Measure Healthcare Data Quality

How to Measure Healthcare Data Quality

Monitoring the costs and efficacy of healthcare requires the ability to carefully analyze all of the related information. Unfortunately, it is much harder to do this analysis in the healthcare industry than it is in industries like manufacturing. That’s because the recording of healthcare data is mostly unstandardized. Even where standards exist, they are unique to certain aspects of the process and hard to transfer to other parts. This makes it hard to figure out how to measure healthcare data quality, let alone actually use that data.

The Difficulty of Standardizing Healthcare Reports

Some of the biggest sources of unstandardized information are doctors, nurses, and others involved in diagnosing problems and writing up reports. Even where attempts are made to improve standardization, professionals often have to resort to writing in the details. This type of reporting, known as “free text,” doesn’t work well with databases and other such organizational tools. Computers rely on the ability to search for exact matches and exclude the rest, but when people write things in free text, they may use any number of terms and phrasings. This makes most of their reports unsearchable.

Image -

Even when an office has devised standardized forms for specific conditions, the problem persists. This is because every case is different, but most forms don’t have enough checkboxes or other options to capture all of the relevant details. The doctor ends up having to write these details in by hand, and the attempt at standardization is at least partially stymied.

Different Branches Have Different Foci

This is another big problem when it comes to healthcare reporting. Consider, for example, a broken leg. An emergency room report may focus on whether bleeding was involved and how much swelling the break caused because those things determine how urgent the response needs to be. The treating doctor who comes next will likely focus on the details of the break itself.

All of this information will also need to be transmitted to the relevant insurance companies. It turns out that these companies are usually the only parts of the healthcare system with true standardization. They use codes that are each mapped to specific things. Alas, this also means that now, there’s yet another layer of difference in the records to deal with when attempting to analyze healthcare data.

Why Not Just Use These Standardized Insurance Reports for Analysis?

Insurance reports are indeed very useful for certain types of analysis, so they are often used. The problem is that they don’t capture all of the nuances needed for other types. For example, there’s an insurance code for recording that a patient is nicotine dependent, but it includes so many possible permutations that it isn’t very useful for anything but high-level analysis. Is the nicotine-dependent person trying to quit smoking, a tobacco chewer with no interest in quitting, or someone who has quit and is now suffering withdrawals? It doesn’t matter to the coding system; the can all be recorded as code 305.1.

Possible Solutions

Image -

Efforts have been made to improve standardization through the years, and the best of these can greatly improve the quality of incoming data. Some promote the establishment of a data dictionary along with a policy that enforces its use. This eliminates the problem of healthcare professionals using different wordings for the same things. It also smooths out problems like people using different definitions for common fields like race and ethnicity.

Another approach tackles the data problem at the tail end of the process. One company promotes the use of what it calls a Late-binding Data Warehouse. This model says that data should be brought into an electronic data warehouse (EDW) in an as-is state. Then, when it’s time to draw information from the aggregated raw data, it is transformed into the exact format that is needed. This prevents problems that arise when regulations or protocols change.

No approach to medical data can be called perfect, but both of these models provide benefits. The pre-formatting model is good for situations where big changes are unlikely, while the post-formatting, or late-binding, model is good for institutions where needs are more dynamic. It is, however, clear that healthcare providers do need to have some sort of actual plan in place for data handling. This alone can cut the level of chaos involved and improve the quality and retrievability of medical data.