The increasing adoption of digital health technologies has unlocked health insights that have been sought after for decades. Wearable devices, in particular, offer a glimpse into individuals’ day-to-day health activities, sleeping patterns, vital signs, and certain contextual information. The ability to continually capture multimodal data is groundbreaking for the pharma industry as it gives context to patients’ lived experiences outside of clinic walls.

As a leader in the collection and disaggregation of person-generated health data (PGHD), Evidation was asked to participate in a workshop put on by the IEEE Engineering in Medicine and Biology Society (IEEE EMBS). The workshop, titled “Measuring Quality of Life with Multimodal Data,” explored how machine learning can convert data into measures for both disease detection and overall quality of life. The meeting report was published in the Journal of Medical Internet Research and provided the following takeaways from the workshop.

Multimodal data and machine learning in early disease detection

PGHD can often discover deviations from normal behaviors before individuals become aware of these deviations on their own. Identifying these fluctuations, according to the IEEE EMBS meeting report, is key in triggering further actions that could lead to disease detection. These actions could include anything from receiving digital interventions from health apps to scheduling a follow-up appointment with a provider.

Evidation has put the idea of early disease detection through multimodal data and machine learning to the test in many research studies. The “Dear Watch, Should I Get a COVID-19 Test? Designing deployable machine learning for wearables” study, in particular, looks at changes in wearable data that could be attributed to either influenza or COVID-19.

For example, increases in resting heart rate, impaired sleep due to illness, and reduced steps due to inactivity have all been correlated with influenza rates at a population level. Taken together across data modes, these changes are more likely reflecting the onset of an infection, as compared to each of the changes taken separately. When a wearable device identifies changes such as these, individuals can be nudged to take a specific further action, i.e. seek out an influenza or COVID-19 confirmatory test.

While the changes in wearable data itself aren’t sufficiently reliable to provide a diagnosis, augmented with symptom reports and then confirmatory tests, they can be triggers and catalysts in the process of achieving better health outcomes.

Multimodal data and machine learning in measuring quality of life and well-being

While multimodal data collection can help with early disease detection, it can also help in measuring individuals’ quality of life and well-being. R&D teams in the pharma industry have seen valid evidence for measures across a range of health-related quality of life-relevant symptoms, including fatigue, depression, stress, anxiety, and independence.

Evidation’s “Predicting Changes in Depression Severity Using the PSYCHE-D (Prediction of Severity Change-Depression) Model Involving Person-Generated Health Data” study is an example of identifying these symptoms and using them to predict changes in an individual’s well-being.

In 2020, an estimated 21.0 million adults in the U.S. had at least one major depressive episode. Getting ahead of the onset of these episodes could bring this number down, which is where PGHD enters the conversation. Researchers in the PSYCHE-D study developed a predictive model using a range of PGHD input features, including objective activity and sleep, self-reported changes in lifestyle and medication from 4000+ individuals that forecast long-term increases in depression severity. This work shows that low-burden PGHD can be the basis of timely warnings that an individual’s mental health may be deteriorating, and hopefully lead to improved engagement and treatment of individuals experiencing depression.

Looking ahead

It should be made clear that multimodal measures are not a replacement for patient-reported outcomes, but rather an additional tool that helps researchers better understand a patient’s lived experience in an unobtrusive manner. Moving forward with multimodal measures, the workshop panel highlighted the need for the industry to conduct more comparative studies and patient-centric research to focus efforts on measures that will be most valuable to the patient.

While there are challenges that lie ahead in multimodal data capture, one thing is certain: there is a lot of optimism for future advancement and progress.

Have questions?

CONTACT US

The increasing adoption of digital health technologies has unlocked health insights that have been sought after for decades. Wearable devices, in particular, offer a glimpse into individuals’ day-to-day health activities, sleeping patterns, vital signs, and certain contextual information. The ability to continually capture multimodal data is groundbreaking for the pharma industry as it gives context to patients’ lived experiences outside of clinic walls.

As a leader in the collection and disaggregation of person-generated health data (PGHD), Evidation was asked to participate in a workshop put on by the IEEE Engineering in Medicine and Biology Society (IEEE EMBS). The workshop, titled “Measuring Quality of Life with Multimodal Data,” explored how machine learning can convert data into measures for both disease detection and overall quality of life. The meeting report was published in the Journal of Medical Internet Research and provided the following takeaways from the workshop.

Multimodal data and machine learning in early disease detection

PGHD can often discover deviations from normal behaviors before individuals become aware of these deviations on their own. Identifying these fluctuations, according to the IEEE EMBS meeting report, is key in triggering further actions that could lead to disease detection. These actions could include anything from receiving digital interventions from health apps to scheduling a follow-up appointment with a provider.

Evidation has put the idea of early disease detection through multimodal data and machine learning to the test in many research studies. The “Dear Watch, Should I Get a COVID-19 Test? Designing deployable machine learning for wearables” study, in particular, looks at changes in wearable data that could be attributed to either influenza or COVID-19.

For example, increases in resting heart rate, impaired sleep due to illness, and reduced steps due to inactivity have all been correlated with influenza rates at a population level. Taken together across data modes, these changes are more likely reflecting the onset of an infection, as compared to each of the changes taken separately. When a wearable device identifies changes such as these, individuals can be nudged to take a specific further action, i.e. seek out an influenza or COVID-19 confirmatory test.

While the changes in wearable data itself aren’t sufficiently reliable to provide a diagnosis, augmented with symptom reports and then confirmatory tests, they can be triggers and catalysts in the process of achieving better health outcomes.

Multimodal data and machine learning in measuring quality of life and well-being

While multimodal data collection can help with early disease detection, it can also help in measuring individuals’ quality of life and well-being. R&D teams in the pharma industry have seen valid evidence for measures across a range of health-related quality of life-relevant symptoms, including fatigue, depression, stress, anxiety, and independence.

Evidation’s “Predicting Changes in Depression Severity Using the PSYCHE-D (Prediction of Severity Change-Depression) Model Involving Person-Generated Health Data” study is an example of identifying these symptoms and using them to predict changes in an individual’s well-being.

In 2020, an estimated 21.0 million adults in the U.S. had at least one major depressive episode. Getting ahead of the onset of these episodes could bring this number down, which is where PGHD enters the conversation. Researchers in the PSYCHE-D study developed a predictive model using a range of PGHD input features, including objective activity and sleep, self-reported changes in lifestyle and medication from 4000+ individuals that forecast long-term increases in depression severity. This work shows that low-burden PGHD can be the basis of timely warnings that an individual’s mental health may be deteriorating, and hopefully lead to improved engagement and treatment of individuals experiencing depression.

Looking ahead

It should be made clear that multimodal measures are not a replacement for patient-reported outcomes, but rather an additional tool that helps researchers better understand a patient’s lived experience in an unobtrusive manner. Moving forward with multimodal measures, the workshop panel highlighted the need for the industry to conduct more comparative studies and patient-centric research to focus efforts on measures that will be most valuable to the patient.

While there are challenges that lie ahead in multimodal data capture, one thing is certain: there is a lot of optimism for future advancement and progress.

Have questions?

CONTACT US

The increasing adoption of digital health technologies has unlocked health insights that have been sought after for decades. Wearable devices, in particular, offer a glimpse into individuals’ day-to-day health activities, sleeping patterns, vital signs, and certain contextual information. The ability to continually capture multimodal data is groundbreaking for the pharma industry as it gives context to patients’ lived experiences outside of clinic walls.

As a leader in the collection and disaggregation of person-generated health data (PGHD), Evidation was asked to participate in a workshop put on by the IEEE Engineering in Medicine and Biology Society (IEEE EMBS). The workshop, titled “Measuring Quality of Life with Multimodal Data,” explored how machine learning can convert data into measures for both disease detection and overall quality of life. The meeting report was published in the Journal of Medical Internet Research and provided the following takeaways from the workshop.

Multimodal data and machine learning in early disease detection

PGHD can often discover deviations from normal behaviors before individuals become aware of these deviations on their own. Identifying these fluctuations, according to the IEEE EMBS meeting report, is key in triggering further actions that could lead to disease detection. These actions could include anything from receiving digital interventions from health apps to scheduling a follow-up appointment with a provider.

Evidation has put the idea of early disease detection through multimodal data and machine learning to the test in many research studies. The “Dear Watch, Should I Get a COVID-19 Test? Designing deployable machine learning for wearables” study, in particular, looks at changes in wearable data that could be attributed to either influenza or COVID-19.

For example, increases in resting heart rate, impaired sleep due to illness, and reduced steps due to inactivity have all been correlated with influenza rates at a population level. Taken together across data modes, these changes are more likely reflecting the onset of an infection, as compared to each of the changes taken separately. When a wearable device identifies changes such as these, individuals can be nudged to take a specific further action, i.e. seek out an influenza or COVID-19 confirmatory test.

While the changes in wearable data itself aren’t sufficiently reliable to provide a diagnosis, augmented with symptom reports and then confirmatory tests, they can be triggers and catalysts in the process of achieving better health outcomes.

Multimodal data and machine learning in measuring quality of life and well-being

While multimodal data collection can help with early disease detection, it can also help in measuring individuals’ quality of life and well-being. R&D teams in the pharma industry have seen valid evidence for measures across a range of health-related quality of life-relevant symptoms, including fatigue, depression, stress, anxiety, and independence.

Evidation’s “Predicting Changes in Depression Severity Using the PSYCHE-D (Prediction of Severity Change-Depression) Model Involving Person-Generated Health Data” study is an example of identifying these symptoms and using them to predict changes in an individual’s well-being.

In 2020, an estimated 21.0 million adults in the U.S. had at least one major depressive episode. Getting ahead of the onset of these episodes could bring this number down, which is where PGHD enters the conversation. Researchers in the PSYCHE-D study developed a predictive model using a range of PGHD input features, including objective activity and sleep, self-reported changes in lifestyle and medication from 4000+ individuals that forecast long-term increases in depression severity. This work shows that low-burden PGHD can be the basis of timely warnings that an individual’s mental health may be deteriorating, and hopefully lead to improved engagement and treatment of individuals experiencing depression.

Looking ahead

It should be made clear that multimodal measures are not a replacement for patient-reported outcomes, but rather an additional tool that helps researchers better understand a patient’s lived experience in an unobtrusive manner. Moving forward with multimodal measures, the workshop panel highlighted the need for the industry to conduct more comparative studies and patient-centric research to focus efforts on measures that will be most valuable to the patient.

While there are challenges that lie ahead in multimodal data capture, one thing is certain: there is a lot of optimism for future advancement and progress.

Have questions?

CONTACT US

The increasing adoption of digital health technologies has unlocked health insights that have been sought after for decades. Wearable devices, in particular, offer a glimpse into individuals’ day-to-day health activities, sleeping patterns, vital signs, and certain contextual information. The ability to continually capture multimodal data is groundbreaking for the pharma industry as it gives context to patients’ lived experiences outside of clinic walls.

As a leader in the collection and disaggregation of person-generated health data (PGHD), Evidation was asked to participate in a workshop put on by the IEEE Engineering in Medicine and Biology Society (IEEE EMBS). The workshop, titled “Measuring Quality of Life with Multimodal Data,” explored how machine learning can convert data into measures for both disease detection and overall quality of life. The meeting report was published in the Journal of Medical Internet Research and provided the following takeaways from the workshop.

Multimodal data and machine learning in early disease detection

PGHD can often discover deviations from normal behaviors before individuals become aware of these deviations on their own. Identifying these fluctuations, according to the IEEE EMBS meeting report, is key in triggering further actions that could lead to disease detection. These actions could include anything from receiving digital interventions from health apps to scheduling a follow-up appointment with a provider.

Evidation has put the idea of early disease detection through multimodal data and machine learning to the test in many research studies. The “Dear Watch, Should I Get a COVID-19 Test? Designing deployable machine learning for wearables” study, in particular, looks at changes in wearable data that could be attributed to either influenza or COVID-19.

For example, increases in resting heart rate, impaired sleep due to illness, and reduced steps due to inactivity have all been correlated with influenza rates at a population level. Taken together across data modes, these changes are more likely reflecting the onset of an infection, as compared to each of the changes taken separately. When a wearable device identifies changes such as these, individuals can be nudged to take a specific further action, i.e. seek out an influenza or COVID-19 confirmatory test.

While the changes in wearable data itself aren’t sufficiently reliable to provide a diagnosis, augmented with symptom reports and then confirmatory tests, they can be triggers and catalysts in the process of achieving better health outcomes.

Multimodal data and machine learning in measuring quality of life and well-being

While multimodal data collection can help with early disease detection, it can also help in measuring individuals’ quality of life and well-being. R&D teams in the pharma industry have seen valid evidence for measures across a range of health-related quality of life-relevant symptoms, including fatigue, depression, stress, anxiety, and independence.

Evidation’s “Predicting Changes in Depression Severity Using the PSYCHE-D (Prediction of Severity Change-Depression) Model Involving Person-Generated Health Data” study is an example of identifying these symptoms and using them to predict changes in an individual’s well-being.

In 2020, an estimated 21.0 million adults in the U.S. had at least one major depressive episode. Getting ahead of the onset of these episodes could bring this number down, which is where PGHD enters the conversation. Researchers in the PSYCHE-D study developed a predictive model using a range of PGHD input features, including objective activity and sleep, self-reported changes in lifestyle and medication from 4000+ individuals that forecast long-term increases in depression severity. This work shows that low-burden PGHD can be the basis of timely warnings that an individual’s mental health may be deteriorating, and hopefully lead to improved engagement and treatment of individuals experiencing depression.

Looking ahead

It should be made clear that multimodal measures are not a replacement for patient-reported outcomes, but rather an additional tool that helps researchers better understand a patient’s lived experience in an unobtrusive manner. Moving forward with multimodal measures, the workshop panel highlighted the need for the industry to conduct more comparative studies and patient-centric research to focus efforts on measures that will be most valuable to the patient.

While there are challenges that lie ahead in multimodal data capture, one thing is certain: there is a lot of optimism for future advancement and progress.

Have questions?

CONTACT US

The case for multimodal data capture: Deeper insights and increased prediction accuracy

September 8, 2022
Thought Leadership

The case for multimodal data capture: Deeper insights and increased prediction accuracy

September 8, 2022
Thought Leadership

September 8, 2022
Thought Leadership

The case for multimodal data capture: Deeper insights and increased prediction accuracy

September 8, 2022
Thought Leadership
Eve: Evidation's brand mark which is a yellow glowing orb

The increasing adoption of digital health technologies has unlocked health insights that have been sought after for decades. Wearable devices, in particular, offer a glimpse into individuals’ day-to-day health activities, sleeping patterns, vital signs, and certain contextual information. The ability to continually capture multimodal data is groundbreaking for the pharma industry as it gives context to patients’ lived experiences outside of clinic walls.

As a leader in the collection and disaggregation of person-generated health data (PGHD), Evidation was asked to participate in a workshop put on by the IEEE Engineering in Medicine and Biology Society (IEEE EMBS). The workshop, titled “Measuring Quality of Life with Multimodal Data,” explored how machine learning can convert data into measures for both disease detection and overall quality of life. The meeting report was published in the Journal of Medical Internet Research and provided the following takeaways from the workshop.

Multimodal data and machine learning in early disease detection

PGHD can often discover deviations from normal behaviors before individuals become aware of these deviations on their own. Identifying these fluctuations, according to the IEEE EMBS meeting report, is key in triggering further actions that could lead to disease detection. These actions could include anything from receiving digital interventions from health apps to scheduling a follow-up appointment with a provider.

Evidation has put the idea of early disease detection through multimodal data and machine learning to the test in many research studies. The “Dear Watch, Should I Get a COVID-19 Test? Designing deployable machine learning for wearables” study, in particular, looks at changes in wearable data that could be attributed to either influenza or COVID-19.

For example, increases in resting heart rate, impaired sleep due to illness, and reduced steps due to inactivity have all been correlated with influenza rates at a population level. Taken together across data modes, these changes are more likely reflecting the onset of an infection, as compared to each of the changes taken separately. When a wearable device identifies changes such as these, individuals can be nudged to take a specific further action, i.e. seek out an influenza or COVID-19 confirmatory test.

While the changes in wearable data itself aren’t sufficiently reliable to provide a diagnosis, augmented with symptom reports and then confirmatory tests, they can be triggers and catalysts in the process of achieving better health outcomes.

Multimodal data and machine learning in measuring quality of life and well-being

While multimodal data collection can help with early disease detection, it can also help in measuring individuals’ quality of life and well-being. R&D teams in the pharma industry have seen valid evidence for measures across a range of health-related quality of life-relevant symptoms, including fatigue, depression, stress, anxiety, and independence.

Evidation’s “Predicting Changes in Depression Severity Using the PSYCHE-D (Prediction of Severity Change-Depression) Model Involving Person-Generated Health Data” study is an example of identifying these symptoms and using them to predict changes in an individual’s well-being.

In 2020, an estimated 21.0 million adults in the U.S. had at least one major depressive episode. Getting ahead of the onset of these episodes could bring this number down, which is where PGHD enters the conversation. Researchers in the PSYCHE-D study developed a predictive model using a range of PGHD input features, including objective activity and sleep, self-reported changes in lifestyle and medication from 4000+ individuals that forecast long-term increases in depression severity. This work shows that low-burden PGHD can be the basis of timely warnings that an individual’s mental health may be deteriorating, and hopefully lead to improved engagement and treatment of individuals experiencing depression.

Looking ahead

It should be made clear that multimodal measures are not a replacement for patient-reported outcomes, but rather an additional tool that helps researchers better understand a patient’s lived experience in an unobtrusive manner. Moving forward with multimodal measures, the workshop panel highlighted the need for the industry to conduct more comparative studies and patient-centric research to focus efforts on measures that will be most valuable to the patient.

While there are challenges that lie ahead in multimodal data capture, one thing is certain: there is a lot of optimism for future advancement and progress.

Download app