Skip Navigation Bar
 

Storing Results Data

Storing Results Data

Clement McDonald, M.D.

Director, Lister Hill Center National Center for Biomedical Comm Communications unications

National Library of Medicine

February 11, 2008

Scope

  • The focus is on summary data of the kind one sees in publication
  • Characteristics of patients at base line
  • Differences between the outcomes of the various study arms
  • Not patient level data

Goals

  • Represent this data in a generalized but “structured structured” format
  • Take advantage of what is in the registry data base to ease entry of results data
  • Minimize entry work
  • Accommodate most study designs

Three parts to the problem

  • Define the variables - used in the study
    • To characterize population
    • To measure outcomes
    • Etc.
    - E.g. Age, creatinine, PHQ PHQ-9

For each variable

  • Name
  • Description
  • Data type (e.g. continuous, categorical, time to event)
  • Additional informatino depending upon data type e.g.:
    • For categorical variables - the categories
    • For continuous variables - the units of measure

Define the study arms and phases.

  • Name
  • Description
  • E.g.
    • Arm 1 - treat with new drug or device X
    • Arm 2 - placebo
  • Statistical analyses – as descriptive text.

Define tables

  • Rows variables
  • Columns study arm phase - and one column for statistical analysis.
  • List the discrete summary observations per cell depends on data type
    - E.g. for continuous
    • N
    • Mean
    • Coefficient of variation
    • (more)

Data structures

  • XML for current testing phase

As relational data base - simplest incarnation

  • Table for study – with record per study
  • Table for variables - one record per variable
  • Table for study arms/phases - one record per study arm phase
  • Table for overview of each tables – one record per cell
  • Table to define each table - one record per cell

Non trivial problem

  • Especially for newer study designs