Doctor on Demand

Empowering patients and improving health outcomes with expanded access to diagnosis and care for neurological disorders.

There is a global health crisis in which the general public needs an affordable system to predict neurological crises. This includes both biological crises, such as brain strokes, and psychological crises, such as depression and anxiety. 

The current lack of such a system is a devastating pain point. Although these neurological crises can be detected by a medical doctor or psychiatrist, there is currently a shortage of doctors globally. This, along with many other factors, has led to exorbitant health prices, especially in the United States. This is particularly devastating for young adults, the population with some of the highest rates of anxiety and depression, but that also has the least income. According to the National Institute of Mental Health, 30.6% of young adults aged 18-25, 25.3% of adults aged 26-49, and 14.5% of adults aged 50+ have some form of a mental condition (NIMH, 2020). (https://www.nimh.nih.gov/health/statistics/mental-illness) Lack of mental health support is also detrimental to older adults, the population with some of the highest rates of depression, suicide, and brain stroke, but also the population that is more likely to lack family to support them through recovery.

Psychological disorders are not the only healthcare problem that are going untreated and undiagnosed. Currently, the only way for people to predict brain stroke without a doctor is by observing visible symptoms, like numbness and slurred speech. However, the problem with this method is that often people are unprepared to react in the case that someone is actually having a brain stroke since they likely do not expect it. Panic ensues, and by the time the patient reaches the hospital, it may be too late. Even after medical attention, they may still end up with a disability, such as post-stroke aphasia. 

Secondly, services exist to help customers deal with depression and anxiety, most notably human therapists, but these are expensive and highly stigmatized among older generations. Therefore, another pain point is the lack of alternatives to face-to-face therapy sessions with another person. Additionally, even if a person has a human therapist, they may not have anyone to talk with in between therapy sessions.

Doctor on Demand, or DoD, is an exciting new app that integrates cutting-edge machine learning technologies for the treatment and prediction of psychological and physiological problems of the brain. DoD's brain stroke prediction model, mental health chatbot, and mental health crisis mode are all incredibly impactful because they provide free healthcare for anyone with an Internet connection. 

A user must complete a comprehensive (though painless!) survey that evaluates their risk for brain stroke with 8 different factors, including age, hypertension, and smoking. The survey relies on an artificial intelligence model that learns how strongly each factor correlates with brain stroke. Then, the algorithm takes these inputs and translates them into a single number prediction with a 93% accuracy rate, far exceeding the gold standard for similar healthcare products. 

Next, our mental health chatbot communicates directly with the patient, alleviating loneliness and offering advice to improve mental health. The chatbot responds like a therapist (after all, the bot was trained with real data of patient-therapist conversations) and engages the patient with different ‘emotions’ and questions. Natural language processing, an applied AI program, is used so the chatbot can understand the intent of the conversation, rather than manufacture generic answers to queries. 

The last part of Doctor on Demand is the mental health crisis mode. While chatting with the chatbot, if the patient uses extremely negative words (indicating depression, self-harm, anxiety, or more) repeatedly in a short amount of time, then emergency contacts are warned of the user’s behavior. To accomplish this, we added a rule bot that searches for keywords amongst thousands of words that are preloaded to trigger the crisis mode. 

Doctor on Demand wants to improve the lives of people who struggle with neurological disorders and suffer from a lack of access to healthcare. Young adults and teenagers, especially Gen Z, are the demographic at the highest risk for developing a mental illness. On the other end of the spectrum, older generations struggle with isolation, dementia, and anxiety disorders, while physically, they are most susceptible to brain stroke. One thing remains consistent–those who need help rarely receive it. Healthcare expenses are climbing, hospitals across the nation have closed, and access is less available than ever. There has never been a greater need for a free product that comprehensively addresses physiological and psychological problems.

Therefore, having an all-inclusive mental healthcare app can be helpful to a wide segment of the population across ages. DoD’s unique emergency contacts feature is particularly helpful for individuals living alone, especially the elderly and teenagers/young adults living away from home. Furthermore, DoD’s self-check-up tool for brain stroke risk will best serve customers ages 55+ who are at the greatest risk of stroke. Finally, our mental health chatbot fights against loneliness while providing the expert perspective of a therapist whose responses were used to train the algorithm. DoD doesn’t neglect the human touch when it comes to patients–rather, it prioritizes care.

Founded in May 2022, the DoD creators formed a strong team of 3 members brought together by a shared interest in technology. We are representative of those we serve because we ourselves are among the many adolescents today who either have been exposed to or have experienced first-hand the mental health disorders that we want to treat. These experiences empowered us to create DoD as a way of healing and coming to terms with our past feelings and struggles, instead of pretending they never happened. 

We all already had a background in AI, and we were motivated to address the neurological health crisis that has long been present. Janvi and Meenal founded the Neuroscience Club at our school, which motivated them to combine technology and neuroscience to make an app for neurological crises. Janvi experimented with brain disorder diagnosis AIs in the past and handled most of the brain stroke AI design, while Meenal was long-interested in NLP and took over the chatbot. Along with being experienced in AI development, Sruti is a talented website and graphic designer who handled website development and video/visual aid design. She also runs her own nonprofit, which gives her ample experience in marketing and fundraising. Our diverse interests and skills were the perfect fit to create DoD.

One of the greatest obstacles we knew we would face is public mistrust of artificial intelligence in healthcare. Much of the public is still hesitant of using AI for medical treatment or diagnoses, which could hinder Doctor on Demand’s growth rate. To study this issue and find ways to improve our marketing of DoD, last summer we collaborated with professors at the Psychology Department of Glendale Community College (a local community college) to conduct research on public trust in AI in healthcare. By doing so, we take the public’s thoughts and suggestions to improve upon our approach towards the marketing and implementation of Doctor on Demand. This has also helped us understand the needs of the diverse populations we want to serve. 

One great example of how we implemented findings from our research into our marketing plans is our special outreach to people in indigenous communities. One of the comments we received from a participant who self-identified as American Indian or Alaskan Native was “The care at [IHS] hospitals are biased because the doctors at these hospitals are temporary. The turnover rate of these doctors makes the Native community see them as untrustworthy.” This comment was echoed by other American Indian or Alaskan Native participants, who also had a lower trust rating of healthcare and human doctors in general than other demographics surveyed. To address this, we organized outreach initiatives with a community center at Fort McDowell Yavapai Nation Reservation, a nearby indigenous community, where we surveyed residents about their thoughts about mental healthcare in general and also helped them install and use our free app in beta. 

Finally, we are also engaging potential users in the development of DoD by recruiting over 150 people from around the world to test our app and answer short surveys about their experience.

DoD incorporates many new features that make it an innovator in the field of mental and neurological health. Firstly, both its AI models use real-life data released by medical institutions, making the model uniquely adept at handling real medical situations. All of our training data is ethically obtained, and we keep all patient data confidential, concerns that have recently arisen with other chatbots, such as Youper and Woebot. These popular mental health apps have been accused of selling patient information to advertisers, with their policy statements even stating "Privacy not Included." [https://foundation.mozilla.org/en/blog/top-mental-health-and-prayer-apps-fail-spectacularly-at-privacy-security/#:~:text=There%20are%20six%20worst%20offenders,Pray.com%2C%20and%20Talkspace.] 

Additionally, they have not made the databases they use to train their bots public, whereas all of our training databases are publicly available, improving our business transparency. In this way, we expect DoD to change the market for medical technologies by drastically improving the privacy of user information and transparency of datasets used to train new AI technologies.

DoD is also new in that it integrates both care for psychological illnesses–common among younger generations– and physiological illness–common among older generations. This uniquely increases DoD's applications in the wider community, changing the market for medical AI technologies. Currently, medical AI models almost always only have one specific function (ex. diagnosis of one specific disease). DoD will enable broader positive impacts in the field of medical AI by showing that multipurpose technologies in medicine are possible. 

Finally, DoD uses cutting-edge new technology such as transformers, a language analysis tool that was only discovered in 2017. They aren't yet mainstream, making our model a leader in NLP. [https://thesequence.substack.c...]

We aim to impact at least 5000 users in the first few months, with at least 3000 active users within the first few months and an engagement rate of at least 60%. Although these may seem like low numbers, we believe this is realistic, considering that we are a startup with no previously-existing connections in the business world and limited money. We will start in our own community of Arizona, where 3227 people out of every 100,000 die from a stroke. [https://www.cdc.gov/nchs/press...] By predicting brain strokes before they happen, DoD will help people take preventative measures against brain strokes. This could save the lives of at least a couple hundred older Arizonans every year, and the number of lives saved will increase exponentially as we expand to other areas, benefiting the quality of life of individuals and mankind in general. Also, our model could speed up stroke diagnosis rates by identifying individuals as high or low-risk, saving doctors time. This can contribute to lowering the approximately $750 million wasted yearly by inefficiencies in medical diagnosis and treatment. [https://khn.org/morning-breako...] Similar healthcare benefits can be expected by improved mental health, which our chatbot works towards. 

One of our target segments is individuals who are seeking support for mental health conditions such as anxiety, stress conditions, depression, stress, and trauma, as well as those who are looking for general wellness and self-care support. DoD’s unique emergency contacts feature is particularly helpful for individuals living alone, especially the elderly and teenagers/young adults living away from home. Patients who have had a stroke and their caregivers may also be able to benefit from DoD’s AI-based stroke diagnostic technologies to help with monitoring, recovery, and rehabilitation.

We also intend to sell group subscription packages of DoD to larger organizations and companies, such as schools/universities, employers, healthcare providers, and insurance companies. Schools and universities can offer the chatbot to support students who may be struggling with their mental health, further increasing our impact on younger people. Similarly, companies may include the chatbot and brain stroke risk assessment as a mental health and wellness benefit for their employees. Furthermore, in addition to being used in the hospital or clinic setting to assist in the early detection and risk-monitoring of brain stroke cases, our brain stroke risk assessment feature can also help emergency responders and paramedics predict whether a patient is experiencing a stroke or another conditions entirely, as many times emergency responders do not know what medical condition the patient has when they are first called. In this way, DoD can be used in the pre-hospital setting to optimize patient health. Finally, insurance companies can also broaden their mental health coverage by providing DoD to their customers, allowing DoD to impact a wider range of people from different age groups.

DoD has 3 key technologies: the brain stroke prediction model, the mental health chatbot, and the mental health crisis warning system. All of our AI models were coded using Python.

Our brain stroke prediction model is an AI model trained with a dataset of information on 5110 real stroke patients. For each patient, this database includes their age, gender, BMI, their history of heart disease, smoking, hypertension, brain strokes, and more. From this information, we trained a convolutional neural network (CNN) to learn which factors correlate with brain stroke. After the patient completes a survey inputting their information (age, gender hypertension, etc), the app produces a percentage showing their chance of getting a brain stroke.

Secondly, DoD's mental health chatbot is an AI model trained on a dataset of 25548 patients talking to a licensed human therapist. The database contains what each patient said to the therapist, the therapist's response, and the topic of the question (ex. depression, anxiety) The AI model learned from this database to interpret and respond to patients' questions in real-time using a transformer. Transformers are a new tool in NLP that learns which words in the input question to pay attention to by taking into consideration the entire past conversation and picking up on keywords to guide its response. This is in contrast to other forms of NLP, such as recurrent neural networks (RNN) and long-short-term memory networks (LSTM), which only have a very short-term memory. Transformers have an encoder-decoder architecture–the encoder encodes the input sequence into a context vector, while the decoder reads from the context vector and predicts what words are commonly associated with responses to the input. Transformers also incorporate multi-level attention, which means the many "heads" of the AI model analyze the input to create weighted values for the response, which are then concatenated to make the final response.

The third layer of our innovation is a simple rule-based bot that picks up on trigger words in a user's conversation with the chatbot. If the patient uses suicide or self-harm words/phrases repetitively within a small period of time, the bot will activate emergency mode and message the users' emergency contacts.

Doctor on Demand currently operates on a global scale, as we’ve implemented it for testing in the US, India, and North Africa, and are now looking for more dedicated chapters. The main objective of DoD was to spread our message of health equity and train others, so launching our product on the large scale will take another step. However, so far, we have 150+ people using our trial product. By the end of this year, and bringing attention to our project, we want to officially launch our app and increase the number of users to at least in the thousands. As more people learn of our innovation, the quicker it spreads, and the more lives it impacts.

The two current dominating, free mental health chatbots are Woebot and Wysa, which will pose significant market barriers for us as DoD tries to break into the healthcare market. However, DoD’s automatic emergency contact feature is quite unique from both. Woebot has a somewhat comparable feature that is triggered by keywords mentioning suicide or self-harm; however, it only suggests suicide hotlines or prompts users to manually call a trusted person. Wysa is similar, except it is triggered by an emergency button. Therefore, DoD is unique from its competitors because it automatically contacts the user’s emergency contacts to most readily manage a severe mental health crisis. Furthermore, there is no free brain stroke detection and risk assessment app available for individuals; rather, they are mostly only available for use in hospitals and clinics. In contrast, DoD is readily accessible to all phone users. It also uses data from thousands of brain stroke patients to train and test its AI, ensuring accuracy across patient types. All of these creative features may help DoD overcome the market barrier. 

Some of the advantages of mental health chatbots may include anonymity, affordability, and providing support in a non-stigmatizing way, helping us overcome the cultural barrier of stigma surrounding mental health. However, some of the disadvantages may include their inability to provide the same emotional connection as a human therapist, creating social barriers as people continue to prefer human interaction.

Some of the advantages of our brain stroke AI include analyzing large amounts of medical data and identifying patterns that may not be immediately apparent to human doctors, which can lead to more accurate diagnoses and better outcomes for patients. It can also help to bring down the overall costs of medical treatment. However, some technical barriers include the fact that since AI-powered systems rely on the data they are trained on, if the data is not complete, diverse, and accurate the system will not be able to perform optimally. 

With all of these benefits and limitations in mind, Doctor on Demand’s positioning is as a complementary tool to professional help, not a standalone.

Currently, Doctor on Demand has partnered with several large nonprofits across the world to promote our message, including NeuroEd, Civication, Psychonovation, Invest in Her, and Lorraine's Project, along with local libraries, community centers, and schools across Arizona. These organizations have helped us with outreach for our mission of spreading knowledge of neurological disorders, as well as provided us with advertising for volunteer product testers.

To maximize our consumer base, we will compete in both the business-to-consumer (B2C) market and the business-to-business (B2B) market. 

One of our target segments is individuals who are seeking support for mental health conditions such as anxiety, stress conditions, depression, stress, and trauma, as well as those who are looking for general wellness and self-care support. We will provide free subscriptions to individuals.

We also intend to sell paid group subscription packages of DoD to larger organizations and companies, such as schools/universities, employers, healthcare providers, and insurance companies. Schools and universities can offer the chatbot to support students who may be struggling with their mental health. Similarly, companies may include the chatbot and brain stroke risk assessment as a mental health and wellness benefit for their employees. 

Furthermore, in addition to being used in the hospital or clinic setting to assist in the early detection and risk-monitoring of brain stroke cases, our brain stroke risk assessment feature can also help emergency responders and paramedics predict whether a patient is experiencing a stroke or another conditions entirely, as many times emergency responders do not know what medical condition the patient has when they are first called. In this way, DoD can be used in the pre-hospital setting to optimize patient health. Finally, insurance companies can also broaden their mental health coverage by providing DoD to their customers.

DoD will not host any ads, neither on the website nor on the app. We will also make our app free for individual users so that more young people will be able to use it.

However, it is unsustainable to make the app free to all and ad-free, since we have many costs. We expect to pay $15-25k a year on app development and updates, hosting our machine learning models on AWS, hosting the site, and designing our app for both Android and iOS. [https://www.businessofapps.com...] We also foresee spending about $10-20k a year on hiring off-shore contractors to help maintain the app if it crashes or there are extremely complex bugs. Our last major cost is social media marketing on platforms such as Youtube, Instagram, Twitter, and Discord, since we expect to hire a part-time social media worker to help the co-founders with creating a posting and engagement plan. With the employee working only about 10 hours a week and 30 weeks a year, we expect the cost of our social media marketing to only be about $5k a year. Total yearly costs for DoD would therefore amount to about $30k-$50k. It is difficult to quantify the cost per app unit delivered, especially since this drops as users increase, but we estimate this cost to be about $8 per unit per year.

To support these costs, our business model will be an embedded model, and the main revenue source will consist of mass subscriptions sold to businesses, schools, and insurance corporations. Doctor on Demand provides free app subscriptions for individuals, specifically young adults and senior citizens, while our paid mass subscription plans will be available for large organizations like schools, insurance companies, and businesses. We plan on charging these large organizations about $30 per employee they register under their business package. $30 would be large enough to support both the employee's subscription and 2-3 additional individuals, allowing us to provide our services for free for users not affiliated with a business, school, or insurance company that can pay for a subscription for them. 

After extensive research on costs for other app-based businesses, we realized it would require between $1k-3k to release DoD, from running social media to getting AWS servers. Luckily, we received the attention of an eager entrepreneur through our AI advocacy at various technology conferences. The investor, who met Janvi, one of our team members,  at an AI conference, is intrigued by our app’s potential and is willing to support it if the platform gains traction, like with a long customer waitlist. This investor, whose company specializes in AI-based ventures, is willing to provide a team of developers for free to release an IOS/Android app to the public. This will reduce our development and design costs to almost zero. Our only cost prior to app release will be for social media marketing; however, this will have low costs as we will do marketing ourselves during the first stages of the app.