ARCHIVING A COMPLEX DATASET: IST-3 A CASE STUDY by Jonathan Drever.

Post on 04-Jan-2016

214 views 0 download

transcript

ARCHIVING A COMPLEX DATASET:IST-3 A CASE STUDY

by Jonathan Drever

Third International Stroke Trial

“The benefits and harms of intravenous thrombolysis with alteplase within 6 h of acute ischaemic stroke”

• 3035 Patients from 156 Hospitals in 12 Countries• Baseline, 7 Day, 6m and 18m CRFs• Pre and Post Randomisation Brain Imaging

IST-3 Results

Favourable shift; adjusted common odds ratio 1·27 (95% CI 1·10- 1·47), p=0·001 or, the odds of surviving with less disability were 27% greater for patients treated with rt-PA

ChallengesCore Data (CRFs)

Scan Housekeeping

DICOM Imaging Imaging

Reviews

Sub- Study

Sub- Study

Sub- Study

Data Eccentricities• Legacy Data

• Blinded / Startup Phase of the trial• Paper Scan reads• Empty fields• Pre and Post Calculated Values

• Duplicated Data

• Administration Data• Form sent and received dates• Centre activation dates

• Missing Data• Form version change• NULL, 9 and Unknown

Documentation

1. Plain English

2. Bullet Points

3. The Devils in the Detail

Annotated Form

Form Versions

Creating a Basic Dataset

• Benefits• Simpler to understand and analyse• Cleanest most complete data • Snapshots• Concentrate your efforts

• IST-3 Dataset• Up to 6m• Very limited scan data• Used for main trial paper and further meta-analysis• More use in future…?

Archiving (Paper)

• Long Term store (15yrs)• Resources (where?)• Huge amounts of paper (114 boxes)

• Only 1 box was the electronic files

Archiving (electronic)

• Bringing it all together• Investigation• Cleaning• Documentation• Basic Dataset

• Archiving Catalogue• All the electronic repositories• Rough view of contents• Points to more detail

Archiving (Technicalities)

Next Steps

• 2nd Round of Archiving

• Further Analysis

• Data Sharing• Updated ‘Basic Dataset’• What documentation is needed?• Where will it do? (Edinburgh DataShare)• Who will get access?

Take Home Messages

• Planning early

• Simple documentation

• Value of a basic dataset

• Factor in your resources

• Be pragmatic

• Its not just for Archiving