J Nurs Can Quat Vol. 18, No. >, pp. 129-138 Wilkin.s, Int ... an Expert System 131 For example, when comparing several nurs-ing texts, nursing interventions for the nursing diagnosis.

J Nurs Can QuatVol. 18, No. >, pp. 129-138© 2O<>3 Uppintott WillLuns & Wilkin.s, Int

Building an Expert SystemA Systematic Approach to Developingan Instrument for Data ExtractionFrom the Literature

Renee Daiuta Feuerhach, MS, ANP;Teresa L. Panniers, PhD, RN, CRNP

When building an expert system that wili be acceptable to clinicians in their practice, it i.s imper-ative that the knowledge engineer identifies, defines, and describes a clinical problem precisely.This can be accomplished by eliciting private knowledge from expert clinicians or by analyzingpublic knowledge available in the scientific literature. This study describes a systematic method forexamining public knowledge found in health care textbooks and practice gtiideliiies surroundingthe concept of oral feeding in premature infants in a neonatal intensive care unit. It includes tbedevelopment of an instrument for extracting data from these sources to standardize definitions ofterminologies. Preliminary' results and plans for further analysis are reported. Tbis method can beapplied to other clinical problems deemed appropriate tor decision support system development.Key words: decision making: decision making, computer-assisted; decision support systems,clinical; decision stipport techniques; expert systems: nursing informatics

EXPERT .systems arc autotnated know-ledge-based systems that represent a spe-

cific expression of medical and nursing in-formatics to help caregivers interpret dataand make better decisiotis. They can improveclinical practice and foster positive healthoutcomes. In order to build an expert sys-tem it is imperative that the knowledge engi-neer identifies, defines, and describes a clin-ical problem precisely. This precision facili-tates acceptance of tbe system by cliniciansand enhances communication when decisi-ons are modeled.''^ Clinical problems can bedefined with precision by eliciting 2 types ofknowledge, private and public. Private knowl-edge consists of heuristic atid experientiallybased information that has not been madeavailable in the literature. This type of knowl-

edge usually comes from expert clinicians.Public knowledge consists of knowledge thathas been made available in the literature andusually comes from textbooks, reports, andjournals. * The purpose of this article is to re-port a systematic analysis of ptiblic knowl-edge fotind in health care textbooks iUid prac-tice guidelines surroimding the concept oforal feeding in premature infants. Walker andAvant's concept analysis and Weber s contentanalysis were adapted for the developmentof an instrument tbat extracts data from text-books and practice guidelines abotit the defi-nition of oral feeding.'''' Ultimately, the resultsof this analysis will be used to build an expertsystem, modeling decisions made by neona-tal ntirse practitioners (NNPs) when itiitiatingoral feeding in premature itifants cared for inthe neonatal intensive care unit *^

From the Division of Nursing, Steinhardt School ofEducation, New York University, New York, NY.

Ciirresponding author: Teresa L Panniers, PhD, RN.CKNP, Dit<ision of Nursing, Steinhardt Scbool of Educa-tion, New York University, 246 Greene Street, New York.NY lOOOJ (email: [email protected]).

SIGNIFICANCE

Despite extensive work on the standard-ization of ntirsing terminologies to describediagnoses, interventions, outcomes, and

129

130 JOURNAL OF NURSING CARE QLIALITY/APRIL-JUNE 2003

patient problems, a concept-oriented ap-proach to their development is lacking/"'^Additional issues related to the developmentof nursing terminologies include (1) the exis-tence of multiple data sets in some practiceareas and the lack of even 1 classificationsystem in other practice areas, and (2) thefact that existing terminologies are usuallydeveloped for human interpretation, withcomputer interpretation as a secondarygoal.** This is significant because knowledgethat is understandable to humans is oftenambiguous when applied to an automatedsystem.

The Nursing Minimum Data Set (NMDS)was designed to organize nursing terminolo-gies, creating a common nursing frameworkto describe nursing practice and evaluate itseffectiveness across practice settings. It in-ciudes the data elements of nursing diag-noses, interventions, and outcomes as well asa nursing care intensity measure and a healthrecord number.'' '" In 1973 the North Ameri-can Nursing Diagnosis Association (NANDA)was formed to develop a classification ofnursing diagnoses. A list of over 100 di-agnoses was compiled representing clinicalproblems in nursing." In 1992 the Nurs-ing Interventions Classification (NIC) team,headed by McCloskey and Bulechek, was in-terested in standardizing nomenclature to linknursing diagnoses and interventions for thepurpose of building information systems.'-They constructed and validated a taxonomyof nursing Interventions called the NIC. TheNursing Outcomes Classification (NOC) teamfollowed development of the NIC systemwith a comprehensive taxonomy of standard-ized patient outcomes, applicable across thecare continuum, to link to these nursinginterventions.'^""

In 1995 the American Nurses Associationestablished the Nursing Information and DataSet Evaluation Center (NIDSEC) to developand disseminate st:mdards pertaining to clini-cal information systems that support the doc-umentation of nursing practice. NIDSEC de-veloped standards to evaluate the following 4

dimensions of nursing data sets and the sys-tems that contain them: nomenclature, clini-cal content, clinical data repository, and gen-eral system characteristics.^' NANDA, NMDS,NIC, and NOC are 4 of the American NursesAssociation recognized nomenclatures. Ad-ditional data sets recognized by the Amer-ican Nurses Association include the HomeHealth Care Classification (HHCC), the Om-aha Problem Classification System, the PatientCare Data Set (PCDS), the Perioperative Nurs-ing Data Set, and the Systematized Nomen-clature of Human and Veterinary Medicine(SNOMED).

Saba's HHCC derived terms from patientrecords in a national sample of home healthcare agencies, producing 147 nursing di-agnoses and interventions.'* This classifica-tion scheme is part of a nursing informa-tion system designed to predict resourcerequirements. ' ' -^ The Omaha System was de-veloped in a community' health setting. It isa problem-oriented scheme and includes aproblem rating scale.'"'•^ Ozbolt's PCDS isfocused on the acute care setting. Standardterms were derived from patient care plan-ning and documentation materials from 9 hos-pitals in the Linited States. This standardiza-tion led to the development of a computerprogram designed to assist nurses in mak-ing decisions in each step of the nursmgprocess.'" i"* 'y-^-^ More recently, the Periop-erative Nursing Data Set, an automated lan-guage, was developed. This data set is com-posed of 4 domains. The domains are safety,physiologic response to surgery, patient andfamily behavioral response to surgery, andhealth system. Each domain includes nurs-ing diagnoses, interventions, and outcomesrelated to perioperative nursing care.'"'

Methodological challenges related to creat-ing these nomenclatures have been reportedin the literature.'"-^'"'' The NIC team re-ported difficulty in completing their initial listof nursing interventions because of the largenumber of available nursing textbooks anddifferences among texts regarding the nurs-ing interventions for a particular condition.

Building an Expert System 131

For example, when comparing several nurs-ing texts, nursing interventions for thenursing diagnosis. Activity Intolerance, var-ied greatly. A systematic method for theselection of sources was developed. After thisselection process was completed, nursing in-terventions related to the diagnosis were col-lected, despite variability among sources.'' Indeveloping their initial list of nursing-sensitivepatient outcomes, the NOC team identified alarge variety of sources. For this reason theydesigned a sampling plan and selection cri-teria to determine which sources would heused.^'' Ozholt discussed issues in determin-ing how to assign order to the standard termsderived from in-patient records. She adaptedSaha s 20 components of nursing diagnosesand interventions in home care as a system oforganization.^*

Nursing concepts must be defined pre-cisely when building automated systems de-signed to commimicate information about pa-tient care, health outc<jmcs, resources, cost,policy, research, and clinical and administra-tive decision-making across multiple healthcare disciplines and settings. Since most nurs-ing classification systems were designed forhuman interpretation, they do not providethe level of precision necessary fbr com-puters. For this reason a concept-orientedstrategy for standardizing nursing terminolo-gies with the precision necessary fbr automa-tion is important. Evaluation criteria relatedto concept-oriented approaches in computer-based systems have been developed.*^ Theseinclude the ability to separate concepts intoconstituent components, combine simpleconcepts into composite concepts, retainconcepts once these have been defined, sup-port multiple linguistic expressions and hier-archies, avoid redundancy, support synonymy,and provide precise, explicit definitions foreach term. These criteria are particularly im-portant when developing expert systems be-cause these systems require more precise,finely granular data than typically exists in theclassification systems discussed previously.SNOMED is an example ofa concept-oriented

expert system.**'^ It is a comprehensive set ofover 150,000 health care terms. Concepts in-clude anatomy, morphology, normal and ab-normal functions, symptoms and signs of dis-ease, chemicals, drugs, enzymes and otherhody proteins, living organisms, physicalagents, occupations, social contexts, diseases/diagnoses, and procedures. Computer-basedtools are used to describe, manipulate, andreason about these concepts.

THE METHOD PHASE I

Determining how to systematicallyapproach the literature

Because of the vastness and variability ofthe literature, a systematic approach to de-termine how and w^here to locate textbooksand practice guidelines was necessary. Text-books :md practice guidelines were used be-cause these included a more comprehensivedefinition of the concept of oral feeding ascompared to journals. In consultation witha health sciences librarian, 3 published listsof books and journals, 2 unpublished lists ofbooks, and 4 electronic databases were se-lected. The published lists included Brandon/Hill's Selected List of Print Nursing Books andJournals, Brandon/Hill's Selected List of Booksand Journals for the Small Medical Library,and the Interagcncy Council's List of Informa-tion Resources for Nursing.- ^" ' The unpub-lished lists were provided by The NationalCertification Board of Pediatric Nurse Prac-titioners and Nurses, Inc. and Ibe NationalCertification Corporation for the Obstetric,Gynecologic and Neonatai Nursing Special-ties (NCC).^"'^' Contact with the AmericanBoard of Pediatrics/Stibboard of Neonatal-Perinatal Medicine resulted in a course out-line, but no list. The National Associationfor Neonatal Nurses (NANN), the Associationof Women s Health Obstetrics and Neona-tal Nurses (AWHONN), a National GuidelinesClearinghouse, and the Agency for HealthcareResearch and Quality were also contacted;however, no additional lists were identified.

132 JOURNAL OF NURSING CARE QUALTTY/APRIL-JUNE 2003

The 4 electronic databases included MEDLINE(1966 to December Week 4 2000), CINAHL(1982 to January 2001), HealthSTAR (1975 toDecember 2000), and EMBASE (all years).

Creating the source list from the 5 booklists and 4 databases (A'—166)

A source list of textbooks and practiceguidelines deemed relevant to the conceptof oral feeding in premature infants was de-veloped from the hook lists and databases.Since the 3 published lists were orgiuiizcdby subject headings, textbooks from the fol-lowing suhject headings were selected: dictio-naries, maternal-child nursing, obstetric andgynecologic nursing, pediatric nursing, gy-necology and obstetrics, nutrition, and pe-diatrics (A'^ 105). Inclusion criteria weretitles related to tlie keyw<jrds premature in-fant, infant nutrition, and feeding meth-ods. These keywords were chosen in anattempt to include all textbooks with somediscussion of oral feeding in prematureinfants. Examples were Mosby's Medical,Nursing, and Allied Health Dictionary', Com-prehensive Neonatal Nursing: A FfxysiologicPerspective, Whaley and Wong's Essentials ofFec/iatric Nursing, Taber's Cyclopedic Med-ical Dictionary, Clinical Nutrition: Enteraland Tube Feeding, and Rudolph's Fediatrics.Exclusion criteria were titles unrelated tothese keywords. Examples were Dictionaryof Medical Acronyms and Abbreviations,Maternity and Women's Health Care, Man-agement of the Infertile Woman, FediatricRehabilitation Nursing, Chnjtiic Pelvic Fain:An Integrated Approach, and ComprehensiveAdolescent Health Care^'^"-'^

The 2 unpublished lists had no subjectheadings. For this reason textbooks w ereselected from these lists by selecting ti-tles related to the keywords premature in-fant, infant nutrition, and feeding meth-ods (A^—58). Inclusion criteria were titlesrelated to these keywords. Examples wereBreast-feeding the Newborn: Clinical Strate-gies for Nurses, Pediatric Nutrition Hand-book, Manual of Neonatai Care, and CoreCurriculum for Neonatal Intensive Care. Ex-

clusion criteria were titles unrelated to thesekeywords. Examples were Sports Medicine:Health Care for Young Athletes, FediatricFrimar;)' Care, Fediatric and Adolescent Gy-necology, and Fhysical Assessment of theNewbom.^^'^ Practice guidelines were se-lected from the 4 electronic databases hy thefbliowing search strategy: keywords infantnutrition, feeding methods were comhinedvia the Boolean operator "OR."This result wascombined with the keyword, infant, prema-ture via the Boolean operator AND"; thenliniited to human and English language andpractice guidelines. An independent searchby the health sciences librarian was com-pleted to assess reliability. Three identical setsof practice guidelines were ohtained fromeach search (A' = 3).- "~ ^ The outcome oftheprocedure for the selection of textbooks andpractice guidelines was an initial source listof 47 Clnitial yield — 47 selections). After re-dundancy among lists was corrected, a sourcelist of 40 selections was compiled (Correctedyield =: 40 selections).

Assessing validity of the source list

The corrected source list was assessed forcontent validity by a multidisciplinary panelof clinical experts, This panel consisted of 1neonatologist and 2 NNPs. One NNP was afull-time clinician in a neonatal intensive careunit. The other NNP and the neonatologistheld clinical faculty positions at academic in-stitutions, with dual clinical and teaching re-sponsibilities, respectively. The procedure forassessing content validity included a letter, ex-plaining the purpose ofthe reference list, anda response rating form. Each expert was askedto respond to the fbliowing 3 questions:

1. Are there other sources you w^oiild in-clude? If yes, please list these sources.

2. Are there more recent editions ofthe ref-erences listed that you are aware of? If so,please list them.

3. Are there sources listed that could bedeleted from the list? If so, piease listthem and indicate why.

One expert suggested the addition of2 neonatal physiology textbooks and a

Building an Expert System 133

Table 1. The source list categorized by strata(A = 43)

Strata

MedicineNiirsiJigNutritionReferencePractice guideline

Number of sources

1714050403

neonatal-perinatal medical text of intantdiseases. These suggestions were added tothe source list and a tinal list consisting of 43selections was compiled (Final yield — 43).Finally, by using the Library of Congressclassification system, the source list of text-books and practice guidelines (A^=43) wascategorized by strata. First, the Library ofCongress subject headings for each sourcewere recorded. The following 4 categorieswere identilied: medicine, nursing, nutri-tion, and reference. Next, each source wasassigned to an appropriate category, creating4 strata. The 3 sets of clinical guidelinescomprised the fifth strata (see Table 1).

THE METHOD PHASE n

Scoring each textbook and practiceguideline from tlie source list

In order to determine whether the contentof eacb textbook and practice guideline wasrelevant to the concept of oral feeding in pre-mature infants, a source rating form was de-veloped. This form was similar to tbat usedby the NIC team when compiling their listof nursing interventions. The specific feedingterms of interest were tbe result of an ethno-graphic study conducted by Panniers.^ Tbeseincluded tbe following 4 oral feeding terms:nipple feeding, breast feeding, gavage feeding(bolus and continuous), and transpyloric feed-ing. Tbe eontent of each textbook and prac-tice guideline was reviewed for informationabout eacb specific feeding term as it relatedto premature infants. Tbe souree rating tbrm(See Table 2) was organized as follows:

1. Is there a reference to at least 1 of tbefeeding terms in this source?

2. Do at least I of the feeding terms in thissource refer to infants ot premature in-fants or both}

Table 2. The source rating form for textbooks and praetice guidelines

Title of source:Author: Edilion:Location of source:1. Refers to :it least 1 of the follow

Nippie feedingHreast feedingGavage feeding

Bolus

Tnuispylorie feeding2. At least 1 ot the feeding terms

Nipple feedingBrt-ast feedingGavage feeding

BolusContinuousInuispyloric feeding

3. Overall score:4. Comments:

Publieaiion data: (

ving feeding terms:V NI

V

Y

V

V

V

N

N

N

N[

N

refers to infants OT premature infants orI P

1

I

T

1

I

P

P

P

P

P

!:hapters/pages:

3oth:

#

#•ff

Note\ Y = yes, N — no, I = infants, P — premature infant, # = page *.

134 JOURNAL OF NURSING CARE QUAUTY/APRIL-JUNE 2003

Table 3. Scoring criteria for each textbook and practice guideline

Score Accept/reject Criteria for accept/reject

Reject No reference to any of the above feeding terms OR no discussion offeeding term as it relates to infants orpivmature infants or both

Accept One reference to one of the above feeding terms with a discussion ofhow the term relates to infants or premature infants or both

Accept Greater than 1 reference to any of the above feeding terms, but not allreferences accompanied by a discussion of how each term relates toinfants or premature infants or botb

Accept Greater than 1 reference to atiy of the above feeding terms with adiscussion of how each term relates to infants or premature infantsor both

After a souree rating form was completedfor each textbook or practice guideline, anoverall score was assigned based on tbe scor-ing criteria oittUned in Table 3. For exam-ple, if the source bad no referenee to any oftbe feeding terms or tbere was no discussionof how this feeditig term related to infantsor premature infants or botb, it received ascore of 1 and was rejected. Therefore, evenif the source listed all 4 terms, it was rejectedif there was no specific reference to infantsor premature infants or botb. If, however,the source included 1 reference to 1 of tbefeeding terms witb a discussion of how theterm related to infants ot premature infantsor botb, it reeeived a score of 2 and was ac-cepted. A source tbat included greater than1 reference to any of the feeding terms, butnot ail references were accompanied by a dis-cussion of bow eacb term related to infantsot premature infants or botb, was acceptedwith a score of 3 Finally, sources that re-ferred to greater than I reference to any of thefeeding tertns witb a discussion of bow eachterm related to infants ot premature infantsor both, received a score of 4 and were alsoaccepted.

Developing the coding systemA coding system was developed that ex-

tracted data trom eacb textbook and prac-

tice guideline about eaeh of 4 specific feedingterms. Walker and Avant's concept analysisand Weber scontent analysis were adapted."'• 'Modifying Walker atid Avant s procedure ofconcept anaiysis, 4 steps were undertaken: aconcept was selected, the purpose of the anal-ysis was determined, the defining attributeswere identified, and a model case was con-structed. Tlie concept to be analyzed was oralfeeditig in premature itifants. The purpose ofthe analysis was to determine a precise defini-tion for use in building an expert systetu, mod-eling decisions made by NNPs when initiatingoral feeding in premature infants cared for inthe NICU.'"' 1 he defining attributes were pro-vided by the participant observation phase ofPamiiers' recent ethnographie study.^ Tlieseinchided tbe 4 oml teediiig terms: nipple feed-ing, breast feeding, gavage feeding (bolus andcontinuous), and transpyloric feeding. Themodel ease was constructed by trial atid errorin an attempt to eticompass all aspects of theattributes of oral feeding in premature infants.Weber s discussion of creating and testing acoding scbeme was belptiil in refining this laststep. Fotir broad categories, designed to eti-sure precision and comprehensiveness in datacollection, were defined. These consistedof a basic definiticm, assessment, nursitigactivitiies/behaviors, and caregiver/prematureinfant dyad activities/behaviors. Entire para-graphs were coded and the coding scheme

Building an Expert System 135

Table 4. The coding system for each prema-ture infant feeding term found in eacb text-book or practice guideline (score > 2)

Premature itifant feeding term:1. Basic deftnition:2- Signs/indications for nursing intervention:3. Nursing activities/behaviors related to the

implementation of the intervention:4. Caregiver/premature infant dyad

activities/behaviors related to theimplementation of the intervention:

Comments:

(see Table 4) for eacb oral feeding term wasas follows:

1. Basie defitiition2. Signs/indications for nursing interven-

tion3. Nursing activities/behaviors related to

tbe implementation of the intervention4. CCaregiver/premature intant dyail activi-

tiesAiehaviors related to the implemen-tation of tbe intervention

THE PILOT STUDY

Results

A pilot study was tindertaken to test tbesource rating fonii and coditig system. Asource title was randomly selected from eachof the 5 strata (see Table 1). Tbe followingsources were piloted: Whaley and Wong's Es-sentials of Pediatric Nursing, Breastfeedingthe Newborn: Clinical Strategies for Nurses,

Williams Obstetrics, Taber's Cyclopedic Medi-cal Dictionary, atid Nutrient Needs and Feed-ing of Premature Infants. A source ratingform was used for each of the 5 sources andthe scoring criteria previoitsly described wereapplied to eacb source. Scoring results (seeTable 5) itidicated tbat the strata of nursingatid nutrition were rated the liighest (bothsources received a seore of 4), followed bymedicine (score of 3), reference (score of3), and practice guideline (score of 2). Dur-ing tbe initial stage of tbe pilot stttdy, an ad-ditional step was identified as necessary be-fore a source rating form could be completedfor each soitrce. Tbis additional step involveda metbod for ensuring tbat all informationabottt oral feeding in each source was iden-tified and copied. Initially, tbe table of con-tents of each souree was used; however, bytrial and error it was determined tbat eacbsource index served as a mure coniprehetisiveguide. Using eacb sotirce s index as a guide,tbe terms infant, premature infant, preterminfant, nutrition, and feeding were scatined.Appropriate pages were recorded and copied.In addition, the specific feeditig term and al-ternative labels were scatined. For example,w hen attempting to locate all the itifbrtna-tion in Whaley and Wong's Essejitials of Pe-diatric Nursing about the term nippie feed-ing, it was necessary to refer to tbe tbilowingadditional terms in the index: infant, prema-ture infant, preterm infant, nutrition, feed-ing, bottle feeding, formula feeding: feeding,nipple: feeding, formula; feeding, bottle. Ifthese additional terms were foiuid, appropri-ate page nttmbers were recorded and copied.Tbis additional step provided rigor by further

Table 5- Pilot study: Using the sotirce rating form to rate I source from eacb stratum

Category Source title Overall score

Nursinj;NutritionMedicineReferencePraetiee guideline

Whaiey and Wong's Essentials of Pediatric Ntirsing 4Breastfeeding tbe Newborn: Ciinicai Strategies for Nurses 4Wiiliams Obstetrics 5Taber's Cyclopedic Medical Dictionary 3Nutrient Needs and heeding of Premature Infants 2

136 JOURNAL OF NURSING CARE QUALITY/APRIL-JIINE 2003

Table 6. Pilot study: Using the coding system for the term nipple feeding

Strata

Nursing

Nutrition

Medicine

Reference

Pnictieeguide-line

Source

W}3aley andWong'sEssentials ofPediatricNursing^''

Breastfeedingthe Newhom:ClinicalStrategies forNurses''''

WilliamsObstetrics'''

TabersCyclopedicMedicalDictionary^"

Nutrient Needsand feedingof PrematureInfcints^^

Basicdefinition

Bottle feedingrefers to theuse of bottlesfor feedingmild formula orhuman milk

Bottles givesu pp lemciitarynourishment Eonewborns andconsist of 3basie styles:standard.preterm, &Nuk

No discussion ofterm

An artificialsubstitute for afemale nipple.used torbottle-feedinginfants

No discussion ofterm

Signs/indications

Normal vitalsigns; strongsuck;coordinationof suck andswallow; a gagreflex

No discussion

No discussion

Activities

Use an angledbottle; feed 6times a day;provide timeand patience

No discussion

No discussion

. . .

Dyadbehaviors

Hold infant closeto body withhack curvedslightly;supportcheeks/chin toenhance nipplecompression

No discussion

No discussion

ensuring that all potentially relevant informa-tion was in fact, collected.

In tbe next stage of tbe pilot study, theportion of each sotirce tbat had been iden-tified and copied for data extraction was re-viewed. The paragraph, the unit of print tobe eoded, was scanned tor inforniatioti ahoutnipple feeditig in premature inf;mts. Relevantinformation was underlined and recorded ac-cording to tbe 4 categories of basic defini-tion, signs/indicatiotis, nursing activities, anddyad bebaviors (see Table 6). Tbe nursingsottrce provided tbe most comprehensive re-sult by including the itidications for readinessto tiipple-feed, nursing interventions specificto the activity of nipple feeding, and behav-iors between caregiver and infatit. For exam-

ple, readiness to nipple-feed included the fol-lowing signs: vital signs within normal limits,presence of a gag reflex and strong stick, andtbe ability to coordinate sucking and swallow-ing. Nursing activities included the use of anangled bottle and feeding schedule. Dyad be-haviors addressed proper positioning of theinfant wbile feeding. It is also interesting tonote that the nursing, nutrition, and referencesources used the term bottle feeding in theirdiscussion of a basic definition.

DISCUSSION

A method of information analysis abouttlie coticept of oral feeding in premature

Building an Expert System 137

inlants bas been described. Wbile this is awork in progress, tbe pilot study has demon-strated that information extractioti tieeded tobuild expert systems is complex, but tbatthe method can be made explicit and canbe shared witb other researcbers and clini-cians. After completing tbe data extraetionfrom textbooks and practice guidelines, peer-reviewed journals will be analyzed.

Concept analysis is a strategy to determinetiie cbaracteristics and meaning of a concept.Tbe method presented in this article was de-signed as a particular type of concept analysisfor the purpose of standardizing nursing ter-

minologies for use in computerized systems.We believe this method is amenable to ex-plaining other clinical problems atid, as such,can add to tbe body of knowledge related tothe construction of expert systems.

ACKNOWLEDGME]\TS

Tbe National Library of Medicine sup-ported this work through NIH-NLM ROl-LM06967-01 Decision Support for tieonatalintensive care. Teresa L. Panniers, PriticipalInvestigator. We acknowledge tbe intellect ttalcontribution of Dr. Karen Soeken in tbis work.

REFERENCES

I I'annicrs TL. Levels of urgency as a criterion tor se-lecting an appropriate problem for decision suppC'tsystem development in the neonatal intensive careunit. In: Proceedings of the 7th ItUernatiojial Con-ference on Nursing Use of Computers and Informa-tion Sciejice. Auckland: New Zealand: 201)0.

2. Rinniers TL. Determining an appropriate clinicalpnibicni for expert system developmenl In theneonatal intensive care unit. .StudHealth Technol In-for?n. 1997:46:241-246.

3. Panniers TL. Refining clinical terminolog>' for expertsystem development: an application in the neona-tal intensive care unit. Pediatr Nurs. September-October 2()()2:28(5):519-523.129.

4. Benfcr RA. Hrent EE, Furbee L. Expert systems. New-bury- Park. CA: Sage; 1991.

5. Walker l-O, Avant KC. Strategies for Theory Con-struction in Nursing. 5riJ ed. Norwalk, Conn: Apple-Ion &Langc: 199S.

(•). Weber Rl* Basic Content Analysis. 2nd ed. London:Sage; 1990.

7. Bakken S, Casben MS, Mcndonca IA, O Brien A.Zieniewicz J. Representing nursing activities withina concept-orientctl terminological system: evalua-tion of a type definition.//!)» Med inform Assoc.2000;7(l):8l-90.

8. Bakken S, Hardiker NR, .Mead CN. Concept-orientedterminological .systems. In: Saba VK, McCormick KA,eds. Essentials of Computers for Nurses. 3rd cd.New York: McGraw-Hill; 2OOt;191-19S.

9. Werley HH, Ryan P. Zorn CR. The Nursing Mini-mum DaEa Set (NMDS): a framework tVir the orga-nization of nursing language. In: American NursesAssociation, ed. Nursing Data Systetns: TJje Emerg-ing Framework. New York: American Nurses As.so-ciation; 1995:19-27.

10. Panniers TL, Jacobs SK. Health care informatics. In:

Creasia JL, Parker li, eds. Conceptual foutidatiims:The Bridge to Professional Nursing Practice. 3rt.l ed.St. Louis: Mosby; 2001:46^-490.

11. Nortb American Nursing Diagnosis Association.Nursing Diagnoses: Definitions and Classification,1'J95-1996. Philadelphia: NANDA: 1994.

12. Mc<:ioskey JC, Docbtekman JM, Bukcbek CiM. Curs-ing liiterfentifms Classification (NIC). Philadel*phia: Mosby; 1992.

13. Johnson M, Ma;is M. Tlie Nursing Outcomes Classifi-cation../ Nurs Care Qual. 199S: 12(5):9-2O.

14. Saba VK. Home Health Care Classification (HHCC)tif Nursing Diagnoses and Interventions. Washing-ton: Georgetown University; 1994.

15. OzboltJG. From minimum data to maximum impact:using clinical data to strengthen patient care. AilrPract Nurs Q. 1996;l(4):62-69.

16. Kleinbeck SVM. Develtipment of the Perioperativenursing data set. Assoc Operating Room Nurses J.1999;7U(l):l=>-27.

17. Martin KS, NorrisJ. Tlie omaha system: applicationstor community health nursing. Holistic Nurs Pract.1996:1 l(l):75-83.

18. Spacktnan K.\, Campbell Kt:. Cote RA. SNOMED RT:a reference terminology tor health care. In: Pro-ceedings of the American Medical Informatics As-sociation Annual fall Symposium. Nash-ville: TN:1997:640-64-L

19. ()/-bo!t J, Abraham 11., Scbultz S. Nursing informa-tion systems. In: Shortliffe Ell, Perrcault LE, eds.Medical Infortnatics Compuier Applications inHealthcare. New York: Addison-Wesley; 1990:244-272.

20. Moorhead S, Clarke M, Willits M, Tomsha KA. Nurs-ing outcomes classification implementation projectsacross the care continuum. J Nurs Caiv Qual.1998;12(5);52-63.

138 JOURNAL OF NURSING CARE QUALITY/APRIL-JUNE 2003

21. Nursing Information & Data Set Evaluation Center.American Nurses Association. Available at: http://www.ana.or^nidsec/iiidt;x.htm. Retrieved October3,2001.

22. Ozbolt JG, Graves JR. Clinical nursing informaticsdeveloping tools for knowledge workers. Nurs ClinNorth Am. (93 A.D);28(2):407-42'>.

23. Ozboll JG, Russon M, Stultz MP Validity and rdiabil-it>' of standard terms and codes for patient care data.In. Proceedin}{s—the Annual Symposium on Com-puter Appticatiotis in Medical Care. New Orleans:Loui.siana; 1995:.37-4l.

24. Ozbolt JG. Fnichtnicht JN. Hayden JR. Toward data.standards for clinicai nursing information..//Im MedInform Assoc. 1994; 1(2):175-185.

25. Baud RH. I.ovis C, Rassinoux AM, Scherrer JR. Al-ternative ways for knowledge collection, indexingand robust language retrieval. Methods InformationMed. 1998:37:315-326.

16. Maas ML Johnson M. Moorhead S. Classifyingnursing-sensitive patient outcomes. IMAGE J NursScholarship. 1996;28(4):295-3O I.

27. Hill DR, Stickell HN. Brandon/Hill selected lisI olprint nursing books and journals. Nurs Outlook.2000:48Cl):10-22.

28. Hill DR, Stickell HN. Brandon/Hill selected list ofbotjks and journals for the small medical library. BullMed Libr Assoc. 1999:87(2): 145-169.

29. Allen M, Barry R, Dubois K et al. Interagency council

on information resources for nursing. Nms HealthCare Perspect. 2000; 21(5): 247-2 54.

30. Tlie National Certification Board of Pediatric NursePractitioners and Nurses. PNP Certification Re-source List. Gaithensburg: The National CertificationBoard (jf Pediatric Nurse Practitioners and NursesInc: 2001.

31. National Certification Corporation for the Obstet-ric, Gynecologic and Neonatal Nursing Specialities(NCO, 2001 Guide to NCC Certification: Neona-tal Nurse Practititmer Bibliography. Chicago: NC:C:2001.

32. Anonymous. Nutrient needs and feeding of prema-ture infants. Can Med AssocJ. 1995:152(11):1765-1785.

33. Anonymous. Professional: hypoglycaemia of thebreast fed newborn. Modem Midwife. 1997;7(1O):31-33.

34. Anonymous. Hypogiycaemia of the newborn. Mid-wives. 1997;110(1317): 248-249.

35. Wong DL, Whaley LF Whaley and Wong's Essentialsof Pediatric Nursing. 5th ed. St. lx)uis: Mosby: 1997.

36. Biancuzzo J. Breastfeeding tbe Newhom: ClinicalStrategies for Nurses. St. Louis: Mosby; 1999.

37. Cunningham FCi. Williams Obstetrics. 20th ed. NewYork: McGraw-Hill; 2(K)1.

38. Venes D, Thomas CL, eds., Taber's Cyclopedic Med-ical Dictionary, I9th ed. Philadelphia: FA. Davis:2001.