Pearson et al1 from RAND Health and the RAND Statistics Group, Santa Monica, Calif, studied the difficulties involved in identifying a population of cancer patients so that the quality of cancer care could be measured. The delivery of high quality cancer care was considered by this group to be a US health care priority. The investigators noted that it is impossible to measure the quality of care unless cases have been properly identified, since failure to detect cases of a particular type would alter the apparent success rates of various cancer interventions.
Case identification, however, remains a basic challenge, and better case-identification strategies are needed. Two particular problems are faced by those attempting to assess breast-cancer treatment outcomes. The first is the identification of all patients who have a breast-cancer diagnosis. The second is the exclusion of patients who do not have that diagnosis, even if they were initially evaluated for suspected cancer.
It may be possible to identify newly diagnosed breast-cancer patients using information from insurance claims and payors’ administrative reporting systems, case registries, pathology reports, and the patient records of oncologists and other physicians. Claims data exhibit a sensitivity of 75% to 95% in detecting cancer cases; certified cancer registries must be at least 90% complete. Unfortunately, both of these sources are flawed because the ability to obtain comparable, valid data across multiple settings of care is limited.
The accuracy and timeliness of data would be improved substantially if existing data-collection systems were staffed by managers and data collectors who had enhanced skills. The adoption of quality-of-care analysis as a major goal by cancer registries (and the agencies that support those registries financially) would be very important in furthering the cause of cancer outcomes assessment.
Registries need to be able to find cases rapidly, collect adequate patient-contact information, and verify the patient’s stage of cancer at the time of diagnosis. Additional support for cancer registries is highly likely to be necessary if they are to meet the increasing demand for quality-of-care information that is accurate and up-to-date.
IMPROVING RELIABILITY
Researchers at UCLA also conducted a critical review2 of studies covering breast-cancer patterns of care. They simultaneously reviewed the sources of data used in those studies to identify cases and to determine patterns of care. In August 2001, the investigators searched MEDLINE (the National Library of Medicine’s online journal index) for research on breast-cancer care published between January 1, 1985, and June 30, 2001. They found 32 studies described in 38 articles.
These studies indicated that axillary lymph-node dissection was not provided to approximately 10% of women with breast cancer. Hormone-receptor status was not reported for 11% to 26% of breast-cancer patients, and 20% of those who underwent breast-conserving surgery received no radiation therapy. Of patients whose lymph nodes were positive for cancer, 30% to 70% did not have tamoxifen prescribed for them.
Of the 32 studies, 25 (78%) identified breast-cancer patients using cancer registries; data on the process of care came from medical records in 38% of the studies, from cancer registries in 47% of studies, and from more than one of these and other sources in 20%.
Breast cancer has been the subject of more research on patterns of care than any other type of cancer. Nonetheless, the authors described the available data as having many limitations.
This indicates the difficulties involved in conducting and interpreting research on outcomes and underscores the urgent need for more reliable data. The speed with which health care delivery is changing calls for sound quality-of-care information on which clinical and financial policies can be based.
OUTCOMES DATA SOURCES
A US national system designed to monitor the quality of the care provided to cancer patients has been proposed by several groups. A lack of research has hampered the ability to assess whether cancer-registry data are adequate for use as the basis of studies of cancer-care quality. Malin et al3 at UCLA used the California Cancer Registry to evaluate the usefulness of registry data. Registry data covering the care received by patients with breast cancer were compared with information from the medical records of the same patients.
For each study subject, a quality-of-care score was calculated based on four evidence-based quality indicators, as reflected in the registry and medical-record information. The medical record was considered to be an accurate data source, for purposes of comparison with the registry record for the same patient.
Registry data were fairly accurate where hospital-based services were concerned, reflecting 95.9% of lymph-node dissections, 95% of mastectomies, and 94.9% of lumpectomies. The accuracy level of registry reporting for ambulatory interventions was considerably lower, with representation of cases involving radiation therapy at 72.2%; chemotherapy, 55.6%; hormone therapy, 36.2%; and biopsy, 9.8%.
Quality-of-care scores based on the registry data were an average of 11% lower than those based on medical records. The quality-of-care scores also varied according to the patient’s cancer stage at diagnosis. Quality scores based on registry data (versus medical records) were 5% lower for patients with stage I breast cancer, 16% lower for those with stage II breast cancer, and 20% lower for those with stage III breast cancer.
Because the difference in quality-of-care scores was larger for patients with more advanced disease, the setting in which care is delivered and the severity of disease are now considered to affect the validity of registry data.
The valid data needed to measure quality of care may not yet be available from cancer registries. Nonetheless, registries are in a position to provide the infrastructure needed for the collection of valid outcomes data. For this reason, the authors urge the agencies concerned to increase funding for cancer registries so that they can improve and expand data collection.
Kris Kyes is technical editor of Decisions in Axis Imaging News.
References:
- Pearson ML, Ganz PA, McGuigan K, Malin JL, Adams J, Khan KL. The case identification challenge in measuring quality of cancer care. J Clin Oncol. 2002;20:4353-4360.
- Malin JL, Schuster MA, Khan KL, Brook RH. Quality of breast cancer care: what do we know? J Clin Oncol. 2002;20:4381-4393.
- Malin JL, Khan KL, Adams J, Kwan L, Laouri M, Ganz PA. Validity of cancer registry data for measuring the quality of breast cancer care. J Natl Cancer Inst. 2002;94:835-844.