Reduction in Mammography Screening for Breast Cancer Among Visually Impaired Women

Abstract

Purpose:

Timely receipt of mammograms to screen for breast cancer in accordance with the United States Preventive Services Task Force (USPSTF) recommendations can substantially reduce morbidity and mortality. The purpose of this study is to assess whether odds of receiving screening mammograms are similar for women with and without visual impairment.

Design:

Retrospective longitudinal cohort study.

Participants:

1044 females, 65–72 years of age, enrolled in fee-for-service Medicare from January 1, 2008 to December 31, 2015.

Methods:

We matched patients with no vision loss (NVL), partial vision loss (PVL), and severe vision loss (SVL) 1:1:1 based on age, race, time in Medicare, urbanicity of residence, and overall health. Women with pre-existing breast cancer were excluded. Multivariable conditional logistic regression modeling compared the odds of receiving screening mammography within a 2-year follow-up period among the 3 groups.

Main Outcomes Measures:

Proportion receiving mammography and adjusted odds ratios (OR) of receiving mammography within 2 years of follow-up among the groups.

Results:

A total of 1044 patients were matched (348 in each group). The mean ± SD age at the index date was 69.0 ± 1.5 years for all 3 groups. The proportion of women receiving ≥ 1 mammogram within the 2-year follow-up period was 69.0% (n=240), 56.9% (n=198), and 56.0% (n=195) for the NVL, PVL and SVL groups, respectively (p=0.0005). The mean ± SD number of mammograms received per patient during the 5-year period (3-year look-back plus 2-year follow-up period) was 3.1 ± 2.0, 2.5 ± 2.0, and 2.3 ± 2.1 for the NVL, PVL and SVL groups, respectively (p<0.0001). Women with SVL had 42% decreased odds (OR=0.58; 95% CI: 0.37–0.90, p=0.01), and those with PVL had 44% decreased odds (OR=0.56; CI:0.36–0.87, p=0.009) of receiving mammography during the follow-up period compared to those with NVL.

Conclusions:

Women with visual impairment were significantly less likely to receive mammography to screen for breast cancer than their non-visually-impaired counterparts. Clinicians should look for ways to help ensure that patients with visual impairment receive mammograms and other preventive screenings as recommended by the USPSTF.

PRECIS

Visually-impaired women are substantially less likely to undergo screening mammography in accordance with United States Preventive Services Task Force guidelines to check for breast cancer than their non-visually-impaired counterparts.

INTRODUCTION

Breast cancer is the second leading cause of cancer-related death among women in the United States (U.S.). In 2018 alone, more than 266,000 women were diagnosed with breast cancer, and nearly 50,000 died from this condition.¹ Routine breast cancer screening has been found to substantially reduce breast cancer mortality.^2–5 The U.S. Preventive Services Task Force (USPSTF) makes specific recommendations for preventive care services for patients based on available evidence. Since 2009, the USPSTF has recommended biennial screening mammography for women aged 50 to 74 years old.^6,7 Professional societies offer similar recommendations based on available evidence balancing the benefits of screening against potential harms of frequent diagnostic testing (e.g. false positives, overdiagnosis, and overtreatment); the American Cancer Society recommends that women 55 years and older who are at average risk of breast cancer to undergo biennial screening mammography;⁸ the American College of Obstetricians and Gynecologists recommends breast cancer screening begin no later than 50 years of age and until at least 75 years of age, with screening mammography performed every 1 to 2 years based on a shared decision making process between the physician and patient, considering the benefits and harms of screening mammography.⁹

Visual impairment (VI) is one of the leading causes of disability in the U.S. The prevalence of VI, defined as best-corrected visual acuity below 20/40 in the better seeing eye was estimated at nearly 4 million persons in 2017 and is projected to increase to more than 9.5 million by 2050.¹⁰ Women are more likely than men to have VI, and impaired visual function has been associated with poor overall physical function.^11,12 Prior studies have found that women with physical disabilities are less likely than women without disabilities to receive preventive services, including screening for cervical cancer and breast cancer.^13–15

A prior study using data from 2000 to 2010 focused on women residing in South Carolina found that those with VI (visual impairment) were significantly less likely to undergo screening mammography in accordance with USPSTF recommendations.¹³ We sought to examine whether the trends noted by this earlier study continue to persist, and whether differences noted in this one particular state were found when assessing women across the entire country utilizing a nationally-representative sample of women enrolled in fee-for-service Medicare throughout the entire U.S.

METHODS

Data Source

Data for Medicare beneficiaries was obtained from a nationally representative 20% sample of enrollees with Medicare Parts A and B coverage enrolled in Medicare from January 1, 2008 to December 31, 2015. This database captures all ocular and systemic diagnoses based on International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) billing codes and information regarding diagnostic and therapeutic procedures performed based on Current Procedural Terminology (CPT) and Healthcare Common Procedure Coding System (HCPCS) codes. Prior studies have utilized this dataset to study utilization and outcomes for patients with ocular and non-ocular diseases.^16–18 This study was approved by the University of Michigan Institutional Review Board and adheres to the tenets of the Declaration of Helsinki. Since patient identifiers are removed from the Medicare database before it is made available to researchers, it is impossible to identify the enrollees to obtain their informed consent.

Inclusion and Exclusion Criteria

Females aged 65–72 years with at least 5 years of continuous enrollment in Medicare were eligible for matching. The USPSTF recommendation for breast cancer screening applies to women aged 50 to 74 years; therefore, males and women outside of this age group were excluded. We excluded women <65 years old, as Medicare enrollees under 65 years old are a unique subset of the population who either are eligible for Social Security disability benefits for at least 24 months or who have end-stage renal disease. Women with ≥1 pre-existing breast cancer diagnosis or evidence of treatment were excluded, as these patients may be receiving mammograms for purposes other than screening. Patients with non-continuous enrollment may have had care during the study interval not captured within Medicare data and were also excluded. Likewise, women enrolled in Medicare Advantage plans were excluded as our database lacks complete data of all healthcare services rendered for them.

Identifying Patients with Vision Loss

Patients with VI were identified using ≥1 of the following ICD-9-CM diagnosis codes: 369.0X, 369.1X, 369.2X, 369.3, 369.4, 369.6X, 369.7X, 369.8, 369.9. (Supplemental Table 1, available at http://aaojournal.org). Partial vision loss (PVL) was defined as one or more records of VI involving only 1 eye and no record of severe vision loss (SVL); SVL was defined as visual impairment involving VI of both eyes as described previously.^19.20 Persons were characterized as no vision loss (NVL) if they had no codes for SVL or PVL, as well as no record of any common chronic ocular disease (including age-related macular degeneration, glaucoma, and diabetic retinopathy) during the study period.

Characterization of the Study Period and Index Date

We assigned each patient an index date, which was defined as the date corresponding to 3 years after entry into Medicare. The first 3 years the patients were in Medicare was used as a look-back period to exclude patients with pre-existing breast cancer, to assess the overall health of each patient (for use in our matching algorithm), and to characterize women as possessing NVL, PVL, or SVL. All eligible women were required to have ≥1 visit to an ophthalmologist or optometrist during the look-back period to give them an opportunity to receive a diagnosis of PVL or SVL. We employed a 2-year follow-up period to check for receipt of mammography among the 3 groups. Mammography testing should be performed once every 2 years based on USPSTF recommendations. Eligible patients could be age 65 to 72 years old at the index date. We set a maximum age of 72 years at the index date so these patients could be followed until age 74 to check to see whether they underwent mammography. USPSTF guidelines do not apply to patients older than 74 years of age given lack of sufficient evidence to support continued screening of these patients. (Figure 1)

Figure 1. Study Design.

NVL = no vision loss (no visual impairment or ocular disease in either eye); PVL= partial vision loss (visual impairment involving 1 eye only); SVL = severe vision loss (visual impairment involving both eyes)

Outcomes of Interest

The outcome of interest was receipt of mammography once or more during the 2-year follow-up period. Mammography receipt was identified by CPT and HCPCS billing codes (Supplemental Table 1, available at http://aaojournal.org).

Statistical Analysis

Statistical analyses were performed using SAS software version 9.3 (SAS Institute, Cary, North Carolina). Participant characteristics were summarized for the entire sample using means and standard deviations (SD) for continuous variables and frequencies and percentages for categorical variables.

Women were matched 1:1:1 across NVL, PVL and SVL groups based on age at index date (±2 years), calendar year at index date (±2 years), race, urbanicity, and Charlson Comorbidity Index (CCI) score, (a measure of overall health;²¹ ±1 point) within the 3-year look-back period. Total number and mean number of mammograms received as well as proportion receiving at least 1 mammogram during the 3-year look-back, 2-year follow-up, and overall 5-year study period were recorded for women in all 3 groups. Continuous variables were compared using 2-tailed t-tests (when comparing any vision loss (PVL or SVL) with NVL) and analysis of variance (when comparing the 3 groups). Chi-squared tests were used for categorical variables (comparing any vision loss to NVL for the 3 groups).

Multivariable conditional logistic regression modeling was used to determine the odds of receiving mammography among visually-impaired women (PVL or SVL) compared to those with NVL. Adjusted odds ratios were calculated after adjustment for receipt of mammograms during the 3-year look-back period and receipt of colonoscopy (another preventive screening test recommended by the USPSTF for this age group) during the 5-year study period for each patient. Comparisons were made of NVL vs. PVL, NVL vs. SVL, and PVL vs. SVL. The models generated odds ratios (OR) with 95% confidence intervals (CI). For all analyses, p<0.05 was considered statistically significant.

RESULTS

We were able to successfully match 348 women with NVL, 348 women with PVL, and 348 women with SVL 1:1:1 across all the matching criteria, for a total of sample of 1044 enrollees. For all three groups, the mean ± SD age at index date was 69 ± 1.5 years, the majority of patients were white (n=855, 81.9%), and most resided in urban areas (n=756, 72.4%, [Table 1]).

Table 1.

Characteristics of Study Sample

	Women with NVLa (n=348)	Women with PVLb (n=348)	Women with SVLc (n=348)	p-value
Age at index date, mean±SD	69 ± 1.5	69 ± 1.5	69 ± 1.5	0.73
CCI in look-back period, mean±SD	3.6 ± 2.0	3.3 ± 2.3	3.3 ± 2.3	0.07
Race, n (%)				0.99
White	285 (81.9)	285 (81.9)	285 (81.9)
Black	51 (14.7)	51 (14.7)	51 (14.7)
Latino	2 (0.6)	2 (0.6)	2 (0.6)
Asian	4(1.2)	4(1.2)	4 (1.2)
Other	6(1.7)	6(1.7)	6 (1.7)
Residence, n (%)				0.99
Urban	252 (72.4)	252 (72.4)	252 (72.4)
Large rural	49 (14.1)	49 (14.1)	49 (14.1)
Small rural	47 (13.5)	47 (13.5)	47 (13.5)

^aNVL = no vision loss (no visual impairment or ocular disease in either eye)

^bPVL= partial vision loss (visual impairment involving 1 eye only)

^cSVL = severe vision loss (visual impairment involving both eyes)

CCI = Charlson Comorbidity Index

SD = standard deviation

The number of patients with PVL and SVL receiving ≥1 mammogram during the 3-year look-back period were 253 (72.7%) and 227 (65.2%), respectively, which were lower than the 265 (76.1%) patients with NVL who received ≥1 mammogram during the look-back period (p=0.005). The number of patients with PVL and SVL receiving ≥1 mammogram during the 2-year follow-up period were 198 (56.9%) and 195 (56.0%), respectively, which were considerably lower than the 240 (69.0%) patients with NVL receiving ≥1 mammogram during this period (p=0.0005). The proportions of patients with PVL and SVL receiving at least 1 mammogram in the entire 5-year study period were 268 (77.0%) and 250 (71.8%), respectively, compared with 283 (81.3%) patients with NVL (p=0.01). The mean ± SD number of mammograms received per patient during the 5-year period was 3.1 ± 2.0, 2.5 ± 2.0, and 2.3 ± 2.1 for the NVL, PVL and SVL groups, respectively (p<0.0001, [Table 2]).

Table 2.

Comparison of Receipt of Mammography Among Women With and Without Vision Loss

	Women with NVLa (n=348)	Women with PVLb (n=348)	Women with SVLc (n-348)	p-value
Number of mammograms received, mean±SD
3-year look-back period	1.8 ± 1.3	1.5 ± 1.2	1.4 ± 1.3	<0.0001
2-year follow-up period	1.2 ± 1.0	1.0 ± 1.0	0.9 ± 1.0	0.0004
5-year study period	3.1 ± 2.0	2.5 ± 2.0	2.3 ± 2.1	<0.0001
Patients receiving ≥ 1 mammogram, n (%)
3-year look-back period	265 (76.1)	253 (72.7)	227 (65.2)	0.005
2-year follow-up period	240 (69.0)	198 (56.9)	195 (56.0)	0.0005
5-year study period	283 (81.3)	268 (77.0)	250 (71.8)	0.01
Proportion of patients receiving colonoscopy during enrollment period, n (%)	246 (70.7)	229 (65.8)	225 (64.7)	0.20

^aNVL = no vision loss (no visual impairment or ocular disease in either eye)

^bPVL= partial vision loss (visual impairment involving 1 eye only)

^cSVL = severe vision loss (visual impairment involving both eyes)

SD = standard deviation

After adjusting for receipt of mammograms during the 3-year look-back period and receipt of another preventive screening test (colonoscopy) during the study period, patients with SVL had 42% decreased odds (OR=0.58; 95% CI: 0.37–0.90; p=0.01), and patients with PVL had 44% decreased odds (OR=0.56; CI: 0.36–0.87; p=0.009) of receiving mammography during the 2-year follow-up period compared with patients with NVL (Table 3). There was no significant difference in adjusted odds of receiving mammography for patients with SVL vs. PVL (OR=1.03; CI: 0.67–1.57; p=0.90). Receipt of mammograms during the look-back period (OR=20.63; CI: 10.40–40.93; p <0.001) and receipt of colonoscopy during the study period (OR=3.17; CI: 1.85–5.45; p<0.001) were associated with increased odds of receiving mammograms in the follow-up period among all groups (Table 3).

Table 3.

Comparison of Odds of Receiving Screening Mammography at Least Once in the 2-Year Follow-up Period Among Women With and Without Vision Loss

	OR (95% CI)	p-value
Women with PVL vs. NVL	0.56 (0.36–0.87)	0.009
Women with SVL vs. NVL	0.58 (0.37–0.90)	0.01
Women with SVL vs. PVL	1.03 (0.67–1.57)	0.90
Receipt of mammogram in 3-year look-back period	20.63 (10.40–40.93)	<0.001
Receipt of colonoscopy during Medicare enrollment period	3.17 (1.85–5.45)	<0.001

NVL = no vision loss (no visual impairment or ocular disease in either eye)

PVL= partial vision loss (visual impairment involving 1 eye only)

SVL = severe vision loss (visual impairment involving both eyes)

OR = odds ratio

CI = confidence interval

DISCUSSION

This longitudinal analysis of a nationally-representative cohort of women enrolled in Medicare found that relative to women with NVL, those with PVL or SVL received fewer total mammograms and were more than 40% less likely to receive mammograms to screen for breast cancer in accordance with USPSTF guidelines. These findings persisted even after accounting for demographic factors, urbanicity of residence, patient overall health, prior receipt of mammography, and utilization of another preventive service. Moreover, since all of these women were insured under Medicare, type of insurance could not account for the differences in mammogram utilization noted among the 3 groups.

One of the factors we adjusted for in our model was prior receipt of mammography during the look-back period. We did not know, a priori, whether differential prior receipt of mammography among the 3 groups would affect the odds of receiving mammography during the 2 year follow-up period so we wanted to account for this in our regression model. We thought prior receipt of mammography could be a surrogate for access to a primary care professional who is familiar with the USPSTF guidelines and this may increase the patient’s likelihood of subsequent mammograms during the follow-up period. We were also concerned that prior receipt of mammography might reduce the likelihood of getting this testing during the period of interest since the patient underwent the testing not too long before, even though the guidelines specify that they are due for testing every 2 years. By our adjusting for prior receipt of mammography, we could account for differences that may have existed among the 3 vision loss groups in utilization of this procedure during the look-back period. As reflected in Table 3, we found that prior receipt of mammography greatly increased the odds of getting mammography during the follow-up period. We also found that the differences in the odds of undergoing mammography among the 3 groups during the follow-up period persisted after accounting for prior receipt of this testing.

An additional factor we accounted for in our model was whether the enrollee had undergone colonoscopy, another preventive screening test that the USPSTF recommends patients undergo at least once every 10 years. Our rationale for including this variable in the model was due to a concern that certain healthcare professionals may be more likely to follow the USPSTF guidelines compared to other providers, and patients with NVL may differentially receive care from those providers. Our model results show that even after we account for this potential confounding factor in our model, we are continuing to observe patients in the NVL group with a higher odds of receiving mammography compared to those with PVL or SVL, suggesting that patient-related factors (such as vision loss) may be responsible for the differences in utilization we are observing.

Past research has found that physical disabilities, increased number of medical comorbidities, lack of health insurance, higher out-of-pocket costs, non-white race, lower education status, and lower frequency of primary care visits are known to be associated with poor adherence to mammography screening guidelines.^13,22–29 Horner-Johnson and colleagues showed that women with disabilities that limit their physical mobility are less likely to receive preventive screenings such as mammograms and Pap smears, compared to others who are not physically disabled; these disparities in utilization were most pronounced for women with more complex physical limitations.²⁴ In the present study, visual impairment may make it more challenging for women with vision loss to access the healthcare system compared to their counterparts with NVL; for instance, women with visual disabilities may require caregivers to transport them to screening appointments and escort them to the clinic for examinations. Given mammography is recommended as often as once every 2 years, even PVL may pose a barrier to transportation and affect a woman’s ability to receive screening in accordance with USPSTF guidelines.

We are aware of one other study that assessed the association between sensory impairment and receipt of preventive services in accordance with USPSTF guidelines. Xu and colleagues found that among a cohort of patients in Medicare residing in South Carolina, women with VI were significantly less likely to have full adherence to mammography screening recommendations.¹³ They found that women with VI had 51% decreased odds of receiving screening mammography compared with non-visually impaired counterparts. Our study demonstrates that the findings from the South Carolina group extend beyond that one particular state; across our nationwide sample, women with SVL had 42% decreased odds of screening mammography relative to women with NVL. Additionally, our results using data from 2008–2015 show that these disparities continue to persist beyond the findings from the earlier study which used data from 2000–2010. Together, these findings underscore the need for effective interventions to address barriers to preventive screening among patients with VI.

We found that even among women with NVL, about 20% did not adhere to USPSTF guidelines during the 5-year study period, suggesting that other factors may be affecting utilization of preventive services among these women. Despite having health insurance, out-of-pocket costs of mammography can vary widely, and copayments as low as $10 can serve as deterrents to screening and result in lower mammography rates, especially among women with lower income and/or lower education levels.^30–32 The elimination of cost-sharing in Medicare and the reduction of out-of-pocket costs of screening mammography under the 2011 Affordable Care Act led to a modest increase in mammography utilization from 2010 to 2013 among Medicare beneficiaries aged 65 to 74 years old, demonstrating the potential for increased participation in preventive services when there is a reduction of financial barriers.³³

Financial incentives alone may not fully address the disparities in utilization of preventive services. Trivedi and colleagues, using data from Medicare Advantage plans, found that elimination of cost sharing resulted in increased mammography utilization in some but not all groups; this effect was attenuated among women residing in areas of lower socioeconomic status and was insignificant among Latinas.³⁴ The authors hypothesized that these disparities may stem from lower health literacy among some patients. In addition to expanding insurance coverage and reducing costs of preventive screening, additional studies suggest that implementation of patient navigation services such as utilization of patient advocates to assist with patient education, language interpretation, guidance with understanding physician recommendations, completion of required documents, and reimbursement for transportation may reduce barriers to preventive care and improve adherence with breast cancer screening guidelines.^35–37

Prior research has demonstrated that presence of VI can affect utilization of healthcare services for patients admitted to the hospital. Among a cohort of 12330 Medicare beneficiaries, those with vision loss had longer mean lengths of stay (6.5 vs 5.3 days), higher readmission rates (23.1% vs 18.7%), and higher hospitalization and 90-day post-discharge costs ($64711 vs $61060) compared with their non-visually impaired counterparts, incurring an estimated $500 million in additional costs annually caring for patients with vision loss requiring hospitalization.²⁰

Opportunities to reduce disparities in utilization of preventive services among persons with and without vision loss may be achieved by employing patient navigation services, reminders to healthcare professionals, and offering financial incentives. Patient navigation services aimed at serving patients with disabilities may be used to encourage participation in preventive health services. Electronic medical records could be programmed to flag patients with diagnoses of VI and alert primary care providers to remember to take extra care to ensure that these patients receive timely preventive services. Health insurers could move towards elimination of cost sharing and reduction of financial barriers to preventive screening for patients with vision loss or other disabilities. Alternatively, reimbursement for ordering mammograms in women with disabilities can be made higher than standard reimbursement to incentivize utilization. Together, efforts to increase participation in preventive screening services among these groups may ultimately reduce healthcare costs and poor health outcomes among vulnerable groups such as those with VI.

Strengths of our study include the use a database that captures a nationally-representative sample of older Americans in Medicare who are eligible for preventive services. Additionally, given that all of the women in our study were covered under Medicare, lack of health insurance was not a factor affecting utilization of care between women in the NVL, PVL, and SVL groups. Limitations in this study include its retrospective design and drawbacks inherent to the use of a large administrative database, including inability to consider variables not captured by Medicare claims data such as whether orders for mammograms were entered but not acted on, and certain patient characteristics including health literacy, education, and income. Second, even though there are many women enrolled in Medicare, as a result of our attempting to match women with NVL, PVL and SVL across multiple different criteria (age, race, urbanicity, overall health), there were only 348 per group who could be successfully matched across all these criteria to study. Third, the billing codes for vision loss may not be used universally by all healthcare professionals, and do not differentiate between certain nuances such as central vs. peripheral vision loss; inability to code for these differences may lead to miscategorization of some patients in the NVL group and bias our findings to the null. Although patients with vision loss were significantly less likely to receive mammograms, we also found that only 81.3% of patients with NVL received at least 1 mammogram in the study period, suggesting that other factors may be at play to explain why nearly 1 of 5 women with NVL were ostensibly not receiving any mammograms during the 5-year study period. Patient attitudes toward screening and social determinants of health may contribute to why utilization rates are lower than desired, and may vary among the 3 groups.^28,29 Finally, our findings may not generalize to those with other types of health insurance, patients in Medicare Advantage plans, and uninsured patient populations or patients under the age of 65 years old. One would expect that uninsured patients would be even less likely to obtain screening compared to those in Medicare, and a larger proportion of women in the country with visual impairment would also lack health insurance.

CONCLUSIONS

In this study, women with visual impairment received fewer mammograms and were significantly less likely to receive mammography to screen for breast cancer compared with women with no visual impairment. Clinicians should look for ways to help ensure that patients with visual impairment or other disabilities receive mammograms and other preventive screenings as recommended by the USPSTF.