ISSN 1470-3947 (print) | ISSN 1479-6848 (online)

Endocrine Abstracts (2019) 63 P641 | DOI: 10.1530/endoabs.63.P641

Digital epidemiology of hypothyroidism: pilot study

Ioannis Ilias, Ioannis Kakoulidis & Eftychia Koukkou


Department of Endocrinology, Diabetes and Metaboism, El. Venizelou Hospital, Athens, Greece.


Introduction: Digital epidemiology uses digital data which was not generated with the primary goal of serving epidemiological research (Life Sci Soc Pol 2018; 14:1). Recently, time and geographical correlation was noted between Google Trends findings and the prevalence of coronary artery disease (JACC, 2018, 71:98). Furthermore, using Google Trends, we have noted that the search terms ‘fatigue/weakness’ were correlated with the terms ‘hypothyroidism/thyroiditis’ worldwide (Cureus 2019; 11: e3965).

Aim: To use Google Trends search results on hypothyroidism versus published epidemiological data.

Methods: We used Google Trends for searches on ‘hypothyroidism’ worldwide for the years 2004–2018. We noted the – relative - popularity of this search term by country and compared it with available published data on the prevalence of hypothyroidism for ten countries (USA, Australia, UK, India, Thailand, Netherlands, Germany, Spain, Italy and Japan; Nat Rev Endocrinol 2018; 14: 301-316) using Pearson’s correlation.

Results: The Google Trends popularity for ‘hypothyroidism’ was highest in the USA at 97% and lowest in Italy and Japan with 1%. The reported prevalence of hypothyroidism ranged from 7.99% (India) to 0.20% (Spain and Italy). Pearson’s R was +0.51 (P=0.10).

Discussion: In this coarse pilot study we noted that internet search data showed some degree of correlation with the prevalence of hypothyroidism. Thus this approach provided information - albeit indirectly - on a disease’s pattern, supporting the use of Google Trends for epidemiology (PLoS One 2014; 9:109583). We have though to acknowledge that this approach has shortcomings, such as the use only of the search term in English. Furthermore, it is limited to internet-literate persons and that internet searches are often prone to be influenced by media reports on diseases and patients. A more robust analysis using disease incidence data would be the next step in the evaluation of this approach.

Article tools

My recent searches

No recent searches.

My recently viewed abstracts