Data sources
We conducted a population-based cohort study using health administrative data from Ontario, Canada. These data include information on all contacts with the health care system for all residents of Ontario who qualify for universal single-payer healthcare (> 99% of the population). Hospitalizations are identified using the Canadian Institute for Health Information’s (CIHI) Discharge Abstract Database. Emergency department visits are identified using CIHI’s National Ambulatory Care Reporting System. Outpatient visits are identified through Ontario Health Insurance Plan physician billing data required for physician renumeration. These health administrative databases are linked deterministically using a unique, encrypted health identification number to the Registered Persons Database (demographic characteristics, including start and end dates of provincial health coverage eligibility), Vital Statistics (date of death), and the Ontario Drug Benefit database. The Ontario Drug Benefit database includes data on all outpatient prescriptions among Ontario residents ≥ 65 years of age, including the type of medication prescribed, the dose of the medication, and the total number of days the medication was prescribed for. Data are maintained by ICES via an agreement with the Ontario Ministry of Health (MOH) and Ministry of Long-Term Care (MLTC), with the full database available to researchers in an uncleaned and unedited format [7]. ICES is an independent, non-profit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement.
Study population
Elderly patients ≥ 65 years with IBD were identified from Ontario health administrative data using previously-validated age-specific algorithms [8]. IBD patients who did not turn 65 years old during the study period were excluded. Individuals diagnosed prior to age 65 required at least 5 outpatient visits or hospitalizations with an International Classification of Diseases (ICD)-9 (555, 556) or ICD-10 code (K50, K51) for IBD within 4 years (sensitivity 76.8–92.3%; specificity 96.2–99.1%; positive predictive value 81.4%; negative predictive value 95.0%) and those diagnosed ≥ 65 years additionally required a prescription for an IBD medication (sensitivity 59.3–78.3%; specificity 98.2–99.0%, positive predictive value 71.1%, negative predictive value 98.3%). We differentiated between subtypes Crohn’s disease and ulcerative colitis based on the last 5 of 9 outpatient visits, with an accuracy of 91.1% noted in validation studies [8].
Study design
We conducted a retrospective cohort study of seniors with IBD who were ≥ 65 years of age between July 1, 1997 and June 30, 2017. Patients younger than 65 at diagnosis contributed person-time to the study beginning on their 65th birthday. When patients were diagnosed ≥ 65 years, they began contributing data on their date of IBD diagnosis (the date with their first health care encounter with an associated diagnosis of IBD). Patients were followed until death, migration out of Ontario, or the end of the follow-up period (June 30, 2017).
We compared life expectancy among individuals receiving the following medications: (1) monotherapy with an immunomodulator (azathioprine, 6-mercaptopurine or methotrexate); (2) biologic monotherapy; (3) combination therapy with a biologic and an immunomodulator; (4) mesalamine; (5) systemic steroids; and (6) no therapy. Life expectancy was calculated separately for males and females due to known differences in life expectancy among males and females in the general population. People with IBD ≥ 65 years contributed to the exposure period for the duration they were on a medication based on prescriptions identified in the Ontario Drug Benefit database. Additional file 1: Table S1 provides a list of all medications and their drug identification numbers (DINs) used in this study. They were assigned to an age group (65–69, 70–74, 75–80, 85–90, and 90+) based on their age when each prescription was dispensed. Patients were assigned to a no-therapy group for time periods when they had no active prescriptions. For non-biologic medications, the duration of therapy was determined from the number of days supplied in the Ontario Drug Benefit database. A 30-day window was allowed between the end of prescriptions for non-biologic medication and the next prescription to account for potential non-adherence. We used a 12-week window for biologics typically administered every 8 weeks (infliximab, ustekinumab, and vedolizumab) and a 6-week window for all other biologics (adalimumab, golimumab, certolizumab, and natalizumab). Patients on combination therapy received a prescription for an immunomodulator and a biologic within 3 months of each other. Otherwise, these patients could contribute person-time to biologic or immunomodulator monotherapy groups. All other groups were non-mutually exclusive, meaning that patients could contribute person time to multiple medication groups simultaneously.
Statistical analysis
We calculated sex-specific mortality rates for each medication in each of the following age groups: 65–69, 70–74, 75–79, 80–84, 85–89, and 90+. These mortality rates were then used to calculate life expectancy at age 65 with a period life-table approach, assuming that age- and sex-specific mortality rates for a specific time period remain constant over a person’s life [9]. Hsieh’s modification was used for the last age group [10]. This approach applies age- and sex-specific mortality rates to a hypothetical cohort of individuals, beginning at age 65, and assumes that these mortality rates remain constant over time [9]. This hypothetical cohort has no underlying characteristics prior to being subjected to the mortality rates associated with each medication. Period life expectancy transforms mortality rates into a measure of longevity, increasing interpretability [11].
We determined the differences between the life expectancy at age 65, comparing patients in each medication group, stratified by sex. The impact of medication on life expectancy was explored in (1) people with IBD; (2) people with Crohn’s disease; and (3) people with ulcerative colitis. Differences were considered statistically significant when the 95% confidence interval (CI) did not include the null value of 0. We plotted survival curves based on the expected proportion of people remaining alive in each age interval for each medication group, beginning at 65 years of age.
We also used age- and sex-specific mortality rates and the age- and sex-specific distribution of prevalent IBD cases ≥ 65 years of age on July 1, 2016 to determine age-standardized sex-specific mortality rates for each medication group. Due to the small number of deaths in some medication groups, mortality rates are only presented for overall IBD.
Statistical analyses were conducted using SAS Version 9.4 (SAS Institute, Cary, NC, U.S.A.).
Sensitivity analysis
To estimate the influence of post-operative mortality on life expectancy estimates, we conducted a sensitivity analysis where individuals requiring surgery were censored at the time of surgery. We identified individuals requiring surgery using previously validated procedural codes for small bowel resection (Crohn’s disease only) and colectomy (Crohn’s disease and ulcerative colitis) (Additional file 1: Table S2) [12, 13]. We repeated disease-specific life expectancy calculations and determined the difference in life expectancy across medication groups. All analyses were stratified by IBD type due to differences in surgical procedures across IBD types.