Ada & TMCG - Increased access to healthcare through AI

Low and Middle Income countries (LMICS) around the world face massive challenges in achieving Universal Health Coverage (UHC) by 2030 due to long standing health system issues. Limited public health financing and a severe shortage of healthcare workers, prevents more than half of the world’s population from having access to essential primary healthcare services. According to the World Health Organization (WHO) over 18 million additional health workers are needed by 2030 in order to meet the United Nations’s Sustainable Development Goal objective of achieving Universal Health Coverage (UHC). 

Currently, over 90% of populations across Sub-Saharan Africa are uninsured. This is  largely attributed to the lack of public health insurance provision. Private health insurance offerings remain not only unaffordable for low and middle income groups of the population, but are also not adapted to the realities of these users. 

In the absence of formal health insurance households tend to bear a high burden of out-of-pocket health expenditure and are susceptible to health shocks, an event of death or disease, that can cause significant adverse economic outcomes for households in LMIC. Health shocks specifically pose a risk to the growing middle class of close to 350 million people across the continent. Empirical evidence shows that the growth and stability of the middle class is associated with better governance, economic growth and poverty reduction. However, 60% of the continent’s middle class, almost 180 million people, are barely out of the poor category. These individuals are greatly at risk of falling back into poverty due to out-of-pocket health related expenditures, per experts at the African Development Bank. 

This existing healthcare challenge is further exacerbated by the COVID-19 pandemic, which has made achieving UHC increasingly challenging due to its crippling economic and health care effects. According to the UN, COVID-19 could in fact reverse decades of healthcare improvements towards UHC. In Uganda, an increase of 2.6 million poor people has been witnessed, falling out of the approximately 15 million Ugandans of the middle class, who are deemed as non-poor but insecure.

Finding new operating models to enable the provision of affordable quality healthcare at scale will be critical to protect millions of people from the adverse effects of the pandemic and beyond. The provision of cost-effective health services via telemedicine already has become a powerful key resource to address this challenge. 

People all over the world are turning to digital forms of healthcare delivery to mitigate their concerns about contracting the virus. The changing behavior patterns of consumers are having a profound effect on healthcare delivery and have resulted in a significant acceleration in the adoption of digital care delivery services like telemedicine. This acceleration has also been witnessed in Uganda where telemedicine provider The Medical Concierge Group (TMCG) has experienced a 500% increase in the number of people who are using their telemedicine service. The dramatic increase in telemedicine usage is creating significant pressure on TMCG to scale up their services to meet the surge in demand from users in Uganda. 

Digital health solutions powered by Artificial intelligence (AI) have the potential to address these urgent shortcomings and facilitate scaling by supporting along three key dimensions: 1) Improved access 2) efficiency gains and 3) ensuring quality. To combat today’s growing health challenges and to achieve UHC, systematic integration of technology into the way healthcare is delivered will be critical to expand access for all. 

Ada is an AI-powered personal health companion whose mission is to improve the health of a billion people with trusted medical guidance that is relevant, actionable, and effective. Founded in 2011 by a team of doctors, engineers, and data scientists, Ada has currently over 12 million users who have completed over 27 million Ada assessments. Ada is the world’s most popular symptom assessment app with more than 250,000 5-star ratings since its launch in 2016.

Ada’s AI chatbot allows users to input basic patient information and then guides the users through a series of symptom-related questions, ultimately providing a probability on the possible condition and the appropriate level of care that the patient should seek. Ada has been clinically validated and independent reports have consistently demonstrated that Ada is the most accurate and complete symptom checker currently available.

Recognising the demand for Ada’s technology in low- and lower-middle income countries, Ada’s Global Health Initiative was launched in 2018. The Global Health Initiative focuses on the implementation of Ada’s AI in low-resource settings to increase access to personalized health information and improve primary healthcare delivery for those who need it most.

As COVID-19 has already begun to compound the significant challenges that Uganda's healthcare system faces, and will continue to do so as additional waves occur. The gravity of this challenge requires effective digital health solutions that can unlock improved primary healthcare delivery at scale in low-resource settings. Through this strategic collaboration between Ada & TMCG, we aim to showcase how the integration of Ada’s AI powered symptom assessment into TMCGs telehealth service can drive more efficient utilization of healthcare resources through improved data collection, enabling a a scalable model of healthcare delivery. 

The Medical Concierge Group (TMCG) is a digital health and telemedicine company that has been incorporated in Uganda (HQ), Kenya and Nigeria since 2012. Since August 2019, TMCG has also been operating its public-facing services under the ‘Rocket Health’ brand name. This collaboration focuses on improving TMCG’s integrated telehealth services, in which customers can access medical consultations via telephone with TMCG’s 13 doctors, 3 nurses, 4 pharmacy technicians and 3 clinical officers through a subscription model. The service provides remote teleconsultations with medical professionals that sometimes is also coupled with a medicine delivery or laboratory sample pick up at the client’s doorstep. For further clinical review when required, Rocket Health also runs an appointment-only clinic where clients are scheduled for in-person assessments. Rocket Health presently serves 60,000-80,000 people on a monthly basis. Leveraging digital solutions and AI will be critical to expand TMCGs services and enable this large scale growth, resulting in improved access to primary healthcare. 

Through the integration Ada’s AI powered symptom assessment technology within TMCGs provider workflow, we aim to: 

• provide TMCG’s users with improved access to personalized health information and guidance and 

• support TMCGs medical providers with more efficient utilization of their limited time and improve the quality of care through the improved data collection of patient history 

In further detail, the integration of Ada’s core symptom assessment will provide the user a report of possible conditions, allowing them to understand their symptoms. The report provides the most appropriate guidance based on 8 advice levels (Self care; Self care, Pharmacy; Primary care, 2-3 weeks; Primary care, 2-3 days; Primary care, Same day; Primary care, 4 hours; Emergency care; Call ambulance) on what potential next steps are most appropriate, reducing the barrier to care and instilling confidence in the user. 

Making Ada available 24/7 to TMCGs users contributes to early detection and treatment of diseases and avoiding higher costs associated with acute health events. As 65% of TMCGs workload consists of dealing with acute symptomatology, the ability to make more relevant primary care routine services available is limited. By introducing Ada and the more efficient navigation to care, TMCG also seeks to expand their routine services (e.g. such as chronic/routine care, vaccinations, antenatal care etc) in particular to at-risk groups such as mothers and children. Ada’s capability to safely triage e.g. identify who actually might need same day care vs. at a later stage/or even self-care can help to create these efficiencies by directing people to appropriate care pathways. 

Ada’s solution allows patients to share their Ada assessment with a clinician ahead of their consultation. As the report covers findings, symptoms present and absent, and likely conditions, the clinician will be able to use the report to structure their consultation. Preliminary evidence from an independent study involving doctors from the Global South suggests that the use of Ada could greatly support clinical decision-making and increase the efficiency and quality of a consultation. Results indicated that Ada has the potential to increase diagnostic accuracy of clinicians from 13% to 52%, thereby having a significantly positive impact on patient outcomes. 

The ability to support clinicians with an Ada assessment will drive efficiency gains during the consultation process as doctors will have a clearer understanding of patients’ symptoms and possible conditions before the start of the consultation, and will be able to utilize the duration of the consultation time in the most effective manner. Implementation of Ada in TMCGs telemedicine workflow will enable a more cost-effective delivery of healthcare services, which is key to achieve significant scale. 

Currently, over 90% of populations across Sub-Saharan Africa are uninsured. This is  largely attributed to the lack of public health insurance provision. Private health insurance offerings remain not only unaffordable for low and middle income groups of the population, but are also not adapted to the realities of these users. 

In the absence of formal health insurance, households tend to bear a high burden of out-of-pocket health expenditure and are susceptible to health shocks, an event of death or disease, that can cause significant adverse economic outcomes for households in LMIC. Health shocks specifically pose a risk to the growing middle class of close to 350 million people across the continent. Empirical evidence shows that the growth and stability of the middle class is associated with better governance, economic growth and poverty reduction. However, 60% of the continent’s middle class, almost 180 million people, are barely out of the poor category. This segment of the population is greatly at risk of falling back into poverty due to out-of-pocket health related expenditures, per experts at the African Development Bank. This existing healthcare challenge is further exacerbated by the COVID-19 pandemic, which has made achieving UHC increasingly challenging due to its crippling economic and health care effects. According to the UN, COVID-19 could in fact reverse decades of healthcare improvements towards UHC.In Uganda, an increase of 2.6 million poor people has been witnessed, falling out of the approximately 15 million Ugandans of the middle class, who are deemed as non-poor but insecure.

TMCG’s service has grown rapidly during the COVID-19 pandemic and now has an estimated 60,000 users per month across its voice, messaging and in-person service portfolio. This user base is expected to grow significantly over the next 6 months as TMCG implements the following three key partnerships: 

• MTN: TMCG has signed a partnership with the youth service of the largest local telecommunications company, MTN, to be accessible for youth-friendly services to its membership base of 2.5 million members

• Leading Local Health Insurers: TMCG’s Rocket Health service has recently been taken up by 5 of Uganda’s largest private health insurers with a combined membership of 250,000 people 

• USAID: TMCG has commenced a 5-year nationwide project, Social Behavior Change Activity, supported by USAID and targeted to reach a large percentage of the 43 million Ugandans by 2025

With these additional collaborations and partnerships, the user-base for the Rocket Health services is set to grow sharply and cover a much wider geographical area than before. The Ada-TMCG partnership is envisioned to be the core technology system that powers this growth and ensures optimal quality of services, providing an opportunity to reach an addressable market over greater than 6 million Ugandans with smartphones directly. Ada’s technology will continue to learn with each TMCG user, to allow for the service to continuously provide the most accurate assessment with appropriate suggested care. 

Leveraging digital solutions and AI will be critical to expand TMCGs services and enable this large scale growth, resulting in improved access to primary healthcare.

Through previous work in collaboration with Fondation Botnar, Ada’s medical knowledge base has already been localized for the East African setting. Ada has also been translated into Swahili and will provide 1.5 million Tanzanians with access to Ada’s technology by 2022. As a consequence, Ada has a strong foundation that can be further built upon to further tailor our technology to address the healthcare challenges across Sub-Saharan Africa. As a part of this collaboration, extensive in person user research was conducted to better understand how the product could be improved in this setting. 

Additionally, through each phase of this project, Ada and TMCG will draw on a range of in-house and external experts, including user experience researchers, epidemiologists, health economists and data scientists, to effectively design, measure and evaluate the impact that will be created through the development of this joint solution. Consistent post market surveillance activities will allow us to capture feedback from both patients and health workers, to enable continuous improvements of the solution. The health economic evaluation will explore the acceptance and use of symptom assessments as part of TMCGs service offering, how care seeking behavior of people who use symptom assessments changes, the utility of the assessment to patients and physicians, as well as the associated changes in health care resource utilization and direct costs. The evaluation will help Ada and TMCG to better understand the use of digital symptom assessment tools as a front door to healthcare and the effects on healthcare resource use and related costs and revenues. 

We are applying to this challenge as the integration and evaluation of Ada’s symptom assessment technology into TMCGs telemedicine workflow service requires significant resources. Financial support is critical in order to be able to carry out the integration and continuously iterate and refine the solution to meet the target groups needs. Donor funding is critical to enable this collaboration, as well as the expert guidance of the MIT Solve judges, which will not only help guide our work in Uganda, but enable us to develop an AI-powered telehealth model, which can scale across LMICs. 

Offering exceptional medical safety and accuracy, Ada’s AI-driven enterprise solutions are based on the same technology, medical knowledge and user experience that helped Ada’s consumer symptom assessment app become the world’s most popular and highest rated. Having been made available to partners in 2020, Ada’s solutions are designed to help organizations deliver better care experiences and connect clinical journeys, helping improve user outcomes while driving up efficiency and user satisfaction. A peer-reviewed and published study in BMJ Open also showed that Ada’s safety performance matched that of GPs, at 97%. Ada’s performance was notably higher than the average of other competitors (90.6%) with some performing well below this with one as low as 80%. Safety was defined as the degree of deviation of advice levels compared to the gold standards set by experienced independent GPs prior to the study. An independent, peer-reviewed study conducted by Stanford University and Sutter Health reviewed thousands of Ada assessments and concluded that Ada’s triage recommendations were comparable to those of nurse-staffed telephone triage lines. With scarce medical resources in LMICs, according to WHO there is a current gap of nearly 18 m health workers worldwide. Ada’s medical AI can help to significantly improve medical resource utilization in the global south and drive catalytic change, making primary healthcare more accessible to underserved populations.

The solution in itself is not only innovative with regards to the developed best in class symptom assessment technology, but also specifically because of the partnership between Ada & TMCG, which allows to integrate Ada’s AI for the first time into a telemedicine flow in Sub-Saharan Africa. Creating efficiencies in the workflow through this collaboration enables our local partner TMCG to offer more affordable subscription based care packages and accelerate the access to primary health care not only in Uganda, but across Sub-Saharan Africa.

The objective of this collaboration is to develop a cost-effective AI-powered telehealth model in order to achieve UHC, e.g. making primary healthcare more accessible and affordable. The ability to scale this platform efficiently beyond Uganda is predicated on the upfront investment in a solution that is optimized for LMICs and is predominately digital. We aim to utilize the learnings from this collaboration and seek to replicate this model across several other low-resource settings through partnerships with telemedicine providers, governments and industry stakeholders like mobile network providers and insurance companies to ensure access to affordable and quality care at scale.

• In the next year our major objective is to have the AI-powered telehealth model up and running successfully as well as having measured the real world impact with regards to the efficiencies and improvements created for patients and providers. 
• In the next 5 years, we seek to have this model further refined and adapted for LMICS, having achieved a blueprint and rolled out the model across Sub-Saharan Africa. 

Ada and TMCG will draw on a range of in-house and external experts, including epidemiologists, health economists and data scientists, to effectively measure and evaluate the impact that will be created through the development of this joint solution. Consistent post market surveillance activities will allow us to capture feedback from both patients and health workers, to enable continuous improvements of the solution. The health economic evaluation will explore the acceptance and use of symptom assessments as part of TMCGs service offering, how care seeking behavior of people who use symptom assessments changes, the utility of the assessment to patients and physicians, as well as the associated changes in health care resource utilization and direct costs. The evaluation will help Ada and TMCG to better understand the use of digital symptom assessment tools as a front door to healthcare and the effects on healthcare resource use and related costs. We will focus on the following categories across both phases to measure the health and economic impact of our work: 

Phase 1a: Key Metrics 

1. Impact on individuals’ well-being and health measured through: Collected feedback from TMCG customer if they found the assessment to be helpful and relevant

2. Health Economic impact for individuals to estimate how using Ada can relieve the economic burden of seeking careTime saved from avoiding unnecessary healthcare consultations

• Income retained as a result from missing fewer days at work
• Costs for transportation to healthcare facilities
• Out-of-pocket expenses for medical fees and medications

3. Health Economic impact for the health system to measure efficiency gains (e.g. improved healthcare resource utilization) through the usage of Ada

• Number of Ada assessments and TMCG services used after introducing Ada’s symptom assessment compared to the time period before introducing Ada
• Proportion of Ada assessments that lead to use of TMCG’s services, separated by service type
• Proportion of Ada assessments that lead to use of TMCG’s services, separated by initial intention to seek care
• Proportion of Ada assessments that lead to use of TMCG’s services, separated by advice level recommendation of the assessment
• Metrics for health economic benefits will be derived from the real world use of Ada within TMCG’s infrastructure and the data routinely collected by Ada and TMCG. This data contains users’ intentions on seeking care before and after the assessment, assessment data including recommended level of care, and reviewed and selected care navigation options. Population statistics and unit costs are used to extend the metrics to healthcare and other cost savings.

Phase 1b: Key Metrics 

1. Efficiency and quality gains for Healthcare provider (HCP) measured through user reported outcomes (for both patient and HCP users of Ada) by assessing: 

1. The overall extent to which the system delivered its stated clinical benefits in terms of delivering conversation support, the potential time-saving and the promotion of HCP and patient rapport 

2. The degree of information relevance to conversation support, the potential time-saving and the promotion of HCP and patient rapport 

3. The degree of information completeness for conversation support, the potential time-saving and the promotion of HCP and patient rapport 

4. Details of information that was not provided and could have been useful for conversation support, the potential time-saving and the promotion of HCP and patient rapport 

5. The degree to which the HCP and patient would recommend the system for conversation support, the potential time-saving and the promotion of HCP and patient rapport.

Ada & TMCG will draw upon routinely collected administrative, clinical and user data, as well as user surveys, conducted within the framework of user reported outcomes and real world data gathering frameworks. In a dialogue with local regulators and ethical approval boards, we will explore the degree to which these assessments will be through the confirmation of individual patient Ada assessments through their HCP, or the collated insights of HCPs alongside  user reported perceptions.  Post analysis, we will look to share our results and learnings with industry stakeholders through M&E reports, case studies and  academic publications.

Inputs: 

• Ada’s and TMCGs expertise with regards to their respective fields e.g. medical AI & telemedicine. Both of our organizations leadership and advocacy capacity to 

• Commitment to working in partnership with community and other relevant actors 

Activities: 

• Integration of Ada’s core symptom assessment technology, care navigation modules and TMCGs EHR system. 

Outputs: 

• We expect, out of the currently  60,000 TMCG users, an estimated 15% to complete an assessment prior to receiving care. Assuming a conversion rate of 70% of the Ada assessment, at least an estimated 6,300 users per month and close to 75,000 users over the course of this 12 month project, will have completed an Ada assessment. Assuming that the 75,000 TMCG  users that use the Ada assessment conduct 3-6 assessments per year (based on data on Ada’s existing user base), it can be estimated that between 225,000 and 450,000 assessments will be completed over the course of the project duration.

Outcomes: 

• Well informed and empowered patients. Efficiency and quality gains for Healthcare provider (HCP) measured through user reported outcomes (for both patient and HCP users of Ada). See previous section on evaluation of our impact goals. 

Impact: 

• Increased access to healthcare in LMICs.

• Improved patient outcomes 

• Strengthened, equitable health systems, reduced gaps in access to primary health care 

Ada’s health assessment technology is based on a custom-built reasoning engine and a comprehensive medical knowledgebase, continuously developed and refined over ten years by our team of leading engineers, scientists and clinicians, delivering the most sophisticated personalized health  assessments on the market. Its core system connects medical knowledge with intelligent technology to help people actively manage their own health and health organizations to deliver effective care. Ada’s medical knowledge is built by experts, and validated by clinical research:

1. Medical experts: more than 50 in-house medical experts represent the breadth and depth of current medical knowledge, ensuring Ada’s high medical quality with a sophisticated disease modeling procedure and extensive testing.Medical quality is updated and optimized every fortnight. Our Medical Knowledge Team consists of doctors from various specialties to ensure wide clinical coverage. We have worked hard to ensure that Ada covers a range of conditions for all service users and covers almost 30 specialty areas, including cardiology, dermatology, emergency medicine, endocrinology, gastroenterology, geriatrics, immunology, nephrology, neurology, surgery, and tropical medicine to name a few. We include infectious diseases and several hundred rare diseases, but we also cover several highly important areas that are often not available in other suppliers technologies, including pediatric, mental health, and obstetric and gynecological related conditions:

2. Sources: Sources include peer-reviewed journals, evidence-based recommendations, and gold-standard clinical practice guidelines. 

3. Condition modeling: Our teams use medical publications, clinical guidelines, reference cases, research papers, and their own clinical expertise to create a list of symptoms and risk factors. These are then associated with conditions in Ada’s medical knowledge base. 

4. Peer review: Each condition is typically created using 15 to 20 content sources and cross-checked by  several medical experts.

5. We’re constantly benchmarking Ada’s reasoning through internal stress testing using real-world clinical case scenarios, competitive comparisons, medical collaborations, external independent review, and user feedback. Summaries of clinical studies researching Ada, both in progress and those published in journals, can be found here: Publications and research https://ada.com/studies/ 

The data Ada collects, as described above, consists of medical literature, published research, consolidated practice and past case feedback. Using a proprietary Medical Description Language, Ada’s medical knowledge is coded into structured information. This knowledge base is used to create a probabilistic model, which is combined with Ada’s intelligence consisting of an inference engine and conversation interface.  Ada asks dynamic questions to the user, and considers reported symptoms, gender, age and other medical risk factors to provide users a personalized health assessment. Ada reviews multiple pieces of data and provides a probabilistic assessment that suggests possible causes of symptoms, helping users to identify the best next steps and navigate to the appropriate care.

Our enterprise solutions can integrate seamlessly with our partners' platforms and user journey. Ada's assessment is available in 11 languages and can be further localized. It can include customizable navigation options to direct users to their preferred or in-network care facilities, providers, digital/remote/virtual services, and other care navigation options. 

Through Ada’s symptom assessments, Ada collects primary health care data, such as age, gender, height, weight, medications, allergies, symptoms and health status, including hypertension, diabetes, and smoking history. This information is leveraged when a user is completing a symptom assessment to determine the appropriate advice level and the most likely cause. The incentive for the collected data is helping users understand and manage their help by guiding users to the most appropriate advice and next step. 

In addition, Ada’s data Insights offered in the form of comprehensive data dashboards, helps our partners understand their users’ health on a more nuanced level to drive strategic decision making, identify gaps and opportunities for improvement and service planning. The aggregated collected data through Ada’s symptom assessments can help TMCG analyze information across a range of factors, including geography, gender, age, time of assessment, completion rate, helpfulness ratings, symptoms and conditions, risk factors, presenting complaints, and next-step actions.

• Humans are widely diverse when it comes to their health. Variation in age, gender, ethnicity or geography can lead to people having different health needs, and also to conditions and symptoms being expressed in a different way. Trustworthy AI health assessment solutions need to take these differences in human health into account and adequately represent their user population. This has been especially a challenge for AI-driven healthcare solutions that are based solely on data-driven machine learning techniques. That’s why right from the start, Ada has been taking a different approach.

• Built by a diverse team of doctors: Ada combines advanced AI technology with a medical knowledge base that is engineered by a team of over 50 human medical experts, drawing upon a wide range of top-tier scientific literature and studies. In Ada’s medical team, there are doctors from over 15 nationalities and a wide range of medical experience to cover diverse medical perspectives and knowledge. At Ada, we have put in place a number of processes to ensure that biases are not incorporated into our technology. As part of our continuous review process, a dedicated team of professionals critically appraises any data incorporated into the knowledge base to identify potential biases and to make sure that our users with their individual backgrounds and health needs are represented adequately. 

• Diverse research & evaluation is key: Ada runs a comprehensive program of studies to validate our solutions. In addition to well-balanced, inclusive clinical studies in the US and Europe, Ada conducts studies in underrepresented populations, for example in Tanzania and South Africa, and these allow us to evaluate and adapt our health assessment technology to healthcare needs in different geographies.

• Collaborating with the health community: We collaborate with independent medical and scientific experts, patient organizations and other experts from the healthcare community to advance research, seek external feedback and to ground our solutions in the complex reality of day-to-day healthcare situations. Through our collaborations with life science companies we get valuable insights and feedback from healthcare and industry specialists and can access medical knowledge that is not yet publicly available, such as interim/preliminary results from clinical studies or expert knowledge on rare diseases. 

Ada Health works across the healthcare value chain and provides value to partners through a B2B as well as B2G business model. We partner with health insurers, health and care systems, life science, governments, and philanthropic foundations to improve health journeys for their users, patients, or members. We offer our enterprise technology as a digital front-end to help navigate people to the most appropriate care.

By partnering with TMCG, we will make Ada’s AI accessible to the target population in Uganda. This integration enables TMCG to further expand its subscription based care packages due to the efficiencies of an AI-powered telehealth model. 

The combined solution by Ada and TMCG holds immense value for the Ugandan health system by offering a high quality, highly scalable solution to address the gaps in access to care. While initial revenue sources will include donor funding, we will be working towards a sustainable funding model that will likely include contributions from the government or commercial partnerships (e.g. private insurers and mobile network operators), both of which are only feasible once some evidence of impact from the integration has been achieved.

Similar to the operational model built by Telenor Health with their product, Tonic, in Bangladesh, we aim to scale the operating model built by Ada and TMCG by collaborating with mobile network operators and insurers to enable improved mobile health insurance offerings. These offerings can help to mitigate financial risk and bring awareness to the benefits of insurance, while serving the user with a health assessment and telemedicine services, if needed. The Ada-TMCG partnership seeks to provide a seamless experience to access affordable micro-insurance with primary care health services, directly through a digital platform on a person’s mobile phone. This solution will allow access to healthcare services in Uganda through scalable networks continuously at a nominal cost to the end user. 

Ada’s Global Health Initiative has successfully received grants from leading philanthropic organizations to advance our work in LMICs. Our supporters include the Botnar Foundation, Bayer as well as the Rockefeller Foundation, enabling collaborations with local stakeholders in the Global South to tailor our solution to the local context and needs.