Breast cancer screening among individuals with a substance use disorder: a retrospective cohort study

Abstract

Purpose

There is limited evidence about whether a substance use disorder (SUD) is a barrier to breast cancer screening. Because SUDs are highly prevalent in the USA, it is important to establish whether this patient population is less likely to obtain screening.

Methods

This retrospective cohort study included 220 227 patients, with 209 132 having no SUD and 11 095 (5.0%) with SUD based on electronic health record data in a multi-state, Midwestern healthcare system (1 January 2018–31 December 2022). The outcome was the receipt of a mammogram in the 5-year follow-up period. Patients were women aged 40–69 years as of 1 January 2018, with ≥ 2 in-person primary care visits between 2018 and 2022. Covariates included demographics, health services utilization, and physical/psychiatric conditions.

Results

Mean age of the sample was 54.7 (± 8.3) years old. After controlling for confounding, women without any SUDs had more than twice the odds of mammogram receipt compared to those with stimulant use disorder (odds ratio [OR] 2.06; 95% confidence interval [CI]: 1.83–2.33). Women with no SUDs had 89% higher odds of mammogram receipt compared to those with opioid use disorder (OR 1.89; 95% CI: 1.76–2.03), followed by “other” SUDs (OR 1.86; 95% CI: 1.69–2.06), sedative use (OR 1.70; 95% CI: 1.43–2.04), cannabis use (OR 1.58; 95% CI: 1.44–1.74), and alcohol use disorders (OR 1.49; 95% CI: 1.41–1.58).

Conclusions

Despite the high prevalence of SUDs, evidence of preventive service delivery among individuals with SUDs is still lacking. Further research is needed to investigate other healthcare disparities in preventive service delivery among individuals with SUDs.

Introduction

In 2022, more than 2.3 million women were diagnosed with breast cancer and 670 000 people died from it globally [1]. Screening for breast cancer through mammography is known to decrease breast cancer mortality and reduce rates of advanced breast cancer in a cost-effective way [2–5]. The US Preventive Services Task Force (USPSTF) previously recommended biennial screening mammography for women 50 to 74 years old and selective screening for women 40–49 years old after shared decision-making [6]. However, this recommendation changed in April 2024, when the USPSTF advised biennial screening mammography for women aged 40 to 74 years old [7–9]. While the 2021 breast cancer screening completion rate in the USA was 75.9% in the general population, the coronavirus disease 2019 (COVID-19) pandemic negatively affected follow-up screening within 24 months after their previous screening, especially among certain vulnerable populations, such as older groups (those who are 65 years or older) and racially marginalized groups [10, 11]. To achieve the Healthy People 2030 goal of 80.5% breast cancer screening rates, it is essential to continuously promote breast cancer screening for eligible individuals and address potential barriers, especially among vulnerable groups, such as people with substance use disorders (SUDs).

Unhealthy substance use and SUDs are associated with lower use of cancer screening, including colorectal cancer [12, 13]. Individuals with SUDs may underutilize cancer screening for several reasons, including a tendency to prioritize immediate needs over long-term health benefits, distrust of healthcare, and stigma around SUDs [14–17]. More than 48 million individuals aged 12 years or older live with SUDs, including 29.5 million with alcohol use disorder (AUD) and 6.1 million with opioid use disorder (OUD) [18]. With such a large prevalence of a potential barrier to screening, further research is warranted to determine if an SUD is associated with decreased rates of breast cancer screening. In addition, we investigated the association between specific types of SUDs (e.g. opioids, alcohol) and receipt of breast cancer screening. Last, race is associated with barriers to breast cancer screening and the prevalence of different SUDs varies by race [19–21]. Thus, we computed race-stratified models of the association between SUDs and breast cancer screening.

Methods

De-identified electronic health records for this study were obtained from Saint Louis University-SSM (SLU-SSM) healthcare System’s Virtual Data Warehouse (VDW). The VDW contains extracted electronic health record data from discrete fields. Notes are not included in the database. The VDW contains diagnostic codes, procedure codes, prescription orders, laboratory results, health care use, and demographics which are used to create study variables. Our data were de-identified by honest brokers before it was released for research purposes. SLU-SSM’s VDW was created and is maintained per Health Care Systems Research Network specifications [22]. The SSM healthcare system includes locations in four states: Missouri, Illinois, Oklahoma, and Wisconsin. The VDW contains de-identified clinical data for over 5 million patients dating back to 2008. Multiple studies have been published using this data source [12, 23–25]. This study was approved as non-human subjects research by the Saint Louis University Institutional Review Board.

Sample definition

For this retrospective cohort study, we selected women who will be eligible for mammography defined as aged 40–69 years as of 1 January 2018. Eligible patients had an in-person primary care visit in two different years between 2018 and 2022 to ensure patients were regular users of the SSM healthcare system to increase the probability of follow-up data. We chose this 5-year observation period to ensure patients had at least 2 years to obtain a mammogram according to the new USPSTF recommendations [7], and to allow adequate follow-up time for all patients following their mammogram receipt. Departments considered primary care included Family Medicine, General Internal Medicine, and Obstetrics and Gynecology. Patients were excluded if records indicated a mastectomy or breast cancer diagnosis prior to 1 January 2018, or the first mammogram during the study period. Full details of the cohort sampling approach are provided in Fig. 1 (cohort flow chart). Our database includes patient records from multiple sites from a large Midwestern and Southern healthcare system, capturing variations in patient demographics and characteristics. These criteria resulted in a cohort of 221 029 women eligible for mammography during the observation period.

Variable definitions

The outcome of interest was the receipt of a mammogram during the observation period as determined by Current Procedural Terminology codes. The primary exposure was a diagnosis of a SUD, defined using diagnostic International Classification of Disease-10 diagnostic codes. Substances included alcohol, cannabis, opioids, sedatives, stimulants, and hallucinogens/other substances. Detailed definitions for all variables are included in Supplementary Appendix A.

Covariates included demographics, health care utilization, and comorbid psychiatric and physical conditions. Demographics included age as of 1 January 2018, sex, self-identified race, and neighborhood socioeconomic status (nSES) (ref) [26]. Charlson Comorbidity Index (CCI) was defined based on diagnoses received between 1 January 2018 and 31 December 2022 [27]. Higher CCI scores indicate greater morbidity and mortality risk. To control for detection bias, we adjusted for healthcare utilization determined by counting unique outpatient visit dates during the observation period, then categorizing based on quartile. The highest quartile was ≥ 40 days with a clinic visit. Other comorbid conditions included depression, anxiety disorder, posttraumatic stress disorder (PTSD), obsessive-compulsive disorder, severe mental illness defined as a diagnosis for bipolar and or schizophrenia, obesity, and hypertension. We controlled for pain diagnoses because non-cancer pain is common in SUDs and may be a barrier to preventive health utilization [28, 29]. Pain diagnoses included arthritis, back pain, muscle pain, neuropathy, headache, fibromyalgia, and chronic pain.

Analytic approach

Continuous variables were summarized by means and standard deviations. Categorical variables were summarized by percent and count. Due to the large sample size, the standardized mean difference percent (SMD%) was calculated to quantify the association between each variable and the outcome. Any SMD% greater than 10 is considered a meaningful difference between groups [30]. Unadjusted and adjusted logistic regression models were computed to estimate the association between each SUD and mammogram receipt. Measures of association were expressed as odds ratios (ORs) and 95% confidence intervals (CIs). An alpha of 0.05 was used for all tests and SAS v9.4 (Cary, NC) was used for all analyses.

Primary logistic regression models measured the association between any SUDs and mammogram receipt. In subgroup analyses, we determined if the association between SUDs and breast cancer screening varied by type of SUDs. Because the prevalence of SUDs and breast cancer screening differs by race, we computed models stratified by race/ethnicity (White, Black, and Other) [31, 32]. We also conducted additional analyses using the number of mammograms received as a count outcome. For these analyses, negative binomial regression models were utilized to account for the inflated proportion receiving zero mammograms during the study period.

Results

As seen in Table 1, the average age for this cohort of primary care patients was 54.7 (± 8.3) years old and the average CCI was 1.4 (± 2.2). Patients were most commonly White race (86.2%). The most prevalent comorbid conditions included hypertension (48.5%) and chronic pain (32.9%). Of the substances included, AUD (2.4%) and OUD (1.6%) were the most prevalent. Overall, 59.9% of the cohort (n = 131 950) received at least one mammogram during the 5-year observation period.

Several demographic factors were meaningfully associated with mammogram receipt, see Table 1. Those receiving a mammogram were older (55.3 years vs. 54.0 years, SMD% 15.6), more likely to be in the highest healthcare utilization quartile (31.8% vs. 14.3%, SMD% 42.5), and more likely to be in the highest neighborhood SES (37.0% vs. 32.0%, SMD% 10.5), Chronic pain (35.9% vs. 28.4%, SMD% 16.1) and obesity (33.2% vs. 26.0%, SMD% 15.8) were associated with a higher likelihood of mammogram receipt In bivariate analyses, SUDs were not meaningfully associated with mammogram receipt (SMD% < 10).

Table 2 presents the unadjusted and adjusted logistic regression results estimating the association between each substance and mammogram receipt. All substances investigated showed an inverse relationship with mammogram receipt. The strongest association with mammogram receipt was seen for stimulant use disorder. In an unadjusted model, those without an SUD had 72% higher odds of mammogram receipt compared to those with stimulant use disorder (OR 1.72; 95% CI: 1.53–1.93). All other substances investigated also showed an inverse relationship with mammogram receipt, with the weakest association observed among AUD (OR 1.19; 95% CI: 1.13–1.26).

After adjusting for potential confounders (see Table 2), the strength of association increased between each SUD and mammogram receipt. Those without an SUD had more than twice the odds of mammogram receipt compared to those with stimulant use disorder (OR 2.06; 95% CI: 1.83–2.33). Those without an SUD had 89% higher odds of mammogram receipt compared to those with OUD (OR 1.89; 95% CI: 1.76–2.03), followed by “other” SUD (OR 1.86; 95% CI: 1.69–2.06), sedative use disorder (OR 1.70; 95% CI: 1.43–2.04), cannabis use disorder (OR 1.58; 95% CI: 1.44–1.74), and AUD (OR 1.49; 95% CI: 1.41–1.58).

Table 3 summarizes the adjusted association between each SUD and mammogram receipt, stratified by race. The association between any SUD diagnosis and mammogram receipt was lower among Black patients (OR 1.36; 95% CI: 1.20–1.53) than among White patients (OR 1.64; 95% CI: 1.56–1.71). The association between sedative use disorder and mammogram receipt remains significant only among White patients (OR 1.74; 95% CI: 1.44–2.09). Among Black patients, an even larger point estimate is observed, but the association is not statistically significant (OR 2.04; 95% CI: 0.98–4.25), and among the other race category, no association is observed (OR 0.59; 95% CI: 0.18–1.89).

Table 4 displays the crude and adjusted association between SUDs and the number of mammograms received during the study period, which are similar to the logistic regression results presented above. The strongest association again was seen with stimulant SUD; those without an SUD received mammograms at a 75% higher rate than those with stimulant SUD after adjusting for confounders (relative risk [RR] 1.75; 95% CI 1.62–1.88). The weakest association was observed with sedative SUD (RR 1.37; 1.24–1.51). All other SUD rate ratios fell between these two point estimates.

Discussion

In a large retrospective cohort study with a 5-year follow-up period, the authors found significant associations between breast cancer screening completion and having any SUD diagnosis, excluding tobacco use disorder. Consistent with the study hypothesis, individuals without SUDs were more likely to receive a mammogram compared to individuals with SUDs. This association was independent of demographics, healthcare utilization, neighborhood SES, and chronic conditions such as obesity, hypertension, depression, and chronic pain. Our study was conducted in a sample of patients who utilize primary care, and the disparity in the general population may be worse given the underutilization of healthcare among individuals with SUDs [18].

Our study findings expand upon the literature addressing lower preventive service utilization by people with SUDs, including flu vaccines, cervical cancer screening, and colorectal cancer screening [12, 13, 33]. Consistent with our findings, a retrospective cross-sectional study of 4804 women eligible for mammograms found that individuals with unhealthy substance use in one healthcare system in the Eastern USA were less likely to receive breast cancer screening (75.4% vs. 83.8%, P < 0.01) [13]. Our retrospective cohort study showed that individuals with SUD were significantly less likely to receive a colorectal cancer colonoscopy screening (relative ratios = 0.81; 95% CI = 0.74, 0.89) [12]. On the other hand, a survey-based study showed that current or previous drug use was not associated with lower mammogram completion rates, which was inconsistent with our research [33]. This discrepancy may come from the small sample size of this survey study as only 298 women were surveyed for this research compared with our sample of 221 029 individuals.

To the best of our knowledge, this is the first study that examined the association between specific types of SUDs and breast cancer screening rates. Individual SUDs, excluding tobacco use disorder, were independently associated with lower breast cancer screening completion rates. Family physicians are well-positioned to recognize and address unique barriers to breast cancer screening among individuals with an SUD in the outpatient setting. Family physicians play a vital role in both screening for and treating SUDs, while also offering preventive services [34, 35]. Although mistrust of the healthcare system can hinder individuals with SUDs from health maintenance, family physicians can effectively offer preventive services through their longitudinal physician-patient relationship [36].

Additionally, multiple research findings showed that racially and ethnically marginalized populations are less likely to receive mammograms, compared with their counterparts [37–39]. A cross-sectional study of 26 476 women showed that Black individuals were 15%–26% less likely to receive mammograms, compared with White women [19]. Our finding that women of color with any SUD were less likely to receive mammogram screening than White women is consistent with the existing literature. Analyses by specific SUD type were not statistically significant between racial groups and were likely due to the small sample size. Nonetheless, these findings underscore the persistent racial and ethnic disparities in access to preventive services that are also present among people with SUDs, a group that experiences exponential stigma and marginalization within the healthcare system [17].

There are several limitations in this study. First, our study can access mammogram orders, which can differ from the completion of mammogram screening tests. Some patients might have received their mammogram orders outside the health system for which this study has recorded. Second, our study utilized data from a large Midwestern and Southern healthcare system, which may not generalize to other geographic regions. Thirdly, unmeasured confounding could be present and bias our results. Misclassification is a potential limitation. If we classified true positives as not having SUDs, results could be biased towards the null. Lastly, this study included women aged 40–49 years who may have decided against screening following patient-provider shared decision-making that may have occurred due to falling within the USPSTF grade C breast cancer screening recommendation during the study observation period [6].

Conclusions

Despite the high prevalence of SUDs, evidence of preventive services among individuals with SUDs is still lacking. Further research is needed to identify potential healthcare disparities among individuals with SUDs in terms of preventive services such as cervical cancer.

Given the high prevalence of SUDs among people eligible for breast cancer screening, family physicians are uniquely poised to harness their relationships with their patients to help identify and overcome barriers to breast cancer screening. Addressing barriers to screening is particularly important for people at high risk of mortality from breast cancer, such as racial minorities.

Acknowledgements

None declarerd.

Conflict of interest

None declared.

Funding

Funds to maintain the Virtual Data Warehouse came from the Saint Louis University Research Institute. The AHEAD Institute at Saint Louis University provided data management and biostatistical support. Dr Gebauer’s time was covered by NIH K23AR079035. Dr Bello’s time was covered by NIH K23DA053433.

Data availability

The data are proprietary. Interested individuals may contact Dr Sonoda for details regarding data access.

Conference presentation

None declared.

References