User login
The power and promise of person-generated health data (Part II)
In Part I of our discussion we introduced the concept of person-generated health data (PGHD), defined as wellness and/or health-related data created, recorded, or gathered by individuals.
Such rich, longitudinal information is now being used in combination with traditional clinical information to predict, diagnose, and formulate treatment plans for diseases, as well as understand the safety and effectiveness of medical interventions.
Identifying a disease early
One novel example of digital technologies being used for early identification of disease was a promising 2019 study by Eli Lilly (in collaboration with Apple and Evidation Health) called the Lilly Exploratory Digital Assessment Study.
[embed:render:related:node:215691]
In this study, the feasibility of using PGHD for identifying physiological and behavioral signatures of cognitive impairment was examined for the purpose of seeking new methods to detect mild cognitive impairment (MCI) in a timely and cost-effective manner. The study enrolled 31 study participants with cognitive impairment and 82 without cognitive impairment. It used consumer-grade sensor technologies (the iPhone, Apple Watch, iPad, and Beddit sleep monitor) to continuously and unobtrusively collect data. Among the information the researchers collected were interaction with the phone keyboard, accelerometer data from the Apple Watch, volume of messages sent/received, and sleep cycles.1
A total of 16 terabytes of data were collected over the course of 12 weeks. Data were organized into a behaviorgram (See Figure 1) that gives a holistic picture of a day in a patient’s life. A machine learning model was used to distinguish between behaviorgrams of symptomatic versus healthy controls, identifying typing speed, circadian rhythm shifts, and reliance on helper apps, among other things, as differentiating cognitively impaired from healthy controls. These behaviorgrams may someday serve as “fingerprints” of different diseases, with specific diseases displaying predictable patterns. In the near future, digital measures like the ones investigated in this study are likely to be used to help clinicians predict and diagnose disease, as well as to better understand disease progression and treatment response.
Leading to better health outcomes
The potential of PGHD to detect diseases early and lead to better health outcomes is being investigated in the Heartline study, a collaboration between Johnson & Johnson and Apple, which is supported by Evidation.2
This study aims to enroll 150,000 adults age 65 years and over to analyze the impact of Apple Watch–based early detection of irregular heart rhythms consistent with atrial fibrillation (AFib). The researchers’ hypothesis is that jointly detecting atrial fibrillation early and providing cardiovascular health programs to new AFib patients, will lead to patients being treated by a medical provider for AFib that otherwise would not have been detected. This, in turn, would lead to these AFib patients decreasing their risks of stroke and other serious cardiovascular events, including death, the study authors speculated.
Presenting new challenges
While PGHD has the potential to help people, it also presents new challenges. It is highly sensitive and personal – it can be as identifying as DNA.3
The vast amount of data that PGHD can collect from interaction with consumer wearable devices poses serious privacy risks if done improperly. To address those risks, companies like Evidation have built in protections. Evidation has an app, Achievement, that has enlisted a connected population of more than 3.5 million members who earn rewards for performing health-related actions, as tracked by wearables devices and apps. Through the Achievement app (See Figure 2.), members are provided opportunities to join research studies. As part of these studies, data collected from sensors and apps is used by permission of the member so that it is clear how their data are contributing to specific research questions or use cases.
This is a collaborative model of data collection built upon trust and permission and is substantially different than the collection of data from electronic health records (EHRs) – which is typically aggregated, deidentified, and commercialized, often without the patients’ knowledge or consent. Stringent protections, explicit permission, and transparency are absolutely imperative until privacy frameworks for data outside of HIPAA regulation catches up and protects patients from discrimination and unintended uses of their data.
Large connected cohorts can help advance our understanding of public health. In one study run on Achievement during the 2017-2018 flu season, a survey was sent to the Achievement population every week asking about symptoms of influenza-like illness and requesting permission to access historical data from their wearable around the influenza-like illness event.4 With the data, it was possible to analyze patterns of activity, sleep, and resting heart rate change around flu events. Resting heart rate, in particular, is shown to increase during fever and at the population level. In fact, through the use of PGHD, it is possible to use the fraction of people with resting heart rate above their usual baseline as a proxy to quantify the number of infected people in a region.5 This resting heart rate–informed flu surveillance method, if refined to increased accuracy, can work in near real time. This means it may be able detect influenza outbreaks days earlier than current epidemiological methods.
Health data generated by connected populations are in the early stages of development. It is clear that it will yield novel insights into health and disease. Only time will tell if it will be able to help clinicians and patients better predict, diagnose, and formulate treatment plans for disease.
Neil Skolnik, M.D. is a professor of family and community medicine at Sidney Kimmel Medical College, Thomas Jefferson University, and associate director of the Family Medicine Residency Program at Abington Jefferson Health. Luca Foschini PhD, is co-founder & chief data scientist at Evidation Health. Bray Patrick-Lake, MFS, is a patient thought leader and director of strategic partnerships at Evidation Health.
References
1. Chen R et al. Developing measures of cognitive impairment in the real world from consumer-grade multimodal sensor streams. KDD ’19. August 4–8, 2019 Aug 4-8.
2. The Heartline Study. https://www.heartline.com.
3. Foschini L. Privacy of Wearable and Sensors Data (or, the Lack Thereof?). Data Driven Investor, Medium. 2019.
4. Bradshaw B et al. Influenza surveillance using wearable mobile health devices. Online J Public Health Inform. 2019;11(1):e249.
5. Radin JM et al. Harnessing wearable device data to improve state-level real-time surveillance of influenza-like illness in the USA: a population-based study. Lancet Digital Health. 2020. doi: 10.1016/S2589-7500(19)30222-5.
In Part I of our discussion we introduced the concept of person-generated health data (PGHD), defined as wellness and/or health-related data created, recorded, or gathered by individuals.
Such rich, longitudinal information is now being used in combination with traditional clinical information to predict, diagnose, and formulate treatment plans for diseases, as well as understand the safety and effectiveness of medical interventions.
Identifying a disease early
One novel example of digital technologies being used for early identification of disease was a promising 2019 study by Eli Lilly (in collaboration with Apple and Evidation Health) called the Lilly Exploratory Digital Assessment Study.
[embed:render:related:node:215691]
In this study, the feasibility of using PGHD for identifying physiological and behavioral signatures of cognitive impairment was examined for the purpose of seeking new methods to detect mild cognitive impairment (MCI) in a timely and cost-effective manner. The study enrolled 31 study participants with cognitive impairment and 82 without cognitive impairment. It used consumer-grade sensor technologies (the iPhone, Apple Watch, iPad, and Beddit sleep monitor) to continuously and unobtrusively collect data. Among the information the researchers collected were interaction with the phone keyboard, accelerometer data from the Apple Watch, volume of messages sent/received, and sleep cycles.1
A total of 16 terabytes of data were collected over the course of 12 weeks. Data were organized into a behaviorgram (See Figure 1) that gives a holistic picture of a day in a patient’s life. A machine learning model was used to distinguish between behaviorgrams of symptomatic versus healthy controls, identifying typing speed, circadian rhythm shifts, and reliance on helper apps, among other things, as differentiating cognitively impaired from healthy controls. These behaviorgrams may someday serve as “fingerprints” of different diseases, with specific diseases displaying predictable patterns. In the near future, digital measures like the ones investigated in this study are likely to be used to help clinicians predict and diagnose disease, as well as to better understand disease progression and treatment response.
Leading to better health outcomes
The potential of PGHD to detect diseases early and lead to better health outcomes is being investigated in the Heartline study, a collaboration between Johnson & Johnson and Apple, which is supported by Evidation.2
This study aims to enroll 150,000 adults age 65 years and over to analyze the impact of Apple Watch–based early detection of irregular heart rhythms consistent with atrial fibrillation (AFib). The researchers’ hypothesis is that jointly detecting atrial fibrillation early and providing cardiovascular health programs to new AFib patients, will lead to patients being treated by a medical provider for AFib that otherwise would not have been detected. This, in turn, would lead to these AFib patients decreasing their risks of stroke and other serious cardiovascular events, including death, the study authors speculated.
Presenting new challenges
While PGHD has the potential to help people, it also presents new challenges. It is highly sensitive and personal – it can be as identifying as DNA.3
The vast amount of data that PGHD can collect from interaction with consumer wearable devices poses serious privacy risks if done improperly. To address those risks, companies like Evidation have built in protections. Evidation has an app, Achievement, that has enlisted a connected population of more than 3.5 million members who earn rewards for performing health-related actions, as tracked by wearables devices and apps. Through the Achievement app (See Figure 2.), members are provided opportunities to join research studies. As part of these studies, data collected from sensors and apps is used by permission of the member so that it is clear how their data are contributing to specific research questions or use cases.
This is a collaborative model of data collection built upon trust and permission and is substantially different than the collection of data from electronic health records (EHRs) – which is typically aggregated, deidentified, and commercialized, often without the patients’ knowledge or consent. Stringent protections, explicit permission, and transparency are absolutely imperative until privacy frameworks for data outside of HIPAA regulation catches up and protects patients from discrimination and unintended uses of their data.
Large connected cohorts can help advance our understanding of public health. In one study run on Achievement during the 2017-2018 flu season, a survey was sent to the Achievement population every week asking about symptoms of influenza-like illness and requesting permission to access historical data from their wearable around the influenza-like illness event.4 With the data, it was possible to analyze patterns of activity, sleep, and resting heart rate change around flu events. Resting heart rate, in particular, is shown to increase during fever and at the population level. In fact, through the use of PGHD, it is possible to use the fraction of people with resting heart rate above their usual baseline as a proxy to quantify the number of infected people in a region.5 This resting heart rate–informed flu surveillance method, if refined to increased accuracy, can work in near real time. This means it may be able detect influenza outbreaks days earlier than current epidemiological methods.
Health data generated by connected populations are in the early stages of development. It is clear that it will yield novel insights into health and disease. Only time will tell if it will be able to help clinicians and patients better predict, diagnose, and formulate treatment plans for disease.
Neil Skolnik, M.D. is a professor of family and community medicine at Sidney Kimmel Medical College, Thomas Jefferson University, and associate director of the Family Medicine Residency Program at Abington Jefferson Health. Luca Foschini PhD, is co-founder & chief data scientist at Evidation Health. Bray Patrick-Lake, MFS, is a patient thought leader and director of strategic partnerships at Evidation Health.
References
1. Chen R et al. Developing measures of cognitive impairment in the real world from consumer-grade multimodal sensor streams. KDD ’19. August 4–8, 2019 Aug 4-8.
2. The Heartline Study. https://www.heartline.com.
3. Foschini L. Privacy of Wearable and Sensors Data (or, the Lack Thereof?). Data Driven Investor, Medium. 2019.
4. Bradshaw B et al. Influenza surveillance using wearable mobile health devices. Online J Public Health Inform. 2019;11(1):e249.
5. Radin JM et al. Harnessing wearable device data to improve state-level real-time surveillance of influenza-like illness in the USA: a population-based study. Lancet Digital Health. 2020. doi: 10.1016/S2589-7500(19)30222-5.
In Part I of our discussion we introduced the concept of person-generated health data (PGHD), defined as wellness and/or health-related data created, recorded, or gathered by individuals.
Such rich, longitudinal information is now being used in combination with traditional clinical information to predict, diagnose, and formulate treatment plans for diseases, as well as understand the safety and effectiveness of medical interventions.
Identifying a disease early
One novel example of digital technologies being used for early identification of disease was a promising 2019 study by Eli Lilly (in collaboration with Apple and Evidation Health) called the Lilly Exploratory Digital Assessment Study.
[embed:render:related:node:215691]
In this study, the feasibility of using PGHD for identifying physiological and behavioral signatures of cognitive impairment was examined for the purpose of seeking new methods to detect mild cognitive impairment (MCI) in a timely and cost-effective manner. The study enrolled 31 study participants with cognitive impairment and 82 without cognitive impairment. It used consumer-grade sensor technologies (the iPhone, Apple Watch, iPad, and Beddit sleep monitor) to continuously and unobtrusively collect data. Among the information the researchers collected were interaction with the phone keyboard, accelerometer data from the Apple Watch, volume of messages sent/received, and sleep cycles.1
A total of 16 terabytes of data were collected over the course of 12 weeks. Data were organized into a behaviorgram (See Figure 1) that gives a holistic picture of a day in a patient’s life. A machine learning model was used to distinguish between behaviorgrams of symptomatic versus healthy controls, identifying typing speed, circadian rhythm shifts, and reliance on helper apps, among other things, as differentiating cognitively impaired from healthy controls. These behaviorgrams may someday serve as “fingerprints” of different diseases, with specific diseases displaying predictable patterns. In the near future, digital measures like the ones investigated in this study are likely to be used to help clinicians predict and diagnose disease, as well as to better understand disease progression and treatment response.
Leading to better health outcomes
The potential of PGHD to detect diseases early and lead to better health outcomes is being investigated in the Heartline study, a collaboration between Johnson & Johnson and Apple, which is supported by Evidation.2
This study aims to enroll 150,000 adults age 65 years and over to analyze the impact of Apple Watch–based early detection of irregular heart rhythms consistent with atrial fibrillation (AFib). The researchers’ hypothesis is that jointly detecting atrial fibrillation early and providing cardiovascular health programs to new AFib patients, will lead to patients being treated by a medical provider for AFib that otherwise would not have been detected. This, in turn, would lead to these AFib patients decreasing their risks of stroke and other serious cardiovascular events, including death, the study authors speculated.
Presenting new challenges
While PGHD has the potential to help people, it also presents new challenges. It is highly sensitive and personal – it can be as identifying as DNA.3
The vast amount of data that PGHD can collect from interaction with consumer wearable devices poses serious privacy risks if done improperly. To address those risks, companies like Evidation have built in protections. Evidation has an app, Achievement, that has enlisted a connected population of more than 3.5 million members who earn rewards for performing health-related actions, as tracked by wearables devices and apps. Through the Achievement app (See Figure 2.), members are provided opportunities to join research studies. As part of these studies, data collected from sensors and apps is used by permission of the member so that it is clear how their data are contributing to specific research questions or use cases.
This is a collaborative model of data collection built upon trust and permission and is substantially different than the collection of data from electronic health records (EHRs) – which is typically aggregated, deidentified, and commercialized, often without the patients’ knowledge or consent. Stringent protections, explicit permission, and transparency are absolutely imperative until privacy frameworks for data outside of HIPAA regulation catches up and protects patients from discrimination and unintended uses of their data.
Large connected cohorts can help advance our understanding of public health. In one study run on Achievement during the 2017-2018 flu season, a survey was sent to the Achievement population every week asking about symptoms of influenza-like illness and requesting permission to access historical data from their wearable around the influenza-like illness event.4 With the data, it was possible to analyze patterns of activity, sleep, and resting heart rate change around flu events. Resting heart rate, in particular, is shown to increase during fever and at the population level. In fact, through the use of PGHD, it is possible to use the fraction of people with resting heart rate above their usual baseline as a proxy to quantify the number of infected people in a region.5 This resting heart rate–informed flu surveillance method, if refined to increased accuracy, can work in near real time. This means it may be able detect influenza outbreaks days earlier than current epidemiological methods.
Health data generated by connected populations are in the early stages of development. It is clear that it will yield novel insights into health and disease. Only time will tell if it will be able to help clinicians and patients better predict, diagnose, and formulate treatment plans for disease.
Neil Skolnik, M.D. is a professor of family and community medicine at Sidney Kimmel Medical College, Thomas Jefferson University, and associate director of the Family Medicine Residency Program at Abington Jefferson Health. Luca Foschini PhD, is co-founder & chief data scientist at Evidation Health. Bray Patrick-Lake, MFS, is a patient thought leader and director of strategic partnerships at Evidation Health.
References
1. Chen R et al. Developing measures of cognitive impairment in the real world from consumer-grade multimodal sensor streams. KDD ’19. August 4–8, 2019 Aug 4-8.
2. The Heartline Study. https://www.heartline.com.
3. Foschini L. Privacy of Wearable and Sensors Data (or, the Lack Thereof?). Data Driven Investor, Medium. 2019.
4. Bradshaw B et al. Influenza surveillance using wearable mobile health devices. Online J Public Health Inform. 2019;11(1):e249.
5. Radin JM et al. Harnessing wearable device data to improve state-level real-time surveillance of influenza-like illness in the USA: a population-based study. Lancet Digital Health. 2020. doi: 10.1016/S2589-7500(19)30222-5.
