PUI2016 Extra Credit Project Proposal: Analysis of Medicaid Hospital Visits in NYC

Zhaohong Niu, EstherNiu[github], zn352

Problem Description

U.S. Medicaid Program benefits uninsured population to receive proper medical treatment. With number of uninsured people in New York decreasing (2013-2015), Medicaid expansion shows a drop over these three years[1][2]. Has Medicaid efficiently helped people in need? Evaluating the effectiveness by looking into Medicaid hospital visits could be an interesting approach, which provides us a new perspective while no existing research has fully developed their scale.

Here I unfold my main question into three dimensions: What are the major chronic conditions do people who enrolled in Medicaid Program visit hospital for? Any observable changes over the year as Medicaid expansion decreases? Furthermore, what pattern or relationship can we find between Medicaid expansion, Medicaid Program enrollment and hospital visits?

This project can help NYC government understand how Medicaid has helped uninsured families in treating severe disease, and what type of disease is most common for people in need of a Medicaid program. It will benefit the process of policy-making to better balance and distribute medicaid resources, funds and vacancies.


To answer these questions, I will examine data from NYC Health Data and Open Data -

  1. Medicaid Chronic Conditions, Inpatient Admissions and Emergency Room Visit (2012-2014, NYC)
  • Contains all Medicaid hospital chronic inpatient visits in NYC. It does not include hospital visits with acute condition. I want to examine only the part where Medicaid helps people with chronic conditions.

  • attribute: major diagnostic category, beneficiaries with condition (number of visits), beneficiaries with admission (number of inpatients), year

  • transformation: groupby method to aggregate patient number by each attribute

  1. Medicaid Enrollment Number by Month (2009-2016, NYC) with race, income-level and other demographic information attached
  • The data is legit as it is issued by US Census.

  • attribute: year, month, number of recipients, race, gender, age group, plan type

  • transformation: groupby method to aggregate number of recipients by year, drop other attributes

  1. Population Without Health Insurance Coverage Report: 2013 to 2015, contain information relating to expansion

Those datasets will be analysis by three parts -

  1. the descriptive facts of Medicaid Program’s effectiveness measured by Medicaid hospital visits and their relating disease category

  2. change of disease category and number of Medicaid enrollment over the years

  3. the correlation between number of Medicaid hospital visits, number of Medicaid enrollments and Medicaid expansion in New York City.


Correlation Analysis and t-test to identify possible change/difference between three years in hospital visits and Medicaid enrollment (alpha=0.05)

Logistic modeling to predict future tendency in Medicaid enrollment, tendency in disease category (to define which disease will probably stay high in hospital visits in the future)


3-4 graphic analysis on descriptive facts.

A statistical conclusion of the correlation results and changes in Medicaid hospital visits over three years (2012-2015), and prediction of future tendency.