The Rehabilitation Intervention for People With Schizophrenia in Ethiopia (RISE) cluster randomized controlled trial evaluated the effectiveness of a 12-month community-based rehabilitation program in improving disability among people with severe mental illness in rural Ethiopia [26-28]. To explore experiences of personal recovery in this setting, RISE conducted 52 narrative interviews with 13 linked pairs of individuals with severe mental illness and their primary caregivers at baseline and 12 months [29]. People with severe mental illness were asked to share their experiences over time, describing preillness, help seeking, and treatment; hopes for the future; and barriers to and enablers of desired changes. Caregivers were usually family members who held a primary role providing care due to the lack of mental health services across the region [30,31]. They were asked to comparatively describe their perceptions of their relatives’ experiences. This case study, therefore, amounts to a narrative inquiry into the experiences of a specific group of people, where the work of the people conducting the narrative inquiry was informed by their knowledge of the specific needs of participants.

The NEON study curated a collection of 659 narratives describing personal recovery from mental health problems for use within research that examined whether access to the narratives in this collection could help people affected by mental health problems. This included 3 clinical trials that evaluated whether web-based access to NEON collection narratives benefited people affected by mental health problems [13,32,33]. NEON collection narratives were donated by individuals and sourced from existing published collections such as books and websites. Almost all the narratives in the NEON collection were already in the public domain. To curate this collection, the NEON study team developed consent procedures, inclusion criteria, and decision-making processes [12]. The NEON team also collected and analyzed 80 recovery narratives from groups underrepresented in research [34]. For both datasets, the NEON team developed approved procedures for sharing narratives with other researchers. These included documents specifying practices required of those researchers, such as the inclusion of people with personal experiences of mental health challenges in analytical processes. All the development described above took place through internal discussion and iterative engagement with people with experience curating narrative collections; people with lived experience of mental health problems; and the study sponsor, who governed the ethical and legal conduct of the study. The portion of the work of NEON presented in this case study, therefore, amounts to an action research exploration of narrative practices that integrated the perspectives of a range of experts.

As a creative and reflexive approach to data interpretation in a PhD thesis [35], a researcher created poems from narrative interview transcripts that had previously been considered through Labovian narrative analysis [36]. This work built on a tradition of poetic transcription, which has been described as “the creation of poem-like compositions from the words of interviews” [37] and has also been termed “poetic condensation” [38]. Poetic transcriptions have been generated from narratives to engage and evoke the imaginations of audiences [39] and provide unique insights into the personal experiences of research participants [40]. An example from the PhD thesis is provided in Multimedia Appendix 1. Through the work described in this case study, the researcher carefully reflected on the ethical and legal issues inherent in producing new creative material from existing narratives.

There are a range of online platforms that enable people experiencing mental health problems to receive support from peers with similar experiences. Discussions on these platforms frequently involve the sharing of personal narratives of mental health problems, frequently in the form of shorter or longer narratives. While such platforms can offer a rich environment for researchers to test out their research questions and hypotheses, the perceived privacy, intimacy, and sensitivity of the content that they hold urges researchers and platforms to uphold the highest ethical standards as they undertake responsible research practices [41] that go beyond the legal requirements within platforms’ terms and conditions. A case study examined how to work responsibly with the Kooth online platform [42] and proposed a series of recommendations [43].

Table 1 summarizes the ethical and legal considerations that we identified across the 4 case studies. These have been grouped into 3 categories reflecting different phases of the research process: consent procedures, narrative analysis and interpretation, and dissemination of data and findings.

In synthesizing these 4 case studies, we did not attend to considerations on data curation as these have been addressed in depth in our systematic review on narrative uses and misuses [15].

Obtaining appropriate consent for the collection of narratives raised challenges. In case study 1 (the RISE trial), narratives were collected directly by researchers for research purposes through procedures that involved the provision of participant information and the collection of informed consent. Although obtaining informed consent is an ethical standard in health research, its significance is perhaps amplified within interview-based narrative research given that interviews elicit detailed information as data about participant experiences that may feel deeply connected to personal identity. There may be a particularly high risk of harm to participants if individuals do not fully understand the implications of a study or how the data collected and conclusions drawn through their narratives will be disseminated. This was amplified for case study 1, which was conducted in a setting (rural Ethiopia) where 40% of the population is not formally educated and, hence, lacks literacy, with consent procedures predominately reliant on verbal explanation by research staff or relatives with some formal education [44]. In this setting, people with severe mental illnesses are often generally disempowered by a lack of autonomy and involvement in decision-making processes about their care or life [45,46], meaning that any problems with research consent procedures risked exacerbating previous inequities. Hence, the researchers invested substantial effort in maximizing the clarity of their consent procedures and the materials they used.

Case studies 2 (NEON study) and 4 (online platforms) demonstrate tensions between the legality of the consent process and individuals’ own expectations for reasonable privacy. In case study 4, researchers accessed sensitive narratives shared publicly online but where service users had not shared them for research purposes. To enable ethical access, the researchers worked with the online service provider to insert functionality to collect additional consent rather than relying on clauses buried in a complex “terms and conditions” document that indicated that data could be used to “improve their services” and shared with third parties. Even though it may have been legal to access consented data via uninformed procedures, the researcher felt it morally wrong to access these data without expressed and informed consent; it is widely recognized that online users may not fully attend to or understand the nature or scope of the activities to which they have legally consented [47], and in at least one case, public concern has been raised about an analysis conducted on the basis of clauses in a social media site “terms and conditions” statement [48].

In case study 2, both collections consisted of narratives describing recovery from mental health problems, with a variety of approaches to narrator and third-party identifiability. For example, the NEON collection received some donations of narratives that were pseudonymous and others that directly identified the narrator or third parties by name or through included events or descriptions that would enable the easy identification of either. Given the mental health focus of these narratives, this means that narrative donations frequently contained sensitive personal data as defined in the GDPR. An additional complexity was that some donations included sensitive personal information about third parties whom the narrator had never met but who had inspired their recovery. For example, some narratives referenced the British actor Stephen Fry, who has described his mental health experiences in a range of influential autobiographical books [1]. The legality of this issue was further complicated because, ethically, the research team had committed to not modifying a narrative and respecting narrator choice on self-identification. Each of these commitments excluded a choice to remove sensitive personal information from these narratives.

To address these considerations, the research team developed comprehensive consent procedures, inclusion criteria, and decision-making processes for narrative inclusion and processing to use their collection in research. Their eventual choice was to ensure that they had valid narrator consent to process and share sensitive personal information, to exclude narratives presenting sensitive personal information about third parties, but not to exclude narratives where the research team had collected direct evidence that third-party information had been “manifestly made public by the data subject.” The latter is a term included in article 9 of the GDPR [21] (clause 2e), which removes a prohibition on the processing of sensitive personal data, including sensitive personal data related to the health of the data subject. This allowed the research team to include narratives in which the narrator described the beneficial impact on the self of learning about others’ mental health experiences (eg, as described in Stephen Fry’s autobiography) but to exclude narratives where that information had not been made public (eg, a named person that the narrator had encountered on a mental health ward). Furthermore, the research team only shared these narratives with other researchers for the purposes of secondary research analysis in cases in which they had documented consent to do so.

For a smaller collection of NEON narrative interviews that were collected through researcher interviews, the team took a different path to enable sharing with other researchers for secondary analyses. They had previously pseudonymized these data for analysis through a protocol approved before they realized the importance of giving narrators control over their identity. To enable sharing with other researchers, they then increased transcript anonymity by redacting information about events described in the narrative where there was a reasonable possibility that the narrator could be identified through comparing or combining public records of those events with content described in the narrative. This process was approved by the supervising ethics committee and study sponsor and was legally correct as it placed the anonymized transcript outside of data protection regulations. The study investigator was responsible for confirming the appropriateness of redactions, with their decisions auditable by the study sponsor.

Case study 1 (the RISE trial) illustrates the ethical importance of situating the analysis and interpretation of narratives within the context in which they were constructed and collected. While this applies to the collection and analysis of qualitative data more broadly, narrative analysis is a person-centered approach that emphasizes the particularity of human experience [49], which requires careful attention to the social-cultural or political context [50] that influences those experiences. Similar concerns were present in case study 4 (online platforms), where narrative exchanges between forum contributors were situated in the unique interactional context of the forum itself, which was governed by a set of rules put in place and enforced by the service provider through its moderation team. Analyses of these narratives, therefore, need to account for this framework of rules.

The imperative of attending to the social context and the coconstruction of meaning in narrative data is further illustrated in case studies 1 (RISE trial) and 3 (poetic transcription). In case study 3, this coconstruction took the form of the creative interpretation of an individual’s experiences through the transcription process of translating narrative interviews into poetry. This process raised unique issues regarding the accuracy of representation in experiences (or truthfulness) and data ownership that are further discussed in the next section. However, case study 1 raised the issue of ensuring that researchers’ narrative analyses gave voice to their participants while acknowledging that knowledge generated through research is coconstructed through a multitude of perspectives. There was an ethical imperative to faithfully represent both the experience of participants and their sensemaking [51]. However, narratives were collected from Ethiopia—a collectivist society in which the focus is on group obligation and interpersonal harmony over and above autonomy and self-expression. In this context, the family, rather than the individual, has been described as the smallest autonomous unit for decision-making [52]. The neglect of collectivist values in understanding recovery in favor of individualistic principles has been highlighted [53].

The research team, therefore, faced a challenge in deciding whether and how to integrate the caregiver data into the analysis, balancing a person-centered approach with the contextual importance of caregivers. It is inherently undermining to seek “corroboration” from a caregiver and suggests that the researchers view their participants as “unreliable narrators.” Researchers should also not assume that details are unintentionally omitted from narratives by participants, but rather that the process of composing a narrative is a creative act of representing experience [54,55]. In this dataset, some sensitive and potentially stigmatizing details of an individual’s life were reported only by the caregiver. What is selected or performed through narratives reveals how an individual wants to be seen and is, in turn, shaped by the master narratives that researchers wish to interrogate. Furthermore, not only does the caregiver not have greater access to a more authentic “truth,” but they are also often distinctly nonneutral with respect to the individual’s story. Finally, overemphasis on caregiver perspectives may lead to skewed insights for future interventions that do not address the needs of people with severe mental illness. Taking together these ethical questions, the researchers decided to conduct the analysis primarily using the data of the person with severe mental illness. The caregivers were conceptualized predominantly as a key “voice” (ie, a character in the story rather than the teller of the story) [26].

Finally, we move to discussing the ethical and legal considerations involved in the publication and dissemination of interpretations and the public sharing of narrative curations with the broader scientific community or public. The ethical issue of data ownership pertaining to whom the data belong to—the participant or the research team—is applicable in several of our case studies but is perhaps best illustrated in case study 3 involving the poetic transcription of mental health narratives. In this context, the narrative is transformed into a new artistic or creative expression, such as a poem. The narrative itself is not disseminated; instead, the poem is. Elliott [56] suggests that the researcher enters, at least tacitly, into a moral relationship with participants that is not solely related to data collection but spans the whole process of research. Within this relationship, consideration should be given to questioning to whom the data belong [57] or, in this case, to whom the poem belongs. For example, in the poem shared in case study 3 (through Multimedia Appendix 1), it could be argued that both the researcher and the participant are the authors. These issues further reinforce the importance of the consent procedures and, ideally, engaging with participants about these issues early in the research cycle before the data have been collected.

In scientific disciplines, the context for case studies 1 (RISE trial) and 2 (NEON study), data sharing is becoming commonplace and more of an expected practice even with sensitive qualitative data. While a research team or study sponsor might own and be responsible for the original data they collect, some research funders and journals encourage these data to be transferred into a form where they can be shared with others for the greater scientific good but with suitable adjustments due to their sensitive content. The ethical and legal issues regarding anonymizing narrative data to protect participants’ identities before sharing them with a broader community outside of the research team have some unique challenges, as illustrated in our earlier discussion on the consent procedures for case study 2. The research team or sponsor must decide on appropriate anonymization measures and communicate these clearly to research participants in their consent procedures. While standard practice with qualitative data is to redact personal information to protect the identities of individuals, this poses some unique challenges with certain narratives, where individuals may view this process as a destructive retelling of their story.

When considering the dissemination of research findings and interpretations through research articles and presentations, several of our case studies raised the importance of engaging with ethical questions regarding the portrayal of truthfulness of human experiences and social realities. In case study 3, the ethical issue of truthfulness was represented in the fact that the construction of the poem arguably involved a partial loss of the original narrative. Poetic transcription does not attempt to include all elements of the interview transcript but rather focuses on the critical elements of the narrative [58], with the aim of representing a more embodied sense of the narrative [38]. In refining the data into stanzas, there is the potential to highlight aspects that might have been ignored when encountered in the data’s original composition [59]. Additionally, this process of poetic transcription may provide unique insights into participants’ narratives that might otherwise have been lost if the data had been subjected solely to thematic forms of analysis [60], wherein the search for shared or contrasting experiences across a group of participants may dissociate the narratives from the individuals [61]. However, there is also the potential for this poetic process to distort the original narrative or for aspects of the narrative to become overemphasized as researchers accentuate the transformative, subversive, or unique qualities of the narrative [62]. Furthermore, the readers themselves may become drawn into an affective fallacy where they conflate the content of the poem with their own emotive response [63].