This study explores provincial epidemiological trends of adolescent idiopathic scoliosis (AIS) over 17-years, by validating a population-based health administrative data (HAD) algorithm for patient identification. This case ascertainment validation study developed 93 algorithms to identify 10 & 17-year-olds with AIS, using a combination of diagnostic, specialty, and/or fee codes related to scoliosis, gathered from physician billings, emergency room records, and discharge data over various look-back periods. A validation cohort was established to evaluate algorithm performance, comprised of 2732 patients with confirmed AIS (positive reference) and 49,049 youth without AIS (negative reference), approximating the population prevalence reported in the literature. Performance was evaluated based on measures of sensitivity, specificity, and positive and negative predictive value (PPV and NPV) with 95% confidence intervals. The top performing algorithm was used to derive a provincial cohort of youth diagnosed with AIS between 2005–21 to estimate annual incidence (2005–21) and prevalence (2012–21). The age/sex-adjusted estimates were directly standardized to the 2016 Census population and rates over time were compared using Chi-squared analyses, with p<0.05 significant. The algorithm selected was ‘two encounters in two years associated with a scoliosis diagnosis code’ with sensitivity: 83.1% (81.6–84.5), specificity: 99.3% (99.2–99.3), PPV: 86.3 (85.0–87.6), and NPV: 99.1 (99–99.1) (Figure 1). The algorithm derived cohort included 27 125 youth, 18 358 females (67%) with a mean age 13.9±1.8 years. Age/sex standardized AIS incidence ranged from 108.8 (2006) up to 147.3 (2021) per 100 000 persons. Age/sex-standardized prevalence ranged from 491.3 (2012) up to 541.9 (2021) per 100 000 persons. Rates over the study period were relatively stable (p=0.72 and 0.49, respectively). Identification of youth with AIS is possible using a HAD algorithm, with strong discriminative ability. Provincial epidemiological trends are relatively stable between 2005–21. Our results support this being a promising avenue for conducting population-level AIS surveillance and longitudinal outcome studies in the future. For any figures or tables, please contact the authors directly.
Dermott et al. (Wed,) studied this question.