PC2070: “googling” Cerebral Palsy: Lessons from Search Engine Machine Learning Algorithms

Background: Cerebral Palsy (CP) is the leading cause of childhood physical disability worldwide, but there is no definitive source of information for families of children who receive this diagnosis. Many parents seek information on the internet, but little is known about the types of questions parents ask regarding this condition. Modern search engines identify frequently used search terms to develop commonly asked questions using complex machine learning algorithms. By analyzing this data and the resources available online, we aim to better define common
parental inquiries regarding CP.

Methods: A Google search of “Cerebral Palsy” was performed; questions suggested by language processing algorithm were identified and recorded. Each question was categorized by Rothwell type (fact, policy, value) & classification and scored according to the JAMA benchmark quality criteria by two independent reviewers. The resources related to the questions were categorized by type (academic, government, commercial) and analyzed for reading grade level. Comparative univariate statistics were performed as appropriate.

Results: The top 100 questions suggested were included for analysis. The most common question type was fact based (70%). Of these, inquiries regarding technical details of CP were most frequent (38%), followed by natural history (31%), specific activities (13%), and etiology (11%). Websites sponsored by law firms or individual lawyers made up 32% of the associated websites, slightly more common than government websites (30%). Commercial and academic websites were less common at 21% and 12.5%, respectively. Academic sources trended toward having higher mean JAMA benchmark quality scores than the remaining sources (3 vs 2.5, p=0.054). Mean reading grade level for all resources was 11th grade, without significant difference between types. The interrater reliability ratings for both question and website type indicated very strong agreement at 0.92 and 0.87, respectively.

Conclusion: The most asked questions regarding CP were related to biological processes and natural history. Parents appear to be interested in understanding the mechanisms behind the disease. Academic resources trended toward achieving high JAMA benchmark quality scores but only made up 12.5% of the websites associated with these questions. Moreover, information provided across resources was written well above the recommended 6th grade reading level. Given the paucity of academically produced, quality, readable resources available for this common orthopaedic diagnosis, efforts need to be made to improve availability of reliable online resources. Furthermore, educating parents on how to evaluate online information should be emphasized during orthopaedic CP clinic visits.