Harm Ledger · disputed · Medicine & Health
A November 2023 STAT investigation and the Lokken class action allege that naviHealth’s nH Predict model was used to cut Medicare Advantage rehab stays; UnitedHealth says the tool is only a guide.
Current score
−1.26
If proven, algorithmic denial of rehab for many elderly patients is historic-scale health harm (tier 4). Attribution and causation are contested (0.40). STAT plus a surviving lawsuit (0.50). Realization is alleged denials with some documented cases. Status is disputed.
Internal targets allegedly pushed stays within 1 percent of the model’s prediction. Plaintiffs claim high overturn rates on appeal and patient harm including death. In 2025 a Minnesota court let contract and good-faith claims proceed while dismissing others. Whether nH Predict is “AI” in the frontier-model sense and whether it caused denials remain disputed.
Predicted post-acute length of stay used, plaintiffs say, to time coverage cutoffs.
Plaintiffs place the model at the center of denials; the company denies it makes coverage decisions.
Attribution 0.4000 · Credit share 70% · UnitedHealth Group
UnitedHealth employees were pressured to align rehab stays with nH Predict outputs, according to STAT and the complaint.
outcome · supported
nH Predict, rather than clinicians, made legally binding coverage decisions.
attribution · disputed
primary sources
U.S. District Court District of Minnesota · regulatory record · 13 Feb 2025
independent sources
STAT · journalism · 14 Nov 2023
Secondary domains: Computer Science
13 Sep 2026 · 0.00 → 1.26
Initial adjudicated seed score under methodology 0.1.