Patterns of cancer centre follow-up care for survivors of breast , colorectal , gynecologic , and prostate cancer

Background Rising demand on cancer system resources, alongside mounting evidence that demonstrates the safety and acceptability of primary care–led follow-up care, has resulted in some cancer centres discharging patients back to primary care after treatment. At the same time, the ways in which routine cancer follow-up care is provided across Canada continue to vary widely. The objectives of the present study were to investigate patterns of routine follow-up care at a cancer centre for breast, colorectal, gynecologic, and prostate cancer survivors; factors associated with receipt of follow-up care at a cancer centre; and changes in follow-up care at a cancer centre over time. Methods We identified all people diagnosed in Nova Scotia with an invasive breast, colorectal, gynecologic, or prostate cancer between 1 January 2006 and 31 December 2013. We linked the resulting population-based dataset, at the patient level, to cancer centre or clinic data and to census data. We identified a nonmetastatic survivor cohort (n = 12,267) and developed decision rules to differentiate routine from non-routine visits during the follow-up care period (commencing 1 year after diagnosis). Descriptive statistics were computed to describe the patterns of routine follow-up care at a cancer centre. Negative binomial regression was used to examine factors associated with visits made and changes over time. Results Nearly half the survivors (48.4%) had at least 1 follow-up visit to the cancer centre, with variation by disease site (range: 30.2%–62.4%). Disease site and stage at diagnosis were associated with receipt of follow-up care at a cancer centre. For instance, compared with breast cancer survivors, survivors of gynecologic cancer had more visits [incidence rate ratio (irr): 1.48; 95% confidence interval (ci): 1.34 to 1.64], and survivors of colorectal cancer had fewer visits (irr: 0.45; 95% ci: 0.40 to 0.51). Year of diagnosis was associated with follow-up at a cancer centre, with each successive calendar year being associated with an 8% increase in visits made (irr: 1.08; 95% ci: 1.07 to 1.10). Conclusions Despite evidence that follow-up care can be effectively and safely delivered in primary care, and despite intensifying demands on oncology services, many survivors continue to receive routine follow-up care at a cancer centre.


BACKGROUND
The U.S. Institute of Medicine defines cancer survivorship as the period from completion of primary treatment to cancer recurrence or death 1 .The prevalence of cancer survivors is rapidly growing 1 : Considering all cancer types, two thirds of individuals diagnosed with cancer today will be long-term survivors, meaning that they will live for 5 years or more after diagnosis.For those patients, routine follow-up care marks the transition from intensive treatment to long-term sur vivorship 2 .As long-term survivors, those individuals require routine follow-up to detect disease recurrence or new primary cancers, to prevent and manage the late effects of the cancer and its treatment, to manage pre-existing chronic conditions, and to address general preventive care needs.Consequently, the follow-up of patients after their cancer treatment has become an established component of medical care 1 .
Although the rising prevalence of cancer survivors represents tremendous success in early detection and Current Oncology, Vol. 24, No. 6, December 2017 © 2017 Multimed Inc.
treatment, it raises important questions about how best to provide the follow-up care that survivors require after intensive treatment.Survivors have traditionally received their follow-up care in specialist cancer centres and clinics 3,4 , but a rising cancer incidence, more complex adjuvant therapies, and constrained cancer system resources 5,6 have led many cancer centres to re-examine their priorities in terms of service delivery.That reality, together with a mounting evidence base showing that primary care-led follow-up is a safe and acceptable alternative to oncology-led care [7][8][9] , has resulted in some cancer centres transfering routine follow-up care back to a patient's primary care provider (pcp) after treatment 10 .
Despite the foregoing context, the ways in which routine cancer follow-up care is provided across Canada continue to vary widely [10][11][12][13] .In Nova Scotia, numerous attempts have been made since the early 2000s to discharge survivors at low risk of recurrence after treatment from their cancer centre to their pcp.The results of those attempts are unknown.The objectives of the present study were therefore to investigate patterns of routine follow-up care at a cancer centre for breast, colorectal, gynecologic, and prostate cancer survivors; factors associated with receipt of follow-up care at a cancer centre; and changes in cancer centre follow-up care over time.

METHODS
This retrospective cohort study used linked administrative health data to examine routine follow-up visits to specialist cancer centres and clinics for all persons diagnosed in Nova Scotia with nonmetastatic breast, prostate, colorectal, or gynecologic cancer over an 8-year period.Ethics approval to conduct the study was granted by the Nova Scotia Health Authority's research ethics board.

Cohort
Using the Nova Scotia Cancer Registry (nscr), we created a population-based dataset of all people in Nova Scotia diagnosed with an invasive breast, prostate, colorectal, or g ynecologic cancer between 1 Januar y 2006 and 31 December 2013 (n = 20,901).International Classification of Diseases for Oncology (3rd edition) diagnosis codes 14 were used to identify the disease sites of interest.We linked the resulting dataset, at the patient level, to the Oncology Patient Information System (opis) and to Canadian census data using Statistics Canada's Postal Code Conversion File.The opis provides complete visit and treatment data for all cancer centres and satellite cancer clinics in the province (hereafter referred to as cancer centres).The Postal Code Conversion File is a program generated by Statistics Canada to convert postal codes to census geographic areas.
Next, we established a survivor cohort that included only patients who actually visited a cancer centre for consultation or treatment (that is, they had at least 1 visit at a cancer centre), given that the focus of the study was to examine routine follow-up visits occurring at the cancer centres.Given the nature of Nova Scotia's cancer system, individuals not visiting a cancer centre during their cancer journey (for example, a patient receiving surgical excision only, without a medical or radiation oncologist consultation) would not visit a cancer centre for routine follow-up care after treatment.To develop the survivor cohort, we excluded individuals who died within 1 year of diagnosis and those with stage iv disease at diagnosis.

Data Sources and Variables
From the nscr, we acquired patient demographics (date of birth, sex, postal code at diagnosis), date of diagnosis, tumour characteristics (histology, collaborative stage at diagnosis, site-specific collaborative stage factors), prior cancer diagnoses, and date of death.From the opis, we acquired data related to cancer centre visits (date of visit, appointment type, visit status, visit duration, visit location, provider type, encrypted physician id, centre location), resource use, and recurrence.
The resource use variable provided information about the use of certain equipment (for example, radiotherapy equipment) or service units (for example, chemotherapy units).We acquired complete opis data for the period 1 January 2006 to 31 December 2015.
Because no specific appointment type in opis identifies follow-up care, we developed a decision rule to differentiate routine follow-up visits from non-routine visits during the follow-up care period (for example, visits to manage complex late effects or for suspicion of recurrence).The decision rule used combinations of variables from the nscr (diagnosis and date of diagnosis) and the opis (appointment type, resource use, and recurrence).Diagnosis and recurrence variables were used to define non-routine visits-specifically, visits associated with patients who had evidence of recurrence or a new primary cancer within 1 year of diagnosis, and patients with a prior cancer diagnosis.Regarding the latter patients, it is plausible that follow-up care for individuals with a prior cancer diagnosis differs in substantive ways (for example, in surveillance protocols) from follow-up care for individuals with a single cancer diagnosis.The iterative development process for the decision rule required multiple consultations with cancer centre staff involved in visit scheduling processes.

Analysis
The follow-up care period was defined as starting 1 year after the diagnosis date in the nscr, and continuing until the individual was censored or the end of the study period was reached.In other words, no visits were considered to represent routine follow-up care until 1 year after diagnosis.Our definition is consistent with definitions in earlier studies that used administrative data to examine cancer follow-up practices 12,13,15,16 , where the follow-up period commenced 1 year after diagnosis to exclude tests and visits associated with postoperative complications, routine postoperative v isits, and adjuvant therapy.Survivors were censored on evidence of a new primary cancer, cancer recurrence, or death.In addition to a discrete variable available in the nscr that indicated cancer recurrence or a new primary cancer, receipt of therapy (that is, chemotherapy or radiotherapy) in the follow-up period was also presumed to be indicative of cancer recurrence.
Descriptive statistics were computed to describe the patterns of follow-up care by year and by patient, tumour, and physician characteristics.Multivariate regression was Current Oncology, Vol. 24, No. 6, December 2017 © 2017 Multimed Inc.
used to test for changes in cancer centre follow-up visits over time, adjusting for other covariates.Specifically, negative binomial regression was used to model the number of visits for each individual where the key analysis variable was year of diagnosis.Because the time period available for follow-up visits was shorter for individuals diagnosed later in the study period, an offset technique was used to normalize visits to exposure time (that is, the time from diagnosis until censoring or the end of the study).Regression coefficients were converted to incidence rate ratios (irrs) that, in this context, can be interpreted as the change in the number of follow-up visits per period of exposure.Using that approach, results are relative to a single previous year.If the number of individuals who visited the cancer centre increased or decreased, and yet the distribution of visits per period remained the same for each diagnosis year, results would remain constant.Thus, the approach we used helps to control for changes in incidence over time.For example, if the number of cases doubled, and if all new patients had precisely the same number of visits, and if their characteristics were the same as those in the original population, the percentage increase per year would remain the same.
The model was run with these covariates: year of diagnosis, age at diagnosis (in years), sex, neighborhood median household income, disease site, stage at diagnosis (i, ii, iii, unknown), distance to closest cancer centre (in kilometres), rural or urban residence, and location of first cancer centre visit.Using postal code data at last visit, median household income was obtained from 2011 Canadian census data.Rural and urban status were determined based on postal code at last visit, with a second digit of 0 indicating rural residence.To help interpret trends in follow-up visits over time, we also applied our model to the total number of visits, given that all visits might also be increasing over time because of a rising cancer incidence and more complex therapies.

RESULTS
The final cohort included 12,267 cancer survivors.Table i presents the demographics for survivors with treatment visits only (n = 6332), survivors who had at least 1 routine follow-up visit (n = 5935), and survivors who had 5 or more routine follow-up visits (n = 2358).The latter two groups are not mutually exclusive; survivors included in the "5 or more visits" group are also included in the "at least 1 visit" group.As shown, the relative proportions for each subgroup change with the disease site.For example, colorectal cancer survivors constituted 30.2% of those receiving treatment visits only, but only 14.0% and 12.0% respectively of those receiving at least 1 and 5 or more follow-up visits.Conversely, prostate cancer survivors constituted 16.2% of those receiving treatment visits only, but 28.6% and 23.0% respectively of those receiving at least 1 and 5 or more follow-up visits.
Table ii depicts the number of follow-up visits received by the study population.As shown, 48.4% of all survivors had at least 1 routine follow-up visit at a cancer centre.The proportion varied by disease site, with 50.1%, 30.2%, 47.2%, and 62.4% of breast, colorectal, gynecologic, and prostate cancer survivors respectively having at least 1 follow-up visit at a cancer centre.Overall, 19.2% of the survivors had 5 or more routine follow-up visits at a cancer centre, with that proportion also varying by disease site (21.2%, 10.3%, 26.8%, and 19.9% of breast, colorectal, gynecologic, and prostate cancer survivors respectively).
The pattern of routine follow-up care volume varied between oncologists.Specifically, 6 oncologists handled 8.4% of the treatment visits to the study population; those same 6 physicians provided 34.7% of the cancer centre follow-up care to the study population.An alternative measure of variability between individual physicians is the distribution of treatment visits compared with followup visits.For example, more than 30% of the visits to 13 individual physicians were for routine follow-up care.
Table iii presents factors associated with follow-up visits at a cancer centre.Compared with breast cancer survivors, survivors of a gynecologic cancer had 48% more visits (irr: 1.48; 95% ci: 1.34 to 1.64), and survivors of a colorectal cancer had 55% fewer visits (irr: 0.45; 95% ci: 0.40 to 0.51).Compared with survivors having a stage i Regression analysis also demonstrated that year of diagnosis was associated with having follow-up visits at the cancer centre.Specifically, each successive diagnosis (calendar) year was associated with an 8% increase in follow-up visits (irr: 1.08; 95% ci: 1.07 to 1.10), meaning that survivors diagnosed later in the study period were expected to have 8% more follow-up visits per exposure period.When the model was applied to the total number of cancer centre visits, the expected number of all visits also increased over time, but at a rate of 14% per year (irr: 1.14; 95% ci: 1.13 to 1.15; complete model data not shown).

DISCUSSION
In the present study, we investigated patterns of routine follow-up visits received by breast, colorectal, gynecologic, and prostate cancer survivors at cancer centres in Nova Scotia.We found that nearly half the survivors who received treatment at one of the province's cancer centres (48.4%) had at least 1 routine follow-up visit after treatment, and that delivery of routine follow-up care varied between oncologists.We also found that the number of follow-up visits at a cancer centre increased for each successive year during the study period (2006-2014).These data represent the first examination of routine follow-up care for prostate and gynecologic cancer survivors in Canada.
Despite focusing only on follow-up care at cancer centres, our findings concur broadly with those reported from other studies that reveal variable patterns of practice for follow-up care across Canada 12,13,17 .For example, Grunfeld and colleagues 13 found that most breast cancer survivors diagnosed in Ontario between 1998 and 1999 received follow-up care from oncologists, either solely or in combination with care from pcps.A more recent investigation examining physician visits in four Canadian provinces (British Columbia, Manitoba, Ontario, Nova Scotia) for breast cancer survivors diagnosed between 2007 and 2010 or 2011, depending on province 18 , revealed that 65% -93% of women had visits to both pcps and oncology specialists during the follow-up period (1-5 years after diagnosis).Although those findings are not directly comparable with ours, the lower proportion of breast cancer survivors receiving specialist follow-up in our study might relate to how researchers compute follow-up visits.In the present study, we used a decision rule to distinguish between routine and non-routine follow-up care.Recent population-based U.S. studies have revealed not only considerable variation in the frequency of oncologist follow-up visits for survivors of breast cancer 19 , but also in the types of providers who participate in breast cancer follow-up care 20 .For example, although 80% of survivors received follow-up care with a medical oncologist, 60% received follow-up visits with other oncologists (surgical or radiation), and 94% regularly saw their pcp during the period of follow-up care 20 .
Interestingly, the findings related to colorectal cancer survivors differ substantively from findings in our prior research in Nova Scotia.In a population-based cohort of colorectal cancer survivors diagnosed between 2001 and 2005, 58% of survivors continued to receive follow-up care at a cancer centre 4 years after diagnosis 12 .In the present study, which examined follow-up care for survivors diagnosed in the 8-year period immediately after the period covered by the prior study, only 30.2% of colorectal survivors received follow-up care at a cancer centre, with 10.3% having 5 or more follow-up visits.That observed decrease in cancer centre follow-up could in part reflect differences in the definition of follow-up visits, with the prior study having made the assumption that all visits at a cancer centre were for routine follow-up care, and the present study having used a decision rule to more precisely identify routine follow-up visits.However, that difference alone is unlikely to explain the decrease, and it therefore appears that oncologists are discharging colorectal cancer survivors from the cancer centre more routinely than they are discharging survivors of other cancers.The reasons for that observation are unknown, but might relate to two events.The first was the broad d issem i nat ion, by va r ious mecha n isms (ta rgeted workshops, oncology rounds, research briefs, and small group meetings), of the prior study to oncologists and cancer centre administrators or managers within the province.That dissemination could have had some effect on the ensuing practice patterns.The second was the implementation of survivorship care plans for colorectal cancer survivors approximately midway through the study period.The rollout of survivorship care plans started with colorectal cancer survivors and continued for several years before expansion to other disease sites.In an unintended way, survivorship care plans might have facilitated discharge by increasing the confidence of oncologists, pcps, and survivors in the abilities of pcps to provide quality follow-up care through transfer of vital information concerning surveillance, psychosocial supports, and late or long-term effects.
No Canadian data related to routine follow-up care for prostate and gynecologic cancer survivors are available with which to compare the present findings.Nonetheless, in the present study, the likelihood of follow-up visits at a cancer centre was greater for survivors of gynecologic cancer than for survivors of breast cancer.It is possible that the higher rate of visits relates to limited evidence about models of follow-up care in this population.Indeed, the randomized controlled trials demonstrating the safety and effectiveness of alternative models of follow-up care (that is, pcp-and nurse-led models) have been conducted in populations of breast, colon, and prostate cancer survivors 21 .
Our findings suggest that a relatively small number of oncologists continue to provide follow-up care for many of their patients and that other oncologists discharge more of their patients to community-based care after treatment.That variation between oncologists is not surprising.Indeed, for some oncologists, the relationships they develop with patients during the treatment period can make it difficult to transfer responsibility for care back to pcps in the community [22][23][24] .Medical oncologists have described a sense of responsibility during the follow-up period with respect to long-term toxicities and the survivor's compliance with ongoing therapies (for example, endocrine therapy) 24 ; other oncologists have questioned the ability of pcps to provide cancer-related care for survivors 25,26 .On the other hand, survivors are often confused over who is primarily responsible for cancer follow-up care [27][28][29] and uncertain about whether community-based providers have the expertise required to manage cancer follow-up 22,30 , both of which can contribute to concerns about discharge after treatment.The fact that follow-up guidelines often do not make specific recommendations about who should provide follow-up care might also contribute to a lack of role clarity for providers.
That the number of follow-up visits increased at a rate of 8% per year during the study period could be of particular concern for decision-makers in cancer care programs.Indeed, insufficiencies of human and material resources are already placing strain on specialist cancer centres, particularly in the face of a rising cancer incidence and new and more complex treatments 5,6 .Although the rate of increase for follow-up visits was less than that for all visits (which rose at a rate of 14% per year), such increases will make additional demands on already stretched cancer system resources.In fact, given evidence that pcp-led follow-up is equivalent to cancer centre follow-up, Choosing Wisely Canada, a national initiative to help clinicians and patients take part in conversations about effective care choices, has identified the provision of follow-up care in specialist cancer centres as 1 of its 10 tests, treatments, or practices that are frequently used in cancer care, but that are not supported by scientific evidence or are potentially harmful to patients.As a result, Choosing Wisely Canada recommends that follow-up care not be provided in a highcost setting (that is, a cancer centre) when it can be delivered just as effectively in a lower-cost setting (primary care) 31 .
The present study has a number of strengths.First, it provides a more precise measure of routine follow-up visits.As noted earlier, researchers examining follow-up care across Canada have, in the past, simply quantified visits to oncologists during the follow-up period and assumed that all visits represent routine follow-up care.In the present study, we used a decision rule to identify follow-up visits to oncologists and thus more accurately evaluate routine follow-up visits.Second, our study examined a much more recent period of interest, permitting exploration of patterns of care after policy shifts encouraging pcp-led follow-up.
The study also has a number of limitations.First, we examined follow-up care at cancer centres only, and thus data about follow-up care provided outside of the formal cancer system are not shown.Obtaining the latter data is the next logical step.
Second, our study was conducted in a single Canadian province, and it is unclear whether and how patterns of follow-up care in other jurisdictions would be similar.A recent multi-province study examining physician visits in the follow-up care period found similar overall patterns of care, and yet breast cancer survivors in Nova Scotia had fewer oncologist visits during the follow-up period than did survivors in British Columbia, Manitoba, or Ontario 18 .Despite that limitation, our findings point to the need for further study in various Canadian jurisdictions to understand whether their patterns of follow-up care are similar and how those patterns are changing over time.For instance, the increased rate of both follow-up visits and overall visits warrants similar investigation in other provinces because that finding has important implications for human health resource planning.
Third, we developed and applied a decision rule that used cancer centre scheduling data to identify routine follow-up visits.Developing that rule was a complicated task given the high number of values per relevant data field (for example, 80+ values for appointment type, 100+ values for resource use).However, working with cancer centre staff in an iterative manner throughout this study allowed us to refine the decision rule.The final rule and the resultant findings had face validity when shown to cancer centre staff.
Finally, we present data about survivors with at least 1 routine follow-up visit at a cancer centre beginning in the period 1 year after diagnosis.Clearly, some survivors might not be discharged back to primary care until at least 1 year after diagnosis; thus, a visit during that period could represent a final or "discharge" cancer centre visit.

CONCLUSIONS
Our findings show that oncologists continue to deliver routine follow-up care for many survivors of breast, prostate, colorectal, and gynecologic cancers.Although continuing follow-up might reflect survivor preferences, other factors are likely driving the ongoing involvement of oncologists in routine follow-up care.Our findings underscore the need for tailored interventions to improve post-treatment discharge for survivors at low risk of recurrence who can be safely and effectively cared for by their pcps.An improved understanding of the decrease in cancer centre follow-up for colorectal cancer survivors might shed light about effective interventions to transition survivors from specialist cancer centres to communitybased providers after treatment.

TABLE I
Survivor demographics, grouped by receipt of cancer centre routine follow-up care a Visited the cancer centres for treatment, but had 0 routine follow-up visits at cancer centres.b These columns are not mutually exclusive.Current Oncology, Vol. 24, No. 6, December 2017 © 2017 Multimed Inc.

TABLE II
Cancer centre routine follow-up visits by disease site

TABLE III
Negative binomial model of rate of cancer centre routine follow-up visits a