Estimating the Number of Opioid Overdoses in British Columbia Using Relational Evidence with Tree Structure

TL;DR

This paper introduces two methods, a weighted multiplier and a Bayesian hierarchical model, to estimate the total number of opioid overdoses in British Columbia using linked, tree-structured health data.

Contribution

It compares and evaluates the effectiveness of a novel Bayesian hierarchical approach against a traditional multiplier method for population size estimation in a complex health data setting.

Findings

Bayesian model provides more accurate estimates.

Weighted multiplier method is simpler but less precise.

Both methods help inform public health policy.

Abstract

In many fields, populations of interest are hidden from data for a variety of reasons, though their magnitude remains important in determining resource allocation and appropriate policy. In public health and epidemiology, linkages or relationships between sources of data may exist due to intake structure of care providers, referrals, or other related health programming. These relationships often admit a tree structure, with the target population represented by the root, and paths from root-to-leaf representing pathways of care after a health event. In the Canadian province of British Columbia (BC), significant efforts have been made in creating an opioid overdose cohort, a tree-like linked data structure which tracks the movement of individuals along pathways of care after an overdose. In this application, the root node represents the target population, the total number of overdose events occurring in BC during the specified time period. We compare and contrast two methods of estimating the target population size - a weighted multiplier method based on back-calculating estimates from a number of paths and combining these estimates via a variance-minimizing weighted mean, and a fully Bayesian hierarchical model.