Global Trigger Tool An assessment of ADHB experience Rob - - PowerPoint PPT Presentation

Global Trigger Tool

An assessment of ADHB experience

Rob Ticehurst, Principal Pharmacist Medication Safety (on behalf of Colin McArthur, Medical Advisor – Quality & Safety) Auckland District Health Board

Setting the scene

• ADHB: 3400 admissions/month
• NZ Adverse Events study (Davis et al)

 Case note review of 6500 patients

• all harms: 12.9%
• significant harm: ~4% of admissions = ~130/month for ADHB
• Preventable significant harms: ~45/month
• ADHB Reported significant harm events (SAC 1/2)



~6/month

• GTT

 IHI suggest AEs in ~30% of admissions

• = ~1000/month for ADHB

There’s quite a range!

• 440 harms (Davis)
• 1000 harms (IHI)
• 6 reported (ADHB SAC1/2)

Can’t review every patient Sample size is going to be important….

Global Trigger Tool: Methods

ADHB ~3400 discharges per month

• Review 20-40 discharges (>24h

admission) per month

 ~1% ADHB discharges

• “unintended physical injury resulting from
• r contributed to by medical care that

requires additional monitoring, treatment

• r hospitalisation, or results in death”
• Acts of commission (not omission)
• Irrespective of preventability

Global Trigger Tool: Outputs

• Adverse events

 Unintended physical injury caused or contributed to

by treatment (from patient’s viewpoint)

 Classification of event type and severity of harm  Multiple (separate) events counted

• Events per 1000 patient-days (typically 90)
• Events per 100 discharges (typically 40)
• % of discharges with any events (typically 30%)

Presenting the results – Run chart

Simple time plot of number of harms identified from a small sample presented as harms per 1000 pt days.

Same Data

• A very brief introduction to control

charts4.flv

A closer look…

Statistical Process Control

• Measure current / optimised process

 mean ± standard deviation (SD)

• Regular small samples (n)
• Sample means normally distributed
• Control limits usually 3 x SD

Adverse event “sampling”

An example

• 1000 discharges
• Random sample of n=20
• Believe our AE rate 20-80%

How much variation is there from sample to sample? Is our sample representative of the population?

Example 1 - 30% AE rate

60% 50% 40% 30% 20% 10%

Example 2 - 30% AE rate

60% 50% 40% 30% 20% 10%

Adverse Events per 1000 patient-days

0.0 20.0 40.0 60.0 80.0 100.0 120.0 25/07/2011 8/08/2011 22/08/2011 5/09/2011 19/09/2011 3/10/2011 17/10/2011 31/10/2011 14/11/2011 28/11/2011 12/12/2011 26/12/2011 9/01/2012 23/01/2012 6/02/2012 20/02/2012 5/03/2012 Events Date

% admissions with adverse events

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

27%

Cummulative data accuracy

95% confidence limits

0% 10% 20% 30% 40% 50% 60% 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 Week

Fortnightly samples of 20, >1000 admissions

Sample size required

• Trust me on this….

Suspected AE rate = 30% With a sample size of 10 we can be 95% confident that our actual value is between 0% and 65% (+/- 35%)

Margin of error Sample size 1% 8281 5% 364 10% 100 20% 29 30% 14 40% 8 50% 5

How much uncertainty are you prepared to accept? How much precision do you require?

Conclusion 1

• Small sample size

 Wide confidence limits – our actual AE rate

could be anywhere between 0 and 65%

• Wide limits on control chart
• Control charts of overall AE rate highly

unlikely to be useful to demonstrate the effectiveness of an intervention

 IHI suggest otherwise….

What if you aggregate the data?

ADHB -196 adverse events from 512 admissions over 20 months

Drill down further

Medication adverse events

n

Hypotension/bradycardia 12 Sedation & delirium 10 Nausea & vomiting 10 Constipation 7 Chemo complications 8 Acute kidney injury 6 Hypoglycaemia 2 Other 10 Nosocomial infections

n

Wound 14 Pneumonia 9 CLAB 4 Urinary tract 6 Bacteraemia 7 Other 10 Procedure - related

n

Bleeding 10 Ileus 5 Other 27 ADHB - 196 adverse events from 512 admissions over 20 months

Adverse Event Severity

• E: Temporary harm to the patient and

required intervention – 58%

• F: Temporary harm and required initial or

prolonged hospitalization – 38%

• G: Permanent patient harm – 0.5%
• H: Intervention to sustain life – 2%
• I:

Patient death – 1.5%

ADHB - 196 adverse events from 512 admissions over 20 months

High frequency, low severity

795 admissions to 3 large tertiary USA hospitals Classen et al, Health Affairs 2011; 30(4):581-589

Conclusion 2

• Aggregated data can give us an idea of

what is happening

• Identifies areas for further investigation
• GTT not suitable for demonstrating
• utcomes of any interventions

Summary

• Small random samples provides trend
• ver time (but with wide control limits)
• Large short term variability
• Accurate population rate can only be

estimated over long periods (12 months +)

• Statistically significant change will take

years to demonstrate

Summary

• Run charts are not useful
• GTT identifies high frequency “minor” events

not detected by other methods

• Interventions require more targeted/accurate

measures (eg CLAB, surgical site infection)

• Future:

 Feed into improvement programme priorities?  Intermittent larger sampling?