De-identification and Obfuscation of Gender Attributes from Retinal Scans Chenwei Wu, Xiyu Yang, Emil Ghitman Gilkes, Hanwen Cui, Jiheon Choi, Na Sun, Ziqian Liao, Bo Fan, Mauricio Santillana, Leo Celi, Paolo Silva & Luis Nakayama. LNCS 14242. Clinical Image-Based Procedures, Fairness of AI in Medical Imaging, and Ethical and Philosophical Issues in Medical Imaging. October 2023
Abstract
Retina images are considered to be important biomarkers and have been used as clinical diagnostic tools to detect multiple diseases. We examine multiple techniques for de-identifying retina images while maintaining their clinical ability for detecting diabetic retinopathy (DR), using gender as a proxy for identifiability. We apply two differential privacy algorithms, Snow and VS-Snow, on the entire image (globally) and on blood vessels only (locally) to obfuscate important image features that can predict a patient’s sex. We evaluate the level of privacy and retained clinical predictive power of these de-identified images by using attacking gender classifier models and downstream disease classifiers. We show empirically that our proposed VS-Snow framework achieves strong privacy while preserving a meaningful clinical predictive power across different patient populations.
Misinformation, Trust, and Use of Ivermectin and Hydroxychloroquine for COVID-19 Roy H. Perlis, MD, MSc; Kristin Lunz Trujillo, PhD; Jon Green, PhD; Alauna Safarpour, PhD; James N. Druckman, PhD; Mauricio Santillana, PhD; Katherine Ognyanova, PhD; David Lazer, PhD JAMA Health Forum. 2023;4(9):e233257 doi:10.1001/jamahealthforum.2023.3257
Abstract
ImportanceThe COVID-19 pandemic has been notable for the widespread dissemination of misinformation regarding the virus and appropriate treatment. ObjectiveTo quantify the prevalence of non–evidence-based treatment for COVID-19 in the US and the association between such treatment and endorsement of misinformation as well as lack of trust in physicians and scientists. Design, Setting, and ParticipantsThis single-wave, population-based, nonprobability internet survey study was conducted between December 22, 2022, and January 16, 2023, in US residents 18 years or older who reported prior COVID-19 infection. Main Outcome and MeasureSelf-reported use of ivermectin or hydroxychloroquine, endorsing false statements related to COVID-19 vaccination, self-reported trust in various institutions, conspiratorial thinking measured by the American Conspiracy Thinking Scale, and news sources. ResultsA total of 13 438 individuals (mean [SD] age, 42.7 [16.1] years; 9150 [68.1%] female and 4288 [31.9%] male) who reported prior COVID-19 infection were included in this study. In this cohort, 799 (5.9%) reported prior use of hydroxychloroquine (527 [3.9%]) or ivermectin (440 [3.3%]). In regression models including sociodemographic features as well as political affiliation, those who endorsed at least 1 item of COVID-19 vaccine misinformation were more likely to receive non–evidence-based medication (adjusted odds ratio [OR], 2.86; 95% CI, 2.28-3.58). Those reporting trust in physicians and hospitals (adjusted OR, 0.74; 95% CI, 0.56-0.98) and in scientists (adjusted OR, 0.63; 95% CI, 0.51-0.79) were less likely to receive non–evidence-based medication. Respondents reporting trust in social media (adjusted OR, 2.39; 95% CI, 2.00-2.87) and in Donald Trump (adjusted OR, 2.97; 95% CI, 2.34-3.78) were more likely to have taken non–evidence-based medication. Individuals with greater scores on the American Conspiracy Thinking Scale were more likely to have received non–evidence-based medications (unadjusted OR, 1.09; 95% CI, 1.06-1.11; adjusted OR, 1.10; 95% CI, 1.07-1.13). Conclusions and RelevanceIn this survey study of US adults, endorsement of misinformation about the COVID-19 pandemic, lack of trust in physicians or scientists, conspiracy-mindedness, and the nature of news sources were associated with receiving non–evidence-based treatment for COVID-19. These results suggest that the potential harms of misinformation may extend to the use of ineffective and potentially toxic treatments in addition to avoidance of health-promoting behaviors.
Community Mobility and Depressive Symptoms During the COVID-19 Pandemic in the United States Roy H. Perlis, MD, MSc; Kristin Lunz Trujillo, PhD; Alauna Safarpour, PhD; Alexi Quintana, BA; Matthew D. Simonson, PhD; Jasper Perlis; Mauricio Santillana, PhD; Katherine Ognyanova, PhD; Matthew A. Baum, PhD; James N. Druckman, PhD; David Lazer, PhD JAMA Network Open. 2023;6(9):e2334945. doi:10.1001/jamanetworkopen.2023.34945
Abstract
ImportanceMarked elevation in levels of depressive symptoms compared with historical norms have been described during the COVID-19 pandemic, and understanding the extent to which these are associated with diminished in-person social interaction could inform public health planning for future pandemics or other disasters. ObjectiveTo describe the association between living in a US county with diminished mobility during the COVID-19 pandemic and self-reported depressive symptoms, while accounting for potential local and state-level confounding factors. Design, Setting, and ParticipantsThis survey study used 18 waves of a nonprobability internet survey conducted in the United States between May 2020 and April 2022. Participants included respondents who were 18 years and older and lived in 1 of the 50 US states or Washington DC. Main Outcome and MeasureDepressive symptoms measured by the Patient Health Questionnaire-9 (PHQ-9); county-level community mobility estimates from mobile apps; COVID-19 policies at the US state level from the Oxford stringency index. ResultsThe 192 271 survey respondents had a mean (SD) of age 43.1 (16.5) years, and 768 (0.4%) were American Indian or Alaska Native individuals, 11 448 (6.0%) were Asian individuals, 20 277 (10.5%) were Black individuals, 15 036 (7.8%) were Hispanic individuals, 1975 (1.0%) were Pacific Islander individuals, 138 702 (72.1%) were White individuals, and 4065 (2.1%) were individuals of another race. Additionally, 126 381 respondents (65.7%) identified as female and 65 890 (34.3%) as male. Mean (SD) depression severity by PHQ-9 was 7.2 (6.8). In a mixed-effects linear regression model, the mean county-level proportion of individuals not leaving home was associated with a greater level of depression symptoms (β, 2.58; 95% CI, 1.57-3.58) after adjustment for individual sociodemographic features. Results were similar after the inclusion in regression models of local COVID-19 activity, weather, and county-level economic features, and persisted after widespread availability of COVID-19 vaccination. They were attenuated by the inclusion of state-level pandemic restrictions. Two restrictions, mandatory mask-wearing in public (β, 0.23; 95% CI, 0.15-0.30) and policies cancelling public events (β, 0.37; 95% CI, 0.22-0.51), demonstrated modest independent associations with depressive symptom severity. Conclusions and RelevanceIn this study, depressive symptoms were greater in locales and times with diminished community mobility. Strategies to understand the potential public health consequences of pandemic responses are needed.
Introduction: Analgesia and sedation are integral to the care of critically ill children. However, the choice and dose of the analgesic or sedative drug is often empiric, and models predicting favorable responses are lacking. We aimed to compute models to predict a patient's response to intravenous morphine. Methods: We retrospectively analyzed data from consecutive patients admitted to the Cardiac Intensive Care Unit (January 2011–January 2020) who received at least one intravenous bolus of morphine. The primary outcome was a decrease in the State Behavioral Scale (SBS) ≥1 point; the secondary outcome was a decrease in the heart rate Z-score (zHR) at 30 min. Effective doses were modeled using logistic regression, Lasso regression, and random forest modeling. Results: A total of 117,495 administrations of intravenous morphine among 8140 patients (median age 0.6 years [interquartile range [IQR] 0.19, 3.3]) were included. The median morphine dose was 0.051 mg/kg (IQR 0.048, 0.099) and the median 30- day cumulative dose was 2.2 mg/kg (IQR 0.4, 15.3). SBS decreased following 30% of doses, did not change following 45%, and increased following 25%. The zHR signifi- cantly decreased after morphine administration (median delta-zHR −0.34 [IQR−1.03, 0.00], p< 0.001). The following factors were associated with favorable response to morphine: A concomitant infusion of propofol, higher prior 30-day cumulative dose, being invasively ventilated and/or on vasopressors. Higher morphine dose, higher zHR pre-morphine, an additional analgosedation bolus ±30 min around the index bolus, a concomitant ketamine or dexmedetomidine infusion, and showing signs of withdrawal syndrome were associated with unfavorable response. Logistic regression (area under the receiver operating characteristic [ROC] curve [AUC] 0.900) and machine learning models (AUC 0.906) performed comparably, with a sensitivity of 95%, specificity of 71%, and negative predictive value of 97%. Conclusions: Statistical models identify 95% of effective intravenous morphine doses in pediatric critically ill cardiac patients, while incorrectly suggesting an effective dose in 29% of cases. This work represents an important step toward computer-aided, per- sonalized clinical decision support tool for sedation and analgesia in ICU patients.
Coronavirus disease 2019 (COVID-19) continues to affect the world, and the design of strategies to curb disease outbreaks requires close monitoring of their trajectories. We present machine learning methods that leverage internet-based digital traces to anticipate sharp increases in COVID-19 activity in U.S. counties. In a complemen- tary direction to the efforts led by the Centers for Disease Control and Prevention (CDC), our models are de- signed to detect the time when an uptrend in COVID-19 activity will occur. Motivated by the need for finer spatial resolution epidemiological insights, we build upon previous efforts conceived at the state level. Our methods—tested in an out-of-sample manner, as events were unfolding, in 97 counties representative of mul- tiple population sizes across the United States—frequently anticipated increases in COVID-19 activity 1 to 6 weeks before local outbreaks, defined when the effective reproduction number Rt becomes larger than 1 for a period of 2 weeks.
BackgroundWe aimed to characterize the prevalence of social disconnection and thoughts of suicide among older adults in the United States, and examine the association between them in a large naturalistic study. MethodsWe analyzed data from 6 waves of a fifty-state non-probability survey among US adults conducted between February and December 2021. The internet-based survey collected the PHQ-9, as well as multiple measures of social connectedness. We applied multiple logistic regression to analyze the association between presence of thoughts of suicide and social disconnection. Exploratory analysis, using generalized random forests, examined heterogeneity of effects across sociodemographic groups. ResultsOf 16,164 survey respondents age 65 and older, mean age was 70.9 (SD 5.0); the cohort was 61.4 % female and 29.6 % male; 2.0 % Asian, 6.7 % Black, 2.2 % Hispanic, and 86.8 % White. A total of 1144 (7.1 %) reported thoughts of suicide at least several days in the prior 2 week period. In models adjusted for sociodemographic features, households with 3 or more additional members (adjusted OR 1.73, 95 % CI 1.28–2.33) and lack of social supports, particularly emotional supports (adjusted OR 2.60, 95 % CI 2.09–3.23), were independently associated with greater likelihood of reporting such thoughts, as was greater reported loneliness (adjusted OR 1.75, 95 % CI 1.64–1.87). The effects of emotional support varied significantly across sociodemographic groups. ConclusionsThoughts of suicide are common among older adults in the US, and associated with lack of social support, but not with living alone.
Using general messages to persuade on a politicized scientific issue Jon Green, James N Druckman, Matthew A Baum, David Lazer, Katherine Ognyanova, Matthew D Simonson, Jennifer Lin, Mauricio Santillana, Roy H Perlis British Journal of Political Science. 2023; 53(2)698-706.
Abstract
Politics and science have become increasingly intertwined. Salient scientific issues, such as climate change, evolution, and stem-cell research, become politicized, pitting partisans against one another. This creates a challenge of how to effectively communicate on such issues. Recent work emphasizes the need for tailored messages to specific groups. Here, we focus on whether generalized messages also can matter. We do so in the context of a highly polarized issue: extreme COVID-19 vaccine resistance. The results show that science-based, moral frame, and social norm messages move behavioral intentions, and do so by the same amount across the population (that is, homogeneous effects). Counter to common portrayals, the politicization of science does not preclude using broad messages that resonate with the entire population.
Atypical dengue prevalence was observed in 2020 in many dengue-endemic countries, including Brazil. Evidence suggests that the pandemic disrupted not only dengue dynamics due to changes in mobility patterns, but also several aspects of dengue surveillance, such as care seeking behavior, care availability, and monitoring systems. However, we lack a clear understanding of the overall impact on dengue in different parts of the country as well as the role of individual causal drivers. In this study, we estimated the gap between expected and observed dengue cases in 2020 using an interrupted time series design with forecasts from a neural network and a structural Bayesian time series model. We also decomposed the gap into the impacts of climate conditions, pandemic-induced changes in reporting, human susceptibility, and human mobility. We find that there is considerable variation across the country in both overall pandemic impact on dengue and the relative importance of individual drivers. Increased understanding of the causal mechanisms driving the 2020 dengue season helps mitigate some of the data gaps caused by the COVID-19 pandemic and is critical to developing effective public health interventions to control dengue in the future.
ImportanceLittle is known about the functional correlates of post–COVID-19 condition (PCC), also known as long COVID, particularly the relevance of neurocognitive symptoms. ObjectiveTo characterize prevalence of unemployment among individuals who did, or did not, develop PCC after acute infection. Design, Setting, and ParticipantsThis survey study used data from 8 waves of a 50-state US nonprobability internet population-based survey of respondents aged 18 to 69 years conducted between February 2021 and July 2022. Main Outcomes and MeasuresThe primary outcomes were self-reported current employment status and the presence of PCC, defined as report of continued symptoms at least 2 months beyond initial month of symptoms confirmed by a positive COVID-19 test. ResultsThe cohort included 15 308 survey respondents with test-confirmed COVID-19 at least 2 months prior, of whom 2236 (14.6%) reported PCC symptoms, including 1027 of 2236 (45.9%) reporting either brain fog or impaired memory. The mean (SD) age was 38.8 (13.5) years; 9679 respondents (63.2%) identified as women and 10 720 (70.0%) were White. Overall, 1418 of 15 308 respondents (9.3%) reported being unemployed, including 276 of 2236 (12.3%) of those with PCC and 1142 of 13 071 (8.7%) of those without PCC; 8229 respondents (53.8%) worked full-time, including 1017 (45.5%) of those with PCC and 7212 (55.2%) without PCC. In survey-weighted regression models excluding retired respondents, the presence of PCC was associated with a lower likelihood of working full-time (odds ratio [OR], 0.71 [95% CI, 0.63-0.80]; adjusted OR, 0.84 [95% CI, 0.74-0.96]) and with a higher likelihood of being unemployed (OR, 1.45 [95% CI, 1.22-1.73]; adjusted OR, 1.23 [95% CI, 1.02-1.48]). The presence of any cognitive symptom was associated with lower likelihood of working full time (OR, 0.70 [95% CI, 0.56-0.88]; adjusted OR, 0.75 [95% CI, 0.59-0.84]). Conclusions and RelevancePCC was associated with a greater likelihood of unemployment and lesser likelihood of working full time in adjusted models. The presence of cognitive symptoms was associated with diminished likelihood of working full time. These results underscore the importance of developing strategies to treat and manage PCC symptoms.
Background: Disease surveillance systems capable of producing accurate real-time and short-term forecasts can help public health officials design timely public health interventions to mitigate the effects of disease outbreaks in affected populations. In France, existing clinic-based disease surveillance systems produce gastroenteritis activity information that lags real time by 1 to 3 weeks. This temporal data gap prevents public health officials from having a timely epidemiological characterization of this disease at any point in time and thus leads to the design of interventions that do not take into consideration the most recent changes in dynamics. Objective: The goal of this study was to evaluate the feasibility of using internet search query trends and electronic health records to predict acute gastroenteritis (AG) incidence rates in near real time, at the national and regional scales, and for long-term forecasts (up to 10 weeks). Methods: We present 2 different approaches (linear and nonlinear) that produce real-time estimates, short-term forecasts, and long-term forecasts of AG activity at 2 different spatial scales in France (national and regional). Both approaches leverage disparate data sources that include disease-related internet search activity, electronic health record data, and historical disease activity. Results: Our results suggest that all data sources contribute to improving gastroenteritis surveillance for long-term forecasts with the prominent predictive power of historical data owing to the strong seasonal dynamics of this disease. Conclusions: The methods we developed could help reduce the impact of the AG peak by making it possible to anticipate increased activity by up to 10 weeks.
Importance: Post-COVID-19 condition (PCC), or long COVID, has become prevalent. The course of this syndrome, and likelihood of remission, has not been characterized. Objective: To quantify the rates of remission of PCC, and the sociodemographic features associated with remission. Design: 16 waves of a 50-state U.S. non-probability internet survey conducted between August 2020 and November 2022 Setting: Population-based Participants: Survey respondents age 18 and older Main Outcome and Measure: PCC remission, defined as reporting full recovery from COVID-19 symptoms among individuals who on a prior survey wave reported experiencing continued COVID-19 symptoms beyond 2 months after the initial month of symptoms. Results: Among 423 survey respondents reporting continued symptoms more than 2 months after acute test-confirmed COVID-19 illness, who then completed at least 1 subsequent survey, mean age was 53.7 (SD 13.6) years; 293 (69%) identified as women, and 130 (31%) as men; 9 (2%) identified as Asian, 29 (7%) as Black, 13 (3%) as Hispanic, 15 (4%) as another category including Native American or Pacific Islander, and the remaining 357 (84%) as White. Overall, 131/423 (31%) of those who completed a subsequent survey reported no longer being symptomatic. In Cox regression models, male gender, younger age, lesser impact of PCC symptoms at initial visit, and infection when the Omicron strain predominated were all statistically significantly associated with greater likelihood of remission; presence of ‘brain fog’ or shortness of breath were associated with lesser likelihood of remission. Conclusions and Relevance: A minority of individuals reported remission of PCC symptoms, highlighting the importance of efforts to identify treatments for this syndrome or means of preventing it.
