# Can AI Fix Electronic Medical Records?

### EHR
1. A YOUNG MAN, let’s call him Roger, arrives at the emergency department complaining[^10] of belly pain and nausea[^11]. A physical exam reveals[^12] that the pain is focused in the lower right part of his abdomen[^13]. The doctor worries that it could be appendicitis. But by the time[^14] the imaging results come back, Roger is feeling better, and the scan shows that his appendix appears normal. The doctor turns to the computer to prescribe[^15] two medications, one for nausea and Tylenol for pain, before discharging[^16] him.
1. This is one of the fictitious[^20] scenarios presented to 55 physicians[^21] around the country as part of a study to look at the usability[^22] of electronic health records[^23] (EHRs). To prescribe[^15] medications, a doctor has to find them in the EHR system. At one hospital a simple search for Tylenol creates a list of more than 80 options[^24]. Roger is a 26-year-old man, but the list includes Tylenol for children and infants[^25], as well as Tylenol for menstrual cramps[^26]. The doctor tries to winnow[^27] the list by typing the desired[^28] dose—500 milligrams—into the search window, but now she gets zero hits. So she returns to the main list and finally selects the 68th option—Tylenol Extra Strength (500 mg), the most commonly[^29] prescribed[^15] dose of Tylenol. What should have been[^210] a simple task has taken precious[^211] minutes and far[^212] more brainpower than it deserved[^213]. This is just one example of the countless[^214] agonizing[^215] frustrations[^216] that physicians deal with every day when they use EHRs.
1. These EHRs—digital versions of the paper charts in which doctors used to record patients’ visits, laboratory results and other important medical information—were supposed[^30] to transform the practice of medicine. The Health Information Technology for Economic and Clinical Health (HITECH)[^31] Act, passed[^32] resident Barack Obama said the shift[^35] would “cut waste, eliminate[^36] red tape[^37] and reduce[^38] the need to repeat expensive medical tests.” He added that it would “save lives by reducing the deadly but preventable[^39] medical errors that pervade[^310] our health care system.”
1. When HITECH was adopted[^40], 48 percent of physicians used EHRs. By 2017 that number had climbed to 85 percent, but the transformative[^41] power of EHRs has yet to be realized. Physicians complain[^42] about clunky[^43] interfaces and time-consuming[^44] data entry. Polls[^45] suggest that they spend more time interacting[^46] with a patient’s file than with the actual patient. Even Obama observed[^47] that the rollout[^48] did not go as planned. “It’s proven[^49] to be harder than we expected,” he told Vox in 2017.
1. Yet EHRs do have the potential to deliver[^50] insights[^51] and efficiencies[^52], according to physicians and data scientists. Artificial intelligence in the form of machine learning—which allows[^53] computers to identify patterns[^54] in data and draw conclusions[^55] on their own[^56]—might be able to help overcome the obstacles[^57] encountered[^58] with EHRs and unlock their potential for making predictions[^59] and improving patient care.
### DIGITAL DEBACLE
6. In 2016 the American Medical Association teamed up with MedStar Health, a health care organization that operates[^60] 10 hospitals in the Baltimore-Washington area, to examine[^61] the usability[^62] of two of the largest EHR systems, developed by Cerner, based in North Kansas City, and Epic, based in Verona, respectively[^63]. Together these two companies account[^64] for 54 percent of the acute[^65] care hospital market. The team recruited[^66] emergency physicians at four hospitals and gave them fictitious[^20] patient data and six scenarios, including the one about Roger, who presented with what seemed[^67] like appendicitis. These scenarios asked the physicians to perform[^68] common duties[^69] such as prescribing[^15] medications and ordering[^610] tests. The researchers assessed[^611] how long it took the physicians to complete each task, how many clicks were required[^612] and how accurately they performed[^68].
1. What they found was disheartening[^70]. The time and the number of clicks required varied[^71] from site to site and even between sites using the same system. And some tasks, such as tapering[^72] the dose of a steroid, proved[^49] exceptionally[^73] tricky[^74] across the board[^75]. Physicians had to manually calculate the doses, which took anywhere from two to three minutes and required[^612] 20 to 42 clicks. These design flaws[^76] were not benign[^77]. The physicians often made dosage[^78] mistakes. At one site the error rate reached 50 percent. “We’ve seen patients being harmed[^79] and even patients dying[^710] because of errors or issues[^711] that arise[^712] from usability[^22] of the system,” says Raj Ratwani, director of MedStar Health’s National Center.
1. But clunky[^43] interfaces are just part of the problem with EHRs. Another stumbling block[^80] is that information still does not flow[^81] easily between providers. The system lacks[^82] “the ability to seamlessly[^83] and automatically deliver data when and where it is needed under a trusted[^84] network without political, technical, or financial blocking,” according to a 2018 report from the National Academy of Medicine. If a patient changes doctors, visits urgent[^85] care or moves across the country, her records might or might not follow. “Connected care[^86] is the goal; disconnected care is the reality,” the authors wrote.
1. In March 2018 the Harris Poll conducted[^90] an online survey[^91] on behalf[^92] of Stanford Medicine that examined[^61] physicians’ attitudes[^93] about EHRs. The results were sobering[^94]. Doctors reported spending, on average, about half an hour on each patient. More than 60 percent of that time was spent interacting[^95] with the patient’s EHR. Half of office-based primary care physicians[^96] think using an EHR actually[^97] diminishes[^98] their clinical effectiveness. Isaac Kohane, a computer scientist and chair of the department of biomedical informatics at Harvard Medical School, puts it bluntly[^99]: “Medical records suck.”
1. It is already happening to some extent[^100]. In 2015 Epic began offering[^101] its clients machine-learning models. To develop these models, computer scientists start with algorithms and train[^102] them using real-world examples with known outcomes. For example, if the goal is to predict[^103] which patients are at greatest risk of developing the life-threatening[^104] blood condition known as sepsis, which is caused[^105] by infection, the algorithm might incorporate[^106] data routinely[^107] collected in the intensive care unit[^108], such as blood pressure, pulse and temperature. The better the data, the better the model will perform[^68].
1. Epic now has a library of models that its customers can purchase[^110]. “We have over 300 organizations either[^111] running[^112] or implementing[^113] models from the library today,” says Seth Hain, director of analytics and machine learning at Epic. The company’s sepsis-prediction model, which scans patients’ information every 15 minutes and monitors more than 80 variables[^114], is one of its most popular. The North Oaks Health System in Hammond, La., implemented[^113] the model in 2017. If a patient’s score reaches a certain threshold[^115], the physicians receive a warning, which signals them to monitor the patient more closely[^116] and provide antibiotics if needed. Since the health system implemented[^113] the model, mortality[^117] caused[^105] by sepsis has fallen by 18 percent.
1. But building and implementing[^113] these kinds of models is trickier[^74] than it might first appear[^120]. Most rely[^121] solely[^122] on an EHR’s structured data—data that are collected and formatted in the same way[^123]. Those data might consist[^124] of a blood-pressure reading, lab results, a diagnosis or a drug[^125] allergy. But EHRs include a wide variety[^71] of unstructured data, too, such as a clinician’s notes about a visit, e-mails and x-ray images. “There is information there, but it’s really hard for a computer to extract[^126] it,” says Finale Doshi-Velez, a computer scientist at Harvard University. Ignoring this free text means losing valuable[^127] information, such as whether[^128] the patient has improved[^129]. Moreover[^1210], Ratwani points out[^1211] that because of poor[^1212] usability[^22], data often end up[^1213] in the wrong spot[^1214]. For example, a strawberry allergy might end up documented in the clinical notes rather[^1215] than being listed[^1216] in the allergies box. In such cases, a model that looks[^1217] for allergies only in the allergy section of the EHR “is built off of inaccurate data,” he adds. “That is probably[^1218] one of the biggest challenges we’re facing[^1219] right now.”
1. Leo Anthony Celi, an intensive care specialist and clinical research director at the Massachusetts Institute of Technology’s Laboratory for Computational Physiology[^130], agrees. Most of the data found in EHRs are not ready to be fed[^131] into an algorithm. A massive[^132] amount[^133] of curation[^134] has to occur[^135] first. For example, say you want to design an algorithm to help patients in the intensive care unit[^108] avoid low blood glucose, a common problem. That sounds straightforward[^136], Celi says. But it turns out[^137] that blood sugar is measured[^138] in different ways[^123], with blood drawn[^139] from either[^111] a finger prick[^1310] or a vein. Insulin is administered[^1311] in different ways, too. When Celi and his colleagues examined all the data on insulin and blood sugar from patients at one hospital, “there were literally[^1312] thousands of different ways they were entered in the EHR.” These data have to be manually sorted and clustered[^1313] by type before one can even design an algorithm. “Health data is like crude[^1314] oil,” Celi says. “It is useless[^1315] unless[^1316] it is refined[^1317].”
### AN INTELLIGENT FIX
14. THE CURRENT PITFALLS[^140] of EHRs hamper[^141] efforts[^142] to use artificial intelligence to glean[^143] important insights[^51], but AI might itself provide a possible solution. One of the main drawbacks[^144] of the existing EHR systems, doctors say, is the time it takes to document a visit—everything[^145] from the patient’s complaint[^146] to the physician’s analysis and recommendation. Many physicians believe that much of the therapeutic value[^147] of a doctor visit is in the interactions[^46], Kohane says. But EHRs have “literally[^1312] taken the doctor from facing[^148] the patient to facing the computer.” Doctors have to type up[^149] their narrative[^1410] of the visit, but they also enter much of the same information when they order[^610] lab tests, prescribe[^15] medications and enter billing[^1411] codes, says Paul Brient, chief product officer[^1412] at athenahealth, another EHR vendor[^1413]. This kind of duplicate work contributes[^1414] to physician frustration[^216] and burnout.

Text: https://www.scientificamerican.com/article/can-ai-fix-electronic-medical-records/

### DICTIONARY

[^10]: complain - жаловаться
[^11]: nausea - тошнота
[^12]: reveals - показывать
[^13]: abdomen - живот
[^14]: by the time - к тому времени
[^15]: prescribe - прописать, назначить
[^16]: discharge - выписать (покинуть место)

[^20]: fictitious - выдуманный
[^21]: physicians - врачи
[^22]: usability - удобство
[^23]: EHRs - электронные медицинские карты
[^24]: options - варианты
[^25]: infants - младенцы
[^26]: cramps - спазмы
[^27]: winnow - отсеивать
[^28]: desire - желать
[^29]: commonly - обычно
[^210]: what should have been - что должно было быть
[^211]: precious - драгоценный
[^212]: far - гораздо
[^213]: deserve - стоить, заслуживать
[^214]: countless - бесчисленный
[^215]: agonizing - мучительный
[^216]: frustrations - разочарования

[^30]: this sth were suppose - предполагалось, что sth ...
[^31]: HITECH - Информационная технология здравоохранения для экономического и клинического здоровья
[^32]: pass - принять
[^33]: incentives - стимулы
[^34]: drive - побудить
[^35]: shift - изменения
[^36]: eliminate - ликвидировать
[^37]: red tape - бюрократия
[^38]: reduce - уменьшить
[^39]: preventable - предотвратимый
[^310]: pervade - проникать

[^40]: adopte - принять
[^41]: transformative - трансформационная
[^42]: complain - жаловаться
[^43]: clunky - неудобный
[^44]: time-consuming - трудоемкий
[^45]: polls - опросы
[^46]: interact - взаимодействовать
[^47]: observe - заметить
[^48]: rollout - развертывание
[^49]: prove - оказываться

[^50]: deliver - дать
[^51]: insights - понимание
[^52]: efficiencies - эффективность
[^53]: allow - позволять
[^54]: pattern - шаблон, закономерность
[^55]: draw conclusions - делать выводы
[^56]: own - собственный
[^57]: obstacle - препятствие
[^58]: encounter - сталкиваться
[^59]: predictions - прогнозы

[^60]: operate - управлять
[^61]: examine - исследовать
[^62]: usability - возможность
[^63]: respectively - соответсвенно
[^64]: account - приходится
[^65]: acute - необходимой
[^66]: recruit - набрать, принять на работу
[^67]: seem - казаться
[^68]: perform - выполнять
[^69]: duty - обязанность
[^610]: order - заказ
[^611]: assess - оценить
[^612]: require - потребовать

[^70]: dishearten - приводить в уныние
[^71]: vary - варьироваться
[^72]: taper - уменьшать
[^73]: exceptionally - исключительно
[^74]: tricky - сложно
[^75]: across the board - по всем направлениям
[^76]: flaws - недостатки
[^77]: benign - безобидный
[^78]: dosage - дозировка
[^79]: being harm - (как) .. причинять вред
[^710]: die - умирать
[^711]: issue - проблема
[^712]: arise - возникать

[^80]: stumbling block - камень преткновения
[^81]: flow - проходить
[^82]: lack - недостаток
[^83]: seamlessly - беспрепятственно
[^84]: under a trusted - под доверенной
[^85]: urgent - неотложная
[^86]: сonnected care - дистанционный контроль

[^90]: conduct - проводить
[^91]: survey - опрос
[^92]: behalf - имя
[^93]: attitude - отношения
[^94]: sober - отрезвлять
[^95]: interact - взаимодействовать
[^96]: primary care physician - терапевт (врач общей практики)
[^97]: actually - фактически
[^98]: diminish - уменьшать
[^99]: to put it bluntly - откровенно говорить

[^100]: extent - степень
[^101]: offer - предлагать
[^102]: train - обучать
[^103]: predict - предсказать
[^104]: life-threatening - опасный для жизни
[^105]: cause - вызывать
[^106]: incorporate - включать
[^107]: routinely - обычно
[^108]: unit - отделение

[^110]: purchase - купить
[^111]: either - или... (или...)
[^112]: run - работать
[^113]: implement - внедрять
[^114]: variable - переменная
[^115]: threshold - порог
[^116]: closely - внимательно
[^117]: mortality - смертность

[^120]: first appear - показаться на первый взгляд
[^121]: rely - полагаться
[^122]: solely - исключительно
[^123]: way - способ
[^124]: consist - состоять
[^125]: drug - препарат
[^126]: extract - извлекать
[^127]: valuable - ценный
[^128]: whether - ли
[^129]: improve - улучшаться
[^1210]: moreover - более того
[^1211]: point out - указывать
[^1212]: poor - плохой
[^1213]: end up - оказываться
[^1214]: spot - место
[^1215]: rather - скорее
[^1216]: list - занести
[^1217]: look - изучать
[^1218]: probably - вероятно
[^1219]: face - сталкиваться

[^130]: MIT's LfCP - лаборатория вычислительной физиологии Массачусетского технологического института
[^131]: feed - подать
[^132]: massive - массовый
[^133]: amount - количество
[^134]: curation - наблюдение
[^135]: occur - происходить
[^136]: straightforward - просто
[^137]: turn out - оказываться
[^138]: measure - измерять
[^139]: draw - тянуть
[^1310]: prick - укол
[^1311]: administer - применять
[^1312]: literally - буквально
[^1313]: cluster - группировать
[^1314]: crude - неочищенный
[^1315]: useless - бесполезный
[^1316]: unless - пока не
[^1317]: refine - очищать

[^140]: pitfalls - подводные камни
[^141]: hamper - препятствовать
[^142]: effort - усилие
[^143]: glean - получать
[^144]: drawback - недостаток
[^145]: everything - все
[^146]: complaint - жалоба
[^147]: value - ценность
[^148]: facing - встреча лицом к лицу
[^149]: type up - ввести
[^1410]: narrative - рассказ
[^1411]: bill - выставлять счет
[^1412]: CPO - главный специалист по продуктам
[^1413]: vendor - поставщик
[^1414]: contribute - способствовать