Assessing medication prescribing errors in pediatric intensive careunits*

Michael A. Cimino, RPh, MS; Mark S. Kirschbaum, RN, PhD; Linda Brodsky, MD;Steven H. Shaha, PhD, DBA; for the Child Health Accountability Initiative

I n 2000, the release of the Instituteof Medicine’s report on medicalerrors (1) galvanized the attentionof health care providers, regula-

tors, and policymakers alike. In 1998, theCommittee on Drugs and Committee onHospital Care of the American Academyof Pediatrics identified medication errorand safety as priorities for practitionerscaring for children (2). Also in 1998, theChild Health Accountability Initiative(CHAI) was formed as a 12-hospital col-laborative, and its first project used their

pediatric intensive care units (PICUs) tostudy medication errors in children.

Although adverse events commandour attention, understanding patterns ofmore ubiquitous “near miss” potential er-rors (3–8) may provide data to evaluatemeaningful system change. The ADE Pre-vention Study Group (3–5, 9), which hasadvanced the understanding of medica-tion safety with adults, reported 6.5 ad-verse drug events (ADEs) and 5.5 poten-tial ADEs per 100 nonobstetricadmissions. In intensive care unit (ICU)settings, the rate of preventable and po-tential ADEs was twice as high. However,when adjusted for the number of drugsordered, the rates were similar (5). Twen-ty-eight percent of all ADEs were judgedpreventable (3).

The pediatric population is at evengreater risk of adverse drug event occur-rences since dose, medication safety, andefficacy are difficult issues in young pa-tients (10–14). Pediatric medication er-ror rates have been identified in a small

number of studies in which error ratereports varied widely, although compari-son is limited due to methodological dif-ferences. A recent report, using an obser-vational approach in a tertiary children’shospital, identified a potential ADE ratethat was three times that of a previousadult hospital study (1.1% of 10,778 med-ication orders), whereas the preventableADE rate was similar (15).

As an initial step to develop both amethodology and a benchmark, CHAImembers identified medication prescrib-ing errors in the PICU as the priority forfocus. Physician members of CHAI advo-cated a concentration of efforts on theprescribing phase because they believethat a) prescribing errors were grosslyunderreported via spontaneous incidentreports; b) physicians have control at thisphase of the medication usage process; c)prescribing errors were likely to result inADEs; and d) lack of information aboutboth the drug and patient was a proximalcause of errors in published reports (3).

*See also p. 000.From Children’s Hospital of Buffalo/Kaleida Health

(MAC), Buffalo, NY; Child Health Corporation of Amer-ica (MSK), Overland Park, KS; State University of NewYork at Buffalo School of Medicine and BiomedicalSciences (MAC, LB); and Center for Pediatric Quality(MAC, LB, SHS), Women and Children’s Hospital ofBuffalo, Buffalo, NY

Copyright © 2004 by the Society of Critical CareMedicine and the World Federation of Pediatric Inten-sive and Critical Care Societies

DOI: 10.1097/01.PCC.0000112371.26138.E8

Objective: To evaluate a matrix for determining the predominanttype, cause category, and rate of medication prescribing errors, andto explore the effectiveness of hospital-based improvement initia-tives among pediatric intensive care units (PICUs).

Design: This study involved the prospective identification ofmedication errors for categorization and evaluation by using amatrix methodology. A pretest-posttest design without a controlgroup was used to explore the impact of initiatives employed toreduce medication error rates and severity.

Setting: PICUs in nine freestanding, collaborating tertiary carechildren’s hospitals that participated in both baseline and postint-ervention analyses.

Methods: We evaluated 12,026 PICU medication orders at baselineand 9,187 orders postintervention for prescribing errors, excluding re-suscitation orders. A standardized tool and process captured error type,cause category, and severity for 2 wks before and after intervention.Three levels of error detection were used and included pharmacy orderentry, PICU nurse order transcription, and team-based overview. Site-specific interventions were implemented, which included predominantlyprovider education as well as informational (47%) and dosing “assists”via preprinted orders, forcing functions, or prompts (39%).

Results: Of baseline orders, 11.1% had at least one pre-scribing error. The interception of prescribing errors improved30.9% (1.6% of all orders at baseline, 2.0% post intervention).Preventable adverse drug events were uncommon (0.6% of allmedication errors) and of low severity at baseline; most werewrong dose errors. The implementation of improvement initi-atives, specific for each facility, resulted in a 31.6% reductionin prescribing errors from 11.1% to 7.6%. However, site resultsvaried considerably.

Conclusions: A benchmark for medication prescribing errors inthe PICU was identified among nine children’s hospitals. Themethodology was successful in accounting for site-specific dif-ferences with regard to identifying and documenting errors aswell as reporting results of improvement initiatives. Furthermore,the methodology employed was generalizable in the identificationof predominant prescribing error types, which helped to trackindividual hospital improvement initiative development and im-plementation. Overall improvement in prescribing error rates wasnoted; however, considerable variation in the success of improve-ment initiatives was noted and bears further attention. (PediatrCrit Care Med 2004; 5:124–132)

124 Pediatr Crit Care Med 2004 Vol. 5, No. 2

Therefore, the purposes of this studywere as follows:

1. Establish a methodology, generaliz-able across a broad range of settings,for identifying, documenting, analyz-ing, and reporting prescribing errors.

2. Determine the overall rate of prescrib-ing errors (i.e., benchmark) amongparticipating pediatric hospital PICUs.

3. Report the relative effectiveness of avariety of hospital-specific, self-se-lected interventions in reducing med-ication errors and ADEs in PICUs.

METHODS

Collaborative Arrangements

The participating CHAI hospitals were ter-tiary centers, and all but one were freestand-ing children’s hospitals. PICUs ranged in sizefrom 6 to 24 beds. CHAI collaborative hospi-tals agreed on a set of standard definitions forthe project. Definitions of medication errors,adverse drug events (preventable and nonpre-ventable adverse drug reactions), actual andpotential errors, medication error cause cate-gory (e.g., prescribing), medication error type(e.g., incorrect dose), and adverse drug eventseverity score were standardized (see AppendixA). One site was designated to manage theentire project, acting as a resource regardingstandardization once the project was initiated.Each site designated its own site coordinatorto manage the project locally, and each siteobtained appropriate internal review board ap-proval per their local policies and procedures.A standardized medication error survey formwas adopted (see Appendix B) that requiredcompletion within 72 hrs of the medicationerror, thus capturing prescribing errors ascompletely as possible. Excluded from reviewwere uses of medications for resuscitation,which do not follow routine processes.

Standardized identification of medicationerrors used three levels of surveillance. Theseincluded a) the pharmacy order review forerrors and computer order entry step; b) thePICU nurse order transcription and review forerrors step; and c) an oversight team check.The anticipated variability among institutionsas to the effectiveness of pharmacy and nurs-ing checks for errors was overcome by theoversight team approach. The oversight teamsat each institution were similar in composi-tion and functioned with close collaborativesupervision. A member of the oversight teamfrom each site participated in the collaborativestandardization process involving the develop-ment of the standardized definitions, erroridentification, and reporting process for thisproject. The oversight team served as a finalstep in the project process to ensure as com-

plete and consistent identification and report-ing as possible, given the potential for vari-ability among participating hospitals inmedication error identification and reportingproficiency, hospital-specific medication errordefinitions, organizational cultures, predomi-nant errors, medication use systems, and pre-existing process changes designed to reducemedication errors. Efforts to support adher-ence with and understanding of the collabora-tive standardized process were enhancedthrough conference calls and communicationwith the project management site to trouble-shoot procedures and interpret events. Someminor variations remained in how sites staffedtheir oversight team and who served as sitecoordinator; however, as much standardiza-tion as possible was instituted to minimizesurveillance bias.

The pharmacy and nursing surveillancelevels used processes already available at eachsite to identify and then reconcile medicationerrors with the prescriber: The type of error,use of floor stock, and a brief narrative wererecorded. The oversight team was responsiblefor the review of all orders, correct identifica-tion of errors, and completion of the standard-ized survey form. Each error was classified asactual or potential. Adverse events were iden-tified and given a severity score. The oversightteam coordinated discussion of each error asneeded.

Data Collection

Procedures. A pretest-posttest design with-out control group was used. Two weeks ofpreintervention data collection was followedby 3 months of site-specific error reductioninterventions. A 2-wk postintervention datacollection then was completed. An order in-consistent with good medical practice in anyone or more of these steps was reconciled withthe prescriber and recorded on the medicationerror report form. Orders that were incom-plete with respect to date and/or time werestratified separately. A member of the medica-tion use oversight team collected medicationerror survey forms daily from the pharmacyand the PICU. Data were extracted from surveyforms and tallied in a summary matrix by asite coordinator (Appendix C). Only summarydata were submitted for inclusion into thecollaborative database by each CHAI hospitalto deal with these sensitive data. For cases ofan actual ADE, the patient was observed forresolution or up to 2 wks postevent to assessADE severity and outcome disposition.

Prescribing errors were categorized ac-cording to perceived potential to result in ad-verse events. Incomplete orders were consid-ered to have the lowest potential for adverseevents since an order with missing informa-tion required resolution before the ordercould be acted upon. Intercepted prescribingerrors that held significant potential to result

in harm to the patient (critical informationwas incorrect in the order) yet were inter-cepted before an opportunity for harm oc-curred were considered to have a higher po-tential for adverse events. Noninterceptedprescribing errors that held significant poten-tial to result in harm to the patient yet did notresult in an ADE were considered to hold evengreater potential for adverse events. Prevent-able ADEs were considered to be of the highestpotential for harm, since some adversity oc-curred, and thus were rated for severity level.

Devising Interventions to Reduce Error.Based on the findings of the baseline surveil-lance summary data, the collaborative met anddiscussed potential interventions at a 2-dayconference. This was facilitated by involve-ment of the Institute for Healthcare Improve-ment (16). The collaborative fostered sharedlearning; no efforts were undertaken to stan-dardize or assign interventions across membersites. Rather, sites used their typical internalperformance improvement approaches to de-ploy interventions expected to offer optimalresults. All but one site used a variety of ap-proaches simultaneously (Table 1).

Interventions implemented to reduce er-rors were categorized into three categories,including dosing assists, communication/educational, and floor stocks. Of all interven-tions reported, 47.2% were categorized ascommunication/educational, 38.9% as dosingassists, and 13.9% as floor stocks. Of the indi-vidual strategies reported, the most commoninvolved the systematic education of physi-cians (e.g., residents; 66.7% of organizations),availability of dosing references, guidelines or“cheat” sheets (55.6% of organizations), andthe implementation of forced-function ordersheets (44.4% of organizations). After imple-mentation of selected interventions, another2-wk surveillance period ensued.

Statistical Analysis

The duration of the data collection periodwas based on the necessary sample size re-quired at each hospital to achieve statisticalvalidity. After we conducted a power analysisusing assumptions of 80% power, an alphacoefficient of .05, a desire to detect a 5%reduction in errors, a minimum sample re-quired for detection of 863, and the baselinenumber of orders and errors, three high-volume sites targeted fewer orders than atbaseline.

The prescribing error rate (percent) wascalculated as the percent of errors relative tototal orders. The percent change in error ratewas determined as follows:

% errors postintervention� % errors and baseline

% errors at baseline� 100 [1]

An order may have contained more than onemedication error. The cumulative impact of

125Pediatr Crit Care Med 2004 Vol. 5, No. 2

improvement initiatives was calculated for allparticipating sites and reported as the meanpercent change. Prescribing error types andoutcome categories were grouped. Pre- andpostintervention comparisons were made ofchanges in medication error incidence acrosssites cumulatively.

We evaluated 12,026 medication ordersduring the baseline period and 9,187 ordersduring the postintervention period. Prescrib-ing errors identified were categorized accord-ing to the list of standardized definitions andgeneralizable reporting format. These errorswere summarized using the summary matrixform allowing for a common basis for discus-sion and rapid identification of predominanterror types.

RESULTS

Among the nine participating hospi-tals, a baseline rate of orders with errors

(excluding those missing only date and/ortime) was 11.1% compared with a rate of7.6% following implementation of hospi-tal-specific error reduction interventions(Z � 10.5, p � .001; Table 2). Thus, a31.6% reduction in orders with prescrib-ing medication errors was noted, repre-senting the cumulative impact of im-provement initiatives. The totalprescribing errors per order decreasedfrom 0.22 at baseline to 0.17 during thefollow-up period, a 24.7% reduction (Z �3.22, p � .05).

The rate of incomplete orders identifiedduring the baseline was 18.7% before imple-mentation of interventions and 13.8%postintervention, a reduction of 26.5%(Z � 13.92, p � .001; Table 3). The greatestreduction occurred for those “missing infor-mation” error types in which a medicationcould not be dispensed or administered un-

til the information was obtained, such asmissing drug, dose, route, dosage form, ordosage interval. The rate of intercepted pre-scribing errors at baseline was 1.6% com-pared with 2.0% during postinterventionperiod, a 30.8% increase with the greatestimpact occurring for wrong dose errors(Z � 4.37, p � .01). A nonintercepted errorrate of 2.0% was evident during the base-line period and 0.8% during the postinter-vention period, a 61.7% decrease with thegreatest reductions observed in the wrongdose, omissions, and wrong drug errortypes (Z � 9.71, p � .001).

Preventable ADEs were uncommon,with a rate of 0.13% of all orders duringthe baseline period and 0.03% during thepostintervention period (Z � 3.04, p �.05). The error type most associated withpreventable ADEs was the wrong dose.The severity level of ADEs was generally

Table 1. Interventions implemented by participating Child Health Accountability Initiative (CHAI) hospitals

126 Pediatr Crit Care Med 2004 Vol. 5, No. 2

low. Preventable ADEs during both ob-servation periods produced temporaryharm at most.

Considerable variability existed inbaseline medication error rates as well asin the relative reduction in error ratesfollowing implementation of the varioussite-specific interventions. Seven of thenine sites showed a relative reduction inerror rates, whereas two of the nine siteshad an increase in error rates (Table 4).

Several attempts were made to iden-tify factors that could account for thevariability in medication error rates be-tween sites during the baseline period.Using responses from critical care unitdescription questionnaires returned fromseven sites, the only factor that ac-counted for differences was the patient toprescriber ratio. Sites that reported a pa-tient to physician ratio of �4 had a meanmedication error rate of 12.3% comparedwith a mean rate of 37.5% for sites hav-ing a ratio �4 (Z � 11.30, p � .01).

DISCUSSION

ADEs (injuries, large or small, causedby the use, misuse, or underuse of a drug)have been the focus of considerable study(3–8, 17–22). ADEs are important be-cause they pose an immediate risk topatients and can be costly for the healthcare system, and some are considered tobe avoidable mistakes (3, 4, 9, 19–21, 23,24). The total injury rate is estimated tobe �1 million patients annually, andmedication use accounts for 19.4% of allinjuries suffered by hospitalized patients(18). Deaths from medication errors thattake place both in and outside of hospitalsexceed 7,000 annually (surpassing deathsfrom workplace injuries) (10).

The Premier Health Alliance with 25tertiary center members undertook ananalysis of �9,000 errors in 1 yr (25).Children’s hospitals within PremierHealth Alliance had an error rate of4.37 per 1,000, “eclipsing” the 1.97 per1,000 of all hospitals. Children �5 yrswere at highest risk for errors thatreached patients, and risk for infants�1 yr was more than double the nextrisk-prone group, aged 65–70. Mistakesin prescribing represented 42% of allerrors. In another study, error rateswithin critical care units of a 160-bedCanadian children’s hospital were 7.1%in the intensive care nursery and 11.7%in the PICU excluding wrong-time er-rors (26). These two reports differ dra-matically from rates reported in a study

of two California hospitals, where theerror rates were 4.9 and 4.5 per 1,000medication orders (12).

Neither a measurement standard nor aconsistent benchmark for errors yet ex-ists. This situation is substantiated byrecommendations advanced through var-ious sources for reducing medication er-rors in pediatric settings, including in-creased involvement of pharmacists andpharmacy satellites, especially in the crit-ical care areas (12, 24, 26–28). Clearly,error rates need to be addressed and con-sistent methods of tabulation and bench-marking adopted.

This study a) demonstrated the useful-ness of a generalizable method used toidentify, document, and analyze prescrib-ing errors across a broad range of pediatric

hospital PICUs; b) identified an overall rateof prescribing errors among participatingpediatric hospital PICUs categorized ac-cording to potential for causing harm(benchmark); and c) described the effec-tiveness of a variety of hospital-specific,self-selected interventions in reducingmedication errors and ADEs. The three-tiered methodology for the identificationand documentation of prescribing errors,the matrix format for categorizing errors asto cause and type (promoting the rapididentification of predominant errors), andthe use of severity scoring of preventableADEs were refined. Although standardizedamong facilities, the approach was flexibleenough to be used among collaborativesites despite intersite differences in medi-cation use systems, fostering a common

Table 2. Summary impact of intervention strategies for Child Health Accountability collaborative

Preintervention PostinterventionPercentChange

Total medication orders, n 12,026 9,187Orders with errors, % 27.1 23.7 �12.6a

Time and date only, % 16.0 16.1 0.6All other errors, % 11.1 7.6 �31.6a

Orders without errors, % 72.91 76.3 4.7a

100.0 100.0Error rates

Per order, n 0.22 0.17 �24.7a

Per order with errors, n 2.0 2.2 10.0

aStatistically significant difference between pre- and postintervention (p � .01).

Table 3. Percent of orders with prescribing errors categorized by potential to result in an adverse drugevent (ADE)

Prescribing Error Category Preintervention PostinterventionPercentChange

Number of orders 12,026 9,187Low ADE potential (missing information), % 18.7 13.8 �26.5a

Intercepted with potential for ADEs, % 1.6 2.0 30.8a

Nonintercepted with potential for ADEs, % 2.0 0.8 �61.7a

Adverse drug events, % 0.13 0.03 �76.9a

aStatistically significant difference between pre- and postintervention (p � .01).

Table 4. Baseline and postintervention with relative reductions in medication error rates by site

HospitalSite

Baseline ErrorRate

PostinterventionError Rate

RelativeReduction

A 2 2 35B 5 4 �24C 19 7 �61D 23 19 �16E 33 17 �50F 37 11 �71G 42 58 38H 21 19 �8I 17 15 �11

All values are percentages.

127Pediatr Crit Care Med 2004 Vol. 5, No. 2

approach and set of definitions for address-ing the problem of medication errors.These errors could be summarized usingthe summary matrix form, allowing for acommon basis for discussion and rapididentification of predominant error types.Use of this methodology in the context ofthe collaborative precipitated improvementinitiatives, the results of which were mea-surable by the same generalizable method-ology. Accomplishment of this aspect of theproject was a major goal of the collabora-tive given the variability among sites.

In our study, a baseline for medicationprescribing errors in a PICU setting wasestablished from these nine children’shospitals nationally. When date/time er-rors were excluded from review, aggre-gate rates of prescribing errors per orderfor these PICUs (11.1% preintervention)were more similar to reports by Tisdale(26) than the extremely low rates of oc-currence reported among California chil-dren’s hospitals (12). Although boththese prior reports are dated, they repre-sent the only published pediatric data.Recognizing the existence of site-specificdifferences, we observed an overall 24.7%decrease in prescribing error rate in ourstudy. Considerable variation in reducingerror rates among participating hospitalswas found. Some hospitals reported dra-matic reduction in error rates, whereasothers did not anticipate such changesbecause a low medication error rate wasevident during the baseline period. Errorrates in two hospitals increased (A and G,Table 4), a finding that cannot be ex-plained, other than the possibility thatparticular interventions used at those fa-cilities did not address the underlyingissues responsible for the medication er-ror causes and types most frequently en-countered at these sites. However, overallreductions in the rate of orders with pre-scribing errors and the prescribing errorrate were noted in the majority of partic-ipating sites and for the hospitals collec-

tively. A positive impact in prescribingerror rates occurred in the categories ofincomplete orders, nonintercepted pre-scribing errors with ADE potential, andpreventable ADEs, suggesting some suc-cess for the improvement initiatives im-plemented at the various sites for reduc-ing medication errors in general. Anincrease in error rate occurred in thecategory of intercepted prescribing errorswith ADE potential, suggesting a positiveimpact of improvement initiatives andproject methodology on interceptingADEs using an oversight team. Resultsfrom this study indicate that the method-ology followed represents a readily adapt-able and generalizable approach to col-lecting comprehensive data regardingmedication-related errors in the PICUsetting. As a result of this process, themajority of facilities participating experi-enced statistically significant reductionsin errors. Although the generalizability ofthe methodology was not evaluated inthis research, organizations seeking tobetter understand and remediate medica-tion-related errors in settings other thanPICUs could readily leverage the method-ology and tools derived. Another impor-tant conclusion is that this kind of ap-proach can yield improvements inmedication errors without the implemen-tation of expensive computerization.Even though it is generally believed thatcomputerization (e.g., computer-basedphysician order entry) represents the de-sired long-term objective for controllingerror rates, this approach is beyond theshort-term capabilities of many healthcare organizations.

ACKNOWLEDGMENTS

The following hospitals participated inthis project:

Children’s Hospital of Buffalo, Buffalo, NYCamcare Health Systems and the Womenand Children’s Hospital, Charleston, WVChildren’s Hospital Medical Center,Cincinnati, OHCook Children’s Medical Center, Fort WorthArkansas Children’s Hospital, Little Rock,ARChildren’s Hospital of Los Angeles, LosAngeles, CALe Bonheur Children’s Medical Center,Memphis, TNChildren’s Hospital of Wisconsin, Milwaukee,WIChildren’s Hospital of the King’s Daugh-ter’s Health System, Norfolk, VALucile Packard Children’s Health Services,Palo Alto, CA

Children’s Hospital and Health Center,San Diego, CAChildren’s National Medical Center, WA,DC

REFERENCES

1. Kohn L, Corrigan J, Donaldson M (Eds). ToErr Is Human: Building a Safer Health Sys-tem. Washington, DC, National AcademyPress, 2000

2. Prevention of medication errors in the pediatricinpatient setting. American Academy of Pedia-trics. Committee on Drugs and Committee onHospital Care. Pediatrics 1998; 102:428–430

3. Bates DW, Boyle DL, VanderVliet MB, et al:Relationship between medication errors andadverse drug events. J Gen Intern Med 1995;10:199–205

4. Bates DW, Cullen DJ, Laird N, et al: Inci-dence of adverse drug events and potentialadverse drug events. Implications for preven-tion. JAMA 1995; 274:29–34

5. Cullen DJ, Sweitzer BJ, Bates DW, et al:Preventable adverse drug events in hospital-ized patients: A comparative study of inten-sive care and general care units. Crit CareMed 1997; 25:1289–1297

6. Lesar TS, Briceland L, Stein DS: Factors re-lated to errors in medication prescribing.JAMA 1997; 277:312–317

7. Lesar TS, Lomaestro BM, Pohl H: Medica-tion-prescribing errors in a teaching hospi-tal. A 9-year experience. Arch Intern Med1997; 157:1569–1576

8. Hunt ML Jr, Rapp RP: Intravenous medica-tion errors. J Intraven Nurs 1996; 19:S9–S15

9. Leape LL, Bates DW, Cullen DJ, et al: Systemsanalysis of adverse drug events. ADE Preven-tion Study Group. JAMA 1995; 274:35–43

10. Koren G, Haslam RH: Pediatric medication er-rors: Predicting and preventing tenfold disas-ters. J Clin Pharmacol 1994; 34:1043–1045

11. Anderson BJ, Ellis JF: Common errors ordrug administration in infants: Costs andavoidance. Pediatr Drugs 1999; 1:93–107

12. Folli HL, Poole RL, Benitz WE, et al: Medi-cation error prevention by clinical pharma-cists in two children’s hospitals. Pediatrics1987; 79:718–722

13. Draft guidelines for preventing medicationerrors in pediatrics. J Pediatr Pharm Pract1998; 3:189–202

14. Meyer GE, Novielli KA, Smith JE: Use of refrac-tive index measurement for quality assuranceof pediatric parenteral nutrient solutions. Am JHosp Pharm 1987; 44:1617–1620

15. Kaushal R, Bates DW, Landrigan C, et al: Med-ication errors and adverse drug events in pedi-atric inpatients. JAMA 2001; 285:2114–2120

16. Leape L, Kabcenell A, Berwick D, et al:Breakthrough Series Guide: Reducing Ad-verse Drug Events. Boston, MA, Institute forHealthcare Improvement, 1998

17. Leape LL: A systems analysis approach to medi-cal error. J Eval Clin Pract 1997; 3:213–222

18. Leape LL, Brennan TA, Laird N, et al: Thenature of adverse events in hospitalized pa-

A benchmark for

medication pre-

scribing errors in

the pediatric intensive care

unit was identified among

nine children’s hospitals.

128 Pediatr Crit Care Med 2004 Vol. 5, No. 2

tients. Results of the Harvard Medical PracticeStudy II. N Engl J Med 1991; 324:377–384

19. Johnson JA, Bootman LJ: Drug-related mor-bidity and mortality. Arch Intern Med 1995;155:1949–1956

20. Bates DW, Spell N, Cullen DJ, et al: The costsof adverse drug events in hospitalized pa-tients. Adverse Drug Events PreventionStudy Group. JAMA 1997; 277:307–311

21. Classen DC, Pestotnik SL, Evans RS, et al:Adverse drug events in hospitalized patients.JAMA 1997; 277:301–305

22. Tinkle R, Wu JK: Errors involving pediatricpatients receiving chemotherapy: A litera-ture review. Med Pediatr Pharmacol 1996;26:344–351

23. PMES Committee: Medication Error Study.Rockville, MD, Physician Insurers Associa-tion of America, 1993

24. Prevention of medication errors in the pediat-ric setting. Pediatrics 1998; 102:428–430

25. Study shows big picture on adverse drugevents. Modern Healthcare July 24, 1995; 76

26. Tisdale JE: Justifying a pediatric critical-care satellite pharmacy by medication-error reporting. Am J Hosp Pharm 1986;43:368 –371

27. Leape LL, Cullen DJ, Clapp MD, et al: Phar-macist participation on physician rounds andadverse drug events in the intensive careunit. JAMA 1999; 282:267–270

28. Levine SR: Guidelines for preventing medi-cation errors in pediatrics. J Pediatr Phar-macol Ther 2001; 6:426–442

APPENDIX A: DEFINITIONS

Medication Errors and AdverseDrug Events (ADEs)

Medication Error. This includes anyerror, large or small, at any point in themedication system from the time thedrug is ordered until the patient receivesit. This differs from some prior defini-tions that considered only deviationsfrom the physician’s order as errors.Since nearly half of all errors occur in theprescribing stage, these must be includedin the definition. A medication error mayor may not result in an ADE.

Adverse Drug Event. An ADE is aninjury, large or small, caused by the use(including nonuse) of a drug. It can be asharmless as a drug rash or as serious asdeath from an overdose. There are twotypes of ADEs: those caused by errors andthose that occur despite proper usage. Ifan ADE is caused by an error it is, bydefinition, preventable. NonpreventableADEs (injury, but no error) are calledadverse drug reactions (ADRs).

Preventable Adverse Drug Event. Apreventable ADR is an injury due to anerror in use of a drug (including failure to

use). These would include an actual med-ication error (actual prescribing error) ifassociated with an adverse outcome.

Adverse Drug Reaction. This is definedby the World Health Organization tocharacterize injuries caused when drugsare used in the usual accepted fashion. Bydefinition then, an ADR does not resultfrom an error. Unfortunately, many haveused this term as synonymous with ADE,which blurs an important distinction.

Potential Adverse Drug Event (PADE).A PADE is a serious medication error—one that has the potential to cause anADE but did not. The PADEs are classifiedas nonintercepted, those ADEs with po-tential to harm but did not (e.g., thepatient was not allergic to the drug de-spite a note in the record stating so), orintercepted (e.g., the nurse recognized anorder for a medication to which the pa-tient was allergic and called the physicianto get it changed). Examining PADEshelps to identify both where the system isfailing (the error) and where it is working(the interception).

Nonintercepted PADE. This is a PADEwith no clinical consequence. A noninter-cepted PADE would include a potentialmedication error (potential prescribingerror) if not associated with an adverseoutcome.

Intercepted PADE. This is a medicationerror that never reached the patient since itwas prevented by existing mechanism. Thisincludes errors that are reconciled, that is,intercepted potential medication errors (in-tercepted prescribing errors).

Medication Error Cause

Prescribing Error. A prescribing erroris an incorrect drug selection (based onindications, contraindications, known al-lergies, existing drug therapy, and otherfactors), dose, dosage form, quantity,route, concentration, rate of administra-tion, or instructions for use of a drugproduct ordered or authorized by a phy-sician (or other legitimate prescriber); il-legible prescriptions or medications or-ders that lead to errors that reach thepatient; or use of nonstandard nomencla-ture or abbreviations.

Other. Other causes of errors such asthose produced by drug administration,transcription, or dispensing errors werenot included because they are beyond thescope of prescribing.

Medication Error Type

Although many error types may not beapplicable to medication errors due toprescribing causes, they are presented forcompleteness.

Omissions. The patient did not receive agiven dose by the time the next dose is due.

Wrong Patient. The medication wasgiven to a patient for whom the physiciandid not write an order.

Wrong Drug. A drug was given to theright patient outside a stated set of clinicalguidelines or protocols (may refer to site-specific treatment protocols, guidelines, orcare maps). Duplication of therapy will beconsidered inappropriate if inconsistentwith the local standard of care.

NumericScale

LetterScale Description

0 No error occurred0.5 A Capacity to cause error, but no error occurred0.5 B Potential error1 C Error occurred without harm to the patient2 D Error occurred requiring increased patient monitoring, but no harm to

patient or change in vital signs3 Error occurred requiring increased patient monitoring and laboratory tests;

there was a change in vital signs, but ultimately no harm3.5 E Error occurred resulting in the need for treatment or intervention and

caused temporary harm4 F Error occurred and resulted in need for treatment with another drug,

increased LOS, patient transfer to a higher level of care (eg, ICU) orrequired intervention to prevent permanent impairment or damage

5 G Error occurred and resulted in permanent patient harm5.5 H Error occurred and resulted in a near death event (eg, anaphylaxis, cardiac

arrest)

Numeric Scale Letter Scale Description

6 I Error occurred and resulted in patient death

129Pediatr Crit Care Med 2004 Vol. 5, No. 2

Wrong Dose. Dose administered dif-fers from dose ordered or dose ordered isincorrect based on a stated set of clinicalguidelines or protocols. These would in-clude the wrong rate of administration(medication is administered at a rateabove or below that which is appropriatefor the medication; this error categoryapplies especially to intravenous dripsand infusions). Minor rounding of dosesto fit standardized dosages for certainmedications will be exempt.

Wrong Dosage Interval. Dosage intervaladministered differs from interval orderedor interval ordered is incorrect based on astated set of clinical guidelines or protocols.

Wrong Dosage Form. Dosage form ad-ministered differs from dosage form or-dered or dosage form ordered is incorrectbased on a stated set of clinical guidelinesor protocols.

Excluded would be accepted protocols(established by the Pharmacy and Thera-peutics Committee or its equivalent) thatauthorize pharmacists to dispense alter-nate dosage forms for patients with spe-cial needs (e.g., liquid formulations forpatients with nasogastric tubes or thosewho have difficulty swallowing), as al-lowed by state regulation.

Wrong Duration of Therapy. Durationof therapy ordered is incorrect based on astated set of clinical guidelines or protocols.

Wrong Route. Drug administered byroute not ordered or route ordered isincorrect based on a stated set of clinicalguidelines or protocols.

Wrong Time. Medication is adminis-tered outside a predefined time intervalfrom its scheduled administration time(e.g., 30 mins). This interval must be estab-lished by each individual health care facility.

Unauthorized Dose. The patient receivesan additional dose or doses after the or-dered medication has been discontinued.

Monitoring. There is a failure to re-view a prescribed regimen for appropri-ateness and detection of problems, or fail-ure to use appropriate clinical orlaboratory data for adequate assessmentof patient response to prescribed therapy.

Drug-Drug Interactions. A drug-druginteraction exists for the medication pre-scribed. Only clinically relevant interac-tions will be considered here (any theo-retical or nonmeasurable/clinicallyinsignificant interaction will be exempt).

Drug-Food Interactions. A drug-foodinteraction exists for the medication pre-scribed.

Compatibilities. A drug incompatibility

exists in the manner in which the medica-tion is prescribed or the manner in whichthe medication is prepared or administered.

Laboratory and Drug Levels. An errorexists when pertinent laboratory includ-ing drug level information is either notordered or not appropriately consideredin the drug-ordering process.

Allergy and Other Clinical Information.An error exists when pertinent allergy andother clinical information is not appropriatelyconsidered in the drug-ordering process.

Procedure Error. There is a breach ofstandard procedure resulting in a medi-cation error (e.g., patients transferredfrom recovery room to PICU with unla-belled intravenous medications resultingin an order for the wrong dose or infusionrate).

Completeness of Prescription. Missinginformation exists such as patient name,identification number, date and time oforder, drug, dose, route, frequency, andprescriber signature.

ADE Severity Level and Outcome.A modified numeric scale will be used forthisproject(subcategorizedusing0.5incre-ments reflective of letter scale differ-ences). The letter scale is provided forreference.

130 Pediatr Crit Care Med 2004 Vol. 5, No. 2

APPENDIX B

131Pediatr Crit Care Med 2004 Vol. 5, No. 2

132 Pediatr Crit Care Med 2004 Vol. 5, No. 2