3.30 Update (Outside Research on Medicare)

Good morning!

As an Honors Student, I have to complete an Honors Thesis Project before graduating as a Commonwealth Scholar. Although I had initially planned on completing my project through Undergraduate Research this summer, I am not too sure if those programs will still be running with everything going on. As such, I have decided to take this GeoSpatial Project and form it into my Honors Thesis Proposal. I’m not yet sure if this will be my final project (I’ll have to see how everything with COVID-19 plays out), but for my HON 301 class I am presenting it as my proposal.

Below are three sources that I have reviewed and found helpful in determining the next steps of analysis for my project.

Geographic Variation in Medicare Spending

Cassidy, Amanda. “Geographic Variation in Medicare Spending.” Geographic Variation in Medicare Spending | Health Affairs, 6 Mar. 2014, www.healthaffairs.org/do/10.1377/hpb20140306.633790/full/.

Medicare is the largest single payer in the United States, as it provides nearly 52 million elderly and disabled beneficiaries. It has long been known that there is a large discrepancy of Medicare spending per beneficiary, but this study in particular looks at the direct causes of this disparity. For the purpose of this study, data was collected from the “Five Hospital Regions (HRRs) with the Highest and Lowest Actual Standardized Per Capita Medicare Spending in 2012”. Essentially the data shows how much Medicare pays per person in each region for an entire illness or injury (from initial diagnosis to full recovery). For example, and one of the disparities shown, is for an injury. Although many patients can handle returning home and come in for physical therapy, there are a few who need to be sent to a nursing home, thus increasing the cost of service. 

Through conducting the research Cassidy found one thing to be clear, “[there are] inconsistent relationships between health care and quality cost”. Similarly, it was found that these spending patterns and disparities are only seen through Medicare. Researchers looked at health care spending for those with private insurance or Medicaid, but did not see that same variation at the state spending level. One thing the study attributes to this is the level of income states have. They believe that low income states have less money to spend towards their Medicare programs.

Unfortunately the conclusion of the study was that “there is no single answer to addressing variation in Medicare spending by region”. Although they considered many factors, including the same geographic factor I plan to analyze in my study, there were still unexplainable differences. As a result the CMS has decided to continue investigating possible payment reforms, rather than setting a geographic standard of how much Medicare should be paying per region.

State-Level Variation in Medicare Spending

Gage, B, et al. “State-Level Variation in Medicare Spending.” Health Care Financing Review, CENTERS for MEDICARE & MEDICAID SERVICES, 1999, www.ncbi.nlm.nih.gov/pmc/articles/PMC4194651/.

After looking at the initial study and the issues it saw on a state level, I found a study that specifically looked at state spending on Medicare. Although this study was done in 1999, many of the policies and criteria the study was based on are still in effect today. For example, “Medicare provides, in theory, one standard benefit package to all its 39 million enrollees (currently 52 million), regardless of health needs, age, or location”. The purpose of this study was to look at what you are expected to get from Medicare versus what you actually receive, depending on your state. 

By looking at national records for patients on Medicare, it was discovered that 35% percent of all beneficiaries live in one of 5 states (California, Florida, New York, Pennsylvania, and Texas), making Medicare an important economic force in these states. Similarly, in more rural states like Iowa and Maine, medicare insures nearly one-fifth of the population. Keeping in mind that some of these individuals may also be eligible for Medicaid can alter the percentage of elderly on Medicare. With this in mind, the study showed that “Medicare payments for those age 85 or over are 44% greater than for the average beneficiary” and the states with the largest number of 85+ residents are Massachusetts and D.C.. 

Through conducting this study it was seen that although Medicare is designed to be universal and standardized, the use of benefit varies greatly between populations. Everyone may be entitled to the same standard of care, but population density, age disparities, and cost of living by state all play an important role in what percentage of a service is actually covered by Medicare. With this information in mind and applying it to my own study, I can see that it may be essential to compare cost variance with population of those on Medicare, opposed to general population.

The Pricing of U.S. Hospital Services: Chaos Behind a Veil of Secrecy

Reinhardt, Uwe E., et al. “The Pricing Of U.S. Hospital Services: Chaos Behind A Veil Of Secrecy.” Health Affairs, 2006, www.healthaffairs.org/doi/full/10.1377/hlthaff.25.1.57.

With the knowledge obtained in regards to regional discrepancies of Medicare, I thought it would be essential to look into the formulation of how hospitals set their prices. This study takes a look at the widespread practice of “price discrimination” and how the same service can be charged at an array of prices for different patients. The study also covers how prices are updated and how although some hospitals will simply increase all services by a certain percentage, others will only update certain items, or vary percentages for each department.

The study begins by analyzing the objective of price discrimination, stating “sellers seek to maximize the total amount of revenue that can be extracted from society for a given volume of output and, thus their profits”. Hospitals use part of this price discrimination to make up for all allocated costs (primarily pro bono services), so it isn’t necessarily all bad. Along the same lines, private insurance companies are often charged more for an operation, than a patient without any medical insurance because realistically the company can easily pay it, whereas the patient may turn away from the service because they can’t. Now how does all of this affect Medicare?

The study looks specifically into how Medicare approaches price discrimination, because essentially, Medicare pays hospitals flat fees, according to each distinct diagnosis-related group (DRG). However, that DRG fee is then “further adjusted for regional variations in the cost of labor and of other hospital inputs, and for other local factors that might affect a hospital’s cost of producing care”. If you think that statement is vague, it’s because it is. Essentially, Medicare outputs a list to their patients of how much is covered per service. However, that amount ends up varying greatly at the hospital’s discretion based on factors that they don’t necessarily have to defend to the public. Whether an extra surgeon is observing the surgery, they need to use name brand medicine, opposed to generic, or even if a double room isn’t available and you are up-charged for a private room, there are infinite factors that change the cost of service that Medicare doesn’t account for, thus the “Chaos Behind a Veil of Secrecy”.

With these sources in mind, I plan to do spatial data on the 100 highest costs and seeing if there is some correlation. Based off of the articles, I plan on taking state wealth and per capita income in mind, but I think it will be interesting to see if the highest costs are around the rural & most poverty stricken areas that the studies suggest, or if there are some hot spots in major cities with reasons unaccounted for in the articles.

3.27 Update (Data is Finally Clean & Functioning!)

This week has been, needless to say, tedious with my Data Set. For the purpose of my project, I went with 2 distinct data sets; the first being of all hospitals in the United States, and second being hospitals that charged for at least one Intracranial Hemorrhage. The problem was, the second set of data points did not include lat & long coordinates, which are essential when trying to create a heat map.

Needless to say, after much trial and error, and some much appreciated help from Professor Davis, I was able to create a pair of coordinates for each data entry.

With that information, I began working on what I believe, will be the most telling aspect of my project, the Heat Map. As you will see in the graphs below, I didn’t start out too glorious, there are many “error” images I have saved, however with each attempt, I am getting closer to what I want to create.

Error1

If you look closely (and I mean really SQUINT), you can see that this is in fact the convex polygon of the United States. However, this doesn’t look appealing to the eye, or even allow for any analysis. If you go through my Notebook attached, you can see what I have tried since then, but I think the most promising aspect has been some code I found on the Wolfram Community that plots a heat map by counties in regards to population. Although it takes a while to run, I am hopeful that I can manipulate it into using my data.

This is the population graph that was created based off of population records & Per Capita Income by county throughout the United States.

Now that my data is finally all set to be manipulated, I plan on spending next week taking a look at all of the different ways I can draw an analysis on this data. This will also help with the writing portion of the project, as I begin looking at other studies that have been done on the disparities between medical costs.

March 27 Update

Post Spring Break Update (All Work Through 3.22.20)

Update of What Has Been Done:

The last time I updated the class was the Monday before Spring Break, so I will try not to repeat myself too much. That being said, I was having difficulties utilizing Geopy in Python, so over Spring Break I switched over to Mathematica and it was much easier. After taking into consideration looking at the distribution of hospitals across the United States, I found a current complete list of all hospitals (~7,500 entries), whereas the Data I was using in regards to medicare costs, had about 2,700 data points.

Here is my graph of every single Hospital’s Data Point:

graph1

Additionally, here is a graph of the population distribution across the United States, according to the last census.

graph2

These two graphs looked nearly identical to me, so I didn’t necessarily find the need to use geometric clustering for this. For my next step, I am going to graph the data that I was given with the medicare costs and see if those 2700 points are a good representation of the overall hospitals.

Next Step:

After noticing that the distribution of hospitals closely resembles the general population, I want to start considering the cost distribution and how it relates to geographic position. In order to do this I will first create a heat map of all points, but then I want to utilize geometric clustering for either the top 100 or 500 highest costs of the surgery and the lowest and look to see how those cluster. Are city areas with the highest costs or is it the more rural areas, with only one or two major hospitals in a 100 mile radius.

What I am Having Trouble With:

I’ve never used Mathematica before this project, and although it is fairly straightforward, I feel like I am missing certain things. For example, the second data I have with the medicare costs, doesn’t have lat & long coordinates. I know how to use FindGeoLocation[ ” ” ] to find one coordinate, but I don’t know how to use my file “HosptialAddresses”  or “HospitalNames” to find every coordinate.

Here is my current Mathematica Notebook, in case you were unable to view the graphs of distribution across the United States:

Plot