Effect of Messaging on Support for Breast Cancer Screening Cessation Among Older US Women

Key Points Question Does a message describing rationales for stopping breast cancer screening affect older women’s support for appropriate screening cessation? Findings This 2-wave randomized clinical trial of older women (3051 in wave 1, 2796 in wave 2) used an online survey to assess support for stopping screening in a hypothetical patient and self-screening intentions. A message about stopping screening from a clinician significantly increased support for appropriate screening cessation, with stronger effects (ie, an increase from 18% to 47%) when the message was delivered over time from multiple sources. Meaning These findings suggest that messaging is a viable strategy for reducing overscreening for breast cancer among older women.


Introduction
0][11][12] National data found that 50.6% of women 75 years or older and 42.8% of women with a life expectancy of less than 10 years received breast cancer screening, suggesting overscreening. 104][15][16] In contrast, there has been little messaging about the harms of overscreening or that stopping screening may be appropriate for some women.Messaging strategies have been used successfully to reduce other unwanted health behaviors such as smoking but are an understudied approach to reduce overscreening. 17,18In prior work, we developed and evaluated messages on breast cancer screening cessation for older women. 19We combined the top-rated message elements from this prior work into a single message in the present study to examine whether message exposure would increase support for and intentions of stopping screening.
When considering effective communication strategies, message source also matters.[16][19][20][21][22][23][24] Therefore, we also sought to experimentally evaluate whether receiving consistent messages about stopping screening from nonclinician sources may amplify the message effect compared with receiving a message from a clinician alone.

Study Overview
We conducted a 2-wave, randomized clinical online survey experiment with women 65 years and older.We included women aged 65 to 75 years because women in this age group with multiple chronic conditions and functional impairments may have a life expectancy of less than 10 years. 25rthermore, the message may prime women to consider stopping screening in the future and/or influence others in their social network.
Participants were randomized into groups that varied in the number of messages they received over 2 points (ie, 0, 1, or 2), the source of the messages (ie, clinician only, a news story plus a clinician, or a family member plus a clinician), and the message content (ie, promoting screening cessation vs screening continuation).In this study, we only report on the effects of the screening cessation message (the effects of conflicting messages wherein participants were exposed to both screening cessation and continuation message contents will be reported elsewhere).This study followed the

Survey Instrument
The survey instrument was developed by the study team and cognitively tested with 6 older women prior to study initiation; we iteratively revised the instrument wordings based on their feedback.The survey described a hypothetical 75-year-old woman with serious health problems and functional limitations but who was not imminently dying (eAppendix in Supplement 2).We used this hypothetical scenario to provide a standardized and relevant context of an older woman with a life expectancy of less than 10 years for whom it would be appropriate to stop screening. 7,8We mentioned that the hypothetical patient had no history of breast cancer and had regular mammograms and that her most recent mammogram results were normal.
We then randomized exposure to a message that included rationales for stopping screening.
Based on our prior work developing and evaluating messages on stopping screening, 19 the message in this study mentioned guideline recommendations, an anecdote about women who experienced false-positive results, and evidence on overdiagnosis.We described this message as from one of 3 sources: (1) the hypothetical patient's primary care physician, who shared information during a clinic visit; (2) a news story that the hypothetical patient read in USA Today; or (3) the hypothetical patient's close family member who shared information during a social visit.We made minor modifications based on message source; for example, the message from the clinician said, "I have several older patients," whereas the message from the family member said, "I know several older women."Complete survey details are provided in the eAppendix in Supplement 2.

Experimental Design and Outcomes
All participants viewed the patient vignette and then were randomized into 4 groups (Figure 1).To ensure matched characteristics across groups, participants were sorted by self-reported race and ethnicity, geographic region, and educational level (using information collected about KnowledgePanel members previously), and then a random number generated by survey programming allocated participants sequentially to each experimental group.The research team was blinded to the allocation until data collection was complete.
Group 1 (the control group) received no message at either wave.Group 2 received the message on stopping screening from the clinician at wave 1 and no message at wave 2. Group 3 received the message on stopping screening from a news story at wave 1 and the same message from the clinician at wave 2. Group 4 received the message on stopping screening from the family member at wave 1 and the same message from the clinician at wave 2. By presenting the nonclinician message first, we aimed to mimic a scenario wherein patients may be exposed to information from their social networks and/or the media before a clinic visit.We also assessed other consequences of message exposure, all measured on a 5-point Likert scale at wave 1.We assessed the extent that the message might cause a general negative perception of mammograms.This was a consequence we aimed to avoid since our goal was to inform women of the potential harms of overscreening rather than dissuade all women from mammograms.We asked whether the message would "discourage" one from getting mammograms, make getting mammograms "unpleasant," or make one "concerned about the health effects" of getting mammograms (1 indicates strongly disagree; 5, strongly agree). 27We also assessed 2 negative emotional responses: whether the message made one feel "annoyed" or "worried" (1 indicates not at all; 5, extremely).These outcomes were only assessed in groups 2 to 4.
The survey instrument also assessed prior breast cancer screening, Gail-model breast cancer risk factors (eg, family history, number of breast biopsies), 28 cancer worry, health literacy, 29 and health or functional status (used to estimate 10-year life expectancy). 25KnowledgePanel provided participants' age, sex, race and ethnicity, educational level, and geographic region.Only deidentified data were provided.For select items, a repeat prompt was used if the item was skipped.

Statistical Analysis
The analysis focused on (1) whether exposure to any message about stopping screening would increase support for stopping screening (ie, comparing groups 2-4 vs 1); and (2) whether exposure to consistent messages from 2 sources was more effective than exposure to a single message from the clinician (ie, comparing groups 3 and 4 vs 2).We focused on results at wave 2 to examine the cumulative effects of multiple message exposures.Informed by preliminary data, we estimated 400  participants per group at wave 2 would allow 80% power to detect a 0.30-point difference on the 7-point scale for the primary outcome at an α level of .05.Assuming 20% attrition between waves, we aimed to recruit 500 women per group.Due to high response rate in a shorter-than-expected time, actual recruitment exceeded planned sample size.
Regarding support for stopping screening for the hypothetical patient (primary outcome) and screening intention for oneself (secondary outcome), we compared the mean scores across groups at wave 2 using analysis of variance with adjusting for multiple comparisons using the Tukey test.For ease of interpretation, we also dichotomized these 2 outcomes where scores 5 to 7 were categorized as supports screening cessation and scores 1 to 4 as does not support screening cessation.8][9] We also explored screening intention for oneself at wave 2 among participants with higher breast cancer risk, defined as 5-year risk of at least 3%, which is the threshold used in guidelines for considering cancer prevention medications. 30,31For the other consequence outcomes, analyses focused on summarizing mean scores across groups.All analyses were performed using Stata, version 17 (StataCorp LLC).Two-sided P Յ .05indicated statistical significance.

Results
This study included 3051 women in wave

Support for Screening Cessation for the Hypothetical Patient
At wave 1, the control group that received no message (group 1) had the lowest support for stopping screening for the hypothetical patient ( among participants in group 3 (Figure 2).

Screening Cessation Intention for Oneself
Messaging effects on screening cessation intentions for oneself followed a similar pattern as that on support for screening cessation for the hypothetical patient, but the magnitude of the effects was smaller.At wave Participants who were 75 years or older or with a life expectancy of less than 10 years had higher intentions to stop screening and larger message effects compared with the entire sample (Table 2 and Figure 2).Among participants 75 years or older, intentions to stop screening at wave 2 had a mean score of 3.83 (95% CI, 3.50-4.16)for group 3 and 3.80 (95% CI, 3.46-4.13)for group 4; these b Includes 2992 participants.Life expectancy was estimated using the Schonberg mortality index. 25cores for participants ranged from 0 to 19.Scores of 10 or greater are associated with a greater than 50% chance of 10-year mortality.Thus, women who score 10 or greater are estimated to have a life expectancy of less than 10 years.
c Includes non-Hispanic multiracial and non-Hispanic other race.
d Includes 3047 participants.
e Includes 3044 participants.
f Includes 2966 participants.
g Includes 3038 participants.Health literacy was assessed in a single validated question: "How confident are you filling out medical forms?" 29 Responses of not at all, a little bit, and somewhat confident were categorized as low health literacy; responses of quite a bit and extremely confident were categorized as normal health literacy.

Other Consequences of Message Exposure
Participants reported low rates of negative perceptions of mammograms after reading the message.
All mean scores were less than 3 (the neutral option on the 5-point scale) regarding whether the There was no significant difference between groups 3 and 4 for any of the comparisons.
b The mean was significantly different when compared with group 1 (P Յ .05).
c The mean was significantly different than group 2 (P Յ .05).

Discussion
This is the first national study, to our knowledge, to rigorously test the effects of messaging about breast cancer screening cessation.We found that message exposure increased support for and intentions of stopping screening among older women for whom stopping screening would be  b Adapted from scale used in health communication. 27Measured on 5-point Likert scale where 1 indicates strongly disagree; 5, strongly agree.
c Measured on 5-point Likert scale where 1 indicates not at all; 5, extremely.
appropriate based on age (Ն75 years) or health (life expectancy of <10 years), and the effect was greatest when the message was delivered from multiple sources.Although messaging strategies have been used to change other health behaviors, 17,18,32 they have not been used to encourage screening cessation to reduce overscreening in older women.These results show that messaging is a viable and potentially effective approach for reducing overscreening.
For a hypothetical older woman with multimorbidity and functional impairment, message exposure increased support for stopping screening by almost 30% (1.4-1.6 points on a 7-point scale).
This was achieved with low rates of negative perception of mammograms or negative emotional response.The message effect was appropriately small on participants' own screening intentions since most participants did not meet guideline criteria for stopping screening.The message effect was also appropriately small among participants with higher cancer risk.In the subgroups of women who met guideline criteria for stopping screening, the message effect was larger.Among those 75 years or older or with life expectancy of less than 10 years, the increase in intention to stop screening was 1.0 to 1.2 points on a 7-point scale.This is a sizeable change; in a randomized clinical trial of a breast cancer decision aid among older women, each 1-point change on a 15-point scale of screening intention corresponded with a 10% decrease in screening rate 33 ; a 1-point change on a 7-point scale may potentially correspond to a 10% or larger decrease in screening rate but needs to be confirmed in future studies that measure screening behavior.Although these analyses were exploratory, our findings suggest that messaging holds promise to selectively reduce screening in older women for whom it would be appropriate to stop screening without significantly reducing screening in older women who may still benefit.
Our study of the cumulative effects from multiple message exposures on screening was novel.
Literature on individualized cancer screening for older adults highlights the importance of multilevel influences. 34Our findings suggest that while clinicians are an important target for efforts to reduce overscreening, leveraging other information sources may increase impact.Although we found low levels of negative reactions to the message, engaging key partners to devise message delivery can help identify strategies to further minimize undesirable outcomes from messaging.

Strengths and Limitations
Strengths of our study include a large national sample, randomized experimental design, and study of message exposure over time.Limitations include sampling bias and nonresponse bias.Compared with national census data, 35 our study participants were younger, more likely to be White, and had a higher level of education.7][38] These differences may affect the generalizability of our findings, but the randomized design helps ensure internal validity.Future studies need to examine whether our results can be replicated in different populations.Our primary outcome focused on a hypothetical scenario where the responses may not fully reflect perspectives in real clinical decisions; we mitigated this by also assessing self-screening intentions.We focused on studying message sources in addition to the clinician and did not include combinations of nonclinician sources.Our recruitment exceeded the planned sample size, and the added statistical power can reveal nonmeaningful differences.However, we followed our analysis plan formulated a priori, and our results have face validity.Last, our study design did not allow assessment of actual change in screening behavior.However, screening intention is a strong predictor of behavior. 39

Conclusions
In this randomized 2-wave clinical trial, we demonstrate that a message describing the rationale for stopping screening can significantly increase support and intention for appropriate screening cessation among older women.This message can help inform public-or patient-facing materials and be incorporated into decision-support tools and electronic medical record prompts to optimize screening.The message can also be delivered as a stand-alone intervention to reduce overscreening.

Figure 1 .
Figure 1.Participant Flow and Study Design

Figure 2 .
Figure 2. Percentages of Participants Who "Supported Screening Cessation" at Wave 2 Regarding the Hypothetical Patient and Themselves

JAMA Network Open | Geriatrics Effect
26 Messaging on Support for Breast Cancer Screening Cessation Reporting Results of Internet E-Surveys (CHERRIES).This study was approved by the Institutional Review Board of Johns Hopkins University School of Medicine, and all participants provided consent via completion of the survey.The trial protocol is provided in Supplement 1.Recruitment used a nationally representative, probability-based online panel (KnowledgePanel).26Panelmembersare recruited by random digit dialing (until 2009) and addressbased sampling (since 2009).Panel members were invited to participate via email if they were women 65 years or older, were English-speaking, and had no history of breast cancer.Each received a unique survey link that could only be used once.The study ended once the target sample size was reached.Panel members received incentives from KnowledgePanel using a point system.The first wave of the survey was administered between May 12 and 30, 2023, and the second wave was administered between May 30 and June19, 2023.

Table 2 )
, with a mean score of 2.66 (95% CI, 2.52-2.81) on the 7-point scale (with higher scores indicating stronger support for screening cessation).Receiving a screening cessation message at wave 1 from any source was associated with significantly higher support for screening cessation, with comparable effects when delivered by a clinician (mean score, and 2 (P < .001for all comparisons) but were not significantly different from each other (P = .70).When the outcome was dichotomized, this translated to an increase from 121 participants (17.8%) who supported screening cessation for the hypothetical patient in group 1 to 331 (47.0%) support for stopping screening was still significantly higher than group 1, with mean score of 3.14 (95% CI, 2.99-3.29;P<.001).Groups 3 and 4, which received 2 stopping screening messages over time, had the highest support for screening cessation; the mean scores were 4.23 (95% CI, 4.09-4.38)forthenews story plus clinician message (group 3) and 4.12 (95% CI, 3.97-4.27)forthefamily member plus clinician messages (group 4).These 2 scores were significantly higher comparedJAMA Network Open | GeriatricsEffect of Messaging on Support for Breast Cancer Screening Cessation JAMA Network Open.2024;7(8):e2428700.doi:10.1001/jamanetworkopen.2024.28700(Reprinted) August 19, 2024 5/12 Downloaded from jamanetwork.comby guest on 08/31/2024 with groups 1

Table 2 .
Mean Outcome Scores a Effect of Messaging on Support for Breast Cancer Screening Cessation Downloaded from jamanetwork.combyguest on 08/31/2024 message discouraged one from getting mammograms, made mammograms seem unpleasant, and made one concerned about mammograms (Table3), with only 3.5% (27 of 763) to 10.2% (77 of 756) who strongly agreed with any statement across groups.Similarly, we found low rates of reported worry or annoyance after reading the message (all means <3), with only 4.0% (30 of 754) to 8.3% (63 of 759) who felt extremely worried and 11.8% (89 of 752) to 14.1% (106 of 754) who felt extremely annoyed across groups.

Table 3 .
Means Scores for Other Consequences From Message Exposure a Data are presented as mean (95% CI) unless noted otherwise.Outcomes were only assessed at wave 1. Group 1 received no messages; group 2, a message from a clinician only in wave 1; group 3, messages from a news story in wave 1 and a clinician in wave 2; and group 4, messages from a family member in wave 1 and a clinician in wave 2. a Effect of Messaging on Support for Breast Cancer Screening Cessation Schoenborn NL, Pinheiro A, Kistler CE, Schonberg MA.Association between breast cancer screening intention and behavior in the context of screening cessation in older women.Med Decis Making.2021;41(2):240-244.doi: 10.1177/0272989X20979108 Comparison of Responders vs Nonresponders eTable 2. Comparison of Participants Who Completed Both Waves 1 and 2 vs Those Who Completed Wave 1 Only and Did Not Complete Wave 2 eTable 3. Mean Outcome Scores Regarding Stopping Screening Intention for Oneself Among Those With Higher JAMA Network Open.2024;7(8):e2428700.doi:10.1001/jamanetworkopen.2024.28700(Reprinted) August 19, 2024 9/12 39.