Let’s Talk Medicare Cost

In the healthcare industry, the days of business as usual are over. Despite countless incremental fixes, the U.S. healthcare system is struggling with rising costs and inequality.

$~$

This report will review and analyze the Medicare Hospital Spending by Claim data. It was published on April 6, 2016 and updated on July 14, 2017. The data was published by the Centers for Medicare & Medicaid Services (CMS). The original data set contains 13 variables, and 69,630 rows. It breaks down the Medicare Spending per Beneficiary (MSPB) by Claim Type file and States. The data collected covers claims paid during the period from 3 days prior to a hospital admission through 30 days after discharge. The data contains “0” values but no missing values which could be explained by the fact that the data is a collection of pre-cleaned data electronically gathered from several healthcare organizations across the nation. In order to provide an accurate visualization of the analyses, an arbitrary cost threshold of $10 was set. The data cleaning process is critical since it allows the minimization of analysis errors and maximization of the analysis accuracy.

Why?

Healthcare represents a sixth of the U.S. economy. With advancement of medicine, the US population ages 65 year old and above is expected to grow significantly. Consequently, healthcare costs related to Medicare are projected to drive our nation debt even higher in the coming years. The current trend of medicare costs is unsustainable. In order to address the cost driving factors, an atomistic approach is necessary. The observations will be sorted and grouped by claim types, regions and states. This approach will shed light on states that are costing less to the Centers for Medicare & Medicaid Services in relation to the state’s population size. This analysis will provide decision makers with insights on states overall Medicare cost performance. Additional researches should be commissioned at each state level to understand the difference of performances.

Tools to be used for Analysis

For extensive analysis of the Medicare Hospital Spending by Claim data, the following packages will be utilized:

dplyr
tidyverse
readxl
ggplot2
gridExtra
DT
rpivotTable

The data set analysis will cover 7 claim types. Please see the claim types and the brief descriptions below:

Carrier claims are non-institutional claims, however this does not mean that they are outpatient claims. Providers, such as physicians, can bill for services provided in the office, hospital, or other sites.
Durable Medical Equipment (DME) is any equipment that provides therapeutic benefits to a patient in need because of certain medical conditions and/or illnesses. It includes, but is not limited to, wheelchairs (manual and electric), hospital beds, traction equipment, canes, crutches, walkers, dialysis machines, ventilators, oxygen tanks, monitors, pressure mattresses, lifts, nebulizers, bili blankets and bili lights.
Home Health Agency are claims for home health services which includes skilled-nursing care, home health aides, physical therapy, speech therapy, occupational therapy, and medical social services.
Inpatient claims are related to patients formally admitted to the hospital with a doctor’s order. They are bills submitted by program level (facility-based) services such as Inpatient Services, Residential Services, Day Treatment Services (Partial Hospitalization Services), Structured/Intensive Outpatient Services (IOP). Additionally, facilities may submit bills for Behavioral Health Assessment in the Emergency Room, Observation Services, and Crisis Services.
Outpatient claims consist of medical cares that do not require an overnight stay in a hospital. The care may be administered in a medical office or a hospital, but most commonly, it is provided in a medical office, laboratory, radialogy, outpatient surgery center, hospital outpatient departments, rural health clinics, renal dialysis facilities, outpatient rehabilitation facilities, comprehensive outpatient rehabilitation facilities, or community mental health centers.
Skilled Nursing Agency claims can include, but is not limited to:
- Periodic nursing assessment of a patient
- Visits for administering medications, including intravenous (IV) medications and chemotherapy.
- Visits for acute illness, postsurgical, and sterile wound care.
- Education of the primary caregiver and patient about the illness processes and medical needs.
- Medical treatments that require the skills of a licensed nurse.
- Transition from an inpatient to a community-based home setting.
Total: Total charges per patient

$~$

Data Cleaning

Applying Threshold of $10

Code for Threshold

#Setting up an arbitrary threshold of $10
new_df10 <- subset (new_df, Avg_Spending_Per_Episode_Hospital > 10)
new_df11 <- subset (new_df10, Avg_Spending_Per_Episode_State > 10)
new_df12 <- subset (new_df11, Avg_Spending_Per_Episode_Nation > 10)

$~$

Variables Retained for Analysis

## [1] "Hospital_Name"                     "State"                            
## [3] "Claim_Type"                        "Avg_Spending_Per_Episode_Hospital"

$~$

The data frame needs to be broken down into different data frames to respond to the need of analysis to be proceeded by Claim Types, Regions in the US, and states.

$~$

Assigning Regions to States

$~$

Regions Description

$~$

Regions Population

$~$

Code Used for Regions Assignment

#Using If statement to group states into their respective region
Up_new_df <- mutate(new_df3, Region= ifelse(new_df3[["State"]] %in% midwest, "Midwest", ifelse(new_df3[["State"]]%in% northeast, "Northeast", ifelse(new_df3[["State"]]%in% south, "South", "West"))))

$~$

Tables Visualization

$~$

Claim Type: Carrier

$~$

Claim Type: Inpatient

$~$

Claim Type: Total

$~$

Claim Type: Skilled Nursing Facility

$~$

Claim Type: Outpatient

$~$

Claim Type: Home Health Agency

$~$

Claim Type: Durable Medical Equipment

$~$

Claim Type: States Population

$~$

Cleaned Data Pivot Table

Graphing Representations

Cost Overview in the US

$~$

First, from the graph below, it can be concluded that the distribution of average spending per hospital in the US by Medicare is not normally distributed. This irregularity necessitates further researches.

$~$

The following graphs represent respectively “Medicare Hospital Average Spending by Claim Type” and “Distribution of Medicare Spending per Claim Types in the US”:

$~$

Second, the two graphs above show obvious differences between box plots and whiskers representing the claim types. Here are some general observations: The box plots for Carrier, outpatient, Skilled Nursing Agency, and Total are comparatively short. This suggests that overall cost for those claims have a high level of similarity across the nation and low variability. However, claims Durable Medical Equipment, Impatient, and Home Health Agency, have larger box plot suggesting higher variability among claim types paid by Medicare nationwide.

$~$

Regions Comparison

$~$

Claim Types in Each Region

A general obversation for all regions is that claim types cost share the same variability, even though the regions do not contain the same proportion of the US population. However, the claim type Home Health Agency has more hospital average costs variability in the West region compared to the remaining regions.

$~$

Region Population Distribution

$~$

When comparing the distribution of hospital cost to Medicare by case types, the graph shows greater spending on Durable Medical Equipment, Inpatient and Carrier. Analyses of claim types per region is necessary to further understand the above observations.

$~$

Claim Types per Region

$~$

From the following visualizations, it can be observed that the Medicare costs for Durable Medical Equipment and Inpatient Claim types have high variability. While the remaining claims have significantly low varibility, with many outliers, for average spendings per hospital.

$~$

Next are visualizations of region total Hospital Average Spending for each claim type. It can be concluded that Medicare spend a higher proportion of its budget in the south region to cover claims previously discussed.

Claims Average Cost per State

$~$

These graphs represents the Medicare average hospital spending in each states for each claim type.

$~$

Analysis Summary

$~$

Initially, states were ranked based on their total hospital average cost. Furthermore, this ranking was compared to states population size rankings. This provides a true sense of how much Medicare is spending on each state based on its population size.

With this further analysis approach, several observations were noted. California is number one in total hospital average cost for most claim types and also represents the state with the biggest population size. This observation is not surprising and is applicable to other states like Texas. However, the state of Illinois although it ranks fifth on the population size chart, for all claim types except home health agency, it does not figure out in the list of the top 5 states which cost the most to Medicare. New York state also cost relatively less to Medicare when taking in account its population size. The same reasoning can be applied to states such as Delaware and Hawaii where the ranks of its hospital average total cost is negatively correlated to its population ranks. In order words, they spend less compared to for instance District of Columbia and Vermont with less population size than theirs.

The ranking is confounding because just because some states such as Wyoming and Florida, are spending more money relatively to its population size does not mean it is inefficient. It could mean they have sicker people or older people. There needs to be further research in those areas.

Please see below a summary of the expenditure of medicare by hospital and how they correlate with the population size in comparaison to the nation population size.

Tables variables explained:

Claims: Claim Types
Cost Rank: States are ranked based on the total of their hospital average cost, the top 5 and buttom 5 states rakings
Population Size Rank: States are ranked based on their population size
Cost vs Population Rank: Does the states hospital average cost rankings match their population size rankings?

$~$

“R Programming Project”

by Sedozan Senou