By Amit Sheth | October 20, 2020
At this year’s TEDxUofSC, Amit Sheth discussed faults in American health care and how emerging research in augmented personalized health can prevent many of system’s problems. Below is the transcript from his presentation.
Healthcare is complicated. Your body is a very complex machine, and when there’s a problem with it, it could be due to several causes. Even after a cause has been determined, in the United States, the patients’ access to their doctors may be limited by the rarity of their disease, health insurance, socioeconomic status, or geographic area.
I believe that it does not have to be that way.
I am an engineer and computer scientist, and I founded the Artificial Intelligence Institute at the University of South Carolina. Today I’d like to talk to you about augmented personalized health, our digital health system and artificial intelligence technologies that can help people with chronic health conditions manage their symptoms and prevent severe episodes, regardless of their in-person access to their doctor. Augmented personalized health provides increasingly more comprehensive support for a patient, starting with self-monitoring, then expanding to self-appraisal, self-management, intervention, and disease progression and tracking. This is patient empowerment. This is what patients deserve.
We believe that it should be easier to stay healthy, even if you are not wealthy, even if you live very far from your doctor, even if you are facing a chronic disease.
- Amit Sheth, Founding Director of the UofSC AI Institute
Augmented personalized health systems are being developed for several types of chronic illnesses, but today we are going to use asthma as our example.
Our system for asthma uses multiple tools, including a Fitbit, a peak flow meter, a mobile app, an indoor air quality monitor and web-based services. These are used to measure, record and monitor a patient’s activity, sleep, and breathing; symptoms and medications, environmental factors in their home like temperature, humidity, and smoking, and outdoor environmental factors like pollen count, ozone level, and the level of air-based particulate matter.
This system can collect up to 1852 data points of 29 types for the patient every single day. On a personal scale, it fits the definition of big data.
These data points are stored and analyzed by the augmented personalized health system we have created and named “kHealth.” The K stands for, “knowledge.” Two types of knowledge are of particular interest. The first is knowledge about the patient such as his asthma-related history. This supports personalization. The second is medical knowledge about the disease. This provides the medical context for interpreting the data. For an asthma patient, kHealth can intelligently synthesize the collected data to identify potential triggers that worsen a patient’s condition; it can gauge the effectiveness of the preventative medication; it can inform the patient and the doctor when a patient fails to follow his or her preventative health protocols for exercise, medication, and diet.
The best way to introduce you to kHealth and its potential to improve lives is to introduce you to Amelia.
Amelia is 10-years-old. She lives with her mother in an urban neighborhood. Amelia’s asthma is classified as severe-persistent, which means that flare-ups occur regularly and usually last several days. Coughing and wheezing may disrupt Amelia’s normal activities and make it difficult to sleep. It can be really scary for a child and her family.
According to the standard protocol, Amelia sees the doctor once every three months. At each visit, she is given a physical examination and an Asthma Control Test: a 7-question, memory-based questionnaire that her doctor uses to gauge whether or not Amelia’s asthma is under control.
Once every three months.
For 90 days at a time, a 10-year-old is responsible for monitoring her asthma – a condition that can quickly turn deadly.
Each day Amelia is supposed to take inhaled corticosteroids as a controller medication, and she has a separate rescue medication to take via an inhaler when symptoms occur. The doctor has advised that regular use of controller medication will reduce the symptoms and the need for rescue medication.
Amelia has been hospitalized twice in the last year for asthma attacks. Those attacks and other less severe asthma episodes have resulted in her missing quite a few days of school. Those absences have put her a bit below the middle of the pack, educationally, even though she is an above-average student.
Amelia loves to play volleyball. She often plays her favorite sport outdoors with her team. One evening last spring, Amelia finished her volleyball practice, went home, ate her dinner and went to bed. A little past midnight, she was rushed to the ER after having an asthma attack. She recovered, but she missed three days of school and her mother is facing another copay for a visit to the emergency room.
With the kHealth system, this could have been prevented.
If Amelia was using the kHealth system, then every day the system would record a broad variety of useful information such as her activity, sleep patterns, medication compliances, exposure and symptoms. That’s the “self-monitoring” aspect of the system. With all the collected data, the system would be able to tell her how well controlled her asthma is. This is the “self-appraisal,” benefit of the system. Remember the Asthma Control Test, or ACT? The 7 question, memory-based questionnaire Amelia takes when she sees the doctor? A ten-year-old is responsible for remembering how often she was feeling out of breath in the past four weeks – it’s not the most reliable data.
Instead, kHealth uses all that daily data collected – including her data from sensors, symptoms and medication called by the mobile app, and the inside and outside air quality – to compute a Digital Phenotype Score for Amelia, which our research has proved to be nearly as good as the ACT score. Hence, in practice, Amelia has a daily ACT score, without needing access to her doctor.
Furthermore, with the use of kHealth over a period of time, it would help Amelia’s doctor determine that in her case, her symptoms seem highly sensitive to higher pollen counts, but less sensitive to other potential triggers. On the day in question last spring, kHealth would have alerted Amelia that the pollen count was expected to be high, and it would have reminded that the doctor has advised that she uses her rescue inhaler if she carries out an outdoor activity during the pollen allergy season. This very likely would have prevented her asthma attack. This is the "self-management" aspect of the system where kHealth assists the patient to make better decisions within the scope of her current medical treatment.
This information and more is available to her doctor via an online dashboard. If Amelia’s asthma does not remain well controlled in between two quarterly visits, using the more timely and accurate data, her doctor can advise changes to her maintenance protocol, like medication and activity, before waiting for the next clinic visit. This is the "intervention" aspect of the system. It’s better-informed care, sooner. It gives Amelia a much better chance of avoiding asthma attacks, ER visits, and missed days of school. Over the long-term, the dashboard allows her doctor to identify seasonal trends, changes as she ages, and more. This is "disease progression and tracking."
kHealth is an incredible improvement for Amelia, and can be for the 250 million people around the world who live with asthma.
The hallmark innovation of kHealth is that it is intelligent -- it can “learn” for each individual patient and continually alter its synthesis of new data based on the patient’s individual history, current health status, relevant medical knowledge and the doctor-determined treatment plan. With other systems, you get data and generic health protocol, not highly personalized actionable information.
In addition to asthma patients, kHealth framework is being adopted to serve children with neutropenia, diabetes, and autism spectrum disorder; and to serve adults with mental health problems or recovery after cancer treatment, and those with hypertension or diabetes. Nearly half of the adults in the US have hypertension, and more than 100 million adults in the US are living with diabetes or prediabetes. We are gratified by the prospect of helping so many people.
kHealth is HIPPA compliant. It has been evaluated with 150 pediatric asthma patients at Dayton Children's Hospital, a clinical partner in our NIH-funded project. Before you can expect to see it widely used, it needs to go through a randomized controlled clinical trial.
While it moves through those review and validation processes, we are working to make it even more user-friendly. Soon, patients will be able to complete their daily questionnaire via chatbot or virtual health assistant on their Alexa or Google Assistant.
We believe that it should be easier to stay healthy, even if you are not wealthy, even if you live very far from your doctor, even if you are facing a chronic disease. We are dedicated to employing artificial intelligence technology to improve healthcare outcomes for everyone.
This article has been republished with permission from Dr. Sheth's LinkedIn profile. Read the orginal article.
Acknowledgements: This work is partly supported by the National Institutes of Health under Grant Number 1 R01 HD087132-01: KHealth: Semantic Multisensory Mobile Approach to Personalized Asthma Care. The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.