I’ve written about aspects of the patient journey in the past, and one thorny difficulty is the problem of inappropriate claim denials. Note, I used the word “difficulty” here, not “challenge” because there is no need for diplomacy here. When someone is seriously ill and needs medical care, and obtains that care within the insurer or plan network, wrongful rejection of claims (whether partial or in full) is a dreaded and wholly avoidable chapter in the patient’s medical journey. From the perspective of the insurer or health plan in 2023, the denial of appropriate, in-network care could be described in a couple of pejorative terms: (1) greedy and/or (2) incompetent.
As reported in Fierce Health Payer, claims filed in 2021 for payment by exchange plans were analyzed by the Kaiser Family Foundation (KFF), finding that some plans erroneously rejected 40% or more of payment claims generated by in-network providers. This is then automatically billed to the patient, who can then appeal the denial. Even after the appeal is filed, plans commonly deny the claim.
Within exclusive provider organizations (EPOs)
and health maintenance organizations (HMOs), referrals are required for nearly any specialty office visit or service. This is a common cause of denial of services (preferred provider organizations usually waive referral requirements when services are delivered in-network). Even if the specialty referral is obtained, this is no guarantee that the insurer will deny the claim: either incorrectly identifying a specialty provider as out of the plan’s network or failing to update the provider directory and listing the specialist as a network provider erroneously.
Across the country, average denial rate in 2021 for in-network services was 17% for these exchange plans. Celtic Insurance Company of Florida denied 42% of these claims. In contrast, Bright Health Insurance Company of Florida denied only 6% of these claims. According to KFF’s data, 56% of all insurers denied at least one-fifth of their incoming claims.
The reasons given for payment denial is of interest, as well. For example, Cigna Health and Life of Tennessee rejected 180,124 claims from members of its silver EPO plan, 37% of those for reasons of medical necessity. Overall, of the 44.7 million claims rejected for in-network services, 8% were for lack of prior authorization/referral, 13% for excluded services/benefits, 2% for medical necessity, and 76% for “all other reasons” (which includes out-of-network services). Examples provided by Kaiser counter each one of these reasons, including denial of epinephrine injections and steroid therapy for an anaphylactic reaction as “not medically necessary” or the insurer categorizing cardiac ablation for arrhythmia as “coverage for injections into the spine” (despite receiving prior approval for the procedure from the insurer). For the former, filing of two appeals did not resolve the issue.
Claims denials is a big revenue generator or saver, depending on how you view it, for insurers. Data from the analysis showed that only 0.2% of all claim denials were appealed (based on initial appeals only), regardless of whether the denial was appropriate. This contributes $11 billion annually in extra revenue to the insurance industry. I cannot estimate the amount of aggravation, waste of time and resources, and medical debt this causes patients. Claims denial rates should be of priority interest to consumers who are choosing an insurance plan.
As reported by KFF, “Consumers are not provided any information about how reliably marketplace plan options pay claims and plans reporting high claims denial rates do not appear to face any consequences.”
We are on the cusp of an artificial intelligence (AI) revolution in this country, and one might expect that AI applied to the process of claims review could improve the situation. Of course, this depends on the algorithm used by the individual insurer. One investigation, published in March by KFF, found Cigna’s use of an automated system allowed medical reviewers to make claims acceptance/denial decisions on 50 charts in 10 seconds. At this speed, a review of the patient’s records is not humanly possible.
Assuming an accurate algorithm is utilized, AI should only be used as a first step in the process—to pick out claims that seem to fail some basic tests, like out-of-network providers for a member of an EPO (which does not cover any nonnetwork services). Following this identification, the claim should be manually reviewed before a claim denial is sent. Clearly, the system is not working in several plans, and the external checks on the system are not being conducted as directed by law.
