Using FRED

FRED is a computer program that uses Big Data to model the dynamics of infectious disease epidemics. FRED has been used by professional epidemiologists to study historic influenza and measles outbreaks, including the effectiveness of vaccinations, school closing, travel restrictions and other factors that might slow an outbreak. Now you will use FRED!

But first, lets review disease modeling. Most people are introduced to infectious disease modeling using the SIR approach. A population of people starts as being susceptible (S) to a disease, and one or two already infected (I) people initiate the spread of the disease through the population. Over time infected people are removed (R) from the population either by recovering or dying.  SIR models consider the statistical likelihood of a mass of people moving from one category to the next. SIR models have just a few equations and can be calculated using an Excel spreadsheet.

FRED is a different type of model – an agent-based model (ABM) – that tracks every single person through time as their infection status changes.  Within an ABM each person (or agent) has a specific age, gender, home location, travel needs, and other realistic characteristics. Once a location for the outbreak is selected, for example, Cleveland, Ohio, the number of people in the model is based on the actual population of Cleveland, and each person is assigned an age and gender, based upon real demographic data for that city. Other data are used to assign each person an average daily travel distance and time, information used to calculate how many people an individual is exposed to who might be infected, or who might be infected by the individual.

Simple FRED output of the occurrence of flu in a Pennsylvania county.

With this realistic population the researcher first models the spread of a disease defined by specific rates of infection and recovery. Once that baseline is established the model can be modified – for example, what is the impact of school closings (so kids don’t easily spread the disease to classmates, who then carry it home to parents) – to investigate the impact of various procedures to slow the disease spread.

For privacy reasons and for ease of statistical manipulation FRED does not use demographic data for actual people, but instead creates a synthetic population based on the actual data. This U.S. Synthetic Population Database (USSPD) retains nearly 100% of the information but none can be tied to specific individuals. This is Big Data that has been manipulated to be more accessible to machine processing and to not infringe on privacy. USSPD was developed specifically for the National Institutes for Health to use in modeling epidemic disease outbreaks.

The second part of FRED is a sophisticated disease model that specifies the disease’s contagiousness, period of infectivity, recovery rate, death rate and similar factors. Changes to these parameters allow different diseases to be studied – for example, measles spreads 6 to 8 times faster than Ebola, and it normally kills fewer than 1% of patients, whereas the mortality rate for Ebola is typically 50% to 90%.

FRED is no longer being developed but three versions are freely available to use. First is the source code (programmed in C++) which computer programmers and epidemiologists can download to run on their own machines. The source code allows experimentation with the full range of variables, and more can be added, but it is not suitable for anyone but experts. That version of FRED takes about two minutes on a high-end laptop to run a model with a population of one million people; a run for the entire US population requires about four hours of supercomputer time.

Two collections of FRED simulation outputs are also available online. These include relatively simple models that have been pre-run and the outputs can be selected for display. One simulation series concerns measles and how vaccination rates affect its spread across a population. The second simulation series demonstrates the impact of the contagiousness (R0) in the spread of influenza. Outputs exist for both of these models for hundreds of counties across the U.S.A. In Critical Thinking with FRED you will use these outputs to explore the effects of various parameters on infectious disease spread.