The future of AI in healthcare depends on clinical data quality assessment

With the drive to integrate artificial intelligence and increase interoperability, clinical architecture believes the need for tools that can assess the quality of healthcare data. Poor quality data can lead to incorrect conclusions and waste resources, hinder the progress of medical research, misleading policy decisions and investments, and ineffective and unparalleled care for payers, providers, investors and governments.

The infrastructure that supports interoperability through the number of TEFCA and QHIN continues to grow, but the quality of data, and even structured data, has changed a lot. There is a need for a way to evaluate the data flowing through these pipelines.

The bets in healthcare cannot be higher. To enable widespread adoption of AI tools in healthcare, most importantly, the pool of data needed to generate AI-driven clinical decision support software reflects good quality, reliable data that can help improve patient outcomes. If adverse events are caused based on the conclusion that poor quality data are not available, progress in the medical industry in the widespread adoption of AI tools will be shaken.

In an interview, Charlie Harp, founder and CEO of Clinical Architecture, explained how the data assessment framework he developed, the Patient Information Quality Improvement (PIQI) framework is a taxonomy of medical data quality that addresses these problems.

Harp founded the company in Indiana in 2007 with a career in life sciences and healthcare software because he believes the healthcare industry needs a company that focuses on data quality, data exchange, and how to make data current and portable.

The Patient Information Quality Improvement (PIQI) framework provides a unified, objective approach to assessing the quality of data with selected titles and highlights the root causes of data quality issues so they can be addressed. Clinical Architecture, in partnership with Leavitt partners, created the PIQI Alliance, which includes a working group that designed the open source framework that can be used throughout the healthcare industry. PIQI is currently using HL7 for the voting process and becomes the standard.

HARP created a tagline for USCDI version 3 that shows the appearance of data quality.

“We are in the watershed of health care,” Hap said. “The number of aging patients is increasing, and with it the number of people with multiple drug comorbidities. We are doing all of these things with genomics and pharmacogenomics, and we are getting to another level of grain in human interaction with the world, which makes us lower the number of providers almost every year. We are on this range, our time is all in our time. Our time is our range. The range of our time is a range of one span. To help everyone, but it wouldn’t be great if we don’t improve the quality of our data.”

PIQI is designed to evaluate the quality of data sent by an organization. HARP compares the PIQI framework to standardized testing, but is used for data. He explained that PIQI evaluated and graded the data on the particle level based on data availability, accuracy, compliance and rationality. Sometimes, the source of the problem is inherent in the collection of data, which is difficult to solve.

One challenge is that in hospitals or health systems, the data contained in EMR is often enough to meet the needs of the application. However, in order to share that data outside of the application, some work is needed to make the data interoperable. When the work is not completed or is not done well, the data will be insufficient. Physicians tend to write clinical notes in ways that make sense to them. Harp said that semantics and the intent behind the data are very important when evaluating the quality of data shared by providers. Another challenge is that each EMR has its own dictionary. Therefore, even with the ICD-10 code and FHIR standard, the clinical data vary greatly depending on the EMR used. There are many data variables that add complexity layers to the transmission of clinical data.

“As The source of most patient data we use in healthcare is no more complex and challenging in terms of data quality than the provider space.

There are multiple groups relying on provider data to support their own data sets, such as payers and their star ratings for their healthcare effectiveness data and information sets (HEDIS) measures. When government departments like Social Security or Centers for Disease Control who want better monitoring of what is happening in public health have poor quality and incomplete data, they cannot use that data to provide an accurate image of their needs.

There is a value chain that lies across the scope of healthcare. Anyone who wants to receive clinical data should have a way to trust the data, whether it is for research or claim processing, or for population health or public health.

“What’s cool about PIQI Alliance is that we’re attracting people from that range. Members include individuals with payers, social security and CMS,” Harp said. “We’re also starting to attract some providers that are doing the same.”

Clinical architecture is currently looking for early adoption partners to test and perfect the PIQI framework in real-world environments.

By choosing to make PIQI framework open source, HARP wants to encourage anyone to use the method to use shared evaluation title evaluation and scoring messages. Measure data objectively using standard frameworks will help improve data quality in the healthcare industry.

Clinical architectures are being tested in beta, with some of the Health Information Exchange (HIE) harp walking through an example of a fictional PIQI framework. He noted that the hospital's allergy data scored 41% and its condition data scored 52%. The demographic is 75%, and its immunity is 77%. But he pointed out that users can drill deeper.

Demographic scores only 75% because birth sexual information was not obtained in 2,761 messages. According to the PIQI framework, the data provided by the hospital is invalid. You can view different data measured in the title. Drug tends to The score is poor, as 89% of the data do not include indications. USCDI version 3 requires instructions.

If you need data to support value-based care, you may be looking for patients who are taking metformin with type 2 diabetes. There is no sign that users can't verify that the reason they take metformin is because they are diabetes. Adding missing information can significantly improve scores. So if you are an organization that buys patient data, then if you can improve the score by 70% to 95%.

HARP sees the PIQI framework and PIQI alliance as a wake-up call to the healthcare industry, highlighting the need to identify and correct different clinical data to improve data quality. When you consider the quality of life science data, data has been curated manually when using the Observational Medical Outcome Partnership (OMOP) standard, many touch on many to step into the OMOP format and normalize. Life science use cases are much narrower because they usually look for patients in specific cohorts.

“There are many reasons we need to drive data quality in healthcare. Ultimately, the economic driver is what makes things happen. If a payer like CMS drives data quality and needs some kind of quality assessment, that will drive adoption,” Harp noted. “If we can push adoption of the minimum data quality threshold, I’d be happy to see what happens next.”

photo: Issarawat Tattong, Getty Images