Re: possible usecase: tables for medical clinical trial data

While this format which is what gets published tends not to follow
relational conventions, the raw format that is used behind the scenes
during clinical trials by companies like MediData, and CROs, as well
as submission data to the FDA is very much relational. Unfortunately,
I don't have any public examples I can share right now. I will try to
dig some out.

Thanks,
Yakov

On Wed, Feb 26, 2014 at 7:53 AM, Tim Finin <finin@cs.umbc.edu> wrote:
> In our research on interpreting tables and generating LOD from them,
> we've been looking tables that represent medical clinical trials.  The
> attached pdf document shows some in medical journal articles.  I've
> not yet found a good source of CSV files for such tables, but I'm
> confident that there we can find many examples.
>
> These represent another class of examples where the table does not fit
> well into the simple relational example where each row in the table
> represents an entity and the columns map to entity properties.
>
> Here's a sketch of the format, which is very common for clinical
> trials. The table describes a clinical trial as a whole.  Each row is
> one subset of the subjects which can overlap with other subsets. The
> first column or two columns specify a property (e.g., sex, age, died)
> and value or value range (e.g., male, <10) of the subsets.  The other
> columns describe different treatments (e.g., drug a, drug b, placibo)
> and their values are often the number of patients and/or their
> percentage of overall population.
>
> The tables can get more complex, but they seem to follow a few common
> patterns which I assume you learn in your first year in medical
> school.

Received on Wednesday, 26 February 2014 13:48:07 UTC