As a component of an ongoing research project to address the gaps in medical care that contribute to delayed or missed diagnosis of serious disease and conditions, AcademyHealth has engaged in a study to examine how the use of a patient’s internet search data can be applied as a data source and tool to inform clinical diagnosis practices. Sponsored by a grant from the Gordon and Betty Moore Foundation, a literature review and comparative analyses were conducted utilizing data from internet searches by patients; the purpose of the review and analyses was to identify correlations (or a lack thereof) between each patient's searches and the corresponding patient's medical diagnoses. While many scientific reviews have previously summarized the utility of population-level internet search data for public health and health outcomes research, this is the first documented assessment of peer-reviewed publications focused on personalized medical diagnostic use. AcademyHealth conducted this study to help guide health services researchers, patient advocates, health care policy makers, and others to learn more about these data resources and their potential to inform new approaches to address diagnostic gaps and opportunities. This approach applies patient-generated data to assess the capability of the internet as a tool to improve understanding of symptoms, illuminate warning signs leading to a medical diagnostic workup, and empower caregivers to more effectively respond to health concerns.
The study design and findings represented in this report were reviewed by an Academy Health multidisciplinary steering committee that approved the report. The report details how researchers obtained and analyzed patient data from Microsoft and Google search engines to then link the data with the corresponding patient’s clinical information. All patient-generated data was provided by patients retrospectively with informed consent for research use. Of the 43 peer-reviewed publications identified and reviewed by AcademyHealth for inclusion in the report, all were retrospective analyses. The analyses included in the report represented a variety of medical and health considerations including cancer, mental health conditions, vulnerability, violence, and personal safety associated with neurodegenerative and aging disorders. Also included in the report were features published by researchers about tools that were developed to enable large data sets to be analyzed and integrated with other data, such as clinical trials and clinical care data. Cumulatively, these publications provide unique insights into the patient information needs, informed consent, and clinical research overview requirements needed to meet standards of conduct for research involving human subjects.
While many of these research publications identified signals of potential clinical use in diagnosis for specific diseases and conditions, none of the studies offered the type of confirmatory evidence that are required for medical use; at a minimum, evidence required would include prospective randomized studies. Therefore, this nascent type of diagnostic application research using large patient internet search data sets needs additional sponsorship to support infrastructure, analytic approaches, and importantly, innovative study design that enables prospective data collections to avoid bias and ultimately better understand the clinical efficacy and utility of this practice.
Further information can be found at the Exploring Consumer & Patient Internet Search Data to Improve Diagnosis Grant Program webpage. Innovation Horizons provided the research and analysis with support from the Gordon and Betty Moore Foundation.