Cardiovascular Diseases
Conditions
Keywords
digital data, digital health
Brief summary
This project seeks to identify and characterize features derived from digital data (e.g. social media, online search, mobile media) which are associated with coronary heart disease (CHD) and related risk factors, and develop models that use digital data and conventional predictive models to predict CHD risk and health care utilization.
Detailed description
Cardiovascular disease is the leading cause of death in the US. While secondary prevention approaches have improved longevity of patients, risk factors and adverse health behaviors (e.g., physical inactivity, smoking) are highly prevalent, and in most contemporary series, less than 1% of adults meet all factors of ideal CV health. The logistics and practicalities of meeting the goal of ideal CV health have not been clearly elucidated. Practice guidelines recommend using the Framingham risk score (FRS) or other risk prediction tools to classify patients' risk of CV disease. These models however are imprecise and there is increasing focus on identifying markers that provide better measures of risk. As digital platforms are increasingly used to document lifestyle and health behaviors, data from digital sources may provide a window into manifestations of novel risk factors and potentially a better characterization of existing risk factors. While it seems like a cliche to mention the profound impact of digital data on everyday lives, there is indeed great substance in the opportunities these new media provide for understanding behavioral, social, and environmental determinants of health. This project seeks to identify and characterize features derived from digital data (e.g. social media, online search, mobile media) which are associated with coronary heart disease (CHD) and related risk factors, and develop models that use digital data and conventional predictive models to predict CHD risk and health care utilization.
Interventions
OTHERSurvey
Interested participants may complete the informed consent online. After informed consent, the participant will be asked to share the digital data types that they use (Facebook, Instagram, Twitter, Google search, step data) and then participants will complete a cross-sectional survey.
Sponsors
University of Pennsylvania
Study design
Observational model
CASE_CONTROL
Time perspective
CROSS_SECTIONAL
Eligibility
Sex/Gender
ALL
Age
30 Years to 74 Years
Healthy volunteers
Yes
Inclusion criteria
* 30 - 74 years of age * Willing to sign informed consent * Primarily English speaking (for language analysis) * Has an account on any of the following digital data platforms (Facebook, Instagram, Twitter Reddit, Google (gmail), or smartphone or wearable device such as Apple Health, Fitbit, Samsung Health, MapMyFitness or Garmin) and willing to share data * If has social media account, Instagram or Facebook, willing to share historical and prospective data (60 days) If has Google (gmail) account, willing to download and share google takeout zip file * If has smartphone or wearable device, willing to share step data * Willing to share access to medical health records * Willing to share healthcare insurance information
Exclusion criteria
* Patient does not meet age inclusion criteria above * Does not use and post on digital data sources we are studying or unwilling to donate data * Patient is in severe distress, e.g. respiratory, physical, or emotional distress * Patient is intoxicated, unconscious, or unable to appropriately respond to questions
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | Through study completion, an average of 3 years | The primary outcome is topics and features (derived using the LDA method for clustering language data). For each participant, we included all available Facebook wall posts from the start of their account history through data collection, regardless of whether they occurred before or after a CHD diagnosis. We examined associations between linguistic features (unigrams, LIWC categories, LDA topics) and cardiovascular case status (CHD presence vs absence) using Pearson correlation and logistic regression. Latent LDA, a systematic method to identify text-based themes, was applied to generate 200 clusters of co-occurring words (topics). For each feature type (unigram, LIWC category, LDA topic), we fit separate logistic regression models and calculated Pearson correlation coefficients to assess predictive value for case status. Each language-derived feature was encoded as a normalized frequency count per user to enable consistent comparison across participants. |
Other
| Measure | Time frame | Description |
|---|---|---|
| CHD Event | Through study completion, an average of 3 years | Reliability in predicting CHD related event in patient as measured by Framingham Risk Score. The Framingham Risk Score (FRS) is a validated means of predicting cardiovascular disease (CVD) risk. Input variables include age, cigarette smoking, total cholesterol, HDL cholesterol, systolic blood pressure measurement and treatment for hypertension. Point values are calculated based on each of these risks. A 10-year risk score can be derived as a percentage. Risk scores range from 0-20%. Low Risk: Less than 10% risk that you will develop a heart attack or die from coronary disease in the next 10 years. Intermediate risk: A 10 to 20% risk that you will develop a heart attack or die from coronary disease in the next 10 years. High Risk: A greater than 20% risk that you will develop a heart attack or die from coronary disease in the next 10 years. |
| Health Care Utilization | Through study completion, an average of 3 years | Prediction of cost for health care utilization between heart disease and non- heart disease subjects measured by insurance claims data |
Countries
United States
Participant flow
Participants by arm
| Arm | Count |
|---|---|
| Case Patients ages 30-74 with and without CHD (IICD 10: I63, I20-I25 ) within the last 5 years.
Survey: Interested participants may complete the informed consent online. After informed consent, the participant will be asked to share the digital data types that they use (Facebook, Instagram, Twitter, Google search, step data) and then participants will complete a cross-sectional survey. | 326 |
| Control Patients aged 30-74 who have non-cardiovascular-related history
Survey: Interested participants may complete the informed consent online. After informed consent, the participant will be asked to share the digital data types that they use (Facebook, Instagram, Twitter, Google search, step data) and then participants will complete a cross-sectional survey. | 455 |
| Total | 781 |
Baseline characteristics
| Characteristic | Control | Total | Case |
|---|---|---|---|
| Age, Continuous | 59.5 years STANDARD_DEVIATION 9.2 | 60.1 years STANDARD_DEVIATION 9.8 | 60.1 years STANDARD_DEVIATION 9.5 |
| Race (NIH/OMB) American Indian or Alaska Native | 1 Participants | 2 Participants | 1 Participants |
| Race (NIH/OMB) Asian | 13 Participants | 23 Participants | 10 Participants |
| Race (NIH/OMB) Black or African American | 152 Participants | 207 Participants | 55 Participants |
| Race (NIH/OMB) More than one race | 0 Participants | 0 Participants | 0 Participants |
| Race (NIH/OMB) Native Hawaiian or Other Pacific Islander | 0 Participants | 0 Participants | 0 Participants |
| Race (NIH/OMB) Unknown or Not Reported | 35 Participants | 78 Participants | 43 Participants |
| Race (NIH/OMB) White | 254 Participants | 471 Participants | 217 Participants |
| Sex: Female, Male Female | 287 Participants | 428 Participants | 141 Participants |
| Sex: Female, Male Male | 125 Participants | 272 Participants | 147 Participants |
Adverse events
| Event type | EG000 affected / at risk | EG001 affected / at risk |
|---|---|---|
| deaths Total, all-cause mortality | 0 / 0 | 0 / 0 |
| other Total, other adverse events | 0 / 0 | 0 / 0 |
| serious Total, serious adverse events | 0 / 0 | 0 / 0 |
Outcome results
Primary
Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease
The primary outcome is topics and features (derived using the LDA method for clustering language data). For each participant, we included all available Facebook wall posts from the start of their account history through data collection, regardless of whether they occurred before or after a CHD diagnosis. We examined associations between linguistic features (unigrams, LIWC categories, LDA topics) and cardiovascular case status (CHD presence vs absence) using Pearson correlation and logistic regression. Latent LDA, a systematic method to identify text-based themes, was applied to generate 200 clusters of co-occurring words (topics). For each feature type (unigram, LIWC category, LDA topic), we fit separate logistic regression models and calculated Pearson correlation coefficients to assess predictive value for case status. Each language-derived feature was encoded as a normalized frequency count per user to enable consistent comparison across participants.
Time frame: Through study completion, an average of 3 years
Population: To evaluate predictive performance for ASCVD risk, we used regression and classification analyses. Pearson correlations assessed how language features predicted continuous risk scores. For classification, we measured AUC for binary risk (≥10% vs \<10%) and standard risk categories. Three logistic regression models were tested: language alone, demographics alone, and combined. Arms/groups were combined because the goal was to predict ASCVD risk continuously and categorically across participants.
| Arm | Measure | Group | Value (NUMBER) |
|---|---|---|---|
| Binary Classification (≥10% vs. <10%) Binary Classification | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | Unigrams only | 0.61 proportion probability AUC (Area Under t |
| Binary Classification (≥10% vs. <10%) Binary Classification | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | LIWC + Demographics (age, sex, race) | 0.79 proportion probability AUC (Area Under t |
| Binary Classification (≥10% vs. <10%) Binary Classification | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | LIWC only | 0.55 proportion probability AUC (Area Under t |
| Binary Classification (≥10% vs. <10%) Binary Classification | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | Unigrams + Demographics (age, sex, race) | 0.61 proportion probability AUC (Area Under t |
| Binary Classification (≥10% vs. <10%) Binary Classification | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | LDA only | 0.64 proportion probability AUC (Area Under t |
| Binary Classification (≥10% vs. <10%) Binary Classification | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | LDA + Demographics (age, sex, race) | 0.81 proportion probability AUC (Area Under t |
| Binary Classification (≥10% vs. <10%) Binary Classification | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | Demographics only | 0.85 proportion probability AUC (Area Under t |
| All Risk Categories | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | LDA + Demographics (age, sex, race) | 0.82 proportion probability AUC (Area Under t |
| All Risk Categories | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | Demographics only | 0.87 proportion probability AUC (Area Under t |
| All Risk Categories | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | LIWC only | 0.67 proportion probability AUC (Area Under t |
| All Risk Categories | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | Unigrams only | 0.71 proportion probability AUC (Area Under t |
| All Risk Categories | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | LDA only | 0.67 proportion probability AUC (Area Under t |
| All Risk Categories | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | LIWC + Demographics (age, sex, race) | 0.80 proportion probability AUC (Area Under t |
| All Risk Categories | Latent Dirichlet Allocation (LDA) Topics - Topics / Themes Discussed Between Patients With and Without Heart Disease | Unigrams + Demographics (age, sex, race) | 0.74 proportion probability AUC (Area Under t |
Other Pre-specified
CHD Event
Reliability in predicting CHD related event in patient as measured by Framingham Risk Score. The Framingham Risk Score (FRS) is a validated means of predicting cardiovascular disease (CVD) risk. Input variables include age, cigarette smoking, total cholesterol, HDL cholesterol, systolic blood pressure measurement and treatment for hypertension. Point values are calculated based on each of these risks. A 10-year risk score can be derived as a percentage. Risk scores range from 0-20%. Low Risk: Less than 10% risk that you will develop a heart attack or die from coronary disease in the next 10 years. Intermediate risk: A 10 to 20% risk that you will develop a heart attack or die from coronary disease in the next 10 years. High Risk: A greater than 20% risk that you will develop a heart attack or die from coronary disease in the next 10 years.
Time frame: Through study completion, an average of 3 years
Other Pre-specified
Health Care Utilization
Prediction of cost for health care utilization between heart disease and non- heart disease subjects measured by insurance claims data
Time frame: Through study completion, an average of 3 years