Min­ing Twit­ter to pre­dict ED vis­its

Modern Healthcare - - OUTLIERS ASIDES & INSIDES -

Don’t think of it as over­shar­ing. Re­searchers at the Uni­ver­sity of Ari­zona at Tucson and Park­land Hos­pi­tal in Dal­las say health-re­lated tweets qual­ify as epi­demi­o­log­i­cal data.

The re­searchers com­bined tweets about asthma with data taken from air-qual­ity sen­sors and elec­tronic health records and were able to pre­dict with 75% ac­cu­racy if the Park­land emer­gency depart­ment staff could ex­pect a high, low or medium num­ber of asthma-re­lated vis­its that day.

Dur­ing the three-month re­search project, Twit­ter saw mil­lions of tweets con­tain­ing the words “asthma,” “in­haler” or “wheez­ing.” But through text-min­ing tech­niques, re­searchers were able to iso­late those tweets gen­er­ated from the ZIP codes found in asthma pa­tients’ EHRs. Wors­en­ing air-qual­ity con­di­tions led to si­mul­ta­ne­ous in­creases in ED vis­its and asthma-re­lated tweets. Asthma-re­lated Google searches couldn’t pro­duce the same level of pre­dictabil­ity.

“You can get a lot of in­ter­est­ing in­sights from so­cial me­dia that you can’t from elec­tronic health records,” Sudha Ram, a Uni­ver­sity of Ari­zona pro­fes­sor of man­age­ment in­for­ma­tion sys­tems, said in a news re­lease.

Ram and col­leagues plan to ex­pand the study to 75 hos­pi­tals in the Dal­las-Fort Worth area. They also want to see whether they can get the same re­sults for di­a­betes­re­lated tweets.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.