Real World Evidence stream - SAS Proceedings and …PhUSE 2016 RW04 Oct 2016 Asthma/COPD Algorithm 1...

Post on 03-Jun-2020

0 views 0 download

Transcript of Real World Evidence stream - SAS Proceedings and …PhUSE 2016 RW04 Oct 2016 Asthma/COPD Algorithm 1...

1 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Real World Evidence stream

An Algorithm to distinguish between COPD and asthma

Berber T. Snoeijer, Jetty A. Overbeek

PHARMO Institute for Drug Outcomes Research,

Utrecht, Netherlands

Berber Snoeijer

October 2016

PhUSE 2016 RW04

2 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Marketing of drugs

• What drug is used

• Where is it used

• To whom is it given

• At what price

3 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Marketing of drugs

• What drug is used

• Where is it used

• To whom is it given

• At what price

4 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

No indication

• Not in de database

• Not allowed to link

• Not available

5 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

6 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Linkage of databases

GP database

Outpatient pharmacy database

7 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Overlap

• GP diagnosis: ICPC R.. (respiratory)

• Pharmacy: ATC R03…. (respiratory)

• > 20,000 patients

Outpatient Pharmacy

data

GP data

8 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

To build a model

• Linkage of databases

• Exploratory analysis

• Logistic regression

– 2 models

• Finetuning

• Estimating the cut-off

• Validation

9 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Exploratory analysis

• Examine Differences between indication groups

– Asthma

– COPD

– Allergic conditions

– Other

• Statistics

– Descriptive

– T-tests / Chi-square

– Correlations / Regression

10 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Exploratory analysis

• Age

• Gender

• Drugs

• Drug classes

• Drug combinations

• Prescriber (GP / specialist)

• Duration of use

• Number of prescriptions

11 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Exploratory analysis

12 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Prepare dataset

• 1 row per patient

• Explanatory variables

– Drug x? –> 1/0

– Age class

– Duration of use

– …

• Outcome variables

– COPD? -> 1/0

– Asthma? -> 1/0

13 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Logistic regression

proc logistic data=lngWrk descending

OUTest=TmpEst outmodel=data.model1;

class agecat gender ;

model COPD= AgeCat gender Med1 Med2 Med3 … /

selection=backward slstay=0.001;

run;

14 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Finetuning

• C-statistic (concordance)

• Significance vs influence

• Correlation between explanatory variables

• Comparison of models for consecutive years

15 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

The COPD model

Age

Gender

Duration of use

R03

tiotropium

R01

R03BA ipratropium

LABA (no FDC)

montelukast

ICS

16 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Estimating the cut-off

PROC LOGISTIC inmodel=data.model1;

score data=InputData OUT=plong

(RENAME=(P_1=pcopd) DROP=p_0);

RUN;

• Sensitivity, specificity, NPV, PPV

• GP ratios

• Expected ratios based on literature

17 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Expected ratios based on literature

R03 users

asthma ≈ 50%

COPD ≈ 30 %

18 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Sensitivity

• Sensitivity 77.3%

COPD GP records

Yes No

Model prediction

Yes

No

19 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Specificity

• Specificity 85.6%

COPD GP records

Yes No

Model prediction

Yes

No

20 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Ratios of patients with COPD

• Ratio relative to all R03 users

2009 2010 2011 2012

GP

22.7% 23.6% 23.7% 23.8%

Model 2010

26.1% 27.5% 28.3% 29.0%

Model 2011

26.3% 27.7% 28.4% 29.1%

Model 2012

26.6% 28.0% 28.7% 29.4%

21 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Other models

• Diabetes Type I/II

• Heart failure

• High blood pressure

22 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Conclusions

• We created an useful model for estimating GPs indication for asthma or COPD.

• This model is used for market analysis for new drugs in the COPD/asthma market.

• COPD model based on age, gender, duration of R03 use and 5 different medication (group)s.

• Asthma model based on age, duration of R03 use and 5 different medication (group)s.

• Yearly update to keep accurate

23 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

Questions

?

24 Asthma/COPD Algorithm PhUSE 2016 RW04 Oct 2016

WWW.PHARMO.COM

+31 (0)30 744 08 00

pharmo@pharmo.nl

Contact details

E

T