This was a population-based cohort study in Alberta, Canada, a province of ~ 4 million people with a single healthcare system, Alberta Health Services (AHS), that included all patients who underwent CIED implantation, including those who subsequently developed a SSI. Patients were identified using a centralized CIED database (Paceart™) and these data were linked to AHS health administrative data that captures patient comorbidities and healthcare costs.

We identified a cohort of adult patients (i.e., age ≥ 18 years) who underwent a new CIED implantation (including pacemaker (PM), implantable cardioverter defibrillator (ICD), or cardiac resynchronization therapy (CRT)) or generator replacement between January 1, 2011, and December 31, 2019. The infection group was identified using Discharge Abstract Database (DAD) and International classification of Diseases 10th revision Canada (ICD-10-CA) codes (T827, T857, I330, I339, I38, I398, L0330, L0339, L038, L039) that were previously validated as able to identify complex CIED SSIs (i.e. deep and organ space but not superficial SSIs) with sensitivity and specificity > 90% [12]. All infections that occurred within one year from the index date of CIED implantation were tracked. Patients who had their device implanted outside of Alberta were excluded from the study.

CIED implantations were identified using the Paceart™ database which contains all device-related clinical encounters for patients followed within the province of Alberta, Canada. Paceart™ contains information regarding indications for device implantation, type of device, date of operation and basic demographic information including sex. Repeat procedures within a two-year period from the index surgical date were censored. This avoided double counting patient encounters months later as only the initial implant was counted as an index procedural date.

AHS analytics data which provides healthcare information on all Alberta residents with an Alberta Health Care Insurance Plan (> 99% of provincial coverage) was used. This data repository provided records from both DAD which was used for tracking infection cases as described above. Both DAD and the national ambulatory care reporting system (NACRS, which contains data for hospital based and community based ambulatory care including day surgery, outpatient and community ambulatory clinics and emergency department visits) were used to obtain demographic information about the patient cohort including comorbidities which were collected using a two-year retrospective review. Rural versus urban location and the Pampalon Deprivation Index was collected as well. The Pampalon Deprivation Index is a composite index using Canadian census data in order to create a measure of socioeconomic disparity, [13] and urban versus rural residence data was determined from patient postal codes.

A mix of gross costing and micro-costing was used to assess economic burden. Gross costing was used where micro-costing was not available (i.e. for any inpatient encounter outside of Calgary or Edmonton or any outpatient encounter). Gross costing is when aggregate resource use items are identified and expenditure data is collected at the organization level. Gross costs were identified from DAD and NACRS using resource intensity weights (RIW) for any healthcare encounter and were multiplied by the cost of a standard hospital stay (CSHS) in Alberta by year from the Canadian Institute for Health Information [14].

Micro-costing data was available from AHS corporate finance for all inpatient admissions in Calgary and Edmonton. This is considered the gold standard of costing data. Patient level costs are provided and each component of resource use is estimated and a unit cost derived providing the most specific costing information possible [15]. This data includes the specific costs for each patient for nursing, operating room expenses, patient supplies, in-hospital drug use, allied healthcare, diagnostic imaging and testing (such as echocardiograms), laboratory testing, equipment costs (including equipment for specialized medical procedures such as hemodialysis), organization supports such as utilities, and housekeeping.

The primary outcome was mean 12-month cumulative healthcare costs for all patients who had a CIED implant or generator replacement. Costs were also stratified into inpatient and outpatient costs. Additionally, healthcare utilization including number of inpatient and outpatient visits to any AHS facility were considered as well as LOS in hospital over 12 months. All outcomes were compared amongst those who did and did not develop a complex CIED SSI. All costs were inflated to 2022 Canadian dollars. The perspective taken was that of the pubic healthcare payer and therefore, patient-borne costs such as outpatient antibiotic prescriptions, were not included. Physician claims were not accessed for this work and thus were not included.

Descriptive statistics, such as frequencies and percentages for categorical variables and means with standard deviation (SD) for continuous variables were used to describe baseline characteristics of patients with and without infections. For our primary method, we assessed the total mean costs, inpatient costs, and outpatient costs for all patients at one year after the index date of their implantation. Number of inpatient admissions, outpatient visits and total LOS were analyzed for all patients over the subsequent one year.

In order to adjust for covariates, a propensity score match was conducted for comparing CIED patients who developed a complex SSI within one year of implantation to those who did not develop an infection. Propensity scores were estimated using a logistic regression model with observed baseline characteristics of age, sex, Elixhauser comorbidity index [16] and device type (i.e., PM, ICD, CRT). Matching was performed using the greedy nearest-neighbor methods without replacement and a caliper of 0.2 of the standard deviation of the log-odds of the propensity score. To assess the balance in baseline characteristics, standardized mean differences in proportions between patients with and without complex SSIs for each covariate were calculated after matching. A weighted standardized difference < 10% indicated good balance and acceptable bias. Incremental healthcare utilization (i.e. costs, number of admissions/visits and LOS) were calculated to determine the effect of infection on these outcomes.

As a sensitivity analysis we conducted generalized linear models (GLM) to identify the relationship between the outcomes and infection. We used GLM with Gamma distribution and a log link function for the costs, GLM with Poisson distribution and a log link for number of hospital admissions, LOS, and number of outpatient visits.

All statistical analyses were performed using R Statistical Software (Version 1.4.0). This research was approved by the University of Calgary Health Research Ethics Board (REB20-2186).