Table 1.

List of all used sources with their number of posts and with the available demographic attributes.

Dataset	No. of posts	Gendera	Agea	Ethnicitya	Locationa	Writing level
TwitterHealth [23]	11,637,888	Gender classifier	NO	Ethnicity classifier	YES	Writing level classifier
Google+Health [24]	186,666	YES	YES	Ethnicity classifier	YES	Writing level classifier
Drugs.com [25]	74,461	Gender classifier	NO	NO	NO	Writing level classifier
DailyStrength/Treatments [26]	1,055,603	YES	YES	NO	YES	Writing level classifier
WebMD/Drugs [27]	122,040	YES	YES	NO	NO	Writing level classifier
Drugs.com/Answers [28]	320,118	Gender classifier	NO	NO	NO	Writing level classifier
DailyStrength/Forums [29]	5,948,877	YES	YES	NO	YES	Writing level classifier
WebMD [30]	1,128,629	Gender classifier	NO	NO	NO	Writing level classifier

^aNO indicates that the demographic attribute is not provided by the source and no classifier is used due to low accuracy. YES indicates that the demographic attribute is provided by the source. More details on the demographic classifiers are available in the paper by Sadah et al [21].