Transcript LUMC-Heelkunde Complicatie- bespreking en Incidentmelding

Integration of predictive and
retrospective risk analysis
in health care
Tjerk van der Schaaf
Leiden University Medical Center
Eindhoven University of Technology
overview
• retrospective method: PRISMA-medical
• predictive method : HFMEA
• 3 examples of possible integration
– direct comparison of predicted vs
“actual”causes (radiotherapy)
– components combined in a Healthcare Safety
Management System (convergent approach)
– evaluating major interventions (impact of IT on
medication safety)
retrospective risk analysis :
PRISMA- medical
• (voluntary) incident reporting and analysis
• learning from actual / reported process
deviations
PRISMA-Medical
• Prevention and Recovery Information System
for Monitoring and Analysis
• Three subsequent steps:
– Description by means of causal trees
– Classification according to the Eindhoven
Classification Model (medical version)
– Determination of countermeasures by means of the
Classification/Action Matrix
Causal tree example
Wrong route
Lines at
same place
Nurses not
informed
Similar lines
O
Catheters
not removed
No coding
T
No protocol
O
Connection
possible
Inadequate
check
T
H
Eindhoven
Classification
Model (medical
version)
Database
• Root causes for failure
failure profile
• Root causes for recovery recovery profile
• Context variables
black-spot analysis
PRISMA failure profile:
hospital medication errors
25
15
10
PRISMA category
X
PRF
HST
HSS
HRM
HRI
HRV
HRC
HRQ
HKK
H-EX
OC
OM
OP
OK
O-EX
TM
TC
0
TD
5
T-EX
Percentage (%)
20
Classification/Action Matrix
ECM
code
Design:
Technolo
gy/workplace
Procedures Information
and Commu
nication
Training
Motiva
tion
Escala
tion
×
T-EX
TD
×
TC
×
TM
×
O-EX
×
OK
×
×
OP
×
OM
×
OC
×
H-EX
×
HK_
NO
×
HR_
HS_
Reflection
×
NO
predictive risk analysis
HFMEA / SAFER
• series of group meetings to build a set of
failure scenario’s for a (small) process of
care : what may go wrong; why; what to do
about it
• pro-active appeal
Healthcare Failure Mode and
Effect Analysis (HFMEA)
• A systematic approach to identify and
prevent product and process problems
before they occur
• Developed by the "VA National Center for
Patient Safety"
(http://www.patientsafety.gov/)
Relevance of predictive risk
analysis
• Retrospective (incident) analysis takes
place after incidents did occur
 hindsight bias
• Because of underreporting, biases can
arise in incident databases
 identification of "missing risks"
Definitions
• Failure Mode: Different ways that a process or
subprocess can fail to provide the anticipated
result (i.e. think of it as what could go wrong)
Prescribing the wrong dose
• Failure Mode Cause: Different reasons as to
why a process or subprocess would fail to
provide the anticipated result (i.e. think of it as
why it would go wrong)
Miscalculation
HFMEA process
•
•
•
•
•
Step 1: Define the topic
Step 2: Assemble the team
Step 3: Graphically describe the process
Step 4: Conduct a hazard analysis
Step 5: Identify actions and outcome
measures
examples of integration (1)
• direct comparison of predicted (HFMEA)
vs reported causes
• user problems with a new radiation
therapy technology
• both types of failure causes expressed in
the same PRISMA-medical classification
(sub-)categories
PRISMA vs HFMEA : main categories
PRISMA
60%
HFMEA : predicted causes
Percentage
50%
40%
30%
20%
10%
0%
Tech
Org Human other
PRISMA main category
PRISMA vs HFMEA : subcategories
35%
33%
PRISMA
HFMEA
25%
21%
17%
13%
4%
4% 3%
2%
2%
1%
PRISMA category
Frequency category HFMEA
Weight-factor (= translation to 9 months)
HRV
HRC
HRQ
0%
HKK
H-EX
OM
OP
O-EX
0% 0%
OC
1%
0%
1%
1%
1%
0% 0%
0%
X
3%
0% 0%
TM
TD
TC
1%
5%
4%
3%
2%
0% 0%
4%
4%
PRF
5%
4%
OK
5%
0%
8%
8%
HST
10%
13%
HSS
15%
16%
HRM
17%
HRI
20%
T-EX
Percentage
30%
less than yearly
yearly
monthly
weekly
0,1
0,89
9
36
examples of integration (2)
• combining retrospective and predictive
components in an overall Healthcare
Safety Management System
• convergent approach of two imperfect risk
identification methodologies
• mutual checks, comparisons, and inputs
possible
examples of integration (2)
continued
• are repeatedly predicted problems (failure
modes) ever being reported?
• can frequently reported problems help to select
suitable processes for HFMEA and generate
realistic failure modes?
• can frequently predicted causes steer the
information gathering after an initial report?
• are proposed interventions for predicted vs
“reported” causes similar?
• etc…
examples of integration (3)
• developing a process-based evaluation
methodology for major (patient safety)
interventions
• predicting and monitoring the impact of
IT on medication safety
Medication safety: definitions
Adverse drug reactions
Harm
Medication error
Harm
Drug
Error
Medication error
No harm
[Van den Bemt et al., 2000]
Medical error
(not drug related)
Medication errors: causes (1)
•
•
•
•
•
•
•
•
Handwritten prescriptions and drug orders
Look-alike drug names
Sound-alike drugs and verbal orders
Use of abbreviations
Similar packaging and labelling
Inadequate training and supervision
Staff shortages
Overwork and fatigue
[Habraken, 2004]
Medication errors: causes (2)
IT: possibilities and problems
IT: possibilities and problems
56%
Prescribing
Physician order entry / Computerised decision support
Transcribing
Electronic order transcription
6%
Dispensing
4%
Robots / Bar coding / Automated dispensing devices
34%
Administering
Bar coding / Automated dispensing devices
Monitoring
Computerised monitoring of adverse drug events
[Bates et al., 1995; Bates, 2000]
Medication administration record
Computerised medication administration record
IT: possibilities and problems
IT application
PROS
CONS
CPOE
Legible prescriptions; no
handwriting required
Possibility of substitution errors
Data entry only necessary once
Failure to warn
Exchange of data is easy
Computerised decision Drug information
support
Patient-specific information and
Risk of low vigilance and
overtrust
advice
Bar coding
Ensure five "rights": right drug,
Degraded coordination and
right patient, right dose, right route, communication
right time
Computerised medical
record
Legible prescriptions; no
handwriting required
Data entry only necessary once
Exchange of data is easy
[Habraken and Van der Schaaf, 2006]
Possibility of substitution errors
Barriers to the implementation of IT
• Significant costs: technical, process
redesign, and implementation and support
• Cultural obstacles: resistance to change
• Privacy and protection of (patient) data
• Lack of data standards
• Lack of (clinical) evaluation
[Habraken, 2004]
Evaluation of effects and impact
of IT: PRISMA and HFMEA
• Not only outcomes of care but also the
mechanisms underlying those outcomes
• Impact of IT on "error recovery " :
– Detection
– Diagnosis
– Correction
of earlier errors / deviations
Evaluation of effects and impact
of IT: PRISMA
•
•
PRISMA can be used to obtain an insight
into the behavioural mechanisms
underlying medication errors
Classification/Action Matrix enables us to
predict which types of human behaviour
will be influenced by IT
Evaluation of effects and impact
of IT: PRISMA
ECM
code
Design/
Technol
Procedures
Information and
Communication
Training
Motivation
Escalation
×
T-EX
TD
×
TC
×
TM
×
O-EX
×
OK
×
×
OP
×
OM
×
OC
×
H-EX
×
HK_
NO
×
HR_
HS_
Reflection
×
NO
Evaluation of effects and impact
of IT: PRISMA
•
•
•
•
IT applications would fall in two
categories: "technology" and "information
and communication"
In case of improved technology 
reduction of skill based human errors
In case of information and
communication support  reduction of
knowledge based errors
BUT: rule based human errors would not
be influenced by IT
Evaluation of effects and impact
of IT: PRISMA and HFMEA
•
•
•
Theoretical predictions could be
reinforced by predictive risk analysis,
such as HFMEA
Empirical evaluation of actual impact of
IT by means of intensified incident
reporting
Comparison of causal patterns of
incidents that occur before, during, and
after the IT intervention
Conclusion (1)
• IT often mentioned as prerequisite for
reduction of medication errors
• Results regarding effects of IT vary greatly
• Effects of IT on behavioural mechanisms
are not/hardly taken into account
• PRISMA and HFMEA offer a framework for
in-depth analysis of impact of IT
Conclusion (2)
• Two types of predictions can be made of
expected effects of IT on error and error
recovery:
– Theoretical predictions by means of PRISMA
– HFMEA scenario-based predictions
• Intensified incident reporting and analysis would
enable a fast comparison between predicted and
actual effects
• On-line corrections of implementation process
could prevent actual adverse events
Thank you for your attention