Association between breast cancer screening intention and behavior in the context of screening cessation in older women

INTRODUCTION

Breast cancer is the most common life-threatening cancer in women and mammography screening is an important intervention that can reduce breast cancer-related mortality and morbidity.¹ Not surprisingly, numerous studies have aimed at promoting the adoption of breast cancer screening.^2–7 Based on the Theory of Planned Behavior, which posits that intention is the most important and proximal predictor of behavior,⁸ studies have found moderate association between breast cancer screening intention and receipt of screening,^2,3 and a number of interventions have targeted screening intention as the outcome.^4–7 All of these studies focused on increasing screening rates.

More recently, there is growing recognition that among women aged 75 and older, the harms and burdens of routine mammography screening may outweigh the benefits.^9–11 Therefore, the National Cancer Institute and others have called for interventions to reduce overuse and address screening cessation in older women.¹² Screening cessation after decades of participation is likely different than the decision to initiate screening and it is not known how older women’s screening intentions relate to screening behavior in the context of de-implementing screening. Given that it can take years and significant resources to learn if an intervention leads to a reduction in overuse of mammography screening in older women, we aimed to examine among women aged 75 and older if lower intentions to be screened predicts lower screening rates in the context of screening cessation.

METHODS

As part of a large cluster randomized controlled trial of a decision aid (DA) on mammography screening for women aged 75 and older,¹³ we examined older women’s breast cancer screening intentions before a visit with their primary care provider (PCP) and then again after receiving the DA or a two-page pamphlet on home safety (the control intervention) and seeing their PCP. We then followed patients for 18 months after their PCP visit. For the current study, we aimed to examine whether lower intentions to be screened were associated with lower mammography screening rates in women ≥75 years after 18 months. Beth Israel Deaconess Medical Center’s and the University of North Carolina’s IRB approved this study.

Details of the trial study design have been previously described.¹⁴ We recruited English-speaking women 75–89 years scheduled to see their PCP in the next 4–12 weeks from 11 diverse primary care practices in the Boston-area or in Chapel Hill, North Carolina from November 2014 to January 2017. Given the focus on screening cessation, we only included women who had been screened in the past two years. To identify women more likely to be contemplating their next mammogram, we excluded women who were screened in the past 6 months so that, with 18 months of follow up, we were able to capture receipt of screening 2 years from the women’s last mammogram – the upper limit of the recommended screening interval per guidelines. We excluded women with dementia, history of breast cancer, incapacity, <7^th grade education, documentation of having stopped screening, first visit with PCP, and who did not see their PCP during the study. Oral informed consent was obtained from all participants.

At baseline and post-visit, participants were asked their intentions to be screened with mammography in the next year using the validated 15-point choice/predisposition scale (1=will not get mammogram to 15=will get mammogram).¹⁵ At baseline, participants were also asked about their socio-demographics, health and functional information to estimate life expectancy,¹⁶ and their breast cancer risk factors to calculate their 5-year absolute risk of breast cancer using Gail’s Breast Cancer Risk Assessment Tool.^{17, 18} The baseline questionnaire was administered a median of 34 days (interquartile range, 34 days) before the PCP visit. The post-visit questionnaire was administered immediately after the PCP visit. To assess receipt of screening, the research team reviewed electronic medical record primary care notes, radiology records, and screening sheets (where mammograms performed outside the medical system are entered manually). Initially, all charts were dually abstracted but after 280 charts there was consistently 100% agreement between chart abstractors. Therefore, the remainder were singly abstracted with 20% randomly dually abstracted to ensure quality. If it was unclear from the medical record whether a participant had been screened (e.g., the participant had moved, and/or there were no notes in the last six months of follow-up and no documentation of death), then participants (or if necessary a proxy) were contacted by phone to assess screening.

We examined the association between women’s intentions to be screened and receipt of mammography screening using two methods. First, using univariable and multivariable general estimating equations we examined whether any decline in intentions to be screened from pre-visit to post-visit was associated with lower screening rates after 18 months compared to women whose screening intentions did not change or increased. Second, we used univariable and multivariable general estimating equations to examine association between post-visit screening intention as a continuous variable and receipt of screening after 18 months. Logistic regression analysis using generalized estimating equations were used to account for the clustering by PCP in the study design. We adjusted for treatment arm, PCP panel size (<25 versus 25 or more women ≥75 years in their panel), and PCP site in multivariable analyses. In secondary analyses, we also adjusted for patient education and race/ethnicity. All analyses were completed using SAS statistical software, version 9.4 (SAS Institute Inc., Cary, NC).

The funding source had no role in the design, methods, subject recruitment, data collections, analysis and preparation of the paper.

RESULTS

Recruitment to the trial has been defined previously.¹³ In brief, of 1,596 women who received a study informational letter, 263 opted-out of initial telephone contact, 349 could not be reached, 421 declined to participate by telephone, and 563 agreed to participate; 17 of those who agreed, subsequently withdrew leaving 546 participants seen by 137 different PCPs. Among participants, 283 (51.8%) were randomized to receive the DA. Since characteristics were similar in the DA and control groups, we present participant characteristics overall (Table 1). Participant mean age was 79.8 years, 78.4 % were non-Hispanic white, 58.8% completed college, 35.2% had <10 year life expectancy, and 20.1% were from North Carolina.

Table 1.

Participant Characteristics (N=546)

Characteristic	Number (%) / Mean (SD)
Received Decision Aid in trial	283 (51.8)
Age, year	79.8 (3.7)
Baseline screening intentions	13.0 (4.2)
Recruitment site a
BIDMC (2 practices)	144 (26.4)
BIDMC community (4 practices)	101 (18.5)
Atrius community (3 practices)	191 (35.0)
UNC academic (2 practices)	110 (20.1)
Race
Non-Hispanic White	428 (78.4)
Non-Hispanic Black	100 (18.3)
Hispanic	8 (1.5)
Other	10 (1.8)
Education
<High school	27 (5.0)
High school	88 (16.1)
Some college	107 (19.6)
College degree or beyond	321 (58.8)
Missing	3 (0.6)
Income
$35K or less	170 (31.1)
>$35K to $65K	117 (21.4)
>$65K or higher	172 (31.5)
Declined to answer	87 (15.9)
Marital status
Currently married	213 (39.0)
Single/divorced/separated/widowed	329 (60.3)
Missing	4 (0.7)
Living arrangement
Lives alone	292 (53.5)
Lives with others	250 (45.8)
Missing	4 (0.7)
Cognitive impairment, assessed using Short-Blessed Test b
0–8 (no impairment)	535 (98.0)
9–18 (mild to moderate impairment)	6 (1.1)
Missing	5 (0.9)
Estimated life expectancy c
≥ 10 years	353 (64.6)
≤ 10 years	192 (35.2)
Missing	1 (0.2)
Medical literacy, assessed using REALM-7 d
7 medical terms correctly pronounced	481 (88.1)
<7 medical terms correctly pronounced	49 (9.0)
Missing	16 (2.9)
Five-year probability of breast cancer based on Gail risk score e	2.2 (1.1)

^aBIDMC: Beth Israel Deaconess Medical Center, Boston, MA. Atrius community: group practice in Boston, MA. UNC: University of North Carolina.

^bShort-Blessed Test:¹⁹ Scores range from 0–28; scores 9–18 suggest some cognitive impairment and scores >18 suggest severe cognitive impairment.

^cSchonberg mortality index:¹⁶ Scores ≥10 are associated with >50% chance of 10 year mortality. Thus, women who score ≥10 are estimated to have <10 year life expectancy.

^dREALM-7 scores range from 0 to 7 with higher scores indicating greater literacy.²⁰

^eThe Gail Breast Cancer Risk Assessment Tool ^17,18 considers a woman’s age, race/ethnicity, first degree family history of breast cancer, age at menarche, age at first birth, and history of breast biopsies (and specifically atypical hyperplasia) in assessing women’s probability of breast cancer in the next 5 years. The average 75 year old white woman has a 2.2% 5-year risk.

From pre to post-visit with their PCP, 21.7% of women lowered their intentions to be screened; 7.9% increased their intentions to be screened, and 70.4% did not change. Compared to those who had no change or increased their screening intentions, women who had a decrease in screening intention were significantly less likely to receive screening after 18 months (Table 2), adjusted RR of 0.42, 95% CI 0.30–0.59. Secondary analyses that compared women with decreased screening intentions to all others or only to those with no change in screening intentions found identical results (Appendix). Mean intentions to be screened post-visit was 11.6 (+/− 5.4). Each one point decrease in intentions to be screened (on the 15 point scale) was significantly associated with a lower screening rate (adjusted RR 0.90, 95% CI 0.88–0.93). Secondary analyses that also adjusted for patient education and race/ethnicity did not change the results (Appendix).

Table 2.

Association between intentions to be screened and receipt of mammography screening after 18 months follow-up.

	Underwent screening (n=316) a	Did not undergo screening (n=217) a	Unadjusted RR (95% CI)	Adjusted RR (95% CI) b
Change in mammography screening intentions from pre to post visit with PCP c
Lower intention to be screened, No. (%)	31 (9.8)	84 (38.7)	0.38 (0.27–0.55)	0.42 (0.30–0.59)
Higher intention to be screened, No. (%)	24 (7.6)	18 (8.3)	0.85 (0.66–1.09)	0.92 (0.75–1.12)
No change, No. (%)	261 (82.6)	115 (53.0)	-	-
Post-visit screening intentions d, mean (SD)	13.9 (3.3)	8.4 (6.2)	0.89 (0.86–0.92)	0.90 (0.88–0.93)

^a18 months follow-up was at least 2 years since women’s last screening mammograms.

^bGeneral estimating equation logistic regression models adjusted for treatment arm, PCP panel size and site to account for the clustering by PCP in the trial design.

^cIntentions to be screened was measured on a 15-point scale, ranging from 1 (will not have mammogram) to 15 (will have mammogram).¹⁵

^dUnadjusted and adjusted RR are for each point decrease in screening intention.

DISCUSSION

As posited by the Theory of Planned Behavior,⁸ we found that both lower level of screening intention and a decrease in screening intention were associated with lower rates of screening after 18 months. Previous literature examining the relationship between screening intention and behavior only did so in the context of increasing screening adoption;^2–7 our results adds to the literature by showing that in the context of screening cessation, intention is also a strong predictor of behavior. These findings have several important implications. First, they suggest that studies testing interventions to de-implement breast cancer screening in older women may use a reduction in women’s intentions to be screened as a proxy for understanding the likely effect of their intervention on receipt of screening. Second, our findings may also be used to help estimate the effect an intervention will have on actual screening rates by the effect of the intervention on women’s intentions to be screened. This is important since it is expensive and resource intensive to follow large numbers of older women for long time periods of time to see whether or not they are screened for breast cancer. Intentions to be screened may be assessed relatively easily and more feasibly through the administration of questionnaires. Third, additional studies are needed in different target populations and for other cancer screening tests to confirm that lower screening intentions are associated with lower screening rates for these groups.

Our primary limitations were that participants were English speaking and while we recruited women from two very different geographical regions, our findings may not generalize to other areas. Also, the strength of association between screening intention and behavior may vary in different patient populations and/or screening contexts. Confirming the specific association between screening intention and behavior in the target population would likely improve the accuracy of using screening intention as a proxy outcome for studies examining the effects of interventions on de-implementing cancer screening.

In summary, lower screening intention is associated with lower breast cancer screening rates among older women, suggesting that screening intention is a reasonable proximal outcome for interventions aimed at reducing over-screening in older women.

Appendix.

Table 1a.

Secondary analysis on the association between intentions to be screened and receipt of mammography screening after 18 months follow-up, comparing those with lower intention to be screened with all others.

	Underwent screening (n=316) a	Did not undergo screening (n=217) a	Unadjusted RR (95% CI)	Adjusted RR (95% CI) b
Change in mammography screening intentions from pre to post visit with PCP c
Lower intention to be screened, No. (%)	31 (9.8)	84 (38.7)	0.39 (0.28, 0.56)	0.42 (0.30, 0.60)
All others, No. (%)	285 (90.2)	133 (61.3)	-	-
Post-visit screening intentions d, mean (SD)	13.9 (3.3)	8.4 (6.2)	0.89 (0.86–0.92)	0.90 (0.88–0.93)

Table 1b.

Secondary analysis on the association between intentions to be screened and receipt of mammography screening after 18 months follow-up, comparing those with lower intention to be screened with those with no change in screening intentions.

	Underwent screening (n=292) a	Did not undergo screening (n=199) a	Unadjusted RR (95% CI)	Adjusted RR (95% CI) b
Change in mammography screening intentions from pre to post visit with PCP c
Lower intention to be screened, No. (%)	31 (10.6)	84 (42.2)	0.38 (0.27–0.55)	0.42 (0.29, 0.59)
No Change, No. (%)	261 (89.4)	115 (57.8)	-	-
Post-visit screening intentions d, mean (SD)	13.8 (3.4)	8.0 (6.3)	0.89 (0.87–0.92)	0.90 (0.88–0.93)

Table 1c.

Secondary analysis on the association between intentions to be screened and receipt of mammography screening after 18 months follow-up, also adjusting for education and race.

	Underwent screening (n=316) a	Did not undergo screening (n=217) a	Unadjusted RR (95% CI)	Adjusted RR (95% CI) e
Change in mammography screening intentions from pre to post visit with PCP c
Lower intention to be screened, No. (%)	31 (9.8)	84 (38.7)	0.38 (0.27–0.55)	0.41 (0.29, 0.58)
Higher intention to be screened, No. (%)	24 (7.6)	18 (8.3)	0.85 (0.66–1.09)	0.83 (0.66, 1.06)
No change, No. (%)	261 (82.6)	115 (53.0)	-	-
Post-visit screening intentions d, mean (SD)	13.9 (3.3)	8.4 (6.2)	0.89 (0.86–0.92)	0.90 (0.88–0.93)

^a18 months follow-up was at least 2 years since women’s last screening mammograms.

^bGeneral estimating equation logistic regression models adjusted for treatment arm, PCP panel size and site to account for the clustering by PCP in the trial design.

^cIntentions to be screened was measured on a 15-point scale, ranging from 1 (will not have mammogram) to 15 (will have mammogram).¹⁵

^dUnadjusted and adjusted RR are for each point decrease in screening intention.

^eGeneral estimating equation logistic regression models adjusted for treatment arm, PCP panel size and site to account for the clustering by PCP in the trial design, education, and race.