Examining the Spatial Varying Effects of Sociodemographic Factors on Adult Cochlear Implantation Using Geographically Weighted Poisson Regression

Abstract

Objective To (i) demonstrate the utility of geographically weighted Poisson regression (GWPR) in describing geographical patterns of adult cochlear implant (CI) incidence in relation to sociodemographic factors in a publicly funded healthcare system, and (ii) compare Poisson regression and GWPR to fit the aforementioned relationship. Study Design Retrospective study of provincial CI Program database. Setting Academic hospital. Patients Adults 18 years or older who received a CI from 2020 to 2021. Intervention(s) Cochlear implant. Main Outcome Measure(s) CI incidence based on income level, education attainment, age at implantation, and distance from center, and spatial autocorrelation across census metropolitan areas. Results Adult CI incidence varied spatially across Ontario (Moran's I = 0.04, p < 0.05). Poisson regression demonstrated positive associations between implantation and lower income level (coefficient = 0.0284, p < 0.05) and younger age (coefficient = 0.1075, p < 0.01), and a negative association with distance to CI center (coefficient = −0.0060, p < 0.01). Spatial autocorrelation was significant in Poisson model (Moran's I = 0.13, p < 0.05). GWPR accounted for spatial differences (Moran's I = 0.24, p < 0.690), and similar associations to Poisson were observed. GWPR further identified clusters of implantation in South Central census metropolitan areas with higher education attainment. Conclusions Adult CI incidence demonstrated a nonstationary relationship between implantation and the studied sociodemographic factors. GWPR performed better than Poisson regression in accounting for these local spatial variations. These results support the development of targeted interventions to improve access and utilization to CIs in a publicly funded healthcare system.