JAMA_2002_Hersh.pdf

8/14/2019 JAMA_2002_Hersh.pdf

1/4

CONTEMPO UPDATES

LINKING EVIDENCE AND EXPERIENCE

Medical InformaticsImproving Health Care Through InformationWilliam R. Hersh, MD

HEALTH CARE IS AN INFORMATION-based science. Much of clini-cal practice involves gather-

ing, synthesizing, and acting oninformation. Medical informatics is thefield concerned with the managementand use of information in health andbiomedicine. This article focuses onthe problems that motivate work in

this field, the emerging solutions, andthe barriers that remain. It also ad-dresses the core themes that underlieall applications of medical informaticsand unify the scientific approachesacross the field.

There is a growing concern that in-formation is not being used as effec-tively as possible in health care. Recentreports from the Institute of Medicinehave reviewed research findings relatedto information use and expressed con-cerns about medical errors and patient

safety,1 the quality of medical records,2andtheprotection of patient privacy andconfidentiality.3 The latest Institute ofMedicine reporton this topic tiesall theseproblems and potential solutions to-gether in a vision for a health care sys-temthatissafe, patient-centered, andevi-dence-based.4 A variety of solutions isrequired to address the information-related problemsinhealthcare;many so-lutions involve theuseof computersandcomputer-related technologies.

The Field of Medical Informatics

Medical informatics is a heterog-eneous field, composed of individualswith diverse backgrounds and levels oftraining. Although virtually all healthscience universities have some entitywith the word informatics in its title,there are fewer than 25 that carry outresearch in medical informatics and

offer educational programs.5 At someinstitutions, medical informatics isviewed as a service (eg, helping clini-cians implement informatics applica-tions), but it is more appropriately con-sidered a science that addresses howbest to use information to improvehealth care. The government leader infunding research and education inmedical informatics has been the Na-tional Library of Medicine (www.nlm

.nih.gov).Some have argued that the adjective

medicalin front of informatics is inap-propriate because it implies the workof physicians and not the remainder ofhealth care and biomedical science.However, this name has achieved wide-spread usage. In the article by Kukafkaet al,6 Shortliffe described medical in-formatics as the broad term represent-ing the core theories, concepts, andtechniques of information applica-tions in health and biomedicine, with

the other adjectives preceding the wordinformaticsdenoting the specific appli-cation area (BOX). Core themes thatemerge from informatics science (stan-dards, terminology, usability,and dem-onstrated value) are relevant across alllevels of medical informatics, not solelyclinical informatics.

Applications of Clinical Informatics

There is a variety of classification typesfor the different applications of clini-cal informatics; one approach is by the

type of information used. There are es-sentially 2 types of information used inclinical informatics: patient-specific andknowledge-based. Patient-specific in-formation is generated by and used inthe care of patients in the clinical set-ting, whereas knowledge-based infor-mation comprises the scientific basis ofhealth care.

Electronic Medical Records

The core application using patient-specific information is the electronicmedical record (EMR). The paper-based medical record has its traditionand virtues; however, research hasshown it can be illegible, incomplete,difficult to access in more than oneplace, and insecure from unautho-rized uses and users.2 Although theEMR overcomes some of these prob-

lems, there are challenges to imple-menting the EMR at the levels of the in-dividual and the organization.

The main challenge to individual useof the EMR has been its integration intothe busy clinical workflow. The fewstudies that have been performed showcomputerized physician order entry(CPOE) adds time for the clinician, al-though other time savings are usuallygained elsewhere through error reduc-tion or the automation brought aboutby other features of the EMR (eg, ac-

cessing test results).7-9 A related chal-lenge is determination of the optimalcomputing device for the clinical set-ting. Handheld computers (also calledpersonal digital assistants) are increas-ingly popular, as documented by theiruse by internal medicine10 and familypractice physicians.11While their port-ability is of great value, most usage hasfocused on entry and retrieval of simpledata (eg, prescription writing and druginformation) and it is unclear whetherother usage (eg, image viewing and lit-

erature access)is amenable to these por-Author Affiliation: Division of Medical Informatics andOutcomes Research, School of Medicine, OregonHealth and Science University, Portland.Corresponding Author andReprints: WilliamR. Hersh,MD, Division of Medical Informatics and OutcomesResearch, School of Medicine, Oregon Health andSci-ence University,BICC, 3181 SW SamJackson ParkRd,Portland, OR 97201 (e-mail: [email protected]).Contempo Updates Section Editor: Janet M. Torpy,MD, Contributing Editor.

2002 American Medical Association. All rights reserved. (Reprinted) JAMA, October 23/30, 2002Vol 288, No. 16 1955


2/4

table devices or whether larger de-vices might be required.

At the organizational level, the keychallenges have been managing com-plex informatics applications and thecomputer networks upon which theyrun. Althoughindividual computers arerelatively inexpensive, maintaining

large networks of them and training themyriad of health care workers who usethem are not. Berg12 has noted that theinterpersonal challenges to large orga-nizations in implementing EMRs ismuch more daunting than managingthe technology itself. Research hasshown thatinvolving users in the imple-mentation process and providingfeatures of benefit to them, such astime-saving measures like specialty-specific order sets, widespread imple-mentation across the organization, and

engaging the clinical leadership, are themost important keys to success.13

Organizational challenges are not lim-ited to hospitals and other large insti-tutions. A particular problem in theout-patient setting is that small practicesusually lack the minimal technical ca-pabilities and financial resources nec-essary to implement EMRs.14 A finalchallenge to all involved with the EMRis the protection of patient privacy andconfidentiality, with the Health Insur-ance Portability and Accountability Act

legislating their protection at substan-tial cost and effort.15

Information Retrieval

The field concerned with the organi-zation and retrieval of knowledge-based information (notlimited to medi-cine) is called information retrieval.16

This area has seen tremendous growth

due to the Internet and World WideWeb. More than 50% of the US popu-lation uses the Web regularly, and ofthose who do, more than 50% searchfor personal health information.17 Phy-sicians have embraced the Web as well,with 90% using it on a regular basis.18

A large variety of online resources are

now available to both patients andphysicians. The oldest of these isMEDLINE, the bibliographic databaseof journal literature with more than 11million references to approximately4000 journals dating from 1966.19 Pro-duced by the US National Library ofMedicine, MEDLINE is freely avail-able to theentire world via thePubMedsystem on theNational Library of Medi-cineWeb site(http://pubmed.gov). Bib-liographic databases only contain titlesand abstracts of articles; however, us-

ers increasingly want the full text ofjournals online. The technical chal-lenge of producing such journals hasbeen supplanted by the economic ques-tion of how to increase availability whenelectronicpublication is cheaper (ie, lessprinting and mailing costs) but still hassome cost (due to value added in edit-ing and production).20 Another chal-lenge is how to improve users abili-ties to use such systems, as a systematicreview has shown physicians do not al-ways achieve optimal results with

them.

21

In addition to the journal literature,other knowledge-based information isavailable on the Web, including manyof the traditional medical textbooks,clinical practice guidelines (NationalGuidelines Clearinghouse [http://guideline.gov]), and a growing num-ber of Web sites aimed at patients and

consumers, many of which canbe foundusing the National Library of Medi-cines MedlinePlus database (http://medlineplus.gov). One concern aboutWeb sites is their quality, as peer-review mechanisms that normally con-

trol print literature are often not pres-ent, and sites purporting to providebalanced views may well be promotinga point of view or a product to sell.22

Concomitant with the growth of on-line information resources hasbeen theemergence of tools to use them moreeffectively. Probably the most impor-tant of these is evidence-based medi-cine.23 The original focus of evidence-based medicine was to train cliniciansto find and critically appraise indi-vidual studies. Few clinicians haveready access or the time required to

search MEDLINE, read articles, andsynthesize their findings in the busyclinicalsetting.As a result, thefocushaschanged toward approaches that pro-vide highly concise information in thecontext of the specific patient and clini-cal problem.24 This has led to a changein the emphasis of evidence-basedmedicinetowardtheproduction of syn-theses of clinicaltopics and concisesyn-opses of their findings.25

Decision Support Systems

Oneclinical informaticsapplication, thedecision support system, crosses theboundary of patient-specific and knowl-edge-based information. These appli-cations, which apply knowledge to pa-tient data, emerged from artificialintelligence and expert system re-search in the 1970s and 1980s that at-tempted to model the clinical diagnos-tician. However, the goal of building anelectronic diagnostician never materi-alized and the resulting systems weretootime-consuming for usein the clini-

cal setting.

26

These applications havere-emerged in the form of decision sup-port systems, which are embedded inthe EMR and aim to detect critical situ-ations and errors in care and then no-tify the clinician, provide appropriateinformation accordingly, or both.27

Several systematicreviews havedocu-mented the effectiveness of decision

Box. The 4 Levels of Medical Informatics and Their Respective Foci

From the Cell to the Population*

Bioinformaticsmolecular and cellularprocesses, such as gene sequences and mapsImaging informaticstissues and organs, such as radiology imaging systemsClinical informaticsclinicians and patients, including applications of nursing,

dentistry, and other clinical specialtiesPublic health informaticspopulations, such as disease surveillance systems

*Adapted from Kukafka et al.6

MEDICAL INFORMATICS

1956 JAMA,October 23/30, 2002Vol 288, No. 16 (Reprinted) 2002 American Medical Association. All rights reserved.


3/4

support applications within the EMR.A variety of approaches, including re-minders that increase adherence withordering preventive measures, hospi-tal admission order sets and display ofcosts, and interventions to detect medi-

cation prescribing errors, has beenfound beneficial.28-30 Additional workhas demonstrated the value of clinicalpracticeguidelinesto standardize care.31

One interesting finding of decisionsup-port systems applications is that theirbenefits do not appear to be educa-tional (ie, they do not result in clini-cians learning how to provide bettercare). This is illustrated by the fact thatwhen decision support systems appli-cations are removed, the adherence tospecific recommendationsreturnsto thepresystem baseline.32 Decision sup-

port systems provide timely remind-ers for busy clinicians.

Core Themes

There are a number of core themes thatunderlie medical informatics, not lim-ited to clinical informatics: standards,terminology, usability, and demon-strated value. As most applications inmedical informatics interact with oth-ers in a federation of different sys-tems, standards are essential to facili-tateintegrationof data, especially across

systems from different vendors.33

Onearea in which standards are particu-larly important is terminology.34Whilemost individuals know that an upperrespiratory tract infection, a cold, anda viral syndrome are similar, comput-ers do not inherently know this. If thebenefits of data aggregation are to beachieved (eg, comparing patient out-comes and resource utilization acrosspractices, institutions, and regions),then standard terminology is essen-tial. Standardizing the structure of clini-

cal documents will also make aggrega-tion and movement of data acrosssystems easier.35

Another core theme in medical infor-matics is usability. Systems must be in-tegrated into the clinical workflow anddemonstrate other benefits when theyrequire more time or effort on the partof the user. One example of research fo-

cused on usability found that a CPOEsystem that allowed physicians to typein orders in free text and mapped themto known order sets was faster andmoreacceptablethana standardsystem basedon a point-and-click interface.36 Re-

lated to usability is the need to demon-strate value. One study of CPOE dem-onstrated thatuseof guidelines anddoseselection menus resulted in signifi-cantly increased adherence to prescrib-ing regimens known to optimize pa-tient safety and reduce cost.37

Future Directions

Although considerable challengesremain, the impact of medical infor-matics will certainly grow. Theimperatives of improving documenta-tion, reducing error, and empowering

patients will continue to motivate useof information technology in healthcare. There is plenty of evidence thatclinical informatics applications canaddress these imperatives to enhancepatient outcomes, reduce costs, andprovide access to knowledge. Evi-dence alone will not be enough toensure their widespread adoption. Forexample, one concern is that most ofthe studies documenting the effective-ness of decision support have beencarried out in academic medical cen-

ters with institution-specific EMRsand may not be generalizable. Anotherconcern is that while the correction ofa small number of clinical anomalies(eg, improper drug doses) is feasible,the increase of this process might leadto cognitive overload of the clinician.It is also too early to know the truecost of widespread use of these appli-cations.38

The optimal use of clinical informat-ics applications will require some re-engineering of the health care system.

It will be crucial for the medical infor-matics fieldto accountfor theneeds andconcerns of all parties who participatein the process: patients, clinicians, pay-ers, and governments. Clinicians willhave to accept some impact on theirpractices, particularly as the indi-vidual physician becomes more ac-countable to document increasingly ex-

pensive care anddemonstrate avoidanceof error.They will have to accept CPOE,because this is the only place to effec-tively apply decision support. Societyas a whole will need to determine whowill pay the costs of EMRs and CPOE,

because even if they save money in thelong term, the up-front investment willbe substantial. The key challenge acrossall applicationswill be adherence to thebasic goals of the science of medical in-formatics: developing systems that areeasy to use and provide demonstrablebenefit.

Funding/Support: Dr Hershs work is supported bygrantsand contractsfromthe National Library ofMedi-cine and the Agency for Healthcare Research andQuality.

REFERENCES

1. KohnLT, Corrigan JM,Donaldson MS,eds. ToErrIs Human:Buildinga Safer HealthSystem. Washing-ton, DC: National Academy Press; 2000.2. Dick RS,Steen EB,DetmerDE, eds. The Computer-Based Patient Record: An Essential Technology forHealth Care, Revised Edition.Washington, DC: Na-tional Academy Press; 1997.3. For the Record: Protecting Electronic Health In-formation. Washington, DC: NationalAcademy Press;1997.4. Crossing the Quality Chasm: A New Health Sys-tem for the 21st Century.Washington, DC: NationalAcademy Press; 2001.5. American Medical Informatics Association. Re-source Center: Academic & Training Programs. Avail-able at: http://www.amia.org/resource/acad&training/f1.html. Accessed April 15, 2002.6. Kukafka R, OCarroll PW, Gerberding JL, et al. Is-sues and opportunities in public health informatics: apaneldiscussion.J Public Health Manag Pract. 2001;

7:31-42.7. Bates DW, Boyle DL, Teich JM. Impact of Com-puterized Physician Order Entry on Physician Time:Proceedings of the 18thAnnualSymposium on Com-puter Applications in Medical Care, 1994.Washing-ton, DC: Heinley & Belfus; 1994:996.8. Shu K, BoyleD, Spurr C,et al. Comparisonof timespent writing orders on paper withcomputerizedphy-sician order entry. Medinfo.2001;10:1207-1211.9. Overhage JM, Perkins S, Tierney WM, McDonaldCJ.Controlledtrialof direct physicianorderentry:ef-fectson physicians time utilization in ambulatorypri-mary care internal medicine practices. J Am Med In-form Assoc.2001;8:361-371.10. American College of Physicians, American Soci-ety of Internal Medicine. ACP-ASIM survey findsnearlyhalf of US members use handheld computers. Avail-able at: http://www.acponline.org/college/pressroom/handheld_survey.htm. Accessed April 15, 2002.

11. Criswell DF, Parchman ML. Handheld computeruse in US Family Practice residency programs. J AmMed Inform Assoc.2002;9:80-86.12. Berg M. Implementing information systems inhealth care organizations: myths and challenges. IntJ Med Inf.2001;64:143-156.13. Ahmad A, Teater P, Bentley TD, et al. Key at-tributes of a successful physician order entry systemimplementation in a multi-hospital environment.J AmMed Inform Assoc.2002;9:16-24.14. Wager KA, Lee FW, White AW, et al. Impact ofan electronic medical record system on community-

MEDICAL INFORMATICS

2002 American Medical Association. All rights reserved. (Reprinted) JAMA, October 23/30, 2002Vol 288, No. 16 1957


4/4

based primary care practices.J Am Board Fam Pract.2000;13:338-348.15. Gostin LO. National health information privacy:regulations under theHealth Insurance Portability andAccountability Act.JAMA.2001;285:3015-3021.16. HershWR. InformationRetrieval:A Health CarePerspective.2nd ed. New York, NY: Springer Verlag;2002.17. Fox S, Rainie L. The online health care revolu-

tion: how the Web helps Americans take better careof themselves. Available at: http://www.pewinternet.org/reports/pdfs/PIP_Health_Report.pdf. AccessedApril 15, 2002.18. Taylor H, Leitman R. The increasing impact ofeHealth on physician behavior. Available at: http://www.harrisinteractive.com/news/newsletters/healthnews/HI_HealthCareNews2001Vol1_iss31.pdf. Accessed April 15, 2002.19. Wood EH.MEDLINE: theoptions forhealthpro-fessionals. J Am Med Inform Assoc. 1994;1:372-380.20. Hersh WR, Rindfleisch TC. Electronic publishingof scholarly communication in the biomedical sci-ences.J Am Med Inform Assoc.2000;7:324-325.21. Hersh WR, Hickam DH. How well do physiciansuse electronic information retrieval systems? JAMA.1998;280:1347-1352.22. Silberg WM, Lundberg GD, Musacchio RA. As-sessing, controlling,and assuring the quality of medi-cal information on the Internet. JAMA. 1997;277:1244-1245.

23. Sackett DL, RichardsonWL, RosenbergW, HaynesRB.Evidence-Based Medicine: How to Practice andTeach EBM. New York, NY: Churchill Livingstone;2000.24. Chueh H, Barnett GO.Just-in-timeclinical in-formation.Acad Med.1997;72:512-517.25. HaynesRB. Of studies, syntheses, synopses,andsystems:the 4S evolution of servicesfor findingcur-rent best evidence.ACP J Club.2001;134:A11-A13.

26. MillerRA. Medical diagnostic decision support sys-temspast, present, andfuture.J Am Med InformAs-soc.1994;1:8-27.27. Bates DW, Cohen M, Leape LL, et al. Reducingthe frequency of errors in medicine using informa-tion technology. J Am Med Inform Assoc. 2001;8:299-308.28. Hunt DL, Haynes RB, Hanna SE, Smith K. Effectsof computer-based clinical decision support systemson physician performance and patient outcomes.JAMA.1998;280:1339-1346.29. Gandhi TK, Bates DW. Computer adverse drugevent (ADE)detection andalerts. In: ShojaniaK, Dun-can B, McDonald K, Wachter R, eds. Making HealthCare Safer:A Critical Analysis of Patient SafetyPrac-tices.Rockville, Md: Agency for Healthcare Researchand Quality; 2001:79-85. AHRQ PublicationNo. 01-E058.30. KaushalR, Bates DW. Computerized physician or-der entry (CPOE) with clinical decision support sys-tems (CDSSs). In: Shojania K, Duncan B, McDonaldK, Wachter R, eds.Making Health Care Safer: A Criti-

cal Analysisof Patient Safety Practices. Rockville, Md:Agency for Healthcare Research and Quality; 2001:59-69. AHRQ Publication No. 01-E058.31. ShiffmanRN, LiawY, BrandtCA, CorbGJ. Com-puter-based guideline implementation systems.J AmMed Inform Assoc.1999;6:104-114.32. McDonald CJ, Wilson GA, McCabe GP. Physi-cian response to computer reminders. JAMA.1980;244:1579-1581.

33. McDonald CJ. The barriers to electronic medicalrecord systemsand howto overcome them.J Am MedInform Assoc.1997;4:213-221.34. Cimino JJ. Desiderata for controlled medical vo-cabularies in the twenty-first century.MethodsInf Med.1998;37:394-403.35. Dolin RH, Alschuler L, Beebe C, et al. The HL7clinical document architecture.J Am Med Inform As-soc.2001;8:552-569.36. Lovis C, Chapko MK, Martin DP, et al. Evalua-tionof a command-line parser-based order entry path-way for the Department of Veterans Affairs elec-tronic patient record.J Am Med Inform Assoc.2001;8:486-498.37. Teich JM, Merchia PR, Schmiz JL, et al. Effectsof computerized physician order entry on prescrib-ing practices. Arch Intern Med. 2000;160:2741-2747.38. OverhageJM, MiddletonB, MillerRA, etal. Doesnational regulatory mandate of provider order entryportend greater benefit than risk for health care de-livery?J Am Med Inform Assoc.2002;9:199-208.

The whole purport of literature . . . is the notation oftheheart. Style is but thecontemptible vessel in whichthe bitter liquid is recommended to the world.

Thornton Wilder (1897-1975)

MEDICAL INFORMATICS

1958 JAMA,October 23/30, 2002Vol 288, No. 16 (Reprinted) 2002 American Medical Association. All rights reserved.

Documents

JAMA_2002_Hersh.pdf