D Jones, Center for Health Quality, Outcomes and Economic Research, Bedford, MA; A Hendricks, Center for Health Quality, Outcomes and Economic Research, Bedford, MA; C Comstock, Center for Health Quality, Outcomes and Economic Research, Bedford, MA; B Chang, Center for Health Quality, Outcomes and Economic Research, Bedford, MA; D Bross, Office of Quality and Performance (10Q), Veterans Health Administration (VHA), Washington, DC; C Sawin, Office of Medical Inspector, VHA, Washington, DC

Objectives: To compare cervical cancer screening rates obtained from administrative data with those obtained from medical record review and estimate the extent to which sampling, coding practices, and exclusion of women who had hysterectomies account for differences.

Methods: We calculated the percentage of female veteran patients who received cervical cancer screening ("Pap tests") one or more times during 1996 to 1998, using Veterans Health Administration (VHA) national administrative data and VHA medical record review. We included patients if they were 21-64 years and received medical care in VHA during fiscal year (FY) 1998. We validated coding for the administrative data with laboratory data from two VHA networks. We estimated non-VHA Pap tests using Medicare claims data for dually-eligible patients. VHA’s External Peer Review Program (EPRP) drew a sample of 3,446 patients from VHA medical centers in FY 1998, including women who had either 3 or more visits to specific primary care clinics or a diagnosis of diabetes, hypertension, or chronic obstructive pulmonary disease. Documentation in the medical record established whether a Pap test was performed at that VHA hospital, another VHA facility, or in the private sector within the last 3 years.

Results: In administrative data, 37% of all female patients seen throughout VHA during FY 1998 (36,250 of 99,044 patients) were screened for cervical cancer at a VHA medical center at least once during FY 1996 to 1998. Adding laboratory data increased the rate 6%. Medicare claims added few tests. The calculated rate for those seen three or more times during FY 1998 was 48% (22,416 of 46,409). With administrative data, excluding women who had had a hysterectomy was impossible. The rate obtained from direct review of medical records was 84% (2,898 of 3,446); this rate was for those with more continuous care and excluded those with hysterectomy. Adding laboratory data increased the rate to an estimated 90%. Administrative data were available for 91% of these patients (3,123 of 3,446); the administrative data-based rate of Pap tests in this group was only 54% (1,699 of 3,123).

Conclusions: Among women seen in VHA during FY1998, the calculated rate of Pap tests ranged from 37% to 90%, depending on administrative coding practices, whether patients had more or less continuous care, and whether those with prior hysterectomy were excluded. Administrative databases, if appropriately constructed, provide information on total populations at relatively low expense; those currently available in VHA gave relatively low rates for cervical cancer screening. Rates based on review of a sample of medical records are costly to obtain and assess only a small fraction of the population; the rate of cervical cancer screening was higher with this method than with the use of administrative data.

Impact: This study has important implications for VHA’s performance measures. EPRP’s sampling strategy maximizes performance measurement by including only high users of VHA primary care. Improvement in VHA coding practices would facilitate performance measurement that is easier, less costly, and more representative of all VHA patients.