Administrative Claims Data and Augmented Data
Proctor and colleagues used the Centers for Medicare & Medicaid
Services Chronic Conditions Data Warehouse to identify TG Medicare
patients in 2013 (n=4098).17 This study utilized a CP
of ICD-9 diagnosis codes that were relevant to TG status, then
subsequently addressed concerns of coding errors by validating their
method through a specific logic. This logic required that patients with
an initial ICD-9 diagnosis code relevant to TG status must have at least
one or more of the following: more than 1 claim of an ICD-9 diagnosis
code relevant to TG status in 2012, 2013, or 2014, an unspecified
endocrine disorder code used by providers for TG patients to avoid TG
stigma through health insurance, a sex hormone prescription in 2013, a
principle diagnosis code from the ICD-9 diagnosis codes relevant to TG
status, or a billing claim condition code 45 modifier or KX modifier.
Proctor and colleague’s CP found a sensitivity of 89.26% though their
internal hierarchal method in identifying TG patients using Medicare
insurance.17 Their study was able to find 66.03% of
TG Medicare patients using ICD-9 diagnosis codes relevant to TG status
in 2013 alone. Approximately forty percent were identified with similar
claims in 2012 and 2014, as well as through sex hormone prescriptions or
a principle diagnosis code specific to TG status.17
Jasuja and colleagues uses a similar approach to Proctor and colleagues
by using a hierarchal method of claims data in order to validate their
CP.17,19 Jasuja and colleague’s retrospective analysis
of administrative claims from OptumLabs Data Warehouse from 2006 to 2017
identified 27,227 potential TG patients. They initially use gender
identity disorder diagnosis codes, then incorporates endocrine
non-specific codes, procedure codes relevant to TG status, then sex
hormone receipts discordant with sex recorded in the claims data as a
validation method to improve their accuracy of identifying their TG
patients. To enhance their positive predictive value (PPV) even further,
they required non-specific endocrine disorder codes to not be followed
by non-diabetes codes such as thyroid or adrenal diseases and utilized a
technical panel of experts to categorize a list of procedural codes that
could be used for TG patients undergoing gender affirmation surgery.
This study specified a minimum dosage for hormone replacement therapy to
exclude non-TG patients who may receive these prescriptions at a lower
dose. Using ICD-9 and ICD-10 gender identity disorder diagnosis codes
alone, they were able to find 69% of TG patients. The added internal
validation method of non-specific endocrine disorder codes with
TG-related procedure codes added 16% TG people, and non-specific
endocrine disorder codes along with sex-discordant hormone prescriptions
added another 15%. They were also able to remove 1.2% of patients from
the overall TG cohort after validation methods revealed their procedure
or hormone prescription codes did not align with gender identity status,
such as an estrogen prescription along with a transmasculine-identified
procedure code (e.g., bilateral mastectomy).19
Yee and colleagues also used a hierarchal method approach within Oregon
Medicaid claims data from 2010 to 2019 to identify TG patients with at
least one ICD-9 or 10 gender identity-related diagnosis
code.18 In their approach for confirming additional
details once a patient was identified as TG, they differentiated sex
assigned at birth (SAB) information from self-reported gender in the
enrollment file. They operationalized this by determining whether a
patient’s procedural or medication codes differed from their recorded
SAB. As this method relied on gender identity-related diagnosis codes as
entry into their study cohort, 100% of their patients had a diagnosis
code. Of the 2,940 beneficiaries identified as TG, they were able to
confirm 92.1% as TG using the hierarchy method described by Proctor et.
al.17 They also used a sensitivity analyses that
included all changes in recorded gender and found an additional 16%
transmasculine and 21% transfeminine patients.
These reviewed studies provide evidence that CPs applied in claims data
with the strongest sensitivity and PPV contained ICD-9 or ICD-10 gender
identity disorder codes along with additional diagnosis codes and
procedural codes, although diagnosis codes were able to identify most TG
patients.