Data4Med -- Analyzing real-life health-data to suggest negative and positive effects of existing treatments |
Main Page | Results so far | About |
How good is Metformin? Asking the question to the OpenDamir -- Hacking Health Camp, Strasbourg, March 2015
Introduction to the research project
In the main page of this website we compute 3 health indicators by age
and we cross them with the use of chronic treatments:
we hope to see if people who take a chronic treatment "A"
have lower or higher health risks than people who instead take chronic treatment "B" under similar conditions,
so that we can build a first estimate of comparative therapeutic effects.
However this is a hope, and there are many potential bias.
Indeed, using the OpenDamir to study health effects of medicines in real-life is ambitious :
the format was not designed for such a study. "We do what we can with what we have",
but the indicators we can compute have a variety of potential bias. For example:
- the data does not allow to follow people over time so
we can not distinguish between causes and consequences
-- we can only interpret due to a given context: for example that people where diagnosed with a disease
before they started taking a corresponding chronic treatment.
- we do not compute risks per person and unit of time, as actuaries do when building mortality tables;
rather we compute risks per medical treatment and unit of time
-- if the considered treatment is bought regularly for the considered population
then it is equivalent to a standard actuarial table;
but if not then ill persons will be overly represented.
Therefore, it is not a priori obvious whether the indicators make any sense.
Instead of theorizing, we thought of asking the question very concretely:
we thought of testing the indicators against specific results that either
seem to make sense or stem from the scientific literature.
Results