Emergency Room Utilization: Admissions, Costs, and Testing Patterns

Introduction

Emergency room care is one of the most expensive components of U.S. healthcare, and costs have risen dramatically in recent years. Understanding the drivers of high ER expenditures and the factors that lead to hospital admission can help policymakers, hospital administrators, and insurers design more efficient care pathways and reduce financial burden on patients.

Research Questions

  1. RQ1: What predictors can help determine whether an ER visit results in inpatient admission?
  2. RQ2: Which characteristics of an ER visit are most strongly associated with the highest total ER expenditure?
  3. RQ3: What impacts the number of tests and procedures a patient gets done in the ER?

Data Source & Description

The dataset is under the 2023 Medical Expenditure Panel Survey, which surveys a geographically diverse population of the US and their medical providers. A sample of households are interviewed and surveyed for information, then their medical providers are contacted for more information. The medical providers include doctors, hospitals, pharmacies, etc. This dataset specifically covers 2023 and reports data on household reported emergency visits. Only households that have reported an emergency room visit were included in the dataset.

The original MEPS ER visit file includes 4,241 unique patients and 53 variables. According to MEPS, this file captures information about each emergency room visit, including: services and procedures performed, diagnostic testing, medications provided, total expenditures, and sources of payment.

Key Variable Dictionary

Variable Description
person_id Unique MEPS identifier (DUPERSID)
admitted Whether the ER visit resulted in inpatient admission
mri_ct MRI or CT scan performed
surgery Minor surgical procedure performed
xray X-ray received
lab_tests Laboratory tests completed
ekg Electrocardiogram performed
ultrasound Ultrasound completed
mammogram Mammogram received
vaccination Vaccination provided
rx_given Prescription medication given
related_condition Visit related to prior condition
total_cost Total ER expenditure (facility + physician)
oop_facility Out-of-pocket cost for ER facility
oop_doctor Out-of-pocket cost for ER doctor
private_fac Private insurance payment (facility)
private_doc Private insurance payment (doctor)
medicaid_fac Medicaid payment (facility)
medicaid_doc Medicaid payment (doctor)
medicare_fac Medicare payment (facility)
medicare_doc Medicare payment (doctor)
oop_total Total out-of-pocket cost (facility + doctor)
insurance_type Derived insurance category
sex Male / Female
poverty_cat Poverty classification (5 levels)

Data Preprocessing

From the original 53 variables, 23 were selected as relevant for the research questions.

  • Diagnostic tests and procedures (MRI/CT, surgery, X-ray, lab tests, EKG, ultrasound, mammogram, vaccination, prescriptions, related condition) used codes like: binary factors (0,1)
    • -8 → Don’t know
    • -7 → Refused
    • 95 → Not received
  • Admission outcome: - binary factor for modeling.
    • 0 = Not admitted
    • 1 = Admitted
  • Insurance: New binary variable created:
    • If all insurance payments (Medicaid, Medicare, Private) were zero or missing → Uninsured (0)
    • If any insurance payment > 0 → Insured (1)
  • Total tests: A composite variable (total_tests) counting the total number of diagnostic tests and procedures received during an ER visit, including MRI/CT, surgery, X-ray, lab tests, EKG, ultrasound, mammogram, vaccinations, and prescriptions.

The ER visit dataset was merged with the MEPS 2023 Full-Year Consolidated Data using the patient identifier to incorporate:
- Sex
- Poverty category (five levels: Poor/Negative, Near Poor, Low Income, Middle Income, High Income)