Healthleap

How does the model work?

The validation paper describes a gradient-boosted decision tree model that uses a large language model for feature extraction. The vendor positions the score upstream of clinical documentation improvement and coding, and lists delirium and toxic metabolic encephalopathy, aspiration pneumonia, heart failure readmissions and pressure injuries as planned or adjacent screening areas.

The vendor refers only to "the EHR" and names none. It lists SOC 2 Type II and HIPAA compliance and makes no FDA claim, describing the product as clinical decision support and not a diagnostic tool.

What did the validation study show?

Bernstein and colleagues published it in Applied Clinical Informatics in 2025 (PMID 40523638). It used EHR data from 106,449 patients over 3.75 years. The AUROC was 0.92 (CI 0.91 to 0.92) on day 1 and 0.95 (CI 0.95 to 0.96) using the maximum risk during the stay, and the model outperformed the nurse-administered modified Malnutrition Screening Tool. Four authors are Healthleap employees and three are advisors. The vendor claims "88% higher sensitivity than nurse-administered screening".

What do hospitals report?

The vendor's evidence page gives deployment results. Penn Medicine's HUP shows an annualized impact of $23.8 million, made up of $6.3 million in reimbursement lift and $17.5 million from length of stay, plus a 21% year-over-year rise in malnutrition diagnoses and a 13% risk-adjusted length-of-stay improvement. Cedars-Sinai shows $11 million in annual impact and 39% more malnutrition diagnoses. The vendor publishes all of these figures.

The company and its contract

Healthleap, Inc. is based in San Francisco and was founded in 2022 by siblings Josiah Meyer, CEO, and Jemima Meyer, a clinician. The vendor states $38 million in combined seed and Series A funding from Sequoia Capital, First Round Capital and Hummingbird Ventures, with no date or lead shown. It says it screens every adult inpatient daily at more than 20 hospitals, naming Penn Medicine, Houston Methodist, Cedars-Sinai and Emory Healthcare, and calls itself, "to our knowledge", the only commercially available peer-reviewed AI malnutrition screening tool.

No price is public. The evidence page describes "a 2:1 ROI floor on hard reimbursement", under which Healthleap pays back any shortfall.

The verdict

It fits a hospital nutrition team that wants a daily risk-sorted list of adult inpatients to assess first, and the vendor's deployment figures concern malnutrition diagnoses, reimbursement and length of stay. It does not fit a pediatric service, since it screens adults only, or a committee that requires validation by authors outside the vendor, since four authors of the validation paper work for the company.