Physiological data from wearable devices improves COVID-19 diagnosis prediction compared to self-reported symptoms alone
The COVID-19 pandemic has presented new challenges to the public health community, including controlling the transmission of SARS-CoV-2 before it can further spread to susceptible individuals. In a study funded by the NCATS, data acquired from wearable fitness devices combined with self-reported symptoms was found to identify cases of COVID-19 with greater success than measuring symptoms alone. In this study, data from smartwatches and activity trackers, such as Fitbit, was found to offer a potential value to predict onset of a viral illness due to its ability to track trends in an individual’s physiological measures, including resting heart rate, sleep, and activity. Although a previous study reported that self-reported symptom data captured via a smartphone-based app could predict COVID-19 diagnose, the ability of sensor data to improve the reliability of these predictions had not previously been tested.
This research study enrolled 30,529 individuals (62.0 percent female, 12.8 percent 65 or more years old) across the United States between March 25 and June 7, 2020. Users connected their Fitbit, AppleHealthKit, or Google Fit device to the study smartphone app called DETECT (Digital Engagement and Tracking for Early Control and Treatment). This mobile app compiled wearable data in addition to self-reported symptoms, physician diagnoses, and electronic health record data with the aim to improve the identification and tracking of individual viral illness, including COVID-19. Of all participants, 3,811 (12.5 percent) reported at least one symptom over the time period of the study, with 54 of these participants testing positive for COVID-19 and 279 testing negative. Certain self-reported symptoms (fever, fatigue, difficulty breathing, loss of smell, cough, and body aches) were significantly more frequent in individuals that tested positive for COVID-19. Researchers once again found that self-reported symptoms, along with age and sex, were able to predict whether the participant later tested positive for COVID-19 at about 71 percent accuracy. However, this was not the sole predictor as many individuals that were negative for COVID-19 also reported these symptoms.
Physiological data from the smartwatches and activity trackers was analyzed to determine if any measurements from these devices were predictive of COVID-19 diagnosis. Although a portion of COVID-19-positive participants had a significantly higher resting heart rate than average baseline, this factor was not sufficient to discriminate between COVID-19 positive and COVID-19 negative participants. Sleep and activity, however, showed a significant difference between testing groups, supporting that sleep and activity data from wearable devices could each predict COVID-19 diagnosis, at about a 69 percent accuracy. Combination of all three physiological measurements (heart rate, sleep, and activity) in a single metric resulted in the best predictor of a COVID-19 diagnosis (73 percent accuracy), which showed a similar reliability as self-reported symptoms alone. Therefore, when participant-reported symptoms and sensor metrics from wearable devices was jointly considered in the analysis, the prediction of accurate COVID-19 diagnosis (80 percent accuracy) was significantly improved, as compared to either type of measurement alone.
As smartwatches and activity trackers are worn by roughly 100 million Americans, mobile apps like DETECT may offer a method to integrate both self-reported data and physiological data from wearable devices to predict a COVID-19 diagnosis. However, as our knowledge about the frequency and prevalence of asymptomatic cases evolves, the reported algorithms in this study may not prove to be as reliable of a predictor for asymptomatic positive cases. Since testing was not as widespread during the timespan of enrollment, many participants may have been infected with COVID-19 without presenting symptoms. Future studies should test whether physiological data from wearable devices can reliably predict COVID-19 infection in individual who don’t experience symptoms.
The global interest in sensor and clinical data to address the COVID-19 health crisis is continuing to grow. Use of DETECT and other similar digital tools represent the transition of health research to remote, direct-to-participant approaches that are now possible with digital technologies. The promise of digital technologies is that they will offer new methods by which to identify and track COVID-19 infection in addition to other infectious pathogens in the future.
Quer G, Radin JM, Gadaleta M, Baca-Motes K, Ariniello L, Ramos E, Kheterpal V, Topol EJ, Steinhul SR. 2020. Wearable sensor data and self-reported symptoms for COVID-19 detection. Nature Medicine. doi: 10.1038/s41591-020-1123-x
Prenatal choline levels in children of Black American women may indicate potential predisposition for future mental illness risk
Rates of schizophrenia and other mental health disorders are higher in the Black American population when compared to the white population. Black American women have been shown to have lower concentrations of choline which is important for healthy infant brain development. Contributing factors for the lower levels of choline include postpartum depression and its impact on cortisol and inflammation, along with interactions with systemic racism, neighborhood environment, partner relationships, and financial insecurity. A study supported by the NICHD, NCATS, NIDDK, the Institute for Children’s Mental Disorders, and the Anschutz Foundation, investigated prenatal choline, prenatal stressors, and their interactions in Black American women. The researchers hypothesized that lower choline levels would negatively affect the outcomes of gestation, beginning with gestational age at birth and extending to early brain development.
In order to investigate prenatal influences on fetal brain development and early childhood behavior, the researcher assessed the differences in 16-week choline plasma concentrations between Black American, American Indian, and white women. Prenatal influences of interest included maternal infection, cannabis use, inflammation, and their interaction with maternal choline plasma concentrations. Infants were tested 1 month post birth (adjusted for gestational age) using a behavioral paradigm, cerebral inhibition response, such that a higher cerebral response to a repeated sound indicates poorer development of inhibition. A deficit in cerebral inhibition has been found to be associated with schizophrenia, bipolar disorder, attention deficit disorder, and autism spectrum disorder. Additionally, choline has been found to contribute to the development of the cerebral inhibitory response.
Pregnant women were recruited from prenatal clinic admissions at 14 to 16 weeks gestation from July 2013 to July 2016 (n = 183). By the 16th week of gestation, 162 of these women completed the initial assessment which included: plasma choline level, demographic information, and mental condition ratings (self-reported race/ethnicity: 23 = Black, 13 = American Indian, 129 = white). Additionally, a total of 166 women in Uganda were also enrolled from the Gulu regional referral hospital in rural northern Uganda at 18 to 25 weeks gestation.
In this study, Black American women had higher self-rating of depression symptoms and adverse childhood experiences, were less likely to have Hispanic heritage, and were less likely to be living with the biological father of their child. Choline levels in Black American women were lower than those of white women from the same neighborhoods (5.48 μM versus 6.58 μM, respectively, P = 0.008). While Black American participants did not indicate more stress through questionnaires, their cortisol levels, an indication of stress, were markedly higher and were associated with higher levels of depressive syndromes. Maternal choline concentration analyses revealed that for Black Americans, maternal cortisol was a significant negative factor (β = −0.477, P = 0.045) and father’s presence was negatively correlated, but did not reach statistical significance, in contrast to white mothers where the father’s presence was a significant positive factor (β = 0.203, P = 0.048), and cortisol was positive but not statistically significant. Choline levels in pregnant Black women in Uganda were significantly higher than Black American levels. These differences point to the need for further study to understand the cross-cultural effects of race, stressors and other environmental factors.
Lower maternal choline was also associated with offspring’s’ lower gestational age at birth and decreased auditory inhibition as measured by cerebral response. A previously published randomized clinical trial of Black American women indicated that the behavioral deficits associated with low prenatal choline can be improved with gestational phosphatidylcholine supplementation, further supporting this study’s finding that choline levels can impact later mental illness in children of Black American women.
This study found that Black American women had the lowest 16-week gestation plasma choline of any self-identified racial/ethnic group in this study. Maternal cortisol may regulate plasma concentrations as maternal stress was the most significant factor associated with Black American choline concentration. Further understanding of the complex interactions of exposure to high levels of systemic racism, discrimination stressors, and other environmental factors in Black American women is warranted.
Hunter SK, Hoffman MC, McCarthy L, D’Alessandro A, Wyrwa A, Noonan K, Christians U, Nakimuli-Mpungu E, Zeisel SH, Law AJ, Freedman R. 2020. Black American Maternal Prenatal Choline, Offspring Gestational Age at Birth, and Developmental Predisposition to Mental Illness. Schizophrenia Bulletin, sbaa171, https://doi.org/10.1093/schbul/sbaa171
Electronic cigarettes may predispose adolescents to smoking cigarettes, even if they have no prior intention to smoke
Do electronic-cigarettes (e-cigarettes) aid in smoking cessation and/or act as a replacement/substitute to cigarettes or do e-cigarettes initiate a pathway toward smoking cigarettes—even if that was not the original intent of the e-cigarette user? Recent research supported by the NICHD, NCI, and NIDA investigated the causal pathway of this relationship by examining smoking intention as a dependent factor for smoking e-cigarettes and cigarette smoking. Despite declines in cigarette smoking, it is still one of the leading preventable cause of morbidity and mortality in the United States. Additionally, the increase in e-cigarette use may lead to a new risk for nicotine use disorder and cigarette smoking among adolescents. Additionally, the recent emergence of newer and potentially highly addictive e-cigarette products may further these risks.
In the current study, the researchers tested the smoking intention pathway—which posits that e-cigarette use may lead to the development of smoking intention especially if there is no firm commitment to not smoke cigarettes. To determine whether smoking intention is a necessary antecedent of cigarette smoking among adolescent e-cigarette users, they used data from the Population Assessment of Tobacco and Health (PATH) Study (Wave 2 [2014–2015], Wave 3 [2015–2016]), which is a prospective longitudinal cohort study of a nationally representative sample of people in the United States investigating tobacco use. The researchers analyzed data of 8,661 U.S. adolescents, ages 12 to 17, who had never smoked.
Descriptively, in data collected during Wave 2, among adolescents that had never smoked cigarettes, 12.8 percent had intention to smoke and 8.5 percent had used an e-cigarette. However, one year later, of the adolescents that had not previously smoked cigarettes when asked during Wave 2, only 3.2 percent had smoked a cigarette when asked again during Wave 3. In assessing whether intention and e-cigarette experience is important, the data show that both smoking intention and ever using e-cigarettes at Wave 2 were positively associated with cigarette smoking at Wave 3 (adjusted odds ratio [aOR] = 3.03, P < 0.001; aOR = 4.62, P < 0.001, respectively). The interaction between smoking intention and ever using e-cigarettes was also significant (aOR = 0.34, P < 0.01).
Researchers found that of the adolescents who had expressed intention to smoke cigarettes at Wave 2, the odds of cigarette smoking at Wave 3 did not significantly differ for e-cigarette users and never e-cigarette users (aOR = 1.57, P = 0.08). However, among adolescents who had no intention to smoke at Wave 2, e-cigarette users compared with never e-cigarette users, had four times the odds of cigarette smoking (aOR = 4.62; 95 percent CI 2.87–7.42; P < 0.0001).
In conclusion, these results partly support the smoking intention pathway from e-cigarette use to conventional cigarette smoking. The use of e-cigarette was associated with higher odds of cigarette smoking in adolescents who did not previously express smoking intention. However, e-cigarette use was not associated with cigarette smoking among adolescents who had the initial intention to smoke. These findings suggest that e-cigarette use may create the intention to smoke which then leads to cigarette smoking. Thus, abstaining from e-cigarette use may also be considered a preventative strategy for cigarette smoking in adolescents.
Owotomo O, Stritzel H, McCabe SE, Boyd CJ, Maslowsky J. 2020. Smoking Intention and Progression From E-Cigarette Use to Cigarette Smoking. Pediatrics.146(6):e2020002881. doi: 10.1542/peds.2020-002881