TB Notes Newsletter
No. 4, 2006
Clinical and Health Systems Research Branch Update
Using a Private Claims Database for TB Health Services Research,
Evaluation, and Analysis
The Institute of Medicine (IOM) report on TB elimination in the
United States highlighted several challenges and opportunities for
TB prevention and control related to the use of managed care systems
and privatization of health services. A DTBE study revealed that
of the 733 patients hospitalized for TB in 1996, incurring total
costs of $11,263,853 (1999 dollars), approximately 12% were covered
by private insurance (which paid 9% [$1,038,759] of the total costs),
21% by Medicaid, 10% by Medicare, and 6% by both Medicaid and Medicare.
However, the insurance status of all TB patients (including outpatients)
and the cost of their care was and remains unknown. The IOM report
anticipated that the percentage of TB patients cared for in the
private sector was likely to grow, because increasingly Medicare
recipients are insured by private for-profit organizations and Medicaid
recipients are cared for by Medicaid managed-care organizations.
In addition, some public health departments contract TB care to
private organizations or companies. Understanding the role that
private insurance plays in TB care can provide valuable guidance
for possible interventions.
When an administrative claims and encounters database of private
employers (MarketScan) became available for use by DTBE, the Health
Systems Research Team accessed it to attempt to answer several research
questions related to TB care provision in the private sector. The
methods and results of this exploratory project are presented below.
In summary, though we were unable to answer the research questions
using the Marketscan data, we describe the challenges we encountered
and suggestions for addressing them.
The MarketScan database is a large, multisource, longitudinal database
of inpatient and outpatient insurance claims and encounter information
of individuals covered by employers’ benefit plans. The database
was created and maintained by Medstat, a for-profit health care
information company (www.medstat.com). While MarketScan does not
include Medicaid managed care providers, it does include Medicare
recipients and others covered by private insurance. Previously,
the MarketScan database has been used by CDC researchers in the
National Immunization Program and the Division of STD Prevention,
among others, to examine costs and service utilization patterns.
Approximately 65 employers and 200 payers, including commercial
insurance companies, contribute data. MarketScan links medical procedures
and prescription data with provider descriptions, patient enrollment,
and benefit plan information. The database retrospectively captures
10 years, which includes approximately 3.6 million persons, 75 million
services, and prescription drug information for 2.8 million covered
The primary objective of this project was to demonstrate the utility
of the MarketScan database in describing TB service provision in
private settings. Five research questions were initially posed:
- Who (in terms of age, sex, region, employment status, and industry)
receives care for active TB disease from private providers?
- What types of private providers perform TB diagnosis and treatment?
- What are the duration of care for TB disease, types of services
provided, procedures performed, drugs prescribed, and insurance
payments to the provider?
- What is the cost of TB care?
- What are the rates of TB screening and LTBI treatment among
persons at risk for TB disease?
The initial steps were to identify which years (1997–2002)
and TB-related “procedural” and “diagnostic”
codes (Current Procedural Terminology (CPT) and International Classification
of Diseases, 9th Revision (ICD-9)) codes to include. We obtained
data through the specialized Medstat software, Dataprobe, and created
SAS datasets. The following are some of the challenges we faced
finding and describing true TB cases from the MarketScan data:
- Enrollment data, the sole source of stable demographic data
and enrollment dates, were available for only a portion of total
patients (50%-60% in the years selected).
- Diagnostic test results were not included in the database.
- Medication data were available for only a portion of patients
(50%-60%) for the specific years included in the analysis.
- The inpatient admissions record listed one principal ICD-9 diagnosis
code and 15 secondary diagnosis codes chronologically (i.e., not
in order of “clinical importance”).
- Any TB diagnostic code may have been applied during patient
evaluation when TB is suspected, and tests are ordered to confirm
or rule out TB.
- We were unable to determine if a TB patient was referred to
the public sector for diagnosis or treatment.
- We were unable to differentiate suspected TB cases from true
TB cases since we could not match to a confirmed TB case registry.
There were no shared identifiers that could be linked to an existing
We used various strategies to identify TB cases and isolate claims
related to TB treatment:
- We completed an extensive process of determining best “criteria”
for identifying TB cases, with expert DTBE physician consultation;
used a claims algorithm with CPT codes, ICD-9 codes, and number
of intervening days (specifically, presence of TB procedure code,
a TB diagnostic code within the 60 days after, and another TB
diagnostic code within 180 days after the first).
- We identified 190 possible TB patients (35 inpatients and 155
outpatients) who received some amount of care (as indicated by
our algorithm) and described them by age, sex, geographic region,
industry, and employment status.
- We randomly selected 10 of the 190 potential TB patients, then
asked DTBE physicians to perform a detailed review of each patient’s
claims history within 3 months of initial and last TB ICD-9 code
to assess the validity of our algorithm, and evaluated the reviewer’s
- In a separate query of the MarketScan database, we identified
100 inpatients having TB as the principal diagnosis on the discharge
record and conducted another in-depth claims review of records
linked with hospital stays.
These strategies were unsuccessful in identifying true TB patients
and TB claims. New and significant limitations were identified at
this stage. First, our algorithms for identifying TB patients could
not be validated. For example, the 35 inpatients identified using
our claims algorithm and the 100 inpatients identified using hospital
discharge records were not the same; there were only 15 overlapping
patient ID numbers. This suggested a lack of validity in our case
criteria, assuming that TB indicated on the hospital discharge record
was the “gold standard” (i.e., that a hospital discharge
record with TB as the principal ICD-9 code should be that of a TB
true patient, even if some true TB patients do not have a TB ICD-9
code on their hospital discharge record). The detailed claims review
(independent of hospital discharge data) yielded concerns about
the validity of the case criteria among outpatients as well; the
DTBE physicians who examined the claims records concluded that 9
out of 10 were unlikely to be true TB patients.
The second challenge was the relationship between the CPT and ICD-9
codes. We originally aimed to extract specific claims associated
with TB care to conduct episode-of-care analysis for both inpatients
and outpatients. It was expected that a claim with a TB-related
CPT code (e.g., 87116, “Culture, tubercle or other acid-fast
bacilli”) would usually be linked to a TB diagnostic code,
but this was not observed in the data. Conversely, the TB ICD-9
codes were not consistently associated with procedures that are
relevant or specific to TB diagnosis or care. Upon examination,
we found no clear and consistent linkage between possible TB CPT
and ICD-9 codes; thus, we did not identify a mechanism to isolate
TB claims of true TB cases.
Because there was no link to a TB registry, we could not determine
when TB treatment started or ended to estimate a relevant time period
for TB-related services or costs. Also, we could not calculate the
TB screening rates among persons at risk for TB disease because
it was not possible to identify an accurate denominator of patients
with diseases that place them at risk for TB (e.g., HIV, silicosis,
diabetes). Rather, we could only ascertain whether someone was receiving
care for that condition while insured. For example, a diabetic patient
may not have had a tuberculin skin test (TST) CPT code because he
or she was tested before or after the claim reporting period (and
enrollment dates were available for only half of the population).
No TST results were available, so we were unable to use these data
to determine TST positivity rates.
We reflected upon the study research questions and concluded that
we were unable to answer them with the MarketScan data. This was
mainly because we could not identify individual TB patients or patients
at risk for TB solely through the MarketScan database.
However, we offer a few suggestions for meeting these challenges.
It might be possible for states or local areas, through negotiation
and confidentiality agreements with large private employers, to
link the MarketScan data to TB patients or TB suspects in their
area. Alternatively, local TB programs might be able to access the
MarketScan data for their geographic area and could attempt to match
TB patients, without linking the databases, by patient employment
status, county, gender, and year of birth, though the combination
may not yield an exact match. After identifying a cohort of TB patients
or TB suspects, TB programs would need to determine if any TB-related
diagnostic codes or CPT codes are documented in the claims histories
for matched patients. If analysis is limited to known TB patients,
programs could describe the characteristics of who is served by
private insurers, types of private insurers that TB patients use,
and duration of and type of care received (i.e., research questions
1, 2, and 3). Some of the limitations mentioned above would still
pose challenges to analysis. However, this strategy would facilitate
analysis of the costs of TB care in the private sector, which is
lacking in the literature, and might permit us to assess whether
the movement towards provision of TB care in the private sector
is cost efficient for society.
1. Institute of Medicine. Ending Neglect: The Elimination of
Tuberculosis in the United States.
Washington D.C.: National Academy Press, 2000:73-74
2. Marks SM, Taylor Z, Miller BI. Tuberculosis prevention versus
hospitalization: taxpayers save with prevention. The Journal
of Health Care for the Poor and Underserved. 2002;13(3):392-401.
3. Draper DA, Hurley RE, and Short AC. Medicaid managed care: The
last bastion of the HMO? Health Affairs. 2004;23(2):155-167.
—Reported by Heather Joseph, MPH
Div of HIV/AIDS Prevention, and
Suzanne Marks, MPH, MA
Div of TB Elimination