Hello All,

Hoping to receive feedback on a plan for an OLS MLR analysis planning in STATA.

Question: What is the effect of health insurance coverage type, or lack thereof, on total number of healthcare visits in a given year?

Motivation: Healthcare in the United States has been a popular and polarizing topic, particularly the argument of a government-run single-payer vs. private insurance scheme. However, even for those lucky enough to have health insurance coverage, not all coverages are created equal. These different types of coverage and their associated differences in cost can dictate how often people seek out care.

Data: Data will come from US Census IPUMS Health Surveys, MEPS. All data will come from one cross-section of individuals in 2019 and provide information on their demographics, racial and ethnic background, number of healthcare visits in last year (both office & ER visits aggregated), income, education level, and a general measure of one’s health, and health insurance type (Private, Public-Military, Public-Medicaid, none)

Model:
Healthcare visits = b0 + b1 health insurance type + b2 income + b3 education + b4 age + b5 gender + b6 race + b7 marital status + b8 dependents + b9 health measure

X-Variable:
  • Health insurance type (either private, public military, public medicaid, none. dummy variable setup)
Y-variable:
  • Healthcare visits in last year (numeric, 0+)
Controls:
  • Total personal income (numeric, dollars annually)
  • Education level (numeric, 0-20)
  • Age (numeric)
  • Gender (numeric, 0 or 1 binary)
  • Race (numeric, dummy variable setup)
  • Marital status (numeric, 0 or 1 binary)
  • Number of dependents (numeric)
  • Health measure (1-5 excellent to poor) This is a subjective measurement. Not sure if I should include it.
Unobservables/Endogeneity/Spurious Correlation:
  • Endogeneity (when x-variable correlated with u)
    • Intellect: endogeneity issue with education and income. Also, smarter people may select more appropriate healthcare for their situation, or seek out healthcare if they didn’t have any
    • Upcoming healthcare: endogeneity issue with health insurance type. People may have a type of healthcare for upcoming medical issues
    • Pre-Existing conditions: endogeneity issue with health insurance type. Sometimes those with pre-existing conditions cannot get access to the healthcare needed for their situation
  • Spurious correlation (when x and y appear correlated, but u is driving x or y)
    • Health status: health status may mandate seeking services regardless of health insurance type. Attempt to control using rating scale, but this won’t be perfect
    • Nuances to health insurance type: Even within a health insurance type (private, public, none) there are further nuances to specific plans and coverages which may dictate how often care is received
Data Filters:
Ages 26-64 (remove all children’s insurance and Medicare) to focus only on general population adult coverages
Other child/Medicare effects removed
Allowed to have b1 = 0