Factors Associated With the Receipt of Female Breast Cancer Treatment Among People Living With Intellectual or Developmental Disabilities: A Population-Based Retrospective Cohort Study.

1
Introduction
People with intellectual or developmental disabilities (IDD) experience worse cancer‐related healthcare and outcomes than those without IDD, including lower rates of cancer screening, more advanced stages at diagnosis and lower survival rates (Cobigo et al. 2013; Mahar et al. 2023; Hansford et al. 2024). These disparities are likely due to a combination of individual‐ and system‐level factors, including characteristics of having IDD as well as how social and health systems interact with cancer patients with IDD (Mahar et al. 2023; Stirling et al. 2021). A recent population‐based Scottish study reported that while breast cancer may be less common among females with IDD than in the general population, breast cancer mortality is higher (Ward et al. 2024). Similarly, a Canadian population‐based retrospective study identified that people with IDD are 2.74 times more likely to die following a breast cancer diagnosis than those without IDD (Hansford et al. 2024). One possible contributor to worse breast cancer outcomes among people with IDD could be a lower likelihood of receipt or completeness of cancer treatment (Boonman et al. 2022).
International, national and local guidelines for breast cancer treatment and decision‐making exist based on decades of research (Gradishar et al. 2017; Balic et al. 2019). Treatment options include surgical resection, radiation and systemic therapy (Smolarz et al. 2022). The diagnosis and staging of breast cancer include a mammogram and core biopsy, assessment of biomarkers relevant to treatment decision‐making, blood work and, in some cases, imaging of the chest, abdomen and pelvis and bone scans (Smolarz et al. 2022). Not all people eligible for treatment will have it offered or completed. Factors unrelated to the patient's health or the cancer itself, including social determinants of health, may influence cancer treatment (Lambert et al. 2023; Adsul et al. 2023). The World Health Organization defines social determinants of health as non‐clinical or medical factors that affect health outcomes, including but not limited to racism, socio‐economic status (SES) and rurality (World Health Organization 2024). Such non‐clinical factors have been associated with breast cancer treatment in the general population. For example, in the United States, non‐Hispanic Black people with breast cancer are less likely to receive breast cancer–directed surgery relative to White people (Markey et al. 2022). Lower SES has been associated with lower rates of breast cancer treatment, including sentinel lymph node biopsy, radiation, axillary surgery and chemotherapy (Neuner et al. 2020; Dreyer et al. 2017). Additionally, having additional comorbidities has been documented with lower cancer treatment rates (Safarti et al. 2014; Houterman et al. 2004; Minicozzi et al. 2019), including increased likelihood of receipt of mastectomy rather than breast‐conserving surgery (Minicozzi et al. 2019).
Inferior survival outcomes among patients with IDD may arise in part due to disparities in breast cancer treatment. While clinical factors, including cancer staging, are associated with treatment decisions, sociodemographic or healthcare access–related variables could influence the type of breast cancer treatment people with IDD receive. Exploring factors associated with receipt of breast cancer treatment among people with IDD could help us identify possible areas to target for improving care. Therefore, we examined factors associated with receipt of breast cancer treatment among people with IDD diagnosed with Stage I–III breast cancer.

2
Methods
2.1
Setting and Study Design
We employed a population‐based retrospective cohort study using the Ontario health administrative data held at ICES Queen's. ICES is an independent, non‐profit research institute whose legal status under Ontario's health information privacy law allows it to collect and analyse healthcare and demographic data, without consent, for health system evaluation and improvement. Datasets were linked using unique encoded identifiers and analysed at ICES. Ontario has a population of approximately 15 million residents (Statistics Canada 2022). Ontario has a universal healthcare system, and the majority of cancer diagnostic procedures, staging investigations and treatments is publicly funded.

2.2
Data Sources
We linked multiple ICES databases using unique encoded identifiers. Data sources included hospitalization records (the Canadian Institute of Health Information Discharge Abstract & Same Day Surgery and the Ontario Mental Health Reporting System [OMHRS]), physician billing records (Ontario Health Insurance Plan [OHIP]), emergency department visit records (National Ambulatory Care Reporting System), home care visit records (Home Care Database), continuing care records (Continuing Care Reporting System, Long‐Term Care), sociodemographic data (Registered Persons Database [RPDB] and Ontario Marginalization Index), the cancer registry (Ontario Cancer Registry [OCR]) and the database of cancer‐specific treatment (Cancer Care Ontario Activity Level Reporting [ALR]). A description of each dataset used is provided in Table S1. Stage data have been captured using the Collaborative Staging system in the OCR database since 2007.

2.3
Study Population
The study population included adult females with IDD (≥ 18 years) eligible for the OHIP, diagnosed with breast cancer between 2007 and 2018 as identified by the OCR (ICD‐O‐3 codes included topography = C500–C509). Various administrative health databases were used to identify IDD status based on the International Classification of Disease (ICD)‐9 and ICD‐10 diagnosis codes (Lin et al. 2013; Ouellette‐Kuntz and Martin 2014). IDD status was determined based on whether the individual had one or more eligible diagnoses in hospital admission records, complex continuing care admission records, emergency department visit records, home care visit records or physician billing data (two or more visits). IDD status was identified at least 6 months before cancer diagnosis (Table S2). We restricted the study population to females due to the small number of males diagnosed with breast cancer in Canada each year (Canadian Cancer Society n.d.). Only adults diagnosed with Stage I–III breast cancer were included in the study, as treatment recommendations are heterogeneous for unknown and metastatic stages at diagnosis (Gradishar et al. 2017). The study cohort was followed until 31 December 2020.

2.4
Cohorts and Outcome Definitions
We examined four treatment outcomes informed by guidelines in three distinct cohorts, described below.

Cohort 1.

This cohort included the whole study population (adults with Stage I–III breast cancer). The primary outcome was overall breast cancer treatment; defined based on international and national breast cancer treatment guidelines (Gradishar et al. 2017; Balic et al. 2019). For those with Stage I–III breast cancer, breast cancer treatment based on guideline recommendations included all of those who had received a mastectomy or breast‐conserving surgery and radiation. Given the variability in systemic therapy recommendations based on patient and disease‐related factors not easily available within the administrative dataset, systemic therapy was not included in the definition of guideline concordance for those with Stage I and II disease. For individuals with Stage III breast cancer, overall treatment included individuals who had received surgical resection (either breast‐conserving surgery or mastectomy), adjuvant chemotherapy and radiation. Only adjuvant chemotherapy was considered based on treatment guidelines available during the study time frame (Korde et al. 2021). We also examined receipt of surgical resection (either mastectomy or breast‐conserving surgery) as a secondary outcome of interest among all those with Stage I–III breast cancer.

Cohort 2.

The second cohort included the study population (Stage I–III) who had received surgical resection. We created this cohort to examine the outcome of mastectomy receipt. Namely, including only those who received surgical resection allows us to directly compare those who had received mastectomy to those who received breast‐conserving surgery.

Cohort 3.

The third cohort included those with Stage I and II breast cancer who had received breast‐conserving surgery. This cohort was created to examine the receipt of radiation. This cohort did not include individuals who were Stages I and II and had received a mastectomy, as these individuals may not be eligible for receiving radiation. We did not include Stage III breast cancer patients in this cohort as we hypothesized that individuals with Stage I/II and Stage III could differ in terms of their cancer, requiring different treatment pathways, in particular, the requirement of chemotherapy prior to radiation for Stage III patients. This particular outcome was not considered among individuals with Stage III breast cancer due to small sample sizes.

2.4.1
Outcome Measurement
We identified receipt of treatment from OHIP and ALR databases. Breast‐conserving surgery and mastectomy were identified from the OHIP database from 30 days before to 365 days after the date of diagnosis. Receipt of chemotherapy was identified from OHIP billings, including at least one receipt of chemotherapy within 180 days of the first surgery date. Neoadjuvant chemotherapy was not considered an outcome of interest due to the study time frame, in which the recommendation of neoadjuvant chemotherapy was primarily recommended following the termination of the study (Korde et al. 2021). We determined radiation receipt from the ALR within 365 days of the first breast‐conserving surgery (Stages I and II) or first chemotherapy (Stage III) date. A summary of these treatment outcome definitions and data sources is provided in Table S3.

2.4.2
Covariates
2.4.2.1
Sociodemographic Variables
We examined age at diagnosis, region of residence, rurality, neighbourhood income and age and labour force measure. These variables were selected as they are associated with treatment receipt in the general population of cancer patients and typically are important social determinants of health in cancer equity research (Neuner et al. 2020; Dreyer et al. 2017; Lemasters et al. 2018; Lambert et al. 2023). Age was captured in the RPDB. Region of residence was categorized based on five transitional health regions in Ontario, including East, West, North, Central and Toronto (Government of Ontario n.d.). Rurality was identified from the postal codes, wherein rural is indicated when the community size is ≤ 10 000 (rural, urban, unknown) (Rural and Small Town Canada Analysis Bulletin 2001). We used neighbourhood income categorized into quintiles as a proxy for individual income, and this was assigned by linking neighbourhood‐level income data from the census with postal codes (1 = lowest income quintile; 5 = highest income quintile). We approximated marginalization at the individual‐level using the age and labour force measure from the Ontario Marginalization Index (Public Health Ontario 2018; Matheson et al. 2012). A region that has been assigned the highest marginalized score for the age and labour force measure will have a higher number of seniors, a lower percentage of persons who are part of the labour force, and a higher ratio of seniors and children to adults aged 15–64 than other areas in Ontario (Public Health Ontario 2018, Matheson et al. 2012). As such, age and labour force quintiles identify neighbourhoods with high proportions of individuals who do not work due to disability (Van Ingen and Matheson 2022) and could possibly serve as a proxy measure for individuals with more severe or profound IDD. We excluded individuals with incomplete demographic data (n < 6).

2.4.2.2
Clinical Variables
We measured comorbidities using the Aggregated Diagnostic Groups (ADGs), which capture hospitalizations, emergency department visits and physician visits 12 months before each individual's cancer diagnosis (Johns Hopkins University 2014). These comorbidities were further classified as major (time‐limiting and unstable, such as diabetes mellitus) and minor (not time‐limiting, acute and low severity, such as allergies). Cancer diagnoses were not included. ADG major comorbidities were grouped into 0, 1 and 2 or more diagnoses. ADG minor comorbidities were categorized as 0–2, 3–4, 5–6 and 7 or more diagnoses.

2.4.2.3
Cancer‐Related Variables
The cancer‐specific variables we captured were stage at diagnosis (I, II, III), year of diagnosis (2007–2012; 2013–2018), having a previous cancer diagnosis (non‐breast cancer diagnosis before breast cancer diagnosis), cancer biomarker status (oestrogen receptor, ER; progesterone receptor, PR; human epidermal growth factor receptor 2, HER2; yes, no, unknown), cancer grade (poorly, moderately, highly differentiated, unknown), cancer tumour size (0.1 to < 20 mm, 20 to < 50 mm, ≥ 50 mm, unknown), lymph node status (yes, no, unknown) and death within 12 months of cancer diagnosis (yes, no). These variables relate to the extent and aggressiveness of the cancer and drive cancer treatment decisions (Gradishar et al. 2017; Balic et al. 2019). Information for identifying HER2 was identified from previous research (Holloway et al. 2018). A summary of how these variables were identified is provided in the Table S4.

2.4.2.4
Health System Factors
We identified consultations and family interviews from 180 days before the cancer diagnosis date to 365 days after (Table S5). Consultations with surgeons, radiation oncologists and medical oncologists, as well as family interviews, were identified from the physician billing records. Family interviews refer to when consultations are conducted between physicians and relatives of persons authorized to make healthcare decisions on someone's behalf based on the Health Care Consent Act (Government of Ontario 2023). This interview was identified from 180 days before the cancer diagnosis to 365 days after to increase the likelihood that the family interview was specific to the cancer diagnosis. We identified if individuals had a primary care physician visit up to 12 months before their cancer diagnosis.

2.5
Data Analysis
We present demographic and clinical descriptive summaries of the three cohorts. Factors associated with each receipt of treatment outcome were explored using modified Poisson regression models with robust error variance. Modified Poisson models with robust error variance are established approaches to estimating a relative risk for a dichotomous outcome when it is not rare (> 10%) (Rothman et al. 2021; Zhou 2004). Individuals were followed forward in time until they received the treatment, passed away, or the end of follow‐up. Relative risks (RR) with 95% confidence intervals (CI) were estimated. Separate adjusted models were estimated for each independent variable explored, adjusted for age group and stage at diagnosis. For example, the adjusted model with family interview as the independent variable included age group and stage for each outcome variable. Age and stage were selected as both factors affect and guide cancer treatment (Lemasters et al. 2017; Lemasters et al. 2018; Gradishar et al. 2017). Additional variables were not adjusted for due to sample sizes and based on preliminary bivariate analyses. Age groups 70–79 and 80+ were combined in the receipt of the radiation model due to smaller sample sizes. Analyses were conducted using SAS Software (version 9.4, SAS Institute, Cary, NC). Statistical significance was set to p < 0.05.

3
Results
Cohort 1, the overall treatment and surgical resection‐eligible cohort (Stages I–III), included 365 people with IDD (Figure 1). Cohort 2 consisted of Stage I–III breast cancer patients who had received surgical resection (n = 333). Lastly, Cohort 3 included 138 Stage I and II breast cancer patients who had received breast‐conserving surgery. Demographic and clinical variables at baseline are displayed in Table 1. Across the cohorts, the mean age at diagnosis was approximately 60 years of age, and people tended to live in areas with the lowest income. Comorbidities were common across groups, with all cohorts having 27.7%–29.7% of individuals experiencing two or more major comorbidities. Approximately 20% of individuals had unknown ER, PR, or HER2 status across all the cohorts.
3.1
Cohort 1
Factors associated with overall receipt of breast cancer treatment informed by guidelines are reported in Table 2. In the adjusted models, the parameter estimates for age, stage and lymph node status were significantly associated with receipt of overall breast cancer treatment. Older individuals were significantly less likely to receive treatment than those 60–69 years of age (RR = 0.55; 95% CI 0.33–0.93). The likelihood of receiving treatment was lower among those with Stage III cancer relative to those with Stage I cancer (RR = 0.26; 0.16–0.43). People with unknown lymph node status were 0.74 times as likely to receive overall treatment that aligns with guidelines (95% CI 0.63–0.88) compared to those with negative lymph node status.
Findings for the receipt of surgical resection analyses are presented in Table 3. In the adjusted models, age, tumour grade, lymph node status and having a radiation consult were significantly associated with receipt of surgical resection. After adjusting for stage, people who were 80+ were significantly less likely to receive surgery than those who were 60–69 years of age (RR = 0.67; 95% CI 0.52–0.86). Individuals with unknown tumour grade, unknown nodal status and receipt of radiation consult were significantly less likely to receive surgical resection, potentially reflecting that these disparities are occurring simultaneously if breast cancer patients with IDD are not being considered for treatment (unknown grade vs. well‐differentiated grade: RR = 0.89; 95% CI 0.81–0.97; unknown nodal status vs. no positive lymph nodes: RR = 0.83; 95% CI 0.76–0.91; no radiation consult vs. radiation consult: RR = 0.84; 95% CI 0.82–0.97).

3.2
Cohort 2
Among individuals who received a surgical resection, factors associated with receipt of mastectomy versus breast‐conserving surgery are presented in Table 4. After adjustment, age, stage, not having a radiation oncologist consult and having a family interview were significantly associated with receipt of mastectomy. Relative to those aged 60–69 years of age, all other age groups were significantly more likely to receive mastectomy than breast‐conserving surgery. Additionally, individuals with Stage III breast cancer were more likely to receive mastectomy than those who were Stage I (RR = 1.82; 1.46–2.27). Individuals who did not receive a radiation oncologist consult were 1.89 times more likely to receive mastectomy (95% CI 1.59–2.24) than those who did have a consult. People who did not receive a family interview were significantly less likely to receive a mastectomy than those who had a family interview (RR = 0.77; 95% CI 0.63–0.95).

3.3
Cohort 3
Receipt of radiation in Stage I and II breast cancer patients among those who had received breast‐conserving surgery is displayed in Table S6. After adjustment, only the parameter estimates for age and lymph node status were significantly associated with receipt of radiation. Relative to those who were 60–69 years old, those who were < 50 were 1.56 times as likely to receive radiation (95% CI 1.05–2.31). Lastly, individuals with unknown lymph node status were significantly less likely to also receive radiation, as were individuals with positive lymph nodes identified.

4
Discussion
This study explored factors associated with receipt of breast cancer treatment among a population of females with IDD. For most outcomes, older individuals and those with a greater extent of anatomic disease spread were less likely to receive guideline‐informed treatment. Additionally, those who did not have a family interview were significantly less likely to receive mastectomy than those who had such an interview.
Our findings align with research exploring how age affects receipt of care in the general breast cancer population. Older age is consistently associated with lower rates of guideline‐recommended care, as well as receipt of mastectomy rather than breast‐conserving surgery in the general population (Lemasters et al. 2018; Lemasters et al. 2017; Punglia et al. 2008; Kimmick et al. 2014; Minicozzi et al. 2019). Similarly, in the general population, clinical factors, including stage, tumour type, lymph node status and receptor status, are directly related to receipt of care (Punglia et al. 2008). In the current study, adults with more advanced stages of cancer were less likely to receive overall treatment and more likely to receive mastectomy rather than breast‐conserving surgery. More research is needed to understand differences in nodal status and receipt of radiation treatment among individuals with IDD, particularly given the small sample size for the radiation cohort.
Some of our findings did not align with previous research exploring risk factors for breast cancer treatment in the general population, including comorbidities and SES. Our findings are interpreted cautiously due to the small sample size and limited statistical power. In the general population, comorbidities are well‐documented as contributing to lower cancer treatment rates (Safarti et al. 2014; Houterman et al. 2004; Minicozzi et al. 2019). Minicozzi et al. (2019) reported that among the general population of breast cancer patients, women who received mastectomy were more likely to have comorbidities than patients who received breast‐conserving surgery (Minicozzi et al. 2019). In the current study, we did not find that comorbidities significantly influenced receipt of care in adjusted models; however, it is possible that there was simply not sufficient power to detect an association, rather than there being no actual effect present. Similarly, neighbourhood income quintile was not significantly associated with treatment outcomes among people with IDD, which contrasts with research in the general population (Dreyer et al. 2017). In addition to challenges with power, it is important to consider how neighbourhood income quintiles are defined. In the general population, there could be misclassification as aggregate levels may not agree with individual‐level income (Davis et al. 2023). This issue may be further amplified when considering individuals with IDD, who sometimes live with family or in group homes and for whom many may receive similar income through the provincial and federal disability income support programs (Residential Information Systems Project [RISP] Annual Survey of State Developmental Disabilities Agencies [RISP] n.d.).
Receipt of mastectomy rather than breast‐conserving surgery was more likely among individuals who had a family interview after adjusting for age and stage at diagnosis. Having a family interview indicates that a family member had spoken with a healthcare professional as part of the treatment consent process, based on the Health Care Consent Act, potentially serving as a proxy for more complex limitations in communication or understanding for the patient. While some oncologists could potentially meet with people with IDD and their families and not bill this specific code, those who have this particular code may have more complex needs and require this billing legally based on the Health Care Consent Act. Providing mastectomy rather than breast‐conserving surgery and radiation may be a strategy to circumvent providing radiation to complex patients while still providing guideline‐recommended care (Gradishar et al. 2017). However, this is not the case for people with Stage III disease, who require radiation regardless of surgery type. Barriers to cancer care identified among adults with IDD include difficulties with communication, understanding and obtaining consent (Stirling et al. 2021; Boonman et al. 2022; Tuffrey‐Winje et al. 2010; Tuffrey‐Winje et al. 2006). In addition to communication challenges, there may also be other factors that could contribute to decisions regarding avoiding adjuvant therapy by providing mastectomy, including challenges with coordinating care, repeat visits to the hospital for adjuvant treatments, as well as needing to remain still during adjuvant treatment (Delany et al. 2023). Multiple studies have also reported the potential for the gatekeeping role of caregivers when attempting to protect individuals with IDD from harm (Stirling et al. 2021; Tuffrey‐Winje et al. 2006; Tuffrey‐Winje et al. 2013). These barriers may be exacerbated when healthcare professionals lack training in working with people with IDD, providing patient‐centred care approaches to cancer treatment, and understanding of stigma and biases prevalent in healthcare settings (Flynn et al. 2016). While these findings are exploratory, they suggest that factors beyond the individual's health and cancer may contribute to differences in receipt of treatment among people with IDD.
Ableism, namely the discrimination of people with disabilities (Goodley 2014), is prevalent in healthcare settings (Janz 2019). Ableism can arise internally (among the individual themself), interpersonally (during their day‐to‐day interactions‐ e.g., with family members, etc.), or at the institutional level (e.g., assumptions in medical settings about the ability to communicate and quality of life) (Mannor and Needham 2024). Assumptions may be about what someone with disabilities would prefer in regard to their treatment, what this patient group can handle, or attitudes regarding what is considered a good quality of life (Cavallo 2018). There could also be assumptions made regarding whether people with IDD would like to keep their breasts, as this is a group that is routinely infantilized and asexualized, with assumptions potentially made regarding their adult status or conceptualizations of gender and sexuality, respectively (Martino 2022a, 2022b). Some healthcare professionals may feel unprepared to support cancer patients with IDD effectively (Flynn et al. 2015). Improving training and education for healthcare professionals and caregivers in using patient‐centred approaches would be beneficial in supporting self‐determination among cancer patients with IDD (Flynn et al. 2016; Stirling et al. 2021; Dibble et al. 2024). Resources for supporting cancer patients with IDD and their caregivers or family members may also be a useful future direction for improving cancer care. Such resources could include visual and or written guides for breast cancer treatment that specifically involve people with lived experience in their creation, in addition to the involvement of more patient navigators trained in supporting individuals with IDD (Chen et al. 2024).
This study explored factors associated with receipt of breast cancer treatment aligning with treatment guidelines among individuals with IDD using high‐quality population‐based health data. While these outcomes are not specifically defined as guideline recommended as they were not defined based on subgroup typing or nodal status, they reflect a population within whom guidelines are clear regarding how treatment should proceed. Overall, this evidence generates important preliminary findings regarding how clinical and non‐clinical factors affect breast cancer treatment among adults with IDD. However, limitations emerged in this study. First, the sample sizes were small, which resulted in challenges with power and limited interpretation. Such low power could mean that where we did not detect an association, a true association may still exist. Future studies spanning longer periods or involving other regions may increase the cohort size to address this limitation. Information on other key considerations for cancer treatment among adults with IDD, such as the underlying condition causing disability, the severity of the disability and accessibility of supportive social resources, was not available in the administrative health databases. We had initially planned to explore the scales available in the interRAI‐ID in the home care database, including measures relating to communication, adaptive daily living skills and living settings (Martin et al. 2007). However, only 10% of individuals had this assessment completed, and as such, this data could not be explored. Therefore, these findings represent the overall estimate of treatment receipt for those with IDD, and the effect could differ depending on each individual's type of disability experience, for example if they require more or less adaptive supports, or their communication modalities. Future research should separately study cancer outcomes for different health conditions included under the umbrella of IDD diagnoses, where sample sizes allow. We also used a specific type of comorbidity score that is based on previous healthcare encounters (Johns Hopkins University 2014). If someone is not able to access such care, they may be incorrectly reported as having no comorbidities. Given that the mean age of the study population is above 50 years of age, it is likely that most individuals would have had an encounter with the healthcare system, but the potential for misclassification should be noted. Lastly, race has been found to be associated with lower levels of guideline‐recommended breast cancer treatment in American studies (Markey et al. 2022; Yedjou et al. 2019); however, race is not available in population‐based administrative data in Ontario. Further exploration of the intersection between IDD status and race and its effect on breast cancer treatment is needed.
This study provides preliminary evidence that beyond age and stage, breast cancer treatment among people with IDD may be influenced by several factors. Given that people with IDD are less likely to receive most types of breast cancer treatment and more likely to die following a breast cancer diagnosis, further emphasis should be placed on exploring factors contributing to treatment decision‐making, including patient‐level factors such as communication and understanding and system‐level factors such as training and education of healthcare professionals and availability of staff to help cancer patients with IDD navigate the cancer system. Working with self‐advocates, family advocates and healthcare professionals supporting individuals with IDD is needed to improve patient‐centred care for people with IDD diagnosed with breast cancer.

Funding
This study was supported by the Canadian Institutes of Health Research (CIHR), grant # 162130.

Ethics Statement
The study received ethical clearance from the Queen's Faculty of Health Sciences and Affiliated Hospitals Research Ethics Board for review (HSREB = 6038034).

Conflicts of Interest
The authors declare no conflicts of interest.

Supporting information

Table S1: Data sources.

Table S2: Algorithm for identifying IDD status. Please note that there is derogatory historical language listed below.

Table S3: Summary of treatment outcomes.

Table S4: Summary of clinical covariates.

Table S5: Data sources, codes and descriptions for identifying consultations and family interviews.

Table S6: Effect estimates for guideline‐recommended radiation Stages I and II.