Estimates and predictors of health care costs of esophageal adenocarcinoma: a population-based cohort study

Background Esophageal adenocarcinoma (EAC) incidence is increasing rapidly. Esophageal cancer has the second lowest 5-year survival rate of people diagnosed with cancer in Canada. Given the poor survival and the potential for further increases in incidence, phase-specific cost estimates constitute an important input for economic evaluation of prevention, screening, and treatment interventions. The study aims to estimate phase-specific net direct medical costs of care attributable to EAC, costs stratified by cancer stage and treatment, and predictors of total net costs of care for EAC. Methods A population-based retrospective cohort study was conducted using Ontario Cancer Registry-linked administrative health data from 2003 to 2011. The mean net costs of EAC care per 30 patient-days (2016 CAD) were estimated from the payer perspective using phase of care approach and generalized estimating equations. Predictors of net cost by phase of care were based on a generalized estimating equations model with a logarithmic link and gamma distribution adjusting for sociodemographic and clinical factors. Results The mean net costs of EAC care per 30 patient-days were $1016 (95% CI, $955–$1078) in the initial phase, $669 (95% CI, $594–$743) in the continuing care phase, and $8678 (95% CI, $8217–$9139) in the terminal phase. Overall, stage IV at diagnosis and surgery plus radiotherapy for EAC incurred the highest cost, particularly in the terminal phase. Strong predictors of higher net costs were receipt of chemotherapy plus radiotherapy, surgery plus chemotherapy, radiotherapy alone, surgery alone, and chemotherapy alone in the initial and continuing care phases, stage III-IV disease and patients diagnosed with EAC later in a calendar year (2007–2011) in the initial and terminal phases, comorbidity in the continuing care phase, and older age at diagnosis (70–74 years), and geographic region in the terminal phase. Conclusions Costs of care vary by phase of care, stage at diagnosis, and type of treatment for EAC. These cost estimates provide information to guide future resource allocation decisions, and clinical and policy interventions to reduce the burden of EAC. Electronic supplementary material The online version of this article (10.1186/s12885-018-4620-2) contains supplementary material, which is available to authorized users.


Background
Esophageal cancer is the eighth most common cancer worldwide [1]. The incidence of esophageal adenocarcinoma (EAC) has increased rapidly in North America and other Western countries over the past several decades [2][3][4][5][6]. In fact, EAC has become the predominant histological subtype of esophageal cancer (relative to squamous cell carcinoma) in North America and Europe, and the sixth leading cause of cancer-related deaths worldwide [1,7,8]. In Canada, the incidence of EAC has risen steadily at 4% per year over the past 30 years (between 1981 and 2009), making it the most common type of esophageal cancer in Ontario [9]. These trends may be attributed to a growing and aging population, and the rise in the prevalence of important risk factors, such as obesity and gastroesophageal reflux disease (GERD) which leads to the development of Barrett's esophagus [7,[9][10][11]. Esophageal cancers symptomatically present late and carry poor prognoses, despite advances in multimodality treatment [12,13]. Esophageal cancer has the second lowest 5-year relative survival rate for people diagnosed with cancer in Canada (i.e., pancreatic cancer 9.5%, esophageal cancer 15.3%, lung cancer 20%, and liver cancer 20.4%) [14]. Therefore, diagnosing esophageal cancers at an early stage before the development of symptoms, is critical for improving prognosis [15].
Recent cancer-related cost estimates placed esophageal cancer patients who survived for more than 1 year post-diagnosis at the top of the cost table at $50,620 (95% CI $47,677-$53,562, 2009 Canadian dollars) [16]. These patients also had the highest cost for hospital admissions of all cancers ($27,506) due to the performance of resource intensive procedures, such as post-surgery esophageal dilation and biopsies to the esophagus or other parts of the gastrointestinal tract (through endoscopies) [16]. Additionally, these patients had frequent post-treatment follow-up visits [17], demonstrated by high costs for physician services ($4757) and home care ($4058) [16]. The costs were higher in the initial and terminal phases, and lower in the pre-diagnosis and continuing phases [18]. However, these studies provide estimates for EAC care that are broad in categorization, and more detailed estimates by specific clinical care elements and characteristic could provide significant data to guide clinical care, policy and future research.
Techniques to reduce EAC incidence, such as endoscopic mucosal resection or radiofrequency ablation of Barrett's esophagus, will likely be more cost-effective than current surveillance strategies that rely on early detection of cancer [19,20]. There is, however, limited relevant evidence in the Canadian context; costs estimates of EAC are needed for use in cost-effectiveness analyses of innovative technologies to inform health care professionals, policy makers, and the public in order to aid prevention and the early detection of EAC.
The purpose of this study was to estimate: i) the phase-specific net direct medical costs of care attributable to EAC for all adults aged 18 years and older, from the perspective of the Ontario Ministry of Health and Long-Term Care; ii) total net health care costs by cancer stage and type of treatment for EAC; and iii) predictors of the total net costs of care for individuals diagnosed with EAC.

Study design and setting
We conducted a population-based retrospective cohort study by linking the Ontario Cancer Registry (OCR) with administrative health data and a reference Ontario population to estimate the phase-specific net costs of care for primary EAC from January 1, 2003, through December 31, 2011. Individuals were followed from the day of diagnosis until death or until 12 months after the end of the study period, i.e., December 31, 2012, whichever came first. We approached costing [21][22][23][24][25] based on three care phases: 1) initial phase, the first 12 months after diagnosis of EAC, which would include diagnostic services, primary therapy, and adjuvant therapy to lower the risk of cancer recurrence; 2) continuing care phase, all months between the initial and terminal phases of care, which would include surveillance activities for detecting recurrences, follow-up treatment to prevent cancer recurrence, and treatment of complications following the initial therapy; and 3) the terminal phase, the final 12 months before death, which applies to care received at the end of life, often palliative in nature. For patients who died within 12 months post-diagnosis, the costs were attributed to the terminal phase only. For patients surviving < 24 months after diagnosis, the final 12 months of observation and costs of care were allocated to the terminal phase first while the remaining months were allocated to the initial phase [21,25]. For patients who did not die during the study period, the first 12 months (and costs) were allocated to the initial phase and all remaining months were allocated to continuing care phase [22]. We estimated phase-specific net costs of care as the difference between the mean costs for EAC cases and for matched controls without cancer [22,24,25]. Additionally, we stratified total net costs by stage at diagnosis and treatment for EAC, and identified predictors of total net costs.

Data sources
We conducted our analyses using population-level administrative health databases with information on all 14 million Ontario residents. Data were provided by the Institute for Clinical Evaluative Sciences, the main data repository for health records in the province of Ontario, Canada. These data have been validated for completeness and accuracy [26][27][28][29][30][31]. This included cancer registry linked to demographic and geographic information, physician billings for outpatient, inpatient, community-based, and laboratory services, hospital and emergency department discharge abstracts, hospital-based ambulatory care data, and prescription drugs (for those over age 65), home care, continuing care, and long-term care [32,33].
All cancer incidence in Ontario and subsequent mortality has been captured by the OCR from 1964 onwards. The Registered Persons Database contains demographic and geographic information for all people registered for provincial government-sponsored health insurance coverage. The Ontario Health Insurance Plan (OHIP) claims database contains the records of all physician billings for outpatient, inpatient, community-based, and laboratory services starting from July 1991. Non-physician procedures with an OHIP billing number (for example, midwife, chiropractor, nurse practitioner, or physiotherapist) are also included. Billings are based on the Ontario Health Insurance Plan fee-for-service rates in effect in the year the services were provided. The Canadian Institute for Health Information Discharge Abstract Database (CIHI-DAD) contains demographic, clinical, and administrative information on inpatient hospitalizations from April 1988 onwards; and CIHI-National Ambulatory Care Reporting System (CIHI-NACRS) contains administrative, demographic, clinical, and financial data for hospital-based and community-based ambulatory care (day surgery, emergency department visits, outpatient and community-based clinics) which is available from April 2003 onwards. OHIP, CIHI-DAD, and CIHI-NACRS fee codes were used to identify surgical resection, chemotherapy and radiotherapy, as well as esophageal dilation, drainage, esophageal stenting, laser debulking of tumor, and palliative care for EAC (see Additional file 1: Table S1). We used previously published and validated fee codes for these procedures [34].
Direct medical costs were determined using the perspective of the public payer. The costing methods followed the guidelines of the Canadian Agency for Drugs and Technology in Health [33] and the Health System Performance Research Network [32], and were based on previous cancer costing work done in Ontario [16,18,35]. Costs associated with physician services, including outpatient visits, laboratory services, diagnostic tests, emergency physicians, and medical and radiation oncologists, were determined through the OHIP claims database. The cost of inpatient hospitalization was determined from the CIHI-DAD database. Costs associated surgical resection, chemotherapy and radiotherapy for EAC were determined using the CIHI and OHIP databases with application of standard provincial unit costs. Individual-level income quintile was not available; therefore, area-level income quintile was used as a surrogate. Area-level income quintile was quantified using median neighbourhood household income, which was determined through linking of postal codes to Canadian census data and categorized into quintiles corresponding to income status of neighbourhoods. The income quintile 1 represents the lowest 20% of neighbourhoods and income quintile 5 represents the most well-off 20% of neighbourhoods.
Ontario has 14 health regions, called Local Health Integration Networks (LHIN) [37] which we used as a factor to explain regional health care service and availability. The Johns Hopkins Adjusted Clinical Groups case-mix system [38][39][40][41] was used for comorbidity adjustment [42][43][44].
Estimates of the net cost of care for EAC patients: Matching cases and controls The net cost method matches cases and controls on socio-demographic and clinical factors associated with resource use and calculates the difference in cost for cancer patients and non-cancer control subjects [22,24,25]. Cases (cancer patients) were identified as all eligible individuals 18 years of age and older in the OCR with an International Statistical Classification of Disease and Related Health Problems (ICD-9) site codes 150.0-150.9 and ICD-10 codes (C15.3-C15.9), in combination with histology International Classification of Diseases for Oncology, Third Edition (ICD-O-3) codes 8140-8575 corresponding to primary cancer (see Additional file 1: Table S2) [45]. Individuals were excluded if the EAC diagnosis was recorded on or after the date of death or individuals whose EAC was not the primary site.
Potential controls were selected from a 5% random sample of the reference Ontario population Registered Persons Database, including all individuals 18 years of age and older with no cancer diagnosis before or during our analysis period. Control subjects who died before the patient's EAC diagnosis date were excluded.
Two sets of cases and controls were used to match 1:1 at two index dates (date of diagnosis and 12 months preceding the date of death) to estimate costs for the initial and continuing care phases. For the latter index date, cases who died were matched 1:1 to controls with similar conditional probability of a diagnosis of EAC given the observed individual covariates [46,47] who died on the same date to estimate costs for the terminal phase. This was derived by fitting a logistic model with EAC status as the dependent variable and the index year (year of EAC diagnosis), age group at index date, gender, urban or rural residence, neighbourhood income quintile, Ontario health region, and comorbidity [18,35]. For each case, the closest non-EAC control was selected that matched the following criteria: age ± 5 years at the index date; same gender; same index year; comorbidity (ADGs), and a propensity score within a caliper width of 0.2 standard deviation [48].

Estimation of health care costs
Cost estimates for inpatient hospitalizations, same-day surgery, and emergency department visits were obtained by multiplying the resource intensity weight (measure of resource utilization intensity) by the cost per weighted case (unit cost) [32,[49][50][51]. Costs for services included in Ontario Health Insurance Plan, Ontario Drug Benefit, and Home Care were obtained by multiplying the number of services by unit cost. Continuing care cost was determined using Continuing Care Reporting System, which contains clinical and demographic information on individuals receiving facility based continuing care. Services include medical long-term care, rehabilitation, geriatric assessment, respite care, palliative care, and nursing home care. Patients are classified into 44 Resource Utilization Groups, and are assigned a Case Mix Index that approximates their per day resource usage. Case Mix Index is reviewed every quarter and can be adjusted multiple times [32]. Continuing care cost per weighted day was derived by dividing the total annual cost by the total annual weighted day. The case cost is the product of weighted days multiplied by the cost per weighted day. The cost of long-term care was obtained through the product of the year-specific length of stay and the Ministry of Health cost per diem. All costs were adjusted to 2016 Canadian dollars using the Consumer Price Index for Health and Personal Care [52]. Costs were undiscounted (i.e., exact costs billed).

Statistical analysis
Sociodemographic and clinical characteristics and health care costs for the EAC cases and non-EAC control cohorts were summarized by phase of care. We presented categorical variables as frequencies and percentages, and continuous variables as means ± standard deviations. For each phase of care, we estimated mean (95% confidence interval [CI]) net costs of care due to EAC (per 30 patient-days) using generalized estimating equations to account for the matched study design. Estimates were bootstrapped 1000 times to obtain CIs. Total net health care costs and by phase of care were analyzed by stage at EAC diagnosis and type of treatment received. Generalized estimation equation model with a logarithmic link and gamma distribution, which specifies the conditional mean function directly, was used to examine unadjusted and adjusted relationships between covariates and total net health care costs per 30 patient-days by phase of care among all EAC cases [53][54][55]. Potential covariates included age at EAC diagnosis, gender, urban or rural residence, birth country, income quintile, Ontario health region, comorbid conditions (ADGs), stage of disease at diagnosis, treatment for EAC, and year of EAC diagnosis. Variables with a significance level of P ≤ 0.2 in the univariate analyses were entered into the multivariate generalized estimation regression analysis and were considered independently significant when P ≤ 0.05 [56,57]. Interactions were considered in the context of regression analysis. The adjusted model was constructed according to a stepwise backward selection methodology and only included those variables that remained significant at the two-sided level of P ≤ 0.05 [57]. Finally, variables that were non-significant in the univariate test were added to see if they became significant when adjusted for other factors [58]. Statistical analyses were conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Sensitivity analysis
A sensitivity analysis was performed where the initial phase was defined as the first 6 months after diagnosis of EAC, the terminal phase was defined as the final 6 months before death, and the continuing care phase was defined as all months between the initial and terminal phases of care.

Characteristics of the study population
A flow chart of the study population is shown in Additional file 2: Figure S1. Overall, 3035 EAC cases and 560,997 control subjects were identified during the study period 2003-2011 (see Additional file 3: Table S3). Over the period, the number of new EAC cases increased from 285 in 2003 to 413 in 2011, and the proportion of those with age group at diagnosis of 50-54, 55-59, 60-64, 65-69, and 70-74 years increased from 5.8, 10.5, 8.3, 8.0, and 10.3% to 12.6, 17.1, 13.9, 18.3, and 12.5%, respectively. Stage at EAC diagnosis was available from 2003 in the data; 126 (4.2%) people were diagnosed with stage 0-I, while 420 (13.8%) were stage II, 455 (15.0%) were stage III, 940 (31.0%) were stage IV, and 1094 (36.1%) were unknown stage. In addition, the proportion of patients with known stages increased from 2003 to 2011; stage 0-I from 1.6 to 21.4%; stage II from 3.1 to 12.6%; stage III from 0.9 to 16.0%; and stage IV from 2.5 to 11.5%. Patients receiving treatment with radiotherapy alone after EAC diagnosis increased from 5.2% in 2003 to 19.3% in 2011. In addition, those not receiving treatment increased from 8.9 to 14.0%. In contrast, the proportion of patients receiving surgery plus chemotherapy decreased over time, from 13.6 to 5.1%. In our cohort, 2490 of EAC patients died during the mean 510 days or median 288 days of follow-up and 18,536 of controls died during the mean 2309 days or median 2373 days of follow-up. Table 1 describes the baseline characteristics of the matched cases and controls by phase of care. Cases that contributed person-time to the initial (259 days) and the terminal phase (242 days) were closely matched to the controls (initial phase: 360 days and terminal phase: 360 days); however, many cases that contributed person-time to the continuing care phase could not be matched with suitable controls (726 versus 1521 days).
Estimates of the average total net costs of EAC care per 30 patient-days were highest in the terminal phase ($8678, 96% of overall EAC net costs), followed by the initial phase ($1016, 11%) and continuing care phase ($669, 7%) of overall EAC net costs (Table 3 and see Additional file 4: Figure S2a-S2d). The net costs of inpatient hospitalization (85-97% of the mean health care costs of inpatient hospitalization in Table 2) and outpatient visits (75-97% of the mean health care of outpatient visits in Table 2) due to EAC accounted for the highest cost categories across all three phases. We reported bootstrap mean and 95% CIs derived from the generalized estimating equations on Additional file 5: Table S4. With large sample sizes, the bootstrap samples results are similar to the original sample.

Sensitivity analysis
In the sensitivity analysis assigning 6 months after the diagnosis to the initial phase and 6 months preceding death to the terminal phase, there was a significant increase (85%) in the total net costs of care in the continuing care phase and a modest increase (42%) in the initial phase compared with the total net costs of care in the primary analysis (Additional file 8: Table S5). The mean net costs of surgery plus radiotherapy (108%) and all treatments significantly increased in the initial phase (23-108%) and continuing care phase (63-639%), respectively compared with the primary analysis (Additional file 8: Table S6). Predictors of total net costs of care associated with EAC were similar to the primary analysis in the terminal phase (Additional file 8: Table S7).

Discussion
This population-based retrospective cohort study examined phase-specific net costs of care per 30 patient-days attributable to EAC from a public payer perspective, total net costs of care by stage and treatment for EAC, and predictors of total net costs of care in individuals diagnosed with EAC by phase of care. The aggregated total net health care costs of EAC care were highest in the terminal phase, next highest in the initial phase, and the lowest cost was in the continuing care phase. Inpatient hospitalizations accounted for the largest share of costs in all phases, followed by outpatient visits and home care. Overall, stage IV at diagnosis and surgery plus radiotherapy for EAC accounted for the highest cost, in particular in the terminal phase. The factors that were associated with higher net costs of care included treatment for EAC, especially chemotherapy plus radiotherapy, surgery plus chemotherapy, radiotherapy alone, surgery alone, and chemotherapy alone in the initial and continuing care phases; intermediate or advanced stage and the latest year of EAC diagnosis in the initial and terminal phases; comorbidity in the continuing care phase; and older age at diagnosis (70-74 years) and Ontario health region (Waterloo Wellington, North East, North West) in the terminal phase. Associations like older age and lower income quintile may reflect medical factors such as comorbidity, or social factors like lesser social support that could lead to higher use of medical services. Finally, lower costs were associated with individuals diagnosed with EAC included 85 years of age and older at index date and rural residence in the initial phase.
Phase-specific costs are useful for estimating incidence-based and long-term care costs, defined as cumulative costs from the date of diagnosis to death [22]. In addition, phase-specific cost estimates constitute an important input for economic evaluation of prevention, screening, and treatment interventions [21,22,25]. Our phase-specific costing approach provided in-depth cost analysis to the specific net phase of care costs for EAC, compared to previous studies which only looked at overall costs. Recent and past studies analyzing hospital costs after complex esophageal surgical procedures indicate that postoperative complications are associated with increased resource utilization and costs [59,60]. Such complications were captured in the phased costing approach we used. According to a large randomized trial, preoperative chemoradiotherapy is safe and leads to a significant increase in overall survival among patients with localized adenocarcinoma or squamous-cell carcinoma of the esophagus compared with those treated with surgery alone [61]. Esophageal cancer is often in an advanced stage when it is diagnosed, however. At later stages, esophageal cancer can be treated but not cured. The selection of prevention and treatment activities at different stages of disease can have significant impact on resource utilization [21,62].
The strengths of our study include comprehensive cost estimation and rigorous propensity score matching between cases and controls, which was based on sociodemographic and comorbidity characteristics, providing unbiased estimates of the net costs of care. Our study results can inform publically funded health care systems on the cost of treatments for patients, considering stage and other sociodemographic and clinical patient characteristics. It can also aid detailed future planning of health care costs.
Our study has some limitations. Our cost estimates did not reflect the overall economic burden of EAC to the society. Because Ontario only provides comprehensive coverage for the elderly and those on social assistance, prescription medication costs were not included