Advancing equity with technology and innovation in obstetrics: A critical approach to obstetric TechQuity

Abstract

Innovative technology, including wearable devices, remote patient monitoring, smartphone‐based applications, and artificial intelligence, has often been suggested as a potential solution to addressing racial disparities and the maternal mortality crisis in the United States. The rapid pace of development of new technologies without deliberate efforts to ensure equity may in fact amplify rather than alleviate disparities. “Techquity” is a concept that encompasses “strategic development and deployment of technology to advance health equity” and has not been well explored as it relates to innovation in obstetrics. While advanced technology can be a powerful tool for achieving equity in healthcare, the irresponsible use of this technology can have devastating consequences for vulnerable communities. In this article, we highlight some of the more commonly used new technologies in obstetric care, examine potential pitfalls, discuss approaches to critical evaluations and proposed solutions, and explore the role of community partnerships in successfully advancing obstetric TechQuity.

1. INTRODUCTION

Severe maternal morbidity and mortality are increasing at an alarming rate in the United States. While all other developed countries have seen a decline in maternal mortality rates, the United States has seen a consistent increase [1, 2]. The etiology of this rise is multifactorial, likely related to advancing maternal age and medical comorbidities and the concomitant limitations of the healthcare system, including limited access to high‐quality obstetric services. The urgency of these trends is amplified by the persistent racial and ethnic inequities in severe maternal morbidity and mortality. These data reinforce the need for a call to action to prioritize improvements in maternal health.

Innovative technology, including wearables, remote monitoring, smartphone‐based applications and artificial intelligence, has often been cited as a potential solution to the maternal mortality crisis in the United States. However, the complexity and rapid pace of development of these technologies without a conscious, deliberate effort to ensure equity may in fact amplify rather than alleviate disparities. The concept of “Techquity” or the “strategic development and deployment of technology to advance health equity” has not been explored in detail as it applies to obstetric innovation [3, 4]. TechQuity calls for accountability for active promotion of health equity and emphasizes the potential role of technology in amplifying disparities in health. It also calls into question the assumption that technological advances can uniformly benefit health without a paradigm shift centering on equity. While advanced technology can be a powerful tool for achieving equity in healthcare, the irresponsible use of this technology can have devastating consequences for vulnerable communities. Figure 1 outlines a framework to approach the development and implementation of new technology in order to assure equity is being centered in any new innovation.

FIGURE 1.

A framework for approaching techquity when developing or implementing new technology or innovations. Guidance is provided for each domain of the innovation lifecycle to support the desired outcome of advancing equitable innovation. Figure adapted from National Academy of Medicine. Toward Equitable Innovation in Health and Medicine: A Framework [5].

In this Narrative Review, we highlight some of the more commonly used new technologies in obstetric care and examine potential pitfalls, approaches to critical evaluations, proposed solutions, and the role of community partnerships in successfully advancing obstetric TechQuity. This is not meant to be a comprehensive, systematic review, but instead highlights some key examples of technologies used in modern obstetric practice and their potential pitfalls through a Techquity lens.

1.1. Smartphone applications

At first glance, smartphone applications (apps) may be an ideal mechanism to improve TechQuity. Indeed, smartphone ownership in the United States is nearly universal among reproductive‐aged patients (95% of individuals aged 18–49 years own a smartphone) [6], and the digital divide of smartphone ownership continues to decline in terms of income and rural versus urban settings [7, 8]. However, apps can contribute to digital inequity for many reasons. First, not all those who own smartphones can afford consistent data plans or home internet [9, 10]. Second, the majority of apps—in particular mental health‐related apps—that claim to improve outcomes are not supported by feasibility data, much less efficacy data [11]. Third, many health‐related apps are in English, which widens the digital divide. Lastly, familiarity with apps—or any technology‐based health intervention—may differ among people with less health literacy.

For those of us who create perinatal apps, there are clear steps we can take to improve equity in the perinatal app space. First, we must formally incorporate target end‐users’ perspectives via qualitative research in the development process. Ensuring that the app is optimized according to the perspectives of those who will eventually use it—and not the opinions of those who create it—will increase user engagement with the app, likely by increasing its user‐friendliness [12]. Second, just like any intervention, apps must be evidence‐based. For example, perinatal mental health apps that included psychotherapy have been shown to be more effective and to have higher user engagement than those with psychoeducation alone [13]. Third, for any app endorsed by a health team, it is the responsibility of the healthcare team to ensure patients understand how to use all app features and, ideally, engage longitudinally to ensure ongoing app usage. Lastly, apps should be available in languages other than English and consider approaches which differ from the standard Western framework. This will require qualitative research conducted in other languages and, potentially, apps with variable content based on feedback from non‐English speaking patients.

We recognize that most perinatal healthcare providers will not create a perinatal app. Nevertheless, we can still help our patients overcome digital inequity. We can recommend specific apps for our patients to use, either by conducting our own research or rely on experts to do so. For example, researchers have published lists of preferred apps for pregnant people and for those looking to learn more about contraception [14, 15]. We can also recommend apps that do not require users to use data or Wi‐Fi to access the app and encourage our patients to utilize the complimentary Wi‐Fi in our clinics (or a local coffee shop) to download the recommended apps without fees. Finally, as noted above, there are multiple apps available related to perinatal health with very few of them supported by robust evidence. Prior to recommending the use of an app for perinatal health, clinicians should critically evaluate the available evidence to support such an app. These efforts seem simple but, collectively, can help all our patients receive app‐based benefits, improving TechQuity.

1.2. Remote patient monitoring

Remote patient monitoring has been deployed broadly across obstetrics, including for management of diabetes, hypertension, and antenatal surveillance. There is robust evidence both in obstetric as well as non‐obstetric populations to support the use of remote monitoring for hypertension and diabetes [16, 17, 18]. These interventions are designed to serve as a bridge between in‐person care and in‐hospital care. On the surface, the use of remote monitoring should overcome many barriers to care, including transportation, childcare and in the case of postpartum care, lack of parental leave. There are specific considerations that need to be intentionally addressed prior to implementation of remote monitoring to promote equity. The cost, validity and accessibility of any remote technology must be critically examined.

For example, in the case of home blood pressure (BP) monitoring, there are few devices that have been validated in pregnancy, which is considered a special population for the purposes of medical device development and validation [19]. The American Heart Association maintains a validated device listing (validate.org), which includes only 3 devices for home BP monitoring in pregnancy with a cost per device ranging from $68 to $129 [20]. If devices are not provided to pregnant or postpartum individuals or not covered by insurance, individuals may purchase a less expensive, and potentially less accurate device. In fact, many automated BP cuffs that are not validated for use in pregnancy may underestimate BP, with significant implications for pregnant and postpartum people and their fetuses [21]. In considering accessibility, the platform from which these data are collected is another consideration, that is, if Bluetooth technology is being used, are individuals being provided smartphone or tablets or is access to reliable broadband internet for transmitting data assessed prior to enrollment? Any remote monitoring programming should also be available to individuals across all languages spoken in the community. By limiting inclusion to only English‐speaking people, these technologies only serve to widen disparities. Interventions may be more effective and equitable if they are designed to be patient‐centered, which is often not the culture in health technology innovation [22].

We can overcome these limitations, by intentional design, development, and deployment of remote monitoring technology keeping in mind these key pitfalls. Intentional, prospective validation of new devices in a pregnant and postpartum population is important to advance equity and ensure new technology is valid and accessible to all. Further, advocacy at the local, state and federal level is critical to develop more equitable pay structures for remote monitoring during pregnancy and postpartum to ensure access across all geographic and income strata.

1.3. Wearable technology

Smartphone‐based wearables have the potential to transform patient monitoring and surveillance during pregnancy and postpartum by providing objective, continuous, and remote monitoring of patients’ vital signs, biomarkers, and health behaviors. Wearables could be used to inform patients of their health status on a day‐to‐day basis or help them recognize early signs of complications [23, 24, 25]. Furthermore, wearables could reduce challenges for clinicians who at times only see patients once per month – or often less in the case of socioeconomically disadvantaged patients. As Obstetrics continues to investigate the benefits and challenges of wearables, it is important that we establish equitable development, investigation, and implementation of wearables in our clinics, which means holding clinicians, researchers, and companies accountable for TechQuity for wearables in Obstetrics. In non‐pregnant populations, wearables have been shown to be effective in promoting physical activity and detection of arrythmias [23, 24]. In obstetrics, the data is more limited. There is a lack of robust, randomized controlled trials, and many studies have been performed in controlled environments [25, 26].

There are a myriad of ways that inclusion, acceptability, and access to wearables may be obscured from various clinical populations, including, but not limited to, socioeconomically disadvantaged, physically or mentally disabled, racial/ethnically diverse groups, and co‐morbidities. Although each group may be affected differently there are a handful of questions that can be asked to jumpstart conversations about TechQuity with companies and engineers, to ensure appropriate validation and evaluation for research teams, and to prevent disenfranchisement and inadvertent harm by clinical care teams. These questions are outlined in Table 1. Although this is not an exhaustive list of patient groups to consider or questions to ask in TechQuity discussions, this is a good place to begin a discussion. Ultimately, if we, strive for TechQuity in Obstetrics, then possibilities for wearables are endless and we are on the right path.

TABLE 1.

Key questions to assess Techquity in regard to wearables in obstetrics.

Key question	How to assess
Was the wearable and/or its sensors tested and validated in a diverse cohort of participants?	Various ages, races/ethnicities, body compositions, physical abilities statuses, and comorbidities should be included, particularly those that reflect our patients.
What type of sensors are used in the wearable?	Given that optical sensors are less effective on darker skin tones, bioimpedance sensors assume symmetry that may be absent in some with disabilities or obesity, and temperature sensors may create bias in some with obesity as the most distal regions will have decreased temperatures, it is important to understand, which sensors are used, and in which metrics the sensor data are included.
What are the direct and indirect costs of the wearable?	It is important to consider direct cost such as the wearable device itself and subscription costs. However, indirect cost such as in‐home wi‐fi or cellular service should be considered as costs. Patients who are employment‐ or housing‐ insecure should be considered.

1.4. Artificial intelligence

Artificial intelligence (AI) is defined as the science and engineering of making intelligent machines [27]. The development of AI technologies for use in medicine and healthcare has increased dramatically in the last ten years with now 1017 FDA‐cleared AI and machine learning (ML)‐based technologies as of March 2025 [28]. The use of AI in Obstetrics is relatively nascent and has focused on advanced ultrasound imaging techniques, fetal heart rate interpretation, and clinical predictive modeling [29, 30]. While both imaging and cardiotocography have the potential for exacerbating inequities in care, our principal focus in considering Techquity will be in the development of predictive models.

Predictive models, whether driven by advanced deep learning algorithms or simple regression techniques, provide risk estimates for the presence of disease (diagnosis) or an event in the future course of disease (prognosis) for individual patients [31]. Developing a clinical predictive model requires choosing a problem, identifying and weighting predictive factors, then using those factors to discriminate between different target outcomes. Predictive modeling in Obstetrics merits special consideration, since the unique physiologic changes of pregnancy have yielded numerous examples of models which were developed and proven effective in non‐pregnant adults faltering when applied to pregnant persons [32]. Even before the advent of AI, critiques of prediction models and decision aids in medicine have been plentiful [33]. Illustrative of how these may increase inequity is the original vaginal birth after cesarean (VBAC) calculator [34, 35]. Historical inequities driven by structural racism have contributed to an excess of cesarean deliveries being performed on Black and Hispanic patients [36]. As a result, a calculator derived from this unequal treatment then recommended in excess that those identifying as Black or Hispanic would require a repeat cesarean. The original VBAC calculator (since revised) thus also represented a potential source of a positive feedback loop in which unequal past treatment yields unequal recommendations thus creating and mathematically validating further unequal treatment [35, 37]. What makes AI‐driven algorithms far more dangerous, though, is that this potentiation of bias may be opaque. An AI algorithm either by intent or by design can often be a “black box” in which predictive features cannot be unpacked in the same way as a simple score or calculator.

Inequities in obstetric outcomes extend beyond race, encompassing factors such as socioeconomic status, English proficiency, rurality, and insurance status [38, 39, 40, 41]. Additionally, historical mistreatment is one potential source of algorithmic bias among many including other sources of statistical bias (representation bias, aggregation bias), measurement bias, and evaluation bias. Despite these multiple sources, the solutions to date are few [42]. It is necessary then that as clinicians we demand greater transparency in the creation and application of AI‐based models. In the same way in which we would not tolerate the introduction of a new drug into our pregnant patients without a comprehensive understanding of the mechanism of action, the pharmacodynamics, and the clinical outcomes from its use, we must expect similar comprehensive evaluation for the implementation of AI‐driven algorithms into the care of our patients. Further, we must recognize that there are often trade‐offs inherent to the development and mitigation of bias in these models: optimizing for perfectly calibrated predictions along lines of race and ethnicity might yield different false positive results that inappropriately withhold treatment for certain groups [43]. Choosing the “right” targets and metrics is ultimately a value judgment, and clinicians and patients must be at the helm in making those choices.  Only by creating such expectations can we ensure Techquity in the application of AI to our patients' care.

1.5. Natural language processing

Clinical documentation via “free text,” or unstructured text, is used to record details about patient visits and encounters with the health system (e.g., clinical notes) and provide interpretation of labs and studies (e.g., radiology reports). Now predominately stored in the electronic health record (EHR), these notes represent a data source that can be harnessed for research, quality improvement, and clinical operations. This free‐text data can be analyzed using natural language processing (NLP), which converts text into a data format that can be analyzed [44, 45]. NLP analyses often utilize machine‐ or deep‐learning methods, given the quantity of data generated and the complicated relationships (e.g., negation) formed between words and phrases in free text.

While NLP is widely used in other industries and applications, investigators have begun to examine its use in clinical medicine. In a 2020 systematic review of NLP of EHR data by Spasic and Nenadic, the authors identified 110 studies using free‐text clinical notes and found that most studies used relatively small data sets for training due to the “annotation bottleneck” (e.g., the process of individually labeling outcomes) required for many supervised machine learning techniques [46]. The authors noted that NLP was being used in a wide range of applications, from disease phenotyping to predicting prognosis [46]. There are fewer studies examining NLP in obstetrics; some examples include the identification of individuals who are candidates for low‐dose aspirin for preeclampsia prevention and for identifying individuals at high risk for severe maternal morbidity during the delivery admission [47, 48, 49, 50]. More applications are likely as these tools become more easily embedded into EHR systems and more user‐friendly, requiring less technical expertise to build and implement. Furthermore, as the technology is rapidly advancing, text analysis via large language models (LLM) presents both promise and challenges (i.e., hallucinations, inaccuracies) for use in clinical medicine [51, 52, 53]. It is imperative for clinicians to be informed and aware of the abilities and limitations of these emerging technologies [47, 48, 49, 50].

Uniquely, NLP of clinical documentation presents challenges and opportunities related to health equity. Like many other forms of AI, inherent inequities in the data used to develop or train an NLP model will be reflected and perpetuated in future applications [54, 55]. For NLP, these inequities may manifest in the components of clinical care that are documented and individual documentation practices (e.g., words, syntax, frequency, length), mainly if documentation practices vary among different subgroups of the population (e.g., sociodemographic variables, disease states). Furthermore, inequities can be introduced if certain groups are inherently underrepresented in the development data, such that accurate or meaningful predictions cannot be made (e.g., lack of clinical documentation for certain groups in the EHR because of underlying variations in care‐seeking patterns due to structural barriers to access, like insurance coverage). The benefits of these technologies may not apply to all groups, and in fact NLP systems may reproduce stereotypes and reinforce linguistic stigmatization and discrimination [56].

As NLP‐based models enter clinical practice and in consideration of TechQuity, it is essential to understand how and why a particular model was developed and its ultimate purpose [42]. Much like all AI algorithms, end users of NLP models should critically evaluate the nuances of the data used to derive and train the model, reflecting how similarities or differences in their patient characteristics may support or challenge the generalizability of its application. External validation of the model outside of the training data is imperative to demonstrate scalability. Among groups with known inequities in health outcomes, the predictive performance of any model should be examined to understand if it may be reinforcing disparities or inequities. Last, in considering TechQuity, NLP models that show promise to improve health outcomes should not be limited to health systems or hospitals with the financial and technical resources to implement them; they should be available to all clinical providers and patients to improve maternal and neonatal health outcomes at a population level.

1.6. Building community partnerships to advance TechQuity in obstetrics

It is imperative that when new technology solutions are developed as outlined above, that we challenge traditional methods of research which largely have not included the voices of communities, especially historically marginalized communities. By employing multi‐faceted, interdisciplinary, and community‐based methodologies in crafting tech solutions, we can systematically address the root factors fueling and intensifying inequities. This strategic approach holds the potential to bring about tangible enhancements in the maternal health crisis.

As researchers we must ensure that community voice is centered and amplified in the development of these innovations. We can do this through research approaches like community‐based participatory research, human centered design (HCD) and/or equity‐centered design (ECD) [57, 58, 59]. These research approaches are collaborative and focus on users’ (community), needs and requirements. However, ECD diverges from HCD through its deliberate selection and engagement with the design target. In the design process, careful consideration is given to the lived experiences of the individuals for whom problem‐solving is undertaken. ECD aims to mitigate assumptions by fostering increased representation throughout the design process, while also acknowledging and addressing systems of oppression that historically marginalized populations have faced. The collaborative design approach of ECD actively involves community members from the project's inception, promoting inclusive participation. The goal is to transcend traditional approaches, placing community/users at the forefront and ensuring their active involvement throughout the design journey.

The initial step in any participatory process focused on users/community is building community partnerships. Building community partnerships demands resilience and adaptability. It requires a high level of commitment, flexibility, and patience. Reflect on both your personal strengths and limitations, as well as those of your organization. Evaluate the advantages your organization stands to gain, the benefits extended to the community, and any supplementary concerns that warrant consideration. Second, it is essential to approach and involve community. This means identifying who the community is, understanding the strengths, assets and resources of that community, respecting the history that came before the project, connecting with gatekeepers, and meeting with community members. Third, you have to formalize and sustain the partnership. This includes defining what partnership looks like and developing structures and processes. This may also require building community capacity to work together. Building community partnerships is not easy. Expect and embrace challenges as integral to the partnership‐building process; they are inherent in creating something of value.

Building strong community partnerships begins with identifying your target community and recognizing who already holds trust and influence. Partnering with trusted leaders, grassroots organizations, and faith‐based groups can amplify efforts through shared decision‐making and relationship‐building [60]. Partnerships with research institutions, in particular, can provide community organizations with credibility, access to funding and expertise, and expanded capacity to meet community needs [61]. Concrete strategies for building partnerships include conducting asset mapping to identify key organizations, leaders, and networks; hosting listening sessions and community forums to gather insights; and establishing advisory boards or co‐leadership structures that engage community members in setting agendas, designing interventions, and evaluating outcomes [60, 62]. To build sustainable partnerships, it is crucial to compensate community partners, involve them directly on project teams, share findings transparently, and prioritize long‐term relationships over short‐term goals. Platforms for collaboration include CBPR networks, faith‐based coalitions, and local health alliances [63]. Critically, the first step is selecting partners thoughtfully—choosing those with authentic community ties and a shared commitment to equity.

1.7. Reimbursement equity

Existing healthcare systems and reimbursement models have not kept pace with the new technologies that are being developed and deployed within patient care spaces. For most health technologies, patients or physicians are not directly purchasing a product, instead access is dictated by healthcare organizations or insurers. The decisions of these organizations and insurers about purchasing and using new technologies therefore impact access and the cost. However, neither of these entities have strong incentives or mechanisms in place to establish and use equity‐focused metrics in these decisions [5, 10]. One avenue to expand this would be through advocacy, as insurers’ coverage decisions are subject to both state and federal law.

To demonstrate these concepts, we highlight the example of postpartum remote blood pressure monitoring. As noted above, there is robust data to support improved outcomes with the use of remote blood pressure monitoring in the postpartum period after a hypertensive disorder of pregnancy [18]. These programs reduce unplanned readmissions and emergency room visits have been shown to be cost saving in prior analyses [64]. Increased participation is cost saving to health insurance companies, however, most health insurance companies are not involved in the payment structures for these types of programs, as most programs to date have been funded by healthcare organizations or extramural grant funding [65]. Wide implementation of these types of programs would require overhauling the financial models that support them. Advocacy from organizations such as the American Heart Association and the American College of Obstetricians and Gynecologists (ACOG) has led to passage of state legislation mandating coverage of BP devices and self‐monitored programs in at least 21 states to date [66]. The lack of federal or state policies on a given technology plays a role in affecting its access and use to certain populations, and equity is unlikely to be systematically incorporated into these types of decisions by healthcare organizations or insurers until reimbursement frameworks require integration of these criteria.

2. CONCLUSION

We have highlighted here some of the commonly used new technologies in obstetric care and examined limitations, approaches to critical evaluations and proposed solutions to overcome those limitations to successfully advance obstetric TechQuity. This list is not meant to be comprehensive and is growing at a rapid pace. For example, other new technologies to be considered and critically evaluated in obstetrics include home fetal monitoring, remote ultrasound technology and the use of continuous glucose monitors for gestational diabetes. To move the needle on development and implementation of technology that reduces maternal morbidity and mortality and address the alarming disparities in outcomes, we must shift to a culture of equity. In 2023, the National Academy of Medicine released a framework to align emerging science, technology, and innovation in medicine with equity [5]. This framework recommends reorienting the culture of innovation, incentivizing equity, expanding participation of underrepresented and underserved communities in innovation, developing equity science, and creating and promoting equity playbooks for particular areas of emerging science. Each stage of an emerging technology's development should re‐align the innovation with the ideal of equitable benefits across society. Elimination of racial and ethnic disparities in maternal morbidity and mortality and improvement in outcomes for our patients will require an intentional, steadfast focus on evaluation of technology through an equity lens.

CONFILCT OF INTERESTS STATEMENT

A.K.L. served on a medical advisory board to Shields Therapeutics in 2021 and to Pharmacosmos Therapeutics in 2022.

FUNDING INFORMATION

NIH/ORWH Building Interdisciplinary Research Careers in Women's Health (BIRCWH); NIH, Grant Numbers: NIH K12HD043441; NIH/NHLBI K23HL168356; NIH/NICHD K23HD103961