Variations in Emergency Service Utilization among Cancer Survivors: Results from the Pan-Canadian Experiences of Cancer Patients in Transition Study Survey

Purpose The objective of this study was to examine variations in emergency service utilization (ESU) among cancer survivors during the first year after completing primary cancer treatment. Methods In 2016, the Canadian Partnership Against Cancer collected survey responses from cancer survivors across Canada about self-reported ESU after completing primary cancer treatment. We included survey respondents diagnosed with nonmetastatic breast, hematologic, colorectal, melanoma, or prostate cancer. Multivariable, multinomial logistic regression analysis was used to examine factors associated with cancer survivors' ESU. Results Of the 5,774 cancer survivors included in our analysis, 22% reported ESU during the first year after completing their primary cancer treatment, 16% reported ESU one to three times, and 6% reported ESU more than three times. Factors significantly associated with frequent ESU included younger age, colorectal and hematologic cancers, more frequent primary care provider and oncology specialist visits, single or retired status, lower income, and self-reported lower quality of life. Conclusion Our study identified factors associated with more frequent ESU among cancer survivors in the first year after completing primary cancer treatment. These factors highlight differences in cancer survivors' demographics, their ability to access and need for healthcare services, and the complexity of using ESU as a metric for quality improvement in survivorship care. These variations must be considered in quality improvement initiatives.


Background
In countries such as Canada, the cancer incidence is estimated to increase by 32% between 2020 and 2040, with more than 60% of adults now living fve years or more [1][2][3][4]. As a result, there are more cancer survivors than ever before. Te growing number of cancer survivors results in greater demand for healthcare that meets their ongoing, specialized needs [1,4].
Cancer survivors have unique healthcare needs due to multiple factors. Tese needs vary depending on individual risk of cancer recurrence, development of new and secondary cancers, ongoing physical and psychosocial efects of treatment, and noncancer-related comorbidities [5].
Research has shown that healthcare for cancer survivors frequently does not meet their needs, owing partly to a lack of coordination, insufcient resources, workforce shortages, and rising healthcare expenses [5,6]. Improving the quality of survivorship care is a priority [6,7].
Quality metrics have been proposed to measure, evaluate, and improve the quality of survivorship care. Healthcare utilization is one such metric [5,8]. Healthcare utilization represents the interrelated complexities of the availability and accessibility of needed healthcare services by patients [9]. Ideally, optimized healthcare utilization would result in improved health, quality of life, and patient experience through efcient use of resources. Certain measures of healthcare utilization can reveal areas of fragmentation and preventable inefciency. In this regard, emergency service utilization (ESU) is a healthcare utilization metric proposed to measure the quality of survivorship care, as it has been associated with high resource utilization that does not always result in improved outcomes [10]. However, studies assessing ESU in cancer survivors and strategies to improve it are lacking.
In 2016, the Canadian Partnership Against Cancer (CPAC) developed and distributed the Experiences of Cancer Patients inTransition Study survey, herein referred to as the Transition Study, to over 40,000 randomly selected cancer survivors to understand experiences with follow-up cancer care one to three years after completing primary treatment. Te primary objective of this study was to understand the factors associated with cancer survivors' selfreported ESU during the frst year after completing primary cancer treatment. Te secondary objective was to assess ESU, primary care provider (PCP), and oncology specialist utilization three years after completing primary cancer treatment.

Methods
Tis article was reported in accordance with the Checklist for Reporting of Survey Studies (CROSS) [11].

Survey Development.
Fitch et al. previously described the Transition Study survey development and dissemination in detail [12]. In summary, CPAC worked with all ten provincial cancer agencies across Canada to develop the survey guided by stakeholders. Te survey was piloted with cancer survivors. Te fnal Transition Study survey included closedand open-ended questions, was available in both ofcial languages of Canada (French and English), took 30 to 45 minutes to complete, and was available on paper or online. A copy of the full survey is publicly available on the CPAC system performance website [13].

Survey
Dissemination. CPAC collaborated with the ten provincial cancer agencies across Canada to disseminate the survey. Te eligibility criteria included adult survivors over 30 years of age with one of the following nonmetastatic cancer types: breast, prostate, colorectal, and melanoma, or selected metastatic hematological cancers (i.e., Hodgkin's lymphoma, difuse B-cell lymphoma, acute myelogenous leukemia, and acute lymphocytic lymphocyte leukemia). Adolescent and young adult (AYA) survivors between 18 and 29 years old with any nonmetastatic cancer types or metastatic testicular cancer were also eligible. Survivors were defned in the Transition Study as those who completed primary cancer treatment.
Each province obtained ethics and privacy approval before data collection. Informed consent was obtained from respondents prior to engaging with the survey. Between June and October 2016, 40,790 survey packages were mailed out. Te number of surveys for distribution was calculated for the adult sample such that 95% confdence intervals would have a width of no more than ±5% by disease sites and provinces for a percentage assumed to be 50% and an assumed response rate of 30%. Estimates of the number of eligible survivors per province were based on national disease site prevalence and incidence by province. As smaller provinces were unlikely to have enough survivors within disease sites to achieve the desired confdence interval precision, all eligible survivors were contacted. A random sample within the cancer disease site was chosen for larger provinces and disease sites where the number of eligible survivors was anticipated to be greater than the required number. Te fnal sample was sufcient to achieve precision of at least ±3% by disease site for all provinces combined.

Study Population.
Te subset of respondents we included was diagnosed with nonmetastatic breast, hematologic, colorectal, melanoma, or prostate cancer. We excluded respondents who did not complete the survey themselves, or whose age, sex, or year of diagnosis were unknown, or whose last treatment date was unknown or more than fve years ago. We also excluded respondents for whom the primary outcome was unavailable (i.e., those who, when asked how many times they used emergency services in the frst year after cancer treatment, responded that they did not know and could not remember or those services did not apply to them).

Outcome.
In a previous study, we examined the infuence of individual and structural factors on patients' experience with survivorship care among respondents who completed the Transition Study [14]. Te primary outcome of this study was cancer survivors' self-reported ESU in the frst year after completing cancer treatment, defned using responses to one of the subquestions from question 23 of the Transition Study survey (Supplementary Material Table 1). Tis subquestion asked respondents how often they visited an emergency room or urgent care center (herein referred to as emergency services) in the frst year after primary cancer treatment.
Te secondary outcome was emergency service, primary care provider (PCP), and oncology specialist utilization within three years after completing primary cancer treatment. Tis information was obtained from other subquestions from question 23 (Supplementary Material  Table 1). PCP was defned as a family doctor, general

Results
A total of 13,319 respondents returned the survey (response rate 33%). Based on our defned inclusion criteria, 5,774 unique respondents were part of this analysis. We excluded the following respondents: those who did not have nonmetastatic breast, hematologic, colorectal, melanoma, or prostate cancer (n � 3,904), those who had primary outcomes missing (n � 3,170), those who did not complete the survey themselves (n � 237), whose age or sex was unknown (n � 34), whose year of diagnosis was unknown (n � 35), or whose last treatment date was unknown or more than fve years ago (n � 165).

Personal and Clinical Characteristics of the Included Study
Population. Most respondents in our analysis were married or partnered (76%), found it easy or very easy to ask their healthcare provider questions about their follow-up cancer care-related concerns (75%), were born in Canada (85%), spoke English most commonly at home (78%), rated their overall quality of life as good or very good (86%), rated their physical health (78%) and emotional health (78%) as good or very good, received cancer treatment that included surgery (71%) and/or systemic therapy (80%), and were not enrolled in a clinical trial (86%) ( Table 1).

ESU in the First Year after Completing Cancer Treatment.
Most respondents (78%) reported not using emergency services during the frst year after cancer treatment; 16% reported using them one to three times, and 6% reported using them more than three times ( Table 2).

Factors Associated with ESU in the Year following Cancer
Treatment. We identifed statistically signifcant factors associated with increased ESU across all levels of the multivariable, multinomial regression analysis (Tables 2 and 3). Tese included younger age, more frequent PCP and oncology specialist utilization, colorectal and hematologic cancers, single or retired status, speaking French most commonly at home, lower income, self-reported lower quality of life, and access to a copy of their medical records (Table 3).
Some characteristics were signifcantly more likely to be associated with ESU one to three times but not more than three times, such as living in a rural area, lower self-reported emotional health, receiving surgery as part of their cancer treatment, and more recent completion of cancer treatment at the time of survey completion (i.e., within the last year versus one to three years ago versus three to fve years ago versus no treatment) ( Table 3).
Some characteristics were signifcantly more likely to be associated with ESU more than three times but not one to three times, such as having an unmet practical need, more comorbidities, and clinical trial enrollment (Table 3).

Emergency Service, PCP, and Oncology Specialist Utilization in the Second and Tird Years after Completing Primary
Cancer Treatment. Compared to ESU in the frst year (78%), more respondents reported not using emergency services in the second (92%) and third (94%) years after completing cancer treatment.  Journal of Oncology

Journal of Oncology
Emergency service, PCP, and oncology specialist utilization in the second and third years after completing cancer treatment by frequency of ESU in the frst year is displayed in Table 4. Respondents who reported less ESU in the frst year after cancer treatment were also less likely to report emergency service, PCP, and oncology specialist utilization in the second and third years.

Discussion
Our study identifed several factors associated with more frequent ESU among cancer survivors in the frst year after completing primary cancer treatment, including younger age, colorectal and hematologic cancers, more frequent PCP and oncology specialist visits, single or retired status, lower income, and self-reported lower quality of life. Tese variations in ESU suggest diferences in cancer survivors' individual needs and ability to access healthcare services. Tese diferences are important to consider when using ESU as a quality metric or to plan initiatives aimed at improving survivorship care.
Younger patients reported higher ESU in our study. Previous studies have shown that younger age at cancer diagnosis is associated with more side efects and disability during and after treatment [19][20][21][22]. In the Transition Study, 80 to 90% of adolescent and young adult (AYA) cancer         survivors, i.e., between 18 and 29 years old, reported physical, emotional, or practical concerns and 85% reported experiencing challenges related to reduced physical capacity, pain, difculty coping distress, and a struggle to return to normal function [19][20][21]. In addition, they experienced difculty accessing nonemergent healthcare services and encountered challenges in obtaining assistance with side efects from their cancer treatment [19][20][21][22]. Te higher needs in this age group, which are often less well met compared to other groups, could be one reason for increased ESU, along with diferences in treatment regimens with unique side efect profles, social functioning, and caregiver support in AYA cancer survivors. Tis population has been historically underrepresented in cancer care quality improvement projects, and gaps exist in our understanding of their needs and experiences [17,23]. Further research will help inform strategies to improve the quality of survivorship care for AYA cancer survivors.
Previous publications have suggested that increased ESU is associated with limited access to routine, outpatient healthcare services [24,25]. Interventions, such as expedited access to knowledgeable providers by phone (telehealth) or to next-day appointments, have been shown to reduce ESU [26]. Interestingly, we found that respondents with increased ESU also reported increased PCP and oncology specialist utilization. Tese fndings suggest that other factors may be driving ESU. For example, the high usage of emergent and nonemergent healthcare services observed in our study could represent care fragmentation in a subpopulation of patients [16]. Care fragmentation occurs when individuals visit multiple healthcare providers for the same reasons and is associated with repetitive investigations, increased healthcare costs, and more frequent ESU among cancer survivors [16]. Higher emergency service, PCP, and oncology specialist utilization may also refect the small portion of patients with the most acute care needs. Tese patients may beneft from more intensive longitudinal follow-ups [27]. Without information on the appropriateness and quality of care provided through emergency services, PCPs, and oncology specialists in the Transition Study, it is difcult to recommend ways to improve. A risk-stratifed approach to cancer survivors' follow-up cancer care, where survivors with more needs have frequent contact with personalized healthcare services, may help provide higher quality survivorship care.
Another key fnding in our study is that ESU difered signifcantly by cancer types. Patients with colorectal or hematological malignancies reported higher ESU than those with breast or prostate cancer or melanoma. In this regard, ESU is likely more clinically appropriate for individuals diagnosed with cancers with delayed treatment toxicities or more acute recurrence presentations, which are especially prevalent in the frst year following treatment completion, when complications and disease relapse rates are highest [4,28]. For example, patients with hematologic cancers may experience signifcant late efects of treatment, including higher needs for access to emergency services to treat recurrent infections or transfusions. Similarly, colorectal cancer patients may present with more acute symptoms related to major surgery or local relapses, such as bowel bleeding and obstruction, which require emergent workup and treatment. Future initiatives aiming to use ESU as a quality metric in survivorship care should consider this to avoid penalizing those who need access most.
Despite high variability in ESU, we see a decline in emergent and nonemergent healthcare utilization after the frst year. Tis fnding is consistent with the existing literature; long-term cancer survivors have diferent healthcare needs than new survivors, especially those within the frst year after treatment completion when recurrence rates and toxicities are highest [29]. Our study provides insights into ESU variations within the frst three years after completing cancer treatment, supporting the need to defne phases of survivorship across settings for meaningful comparison of quality metrics.
Even though Canada has public, universal healthcare coverage, we found that respondents with lower income, living in rural areas, single status, and lower quality of life used emergency services more frequently. Te most recent systematic review of cancer survivors' healthcare utilization published in 2012 similarly found that preventable hospitalization was higher among those with lower income, of minority race or ethnicity, and widowed or divorced [15,[30][31][32]. While universal healthcare reduces direct costs related to healthcare, it does not always alleviate indirect costs, such as those related to transport, childcare, and supportive care services [30][31][32]. In some instances, services are available to help with indirect costs, but these may not be equitably accessible. Tese fndings highlight the need for personalized survivorship care models to ensure a healthcare system that meets everyone's needs, regardless of difering social determinants of health.
Determining whether the ESU reported in our study represents an efcient use of healthcare resources requires an understanding of the factors infuencing ESU among oncology patients [33]. Research into these factors has shown that most ESU is for treatment-related side efects and complications, of which approximately 40% may be preventable [34,35]. Still, few studies have assessed why oncology patients decide to use emergency services compared to other healthcare services. Nguyen et al. identifed that oncology patients over the age of 70 in Canada who accessed emergency services during cancer treatment did so primarily because other cancer care options were unavailable or they had a life-threatening health condition [36]. Tus, some needs of cancer survivors driving ESU may be better met for certain patients through alternative healthcare services, if available. However, as seen in our study, factors contributing to ESU are varied, and the reasons behind ESU are complex.

Clinical Implications.
Quality in healthcare is a complex and interconnected concept that can be difcult to capture, yet measuring quality in healthcare is paramount to compare and improve it. Te existing defnitions and frameworks can be challenging to operationalize locally driven changes [5,37]. We demonstrate that using ESU as a quality metric alone, without a deeper understanding of contextual factors contributing to ESU, is unlikely to provide information for meaningful improvements in the quality of care [38].
Another important factor to consider when using quality metrics to guide quality improvement is how using diferent measures for the same concept can refect diferent aspects of healthcare quality. For example, a recent review identifed 14 unique ESU measures from 29 studies [39]. Te authors concluded that the number or frequency of use of emergency services was believed to measure accessibility, whereas the time between arrival to the emergency department and diagnosis or treatment was believed to measure timeliness [39]. A review of cancer survivors' healthcare utilization patterns identifed that it was not just the number of healthcare visits that were associated with the quality of healthcare but also the type of provider and the number of patients shared between providers [16]. In our study, we used the number of times emergency services were used over a 12-month period, which is hypothesized to refect differential access to inpatient, acute care versus outpatient, and routine care [39]. It is important that future research should aim to understand the dimensions of healthcare quality refected by ESU among cancer survivors across diferent contexts to be able to make meaningful changes with these results. Ultimately, while quality metrics are important to improve healthcare, how these metrics are defned, prioritized, and interpreted must be determined in the local context.

Study
Limitations. Limitations of our study are the selfreported survey design and limited data generalizability. Te accuracy of self-reported healthcare use varies by healthcare service types (e.g., routine healthcare, emergency services, and invasive testing), time elapsed since use, utilization frequency, and questionnaire design [40]. Memorable or highly emotional events, as is common in ESU, are more accurately recalled [40]. Despite these limitations, selfreported healthcare service utilization remains a commonly used data source in health service research. Validated self-reported healthcare use questionnaires have been developed but not yet adapted for cancer survivors, which should be done to improve the quality of future research on healthcare utilization among cancer survivors [41].
Our study fndings may not be generalized to all cancer survivors in Canada because weighting was not applied due to the confdential nature of the data and the use of an external vendor for survey dissemination. Furthermore, the response rate of the Transition Study survey was 33%, and we excluded additional 57% of respondents from our analysis. In our opinion, considering the length and detail of the Transition Study survey, this response rate represents a strength compared to other surveys of this magnitude. In addition, we presented data by comparing respondents included in versus excluded from our analysis to provide transparency about to whom our fndings may not apply.
We recommend caution in generalizing our fndings to other healthcare systems, particularly with diferent insurance or survivorship care models. Factors infuencing ESU are likely to difer in jurisdictions with predominately private insurance or more geographically accessible specialized oncology care. Current survivorship care is not standardized in Canada. While we controlled for some diferences in survivorship care (e.g., receipt of a care plan, a copy of medical records, and providers responsible for follow-up cancer care), some aspects remain unaccounted for, and the infuence of these aspects on ESU is unknown. Tese unaccounted aspects refect an ongoing challenge with cancer survivorship research. Future eforts should aim to standardize data collection on survivorship care, which is of particular importance given that one of the solutions to improving the quality of survivorship care is likely to involve leveraging personalized survivorship care models.
Other limitations of our study represent future directions for research. We could not provide insights into the appropriateness and outcomes of ESU among cancer survivors in our study. Due to these limitations, it is difcult to make recommendations to improve ESU among cancer survivors as we cannot determine if variations in use represent low-or high-quality care.

Conclusion
Tere are an increasing number of cancer survivors who require specialized survivorship care to improve their quality of life and long-term outcomes. Our study identifed several factors associated with more frequent ESU among cancer survivors in the frst year after completing primary cancer treatment. Tese variations in ESU highlight the diferences in cancer survivors' individual predisposition, ability to access, and need for survivorship care. Additional research is needed to understand if these diferences represent clinically appropriate care consistent with patient preferences or result from inequities. Tis foundational work is necessary before ESU can be used as a quality metric in survivorship care.

Data Availability
Te data that support the fndings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.