1395 Racial Differences in Timeliness of Follow-Up after Abnormal Screening Mammography Sophia W. Chang, M.D., M.P.H.’ Karla Kerlikowske, M.D? Anna Napoles-Springer, M.P.H.’ Samuel F. Posner, P h n ’ Edward A. Sickles, M . D . ~ Eliseo J. Perez-Stable, M.D.’ ’ Medical Effectiveness Research Center for Diverse Populations and Division of General Internal Medicine, Department of Medicine, University of California, San Francisco, California. Department of Epidemiology and Biostatistics, and General Internal Medicine Section, Department of Veterans Affairs, University of California, San Francisco, California. Department of Radiology, University of California, San Francisco, California. This work was supported by an AHCPR grant, no. HS 07373 and NCI-funded Breast Cancer SPORE grant, P50 CA58207. Dr. Kerlikowske is supported, in part, by an American Cancer Society Career Development Award for Primary Care Physicians. A portion of this manuscript has been presented in poster form at the 5th Annual Society for General Internal Medicine Meeting, April 30, 1996, Washington, DC. The authors thank Diane Emler, Cristina Amaya, Stac:y Jones, John Barclay, and the staff of the UCS8FMobile Mammography Program for their help in collecting the information for our analyses. The authors also thank Anita Stewart for her comments on earlier drafts of this manuscriDt. Address for reprints: Sophia W. Chang, M.D., M.P.H., Medical Effectiveness Research Center for Diverse Populations, University of California, San Francisco, 3333 California St., Suite 300 M, San Francisco, CA 94143-0856. Received March 7, 1996; revision received May 28, .1996; accepted May 28, 1996. 0 1996 American Cancer Society BACKGROUND. To determine whether patient race was associated with timeliness of follow-up after abnormal screening mammography, a retrospective record review of diagnostic tests for women with abnormal screening mammography from a Northern California mobile van program was conducted. METHODS. The study included 317 women between the ages of 33 and 85 who were reported to have abnormal screening mammography between July 1993 and May 1994. Measurements included patient demographics, screening mammography interpretation, follow-up diagnostic tests, and dates of diagnostic evaluation. RESULTS. Women with abnormal screening mammography underwent a wide variety of diagnostic evaluations. Nonwhite women had significantly longer time (median time, 19 days) from date of index abnormal screening mammography to final disposition compared with white women (median time, 12 days). This racial difference was primarily due to the longer interval between index abnormal screening mammography and first diagnostic test (median time, 15 days for nonwhite women versus 7 days for white women, P < 0.001). The difference persisted when adjusting for patient age, family history of breast cancer, report of palpable mass, and income. The racial difference was similarly significant for each nonwhite subgroup (African American, Latina, and Asian) when compared with white women ( P < 0.01). CONCLUSIONS. Reasons for less timely follow-up of abnormal mammography among nonwhite women need to be identified. Delays that may be instigated by the patient or be due to her physician or system of care need to be explored further. Cancer 1996;781395-1902. 0 1996 American Cancer Society. KEYWORDS: breast neoplasms-diagnosis, mammography-prevention trol, follow-up studies, time factors, ethnic groups. B and con- reast cancer is the most common cancer diagnosed among all women across all ethnic groups in the United States and is the second leading cause of cancer death.’ Differences in breast cancer mortality by race/ethnicity are striking, notably the higher mortality rates for African-American women‘-5 despite higher incidence rates among white, non-Hispanic (henceforth white) ~ o r n e n .The ~ ’ ~reasons for these mortality differences across race/ethnicities are not completely understood. Studies have noted a strong correlation between mortality and socioeconomic tatu us,^" pointing to factors that may affect mortality rates, e.g., access to health care and comorbid conditions. Other studies have found that breast cancer tends to be diagnosed at a later stage among African-American ~ o m e n , ~ fre’O sulting in a higher breast cancer mortality rate.” There are also indications that tumor biology may differ by race, with African Americans more likely to be estrogen and progesterone receptor negative and to 1396 CANCER October 1, 1996 / Volume 78 / Number 7 have a high S-phase, both prognostic indicators of more aggressive tumors and poorer survival.” Within the past decade, increased rates of screening have been reported across all racial/ethnic g r o ~ p s . ’ ~Recently -’~ published data from the National Cancer Institute show that there has been a decrease in breast cancer mortality of 5.5% among white women from 1989-1992 that may, in part, be due to an increase in mammography utili~ation.~ There has not, however, been a concomitant decrease in breast cancer mortality among African-American women5 despite similar increased rates of mammography screening.14It is unknown if there is differential followup after abnormal mammography screening among racial/ethnic groups and, if differences exist, whether this accounts for later breast cancer stage at diagnosis and higher mortality rates. Given that the diagnosis of breast cancer at an early stage is generally accepted to be one of the most important prognostic indicators for prolonged survival, the timeliness of evaluating abnormal screening results becomes important. This study examined whether raceiethnicity was a predictor of time to follow-up after abnormal screening mammography. METH0DS Subjects and Measurements Data were abstracted from records of the University of California at San Francisco (UCSF) Mobile Mammography Screening Program, which has performed more than 78,000 screening examinations since 1985. The program began recording patient race/ethnicity in mid-July 1993. The program methodology and definitions of database variables collected were wellalong with the screening chardescribed elsewhere,*’ acteristics of the population.18 In brief, the program is an efficient, high-quality screening program with mammographic interpretations performed by radiologists with specialized training in interpreting mammography.lgPatient breast cancer risk profile and clinical history were collected at the time of mammography screening via interview and entered directly into a database. The information collected included age, race/ethnicity, family history of breast cancer, report of palpable mass, zip code, and referring provider. Race/ethnicity was recorded based on staff ascertainment; when a staff member was unsure, race was clarified by patient self-identification. Categories were recorded as white, black, Hispanic, Chinese, Filipina, Japanese, and other. Family history of breast cancer was defined as having at least one first-degree relative with breast cancer (mother, sister, daughter). The presence of a palpable mass was based on personal history or physical examination. As this study was a retrospective review of patient records and there was no direct participant contact with study investigators, informed consent was not received per approval of the institution’s committee on human subjects. The database also included screening mammography interpretation with mammography read as normal or abnormal. Abnormal mammography readings were further classified according to the BI-RADS lexicon categories adapted by the American College of Radiology‘”: (1) additional examinations needed; (2) suspicious for malignancy, biopsy recommended; and (3) malignant by radiologic criteria, biopsy recommended. Separate interpretations were made for each breast. For purposes of our analysis, the more abnormal reading was coded for women who have discordant interpretations between breasts. Results of subsequent diagnostic tests were also coded for each breast. Follow-Up As part of the screening process, systematic follow-up of patients with identified mammographic abnormalities was conducted using a computerized tracking sysOne month after an abnormal screening examination, a computer-generated letter requesting information about diagnostic procedures and clinical outcome was mailed to the woman’s identified physician, unless follow-up information had already been provided to the screening program. If physicians did not respond to the mailed request within 1 month, they were contacted by telephone. Standardized requests for follow-up information provided information on the type, timing, and interpretation of subsequent diagnostic tests as well as clinical outcome. Classification of breast cancer, if diagnosed, was based on the American Joint Committee on Cancer staging system.” Time to first diagnostic test was defined as the number of days from index abnormal screening examination to first diagnostic test. Time to final disposition was defined as the number of days from index abnormal screening examination to final disposition, i.e., cancer diagnosis or last diagnostic test that allowed the physician to decide no further workup was necessary. Data and Statistical Analysis A total of seven women (2%)were missing the date of their last diagnostic test. Three of these women were white, and four were nonwhite. For these patients, the overall sample’s median time to last diagnostic test was used to calculate a final test date. This method of imputation will generally underestimate any racial differences in time to final diagnosis or disposition. Income information (the only available socioeco- Race Difference in Mammography Follow-UpKhang et al. noinic status indicator) was operationalized in three different ways using 1990 U S . Bureau of the Census da1.a. First, patient income was estimated using the median income of residence zip code. This method has been commonly used to estimate income." To test the impact of income further. a second indicator (percentage of zip code area households with incomes below poverty level) was also used." Income was also dichotomized (as less than $20,000 annually vs. $20,000 or more annually). Two women were missing zip code/income matches and were therefore not included in the multivariate analyses described subsequently. Comparisons of categorical demographic characteristics between white and nonwhite women were tested for statistical significance using chi-square analysis. Continuous variables were compared between groups using Student's t test or nonparametric (Wilcoxon rank-sum) methods in cases when measures had nonnormal distributions. Multivariate analyses were performed to determine significant predictors of time to diagnostic testing. The major dependent measures of time to diagnosis and time to first test were skewed to the right and were therefore transformed using a natural log function to meet assumptions of normality. Two women underwent same-day additional tests, and their time to first test was set equal to 0.5 days to allow log transformation. Linear regression analysis was then performed to identify important predictors of log-transformed time to either first test or diagnosis/disposition. All analyses were performed using SAS for a personal computer. The variables included in the model were those tlhat were available and that the authors presumed would affect time to diagnostic follow-up. These variables include: age, raceIethnicity, family history of breast cancer, presence of a palpable mass, income, and screening mammography interpretation. In a recent review, older women have been shown to be more likely to have longer time to diagnostic e~aluation.'~ Having a family history of breast cancer or a history of a palpable mass might also affect the timing of diagnostic evaluation, with the likelihood of more aggressive (i.e., timely) follow-up. A suspicious or malignant interpretation of the index screening mammography was also included in the model, as it could similarly act as an alert to the physician to pursue more timely (diagnosticevaluation. RESULTS Final Study Group .Among 6626 screening examinations performed between July 1993 and May 1994,337 were interpreted as 1397 abnormal at the end of the study period. Twelve records were not entered in our analytical database; 10 records were missing, and 2 were pending final disposition. Eight additional women were not included in the analysis. Four women had no documented follow-up as of 6 months after index abnormal mammography, two were white women and two were Latinas (one of whom reportedly refused additional tests). One woman had missing raceIethnicity information, and three women were excluded because they had a previous history of breast cancer. Because women with a previous history of breast cancer are eligible for mammography screening, a repeat analysis was conducted including these women; their inclusion did not alter our study findings. The final study sample, therefore, consisted of 317 women with an abnormal result of a screening mammography examination during the study period with documented follow-up. In all, 95% of those eligible were included in the analysis. Baseline Characteristics The mean age of the women studied was 52 years (range, 33 to 85), with 48% of the sample aged 50 or older. The majority of the sample was white (64%), with 16% Latinas, 12% Asians, and 8%African Americans, which closely reflects the San Francisco Bay Area population statistics of 64% whites, 11% Latinas, 17% Asians, and 8%African Americans.25Fifty-four percent were first-time mammography screeners, and 8%were noted to have a palpable mass by history or physical examination at the time of screening. Compared with white women, nonwhite women (Table 1) were less likely to have a family history of breast cancer, had a lower median income, and were more likely to be undergoing their first screening mammography. Abnormal screening mammography interpretation did not differ significantly by race (Table 2). Median time to final disposition differed significantly ( P < 0.001) by race, with medians of 12 days (range, 1 to 192) for white women and 19 days (range, 2 to 176) for nonwhite women. Median time to first diagnostic test also differed significantly ( P < 0.001) for white versus nonwhite women at 7 days (range, 0 to 161) and 15 days (range, 2 to 1261, respectively. The overall difference in time to final disposition appears to be explained by the lag time to first diagnostic test. That is, the median times from first diagnostic test to final disposition did not differ significantly ( P = 0.06) between white and nonwhite women. Among women evaluated for abnormal mammography, 30 were diagnosed with breast cancer. Ten cases were ductal carcinoma in situ, 2 were Stage I1 cancers, and the remaining 18 were Stage I cancers, Four nonwhite women were diagnosed with cancer, 1398 CANCER October I , 1996 I Volume 78 I Number 7 TABLE 1 Characteristicsof Women with Abnormal Screening Mammography, University of California San Francisco Mobile MammographyProgram, July 1993-May 1994 Mean age (yr) Family history of breast cancer Report of palpable breast mass First mammography screen (baseline) Median income White (n = 203) Nonwhite (n = 114) Test statistic 52.5 32 (16%) 15 (7%) 87 (43%) $30,656 50.5 5 (4%) 9 (8%) 85 (75%) $27,708 x2 = 9.2 xz = 0.03 xz = 29.6 t= t= ~ ~ 1.51 4.72 P value NS <0.01 NS <0.001 <0.001 ~~~~ NS: not statistically significant. TABLE 2 Outcomes of Women with Abnormal Screening Mammography,University of California San Francisco Mobile Mammography Program, July 1993-May 1994 White (n = 203) Nonwhite (n = 114) Test statistic P value ~~ Suspicious or malignant screening mammography interpretation Mean number of diagnostic tests Median time from abnormal mammography to final disposition (days) Median time from abnormal mammography to first diagnostic test (days) Number of carcinomas 10 (5%) 1.4 3 (3%) 1.3 x2 = 0.97 t = 25G.5 NS NS 12 19 Z = 2.98 <0.01 7 15 4 (4%) Z = 5.63 x? = 7.37 <0.001 10.01 26 (13%) ~~~~~ NS: nnt statistically significant. one Asian woman at Stage 11, one Latina at Stage I, and two Latinas with ductal carcinoma in situ. Overall, white women in our study group were more likely to be diagnosed with cancer ( P < 0.01). Diagnostic Tests The vast majority of women (85%) underwent additional mammography views as the first follow-up test. A total of 265 women (84%)underwent spot-compression magnification views, and 24 women (8%) began their evaluation with ultrasound. Twenty-four women (8%) began their diagnostic evaluation with an invasive procedure, 9 with fine-needle aspiration and 15 with biopsy (excisional or core biopsy). Of these 15 women who went directly to biopsy, 6 were diagnosed with breast cancer. Among the 13 women who had screening mammography interpreted as “suspicious” or “malignant,” 4 went directly to fine-needle aspiration or biopsy. Eight were diagnosed with cancer. Three women completed their diagnostic evaluation within 1 week, five within 1 month, and five within 2 months (range, 51 to 65 days). Three of these women were nonwhite; their diagnostic evaluations were completed between 20 and 51 days after the date of index abnormal screening mammography. Evaluation of Mammographically Detected Abnormalities It appeared that a wide range of evaluative tests were performed on these women with abnormal mammography. To more closely examine this variation, we focused on women who had screening mammography interpreted as “additional evaluation needed” and had no palpable lesion. In other words, we looked at the range of evaluations among 283 women who presumably had abnormalities detected only by screening mammography. Table 3 illustrates both the range of final diagnostic tests conducted on these women and the various routes used to reach a final diagnosis or disposition. The majority of these women (61%) underwent spot-compression magnification mammography (magnification views) alone for resolution of their abnormality. Six women underwent fineneedle aspiration alone, and 10 women went directly to biopsy. In all, 67 of these 283 women (24%) with abnormal mammography interpreted as “additional Race Difference in Mammography Follow-UplChang et al. TABLE 3 Description of Diagnostic Tests for Women with Mammographically Detected Lesions Interpreted as “Additional Evaluation Needed” Diagnostic tests Repeat screening mammography Spot-compression magnification mammography only U1trasonography Only With spot-compression magnification mammography Fineneedle aspiration Only With spot-compression magnification mammography and/or ultrasonography Biopsy 0nly With spot-compression magnification mammography With spot-compression magnification mammography and ultrasonography With fine-needle aspiration with or without ultrasonoRraphv White (n = 180) Nonwhite (n = 103) 1 (0.6%) 1 (1.0%) 104 (57.8%) 18 (10,0%) 10 69 (67.0%) 14 (13.6%) 9 8 5 4 (3.9%) 5 (2.9%) 3 3 2 52 (28.9%) a 1 15 (14.6%) 2 36 9 7 3 1 1 ~ P value F value ~ ~~ The same statistical results were found when substituting two other income variables, percentage of zip code area households with incomes below poverty level and income dichotomized at $20,000. Analysis specifymg nonwhite racial/ethnic groups found that each group (African American, Latina, and Asian) had significantly longer log time to first diagnostic test compared with white women ( P < 0.01 for each group) but not compared with one another ( P > 0.05). The same statistical results were found when performing multivariate analysis of predictors of time to final diagnosis/disposition (log transformed). Because prevalence of first screening mammography differed by race, we tested a model (not shown) that included whether the index mammography was a baseline examination; this did not alter the multivariate model results. Similarly, the age variable was categorized to compare women younger than 65 to those age 65 and older (based on previous reports of later cancer diagnostic stage for older womenlo).Using this cut point and a continuous variable, patient age was not a significant predictor of log time to first test in the multivariate model. DISCUSSION TAIILE 4 Multivariate Analysis of Predictors of Time from Abnormal Mammographyto First Diagnostic Testa (n = 315Ib Independent variable 1399 ~ Age 50 yr or older <0.01 0.96 Nonwhite Palpable mass Family history of breast cancer Suspicious or malignant interpretation ot screening mammography Median income 25.89 0.30 0.07 0.59 0.79 0.70 0.33 <0.001 0.41 0.57 Log transformation of time (to meet assumptions of normality:. Nor inclriding two women for whom there was missing incomc! data (no zip code match). Mod&adiusted R-square = 0.087, F value = 4.88. evaluation needed” and no reported palpable mass underwent biopsy. White women were more likely to undergo biopsy than nonwhite women ( P < 0.02). Among this group of 283 women, 22 (8%)were diagnosed with cancer; all but 1 were white. Multivariate Model Time to first diagnostic test (log transformed) differed significantly for white compared with nonwhite women ( P < 0.001) controlling for patient age, presence of a palpable breast mass, family history of breast cancer, screening mammography interpretation, and median income of residence zip code area (Table 4). Definitions of timeliness of follow-up have varied, with published reports measuring either time from index abnormal screening mammography to first subsequent evaluative test or to final test/disposition. This report focuses on absolute time of follow-up as opposed to rates of follow-up based on a defined cut point for timeliness (i.e., within 8 weeks) because the absolute timing is more specific and clinically relevant. We found that there may be significant differences between white and nonwhite women in the process of evaluating abnormal screening mammography. Although rates of mammography screening have improved among women of ail raciallethnic group^,'^ there may still be barriers to timely performance of subsequent evaluation of abnormal mammography that differ by raceIethnicity. Although there is little published literature on the timing of follow-up in the evaluation of abnormal mammography, a recent review of unpublished data has shown some significant bivariate predictors of longer follow-up times, including age (women older than 65) and lower socioeconomic tatu us.'^ Our data, using multivariate analysis, did not find age to be a significant independent predictor of timeliness of follow-up. Data from the Henry Ford Health System found racial differences in timely follow-up that could be accounted for by adjusting for income, with women of lower socioeconomic status (median income less 1400 CANCER October 1,1996 / Volume 78 / Number 7 than $20,000) being less likely to have timely followup compared with women with higher incomes.2"Our results-using a continuous outcome variable (log time to first diagnostic test)-detected such a racial/ ethnic difference when controlling for income, most likely due to greater statistical power. Using less powerful logistic regression techniques to examine followup within 30 days (a dichotomized outcome), we found that race lost its statistical significance in multivariate analysis. The time involved in completing follow-up tests will be affected by a variety of conditions but may reflect delays related to the health-care system, patient, or p r ~ v i d e r . ~First, ~ , ~patient ~ , ~ ~-induced delay may be related to individual fears or anxieties. A study of women who have had abnormal screening mammography found significant amounts of anxiety associated with the test, with higher rates among those with more suspicious reading^.^' RaciaUethnic differences in rates of fear and anxiety have also been reported in the context of routine mammography ~ c r e e n i n g . ~Similar ' ~ ~ ~ effects in the context of abnormal mammography screening may contribute to delay in seeking or scheduling additional diagnostic tests. Once women experience breast symptoms, there have been reports of significant racial/ethnic differences in the timing of seeking care. Vernon et al.32 found that Latina and African American women had significantly longer delays from the time of noticing symptoms to time of first physician consultation. A more recent study examining patient delay in seeking breast care after onset of symptoms focused on white versus African American women and failed to reveal any significant d i f f e r e n ~ e . ~ ~ Second, provider-related delay may contribute to our findings. Timely communication may be more difficult for nonwhite women if they predominantly receive their care through overburdened community and public health clinics. Although women were required to provide the name and address of a primarycare provider before undergoing screening mammography, these names and addresses were not validated at that time. The UCSF Mobile Mammography Program, however, has a tried-and-tested computer system to provide written notification of abnormal test results to both referring providers and women who have undergone screening. Referring providers also receive telephone notification of all screening-detected abnormalities on the day after examination. The effectiveness of communication between providers and patients may also differ, especially for those patients whose primary language is not English but who are likely to have English-speaking providers. It is noteworthy that all women and their primary physi- cians in our study received identical information (in English only), emphasizing the need for prompt follow-up after abnormal mammography screening. It is also important to note that we found no significant racial/ethnic differences in the time from first to last diagnostic test. This suggests that follow-up care proceeded in a timely manner for all women once diagnostic testing was begun. Third, nonwhite women may have poorer access to care (e.g., more likely to have public or no insurance, no regular care site or provider, and/or difficulty maneuvering within a health-care system). All the women included in our study had access to screening services, but we had no information on their access to subsequent diagnostic services or the continuity of their care. Access difficulties may similarly be reflected by the fact that nonwhite women were more likely to be undergoing a baseline (or first-time) screening mammography than the white women in our sample. We have also described a wide range of practice variation in the evaluation of abnormal screening mammography cases. Whereas women were screened as part of a single program, subsequent diagnostic evaluation occurred in many facilities throughout the San Francisco Bay Area. This variation therefore reflects clinical practice in both private and public systems of health-care delivery. The racial difference described, with white women being more likely to undergo biopsy, may be more appropriate given their higher observed cancer incidence rate in our series. The differing patterns of follow-up care by race maj. also be related to women's socioeconomic status or to differences in practice style by site or provider. A lower rate of invasive testing (i.e., biopsy) among nonwhites in our series is consistent with reports of racial differences in the evaluation of other disorders, namely cardiovascular disease34and cerebrovascular disease.35 These studies have primarily focused on white versus African American populations and have raised questions concerning potential differences in patient preferences, insurance coverage, and discriminatory practices among providers. This report is based on a limited series of patients accrued over a 10-month period. Although follow-up information was collected on almost every woman with abnormal mammography during the study period, many reports of follow-up testing were supplied by primary-care providers. With the emphasis on the patient's final diagnosis or disposition, they niay have underreported the total number of diagnostic tests. 'Therefore, our report may underestimate the relative differences in the timeliness of follow-up based on race/ ethnicity. The statistical findings of our multivariate results, however, did not change when we excluded Race Difference in Mammography Follow-UplChang et al. from the analysis those women for whom we estimated (rather than ascertained) the dates of final diagnosis or disposition. The database does not provide reliable, individual-level insurance information, so income (estimated by zip code area) was used as a proxy measure for slocioeconomic status. Although similar methodologies have been used in population-based studies,’”the range of incomes within the San Francisco Bay Area may not be reflected adequately by using median income values. Other studies have used percentage of the population within the zip code area with incomes below poverty level as a predictor of health service ~itilization.~~ We used this variable as well, with no change in the results of our multivariate analysis. The San Francisco Bay Area is not reflective of the IJnited States population as a whole because of its raciallethnic diversity and its generally higher rates of mammography screening. Programs for both mammography screening and subsequent diagnostic testing, when necessary, are also available to women with low incomes and with no health insurance. The setting of our report, therefore, may reflect a “better case scenario” in terms of service availability and timeliness of follow-up. The clinical implications of longer follow-up times after abnormal mammography screening are not known. Screening mammography detects a higher proportion of early stage breast and the median doubling time for mammographically detected tumors is long (median, 260 days).36 Consequently, it is likely that the relatively small, albeit significantly longer, time to completion of diagnostic workup for nonwhite versus white women in our study (median times, 19 days versus 12 days, respectively) has had no impact on breast cancer stage or mortality. It must be emphasized, however, that this difference in workup time occurred in the setting of a well-organized system of follow-up and reporting of results. The finding of raciaUethnic differences in our setting is worrisome. It is possible that larger, potentially more important raciallethnic differences may be observed in clinical settings where reporting and follow-up procedures are not so complete. REFERENCES Devesa S, Blot W, Stone B, Miller B, Tarone R, Fraumeni J. Recent cancer trends in the United States. J Nut1 Cancer Inst 1995;87(3)1175-82. Bain R, Greenberg R, Whitaker 1. Racial differences in survival of women with breast cancer. J Chronic Dis 1986;39(8):631-42. Chen F, Trapido E, Davis K. Differences in stage at presentation of breast and gynecologic cancers among whites, blacks and hispanics. Cancer 1994;73(11):2838-42. 1401 Howard J, Hankey B, Greenberg R, Austin DF, Correa P, Chen VS, et al. A collaborative study of differences in the sunrival rates of black patients and white patients with cancer. Cancer 1992;69(9):2349-60. 5. Smigel K. Breast cancer death rates decline for white women. J Nut1 Cancer Inst 1995;87(3):173. 6. Northern California Cancer Center. Greater Bay Area Cancer Registry Report 1994;1:6. 7. Swanson G, Ragheb N, Lin C, Hankey BF, Miller B, HornRoss P, et al. Breast cancer among black and white women in the 1980s. Changing patterns in the United States by race, age, and extent of disease. Cancer 1993;72(3):788-98. 8. Gordon N, Crowe J, Brumberg 1, Berger N. Socioeconomic factors and race in breast cancer recurrence and survival. Am J Epidemiol 1992;135(6):609-18. 9. Ansell D, Whitman S, Lipton R, Cooper R. Race, income, and survival from breast cancer at two public hospitals. Cancer 1993;72(10):2974-8. 10. Mandelblatt J, Andrews H, Kerner 1, Zauber A, Burnett W. Determinants of late stage diagnosis of breast and cervical cancer: the impact of age, race, social class, and hospital type. Am J Public Health 1991;81(5):646-9. 11. Eley J, Hill H, Chen V, Austin DF, Wesley MN, Muss HB, et al. Racial differences in survival from breast cancer: results of the National Cancer Institute BlackiWhite cancer survival study. JAM4 1994;272(12):947-54. 12. Elledge R, Clark G, Chamness G, Osborne C. Tumor biologic factors and breast cancer prognosis among white, Hispanic, and black women in the United States. J Natl Cancer Inst 1994;86(9):705- 12. 13. Fox S, Roetzheim R. Screening mammography and older hispanic women. Cancer 1994;74:2028-33. 14. Breen N, Kessler L. Changes in the use of screening mammography: evidence from the 1987 and 1990 National Health Interview Surveys. Am J Public Health 1994;84(1):62-7. 15. Coleman E, Feuer E, NCI Breast Cancer Screening Consortium. Breast cancer screening among women from 65 to 74 years of age in 1987-1988 and 1991. Ann Intern Med 1992; 117(11):961-6. 16. Rakowski W, Fulton J, Feldman J. Women’s decision-making about mammography: a replication of the relationship between stages of adoption and decisional balance. Health PSychol 1993;12(3)1209- 14. 17. Sickles E. The use of computer in mammography screening. Radio1 Clin North Am 1987;25(5):1015-30. 18. Kerlikowske K, Grady D, Barclay J, Sickles E, Eaton A, Ernster V. Positive predictive value of screening mammography by age and family history of breast cancer. JAMA 1993; 270(20):2444- 50. 19. Sickles E, Ominsky S, Sollitto R, Galvin H, Moticciolo D. Medical audit of a rapid-throughput mammography screening practice: methodology and results of 27,114 examinations. Radiology 1990;175(2):323-7. 20. D’Orsi C, Kopans D. Mammographic feature analysis. Semin Roentgenol 1993;28(3):204-30. 21. Monticciolo D, Sickles E. Computerized follow-up of abnormalities detected at mammography screening. Am J Roentgenol 1990;155(4):751-3. 22. American Joint Committee on Cancer. Manual for staging of cancer. Philadelphia: JB Lippincott, 1988. 23. Bindman A, Grumbach K, Osmond D, Komaromy M, Vranizan K, Lurie N, et al. Preventable hospitalizations and access to health care. JAMA 1995;274:305-11. 4. 1402 CANCER October 1, 1996 I Volume 78 I Number 7 24. Kerlikowske K. Timeliness of follow-up after abnormal screening mammography. Breast Cancer Res Treat 1996;40(1):53-64. 25. Gey F, Oliver J, Highton B, Tu D, Wolfinger R. California LatinalLatino demographic data book. Berkeley: California Policy Seminar, University of California, 1993. 26. McCarthy B, Ulcickas Yood M, Bookhaker E, Ward R, Beboer M, Johnson C. Inadequate follow-up of abnormal manimograms. Am I Preu Med. In press. 27. Facione N. Delay versus help seeking for breast cancer symptoms: a critical review of the literature on patient and provider delay. SOCSci Med 1993;36(12):1521-34. 28. Katz S, Hislop G, Thomas D, Larson E. Delay from symptom to diagnosis and treatment of breast cancer in Washington state and British Columbia. Med Cure 1993;31(3):264-8. 29. Lerman C, Trock B, Rimer B, Boyce A, Jepson C, Engstrom P. Psychological and behavioral implications of abnormal mammograms. Ann Intern Med 1991; 114(8):657-61. 30. AMC Cancer Research Center. Breast and cervical screening: barriers and use among specific populations. Atlanta, GA U.S. Centers for Disease Control, 1992 Dec. report no: 1J50i CCU806186-02. 31. C o p e C, Hohman K, Levinson A. Reaching special populations with breast and cervical cancer public education. J Cancer Educ 1992;7(4):293-303. 32. Vernon S, Tilley B, Neale A, Steinfeldt L. Ethnicity, survival, and delay in seeking treatment for symptoms of breast cancer. Cancer 1985;55(7):1563-71. 33. Coates R, Bransfield D, Wesley M, Hankey B, Eley J, Greenberg R, et al. Differences between black and white women with breast cancer in time from symptom recognition to medical consultation. BlacklWhite Cancer Survival Study Group. J Natl Cancer Inst 1992;84(12):936-50. 34. Ford E, Cooper R. RaciaUethnic differences in health care utilization of cardiovascular procedures: a review of the evidence. Health Serv Res 1995;30(1):237-52. 35. Horner R, Oddone E, Matchar D. Theories explaining racial differences in the utilization of diagnostic and therapeutic procedures for cerebrovascular disease. Milbank Q 1995; 73(3):443-62. 36. Spratt J, van Fournier D, Spratt J, Weber E. Mammographic assessment of human breast cancer growth and duration. Cancer 1993;71(6):2020-6.