==== Front Emerg Infect Dis Emerg Infect Dis EID Emerging Infectious Diseases 1080-6040 1080-6059 Centers for Disease Control and Prevention 23-0617 10.3201/eid2907.230617 Letters to the Editor Letters to the Editor Challenges in Forecasting Antimicrobial Resistance (Response) Challenges in Forecasting Antimicrobial Resistance (Response) Challenges in Forecasting Antimicrobial Resistance (Response) Pei Sen Columbia University, New York, New York, USA Address for correspondence: Sen Pei, Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, NY 10032, USA; email: sp3449@cumc.columbia.edu 7 2023 29 7 14961497 2023 https://creativecommons.org/licenses/by/4.0/ Emerging Infectious Diseases is a publication of the U.S. Government. This publication is in the public domain and is therefore without copyright. All text from this work may be reprinted freely. Use of these materials should be properly cited. AldeyabMA, LattyakWJ. Challenges in forecasting antimicrobial resistance.Emerg Infect Dis. 2023;29 :1496.Keywords: health-associated infection real-time forecasting antimicrobial resistance bacteria United States ==== Body pmcIn Response: Real-time evaluation of predictive models for antimicrobial resistance (AMR) is critical for real-world applications, as indicated in our recently published article (1). Aldeyab and Lattyak introduced a threshold-logistic regression model that links antimicrobial drug use to AMR prevalence in hospital settings (2). The authors advocate implementing and testing this model in hospitals to assess operational utility. I agree that this is a practical starting point to challenge time-series model use for real-time AMR predictions. Most time-series models have been validated in retrospective analyses. Translational research is needed to promote the use of those models for real-world AMR control. The authors mention several practical considerations when applying time-series models in real time, including stationarity of both predictor and target variables and criteria for model recalibration. Evaluating methods to address those issues is crucial to achieve desirable performance in hospital settings. In addition to those technical challenges, several broader questions remain regarding model design and utility. First, how much AMR prevalence variation can be explained by antimicrobial drug use? Are there other essential factors (e.g., community introduction) that should be included in the model? Second, how will healthcare providers and hospitals use AMR forecasts? What policies will be informed by forecasts, and what are the downstream effects? Answers to those questions will help determine the eventual real-world utility of predictive models. Evaluating real-time AMR prediction is a complicated task. By drawing experience from computer vision (3) and forecasts for other infectious diseases (4–6), open-access challenges with transparent and fair evaluation methods run in a common task framework (7) can substantially stimulate the advance of predictive methods and might produce robust application models. Such collaborative efforts are needed to evaluate existing methods, identify difficulties and solutions, and push the operational use of AMR predictive models forward. Acknowledgments This work was supported by the US Centers for Disease Control and Prevention, grant nos. U01CK000592 and 75D30122C14289. Suggested citation for this article: Pei S. Challenges in forecasting antimicrobial resistance. Emerg Infect Dis. 2023 Jul [date cited]. https://doi.org/10.3201/eid2907.230617 ==== Refs References 1. Pei S, Blumberg S, Vega JC, Robin T, Zhang Y, Medford RJ, et al. ; CDC MIND-Healthcare Program. Challenges in forecasting antimicrobial resistance. Emerg Infect Dis. 2023;29 :679–85. 10.3201/eid2904.221552 36958029 2. Aldeyab MA, Lattyak WJ. Challenges in forecasting antimicrobial resistance. Emerg Infect Dis. 2023 Jul [date cited]. 3. Russakovsky O, Deng J, Su H, Krause J, Satheesh S, Ma S, et al. ImageNet large scale visual recognition challenge. Int J Comput Vis. 2015;115 :211–52. 10.1007/s11263-015-0816-y 4. Cramer EY, Ray EL, Lopez VK, Bracher J, Brennen A, Castro Rivadeneira AJ, et al. Evaluation of individual and ensemble probabilistic forecasts of COVID-19 mortality in the United States. Proc Natl Acad Sci U S A. 2022;119 :e2113561119. 10.1073/pnas.2113561119 35394862 5. Reich NG, Brooks LC, Fox SJ, Kandula S, McGowan CJ, Moore E, et al. A collaborative multiyear, multimodel assessment of seasonal influenza forecasting in the United States. Proc Natl Acad Sci U S A. 2019;116 :3146–54. 10.1073/pnas.1812594116 30647115 6. Johansson MA, Apfeldorf KM, Dobson S, Devita J, Buczak AL, Baugher B, et al. An open challenge to advance probabilistic forecasting for dengue epidemics. Proc Natl Acad Sci U S A. 2019;116 :24268–74. 10.1073/pnas.1909865116 31712420 7. Donoho D. 50 years of data science. J Comput Graph Stat. 2017;26 :745–66. 10.1080/10618600.2017.1384734