Demand for digital mental health interventions (DMHIs) has been increasing in many countries [1]. For example, 76% of the Australian general public indicated an interest in using DMHIs [2], and US $2.4 billion were invested in these interventions in 2020 alone in the United States [3]. A Singaporean mental health app, Intellect, is now used by over 3 million people in 20 countries, mostly in Asia [4]. A large-scale education program about ethics in DMHIs has been launched [5], and more active use of DMHIs is expected in Africa [6]. The advantages of DMHIs include accessibility, cost-effectiveness, and personalization [7], which can address key barriers for mental health recovery, such as low help-seeking and the stigma associated with mental health problems [8]. Effectiveness of DMHIs has been generally reported in diverse populations (eg, children, young people, older adults, university students, health care workers, people with neurodevelopmental disabilities) and on diverse mental health experiences [2,9-12]. DMHIs have become an important domain in medical internet research [13-15].

Mental health recorded recovery narratives (RRNs) have been used in DMHIs. Mental health recovery narratives can be defined as “first-person lived experience accounts of recovery from mental health problems, which refer to events or actions over a period of time, and which include elements of both adversity/struggle and of self-defined strengths/successes/survival” [16]. Although narratives can be shared in person, such as when a peer support worker tells their story [17], RRNs are those presented in recorded formats such as written text, audio, and video. RRNs are often disseminated online [18] and have been used in a range of clinical and societal interventions [19]. For example, the Narrative Story Bank was created by the Scottish Recovery Network to inspire hope and offer tools and technologies for recovery [20]. Content from the Narrative Story Bank helped to inform Scotland’s mental health strategy [21]. In Scotland, the use of recovery narratives has been incorporated into clinical practice, such as the production of written RRNs to support self-reflection [21]. Similarly, RRNs have been used in clinical training to enhance communication skills and empathy of health care workers [22]. RRNs are used in national antistigma campaigns [23], which aim to reduce stigma associated with mental health problems [24].

The impact of RRNs on recipients has been evaluated. For example, students who listened to recovery narratives of people with anorexia nervosa showed more understanding of the mental illness and a less stigmatized view toward these individuals [25]. Eating disorder memoirs offered validation of personal experience and created a positive affective response among people with eating difficulties [26]. Helpful outcomes of accessing both live and recorded recovery narratives include connection with others, validation of own experiences, empowerment, hopefulness, gratitude, and stigma reduction [27]. Mechanisms of impact from RRNs include comparison with the narrative and narrator, learning about the experiences of others, and feeling empathy, all of which create a sense of connection [28]. Accessing RRNs can thus support personal recovery by increasing hope, meaning, and a sense of fulfillment [18,29].

RRNs are often grouped and presented as collections [19], which have been provided on bespoke websites [30] and in books composed of recovery narratives linked by a theme such as psychosis [31] or men’s eating disorders [32]. The people who assemble RRN collections do so for a range of reasons, including providing education about mental health and recovery, supporting others in their recovery journey, and campaigning for change in health service provision [33].

Clinicians who use RRNs as part of treatment need to know the possible impact of the RRN collections they recommend to their service users [34]. Given their broad range of clinical applications, the ability to characterize a narrative collection is important because not all RRNs are helpful to all people. An interview study reported that recovery narratives can be harmful if recipients feel disconnected from the narrative, resulting in distancing themselves from the narrator [27]. For instance, if a “right” way to recover is communicated by the RRNs, this implicit message can cause a recipient to feel disconnected [35,36] if that particular pathway is in some way “outside” that person’s scope, experience, and perspective. A feeling of being disconnected is detrimental because connection is the central mechanism supporting recovery after access to RRNs [28,37].

Providing diversity in RRN collections may be one approach to maximizing benefit and minimizing harms, enabling a sense of connection to people with different backgrounds and experiences. A diverse set of narratives will increase the opportunity for the recipient to feel connected to a narrative [27]. Further, a lack of diversity in RRN collections may give rise to stereotyping [38], leaving recipients feeling excluded and disconnected [18], or suggest that there is only one way or a “right” way to recover [35,39], which may not fit with their own beliefs, experiences, and needs. Recipient characteristics and narrative characteristics moderate the impact of a narrative on a sense of connection [28]. For example, people in ethnic minority groups felt a lower level of connection when an RRN collection marginalized these groups [40]. Diverse narratives are needed to ensure that recipients from different backgrounds and with different identities have the greatest opportunity to feel connected with both a narrative and a narrator, as both types of connection are known to maximize the positive effects of RRNs [38]. One exception might be collections of RRNs targeted at specific groups experiencing structural inequalities or marginalization (eg, refugees), where the selection of narrators who belong to those groups might be an appropriate strategy for maximizing beneficial impact.

Two types of variation in a narrative collection can be differentiated: diversity and inclusivity [41]. Diversity is defined as heterogeneity in narrative characteristics, such as a spread of narrator demographics and “protected characteristic” identities, as defined by the Equality Act 2010 (eg, age, gender, sexual orientation), or types of narrative content (eg, trajectory, genre). Inclusivity is defined as representativeness of narrative characteristics in relation to a broader population [42].

Diversity and inclusivity concepts can be applied to a specific characteristic of an RRN collection, such as narrator age or a narrative focused on a certain demographic characteristic (eg, race, disability). Diversity is present when there are a meaningful number of narratives within each subcategory [43]. For example, diversity in narrator age is present when there is a spread of younger, middle-aged, and older narrators. Since diversity is a feature of the collection, a diversity metric will be a constant for a given collection. By contrast, inclusivity is present when the proportion of narratives within each subcategory is similar to the proportion in a wider comparison population at a given time [44]. For example, inclusivity in narrator age is present when there are similar proportions of each narrator age group in the collection and in a comparison population, such as other people on the caseload of the mental health service or in the general population. Since inclusivity is a function of the collection when used in a specific context, an inclusivity metric will vary based on context.

We argue that assessing diversity and inclusivity of narrative collections is the next frontier in mental health recovery narrative research and practice. A tool to characterize individual RRNs has been developed. The Inventory of Characteristics of Recovery Stories (INCRESE) is a standardized 77-item instrument characterizing narrative mode, narrator and narrative characteristics, content warnings, turning points, and narrative content [45]. Although INCRESE is used to characterize individual RRNs, no method currently exists to assess the diversity and inclusivity of an RRN collection.

We here discuss aspects of measuring diversity and inclusivity in recovery narrative collections. The Narrative Experiences Online (NEON) study used INCRESE to characterize a large collection of 687 recovery narratives. Because we had a large data set of narrative characteristics measured using INCRESE available, we used this data set for the present analysis.

The NEON study received approval from a UK National Health Service Research Ethics Committee (West London and GTAC, 18/LO/0991).

We present one option for measuring diversity and two options for measuring inclusivity for recovery narrative collections. The two options for measuring inclusivity have different properties, enabling people assessing inclusivity to make a choice over which to use.

The Simpson Diversity Index (SDI) is an established index used in the natural sciences to assess biodiversity [43]. The SDI considers the number of species present and the abundance in each species to indicate the variance in species. The SDI is calculated by deducting the Simpson Index (SI) from 1, where SI=Σn(n–1)/N(N–1). When used to assess narrative collection diversity in relation to a particular characteristic (eg, “narrator gender”), n refers to the total number of narratives within each option of the characteristic (eg, “female narrator”) and N refers to the total number of narratives across all options. See Multimedia Appendix 2 for example calculations. The SDI ranges from 0 (low diversity) to 1 (high diversity).

Inclusivity captures the extent to which minority groups in a comparison population (eg, a country population, a service user cohort at one mental health service) are included in a collection [51]. One approach to assess inclusivity is to identify the categories that are minoritized in the comparison population and compare their proportion in the narrative collection. For example, the Origin characteristic is measured by INCRESE item 13 “Ethnicity,” with categories of “Not identifiable,” “Asian,” “Black/African/Caribbean,” “Dual/multiple ethnic group,” “Other ethnic group,” and “white.” When used in the United Kingdom, all choices apart from “white” are minority groups (a limitation of this categorization is that some “white” communities such as the Traveler, Gypsy, and Roma communities are also very socially excluded). In the United Kingdom, 13% of the population are nonwhite [52]. In a collection, among all narratives, if the ratio of nonwhite narratives is higher than 13%, the collection can be considered as inclusive with respect to ethnicity in the United Kingdom. If inclusivity is being assessed in a different comparison population such as a different country, then alternative choices for minority categories would be made.

A second approach to measuring inclusivity is to set an arbitrary benchmark. One approach is to decide that five narratives are sufficient to satisfy a benchmark for a certain characteristic. A stronger approach, as often used for external examination in the university sector [53,54], is to decide both a minimum rate (eg, 10%) and number (eg, 5), and choose whichever is greater.

The three methods above were applied to the NEON Collection as a case study. Four characteristics matching INCRESE items were considered: Sex and gender, Origin, Disability, and Positioning. The NEON Collection is a curated collection of mental health RRNs. All narratives are included in the NEON Collection with permission [55]. Each narrative is characterized using INCRESE by multiple raters [45].

The candidate approaches were applied to four groups: the entire NEON Collection, two of the larger collections chosen for difference in source, and the individual donations contained in the NEON Collection. In September 2022, the NEON Collection comprised 687 narratives compiled from 34 public collections and from individual donations. One of the two larger collections, which we here refer to as “statutory service” (78 narratives), was compiled by a statutory mental health service. The other, which we refer to as “ethnic minority book” (19 narratives), was published as a book focusing on the mental health of ethnic minority groups. Individual donations (n=29) comprised narratives collected directly from individuals as donations to the NEON Collection.

To evaluate diversity using the SDI, because the Origin and Disability characteristics consist of multiple INCRESE items, the mean SDI scores were calculated. The SDI scores for each collection group are shown in Table 4.

Of the four groups assessed, the NEON Collection is the most diverse in terms of Positioning, Origin, and Disability. Individual donations are the most diverse in terms of Sex and gender. Statutory service is the most diverse in the Disability subcategory of cognitive difficulties. Ethnic minority book is the most diverse in the Original subcategory ethnicity.

To evaluate inclusivity using option 1 (demographic rates), the comparison population used was the UK general population. The Positioning characteristic was excluded as there are no demographic data of narrative positioning available. Minority groups in each characteristic were identified and the proportion of narratives from each minority group was calculated. The proportions of minority groups in the comparison UK general population were obtained for gender [56], ethnicity [52], and each Disability component [57-61]. For location, the number of British nationals living outside Europe was identified and then the proportion against the UK population was calculated [62]. For self-care, the prevalence of self-neglect was identified [61]. The findings are shown in Table 5.

The NEON Collection met the inclusivity benchmark of being above the UK general population for the location subcategory in the Origin characteristic and for two subcategories in the Disability characteristic: cognitive difficulties and self-care. Likewise, statutory service met the inclusivity benchmarks for cognitive difficulties and self-care. Ethnic minority book met the inclusivity benchmarks for ethnicity in the Origin characteristic and self-care in the Disability characteristic. Individual donations met the inclusivity benchmark for cognitive difficulties. No collections met the inclusivity benchmarks for the Sex and gender characteristic or the Disability subcategories of visual difficulties, hearing difficulties, and mobility. The largest inclusivity score was for ethnicity in the ethnic minority book, followed by location in the NEON Collection.

To evaluate inclusivity using option 2 (arbitrary threshold), benchmarks were set at a minimum proportion of 10% of the number of narratives in a collection and a minimum number of 5 narratives. Whichever was the higher number was used as the threshold, as shown in Table 6.

The NEON Collection met the inclusivity benchmarks for both ethnicity and location in the Origin characteristic and for the Positioning characteristic. Ethnic minority book met the inclusivity benchmark for ethnicity in the Origin characteristic. Neither statutory service nor individual donations met any of the characteristics or subcategories. Similar to inclusivity option 1, the scores for location in the NEON Collection and ethnicity in the ethnic minority book markedly exceeded the benchmarks.