IN­FOR­MA­TION TECH­NOL­OGY: Ar­ti­fi­cial in­tel­li­gence seen as de­ci­sion sup­port aid

Ar­ti­fi­cial in­tel­li­gence seen as aid­ing de­ci­sion sup­port

Modern Healthcare - - NEWS - Joseph Conn

The decades-old quest to bring ar­ti­fi­cial in­tel­li­gence to bear on ev­ery­day health­care de­ci­sion­mak­ing is mov­ing in­ex­orably for­ward, ac­cord­ing to health in­for­mat­ics ex­perts. So what’s to be made of re­cent an­nounce­ments by In­di­ana Univer­sity and IBM Corp. about ad­vances in health­care AI? The ap­par­ent progress is just great, the ex­perts said, but it’s still very early in the game.

“I think there is a new gen­er­a­tion of tech­nol­ogy coming out in AI, in com­pu­ta­tional power, that will make a big dif­fer­ence,” said Casey Ben­nett, an IU com­puter sci­en­tist. “So, I think over the next 10 years, you’ll see a big change.”

“I think we’re start­ing to move to the next do­main within clin­i­cal de­ci­sion sup­port,” said Dr. Wil­liam Bria, a veteran physi­cian in­for­mati­cist, pres­i­dent of the As­so­ci­a­tion of Med­i­cal Direc­tors of In­for­ma­tion Sys­tems and chief med­i­cal of­fi­cer of Dataskill, a San Diego sys­tems in­te­gra­tor where Bria is help­ing de­velop clin­i­cal de­ci­sion-sup­port, or CDS, sys­tems that lever­age the com­put­ing power of IBM’s Wat­son su­per­com­puter sys­tem.

How long be­fore so­phis­ti­cated CDS sys­tems such as those will be in wide­spread rou­tine use by clin­i­cians? “An­other decade,” Bria said. Of the two an­nounce­ments, the one with the big­ger bal­ly­hoo came from IBM, with con­tri­bu­tions also by in­surance gi­ant Wel­lPoint and New York’s renowned Me­mo­rial Sloan-Ket­ter­ing Can­cer Cen­ter, all re­lated to Wat­son.

Fa­mous for its 2011 vic­to­ries on the TV game show “Jeop­ardy!”, Wat­son has since been crammed with 1.5 mil­lion pa­tient med­i­cal records, 2 mil­lion med­i­cal jour­nal ar­ti­cles and more than 600,000 “pieces of med­i­cal ev­i­dence,” ac­cord­ing to an IBM news re­lease tout­ing its lat­est “break­through” de­vel­op­ments.

One is a Wat­son-pow­ered on­col­ogy clin­i­cal-de­ci­sion sup­port tool. In cre­at­ing the tool, Wat­son re­ceived about a year of tu­tor­ing from on­col­o­gists at Me­mo­rial Sloan-Ket­ter­ing.

But the CDS tool it­self is still “in train­ing,” ac­cord­ing to a Me­mo­rial Sloan-Ket­ter­ing spokes­woman. It is not yet linked to, much less in­te­grated into, the can­cer cen­ter’s elec­tronic health-record sys­tem, it won’t be tested in a clin­i­cal set­ting un­til fall and a com­mer­cial prod­uct is not ex­pected for re­lease un­til the end of 2013, the spokes­woman said.

Mean­while, in an­other Wat­son ap­pli­ca­tion, nurses and tech­ni­cians at Wel­lPoint have worked with the com­puter for more than a year, with more im­me­di­ately ser­vice­able re­sults, ac­cord­ing to IBM. They have devel­oped a Wat­son-lever­aged sys­tem that, in De­cem­ber 2012, be­gan pro­cess­ing provider re­quests to Wel­lPoint for ap­proval of com­mon med­i­cal pro­ce­dures. Wel­lPoint plans to have more than 1,600 providers us­ing it by the end of the year, ac­cord­ing to a news re­lease.

Ben­nett is a doc­toral stu­dent and in­struc­tor at IU’s School of In­for­mat­ics and Com­put­ing and a re­search fel­low at Cen­ter­stone Re­search In­sti­tute, an arm of Cen­ter­stone, a be­hav­ioral health­care provider based in Colum­bus, Ind., and Nashville. Ben­nett and col­league Kris Hauser, an as­sis­tant pro­fes­sor of com­puter sci­ence at IU, are co-au­thors of a re­cently pub­lished re­search report, “Ar­ti­fi­cial in­tel­li­gence frame­work for sim­u­lat­ing clin­i­cal de­ci­sion­mak­ing: A Markov de­ci­sion process ap­proach,” in the jour­nal Ar­ti­fi­cial In­tel­li­gence in Medicine.

The report tells the story of an AI sys­tem they devel­oped that sec­ond-guessed the de­ci­sions of clin­i­cians in 500 cases in­volv­ing be­hav­ioral-health pa­tients with a va­ri­ety of other chronic ill­nesses, such as hy­per­ten­sion, di­a­betes and car­dio­vas­cu­lar disease. Their AI sys­tem came up with alternative, ev­i­dence­based care paths that out­per­formed those of the hu­man de­ci­sion­mak­ers in terms of cost while boost­ing pa­tient out­comes 30% to 35%, the au­thors said.

Since publi­ca­tion, they have re­ceived some “an­gry re­sponses” in “some random, un­so­licited e-mail,” Ben­nett said. Ac­cept­ing a greater level of re­liance on com­puter sup­port will be “a bit of a cul­ture shift,” for some physi­cians, he said.

“I think the younger gen­er­a­tion is more open to it,” he said.

Hard or easy, a shift to more re­liance on clin­i­cal sup­port tools is go­ing to come, Ben­nett said.

“With mean­ing­ful use and Oba­macare, there is go­ing to be a po­lit­i­cal shift that’s go­ing to force change,” he said. With­out it, he be­lieves, soar­ing health­care costs will bank­rupt the coun­try.

“So, you won’t be able to quib­ble about it,” Ben­nett said. “It will hap­pen whether peo­ple like it or not.”

“I think it’s all good news, it’s all pos­i­tive,” said Dr. Mitchell Feld­man, who has worked for more than two decades on a com­put­er­ized di­ag­nos­tic sup­port tool, DX­plain, first devel­oped at the Lab­o­ra­tory of Com­puter Sci­ence at Mas­sachusetts Gen­eral Hospi­tal, Bos­ton. “It’s try­ing to bring more ev­i­dence and more data and more sci­ence and a lit­tle less art to medicine,” Feld­man said.

For now, though, “No one I know in in­for­mat­ics is ready to trust their pa­tient care to a com­puter, but is man plus ma­chine bet­ter than man alone? The an­swer is def­i­nitely yes.”


Dr. Mark Kris of Me­mo­rial Sloan-Ket­ter­ing Can­cer Cen­ter, left, and Manoj Sax­ena of IBM demon­strate med­i­cal ap­pli­ca­tions con­nected to IBM’s Wat­son su­per­com­puter.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.