---
id: "coronavirus-smarter-crowdsourcing:strategies/Develop,use,and-refine-multiple-predictive-models-"
type: "GovLab policy strategy"
title: "Develop,use,and refine multiple predictive models to help with pandemic planning,preparedness,and response."
url: "https://coronavirus.smartercrowdsourcing.org/strategies/develop,use,and-refine-multiple-predictive-models-"
summary: "Develop,use,and refine multiple predictive models to help with pandemic planning,preparedness,and response."
source: "The GovLab (coronavirus.smartercrowdsourcing.org)"
licence: "Creative Commons Attribution-ShareAlike 4.0 International License (https://creativecommons.org/licenses/by-sa/4.0/)"
---

# Develop,use,and refine multiple predictive models to help with pandemic planning,preparedness,and response.

Develop,use,and refine multiple predictive models to help with pandemic planning,preparedness,and response.

- Timeframe: 1-3 Months (immediate)
- Audience: Academia, Government Workers
- Type: Research, Data, Technology/Innovation
- Recommendation: [Increase Government Capacity](https://coronavirus.smartercrowdsourcing.org/recommendations/increase-government-capacity)
- Part of: [Smarter Crowdsourcing in the Age of Coronavirus](https://coronavirus.smartercrowdsourcing.org/)

## Actions governments should take

1. Estimate preparedness of and potential impact on healthcare infrastructure.
2. Use open-source data to predict and model viral spread.
3. Be creative in finding new data sources such as government hotline data.
4. Get the public to collect and contribute data. Citizen participation can be leveraged in data collection. When the community’s contribution is built on trust, data collection can be increased and improved, especially if the community can see a correlation to useful public services provided as a result.
5. Establish public-private data partnerships (data collaboratives) with the private sector. Data sources that illustrate human movement are especially valuable and those can be sourced from telephone companies. Data sources that illustrate behavior are also valuable and these can be sourced from search engine companies.
6. Incentivize engagement with the data-science community through prize-backed challenges. Use challenges to rapidly develop alternative predictive models and leverage outside expertise.
7. Assess the risks and biases inherent in the data in order to plan for and mitigate unintended consequences.
8. Increase data analytics capacity of public health officials in partnership with research institutions and universities through training and other knowledge transfer opportunities.

## Examples from around the world

1. Open source models like [Facebook AI](https://ai.facebook.com/blog/using-ai-to-help-health-experts-address-the-covid-19-pandemic/) model or the [UPenn Chime Model](https://oto.med.upenn.edu/2020/03/31/covid-19-hospital-impact-model-for-epidemics-chime/) can be used to predict impacts on specific populations, such as elderly populations or those living in [long term care facilities](https://jamanetwork.com/journals/jama/article-abstract/2767062). They have also been used to 1) predict cases and deaths, 2) predict hospital level demand, 3) support bed capacity planning, 4) predict peak estimation, 5) support PPE estimation, and 6) estimate second wave modeling.