This marks the first post in the Dataset of the Week series [github] – where I take a little time each week to explore a dataset that I’ve never seen before. I’m not sure why this dataset stood out to me as something to explore. I am sure that my wife’s fascination with the lackluster care that we so often provide our elderly played no small part. With a fairly regular cadence, she espouses the need for a great shift in elderly care and ponders making a career move to take up the cause. Perhaps when I found this dataset, I saw it as a way to showcase a small part of that story of waning care for our older family members and neighbors in need. Perhaps it was a way to prepare myself to be more supportive and proactive in her career shift. Whatever the reason, as soon as I saw a dataset that highlighted the deficiencies of nursing home care, I knew I wanted to dig in and explore. Honestly, I could have spent a lot more time with this data. I found it fascinating and it kept bringing up new questions. I even went so far as to find an additional dataset that tells a little more about the financial side of this story (see: https://data.medicare.gov/Nursing-Home-Compare/Penalties/g6vv-u9sr).
The data this week comes primarily from a ProPublica investigation that has been ongoing for years. They provide a number of Excel sheets containing information on the facility (including name and location) and the deficiency (severity, date, inspection notes, and tag code). In order to do some quick analyses on state-level data, I also pulled in population data from the US Census Bureau. This is simply a state-level population estimate by year. If you want a little more information about the datasets, check out the Jupyter notebook which contains all of the code for this analysis.
I had very little idea what the dataset would include when I first loaded it, so I did not go into this with any hypotheses. I just went were the data exploration took me. I was initially drawn to the question of where do things look the worst? Why not look at where things look the best? I guess I like to look at the low hanging fruit of where we can make the most impact. There are several ways to answer that question, so I first focused on the total number of deficiencies by facility.
Facilities to Avoid
| Facility Name | Total |
| MILLER’S MERRY MANOR | 427 |
| MANORCARE HEALTH SERVICES | 210 |
| GROVE AT NORTH HUNTINGDON, THE | 175 |
| APERION CARE BLOOMINGTON | 171 |
| COMMUNITY CARE CENTER | 169 |
At first, it looks like Miller’s Merry Manor is an absolute nightmare with twice as many deficiencies as any other facility. But then I realized that facilities are more accurately described as companies. While Miller’s Merry Manor has twice as many deficiencies, they also have twice as many unique location (27, with Manorcare in second place at 10). To get a true count by unique location, I combined the facility name and city. When we do this, we see that #3 on the list above is the true winner loser with only one location and 175 deficiencies.
| Facility Name | Total |
| GROVE AT NORTH HUNTINGDON, THE (N. HUNTINGDON, PA) | 175 |
| APERION CARE BLOOMINGTON (BLOOMINGTON, IL) | 171 |
| LAKEWOOD HEALTHCARE CENTER (DOWNEY, CA) | 150 |
| RIVERSIDE POSTACUTE CARE (RIVERSIDE, CA) | 149 |
| CORONA POST ACUTE (CORONA, CA) | 143 |
| GARDENS AT WEST SHORE, THE (CAMP HILL, PA) | 141 |
| CHAMPAIGN URBANA NRSG & REHAB (SAVOY, IL) | 137 |
| PARAMOUNT REHABILITATION AND NURSING (SEATTLE, WA) | 136 |
| ASHLAND NURSING AND REHABILITATION (ASHLAND, VA) | 128 |
| GARDENS ON UNIVERSITY, THE (SPOKANE, WA) | 127 |
While The Grove and Aperion have been flagged by the federal government for serious quality issues, it is worth noting that nearly all of their deficiencies have a low severity score. The Grove only has two severe deficiencies and Aperion just one. Maybe we also need to avoid facilities with severe deficiencies.
It will help give us some context if we have an idea of the distribution of deficiencies by severity score.
Now we can see that the 23 severe deficiencies from River Haven is a real outlier with nearly all facilities having no severe deficiencies. It also puts the seemingly small counts from The Grove and Aperion into context: although small, they are still outliers. It should be expected that no severe deficiencies should be accepted. Further investigation into the Centers for Medicare and Medicaid Services definitions of severity scope revealed that these severe scores indicate a pattern or pervasive “immediate jeopardy to resident health or safety [which] is likely to cause, serious injury, harm, impairment or death to a resident receiving care in a facility or an employee of the facility. ” So let’s give severe deficiencies a one strike rule.
States to Avoid
We can zoom out real quick and ask: are there states that seem to be worse for those in nursing homes? If the answer is yes, then additional pressure can be put on the state-level legislators to make changes to improve care and enforcement in nursing homes. The cheap question is which states have the most number of deficiencies?
But you can clearly see that these counts are biased by a states population. And it stands to reason that a state with more people has more elderly people, which means more nursing homes. Although New York already seems to be doing well for itself, lets look at the number of deficiencies per 10,000 residents to get a relative count.
And there you have it. Those states with larger populations moved more toward the center and we can see that New York is doing really well when it comes to a relative count compared to other states. That leaves us with 9 states that are more than one standard deviation from the mean: North Dakota, Nebraska, Kansas, Iowa, Indiana, West Virginia, Missouri, Wyoming, and Montana. And it looks like the place to be in a nursing home might be Puerto Rico. But there are also six other states that fall well below the mean if you aren’t looking to leave the mainland.
Are Things Getting Better?
One would hope that with increased scrutiny and penalization, that the number of deficiencies would decrease over time. However, it could stand to reason that as our population ages and nursing homes fail to keep up with staffing needs that perhaps safety standards are also falling behind. So let’s run a quick regression.
Phew! It looks like we do appear to be improving over time, although I do wish that confidence interval were a little tighter. Let’s keep the pressure up and do our part to make sure that care for the elderly continues to improve.
While I still have a lot of unanswered questions, I want to keep moving on to explore more datasets. Perhaps in the future I will circle back to nursing home data, but for now:
