To properly account for the acuity of risk-adjusted patient populations, organizations typically turn to a combination of retrospective review and prospective suspecting. The goal being to ensure accuracy and safeguard against incomplete submissions. While this approach works, it does have its drawbacks (total duration, lack of timely information impacting care, etc.). Fortunately, there is another route available. Through technology assisted concurrent coding, organizations can obtain greater benefits and circumvent the limitations of retrospective and prospective analysis. This also builds a better overall picture of a patient’s health burden earlier in the care cycle. Operationally, more accurate coding also leads to more accurate reimbursement for risk taken on and the subsequent care provided.
In as simple of terms as possible, concurrent coding involves moving that retrospective review upstream to just after a provider visit. This allows for new levels of accuracy in documentation and review, both supporting the financial needs of an organization, and in reflecting that true patient disease burden. In a broader sense, it also results in more accurate benchmarking and greater available resources to allocate and improve the quality of care without putting additional coding burden on the physicians.
Concurrent Coding in the Clinical Setting
Concurrent coding is an alternative audit process to correct documentation and coding errors that immediately follows the physician visit, before a claim is submitted for payment. Physicians are not coders, so concurrent coding finds billing oversights that physicians might overlook in the rush to complete electronic paperwork.
In the span of a few days following a patient encounter, there is a window of time where physicians are able to amend clinical notes before they are locked in the EHR. That window gives coders an opportunity to review a visit to ensure that the claim, when submitted, more accurately reflects the patient’s experience. Doing so without putting additional work directly on the physician requires a unique assist in the workflow: an AI equipped with natural language processing (NLP) not just capable, but purposefully “raised,” to process and understand clinical data.
While valuable across the board, this is especially critical for patients with complex conditions. A physician might not code all of the diagnoses on every visit, even though evidence of active management and treatment of those conditions occurred during the consultation and was well documented by that physician. For example, a patient might present with diabetes and congestive heart failure as the primary reasons for their visit. However, the patient is also obese and suffers from hypertension, issues that were also addressed as part of the patient’s visit as part of the treatment plan. When performed properly, concurrent coding would pick up all four conditions, producing a claim that more accurately reflects the depth and complexity of the patient visit. With a well deployed NLP solution, that pick up produces virtually no additional effort for the physician. Doctors are only engaged in cases where an opportunity is accurate but needs better documentation, even then, this only occurs when coders have recognized the most productive opportunities.
While this process can add a day or two before a claim is submitted, the value of getting the coding right on the first submission means a timelier reimbursement and a measurable drop in accounts receivables (A/R) cycles. Additionally, as more physicians share risk and receive compensation based on the complexity of patient populations, concurrent coding helps keep documentation accurate, which in turn allows for more comprehensive care and disease management programs to help improve patient outcomes. This is not only significant in the short term, but it accrues over time as subsequent contract years are built on the prior year’s performance and disease burden benchmarking.
Technology to Best Support the Concurrent Coding Workflow
With Health Fidelity’s Lumanent Post-Encounter Review, information is extracted directly from the EHR and analyzed via the oldest clinically raised natural language processing (NLP) engine. The NLP quickly examines each chart, returning questionable coding incidents to a human coder for further review. A coder works through the prioritized coding queue, adding or removing diagnosis codes where necessary. A physician may be queried in cases where further documentation is required. Once coding gaps are closed, the claim is submitted for payment.
The outcomes are already very optimistic. Based on client experience, 25% of encounters have an NLP identified opportunity (upside or downside) requiring coder review. Because our NLP identifies those encounters and presents them to the coder, the coder does not have to seek them out and risk coming up empty 75% of the time. At the same time, the NLP automatically prioritizes the coder work queue to find the most significant coding opportunities, bringing a more accurate picture to the overall disease burden and ensuring reimbursement is appropriate for the work being provided. In dollars and cents, the ROI is already coming back at a 5:1 ratio with development partners.
Put another way, from a productivity scale, for health plans and provider groups that utilize their own coders, the coder-to-member/patient ratio can rise from one coder to 3,000 members to one coder for 15,000 to 20,000 members. Beyond that prioritization mentioned above, reviews and claim edits take 3-5 minutes, and 60% of encounters flagged for review have a resultant action- to either add or remove a code.
Ultimately, it comes down to this: moving chart review closer to the patient encounter gives providers a truer picture of their patient populations than can be accomplished months afterwards in a year-end retrospective review. Better understanding a patient population helps with risk-sharing care arrangements while increasing reimbursement and lowering accounts receivable days through more accurate claims.