Grant Details
Grant Number: |
5R01CA071533-04 Interpret this number |
Primary Investigator: |
Penberthy, Lynne |
Organization: |
Virginia Commonwealth University |
Project Title: |
Using Claims Data for Cancer Surveillance |
Fiscal Year: |
2000 |
Abstract
DESCRIPTION: The overall purpose of this study is to examine the utility
and validity of linking data from three claims-based sources to the Virginia
Cancer Registry (VCR) for cancer surveillance. The study will focus on the
five leading cancers in Virginia: breast, cervical, colorectal, lung, and
prostate. The three claims-based files to be linked to the VCR are:
Medicare, Medicaid, and the statewide hospital discharge summary files.
Because of its high incidence, devastating impact, and potential
preventability, monitoring cancer epidemiology is essential. An effective
cancer surveillance program could help track groups at high risk for the
disease and assess the value of interventions, such as screening.
Surveillance mechanisms must produce information in a timely fashion to be
useful to policy makers who are deciding about allocation of limited health
care resources.
Claims files offer an important potential source of routinely available,
population-based, computer readable information that could supplement the
cancer surveillance activities of statewide registries. These databases,
however, have the limitations of minimal clinical content and association
with billing activities. Accuracy of diagnosis coding is a particular
concern, although one study indicated good accuracy for cancer diagnoses.
Despite these limitations, linking claims files to cancer registries could
capture more cancer incident cases and add to understanding of cancer care.
Four databases will be linked in this study: 1. The Virginia Cancer
Registry (VCR). Until 1990, reporting to the VCR was voluntary and included
half the hospitals in Virginia; starting in 1990, reporting of cancer
incident cases became mandatory. About 85% of the cases are reported by
hospitals that include complete staging data; 2. Medicare files for Parts A
and B, including all institutional and noninstitutional bills; 3. Virginia
Medicaid files, including inpatient, outpatient, and pharmacy claims. The
Medicaid files contain a large number of minority (46% black) and high risk
patients; and 4. Virginia Health Information (VHI) files. Since 1993, VHI
has maintained inpatient claims for all admissions to Virginia hospitals.
To address the first Specific Aim, the latter three claims-based files will
be linked to the VCR at the person level using AUTOMATCH, a product of Match
Ware Technologies for probabilistic record linkages.
The second specific Aim involves validating methods for identifying cancer
incidence from the three claims files and assessing the reliability and
validity of incidence and treatment data in the four data sources. Data
will be abstracted from inpatient medical records and outpatient health care
providers, and from reviewing laboratory, radiation treatment, and
outpatient chemotherapy logs. A stratified random sample of 2,750 patients
will be drawn, with the sampling protocol considering the likelihood that
patients will be flagged as a cancer case by more than one of the data
sources. Strata were created based on the data source of the sampled case
(Figure 1, p. 52); sampled cases will be clustered within hospitals.
Abstraction of inpatient records will be performed by the Virginia Health
Quality Center, the state peer review organization. Collection of
outpatient information will have four approaches: 1. A survey will be
mailed to physicians from the VCR asking about the treatment given to a
specified patient; 2. When physicians fail to respond to the mail survey,
the hospital tumor registrar will contact physicians to get them to respond
by mail or agree to a telephone interview; 3. Outpatient medical records
will be abstracted for a random 10% sample of the 2,750 patients. The
purpose of the validation is to judge the accuracy of the physicians' mailed
responses. A VCR representative (a trained RN nurse abstractor) will
conduct these reviews; and 4. For each hospital in the cluster sample, logs
from pathology, radiotherapy, and chemotherapy units will be reviewed.
Information from the primary data collection will be compared to that of the
four data sources to assess their accuracy. The primary data will serve as
the "gold standard" in determining the sensitivity and predictive values
positive and negative of the four data files. The areas that will be
examined include the definition of incident cases and initial surgical and
nonsurgical treatment. An analysis will be performed to assess the
representativeness of the different data sources in identifying cancer
cases, using as a framework a Venn diagram showing potential overlap and
discordance. The completeness of a surveillance approach combining all data
sources will also be assessed by estimating the frequency of missed cases
using capture-recapture techniques developed by naturalists. To assess
further false negatives and the value of the capture-recapture method, a
pilot study will be performed at Medical College of Virginia and three
associated rural hospitals, combing all primary data sources and billing
data to identify all cancer cases.
The outcome of this project will be a clear sense, at least for Virginia,
about the utility of linking administrative data files to a cancer registry
for examining the incidence and initial treatment of cancer.
Publications
The added value of claims for cancer surveillance: results of varying case definitions.
Authors: Penberthy L.
, McClish D.
, Manning C.
, Retchin S.
, Smith T.
.
Source: Medical care, 2005 Jul; 43(7), p. 705-12.
PMID: 15970786
Related Citations
Using Medicare claims to identify second primary cancers and recurrences in order to supplement a cancer registry.
Authors: McClish D.
, Penberthy L.
, Pugh A.
.
Source: Journal of clinical epidemiology, 2003 Aug; 56(8), p. 760-7.
PMID: 12954468
Related Citations
Using hospital discharge files to enhance cancer surveillance.
Authors: Penberthy L.
, McClish D.
, Pugh A.
, Smith W.
, Manning C.
, Retchin S.
.
Source: American journal of epidemiology, 2003-07-01; 158(1), p. 27-34.
PMID: 12835284
Related Citations