Table 2 Areas where machine learning could be leveraged for improving the accuracy of estimations or projections and potential data sources identified through the review of traditional approaches.

Infectious disease dynamics: R₀, reproduction number¹; peak time, intensity, and duration²; proportion of asymptomatic infections³; transmission rate⁴; household transmission⁵; spatiotemporal dynamics with GIS⁶; population immunity⁷

Published literature: seroconversion⁴⁰; vaccine effectiveness⁴²; estimates from other diseases (e.g., SARS)⁴⁶

Publicly available data: COVID-19 global cases from the Center for Systems Science and Engineering at Johns Hopkins University⁴⁷; Worldometer COVID-19 pandemic updates⁴⁸

Health outcomes: infections⁸; severe cases⁹; susceptibility¹⁰; deaths¹¹; latency and infectious period¹²; undetected cases¹³; infections among healthcare workers¹⁴; infections among incarcerated populations¹⁵; infections among homeless populations¹⁶

Government data: government reported case data¹; census data⁴⁹; World Health Organization case reports and data on respiratory risk factors⁵⁰; Ministry of Civil Aviation (i.e., airline passenger data)⁶; airport transportation data⁵¹; Armed Forces health surveillance data⁵²; board of education school data (e.g., absentee reports that document H1N1)⁵³; hospital/outpatient sentinel surveillance²¹

Impact on healthcare systems and demand for: inpatient beds¹⁷; intensive care unit beds¹⁸; N95 respirators and surgical masks¹⁹; ventilators¹⁸; medical supplies²⁰; staffing²¹

Effects of NPIs: social distancing²²; self-isolation/quarantining²; workplace closures²³; contact tracing²⁴; wearing masks by general population²⁵; handwashing²⁶; optimal assay test pooling strategies for efficient testing²⁷; frequency of routine testing for COVID-19 in high-risk environments to reduce workplace outbreaks²⁸; mass testing/using drones to deliver tests²⁹; periodic testing of health workforce³⁰; impact of NPIs in residential care facilities³¹; closing borders³²; effectiveness of airport thermal screening³³; restrictions to sea, land and air travel³⁴; school closures³⁵; control measures for children if schools open (e.g., better ventilation, mask wearing)³⁶; public risk communication³⁷; media/news reports³⁷; impact of available information on behavior and vaccination uptake³⁸; ventilation of indoor spaces³⁹

Mobile phone mobility data⁵⁴

Daily news reports³⁷

Purchasing information⁵⁵

Google mobility reports using five different categories: retail and recreation, grocery and pharmacy, transit stations, workplace, and residential⁵⁶

Google query and news data⁵⁷

Google Flu Trends data⁵⁸

Apple Maps COVID-19 mobility trends⁵⁹

Publicly available data on infection rates from cruise companies^60,61

Impact of lifting NPI-related restrictions⁴⁰

Effects of pharmaceutical interventions: use of antiviral medications⁴¹; vaccination strategies⁴²; disease spread given vaccination availability constraints⁴³; transmission risk at immunization clinics⁴⁴

National Oceanic and Atmospheric Administration: temperature, humidity, wind speed^62,63

Home television watching (i.e., proxy for time spent at home)⁶⁴

Humanitarian assistance such as food distribution planning⁴⁵