Predictive Medicine: Advancing Healthcare Through Better Data Governance

Healthcare is facing unsustainable challenges in today’s economic environment and Steve Holloway from Signify Research noted several drivers in his post-Radiological Society of North America (RSNA) conference wrap-up: global staff shortages, fewer physicians, increasing workloads and rising patient volumes.¹ Consider the current economic pressures of rising interest rates and on-going supply chain challenges, and it’s evident that vendors and providers must find new ways to solve lingering problems because people must be able to rely on quality, affordable and accessible healthcare every day. But is our industry doing enough to improve the patient experience and their outcomes while easing the burden placed on clinicians? Many of today’s workflows require significant focus on mundane tasks, adding little value to overall patient care, and this is where modernization of technology becomes critical in solving today’s most pressing challenges.

Yet to occur at scale is the consolidation of data across disparate silos created from outdated departmental focused workflows. Organizations must look beyond near-term operational challenges and focus on better care pathways. This means merging imaging and clinical data, which provides advanced insight into that data, and creates the basis for more predictive, personalized medicine. Healthcare data, after all, is very valuable and a goldmine to researchers. Once we effectively tap into it, gain unrealized insights, and effectively act upon it, we will begin to change healthcare of the future. 

Artificial Intelligence (AI) is one technology that has shown promise in bringing insight to massive amounts of data that would take humans far too long to analyze. Its adoption is most prevalent in radiology, where point-based algorithms recognize clinical conditions that radiologists are already trained to do. AI as a screening tool provides value to over-worked radiologists, but in the long-term, AI can provide much more. Solutions that impact workflow, rapid triage and care team engagement will have the most downstream value. But for this to work, organizations need to adopt a strong stance on data governance, as this decades-long problem has been ignored far too long by adopting workaround solutions but not solving the core problem.

Consider the Electronic Health Records (EHR) rollout and adoption from government mandates like the Health Information Technology for Economic and Clinical Health (HITECH) Act and Meaningful Use from 2009 that drove EHR adoption. There was no time for vendors to create ideal workflows that focused on best-practices or mimic how clinicians delivered care to their patients. The motivating principle then was to comply with specific regulations. Initially, there wasn’t even a standard method to exchange data between EMRs. The rush to adopt the technology and avoid financial penalties drove the market a disjointed patchwork of solutions, creating as many problems as it solved. For healthcare to evolve and drive costs down, the goal must be on modernization that improves workflows, consolidates all patient data, drives better communication and engagement, and manage data governance.   

Interoperability vs. Data Governance

Modernization requires a strong data governance foundation with a focus on data standardization to drive better interoperability. HIMSS defines interoperability as “the ability of different information systems, devices and applications (systems) to access, exchange, integrate and cooperatively use data in a coordinated manner, within and across organizational, regional and national boundaries, to provide timely and seamless portability of information and optimize the health of individuals and populations globally.”

Data governance is defined as “a set of processes that ensures important data assets are formally managed throughout the enterprise. Data governance ensures that data can be trusted, and that people can be made accountable for any adverse event that happens because of low data quality.  It’s the management of the availability, usability, integrity, and security of your organization’s data asset(s).”²

Interoperability and data governance get a lot of attention. Today, health systems can share and exchange much of their data using accepted standards, to name a few: Health Level-7 (HL7), clinical and message format standards, Digital Imaging and Communications in Medicine (DICOM), standard for storing and exchanging medical images, International Classification of Diseases, Tenth Revision (ICD-10), classification of disease standard, and Fast Healthcare Interoperability Resources (FHIR), newer data exchange interoperability standard. The biggest difference between interoperability and data governance, however, is interoperability defines how data gets shared, not what the data should be labeled when it is shared. Interoperability is like a phone line connecting two nations together, like China and the U.S. You can call back and forth between the two countries—that’s interoperability (standards exist to connect the two countries to exchange/share information). However, if neither side speaks the other’s language nor can translate what is being said, you have a data governance issue. In essence, you need a translator so the other side understands and can draw actionable insights on what is being communicated. 

Data governance opens the door to better analysis, analytics, understanding, deep search and improved insights on the data; it can impact clinical decisions and drive population health initiatives. Data governance provides the translation so proper cataloging and analysis can be applied while gaining deeper insight into your data. Consider the importance of data governance in radiology. Without data governance being applied to the massive amount of medical imaging data, we’re missing out on a wealth of clinical insight!

Radiology Data Standardization Example

A computed tomography (CT) abdomen/pelvis can take on multiple permutations: CT Abd/Pelvis, CT Abdomen/Pelvis, CT Intravenous Pyelogram (IVP), plus many more. This procedure also includes multiple series, such as 1) Scout image 2) Axial no contrast 3) Axial with contrast 4) Sagittal reconstruction and 5) Coronal reconstruction. There are fields in the DICOM standard to capture information that is shared from the modality to the Picture Archiving Communication System (PACS) or Vendor Neutral Archive/Enterprise Imaging System (VNA/EI), but how the data is labeled varies across every vendor and modality. The ability for the technologist to further edit descriptions makes standardization nearly impossible. Two identical studies could look like this:

CT Abdomen/Pelvis
Series	CT Scanner #1	CT Scanner #2
1	(empty)	Localizer
2	Axial thin no contrast	Axial 5MM w/o contrast
3	Axial thin soft tissue intravenous	Axial 5MM w/contrast
4	Sagittal intravenous thin soft tissue	Sagittal Recon w/ contrast
5	Coronal intravenous thin soft tissue	Coronal Recon Thin C+

The studies are the same but labeled differently. Most premature atrial contractions (PACS) administrators build reference tables to align all study/series permutations, so PACS display protocols work properly. But mis-keyed information or imported outside studies can break even the best matching matrixes, resulting in significant time wasted for the radiologist to manually manipulate the images, compare exams and read the study. With data governance, the CT study would be labeled in a meaningful methodology, consistent every time, regardless of how the data was exported from the modality or any label changes applied by the technologist, automatically. The standardized output could look something like this:

CT Abdomen/Pelvis
Series	Study #1 Description	Study #2 Description
1	Localizer	Localizer
2	Axial Phase 1 Pre Thin Abdomen/Pelvis	Axial Phase 1 Pre Thin Abdomen/Pelvis
3	Axial Phase 2 Post Thin Abdomen/Pelvis	Axial Phase 2 Post Thin Abdomen/Pelvis
4	Sagittal Phase 1 Pre Thin Abdomen/Pelvis	Sagittal Phase 1 Pre Thin Abdomen/Pelvis
5	Coronal Phase 1 Pre Thin Abdomen/Pelvis	Coronal Phase 1 Pre Thin Abdomen/Pelvis

When proper data governance is followed, the study and series names are archived consistently every time, including any outside exams. With standardized imaging sets, advanced search and analysis can be performed with high confidence that all data has been included. The insight and intelligence from the data provides the content to create a real-world evidence database on imaging data, which can impact different workflows, and open the door to better clinical decision making.

The following is a framework for a good data governance policy. Here are the key elements and characteristics of data governance and how it applies to the data throughout its lifecycle:

1. Accuracy: The data should be free of errors, is correct.
2. Accessibility: Proper safeguards established to ensure data is available when needed.
3. Comprehensiveness: The data contains all required elements.
4. Consistency: The data is reliable and the same across the entire patient encounter.
5. Currency: Data is current and up to date.
6. Definition: All data elements are clearly defined.
7. Granularity: The data is at the appropriate level of detail.
8. Precision: The data is precise and collected in their exact form.
9. Relevancy: Data is relevant to the purpose it was collected.
10. Timeliness: Documentation is entered promptly, is up-to-date and available within specified and required time frames.³

Establishing a culture that applies a strong data governance framework makes the data more valuable when it comes to clinical research and population health, furthering advancements in healthcare. 

Choosing the Right Technology

Young professionals in healthcare are no doubt more adept at adopting technology than their predecessors. If you compare baby boomers (born between 1946 and 1967) vs. millennials (born between 1981 and 1986) the comfort level and availability of technology is more ingrained in the latter generation. This dynamic is increased further with future generations, (Generation Z and Generation Alpha)⁴ as the technology curve continues to advance, and information is immediately accessible on personal devices regardless of time or location.

Organizations should consider their users and how new technology impacts them, along with their willingness to use that technology. Poor execution leads to patched workflows and some older providers may leave the field prematurely to avoid using the technology, adding additional burden to an already overworked staff. But the right technology provides significant improvements in clinical workflows, and the adoption rate will be greater when there is a direct impact on improved workloads and life balance.

With so many exciting technologies available, where should we start? Solutions that impact enterprise workflow and entire care pathways may provide the best Return on Investment (ROI), as isolated solutions may be disjointed from the enterprise. Enterprise Imaging (EI) solutions (and enterprise strategy) along with AI are two areas that can provide enterprise value and positively impact patient care.

An enterprise imaging strategy aligns with the goals of value-based care, because having access to all patient information improves better decision making. Meanwhile, AI can optimize and streamline workflow and care path processes allowing physicians to focus on delivering the best patient care with new clinical insights. Initially, AI focused on solving “point” challenges or identifying single anomalies and was not enterprise focused. However, as noted by Sanjay Parekh, PhD, from Signify Health, the focus of AI is moving away from point-based solutions to more service-line solutions (impacting complete workflows).⁵ Expanding AI beyond a single department, like radiology, and connecting it to the enterprise will drive the biggest impact for any health system. But to effectively achieve the value at the enterprise level, it’s imperative that organizations get control of their data.

Although enterprise imaging (EI) and artificial intelligence can work seamlessly together, there remains a gap: data governance and care collaboration. Every day, real-life scenarios occur that a solid EI and AI strategy can help improve care outcomes and connecting that insight and understanding to the entire care team in real-time opens the door to what real cross-departmental collaboration can bring. Today’s clinicians face an avalanche of information and data, making it challenging to manage and make split-second decisions without comprehending all the data available. This is where EI and AI solutions can help. When data is consolidated, connected and better collaboration and insight are enabled, they will drive better decisions, reduce costs, and improve patient and population health.

Organizations Must Invest in Modern Technology

According to Medscape’s 2022 Physician Burnout and Depression Report, about 49% of radiologists experience burnout. The highest group are Emergency Physicians (around 60%) and the lowest are Public Health physicians (around 26%). The leading cause of burnout for physicians are too many bureaucratic tasks, too many hours worked and lack of respect.⁶ 

Physician burnout is a real challenge.  That top-rated physicians are quitting mid-career at an alarming rate and few are joining to replace them spells dire ramifications ahead. Modernization is one of the best ways to help alleviate burnout by improving workflows and efficiency to allow physicians to spend more time with their patients. Not only are modern EI and AI solutions meant to manage all imaging data at scale and provide better workflows and clinical insight, but their integration to the EHR provides clinicians with all information in a single place. Consolidating your imaging infrastructure and integrating it to your patient records is the first step. Next, AI helps drive better insight to aid in real-time decisions.

With over 300 U.S. Food and Drug administration (FDA) approved algorithms (and growing!), AI continue to show real promise in many patient care situations. But the real benefit comes in workflow optimization and care team collaboration.  To date, radiology is leading AI adoption, but new solutions are constantly emerging, each improving different clinical care pathways. But how does an organization choose the best solutions to meet their needs? 

Deploying many “point” solutions becomes costly in terms of time, management and ultimately scale, and most replicate what radiologists already do—interpret images and diagnose the problem. Clinical AI solutions that improve screening (like lung or breast screening) are more ideal because they free radiologist’s time to focus on more critical cases, improving their productivity.

To achieve success when selecting AI solutions, organizations must plan the same as they would for any other major Information Technology (IT) initiative, and HIMSS provides several Maturity Models for technology adoption:

• The Adoption Model for Analytics Maturity (AMAM)
• The Continuity of Care Maturity Model (CCMM)
• The Clinically Integrated Supply Outcomes Model (CISOM)
• The Digital Imaging Adoption Model (DIAM)
• The Infrastructure Adoption Model (INFRAM)
• The Outpatient Electronic Medical Record Adoption Model (O-EMRAM)
• The Electronic Medical Record Adoption Model (EMRAM)⁷

Up Next: Precision Medicine

In President Obama’s 2015 State of the Union Address, he launched a Precision Medicine Initiative, aimed at improving health and treating disease.⁸ Imagine a world where medical treatments are customized for individual users at the genetic level, offering far more effective and timely outcomes that work with an individual’s genetic makeup. Precision medicine is an emerging field where new discoveries are made every day. Understanding the various aspects of a patient’s genome, behavioral and economic lifestyles require massive storage and analysis of a significant amount of data. To analyze that much data (aka Big Data), researchers first need to standardize it, which is what a group was tasked to do with over 1 million volunteers in a national study.⁹ 

When imaging data is included, the standardization challenge amplifies. Healthcare data is the most valuable asset hospitals possess, but without standardization, its value is diminished. Virtually impossible today at most organizations is a complex search such as, “all female patients between the age of 39 and 42, with a lung nodule between .30 and .50mm in size, a smoker for 10 years, is on Coumadin, lives in greater Chicago, is of Asian descent, and has high blood pressure readings over 150/90.” Search any PACS archive for all “CT abdomen” studies and there is a very high probability you will not see every result due to a lack of data standardization.

Precision medicine offers great hope for the Life Sciences industry to create novel treatments that are far more effective at treating and curing disease. But without data standardization, and AI to provide advanced insight, the journey will be fraught with countless man hours of effort to categorize the data. The real value of precision medicine is that not only can data be analyzed on an individual basis, but it can be analyzed at the population level. This means that a clinician can predict and intervene before a medical condition begins. How does AI help with precision medicine? It can:

1. Classify problems using different algorithms to solve precision medicine problems like accurate disease diagnosis, disease detection and prediction, and treatment optimization;
2. Predict the risk of a disease, identify the disease response and outcomes on the individual patients based on their own characteristics; and
3. Be used for clinical data extraction, aggregation, management, and analysis to support clinicians by efficiently stratifying subjects to understand specific scenarios and optimize decision-making.¹⁰

AI can be used to process historical and current data and relate it to medical reports for new insights. This could be a game changer in patient treatments, and it’s possible only with standardized and connected imaging records, EHR data and longitudinal genomics information.  Consistent standardized data is a must. The algorithms of EHR, EI solution and AI must work seamlessly together to provide researchers with advanced data analysis to advance medicine. This highly valuable data can also potentially open new revenue streams to fund enhanced patient outreach and health equity initiatives.

Let’s Create the Healthcare of the Future

Clinicians and healthcare organizations across the world face real-world problems, but these daily challenges are frequently met with limited insight, the right technology, or even the right data. New technology and modernization will help solve difficult challenges, still, many organizations lack the ability to acquire this needed newer technology to personalize healthcare, improve outcomes, and to apply new treatment programs to populations. 

We must start with data governance and standardization, image and data consolidation, then using AI, apply advanced insight to the data. We must improve access and insight to the patient’s complete clinical record, and invest in the tools to provide advanced workflow, analysis, and care team collaboration. These steps will have a major impact on patient care and managing an aging population.

Preparing organizations to offer personalized medicine will further advance the outcomes precision medicine can offer. It also adheres to the principles of value-based care initiatives. Ultimately, modernization of technology is the key to promote a rewarding journey to provide cutting-edge, quality healthcare to everyone.   

References

1. Holloway, Steve. “RSNA 2022: Predictions vs Reality.” Signify Research, Signify Research, 13 Dec. 2022, https://www.signifyresearch.net/healthcare-it/rsna-2022-predictions-vs-….
2. Gaston, James. “A Recipe for Analytics, Key Ingredient #3 – Data Governance.” A Recipe for Analytics, Key Ingredient #3 – Data Governance, HIMSS, 3 Nov. 2017, https://www.himssanalytics.org/news/analytics-key-ingredient-data-gover….
3. Patty, Buttner, et al. “Healthcare Data Governance - Ahima.” Healthcare Data Governance Practice Brief, AHIMA, Jan. 2022, https://www.ahima.org/media/pmcb0fr5/healthcare-data-governance-practic….
4. Debczak, Michele. “Revised Guidelines Redefine Birth Years and Classifications for Gen X, Millennials, and Generation Z.” Mental Floss, Mental Floss, 6 Dec. 2019, https://www.mentalfloss.com/article/609811/age-ranges-millennials-and-g….
5. Fornell, Dave. “Video: Radiology AI Trends at RSNA 2022.” Radiology Business, Innovative Healthcare, 13 Dec. 2022, https://radiologybusiness.com/topics/artificial-intelligence/video-radi….
6. Kane, Leslie. “Physician Burnout & Depression Report 2022: Stress, Anxiety, and Anger.” Physician Burnout & Depression Report 2022, Medscape, 21 Jan. 2022, https://www.medscape.com/slideshow/2022-lifestyle-burnout-6014664.
7. “Maturity Models.” HIMSS, HIMSS, 17 Nov. 2022, https://www.himss.org/what-we-do-solutions/digital-health-transformatio….
8. “White House Precision Medicine Initiative.” National Archives and Records Administration, National Archives and Records Administration, https://obamawhitehouse.archives.gov/precision-medicine.
9. “What Are Some of the Challenges Facing Precision Medicine and the Precision Medicine Initiative?: Medlineplus Genetics.” MedlinePlus, U.S. National Library of Medicine, 17 May 2022, https://medlineplus.gov/genetics/understanding/precisionmedicine/challe….
10. Jain, Kamal. “Artificial Intelligence for Precision Medicine and Better Healthcare.” Medium, Analytics Vidhya, 22 Aug. 2020, https://medium.com/analytics-vidhya/artificial-intelligence-for-precisi…;