Under‐reporting of COVID‐19 in the Northern Health Authority region of British Columbia

Abstract

Asymptomatic and pauci‐symptomatic presentations of COVID‐19 along with restrictive testing protocols result in undetected COVID‐19 cases. Estimating undetected cases is crucial to understanding the true severity of the outbreak. We introduce a new hierarchical disease dynamics model based on the N ‐mixtures hidden population framework. The new models make use of three sets of disease count data per region: reported cases, recoveries and deaths. Treating the first two as under‐counted through binomial thinning, we model the true population state at each time point by partitioning the diseased population into the active, recovered and died categories. Both domestic spread and imported cases are considered. These models are applied to estimate the level of under‐reporting of COVID‐19 in the Northern Health Authority region of British Columbia, Canada, during 30 weeks of the provincial recovery plan. Parameter covariates are easily implemented and used to improve model estimates. We compare two distinct methods of model‐fitting for this case study: (1) maximum likelihood estimation, and (2) Bayesian Markov chain Monte Carlo. The two methods agreed exactly in their estimates of under‐reporting rate. When accounting for changes in weekly testing volumes, we found under‐reporting rates varying from 60.2% to 84.2%.

Multi-site disease analytics with applications to estimating COVID-19 undetected cases in Canada

“…We have built on the Parker et al (2021) model to allow for multiple sites, which in our B.C. case study corresponds to the five Health Authority Regions of B.C., and which in our Canada-wide case study corresponds to the provinces and territories of Canada.…”

“…We compare the Northern Health Authority Region results from the single-site model of Parker et al (2021) with results from our multi-site model M2 (Figure 7). The single-site results were obtained using maximum likelihood methods, and so the variability shown indicates the 95% confidence interval, whereas the multi-site model uses Bayesian MCMC to obtain 95% credible intervals.…”

“…The blue band shows the 95% credible interval for total active cases as estimated using model M2. The green bands show the 95% confidence intervals for total active cases as estimated using the single site model from Parker et al (2021), denoted Single Site. The black dotted line shows the newly detected active cases each week.…”

“…These methods include meta-analyses of asymptomatic prevalence (He et al, 2021; Alene et al, 2021), extensions to susceptible-infectious-recovered type modelling (Li et al, 2021; Subramanian et al, 2021; Huo et al, 2021), and seroprevalence studies (Bendavid et al, 2021; Saeed et al, 2021; Halili et al, 2022). Alternative models include integer-autoregressive models (Fernández-Fontelo et al, 2020) and case fatality rate (CFR) models (Dougherty et al, 2021), and estimating under-reporting using discrete count models (Parker et al, 2021).…”

“…Our multi-site hidden Markov model has several new mathematical contributions in comparison with the single-site model (Parker et al, 2021): We provide a multi-site framework for disease analytics, we add a saturation effect to model limitations on disease spread, and we model the disease spread rate separately for the detected and the undetected disease cases. The latent states in our model are identifiable due to parametric assumptions.…”