Digital Mental Health Through an Intersectional Lens: A Narrative Review

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for the opportunity to review Digital Mental Health Through an Intersectional Lens: A Narrative Review. I really enjoyed reading it. The manuscript is clear, engaging, and very up to date, and it does a great job connecting AI in mental health with concrete equity and intersectionality concerns. Bringing together work on racial/ethnic minorities, LGBTQ+ communities, and neurodivergent people in this way feels both necessary and timely, and I see this as a meaningful contribution to the literature. In my view, the paper is suitable for publication after some minor revisions to sharpen a few points and clarify the methods, as outlined in my comments.

  Comments for the authors:

1) You say “This narrative review analyzes the current literature on digital mental health through an intersectional framework” and “we highlight several culturally responsive strategies to improve community outcomes,” but you never spell out what was actually found or what kinds of strategies. I suggest the authors to add 1–2 concrete clauses, for example after “we highlight several culturally responsive strategies…

2) You open with “The United States is suffering a profound mental health crisis” and then quickly move to “Globally, approximately 970 million people are living with a mental disorder” and “61.5 million individuals in the United States…” As a reader, I’m not entirely sure whether the review is primarily US-centric with some global references, or genuinely global in scope. You could add one clarifying sentence at the end of the first paragraph (around line 39).

3) The Methods section is currently quite high-level. For a narrative review, that’s acceptable, but right now it is hard to see what was actually included versus what was screened out. 

You state:

• “A total of 500 references were originally considered…”

• “Studies that did not contain appropriate methodology or design quality were excluded.”

• “The inclusion criteria focused on the impact of digital mental health on racial/ethnic minorities, LGBTQ+ individuals, and neurodivergent populations.”

But:

• We never learn how many studies ended up forming the core of the narrative synthesis.

• “Appropriate methodology or design quality” is vague.

• It’s also unclear whether opinion pieces and blogs (which you do include) were judged by different standards.

Consider adding one sentence stating the approximate number of empirical/primary studies versus conceptual or opinion pieces that were ultimately included (even a range is fine).

You should replace “appropriate methodology or design quality” with 1–2 examples, e.g.:

“We excluded studies that did not report basic methodological details (e.g., sample characteristics, analytic approach) or that relied solely on non-systematic anecdotal evidence.”

• Explicitly note that opinion pieces and major news/blog sources were included as contextual rather than primary evidence.

4) You say the narrative review was “conducted using an intersectional framework [27]” and later define intersectionality conceptually in Section 4.1. However, it is not yet clear how intersectionality affected the review process itself (beyond being a lens for interpretation).

5) Section 3.1 offers a strong, largely optimistic view of AI’s potential, then Section 3.2 suddenly shifts to bias. The logic is there, but the hinge is implicit. Consider adding, at the end of Section 3.1 (around line 131–133), bridging sentence explicitly pointing to the equity problem. 

6) In section 3.2, lines 143–152 and 156–162, you very effectively summarize historical disparities in psychiatric diagnosis and coercive interventions for Black and other minoritized patients, and then say: “Comparable disparities are now also evident in digital contexts” and move into digital tools. Right now, those two domains (offline vs. digital) are blended in the same paragraph. I suggest the authors to break this into two paragraphs: 1) one on historical and current clinical disparities (diagnosis, coercive interventions); 2) a second that explicitly shows how similar patterns re-emerge in digital tools, tying references [41–48] directly to digital processes (NLP misinterpretation, biased EHR data, speech-based models misclassifying Black women, etc.).

7) Section 4.3 “LGBTQ+”, lines 235–265 is strong but mostly treats “LGBTQ+ individuals” as a single group, even though the whole paper is intersectional. You already cite highly relevant work (e.g., WinoQueer benchmark, higher prevalence of anxiety, depression, suicidal ideation, and data on transgender youth). What’s missing is an explicit acknowledgment that queer and trans people of color, disabled LGBTQ+ people, and neurodivergent LGBTQ+ people may experience compounded bias from LLMs and DMHAs. You could implement 1–2 sentences near the end of the section noting that most current AI evidence focuses on LGBTQ+ as a single category, and that very little work has examined intersectional subgroups (e.g., Black trans youth or autistic queer people) in digital mental health systems.

8) In Table 1, WinoQueer is listed under “AI Mental Health App” alongside Wysa-Spanish, Shine, and Hazel. But WinoQueer is explicitly a benchmark for anti-LGBTQ+ bias in LLMs, not a DMHA or wellness app. That could confuse readers. Either rename the first column to something like “AI mental health–related tool / resource” or split the table into “Applications” versus “Benchmarks/Research tools.”

11) To ground the intersectional claims in empirical youth research, you might consider citing recent intersectional studies showing how multiple stigmatized identities cluster mental health risk among SGM and LGBTQ+ youth of color (e.g., Mereish et al., 2025; Gower et al., 2023; Amadori et al., 2025). This could fit well in Section 4.1 or 4.3: 

• Mereish, E. H., Abramson, J. R., Lee, H., & Watson, R. J. (2025). Intersectional Oppression-Based Stress, Drinking to Cope Motives, and Alcohol Use and Hazardous Drinking Among Sexual and Gender Minority Adolescents Who Are Black, Indigenous, and People of Color. LGBT health, 12(2), 125–133. https://doi.org/10.1089/lgbt.2024.0023
• Amadori, A., Real, A. G., Brighi, A., & Russell, S. T. (2025). An intersectional perspective on cyberbullying: Victimization experiences among marginalized youth. Journal of Adolescence, 97(4), 931–940. https://doi.org/10.1002/jad.12466  
• Gower, A. L., Rider, G. N., del Río-González, A. M., Erickson, P. J., Thomas, D., Russell, S. T., Watson, R. J., & Eisenberg, M. E. (2023). Application of an intersectional lens to bias-based bullying among LGBTQ+ youth of color in the United States. Stigma and Health, 8(3), 363–371. https://doi.org/10.1037/sah0000415

10) The limitations read as somewhat generic and do not fully exploit your intersectional framing. Consider adding one explicit point about publication bias and data gaps for intersectional populations (e.g., lack of studies specifically on Black neurodivergent people or Spanish-speaking trans users of DMHAs). Also, you could add one sentence acknowledging that most studies you cite examine single-axis categories (race or LGBTQ+ or neurodivergence), which limits your ability to draw strong conclusions about truly intersectional experiences.

 

Author Response

Hello! Please find attached template for revisions below. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Excellent work: this will bring valuable insights into digital mental health domains. I have a few suggestions for your consideration:

• In the introduction, at the beginning, you mentioned the United States. Could you please clarify in the title or abstract whether the focus is solely on the United States or the global context? If it is global, please start with some global information; otherwise, readers may become confused.
• Could you list some of these Smartphone-Based AI Chatbot apps and identify their role, such as CBT, dialectical behaviour therapy, positive behaviour support, and behavioural reinforcement?
• Line 131-133, you have mentioned that these tools eliminate the barriers. Could you explain how they eliminate these historical barriers?
• What steps have been taken to reduce it? If they have co-designed an algorithm or language model, please add a list of these co-designed algorithms.
• Could you please add a current limitation and future directions paragraph in this study, which might provide valuable insight for future research?

Author Response

Hello! Please find attached template for revisions below. 

Author Response File: Author Response.pdf