How GenAI is Helping Reimagine Antenatal Care in A Low-Resource Setting: From Provider Enablement to Patient Empowerment

Amna Shahnawaz; Bilal A Mateen; Imaan Hameed; Maryam Mustafa

arxiv: 2604.22610 · v1 · submitted 2026-04-24 · 💻 cs.HC

How GenAI is Helping Reimagine Antenatal Care in A Low-Resource Setting: From Provider Enablement to Patient Empowerment

Maryam Mustafa , Imaan Hameed , Amna Shahnawaz , Bilal A Mateen This is my paper

Pith reviewed 2026-05-08 10:26 UTC · model grok-4.3

classification 💻 cs.HC

keywords maternal healthspeech AIelectronic medical recordspatient empowermentlow-resource settingsantenatal careWhatsAppPakistan

0 comments

The pith

A speech AI system in Pakistan shifts antenatal care by letting pregnant women generate their own EMRs and receive guidance on WhatsApp.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper traces the three-year evolution of Awaaz-e-Sehat from a clinician tool that turned Urdu speech into electronic records to a patient-facing platform where women voice their notes, get AI risk guidance, and share QR-coded data with any provider. It argues this addresses Pakistan's high maternal mortality from paper records, low literacy, and access barriers by making patients active generators and owners of their health information. A sympathetic reader would care because the approach proposes a model where limited clinician time no longer blocks continuous, informed care. The work shows how real-world constraints like linguistic nuance and infrastructure forced a redesign that centers patients in the decision-support loop. This turns static records into tools that promote self-advocacy and shared accountability rather than purely institutional documentation.

Core claim

In settings where clinicians face heavy workloads and patients encounter literacy and access hurdles, Awaaz-e-Sehat demonstrates that a WhatsApp-based speech AI can convert fragmented care into a continuous process by enabling women to create structured clinical notes, receive personalized antenatal guidance, and carry portable records that integrate them into clinical decision support.

What carries the argument

Awaaz-e-Sehat, the WhatsApp-integrated speech AI platform that converts voice inputs into structured EMRs while generating risk-based guidance for patients.

If this is right

Patients move from passive recipients to active participants who own and act on their antenatal data.
Providers gain access to up-to-date records from any location via QR codes without relying on institutional systems.
Care continuity improves in areas with fragmented paper records and high clinician turnover.
Decision support becomes shared rather than top-down, raising patient awareness of personal risk factors.
EMRs and clinical decision systems shift from static archives to dynamic tools supporting accountability between women and providers.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Similar voice-first designs could be adapted for other chronic conditions in low-literacy populations where follow-up visits are unreliable.
Policy makers might explore legal recognition of patient-generated records as official documentation to accelerate adoption.
Adding optional clinician review loops could be tested as a safeguard without undermining the patient-empowerment goal.
The approach highlights a path to reduce gender-related barriers by letting women control when and with whom they share health information.

Load-bearing premise

Women with low literacy can reliably use the speech interface to create accurate self-notes and follow AI guidance without new errors or clinician verification.

What would settle it

A field test in which a measurable share of patient-generated notes contain clinically significant inaccuracies or users misinterpret the AI guidance in ways that alter care decisions.

read the original abstract

Despite steady global advances, maternal mortality remains alarmingly high in Pakistan (155 deaths per 100,000 live births in 2023); largely as a consequence of fragmented paper records, low literacy, poor access to quality healthcare, and gendered barriers that compromise care continuity. Over three years, we designed, deployed, and iteratively developed Awaaz-e-Sehat, a speech-based artificial intelligence (AI) system that generates electronic medical records (EMRs) and supports decision-making in maternal health. The tool evolved from a clinician-facing AI assistant that automated Urdu speech-to-EMR generation into a patient-centred WhatsApp-based platform, enabling women to generate their own structured clinical notes, receive AI-generated antenatal guidance, and share QR-coded records with providers anywhere in the country. This case study documents that translational journey, i.e., how the ground realities of workload, linguistic nuance, and infrastructural constraints reshaped our design. The result is not merely a new method of record-keeping, but a reimagining of antenatal care and electronic medical records themselves. In settings where clinicians are time-constrained and have little institutional incentive to document, Awaaz-e-Sehat proposes a model of care that centres patients as active participants in generating and owning their health data. By keeping patients informed about their own risk factors and integrating them into the clinical decision-support loop, the system transforms EMRs and CDSS from static institutional artefacts into dynamic tools for self-advocacy and shared accountability in maternal health.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This is a three-year case study of shifting a speech AI tool from clinician notes to patient-generated antenatal records via WhatsApp in Pakistan, but the transformation claims rest on design description alone.

read the letter

The core of the paper is a grounded account of how Awaaz-e-Sehat evolved over three years. It started as a clinician-facing Urdu speech-to-EMR system and moved to a patient-facing WhatsApp platform where women can dictate notes, get AI risk guidance, and share QR-coded records. The authors walk through the practical constraints that drove the changes: clinician time pressure, low literacy, linguistic details in Urdu, and patchy infrastructure. That part feels honest and specific to the Pakistani setting, and it shows how those realities forced a patient-ownership model rather than a top-down one. Credit to them for documenting the iterative process instead of just presenting a finished product. The shift to letting patients generate and own their data is presented as a reimagining of EMRs and decision support, which is the main novelty claim. The abstract and description make clear this is not just another AI tool but an attempt to change accountability in a high-maternal-mortality context. What is actually new is the concrete trajectory in this low-resource environment, including the move to patient self-generation. The paper does a decent job laying out why standard clinician-only approaches fall short there. The soft spots are straightforward and fairly large. There are no usage numbers, no accuracy checks on the generated notes against clinician versions, no error rates for the speech interface with low-literacy users, and no measures of whether patients actually act on the AI guidance or see better continuity. The central claim that this turns EMRs into tools for self-advocacy therefore stays at the level of design narrative. The assumption that women can reliably produce accurate structured notes and use the guidance without new problems or mandatory verification is stated but not tested. That leaves the accountability shift unsupported by evidence. This paper is for readers in HCI, global health informatics, or AI deployment in low-resource maternal care who want to see how real constraints reshape a system. It is not for someone looking for outcome data or controlled evaluation. It deserves a serious referee because the deployment story is real and the context matters, even though reviewers will almost certainly push for metrics or a stronger limitations section. I would send it for review with the expectation that the authors add some validation data or clarify what the paper can and cannot claim.

Referee Report

2 major / 2 minor

Summary. The manuscript is a descriptive three-year design case study of Awaaz-e-Sehat, a speech-based GenAI system for maternal health in Pakistan. It details the evolution from a clinician-facing Urdu speech-to-EMR tool to a patient-centered WhatsApp platform enabling women to generate structured clinical notes, receive AI antenatal guidance on risk factors, and share QR-coded records. The central claim is that this reimagines EMRs and CDSS as dynamic tools for patient self-advocacy and shared accountability rather than static institutional records.

Significance. If the transformation claims were empirically supported, the work would offer valuable design insights for AI interfaces addressing literacy, access, and continuity barriers in low-resource maternal health settings. It documents real-world adaptations to workload, linguistic, and infrastructural constraints. However, the purely narrative presentation without metrics limits its contribution to speculative design implications rather than demonstrated improvements in accountability or care.

major comments (2)

[Abstract] Abstract: The assertion that the system 'transforms EMRs and CDSS from static institutional artefacts into dynamic tools for self-advocacy and shared accountability in maternal health' is unsupported. No quantitative data on patient-generated note accuracy, inter-rater agreement with clinician notes, usage logs showing independent patient action on guidance, or health outcome measures are provided to substantiate the reimagining or accountability shift.
[The case study description] The case study description: The manuscript does not report error rates, verification needs, or reliability metrics for the WhatsApp speech interface among low-literacy users, leaving the core assumption that patients can reliably generate accurate self-notes and act on AI guidance without introducing new errors or requiring clinician oversight unaddressed and load-bearing for the empowerment claim.

minor comments (2)

Consider adding a summary table of the iterative design phases, key constraints encountered, and specific changes made to the interface.
[Abstract] The title and abstract use 'GenAI' without initial expansion; clarify the specific generative AI components (e.g., models for speech-to-text or guidance generation) for technical readers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments on our design case study of Awaaz-e-Sehat. We clarify that the work is a narrative account of a three-year iterative design process rather than a quantitative evaluation study, and we address each major comment below with proposed revisions.

read point-by-point responses

Referee: [Abstract] Abstract: The assertion that the system 'transforms EMRs and CDSS from static institutional artefacts into dynamic tools for self-advocacy and shared accountability in maternal health' is unsupported. No quantitative data on patient-generated note accuracy, inter-rater agreement with clinician notes, usage logs showing independent patient action on guidance, or health outcome measures are provided to substantiate the reimagining or accountability shift.

Authors: We agree that the manuscript does not include quantitative metrics to empirically validate the transformation in accountability or care processes. As a descriptive design case study, the contribution centers on documenting real-world adaptations to workload, linguistic, and infrastructural constraints in Pakistan's antenatal care context, along with the resulting shift toward patient-generated records via WhatsApp. The reimagining claim is presented as a design implication arising from this evolution, not as a measured outcome. We will revise the abstract to explicitly frame the proposed model as an interpretive outcome of the case study observations, without implying empirical substantiation of improved self-advocacy or shared accountability. revision: yes
Referee: [The case study description] The case study description: The manuscript does not report error rates, verification needs, or reliability metrics for the WhatsApp speech interface among low-literacy users, leaving the core assumption that patients can reliably generate accurate self-notes and act on AI guidance without introducing new errors or requiring clinician oversight unaddressed and load-bearing for the empowerment claim.

Authors: The referee accurately identifies that no technical reliability metrics, such as speech recognition error rates or inter-rater agreement for patient-generated notes, are reported. The manuscript's scope is the translational design journey and how constraints reshaped the system from clinician-facing to patient-centered, including the choice of WhatsApp for accessibility among low-literacy users. We did not collect or analyze such metrics in this study. We will add a limitations subsection that directly addresses this gap, discusses the role of provider verification via shared QR-coded records, and notes the assumption of reliable patient action as an area requiring future empirical investigation. revision: yes

Circularity Check

0 steps flagged

Descriptive case study with no derivations, predictions, or self-referential steps

full rationale

The manuscript is structured as a three-year design case study that narrates the iterative evolution of Awaaz-e-Sehat from clinician-facing speech-to-EMR automation to a patient-centered WhatsApp platform. It reports design choices driven by workload, linguistic, and infrastructural constraints but contains no equations, fitted parameters, quantitative predictions, or load-bearing self-citations. The central claim that the system transforms EMRs and CDSS into tools for self-advocacy is presented as the outcome of the described design journey rather than a result derived from prior inputs by construction. None of the six enumerated circularity patterns are present; the paper is self-contained as an empirical design narrative.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on domain assumptions about healthcare infrastructure and patient capabilities in Pakistan without introducing new mathematical constructs or fitted parameters.

axioms (1)

domain assumption Fragmented paper records, low literacy, poor healthcare access, and gendered barriers compromise care continuity in Pakistan.
Stated directly in the abstract as the motivating context for the system design.

pith-pipeline@v0.9.0 · 5592 in / 1288 out tokens · 69502 ms · 2026-05-08T10:26:14.281356+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1]

Koyi Sawaal Nahi Hai

Meunier M, Guerin L, Morin L, et al. Barriers and facilitators to the implementation of clinical decision support systems in hospitals: a mixed-methods systematic review. J Am Med Inform Assoc. 2023. 13. BBC News. Pakistan shut down the internet — but that didn’t stop the protests. BBC News. November 25, 2024. Accessed December 14, 2025. https://www.bbc.c...

work page doi:10.48550/arxiv.2510.27401 2023
[2]

Your AI tool is subject to medical device regulation

Mateen BA, Ordish J, Reid MJ. Your AI tool is subject to medical device regulation. Nature Health. 2026 Jan;1(1):9-10. 25. Fuad M, Rahayu S, Halimah E, et al. Introducing a regulatory sandbox into the health sector: a scoping review. J Med Internet Res. 2023. 26. World Bank. Regulatory Sandboxes for Digital Health. Washington, DC: World Bank; 2025. 27. SA...

work page 2026

[1] [1]

Koyi Sawaal Nahi Hai

Meunier M, Guerin L, Morin L, et al. Barriers and facilitators to the implementation of clinical decision support systems in hospitals: a mixed-methods systematic review. J Am Med Inform Assoc. 2023. 13. BBC News. Pakistan shut down the internet — but that didn’t stop the protests. BBC News. November 25, 2024. Accessed December 14, 2025. https://www.bbc.c...

work page doi:10.48550/arxiv.2510.27401 2023

[2] [2]

Your AI tool is subject to medical device regulation

Mateen BA, Ordish J, Reid MJ. Your AI tool is subject to medical device regulation. Nature Health. 2026 Jan;1(1):9-10. 25. Fuad M, Rahayu S, Halimah E, et al. Introducing a regulatory sandbox into the health sector: a scoping review. J Med Internet Res. 2023. 26. World Bank. Regulatory Sandboxes for Digital Health. Washington, DC: World Bank; 2025. 27. SA...

work page 2026