Study setting and data sources
We conducted a retrospective observational cohort study in Alberta, a Canadian province with a single publicly funded universal healthcare system. Six administrative data sources housed within Alberta Health Services, the corporation that manages most of the healthcare services in the province, were used for this study along with the utilization of diagnostic codes from the International Statistical Classification of Diseases and Related Health Problems, 10th revision, with Canadian enhancements (ICD-10-CA). The six administrative databases consulted were: 1) The Discharge Abstract Database (DAD), which tracks all acute care inpatient discharges from Alberta hospitals, 2) The National Ambulatory Care Reporting System (NACRS) that tracks all Emergency Department (ED) visits, 3) The Practitioner Claims database, which tracks all health service claims submitted for payment by healthcare providers using the International Statistical Classification of Diseases and Related Health Problems, 9th revision (ICD-9), 4) The National Rehabilitation System (NRS) database that tracks inpatient rehabilitation stays, 5) The Provincial Registry that identifies date of death and Alberta Health Care Insurance Plan (AHCIP) coverage for Alberta Residents, and 6) The Vital Statistics database, which also tracks dates of death of Alberta residents. This study received ethics approval and a waiver of informed consent from the University of Alberta Health Research Ethics Board (Pro00097538).
We created a prevalent SCI cohort by looking for cases between April 1, 2002 and December 31, 2017 who were over 18 years of age, alive, and residents of Alberta on January 1, 2018 using a validated administrative case definition with some modifications . This population was then followed for outcomes until December 31, 2019. Two main cohorts were created, one for traumatic etiologies (i.e., TSCI cohort) and one for non-traumatic etiologies (i.e., NTSCI cohort) as identified using ICD-10-CA diagnosis codes (Supplementary Table 1). In cases where individuals met both TSCI and NTSCI case definitions they were classified according to which definition they met first. Individuals were excluded from the study if they were younger than 18 on January 1, 2018, if they died before this date, were not Alberta residents (based on the Provincial Registry), or if they did not have provincial healthcare coverage for any time during the follow-up period. Prevalence was calculated as of January 1, 2018. Mortality rates were estimated based on deaths during the two-year follow-up period from January 1, 2018 to December 31, 2019. Those who died during the follow-up period were included for prevalence and mortality rate estimates, but excluded from the follow-up outcomes (visit and complication rates).
Individuals were flagged as First Nations or non-First Nations by the Alberta Ministry of Health based on the First Nations Status Registry, which includes all AHCIP registrants associated with a First Nations group number since 1983 and individuals who register under a main AHCIP account for which the main registrant had a First Nations group number. First Nations status is given to all individuals who are on the First Nations Status Registry at the mid-year registry file (June 30 each year).
The Charlson comorbidity index was calculated using Deyo weights [17, 18] by reviewing 2 years prior to the index date in DAD, NACRS, and Practitioner Claims where comorbidities identified in Claims required at least two visits with the appropriate diagnostic code. All individuals in the cohort were assigned 2 points in the Charlson index for the comorbidity of SCI. Location of residence was based on the Alberta Health Services classification of sub-local geographic areas as Metropolitan, Urban, or Rural.
We quantified healthcare utilization and the occurrence of common SCI complications during the 2-year follow-up period from January 1, 2018 to December 31, 2019. Healthcare utilization measures included the number of visits to a general practitioner (GP), specialist, ED, inpatient visits, and inpatient days. GP and specialist visits were based on unique physician-days from the Practitioner Claims database in a community setting (physician office, clinic, or long-term care delivery site). Specialist visits were limited to those with SCI-related practitioner specialties and included: Physical Medicine and Rehabilitation, Orthopedic Surgery, Urology, Cardiology, Respiratory Medicine, Nephrology, General Surgery, Infectious Diseases, Neurology, Neurosurgery, Plastic Surgery, Gastroenterology, and Podiatry. The occurrence of SCI-related complications were quantified for community, emergency, and inpatient settings. Complications were grouped as pulmonary, gastrointestinal (GI), genitourinary (GU), skin, cardiovascular/autonomic, and other (Supplementary Table 2).
We compared baseline characteristics between First Nations and non-First Nations groups. TSCI and NTSCI were analyzed separately, using a t-test for continuous variables and a chi-square test for categorical variables to compare First Nations versus non-First Nations populations. Prevalence ratios and mortality rate ratios were estimated using a log-binomial model. The conditional mean is the median number of visits (Table 2) or complications (Table 3) of those who had at least one visit/complication. Rate ratios and associated p values comparing First Nations and non-First Nations healthcare utilization and the occurrence of complications were estimated using a negative binomial model. Adjusted rate ratios were calculated by including the following predictors in the model: injury type (TSCI/NTSCI), residence (metropolitan/urban/rural), sex, age (under 40/40–65/65+years), level of injury (paraplegic/quadriplegic/unknown) and Charlson index (Mild: ≤2/Moderate: 2–4/Severe: ≥5).
A secondary analysis using matched First Nations and non-First Nations cohorts was performed. Non-First Nations controls were matched to the First Nations cohort with a ratio of 2:1 using exact match on sex, residence location (metropolitan/urban/rural), injury type (TSCI/NTSCI), injury level (tetraplegia/paraplegia/unknown), age +/−5 years and Charlson index +/−2. A greedy matching algorithm was used so those with the closest matches were selected. Chi-square testing was used to determine significant differences between First Nations and non-First Nations cohorts when data values were at least 5 and a Fisher’s exact test was used when values were less than 5. All statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC).