Wait times
Healthcare costs for medically indicated procedures for the 13 million Ontarians (including costs of surgery and hospitalization) are covered by a government single payer, the Ministry of Health and Long-Term Care. Little surgical care is provided outside the province.
The Ontario Wait Times (WTIS) database contains patient level wait times data from 98% of Ontario hospitals. Surgical wait times are calculated as the time from decision to treat until surgery. The database also contains demographic data, postal codes, and information related to the type of surgery. The study included all adult patients with complete data in the WTIS database from 2006–2015. We excluded patients < 18 years old at the time of surgery, missing wait time data, and missing or non-Ontario health card number.
Access targets for urgent and nonemergent care were previously established through expert consensus. Surgeons assign patients a priority 1 through 4 (P1-P4); with access targets ranging between 7–182 days [20,21,22,23,24]. Periods of delay due to “patient related” reasons, such as a change in medical status or patient rescheduling, are recorded as Days Affecting Readiness to Treatment (DART), and the calculated waiting time is adjusted for any DART days. During the study period, hospitals received extra targeted funds to increase surgical volume and reduce wait times for hip and knee replacements, cataracts, oncology procedures and cardiac surgery.
The following variables were obtained: Ontario health card number (unique individual identifier), birth year, sex, postal code, hospital, surgical specialty, procedure type, priority level, access target and DART reason.
Socioeconomic status and rurality
The 2015 Postal Code Conversion File (PCCF, obtained from Statistics Canada) contains all the postal codes in Canada. Postal codes are combined by Statistics Canada into larger census groupings called Dissemination Areas (DAs). The PCCF also includes the relative rurality of each census subdivision. The Material and Social Deprivation Index (MSDI), developed by the Institut national de santé publique du Québec, uses Statistics Canada census data to calculate both material and social deprivation scores for each DA [25,26,27]. The MSDI separates DAs into deprivation quintiles containing 20% of Ontario’s population based on six indicators: average income, proportion of individuals without a high school diploma, proportion of employed individuals (material deprivation), and proportion of individuals living alone, proportion of lone parent families, and the proportion of separated, divorced or widowed individuals (social deprivation). We used the 2011 MSDI, the most recent version centered on the study period.
The PCCF was linked to the MSDI to obtain the material deprivation quintiles of each postal code in Ontario based on their DA. We merged the resulting database with the WTIS to link individual patients to their deprivation quintiles based on postal code. Rurality was classified as metropolitan areas (population at least 100,000), census agglomeration areas (CA) (two levels, each with population at least 10,000), or areas with strong to no metropolitan influence based on the proportion of commuting population. This measure has been previously used to examine correlations with health status [28].
The primary question was whether patients’ material deprivation levels were associated with surgical waiting times, controlling for the number of visits, age at procedure, material deprivation, rural status, sex, surgical specialty, priority level, and year of procedure. Social deprivation was found to have nearly the same distribution as material deprivation and so was not included in further analysis. Since patients underwent multiple procedures a fixed-effects, repeated measures least squares regression model using the Mixed Procedure in SAS was used. We used an Autoregressive AR(1) correlation matrix to model the correlation of the repeated waiting times. The correlation for this matrix decreases the further apart the waiting time within an individual. This structure made the most clinical sense. There was no concern with power of the study since the sample size was close to 4.5 million.
No individual patients were identified in this study. All data were presented with patient identifiers removed. The SAS software package V9.4 and Stata V15 was used for statistical analysis, and all data was analyzed by the authors. The research protocol was approved by the SickKids REB.