errors in radiotherapy actions - american association of ... · sp e ci f y what how why...

Errors in RadiotherapyErrors in Radiotherapy

Bruce ThomadsenBruce ThomadsenShi-Woei LinShi-Woei Lin

University ofUniversity ofWisconsin -Wisconsin -

MadisonMadison

Slides © Bruce Slides © Bruce ThomadsenThomadsen

RasmussenRasmussen’’s Performance-baseds Performance-basedActionsActions

ErrorsErrorsll Systematic Errors:Systematic Errors:

–– Usually one mistake tucked into the procedureUsually one mistake tucked into the procedure–– Affects all, or a large class of patients.Affects all, or a large class of patients.–– Often found in Process AuditOften found in Process Audit–– Must be rooted outMust be rooted out

ll Random Errors:Random Errors:–– Happen on a per-patient basisHappen on a per-patient basis–– May be caught through QMMay be caught through QM–– Will never be eliminated (because of creativity)Will never be eliminated (because of creativity)

OneOne Example of ErrorExample of ErrorAnalysisAnalysis in Radiotherapyin Radiotherapy

ll We did a study of brachytherapy errorsWe did a study of brachytherapy errorsbased on all misadministrations reported tobased on all misadministrations reported tothe NRC.the NRC.

ll WeWe performed several analyses of the events.performed several analyses of the events.

CaveatCaveat

ll The NRC does NOT keep any records ofThe NRC does NOT keep any records ofphysician errors in diagnosing orphysician errors in diagnosing orprescribing. That, they say, would beprescribing. That, they say, would bedictating medicine.dictating medicine.

ll The only data is on deviations fromThe only data is on deviations fromprescriptions.prescriptions.

AnalysisAnalysis

ll We constructed a process tree for theWe constructed a process tree for theprocedure.procedure.

ll We constructed a fault tree for the procedure.We constructed a fault tree for the procedure.ll For each event, we:For each event, we:

–– Contacted the principal and got the facts.Contacted the principal and got the facts.–– Constructed a root-cause analysis tree.Constructed a root-cause analysis tree.–– Marked the position of the failure on the fault andMarked the position of the failure on the fault and

process trees.process trees.–– Classified the events using three taxonomies.Classified the events using three taxonomies.

Unintended Area

The step size(parameter) was

wrong

Fail to identify theerror

The arrow keyrotate through

different sizes inthe step field

The physicist didnot notice it

He wasinterrupted

The dosimetristwho check was

not familiar withthe program

enough

The step size stayed atthe wrong size when

the physicist moved itwith mouse

Requirement for amanual entry

The computerwould not

transfer the file atthat size

(P) HRV(R) Identification not correct

(P) OK(R) Excessive demand onknowledge/training(S) Bounded Rationality

(R) Distraction from otherperson

(P) OK(R) Excessive demand onknowledge/traning(S) Incomplete rule(S) Bounded Rationality

(P) TM

(P) HRI(R) Execution iserroneous

The physicist notfamiliar with the

program?

(R) Information not seen orsought(S) Distraction

(R) Procedure isincorrect

The normaldosimetrist who

check the plan leftwith an

emergency(S) ExternalInterference

(P) TD

The physicianjust missed the

error

(R) Manual variability(R) Spontaneoushuman variability(S) Inadequate SearchBehavior

ExampleExampleRoot-cause-Root-cause-

analysisanalysisTreeTree

SuccessfulTreatment

Treatmenttermination

Emergency response

Post Txmonitoring

Monitoring

Applicatorcheck

Source verification

Patientidentification

Dose / timecalculation

Specification

Prescription

Limitations ofalgorithms

Dosimetrycalculation

ReconstructionApplication

Calibration

clinicalstage

Protocol

Anatomicinformation

Planning qualityassurance

Correctness

Consistence

Completion

Localization

Geometry

Identification

Targeting

Record Setup

DummiesRecord

Image quality

Correct films Interpretation

Image quality

Geometry

Identity

Data entry

Hardware operation

Software operation

Procedure

Peralgorithm

Per localization

Fiducials

Anatomy

DummiesPlacement

Applicator

correcttarget

Correctanatomy Correct

applicatorplacement

Identification Correctselection

Entering datain computer

Calculation of strength

Reading

Placement in well

Calibration factors

factorsdate

strength

Measurement

Set chamber

Set source

LDR Brachytherapy Process Tree 1:Placement followed by dosimetry

Proceduresleading to anLDR Patienttreatment Optimization

Duration calculation

Satisfaction

Sourceloading

Sourcepreparation

Sourceselection

Insertion incarrier

sourceintegrity

Source loading

Time recording

Removalcalculation

Removal timeverification

Removal preparation

Source removal

Sourcecount

Radiationsurvey

Recording

Doseinformation

Other Txinformation

treatment durationexecuted

Source fixation

Applicatorfixation

sourcestrength

factors patientdatainput

1

1

10

4

1

1

9

1

1

1

2

10383

1

2

1

2

1

1

Conclusions fromConclusions fromProcess Tree Analysis 1Process Tree Analysis 1

ll For HDRFor HDR–– By far the most common step with failure wasBy far the most common step with failure was

entering the treatment distance, usually notentering the treatment distance, usually notchanging the default value.changing the default value.

–– Almost all steps in treatment unit programmingAlmost all steps in treatment unit programmingor delivery had some errors.or delivery had some errors.

–– Dose specification accounted for several errors.Dose specification accounted for several errors.–– The only problems with source calibration wereThe only problems with source calibration were

in entering the calibration data into the treatmentin entering the calibration data into the treatmentplanning computer.planning computer.


ll For LDR (placement followed by dosimetry)For LDR (placement followed by dosimetry)–– Errors in four steps accounted for most of the events:Errors in four steps accounted for most of the events:

»» Selection of the sources,Selection of the sources,»» Loading of sources into the applicator,Loading of sources into the applicator,»» Using the required units when entering data into theUsing the required units when entering data into the

computer, andcomputer, and»» Fixing the sources in the applicator, or applicator in theFixing the sources in the applicator, or applicator in the

patient.patient.–– Most steps in Most steps in ““Source Loading,Source Loading,”” ““Dose/timeDose/time

calculation,calculation,”” and and ““Treatment terminationTreatment termination”” had haderrors.errors.


ll For LDR (Dosimetry followed by placement),For LDR (Dosimetry followed by placement),errors occurred only inerrors occurred only in–– source preparation (usually ordering), andsource preparation (usually ordering), and–– source delivery (usually a failure to monitor).source delivery (usually a failure to monitor).

Deviation fromadequatetreatment

or

Wrong dosedistribution or

site

Wrong patienttreated

or

and

Fractionationfailure

Wrongapplicator used

Treatmentplanning failure

Treatmentimplementation

failure

Wrong patientselected

Failure toidentify patient

Prescriptionerror

Accountingerror

Error intreatmentplanning

Verificationerror

Applicatorpositioning

error

Applicatorconnection

error

Treatmentprogramming

failure

Treatmentdelivery error

Treatmentterminated

prematurely

or

and

or

A

B

C

D

G

H

I

J

L

Go to Page 2

Go to Page 2

Go to Page 2

Go to

Go to Page 6

Go to Page 7

Go to Page 7

Go to Page 8

Go to Page 10

HDR Fault Tree

Page 1

0

0

13/44

13/44

3/44

2/44

21/44

4/44

0

1/44

1/44

30/44

1/44

13/44

0

0

43/44

44/44

F Dose calculationerror

Incompatiblefactors for

calibration anddose calculation

Software error

or

Inconsistent stepsize

Incorrect shapeof dose

distribution(incorrect

optimization)

Incorrect dwelltimes entered(manual, notoptimized)

Wrong dwellpositionsactivated

Dosespecification towrong points

Wrong locationof dose

distribution

Wrong dose

Incorrect dataentry

Incorrect datatransfer

Wrong patient'sdata used

Wrong orincompatible

units

or

Error in transfer(transcription)

Datatransposition

Interpretationerror

or

Wrong chartreferenced

Physician's error

or

Algorithm error

Software versionincompatibility

Corrupt file

or

Entry error

Inappropriatemarker

Marker in wrongposition

orIncorrect entry

Physician's error

or

Error inspecification

Incorrect markerused

or

Inaccuratesource position

entry

QM failure

andFile corruption

QM failure

and

Programmingerror

Acceptancetesting error

Page 5

2/44

2/44

2/44

1/44

2/44

2/44

7/44

SomeSomeParts ofParts ofthe Treethe Tree

areareBroadBroad

E Dosimetry Error

Source strengtherror

Dose calculationerror

or

or

Wrongcalibration

Wrong sourcedata

Wrong data(wrong decay

factor)

F

Calibration error

Failure ofverification

and

Erroneousstrength forsource data

Failure ofverification

and

Wrong dataformat (US/Euro)

Incorrect entry

and

or

Wrong units

Measurementerror

Calculation error

or

Error in dataentry

Failure to enteror alter data (unit

default)

Wrong source indevice

Discrepency instrength between

device andplanning system

Go toPage 5

3/443/44

3/44

3/443/44

7/44

10/44

Some PartsSome Partsof the Treeof the Treeare Deepare Deep

Summary from HDR FaultSummary from HDR FaultTree TabulationTree Tabulation

ll 2/3 errors in delivery2/3 errors in delivery–– 40% of errors due to default value for distance not40% of errors due to default value for distance not

being changed.being changed.–– Applicator shifting in patient was only otherApplicator shifting in patient was only other

frequent problem in delivery.frequent problem in delivery.ll 1/3 errors in treatment planning1/3 errors in treatment planning

–– 16% of errors in calculation, but of various types.16% of errors in calculation, but of various types.–– 7% of errors due to incorrect source strength entry.7% of errors due to incorrect source strength entry.

ll Almost all events had failures in verification.Almost all events had failures in verification.

Summary from LDR FaultSummary from LDR FaultTree Tabulation 2Tree Tabulation 2

ll 1/4 errors in treatment planning1/4 errors in treatment planning–– 15% of errors in calculation,15% of errors in calculation,

»» 11% due to incompatible units.11% due to incompatible units.–– 8% of errors due to incorrect source strength entry.8% of errors due to incorrect source strength entry.

ll Again, almost all events had failures inAgain, almost all events had failures inverification.verification.

Summary from LDR FaultSummary from LDR FaultTree TabulationTree Tabulation

ll 3/4 errors in delivery3/4 errors in delivery–– 11% because the patient removed the sources and11% because the patient removed the sources and

the staff didnthe staff didn’’t notice or correct.t notice or correct.–– 12% because the sources were never placed in the12% because the sources were never placed in the

applicator correctly.applicator correctly.–– 14% because the wrong source strengths were14% because the wrong source strengths were

used.used.–– 8% because the physician placed the applicator8% because the physician placed the applicator

incorrectly.incorrectly.

Analysis Based onAnalysis Based onTaxonomiesTaxonomies

ll Taxonomies, as you have heard, are orderedTaxonomies, as you have heard, are orderedand organized classifications.and organized classifications.

ll They often can give insight into the nature ofThey often can give insight into the nature ofthe errors occurring.the errors occurring.

ll While we looked at several, and developedWhile we looked at several, and developedour own, we will justour own, we will just present twopresent two today.today.

RasmussenRasmussen’’ssWhatWhat

HappenedHappenedPathwayPathway

RasmussenRasmussen’’ssWhy ItWhy It


RasmussenRasmussen’’ssWhy ItWhy It


Rasmussen Human Error Model (HDR)

10

22

2

0

3

8

20

15

43

7

1

3

1

34

13

9

56

12

10

3

1

12

12

15

00

5

10

15

20

25

Det

ectio

n

Iden

tific

atio

n

Goa

l

Targ

et

Task

Pro

cedu

re

Exe

cutio

n

Man

ual v

aria

bilit

y

Topo

grap

hic

diso

rient

atio

n

Ste

reot

ype

take

over

Ste

reot

ype

fixat

ion

Fam

iliar

pat

tern

not

reco

gniz

ed

Forg

ets

isol

ated

act

Mis

take

s al

tern

ativ

es

Oth

er s

lip o

f mem

ory

Fam

iliar

ass

ocia

tion

shor

tcut

Info

rmat

ion

not s

een

Info

rmat

ion

assu

med

Info

rmat

ion

mis

inte

rpre

ted

Sid

e ef

fect

s no

t ade

quat

ely

cons

ider

ed

Oth

er

Dis

tract

ion

from

sys

tem

Inte

rferr

ing

task

Dis

tract

ion

from

oth

er p

erso

n

Exc

essi

ve p

hysi

cal d

eman

d

Exc

essi

ve d

eman

d on

kno

wle

dge

Inst

ruct

ion

inco

rrec

t

Ope

rato

r inc

apac

itate

d

Spo

ntan

eous

hum

an v

aria

bilit

y

Oth

er, s

peci

fy

WhatWhat HowHow WhyWhy

Conclusions fromConclusions fromTaxonometric Taxonometric Analysis 1Analysis 1

From From RasussenRasussen’’ss Model: Model: ““WhatWhat””ll For both HDR and LDR, noticing the problemFor both HDR and LDR, noticing the problem

was the most significant variable.was the most significant variable.–– For HDR, mostly the problem was identifying theFor HDR, mostly the problem was identifying the

problem using verification procedures in placeproblem using verification procedures in place(either they were not adequate or not performed).(either they were not adequate or not performed).

–– For LDR, mostly there were no procedures in placeFor LDR, mostly there were no procedures in placeto look for problems.to look for problems.

ll For both HDR and LDR, the next rankingFor both HDR and LDR, the next rankingfailure was in the execution of procedures,failure was in the execution of procedures,

ll Followed by the procedures being wrong.Followed by the procedures being wrong.


From From RasussenRasussen’’ss Model: Model: ““HowHow””ll From both HDR and LDR, the single mostFrom both HDR and LDR, the single most

common failure is common failure is ““manual variabilitymanual variability””..–– This is probably an artifact of the model, whichThis is probably an artifact of the model, which

expects a human reactions to a plant problem.expects a human reactions to a plant problem.–– This reflects that the initiating events in medicine isThis reflects that the initiating events in medicine is

usually some personusually some person’’s action.s action.ll Grouped, Stereotype responses come close.Grouped, Stereotype responses come close.ll Information not seen, assumed orInformation not seen, assumed or

misinterpreted also was significant; for HDRmisinterpreted also was significant; for HDRthey formed the dominant failure modes.they formed the dominant failure modes.


From From RasussenRasussen’’ss Model: Model: ““WhyWhy””ll These categories were not These categories were not codablecodable for many for many

events.events.ll The most common classification was theThe most common classification was the

catchall catchall ““Spontaneous human variabilitySpontaneous human variability””..ll ““Excessive demand on knowledgeExcessive demand on knowledge”” was was

significant, particularly for HDR, which is moresignificant, particularly for HDR, which is moretechnical.technical.

ll Interfering tasks were also important in HDR,Interfering tasks were also important in HDR,which in more intensive at a given time.which in more intensive at a given time.

SMARTSMARTPathwayPathway

van van der Schaaf der Schaaf et al.et al.

SMART Human Error Model (Pinball Method)

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

Ext

erna

l (Te

chni

cal)

Des

ign

Con

stru

ctio

n

Mat

eria

ls

Ext

erna

l(O

rgan

izat

iona

l)

Kno

wle

dge

trans

fer

Pro

toco

ls

Man

agem

ent

prio

ritie

s

Cul

ture

Ext

erna

l (H

uman

beha

vior

)

Kno

wle

de(K

now

ledg

e ba

sed)

Qua

lific

atio

ns

Coo

rdin

atio

n

Ver

ifica

tion

Inte

rven

tion

Mon

itorin

g

Slip

s

Trip

ping

Pat

ient

rel

ated

failu

re

Unc

lass

ifiab

le

HDRLDR

SMARTSMART’’s s Suggested ActionsSuggested Actions


From the SMART model: The results are veryFrom the SMART model: The results are verysimilar for both LDR and HDR.similar for both LDR and HDR.

ll By far, the dominant failure mode wasBy far, the dominant failure mode was““Verification failureVerification failure””, followed by, followed by““InterventionIntervention”” (which was scored if someone (which was scored if someonejust goofed).just goofed).

ll Inadequate Inadequate ““ProtocolsProtocols”” (i.e., procedures) were (i.e., procedures) wereimportant, particularly in LDR. Also in LDR,important, particularly in LDR. Also in LDR,lack of lack of ““MonitoringMonitoring”” was a common problem. was a common problem.


From the SMART model: (Continued)From the SMART model: (Continued)ll Of about equal importance, Of about equal importance, ““KnowledgeKnowledge

transfer transfer ““ (training), (training), ““ManagementManagementprioritiespriorities”” (lack of staffing), and (lack of staffing), and ““CultureCulture””(disregard for safety procedures) each(disregard for safety procedures) eachshowed up as important.showed up as important.

ll Design was a common problem, as as wasDesign was a common problem, as as wasnoted by all the other analyses.noted by all the other analyses.


ll Very few of the events involved knowledge-Very few of the events involved knowledge-based errors.based errors.

ll While the taxonomies tested did give usefulWhile the taxonomies tested did give usefulinformation, they obviously did not matchinformation, they obviously did not matchthe medical setting well.the medical setting well.

Overall Conclusions 1Overall Conclusions 11.1. Evaluation of a medical procedure using risk analysisEvaluation of a medical procedure using risk analysis

provides insights.provides insights.2.2. Failure to consider human performance in the design ofFailure to consider human performance in the design of

equipment led to a large fraction of the eventsequipment led to a large fraction of the eventsreviewed.reviewed.•• While the equipment per se did not fail, the designWhile the equipment per se did not fail, the design

facilitated the operator to make mistakes that resultedfacilitated the operator to make mistakes that resultedin the erroneous treatments.in the erroneous treatments.

•• Of particular danger were those situations whereOf particular danger were those situations whereequipment malfunctions force operators to performequipment malfunctions force operators to performfunctions usually executed automatically by machines.functions usually executed automatically by machines.

•• Entry of data in terms of units other than thoseEntry of data in terms of units other than thoseexpected by a computer system also accounted forexpected by a computer system also accounted forseveral events.several events.

Overall Conclusions 2Overall Conclusions 23.3. HDR brachytherapy events tended to happenHDR brachytherapy events tended to happen

most with actions having the least timemost with actions having the least timeavailable.available.

4.4. LDR brachytherapy, the most hazardous stepsLDR brachytherapy, the most hazardous stepsin the procedure entailed:in the procedure entailed:–– selecting the correct sources to place in the patient,selecting the correct sources to place in the patient,–– setting the sources in place properly in the patientsetting the sources in place properly in the patient

and keeping them in place.and keeping them in place.–– These events mostly result from lack of attention atThese events mostly result from lack of attention at

critical times.critical times.

Overall Conclusions 3Overall Conclusions 3

5.5. Many events followed the failure of personsMany events followed the failure of personsinvolved to detect that the situation wasinvolved to detect that the situation wasabnormal, often even though manyabnormal, often even though manyindications pointed to that fact.indications pointed to that fact.

6.6. Once identified, the response often includedOnce identified, the response often includedactions appropriate for normal conditions,actions appropriate for normal conditions,but inappropriate for the conditions of thebut inappropriate for the conditions of theevent.event.

Overall Conclusions 4Overall Conclusions 47.7. Lack of training (to the point that personsLack of training (to the point that persons

involved understand principles) andinvolved understand principles) and8.8. Lack of procedures covering unusualLack of procedures covering unusual

conditions likely to arise (and sometimes, justconditions likely to arise (and sometimes, justroutine procedures) frequently contributed toroutine procedures) frequently contributed toevents.events.

9.9. New procedures, or new persons joining aNew procedures, or new persons joining acase in the middle also present a hazard.case in the middle also present a hazard.

–– 7/46 7/46 evaluable evaluable in LDR.in LDR.–– 12/38 12/38 evaluable evaluable in HDR.in HDR.

Overall Conclusions 5Overall Conclusions 5

10.10. Most of the events suffered from ineffectualMost of the events suffered from ineffectualverification procedures, a failure noted byverification procedures, a failure noted byall three taxonomies. For the most part,all three taxonomies. For the most part,improved quality management would serveimproved quality management would serveto interrupt the propagation of errors byto interrupt the propagation of errors byindividuals into patient events.individuals into patient events.

Observations on CommonObservations on CommonCauses of EventsCauses of Events

ll Failures in medicine parallel those in industry.Failures in medicine parallel those in industry.ll Errors donErrors don’’t just happen from a single cause,t just happen from a single cause,

but are surrounded by complicating situations.but are surrounded by complicating situations.ll Distraction (due to pressures and otherDistraction (due to pressures and other

assignments)assignments)ll Rushing (due to pressures and Rushing (due to pressures and lack of staffinglack of staffing))ll Lack of communication (between parties)Lack of communication (between parties)

Analysis of External-beamAnalysis of External-beamEventsEvents

The events fall clearly into categories:The events fall clearly into categories:ll Random errors in a patient treatmentRandom errors in a patient treatment

–– Few calculation errors (where much of QA falls)Few calculation errors (where much of QA falls)–– Frequent errors when treatments are odd (e.g., oddFrequent errors when treatments are odd (e.g., odd

angles used in the wrong direction)angles used in the wrong direction)–– Not uncommon following a change in prescription mid-Not uncommon following a change in prescription mid-

course.course.–– Not checking patient set-up after pause or interruption.Not checking patient set-up after pause or interruption.

Analysis of External-beamAnalysis of External-beamEvents (continued)Events (continued)

ll Systematic errorsSystematic errors–– Errors in commissioning or calibration (note: theErrors in commissioning or calibration (note: the

errorserrors themselves are random, but propagate themselves are random, but propagate asassystematic).systematic).

–– Errors in formulaeErrors in formulae–– Errors in data entry or use of incorrect unitsErrors in data entry or use of incorrect units–– Usually there has been no verification or check ofUsually there has been no verification or check of

the data (strange, that we now always check athe data (strange, that we now always check asinglesingle patientpatient’’s calculation, sometimes severals calculation, sometimes severaltimes)times)

Commonalty in Most EventsCommonalty in Most Events

The persons involved often fall into traps, setThe persons involved often fall into traps, setby the practice environment, and respondby the practice environment, and respondlike human beings.like human beings.

errors in radiotherapy actions - american association of ... · sp e ci f y what how why...

Documents