3.5 Identifying Overprescription in Dual Enrolled VA Patients

3.5.1 Materials/Data

3.5.2 Competencies

Modules: 1 (Importing/Reading Data); 4 (Pandas Intro); 5 (Relational Data/Data Merging); 6 (Data Cleaning); 7 (Data Visualization)
Competencies: Pandas (general); Data merging; Matplotlib/Visualization (basic, general); Cleaning string data; Cleaning inconsistently coded data

3.5.3 Background

This use cases borrows from both a publication and Presentation by Dr. Walid Gellad within Pittsburgh’s Veterans Affairs Hospital system. The simulated data emulates a simplified set of prescription data for Veterans Affairs (VA) and Center for Medicaid/Medicare Services (CMS) patients, with overlap of healthcare utiliation in both systems. You will work with this data to merge the VA and CMS data sources and identify prescription (and overprescription) of opioids by patients who utilize VA and both VA & CMS healthcare systems.

*Preface Notes & Assumptions:**
- Assume all buprenorphine medications are equivalent to buprenorphine tablets in the given reference table for morphine mg equivalents

Data Dictionary
Data Element VA Data Variable CMS Data Variable
Patient ID Patient ID Patient ID
Date of Hospital Visit/Encounter Visit date N/A
Age at Visit/Encounter Age N/A
Height at Visit/Encounter Height N/A
Weight at Visit/Encounter Weight N/A
Medication Name Medication Medication
Dose (Daily) Medication Dose Medication Dose
Unit of Medication Dose Medication Dose Unit Medication Dose Unit
Length of Prescription Duration Medication Duration Value Medication Duration
Unit of Time for Length of Prescription Duration Medication Duration Unit Duration Unit

3.5.4 Tasks

3.5.4.1 Data Cleaning & Preparation

  1. Import CMS and VA data into your Jupyter Notebook
    1. Examine the data. Are the variables consistently names between the two data frames? Is the same information collected? What differences exist?
  2. Create opioid administration outcome variable of interest within both the VA and CMS data
    1. We want to simply capture whether a given patient was given an opioid, essentially whether that patients medication was an opioid or not. What could a variable that codes this information look like?
  3. Standardize coding of relevant variables within data frames and between VA and CMS data sources (Examine units and/or variables that may be included as a singular variable in one data set and as a pair or set of values in a different data set).
    1. Dose Value (by daily intake)
    2. Social Security Number (SSN)
    3. Medication Duration
  4. Identify total, daily intake of mg morphine equivalents
    1. A PDF of the Center for Medicare and Medicaid Service’s “Conversion Factors” Document and a screenshot (accessed 9/22/2020) included below for your reference.

MG Morphine Equiv Chart

  1. Merge data frames together on common identifier (SSN)
  2. Create and indicator variable to identify individuals dually enrolled in VA and CMS healthcare systems.
  3. Create a sum variable for total mg-morphine-equivalents within dual-use patients

3.5.4.2 Exploratory Analysis & Visualization

  1. How many patients are dual enrolled and how many were solely accessing VA healthcare resources?
  2. Visualize MME among patients using histograms and/or density curves:
    1. Separately for VA patient visits and CMS patient visits (using only VA and CMS prescriptions respectively)
    2. For dual-enrollment patients (using our MME total/sum variable created in Data Cleaning Step 8)
    3. Analyze the comparative visualizations (i.e. differences you do (or do not) see).
  3. What proportion of dual enrolled patients were prescribed an opioid? Similarly, what proportion of “VA-only” patients?
    1. What proportion of dual enrolled patients were prescribed an opioid during both their VA and CMS visits?
  4. Create a visual that illustrates this difference

3.5.4.3 Bonus Questions

The below questions demonstrate additional skills that, while beyond the scope of building foundational Python competencies/skills, may be of interest to explore if you found the earlier assessment simple!

  1. Create a density plot of total mg morphine equivalents among dual-enrolled patients, among VA-only patients, and among CMS only patients. Shade in the plot for patients with greater-than-or-equal-to 120 mg morphine equivalents
    1. Hint: This StackOverflow post may be useful in extracting densities using scipy)