Bhawanipatna Call Girls ๐9332606886 Call Girls in Bhawanipatna Escorts servic...
ย
Ian Blunt: PARR-30: a predictive model for readmission within 30 days
1. PARR-30: a predictive model for
readmission within 30 days
Presenter: Ian Blunt
19 June 2012 ยฉ Nuffield Trust
2. Development of a predictive model for readmission within 30
days of discharge (PARR-30)
Model developed by Billings, Blunt, Steventon, Georghiou, Lewis and
Bardsley
โข Motivation
โข Development
โข Model performance
โข Testing in hospitals
โข Conclusions
ยฉ Nuffield Trust
3. Why predict readmissions within 30 days?
โข Readmissions are costly, suboptimal health care - costs to the NHS
estimated at ยฃ1.6 billion each year
โข DH guidance for the NHS proposes commissioners do not pay
provider hospitals for emergency readmission within 30 days of a
selected index elective admission
โข Rate of readmissions will also play an important part in monitoring
health system performance, as one of the new English Public Health
โoutcome indicatorsโ
ยฉ Nuffield Trust
4. Not first to try this, butโฆ
Number of international 30 day models Model From C
statistic
Predictive tools built in one setting Halfon et al 2008 0.67
may not necessarily be accurate when
used in other health care settings Silverstein et al 0.65
2008
Used hospital episode statistics (HES)
data to develop model for NHS in Van Walraven et al 0.68
England 2010
Howell et al 2009 0.65
Make PARR-30 freely available for
use across the NHS in England
And othersโฆ
(possibly tablet/smartphone app)
See Kansagara et al JAMA 2011
ยฉ Nuffield Trust
5. How is PARR30 different from PARR++?
โข Readmission in next 30 days vs next 365 days
โข Tools operate in different ways, trigger different responses
โข Next year โ longer for clinicians and care managers/coordinators to contact
and engage with high-risk patients, effect behavioural change
โข 30 days โ highest likelihood of an unplanned admission, focussing their
discharge planning efforts and post-discharge support for high-risk patient
Model Timescale Run by Input Data Data lag
variables source
PARR++ 12 months PCT ~250 SUS ~ 3 months
PARR30 30 days Acute 17 PAS/notes None
โข Aim for speed of low-variable models with accuracy of PARR
ยฉ Nuffield Trust
6. How is PARR30 different from PARR++?
PARR30
Hospital provides SUS
Patient nears discharge
PCT runs PARR++
Risk score calculated on
ward
Patients selected for
intervention (via GP)
Any extra intervention put
in discharge plan
Predicts readmission in next year โ Predicts readmission in 30 daysยฉ Nuffield Trust
PPV 65% โ PPV ???%
7. Model development
Developed using 10% sample HES
from April 2006 to May 2009
Index discharges in FY 2008/09
Readmissions within 30 days
reflected 2011-12 operating
framework
Logistic regressions identify variables
that contributed most to predictions
Validated with split sample
ยฉ Nuffield Trust
8. Model development
Hospital of current admission
Patient age
Deprivation (via post code)
History of emergency admissions:
Current? Last 30 days? Past year?
History in the prior two years of eleven
major health conditions drawn from
the Charlson co-morbidity index
ยฉ Nuffield Trust
9. Results
The performance of the model
was respectable, with a positive
predictive value (PPV) of
59.2% and area under the ROC
curve (โc-statisticโ) of 0.70.
For the higher-risk patients (risk
score > 50%), readmission rates
ranged from 47.7% up to 88.7%.
However, these patients only
represented a small share
(1.1%) of all patients analysed. Receiver Operating Characteristic Curve (ROC) for the
bootstrapped central estimate (red line) and 95%
confidence Intervals (shaded area) ยฉ Nuffield Trust
10. Results
ยฃ3,000
Predictive modelling only as effective
as the intervention it is used to
ยฃ2,500
trigger. Providers need to know
Mean cost of readmission
potential costs of readmission to ยฃ2,000
build business case for intervention
ยฃ1,500
For patients risk score > 50%, mean ยฃ1,000
readmission cost per patient was
ยฃ500
ยฃ1,088. Assuming that an
intervention can reduce the number
ยฃ0
of readmissions by 10% for this Risk score
group, ยฃ109 per patient could be Mean of cost readmission (readmitted patients only)
spent at breakeven Mean of cost readmission (all patients) ยฉ Nuffield Trust
11. Testing PARR-30 in hospitals
Testing: Royal Berkshire Hospital using
spreadsheet version of tool on wards:
โข Is the tool easy to use?
โข Completed by junior doctors
โข Bedside info vs admin systems?
โข Test tool stored its output
โข Does ward PPV reflect national?
โข Later reconciled with admin
systems for analysis
Chelsea & Westminster Hospital
running tool direct from their data
warehouse: Applied on four care of the elderly
wards in Feb/March 2012
โข Proved it can be done easily
โข Looking into PPV and clinical
engagement ยฉ Nuffield Trust
12. Testing PARR-30 in hospitals
Results from using spreadsheet on the wards:
โข Tool was used 88 times โข Low risk scores โ max 39%
โข Median time to complete 1m 41s โข Em admit in last 30 days diff 10%
โข Median patient age was 86, โข Em admits last year diff 20%, ยฑ1,2
mostly emergency admissions
โข Even split whether tool or system
โข Average 1.3 co-morbidities, max 4 has more
โข 10 patients had emergency โข 14% where system has diagnosis
readmission within 30 days not ticked as co-morbidity
ยฉ Nuffield Trust
13. Conclusions
Built a predictive model using a limited set of variables that were
generated from hospital episode statistics
Variables easily available from patientsโ notes or from the patients
themselves โ can calculate from spreadsheet or in PAS
The performance of the model was respectable - highest risk patients
had a 88.7% chance of hospital readmission within 30 days โ but high
risk patients relatively rare
Cost data suggests interventions need to be lower-cost to break even
Easily used on wards in trials - less than 2 minutes per application
Some differences in data on ward, but not huge
ยฉ Nuffield Trust
14. www.nuffieldtrust.org.uk
Sign-up for our newsletter
www.nuffieldtrust.org.uk/newsletter/login.aspx
Follow us on Twitter
(http://twitter.com/NuffieldTrust)
19 June 2012 ยฉ Nuffield Trust