Validation of a Machine Learning Model to Predict Childhood Lead Poisoning.

Importance: Childhood lead poisoning causes irreversible neurobehavioral deficits, but current practice is secondary prevention. Objective: To validate a machine learning (random forest) prediction model of elevated blood lead levels (EBLLs) by comparison with a parsimonious logistic regression. Design, Setting, and Participants: This prognostic study for temporal validation of multivariable prediction models used data from the Women, Infants, and Children (WIC) program of the Chicago Department of Public Health. Participants included a development cohort of children born from January 1, 2007, to December 31, 2012, and a validation WIC cohort born from January 1 to December 31, 2013. Blood lead levels were measured until December 31, 2018. Data were analyzed from January 1 to October 31, 2019. Exposures: Blood lead level test results; lead investigation findings; housing characteristics, permits, and violations; and demographic variables. Main Outcomes and Measures: Incident EBLL (≥6 µg/dL). Models were assessed using the area under the receiver operating characteristic curve (AUC) and confusion matrix metrics (positive predictive value, sensitivity, and specificity) at various thresholds. Results: Among 6812 children in the WIC validation cohort, 3451 (50.7%) were female, 3057 (44.9%) were Hispanic, 2804 (41.2%) were non-Hispanic Black, 458 (6.7%) were non-Hispanic White, and 442 (6.5%) were Asian (mean [SD] age, 5.5 [0.3] years). The median year of housing construction was 1919 (interquartile range, 1903-1948). Random forest AUC was 0.69 compared with 0.64 for logistic regression (difference, 0.05; 95% CI, 0.02-0.08). When predicting the 5% of children at highest risk to have EBLLs, random forest and logistic regression models had positive predictive values of 15.5% and 7.8%, respectively (difference, 7.7%; 95% CI, 3.7%-11.3%), sensitivity of 16.2% and 8.1%, respectively (difference, 8.1%; 95% CI, 3.9%-11.7%), and specificity of 95.5% and 95.1% (difference, 0.4%; 95% CI, 0.0%-0.7%). Conclusions and Relevance: The machine learning model outperformed regression in predicting childhood lead poisoning, especially in identifying children at highest risk. Such a model could be used to target the allocation of lead poisoning prevention resources to these children.

Tweeting for and against public health policy: response to the Chicago Department of Public Health's electronic cigarette Twitter campaign.

BACKGROUND: In January 2014, the Chicago City Council scheduled a vote on local regulation of electronic cigarettes as tobacco products. One week prior to the vote, the Chicago Department of Public Health (CDPH) released a series of messages about electronic cigarettes (e-cigarettes) through its Twitter account. Shortly after the messages, or tweets, were released, the department's Twitter account became the target of a "Twitter bomb" by Twitter users sending more than 600 tweets in one week against the proposed regulation. OBJECTIVE: The purpose of our study was to examine the messages and tweet patterns in the social media response to the CDPH e-cigarette campaign. METHODS: We collected all tweets mentioning the CDPH in the week between the e-cigarette campaign and the vote on the new local e-cigarette policy. We conducted a content analysis of the tweets, used descriptive statistics to examine characteristics of involved Twitter users, and used network visualization and descriptive statistics to identify Twitter users prominent in the conversation. RESULTS: Of the 683 tweets mentioning CDPH during the week, 609 (89.2%) were anti-policy. More than half of anti-policy tweets were about use of electronic cigarettes for cessation as a healthier alternative to combustible cigarettes (358/609, 58.8%). Just over one-third of anti-policy tweets asserted that the health department was lying or disseminating propaganda (224/609, 36.8%). Approximately 14% (96/683, 14.1%) of the tweets used an account or included elements consistent with "astroturfing"-a strategy employed to promote a false sense of consensus around an idea. Few Twitter users were from the Chicago area; Twitter users from Chicago were significantly more likely than expected to tweet in support of the policy. CONCLUSIONS: Our findings may assist public health organizations to anticipate, recognize, and respond to coordinated social media campaigns.