Query+Health+Hypothesis+Generation+User+Story+Proposal

include component="page" wikiName="siframework" page="Query Health Header"



This proposal was submitted by David McCallie. [UPDATED 10/4/2011 by DPM, to replace my notions with a real article from the literature]

This use-case is around what some would call "hypothesis generation" -- usually done for research purposes, but also applicable to safety monitoring, comparative effectiveness research, outcomes research, etc. Some examples that might fit: Earlier detection of the birth defect side effects of Thalidomide, or seeking evidence of suspected cardiac side effects of COX inhibitors, etc.

The core technical capability necessary is the ability to find **temporal patterns** in the data where the patterns are spread over longer periods of time than just a single encounter. These time-based-pattern queries can be challenging for some kinds of data models, so I think we ought to ensure that we have at least one example like this in our list of use-cases.

Rather than make up a fake user story, I suggest we use a recently published clinical study done at Partners, in Boston. This study was referred to me by Sean Murphy, one of the physicians behind the i2b2 work that was presented during the Summer Concert Series. Their study was performed against Partner's clinical data repository, and represents an achievable (though challenging) goal for Query Health.

The paper's title is: //Rapid Identiﬁcation of Myocardial Infarction Risk Associated With Diabetes Medications Using Electronic Medical Records// and is [|available here.]

In the referenced study, a retrospective cohort of patients were identified who had received one of four particular diabetes drugs and had subsequently sustained a myocardial infarction. After excluding various non-drug risk factors, the investigators were able to find evidence that one of the four drugs was associated with an increased relative risk of heart attack, compared to the other three drugs.

This is a **hypothesis**, of course, and in general would need to be confirmed with additional studies, which might be outside the scope of Query Health.

Here are some excerpts (in blue) that define various ways that they used the CDR to perform the study:

The retrospective cohort analysis (n=34,253) included all patients aged 18 years identiﬁed by an ICD-9 code for Diabetes Mellitus (250.XX) or an A1C of 6.0% and at least one record of prescription of an oral diabetes medication as an outpatient or dispensation as an inpatient, between 1 January 2000 and 31 December 2006. Analyses focused on three classes of diabetic medications: sulfonylureas, the biguanide metformin, and the thiazolidinediones, rosiglitazone and pioglitazone. Evidence of insulin therapy did not exclude patients but was adjusted for in multivariate models and used for stratiﬁed analysis (described below). We excluded patients receiving either metformin or thiazolidinedione who had a diagnosis of polycystic ovaries but not diabetes. We used a cumulative temporal approach to ascertain the calendar date for earliest identiﬁable risk associated with rosiglitazone compared with that for other therapies.

For each patient, duration of exposure to individual diabetes medications was assessed in 6-month increments during which only one of the four medications was prescribed. Patients receiving multiple medications under consideration were excluded.

Events were associated with a particular medication only when the prescription or dispensation occurred within 6 months before the documented myocardial infarction. If a patient did not have any activity for a 6-month observation period but resumed activity in the following period, than the particular 6-month observation period with no activity was excluded from analysis.

Analysis was repeated considering only patients having been prescribed one of the four medications, considered to be monotherapy. Finally, we also performed stratiﬁcation of our data to analyze patients who had not received insulin as outpatient therapy

I think the cited study is a good example of what I mean by "hypothesis generation" and clearly demonstrates the power of a properly-enabled Query Health model.

include component="page" wikiName="siframework" page="space.template.inc_contentleft_end"