Introduction
Real world data (RWD) analysis of electronic medical records (EMR) for
comparative efficacy and safety assessment is
challenging.1 Two key challenges are factors relating
to the data itself and the analytical methods applied on the data.
Predominantly observational in nature, RWD is rarely collected and
organized in a form suitable for analysis. Disparate data coding
standards, database schema architectures and vocabularies – observable
even within one organization – can hinder accurate
analysis.2 Adopting a common data model (CDM) may
address some of these data challenges.
Converting source data into a CDM creates a copy of the original data,
reshapes it to fit the CDM structure. Source data elements are
translated to the standard vocabularies (e.g. RxNorm for medication data
and SNOMED-CT for diagnosis data, as specified by the CDM) and columns
from various source data tables are split or merged, to fit into target
table columns in the CDM.3,4 CDM conversion of EMR has
several merits – the most obvious of which being the ability to conduct
multi-database studies and pool results for obtaining valid inferences
to important clinical questions.5,6
However, converting data from EMR sources to CDM can be challenging. The
value of conversion may not be as apparent to individual institutions as
it is to the research community intending to carry out multi-centre
analyses. There are however potential benefits of locally analysing a
single healthcare system’s data – particularly if the system serves a
unique patient population. Healthcare providers often struggle to apply
new medical findings to their own patients because the evidence
regarding efficacy and safety of a given drug would have been generated
by studies involving patients with different clinical characteristics
and who were studied under highly controlled research
settings.7,8 An introspective analysis of an
institution’s EMR data in the post-market setting can generate insights
that are directly relevant to the patients that the institution serves.
In this study we explored the usefulness of the CDM by converting EMR
data from a tertiary care facility in Singapore, into the Observational
Medical Outcomes Partnership-Common Data Model (OMOP-CDM). Potential
advantages of converting to the OMOP-CDM include its large and active
user community and its emphasis on open source software use and analytic
code sharing and peer review. This could address another key challenge
of RWD – its appropriate analysis. Upon conversion, we illustrated the
possibilities that CDM conversion can potentially offer by delving
deeper into a use case involving oral anticoagulants (OAC) in patients
with atrial fibrillation.