Preventive Medicine 43 (2006) 178 – 182 www.elsevier.com/locate/ypmed
Mammography screening rates decline: A person-time approach to evaluation Adrianne C. Feldstein a,b,⁎, Thomas M. Vogt c , Mikel Aickin d , Weiming R. Hu a a
Center for Health Research, Kaiser Permanente, Portland, OR 97227-1110, USA b Northwest Permanente, Portland, OR 97232-2099, USA c Center for Health Research, Kaiser Permanente, Honolulu, HI 96817, USA Program in Integrative Medicine, The University of Arizona, Tucson, AZ 85721, USA Available online 3 May 2006
Abstract Objectives. Early detection through mammography can reduce breast cancer mortality. This cohort study evaluated trends in mammography screening, demonstrating a person-time approach. Methods. Included were women HMO members aged 50–69 from 1999 to 2002 who had not had breast cancer, dysplasia, fibrocystic disease, or implant. The amount of person-time covered by mammography as a percent of the time eligible for mammography screening (the prevention index (PI)) was calculated using electronic data. The denominator was the time during which the guidelines recommended that each participant should have been covered by a mammogram (every 24 months), excluding times when breast mass, abnormal mammogram, galactorrhea, or other breast disorders were under evaluation. The numerator was the time during which she was covered by a mammogram. Results. The number of women who contributed person-time increased from 43,283 to 49,512 and the number of screening mammograms declined from 23,586 to 22,719. The overall PI for screening mammography declined from 67.0 (67% of eligible person-time was appropriately covered by a mammogram) to 62.5, and the proportion of women with no coverage during a given year increased 16%. Conclusions. This study shows a declining pattern of mammography screening using a person-time approach, a decline greater than that shown by methods that include diagnostic mammograms. The study highlights opportunities for use of the PI and quality improvement initiatives to improve breast cancer outcomes. © 2006 Elsevier Inc. All rights reserved. Keywords: Mammography; Prevention index; Quality improvement
Introduction Breast cancer is the most common cancer among women in the United States (Overmoyer, 1999). About 1 in 8 women will develop breast cancer during their lifetime, and annually, 46,000 women will die from it (George, 2000). The early detection of breast cancer through mammography screening can reduce mortality from breast cancer approximately 15–40% in women aged 40 and older, with greater absolute risk reduction in older women (ACOG, 2003; Humphrey et al., 2002; Paquette et al., 2000).
⁎ Corresponding author. Center for Health Research, 3800 N. Interstate Ave., Portland, OR 97227-1110, USA. Fax: +1 503 335 6311. E-mail address: [email protected]
(A.C. Feldstein). 0091-7435/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.ypmed.2006.03.009
Multiple evidence-based clinical guidelines recommend regular screening mammograms (ACS, 2005; ACOG, 2003; Humphrey et al., 2002). The United States Preventive Services Task Force (USPSTF) recommends screening women every 1– 2 years at age 40 and older. The strength of the evidence is highest in women aged 50–69; women aged 40–49 were added to the USPSTF recommendation in 2002 when the strength of the evidence to include them was upgraded from “insufficient” to “probably effective” (USPSTF, 2005). Data based upon Medicare claims reveal that the prevalence of screening mammography in women over the age of 65 increased substantially between 1993 and 1998 (Randolph et al., 2002), and that the proportion of physicians reporting referrals for mammography increased from 37% in 1988 to 64% in 1995 (Lane and Messina, 1999). Reports in the lay literature have suggested that screening rates have been declining in
A.C. Feldstein et al. / Preventive Medicine 43 (2006) 178–182
multiple states among Medicare beneficiaries (Tampa Bay Business Journal, 2004). For example, Florida reported that 63% of female Medicare beneficiaries between the ages of 50 and 67 received a mammogram in 2003 as compared to 65% in 2001 and noted that 33 other states were experiencing declines (TBBJ, 2004). Recent trends have not been reported or discussed in the scientific literature, however. This retrospective cohort study evaluated current trends in mammography screening. It demonstrated the use of data from electronic medical records and a person-time approach to coverage for breast cancer screening. This approach, called the Prevention index (PI), has been described elsewhere (Vogt et al., 2004) as have its potential benefits over the current National Center for Quality Assurance (NCQA) Health Employer Data and Information Set (HEDIS®) cross-sectional methods (NCQA, 2005; Vogt et al., 2004). Methods The protocol for this study was approved by the institutional review board of the health maintenance organization (HMO) in which it was conducted.
Setting and databases The study was conducted in a non-profit, group-model HMO in the Pacific Northwest with about 454,000 members, 35% of whom are over age 50. The HMO covers all mammograms ordered by clinicians and allows patients to selfrefer for screening up to annually. Mammogram frequency was not driven by clinician reimbursement; all physicians were salaried. The HMO encompasses 20 medical offices in 2 states, includes 797 physicians and 395 allied health care providers, and maintains electronic inpatient and outpatient medical and, membership records.
Study population Participants eligible for the study were women 50–69 years old who were HMO members for at least 1 month between 1/1/1999 and 12/31/2002. Women aged 40–49 were not included because they were not included in USPSTF screening recommendations until 2002.
Analysis variables and dataset construction For each study-eligible woman, several data points were extracted from the electronic records. The first was the beginning and ending dates for HMO membership intervals that overlapped the target measurement period (1/1/1999– 12/31/2002). The second was dates of completed mammograms (with separate notations for diagnostic mammograms and screening mammograms) between 1/ 1/97 and 12/31/2002. The third was dates of temporary diagnostic exclusions
that occurred during the target measurement period and permanent diagnostic exclusions that occurred during or prior to the target measurement period. Mammograms done for a period of time after the exclusion dates were assumed to be diagnostic rather than screening mammograms. Temporary exclusions included breast mass, abnormal mammogram, galactorrhea, other breast disorders, and all diagnostic breast procedures, such as ultrasound, MRI, CT, biopsy, aspiration, and thermography. Permanent exclusions included diagnosis of breast cancer, dysplasia, or fibrocystic disease of the breast; mastectomy; breast reconstruction; and breast implant. Data extracted from the electronic medical record were used to create an analysis dataset with four variables: patient identification number, type of event, event details, and date of event. The event types included start and end of HMO membership periods, temporary exclusions, permanent exclusions, and screening and diagnostic mammograms.
Data management and analysis Data were extracted and analyzed in 2004. A programming approach called “event-stream analysis” was applied to determine periods during which a woman was eligible for a screening mammogram and periods following a screening mammogram during which she was considered to have been “covered.” Following the recommendations of the USPSTF (Humphrey et al., 2002), appropriate screening intervals were defined to be 2 years. This allowed us to define for the cohort risk period/s for each woman where coverage by a screening mammogram was applicable. These periods took the form of [sum(c1/ e1 + c2/e2…cn/en)] / n (c = covered person-time; e = eligible person-time). For each year in the target period, all time for every woman was classified as (1) eligible for a mammogram and appropriately covered; (2) eligible for the service but not covered; or (3) excluded (not counted in the PI calculations in either the numerator or the denominator). For each participant who was eligible at any time during the study period, a mammography prevention index was calculated as person-time covered divided by person-time eligible for coverage multiplied by 100 (Vogt et al., 2004). For example, if a 51-year-old woman had a screening mammogram in January of 1999 and was due for a repeat exam in January of 2001 but instead had it in June of 2001, her PI for 1999 and 2000 is 100%. Her PI for 2001 is 50% (only 6 months are “covered”). This is graphically displayed in Fig. 1. Person-time during which a woman was not eligible on the basis of age or HMO membership was removed from both the numerator and denominator. Since the goal was to create a mammography screening PI, person-time during which mammography services were delivered for diagnostic reasons also was excluded from both the numerator and denominator. A mammogram designated as screening was categorized for study purposes as a diagnostic mammogram if a temporary exclusion occurred during the prior 2 years. Person-time covered by mammography during a period of temporary exclusion (for any diagnosis) was excluded from the date of the diagnosis for 2 years or the next mammogram, whichever came first. Person-time for women with permanent exclusions was excluded from the date of diagnosis forward. An early screening (within 2 years of a prior screening) extends the coverage interval and creates a period of “double coverage.” Because the PI is the percent of the target interval covered by any service, periods of overlapping coverage do not affect the PI score.
Fig. 1. PI “Covered Time” concept.
A.C. Feldstein et al. / Preventive Medicine 43 (2006) 178–182
To compare the PI approach to the HEDIS mammography measure, we calculated the proportion of women aged 52–69 by December 31st of each measurement year who had a mammogram during the measurement year or the year before. This analysis was restricted to individuals who had continuous membership during the measurement year and the year before.
N 43,283 45,304 47,547 49,512 PI total (SD) 67.0 (43.5) 65.3 (43.9) 64.1 (44.4) 62.5 (45.5) Covered/ 13,050/19,477 14,200/21,746 16,153/25,200 18,567/29,707 Eligible a
Results Table 1 presents the characteristics of the study population. The number of women contributing person-time to the PI went up each year, from a low of 43,283 in 1999 to a high of 49,512 in 2002. The number of women contributing person-time was consistently higher in the 50–59 year old age group than the 60–69 year old group, with the 50–59 group contributing nearly two-thirds of the person-time. The proportion of eligible women Table 1 Study population characteristics: age, HMO membership*, screening and diagnostic mammograms, and exclusions, by year Measure
Age Total N a 50–59 N (%) 60–69 N (%)
43,283 45,304 47,547 49,512 28,182 (64.2) 29,798 (64.8) 31,442 (65.2) 32,559 (64.7) 15,706 (35.8) 16,167 (35.2) 16,809 (34.8) 17,736 (35.3)
HMO membership <5 years N (%) 13,980 (32.3) ≥5 years N (%) 29,303 (67.7) Mammograms 34,856 total Total screening 23,586 mammograms Total diagnostic 11,270 Coded 8777 diagnostic Screening to 2493 diagnostic b Diagnoses Total diagnostic exclusions Women with temporary exclusions c N (%) Women with permanent exclusions d N (%)
Table 2 Mammography prevention index results by year: total PI, by age, and by eligibility period at HMO in Pacific Northwest
15,449 (34.1) 16,927 (35.6) 18,072 (36.5) 29,885 (65.9) 30,620 (64.4) 31,440 (63.5) 35,681 34,428 34,719 23,952
N = Number of persons in category. HMO* = health maintenance organization (Group-model HMO in the Pacific Northwest). a Note that summing the N values for the age strata does not equal the total N because women may contribute person-time for both strata during a measurement year. b A screening mammogram within 2 years of a temporary exclusion was converted to a diagnostic mammogram. c Temporary exclusions—breast mass, abnormal mammogram, galactorrhea, other breast disorders, and all diagnostic breast procedures (e.g., ultrasound, MRI, CT, biopsy, aspiration, thermography). d Permanent exclusions—current of historical breast cancer or dysplasia, fibrocystic disease of the breast, mastectomy, breast reconstruction, or breast implant.
Age 50–59 N PI (SD) Covered/ Eligiblea 60–69 N PI (SD) Covered/ Eligiblea
28,182 64.9 (44.3) 8231/12,682
29,798 63.4 (44.7) 9068/14,303
31,442 32,559 62.1 (45.1) 60.6 (45.2) 10,349/16,664 11,838/19,535
15,706 70.9 (41.9) 5011/7068
16,167 69.3 (42.3) 5378/7760
16,809 68.5 (42.8) 6103/8909
17,736 66.8 (43.2) 7109/10,642
41.7 49.2 53.9 56.4 58.1
38.9 47.6 51.9 54.6 56.6
37.9 46.6 51.1 56.6 55.3
PI by length of eligibility b 1 year 44.3 2 years 52.4 3 years 56.1 4 years 58.2 5 years 59.8
PI = Prevention index. N = Number of persons in category. SD = Standard deviation. a Covered person-years/ eligible person years. b Length of eligibility period—length of time individual is eligible to be screened.
who had fewer than 5 years of membership increased from 32.3% in 1999 to 36.5% in 2002. As the number of women contributing person-years increased, the total number of mammograms performed remained flat. The total number of diagnostic mammograms increased from 11,270 to 12,000 during the study period, while the number of screening mammograms declined from 23,586 to 22,719. Time was excluded due to temporary or permanent exclusions in a small proportion of the women included in the analyses each year (<8% in 1999 and <6% in 2002). Table 2 presents the mammography PI results by year for the total group, by age and by length of eligibility period strata. From 1999 to 2002, the overall PI for mammography declined from 67.0 (i.e., 67% of eligible person-time was appropriately covered by a mammogram) to 62.5 (P < 0.0001). Women in their 60s were more likely to be covered than women in their 50s (66.8% vs. 60.6% of person-time covered in 2002). The PI increased progressively as initial screening eligibility increased from 1 to 5 years (37.9 in first year of screening eligibility vs. 55.3 in fifth year for 2002). Importantly, the decline in the PI over the 4-year period was greater for those with fewer years of screening eligibility. Table 3 presents the proportion of women in various mammography prevention index strata by year. From 1999 to 2002, there was a 16% increase in the proportion of women who had a PI of zero – i.e., those with no coverage at all during a given year (25% in 1999; 29% in 2002) – and a 9% decrease in
A.C. Feldstein et al. / Preventive Medicine 43 (2006) 178–182
Table 3 Proportion of women in mammography prevention index strata by year at HMO in Pacific Northwest
Table 5 Comparison of PI to HEDIS scores for mammography at HMO in Pacific Northwest
PI = 0 0 < PI < 100 PI = 100
25 21 54
26 21 53
28 21 51
29 22 49
Prevention index HEDIS
PI = Prevention index.
those who had a PI of 100% – i.e., those with all eligible persontime appropriately covered (55% in 1999; 50% in 2002). Table 4 presents the percent never served (PI = 0) by decade of age and year. Women in their 50s were about 24% less likely to be served at all than women in their 60s (31% vs. 25% for 2002). The increase in the proportion of women never served increased similar amounts for both age strata. Table 5 compares the PI scores for 1999–2002 to HEDIS scores for mammography. Although both scores decline, the PI decrease (4.5 units/67.0 = 6.7%), which represents the proportion of covered person-time, is 68% more than the percent decline in the HEDIS score (3.2 units/79.5 = 4.0%). Discussion The prevention index, a person-time approach to assessing mammography screening trends, found a decline in this insured population in covered time from 67.0% in 1999 to 62.5% in 2002 and a 16% increase in the proportion of women who had no coverage during a given year. Most regions of the study HMO are experiencing declining mammography screening rates (Domby, 2004). Overall, commercial mammography screening rates for health plans participating in the NCQA have remained flat in recent years. Many health plans are reporting declines, however, and mammography rates for those insured through Medicare are declining overall (NCQA, 2004). The findings from the study HMO are particularly concerning given that this institution initiated reminder letters to patients (Vogt et al., 2003) and electronic reminders to clinicians within the HMO's electronic medical record targeting those who had not had a mammogram for over 24 months. Study site managers felt that mammography radiology capacity and co-pays for patients had remained stable, and the use of other breast cancer screening exams such as magnetic resonance imaging remained negligible during the study period. Most importantly, the decline in the screening rate was steeper when only screening mammograms were considered. Generally, assessments of mammography screening (e.g., HEDIS measures) do not differentiate between diagnostic and screening examinations. Comprehensive electronic medical Table 4 Mammography prevention index: percent never served by decade of age and year at HMO in Pacific Northwest Age group
50–59 never served (PI = 0) 60–69 never served (PI = 0)
PI = Prevention index.
PI = Prevention index. HEDIS® = Health Plan Employer Data and Information Set
records, with coded fields completed through clinician order entry, efficiently allow for this important distinction. This is important because a diagnostic mammogram should not be considered a preventive examination. For example, a mammogram done to evaluate a clinically detected breast mass and confirm an advanced breast cancer is not a screening success. Our findings using the prevention index thus raise more concern about current gaps in mammography screening than more traditional measures would have. Nearly all of the uncovered person-time for mammography is attributable to women who rarely or never receive screening or those who recently became eligible by virtue of age or joining the health plan. The remainder is attributable to women who are “late” for screening, i.e., attend less frequently than every 2 years. These findings point out areas where barriers to screening and potential interventions to improve covered person-time should be further explored. It will be particularly important to explore patient and health care system barriers to mammography for the newly insured, new health plan members, and “late” attenders of mammography, as their barriers may be different from those previously described for the rarely screened (Valanis et al., 2003). Also, although many barriers to mammography have been described (Valanis et al., 2003) system barriers such as uncertainty about how to best access services, excessive wait times, difficulty getting appointments, inconvenient screening locations, and cost of care need further evaluation among the insured (Smith et al., 2002). In the last 5–10 years, controversies about breast cancer screening through mammography have been highlighted in the medical literature and in the lay press (Elliott, 2003). For example, randomized trials of mammography have produced conflicting results regarding the extent to which screening reduces morbidity and mortality (Overmoyer, 1999). Women have also received conflicting messages regarding the age at which screening should optimally occur Multiple references including Victoria Stagg Elliott, 2003; Overmoyer, 1999 and the extent to which screening benefits exceed the harms (Elmore et al., 1998; Harris, 1997). This conflicting information suggests that research evaluating barriers to breast cancer screening needs to be updated. Although letters of invitation and phone call reminders have been shown to be effective in community settings (Bonfill et al., 2004), given the study HMO experience, critical assessment of the reach of these procedures to increasingly diverse populations in real-world implementation settings is necessary. Strategies addressing contemporary barriers such as the recommended age for initiation of screening, screening intervals, and methods to access care most efficiently should also be evaluated.
A.C. Feldstein et al. / Preventive Medicine 43 (2006) 178–182
This study found that the PI for mammography declined more than the HEDIS measure for mammography during the study period. This suggests that the HEDIS scoring method (proportion of eligible women screened in the last 2 years) is less sensitive to overall changes in screening pattern, and observed changes in HEDIS scores may understate actual changes occurring in the population. The PI method allows complete data to be readily accessed using electronic medical records, permitting aggregation at the patient, provider, clinic, and health care system levels, and also allows for using patient records that may be discontinuous over time. This study has several limitations. The data were from a single health plan. The membership of the HMO is representative of the demographics of its service area in two states (Freeborn, 1994), but it is unclear if the findings are generalizable to other communities. This fairly large decline in screening occurred in a stable population where all women have insurance coverage for screening, highlighting the need to include a broader population base in future studies (where we might expect an even greater decline). The data utilized included clinician-coded diagnoses that could result in under- or over-ascertainment of screening mammograms when compared to the medical record text. The study also depended upon an internal radiology database to determine completed mammograms and therefore could not ascertain mammograms that were done in facilities outside of the HMO. It is likely that few mammograms would be performed outside the HMO on HMO members, however, because such would not be a covered benefit. The 4-year study period did not ascertain longer population trends in screening. We did not associate changes in the mammography PI with changes in breast cancer morbidity and mortality; this should be a topic for future research.
Conclusions A person-time approach reveals a declining pattern of mammography screening. The study highlights opportunities for further evaluation and quality improvement initiatives to improve breast cancer outcomes. Acknowledgments This work was funded through a research grant from the Garfield Memorial Fund. The study was investigator initiated and the funder had no role in the study design or analysis. We would like to thank Debra Burch and Elizabeth Sheeley for their expert secretarial support, Martha Swain for editorial support, and Dr. Radhika Breaden for her review of the manuscript and insight into quality improvement efforts for mammography at the study HMO.
References American Cancer Society, 2005. Detailed Guide: Breast Cancer Can Breast Cancer Be Found Early? http://www.cancer.org/docroot/CRI/content. American College of Obstetricians and Gynecologists, 2003. ACOG practice bulletin. Clinical management guidelines for obstetrician-gynecologists. Number 42, breast cancer screening. Obstet. Gynecol. 101 (4), 821–831. Bonfill, X., Marzo, M., Pladevall, M., Marti, J., Emparanza, J.I., 2004. Strategies for increasing women participation in community breast cancer screening. Cochrane Database Syst. Rev. 3. Domby, M., 2004. HMO HEDIS lead (personal communication, 2-18-04). Elliott, V.S., November 10, 2003. Mammography debate: who should get screened and when. American Medical News, An AMA Publication. Volume 10, number 42 pages 35–37 (http://www.amednews.com). Elmore, J.G., Barton, M.B., Moceri, V.M., Polk, S., Arena, P.J., Fletcher, S.W., 1998. Ten-year risk of false positive screening mammograms and clinical breast examinations. N. Engl. J. Med. 338 (16), 1089–1096. Freeborn, D.K., 1994. Promise and Performance in Managed Care. The Johns Hopkins Univ. Press, Baltimore, MD. George, S.A., 2000. Barriers to breast cancer screening: an integrative review. Health Care Women Int. 21 (1), 53–65. Harris, R., 1997. Variation of benefits and harms of breast cancer screening with age. J. Natl. Cancer Inst. Monographs 22, 139–143. Humphrey, L.L., Helfand, M., Chan, B.K., Woolf, S.H., 2002. Breast cancer screening: a summary of the evidence for the U.S. preventive services task force. Ann. Intern. Med. 137 (part 1) (5), 347–360. Lane, D.S., Messina, C.R., 1999. Current perspectives on physician barriers to breast cancer screening. J. Am. Board Fam. Pract. 12 (1), 8–15. National Committee for Quality Assurance, 2004. The state of health care quality 2004: Industry trends and analysis. NCQA, WA, DC. National Committee for Quality Assurance, 2005. The Health Plan Employer Data and Information Set HEDIS®. http://www.ncqa.org/programs/hedis/ WA, DC, NCQA. Overmoyer, B., 1999. Breast cancer screening. Med. Clin. North Am. 83 (6), 1443 (1vii). Paquette, D., Snider, J., Bouchard, F., Olivotto, I., Bryant, H., Decker, K., Doyle, G., 2000. Performance of screening mammography in organized programs in Canada in 1996. The Database Management Subcommittee to the National Committee for the Canadian Breast Cancer Screening Initiative. CMAJ 163 (9), 1133–1138. Randolph, W.M., Mahnken, J.D., Goodwin, J.S., Freeman, J.L., 2002. Using medicare data to estimate the prevalence of breast cancer screening in older women: comparison of different methods to identify screening mammograms. Health Serv. Res. 37 (6), 1643–1657. Smith, K.S., O'Connor, E.A., Valanis, B.G., Glasgow, R.E., Whitlock, E.P., 2002. Intermittent and nonusers of mammography: comparison of demographics, health beliefs, and health care use. Clin. J. Women's Health 2 (1), 19–28. Tampa Bay Business Journal (6–14) 2004. FMQAI: mammography rates decline in Florida. Tampa Bay Business Journal. American City Business Journals, Inc. US Preventive Services Task Force, 3–1–2005. Screening for Breast Cancer. http://www.ahrq.gov/clinic/uspstf/uspsbrca.htm. Valanis, B., Whitlock, E.P., Mullooly, J., Vogt, T., Smith, S., Chen, C., Glasgow, R.E., 2003. Screening rarely screened women: time-to-service and 24-month outcomes of tailored interventions. Prev. Med. 37 (5), 442–450. Vogt, T.M., Glass, A., Glasgow, R.E., La Chance, P.A., Lichtenstein, E., 2003. The safety net: a cost-effective approach to improving breast and cervical cancer screening. J. Women's Health (Larchmt.) 12 (8), 789–798. Vogt, T.M., Aickin, M., Ahmed, F., Schmidt, M., 2004. The prevention index: using technology to improve quality assessment. Health Serv. Res. 39 (3), 511–530.