JMIR Publications

ABSTRACT

Background:

Mobile health and artificial intelligence (AI) tools are increasingly used in perinatal care, yet their development has largely overlooked women with disability. Across different countries, women with disability continue to face inequitable perinatal care, including inaccessible services, stigma, and a lack of disability-specific information and support. As mHealth and AI tools become more popular in perinatal care delivery, the absence of disability-inclusive design risks deepening existing inequities rather than addressing them. No consensus exists on the features, functions, and ethical safeguards needed to make such tools accessible and relevant for this population.

Objective:

In this Delphi study, we aimed to establish consensus on important AI-assisted mobile app features for perinatal care of women with disability across the perinatal continuum, drawing on lived experience and interdisciplinary expertise.

Methods:

A three-round Delphi study was conducted with two panels: birthing people with lived experience of disability and multidisciplinary experts in perinatal health, AI development, and health informatics. Round 1 combined pre-identified functions from four published systematic reviews with open-ended questions. Round 2 used structured Likert-scale ratings 31 items across six domains: AI functions, information content, design and display, peer connection, safety and ethics, and affordability and access. Consensus was defined as ≥70% of participants rating an item as important or very important. Items that didn’t reach consensus in round 2 were reassessed in round 3

Results:

Nineteen women with disability and nine multidisciplinary experts completed Rounds 2 and 3. In Round 2, women with disability reached consensus on 27 of 31 items, while experts reached consensus on 28 of 31 items. In Round 3, two additional items reached consensus across both panels. Five items did not reach consensus after three rounds and will be revisited during prototyping and iterative app development. Both panels prioritized AI-assisted monitoring responsive to disability-specific needs, disability-affirming information, moderated peer connection, and strong safety and ethical safeguards. Free or very low-cost access received the highest consensus across both groups. The only item reaching consensus in the unimportant range was AI-based chatbots for mental health support, rejected by 70% of experts.

Conclusions:

This study provides a validated, experience-grounded feature set to guide the development of an AI-assisted perinatal health app for women with disability. Findings highlight the need for disability-affirming design, ethical AI governance, and meaningful engagement with disabled communities throughout app development. The results contribute to a nascent evidence base at the intersection of disability, digital health, and AI ethics in perinatal care.