Alignment Healthcare AVA AI Flags 70% of Near-Term Admissions

Alignment Healthcare, Inc. (ALHC) said on its July 30, 2026, second-quarter earnings call that the latest version of its AVA AI member risk-stratification model had been deployed. The Alignment Healthcare AVA AI risk model identifies about 10% of members who, according to the company, account for nearly 70% of hospital admissions over the next 30 days. That figure demonstrates risk identification, not that admissions or medical costs have fallen.[1]

Alignment Healthcare operates Medicare Advantage health plans for seniors. It primarily receives monthly premiums from the US federal government for enrolled members and bears the associated medical-service and administrative costs. AVA is the company's proprietary data and technology platform. Its AI stratification model supports the Care Anywhere clinical and medical-management teams by ranking members according to their risk of hospitalization, helping doctors, nurses and other staff prioritize outreach, chronic-disease management and care at home. Because this workflow directly guides the allocation of care resources for high-risk members, AVA member risk stratification is part of core clinical operations, not a general back-office tool.

From Planned Upgrade to Deployed Model

AVA member risk stratification moved from an upgrade concept to a new production version over four quarters. On July 31, 2025, management said it was evolving the machine-learning algorithms in its care model and applying AI techniques to generate insights and act faster, but it provided neither a deployment schedule nor results.[4] In February 2026, the company said more specifically that it was revisiting AVA's original stratification model and using new tools to improve stratification for Care Anywhere members.[3] In April, management described the upgrade as the "next leap forward" in making clinical engagement more precise, again without a result metric.[2] On July 30, 2026, the latest AVA AI member-stratification model was deployed and the company disclosed its first application-specific figure.[1]

What the 10% and 70% Figures Show

The latest model dynamically identifies about 10% of members who, the company says, account for nearly 70% of hospital admissions over the following 30 days.[1] The two figures describe a single prediction: admission risk is concentrated in a small member cohort, allowing clinical teams to focus limited proactive outreach and care capacity on those people. They do not measure an actual intervention rate, avoided admissions or dollar savings. The company also did not disclose a baseline for the previous model, so the new model's incremental improvement cannot be quantified. An earlier disclosure that 10% of members accounted for 78% of spending concerns a different population and metric and cannot be combined with this admission forecast to imply a trend.

The Financial Link Still Needs Evidence

This workflow could eventually affect medical costs and the medical benefit ratio (MBR). If teams identify members most likely to be admitted sooner and effective interventions reduce avoidable hospitalizations, medical spending could decline. However, the company said only that deployment of the new model supported second-quarter performance; it did not isolate any contribution to admission volume, medical costs or MBR. Seasonality, member mix, disease-recording systems and clinical execution can also affect results in the same period. Two layers of evidence—intervention and outcome—still separate risk-identification performance from financial improvement.

AVA member risk stratification has progressed from a model under development to a production tool with a quantified reading. The confirmed operational change is that the Care Anywhere team now has a more focused ranking of members' hospitalization risk over the next 30 days. Establishing financial value will require separate disclosure of intervention rates among identified members, reductions in admissions and the related medical-cost savings.

Application Assessment

• AVA member risk stratification | Business position: Core operations | Application stage: Limited production | Deployment scope: Single business unit | Value type: Cost reduction

Sources

[1] Drillr · Alignment Healthcare, Inc. (ALHC) · July 30, 2026 · Earnings call

Original: Our model now accurately and dynamically predicts the 10% of members who account for nearly 70% of hospital admissions over the next 30 days.

[2] Drillr · Alignment Healthcare, Inc. (ALHC) · April 30, 2026 · Earnings call

Original: Meanwhile, we are also deploying contract management solutions that leverage AI to create a more dynamic contract management platform and taking the next leap forward in our AVA AI risk gratification models to create even greater precision in our clinical engagement efforts.

[3] Drillr · Alignment Healthcare, Inc. (ALHC) · February 26, 2026 · Earnings call

Original: we talk about AVA, and we're looking at using the new tools to make the stratification model even better for our Care Anywhere members.

[4] Drillr · Alignment Healthcare, Inc. (ALHC) · July 31, 2025 · Earnings call

Original: I think we are looking very carefully at evolving our machine learning algorithms and applying those AI techniques into the next generation of AI.