eagvasfifirb

Expert Reviews www.AJOG.org

PATIENT SAFETY SERIES

A systematic approach to the identification and classificationof near-miss events on labor and deliveryin a large, national health care system

Steven L. Clark, MD; Janet A. Meyers, RN; Donna R. Frye, RN; Kathryn McManus, MBA; Jonathan B. Perlin, MD

Avoidance of medical errors is akey component of quality medical

care.1-3 However, because the completelimination of medical error is notchievable, effective patient safety pro-rams seek to construct barriers that pre-ent errors from reaching the patientnd causing harm.1-4 The creation ofuch barriers is facilitated by the identi-cation of near-misses or close calls, de-ned as unplanned events caused by er-or that do not result in patient injuryut have the potential to do so.5,6 Such

near-miss events are, in most systems,many times more common than injuryevents, yet are generally much more dif-ficult to identify and quantify.1

We report our experience with a sys-tem of near-miss identification and anal-ysis on labor and delivery units in thelargest obstetric health care delivery sys-tem in the United States. This system hasallowed us to identify and prioritize ef-forts to reduce patient harm, and repre-sents the first reported comprehensiveanalysis of near-miss events in obstetrics.

Materials and methodsThe Hospital Corporation of America(HCA) has an annual delivery volume ofover 200,000 deliveries in 104 hospitalsin 21 states. Such facilities range fromsmall, rural hospitals to major academicmedical centers and are reasonably rep-resentative of the overall US health caresystem.7-9 In 2008, HCA developed and

From the Hospital Corporation of America,Nashville, TN.

Received May 25, 2012; revised Aug. 17,2012; accepted Sept. 12, 2012.

The authors report no conflict of interest.

Reprints are not available from the authors.

0002-9378/$36.00© 2012 Mosby, Inc. All rights reserved.

http://dx.doi.org/10.1016/j.ajog.2012.09.011

deployed an electronic voluntary report-ing system aggregating deidentified closecall and event reports into a centralrepository called the Patient Safety Im-provement Program (PSIP) databasethat is among the largest event reposito-ries in the United States.10 Using a stan-dard taxonomy derived from existingexternal and internal programs and nar-rative reports, this reporting system be-came a resource for collective learningand identification of high-risk processesthat have the potential to cause patientharm. This program encourages all pro-viders to report both unplanned events,(defined as any unplanned occurrence,which is undesirable, untoward, or un-anticipated) and near-misses, (definedas unplanned events that have the poten-tial to cause harm, but do not do so). Theclear overlap in these definitions is pur-poseful, and directed at the identifica-tion of all events that could, under closerscrutiny, possibly represent actual or po-tential harm to a patient. Although thisis a nonpunitive system that allows thereporting of near-miss events withoutfear of individual sanction or disciplin-ary measures, every effort is made to ad-

We describe a systematic approach to theevents on labor and delivery in a large, natnear-miss events were prospectively collecreports were analyzed according to frequewere reported in 0.69% of deliveries. Medicmost common near-miss events. Howeveeffective in preventing such errors from rpotential for causing harm involved physiciless effective existing barriers between thidentified. Use of a comprehensive systemand delivery units have proven useful in alloof greatest need.

Key words: mediation error, near-miss, pa

dress contributing factors related to hu-

DECEMBER 2012 Am

man error and system flaws identifiedthrough in-depth report analysis.

After collection of all reported eventsfrom labor and delivery units during 2010,a random sample was selected for analysisusing Cochran’s sample size formula setfor a 5% maximum error.10,11 All reportswere initially reviewed and preliminarilycategorized as to the potential for patientharm by a team consisting of nursing andadministrative personnel. Each report wasthen additionally reviewed and analyzedby the principle physician author. The se-lected reports were divided into 3 groups:unplanned events not related to medicalerror, medical errors resulting in tempo-rary or permanent patient harm, and near-misses, as defined above. The near-missgroup was then subjected to detailed anal-ysis and is the subject of this report.

Each near-miss event was categorizedas to absolute and relative frequencies asoutlined in the Table. Events were thengiven a score for each of several definedparameters in a modification of a previ-ously described system known as failuremodes and effects analysis.12,13 In eachcase, a low score is desirable and a highscore undesirable. These parameters

ntification and classification of near-missl health care system. Voluntary reports ofduring 2010 in 203,708 deliveries. Theseand potential severity. Near-miss events

n and patient identification errors were thexisting barriers were found to be highly

hing the patient. Errors with the greatestesponse and decision making. Fewer anderrors and potential patient harm were

identification of near-miss events on laborg us to focus patient safety efforts on areas

t safety

ideionatedncyatior, eeacan reseforwin

tien

were defined as follows:

erican Journal of Obstetrics & Gynecology 441

oJSpnpoteemmd

rarres

s on l

Expert Reviews Patient Safety Series www.AJOG.org

A. Worst possible outcome. Each eventwas given a score from 1 (negligiblechance of harm had the eventreached the patient) to 5 (high likeli-hood of serious harm had the eventreached the patient.) An example ofthe former would be a charting errorindicating that 1 g of cefazolin hadbeen given for cesarean prophylaxiswhen in fact 2 g had been given. Anexample of the latter would be the or-dering of penicillin in a patient witha history of penicillin anaphylaxis,which was caught by standing nurs-ing and pharmacy protocols mandat-ing referral to allergy history beforeadministration of any drug.

B. Method of detection. Each event wasgiven a score of 1 (error detected byexisting monitoring systems) or 2(error detected fortuitously).

C. Number of barriers in place. Eachevent was given a score of 1 (singlebarrier in place aimed at preventingerror from causing patient harm) or2 (more than 1 barrier in place pre-venting error from causing harm).

D. Quality of barriers in place. Each

TABLENear-miss events in labor and deli

Error type

Medication error...................................................................................................................

Patient identification...................................................................................................................

Failure/delay in obtaining laboratory...................................................................................................................

Failure to respond...................................................................................................................

Failure to follow policy/protocol...................................................................................................................

Charting error...................................................................................................................

Equipment failure...................................................................................................................

Slip/fall...................................................................................................................

Information transfer...................................................................................................................

Physician performance error...................................................................................................................

Incorrect test ordered...................................................................................................................

Inadequate staffing...................................................................................................................

Error in test interpretation...................................................................................................................

Wrong procedure ordered...................................................................................................................

Laboratory error...................................................................................................................

Retained foreign body...................................................................................................................

Clark. Identification and classification of near-miss event

event was given a score of 1 (high

442 American Journal of Obstetrics & Gynecology

quality barrier[s]) in place or 2 (lowquality barrier[s]) in place. This wasbased on our knowledge of our inter-nal error-prevention systems, as wellas our observation of reported actualharm events for the various catego-ries of errors in the broader database.

After the initial scoring of each event, amean score for events in each categorywas calculated (termed A, B, C, and D)and used to create a new parameter foreach category of event, the hazard score(A � B � C � D) (Table) (individualscores for each near-miss event categoryavailable on request).

This analysis included only deidenti-fied data examined for purposes of qual-ity improvement and was thus exemptfrom institutional review board reviewbased on 45CFR46.101(b)2 and 46.102(f) and 45CFR164.514(a)-(c) of theHealth Insurance Portability and Ac-countability Act.

ResultsDuring 2010, 3959 unplanned eventswere reported on labor and delivery of203,708 total deliveries (1.9%). More

y

No. (%) Hazard score

120 (33) 3..................................................................................................................

95 (19) 2..................................................................................................................

57 (11) 6..................................................................................................................

49 (10) 15..................................................................................................................

46 (9) 9..................................................................................................................

31 (6) 2..................................................................................................................

28 (6) 5..................................................................................................................

27 (5) 11..................................................................................................................

23 (5) 10..................................................................................................................

14 (3) 17..................................................................................................................

7 (1) 1..................................................................................................................

4 (�1) 12..................................................................................................................

2 (�1) 3..................................................................................................................

3 (�1) 4..................................................................................................................

3 (�1) 1..................................................................................................................

3 (�1) 5..................................................................................................................

abor and delivery. Am J Obstet Gynecol 2012.

than 95% of reports originated from

DECEMBER 2012

nursing or laboratory staff. Using themethodology described above, 1248events were randomly selected for anal-ysis. Seven hundred eighty-five events ei-ther did not involve medical error (forexample, precipitous delivery not at-tended by a physician or certified nursemidwife within a few minutes of patientarrival in the hospital) or was a medicalerror resulting in patient harm (for ex-ample, fetal scalp injury resulting frommisuse of forceps). Four hundred sixty-three reports involved near-miss eventsas defined above, and were analyzed indetail. When extrapolated to the entiredelivery population during 2010, near-miss events were reported in 0.69% ofdeliveries.

The Table demonstrates the types andfrequency of near-miss events on laborand delivery as well as the compositehazard score for each type of event.

CommentThis report is unique and represents thefirst systematic analysis of true near-missevents in obstetrics. In this analysis, wehave used the definition of near-missevents developed by Reason4 and ad-

pted by the Institute of Medicine, theoint Commission, and the Institute forafe Medication Practices, namely: un-lanned events caused by error that doot result in patient injury but have theotential to do so.1,5,6 This differentiatesur data from existing series that use theerm “near miss” to refer to a general cat-gory of patients who experience adversevents, or nonfatal critical illness thatay or may not involve error.14-16 Near-iss events were reported in 0.69% of

eliveries in this series.Near-misses are not as dramatic as er-

ors that result in patient harm and thusre less likely to provoke spontaneousesponses aimed at the prevention ofecurrence than are actual harmfulvents.1 However, as described by Rea-on,4 medical errors generally result

in patient harm only when (1), barriersto patient harm involve weaknesses,or “holes” and (2), the “holes” fortu-itously align, allowing potentially harm-ful events to result in actual harm. Be-cause such an alignment of weaknesses is

ver

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

uncommon, systems change introduced

www.AJOG.org Patient Safety Series Expert Reviews

only as a result of actual harmful eventswill fail to identify common and impor-tant systems defects, leaving patients stillvulnerable to harm.1 In addition, institu-tions often react to occasional harmfulevents by instituting disruptive work-flow changes in a reactive effort to dosomething in response to a tragic but rareadverse event. Although such changesare sometimes necessary, a systematic fo-cus on near-miss events allows leaders tobe proactive, focusing their patient safetyefforts on those events likely to cause re-petitive harm in the future, rather thanreactive, responding only to events thathave already occurred and may not hap-pen again in decades. The Institute ofMedicine encourages the developmentof voluntary reporting systems for near-miss events and suggests that mandatoryreporting systems focus on errors thatactually result in serious patient harm ordeath.1

An analysis of the Table suggests thatnear misses may be divided into 4 groupsbased on frequency and potential hazardscore.

High frequency-low hazard eventsSuch events represent areas of frequenterror in which effective safety barriersare already in place preventing such er-rors from reaching the patient and caus-ing harm. Examples in our series aremedication errors and patient identifica-tion and labeling error (Table). In oursystem, the use of the electronic medicaladministration record (EMAR) and pro-tocols requiring multiple checks of pa-tient identification bands before any in-tervention appear to be highly effectivein preventing these relatively frequentevents from reaching the patient andcausing harm.

High frequency-high hazard eventsThese events represent areas of relativelyfrequent error in which effective barriersto prevent the error from reaching thepatient are currently suboptimal. Failureof a provider to respond or failure of aprovider to follow protocol are the 2most prominent examples of this type oferror (Table). In our series, the vast ma-

jority of such errors involved physicians

on independent medical staffs at ourfacilities.

Low frequency-high hazard eventsOn a labor and delivery service, physi-cian performance error in practices orprocedures not governed by hospitalprotocol appears to be the most com-mon example of this error type (Table).Such errors generally involve an uncom-mon but potentially hazardous physi-cian-specific action for which few effec-tive barriers exist. Barriers to harm thatdo exist primarily involve nurse-physi-cian confrontation and accessing thechain of command. Ongoing educa-tional programs, local peer review andinsistence on recertification and con-tinuing medical education may all re-duce the frequency of these already low-frequency events, but do little withrespect to the construction of true barri-ers to harm in the presence of a physi-cian-mediated medical error. Barringimposition of a pilot/copilot model forpatient management (an unrealistic ex-pectation given national physician avail-ability) nurse recognition of perfor-mance errors in progress and, whennecessary, access of chain of command,will probably remain the primary, albeitimperfect barrier preventing some ofthese types of errors from causing harm.Inadequate nurse staffing is another ex-ample of this category of event, and inmost instances involved unforeseeablecircumstances, such as the rare need tosimultaneously perform multiple emer-gency cesarean deliveries.

Low-frequency-low hazard events areexemplified by cases of equipment fail-ure (Table). These events represent er-rors that are both uncommon, and inour system are generally readily and ef-fectively detected before harm reachesthe patient. Although ongoing vigilanceis warranted, such events should takelower priority and deserve less resourceallocation than the other categories ofnear-miss events.

Slip and fall events were surprisinglycommon in this series, and appear topose a unique challenge to a labor anddelivery service. Current systems in placeinclude instruction of both mother and

father in proper handling of an infant

DECEMBER 2012 Am

and ready assistance in ambulation forall women, particularly those with im-paired sensorium, mobility, or balance.However, in most cases, these events oc-curred in patients whose clinical statusdid not appear to warrant unusual pre-cautions. Given the current emphasis ona private, family-centered peripartumexperience, it is unrealistic to insist that apostpartum woman not get out of bedunless assisted by a nurse. Further, we areuncertain how a facility can ever elimi-nate the possibility of a father fallingasleep and dropping his infant. Indeed, apostpartum rooming-in mother may bethe only type of patient in all of medicinewho is expected to simultaneously fulfillthe roles of patient and provider (to herinfant). Although continued vigilanceand patient education are certainly war-ranted, achievement of “never event”status for maternity slip and falls is prob-ably not consistent with the provision ofa family-centered birthing experience.Fortunately, we found both mothers andinfants to be resilient to such events; noserious injuries were seen due to slip andfall events on labor and delivery in theentire broader series.

This study has several limitations.First, given the extensive perinatal pa-tient safety efforts instituted in our hos-pital system over the past decade, ournear-miss experience may not be gener-alizable to all US hospitals. For example,our efforts to construct effective barriersto prevent harm because of medicationerrors or errors in patient identificationhave been highly effective but have comeabout only because we have dedicatedimmense resources to this area of care.Second, reporting was voluntary. Thisapproach is encouraged by the Instituteof Medicine and modeled after NASA’shighly successful Aviation Safety Report-ing System.1,17 Although we were en-couraged by the observation that re-ported events (nonerror, error leading toharm, and near-misses) included bothtrivial and fatal events, we can think ofgood reasons to speculate that events oneither extreme of this spectrum mighttheoretically be either over or underreported in our series. Unfortunately,to avoid the potential errors involved

in voluntary reporting, one would have

erican Journal of Obstetrics & Gynecology 443

fcmnc

ahsit

bhehiim

bacchs

tpfpfwvaiHqleilvendtmgsiiwp

sIsaS6aC

Expert Reviews Patient Safety Series www.AJOG.org

to systematically scrutinize tens ofthousands of medical records for sub-clinical near-miss errors. Our data sug-gests that even if only half the near-miss events that actually occurred werevoluntarily reported, a detailed analysisof over 100,000 medical records wouldbe necessary to generate a series of com-parable size. Given the time-intensivenature of such an analysis, voluntary re-porting probably represents the best datathat will ever be available on this scale.Clearly, however, our data are more reli-able in describing the relative than theabsolute frequency of near-miss events.Finally, although the analysis by ourteam included a detailed rereview of eachindividual report by an experienced ma-ternal-fetal medicine subspecialist, a de-gree of subjectivity was inevitable in as-signing the potential for harm that didnot, in fact, occur, as well as in assessingthe strength of barriers to harm.

Our analysis has led to a numberof important observations that are assist-ing us in improving patient safety withinour organization. First, although we arepleased with the effectiveness of our ex-isting patient safety barriers to preventmedication and patient identification er-rors from reaching the patient, we werereminded by this analysis that potentialvulnerabilities in these areas persist, andare common. This realization has led usto a major initiative, spearheaded by ourlocal nursing leaders to continually re-mind all staff of the importance of ongo-ing vigilance and compunction in adher-ence to protocol in areas in which therarity of adverse outcomes could lead tocomplacence. We are also seeking tosimplify and streamline current medica-tion delivery systems that require repet-itive human input and interaction, thusmultiplying the opportunity for error.For example, a critical analysis of workflow in the provision of medication atHCA demonstrated the need for morethan 30 individual steps, each involvingthe potential for human error, betweenthe physician decision to order certainmedications and the medication enter-ing the patient’s body. Work flow rede-sign has resulted in significant reduc-tions in the number of such steps, thus

reducing the potential for error. h

444 American Journal of Obstetrics & Gynecology

Second, most high-frequency/highhazard errors involved the failure to re-spond, or to follow established protocolby staff physicians, less than 5% of whomare employed. Solutions to this problemare problematic, and suggest a signifi-cant flaw in the independent medicalstaff model of practice present in the ma-jority of facilities in the United States.The most effective solutions to these se-rious problems would involve nationalefforts to strictly define and implementuniform standards of physician practicefor high risk areas.1,2 We have success-ully promoted such a system for selectare processes within our facilities forany years and have seen an accompa-

ying dramatic decline in adverse out-omes.18-22 Our data confirm the need

for a similar approach by physician orga-nizations as recommended over a decadeago by the Institute of Medicine.1,2 Ourdata also support the importance of, andjustification for such efforts by otherprovider and payer groups to standard-ize evidence-based physician care pend-ing effective efforts by professional orga-nizations. We note, however, that in arecognized hierarchy of safety barriers,education, instruction, drills, protocols,and policies are relatively weak com-pared with barriers involving forcingfunctions or electronic barriers such asthose available for medication and pa-tient identification errors.4,13 The lack ofpotential for implementation of thesemore effective types of safety barriers toprevent physician practice error furtherilluminates the critical need for ex-panded efforts in practice standardiza-tion and drills, because these efforts are,by nature, intrinsically less effective.4 Inddition, the 24-hour presence of an in-ouse obstetrician, when feasible, woulderve as a more effective forcing functionn cases of failure to respond by the at-ending physician.

Finally, although chain of commandarriers proved effective in preventingarm because of physician performancerror in our series of low-frequency/highazard near misses, we feel less confident

n the universal effectiveness of barriersnvolving events that are both uncom-

only encountered by nurses, and are

ighly dependent on the personalities of 2

DECEMBER 2012

oth providers involved. The continuedvailability and emphasis of an effectivehain of command for all providers in alllinical situations seems imperative andas been a major focus of our patientafety efforts.

In conclusion, our analysis suggestshat the construction of effective barriersreventing errors on labor and delivery

rom causing harm to the patient areossible. Near-miss error analysis is use-

ul in identifying both strengths andeaknesses in an individual system in-olving physicians and nurses. The use offrequency/hazard score matrix (Table)

s of value in prioritizing safety efforts.igh hazard, high frequency events re-

uire intense and immediate resource al-ocation to reduce both the frequency ofrror and the quality of barriers prevent-ng error from causing harm. Events ofower frequency or hazard may be indi-idually addressed by efforts to reduceither the rate of error or the quality orumber of effective barriers to harm, asetermined by this matrix analysis. Thisype of analysis facilitates resource use

ore efficiently than “shotgun” pro-rams broadly directed at general patientafety. Such analysis is useful in facilitat-ng the efficient direction of resourcesnto areas in which additional barriersill be maximally effective in advancingatient safety. f

REFERENCES1. Institute of Medicine. To err is human. Build-ing a safer health system. Washington, DC: Na-tional Academy Press; 2000.2. Institute of Medicine. Crossing the qualitychasm. A new health system for the 21st cen-tury. Washington, DC: National AcademyPress; 2001.3. Institute of Medicine. Preventing medicationerrors. Washington, DC: National AcademyPress; 2007.4. Reason J. Managing the risks of organiza-tional accidents. Burlington, VT: Ashgate Pub-lishing; 1997.5. Institute for Safe Medication Practices. ISMPurvey helps define near miss and close call.nstitute for Safe Medication Practices web-ite. Available at: www.ismp.org/newsletters/cutecare/articles/20090924.asp. Accessedept. 24, 2009.. Joint Commission. Sentinel events. Availablet: www.jointcommission.org/assets/1/6/2011_AMOBS_SE.pdf Jan. 6, 2011. Accessed May

3, 2012.

www.AJOG.org Patient Safety Series Expert Reviews

7. Clark SL, Belfort MA, Hankins GDV, et al.Variation in the rates of operative delivery in theUnited States. Am J Obstet Gynecol 2007;196:526-7.8. Clark SL, Belfort MA, Dildy GA, et al. Maternaldeath in the 21st century: prevention and rela-tionship to cesarean delivery. Am J Obstet Gy-necol 2008;199:36.e1-5.9. Clark SL, Belfort MA, Dildy GA, et al. Emer-gency department utilization during the post-partum period–implications for current man-agement of the puerperium. Am J ObstetGynecol 2010;202:e35-7.10. Cochrane WG. Sampling techniques, 3rded. New York: John Wiley & Sons; 1977.11. Bartlett JE, Kotrlik JW, Higgins CC. Org-anizational research: determining appropriatesample size in survey research. Available at:http://www.osra.org/itlpj/bartlettkotrlikhiggins.pdf. Accessed May 23, 2012.12. American Society for Quality. Process anal-ysis tools: failure modes and effects analysis(FMEA) American Society for Quality Website.

Available at: http://asq.org/learn-about-quality/

process-analysis-tools/overview/fmea.html. Acc-essed May 23, 2012.13. Thornton RH, Miransky J, Killen AR, et al.Analysis and prioritzation of near-miss adverseevents in a radiology department. AJR 2011;196:1120-4.14. Bilimora KY, Kmiecik TE, DaRosa DA, et al.Development of an online morbidity, mortalityand near-miss reporting system to identify pat-terns of adverse events in surgical patients.Arch Surg 2009;144:305-11.15. Souza JP, Cecatti JG, Parpinelli MA, et al.Appropriate criteria for identification of near-miss maternal morbidity in tertiary care facilities:a cross sectional study. BMC Pregnancy Child-birth 2007;7:20-31.16. Mhyre JM, Baterman BT, Leffert LR. Influ-ence of patient comorbidities on the risk ofnear-miss maternal morbidity or morality. Anes-thesiology 2011;115:963-72.17. NASA Aviation Safety Reporting System.Available at: http://asrs.arc.nasa.gov/. AccessedMay 23, 2012.18. Clark SL, Belfort MA, Saade GA, et al. Im-

plementation of a conservative, checklist based

DECEMBER 2012 Am

protocol for oxytocin administration: maternaland fetal outcomes. Am J Obstet Gynecol2007;197:480.e1-5.19. Clark SL, Belfort MA, Meyers JA, et al. Im-proved outcomes, fewer cesarean deliveriesand reduced litigation: results of a new para-digm in patient safety. Am J Obstet Gynecol2008;199:105.e1-7.20. Mah MP, Clark SL, Akhigbe E, et al. Reduc-tion of severe hyperbilirubinemia following insti-tution of universal predischarge bilirubin screen-ing in a large private hospital system. Pediatrics2010;125:e1143-8.21. Clark SL, Frye DR. Meyers JA, et al. Reduc-tion in elective delivery prior to 39 weeks gesta-tion–comparative effectiveness of 3 approachesto change and the impact on neonatal outcomeand stillbirth Am J Obstet Gynecol 2010;203:449.e1-6.22. Clark SL, Meyers JA, Frye DK, Perlin JA.Patient safety in obstetrics: the hospital corpo-ration of America experience. Am J Obstet Gy-

necol 2011;204:283-7.

erican Journal of Obstetrics & Gynecology 445