publications

2021

1. CODS-COMAD

   Adaptive COVID-19 Forecasting via Bayesian Optimization

   Nayana Bannur, Harsh Maheshwari, Sansiddh Jain, Shreyas Shetty, Srujana Merugu, and Alpan Raval

   In 8th ACM IKDD CODS and 26th COMAD

   Abs medRxiv HTML

   Accurate forecasts of infections for localized regions are valuable for policy making and medical capacity planning. Existing compartmental and agent-based models for epidemiological forecasting employ static parameter choices and cannot be readily contextualized, while adaptive solutions focus primarily on the reproduction number. The current work proposes a novel model-agnostic Bayesian optimization approach for learning model parameters from observed data that generalizes to multiple application-specific fidelity criteria. Empirical results point to the efficacy of the proposed method with SEIR-like models on COVID-19 case forecasting tasks. A city-level forecasting system based on this method is being used for COVID-19 response in a few impacted Indian cities.

2. ICLR

   Managing an SIR Epidemic System via Optimal Control of Transmission Rate

   Harsh Maheshwari, Shreyas Shetty, Nayana Bannur, and Srujana Merugu

   In ICLR 2021 Workshop on Machine Learning for Preventing and Combating Pandemics

   Abs

   Shaping an epidemic with adaptive contact restriction policies has been the holy grail during the COVID-19 pandemic. In this paper, we explore the problem of determining the optimal control policy for transmission rate assuming SIR dynamics. We first demonstrate that the SIR model with infectious patients and susceptible contacts (i.e., product of transmission rate and susceptible population) interpreted as predators and prey respectively maps to a Lotka-Volterra (LV) predator-prey model. The modified SIR system (LVSIR) has a stable equilibrium point, an “energy” conservation property, and exhibits bounded cyclic behaviour similar to an LV system. We exploit this mapping using a control-Lyapunov approach to design a novel, practical control policy (CoSIR) that nudges the SIR model to the desired equilibrium. Empirical comparison with periodic lockdowns on simulated and real COVID-19 data demonstrates the efficacy and adaptability of our approach.

3. ICLR

   Interpretability of Epidemiological Models: The Curse of Non-Identifiability

   Ayush Deva, Siddhant Shingi, Avtansh Tiwari, Nayana Bannur, Sansiddh Jain, Jerome White, Alpan Raval, and Srujana Merugu

   In ICLR 2021 Workshop on AI for Public Health

   Abs arXiv PDF Poster

   Interpretability of epidemiological models is a key consideration, especially when these models are used in a public health setting. Interpretability is strongly linked to the identifiability of the underlying model parameters, i.e., the ability to estimate parameter values with high confidence given observations. In this paper, we define three separate notions of identifiability that explore the different roles played by the model definition, the loss function, the fitting methodology, and the quality and quantity of data. We define an epidemiological compartmental model framework in which we highlight these non-identifiability issues and their mitigation.

4. Evaluation of individual and ensemble probabilistic forecasts of COVID-19 mortality in the US

   Estee Y Cramer, Evan L Ray, Velma K Lopez, Johannes Bracher, Andrea Brennen, Alvaro J Castro Rivadeneira, Aaron Gerding, Tilmann Gneiting, Katie H House, Yuxin Huang, Dasuni Jayawardena, Abdul H Kanji, Ayush Khandelwal, Khoa Le, Anja Mühlemann, Jarad Niemi, Apurv Shah, Ariane Stark, Yijin Wang, Nutcha Wattanachit, Martha W Zorn, Youyang Gu, Sansiddh Jain, Nayana Bannur, Ayush Deva, Mihir Kulkarni, Srujana Merugu, Alpan Raval, Siddhant Shingi, Avtansh Tiwari, Jerome White, et al.

   medRxiv

   Abs medRxiv HTML

   Short-term probabilistic forecasts of the trajectory of the COVID-19 pandemic in the United States have served as a visible and important communication channel between the scientific modeling community and both the general public and decision-makers. Forecasting models provide specific, quantitative, and evaluable predictions that inform short-term decisions such as healthcare staffing needs, school closures, and allocation of medical supplies. In 2020, the COVID-19 Forecast Hub (https://covid19forecasthub.org/) collected, disseminated, and synthesized hundreds of thousands of specific predictions from more than 50 different academic, industry, and independent research groups. This manuscript systematically evaluates 23 models that regularly submitted forecasts of reported weekly incident COVID-19 mortality counts in the US at the state and national level. One of these models was a multi-model ensemble that combined all available forecasts each week. The performance of individual models showed high variability across time, geospatial units, and forecast horizons. Half of the models evaluated showed better accuracy than a naive baseline model. In combining the forecasts from all teams, the ensemble showed the best overall probabilistic accuracy of any model. Forecast accuracy degraded as models made predictions farther into the future, with probabilistic accuracy at a 20-week horizon more than 5 times worse than when predicting at a 1-week horizon. This project underscores the role that collaboration and active coordination between governmental public health agencies, academic modeling teams, and industry partners can play in developing modern modeling capabilities to support local, state, and federal response to outbreaks.

5. A Flexible Data-Driven Framework for COVID-19 Case Forecasting Deployed in a Developing-world Public Health Setting

   Sansiddh Jain, Avtansh Tiwari, Nayana Bannur, Ayush Deva, Siddhant Shingi, Vishwa Shah, Mihir Kulkarni, Namrata Deka, Keshav Ramaswami, Vasudha Khare, Harsh Maheshwari, Soma Dhavala, Jithin Sreedharan, Jerome White, Srujana Merugu, and Alpan Raval

   In progress

   Abs

   Forecasting infection case counts and estimating accurate epidemiological parameters are critical components of managing the response to a pandemic. This paper describes a modular, extensible framework for a COVID-19 forecasting system, primarily deployed in Mumbai and Jharkhand, India. We employ a variant of the SEIR compartmental model motivated by the nature of the available data and operational constraints. We estimate best fit parameters using Sequential Model-Based Optimization (SMBO), and describe the use of a novel, fast and approximate Bayesian model averaging method (ABMA) for parameter uncertainty estimation that compares well with a more rigorous Markov Chain Monte Carlo (MCMC) approach in practice. We address on-the-ground deployment challenges such as spikes in the reported input data using a novel weighted smoothing method. We describe extensive empirical analyses to evaluate the accuracy of our method on ground truth as well as against other state-of-the-art approaches. Finally, we outline deployment lessons and describe how inferred model parameters were used by government partners to interpret the state of the epidemic and how model forecasts were used to estimate staffing and planning needs essential for addressing COVID-19 hospital burden.

2020

1. ACM CHIL

   CoSIR: Managing an Epidemic via Optimal Adaptive Control of Transmission Policy

   Harsh Maheshwari, Shreyas Shetty, Nayana Bannur, and Srujana Merugu

   Poster at ACM CHIL

   Abs medRxiv HTML Poster

   Shaping an epidemic with an adaptive contact restriction policy that balances the disease and socioeconomic impact has been the holy grail during the COVID-19 pandemic. Most of the existing work on epidemiological models focuses on scenario-based forecasting via simulation but techniques for explicit control of epidemics via an analytical framework are largely missing. In this paper, we consider the problem of determining the optimal policy for transmission control assuming SIR dynamics, which is the most widely used epidemiological paradigm. We first demonstrate that the SIR model with infectious patients and susceptible contacts (i.e., product of transmission rate and susceptible population) interpreted as predators and prey respectively reduces to a Lotka-Volterra (LV) predator-prey model. The modified SIR system (LVSIR) has a stable equilibrium point, an energy conservation property, and exhibits bounded cyclic behaviour similar to an LV system. This mapping permits a theoretical analysis of the control problem supporting some of the recent simulation-based studies that point to the benefits of periodic interventions. We use a control-Lyapunov approach to design adaptive control policies (CoSIR) to nudge the SIR model to the desired equilibrium that permits ready extensions to richer compartmental models. We also describe a practical implementation of this transmission control method by approximating the ideal control with a finite, but a time-varying set of restriction levels and provide simulation results to demonstrate its efficacy.

2. Synthetic Data Generation for Improved COVID-19 Epidemic Forecasting

   Nayana Bannur, Vishwa Shah, Alpan Raval, and Jerome White

   medRxiv

   Abs medRxiv HTML

   During an epidemic, accurate long term forecasts are crucial for decision-makers to adopt appropriate policies and to prevent medical resources from being overwhelmed. This came to the forefront during the COVID-19 pandemic, during which there were numerous efforts to predict the number of new infections. Various classes of models were employed for forecasting including compartmental models and curve-fitting approaches. Curve fitting models often have accurate short term forecasts. Their parameters, however, can be difficult to associate with actual disease dynamics. Compartmental models take these dynamics into account, allowing for more flexible and interpretable models that facilitate qualitative comparison of scenarios. This paper proposes a method of strengthening the forecasts from compartmental models by using short term pre-dictions from a curve fitting approach as synthetic data. We discuss the method of fitting this hybrid model in a generalized manner without reliance on region specific data, making this approach easy to adapt. The model is compared to a standard approach; differences in performance are analyzed for a diverse set of COVID-19 case counts.