Direct Medical Spending on Young and Average-Age Onset Colorectal Cancer before and after Diagnosis: a Population-Based Costing Study

Colorectal cancer is the third most common cancer worldwide (1). Although age is an established risk factor, with marked onset of disease after 50 years of age (2), the incidence of young-onset colorectal cancer (yCRC) among individuals less than 50 years old has drastically risen over the past two decades (3–5). Despite the rising incidence of yCRC there remains a scarcity of information related to its economic burden. Although a 2022 global systematic review synthesized the average cost incurred during the 12-month period following a yCRC diagnosis, ranging from $23,368 to $89,945 (inflation adjusted to 2020 USD), a majority of the included studies used an age cutoff value of 65-years-old (6). However, emerging evidence indicates a marked increase in colorectal cancer cases as individuals shift from 49 to 50 years of age (3, 4). Therefore, it becomes prudent to estimate the cost of colorectal cancer at more frequent age intervals (i.e., <45 and <50) to better understand the economic burden of yCRC.

Because of the rising prevalence of yCRC, in 2021 the US Preventative Services Task Force lowered the recommended age of asymptomatic screening from 50 to 45-years-old (7), based on evidence that suggested that screening for colorectal cancer at age 45-years-old provided an efficient balance between increased colonoscopy burden and life-years gained (8). Another important metric to consider when evaluating the age of asymptomatic colorectal cancer screening is its cost effectiveness, which may include cost savings and life years gained through identification of early- versus late-stage yCRC. However, such an economic evaluation requires updated and accurate cost estimates for healthcare spending on yCRC that is representative of the current yCRC age demographic (i.e., age ≤50 years). Therefore, we aimed to accurately compute the economic burden of yCRC, based on onset below 50-years-old, from a public payer perspective. Specifically, we aim to estimate: (i) direct medical spending on yCRC before diagnosis; and (ii) direct medical spending attributable to yCRC after diagnosis, according to phases of care. In addition, we conducted a parallel assessment of direct medical spending on average age-onset colorectal cancer (aCRC, age at diagnosis ≥50 years) to facilitate comparison with yCRC.

We leveraged linked administrative health databases in British Columbia (BC), which capture longitudinal and de-identified patient-level health services data for all BC residents (9–15). Population Data BC facilitated data access to the following databases: (i) Discharge Abstract Database for inpatient visits (15); (ii) Medical Services Plan database for information pertaining to provision of services by healthcare practitioners under BC's public health insurance program (14); (iii) PharmaNet database for prescriptions dispensed from community and outpatient hospital pharmacies (12); (iv) Consolidation File on demographic characteristics (10); and (v) Vital Statistics File on deaths (13). These databases were linked to the BC Cancer Registry, which captures information on diagnosis (i.e., date, site, stage), and dispensing records from the Provincial Systemic Therapy Program and radiotherapy treatment records at BC Cancer (9), the sole public provider of radiotherapy and systemic cancer therapies across BC. Use of data provided by Population Data BC does not require informed patient consent. This study was approved by the University of British Columbia Research Ethics Board (H17–03530).

We used International Classification of Diseases for Oncology, Third Edition codes (Supplementary Table S1) to identify incident cases of colorectal cancer, diagnosed between January 1, 2010 and December 31, 2016. An index date, corresponding to date of colorectal cancer diagnosis was assigned. Next, we stratified colorectal cancer cases by age at diagnosis (i.e., index date). Specifically, those diagnosed with colorectal cancer at <50-years-old were defined as yCRC cases, whereas those diagnosed at ages ≥50-years-old were defined as aCRC cases. For each colorectal cancer case, data stewards at BC Cancer provided up to 10 cancer-free controls who were the same age, sex, and had at least one healthcare interaction (i.e., outpatient visit, inpatient visit, or prescription fill) within the same year the colorectal cancer case was diagnosed. Cancer-free controls were assigned an index date that corresponded to the date of entry into our study population and were not diagnosed with cancer before the index date or throughout the study period (January 1, 2009 to December 31, 2017). Finally, we excluded individuals without 12 months of observation before their index date within the aforementioned health databases, and cases without information on stage at diagnosis (Fig. 1 illustrates identification of colorectal cancer cases and cancer-free controls).

For estimation of direct medical spending before a colorectal cancer diagnosis, we defined a 1-year observation period preceding case index date. To facilitate comparison, we delineated a similar 1-year observation period that preceded the index date for cancer-free controls.

For estimation of direct medical spending attributable to yCRC after diagnosis, we used methods outlined by Mariotto and colleagues (16), who estimated the cost attributable to colorectal cancer across different phases of care among individuals 65 years and older. We followed cases and cancer-free controls from their index date until the end of follow-up (December 31, 2017), the last date of observation or death, whichever occurred first. Phases of observation for colorectal cancer cases were defined as (i) initial (first 12 months after diagnosis); (ii) continuing (months between the initial phase and end-of-life cancer/non-cancer phases or end of follow-up); (iii) end-of-life cancer (12 months before death due to cancer related reasons as identified using Vital Statistics File ICD-10 codes; Supplementary Table S1); and (iv) end-of-life non-cancer (12 months before death, due to non-cancer related reasons). For cases who died within 24 months of a colorectal cancer diagnosis, months of observation were first assigned to the applicable end-of-life phase, and then to the initial phase. Next, we assigned the following phases of care to cancer-free controls: (i) continuing phase (months between index date and end-of-life non-cancer phase or end of follow-up); and (ii) end-of-life non-cancer (12 months before death). The unit of analysis was matched month of observation—whereby we matched a given month of observation from a colorectal cancer case to a cancer-free control based on sex, age, and phase of care. Specifically, months of observation for colorectal cancer cases in the initial, continuing, and end-of-life cancer phase were matched to cancer-free control months of observation in the continuing phase, whereas colorectal cancer case months observed within the end-of-life non-cancer phase were matched to cancer-free controls in the end-of-life non-cancer phase (Supplementary Fig. S1 illustrates matching process).

We estimated direct medical healthcare spending from a public payer perspective based on the following healthcare components: Hospital visits, healthcare practitioners, medications provided by BC Cancer (e.g., chemotherapy and targeted therapy), and radiotherapy. We used the service date for each healthcare interaction to allocate costs toward the relevant phase of observation. For brevity, we used the term “costs” to refer to direct medical spending. Costs for inpatient visits, defined as a hospital visit resulting in at least one overnight stay, were calculated using the Canadian Institute for Health Information's Case-Mix Groups+ methodology (15), which included the cost of surgical treatments, overhead costs related to hospital operations and hospital staff excluding physicians. An estimated resource intensity weight was assigned to each inpatient record in the Discharge Abstract Database based on the case-mix group, patient characteristics, and presence of high-cost factors. We calculated the cost of each inpatient stay as the product of the resource intensity weight and cost of a standard hospital stay in BC for the year of interest. Costs related to provision of services by a healthcare practitioner were estimated through billing records in the Medical Services Plan database, which captures the dollar amount paid to a healthcare practitioner for a given visit through fee-for-service and shadow billings. Specifically, the Medical Services Plan database captures physician and surgeon remunerations, along with visits to allied healthcare professionals (i.e., optometrists, dentists, physiotherapists) that may be eligible for coverage under BC's publicly funded healthcare program. Costs for prescription medications were obtained from PharmaNet, as the dollar amount covered by PharmaCare, an income-based provincial program providing prescription medication coverage, which includes the drug cost and pharmacist dispensing fee. We excluded the cost of prescription medications paid for by the patient and/or third-party private insurance, as our study is from the public payer perspective.

To estimate the cost associated with the provision of prescription cancer therapies (e.g., chemotherapy and targeted therapy), we obtained a unit cost price list for publicly funded agents provided by BC Cancer, which was available for 2022 and inflation adjusted to 2017 CAD (Supplementary Table S2). We then calculated prescription cancer therapy costs as the product of unit cost and milligrams dispensed, as captured in the BC Cancer Registry. We restricted our cost estimates to medication indicated for colorectal cancer treatment, as defined by BC Cancer chemotherapy protocols (ref. 17; confirmed by co-author S.S.T. Yeung, a clinical pharmacist at BC Cancer) and excluded agents not funded by BC Cancer (i.e., experimental drugs).

Finally, the cost of radiotherapy was estimated using a previously developed activity-based costing model (18). Modifications were made to reflect allocation of planning and treatment times for colorectal cancer radiotherapy and the BC Cancer payment structure for radiotherapy oncologists and allied healthcare providers. We consulted with providers at BC Cancer to determine the number of hours allocated for planning and delivery and confirmed hourly wages. Unit cost estimates for treatment planning (cost per course of radiotherapy), treatment delivery (cost per fraction), and supporting infrastructure (i.e., overhead costs) were applied to individual-level data from the BC Cancer radiotherapy database (Supplementary Table S3). Costs related to clinical infrastructure and equipment acquisition and maintenance were not included as this was outside the scope of our study.

We estimated the average cost incurred during the 1-year period preceding the index date for yCRC and aCRC cases along with their cancer-free controls, stratified by age at index date (i.e., <50 or ≥50 years old). Cost incurred before diagnosis included the following healthcare components: inpatient visits, healthcare practitioners, and prescription medications.

Next, we estimated the average annualized cost attributable to yCRC from the index date until December 31, 2017 or death, according to the following healthcare components: inpatient visits, healthcare practitioners, prescription medications, along with colorectal cancer treatments, including chemotherapy and radiotherapy. We calculated the average monthly direct medical spending for each yCRC case and matched cancer-free controls according to phases of care, and estimated cost attributable to yCRC as the net difference between cases and controls. The numerator and denominator for calculation of the average monthly cost was defined as the sum of the respective cost components for cases and controls across the entire phase of care and number months observed within the phase of interest, respectively. Costs were then annualized. We stratified cost estimates before and after diagnosis according to cost components, age group, sex, cancer site and stage. A parallel analysis was conducted to calculate average annualized cost attributable to aCRC. Next, the Wilcoxon rank-sum test was used to compare overall annualized cost attributable to yCRC and aCRC for each phase of care, and stratified cost components. Finally, 95% non-parametric bootstrap percentile confidence intervals were computed to account for data skewness. All cost estimates were inflation adjusted to 2017 Canadian dollars (CAD) using the Medical Consumer Price Index, Healthcare (19). Analyses were conducted using SAS Version 9.4 (SAS Institute Inc.; Statistical Analysis System, RRID:SCR_008567).

The data that support the findings of this study are available from Population Data BC https://www.popdata.bc.ca/. Restrictions apply to the availability of these data, which were used under license for this study. Access to data provided by the Data Steward(s) is subject to approval but can be requested for research projects through the Data Steward(s) or their designated service providers. All inferences, opinions, and conclusions drawn in this publication are those of the author(s), and do not reflect the opinions or policies of the Data Steward(s).

The study population included 1,058 (45.4% females) yCRC and 12,619 (44.8% females) aCRC cases with a mean age at diagnosis of 42.4 ± 6.2 years and 68.1 ± 9.2 years, respectively (Table 1). yCRC cases were most frequently diagnosed with cancer in the rectum (41.3%) or left colon (41%) and disease at Stage III (35.7%) or IV (26.0%). Whereas aCRC cases were most frequently diagnosed with cancer in the left colon (42.3%) and disease at Stage III (29.6%) and I (21.9%). Finally, the mean Charlson–Romano comorbidity index score was greater for yCRC and aCRC cases in comparison with their corresponding cancer free controls.

Average costs incurred during the one-year period preceding the index date for yCRC and aCRC cases, along with corresponding cancer-free controls are shown in Table 2. The overall average cost incurred before a yCRC diagnosis was $6,711, which was greater in comparison with cancer-free controls ($2,343). Inpatient visits ($5,131) accounted for 76.5% of the overall average cost preceding a yCRC diagnosis. Stratification by stage at diagnosis showed an increase in the average cost incurred before a yCRC diagnosis according to stage, ranging from $2,784 for stage I to $9,140 for stage IV.

The overall average annualized cost attributable to yCRC during the initial phase was $50,216, which decreased to $8,361 during the continuing phase and later peaked during the end-of-life phases—whether due to cancer ($86,125) or non-cancer ($77,273; Table 3). The cost attributable to yCRC was largely driven by inpatient visits and provision of prescription cancer therapies across all phases of care. Specifically, inpatient visits accounted for 41.5% ($20,859), 31.6% ($2,640), 49.3% ($42,466), and 73.7% ($56,988) of the overall average cost attributable to yCRC, whereas prescription cancer therapies accounted for 38.1% ($19,125), 50.6% ($4,227), 33.8% ($29,100), and 19.4% ($14,972) during the initial, continuing, end-of-life cancer and end-of-life non-cancer phase, respectively. When cost estimates were stratified according to site, attributable costs were highest for rectal cancer within the initial and continuing phases. Finally, when stratified according to stage at diagnosis, we found the lowest costs for Stage I and highest for Stage IV, across all phases of care. Specifically, the average annualized cost estimates for Stage I were $23,707, $1,629, and $80,719, compared with $83,566, $31,642, $91,526 for Stage IV during the initial, continuing, and end-of-life cancer phases, respectively.

The overall average annualized cost attributable to aCRC was $37,842 during the initial phase, which decreased to $5,014 during the continuing phase and later increased to $61,512 and $23,316, during the end-of-life cancer and non-cancer phases, respectively (Table 4). Similar to yCRC, costs attributable to aCRC were primarily driven by inpatient visits and prescription cancer therapies. Inpatient visits accounted for 53.1% ($20,078), 35.8% ($1,797), 57.6% ($35,420), and 73.3% ($17,102), whereas provision of prescription cancer therapies accounted for 25.5% ($9,652), 43.0% ($2,157), 25.4% ($15,641), and 10.1% ($2,365) of the overall annualized cost attributable to aCRC during the initial, continuing, end-of-life cancer and end-of-life non-cancer phase, respectively.

The overall average annualized cost attributable to yCRC significantly differed in comparison with aCRC across the initial (yCRC $50,216 vs. aCRC $37,842, P < 0.001), continuing (yCRC $8,361 vs. aCRC $5,014, P < 0.001), and end-of-life cancer phases (yCRC $86,125 vs. aCRC $61,512 P < 0.001), but not end-of-life non-cancer (yCRC $77,273 vs. aCRC $23,316, P = 0.372). When comparing cost estimates stratified by age group, the absolute difference in the cost attributable to yCRC versus aCRC was greatest among individuals diagnosed with colorectal cancer between the ages of 30 and 39 years, as compared with those ages ≥80-years-old as cost estimates decreased with each 10-year increase in age at diagnosis. Although we noted significant differences between the cost attributable to yCRC and aCRC across all healthcare components, the absolute difference was greatest for the provision of prescription cancer therapies across all phases of care: initial (yCRC $19,125 vs. aCRC $9,652, P < 0.001); continuing (yCRC $4,227 vs. aCRC $2,157, P < 0.001); end-of-life cancer (yCRC $29,100 vs. aCRC $15,641, P < 0.001); and end-of-life non-cancer (yCRC $14,972 vs. aCRC $2,365, P < 0.001).

We used population-based administrative health data to provide a better understanding of direct medical spending on yCRC before and after diagnosis. The average cost incurred in the year preceding a yCRC diagnosis was $6,711. The overall average annualized cost attributable to yCRC after diagnosis was $50,216, $8,361, $86,125, and $77,273, during the initial, continuing, end-of-life cancer, and end-of-life non-cancer phases, respectively. To our knowledge, our study provides the first comprehensive estimate of direct medical spending on individuals with yCRC, that is, contextualized with parallel analyses for aCRC. Cost estimates reported in our study may be used as cost inputs in future economic evaluations, which may evaluate the cost effectiveness of lowering the age of asymptomatic screening for colorectal cancer.

The cost of phase-specific colorectal cancer treatment among older adults has been previously established. Using US Medicare data, Mariotto and colleagues (16) estimated the average cost (2014 USD) attributable to colorectal cancer among individuals 65 years or older to be $55,845, $5,313, $92,476, and $24,235 during the initial, continuing, end-of-life cancer and end-of-life non-cancer phases, respectively. Although prior colorectal cancer costing studies may have included individuals less than 50-years-old, many did not stratify cost by age at diagnosis or used age cutoff values that were not consistent with the current definition of yCRC (ref. 6; i.e., diagnosis at less than 50 years old; refs. 6, 20). Our study adds to the literature through evaluation of direct medical spending on yCRC before and after diagnosis, which is also stratified by cost components, stage at diagnosis, and cancer site at diagnosis along with a parallel analysis for aCRC for comparison. We found that cost attributable to yCRC was greater than aCRC, which was primarily driven by prescription cancer therapies as the cost of inpatients visits, healthcare practitioners, prescription therapy, and radiotherapy were comparable between yCRC and aCRC cases. The higher average cost of prescription cancer therapies among yCRC cases was likely due to the more frequent use of aggressive systemic treatment among patients with yCRC, in comparison with those with aCRC (2, 21, 22). Given the substantial economic burden of chemotherapy on healthcare spending, a future economic evaluation on the cost-effectiveness of aggressive systemic therapy use among patients with yCRC is warranted.

In alignment with current evidence, our study results demonstrate that individuals with yCRC are more likely to be diagnosed at later stages in comparison with those with aCRC (23, 24), as 61.7% of patients with yCRC in our study cohort were diagnosed at Stage III/IV, compared with 48.6% of patients with aCRC. Although aCRC accounted for 92.3% of colorectal cancer cases in our study cohort, given the rising incidence of yCRC and its substantial cost of treatment particularly those diagnosed with late-stage colorectal cancer, yCRC has the potential to place a significant burden on healthcare spending. Specifically, these findings underscore the importance of improving earlier identification of yCRC given implications for improved patient outcomes along with lower healthcare spending.

Finally, although lower than yCRC, 19.0% of aCRC cases in our study were diagnosed with stage IV colorectal cancer, with attributable costs estimated at $69,416, $28,881, $67,349, and $48,948 during the initial, continuing and end-of-life cancer and end-of-life non-cancer phases. The economic burden of late-stage aCRC may be reduced with increased participation in asymptomatic colorectal cancer screening programs, with a 2021 Canadian cohort study reporting lower cost of aCRC treatment among individuals who underwent colorectal cancer screening six months before diagnosis in comparison with those without an asymptomatic screening history (25). Despite implications for improved survival and lower healthcare spending, participation in asymptomatic colorectal cancer screening programs for individuals 50–74-years-old continue to fall below target in Canada (26) and the US (27). Requirements for up-to-date participation in asymptomatic screening programs to individuals 50 years or older must be emphasized.

Strengths and limitations of our study warrant discussion. Our use of population-based administrative health data provided generalizable and comprehensive estimates for direct medical spending on yCRC. To obtain accurate cost estimates, we matched observation months from individuals with yCRC to cancer-free controls by age, sex, and phase of care. We were unable to match by other covariates that may impact healthcare spending, namely race due to lack of data availability, and comorbidity status as it was not feasible to match on variables that may change over time. Although comorbidity status influences healthcare utilization, in 2018, Chen and colleagues (28) demonstrated that matching on comorbidity scores had minimal impact on attributable cancer cost estimates. Next, although we attempted to account for various healthcare components, we encountered several data limitations. First, we were unable to account for costs related to emergency department visits, as BC did not begin use of the National Ambulatory Care Reporting System until 2013. Second, we used the prescription medication dispensing date to assign associated costs to a month of observation, which may underestimate costs related to prescription medications in subsequent months. Third, although we accounted for the purchase price of prescription cancer therapies, we were unable to account for costs related to healthcare provider compensation and medical supplies, which may underestimate the true cost of treatment. In addition, due to data availability limitations we were only able to access the 2022 price list for prescription cancer therapies provided by BC Cancer. Although cost estimates were adjusted to 2017 CAD, due to the rapidly rising cost of cancer therapies (29), cost estimates reported in our study may not align with the cost of treatment in 2017. Fourth, approximately 30% of BC physicians receive remuneration through alternative payment plans (i.e., service contracted and salaried physicians), which is estimated to represent 20% of total physician payments in BC. Thus, our estimate of physician remuneration captured through the MSP was likely underestimated. Finally, cost estimates for radiotherapy were derived from an activity-based costing model developed to estimate cost of prostate cancer in Ontario and has not been validated for use in BC (18), though we leveraged clinical expertise to modify the model and capture cost of colorectal cancer radiotherapy in BC.

Our population-based estimate of direct medical spending on yCRC before and after diagnosis provides better understanding of the economic burden of this disease. We found higher attributable costs for yCRC after diagnosis across all phases of care in comparison with aCRC, which was primarily driven by the higher cost of prescription cancer therapies among yCRC cases and found cost attributable to yCRC and aCRC to vary by stage and tumor site at diagnosis. The reported cost estimates may be leveraged by future economic evaluations pertaining to yCRC such as asymptomatic screening programs targeted toward a younger demographic and the cost effectiveness of aggressive treatment regimens among patients with yCRC.

J.M. Loree reports personal fees from Amgen, grants and personal fees from Ipsen, Novartis, and Bayer, and grants and non-financial support from Personalis, and non-financial support from SAGA Diagnostics, as well as personal fees from Merck and Eisai, and non-financial support from AstraZeneca outside the submitted work. C.J. Brown reports personal fees from Ethicon outside the submitted work. No disclosures were reported by the other authors.

R. Garg: Conceptualization, supervision, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. E.C. Sayre: Formal analysis, visualization, writing–review and editing. R. Pataky: Visualization, methodology, writing–review and editing. H. McTaggart-Cowan: Visualization, methodology, writing–review and editing. S. Peacock: Conceptualization, visualization, methodology, writing–review and editing. J.M. Loree: Conceptualization, visualization, methodology, writing–review and editing. M. McKenzie: Visualization, methodology, writing–review and editing. C.J. Brown: Conceptualization, visualization, methodology, writing-review and editing. S.S.T. Yeung: Visualization, methodology, writing–review and editing. M.A. De Vera: Conceptualization, data curation, supervision, funding acquisition, investigation, visualization, methodology, writing–review and editing.

This research was funded by a Project Grant from the Canadian Institutes of Health Research, “Examining the epidemiology, treatment, and outcomes in young-onset colorectal cancer” (Funding reference number: PJT-159467), which was awarded to M.A. De Vera. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the article. We would like to acknowledge the following individuals for assisting with costing information: Dr. S. Lefresne, Dr. C. Duzenli, and Alison Giddings.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.