Welcome to the Class

Ian McCarthy | Emory University

Motivating question

Does health care spending improve health?

Let’s motivate the course with this simple question…does health care spending improve our health?

Spending and Health

R Code 
ggplot(data = (dartmouth.data %>% filter(Year==2015)), 
       mapping = aes(x = Expenditures, y = Total_Mortality)) + 
  geom_point(size = 1) + theme_bw() + scale_x_continuous(label = comma) +
  geom_smooth(method="lm", se=FALSE, color="blue", size=1/2) +
  labs(x = "Spending Per Capita ($US)",
       y = "Mortality Rate",
       title = "Mortality and Health Care Spending")

Spending and Health

  • Does medical spending make us sicker?
  • What else might explain this relationship?

Goals of this course.

  1. Understand and implement selected methods for causal inference
  2. Along the way…data management and version control with real data
  3. Summarize, visualize, and explain research results

Syllabus highlights

Why this course?

  1. Major problems that need solutions
  2. Need good, convincing empirical work for policy
  3. Working with data is hard, particularly health care data
  4. Your work should be transparent and reproducible

Structure

  • Very applied in nature
  • Methods for causal inference
    • Selection on observables (regression, re-weighting, matching, propensity scores)
    • Instrumental variables
    • Regression discontinuity
    • Difference-in-differences

Structure

  • Substantive areas
    • Hospital pricing, policy, and competition
    • Cigarette taxes and demand
    • Medicare Advantage and quality disclosure
    • Medicaid expansion and health insurance

Structure

  • Datasets from the real world
    • Hospital Cost Report Information System (HCRIS)
    • Centers for Disease Control (CDC)
    • Medicare Advantage data
    • Behavioral Risk Factor Surveillance System (BRFSS), Medicaid, Health Insurance Exchanges

Assignments

  • Homework (x5)
  • Peer Reviews (x4)
  • Participation

Grading

Component Weight
5 × homework assignments (15% each) 75%
4 x peer review (5% each) 20%
Participation 5%