If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Division of General Surgery, Department of Surgery, Queen's University, Kingston, Ontario, CanadaDepartment of Public Health Sciences, Queen's University, Kingston, Ontario, Canada
Department of Public Health Sciences, Queen's University, Kingston, Ontario, CanadaICES, Queen's, Queen's University, Kingston, Ontario, CanadaDivision of Cancer Care and Epidemiology, Queen's Cancer Research Institute, Queen's University, Kingston, Ontario, Canada
Timely cancer treatment improves survival and anxiety for some sites. Patients with esophageal cancer require specific workup before treatment, which can prolong the time from diagnosis to treatment (treatment interval [TI]). The geographical variation of this interval remains uninvestigated in patients with esophageal cancer.
Methods
This retrospective population-level study conducted in Ontario used linked administrative health care databases. Patients treated for esophageal cancer between 2013 and 2018 were included. The TI was time from diagnosis to treatment. Patients were assigned a geographical Local Health Integration Network on the basis of postal code. Covariates included patient, disease, and diagnosing physician characteristics. Quantile regression modeled TI length at the 50th and 90th percentile and identified associated factors.
Results
Of 7509 patients, 78% were male and most were aged between 60 and 69 years. The 50th and 90th percentile TI was 36 (interquartile range, 22-55) and 77 days, respectively. The difference between the Local Health Integration Network with the longest and shortest TI at the 50th and 90th percentile was 18 and 25 days, respectively. Older age (P < .0001), greater comorbidity (P = .0005), greater material deprivation (P = .001), rurality (P = .03), histology (P = .02), and treatment group (P < .0001) were associated with a longer median TI. Older age (P = .03), greater comorbidity (P = .003), greater material deprivation (P = .005), rurality (P = .04), and treatment group (P < .0001) were associated with a longer 90th percentile TI.
Conclusions
Geographic variability of time to treatment exists across Ontario. Investigation of facility-level differences is warranted. Patient and disease factors are associated with longer wait times. These results might inform future health care policy and resource allocation.
Esophageal cancer management is a complex, multistep process. In Ontario, health regions coordinate the care of their own patients. We found differences in time to first health care encounter and time to treatment between health regions, despite adjusting for numerous covariates. Older, comorbid, and rurally located patients waited longer than others.
Timely access to cancer treatment has improved survival outcomes for many disease sites,
and symptom progression while patients await treatment. Patients require time to accommodate staging investigations and specialist visits before treatment begins. Expediting these is crucial in patients with esophageal cancer because of the number of patients who present with locally advanced disease.
Between 2005 and 2010, Ontario Health Cancer Care Ontario regionalized thoracic cancer management. Only hospitals that maintained adequate surgical volumes for lung and esophageal resection, including the appropriate personnel and equipment, received funding to manage these patients.
They postulated that having 1 institution centralize the workup and treatment of esophageal cancer in that region might reduce the number of missed appointments, repeat investigations, and therefore reduce the overall time between diagnosis and treatment (treatment interval [TI]).
partitioned the TI of patients with breast, lung, colon, or rectal cancer into time from diagnosis to the first oncologist consult, and time from the first oncologist consult to treatment. These subintervals have not yet been investigated for esophageal cancer in Canada.
A detailed understanding of the esophageal cancer TI might help improve equitable access to necessary investigations and treatments. Knowledge of the subinterval that contributes most to the TI might inform refinements to the patient pathway and resource allocation. In this study, we aimed to describe the lengths of the TI and subintervals, to investigate the geographical variation of the TI across Ontario, and to evaluate factors associated with the length of those intervals in Ontario esophageal cancer patients.
Methods
Study Design
We conducted a population-level cross-sectional study using linked administrative health care databases housed at Institute for Clinical Evaluative Sciences (ICES). ICES is an independent, nonprofit research institute funded by an annual grant from the Ontario Ministry of Health and the Ministry of Long-Term Care. As a prescribed entity under Ontario's privacy legislation, ICES is authorized to collect and use health care data for the purposes of health system analysis, evaluation, and decision support. Secure access to these data is governed by policies and procedures that are approved by the Information and Privacy Commissioner of Ontario. In Canada, health care is delivered under a universal government-funded system. A population of 14.7 million residents makes Ontario the most inhabited Canadian province. This study was approved by the Research Ethics Board of Queen's University (approval number 6030561; approval date: October 5, 2020).
Data Sources
Patients were identified in the Ontario Cancer Registry (OCR), a province-wide database that captures >96% of all incident cancers.
The OCR was linked to other health administrative databases to obtain demographic, disease, billing, and outcomes data. We used the Registered Persons Database, National Ambulatory Care Reporting System (NACRS), Discharge Abstract Database, Ontario Health Insurance Plan (OHIP), Same Day Surgery, Postal Code Conversion File (PCCF), Local Health Integration Network (LHIN),
Ontario Marginalisation Database, Immigration, Refugees, and Citizenship Canada Permanent Resident Database, Activity Level Reporting, and ICES Physician Database (Table E1). These databases were linked using unique encoded identifiers at ICES.
Study Population
Adult patients diagnosed with incident esophageal cancer between 2013 and 2018 who received treatment were included. Cancer site was identified using topography codes; histology was not restricted (Table E2). Patients were excluded if there was no biopsy procedure, if there was no investigation or consultation between diagnosis and treatment, or if treatment was <4 days or 6 months after diagnosis (Figure 1). Less than 4 days was chosen to exclude patients who presented emergently and had expedited treatment and so were unlikely to have followed the Cancer Care Ontario treatment pathway,
TI length was defined as the number of days from diagnosis to the first treatment. Secondary outcomes were the length of subinterval 1 (time from diagnosis to the first cancer-related event thereafter) and subinterval 2 (time from the first cancer-related event to treatment). The first cancer-related event could be either a specialist visit or an investigation (Table E3).
Date of Diagnosis
We first identified the diagnosis date in the OCR, and then used NACRS, Canadian Institute for Health Information (CIHI), and Ontario Health Insurance Plan billing date to identify the date of an endoscopic biopsy within 2 weeks of the OCR date. For those with a biopsy record on the same day as the OCR, this date was assigned the diagnosis date. For the remainder, the earliest biopsy date was used. For those for whom the biopsy date was >2 weeks before or after the corresponding OCR date, we used that OCR date as the diagnosis date.
Covariates
Age and sex were categorized. Comorbidity information was gathered from 6 to 30 months before the diagnosis date and categorized on the basis of the Johns Hopkins Aggregated Diagnosis Groups (ADGs). The ADGs were created using the Johns Hopkins ACG System v10.0.1 (build 879). Rurality was dichotomized into urban/rural using the PCCF. LHINs are geographical health regions within Ontario tasked to fund and distribute health care to residents living within their borders.
During the study period, there were 14 LHINs. Each patient was assigned a LHIN depending on their postal code at diagnosis using the PCCF and LHIN databases. Material deprivation is an objective marker of socioeconomic status
and is widely used in health services research. We used the Ontario Marginalisation Database to assign each patient a dissemination area via the PCCF using their postal code on the day of diagnosis. Each patient was given a score, and then categorized into quintiles, with quintile 1 being the least deprived. Recent immigration was labeled as yes or no depending on whether the number of years from the date of landing to the diagnosis date was 5 years or less. Histology and tumor location were categorized. Stage was defined using the American Joint Committee on Cancer eighth edition.
First, we used the OCR to identify the best stage information for each patient. The OCR uses an algorithm that provides the stage from a pathological diagnosis if available. If no such diagnosis exists, the algorithm assigns stage on the basis of radiology results, followed by cancer center patient chart entries. Second, we created a separate stage variable for those with missing OCR stage using individual American Joint Committee on Cancer eighth edition T, N, and M categories from the Activity Level Reporting. Diagnosing physician characteristics included specialty (if there was more than one specialty then “mainspecialty” was used), years in practice, and academic affiliation. We operationalized health care utilization as the use of the emergency department (ED) and/or a hospital admission between diagnosis and treatment.
Statistical Analyses
Descriptive statistics were used to describe baseline demographic characteristics. We conducted bivariate analyses of each independent variable against the 50th and 90th percentile of the TI and subintervals using nonparametric tests. We used multivariable quantile regression models, adjusting simultaneously for patient factors, disease factors, treatment group, and LHIN. We used stage in the sensitivity analyses described in the following paragraph. All data processing and analyses were performed at ICES Queen's using the SAS software version 9.4 (SAS Institute Inc).
We conducted 6 sensitivity analyses on the adjusted quantile regression analysis. The first removed LHIN from the original model to determine if LHIN-based patient characteristic variations had distorted the patient characteristic associations. For the second, third, and fourth, immigration and rurality, separately then combined, were removed from the original model to assess whether those variables influenced the LHIN effects. The fifth removed the treatment group from the original model to assess the independence of the other factors from treatment. Last, we added stage to the original model to assess its effect on those associations.
Results
Of 7822 patients diagnosed with esophageal cancer, 7509 patients had a recorded biopsy procedure, and 6042 received at least 1 treatment modality. After exclusions, the final study cohort comprised 5759 patients (Table 1). Most patients were male (77.6%), had a total ADG of between 4 and 6 (32.3%), were not recent immigrants (93.5%), lived in an urban area (84.6%), had adenocarcinoma (AC; 71.6%), and had lower esophagus (39.9%) or gastroesophageal junction (39.8%) tumors. Staging was as follows: I: 5.1%, II: 9.6%, III: 14.8%, IV: 23.9%, and missing: 46.7%. Gastroenterologists diagnosed the most cancers (41.3%). Chemoradiotherapy was the most common first treatment modality (26.8%).
Table 1Patient, disease, diagnosing physician, health care system, and health care utilization characteristics of Ontario patients with esophageal cancer between 2013 and 2018
Cohort characteristic
Number of patients (%)
Age group, y
18-49
296 (5.1)
50-59
1105 (19.1)
60-69
1923 (33.3)
70-79
1596 (27.6)
≥80
855 (14.8)
Sex
Female
1293 (22.4)
Male
4482 (77.6)
Sum of minor AGDs
0
338 (5.9)
1-2
1054 (18.2)
3-4
1481 (25.7)
5-6
1421 (24.6)
≥7
1481 (25.7)
Sum of major ADGs
0
2097 (36.3)
1
1799 (31.2)
2
1048 (18.2)
≥3
831 (14.4)
Total number of ADGs
0
288 (5.0)
1-3
1346 (23.3)
4-6
1866 (32.3)
7-9
1309 (22.7)
≥10
966 (16.7)
Recent immigration
No
5401 (93.5)
Yes
374 (6.5)
Material deprivation
Least deprived
1075 (18.6)
2
1155 (20.0)
3
1130 (19.6)
4
1176 (20.4)
Most deprived
1198 (20.7)
Unknown
41 (0.7)
Rurality
Rural
884 (15.3)
Urban
4885 (84.6)
Unknown
6 (0.1)
Calendar year of diagnosis
2013
930 (16.1)
2014
892 (15.5)
2015
957 (16.6)
2016
948 (16.4)
2017
1002 (17.4)
2018
1046 (18.1)
Histology
Adenocarcinoma
4133 (71.6)
Squamous cell carcinoma
1195 (20.7)
Other
447 (7.7)
Tumor site
Cervical esophagus
94 (1.6)
Upper esophagus
192 (3.3)
Middle esophagus
564 (9.8)
Lower esophagus
2305 (39.9)
Gastroesophageal junction
2298 (39.8)
Other
322 (5.6)
Stage
I
294 (5.1)
II
552 (9.6)
III
852 (14.8)
IV
1382 (23.9)
Unknown
2695 (46.7)
Diagnosing physician main specialty
Gastroenterology
2384 (41.3)
General surgery
1777 (30.8)
Thoracic surgery
584 (10.1)
Other
554 (9.6)
Unknown
476 (8.2)
Diagnosing physician years in practice
1-9
247 (4.3)
10-14
835 (14.5)
15-19
592 (10.3)
20-24
512 (8.9)
25-29
452 (7.8)
≥30
451 (7.8)
Unknown
2686 (46.5)
Diagnosing physician academic affiliation
No
3462 (60.6)
Yes
1484 (25.7)
Unknown
829 (14.4)
LHIN of residence at diagnosis
01
345 (6.0)
02
523 (9.1)
03
323 (5.6)
04
832 (14.4)
05
246 (4.3)
06
375 (6.5)
07
374 (6.5)
08
496 (8.6)
09
674 (11.7)
10
307 (5.3)
11
517 (9.0)
12
281 (4.9)
13
353 (6.1)
14
129 (2.2)
Treatment group
Endoscopy with or without subsequent treatment
543 (9.4)
Chemotherapy only
792 (13.7)
Radiotherapy only
1177 (20.4)
Surgery with or without subsequent treatment
571 (9.9)
Chemotherapy and radiotherapy
1550 (26.8)
Chemotherapy or radiotherapy then surgery
164 (2.8)
Chemotherapy and radiotherapy then surgery
733 (12.7)
Other
245 (4.2)
ED visits between diagnosis and treatment
0
4911 (85.0)
1
661 (11.4)
>1
203 (3.6)
Hospital admissions between diagnosis and treatment
0
4228 (73.2)
1
1323 (22.9)
>1
224 (3.9)
ADG, Aggregate diagnostic group; LHIN, Local Health Integration Network; ED, Emergency Department.
The median TI length was 36 days (interquartile range [IQR], 22-55 days) and the 90th percentile was 77 days (Figure 2). The subinterval 1 median length was 2 days (IQR, −3 to 10 days; 90th percentile, 20 days); the subinterval 2 median length was 34 days (IQR, 20-51 days; 90th percentile, 73 days).
Figure 2Box and whisker plot depicting the distribution of the Ontario esophageal cancer treatment interval between 2013 and 2018. Upper whisker = maximum observation excluding outliers; lower whisker = minimum observation excluding outliers; upper box bar = 75th percentile; lower box bar = 25th percentile; middle box bar = 50th percentile; dots = outliers (observations outside 1.5 times interquartile range).
Geographical differences were seen (Figures 3 and 4, Table 2). The difference between the LHINs with the longest and shortest TI at the 50th and 90th percentile was 18 and 25 days, respectively. Except LHIN 14, all exhibited similar distributions of width and skew, suggesting similar variability within each LHIN. Both subinterval lengths differed across LHINs (P < .0001).
Figure 3Box and whisker plot showing the comparison of the esophageal cancer treatment interval length distribution among LHINs in Ontario between 2013 and 2018. Upper whisker = maximum observation excluding outliers; lower whisker = minimum observation excluding outliers; upper box bar = 75th percentile; lower box bar = 25th percentile; middle box bar = 50th percentile; dots = outliers (observations outside 1.5 times interquartile range). LHIN, Local Health Integration Network.
Figure 4Different distributions of the esophageal cancer treatment interval length among Local Health Integration Networks in Ontario between 2013 and 2018. Upper whisker = maximum observation excluding outliers; lower whisker = minimum observation excluding outliers; upper box bar = 75th percentile; lower box bar = 25th percentile; middle box bar = 50th percentile; dots = outliers (observations outside 1.5 times interquartile range).
Table 2Lengths of the treatment interval, subinterval 1, and subinterval 2 at the 50th and 90th percentile according to category of associated factors in Ontario patients with esophageal cancer between 2013 and 2018
Treatment group: A = endoscopy with or without subsequent treatment; B = chemotherapy only; C = radiotherapy only; D = surgery with or without subsequent treatment; E = chemotherapy and radiotherapy; F = chemotherapy or radiotherapy then surgery; G = chemotherapy and radiotherapy then surgery; and H = other.
∗ Subinterval 1 = diagnosis to first health care encounter.
† Subinterval 2 = first health care encounter to treatment start.
‡ Treatment group: A = endoscopy with or without subsequent treatment; B = chemotherapy only; C = radiotherapy only; D = surgery with or without subsequent treatment; E = chemotherapy and radiotherapy; F = chemotherapy or radiotherapy then surgery; G = chemotherapy and radiotherapy then surgery; and H = other.
Differences remained between LHINs after adjusting for confounding. The biggest change was seen in LHIN 12, which had a 5-day longer median TI (−11 to −6 days longer than the referent group). The remainder demonstrated change of 3 days or less in median TI, suggesting minimal confounding by other covariates in those LHINs.
Bivariate Analysis of Associated Factors
Younger patients (18-49 years) had shorter median TIs than older patients (70-79 years); 30 days (IQR, 15-45) versus 38 days (IQR, 23-57 days; P = .01). Those with a total ADG score of ≥10 waited a median of 40 days (IQR, 23-62 days) versus 33 days (IQR, 18-49 days) for those with no comorbidity. The median TI did not differ statistically on the basis of sex (P = .30), immigration (P = .64), rurality (P = .46), nor deprivation (P = .49).
Those with histology other than AC or squamous cell carcinoma (SCC) had a shorter median TI (29 days; IQR, 16-51 days) than those with AC (37 days; IQR, 22-56 days) or SCC (35 days; IQR, 22-52 days). Cancer stage was inversely proportional to the median and 90th percentile of the TI length; stage I, 48 days (IQR, 35-66 days) versus stage IV, 28 days (IQR, 17-44 days; P < .0001).
Patients diagnosed by a thoracic surgeon had a shorter median TI; 33 days (IQR, 18-49 days) compared with those diagnosed by a general surgeon; 38 days (IQR, 24-56 days) or gastroenterologist; 37 days (IQR, 22-56 days). However, those diagnosed by a physician in the “other” category had the shortest median (29 days) and the 90th percentile (71 days) TI. There was no statistical difference in median TI regarding the number of years the diagnosing physician had been in practice (P = .13), however there was a difference at the 90th percentile (P = .0004; 10-14 years, 70 days vs 20-24 years, 82 days). Academic affiliation was not associated with the median (P = .08) or 90th percentile (P = .26) TI length. The median and 90th percentile TI was longer in patients who had one or more ED visits or hospital admissions between diagnosis and treatment; >1 ED visit, 54 days (IQR, 31-76 days) versus 0 ED visits, 35 days (IQR, 21-52 days); >1 admission, 48 days (IQR, 29-73 days) versus 0 admissions, 36 days (IQR, 22-54 days).
Adjusted Regression Analysis of Associated Factors
In the adjusted models (Table 3), age, comorbidity, deprivation, rurality, histology, and LHIN were associated with statistical differences in the median TI length. There was a 9-day difference in the age variable between those with the longest (≥80 years old) and shortest TI (18-49 years old). The remainder of the variables that showed statistical adjusted differences had differences of <5 days.
Table 3Unadjusted and adjusted differences of the treatment interval at the 50th and 90th percentile according to category in Ontario patients with esophageal cancer between 2013 and 2018
Treatment group: A = endoscopy with or without subsequent treatment; B = chemotherapy only; C = radiotherapy only; D = surgery with or without subsequent treatment; E = chemotherapy and radiotherapy; F = chemotherapy or radiotherapy then surgery; G = chemotherapy and radiotherapy then surgery; and H = other.
P < .0001
P < .0001
P < .0001
P < .0001
Unadjusted Intercept
41 (39-43)
64 (60-68)
A
−17 (−20 to −14)
−16 (−19 to −14)
−9 (−18 to −1)
−16 (−22 to −10)
B
0 (−3 to 3)
0 (−3 to 3)
27 (20-34)
21 (14-27)
C
−10 (−13 to −7)
−11 (−13 to −8)
13 (7-19)
6 (0-13)
D
15 (12-18)
15 (12-18)
45 (35-55)
37 (29-44)
E
−4 (−7 to −2)
−3 (−5 to −1)
10 (5-15)
8 (4-12)
F
−1 (−6 to 4)
−2 (−7 to 3)
−3 (−13 to 7)
0 (−9 to 9)
G
Referent
Referent
Referent
Referent
H
−2 (−6 to 2)
−2 (−5 to 2)
0 (−8 to 8)
−4 (−10 to 2)
CI, Confidence interval; ADG, Aggregate Diagnostic Group; LHIN, Local Health Integration Network.
∗ Treatment group: A = endoscopy with or without subsequent treatment; B = chemotherapy only; C = radiotherapy only; D = surgery with or without subsequent treatment; E = chemotherapy and radiotherapy; F = chemotherapy or radiotherapy then surgery; G = chemotherapy and radiotherapy then surgery; and H = other.
At the 90th percentile, age, comorbidity, material deprivation, rurality, LHIN, and treatment group were associated with differences in the TI length. Those who underwent endoscopic resection (alone or initially) waited 53 days less for treatment compared with those who underwent surgery (alone or initially). Patients aged ≥80 years waited 12 days longer for treatment than those aged 18-49 years old. Those with 3 or more comorbidities waited 11 days longer than those without any comorbidities. Patients living in the most materially deprived areas waited 6 days longer than those in the least deprived areas, and those in rural locations waited 6 days longer than their urban counterparts. Rurality and material deprivation variables became significant in the adjusted analysis at both percentiles.
Sensitivity Analyses
At the 50th percentile, only the exclusion of LHIN and immigration affected other variables, resulting in rurality no longer being significant compared with the original model (Table E4). At the 90th percentile, in 4 of the 6 sensitivity analyses, deprivation and rurality became insignificant, whereas number of minor comorbidities and disease histology and site became significant. The addition of stage resulted in age and rurality no longer being significant but did not affect other variables (Table E5).
Discussion
The key finding of this study was an absolute difference of 18 and 25 days between the LHINs with the longest and shortest median and 90th percentiles, respectively. Furthermore, we identified those who are older, more comorbid, and diagnosed by a physician other than a thoracic surgeon to be vulnerable patient populations that might be more at risk of a prolonged TI.
it was reported that the median time to treatment was 46 days (IQR, 29-66 days) in 79% of their patients. Our cohorts were created differently, which might explain the difference. We labeled the day of diagnosis as the date of endoscopic biopsy if one was available (>80%), and the date of diagnosis in the OCR otherwise, whereas those authors used the OCR date as the day of diagnosis for all. In an older study
a median wait time from esophageal cancer diagnosis to surgery of 32 days, was reported, but that study's cohort was restricted only to patients who underwent surgery, and the study period was 1984 to 2000, which preceded the provincial regionalization of thoracic cancer services. In contrast, our cohort included patients who had treatment modalities other than surgery. This difference might explain why our median TI was shorter, because the patients in our study who underwent surgery first had a longer TI than those who had another treatment before surgery. A more recent study from the United States
calculated a median time to surgery of 54 days in patients with cT1N0M0 esophageal carcinoma who underwent surgery from 2004 to 2015. This also corroborates our findings that patients with an early-stage cancer, or patients having surgery as their first treatment modality, have a longer TI than others.
We found variability across LHINs in all time intervals, at the 50th and 90th percentile. The goal of regionalization was to provide optimal patient care for those who require specialist services, regardless of their location in the province.
demonstrated that median wait times for lung cancer treatment did not shorten over the period from 2007 to 2011, but there was a reduction in 30-day mortality after pneumonectomy. All LHINs have a thoracic center located within their borders except one. One LHIN contains 3 thoracic centers. Neither of these 2 LHINs had the shortest or longest TI, suggesting the difference is explained by factors other than regionalization. Our sensitivity analyses (Tables E4 and E5) showed persistent LHIN differences at the 50th and 90th percentiles, suggesting there might be systemic inefficiencies meriting further study. Table E6 shows the distribution of patient factors within each LHIN.
At the 50th percentile, older, more comorbid, nonurban, and patients living in the most deprived areas waited up to 9 days, 3 days, 2 days, and 4 days longer than their counterparts, respectively. These differences were greater at the 90th percentile (12 days, 11 days, 6 days, and 5 days, respectively). Despite being statistically significant, these differences might not be clinically meaningful on survival
reported that older patients, those living in rural areas, and those with a lower income had a longer wait time to colorectal cancer surgery in Ontario than others. Kulkarni and colleagues
also reported that older age and more severe comorbidity burden were associated with a longer wait time for urology cancer treatment. Bardell and colleagues
also reported that increasing age, decreasing household income, and female sex were predictors of longer wait times between diagnosis and surgery for a cohort comprised of 12 different cancers, but did not stratify their analysis on the basis of cancer type. Possible reasons for differences according to patient characteristics have been postulated. Elderly patients might have more missed or rescheduled appointments, which might contribute to a longer TI.
Patients living in an area of higher material deprivation, which we used as a surrogate for individual-level socioeconomic status, might miss appointments because of difficulty getting time off work or paying for transport to their appointments.
It is unclear why histological subtypes other than AC or SCC would have a shorter TI. Rare diagnoses are more likely to be brought to the multidisciplinary team for discussion and this might expedite pretreatment investigations and specialist visits. Patients with a stage IV cancer had a TI that was 20 days shorter than those with a stage I cancer. At the 90th percentile, this difference increased to 38 days. Previous studies conducted in Ontario have shown the same phenomenon in other cancer sites.
Possible explanations exist. First, patients with a later stage are more likely to have symptoms from their disease than those with early-stage cancers. At the system level, symptomatic patients might have their investigations and specialist visits expedited because of the concerning symptom severity. Second, there might be a lower sense of urgency with lower-stage cancers, and in a resource-constrained health care system, those with a higher stage will likely take priority for investigations and treatment. Third, possible treatment options vary between stage I and stage IV. Most stage IV patients will undergo palliative treatment that does not include surgery.
Our results have shown that patients who are receiving radiotherapy alone have a much shorter TI length than those who undergo surgery.
Patients diagnosed by a thoracic surgeon had a shorter TI than those diagnosed by a gastroenterologist or a general surgeon. Those latter patients will be referred to a thoracic surgeon for a consultation; patients diagnosed by a thoracic surgeon might have that consultation at the same time as the diagnosis, thereby skipping a step on the clinical pathway and shortening the interval.
Patients who had one or more ED visits or hospital admissions between diagnosis and treatment had a longer median TI than those who had neither. These patients might have been too sick to undergo their cancer treatment and require treatment of the illness that prompted the ED visit or admission first.
Strengths and Limitations
To our knowledge, this is the first Ontario-wide population-level study to include such an extensive number of risk factors for a prolonged TI including the assessment of its geographical variation. Previous Ontario studies were either performed on a heterogenous cohort of cancer patients
By partitioning the TI into 2 distinct subintervals, we also identified other potentially modifiable risk factors that were not present on analysis of the overall TI. We used routinely collected health administrative data that allowed us to study the entire esophageal cancer population in Ontario during our time frame. Mandatory submissions from all hospitals in the province to CIHI and NACRS decreases the likelihood of institutions being over-represented. Our definition of the diagnosis date is more refined than in previous studies that used the cancer registry date as the diagnosis date. We used the date of endoscopic biopsy to create a TI definition that was as accurate as possible, and is in line with national efforts to standardize time intervals.
Last, our results are generalizable to other countries because we have found specific patient groups more at risk of longer intervals that transcend geography. Although the magnitude of differences might be specific to Ontario, the wait time variation is unlikely to be on clinical grounds and is generalizable to regions with similar health care models.
Stage was only 54% complete despite capturing data from several databases. A recent study using the same databases
had similar completeness. The sensitivity analysis that included stage showed that stage had no effect on the association of the other variables at the 50th percentile. The unknown group had a longer TI than stage IV patients, but shorter than the others (stage I-III) and are likely to be stage IV patients, receive nonsurgical treatment,
and were equal across all LHINs. There was uncontrolled confounding by using administrative databases. Patient factors not included that might have affected the TI length include a patient's social situation (eg, access to reliable public transportation).
Conclusions
To our knowledge, this population-level study is the first to investigate the esophageal cancer TI length across different LHINs and examine numerous factors. We identified geographical variation despite adjusting for several factors. Patients who are older, more comorbid, or in rural areas are at greater risk for protracted wait times. Future research will be aimed at investigating an association between wait times and survival in our study population.
Conflict of Interest Statement
The authors reported no conflicts of interest.
The Journal policy requires editors and reviewers to disclose conflicts of interest and to decline handling or reviewing manuscripts for which they may have a conflict of interest. The editors and reviewers of this article have no conflicts of interest.
Parts of this material are on the basis of data and information compiled and provided by CIHI, Ontario Health, and Immigration, Refugees and Citizenship Canada current to May 2017. The analyses, conclusions, opinions, statements, results, and views expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred.
Appendix E1
Table E1Health administrative databases used in this study to obtain demographic, disease, billing, and outcomes data
Database
Description
Ontario Cancer Registry
Cancer information, including site, histology, and diagnosis date
Registered Persons Database
Patient demographic data including age, sex, vital status, and dates of last health care encounter
Ontario Health Insurance Plan Database
Physician billing database for inpatient and outpatient services, including diagnoses, services provided, and dates
Discharge Abstract Database
Mandatory submissions from hospitals to the Canadian Institute for Health Information; includes information on hospital admission such as dates and diagnoses
Same Day Surgery Database
Stores information such as date and service for same-day procedures
National Ambulatory Care Reporting Database
Receives mandatory submissions from institutions for visits made to hospital and community ambulatory care centers
PCCF
Converts a patient's postal code into a dissemination area to ascribe certain characteristics to each patient such as rurality and median household income
Activity Level Reporting
Stores information on chemotherapy and radiotherapy dates and services, at regional centers and outreach clinics
LHIN
Stores information including population and number and type of hospitals within each LHIN
Ontario Marginalisation
This database comprises separate elements (eg, material deprivation) and is used in conjunction with PCCF to assign patients a score
IRCC Permanent Resident Database
This includes information on people who applied to land in Ontario such as country of citizenship and date of landing
ICES Physician Database
Demographic information on Ontario physicians including age, specialty, location of work, and year of graduation
PCCF, Postal Code Conversion File; LHIN, Local Health Integration Network; IRCC, Immigration, Refugees, and Citizenship Canada; ICES, Institute for Clinical Evaluative Sciences.
Treatment group: A = endoscopy with or without subsequent treatment; B = chemotherapy only; C = radiotherapy only; D = surgery with or without subsequent treatment; E = chemotherapy and radiotherapy; F = chemotherapy or radiotherapy then surgery; G = chemotherapy and radiotherapy then surgery; and H = other.
P < .0001
P < .0001
P < .0001
P < .0001
P < .0001
P < .0001
A
−16 (−19 to 14)
−18 (−20 to 15)
−16 (−18 to 14)
−16 (−19 to 13)
−16 (−18 to −13)
−12 (−15 to −10)
B
0 (−3 to 3)
1 (−2 to 4)
0 (−3 to 3)
0 (−3 to 3)
0 (−3 to 3)
6 (3-9)
C
−11 (−13 to −8)
−11 (−14 to −9)
−11 (−13 to −8)
−11 (−14 to −9)
−11 (−14 to −9)
−5 (−8 to −2)
D
15 (12-18)
14 (11-17)
15 (11-19)
14 (11-18)
15 (11-19)
17 (13-20)
E
−3 (−5 to −1)
−4 (−6 to −2)
−3 (−5 to −1)
−3 (−5 to −1)
−3 (−5 to −2)
1 (−1 to 3)
F
−2 (−7 to 3)
0 (−5 to 6)
−2 (−7 to 3)
−2 (−7 to 3)
−2 (−6 to 2)
1 (−5 to 6)
G
Referent
Referent
Referent
Referent
Referent
Referent
H
−2 (−5 to 2)
−3 (−7 to 0)
−2 (−6 to 2)
−2 (−6 to 1)
−2 (−6 to 1)
0 (−4 to 4)
Stage
P < .0001
I
3 (−1 to 6)
II
Referent
III
−1 (−4 to 1)
IV
−12 (−15 to −10)
Unknown
−8 (−10 to −5)
All values are difference (95% CI) in treatment interval length at the 50th percentile. SA, Sensitivity analyses; ADG, Aggregate Diagnosis Group; LHIN, Local Health Integration Network.
∗ SA.1 = removal of LHIN.
† SA.2 = removal of immigration.
‡ SA.3 = removal of rurality.
§ SA.4 = removal of immigration and rurality.
‖ SA.5 = removal of treatment group.
¶ SA.6 = addition of stage.
# Treatment group: A = endoscopy with or without subsequent treatment; B = chemotherapy only; C = radiotherapy only; D = surgery with or without subsequent treatment; E = chemotherapy and radiotherapy; F = chemotherapy or radiotherapy then surgery; G = chemotherapy and radiotherapy then surgery; and H = other.
Treatment group: A = endoscopy with or without subsequent treatment; B = chemotherapy only; C = radiotherapy only; D = surgery with or without subsequent treatment; E = chemotherapy and radiotherapy; F = chemotherapy or radiotherapy then surgery; G = chemotherapy and radiotherapy then surgery; and H = other.
P < .0001
P < .0001
P < .0001
P < .0001
P < .0001
P < .0001
A
−16 (−22 to −10)
−15 (−21 to −8)
−15 (−21 to −10)
−16 (−22 to −9)
−15 (−21 to −9)
−11 (−18 to −5)
B
21 (14-27)
19 (13-26)
21 (15-27)
20 (13-27)
20 (13-27)
27 (21-33)
C
6 (0-13)
5 (0-11)
7 (1-12)
6 (0-12)
7 (0-13)
11 (5-17)
D
37 (29-44)
38 (30-46)
37 (29-45)
38 (30-46)
38 (30-47)
34 (27-42)
E
8 (4-12)
7 (3-12)
8 (4-12)
8 (4-12)
8 (4-12)
11 (6-15)
F
0 (−9 to 9)
−3 (−12 to 6)
0 (−8 to 9)
−0 (−9 to 9)
0 (−9 to 9)
−2 (−8 to 5)
G
Referent
Referent
Referent
Referent
Referent
Referent
H
−4 (−10 to 2)
−4 (−10 to 2)
−3 (−9 to 2)
−3 (−9 to 3)
−4 (−11 to 3)
−4 (−10 to 3)
Stage
P < .0001
I
13 (4-22)
II
Referent
III
−4 (−8 to 1)
IV
−16 (−21 to −11)
Unknown
−1 (−5 to 3)
All values are difference (95% CI) in treatment interval length at the 90th percentile. SA, Sensitivity analyses; ADG, Aggregate Diagnosis Group; LHIN, Local Health Integration Network.
∗ SA.1 = removal of LHIN.
† SA.2 = removal of immigration.
‡ SA.3 = removal of rurality.
§ SA.4 = removal of immigration and rurality.
‖ SA.5 = removal of treatment group.
¶ SA.6 = addition of stage.
# Treatment group: A = endoscopy with or without subsequent treatment; B = chemotherapy only; C = radiotherapy only; D = surgery with or without subsequent treatment; E = chemotherapy and radiotherapy; F = chemotherapy or radiotherapy then surgery; G = chemotherapy and radiotherapy then surgery; and H = other.
This study received funding from United Hospitals Kingston Foundation. This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health and the Ministry of Long-Term Care.
00314-X&id=fx1.jpg){kind=link}
00314-X&id=fx2.jpg){kind=link}
00314-X&id=gr1.jpg){kind=link}
00314-X&id=gr2.jpg){kind=link}
00314-X&id=gr3.jpg){kind=link}
00314-X&id=gr4.jpg){kind=link}