Cephalosporin.” More specifically, this is a process measure in
the domain of patient safety which measures patients aged 18
years and over that are undergoing certain procedures that have
an indication for a first- or second-generation cephalosporin
antibiotic, who had one ordered for antimicrobial prophylaxis.
The numerator of this metric measures the number of surgical
patients that have had a first- or second-generation cephalosporin for antimicrobial prophylaxis ordered. As such, a healthcare
organization must have a way to identify surgical patients by the
antibiotics they are taking within the timeframe listed.
In the denominator, the metric requires all surgical patients 18
years of age and older undergoing procedures with the indications for a first- or second-generation cephalosporin prophylactic antibiotic. In addition, healthcare organizations must be able
to factor in exclusions that include:
Patients enrolled in clinical trials
Patients with physician/advanced practice nurse/physi-cian assistant (physician/APN/PA)–documented infection
prior to surgical procedure of interest
Patients taking antibiotics more than 24 hours prior to surgery except colon surgery patients
Other medical reasons (G9196)
This measure can be reported through claims data or certified
registries. The claims data must contain a specific G code (G9197)
that indicates the patient has had an appropriate order written.
In many cases, this requires a manual assignment of the G code
by a certified coding professional. In order to automate this process and reduce errors from human intervention, it is important
to identify the presence of the appropriate order at the appropriate time or evidence of the patient taking antibiotics more than
24 hours prior to surgery. This information must be captured in a
structured way and then translated into the appropriate G code to
be captured in the patient claim data.
In the medical record, medication data can be captured data using proprietary drug databases such as Medi-Span or FDB. It could
also be captured using RxNorm or NDC codes. The question becomes: how do you identify the medication order, understand if it
fits the description of first- or second- generation cephalosporin,
and ensure that G9197 is entered on the patient claim?
Additionally, healthcare organizations often struggle to efficiently identify qualitative information such as evidence of prior
infection. While much of the data found within the EHR are captured through the use of industry standards such as ICD- 10 and
SNOMED CT, this information is often located in free text. Note
that the requirement is for a clinician to document prior infection. This information might (or might not) be included in a problem list that may or may not be codified to SNOMED.
The Free Text Challenge: What Can Be Missed
Many data governance strategies lack an effective way to extract
unstructured patient data. One study found that when only struc-
tured EHR data was used to derive quality measures, practice
performance was undercut when compared to a manual review
of electronic charts that included unstructured patient narrative. 1
There are many examples where providers and payers can
miss patient reporting opportunities. For example, MIPS measure 005 (NQF 0081) considers the use of angiotensin converting enzyme (ACE) inhibitor or angiotensin receptor blocking
(ARB) therapy for patients with documented ejection fractions
of less than 40 percent. The quantified ejection fraction is rarely
documented in a structured form, and the inability to find this
data will skew the measure reporting.
Quality measure PQRS 116 (NQF 58) is another example,
where providers risk lower performance scores when free text
is not factored into the measurement. Used to measure patients who inappropriately receive antibiotics for acute bronchitis, the metric includes exclusion criteria for patients who
have a secondary condition, such as cystic fibrosis or HIV. Often, documentation demonstrating the secondary diagnoses is
found in free text as opposed to structured areas of the EHR.
Health insurers reporting quality measures through the
Healthcare Effectiveness Data and Information Set (HEDIS)
program also want to ensure they are accurately identifying
all inclusion and exclusion criteria in free text to obtain the
highest CMS star ratings and reimbursement. This requires
access to clinical data found in EHRs as well as free text notes.
Improving the Outlook
Many reasons exist for defining patient cohorts beyond quality
measures reporting. The success of any of these efforts rests with
a healthcare organization’s ability to accurately and completely
identify all patients with the pre-defined attributes within a cohort.
Data normalization strategies help healthcare organizations
overcome these challenges. Otherwise, providers and payers
have no way of identifying patients who fit pre-determined criteria without manually combing charts. Technology is an important consideration, and the right platform can address both
structured and unstructured patient data, ensuring patients
are not excluded from patient cohort analytics. Advanced solutions exist that automate and streamline the complexities of
data normalization by addressing the following:
Content—establish a single source of truth for all terminol-ogy-related maps, value sets, and code sets
Applications—enable interoperability and increase the
quality of analytics
Web-based APIs—integrate reference data into existing
data warehouses or analytics platforms ¢
1. Parsons, Amanda et al. “Validity of electronic health record-derived quality measurement for performance monitoring.”
Journal of the American Medical Informatics Association 19,
no. 4 (July-August 2012): 601-609. www.ncbi.nlm.nih.gov/
Cheryl Mason ( firstname.lastname@example.org) is director of clinical informatics consulting at Wolters Kluwer, Health Language.