Tool

FREM Explorer

  • Tool
  • Multidisciplinary Analysis

The FREM Explorer is an easy-to-use point-and-click tool to visualize longitudinal Full Random Effect Models (FREM) built from Ki data. The tool enables the user to visualize and perform prospective and longitudinal modeling and prediction of physical growth and cognitive development.

The Challenge

A key challenge for Ki, the Foundation, and the global health community is to understand risk profiling and prevention in terms of: who is at greatest risk, what risk factors drive most of the burden, and when in the life cycle does a person have greatest risk. As a response to this challenge, the FREM model was developed to address these questions in terms of risk profiling and interventions, and to improve the efficiency of identifying: who will derive maximum benefit from intervention, what risk factors are most responsive to intervention, and when in the life cycle will targeted interventions be most effective. The FREM Explorer was developed with an easy-to-use and intuitive visualization to facilitate collaboration between data scientists and domain experts, and to improve understanding and use of this complex model.

Ki leveraged one of FREM’s strengths which accounts for missing measurements during the course of a longitudinal study. In neurocognitive studies of children with incomplete measurements, the team identified an unexpected result that poor sanitation was a major risk factor for worsened cognitive development.

Start Date

September 2016

Stage of Development

Always in Progress

Working Teams

HBGDki Members

Key Features

Predefined or user-specific scenarios

The user can select a set of predictor variables from a predefined list characterizing the sites of the COHORTS study. The approach addresses partial or missing data. The user does not need to input all the variables – just the collected or available variables – and may leave the other variable cells empty.

Prediction over a population

The user can explore the population heterogeneity for a scenario of interest. The output includes summaries of interest such as percent of children stunted at a specific age.

Prediction for a specific individual

The caregiver can input a specific child age and corresponding length- or height-for-age z-score (HAZ). The tool will show a prediction of the most probable trajectory based on the individual data, suggesting when to collect the next data point to maximize the information.

Prediction of cognitive function

Share

Last Updated

October, 2020