From Web Search to Healthcare Utilization: Privacy-Preserving Studies from Mobile Data

OBJECTIVE

We explore relationships between health information seeking activities and engagement with healthcare professionals via analysis of geo-tagged data from mobile devices.

MATERIALS AND METHODS

We analyze logs of mobile interaction data stripped of individually identifiable information and location data. The data analyzed consists of time-stamped search queries and distances to medical care centers. We examine search activity that precedes the observation of salient evidence of healthcare utilization (EHU), taken as queries occurring at or near medical facilities.

RESULTS

We show that the time between symptom searches and observation of salient evidence of seeking healthcare utilization (EHU) depends on the acuity of symptoms. We construct statistical models that make predictions of forthcoming EHU based on observations about the current search session, prior medical search activities, and prior EHU. The predictive accuracy of the models varies (65-90%) depending on the features used and the time frame of the analysis.

DISCUSSION

We provide a privacy-preserving methodology that can be used to generate insights about the pursuit of health information and healthcare. The findings demonstrate how large-scale studies of mobile devices can provide insights on how concerns about symptomatology lead to the pursuit of professional care.

CONCLUSION

We present new methods for the analysis of mobile logs and describe a study that provides evidence about how people transition from mobile searches on diseases and symptoms to the pursuit of healthcare in the world.