Using Data to Analyze: How Caring is Home Health Care?

Posted on Nov 4, 2021
An exploration of Home Health Agency data to determine factors associated with patient health outcomes


Home Healthcare provides medical care to patients in the comfort of their own homes. It is typically used for transitional care, allowing patients to return home early after invasive surgery.  But it also plays a role in delivering preventive care to the elderly and frail and has been shown to be effective in reducing mortality and admission to long-term care institutions for the elderly.

Home healthcare spending has grown rapidly. According to the Centers for Medicare and Medicaid Services (CMS), expenditures on home health increased 7.7% in 2019, outpacing the total healthcare spending increase of 4.6% for the same period. However, as evident in the CMS Quality indicators , there are large discrepancies in the quality of care provided by different care for the elderly and frail.

My project begins by exploring the geographic and cost variations among home health agencies before digging deeper to determine factors associated with their patient health outcomes. Full results, interactive maps and charts can be found on the project Shiny App.

Data and Methods

The CMS provides data on its 11,000 Medicare-certified agencies across the country. It includes information on the type of ownership, star-rating, and other quality measures. I combined that with a separate dataset on agency patient experience surveys.

In order to conduct geographic utilization analysis, census data provided me with age-segmented state populations, as well as disability prevalence rates.

Exploratory descriptive analysis was used to examine current geographic utilization rates, costs and agency star-ratings. Kruskal-Wallis ANOVA and the Bartlett test were used to discern significant differences among the types of agencies (privately owned, government owned, or non-profit). Finally, linear regression was used to determine factors associated with patient outcomes, namely acute care admissions rate and discharge to community (DTC) rate.

Geographic Variations in Utilization & Cost

In the 1990s when Home Health Services just started to become more popular, there was dramatic geographic variation in its utilization. The southeast region (TN, MS, LA, AL, GA) had the highest concentrations of home visits. Today, the situation hasn’t changed much. The highest use of Home Health still occurs in the southeast (figures 1 & 2). Some northeast states like Massachusetts, Vermont, and New Hampshire have increased utilization, perhaps due to dense senior populations in those states.

Figure 3 below maps differences in average state costs (as a ratio to the national average). As you can see, state averages range from 0.87 to 1.07, while individual agency costs can range from 0.5 to even 2x the national average. Additional interactive density maps are on the project’s Shiny App.

Figure 3: Agency Costs as a Ratio vs. National Average

These cost differences may be due to the competitive environment in the agency’s home state, and past studies have also shown regulations called certificate of need (CON) laws greatly affect agency costs. In a later part of the analysis, I examine if patient outcomes were associated with cost discrepancies.

Star Rating

The CMS uses a curved rating system; only top performers in its 9 measurements of quality can receive high ratings. Only 15% of agencies nationwide received 4.5 or 5 stars (figure 4).

For agencies with a 3-star rating or better, the Discharge to Community rates are equal (~80%). However, outcomes start to worsen for poorly rated agencies. We also observe a slight drop in Patient Experience survey results for poorly rated agencies (figures 5 & 6).

Interestingly, there was a bimodal distribution in the type of agency in relation to star-rating (figure 7). Agencies with both high and low star ratings are mostly privately owned, while those with medium ratings are not. The next sections involve statistical analysis and we begin with determining how costs and patient outcomes differed between types of agencies.

Figure 7: Percentage of Privately Owned Agency by Star Rating

Type of Agency Ownership

The variance of costs differed significantly between types of ownership. Private and government owned agency costs varied greatly, while non-profit agency costs ranged closer to 1.0. The Kruskal-Wallis test showed means were significantly different, and we can observe from the ridge-line chart on the right that private agencies had higher mean costs (figure 8).

The mean discharge to community rates differed significantly among the different types of agency ownership. Non-profit agencies typically performed better. Mean acute care admissions rates between agency types were also significantly different, with private agencies performing better (figures 9 & 10). 


Patient Outcome Measures

Linear regression was used in an attempt to determine factors associated with home health agency patient outcomes, specifically acute care admissions rates and discharge to community rates.

Using nation-wide data points resulted in extremely low R-squares. Perhaps the regulatory environment or market conditions are too different between states to conduct any inter-state comparisons. Instead, I ran separate regressions for data from two individual states. I chose Texas as it had the most agencies (1400+), and Florida, the state with the 3rd most agencies (~800), and known for its high senior population density. Here I will summarize results from Florida, for full results, please refer to the Shiny App.

Notes: Step-wise feature selection (Forward BIC) was used to find appropriate independent variables for each model. When star-rating was included as an independent variable, other features that went into star-rating calculation were excluded and vice versa.

Acute Care Admissions

With acute care admissions as the target dependent variable, forward BIC selected Cost and Communication (% patients that reported their health team communicated well with them) as features. The β coefficient for Communication was -0.055 (p<0.01). That means for each 1 percent improvement in that factor, admission rate was reduced by 0.05 percent. The relationship is not large, which explains the low R2 of 0.122 (although it was still an improvement from the nation-wide regression).

The β coefficient of Cost was 10.15, which is not ideal for patients.  A 1 unit increase in cost, meaning starting from 1 and going to 2 - double the national average - actually increased admissions rates. Prediction bands are shown in figure 11 below.

This result actually points to a major limitation to my analysis and deficiencies of the data set. The positive relationship between an increase in costs and increase in admissions risk may be due to a home health agency’s patient case-mix. Patients with complex disease and higher risk are charged more for services. If agency data included information on patient disease, it would be better to conduct separate analysis for each patient risk level.

Figure 11: Acute care admission regressed on Cost, with prediction lines

Discharge to Community Rate

As for discharge to community (DTC) rate, forward BIC selected Cost and Bed (how often patients got better getting in & out of bed independently). The β coefficient for Bed was 0.24 (p<0.01) which meant a positive association between improving patient bed mobility and improved agency DTC rate. The β for Cost (-28.8) was again pointing to a lack of data on patient risk profile, as increasing agency costs worsened the expected DTC rate. This could also mean agencies were used by some patients as a replacement for institutionalised long-term care, thus delaying discharge back to the community.


Our within-state regression analysis showed some association between patient experiences regarding caregiver communication and acute care admissions. Home healthcare teams should emphasize better communication with patients and households. Healthcare teams could also focus on improving patients' ability to get in and out of bed as this factor was correlated with better discharge to community rates.

The within-state analysis showed that more expensive agencies correlated with higher acute care admissions and lower DTC rates. This means at-risk patients with complicated illnesses pay more and may even be using home health as an alternative to institutional long-term care. This could be a desirable alternative.

Recommendations to Home Healthcare Patients

Agency Star Ratings are given by the CMS and not by patients. It correlates well with quality of care. Patients are advised to hire an agency with at least 3 stars

Patients and family members should make sure agency caregivers spend time communicating with them about their illness, pain, and medication. If the patient or caregiver feels dissatisfied, they should  ask the agency or doctor to request a replacement for the home health team.

Limitations & Further Analysis

Regression analysis did not yield strong results. This could be due to missing detailed agency data, such as: (1) the exact number of home health staff or teams, (2) team composition, e.g. skilled nurses, home aides, etc., and most importantly, (3) agency patient case-mix or risk profiles.

As next steps to the study, if agency patient information is available, studying agency features against patient outcomes for specific diseases, like diabetes mellitus, COPD, or cancer, could yield valuable results.

The skills I demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.

About Author

Daniel Nie

Data Scientist with background experience in both Healthcare Administration and Finance. A versatile thinker that enjoys deep data exploration and generating business value with machine learning in both Python and R
View all posts by Daniel Nie >

Related Articles

Leave a Comment

No comments found.

View Posts by Categories

Our Recent Popular Posts

View Posts by Tags

#python #trainwithnycdsa 2019 2020 Revenue 3-points agriculture air quality airbnb airline alcohol Alex Baransky algorithm alumni Alumni Interview Alumni Reviews Alumni Spotlight alumni story Alumnus ames dataset ames housing dataset apartment rent API Application artist aws bank loans beautiful soup Best Bootcamp Best Data Science 2019 Best Data Science Bootcamp Best Data Science Bootcamp 2020 Best Ranked Big Data Book Launch Book-Signing bootcamp Bootcamp Alumni Bootcamp Prep boston safety Bundles cake recipe California Cancer Research capstone car price Career Career Day citibike classic cars classpass clustering Coding Course Demo Course Report covid 19 credit credit card crime frequency crops D3.js data data analysis Data Analyst data analytics data for tripadvisor reviews data science Data Science Academy Data Science Bootcamp Data science jobs Data Science Reviews Data Scientist Data Scientist Jobs data visualization database Deep Learning Demo Day Discount disney dplyr drug data e-commerce economy employee employee burnout employer networking environment feature engineering Finance Financial Data Science fitness studio Flask flight delay gbm Get Hired ggplot2 googleVis H20 Hadoop hallmark holiday movie happiness healthcare frauds higgs boson Hiring hiring partner events Hiring Partners hotels housing housing data housing predictions housing price hy-vee Income Industry Experts Injuries Instructor Blog Instructor Interview insurance italki Job Job Placement Jobs Jon Krohn JP Morgan Chase Kaggle Kickstarter las vegas airport lasso regression Lead Data Scienctist Lead Data Scientist leaflet league linear regression Logistic Regression machine learning Maps market matplotlib Medical Research Meet the team meetup methal health miami beach movie music Napoli NBA netflix Networking neural network Neural networks New Courses NHL nlp NYC NYC Data Science nyc data science academy NYC Open Data nyc property NYCDSA NYCDSA Alumni Online Online Bootcamp Online Training Open Data painter pandas Part-time performance phoenix pollutants Portfolio Development precision measurement prediction Prework Programming public safety PwC python Python Data Analysis python machine learning python scrapy python web scraping python webscraping Python Workshop R R Data Analysis R language R Programming R Shiny r studio R Visualization R Workshop R-bloggers random forest Ranking recommendation recommendation system regression Remote remote data science bootcamp Scrapy scrapy visualization seaborn seafood type Selenium sentiment analysis sentiment classification Shiny Shiny Dashboard Spark Special Special Summer Sports statistics streaming Student Interview Student Showcase SVM Switchup Tableau teachers team team performance TensorFlow Testimonial tf-idf Top Data Science Bootcamp Top manufacturing companies Transfers tweets twitter videos visualization wallstreet wallstreetbets web scraping Weekend Course What to expect whiskey whiskeyadvocate wildfire word cloud word2vec XGBoost yelp youtube trending ZORI