ChatDoc

Overview

ChatDoc is a chatbot for personal health records that empowers patients to have control over and easy access to their medical records. The MVP is a chatbot that asks you how you are feeling before you see a doctor and securely share that information with your physician.

Key Takeaways

• Khyati and I submitted this to a Facebook Messenger hackathon. Since we didn’t win, we would have really appreciated feedback from the judges :(
• It gave me the excuse to work with one of my best friends :) and an excuse to pick my parents’ brains (they got really excited talking about this idea).
• I now feel pretty strongly about personal agency over personal health records!
• It seems like people feel uncomfortable sharing their personal information (health, SS, address, etc.) via a platform related to Facebook.

Motivation

• Medical records are annoying to access :(
• Chatbots are friendly and conversational.
• Geographic diversity in ML: Born and raised in Tennessee, I care deeply about geographic diversity and increasing access to healthcare. According to Stanford, “patient data from just three states trains most AI diagnostic tools.” Can we expand access to healthcare records while also using geographically diverse health data to train ML?

PROPOSAL

What is the MVP?

The MVP is a Facebook Messenger chatbot for personal health records. Like how the Common Application both cuts out redundancy and empowers high school seniors to have control over and have easy access to their personal information for college applications, our product offers the same value propositions to both patients and doctors. By seamlessly integrating with doctors’ and patients’ existing workflows, our MVP will cut costs and time for doctors as well as empower patients to have control over and easy access to their medical records.

Our idea leverages the spread and accessibility of the internet to enable seamless communication between clinics and their patients. First, it saves a patient effort and time in the waiting room by auto-filling the patient’s personal and medical information across visits, Second, it gives doctors nurses a way to prepare better for the patient’s visit by giving them the necessary information ahead of time, and third, and most importantly, it gives patients access to their full history of medical records and responses with the click of a button.

Long term, we want to change how medical records are stored and accessed, putting the agency in the hands of the patients. Currently, access to personal health and immunization records are distributed and controlled by hospital systems and private physicians. While it is imperative for doctors to have access to this information, it is also in the interest of the patient to easily have access to their medical records. Additionally, future versions will include translation, as language can prevent immigrants from getting medical help.

Much later, we hope to give patients the ability to opt-in to and be connected with virtual and physical communities, called “motivation groups,” for health conditions (e.g. stage 4 terminal cancer, mental health, etc.). Additionally, we hope to eventually give patients the ability to opt-in to sharing their health data (anonymized and only in aggregate) with researchers.

Facebook Messenger is the ideal platform to launch our product because sharing contact, identification, age, gender, and marital status information with your doctor is redundant and users typically already share this information publicly on Facebook.

How It Works

These questions and decision trees were inspired by real medical questionnaires.

LEGEND

Bot Question/Reply is bolded and denoted by [B]

User Question/Reply is bulleted an denoted by:

• Quick reply [UQR]
• User-generated text [UGT]
• User information field from Facebook [UIF]

Explanatory notes are italicized

Section I: Introduction

[B] Hi <First Name>! Welcome to <Bot Name>. Save time in the waiting room by sharing your medical information directly with your doctor.

[B] Who are you here to see today?

• [UQR] General Practitioner
• [UQR] Dermatologist
• [UQR]Other

[B] What’s bothering you today?

• [UGT] At this point, the user replies with a long form free text

[B] Sorry to hear that you’re dealing with <UGT Reply to What’s bothering you today>. Can we share your personal health information with your doctor to cut the waiting room time?

This question functions as CONSENT needed to share Hippa/PPI information

• [UQR] Yes, I want to share my personal health information

If Yes is selected, jump down to Section II

• [UQR] No, Please connect me directly with my doctor instead

If No is selected, we output the Doctor’s contact information and then end the conversation. Such as:

[B] Sorry to see you go! You can contact <Doctor Name> at <phone number> or <website> or <email>

Section II: Confirm Personal Information

[B] Let’s double-check your personal information.

• [UIF] First name
• [UIF] Last name
• [UIF] Middle name
• [UIF] Date of Birth
• [UIF] Marital status
• [UIF] Sex
• [UIF] Race (not available on Facebook API)

After the bot displays the User information field from Facebook [UIF] we have pulled from Facebook, the bot asks

[B] Does this information look correct?

If no is selected, give the user the option to edit:

[B] What’s incorrect?

The following Quick Replies are Generated.

• [UQR] First name
• [UQR] Last name
• [UQR] Middle name
• [UQR] Date of Birth
• [UQR] Marital status
• [UQR] Sex
• [UQR] Race

If information is missing, additionally ask:

[B] Looks like we are missing <missing fields, eg: Date of Birth, Marital Status, and Race>. What’s your <missing field>?

The reply can be user-generated text for now, and we can include quick reply options later

Section III: Contact Information

[B]: Your doctor would like to get in touch with you. Does this contact information look correct?

• Phone Number
• Mailing Address
• Email Address

The user can select “yes” or “no” using the quick reply feature

If no is selected, give the user the option to edit:

[B] What’s incorrect?

The following options are presented as a quick reply.

• Phone Number
• Mailing Address
• Email Address

If information is missing, additionally ask:

[B] Looks like we are missing <missing fields>.

Depending on the missing field, ask:

• [B] What’s your phone number? → reply is user-generated text
• [B] What’s your mailing address? → reply is user-generated text
• [B] What’s your email? → reply is user-generated text

[B] You’re set to see doctor <Doctor Name>.

If a specialist is selected, also ask:

[B] Who is your primary care doctor?

• Reply is user-generated text

[B] What’ is their phone number?

• Reply is user-generated text

Section IV: Employer Information

[B] Does this information about your employer look correct?

The following information is pulled from Facebook and messaged to the user

• Employer Name (full, part time)
• Address

The user can select “yes” or “no” using the quick reply feature

If no is selected, give the user the option to edit:

[B] What’s incorrect?

Offer the following quick replies:

• Employer Name (full, part-time)
• Address

If information is missing, additionally ask:

[B] Looks like we are missing <missing fields>.

Depending on the missing field, ask:

• [B] What’s your Employer’s Name? → reply is user-generated text
• [B] Do you work full-time or part-time? → reply is user-generated text
• [B] What’s your employer’s Address?→ reply is user-generated text

Section VII: Insurance

[B] How many insurance plans do you have?

Offer the following quick replies

• 1
• 2
• 3+

[B] What’s your Insurance company name? → reply is user-generated text

[B] Who is the primary holder?

Offer the following quick replies

• Me
• Someone else

If someone else is selected, ask the following question (with user-generated text as the reply)

• What’s the name of the Subscriber/employee (name)?
• What’s their date of birth?
• What’s their social security number?
• What your Relationship to them?

The bot continues:

[B] Who is the Employer who provides the insurance?

[B] What is the Policy number/ID?

[B] What is the Group #?

[B] Do you have another insurance plan?

Section VIII: Miscellaneous

If the person is under 18 ask the following questions, offering user-generated text as the reply

• Please share information about your “guarantor”?
• What’s their name?
• What’s their address
• What’s their phone?
• What’s their Social Security?
• What’s their DOB
• What’s their relation to you, the patient?
• Who’s their employer?

For visits to a specialist, we may need to ask more specific questions. For example, if the patient is visiting the dermatologist, they would encounter the following workflow:

[B] Have you seen a dermatologist in the last 3 years?

Quick Reply: yes or no

If yes, ask:

• What is the name of that dermatologist
• Referral to office
• Reason for visit

[B] Do you have any dermatology history?

Offer the following quick replies

• Personal
• Skin cancer
• Other disease of the skin (type, date)

[B] Do any skin diseases run in your family

Offer the following quick replies

• eczema
• psoriasis
• moles
• acne

[B] Social history: Do you use sunscreen

Offer “yes” and “no” as quick replies

If yes, ask:

• [B] What type of sunscreen do you use (Brand and SPF)

If patient is female, ask

• [B] What’s the date of your last menstrual cycle?
• [B] Are you nursing?

Why does our idea matter?

Communication through the internet has allowed businesses and ideas to advance. However, the field of medicine and healthcare services remains largely traditional, with reliance on paper forms and physical records of the patient’s history.

Previously, patients filled out the same basic information in every doctor’s (whether it be a general practitioner, dermatologist, or orthopedic surgeon) waiting room right before an appointment. This is frustrating for the patient and wastes time. At the same time, current Electronic Medical Records cause doctors to sink precious time during the physical examination and conversation into typing, which inhibits empathy. Our app gives the patient the ability to document their symptoms and will both provide the information to the doctor in both a legible manner and integrate into the EMR system. This is more efficient, builds empathy, and provides the patient with the opportunity to record their point of view.

Currently, physicians and nurses transcribe patients’ pain and symptoms. According to our medical advisors, a common medical entry takes the form: “The patient complains of dizziness and headache. But from looking at the patient’s history and CT scan, I am unclear as to why the patient continues to have these symptoms.” This anecdote exemplifies how the patient’s voice, which is incredibly important, is lost in translation. However, our product offers patients a way to give their perspective in detail. We are building a new way for patients to communicate with the intricate system of doctors, by efficiently sharing their personal medical and contact information. By being conversational in nature, our MVP generates more empathy than online forms.

Who are our potential users?

Our potential users are patients and doctors who need to share and receive personal, health, and contact information. We have segmented the doctor persona into a general practitioner and specialist. Through our research speaking with physicians in multiple types of practices and comparing forms between specialties, we have found the following differences that require different versions of our core product.

General practitioners are responsible for understanding the patient’s holistic health and thus require extensive social and medical history. For example, a form at a general practitioner’s office asks about any past surgeries, previous medications, and information from specialists. On the other hand, forms at specialists’ offices ask for a brief overview of the patient’s general medical history and often focus on specific parts of the body or medical problem. For example, a dermatology office typically asks: Details of last dermatology visit, Personal and Family Dermatology history (cancers, diseases of the skin), and use of sunscreen.

Finally, the scope of this product is international. Users speaking many languages can use our app abroad.

Competitors

Our competitors are large EMR systems such as Epic, ChartLogic, MyChart, Fusion, NextGen, and more. However, we plan on eventually integrating with these competitors. No direct competitors exist in the Facebook Messenger chat-bot/personal health records space.

Attracting Users

Since we will be using Facebook Messenger as our core platform for messaging and communication, user enrollment will not be an issue. Our plan of attack is getting clinics to adopt our service and allow patients to fill out forms via our service. With the addition of each clinic, we anticipate an addition of 1000 users. Facebook gives us increased access and reach to users.

What would it take to build this?

• Messenger API for messaging and quick data collection.
• Google Cloud for database storage (if the clinic requests data storage service). Languages- Javascript, SQL, HTML.
• React.js with python for separate website creation (later steps)

Challenges

There are no serioussss technical barriers to entry for this solution. However, one challenge we anticipate is the building of psychological trust with our patients. We understand that certain patients might be skeptical of sharing their data via a social media tool like Facebook Messenger. While we will ensure data privacy and full consent for data sharing, we understand that concerns might be raised by clinics and/or patients. To combat this challenge, we plan on being 100% transparent and show how our privacy settings are more rigorous than customizable than traditional Facebook privacy settings. For example, updating one’s medical data will not “share” to a “feed” that their “friends” can view. Since patients have control over their data, we always give patients the ability to withdraw from the program at any time and we will promptly remove their data. We plan on following strict privacy and security protocols.

Next Steps:

• Inspiring trust between our product and users is imperative!
• Getting healthcare providers to accept this is challenging :(

This proposal was written by Ayushi Sinha and Khyati Agrawal.