Review 2: "Estimating The Uncertain Effect of the COVID Pandemic on Drug Overdoses"

The preprint titled “Estimating the Uncertain Effect of the COVID Pandemic on Drug Overdoses” analyzes national and state-level CDC drug overdose mortality data and unique sources of emergency department (ED) visit data to evaluate the claim that drug overdoses rose in 2020- 2021 due to the COVID-19 pandemic. The authors construct a counterfactual of expected overdose deaths during the pandemic period, using observed pre-pandemic trends, and compare it to the observed trends during the first pandemic year. The authors do not find support for the pandemic driving drug overdose mortality increases and conclude that these increases are likely explained by a continuation of pre-pandemic trends in synthetic opioid deaths. Overall, the results are potentially informative.

The strengths of this work include using more comprehensive data over a longer time period than the existing research on the topic, rigorous methodology, and well-structured and clear writing. The authors do an outstanding job positioning their research paper within the current literature and contextualize their research question and methodological approach well.

There are several potential limitations and suggestions for the authors to consider as they prepare to publish their manuscript. First, a more detailed explanation of how model fit was assessed and why the particular polynomial was chosen would be beneficial. The current model predicts a dramatic increase in overdose mortality in 2020, which can be hard to believe (especially the 2015 model for all overdoses). Would a different polynomial predict a lower slope for 2020?

Second, I wonder how the authors might interpret the positive difference between the actual and predicted mortality between April and November 2020 (Figure 1, Panels B and C)? Could this indicate a pandemic effect? This should be discussed.

Third, I believe that the outlying states may be informative and thus should not be omitted from Figure 2. Additionally, a U.S. map with states color-coded according to actual-minus-predicted may be a better visualization. Although the authors did not visually identify a pattern, state heterogeneity may be meaningful in that the pandemic may have played a greater role in some states than others. If the same polynomial model was fit on each state, perhaps model fit could be inappropriate in some states and state-specific models might be beneficial.

Fourth, the conclusion at the end of the Discussion claiming that the data is unlikely to support a more definitive message than in the preprint may be an overstatement. The authors could instead discuss how future research, for example using the CDC multiple-cause-of-death data, could extend their work and potentially elucidate the role of the pandemic in drug overdose mortality.

Additionally, I would recommend adding to the limitations that counts of synthetic opioids may be measured with error given that the percentage of unclassified drug overdoses decreases over time.

In conclusion, the preprint provides an important analysis that challenges the prevailing narrative that the COVID-19 pandemic has played a major role in the growth of drug overdose mortality. The findings should draw public health professionals’ and academics’ attention to the need to continue addressing the opioid epidemic, particularly synthetic opioid use and overdoses, through targeted policies and interventions, rather than hope that drug overdoses will naturally decline as the pandemic wanes.