Information Technology Reference
In-Depth Information
10-1
10-1
0 1
-1
EMR
Genomic
Claims
Quantified Self
P-1
1
0
-1
P-2
:
:
:
:
P-n
thousands
millions
hundreds
engineered
Features
Fig. 7.2
Data can be “big” along three dimensions; the number of samples (
rows
), the number of
features (
columns
), and temporal coverage (frequency at which data are collected, and how far
back the data go). Different sources have different shape along these three axes. For example,
genomic data have millions of features (e.g., SNPs) but are available for a small number of people
(currently in the thousands). Medical billing and claims data have fewer features (e.g., ICD-9
codes, laboratory codes) but are available for millions of individuals over a much longer period of
time
and adverse events in electronic medical records could identify six out of nine drugs
recalled in the past decade, roughly 2 years ahead, when incorporating the time
dimension into the analysis [
13
].
When we cross a certain data-threshold the kinds of questions we can ask
changes, consider the example by Frankovich et al. [
14
]. Where the existing litera-
ture and evidence based guidelines were insuffi cient to guide the clinical care of a
patient, Frankovich et al. applied trend analysis to the EMR data from 98 patients to
“learn” a data-driven guideline on how to provide care for a 13-year-old girl with
systemic lupus erythematosus (SLE) [
14
]. In terms of size, the Electronic Health
Record data from 98 patients are certainly not “big” as would be the case with full
genome sequences from 98 individuals. However, such approaches, which analyze
data that are already routinely collected, are particularly valuable when a formal
guideline is not available or feasible from a practical standpoint—we refer to such
approaches as practice-based evidence [
15
].
As another example, Leeper et al. show that it is possible to examine the out-
comes of decisions made by doctors during clinical practice to identify patterns of
care that are optimal—generating evidence based on the collective practice of
experts. The authors examined the validity of a black-box warning on an effective
drug (Cilostazol) for peripheral vascular disease. Cilostazol is the same kind of drug
as Milrinone, which in a study in 1991 was found to be associated with sudden car-
diac death when used to treat heart failure; as a result, Cilostazol “inherited” a
Search WWH ::
Custom Search