risk models to improve safety of dispensing high-alert
TRANSCRIPT
ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org584 • JAPhA • 52:5 • S e p /O c t 2012
Received September 30, 2010, and in revised form June 17, 2011. Accepted for publication July 22, 2011.
Michael R. Cohen, BSPharm, MS, FASHP, is President; and Judy L. Smetzer, RN, BSN, FISMP, is Vice President, Institute for Safe Medication Practices, Horsham, PA. John E. Westphal, BS, is a consultant; and sharon Conrow Comden, BS, MPH, DrPH, is a con-sultant, Outcome Engenuity, LLC, Plano, TX. Donna M. Horn, BSPharm, is Director of Pa-tient Safety for Community Pharmacy, Insti-tute for Safe Medication Practices, Horsham, PA.
Correspondence: Judy Smetzer, RN, BSN, FISMP, Institute for Safe Medication Practic-es, 200 Lakeside Dr., Suite 200, Horsham, PA 19044. Fax: 215-914-1492. E-mail: [email protected]
Disclosure: The authors declare no con-flicts of interest or financial interests in any product or service mentioned in this article, including grants, employment, gifts, stock holdings, or honoraria.
Funding: Agency for Healthcare Research and Quality contract no. 1P20HS017107.
Previous presentations: 10th International Probabilistic Safety Assessment & Manage-ment Conference, Seattle, WA, June 7–11, 2010, and Agency for Healthcare Research and Quality Annual Conference, Bethesda, MD, September 14, 2009.
abstract
Objectives: To determine whether sociotechnical probabilistic risk assessment can create accurate approximations of detailed risk models that describe error path-ways, estimate the incidence of preventable adverse drug events (PADEs) with high-alert medications, rank the effectiveness of interventions, and provide a more infor-mative picture of risk in the community pharmacy setting than is available currently.
Design: Developmental study.Setting: 22 community pharmacies representing three U.S. regions.Participants: Model-building group: six pharmacists and three technicians.
Model validation group: 11 pharmacists; staff at two pharmacies observed.Intervention: A model-building team built 10 event trees that estimated the inci-
dence of PADEs for four high-alert medications: warfarin, fentanyl transdermal sys-tems, oral methotrexate, and insulin analogs.
Main outcome measures: Validation of event tree structure and incidence of defined PADEs with targeted medications.
Results: PADEs with the highest incidence included dispensing the wrong dose/strength of warfarin as a result of data entry error (1.83/1,000 prescriptions), dis-pensing warfarin to the wrong patient (1.22/1,000 prescriptions), and dispensing an inappropriate fentanyl system dose due to a prescribing error (7.30/10,000 prescrip-tions). PADEs with the lowest incidence included dispensing the wrong drug when filling a warfarin prescription (9.43/1 billion prescriptions). The largest quantifiable reductions in risk were provided by increasing patient counseling (27–68% reduc-tion), conducting a second data entry verification process during product verification (50–87% reduction), computer alerts that can't be bypassed easily (up to 100% re-duction), opening the bag at the point of sale (56% reduction), and use of barcoding technology (almost a 100,000% increase in risk if technology not used). Combining two or more interventions resulted in further overall reduction in risk.
Conclusion: The risk models define thousands of ways process failures and be-havioral elements combine to lead to PADEs. This level of detail is unavailable from any other source.
Keywords: Risk assessment, high-alert medications, preventable adverse drug events, event trees.
J Am Pharm Assoc. 2012;52:584–602.doi: 10.1331/JAPhA.2012.10145
Risk models to improve safety of dispensing high-alert medications in community pharmaciesMichael R. cohen, Judy L. smetzer, John e. Westphal, sharon conrow comden, and Donna M. horn
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 585
Adverse drug events (ADEs), which are defined as inju-ries from drug therapy,1 are among the most common causes of harm during the delivery of health care.2 At
least a quarter of these events are preventable.1,3,4 On an an-nual basis, up to 450,000 inpatients experience a preventable ADE (PADE).3–5 PADEs lead to about 4% (range 1.4–15.4%) of hospital admissions.6–12
Few prospective data detail the incidence of PADEs in am-bulatory patients.12 Four retrospective studies that examined community pharmacy dispensing errors using similar defini-tions, detection methods, and expression of incidence rates re-ported a wide range of errors (from 1.7% to 24%).13–16 The low-est dispensing error rate (1.7%) translates to approximately four errors per 250 prescriptions per pharmacy per day2 or to
60 million errors during the dispensing of 4 billion prescrip-tions annually.2,17
Few studies have reported the frequency of harm caused by PADEs in the community pharmacy setting. Ghandi et al.12 found that 5% of ambulatory patients experienced a PADE with medications dispensed from community pharmacies. Gurwitz et al.18 identified that almost one-half of serious, life-threaten-ing, or fatal ADEs related to medications dispensed from phar-macies were preventable. Several studies suggested that dos-ing errors occur frequently and have the highest rate of clinical significance among types of medication errors.19–22 An estimate in 2000 determined that hospital admissions caused by PADEs accounted for $121.5 billion or 70% of total costs of drug-re-lated problems in the United States.23
The drugs associated with the most harmful PADEs in acute care settings were first coined “high-alert” medications by the Institute for Safe Medication Practices (ISMP) in 1998.24 High-alert medications carry a major risk of causing serious injuries or death to patients if misused. Errors with these drugs are not necessarily more common, but the consequences are devastating.25–27 Appendix 1 (electronic version of this article, available online at www.japha.org) notes the characteristics, medications, and types of errors involved in patient harm from PADEs in the ambulatory setting.
Traditionally, health care systems have relied on root cause analysis (RCA) and failure mode and effects analysis (FMEA) to understand the risks involved in prescribing, dis-pensing, and administering medications.28–30 RCA and FMEA are the most basic types of risk analysis that focus largely on system and process errors.30–35 Both offer qualitative informa-tion about risk and error, but neither helps quantify the level of risk or model the dependencies and effects of combinations of failures.36 Sociotechnical probabilistic risk assessment (ST-PRA) is a prospective technique that advances the qualitative work of FMEA and RCA into a quantitative realm by linking process failures with estimates of human error and behavioral norms, yielding a more accurate picture of why and how often these failures affect patient outcomes.28,30,36,37 Online Appendix 2 summarizes the advantages of ST-PRA over FMEA and RCA.
ST-PRA, which is derived from a probabilistic risk assess-ment (PRA) tool that originated in the mid-1970s to improve safety in nuclear power plants, allows all possible combina-tions of task or system failures to be considered in combina-tion with one another.36–40 Although PRA is predominantly used to model mechanical systems, ST-PRA is especially suited for modeling human systems and is the more appropriate tool for health care.28,30,36,37 Although ST-PRA use in health care remains relatively new,28,30,36,37,41–44 a previous study using ST-PRA to model medication system risk in long-term care strongly suggests that application of this process to high-alert medications dispensed from community pharmacies will be successful in assessing risks and gauging the impact of system and behavioral changes on these risks.30,36
At a GlanceSynopsis: A model-building team built 10 event
trees that estimated the incidence of preventable ad-verse drug events (PADEs) for four high-alert medica-tions and found that sociotechnical probabilistic risk assessment (ST-PRA) was able to define thousands of ways process failures and behavioral elements com-bine to lead to PADEs. PADEs with the highest inci-dence included dispensing the wrong dose/strength of warfarin as a result of data entry error (1.83/1,000 prescriptions) and dispensing warfarin to the wrong patient (1.22/1,000 prescriptions). The greatest quan-tifiable reductions in risk were provided by factors such as increasing patient counseling (27–68% reduc-tion) and conducting a second data entry verification process during product verification (50–87% reduc-tion).
Analysis: The ST-PRA models created in this study were effective at identifying dispensing system vulnerabilities that were largely correctable before reaching patients given environmental, technologi-cal, system/process, and behavioral conditions that are within the reach of most community pharmacies and pharmacy staff. A sensitivity analysis identified that using automated dispensing and barcoding tech-nology, conducting a second data entry verification process during final product verification, counseling patients more frequently and effectively, opening the bag at the point of sale to view all filled prescriptions, and other interventions will reduce prescribing and dispensing errors that reach patients by as much as 87%, or more when combining several interventions. The authors noted that the interventions discussed here are not meant to suggest and/or establish a standard of care for community pharmacies; instead, they are intended to represent future patient safety improvements.
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org586 • JAPhA • 52:5 • S e p /O c t 2012
ObjectivesThe objectives of this study were to (1) identify a list of high-alert medications dispensed from community pharmacies; (2) determine whether ST-PRA can create accurate approxi-mations of detailed risk models that describe error pathways, estimate the incidence of PADEs involving high-alert medica-tions dispensed in community pharmacies, and rank the effec-tiveness of interventions to prevent PADEs; and (3) determine whether ST-PRA provides a more informative picture of risk in the community pharmacy setting than currently available through typical sources, such as retrospective event reporting, RCA, and FMEA.
MethodsThe Temple University Office for Human Subjects Protections Institutional Review Board approved the study before initia-tion. During 2007–08, ISMP partnered with several community pharmacy organizations, from which a purposive sample of 22 pharmacies from three regions was selected to ensure diversi-ty in setting, prescription volume, staffing, hours of operation, and population served.
Model-building sampleThe modeling team consisted of two trained ST-PRA facilitators from Outcome Engenuity, LLC, two clinical research staff from ISMP, and six pharmacists and three pharmacy technicians from nine of the sample pharmacies in the same central south-west region of the United States. The participating pharmacies served urban, suburban, and small community areas and were diverse regarding prescription volumes, hours of operation, and access to drive-through services. Pharmacy staffing pat-terns ranged from a single pharmacist on duty to multiple phar-macists and technicians on duty. The participants had diverse ethnic backgrounds, 5 to 18 years of experience (median 10), and included both genders.
Model validation samplePharmacists who worked in 11 community pharmacies in the New England and mid-Atlantic regions participated in struc-tured interviews to validate the risk models. The pharmacies and participants were diverse in regards to experience, prac-tice settings, daily prescription volume, and gender/ethnic backgrounds. Observations were also conducted at two phar-macies selected from a convenience sample.
Identifying high-alert medicationsA list of high-alert medications dispensed from community pharmacies was compiled using qualitative methods, including analysis of data about PADEs from the following sources: ISMP National Medication Errors Reporting Program,45 the Pennsyl-vania Patient Safety Reporting System,46 the Food and Drug Administration MedWatch database,47 databases from partici-pating pharmacies, community pharmacy survey data,48 public litigation data,49 and literature review.7,9–18,49,50
steps in the modeling processRecruit the modeling team. A voluntary modeling team was recruited using a noncoercive protocol; all members were from different pharmacies.
Build a process and control system map. A process and control map of the pharmacy dispensing process was cre-ated. Observations in pharmacies and discussion with pharma-cy staff verified that any differences in the workflow among the sample pharmacies were minimal, allowing agreement upon one standard process map. The map is a visual aid that clear-ly shows how work inputs, outputs, and tasks are linked and shows the embedded control systems that aid in the prevention and detection of errors.
Mapping occurred at two levels. First, using an iterative progression, the dispensing process steps and decision points were identified and linked according to the current workflow. Then, control systems were identified and added to the map. Active control systems are deliberate steps in the process that specifically help manage the risk of errors, such as data entry verification of prescriptions entered into the computer. Pas-sive controls are features inherent in the system that might help control risks but are not specifically set up for that pur-pose, such as differences in tablet appearance that may alert a pharmacist to an incorrect medication.28
Identify failure modes. An abbreviated FMEA process was used as a hazard identification technique to describe pos-sible failure points (e.g., errors, at-risk behaviors, equipment failures) during the dispensing process and for each targeted high-alert medication. Online Appendix 3 shows small cross sections of the FMEAs related to warfarin and fentanyl trans-dermal systems. The FMEAs were used to determine PADEs to be modeled for the targeted drugs. Six PADEs for warfarin and one PADE each for fentanyl systems, methotrexate, and insulin analogs were selected (Table 1).
Build the risk models. During February to April 2007, the modeling team met six times and created 10 event trees for the PADEs. An event tree is a graphical quantitative risk model that represents the complex relationships among pro-cess steps, organizational culture, human errors, equipment failure, behavioral norms, and undesirable outcomes.28,30,36,37 The event trees decompose the dispensing system as a whole into subsystems and components. The process and control map was used to guide this step. Each individual event tree defined the event sequences that could lead to the specific PADE of in-terest, based on what was currently known about the dispens-ing process and the behavior of systems and pharmacy staff under given conditions.
The risk model building process starts with an initiating er-ror that could lead to a PADE. Each initiating error then was followed through the dispensing process steps, which were called “basic events” in the event trees. The basic events repre-sent (1) exposure rates, or how often certain activities occur; (2) fundamental failures, such as human error, at-risk behav-ior,1 or equipment failure rates; or (3) capture opportunities when errors can be detected and corrected (online Appendix 4). These basic events flowed through the branches of the event tree, linking them together (Figure 1)
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 587
Tabl
e 1.
Sel
ecte
d PA
DEs f
or w
arfa
rin, m
etho
trexa
te, f
enta
nyl s
yste
ms,
and
insu
lin a
nalo
gsH
igh-
aler
t med
icat
ion
Sele
cted
PA
DE
Initi
atin
g er
ror(s
)Ra
tiona
le fo
r sel
ectio
nW
arfa
rin
Wro
ng d
rug
dis-
pens
ed
Disp
ensi
ng e
rror
s: (1
) Wro
ng d
rug
sele
cted
whe
n m
anua
lly fi
lling
a w
arfa
rin p
resc
riptio
n. (2
) Wro
ng
drug
sele
cted
whe
n fil
ling
an a
utom
ated
dis
pens
ing
cabi
net w
ith w
arfa
rin.
Drug
: War
farin
is a
com
mon
ly p
resc
ribed
ora
l med
icat
ion
(con
sist
ently
am
ong
the
top
200 d
rugs
dis
pens
ed e
ach
year
)72,7
3 that
exp
oses
pa-
tient
s to
blee
ding
or t
hrom
bosi
s (su
bthe
rape
utic
dos
es) i
f use
d in
err
or.
Initi
atin
g er
rors
: Sel
ecte
d to
repr
esen
t a va
riety
of p
resc
ribin
g an
d di
spen
sing
err
ors a
ssoc
iate
d w
ith w
rong
pat
ient
, wro
ng d
rug,
wro
ng
dose
, and
wro
ng d
irect
ions
; thr
ee c
ause
s of w
rong
dos
e er
rors
sele
ct-
ed b
ecau
se d
osin
g er
rors
repr
esen
t the
gre
ates
t ris
k to
patie
nts.
19–2
1
Wro
ng d
ose/
stre
ngth
of
war
farin
dis
pens
edPr
escr
ibin
g er
ror:
Wro
ng d
ose/
stre
ngth
tabl
ets
pres
crib
ed.
Wro
ng d
ose/
stre
ngth
of
war
farin
dis
pens
ed
Disp
ensi
ng e
rror
s: (1
) Wro
ng w
arfa
rin d
ose/
stre
ngth
sele
cted
whe
n m
anua
lly fi
lling
a w
arfa
rin
pres
crip
tion.
(2) W
rong
war
farin
dos
e/st
reng
th
sele
cted
whe
n fil
ling
an a
utom
ated
dis
pens
ing
cabi
net w
ith w
arfa
rin.
Wro
ng d
ose/
stre
ngth
of
war
farin
dis
pens
edDi
spen
sing
err
or: W
rong
dos
e/st
reng
th se
lect
ed o
r en
tere
d du
ring
data
ent
ry o
f a w
arfa
rin p
resc
riptio
n.W
arfa
rin d
ispe
nsed
w
ith w
rong
dire
ctio
ns
for u
se
Pres
crib
ing
erro
r: W
arfa
rin p
resc
riptio
n in
clud
ed
dire
ctio
ns to
take
the
drug
mor
e of
ten
than
dai
ly.
War
farin
dis
pens
ed
to th
e w
rong
pat
ient
Disp
ensi
ng e
rror
: War
farin
pre
scrip
tion
ente
red
into
the
wro
ng p
atie
nt’s
dru
g pr
ofile
.W
arfa
rin d
ispe
nsed
to
the
wro
ng p
atie
nt
Disp
ensi
ng e
rror
s: (1
) War
farin
vial
pla
ced
in a
bag
co
ntai
ning
ano
ther
pat
ient
’s m
edic
atio
ns. (
2) W
rong
pa
tient
’s m
edic
atio
n(s)
sele
cted
from
the
will
-cal
l ar
ea a
t the
poi
nt o
f sal
e.M
etho
trexa
te
Met
hotre
xate
dis
-pe
nsed
with
dire
c-tio
ns to
take
dai
ly
Pres
crib
ing
erro
r: Or
al m
etho
trexa
te fo
r non
onco
-lo
gic
use
pres
crib
ed w
ith d
irect
ions
to ta
ke th
e dr
ug
daily
.
Drug
: Met
hotre
xate
is a
n or
al a
ntin
eopl
astic
age
nt a
lso
com
mon
ly
used
in le
ss fr
eque
nt d
ose
inte
rval
s to
treat
oth
er im
mun
e-m
odul
ated
di
seas
es (e
.g., r
heum
atoi
d ar
thrit
is, p
soria
sis)
. Ini
tiatin
g er
ror:
Sele
ct-
ed b
ecau
se d
aily
dos
es th
at e
xcee
ded
5 day
s hav
e be
en fa
tal.63
,64
Fent
anyl
pat
ches
Inco
rrec
t or i
napp
ro-
pria
te d
ose
of fe
ntan
-yl
pat
ches
dis
pens
ed
to a
pat
ient
Pres
crib
ing
erro
r: In
corr
ect d
ose
or in
appr
opria
te
dose
pre
scrib
ed fo
r a p
atie
nt b
ased
on
opio
id to
ler-
ance
and
type
/dur
atio
n of
pai
n.
Drug
: Fen
tany
l tra
nsde
rmal
syst
em o
f del
iver
ing
opio
id p
ain
med
ica-
tion
expo
ses p
atie
nts t
o ov
er se
datio
n, re
spira
tory
dep
ress
ion
and
arre
st. I
nitia
ting
erro
r: Se
lect
ed b
ecau
se fa
talit
ies h
ave
happ
ened
re
peat
edly
afte
r pre
scrib
ing
dose
s too
hig
h fo
r opi
oid-
naiv
e pa
tient
s or
whe
n us
ing
the
drug
to tr
eat a
cute
, not
chr
onic
, pai
n.80
,81
Insu
lin a
nalo
gs
Wro
ng in
sulin
ana
log
disp
ense
d to
a p
atie
nt
Disp
ensi
ng e
rror
: Wro
ng in
sulin
ana
log
sele
cted
fro
m th
e sc
reen
dur
ing
data
ent
ry o
f an
insu
lin p
re-
scrip
tion.
Drug
: Ins
ulin
is a
com
mon
ly p
resc
ribed
inje
ctab
le d
rug
used
to tr
eat
type
1 di
abet
es, a
pre
vale
nt c
hron
ic ill
ness
that
affe
cts 7
00,0
00 A
mer
i-ca
ns.82
It is
one
of t
he m
ost c
omm
on c
hron
ic d
isea
ses i
n ch
ildre
n an
d ad
oles
cent
s.83
Initi
atin
g er
ror:
Sele
cted
due
to fr
eque
ncy o
f rep
orte
d m
ixup
s bet
wee
n in
sulin
pro
duct
s with
look
-alik
e na
mes
and
the
seri-
ous a
dver
se e
ffect
s of t
hose
mix
ups.
84,8
5
Abbr
evia
tion
used
: PAD
E, p
reve
ntab
le a
dver
se d
rug
even
t.
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org588 • JAPhA • 52:5 • S e p /O c t 2012
Event trees are particularly well suited for displaying the order of events and the dependencies between events, such as when the failure of subsystem B may depend on the status of subsystem A.51 To cite a simplified example, subsystem F fails (e.g., patient receives wrong medication) given that the initi-ating event has happened (e.g., pharmacy technician selected
wrong product when filling prescription) and subsystem A has succeeded (e.g., pharmacist checked final prescription), sub-system B has failed (e.g., pharmacist did not capture error dur-ing product verification), subsystem C has succeeded (e.g., pa-tient was counseled when picking up prescription), subsystem D has failed (e.g., prescription vial not opened to view tablets
Medication given to wrongcustomer
Gate 2
Q: 0.0034
Wrong customer’s medications selected by pharmacy atthe point of sale when dispensing medication
Event 1
Q: 0.003
Medication was placed inwrong customer’s bag
Event 2
Q: 0.0004
Identification error not caught whenfollowing customer identification process
Gate 4
Q: 0.0005
Exposure rate for followingcustomer identification process
Event 4
Q: 0.5
Pharmacy staff fail to detect the error whenfollowing customer identifcation process
Event 5
Q: 0.001
Exposure rate for not followingcustomer identification process
Event 6
Q: 0.5
Pharmacy staff fail to detect the error whencustomer identifcation process does not occur
Event 7
Q: 0.9
Identification error not caught when customeridentification process not followed
Gate 5
Q: 0.45
Pharmacy staff do not detectidentification error at point of sale
Gate 6
Q: 0.45
Customer does not catchidentification error at point of sale
Event 3
Q: 0.9
Wrong customer not detectedat the point of sale
Gate 3
Q: 0.405
Medication dispensed to wrongcustomer at the point of sale
Gate 1
Q: 0.00138
Top level event
And gate
Or gate And gate
Initiating errors Or gate Basic event
And gate And gate
Exposure rate Basic event Exposure rate Basic event
Figure 1. Illustration of fault tree for dispensing medication to wrong customerAbbreviations used: PADE, preventable adverse drug event. Example of a small section of a fault tree associated with dispensing a prescription to wrong customer (e.g., patient, family member, friend, caregiver). Events 1 and 2 represent the initiating errors for one pathway leading to the PADE (top-level event in gate 1) of dispensing a medication to the wrong customer. The probability of selecting the wrong customer’s medications (event 1) was estimated to occur with 3 of 1,000 (0.003) prescriptions. The probability of placing the medication in the wrong patient’s bag (event 2) was estimated to oc-cur with 4 of 10,000 (0.0004) prescriptions. These initiating errors were combined with an “or” gate (gate 2), meaning that one or the other must happen for the medication to be given to the wrong customer. Reading from the bottom of the tree, from left to right, events 4 and 6 represent exposure rates for adherence to the patient identification process. In the example, the probabilities were set at 0.5 for each, meaning that 50% of the time, the patient identification process is followed and 50% of the time it is not followed. Events 5 and 7 describe how often pharmacy staff fail to notice that the wrong patient’s medications are in the bag. When following the identification process (event 5), the probability of failing to notice the error was estimated to occur in 1 of 1,000 (0.001) prescriptions. But when the patient identification process is not followed (event 7), the probability of failing to notice the error was estimated to occur in 9 of 10 (0.9) prescriptions. Events 4 and 5 and events 6 and 7 are connected with “and” gates (gates 4 and 5) because both of the basic events below them must occur for the gates to be true. The fault tree software calculated the combined effects of how often the identification process is followed (events 4 and 6) and the estimated rate of failing to detect the error (events 5 and 7). Gates 4 and 5 are connected with an “or” gate (gate 6) because the error was not detected either when following the patient identification process or not following the identification process. Again, the fault tree calculated the combined effects of gates 4 and 5 to arrive at the probability expressed in gate 6. Event 3 to the right of gate 6 shows the probability that the customer will fail to catch the error at the point of sale: 9 of 10 (0.9) opportunities. Pharmacy staff inability to capture the error (gate 6) was then combined with the customer’s inability to capture the error (event 3) through an “and” gate (gate 3) because both failures need to happen for the error to continue through the dispensing process and reach the patient. Gate 3, which expresses the combined effects of the two initiating errors, and gate 4, which expresses the combined effects of inability by pharmacy staff and customers to capture the error, are then combined with an “and” gate to reach the top level event (gate 1). In this example, for illustrative purposes only, the medication dispensed for the wrong patient at the point of sale is estimated to occur in 1.4 of 1,000 (0.00138) prescriptions.
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 589
to aid in error detection), and subsystem E has failed (e.g., pa-tient does not detect error at point of sale).
The event trees went through multiple iterations until the modeling team was satisfied that they accurately captured the components of the dispensing process and dependencies among the different tasks.
Quantify event rates. The modeling team quantified the probability of failure or frequency of occurrence for each basic event in the event trees. The data used to support the quantifi-cation process came from documented component-specific in-formation (e.g., rates of barcode scanning overrides for a par-ticular drug), generic reference points (e.g., well-established
Table 2. Human error probabilitiesDescription of error probabilities Error probabilityHigh probability of error
Unfamiliar task performed at speed with no idea of likely consequences86 0.5Failed task involving high stress levels54 0.3Inspection/verification of tasks with moderate stress87,86 0.2Failed complex task requiring high level of comprehension and skill86 0.15Failed task involving complex math computation88 0.15Failed task conducted in the first 30 minutes of an emergency54 0.1Failure to detect an error after it has happened89 0.1Fairly simple task performed rapidly or given scant attention86 0.1
Moderate probability of error Misidentify/misdiagnose given like symptoms/appearance86 0.05Failure to select ambiguously labeled control/package89 0.05Failure to perform a check correctly90 0.05Wrong conclusion drawn with competing/unclear information86 0.05Failed execution of maintenance/repair86 0.04Failed task with cognitive or task complexity86 0.03Failure to act correctly after the first few hours in a high-stress situation54 0.03 Symptoms noticed, but wrong interpretation86 0.03Failed task related to values/units/scales/indicators86 0.02Failed task related to selection of items from among groups of items88 0.02Failed routine, highly practiced, rapid task, involving a relatively low level of skill86 0.02General mental slip without knowledge deficit91 0.02Failed task related to known hazards/damage86 0.02Failed communication among workers91 0.02Failed task involving both diagnosis and action86 0.01Failed diagnosis task86 0.01Error in a routine operation where care is required54 0.01 Set a switch in wrong position89 0.01
Low probability of error Procedural omission86 0.006Errors during read-back91 0.005Counting/volume errors88 0.004Selection of the wrong control/package (well labeled)89 0.003Operate spring-loaded switch until proper position reached89 0.003Selection of the wrong switch/package (dissimilar in shape/appearance)54 0.001
Lowest limits of human error Completely familiar, well-designed, highly practiced, routine task occurring several times per hour, performed to high-est possible standards by a highly motivated, highly trained, and experienced person, totally aware of implications of failure, with time to correct potential error but without the benefit of important job aids86 0.0004Human-performance limit: single person working alone54 0.0001Responds correctly to system command when there is an augmented or automated supervisory system providing ac-curate interpretation of system state86 0.00002Human performance limit: team of people performing a well-designed task54 0.00001
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org590 • JAPhA • 52:5 • S e p /O c t 2012
equipment failure rates that could be reasonably assumed for pharmacy dispensing equipment), and expert opinion.
The failure rate estimates needed as inputs in the event trees were obtained using Bayesian statistical methods, which work directly with estimated probabilities, rather than classi-cal statistical methods, which work primarily with counts of data. Judgment and expert opinions are required with all PRAs and ST-PRAs because the available data about components of the systems are not of the precise form required for use of clas-sical statistical methods.52
Most health care practitioners do not have actual rate data for the underlying basic events.2 At best, data collection sys-tems only capture the end result, with the rate of intermedi-ate failures relatively unknown.31 Failures, at-risk behaviors, and adverse events are so underreported in health care that using any data sources relying on reporting systems is sus-pect. In addition, PADEs can be masked by the patient's illness and thereby underestimated in occurrence data.2 Thus, some components of the event trees were developed purely through Bayesian methods, which provided a formal and rigorous way of combining expert judgments with observed data to obtain a probability. These probabilities then were propagated in the ST-PRA models to express the likelihood that a particular event would happen and the conditional uncertainty associated with that event.28,31,51,53
The substantive accuracy of the ST-PRA depends on how well the assessors know the problem under consideration. Thus, the modeling team was led by experts in human fac-tors, probability theory, ST-PRA modeling, and medication safety. The pharmacists and technicians possessed deep do-main knowledge of the processes under assessment. Internal pharmacy operational data verified the team's estimates of ex-posure rates (e.g., how often technicians enter prescriptions into the computer, how often prescriptions are received via fax, percent of a specific drug filled via automation). Pharmacists and technicians relied on their work experiences regarding frequencies of at-risk behaviors. Evidence shows that expert opinion–based probabilities are biased toward the low values of failure rates.52 Therefore, team facilitators anchored the group estimates of human error rates on data reported in the literature, setting lower- and upper-bound human error prob-ability limits as reference points for specific conditions. Table 2 summarizes the human error probabilities that helped inform and verify team estimates.
Numerous sophisticated techniques have been used since the early 1980s to estimate probabilities of human error.54 Although the formulas and tables for estimating human error probabilities vary from technique to technique, each factors in error type and performance-shaping factors (PSFs) to make judgments about error rates. Examples of common PSFs can be found in online Appendix 5. PSFs have a positive or nega-tive effect on performance. For example, staff training can influence performance either positively (e.g., when training emphasizes the appropriate learned responses) or negatively (e.g., when training is absent). The modeling team referenced the FMEAs to uncover the most relevant PSFs before making estimates.
In very general terms, given a human performance limit of 0.0001 (10−4 or 1/10,000) for a single worker operating in absolutely ideal conditions,54 the modeling team often started with an error rate of 0.001 (10−3 or 1/1,000) to account for the negative influence of a single PSF such as time constraints. Identification of additional PSFs (e.g., illegible prescriptions, look-alike product names, complex tasks, minimal worker training) was part of the group process. The number of PSFs and their degree of influence helped the team adjust its esti-mates upward or downward through an iterative process be-fore deciding on a final probability. The team quickly gained comfort in the task of estimating error and at-risk behavior probabilities, which is typical of ST-PRA modeling teams.30,36,37 Experience indicates that these team estimates are more ac-cepted than rates derived from event data and are often more accurate than rates predicted by senior management.36 More information on ST-PRA and the risk modeling process are available in Marx and Slonim36 and Comden et al.30
Model validation processObservations were conducted at two pharmacies to validate the event trees’ representation of the dispensing process and confirm the presence or absence of visible process risks (e.g., infrequent patient counseling) and PSFs (e.g., look-alike prod-ucts next to each other). The observations also served to un-derstand the relationship and dependencies among the various components of the dispensing systems and to validate that the structure of the event trees accurately represented the “as is” dispensing process at the participating pharmacies.
A survey instrument about exposure rates, capture oppor-tunities, at-risk behaviors, and failures most predictive of the PADEs was constructed and tested for interviewing pharma-cists who did not participate in the modeling sessions. Prob-abilities estimated by the modeling group were not shared with the validation group. Well-constructed published studies of drug mishaps were examined to provide, where possible, evidence to support the probability estimates derived for the PADEs and initiating errors in the event trees.
Quantifying the impact of risk-reduction interventions (sensitivity analysis)Event tree software55 calculated combinations of failures and the total combined probability of occurrence of each PADE. All unique combinations that could lead to PADEs were identified and ranked, producing cut sets or a “risk portfolio” for each event tree.30 The portfolios defined which components of the dispensing system were truly important to risk in that they con-tributed most frequently among all of the different sequences of events that could lead to PADEs. The portfolios then were used to identify the best interventions to reduce the probability of errors and at-risk behaviors or to change the tree's structure by building into the process new opportunities for capturing errors. After the interventions were identified, the event trees were updated to test and quantify the impact of each strategy.
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 591
Resultscommunity pharmacy high-alert medicationsTable 3 lists the drugs that were identified as high-alert medi-cations dispensed from community pharmacies. Warfarin, fentanyl systems, oral methotrexate, and insulin analogs were selected for ST-PRA modeling. Examples of risk factors with these targeted drugs can be found in online Appendix 6.
Validation of event treesBased on observations and survey findings, 2 of 52 probability estimates associated with at-risk behaviors were adjusted be-cause of minor differences between modeling team and valida-tion group estimates. No changes were made to 306 exposure rates or 211 failure rates, as modeling team and validation group estimates were very similar. No changes occurred as a result of comparison with error rates in published studies. No changes were made to the structure of the event trees, which were determined to be accurate.
Risk of PaDes reaching patientsThis study produced 10 event trees for PADEs associated with warfarin, fentanyl systems, oral methotrexate, and insulin analogs. These 10 event trees produced more than 200,000 failure pathways that could lead to PADEs. Table 4 shows the estimated rate of PADEs reaching patients for each event tree. These rates include all errors that reach patients after they leave the pharmacy counter, even if patients discover the error after leaving the pharmacy and do not take or use any of the erroneous medications. PADEs with the highest incidence in-cluded dispensing the wrong dose/strength of warfarin because of a data entry error (1.83/1,000 prescriptions), dispensing
warfarin to the wrong patient (1.22/1,000 prescriptions), and dispensing an inappropriate fentanyl system dose because of a prescribing error (7.30/10,000 prescriptions). PADEs with the highest incidence were associated with single-pathway failures, meaning that no key opportunities were available to capture the error from the time it was made until it reached the patient. PADEs with the lowest incidence included dispensing the wrong drug when filling a warfarin prescription (9.43/1 bil-lion prescriptions) and dispensing the wrong dose when filling a warfarin prescription (9.25/10 million prescriptions). PADEs with the lowest incidence were associated with consistent use of barcode scanning technology.
A sensitivity analysis conducted to evaluate the impact of selected interventions (Table 5) showed that the largest quan-tifiable reductions in risk were provided by (1) consistently using barcoding technology (up to 100% reduction), (2) build-ing computer alerts that can't be bypassed easily (up to 100% reduction), (3) conducting a second data entry verification process during product verification (50–87% reduction), (4) increasing patient counseling (27–68% reduction), (5) open-ing the bag at the point of sale (56% reduction), and using tall man letters to distinguish insulin products with similar names (50% reduction). Combining two or more interventions result-ed in further overall reduction in risk. Further description of the analyses of several PADEs follows.
Wrong warfarin dose/strength dispensed because of data entry errorEvent tree analysis: High vulnerability of data entry er-rors. Initially entering a wrong dose or strength of warfarin in-to the patient's profile during data entry was estimated to occur
Table 3. High-alert medications in community pharmacyDrug class/category ExamplesAntiretroviral agents Abacavir, atazanavir, diaveridine, lamivudine, ritonavir, zidovudine. Combination
products such as Combivir, Atripla, Epzicom, KaletraChemotherapy, oral (exclusion: hormonal agents) Busulfan, chlorambucil, cyclophosphamide, lomustine, melphalan, mercaptopu-
rine methotrexate, procarbazine, temozolomideHypoglycemic agents, oral Chlorpropamide, glipizide, glyburide, repaglinideImmunosuppressant agents Azathioprine, cyclosporine, daclizumab, mycophenolate, sirolimus, tacrolimusInsulin NPH/regular, aspart, detemir, glargine, glulisine, lisproOpioids, all formulations Butorphanol, fentanyl, hydromorphone, meperidine, methadone, morphine,
opium tincture, oxycodonePregnancy Category X drugs Atorvastatin, bosentan, estazolam, isotretinoin, simvastatin, temazepamPediatric liquid medications that require measurementIndividual drugsCarbamazepineChloral hydrate liquid (for sedation of children)Heparin (unfractionated and low molecular weight)MetforminMethotrexate (nononcologic use)Midazolam liquid (for sedation of children)PropylthiouracilWarfarin
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org592 • JAPhA • 52:5 • S e p /O c t 2012
with 1 in 10 warfarin prescriptions. Of these, 1.83 data entry errors per 1,000 prescriptions reached patients. The dispens-ing system is vulnerable to this type of data entry error because nine different strengths of warfarin tablets exist from which to choose. These errors are rarely corrected if data entry verifica-tion and patient counseling do not occur.
During data entry, pharmacy staff have a greater chance of detecting the error if the patient had previous warfarin pre-scriptions filled at that pharmacy. But failure to detect the er-ror is high (75%) given a 90% probability that these patients will have multiple strengths of warfarin in their drug profile history.
An independent data entry verification process by a phar-macist who has not entered the prescription was estimated to capture up to 99% of errors if an out-of-range dose alert oc-curs and is not bypassed. However, dose alerts are not likely to occur if the wrong strength tablets are selected during data entry. Duplicate therapy alerts occur during data entry with about 80% of warfarin prescriptions but are not reliable as a means of detecting a data entry error. Patient counseling was estimated to occur with 30% of patients picking up warfarin
prescriptions. If the prescription bottle is opened during coun-seling, a patient who knows what color tablets to expect has a 99% chance of capturing the data entry error. However, the bottle is only opened about 30% of the time.
Sensitivity analysis: Impact of data entry verification and patient counseling. We determined the impact of four in-terventions on the incidence of dispensing the wrong warfarin dose because of a data entry error: (1) reducing the incidence of a skipped, rushed, or inattentive data entry verification pro-cess by 50%, (2) increasing patient counseling from 30% to 80%, (3) more frequent (90%) independent verification by an-other pharmacist of prescriptions entered by pharmacists, and (4) the addition of a second data entry verification process dur-ing the product verification step. The most effective interven-tions involved the second data entry verification process and patient counseling.
Increasing patient counseling to 80% resulted in a 67% re-duction in dispensing the wrong warfarin dose because of data entry error; errors that reached patients decreased from 1.83 to 0.6 per 1,000 prescriptions. Conducting a second data entry verification process during product verification by comparing
Table 4. Probabilities of PADEs for warfarin, fentanyl systems, methotrexate, and insulin analogs
Medication, PADEInitiating error rate per
1,000 prescriptions
Capture before reaching patients
%
Rate of PADEs reach-ing patients per 1,000
prescriptions
No. PADEs reaching patients annually among all U.S. com-
munity pharmacies (n = 56,000)Warfarina Prescribing error: wrong dose 10 94.3
0.569 (5.69/10,000) 15,022
Prescribing error: wrong direc-tions
2 99.9 0.0001 (1.34/10 million) 4
Data entry error: wrong dose 100 98.2 1.83 (1.83/1,000) 48,312Filling error: wrong drug
Automated dispensing, 0.1; manual dispensing, 1; combined rate, 1
99.9
0.000009 (9.43/1 billion)
0.25 (once every 4 years)
Filling error: wrong dose
Automated dispensing, 0.1; manual dispensing, 100; combined rate, 100
99.9
0.0009 (9.25/10 million)
24
Fentanyl transdermal patchesb Prescribing error: wrong dose 1 27.0 0.730 (7.30/10,000) 3,431
Methotrexatec Prescribing error: wrong direc-tions
1 99.9 0.0009 (9.64/10 million) 4
Insulin analogsd Data entry error: wrong drug 10 96.9 0.306 (3.06/10,000) 6,426
All prescription medicationse Data entry error: wrong patient 5 99.0
0.052 (5.15/100,000) 197,849
Point-of-sale error: wrong patient
Select wrong patient’s bag, 3; place in wrong patient’s bag, 0.4; com-bined rate, 3.4
64
1.22 (1.22/1,000)
4,641,856
Abbreviation used: PADE, preventable adverse drug event. aAnnual prescription volume for all U.S. community pharmacies: 26,400,000 (2007).72–74 bAnnual prescription volume for all U.S. community pharmacies: 4,700,000 (2007).72–74
cAnnual prescription volume for all U.S. community pharmacies: 4,400,000 (2007).72–74
dAnnual prescription volume for all U.S. community pharmacies: 21,000,000 (2007).72–74
eAnnual prescription volume for all U.S. community pharmacies: 3,804,800,000 (2007).72–74
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 593
Table 5. Sensitivity analysis of selected interventions to reduce PADEs
Evaluated interventionsa
Errors before action, per 1,000
prescriptionsErrors after action, per
1,000 prescriptions
Decrease in risk (increase in risk)
%PADE: Medication dispensed to the wrong patient due to a bagging er-ror or bag selection error at the point of sale 1.22 (1.22/1,000)A: Open the bag at the point-of-sale to view all filled prescriptions 0.534 56B: Increase adherence with following a patient identification process from 50% to 80% 0.804 34C: Increase patient counseling from 30% to 50% 0.889 27D: Reduce at-risk behavior of working on more than one patient’s medi-cations during product verification and bagging (which lowers the bag-ging error rate from 0.4 to 0.1 per 1,000 prescriptions) 1.11 9Action A and action B 0.233 81Action A and action B and action C 0.169 86Action A and action B and action C and action D 0.154 87PADE: Medication dispensed to the wrong patient caused by entering the prescription into the wrong profile
0.052 (5.15/100,000)
A: Reduce at-risk behavior of conducting inattentive data entry verifica-tion from 1 in 10 to 5 in 100 prescriptions (requires changes in the system/environment to support a consistent, cognitive checking process) 0.034 35B: Increase patient counseling from 30% to 50% 0.037 29C: Increase the frequency of an independent double-check for data entry verification when a pharmacist enters prescriptions from 50% to 90% 0.024 17D: Conduct a second redundant data entry verification during the final product verification step 0.007 87E: Reduce the incidence of entering prescriptions into the wrong patient profile from 5 to 1 per 1,000 prescriptions by requiring entry of two unique patient identifiers (name, birth date) in the patient profile 0.010 81Action A and action B 0.024 54Action A and action B and action C 0.020 62Action A and action B and action C and action D 0.002 96Action A and action B and action C and action E 0.004 92Action A and action B and action C and action D and action E 0.0004 99PADE: Wrong or inappropriate dose of fentanyl patches dispensed due to a prescribing error
0.730 (7.30/10,000)
A: Conduct an intake history of opioids when receiving a prescription for fentanyl patches; pharmacist review history before data entry (which results in capture of 40% of prescribing errors) 0.439 40B: Increase patient counseling from 10% to 80%, and increase the ability to recognize inappropriate doses from 10% to 80% by reviewing the pa-tient’s opioid history during counseling session 0.263 64Action A and action B 0.159 78PADE: Methotrexate dispensed with directions to take daily instead of weekly due to a prescribing error
0.0009 (9.64/10 million)
A: Include a diagnosis or indication for use on the prescription 0.0007 22B: Set dose alert as a hard stop that does not allow the entry of metho-trexate prescriptions with daily dosing for more than 1 consecutive week 0.00000001 (1/100 billion) 100C: Eliminate computer warning about daily dosing of methotrexate 0.006 (522)Action A and action B 0.00000001 (1/100 billion) 100PADE: Wrong insulin analog dispensed due to selecting the wrong drug during data entry
0.306 (3.06/10,000)
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org594 • JAPhA • 52:5 • S e p /O c t 2012
Table 5. Sensitivity analysis of selected interventions to reduce PADEs
Table 5 continuedA: Reduce the frequency of misreading prescriptions for insulin products with similar names by increasing electronic prescriptions by 20% (lowers initiating error rate from 10 to 8 per 1,000 prescriptions) 0.245 20B: Reduce the rate of selecting the wrong insulin product during order entry by using tall man letters to distinguish products with similar names (lowers initiating error rate from 10 to 5 per 1,000 errors) 0.153 50C: Increase frequency of patient counseling from 30% to 80% 0.100 67D: Conduct a second redundant data entry verification during the final product verification step 0.153 50Action A and action B and action C 0.028 91Action A and action B and action C and action D 0.014 95PADE: Wrong drug or dose dispensed due to a selection error while filling a prescription for warfarin
(1) Wrong drug, 0.000009 (9.43/1 billion); (2) wrong dose, 0.0009 (9.25/10 million)
A: Eliminate barcoding technology during the dispensing process (1) 0.009 (9/1 million), (2) 0.900 (9/ 10,000)
(1) (95,340), (2) (97,197)
B: Use a cheat sheet to scan a barcode for warfarin 30% of time rather than scanning the bar code on the stock bottle/carton
(1) 0.025 (2.5/100,000, (2) 0.200 (2/10,000)
(1) (265,011), (2) (21,521)
C: No pill image on the product verification screen (and label) (1) 0.00004 (4 /100 million), (2) 0.004 (4/1 million) (1) (324), (2) (332)
D: Increase the automated filling of warfarin prescriptions from 20% to 50%
(1) 0.000007 (7/1 billion), (2) 0.0006 (6/10 million) (1) 25, (2) 35
E: Increase the frequency of patient counseling from 30% to 80% (1) 0.000003 (3/1 billion, (2) 0.0003 (3/10 million) (1) 68, (2) 67
Action A and action C (1) 0.042 (4.2/100 thou-sand), (2) 4.20 (4.2/1,000)
(1) (445,287), (2) (453,954)
Action D and action E (1) 0.000002 (2/1 billion) (2) 0.0002 (2/10 million) (1) 78, (2) 78
PADE: Wrong dose/strength of warfarin tablets dispensed due to a pre-scribing error
0.569 (5.69/10,000)
Increase patient counseling from 30% to 80% 0.274 52PADE: Wrong warfarin dose dispensed due to a data entry error 1.83 (1.83/1,000)A: Reduce at-risk behavior of conducting inattentive data entry verifica-tion from 1 in 10 to 5 in 100 prescriptions (requires changes in the system/environment to support a consistent, cognitive checking process) 1.19 35B: Increase patient counseling from 30% to 80% 0.600 67C: Increase the frequency of an independent double-check for data entry verification when a pharmacist enters prescriptions from 50% to 90% 0.865 53D: Conduct a second redundant data entry verification during the final product verification step 0.366 80Action A and action B 0.393 79Action A and action B and action C 0.283 85Action A and action B and action C and action D 0.174 91PADE: Warfarin prescription dispensed with the wrong directions due to a prescribing error
0.0001 (1.34/10 million)
Make data entry for more frequent than daily dosing of warfarin result in an alert with a hard stop 0.000000001 (1/1 trillion) 100Increase patient counseling when picking up prescriptions from 30% to 80% 0.00005 (5/100 million) 50Abbreviation used: PADE, preventable adverse drug event. aMost evaluations measure the positive impact of increasing an existing risk-reduction strategy or implementing a new risk-reduction strategy. A few of the evaluations measure the negative impact of reducing or eliminating an existing risk-reduction strategy.
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 595
the scanned prescription in the computer with the prescription label reduced the risk of this error reaching patients by 80%. More frequent independent checks and less skipped, rushed, or inattentive checks during data entry verification reduced the risk of PADEs by 53% and 35%, respectively. All four interven-tions together lowered the risk of dispensing the wrong warfa-rin dose from 1.83 to 0.174 per 1,000 prescriptions.
Wrong drug or dose dispensed because of selection error while filling prescription for warfarinEvent tree analysis: Vulnerability of selecting wrong dose higher than wrong drug. Initially selecting the wrong drug while manually filling a warfarin prescription was estimated to occur in 1 of 1,000 prescriptions. Mixups between warfarin and another medication have been reported rarely, although a risk exists with the branded warfarin product Jantoven (Upsh-er-Smith), which could be confused with Januvia or Janumet.56 However, pharmacies often stock warfarin on shelves accord-ing to generic names, thereby lessening the risk of such an er-ror.
Initially selecting the wrong dose while manually filling a prescription for warfarin was estimated to occur more fre-quently (1 of 10 prescriptions). This estimate is in line with a study that found that more than 5% of medications first se-lected to fill a prescription were wrong,57 as more errors are expected with warfarin doses because of nine different tablet strengths. The 1-mg and 10-mg strengths also are prone to mixups, particularly if a trailing zero is used to express the 1-mg dose (1.0 mg) on pharmacy or product labels.58
The probability of an error reaching the patient is 9.43 per 1 billion warfarin prescriptions for wrong drug errors and 9.25 per 10 million prescriptions for wrong dose errors. These low estimates are primarily the result of using barcoding technol-ogy while manually filling prescriptions and the availability of a tablet image for product verification. We estimated that technology would fail to capture the error in 1 of 100,000 op-portunities to account for an occasional problematic barcode or scanner malfunction. If the error is not picked up through barcoding, an image of the correct drug and dose on the screen during product verification facilitates capture of the error in 99 of 100 occurrences. The final opportunity to capture these er-rors is during patient counseling, which was estimated to occur 30% of the time, mostly for patients with a new prescription or dose change. If the bottle label is viewed and the bottle is opened during the counseling session, the chance of capturing the error during this process step was estimated to increase from 90% (bottle not opened) to 99% (bottle opened), as phar-macists and patients often know what color tablet to expect for a given strength. The impact of this intervention is lessened in the overall estimate of PADE occurrence because patient counseling does not always occur.
Sensitivity analysis: Impact of barcode product veri-fication, automated dispensing, tablet imaging, and pa-tient counseling. Our event trees for warfarin drug and dose selection errors add evidence to existing knowledge about the
effectiveness of using barcoding technology during dispens-ing.16,17,59–61 With the technology, 99.9% of selection errors were detected and corrected. However, without it, the prob-ability of dispensing the wrong drug increased from 9.43 per 1 billion to 9 per 1 million, and the probability of dispensing the wrong dose increased from 9.25 per 10 million to 9 per 10,000. Similar increases were seen if an image of the correct tablet was not available during product validation. When barcoding and tablet imaging are absent, the probability of dispensing the wrong drug increased from 9.43 per 1 billion prescriptions to 4.2 per 100,000 prescriptions and the probability of dispens-ing the wrong dose increased from 9.25 per 10,000,000 pre-scriptions to 24.2 per 1,000 prescriptions. We also evaluated the impact of increasing automated dispensing of warfarin from 20% to 50% and increasing patient counseling from 30 to 80%, which reduced the risk of allowing either a wrong drug or wrong dose dispensing error to reach the patient by 78%.
Incorrect or inappropriate dose of fentanyl systems dispensed because of prescribing errorEvent tree analysis: Dispensing system unreliable in de-tecting prescribing errors. Prescribing an incorrect or inap-propriate dose of fentanyl systems was estimated to occur in 1 of 1,000 prescriptions. Fentanyl systems that are prescribed to treat acute pain (not an approved indication) and/or prescribed to opioid-naive patients were classified as an incorrect or inap-propriate dose, as were doses that exceeded safe limits based on the patient's previous opioid use, age, general medical con-dition, conditions associated with hypoxia or hypercapnia, and/or concomitant analgesics. Including all of these wrong dose prescribing errors resulted in dispensing 0.73 wrong doses per 1,000 prescriptions.
The dispensing system in participating pharmacies was largely unreliable in its ability to detect this prescribing er-ror; only 27% of the errors were estimated to be captured and corrected. The low capture rate was primarily associated with inadequate knowledge about the patient's prior opioid use, un-derlying health conditions, and type of pain for which the fen-tanyl system had been prescribed.
Drug use review and patient counseling are two steps dur-ing which fentanyl system prescribing errors can be captured, but with limited success. The acceptable dose range for fentan-yl systems is wide, from 12.5 μg/hour to 100 μg/hour or more, depending on the patient's opioid tolerance and pain level. Doses up to 300 μg/hour are recommended for patients with a 24-hour intake history of oral morphine in doses from 1,035 to 1,124 mg/day.62 An out-of-range dose alert would occur in just 1 of 100 prescriptions, as the computer cannot detect an inappropriate dose within such a wide acceptable therapeutic range. If the computer issues an out-of-range dose alert, the modeling team estimated that the error would be detected 98% of the time, but the low rate of dose alerts makes this an unreli-able capture opportunity.
The frequency of patient counseling was estimated to be 10% because many patients on fentanyl systems do not pick up their prescriptions and counseling is often declined by the
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org596 • JAPhA • 52:5 • S e p /O c t 2012
caregiver or friend picking up the prescription. When counsel-ing occurs, focusing on how to use the system correctly rather than appropriateness of the dose leads to an estimated capture of the dosing error just 10% of the time.
Sensitivity analysis: Impact of obtaining opioid use history. We determined the impact of two interventions on the incidence of dispensing an inappropriate dose of fentanyl sys-tem because of a prescribing error. Both strategies involved obtaining a history of opioid use from the patient or health care provider. In the first case, we added a process step during re-ceipt of the prescription for pharmacists to request informa-tion from patients, caregivers, and/or health care providers about prior opioid use and for pharmacists to review the infor-mation (and the patient's drug profile as necessary) before the prescription is entered into the computer. The team estimated a 40% capture rate of inappropriately prescribed doses with this intervention. Even with this modest capture rate, the pro-cess change allowed capture of 56% of prescribing errors.
The second intervention involved collecting an opioid his-tory during face-to-face or phone counseling. First, the fre-quency of patient counseling was increased from 10% to 80%. The pharmacist's ability to detect an inappropriate dose dur-ing the counseling session was increased from 10% to 80% by estimating the impact of using a checklist that gathered in-formation about the patient's opioid use history. This reduced the probability of errors reaching patients from 0.730 to 0.263 per 1,000 prescriptions, which represents a prescribing error capture rate of 74%.
Oral methotrexate dispensed with directions to take daily due to a prescribing errorEvent tree analysis: Computer alert leads to a high capture rate of prescribing errors. Prescriptions for oral methotrexate with incorrect directions to take the drug daily instead of a weekly dosing schedule were estimated to occur in 1 of 1,000 prescriptions. Of these, 0.0009 errors per 1,000 prescriptions (9.64/10 million prescriptions) were estimated to actually reach patients.
Oral methotrexate intended for nononcologic use typically entails a single weekly dose sometimes spread over three dos-es every 12 hours instead of daily doses. Given that methotrex-ate doses for oncologic use do not typically exceed courses of 5 days repeated after a week with no therapy, the PADE includes all prescriptions for methotrexate that were prescribed daily for more than 1 consecutive week. Pharmacies in our study have a robust dispensing process that captures and corrects 99.9% of methotrexate prescriptions with directions to take the drug daily. The reasons for a high capture rate included computer alerts that appeared when entering a daily dose of oral methotrexate and heightened pharmacy staff awareness of this type of PADE. Such conditions may not exist in other pharmacies.
Sensitivity analysis: Impact of including indication on prescription and computer alerts. Knowing the indication could help detect a methotrexate prescription with the wrong directions for use.63,64 Estimating 80% prescriber adherence
with including the indication on the prescription and estimat-ing modest increases (≤50%) in the ability of pharmacy staff to detect the error when the indication is provided, the rate of the PADE was reduced from 0.0009 to 0.0007 of 1,000 methotrex-ate prescriptions.
We also tested the effect of making the computer alert a hard stop that does not allow entry of oral methotrexate pre-scriptions with daily dosing for more than 1 consecutive week. This lowered the incidence of dispensing methotrexate with di-rections for daily use to once every 100 billion prescriptions.
We evaluated the reliability of the dispensing process to detect the error if the computer system does not warn pharma-cy staff about daily dosing of methotrexate. Removing this im-portant safety feature resulted in more than a 500% increase in risk, with the probability of capturing the error decreasing from 0.0009 to 0.006 events per 1,000 methotrexate prescrip-tions.
Wrong insulin analog dispensed because of selecting wrong drug during data entryEvent tree analysis: Look- and sound-alike product names contribute to errors. Dispensing the wrong insulin analog as a result of data entry error was estimated to occur in 1 of 100 insulin analog prescriptions. The insulin analogs con-sidered were Humalog (Eli Lilly), Humalog Mix 75/25, Huma-log Mix 50/50, NovoLog (Novo Nordisk), NovoLog Mix 70/30, Apidra (sanofi-aventis), Lantus (sanofi-aventis), and Levemir (Novo Nordisk). The participating pharmacies were capable of capturing 96.9% of these data entry errors, leaving 0.306 er-rors per 1,000 prescriptions that reached patients. The error rate for selecting the wrong insulin analog during order entry was estimated to be higher than a typical human error rate due to several well-known PSFs (online Appendix 5), includ-ing look- and sound-alike product names, similar strengths and mixtures, handwritten prescriptions susceptible to being mis-read, and long insulin analog pick lists in the computer from which to select the correct product.
Data entry verification was found to be the most opportune time in the dispensing process to detect these errors. However, only 10% of the errors would be recognized if the pharmacist who initially entered the prescription in error also verified the prescription entry. Drug use review is not a reliable step for capturing this particular type of error. Dosing ranges overlap; therefore, out-of-range dose alerts occur infrequently. Dupli-cate therapy alerts are common and occur with about 80% of insulin prescriptions because many patients with diabetes re-quire more than one type of insulin. The frequency of duplicate insulin therapy alerts led the modeling team to estimate a 30% rate of bypassing the alert without sufficient attention and the frequency of patients receiving more than one type of insulin led to an estimated 90% failure rate for detecting the error even if the duplicate therapy alert was investigated.
The ability to detect the data entry error would increase from 90% to 99% if the counseling session included visual in-spection by the patient of the insulin carton, vial, or pen. To be
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 597
effective, this requires counseling at the counter, not the drive-through window.
Sensitivity analysis: Impact of electronic prescrip-tions, tall man letters, and data entry verification. Two of the risk-reduction interventions that were evaluated focused on reducing the initial data entry error by changing PSFs. We evaluated the impact of increasing electronic or computer-printed prescriptions by 20% to reduce the risk of misreading handwritten prescriptions for insulin products with similar names. This was estimated to lower the initiating error rate from 10 to 8 per 1,000 prescriptions, yielding a reduction in errors reaching patients from 0.306 to 0.245 per 1,000 pre-scriptions. We also evaluated the use of tall man letters to draw attention to the differences in insulin names (e.g., HumaLOG, HumuLIN, NovoLOG, NovoLIN) on computer screens, which lowered the rate of selecting the wrong insulin during data en-try from 10 to 5 per 1,000 prescriptions and lowered the prob-ability of dispensing the wrong insulin to a patient from 0.306 to 0.153 events per 1,000 prescriptions. Conducting a second data entry verification during the product verification step had a similar effect.
By increasing the frequency of patient counseling from 30% to 80%, the probability of capturing the error increased from 96.9% to 99%, which reduced the risk of dispensing the wrong insulin to a patient from 0.306 to 0.100 per 1,000 pre-scriptions.
Dispensing medication to wrong patient at point of saleEvent tree analysis: High vulnerability to dispensing pre-scriptions to the wrong patient. Pharmacies are vulner-able to dispensing correctly filled prescriptions to the wrong individual at the point of sale, which is a risk that has been substantiated in the literature.21,65–67 This PADE is not influ-enced by the attributes of a specific medication; dispensing any prescription medication to the wrong patient at the point of sale carries a similar risk of occurrence. Infrequent patient counseling and flawed patient identification procedures were the most frequent contributory factors leading to an estimated error rate of 1.22 per 1,000 prescriptions. With close to 4 bil-lion prescriptions filled annually, this error rate suggests that 386,821 prescriptions will be dispensed to the wrong patient each month or an average of seven errors per month for every U.S. pharmacy.
Two initiating errors most often led to this PADE: errors when placing the products into a bag for pick up (0.4/1,000 prescriptions) and errors when retrieving the medications at the point of sale (3/1,000 prescriptions). Bagging errors often stem from working on more than one patient's medications during the product verification and bagging process. Because the bag is not opened at the point of sale, customers rarely cap-ture the errors before leaving the pharmacy.
A flawed or absent patient identification process most of-ten led to errors when retrieving the medications at the point of sale. Although patient verification is expected at the point of sale, the modeling team reported difficulty obtaining a birth
date when prescriptions were picked up by caregivers, family, or friends. Using an address as a second identifier is subopti-mal, as patients with the same last name often live together. Modeling team members who worked in stores with lower pre-scription volumes felt that they were able to visually identify most patients; however, they were also more likely to skip a formal verification process, during which errors could occur. Unless patient counseling occurs at the point of sale, dispens-ing a prescription to the wrong patient is a single-pathway failure, meaning that no key opportunities to capture the error occurred from the time it was made to when the prescription reached the patient.
Sensitivity analysis: Impact of opening the bag at the point of sale, adherence to patient identification pro-cess, and patient counseling. A simple process change such as opening the bag at the point of sale to view the products yielded improvement, from 1.22 to 0.534 errors per 1,000 prescriptions. We tested opening the bag to view its contents, with a modest rate (90%) in capturing a wrong patient error compared with when the bag was not opened; the intervention was very effective in reducing the risk of an error reaching the patient.
The modeling team estimated that pharmacy staff followed a patient identification process at the point of sale only half of the time. The process included verifying the patient's last name along with one other unique identifier: birth date or address. With an increase from 50% to 80% in the frequency of carry-ing out the patient identification process at the point of sale, the incidence of dispensing a prescription to the wrong patient decreased from 1.22 to 0.804 errors per 1,000 prescriptions. Combining 80% adherence with following the patient identifi-cation process and opening the bag at the point of sale further lowered the incidence of this error to 0.233 errors per 1,000 prescriptions, representing 81% improvement.
Increasing the frequency of counseling patients who pick up prescriptions from 30% to 50% reduced the incidence of the error from 1.22 to 0.899 per 1,000 prescriptions. Adding this intervention to the previous two interventions—opening the bag and following the identification process 80% of the time—resulted in a further reduction to 0.169 per 1,000 pre-scriptions (86% improvement). This reduction in risk changes the rate of dispensing a prescription medication to the wrong patient at the point of sale at a U.S. pharmacy from seven per month to less than 1 per month.
DiscussionData reported in the literature are not directly comparable with the estimates in our event trees or the calculated prob-abilities of occurrences for each PADE. The incidence of com-munity pharmacy errors and PADEs for each of our targeted high-alert medications and the specific types of errors is not readily available in the literature. Most studies have identified general rates of medication errors, ADEs, and/or PADEs.
To cite one example, in 2003, Flynn et al.16 identified a community pharmacy dispensing error rate of 17.2 errors per 1,000 prescriptions. Dispensing error rates per 1,000 pre-
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org598 • JAPhA • 52:5 • S e p /O c t 2012
scriptions were also reported for wrong drug (1.2) and wrong strength (1.9), but the rates were not categorized according to the class of drugs involved in the errors. Data entry errors accounted for the largest type of dispensing error (11.2/1,000 prescriptions), but details regarding the type of data entry er-ror (e.g., wrong strength, wrong patient) were not provided, making comparison with our results unworkable. The study of Flynn et al. also showed that about 48% of data entry errors and 45% of drug selection errors were captured and corrected before reaching patients. Our results differed in that 96.9% to 99.0% of data entry errors, and virtually all drug selection errors, were captured during the dispensing process. Differ-ences in the capture of data entry errors may be explained by our exclusionary focus on four high-alert medications for which computer alerts were functional to help capture data entry er-rors. Differences in the capture of data entry errors also can be explained by our sole focus on wrong medication, wrong dose, and wrong patient data entry errors. Our study did not evaluate data entry errors associated with entering the wrong directions for use, which was an important source of errors in the Flynn et al. study. Differences in the capture of drug selection errors between the Flynn et al. study and our study were expected be-cause of consistent use of barcoding systems in all pharmacies participating in our study, which was not present in all pharma-cies in the study of Flynn et al.
Direct comparisons of our data with previous studies also were limited by differences in event definitions, event catego-ries, study settings, detection methods, and outcomes evalu-ated. Further, many previous studies do not distinguish be-tween inpatient and outpatient prescriptions; errors, PADEs, and ADEs; types of errors with each drug; or forms of the drugs involved in the errors (e.g., oral, parenteral, transdermal).
The event trees created during our study enhance findings from previous studies on medication errors by demonstrating important and largely correctable community pharmacy dis-pensing system vulnerabilities, identified by the people who work within those systems. The event trees represent the ex-pert opinions of experienced, practicing pharmacists and phar-macy technicians and define thousands of ways process fail-ures and behavioral elements combine to lead to each PADE. They provide insight into deep system weaknesses, human er-rors, and behavioral choices that define medication dispensing risks because the models incorporate operational understand-ing of system design, interdependencies, and human reliabil-ity concepts not easily visualized with traditional quantitative studies. This level of detail, which is not available from any other source, identified dispensing system vulnerabilities and a remarkable capacity to detect and correct errors before reaching patients given certain environmental, technological, system/process, and behavioral conditions that are well within the reach of most community pharmacies and pharmacy staff.
Prescribing errorsWith prescribing errors, the event trees suggest that commu-nity pharmacy dispensing systems may be designed to capture straightforward mistakes, such as prescribing warfarin twice
a day, more often than errors associated with inappropri-ate drugs or doses, such as prescribing fentanyl systems for an opioid-naive patient or warfarin in a dose not appropriate for the patient based on international normalized ratio values. Detecting the former type of error requires knowledge about the drug and/or system safeguards, such as computer alerts, to help capture these errors. Detecting the latter type of er-ror requires knowledge of the drug and knowledge about the patient's medical history and laboratory values—important information rarely accessible in community pharmacies. Lack of information about patients’ diagnoses and/or prescriptions without a specified indication for the drug further hamper the pharmacist's ability to detect errors. It is not surprising that the event trees demonstrated a 99.9% capture rate for pre-scribing errors associated with warfarin directions for use and only a 27% capture rate for prescribing errors associated with the dose of fentanyl systems, which is dependent on the pa-tient's type of pain and prior opioid tolerance.
Sensitivity analysis identified that more frequent and ef-fective patient counseling will reduce prescribing errors that reach patients by as much as 64%. ISMP has long been a staunch supporter of the need for community pharmacy access to clinical patient information and reimbursement systems that compensate the time pharmacists spend on clinical review of prescriptions and counseling.68 Improvement in these areas will considerably improve the success of detecting prescribing errors before they reach patients in the outpatient setting.
Dispensing errorsCommunity pharmacies in the study exhibited highly reliable systems for preventing and detecting drug or dose selection errors when filling prescriptions for warfarin. Reliability was driven by available technologies, including barcoding, auto-mated dispensing, scanned prescription images, and comput-er-generated tablet images. Selecting the wrong drug while filling a warfarin prescription was estimated to reach patients once every 4 years among U.S. pharmacies if barcoding tech-nology was used for verification. On the other hand, wrong dose selection errors were estimated to reach patients 24 times each year. Increasing automated dispensing of warfarin to 50% and the frequency of patient counseling to 80% lowered the probability of wrong drug selection errors to once every 20 years and wrong dose selected and dispensed to once every 5 years among U.S. pharmacies if barcoding technology was used for verification.
Community pharmacies in the current study were vulner-able to wrong drug and wrong dose data entry errors, which is consistent with previous reports.16,69,70 Sensitivity analysis re-vealed that conducting a second data entry verification process at the time of product verification reduced this risk by one-half (50%) for wrong insulin analog data entry errors and by 80% for wrong warfarin dose data entry errors.
Wrong patient errors also present a risk, with about 5 er-rors per 100,000 prescriptions originating from data entry into the wrong patient's profile and 1 error per 1,000 prescriptions originating at the point of sale. These findings are higher than
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 599
rates reported in the literature,16,21,65–67 although errors cap-tured at any point after the patient leaves the pharmacy coun-ter were included in the event tree incidence rates. Opening the bag at the point of sale to view products reduced the error risk by more than one-half (56%). In practice, privacy at the point of sale is difficult; thus, confidentiality concerns and the Health Insurance Portability and Accountability Act of 1996 Privacy Rule will affect the degree to which this intervention can be applied.
The interventions discussed and evaluated in the study are not meant to suggest and/or establish a standard of care for community pharmacies. Rather, the interventions are intended to represent future improvements that will affect patient safety positively and raise the bar in community pharmacies.
LimitationsThis study produced event trees that are representations of the dispensing systems used in community pharmacies. Although community pharmacy dispensing systems share many ele-ments in common, the study results are not generalizable to all community pharmacies because of potential differences in process steps, technology, frequency of at-risk behaviors, and PSFs that increase or decrease risk. Samples sizes were con-strained by funding limits and to avoid disrupting operations at participating pharmacies. The modeling sessions required a manageable number of participants, limiting the sample size. The small sample sizes of the model-building team and the vali-dation group could reduce the certainty of the probability esti-mates based on expert opinions, particularly for components of the dispensing system for which failure was rare and no preex-isting reference data were available to assist with estimations.
Keeping reasonable travel times for the modeling team necessitated a convenience sample from a single geographic location. To ensure diversity among the model-building and validation samples, staff representing minorities may have been overrepresented. The event trees may not include im-portant community pharmacy systems or practice variations, and estimates of PADEs could differ in pharmacies not in the study; however, the event trees represent a starting point for understanding how and why serious errors occur. We view these event trees as living models of risk that can be expanded to accommodate a broader scope of pharmacy practices and refined by replacing expert opinion estimates with known prob-abilities.
Use of expert opinions and uncertainty with sT-PRa probabilitiesMisunderstandings surround the use of ST-PRA, which uses a Bayesian statistical model that relies in part on expert opin-ions to determine probability distributions during the analytic process.51,53 Although users of conventional statistical methods rarely dispute the mathematical foundations of the Bayesian approach,71 they are unaccustomed to mixing objective data with subjective judgments and may feel that ST-PRA lacks sci-entific rigor or validity to guide decision making.52,71 This view-point is misguided. First, many important patient safety deci-
sions cannot wait until all questions can be answered with em-pirical data. Next, ST-PRAs are built on an accepted probability model and provide important information regarding the condi-tional occurrence of a particular event, given what is currently known and accepted assumptions.51–54,71 All ST-PRAs and PRAs require the use of subjective data. It is the uncertain and prob-abilistic nature of risk that requires the inclusion of subjective data.51 Instead of making inferences based purely on empirical data, with ST-PRA, inferences must be made despite uncertain parameters and missing data. Uncertainty and risk are highly interconnected. Thus, estimates of risk resulting from ST-PRAs will be uncertain, in large part because of the sparse nature of data on system components.51,53 Nevertheless, the use of sub-jective and uncertain probabilities in making decisions is the-oretically founded; it is rational for people to make decisions based on the best information available.53
All risk modeling represents a rough approximation of true risk. The uncertainty attached to individual failure rates and the presence of unidentified process variations speak to the uncertainties in risk models as a whole. The task for the risk-modeling team was to determine whether, in a restricted time frame, the ST-PRA technique would provide a more informative picture of actual risk in the community pharmacy setting than that currently available through typical sources, such as retro-spective event reporting, RCA, and FMEA. Through this work, we have provided additional knowledge and understanding of the risks imposed on the pharmacy dispensing system and the effects of various interventions on reducing these risks.
Further discussion on the Bayesian statistical model, un-certainty, and arguments that prove the mathematical rigor in using probability distributions to describe uncertainties is beyond the scope of the current report; however, interested readers are referred to Zimmerman and Bier,51 Bier,53 and Ber-nardo71 for further information.
conclusionThe ST-PRA models created during the study were exception-ally robust for identifying process and/or behavioral failures, estimating the frequency of adverse outcomes, and evaluating the effectiveness of interventions. The event trees revealed im-portant systems relationships, unintended consequences of behavioral choices, and valuable risk-reduction interventions that can guide and accelerate community pharmacy safety improvements. They serve as visible diagrams for shared un-derstanding of the failure pathways that lead to harm, which facilitates communication, shared goals, trust, and agreement between stakeholders because everyone owns the same risk model. The greatest value of the ST-PRA process often lies less in the quantitative estimates of potential adverse events and more in the qualitative, robust risk models that define the in-terdependencies and combination of failure pathways in com-plex systems that lead to adverse events. Each of the ST-PRA risk models we developed identified tens of thousands of failure pathways that could lead to a single adverse event. The risk models were then used to prioritize the process steps, behav-ioral choices, and sequence of events that most often contrib-
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org600 • JAPhA • 52:5 • S e p /O c t 2012
uted to the adverse outcome and to evaluate the impact of se-lected risk-reduction strategies. This study demonstrates the strength and value of ST-PRA and its application in health care, its advantages over current qualitative risk assessment meth-ods, its capabilities to forecast combinations of risk leading to PADEs, and how it might be used to facilitate development and implementation of high-leverage interventions to reduce medi-cation errors.
The techniques required to model human error and behav-ioral risks with ST-PRA are still advancing, and the process can be lengthy, complex, and require expert facilitation. However, the alternative to ST-PRA is to assume the system is safe, wait for events to happen, investigate, and remove the newly seen risk. This study demonstrates that ST-PRA, even with its un-certainties, can lead to learning and improvements not achiev-able by current prospective risk assessment processes or ret-rospective event investigations. The process is proving to be highly useful in health care, although the elicitation and use of expert opinions in safety studies and risk management is an area for further study.52
Although every community pharmacy conducting its own ST-PRA modeling may not be practical, application of the lessons learned from these existing models can lead to wide-spread improvement in community pharmacies nationwide. Logic dictates that the event trees will be useful across a broad range of community pharmacies, as they inform pharmacists about the systems design and behavioral elements that can produce or prevent a dispensing error.
We anticipate that the results of this study will contribute greatly to the growing body of knowledge about the application of ST-PRA in health care and lead to further exploration regard-ing how the process can be demystified and used as a practical tool in health care settings. Since 2008, ISMP and Outcome Engenuity, LLC, have been developing a streamlined, com-puter-based, community pharmacy risk assessment tool that uses the risk models developed during this study to compute quantifiable risks associated with various types of prescribing and dispensing errors. The user will be required to provide eas-ily accessible data gathered during a series of approximately one dozen to three dozen online questions. A scorecard will estimate the rate of PADEs that reach patients based on each pharmacy's unique processes, computer support, and staff be-haviors. The scorecard will also suggest risk-reduction strate-gies and allow the user to recalculate the frequency of PADEs that reach patients to demonstrate the anticipated reduction in risk with the selected risk-reduction strategies. The tool will be freely available in 2012.
Note added in proofFollowing acceptance of the manuscript, the tool High-Alert Medication Modeling and Error-Reduction Scorecards (HAM-MERS) was completed and is now available on the ISMP web-site at www.ismp.org/tools/HAMMERS.
References1. Bates DW, Spell N, Cullen DJ, et al. The cost of adverse drug
events in hospitalized patients. JAMA. 1997;277:307–11.
2. Aspden P, Wolcott JA, Bootman JL, Cronenwett LR, Prevent-ing medication errors: quality chasm series. Washington, DC: National Academies Press; 2007.
3. Bates DW, Boyle DL, Vander Vliet MB, et al. Relationship be-tween medication errors and adverse drug events. J Gen In-tern Med. 1995;10:199–205.
4. Bates DW, Cullen DJ, Laird N, et al. Incidence of adverse drug events and potential adverse drug events: implications for prevention. JAMA. 1995;274:29–34.
5. Classen DC, Pestotnik SL, Evans RS, et al. Adverse drug events in hospitalized patients: excess length of stay, extra costs, and attributable mortality. JAMA. 1997;277:301–6.
6. Porter J, Jick H. Drug-related deaths among medical inpa-tients. JAMA. 1977;237:879–81.
7. Phillips J, Beam S, Brinker A, et al. Retrospective analysis of mortalities associated with medication errors. Am J Health Syst Pharm. 2001;58:1835–41.
8. Lazarou J, Pomeranz BH, Corey PN. Incidence of adverse drug reactions in hospitalized patients: a meta-analysis of prospec-tive studies. JAMA. 1998;279:1200–5.
9. Howard R, Avery A, Bissell P. Causes of preventable drug-re-lated hospital admissions: a qualitative study. Qual Saf Health Care. 2008;17:109–16.
10. Howard RL, Avery AJ, Slavenburg S, et al. Which drugs cause preventable admissions to hospital? A systematic review. Br J Clin Pharmacol. 2007;63:136–47.
11. Budnitz DS, Pollock DA, Weidenbach kN, et al. National sur-veillance of emergency department visits for outpatient ad-verse drug events. JAMA. 2006;296:1858–66.
12. Ghandi Tk, Weingart SN, Borus J, et al. Adverse drug events in ambulatory care. N Engl J Med. 2003;348:1556–64.
13. Buchanan TL, Barker kN, Gibson JT, et al. Illumination and er-rors in dispensing. Am J Hosp Pharm. 1991;48:2137–45.
14. Flynn EA, Barker kN, Berger BA, et al. Dispensing errors and counseling quality in 100 pharmacies. J Am Pharm Assoc. 2009;49:171–80.
15. Allan EL, Barker kN, Malloy MJ, et al. Dispensing errors and coun-seling in community practice. Am Pharm. 1995;NS35:25–33.
16. Flynn EA, Barker kN, Carnahan BJ. National observational study of prescription dispensing accuracy and safety in 50 pharmacies. J Am Pharm Assoc. 2003;43:191–200.
17. Szeinback S, Seoane-Vazquez E, Parekh A, et al. Dispensing errors in community pharmacy: perceived influence of socio-technical factors. Int J Qual Health Care. 2007;19:203–9.
18. Gurwitz JH, Field TS, Harrold LR, et al. Incidence and prevent-ability of adverse drug events among older persons in the am-bulatory setting. JAMA. 2003;289:1107–16.
19. Lesar TS, Briceland L, Stein DS. Factors related to errors in medication prescribing. JAMA. 1997:277:312–7.
20. Leape LL, Bates DW, Cullen DJ, et al. Systems analysis of ad-verse drug events. JAMA. 1995;274:35–43.
21. Dean BS, Allan EL, Barber ND, et al. Comparison of medication errors in an American and a British hospital. Am J Health Syst Pharm. 1995;52:2543–9.
22. Field TS, Gilman BH, Subramanian S, et al. The costs associ-ated with adverse drug events among older adults in the am-bulatory setting. Med Care. 2005;43:1171–6.
RISk MODELS IN COMMuNITy PHARMACIES ReseaRch
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o n www.japha.org S e p /O c t 2012 • 52:5 • JAPhA • 601
23. Ernst FR, Grizzle AJ. Drug-related morbidity and mortal-ity: updating the cost-of-illness model. J Am Pharm Assoc. 2001;41:192–9.
24. Cohen MR, Proulx SM, Crawford Sy. Survey of hospital sys-tems and common serious medication errors. J Healthc Risk Manag. 1998;18:16–27.
25. Davis NM, Cohen MR. Today's poisons: how to keep them from killing your patients. Nursing. 1989;19:49–51.
26. Institute for Safe Medication Practices. ISMP 2007 survey on high-alert medications; difference between nursing and phar-macy perspectives still prevalent. ISMP Medication Safety Alert! 2007;12(10):1–3.
27. Smetzer JL, Cohen MR. Medication error reporting systems. In: Cohen MR, Ed. Medication errors. 2nd ed. Washington, DC: American Pharmacists Association; 2007:513–50.
28. Battles JB, kanki BG. The use of socio-technical probabilistic risk assessment at AHRQ and NASA. In: Spitzer C, Schmocker u, Dang VN, Probabilistic safety assessment and manage-ment. 4th ed. Berlin: Springer; 2004:2212–7.
29. Burkhardt M, Lee C, Taylor L, et al. Root cause analysis of medication errors. In: Cohen MR, Ed. Medication errors. 2nd ed. Washington, DC: American Pharmacists Association; 2007:513–50.
30. Comden SC, Marx D, Murphy-Carley M, Hale M. using proba-bilistic risk assessment to model medication system failures in long-term care facilities. Accessed at www.ahrq.gov/down-loads/pub/advances/vol2/Comden.pdf, August 15, 2010.
31. Senders JW, Senders SJ, Smetzer JL, et al. Heath care failure mode and effects analysis. In: Cohen MR, Ed. Medication er-rors. 2nd ed. Washington, DC: American Pharmacists Associa-tion; 2007:561–86.
32. The Joint Commission. Failure mode and effects analysis in health care: proactive risk reduction. 2nd ed. Oakbrook Ter-race, IL: Joint Commission Resources; 2005.
33. Helmreich RL. On error management: lessons from aviation. BMJ. 2000;320:781–5.
34. Van der Schaaf TW. Medical applications of industrial safety science. Qual Saf Health Care. 2002;11:205–6.
35. Spencer FC. Human error in hospitals and industrial accidents: current concepts. J Am Coll Surg. 2000;191:410–8.
36. Marx DA, Slonim AD. Assessing patient safety risk before the injury occurs: an introduction to socio-technical proba-bilistic risk assessment in healthcare. Qual Saf Health Care. 2003;12(suppl 2):33–8.
37. Marx DA, Westphal JE. Socio-technical probabilistic risk as-sessment: its application to aviation maintenance. Int J Aviat Psychol. 2008;18:51–60.
38. Stamatelatos M, Apostolakis G, Dezfuli H, et al. Probabilistic risk assessment procedures guide for NASA managers and practitioners. Accessed at www.hq.nasa.gov/office/codeq/doctree/praguide.pdf, June 3, 2010.
39. Garrick BJ, Christie RF. Probabilistic risk assessment practices in the uSA for nuclear power plants. Saf Sci. 2002;40:177–201.
40. Forester JA, Whitehead DW, kolaczkowski AM, et al. Results of a nuclear power plant application of a new technique for human error analysis (ATHEANA). Accessed at www.osti.gov/bridge/servlets/purl/543650-eNlI7i/webviewable/543650.pdf, June 3, 2010.
41. Reason J. Hazards, defences and losses. In: Reason J. Manag-ing the risks of organizational accidents. Burlington, VT: Ash-gate; 2000:1–18.
42. Stockwell DC, Slonim AD. Quality and safety in the intensive care unit. J Intensive Care Med. 2006;21:199–10.
43. Wreathall J, Nemeth C. Assessing risk: the role of probabilistic risk assessment (PRA) in patient safety improvement. Qual Saf Health Care. 2004;13:206–12.
44. Cook RI, Wreathall J, Smith A, et al. Probabilistic risk as-sessment of accidental ABO-incompatible thoracic organ transplantation before and after 2003. Transplantation. 2007;84:1602–9.
45. Institute for Safe Medication Practices. The National Medica-tion Errors Reporting Program (ISMP MERP). Accessed at https://www.ismp.org/orderforms/reporterrortoismp.asp.asp, July 1, 2012.
46. Pennsylvania Patient Safety Authority. Pennsylvania Patient Safety Reporting System. Accessed at http://patientsafetyau-thority.org/PA-PSRS/Pages/PAPSRS.aspx, August 15, 2010.
47. Food and Drug Administration. FDA MedWatch: the FDA safety information and adverse event reporting program. Ac-cessed at www.fda.gov/medwatch, February 7, 2010.
48. Institute for Safe Medication Practices. ISMP survey on high-alert medications in community/ambulatory settings. Accessed at www.ismp.org/survey/newsletter/comsur-vey200606.asp, June 10, 2010.
49. Rothschild JM, Federico FA, Ghandhi Tk, et al. Analysis of medication-related malpractice claims: causes, preventability, and costs. Arch Intern Med. 2002;162:2414–20.
50. Moore TJ, Cohen MR, Furberg CD. Serious adverse drug events reported to the Food and Drug Administration, 1998–2005. Arch Intern Med. 2007;167:1752–9.
51. Zimmerman R, Bier VM. Risk assessment for extreme events. Ac-cessed at www.ldeo.columbia.edu/chrr/documents/meetings/roundtable/white_papers/zimmerman_wp.pdf, June 10, 2010.
52. Apostolakis G. The concept of probability in safety assess-ments of technological systems. Science. 1990;250:1359–64.
53. Bier VM. Challenges to the acceptance of probabilistic risk analysis. Risk Analysis. 1999;19:703–9.
54. Shelton CP. Human interface/human error. Accessed at www.ece.cmu.edu/~koopman/des_s99/human, February 10, 2010.
55. Relex software [computer program]. Version 7.6. Greensburg, PA: Relex Software Corporation; 2003.
56. Institute for Safe Medication Practices. Look-alike names: safe-ty briefs. ISMP Medication Safety Alert! 2008;13(6):1–2.
57. koutnik E. Bar coding: working toward error-free phar-macy. Accessed at www.pharmacytimes.com/issues/arti-cles/2003--09_690.asp, January 10, 2010.
58. Institute for Safe Medication Practices. ISMP's list of er-ror-prone abbreviations, symbols, and dose designa-tions. Accessed at www.pharmacytimes.com/issues/arti-cles/2003--09_690.asp, January 22, 2010.
59. Poon EG, Cina JL, Churchill W, et al. Medication dispensing errors and potential adverse drug events before and after im-plementing bar code technology in the pharmacy. Ann Intern Med. 2006;145:426–34.
ReseaRch RISk MODELS IN COMMuNITy PHARMACIES
J o u r n a l o f t h e A m e r i c a n P h a r m a c i s t s A s s o c i a t i o nwww.japha.org602 • JAPhA • 52:5 • S e p /O c t 2012
60. ukens C. New study sheds light on medication errors. Drug Topics. 2002;146(21):33.
61. Maviglia SM, yoo Jy, Franz C, et al. Cost-benefit analysis of a hospital pharmacy barcode solution. Arch Intern Med. 2007;167:788–94.
62. Lexi-Comp. FentaNyL monograph (Lexi-Drugs). Ac-cessed at http://online.lexi.com/lco/action/doc/retrieve/docid/patch_f/6903, August 16, 2010.
63. Institute for Safe Medication Practices. Methotrexate overdose due to inadvertent administration daily instead of weekly. Ac-cessed at www.ismp.org/hazardalerts/ha.pdf, August 16, 2010.
64. Moore TJ, Walsh CS, Cohen MR. Reported medication er-rors associated with methotrexate. Am J Health Syst Pharm. 2004;61:1380–4.
65. Schulmeister L. Chemotherapy medication errors: descrip-tions, severity, and contributing factors. Oncol Nurs Forum. 1999;26:1022–42.
66. Chassin MR, Becher EC. The wrong patient. Ann Intern Med. 2002;136:826–33.
67. Institute for Safe Medication Practices. Oops, sorry, wrong patient! Applying the JCAHO “two-identifier” rule beyond the patient's room. ISMP Medication Safety Alert! 2004;9(11):1–2.
68. Institute for Safe Medication Practices. “20/20” captures com-munity pharmacy errors, but vision seems to be shortsighted. ISMP Medication Safety Alert! 2007;12(7):1, 3.
69. knudsen P, Herborg H, Mortensen AR, et al. Preventing medica-tion errors in community pharmacy: frequency and seriousness of medication errors. Qual Saf Health Care. 2007;16:291–6.
70. Bobb A, Gleason k, Husch M, et al. The epidemiology of pre-scribing errors. Arch Intern Med. 2004;64:785–92.
71. Bernardo JM. Bayesian statistics. Accessed at www.uv.es/ber-nardo/BayesStat.pdf, August 9, 2009.
72. Top 200 generic drugs by units in 2007. Accessed at http://drugtopics.modernmedicine.com/drugtopics/data/articlestan-dard//drugtopics/072008/491181/article.pdf, August 9, 2009.
73. Top 200 brand drugs by units in 2007. Accessed at http://drugtopics.modernmedicine.com/drugtopics/data/articlestan-dard//drugtopics/072008/491207/article.pdf, August 9, 2009.
74. IMS Health. 2009 channel distribution by u.S. dispensed pre-scriptions. Accessed at www.imshealth.com/deployedfiles/imshealth/Global/Content/StaticFile/Top_Line_Data/Chan-nel%20Distribution%20by%20u.S.%20RXs.pdf, August 15, 2010.
75. Wu JS, Apostolakis GE. Experience with probabilistic risk assessment in the nuclear power industry. J Hazard Mater. 1992;29:313–45.
76. Forester JA, Whitehead DW, kolaczkowski AM, et al. Results of nuclear power plant application of a new technique for human error analysis. Risk Analysis. 1996;16.
77. Hsueh kS, Mosleh A. The development and application of the accident dynamic simulator for dynamic probabilistic risk assessment of nuclear power plants. Reliab Eng Syst Saf. 1996;52:297–314.
78. Sues RH, Chen HC, Oswald EA, et al. Integrating internal events in an external probabilistic risk assessment: Tornado PRA case study. Reliab Eng Syst Saf. 1993;40:173–86.
79. System design and analysis harmonization and technology update. Fed Regist. 1996;61:26246–7.
80. Moore TJ, Cohen MR, Furberg CD. Quarter watch: 2008 quar-ter 1. Accessed at www.ismp.org/QuarterWatch/2008Q1.pdf, February 10, 2010.
81. Moore TJ, Cohen MR, Furberg CD. QuarterWatch: 2008 quar-ter 2. Accessed at www.ismp.org/QuarterWatch/200901.pdf, August 10, 2010.
82. Vertical Health LLC. Type 1 diabetes: symptoms, diagnosis, & treatment of type 1 diabetes. Accessed at www.endocrineweb.com/diabetes/1diabetes.html, February 18, 2010.
83. Centers for Disease Control and Prevention. Diabetes public health resource. Accessed at www.cdc.gov/diabetes/projects/cda2.htm, February 18, 2010.
84. Institute for Safe Medication Practices. Proliferation of insulin combination products increases opportunity for errors. ISMP Medication Safety Alert! 2002;7(24):1–2.
85. Institute for Safe Medication Practices. Getting the right in-sulin is becoming a real crapshoot. ISMP Medication Safety Alert! 2004;9(14):2–3.
86. Gertman D, Blackman H, Marble J, et al. The SPAR-H human reliability analysis method. Washington, DC: Nuclear Regula-tory Commission; 2005.
87. System Reliability Center. Technique for human error rate pre-diction (THERP). Rome, Ny: Alion Science and Technology; 2005.
88. Garnerin P, Pellet-Meier B, Chopard P, et al. Measuring hu-man-error probabilities in drug preparation: a pilot simulation study. Eur J Clin Pharmacol. 2007;63:769–76.
89. Lewis M. THERP: technique for human reliability analysis. Ac-cessed at www.pitt.edu/~cmlewis/THERP.htm, June 15, 2010.
90. Institute of Petroleum. Human reliability analysis. London: In-stitute of Petroleum; 2003.
91. Gibson WH, Hicking B, kirwan B. Feasibility study into the col-lection of human error probability data. Brussels, Belgium: European Organisation for the Safety of Air Navigation, Euro-control Experimental Centre; 2006.
Appendix 1. Medications involved in harmful errors in ambulatory settings Studies
Phillips et al. 20017
Rothschild et al. 200249
Ghandhi et al. 200312
Flynn et al. 200316
Gurwitz et al. 200318
Budnitz et al. 200611
Budnitz et al. 200611
Howard et al. 200710
Moore et al. 200750
Howard et al. 20089
Medications
or Medication
Class Reported fatal drug errors in hospitals, ambulatory care settings, patients’ homes
Analysis of malpractice claims for preventable adverse drug events (inpatient and outpatient)
Preventable adverse drug events in adults 18 and older self-administered prescription medications
Medication dispensing errors in ambulatory pharmacies considered clinically significant
Preventable adverse drug events in older individuals in ambulatory clinical settings
Emergency department visits classified as an uninten-tional overdose
Emergency department visits for an adverse drug event that led to hospital-ization
Identified the drugs most often responsible for hospital admission (review of 17 studies)
Increase in reported deaths and serious injuries associated with drug therapy
Causes of preventable drug-related hospital admissions
Analgesics X opioids and nonnarcotic
X nonnarcotic
X nonnarcotic
X opioids
X opioids
X opioids
Anticonvulsants X X
Antidiabetic agents, including insulin
X insulin
X X insulin, oral hypoglycemics
X insulin
X X insulin
X
Antiinfective/ antibiotics
X X X penicillins
X wrong label information, clindamycin
X amoxicillin-‐containing agents
Antineoplastics X X X X
Anticoagulants X warfarin, heparin
X X
X warfarin
X
X X warfarin X warfarin
Antiplatelets X aspirin, Plavix
Antipsychotic/ antidepressants
X X
Antirheumatic X
Antitussive X promethazine with codeine
Cardiovascular drugs
X X X calcium channel blockers, beta blockers, ACE inhibitors
X wrong form (Norpace)
X X digitalis glycosides
X beta blockers, angiotensin inhibitors, inotropes
X digoxin, beta blockers
Diuretics X X X
Electrolytes X
Hormones and synthetic substitutes
X Depo-‐Medrol
X estrogens
Ipratropium bromide inhaler
X
Lithium X
Nonsteroidal antiinflammatory agents
X X X COX-‐2 inhibitors
Theophylline X
Appendix 2. Advantages of ST-‐PRA over FMEA and RCA30,37–43,75–79 • Presents a comprehensive risk assessment of the “as is” state of predominantly human
processes Accommodates latent failures and active risks Includes the identification and effects of:
o Human error* o At-‐risk behaviors** o Intentional procedural deviations o Mechanical failures
Permits examination in complex systems of high impact events of both low and high frequency
Models unique combinations of events, processes, and behaviors that increase or decrease risk
Models the effect of a combination of more than one failure point in complex systems that are most likely to lead to errors
Objective evaluation and comparison of different configurations are possible Models can incorporate the cumulative knowledge of operations experts when
complete data sets from other sources are not available
• Quantifies risk and makes it visible Uses probability estimates culled from an expert panel Uses software-‐facilitated calculation to quantify combinations of risks Predicts the quantitative impact of specific system or practice changes Provides a visual depiction of the dispensing system and ways that errors reach
consumers
• Informs decision making and promotes rapid assessment/improvement Predicts the most common error pathways to ensure risk reduction or mitigation
efforts are effective Permits ranking of relative risks to determine where to concentrate limited
resources for maximum risk reduction Models different interventions to identify the variation with the most benefit Identifies immediate high-‐impact changes Prioritizes interventions based on quantitative risk
• Facilitates the safety culture Key stakeholders in healthcare can see their roles in the error process, their
interdependencies, and how their actions increase or decrease risks for each other and ultimately, for patients
*Human error is defined as: inadvertently doing other than what should have been done; the failure of a planned action to be completed as intended (error of execution) or the use of a wrong plan to achieve an aim (error of planning). **At-risk behavior is defined as: behavioral choice that increases risk where risk is not recognized or is mistakenly believed to be justified. At-‐risk behaviors are often employed and tacitly encouraged as a workaround for various system, process, technological, or environmental weaknesses.
Appendix 3. Sample section of FMEA for warfarin and fentanyl transdermal patches Process Failure
Mode Causes of Failure/ Performance Shaping Factors
Effect of Failure
Upstream/Downstream Controls
Example for Warfarin Prepare medication
Select the wrong drug or dose
• Look-‐alike products stored near each other (e.g., different strengths)
• Look-‐alike drugs mistakenly sent by wholesaler and/or misplaced in pharmacy stock
• Label ambiguity • Knowledge deficit
Allergic reaction Overdose: bleeding, death Subtherapeutic dose: thrombosis
Upstream controls • Separate look-‐alike products • Warning messages in computer system for serious product labeling issues or look-‐alike packages
• Checking process to verify stock upon arrival from wholesaler
Downstream controls • Independent double check before anticoagulants are dispensed from the pharmacy
• Reviewing the medication with the patient before dispensing it
• Patient inspection at the point-‐of-‐sale
Example for Fentanyl Transdermal Patches Prescribe the drug
Prescribe the wrong dose
• Knowledge deficit about dose conversions from opioids to fentanyl patches
• Knowledge deficit regarding use of drug in opioid-‐naïve patients, elderly patients, or patients who do not require a total daily dose of opioids equivalent to 25 mcg/hour
• Patient’s clinical situation not known or considered
• Confuse size of patch as dose (e.g., 10 cm2 for a 25 mcg/hour patch)
• Mental slip • Wrong dose selection from list of doses if prescribed electronically
• Increasing the dose before peak fentanyl levels achieved in 24-‐72 hours (change recommended no sooner than 3 days after the current dose has been administered)
Overdose: respiratory arrest, death Subtherapeutic dose: uncontrolled pain
Upstream controls • Dose conversion guidelines readily accessible
• Warning messages in computer when prescribing the drug using electronic system
Downstream control • Warning messages in computer when entering prescription into the patient profile
Appendix 4. Elements of event trees Event Tree Element Description Top Level Event The adverse outcome that occurs without capture before
reaching the victim. Example: • Dispense the wrong dose of fentanyl patch to a patient
due to a prescribing error that was not captured. Initiating Error The initiating error under study that may result in the
adverse outcome or be captured before it occurs. Example: • Prescribe the wrong dose of fentanyl patch.
Basic Event The fundamental failures, loss of function, unavailability, exposure rates, or capture opportunities that create the branches in the event tree when combined with AND or OR gates (see description of AND and OR gates below). Example: • Pharmacy technician does not catch prescribing error
during data entry into an existing patient profile. Exposure Rate The frequency with which certain process steps or
conditions occur; used to split sections of the tree into appropriate categories that may have differing rates of contribution to the top-‐level event. Example: • Categories based on how the prescription was received
in the pharmacy (e.g., 50% fax, 10% telephone, 39% patient delivery, 1% electronic).
Human Error
Inadvertently doing other than what should have been done; the failure of a planned action to be completed as intended (error of execution) or the use of a wrong plan to achieve an aim (error of planning). Example: • Physician orders wrong dose of a fentanyl transdermal
patch due to knowledge deficit. At-Risk Behavior Behavioral choice that increases risk where risk is not
recognized or is mistakenly believed to be justified. At-‐risk behaviors are often employed and tacitly encouraged as a workaround for various system, technological, or environmental weaknesses. Example: • Patient identification process not followed at point-‐of-‐
sale. Equipment Failure Any equipment failure that can affect the top-‐level event.
Example: • Failure of the barcode reader to accurately scan the
barcode. Capture Opportunities Activities or conditions that actively or passively help staff
detect and correct the initiating error. Failed capture opportunities may be caused by human error or at-‐risk behavior. Example: • Pharmacist fails to capture a prescribing error during
drug use evaluation. Many missed capture opportunities do not represent human error, but the opportunity to catch a mistake that was not realized. Example: • Patient fails to capture a dispensing error at point-‐of-‐
sale. AND Gates A method of combining basic events in which all the basic
events are required to satisfy the condition above the gate. OR Gates A method of combining basic events in which only one of the
conditions below the gate is required to satisfy the condition above the gate.
Appendix 5. Performance-‐shaping factors (PSFs)38–41 EXTERNAL
Task complexity Information complexity Ergonomics/Human-‐machine interface Procedures, including job aids Work environment Communication/Information exchange Workflow/Work processes Stress Time available/Time urgency Design of products and labels Organizational culture/Management
INTERNAL Training Experience Familiarity with task Mental and physical health/Fitness for duty Task tension and engagement Stress Motivation Previous actions
Appendix 6. Risk factors with warfarin, fentanyl patches, methotrexate, and insulin analogs Medication Examples of Known Risk Factors Warfarin
Warfarin is a narrow therapeutic index drug for which small changes in systemic concentration can lead to significant changes in pharmacodynamic response. This may result in potentially subtherapeutic or toxic effects, particularly in patients with advanced age, comorbid illness, or those receiving multiple medications. Warfarin requires close monitoring of INR values for proper dosing. Too frequent dose changes can lead to fluctuating and suboptimal levels of anticoagulation. Due to frequent dose changes, warfarin prescription labels often do not list the most current dosing directions, or prescribers list “take as directed” on prescriptions without explicit directions. Serious drug interactions are common with warfarin.
Fentanyl Patches
Fentanyl should only be used to treat moderate to severe chronic pain, not acute pain. Fentanyl should only be prescribed for patients who are opioid-‐tolerant (have taken other opioids previously for pain). Dose conversion from another opioid to fentanyl can result in overestimation of the dose. Each patch must be removed before applying a subsequent patch. Patches that fall off or are not disposed securely can be lethal to children because a significant amount of drug remains in the patch after disposal. Application on open skin or exposure to heat (e.g., sun, heating pad) can result in overdoses.
Methotrexate
When used for rheumatoid arthritis, the drug should be prescribed as a single weekly dose or in smaller doses every 12 hours for 3 doses/week, but never daily. Weekly versus daily dosing for most drugs is uncommon and thus prone to error. Serious drug/drug and drug/disease interactions and adverse effects can occur with methotrexate. Monthly hematology and bimonthly renal and hepatic lab studies are required during treatment. Handwritten orders for methotrexate 2.5 mg can be mistaken as minoxidil 2.5 mg, and vise versa.
Insulin analogs
There are about a dozen different types of insulins and several dozen different brands, many of which have names and/or packages that look or sound alike. The onset of action for insulin types varies widely. Depending on the product, the onset may vary from mere minutes to 8 hours. This makes the typical time for insulin administration and its relationship to meals confusing. Insulin is available in multiple concentrations (100 units/mL and 500
units/mL). Patients often receive widely variable doses and more than one type of insulin concurrently. Many insulin products are available over-‐the-‐counter. The abbreviation “u” for units can be misread as a zero, risking a 10-‐fold error.