Companies Focused on Using AI and NLP for Patient Education, Behavior Modification: Part 2
In the last article
, I took a look at 2 of the companies that creating virtual health assistants for direct-to-person coaching, lifestyle change, and adherence: CodeBaby and Intelligent Digital Avatars. This time, we’ll delve into 2 more companies in this space. (In alphabetical order.)
MedRespond’s Healthcare Guide system, developed by researchers at Carnegie Mellon University over the past decade, invites patients to engage in a dialogue with an expert on video—healthcare providers, researchers, and fellow patients. This company uses natural language processing (NLP) to understand the intent of the patient’s need(s), and then triggers the human expert in a video recording to discuss this exact need, which lets patients see and hear real healthcare professionals answer their questions via simulated conversations. It is MedRespond’s feeling that this approach gives patients the information they need with a very human touch.
MedRespond’s agents can handle a variety of diseases, including asthma and heart disease, surgical procedures, as well as very sensitive issues such as palliative and end-of-life care. For the latter, its system helps patients understand their options, and then allows them to explore their concerns—genuine fears of death and how they can help their families cope. In this way MedRespond feels that a Virtual Healthcare Guide can provide a safe, private place for patients to work through their thoughts and priorities around some of the most personal and challenging healthcare choices.
MedRespond has also created a surgical preparation application. As an example, patients preparing for open-heart surgery are given instructions and answers to their questions about their surgery the week and day before surgery as well as during recovery. The Virtual Healthcare Guide explains every aspect of the procedure— the medications used, surgical prep, what to eat and drink, the importance of using a spirometer (with instructions), etc. In addition to providing information and instruction around cardiac procedures, the Guide helps patients deal with a range of personal and emotional issues associated with the surgery, such as how frightening it can be when they first awake after surgery.
Patients can access the Guide online, anytime, anywhere. They can revisit the Guide as often as they choose and can share the information with family and friends. The average time on the site during the rollout has been 1 hour and 20 minutes, although some patients actually “conversed” for well over 2 hours. The most common question asked was “Will I be in pain?”
Of interest is how patients respond to the Virtual Healthcare Guide—especially their acceptance that such guides are comforting resources that add to the patient experience. It is MedRespond’s belief that patients welcome the availability of a Virtual Healthcare Guide and have positive responses when they engage with a Virtual Healthcare Guide.
The patient experience is reflected in the way that patients interact with the Guide. When the Guide asks if the patient would like to see a tour of the intensive care unit, patients politely respond, “No, thank you.” Patients who have to exit the program tell their Guide, “I’ll be right back.” They are treating the agent not as a machine, but as they would treat a human.
Through post-surgery interviews with patients using the program, MedRespond discovered that patients who were initially nervous and apprehensive before surgery, reported being comforted and calmed by the information and support from the Virtual Healthcare Guide.
Next IT Corp
(Disclosure: I work for Next IT as the chief medical officer.)
Next IT, another NLP and artificial intelligence company uses the term “human emulation” to describe its overall approach to developing custom-built intelligent virtual health assistants (VHA). Housed on the sponsoring company’s servers, Next IT’s agents emulate the actions of dozens of human jobs in a variety of industries such as human resources, banking, investment, insurance company customer service, airline reservation agents, US Army recruiters, railroad ticket agents, health insurance enrollment, and, most recently, healthcare. Their Alme platform is a VHA or virtual health coach for direct-to-person (ie, patient) interaction.
The Alme VHAs can literally enter into conversations with patients and since they have names, personalities, a sense of humor, favorite colors, actors, food dishes and more, people actually develop an relationship with them and have been known to flirt! The VHAs can remind people of a medication dose, ask how they are doing, perform a validated depression screen (or other clinical assessment such as a disease activity score or health assessment questionnaire score), activate the camera on a smart device to photograph a drug reaction or rash, assist with financial assistance programs, help with a prior authorization process, and, if needed, transfer a patient to a live professional in real time. The VHA could literally, on command, dial 911 and transmit the actual GPS coordinates and “talk” to the 911 operator about the emergency.
The Alme VHA platform is able to basically become the brain for a variety of other devices and adherence technologies by sweeping data from apps, intelligent pill bottles, glucometers, exercise and sleep monitors, predictive modeling software, pharmacy dispensing records, claim systems, and a myriad of other data sources so that a complete dashboard of health data can be presented to a patient, physician, pharmacist, or health plan (all in a HIPAA-compliant manner). The platform can then discuss pertinent findings, deliver incentives, explain the need for adherence, encourage and trigger the desired behaviors, and even notify accountability partners and selected social groups, such as competing exercise or weight loss teams, all with user and sponsor configurability for each function.
Programmed with a “motivational interviewing” mode of conversing, the agents can engage in numerous ways to promote motivation, break down complex behaviors into smaller pieces and trigger their behavior using numerous health behavior theories and models such as the transtheoretical, health belief, and social theories mentioned earlier in this series.
Once programmed and integrated with the sponsoring organization’s specific domain, the VHA can literally understand 10,000 different questions on a given disease and answer the questions with an appropriate answer (both technically and at grade level) due to the ability to understand the intent of the human conversing with the VHA.
As the patient “talks” to the agent, the dialogue is scrolled on the screen so the patient can correct any errors in voice recognition, something that is certain to happen in any NLP system on the market. And because of the way the VHA is created, numerous languages can be incorporated into the same platform. The user experience is identical across platforms such as smartphones, tablets, and computers as the VHA can accept the patient input using type, talk, or tapping functions all the while recording into the same data system.
Next IT feels that the use of a VHA is not only needed but also inevitable in promoting adherence and lifestyle changes citing 100 million people with prediabetes as a very large proof point. The company asks: “Do we have any other viable option?”
Next, bringing it all together.