Nowcasting Sexually Transmitted Infections in Chicago: Predictive Modeling and Evaluation Study Using Google Trends

Area of Science:

• Public Health
• Epidemiology
• Health Informatics

Background:

• Sexually transmitted infections (STIs) present a significant public health challenge in the U.S.
• Traditional STI surveillance systems face limitations including data quality issues, underreporting, and reporting delays.
• Search engine data offers a potential for efficient and cost-effective enhancement to existing STI surveillance systems.

Purpose of the Study:

• To develop and train a predictive model for STIs using reported case data from Chicago.
• To evaluate the model's predictive capacity, timeliness, and ability to target interventions to specific subpopulations.
• To investigate the utility of Google Trends data in conjunction with traditional surveillance methods.

Main Methods:

• Utilized deidentified STI case data (chlamydia, gonorrhea, syphilis) from Chicago (2011-2017).
• Employed Google Correlate to identify search terms related to "STD symptoms" and collected search volumes via Google Health API.
• Applied elastic net regression for prediction accuracy and cross-correlation analysis for timeliness, with subgroup analyses for race, sex, and age.

Main Results:

• High correlations between predicted and actual STI values were observed for chlamydia and gonorrhea from 2012-2017 (r=0.90-0.94).
• Primary and secondary syphilis showed high correlations in specific years (2012, 2013, 2016, 2017), with moderate correlations in others.
• Model performance was most accurate for gonorrhea, and subgroup analyses improved overall model fit and prediction accuracy.

Conclusions:

• Integrating nowcasting with Google Trends can enhance STI outbreak detection, prediction, and response timeliness.
• This approach allows for more targeted interventions to specific subpopulations.
• Future research should prospectively assess the utility of Google Trends in STI surveillance and response efforts.