For data-driven project on hospital ERs and psychiatric care, flexibility proves key

I thought the plan was ambitious. My editor called it “sorta crazy.”

I took on two projects for my 2018 Data Fellowship. The first looked at whether emergency rooms drive hospital profitability, while the second examined long waits for psychiatric patients in need of emergency, inpatient and post-hospital care.

Those are two big topics. My editor and I wondered if I bit off too much.

But tackling two ideas worked well, because one didn’t unfold as envisioned. Therein lies one of my biggest lessons: Be flexible. Work on multiple fronts. I was locked into the mindset that I would complete the ER project and then move on to psychiatric services.

My determination to go in that order cost me time. I should have recognized the ER project was crawling along and dove into behavioral health sooner.

My reporting journey came with a few other lessons along the way.

Last November, fresh off the fellowship conference, I sought to get at the financial impact of more ER visits. In hand was data showing an 18% jump in ER visits in San Diego from 2012 to 2017, reflecting a statewide trend.

Since my outlet is a business journal, we wanted to know, does a busy ER boost the bottom line?

The answer appeared to be in a dataset from California's Office of Statewide Health Planning and Development, or OSHPD. Except OSHPD — and other experts — told me the state data gives an incomplete picture of ER finances.

While I was discouraged, hope came from a 2003 paper that illuminated how ERs influence California hospital finances.

The paper smoothed over state data, which misses key sources of revenue and costs, through regression analysis and other techniques.

Could I re-create this methodology? A researcher who led the paper’s data analysis proved difficult to track down, in part because he shared a name with another university researcher, and traces of him on Google were false leads.

After finally getting ahold of him through some detective work, he said I’d need at least a year and advanced data skills.

I possessed neither. After all, I entered this fellowship as an Excel novice. Plus, the 2003 paper may not have factored in some physician costs.

Other data sources don’t break finances down to the ER level, experts told me. Additional ER data was missing, too. It was December at this point.

Time to get going on the other project, I decided.

The idea for it came after Tri-City Medical Center in San Diego closed its psychiatric units, sparking a larger conversation about deficient mental health services in San Diego County.

A number of hospital, nonprofit and county officials told me it wasn’t just about the lack of inpatient beds. Behavioral health patients struggled to access emergency services and continuing care at places such as board-and-care facilities.

There was a larger story about a broken continuum of care. How long do patients wait at various stages, and what’s the financial impact on hospitals and other facilities?

These were timely questions. The closure of Tri-City’s units prompted a growing community outcry, another reason for pivoting to the topic.

Answers would require public records requests. I kicked myself for failing to file those earlier. This could have been done in October and November during lulls in the ER project. I was too rigid in my thinking.

One mistake led to another. Feeling a need to make up for lost time, I fired off records requests to gauge bottlenecks in care.

Some of the requests should have been better thought out.

No surprise, then, that they yielded data that wasn’t useful. Compounding the error, these requests went to several counties.

Granted, psychiatric care is a complex topic. But this was an example where taking a little more time can save a great deal of time down the line.

For knotty topics, I now sit on a records request for a day or two before submitting. Asking myself, do I understand this well enough to put in a request? Does this fit with my reporting goals? What can be clarified or added?  

One thing I did right, however, was keep a data diary, among the valuable tips from the fellowship conference. A day’s entry might contain who I spoke with, their contact information, data requested, and work done on an Excel sheet. Key definitions would be in there too.

Like many other fellows, I was juggling daily reporting demands while reporting on this project, making the diary invaluable. So many times I turned to the diary, saying aloud something like, “Ah yes, that’s what an IMD bed is.”  

My digging on psychiatric services ultimately generated a three-part series. In the interim, I found more ER data.

The data didn’t show the financial impact of more ER visitors. But based on the data obtained, there was a story there on how hospitals have responded to increased ER volumes.

The project gave me a chance to work with larger datasets. Outside of the fellowship conference, I had yet to use key Excel functions like “VLOOKUP,” which combines matching items in different datasets into columns.

But I was rusty. That brings me to another big takeaway: seek and get help. Don’t be ashamed to do so. YouTube videos demonstrated VLOOKUP and other functions. And my mentor, Meghan Hoyer, guided me through tricky data questions. Anytime there’s uncertainty about the integrity of data, it doesn’t hurt to run it by someone.

Lastly, academic and industry researchers are often happy to check your work. Hat tip to past fellow Jenna Flannigan for sharing that advice, something I used in my data reporting.

The ER article didn’t uncover as much as hoped, but it taught me some valuable Excel skills.