The U.S. healthcare ecosystem is quite complex, and there is a focus on healthcare outcomes, controlling costs and improving data. The importance of healthcare data quality and how "healthy data" can improve overall healthcare quality cannot be overstated.

This has been of paramount importance to Dr. Oleg Bess, a practicing physician in Los Angeles for 25 years and founder and CEO of 4medica, a master patient index and patient-matching health IT vendor. He leads the company's product development strategy across inpatient, ambulatory and other new care settings.

Healthcare IT News sat down with the doctor because he offers a unique perspective, being on the informatics side for 20 years, and being on the patient care side. He saw first-hand how many patients are unaware of what is happening behind the scenes with their data – and ultimately saw that flawed data can lead to errors in decision-making.

Q. Why is healthcare data quality so important?

A. Medicine is a highly data-intensive discipline. As doctors, we make decisions based on clinical information available to us, and 80% of this information is test results. However, unlike other industries, where data is carefully cultivated to achieve greater efficiency, better outcomes, or streamline production, in healthcare data seems to be almost an afterthought.

While doctors should make better decisions based on a historic dataset about a patient, we make life-and-death decisions based on episodic test results, the current set of symptoms and current complaints.

As a doctor, I should be able to trend how a patient's complaints, medications and test results reported to another doctor three months ago correlate with complaints, medications and test results today in my office. This can give me great insight about how this patient will respond today with slightly different medications and complaints.

Today we cannot easily electronically trend this patient's test results, because most likely the other doctor ordered tests at a different laboratory. These results are not electronically shared, patient identity is poorly established, and each lab uses a different set of codes for tests. Therefore, it's impossible to see a simple graph of cholesterol over time while correlating how medications and behavior [have] affected this trend.

The same improvement that other industries can achieve when quality data is widely available is not possible in healthcare. Our patients deserve better, providers need more effective processes, and our leaders need better tools to make decisions.

Q. What happens when flawed data gets in the healthcare system?

A. "Garbage in, garbage out" has never been so true. Achieving electronic data sharing is hard, but when this data contains poorly identified patients, many complications arise when managing entire populations or individual patients. Continuous feeds of discreet digital data rapidly compound the problems.

Two main complications occur with poor record identification – duplication and overlay. Duplicated patients are created most often, resulting in incomplete data for each patient, hindering point-of-care decision-making and quickly invalidating population-wide analytics.

After all, when the same patient has an abnormal test several times, this test could have easily been performed in a different lab, and 30% of the time this patient will be duplicated when results are aggregated.

Population analysis is greatly skewed, appearing as if multiple patients have the incidence of this abnormal result. Additionally, when managing this patient individually, the doctor has incomplete information about the progression of the disease, hindering the decision-making and the treatment. Tests and procedures often are repeated because of these incomplete, duplicated records.

The overlay is a vastly more serious problem, resulting from one patient's record being placed in the wrong chart. It can be caused by either incorrectly merging two patients or when new reports are filed incorrectly.

In this case, the doctors will make erroneous diagnoses, prescribe incorrect medications, and perform wrong procedures, in extreme cases leading to a patient's death. Incorrect decisions caused by wrong information in a patient's chart is one of the more frequent causes of medical errors, resulting in incalculable suffering and high cost of litigation in healthcare.

Electronically duplicated patients also result in billing and collection inefficiencies. It is one of the main reasons for payment denial by payers, in many cases resulting in complete loss of revenue.

Following up on each claim is almost impossible when the denial rate approaches 30%, as seen in many laboratories. It is very important to have the tools and processes in place to assure the ongoing quality of your patient identity database. Manual cleanup is exceedingly expensive when performed at a late stage of the billing process.

Q. What is "healthy data," and how is it improving overall healthcare quality?

A. They say "there is an app for everything," and in healthcare this may also soon become true. Today's technology allows sophisticated analysis, predictive analytics and convenient tools to help doctors better manage patients' health.

Artificial Intelligence can make a diagnosis, provide early alerts based on worsening clinical trends, identify patients for clinical trials and become a patient's personal medical concierge. Unfortunately, the promise of healthcare revolution brought about by these state-of-the-art tools is still only wishful thinking.

In many cases these tools are powerless because of the poor medical data quality. Both population health and individual patient management can be greatly enhanced by using available technology when a quality dataset is readily available.

Q. How can healthcare provider organization CIOs get rid of flawed data and make data healthy?

A. Providing up-to-date and accurate analytics is a major part of any CIO's job. In some cases, for an accountable care organization or other risk-carrying provider group, appropriate reporting will make or break an organization. In other cases, it has the potential to significantly improve how a provider organization operates.

Today, all CIOs understand the value of data, and many have plans to integrate disparate data sources to drive clinical, financial and operational improvements. However, many fail to see how the lack of data quality will affect the usability of data they have collected. Many proceed with introducing an analytics component without assuring the data quality.

After the data sources have been connected, healthcare data quality must continue with two additional steps: patient identity resolution by establishing a single record for each patient and normalizing the data to achieve meaningful reporting.

Identity resolution includes deduplication of the existing database as well as establishing a process for continuous attribution of newly incoming records in the correct patient chart. A master person index engine, capable of real-time production mode, is a must to achieve both the initial clean-up and the ongoing records management.

It's virtually impossible to achieve this task manually as 30% of the demographics data on clinical and claims documents may have errors or updated information. A good MPI will be able to reconcile these changes and either make correct merging decisions or create a convenient work list to resolve these manually.

Even when all records are placed into the right electronic patient chart, the coding principles in healthcare are not rigorous enough to correlate data coming from disparate doctors, hospitals, laboratories and imaging centers.

The other, often overlooked step is providing a formal process to normalize the data. If an organization wants to view outliers of a certain test, say HgA1C > 7, and the test results are performed by multiple labs, the test names in the database will have different test codes and test names for HgA1C.

It is impossible to obtain a true list of all performed HgA1C tests and compare, trend or alert on their values unless the test codes are cross-referenced (normalized) to the same code.

All of this stems from the basic inability in healthcare to:

1. Easily access clinical data electronically.
2. Identify the patient positively, even when the electronic dataset is available, resulting in duplicate patients, each with incomplete records.
3. Even when all records are placed into the same electronic patient chart, the coding principles in healthcare are not rigorous enough to correlate data coming from disparate doctors, hospitals, laboratories and imaging centers.