The document describes a study piloting a preoperative risk assessment tool for predicting postoperative delirium (POD) risk. The study was conducted at a hospital in Switzerland over 6 months, prospectively collecting data from 55 patients undergoing hip or knee surgery aged 65 or older. The tool classified 7 patients as high risk for POD (4.5% incidence). Compliance with using the tool and delirium prevention strategies was high. The results provide support for further evaluation of the tool in reducing POD incidence through targeted prevention and treatment strategies.

1. Kelly Reeve1
, Nayeli Schmutz Gelsomino2, 3
, Benjamin T. Dodsworth2
1
Institute of Data Analysis and Process Design, Zurich University of Applied Sciences, Switzerland
2
PIPRA AG, Zurich, Switzerland
3
Department of Anaesthesia, University Hospital Basel, Switzerland
PreOp Risk Assessment Software for POD
BACKGROUND
DESIGN
RESULTS
CONCLUSIONS
Success of the tool relies on more
than just one person (take a more
holistic approach)
Involve the nursing staff
from day one
Help the clinic to integrate the tool into
the current decision-making, workflow
and infrastructure
Connect with Kelly
kelly@pipra.ch
Piloting of a
PREOPERATIVE RISK
ASSESSMENT TOOL
for postoperative delirium risk www.pipra.ch
Patient Data Risk Score
Goal of Pilot Study: Prospectively test feasibility and performance in real-life hospital setting
55 patients included
7 DOS+ for POD (4.5%)
PIPRA compliance: 153 (98.7%)
DOS compliance: 107 (69%)
Steps for model development, assessment and implementation
Result
43%
Prevention/
Treatment
Adapt operation
and nursing
Cost-effective
reduction of delirium
1 Cochrane Review:
https://doi.org/10.1002/14651858.CD013307.pub3
AI
External Validation
Post Implementation
Evaluation
Clinical Impact
Evaluation
Piloting
%
Risk
Internal
Validation
Model
Development
Data
• 6 months of prospective collection
starting in September 2021
• Kantosspital Baden
• Pre-anaesthesia consultation
Questions that arose
in the survey:
Who needs to be
informed ?
How do we document
and flag this in the
patient database?
How do we communicate
this to patients and family
members?
What should we do when
someone is identified as
high risk?
Inclusion/Exclusion
• 65 years of age or older
• Hip or knee surgery
• No emergency surgeries
Outcomes
• Proportion with POD
• Model performance
• Compliance with use of DOS and
PIPRA
• Usability survey
0.00
0.01
0.02
0.03
0.04
0% 10% 20% 30% 40% 50%
Threshold Probability
Net
Benefit
Treat All
Treat None
Model
0.0
0.1
0.2
0.3
0.4
0.5
No POD POD
Predicted
probability
0.00
0.25
0.50
0.75
1.00
0 25 50 75 100
PIPRA score threshold
Proportion
Sensitivity
Specificity
AUC = 0.9491
95% CI: 0.8853
to 0.9491
?
1

2. Benjamin T. Dodsworth1
, Lisa Falco2
, Tom Hueting3
, Behnam Sadeghirad4
, Lawrence Mbuagbaw4
, Nicolai Goettel5
, Nayeli Schmutz Gelsomino1,6
1
PIPRA AG, Zurich, Switzerland, 2
Zühlke Engineering AG, Schlieren, Switzerland, 3
Evidencio, Haaksbergen, Netherlands, 4
Department
of Health Research Methods, Evidence, and Impact, McMaster University, Hamilton, Canada, 5
Department of Anesthesiology,
University of Florida College of Medicine, Gainesville, USA, 6
Department of Anaesthesia, University Hospital Basel, Basel, Switzerland
A robust pre-surgical risk assessment for postoperative
delirium (POD) is required to find patients at risk.
POD underdiagnosed
= data quality problem!
1
Single centre studies
are not robust enough
2
1. THE PROBLEM
We performed an individual patient
data meta-analysis collecting over
20’000 patients from over 22 studies.
8 studies passed our quality control
(every patient assessed for POD) and
were included in algorithm
development.
To externally validate our algorithm,
we used data from a prospective
quality control study performed at a
hospital in Switzerland.
2. DATA COLLECTION
Coordinators:
Hospitals:
3. THE RESULTS
4. CLINICAL CONSEQUENCES
Algorithm has been developed into a software as a Medical
device and is approved for clinical use.
The evaluated risk of
neurocognitive
complications allows
clinics to adopt
protective personalized
peri-operative measures.
Classification plot
Specificity
Sensitivity
Low
Risk group stratification
thresholds:
Medium
High
Very
40.6% Low risk
25.1% Medium risk
16.6% High risk
Distribution of patients according to risk groups
17.6% Very high risk
Calibration plot:
external validation
Predicted risk
Average
risk
in
group
Calibration plot:
internal validation
Predicted risk
Average
risk
in
group
No further action
Example actions
For awareness (e.g. monitoring of precipitating factors)
Extra nursing time allocated e.g. for reorientation
Allocate rare resources (geriatrician) to most at-need patients
Low
Medium
High
Very high
Performance: Cross-validation AUC of 0.80 (95% CI: 0.77-0.82).
Three thresholds for patient stratification are shown, together
with sensitivity and specificity. External validation AUC 0.76
(95% CI: 0.69-0.83). Calibration plots below:
Connect
with Ben
ben@pipra.ch
For all inpatients
over age 60
Excludes cardiac &
intracranial surgery
Approved for
clinical use in Europe
Development and validation of an
INTERNATIONAL PREOPERATIVE
RISK ASSESSMENT MODEL
for postoperative delirium
This example
patient has 18%
risk of developing
delirium
www.pipra.ch

3. Nayeli Schmutz Gelsomino1,6
, Lisa Falco2
, Tom Hueting3
, Behnam Sadeghirad4
, Lawrence Mbuagbaw4
,
Nicolai Goettel5
, Benjamin T. Dodsworth1
1
PIPRA AG, Zurich, Switzerland, 2
Zühlke Engineering AG, Schlieren, Switzerland, 3
Evidencio, Haaksbergen, Netherlands, 4
Department of
Health Research Methods, Evidence, and Impact, McMaster University, Hamilton, Canada, 5
Department of Anesthesiology, University of
Florida College of Medicine, Gainesville, USA, 6
Department of Anaesthesia, University Hospital Basel, Basel, Switzerland
A robust pre-surgical risk assessment for postoperative delirium (POD) is required
to find patients at risk.
POD underdiagnosed
= data quality problem!
1
Single centre studies
are not robust enough
2
1. THE PROBLEM
We performed an individual patient data
meta-analysis collecting over 20’000 patients
from over 22 studies.
8 studies passed our quality control (every
patient assessed for POD) and were included
in algorithm development.
To externally validate our algorithm, we
used data from a prospective quality control
study performed at a hospital in Switzerland.
2. DATA COLLECTION
Hospitals: Coordinators:
3. THE RESULTS
4. CLINICAL CONSEQUENCES
The evaluated risk of
neurocognitive complications
allows clinics to adopt
protective personalized
peri-operative measures.
No further action
Example actions
For awareness (e.g. monitoring of precipitating factors)
Extra nursing time allocated e.g. for reorientation
Allocate rare resources (geriatrician) to most at-need patients
Low
Medium
High
Very high
Algorithm has been developed into a software
as a Medical device and is approved for
clinical use in Europe.
Performance: Cross-validation AUC of 0.80
(95% CI: 0.77-0.82). Three thresholds for
patient stratification are shown, together with
sensitivity and specificity. External validation
AUC 0.76 (95% CI: 0.69-0.83).
Connect with Nayeli
nayeli@pipra.ch
For all inpatients
over age 60
Excludes cardiac &
intracranial surgery
Approved for
clinical use in Europe
Development and validation of an
INTERNATIONAL PREOPERATIVE
RISK ASSESSMENT MODEL
for postoperative delirium www.pipra.ch
Classification plot
Specificity
Sensitivity
Low
Risk group stratification
thresholds:
Medium
High
Very
40.6% Low
25.1% Medium
16.6% High
Distribution of
patients according
to risk groups
17.6% Very high
Calibration plot:
external validation
Predicted risk
Average
risk
in
group
Calibration plot:
internal validation
Predicted risk
Average
risk
in
group
This example
patient has 18%
risk of developing
delirium