U of T researchers develop tool to tell if tremors are real or a ruse by drug-seeking addicts

The most com­mon clin­i­cal sign of with­draw­al is tremor, espe­cial­ly in the hands and arms. But judg­ing tremor sever­i­ty is hard­er than it sounds; it requires con­sid­er­able med­ical exper­tise and even expe­ri­enced doc­tors’ esti­mates can vary wide­ly.

To assist physi­cians in deter­min­ing the sever­i­ty of a patients with­draw­al, researchers from the Uni­ver­si­ty of Toron­to have devel­oped the world’s first app to mea­sure tremor strength, pro­vid­ing objec­tive guid­ance to direct treat­ment deci­sions. The app also shows promise in mak­ing sol­id pre­dic­tions about whether the tremor is real or fake.

“There’s so much work to do in this field,” said Narges Norouzi, a PhD can­di­date in the Edward S. Rogers Sr. Depart­ment of Elec­tri­cal & Com­put­er Engi­neer­ing (ECE). “There is oth­er work out there on Parkinson’s tremors, but much less on tremors from alco­hol with­draw­al.”

Although with­draw­al is a poten­tial­ly fatal con­di­tion, physi­cians are often reluc­tant to pre­scribe ben­zo­di­azepines – a class of seda­tives used to treat con­di­tions such as alco­hol with­draw­al, anx­i­ety, seizures and  insom­nia. That’s because they’re fre­quent­ly abused and can be dan­ger­ous when mixed with oth­er drugs, espe­cial­ly alco­hol and opi­ates.

“The excit­ing thing about our app is that the impli­ca­tions are glob­al,” said Bjug Borgund­vaag, a pro­fes­sor at U of T’s Temer­ty Temer­ty Fac­ul­ty of Med­i­cine and an emer­gency physi­cian at the Schwartz/Reisman Emer­gency Cen­tre at Mount Sinai Hos­pi­tal.

“Alco­hol-relat­ed ill­ness is com­mon­ly encoun­tered not only in the emer­gency room, but also else­where in the hos­pi­tal, and this gives clin­i­cians a much eas­i­er way to assess patients using real data,” he added.

Experts say chron­ic alco­hol abusers often go to the emer­gency depart­ment claim­ing to be in with­draw­al in an effort to obtain ben­zo­di­azepines and it can be dif­fi­cult for inex­pe­ri­enced clin­i­cians to deter­mine if the patient is actu­al­ly in with­draw­al or “fak­ing” a with­draw­al tremor. Front-line health­care work­ers have no objec­tive way to tell the suf­fer­ers from the fak­ers. But researchers hope to change that.

“Our app may also be use­ful in assist­ing with­draw­al man­age­ment staff, who typ­i­cal­ly have no clin­i­cal train­ing, and deter­min­ing which patients should be trans­ferred to the emer­gency depart­ment for med­ical treat­ment or assess­ment. We think our app has great poten­tial to improve treat­ment for these patients over­all,” said Borgund­vaag.

Researchers test­ed their app on 49 patients expe­ri­enc­ing tremors in the emer­gency rooms at Toronto’s Schwartz/Reisman Emer­gency Med­i­cine Insti­tute at Mount Sinai Hos­pi­tal, St. Michael’s Hos­pi­tal and Women’s Col­lege Hos­pi­tal, as well as 12 nurs­es try­ing to mim­ic the symp­tom.

Their study shows that three-quar­ters of patients with gen­uine symp­toms had tremors with an aver­age peak fre­quen­cy high­er than sev­en cycles per sec­ond. Only 17 per cent of nurs­es try­ing to “fake” a with­draw­al tremor were able to pro­duce a tremor with the same char­ac­ter­is­tics, sug­gest­ing that this may be a rea­son­able cut-off for dis­crim­i­nat­ing real from fake. The app uses data from an iPod’s built-in accelerom­e­ter to mea­sure the fre­quen­cy of tremor for both hands for 20 sec­onds.

In the emer­gency room, clin­i­cians filmed their patients’ hand tremors while using the app and showed the footage to doc­tors after­ward. Norouzi found that her app’s abil­i­ty to assess tremor strength matched that of junior physi­cians, while more senior doc­tors were able to judge symp­toms with bet­ter accu­ra­cy. Norouzi’s next move is to con­tin­ue hon­ing the tool and com­par­ing its per­for­mance to doc­tors’ sub­jec­tive assess­ments, and to fur­ther study the effects of left- or right-hand­ed­ness.

“We have just begun to scratch the sur­face of what is pos­si­ble by apply­ing sig­nal pro­cess­ing and machine learn­ing to body-con­nect­ed sen­sors,” said Pro­fes­sor Parham Aara­bi of ECE. “As sen­sors improve and algo­rithms become smarter, there’s a good chance that we may be able to solve more med­ical prob­lems and make med­ical diag­no­sis more effi­cient.”

Norouzi and the team pre­sent­ed this work on August 29, 2014 at the Inter­na­tion­al Con­fer­ence of the IEEE Engi­neer­ing in Med­i­cine and Biol­o­gy Soci­ety in Chica­go.

-30-

U of T Media Rela­tions
Tel: 416–978-0100
media.relations@utoronto.ca