Future of Data Analytics in Healthcare: Part II
In my previous blog, I reviewed the framework of data analytics and how changing clinical practice requires first measuring it. Not only do you have to measure the thing you want to change, but you must be willing to accept the current reality. We’ve heard many providers say “the data is not accurate and does not apply to my patients.” The biggest obstacle to change is the unwillingness to face current reality.
As previously mentioned, before you start building a database, you need to know what questions or problems you’re trying to solve. For each question or problem, you need to ask key questions such as:
- What type of data output or view of the data do I need to measure the change?
- What data elements should I collect?
- How often do I need to collect these data elements?
- What is the structure of the data elements? Are they “fixed” or are any of these “free text?”
- Can I collect this data element in the normal course of patient care using the Electronic Health Records (EHR)?
What type of data output or view of the data do I need to measure the change?
For example, if I’m looking at a numeric data element over time, I need to view the data as a graphical chart. This allows me to view changes or trends in data over time. Another example could be related to simple “yes-no” questions – like asking the patient if they smoke. In this case, you want to know the percentage of patients asked the question. You are looking for a simple bar chart showing percentage of patients asked versus total patients on the vertical axis (hopefully driving toward 100 percent). So, before collecting data, think about the data view necessary to implement the desired change.
What data elements should I collect?
What seems like a simple question can be deceiving. There is good news and bad news when it comes to an EHR. The good news is the EHR can be setup to collect many data elements. Unfortunately, that’s the bad news as well. With too many data elements, you may introduce errors into the database, since you are unable to review and correct aberrant data. For a database to be useful, you must dedicate time to keep the data clean. If you are just starting out in data analytics, I recommend you be specific on exactly what data elements you want to collect to solve a problem. Be focused and you will have better results.
How often do I need to collect these data elements?
Again, depending on the problem to be solved, the frequency of data collection can vary. The temptation with an EHR is to collect all of the data elements all of the time. The problem with this approach is you will quickly populate a database, causing the unnecessary purchase of more electronic storage. Oftentimes, you find out later you have data elements you’ll never use. This is a waste of time and money. So, if you have answered the first question above accurately, you will have a much better idea of how often to collect the data element.
What is the structure of the data elements? Are they “fixed” or are any of these “free text?”
This is a basic question to ask, since you can only do data analytics on “structured fixed” data elements. Many EHR’s allow for the entry of “free text.” I highly recommend you convert as many free text fields as possible to structured data. A major benefit of using structured data is it forces more standardization with physicians/providers. That is, instead of allowing five terms for the same clinical condition, you have structured the data to allow for only one term. In the long run, this greatly simplifies the ability to extract meaningful data from the database.
Can I collect this data element in the normal course of patient care using the EHR?
The biggest problem with clinical healthcare data is the process of collecting the data. I’ve heard many times “just show me the outcomes,” without any thought as to what it takes to collect the inputs. There are more obstacles when collecting data than in analyzing data. I recommend looking at your EHR and determining if the data can be collected in the normal course of using the system during patient care. It may be that to answer or solve a particular problem, you have to add data elements to a particular screening routine (i.e. to the computer screen). So make sure your EHR system has the ability to easily modify screens if this becomes necessary.
As experts have noted, without analytics we will neither know where we are, where we’re going or how to get there. Make sure you ask the right questions before jumping into the data analytics pool!