When Denominators Collapse: Institutional Shock and the Measurement of Treatment Need

• In Event: Substance Use Research in Times of Social Transformation

Abstract

Substance use disorder (SUD) is a high-priority public health issue due to its substantial social, economic, and health consequences. Yet measuring community-level treatment need remains conceptually and empirically challenging. Survey-based estimates, while nationally representative, are vulnerable to recall, social desirability, and selection biases—limitations that are particularly acute for stigmatized behaviors such as substance use. Administrative data, including Emergency Department (ED) visit claims, provide a more objective alternative and are frequently operationalized as proportion-based “burden” measures (e.g., drug-related visits per 1,000 total ED visits). These measures are widely used in surveillance and research as proxies for local treatment need.
We argue that this approach relies on a strong assumption: that variation in drug-related ED burden reflects changes in underlying need rather than shifts in the broader healthcare system. Using data from hospitals in 10 states, we show that this assumption breaks down during periods of systemic disruption. During the COVID-19 pandemic, overall ED utilization declined sharply, while certain drug-related visits, particularly poisonings, remained relatively stable. As a result, burden measures shifted dramatically relative to prior years, producing changes that risk generating spurious inferences when used uncritically as outcomes or predictors in models assessing treatment need.
To address this construct instability, we reconceptualize drug-related ED burden as a function of both treatment need and institutional constraints. We propose counterfactual denominator approaches to adjust for disruptions in utilization patterns, while cautioning that such adjustments may understate the pandemic’s direct effects on people with SUD. Triangulation with overdose mortality data further reveals substantial increases in unmet need during 2020–2021. We conclude by outlining best practices for researchers using administrative burden measures in periods of structural disruption.

Authors

• Mireia Triguero Roura, Rutgers University

• Aabha Vora, Center for Pharmacoepidemiology and Treatment Science (PETS), Rutgers Institute for Health

• Pia M Mauro, Department of Biostatistics and Epidemiology, Rutgers School of Public Health/Center for Pharmacoepidemiology and Treatment Science (PETS), Rutgers Institute for Health