HER Heard

First: Increase access to and quality of health services for medically underserved groups around the world (such as refugees and other displaced people, women and children, older adults, and LGBTQ+ individuals).

Also: Ensure health-related data is collected ethically and effectively, and that AI and other insights are accurate, targeted, and actionable. 

For trustworthy AI there must be trustworthy partnership with marginalized, underserved, and vulnerable communities - ones experiencing historic and ongoing system harms - to build the source of data. Any AI or analysis is only as good as the quality of the data. Most commonly used data sources and big data either are missing data on marginalized groups or else have mislabeled those groups, thus our solution begins with first inviting women and femmes onto a trustworthy health literacy platform that promotes better shared decision making, invites them to share data on a platform that uses self-hosted data for greater privacy and less risk than the cloud, then builds new and more accurate algorithms, using a federated model (similar to Apple where the data does not leave the individual's own device). This is a phased approach envisioned over a decade.

We aim to build a highly engaging patient-centric platform for shared decision making for the women’s health journey, to redirect away from social media misinformation, while reducing the digital divide for marginalized communities, with a novel approach to user profile hyper-personalization applied to content, chatbot, patient diary, symptom checker (patient reported outcome measures) and navigation functions. Initial use cases: pelvic pain, well woman cancer screening, menopause.

Our technology seeks to be uniquely patient-facing to create the “activated informed patient” per the Wagner Chronic Care model, to allow women to self-assess, learn, and receive patient navigation support for earlier care and better outcomes. We aim to develop a novel digital health app to redirect away from misinformation online for better health information.

Prototype: https://app.herheard.com/

1 in 5 women report poor communication with their providers; with poor communication a known cause of health inequity especially for sensitive issues -in marginalized groups. Increasingly, women go online, evidenced by the $15 million “menopause TikTok” industry, thus get exposed to misinformation. Each year, 1.3 million women enter menopause, yet only 1 in 5 gynecologists have been trained in the care of this condition. This is one of several conditions where women seek information online and underscores the need to develop effective pro-science tech solutions, that engage women as well as social media does.

 Advances in generative AI show better empathy than human doctors, as rated by patients online. This has yet to be applied specifically to midlife women or those of marginalized groups. AI can personalize communication; it can “learn” with each new patient interaction with the model. Current online health information is offered by disease, yet women seek information for non-specific symptoms like “brain fog”, “fatigue”, “heavy period.” These can indicate any number of conditions, some that co-occur. We aim to personalize information for the whole woman’s unique lived experience and her unique combination of communication preferences and needs. Non-specific symptoms are tracked and categorized by NLP and creation of individualized symptom diaries, then sorted into “Quality of Life” (QoL) data fields utilized by clinicians and payers. Women’s narrative, converted into digitized versions of these QoL measures, can help triage care.

The age of perimenopause coincides with when women also develop pelvic floor dysfunction, pelvic pain, and symptomatic fibroids cause reduced quality of life. Minoritized women are often not offered the full range of options. At midlife, women must start routine mammography screening. This age is also when gaps in cervical cancer screening occur. Meanwhile, in women’s health, the wave of closures of outpatient clinics and hospital units decreases both information and care access, especially for well woman exams and cancer screenings. Taken together, this calls for highly scalable digital health solutions providing hyper-personalized, private, patient-centered health education for midlife women’s health across disease conditions for diverse populations.

Umbereen S. Nehal, MD, MPH, Sloan Fellow at MIT, is published in medical AI, refugee health, and patient-centered communication. Dr. Nehal was a Chief Medical Officer of a 14-site agency that included behavioral health integration and telehealth. As Associate Medical Director of Massachusetts Medicaid, Dr. Nehal oversaw state-wide information “HIway” optimization for interoperability. She designed novel HIT for Medicare Advantage plans to track and rank new forms of data used for payment by CMS for integrated cloud-based platform enhanced by AI and natural language processing (NLP). Dr. Nehal served as co-chair of the Patient-Centered Outcomes Research Institute (PCORI) Healthcare Delivery and Disparities Research advisory panel, for a >$300 million portfolio. Dr. Nehal served as the Principal Investigator of a community health grant for low-resourced communities, aligned with the Culture of Health. Dr. Nehal’s national leadership won recognition from President Obama.

Benita Chumo is a product development professional with experience in leading cross-functional teams to conduct market research, performing product and market analytics, and managing products over their lifecycle from ideation to end-of-life phases. She is passionate about building products that appropriately address customer pains/gains and executing on operational needs to get product to users.

Haruka Yamashita is an MIT Sloan MBA student with experience from Morgan Stanley with a passion for finance, technology, female empowerment, and beauty.

Hua Yang, PhD is a Data Science leader specializing in AI and machine learning, with over ten years of hands-on experience in biomedical start-ups.

Denise Howard, MD, MPH is chief of the Department of Obstetrics and Gynecology at NewYork-Presbyterian Brooklyn Methodist Hospital with expertise in shared decision-making.

Leo Anthony Celi, MD, MPH, MSc, co-directs Sana at the Institute for Medical Engineering and Science at MIT, whose objective is to leverage information technology to improve health outcomes in vulnerable populations.

Mark Paxton, JD, as the Sponsor Representative for the Surgeon General of the Army, US Army Medical Research and Material Command, provides regulatory leadership to a staff of approximately 70 Army civilian, military and contractors in support of advanced product development.

We have a functional MVP (minimum viable product) and prototype with 478 users, some of whom from which we have received feedback. We have not yet built sufficient features to drive daily use of the app nor to monetize in D2C business model. 

Our initial app was web-based and build with a low code solution using Flutter. Now we have rebuilt the original app, per feedback, on React, as this better matches the tech stack used in the industry and will be more trusted by healthcare entities when we switch to a B2B model after proving we can engage a sufficient number of women via health literacy content, self-assessment, medical navigation support, and nudges on healthy behaviors. 

We are seeking support of mentorship, industry relationships to better understand how to plan to transition to B2B for the longterm fiscal sustainability, and to establish pilot sites for validating the app.

Given the longterm plan to become a SaMD or software as a medical device that can be used in tribal, rural, and military health as a triage tool, then the impact measurement coaching, mentorship and support are particularly of interest and differentiates SOLVE from other venture incubators or accelerators. Others define metrics only in financial terms whereas SOLVE combines social impact with fiscal sustainability for true social entrepreneurship. 

We also understand the importance of being part of communities and peer support. We hope to both contribute to and benefit from the shared learnings in the Solver community. 

We combine 1-human-centered design for engaging and culturally matched content, 2-use the newest forms of generative AI for cost-effectiveness and customized content, 3-are built by women for women and by a physician who is also a patient 4-use science-based "patient reported outcome measures" as compared most wellness apps that do not use the metrics also found in research 5-designed in stages for lowest cost build but with the plan, over stages, to build into an FDA-approve software as a medical device that can serve urban women in the developed world or rural women in the developing world. 

We are intentional in a phased approach to ensure we are starting by centering the end user's needs and respectful of what women and femmes want via  freemium and D2C then transitioning to a B2B to reduce financial burden on women once we have validated sufficient engagement for insurance to purchase in a B2B model. Most tech is built initially with the economic buyer and the MVP serves insurance or hospitals first whereas we aim to build more engaging human-centered technology to serve women and femmes. 

Per our initial testing of advanced prompts, current LLMs do not suffice for personalized health care journey support for women to self-assess, learn, and navigate care, thus we will finetune a model and compare to the existing baseline models. For our LLM to serve as a customized pre-visit preparation tool, similar to navigation support of a culturally matched community health worker, we will use RAG for finetuning a completion model chatbot, finetuning temperature for accuracy, precision, empathy, matched language, relevant and satisfying response for that user. Health data privacy, security, and HIPAA compliance require a self-hosted model.

We develop a novel yet reliable women’s health-specific knowledge database, compliant with intellectual property law, creating a unique combination of licensed shared decision-making tools, academic research, and expert social media content, using network graph analysis validated by specialist reviewers through a voting process. For vetted social media health experts, we will identify influential nodes, assess their reliability, then analyze distribution with the top 5% of nodes as “influencers.” For medical literature we will link summaries to the names of authors and co-authors, with a fuzzy string matching algorithm, similar to German bank receipt matching, to enhance the accuracy of our health content.

For health literacy, we would start with a rules-based method using a unique combination of metadata from the patient profile (age, ethnicity, etc.) plus categories like reading level, language style, jargon/slang, content type, use of numbers, sentiment, engagement patterns. This differs from current approaches that use readability formulas for educational level to assume comprehension. If rules-based proves insufficient, we would use machine learning (ML) trained on a large dataset of labeled data to decipher patterns within context, engaging a partner site, with a representative population for underserved communities, for training data.

We are designed to address the following system barriers and maladaptive behaviors instead to invoke system incentives to redirect behaviors to ones that are pro-science, pro-health, and deliver better outcomes for individuals and at scale:

1-Problem: Women are not "heard" - on average a woman is interrupted in a doctor's/provider visit within 11 seconds. Stigma and lack of knowledge exist in women's health. This causes women to seek information online on Reddit, TikTok, Facebook, exposed to misinformation and addictive algorithms with personalized feeds. It has been shown that >74% of online content is misleading or overt misinformation.
Solution: create highly engaging content, customized to the user, using genAI to create inexpensive content in multiple forms from existing validated, clinician-approved content that is evidence-based and utilize algorithms currently used in e-commerce and social media to customize feeds to the user profile. 

2-Problem: women feel alone and seek networks online but may encounter misinformation, bullying, loss of privacy on health data. Moderation is needed yet sometimes the most marginalized groups are mislabeled or not included.
Solution: Create online peer groups that women can invite peers to and choose from within recommended moderation guidelines. There will be light moderation for removal of overtly unsafe or false content.

3-Problem: Women are not heard or believed in provider's offices and often speak in narrative form that is too much for a 15 minute visit
Solution: use patient-reported outcomes and narrative with AI to surface insights for data-driven pre-visit preparation. Just like the doctor's office has a medical assistant, the woman needs her own digital medical assistant to assist her in being prepared for the visit

4-Problem: Women are diagnosed 3-10 years late for conditions and were not included in clinical trials until 1993. Current algorithms are not inclusive of women, femmes, women of color and may underdiagnose, overdiagnose, or misdiagnose. As a result, EHR or electronic health records can be their own source of "misinformation" if contain missed diagnoses or misdiagnoses.
Solution: Develop a new data source, directly from women to train algorithms that are inclusive and more accurate across the full diversity of humanity by gender, race, ethnicity, and culture. 

Impact goal progress is measured both with quantitative and qualitative data.

For effectiveness of engagement: number of users, frequency of use, tracking of use of specific parts of the app + focus groups and user feedback to assess qualitative feedback on user experience, areas for improvement

For clinical effectiveness: tracking of patient-reported outcomes to assess quality of life and functionality and how many well woman activities completed (mammograms, HPV vaccination, pap smears, etc).

For cost-effectiveness at a system level: metrics on appropriate use of care like emergency room visits, hospitalizations, etc (This is a long term goal in years 3-5 and beyond) 

For market uptake: number of pilot sites, conversion of pilots to paid customers, revenue

Our technology uses generative AI like large langauge models. For our pilot we are using custom GPT and have built three custom GPT to date. In the future, to reduce cost and ensure adequate privacy of sensitive of women's health data, we will use self-hosted options. As compared to the recent difficulty with the WHO chatbot, Sarah, we will keep ours to narrow use cases initially. 

We also make use of generative AI to convert existing long form text into new forms of content like short videos and instagram style posts that are more engaging. 

We use AI that is commonly used already, for customizing a feed and content to user profiles and suggesting content

Longer term we will be building clinical models as clinical decision support and triage support for insurance and clinical systems. 

Our solution has had a number of part-time contributors, interns working full-time for a semester, and advisors. Please see the team slide for full team. Advisors put in 2-20 hours a month of work. We have had two full time interns devoting a full semester to building the prototype. Other interns have put in 2-8 hours a week. The founder is full-time on this while a mid-career student at MIT Sloan, utilizing her classes to work on the venture. 

Our founder has been living this for her entire reproductive life. The form start to this social entrepreneurship venture idea was in November 2022 with an application to StartMIT. The initial team was formed in January 2023 in StartMIT and has been continued via MIT Sandbox to date. The LLC was formed in September 2023.

Our team has diversity of genders, races, countries of origin, regions within the U.S. religion, industry experience, levels of experience, neurodiversity and different lived experience of menstrual health and different stages of the reproductive health journey. Our teamwork is always grounded in values and our "why." We begin meetings with a practice of gratitude and centering. We work both in person and asynchronously to accommodate the range of needs for the team.

We are committed to user-centered design and using iterative building with multiple sources of user feedback to ensure we are building to serve communities and the end user. While our scope is to serve all of women's health in the future, we are starting with the use case of fibroids as this is an underserved population experiencing disparities as well as an underserviced market that could be a blue ocean of opportunity if approached with a novel take on the business case. The lack of shared decision making offered to BIPOC women, women of color, and Black women that results in not being offered uterus-sparing options and excessive use of hysterectomy represents a reproductive justice issue. While hysterectomy may be the right answer for some women, all women deserve to be offered the full range of choices. 

We have a three-phased approach for the 10-15 year plan, starting with health literacy and patient engagement to becoming an FDA-approved SaMD or software as a medical device

Phase 1 (Years 1-2) Initially, to drive engagement by matching current behavior of going to social media, we will be D2C or direct to consumer to re-direct current behavior to our engaging platform of expert-vetted evidence-based content. We will start as a freemium then add $5/month and $10/month tiered fees to access additional features in addition to the freemium. While we could also charge physicians, we seek to center the user and avoid conflicts of interest that occur when the economic buyer is not the end user and the most vulnerable party. 

Additional features will be self-assess features. Similar to other apps, access to data past 30 days will require a subscription compared to the freemium

Phase 2 (Years 3-5) Once sufficient users have been onboarded with proven engagement, we would switch to a B2B model selling to insurance for a $10 per member per month (PMPM) as a patient engagement platform for population health. The key performance indicators (KPIs) and value proposition would include better user satisfaction, better population health, more appropriate health utilization (not missing needed and not excessive visits for avoidable and low value care), and better health outcomes. Health plans are also rated based on their routine cancer screening rates for mammography and cervical cancer screening, both of which are currently declining, especially with loss of access to women's health from clinic closures in the U.S.

Phase 3 (Years 5-10+) Develop algorithms as clinical decision support tools from the new data built from women's self-assessments and patient-reported outcomes data that can be validated via clinical trials and become FDA-approved "SaMD" or software as a medical device. This can be used by payers and government agencies including tribal health, military health, and rural health to triage and improve resource allocation. These algorithms can also be white-labeled and sold to health systems for a B2B model of licensing fees.

We have the promise of the first angel check for $25K from a Black woman who works in the pharmaceutical industry who personally experienced a harrowing hysterectomy without shared decision-making or respect. We are in talks with 3 other potential angel investors.

We submitted a pitch to the National Science Foundation (NSF) and received approval of the initial step, now invited to submit the full grant submission. 

We have a first LOI from a healthcare provider serving rural Americans, primarily in Southern states.

We have a waitlist of 37 doctors seeking to join the platform to be better able to communicate with patients on a trusted platform free of misinformation or the usual online bullying. 

From MIT Sandbox we have received $7500, approved by the MIT Sandbox funding board, to build the prototype, gain traction, and do customer discovery.

From the MIT Mind Hand Heart initiative, the Chancellor's Innovation Fund, we received over $3K to conduct peer groups for women and femmes with women's health and reproductive health conditions at MIT and have had regular attendance of these lunchtime meetings.