ashutosh-singhal / cancer_social_health_determinants

Using Machine Learning Approach to Identify Determinants of Health associated with Cancer in the Southern USA

Cancer and Determinants of Health in the Southern USA : A Machine Learning Approach

Executive Summary

Determinants of Health (DoH)-- are the factors that influence how likely we are to stay healthy or to become ill or injured. The circumstances in which we grow, live, work and age are called as social determinants include factors such as income, education, employment and social support. A person's health is also influenced by biomedical factors and health behaviors that are part of their individual lifestyle and genetic make-up. These factors can be positive in their effects (for example, being vaccinated against disease), or negative (for example, consuming alcohol at risky levels).

In context to cancer, components of socio-economic framework affect the overall nutrition, stress level, life style and behaviors etc., which may affect the cancer prevalence and mortality. In this project, I will attempt to understand the relationship between determinants of health and cancer incidence and cancer deaths in the southern states of USA using machine learning approach.

My hypothesis is that “Socio-economic and risky health behavior can predict the cancer incidence and death rate in southern state of USA at county level”.

To test this hypothesis, I downloaded cancer death and new cases data (age-matched and normalized to 100,000 people) of years 2011-2015 from CDC. The DoH factors data was downloaded from County Health Ranking (http://www.countyhealthrankings.org/).

Motivation

A recent report published in JAMA (Connor et al 2018, see reference) pointed out that some counties particularly in southern states of USA have relatively very high rate of cancer deaths and authors described those areas as "Hot Spots" of cancer deaths. A quick overview on CDC website validated these points and I learned that most of southern states have poor ranking in cancer deaths at national level. As many parts of southern states have poor socio-economic framework I was curious to finding socio-economic and health behavior factors, which can predict the trend of cancer incidences as well as deaths at county level. This will highlight the need of improving those determinants and reducing cancer deaths.

Data Question

My hypothesis is that "Socio-economic and risky health behavior can predict the cancer incidence and death rate in in southern state of USA at county level”--addressed by the following objectives:

Mapping cancer incidence and deaths at county level in Southern USA.

Analyzing association between determinants of health (DoH) and cancer incidence or deaths at county level?

Building model to predict number of new cancer cases or deaths based on DoH.

Modeling to identify counties with greater burden of new cancer cases and deaths?

Data Sources

Centers for Disease Control & Prevention (CDC) http://www.cdc.gov/cancer/dcpc/data/index.htm

County health ranking http://www.countyhealthrankings.org/

Reference:

https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2705856