-
Notifications
You must be signed in to change notification settings - Fork 9
Home
Mortality rates are rising in the United States with significant community and regional variations. MortalityMinder (MM) is a web-based visualization tool that enables interactive exploration of social, economic and geographic factors associated with premature mortality among mid-life adults ages 25-64 across the United States. Using authoritative data from the CDC and other sources, MM is a freely available, publicly-accessible, open-source, and easily maintained tool. The goal of MortalityMinder (MM) is to enable healthcare researchers, providers, payers, and policy makers to gain actionable insights into how, where, and why midlife mortality rates are rising in the United States (US). It is designed to help healthcare payers, providers and policymakers at the national, state, county and community levels identify and address unmet healthcare needs, healthcare costs, and healthcare utilization.
MM can be found at
- http://mortalityminder.idea.rpi.edu (primary site)
- https://olyerickson.shinyapps.io/mortalityminder/ (backup site)
MortalityMinder uses county-level mortality rates and social and economic factors measurements available from well-known public portals. Mortality rates from 2000-2017 are obtained through the CDC WONDER portal, the definitive source of mortality information in the United States. Social factors data for 2015-2017 are obtained through County Health Rankings (CHR), an aggregate of county-level data curated by the Robert Wood Johnson Foundation. MM considers 168 factors from twenty (20) sources, including datasets such as AHRF, BRFSS, the Bureau of Labor Statistics, the FBI, and many others. This version of MM focuses on midlife deaths attributed to leading causes of death including 'Deaths of Despair', 'Cardiovascular', 'Cancer' and 'All Causes', but the approach can be readily generalized to any health problem of interest.
Age-specific mortality rates were calculated in three-year chunks for each cause of death at the county, state, and national levels. To privacy CDC Wonder suppresses rates for counties with too few deaths, so calculating rates over three year periods ensures more rates are reported. MM aims to capture the actual experience of mortality of ages 25 by Americans in a community, so our analysis does not age-adjust. In the future, age-adjustment could easily be added to MM. To provide for more complete data for effective visualizations, county mortality rates that are suppressed to preserve privacy by CDC WONDER were imputed using mortality rates within a state and the Amelia package for R. Multiple imputation could be added to the analysis in the future. Details of data sources and preparation are available at: https://github.com/TheRensselaerIDEA/MortalityMinder/wiki/data/
We gathered factors addressing health behaviors, clinical care, education, employment, social supports, community safety and physical environment domains from CHR. We filtered the original 168 factors from CHR to a set of approximately 70 factors that were relevant to at least one cause of death at the national level. We first kept only factors that were rates or other measurements that represented rates on other measurements that did no directly reflect county population size. Then we performed at factor association study to determine the associations with all 4 causes of mortality, corrected for multiple hypothesis testing using the Benjamin-Hochberg Method. Thus only potential relevant factors to at least one cause of death were deployed in the app.
MortalityMinder enables users to dynamically investigate county and regional variations within mortality trends at the national, state, and local levels and the social determinants and other factors associated with these trends. MM prioritizes and visualizes the most significant factors associated with higher risks for each cause of death and allows the user to explore individual factors include its relation to the selected cause at a county level for each state and the distribution of those factors within each state.
With MM, the user can visually explore the growth of mortality rates in the United States using county-level three-year averages from 2000-2017, providing both national and state views. For each cause of death, MM clusters counties in each state, categorizing each county into high/medium/low mortality clusters; visualizes the geographic distribution of those clusters; and illustrates the mortality trends of those clusters and user-selected counties over time. MM does a similar analysis for the entire United States, only using six clusters. The app determines the correlation of social factors and visualizes the most significant, highlighting whether those factors are protective or destructive. Users can select individual factors to explore more deeply at the state level.
Demonstrations and user testing by our advisory panel of health care experts and other testers have shown that MortalityMinder is compelling and highly engaging. The page-oriented organization of the app lends itself to user-driven investigation and storytelling, like a highly interactive slideshow.
MortalityMinder provides striking insights into the disturbing rise in 'Deaths of Despair' due to suicide, overdose, and self-harm. The app was inspired by 'Mortality and morbidity in the 21st century' (Case and Dean, 2017) which described the disturbing rise in mortality rates by 'deaths of despair.' The app highlights significant disparities in mortality rates between counties across and within states with both common and distinct associated factors. Such insight can be used to create regional specific interventions and determine best practices.
MortalityMinder further provides a county-level confirmation of the trends and hypothesized causes in 'Life Expectancy and Mortality Rates in the United States, 1959-2017' (Woof and Schoomaker 2019). MM shows, however, that rates due to other causes including cancer, cardiovascular disease and all causes are also on the rise. MM confirms the midlife mortality rate increases in (Woolf and Schoomaker), but it provides much greater insight into community-level variations and their associated factors that may be actionable.
The focus of MortalityMinder development from July 2019 thru December 2019 was on data source identification for multiple causes, beyond our Phase 1 example of 'Deaths of Despair'; social factor selection; refinement of our analytic techniques; expansion of our visualization choices; implementation of our interactive, web-based framework; and refinement of the user experience, with input from our external experts, student graphic designers, and a formal usability study.
The team's overarching development objective, beyond analytical credibility, was to provide a compelling and engaging tool for health policy and other decision-makers. MortalityMinder can be used as an investigative tool, but may also be used for interactive storytelling, e.g. for lectures on mortality across the United States.
MortalityMinder is currently published via two publicly-accessible web locations. Our open-source R code is freely available via a github repository. Source data and generated results may be downloaded from within the app. MM developed using the R language and environment for statistical computing and graphics, incorporating best practices and using well-known packages whenever possible. Data Loader scripts enable additional years and types of data to be easily integrated. The result is a robust, extensible package that can be maintained and grown over time as either an open-source package or within organizations such as AHRQ.
MM can be run from the public web locations; no user installation is required to test the application. Alternatively, the GitHub repository may be cloned and run immediately in the user's [RStudio environment (https://rstudio.com/), either on a server or on a personal machine.
MM utilizes the R Shiny platform for web interactivity; most of the visualizations presented in MM are generated in real-time based on data loaded when the app is launched. Data analysts and software engineers familiar with the R language and reactive coding via Shiny will have no problem customizing the code as required. MM's overall web layout utilizes a standard Javascript UI framework which is easily customized to modify the overall style of the app.
MortalityMinder was created by undergraduate participants in the Data INCITE Lab at Rensselaer Polytechnic Institute with support from the United Health Foundation and the Rensselaer Institute for Data Exploration and Applications. MortalityMinder development was directed by Kristin P. Bennett and John S. Erickson.
The MortalityMinder Team would like to thank Jim Hendler, Director, The Rensselaer IDEA, and Curt Breneman, Dean of the School of Science at Rensselaer for their support and encouragement. Please send questions and comments about MortalityMinder to: erickj4@rpi.edu.