HIM Challenges: Patient Data Integrity and AI/Advanced Analytics
By Rachel Podczervinski, MS, RHIA
For The Record
Vol. 35 No. 2 P. 8
Avoiding the Garbage In, Garbage Out Curse
With revenue cycle (76%), clinical administration (55%), and finance (36%) among the top five areas benefiting most from automation and artificial intelligence (AI), it’s easy to see why 98% of health care leaders say their organizations have implemented or are planning to implement AI. That includes nearly 48% that have already taken the leap, according to Optum’s annual survey on AI in health care.
However, it makes no difference how large the investment or advanced the technology solution is if the patient data flowing through it lacks integrity. In other words, when it comes to AI and patient data, the adage “garbage in, garbage out” still applies. It’s also a very real problem, as evidenced by an average duplicate patient record rate that runs as high as 18% in the typical facility and patient misidentification issues that cost the health care industry more than $6 billion in denied claims.
A clean master patient index (MPI) or enterprise MPI (EMPI) is foundational to a health care organization’s clinical and financial operations and a critical element of any AI/automation strategy. Eliminating duplicates and overlays upfront and implementing complementary technologies to ensure the integrity of patient data going forward are imperative to realizing maximum return on investment for AI/automation and advanced analytics technologies—and to optimize their impact on the clinical and financial bottom lines.
HIM and AI
Health IT has already had a significant impact on the HIM landscape. The advent of computer-assisted coding in the early 2000s gave coders their first significant taste of automation with the use of natural language processing (NLP) to analyze clinical documentation and suggest diagnosis and procedure codes. Its full effect was realized in 2015 with the transition to the ICD-10 code set, allowing coders to code 22% faster despite the more expansive code set and detailed supporting documentation requirements.
ICD-10 also exposed documentation issues, which in turn gave rise to the next automation milestone for HIM—clinical documentation integrity (CDI) tools designed to improve provider documentation for enhanced patient outcomes and data quality and more accurate reimbursement. The importance of CDI technology has grown with the transition to value-based care models, which rely heavily on risk-adjusted coding and publicly reported quality data. Along with prompting providers to improve documentation specificity, these tools also suggest query opportunities based on automated scans of a patient’s record. This allows CDI professionals to identify additional reimbursement prospects and problem areas that can be corrected before claims are submitted.
Adding AI technologies to HIM also automates the analysis of patient chart contents and prioritizes for CDI specialists those with the highest likelihood of requiring clinician queries. Further, when embedded into encoder and computer-assisted coding software, AI can suggest the most likely codes based on clinical indicators, which in turn allows coders to focus on validating or adjusting recommendations based on their review of appropriate chart elements.
The value AI can bring to HIM goes far beyond helping to address documentation and coding issues that cost the United States health care system an estimated $54 billion annually, however. It also holds the power to alleviate some of health care organizations’ most significant administrative pain points by automating repetitive functions such as insurance verification, claims submissions, and billing follow ups.
Harnessing the Power of Analytics
Equally important, if not more so, is the ability of advanced AI, NLP, and machine learning to leverage the massive and growing volume of health care data—which already accounts for about 30% of the world’s data volume—being collected by typical health care organizations.
Every stakeholder group has its own overarching goal for how these data can be leveraged. For clinicians, it provides a comprehensive view of the patient to enable precision treatment and optimal outcomes. For administrators, it can help predict everything from length of stay to utilization for more appropriate resource allocation.
For HIM, patient data can be used to improve coding and billing productivity and accuracy, which in turn can enhance revenues. For example, analytics can reveal if a coding team has achieved optimal productivity by monitoring a wide range of metrics, including start/end times, average number of charts coded per hour, case assignments, physician query turnaround times, and the volume of coding and noncoding tasks assigned to each coder.
Overall, data analytics can be a highly effective strategy for reducing claim denials due to coding and documentation errors, which are the driving force behind an average annual loss of $5 million for hospitals and write-offs of up to 5% of a physician practice’s net patient revenue. Correcting these errors is urgent, as denial rates are trending upward, having already risen more than 20% over the past five years. Not to mention the high cost of reworking denied claims, which averages $25 per claim for practices and $181 per claim for hospitals.
What’s more, leveraging historical data, AI, machine learning, and NLP to perform predictive analytics can identify performance gaps and coding/documentation issues that drive down reimbursements, drive up denials, and increase costs across the board—while reducing audit risks and accelerating the revenue cycle.
Avoiding the Garbage In, Garbage Out Curse
When it comes to AI in health care, it makes no difference how powerful or advanced the technology is if the data flowing through it are compromised by ineffective patient identification and matching. We see its impacts already, most recently in the findings of a patient identification survey from HIMSS and Patient ID Now, in which 70% of respondents agreed that patients undergo or receive duplicative or unnecessary testing or services due to difficulties in managing patient identities, while 67% agreed that a lack of clear identities for patients put their organization at a higher risk for fraud.
There’s no question that AI can have a significant impact; a new report from the Council for Affordable Quality Healthcare concluded that automation saved the health care industry $187 billion in 2022 and could increase that by another $25 billion by transitioning to fully electronic administrative transactions alone. But that can only happen if duplicates and overlays are eliminated and subsequently prevented from recontaminating MPIs and EMPIs and, in turn, downstream systems.
Thus, a critical early step in any AI strategy is to undertake a comprehensive MPI/EMPI cleanup coupled with deployment of technology tools capable of ensuring the integrity of patient data—prior to implementing AI-enabled technologies.
The ideal approach to the pre-AI cleanup involves a combination of professional services and advanced technology to identify and resolve duplicate medical records that already exist within the MPI/EMPI and establish processes to help prevent new ones from occurring. This includes a thorough data assessment, followed by an expert determination of the most appropriate approach for identifying and correcting duplicates and other data integrity issues.
Once the MPI/EMPI is clean, it’s necessary to deploy technology capable of catching and correcting future errors before they can lead to the creation of new duplicate and overlaid records. The goal should be technology that enables end-to-end MPI/EMPI protection by operating in multiple environments and at multiple stages throughout the patient record process—starting at registration, where one study found that 92% of errors take place. Such a system could leverage biometrics to collect a photo along with the information needed to create a patient record in the MPI/EMPI and advanced deterministic and probabilistic matching algorithms to analyze and clean patient data before a record is updated. Text messaging could be leveraged to send the patient a link to take and submit a selfie and photo of their driver’s license and use facial recognition software to validate their identity and search for any record matches before assigning biometric credentials to new patients.
These tools, which ideally will also address privacy and security concerns, go beyond the capabilities of current EHR systems, which are limited by patient lookup functionality that requires specific processes and data to be precisely entered field by field. A single missed or incorrect detail can result in a search yielding invalid results, which in turn can lead to the creation of a new duplicate record, or worse, an overlay if the wrong patient is selected—adding to, rather than alleviating, the problem.
Getting It Right
The potential AI holds for vastly improving the clinical and administrative sides of the health care house is significant. But like any data-driven technology, information integrity is crucial to realizing AI’s full promise.
By starting off on the right foot with a clean MPI/EMPI, the return on a health care organization’s investment in AI will be accelerated and optimized.
— Rachel Podczervinski, MS, RHIA, is vice president of professional services with Harris Data Integrity Solutions and a member of the AHIMA Board of Directors. The viewpoints expressed in the article are personal and are not being made on behalf of AHIMA.