Diagnoses reported are used to determine a “Risk Adjustment Factor” (RAF) score, which determines the amount paid to the organization per beneficiary during the corresponding payment year. Insurance organizations are paid at a higher rate for patients that have multiple conditions and conditions with greater levels of severity, as their anticipated costs of care will be higher.
Prospective vs. retrospective risk adjustment
Two of the more commonly used approaches for identifying unreported diagnoses include “retrospective” chart reviews, and “prospective” information reviews. Retrospective reviews use strict criteria to identify diagnoses that are eligible for reporting. Prospective reviews have emerged more recently as a method to identify additional unreported diagnoses using a significantly different process. The two approaches are complementary but have key differences which I’ll outline in this article.
What is retrospective risk adjustment?
Retrospective reviews may identify conditions that can be directly reported to CMS without additional health care provider involvement. This has been a popular approach with insurance organizations as it eliminates the need to engage providers and request additional documentation or clarifications. However, retrospective reviews are relatively constrained. A diagnosis may only be reported if the following three general criteria are met:
- The source document for the diagnoses must meet several eligibility requirements, including but not limited to, a face-to-face encounter during a date of service in the targeted year, an eligible location of service, provider eligibility, a valid signature, valid credentials, an acceptable specialty type, and an acceptable document type.
- The source document must contain the diagnosis as documented by the provider ( it cannot be inferred by the coding professional).
- The source document must include supporting documentation that demonstrates each reported condition was evaluated and/or managed during the encounter (with certain exceptions for certain chronic conditions). The acronym “MEAT,” which stands for “Monitor, Evaluated, Assess and Treat” is used by some reviewers. This and other approaches (e.g., TAMPER™) may be useful but are not officially recognized. Instead, CMS requires that all conditions reported meet ICD-10-CM reporting requirements.
What is prospective risk adjustment?
In contrast, prospective reviews hunt for clues referred to as “Clinical Indicators” in any available information source. They have none of the above constraints but cannot be used to report diagnoses directly. If the clinical indicators are supportive of an underlying diagnosis, they may be presented to a clinician during a subsequent face-to-face encounter. This allows the condition to be addressed by the clinician. If it is validated, it may be reported using the same criteria used for retrospective reporting.
The role of clinical indicators in prospective risk adjustment
Clinical indicators give an indication of the quality of the patient care delivered and may include information identified in any of the following sources:
- Chronic diagnoses reported in prior years (identified through claims data or review of prior year documents)
- Diagnoses in the targeted year that lack supporting documentation
- Relevant procedures, including operations and amputations (claims data or documentation)
- Medications (medication data or documentation)
- Lab values (data or documentation)
- Diagnostic study reports (e.g., EKG, EEG, etc.)
- Diagnostic imaging reports
- Specialist visits or referrals
- Clinical assessment forms (e.g., PHQ-9 scores)
- Notes created by noneligible healthcare staff members
- Patient correspondence with the clinic (e.g., patient portal communications)
- Patient education materials
- Physical examination findings
- Durable medical equipment (claims data or documentation)
- Patient generated data
- Prior authorizations
However, identifying clinical indicators is just a starting point in the prospective review process. A large number of clinical indicators for a wide range of potential conditions may be identified per patient. It would be burdensome to present them all to a provider during an encounter with the patient. The prospective process relies upon prioritizing clinical indicators that are strongly suggestive of an underlying condition. In addition, the number of clinical indicators that point to a specific condition may be used. A skilled risk adjustment professional may help to adjudicate this process by determining if the clinical indicators are strong enough to be flagged for further review by a clinician.
A common practice is to address the patient’s health status during an Annual Wellness Visit or a home visit. The clinician may review clinical indicators and other information to determine if a suspected condition, a different diagnosis, or no related diagnosis is warranted based strictly on clinical criteria. This structured process allows diagnoses that would have otherwise not been reported to be confirmed and reported. It also has the benefit of establishing or reestablishing care for conditions that may benefit from additional evaluation and/or treatment.
How technology can improve the risk adjustment process
Prospective reviews are more challenging than retrospective reviews as they require review of a much wider range of documents and data types. For example, diagnostic radiology reports cannot be used to report diagnoses retrospectively, but they may contain information that strongly suggests an underlying, unreported condition. This information can be abstracted through manual review but the process is greatly enhanced by natural language processing (NLP) algorithms that can identify relevant information in unstructured text, billing codes, and data. This process is augmented further through machine learning and artificial intelligence applications.
The benefits of an automated approach include the ability to improve the efficiency of the query process and in some cases aid reviewers by pointing to clinical indicators that may have otherwise been missed. For example, the drug “Siponimod” (brand name Mayzent) is a relatively new drug used exclusively to treat multiple sclerosis. If this drug was found in a pharmaceutical database or in documentation from a current or prior year it would strongly suggest that the patient has multiple sclerosis. Since the medication is relatively new some reviewers may not be familiar with the drug and its indications. My colleagues went in-depth into how technology can assist in prospective risk adjustment in a recent webinar, which I highly recommend viewing.
Don’t underestimate the importance of quality coding terminology
A comprehensive content library of coding terminology supported by varying levels of specificity allows for improvements in the efficiency and potentially the accuracy of prospective searches. Other clinical indicators, such as certain lab values, physical examination findings, and symptoms may be relatively nonspecific; however, when presented together in context, they may help the provider confirm or refute the presence of an unreported condition.
The majority of the more than 11,000 ICD-10-CM codes that risk adjust in the Medicare Advantage and Commercial Insurance/Accountable Care Act programs have hundreds of relevant clinical indicators. A robust content repository curated by clinicians and augmented by machine learning is needed to optimize the prospective risk adjustment process.
Looking ahead at potential impacts to prospective risk adjustment
It is difficult to predict the future of prospective risk adjustment. One emerging opportunity in this space may be increased access to patient information through interoperability mandates. Under the 21st Century Cures Act and Final Rules published by CMS and the Office of the National Coordinator of Health Information Technology, provider organizations are required to provide patients with access to an electronic version of their medical records stored within electronic health records. Insurance organizations will also be required to share the requesting patient’s claims information upon request. Standardization of Application Programmer Interfaces (APIs) that must support the Fast Interoperability for Healthcare Resource (FHIR) standard will improve access to electronic patient data (if approved by the patient in most scenarios).
There are several unknowns, but these changes have the potential to increase access to relevant patient data and information and further increase the power and sensitivity of prospective searches.
Health Language solutions help healthcare providers and payers manage and benefit from the volumes of clinical patient data that’s generated within the healthcare ecosystem. To learn more, watch our recent webinar or connect with us to speak with an expert.