Linguistic analysis of empathy in medical school admission essays.

OBJECTIVES: This study aimed to determine whether words used in medical school admissions essays can predict physician empathy. METHODS: A computational form of linguistic analysis was used for the content analysis of medical school admissions essays. Words in medical school admissions essays were computationally grouped into 20 'topics' which were then correlated with scores on the Jefferson Scale of Empathy. The study sample included 1,805 matriculants (between 2008-2015) at a single medical college in the North East of the United States who wrote an admissions essay and completed the Jefferson Scale of Empathy at matriculation. RESULTS: After correcting for multiple comparisons and controlling for gender, the Jefferson Scale of Empathy scores significantly correlated with a linguistic topic (r = .074, p < .05). This topic was comprised of specific words used in essays such as "understanding," "compassion," "empathy," "feeling," and "trust." These words are related to themes emphasized in both theoretical writing and empirical studies on physician empathy. CONCLUSIONS: This study demonstrates that physician empathy can be predicted from medical school admission essays. The implications of this methodological capability, i.e. to quantitatively associate linguistic features or words with psychometric outcomes, bears on the future of medical education research and admissions. In particular, these findings suggest that those responsible for medical school admissions could identify more empathetic applicants based on the language of their application essays.

Evaluating the predictability of medical conditions from social media posts.

We studied whether medical conditions across 21 broad categories were predictable from social media content across approximately 20 million words written by 999 consenting patients. Facebook language significantly improved upon the prediction accuracy of demographic variables for 18 of the 21 disease categories; it was particularly effective at predicting diabetes and mental health conditions including anxiety, depression and psychoses. Social media data are a quantifiable link into the otherwise elusive daily lives of patients, providing an avenue for study and assessment of behavioral and environmental disease risk factors. Analogous to the genome, social media data linked to medical diagnoses can be banked with patients' consent, and an encoding of social media language can be used as markers of disease risk, serve as a screening tool, and elucidate disease epidemiology. In what we believe to be the first report linking electronic medical record data with social media data from consenting patients, we identified that patients' Facebook status updates can predict many health conditions, suggesting opportunities to use social media data to determine disease onset or exacerbation and to conduct social media-based health interventions.

Facebook language predicts depression in medical records.

Depression, the most prevalent mental illness, is underdiagnosed and undertreated, highlighting the need to extend the scope of current screening methods. Here, we use language from Facebook posts of consenting individuals to predict depression recorded in electronic medical records. We accessed the history of Facebook statuses posted by 683 patients visiting a large urban academic emergency department, 114 of whom had a diagnosis of depression in their medical records. Using only the language preceding their first documentation of a diagnosis of depression, we could identify depressed patients with fair accuracy [area under the curve (AUC) = 0.69], approximately matching the accuracy of screening surveys benchmarked against medical records. Restricting Facebook data to only the 6 months immediately preceding the first documented diagnosis of depression yielded a higher prediction accuracy (AUC = 0.72) for those users who had sufficient Facebook data. Significant prediction of future depression status was possible as far as 3 months before its first documentation. We found that language predictors of depression include emotional (sadness), interpersonal (loneliness, hostility), and cognitive (preoccupation with the self, rumination) processes. Unobtrusive depression assessment through social media of consenting individuals may become feasible as a scalable complement to existing screening and monitoring procedures.

Natural Language Processing of Social Media as Screening for Suicide Risk.

Suicide is among the 10 most common causes of death, as assessed by the World Health Organization. For every death by suicide, an estimated 138 people's lives are meaningfully affected, and almost any other statistic around suicide deaths is equally alarming. The pervasiveness of social media-and the near-ubiquity of mobile devices used to access social media networks-offers new types of data for understanding the behavior of those who (attempt to) take their own lives and suggests new possibilities for preventive intervention. We demonstrate the feasibility of using social media data to detect those at risk for suicide. Specifically, we use natural language processing and machine learning (specifically deep learning) techniques to detect quantifiable signals around suicide attempts, and describe designs for an automated system for estimating suicide risk, usable by those without specialized mental health training (eg, a primary care doctor). We also discuss the ethical use of such technology and examine privacy implications. Currently, this technology is only used for intervention for individuals who have "opted in" for the analysis and intervention, but the technology enables scalable screening for suicide risk, potentially identifying many people who are at risk preventively and prior to any engagement with a health care system. This raises a significant cultural question about the trade-off between privacy and prevention-we have potentially life-saving technology that is currently reaching only a fraction of the possible people at risk because of respect for their privacy. Is the current trade-off between privacy and prevention the right one?

Correction: Seeing the "Big" Picture: Big Data Methods for Exploring Relationships Between Usage, Language, and Outcome in Internet Intervention Data.

[This corrects the article DOI: 10.2196/jmir.5725.].

Variations in Facebook Posting Patterns Across Validated Patient Health Conditions: A Prospective Cohort Study.

BACKGROUND: Social media is emerging as an insightful platform for studying health. To develop targeted health interventions involving social media, we sought to identify the patient demographic and disease predictors of frequency of posting on Facebook. OBJECTIVE: The aims were to explore the language topics correlated with frequency of social media use across a cohort of social media users within a health care setting, evaluate the differences in the quantity of social media postings across individuals with different disease diagnoses, and determine if patients could accurately predict their own levels of social media engagement. METHODS: Patients seeking care at a single, academic, urban, tertiary care emergency department from March to October 2014 were queried on their willingness to share data from their Facebook accounts and electronic medical records (EMRs). For each participant, the total content of Facebook posts was extracted. Using the latent Dirichlet allocation natural language processing technique, Facebook language topics were correlated with frequency of Facebook use. The mean number of Facebook posts over 6 months prior to enrollment was then compared across validated health outcomes in the sample. RESULTS: A total of 695 patients consented to provide access to their EMR and social media data. Significantly correlated language topics among participants with the highest quartile of posts contained health terms, such as "cough," "headaches," and "insomnia." When adjusted for demographics, individuals with a history of depression had significantly higher posts (mean 38, 95% CI 28-50) than individuals without a history of depression (mean 22, 95% CI 19-26, P=.001). Except for depression, across prevalent health outcomes in the sample (hypertension, diabetes, asthma), there were no significant posting differences between individuals with or without each condition. CONCLUSIONS: High-frequency posters in our sample were more likely to post about health and to have a diagnosis of depression. The direction of causality between depression and social media use requires further evaluation. Our findings suggest that patients with depression may be appropriate targets for health-related interventions on social media.

Seeing the "Big" Picture: Big Data Methods for Exploring Relationships Between Usage, Language, and Outcome in Internet Intervention Data.

BACKGROUND: Assessing the efficacy of Internet interventions that are already in the market introduces both challenges and opportunities. While vast, often unprecedented amounts of data may be available (hundreds of thousands, and sometimes millions of participants with high dimensions of assessed variables), the data are observational in nature, are partly unstructured (eg, free text, images, sensor data), do not include a natural control group to be used for comparison, and typically exhibit high attrition rates. New approaches are therefore needed to use these existing data and derive new insights that can augment traditional smaller-group randomized controlled trials. OBJECTIVE: Our objective was to demonstrate how emerging big data approaches can help explore questions about the effectiveness and process of an Internet well-being intervention. METHODS: We drew data from the user base of a well-being website and app called Happify. To explore effectiveness, multilevel models focusing on within-person variation explored whether greater usage predicted higher well-being in a sample of 152,747 users. In addition, to explore the underlying processes that accompany improvement, we analyzed language for 10,818 users who had a sufficient volume of free-text response and timespan of platform usage. A topic model constructed from this free text provided language-based correlates of individual user improvement in outcome measures, providing insights into the beneficial underlying processes experienced by users. RESULTS: On a measure of positive emotion, the average user improved 1.38 points per week (SE 0.01, t122,455=113.60, P<.001, 95% CI 1.36-1.41), about an 11% increase over 8 weeks. Within a given individual user, more usage predicted more positive emotion and less usage predicted less positive emotion (estimate 0.09, SE 0.01, t6047=9.15, P=.001, 95% CI .07-.12). This estimate predicted that a given user would report positive emotion 1.26 points (or 1.26%) higher after a 2-week period when they used Happify daily than during a week when they didn't use it at all. Among highly engaged users, 200 automatically clustered topics showed a significant (corrected P<.001) effect on change in well-being over time, illustrating which topics may be more beneficial than others when engaging with the interventions. In particular, topics that are related to addressing negative thoughts and feelings were correlated with improvement over time. CONCLUSIONS: Using observational analyses on naturalistic big data, we can explore the relationship between usage and well-being among people using an Internet well-being intervention and provide new insights into the underlying mechanisms that accompany it. By leveraging big data to power these new types of analyses, we can explore the workings of an intervention from new angles, and harness the insights that surface to feed back into the intervention and improve it further in the future.

Linking social media and medical record data: a study of adults presenting to an academic, urban emergency department.

BACKGROUND: Social media may offer insight into the relationship between an individual's health and their everyday life, as well as attitudes towards health and the perceived quality of healthcare services. OBJECTIVE: To determine the acceptability to patients and potential utility to researchers of a database linking patients' social media content with their electronic medical record (EMR) data. METHODS: Adult Facebook/Twitter users who presented to an emergency department were queried about their willingness to share their social media data and EMR data with health researchers for the purpose of building a databank for research purposes. Shared posts were searched for select terms about health and healthcare. RESULTS: Of the 5256 patients approached, 2717 (52%) were Facebook and/or Twitter users. 1432 (53%) of those patients agreed to participate in the study. Of these participants, 1008 (71%) consented to share their social media data for the purposes of comparing it with their EMR. Social media data consisted of 1 395 720 posts/tweets to Facebook and Twitter. Participants sharing social media data were slightly younger (29.1±9.8 vs 31.9±10.4 years old; p<0.001), more likely to post at least once a day (42% vs 29%; p=0.003) and more likely to present to the emergency room via self-arrival mode and have private insurance. Of Facebook posts, 7.5% (95% CI 4.8% to 10.2%) were related to health. Individuals with a given diagnosis in their EMR were significantly more likely to use terms related to that diagnosis on Facebook than patients without that diagnosis in their EMR (p<0.0008). CONCLUSIONS: Many patients are willing to share and link their social media data with EMR data. Sharing patients have several demographic and clinical differences compared with non-sharers. A database that merges social media with EMR data has the potential to provide insights about individuals' health and health outcomes.

Individual differences in memory search and their relation to intelligence.

Attempts to understand why memory predicts intelligence have not fully leveraged state-of-the-art measures of recall dynamics. Using data from a multisession free recall study, we examine individual differences in measures of recall initiation and postinitiation transitions. We identify 4 sources of variation: a recency factor reflecting variation in the tendency to initiate recall from an item near the end of the list, a primacy factor reflecting a tendency to initiate from the beginning of the list, a temporal factor corresponding to transitions mediated by temporal associations, and a semantic factor corresponding to semantically mediated transitions. Together, these 4 factors account for 83% of the variability in overall recall accuracy, suggesting they provide a nearly complete picture of recall dynamics. We also show that these sources of variability account for over 80% of the variance shared between memory and intelligence. The temporal association factor was the most influential in predicting both recall accuracy and intelligence. We outline a theory of how controlled drift of temporal context may be critical across a range of cognitive activities.

Depth of the olfactory sulcus: a marker of early embryonic disruption in schizophrenia?

Embryonic insults during early gestation increase the risk of schizophrenia. Abnormal forebrain development during this period is often characterized by a shallow olfactory sulcus. The adjacent orbital sulcus does not develop until the third trimester and so is immune to early intrauterine insults. We measured olfactory and orbital sulcal depths in 36 patients and 28 control subjects. Patients had shallower olfactory sulci, but normal orbital sulci. Olfactory and orbital sulcal depths were correlated in controls, but not in patients. Olfactory sulcal depth may therefore be a biomarker denoting an early embryonic disruption in individuals at risk for schizophrenia.