Artificial intelligence is helping physicians move the bar on clinical value — Here's how

Clinical documentation is key to demonstrating value in outcomes-based medicine. Complete documentation helps to validate patient outcomes by reflecting the severity of a patient's medical condition, sharing key data with subsequent caregivers and optimizing claims processing and reimbursement.

This content is sponsored by Nuance

Although clinicians may think they are writing excellent clinical notes, physicians' unfamiliarity with ICD-10 coding often means their notes fail to meet heightened standards for specificity. When this happens, patient outcomes might not accurately reflect the quality of care provided, and may even negatively influence provider performance scores.

"Once the final [patient] bill is established and sent out, that becomes what the rest of the world sees about the care you provided for that patient," Anthony Oliva, MD, vice president and CMO at Nuance, said during a discussion April 18 at Becker's Hospital Review's 8th Annual Meeting in Chicago.

The importance of accurately capturing patient acuity in clinical documentation and coding cannot be understated, especially as a growing number of government and commercial payers use risk adjustment in determining provider reimbursement. Consider the Medicare value-based purchasing program. By 2018, CMS expects use risk adjustment as a factor to determine at least 75 percent of payment criteria for program reimbursement.

Documentation that accurately captures patient severity can reduce variability in patient outcomes and move the bar on value. To improve physicians' documentation processes, hospitals are using artificial intelligence at the point of care to ensure clinical notes are complete, compliant and correct. This article discusses the role of clinical documentation in risk adjustment and quality reporting, inefficiencies in traditional clinical documentation improvement and AI as a tool to support clinical documentation under ICD-10.

Capturing risk severity starts with clinical documentation

In outcomes-based medicine, government and commercial payers refer to clinical documentation and coding to determine value through risk adjustment. Specifically, payers use risk-adjustment formulas to account for baseline differences in patient case mix.

For instance, a rural hospital treating a predominantly government-insured population will likely have worse outcomes than a hospital in an affluent, suburban region with a high number of commercially insured patients. By eliminating case mix bias, government and commercial payers can begin to assess and compare quality and performance between providers.

Risk adjustments are based on ICD-10 codes recorded in the final patient bill. To ensure patient risk is accurately captured during coding, providers' clinical documentation must reflect the complexity of the patient's condition — including the number and severity of comorbidities, the relationships between conditions and the results of treatments and interventions — and capture that information in a way that is codeable.

Clinical documentation practices directly influence physician quality scores. When clinical documentation is incomplete, providers face a number of clinical and financial setbacks. For instance, physicians may fall short on quality standards not because of their clinical care, but because of incomplete EHR data entry. Poor quality scores may also reflect unspecific documentation that didn't accurately capture medical acuity or treatment in a codeable way under ICD-10.

"If you don't document correctly, you will have two different mortality rates — the actual mortality rate and the mortality rate that was coded for based on the note," Dr. Oliva said. "If the risk of mortality is actually higher than what was coded, you put yourself at a disadvantage right out of the gate in terms of demonstrating clinical value."

Traditional clinical documentation improvement is burdensome to physicians

To improve physicians' documentation, hospitals traditionally employed specialists to review clinical notes before submitting documentation to coders for billing. This retrospective approach makes coding more efficient for coders downstream. But because review often happens after the patient is discharged, it can also require CDI specialists to chase after physicians for information on patient cases from weeks prior.

"It is incredibly irritating for physicians to get sidelined by a [clinical documentation specialist] on their way to the bathroom or the cafeteria or in the elevator," said Reid Coleman, MD, CMIO for evidence-based medicine at Nuance. Physicians often face queries for information after they entered the note or after the patient is discharged, and more than 80 percent of physicians find these requests disruptive and time consuming, according to a 2014 survey from Nuance.

"Comprehensive, clear clinical documentation can actually improve physicians' baseline performance, which helps physicians more accurately stack up against their peers and regional competitors on quality comparison," Dr. Oliva said. To support physician documentation, some hospitals have seen value in automated intelligence solutions at the point of care to improve coder accuracy. With such tools, clinicians can ensure their notes are complete, detailed, specific and adhere to ICD-10 and quality reporting standards.

How AI technology can improve clinical quality scores

"Automated intelligence software gives physicians clerical support to help them meet the pressing needs of a value-based world," Dr. Coleman said. Different categories of AI software deployed thoughtfully in the clinical environment can help physicians improve documentation and ensure all diagnoses are captured accurately. 

Dr. Coleman recommended hospitals invest in two specific types of AI technology.

1. Natural language software. Conceptually similar to "spell check," AI software that uses natural language processing can analyze physician documentation and advise physicians on ways to improve the specificity of their notes on the patient's medical condition. It does so by analyzing the clinical note and looking for certain combinations of words, their relationship to one another and their position in the note to extract clinical concepts. The software advises a physician on documentation improvement in real-time so he or she can resolve common coding problems on the front-end of the clinical documentation lifecycle. With natural language software, CDS don't have to waste time chasing down physicians to clarify clinical notes that lack information or specificity, which helps make both professionals' lives easier. 

2. Computer-assisted physician documentation. CAPD software analyzes patient medical information captured in a physician's note and draws out diagnoses that are supported by medical evidence but not explicitly documented. By providing feedback in real time, AI software helps physicians complete documentation in a way that enables coders to fully capture the complexity, severity and quality of medical care in ICD-10 terminology. When CAPD software is implemented effectively, physicians can realize substantial gains in clinical quality scores, continuity of patient care and reimbursement rates.

"Technology should be simple and do work for physicians — not the other way around," Dr. Coleman said. "Through a more natural approach to creating clinically accurate information, everyone wins — the physician, the institution and most importantly, the patient."

More articles on health IT: 

Accumen adds Andrew Azan as CVP of sales: 3 things to know
NSA, NSF sponsor cybersecurity summer camps for kids
Healthcare comprises 15% of all ransomware attacks, report finds

© Copyright ASC COMMUNICATIONS 2020. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.

 

Top 40 Articles from the Past 6 Months