While our critical analysis of ASE QLR is largely consistent with previous research in health, nursing, and HPE more broadly, ¹² , ¹⁹ , ²² we provide recommendations for ASE researchers for the conduct of QLR based on this analysis (see Box 2). Ultimately, we encourage researchers to employ QLR to explore change and continuity in ASE and hope that this methodology paper will act as a springboard to enable ASE researchers to take the plunge into QLR.

BOX 2. Recommendations for qualitative longitudinal research (QLR) studies in anatomical sciences education.

• Conceptualizing time: Think about how you are conceptualizing and operationalizing time in your study (e.g., over time or through time) and across three planes (i.e., prospective/retrospective, micro/macro, intensive/extensive). Ensure that you have internal coherence between your philosophies and temporal theory, with your chosen methodology and methods, and that you report this clearly.

• Methodology: Explicitly identify your QLR study as QLR (or LQR), including citing the methodology in your article title and keywords.

• Sampling: Purposively sample your study participants (and/or documents). Also consider your sampling throughout time, and state explicitly your sampling strategy.

• Retention/attrition: Consider diverse retention strategies to minimize attrition (but adopt these ethically) and report these, as well as attrition across time clearly.

• Data collection: Consider diverse qualitative data collection methods, think about employing pre‐planned and standardized approaches for researching over time, but more adaptive and flexible approaches for researching through time.

• Data analysis: Employ three‐strand qualitative data analysis strategies, conducting cross‐sectional and longitudinal analyses drawing on computer‐assisted qualitative data analysis software (CAQDAS), and report analytical approaches clearly.

• QLR considerations: Discuss ethical, relationship, analytical, dissemination, and funding issues (pertinent to your QLR study) among your research team throughout the study, and report key considerations influencing quality.

• Qualitative quality: Consider trustworthiness elements of QLR, and report credibility, transferability, confirmability, and dependability in your dissemination.

• Temporal quality: Ensure that quality elements relating to time are also reported clearly (e.g., timing/tempo, retention/attrition, and longitudinal analyses).

CONFLICT OF INTEREST STATEMENT

None, although CER is co‐author of two papers critically reviewed in Table 1.

Biographies

Charlotte E. Rees, BSc(Hons), MEd, GradCertTerEd(Mgt), PhD, PFHEA, FRCP(Edin) is Head of School of Health Sciences, The University of Newcastle, Australia, and Adjunct Professor in the Monash Centre for Scholarship in Health Education, Monash University, Australia. Her research interests include workplace learning, professionalism, professional identities, and educational transitions. She has conducted numerous longitudinal studies, including qualitative longitudinal research exploring educational transitions.

Ella Ottrey, PhD, BNut(Hons), BNutrDietet is Senior Research Fellow in the Monash Centre for Scholarship in Health Education, Monash University, Australia. Her research interests include qualitative methodology, interpretivist approaches, preparedness for practice, transitions into practice, and health workforce development. She has conducted qualitative longitudinal research exploring educational transitions with students and graduates from medicine, nursing, dietetics, and pharmacy.

DATA AVAILABILITY STATEMENT

The nine papers critically reviewed as part of this discursive article are all publicly available.