Data Quality - A Tale of Two Hospitals
I’ve talked about how it’s important to understand the deal when using software and how you also have to understand if using the software comes with a catch. That’s great for programs and apps that you use by yourself. What about larger systems that are used by lots of people. Those that help run whole business or organizations? An important consideration is to know how your work in the system impacts others. People do better with large systems if they know what they are doing and why. This was brought home to me years ago when I was working at a hospital that got taken over by another.
Hospitals in Ontario are independent organization that are government funded. Over the last few decades there has been a push to amalgamate hospitals into larger single regional organizations. This is supposed to save money and be more efficient. I’m not sure if that ever worked out or not.
I was working at a small hospital in north Toronto. I’ll call it hospital A. The politics and history are complicated but it turned out that essentially hospital A was merged with hospital B. A larger busier hospital about 10km away. I say merger since that was the official stance at the time - but in reality it was a government mandated takeover of A by B. Not that I’m bitter.
After the takeover hospital A was slowly turned from an inpatient facility with its own emergency department into a very large outpatient and ambulatory care centre. This took years to happen - and many more years than expected. One of the reasons it took so long was the hospital wide computer systems at A and B.
Small hospital A had a hospital wide computer system that had been installed and running for years. It tracked everything from important lab tests to less important things like which beds and rooms needed to be cleaned and prepared for the next patient. It hummed along and worked like a charm.
Hospital B had several older and not well implemented systems. Hospital B had also just purchased its own system. The idea was that hospital A would transition to hospital B’s new system as it was implemented. As the new system at B had enough functionality to replace the system at A - it would take over and system A would be decommissioned. This took a lot longer than expected. Replacing a system that had been implemented over many years took a while.
When the transition finally happened I was part of the small group handling the migration of the historical patient information from system A to system B. That’s when I noticed the complete and marked difference in data quality between the two hospitals. It all had to do with the numbers assigned to patients.
Here in Ontario citizens have a few government assigned numbers. We have our Social Insurance Number (roughly the equivalent of a Social Security Number in the US). We also have a number on our Ontario Health Card. Hospitals use the Health Card Number to bill the government for services provided and to report on patient care - but hospitals do not use the Health Card Number as the primary way of identifying a person. The Health Card Number is meant to be kept to yourself and only used when necessary. And some patients don’t have Health Card Numbers. If a patient is from another country or another province they won’t have one. Newborns in the Labour and Delivery ward aren’t assigned a number until after they’re born and even then it takes a while. Yet every one of these patients need to be tracked inside the computer. When the hospital prints out a wristband with your name, a number, and a barcode - they don’t use the Health Card Number - the hospital computer system assigns its own number to you.
The number assigned to you is usually just the next available number. If the last number assigned by the hospital was 5,678 then the next patient gets 5,679. Different hospitals have their own separate numbering systems. Patient 1,000 at one hospital is going to be a different person than patient 1,000 at another.
The issue I saw at Hospital B was that they were not good at re-using a patient’s number when the patient showed up for another visit.
Here’s how things should work:
- You show up to a hospital for the first time.
- The admitting clerk checks the system and doesn’t find you.
- The clerk goes through the process of entering all your information and the system assigns you a number. Let’s say 100.
- All activity related to you during your visit is attached to patient number 100.
So far so good.
- You show up for a second visit.
- The admitting clerk checks the system and finds you listed as patient 100.
- The clerk does a quick check to make sure your name, information, and address are correct - and updates the system if they aren’t.
- All activity related to you during this second visit is attached to patient number 100.
That’s in a perfect world. Turns out that there are times when you may get admitted into the hospital and given a second number.
If you are rushed to the emergency department unconscious and in need of immediate medical help - no one may be able to figure out that you are patient 100. In that case a new number is assigned. Let’s say 500. All the lab tests, medical imaging, and drugs you need are tracked using patient number 500. Sooner or later, when staff figure out who you are, they will find that you are currently patient 500 but also patient 100. The two numbers will then be merged.
What this means is that the system is told that number 500 should really be number 100. From now on all activity is tracked against number 100. Any information tracked against number 500 is treated as belonging to number 100 instead. The system tracks the two numbers something like this:
- 100 - Patient Jane Doe
- 500 - (Not in use - use #100 instead)
or like this:
- 100 - Patient Jane Doe
- 500 - secondary number for this patient (who is #100)
No matter how conscientious the admitting clerks are - new numbers will be assigned to patients who are already in the system. It happens. Between the Admitting department and the Health Record department - these duplicates are usually caught and merged together. Until they are though - things are a bit wonky.
If you are in the system as patient 500 - then all the history you have as patient 100 isn’t available to clinicians. Until those two numbers are merged your history is in two separate chunks. When they are merged the doctors can see all your information in one place.
Going back to the two hospitals - Hospital A had a very clean set of data. There were duplicate numbers that had been created - but not many. Yes - some patients had two numbers. A rare few had three. Those were the exceptions.
Hospital B on the other hand was a mess. Two patient numbers? Sure - but also three, four, and five different numbers merged together. Patients had been admitted into the system five times. And this wasn’t just for a few cases. Duplicate admissions were rampant.
What’s the problem? The computer can handle the merges without difficulty. After patients are merged their history is in one place. I can ask for the history of patient 100 or patient 500 - and I’ll get the same result - the entire hospital history of patient 100. Well - not so fast. The biggest issue is that if there are that many patient records that had to be caught as duplicates and then merged - how many have not been caught yet? How many duplicates are still in the system? What information are clinicians not able to see because the records are not clean?
At the time I wondered why the two hospitals had such different levels of data cleanliness. One was practically pristine the other was… well slovenly.
It came down to a number of things:
- Hospital B was busier than A and saw more patients.
- Hospital B saw many more patients in its emergency department. Increasing the chance of needing to use a temporary number.
- Hospital A had centralized almost all the admitting. There were only one or two places in the hospital patients went to be admitted. Before your appointment you would go to those places to be admitted into the system. The admitting staff did most of the work checking you in to the system.
- Hospital B had a central admitting area - but it also had staff all across the hospital admitting patients. You would get admitted at the clinic or area you were visiting in many cases. These were not staff that were part of the admitting department. They were the clerks working in those various clinics around the facility.
- Hospital A made properly selecting a patient’s existing account part of the job of admitting patients. Staff were trained to check and double check. If a patient said they had been at the hospital before - there was no reason a new number should be assigned.
- Hospital B didn’t make this as much a part of the admissions process and didn’t hold their staff to the same level of correct data entry.
But there was one more reason. I think it was the most important one of all. Hospital A admitting staff were told why re-using a patient’s existing number was so important. It was part of their training, part of the expectation of doing their job correctly, but it was also made clear why they needed to admit patients properly.
When admitting staff were trained at Hospital A the class would start with an overview of how the hospital wide computer system worked.
- A quick diagram was drawn on the whiteboard showing how the list of patients, each with their own number, drove everything else in the system. From lab tests, to orders for care, to medical imaging, to… well… you name it. They were shown how everything from patient care to reporting to the Provincial authorities relied on the patient list.
- They were shown the work required to merge patient records. Whether it was someone from Health Records or one of their colleagues in Admitting - they were shown that it was a finicky process that took time and effort to do correctly and that the work could be avoided if a patient was admitted using their existing number.
- The clinical impact of not having all a patient’s information available to clinicians was stressed. They were shown the clinicians view of a patient. How all the information, from this visit and previous ones, was instantly available. It was stressed that a person’s medical history could be incredibly important for their care.
The result? Staff knew what was expected of them and why. They understood how doing their job well helped. Helped patients, clinicians, and other staff. They could help all these people by keeping the patient list as clean and accurate as possible. This wasn’t just something they were supposed to do or something they were told to do. It became something they did willingly because they wanted to. People want to help. People will do work if they understand why they are doing it. It became a source of pride that duplicates almost never happened.
We tend to think of people as only wanting to do the bare minimum. That they won’t go above and beyond. It’s just a job and that’s all there is to it. Not true. Give people a reason and they will be more willing to do what is asked of them. Let people know they are helping others and they will rise to the occasion.
Please - let people know why they are doing their jobs. Give them a reason for the work. Give them the context and let them understand. People want their work to be useful. They want to help. If they don’t understand the why then it just becomes a job that gets them paid - instead of a paying job that has a purpose and is useful.
It doesn’t take much. Just let people know the reasons their work is important and useful.