Is your local hospital stingy or generous with charity care?

For decades, study after study has tried to get at the question of how much charity care hospitals were providing, or whether they were providing enough.

But whether they were from Congress, academic researchers, or the hospitals themselves, the studies have nearly always been forced to use incomplete datasets that were based on either anecdotal or voluntary reporting, or the data comingled charity care with bad debt in a figure termed “uncompensated care.”

Because few states mandated reporting on charity care, it was impossible to truly compare one state to another, or properly examine hospitals within a state, with rare exceptions. There was no national dataset on every hospital’s charity care spending.

My home state of Pennsylvania was one of those states where a good state data set on charity care was impossible to come by, too. I wrote my part of our 2014 series Poor Health in the Pittsburgh Post-Gazette, in part, to get some sense of how hospitals in our region were — or were not — providing charity care to the poor.

The Poor Health series, funded by a grant from the Dennis A. Hunt Fund for Health Journalism and the 2013 Center for Health Journalism National Fellowship, used a publicly available dataset for hospitals on uncompensated care.

That dataset alone was informative. We found that all but one of the 11 hospitals that closed in the Pittsburgh region between 2000 and 2010 provided significantly more uncompensated care than the average hospital here, reflecting the burden those hospitals shouldered to care for the poor.

My new project — Counting Charity Care — is a direct descendant of Poor Health. I pursued it after first getting a state-based set of data, and, later, a national set. Getting and analyzing the data was a struggle, but well worth the effort. These new data sets — publicly available through links at the end of this story — allow journalists, and readers, to sort and analyze the data for hospitals in Pennsylvania in the state set, and every hospital that has filed data from 2011 through 2015 with the federal government. I’m sharing our work with the datasets in the hopes of inspiring others to try similar project in their own communities. It allowed me to show that our state’s hospitals provide very little in the way of charity care, and the nonprofits provide significantly more charity care than for-profits. Hopefully others will find something just as impactful for their communities.

A long road to missing data on charity care

It started early in the summer of 2014, when I was deep into writing Poor Health and needed to check a piece of data in my story on uncompensated care. I went back to one of the original financial reports from the Pennsylvania Health Care Cost Containment Council (PHC4), a state agency, to make sure I got it right.

For the first time, I noticed in PHC4’s annual report on hospital finances that it had annual figures of the share of uncompensated care — both for charity care and the share that was bad debt — for all the state’s general acute care hospitals combined.

I wondered: If they had those two figures for all the hospitals combined, did they have it for each hospital individually?

As far as I knew at the time, the only national data on charity care was on the IRS’s 990s Schedule H that nonprofit hospitals fill out. But even then, the IRS data did not include the rising number of for-profit hospitals. In addition, some nonprofits do not file for individual hospital buildings, but rather as systems that combine hospital finances, including charity care, making it impossible to compare one hospital to another within the same system.

Reviewing the 20 years of other hospital financial data PHC4 had online, I had never seen the council list an individual hospital’s charity care or bad debt figures, though it listed lots of other hospital-specific information, such as net patient revenue and operating expenses. Maybe this annual combined figure for charity care and bad debt was just some kind of industry estimate?

I was too close to publishing Poor Health to hope to get such data. But after I finished writing the that series, I sent an email to the agency’s spokesman asking if they had hospital-specific information on charity care and bad debt and, if so, can I have a copy?

“I’ll investigate and get back to you,” he said.

It took awhile for them to decide. But in the fall of 2014, I filed a Right-to-Know request, and PHC4 charged us $200 for our “special request.”

In early December 2014, I received the data, becoming the first reporter to ever have access to this unique data set that I hoped would reveal stories about how our hospitals were — or were not — providing charity care to the poor in our state. At first the agency provided data from 1995 to 2012, and over the next few months I also got data from 2013 and 2014, forming the statistical basis for my Counting Charity Care series.

Of course, the first problem I encountered, as is too often the case with hospital-reported data, was that the nearly 200 hospitals that have operated here since 1995 reported charity care and bad debt data to the state using their chargemaster price. That’s the overly inflated figure that hospitals use to negotiate with insurers, but that few people actually ever pay for a procedure.

I had to find some way to convert it to a more realistic figure. And using a combination of the publicly available data and what the state sent me, I was able to do just that. (There is an explanation of the formula at the end of the second story in the series.)

The end result was that I could say what percentage of net patient revenue each hospital in the state spent on charity care using realistic figures. This would allow me to compare hospitals big and small, urban and rural, to each other.

The state data, with its longer timespan, also led me to two crucial stories that we have yet to publish: First, how the use of a relatively new credit score-like algorithm known as “presumptive eligibility” is increasingly being used by hospitals here and around the country to increase the amount of uncompensated care they claimed as charity care. Second, a story about the rise of for-profit hospitals in Pennsylvania and how they’ve decreased charity care when they have bought formerly nonprofit hospitals.

With state data in hand, what about the national view?

With all of these stories, though, I struggled to put our state in perspective. How did Pennsylvania compare to other states in our region or around the country?

The search for good data was daunting.

But a source told me that because of the Affordable Care Act, the Centers for Medicaid and Medicare Services (CMS) had begun collecting charity care and bad debt data from every hospital in the country in 2011 – and the dollars were not “at charges” but using a cost-to-charge ratio similar to Pennsylvania’s revenue-to-charge ratio.

The numbers would not be the same as our state-collected data, but they would be standardized across the country and good enough to compare states and types of hospital systems with a true national set of data for the first time.

The data was collected via a form called – in a name only a government bureaucrat could love — CMS-2552-10.

My first attempt at getting all of the national data for all of the hospitals led me to CMS’s website, where I found a page that appeared to contain all the data I hoped to use to explore charity care.

I immediately struggled to get the data open and searchable, though. I use a lot of data for my stories, and confess to have gotten used to easily downloadable sources that take a few clicks and merges before you have massive files telling you stories.

This was not one of those easily downloadable sources. I asked for help from a web content producer at the paper. She spent days trying to work the various links in the CMS site to open and connect the different files with no luck and the agency did not answer our requests for help.

Private firm offers rope out of data quagmire

Somewhat in desperation, I reached out to Research & Planning Consultants (RPC) of Austin, Texas, one of several private, commercial health data operators out there that sell raw, federal data primarily to health care companies and litigators.

I had found them earlier when I was looking for comparative state data. They initially provided some of the basic CMS data to me for free, and said to check back if we wanted more — for a fee.

This time I was asking for more extensive data and their minimum price was $500. I was convinced we would eventually figure out the CMS site, but now I was getting closer to the end of my reporting on the state-based data and several more days of trying to get it to work for us by ourselves was looking less and less worthwhile.

So, for $500 we were able to get the raw data on every hospital in the United States for charity care, bad debt, operating expenses and net patient revenue. I had never dealt with the company before, so, to double-check their data set, I used a subset of the CMS data set I had from a source to see if it matched up to what RPC had sent me. It did. Down to the last dollar on every number.

Confidant in the accuracy of the data, I dug in. To refine the data and compare apples to apples, we excluded any hospital that did not file all the data each of the four years. That gave us about 4,500 hospitals nationwide from the more than 5,000 that had reported at least one year’s worth of data.

(Note to reporters: We have both the Pennsylvania data from the state, and the data on every hospital who reported to CMS, in searchable and downloadable Google docs in links with each story if you’d like to utilize either or both for your own work. This searchable dataset shows the raw data for every hospital that reported charity care and bad debt to the federal government. This searchable data set shows annual averages by state and by year.)

What does the data actually tell us?

Once we had that, my first goal was to see how Pennsylvania compared to other states in 2014, the most recent year I had for state data. It showed Pennsylvania was, as most experts believed, surprisingly low in charity care costs compared to other states – in fact, in the bottom nine lowest average charity care costs per hospital for all four years that CMS had data (2011-2014).

That was incredibly helpful, but along the way I noticed something else: When I got the national averages for each year, there was a steady rise in the first three years, and then a fall in 2014. It seemed possible that that was because of the Affordable Care Act, which began offering marketplace insurance and expanded Medicaid at the beginning of 2014.

We then took a closer look at the averages for states that adopted the ACA at the beginning of 2014 versus those that did not. The results were even more definitive: While states that did not adopt the ACA at the beginning of 2014 saw their charity care and bad debt continue to increase, those that did adopt it saw bad debt and charity care fall significantly.

It confirmed multiple anecdotal studies, both from government analysts and private researchers who believed that that ACA had had such an impact.

And once again it demonstrated how doing the work to get to the best possible data can lead to stories you never intended.

Sean D. Hamill is a reporter at the Pittsburgh Post-Gazette who covers health care. 

Explore the data!

1) Click here to access a searchable dataset with raw data for every hospital that reported charity care and bad debt to the federal government.

2) Click here for a searchable data set of annual averages of charity care and bad debt by state and by year.