How the Right AI Can Improve Health Risk Assessments

The Department of Health and Human Services made headlines in October with a report detailing problems with health risk assessments used by Medicare Advantage insurers.

The Office of Inspector General (OIG) investigation found that Medicare Advantage (MA) insurers received an estimated $7.5 billion in MA risk-adjusted payments for 2023 for diagnoses reported only on enrollees’ home risk assessments (HRAs) and HRA-linked chart reviews, and not on any other service records. OIG raised concerns regarding the accuracy of the diagnoses and that enrollees did not receive needed care for serious conditions reported only on HRAs or HRA-linked chart reviews.

The report comes on the heels of a recent Wall Street Journal series showing that MA insurers between 2019 and 2021 earned an average of $1,818 per home HRA visit based on diagnoses for which people received no other treatment.

The OIG report and Wall Street Journal investigation both found that diagnoses made during HRAs often did not result in treatment for those conditions which, in some cases, could be serious.

An avalanche of data

HRAs are typically conducted in a patient’s home by a physician, nurse practitioner or physician’s assistant to collect information on a patient’s health history, medications and overall physical and mental well-being. They can yield valuable information on social and environmental determinants of health that might go undetected in an office setting. These are usually done at the behest of health plans, who often hire third-party vendors to perform them. While similar in scope, HRAs are different from the annual wellness visits usually done in a medical office by a clinician without the involvement of health plans.

Insurers and HRA vendors should share HRA findings with the patient’s care providers, but don’t always. That can leave providers in the dark about their patient’s condition, particularly for patients who skip wellness visits and don’t see their physician often. This could be partly responsible for the missing information in medical records about treatments for conditions diagnosed during HRAs. 

However, even when insurers and vendors share HRAs with providers and patients, the information doesn’t arrive in usable form.

HRA reports typically consist of many pages of data with little organization or structure. They’re usually mailed to the provider, arriving as thick envelopes full of printouts. Hard-pressed clinicians simply do not have the time or patience to wade through the reports looking for diagnoses and other critical information. As a result, care can be compromised.

Lost in transition

This is typical of many patient care transactions, not just HRAs. Valuable data is lost as a patient moves from one provider to another and from one care setting to the next because it is not easily summarized and prioritized. Instead, the new providers are buried under nearly endless pages of transition reports full of diagnoses, observations, lab results, vital signs and more. These reports usually are spat out by EHRs with little structure or organization to them. It can feel like paging through a non-alphabetized phonebook to find a single number.

Healthcare is the only industry where documentation is divorced from the need to communicate with other people. Lots of data, but not much messaging.

Healthcare needs a better way to collect information from all sources, organize it, summarize it and present it to providers in a way that lets them deliver the necessary treatment, confident that the material is sound and verifiable.

Clinically validated AI solution

The answer is a care transition solution guided by responsible AI that is developed by physicians and grounded in clinical evidence. This reasoning-based approach is a step beyond typical healthcare AI, which relies heavily on complex algorithms and generic prompts while providing little transparency into the evidence behind recommendations.

Care transition and management guided by clinically validated AI rests on three pillars:

1. Physician-developed AI logic to structure outputs. This approach is based on a comprehensive set of risk factors and interventions defined and proven by physicians. Risk factors are rooted in specific diagnostic criteria and medical metadata, backed by the latest clinical guidelines and peer-reviewed literature. This ensures that its recommendations reflect medical standards and are well-suited to the complexities of patient care. 
2. Structured intervention framework. This presents interventions as specific, actionable steps. Each intervention action includes a priority level, clear description, interaction types, the healthcare professionals responsible for the action, and recommended tools.
3. Multidisciplinary integration. This approach acknowledges the complexity of patient care and integrates input from the necessary range of medical professionals. This supports the use of coordinated and comprehensive treatment plans.

Clinically validated AI solutions improve care transitions, whether it’s making sure an HRA report doesn’t get ignored or managing a high-risk move from a hospital to a post-acute setting. It improves coordination, activates appropriate interventions, and prevents adverse events. It ultimately creates documentation with clear communication and messaging aimed at facilitating patient care.

Photo: Asawin_Klabma, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.