Correlation between cancer incidences and Google searches in the United States

Abstract

IntroductionDespite being highly prevalent, keratinocyte carcinomas (basal cell and squamous cell carcinomas lack nationwide registries. Internet search data has emerged as a new method to evaluate previously difficult to quantify public health outcomes and may be useful in keratinocyte carcinoma research.ObjectiveWe aimed to evaluate whether Google search density correlated with known incidences of common cancers in the United States.MethodsWe used the Center for Disease Control’s National Program of Cancer Registries ageTadjusted cancer incidences (2008T2012 . We collected Google search data, normalized for total search volume, using Google trends (google.com/trends . We collected data on the ten most incident cancers in the United States: lung, breast, colon, prostate, melanoma, endometrial, bladder, thyroid, NonTHodgkin’s lymphoma, kidney/renal pelvis. We utilized Pearson’s correlation coefficient to evaluate the relationship between known cancer incidence and Google search density by state.ResultsFour cancers (endometrial, bladder, thyroid, kidney/renal pelvis had insufficient Google search quantity among individual states to be evaluated. Lung cancer (R2=0.70, p<0.001 , colon cancer (R2=0.60, p<0.001 , melanoma (R2=0.42, p=0.002 , and NonTHodgkin’s lymphoma (R2=0.47, p=0.006 had statistically significant correlations between actual incidences and Google searches. Breast and prostate cancer incidences were not correlated (p>0.05 .DiscussionFour of the six highly incident cancers evaluated had statistically significant correlations between known incidence and Google search density. Internet search data may be a novel tool to estimate geographical incidence and prevalence of disease. This methodology may be particularly useful for keratinocyte carcinomas, which currently lack nationwide registries