Mednition

Everything sold here is model output. KATE reads the structured data a triage nurse enters together with the free text of the triage note, using machine learning with clinical natural language processing (the method named in the company's own published research), and returns a risk assessment and acuity recommendation in real time; KATE Sepsis applies the same approach to sepsis at triage. Standard triage runs on the Emergency Severity Index applied by a nurse, which the company builds into KATE through its partnership with the Emergency Nurses Association.

KATE recommends and the nurse decides. The company describes KATE as supporting the triage nurse's judgment in real time inside the existing record workflow with no new screens, and its partnership with the Emergency Nurses Association builds the Emergency Severity Index framework into the tool; the nurse assigns the acuity. The home page adds that KATE helps nurses find high risk patients "without increasing effort or oversight", which describes workload, not a promise that each recommendation is right.

What is missing is the disagreement case: nothing says what happens when the recommendation and the nurse disagree, whether that is recorded, or what the nurse sees to justify a recommendation weighed in seconds.

The method is on the record; the models are not. The company's research page lists its own papers on the method, including a 2021 Journal of Emergency Nursing article on improving Emergency Severity Index acuity assignment using machine learning and clinical natural language processing, a June 2023 preprint on sepsis detection at triage using the same approach, and a March 2025 paper on detecting Sepsis-3 in the emergency department. The customer application data privacy policy says the service ingests record data from each customer by bulk transfer or stream messaging and stores transformations of it for machine learning.

No model architecture, version, training data composition or update practice is published, and nothing says whether models are trained per hospital or across customers.

The chain is short and partly named. The customer application data privacy policy says KATE is Mednition's own cloud based machine learning service, hosted on Google Cloud Platform, and that Mednition signs service and business associate agreements with the third parties that host application data or help deliver the service, requiring them to use the data only for those services, keep it confidential and report unauthorized access. The models are the company's own, per its published research, so no outside model provider sits in the chain.

The other third party processors are not listed, nor is it said where Google Cloud hosts the data or whether customer data trains models used by other hospitals. The policy's statement that application data may be used for enhancing and improving the service leaves that last question open. Triage notes are fast, unreviewed free text that often carries social detail, so the training question matters.

Retrospective multi site evidence and named deployments, without a prospective outcome study. The company's research page lists a March 2025 paper reporting sepsis detection on 540,884 patients across 16 sites, a 2021 Journal of Emergency Nursing article on acuity assignment, and nursing studies on triage accuracy published in 2024 and 2025, all with company authors. The home page says KATE has supported more than five million emergency department visits, names Adventist Health White Memorial and Adventist Health Glendale as sites, and cites a HIMSS 2025 Emerge Innovation "Best in Show" in the capacity crisis challenge.

What is not shown is that deploying KATE changes outcomes: the length of stay, capacity and revenue gains the home page and ROI calculator describe carry no study, baseline or period, and the calculator's figures are estimates from sample inputs.

Categorical commitments are published; retention is not. The customer application data privacy policy says Mednition processes protected health information only to provide the service and on customer instructions, does not sell or rent application data, binds its hosting and service providers to confidentiality and breach notification, and requires any successor in a sale of the business to honor existing customer agreements or have the data destroyed. It also says application data is used for administering, deploying, enhancing and improving the service, and that the service stores transformations of record data for machine learning.

No retention period, deletion practice or de identification position is published, and the improvement clause leaves open whether one hospital's data trains models others use; triage notes are fast, unreviewed free text that often carries social detail. Nothing is published on safety engineering.

A service privacy document backs the stated business associate status. The customer application data privacy policy says record data received from customer hospitals is protected health information, that Mednition acts as a service provider under a master subscription agreement and business associate agreement with each customer, processes that information only to provide the service and on the customer's instructions, and signs business associate agreements with the third parties that host application data. It names the compliance officer and describes how Mednition cooperates with customers on patients' access, amendment and accounting rights.

The business associate agreement itself is not published, so its terms and any tier distinctions cannot be read. The policy also allows application data to be used to enhance and improve the service, and how that squares with the agreement is not shown.

No security certification is published. The customer application data privacy policy commits Mednition to reasonable and appropriate administrative, physical and technical safeguards under HIPAA and to monitoring its service providers; the website privacy notice describes encryption and need to know access for website information; and hosting is on Google Cloud Platform, whose certifications belong to the host rather than to Mednition. No SOC 2 report, ISO 27001 certificate, HITRUST certification, penetration testing statement or trust page is published.

Controls are asserted with nothing independent behind them.

A device pathway is stated and in progress for the sepsis model. The home page announces that FDA granted KATE Sepsis Breakthrough Device Designation (November 2023), a program that speeds review of a device that has not yet been authorized; no clearance or authorization for KATE is claimed. KATE Triage, which recommends an acuity level, is positioned as support for the nurse's judgment.

One tension is unresolved: the site sells KATE Sepsis now ("Stop Sepsis at Your Front Door"), without calling it cleared, while the sepsis model is under a designation that anticipates FDA review.

Equity is stated as an aim and no subgroup results for KATE are published. The site sells a "Deliver True Health Equity" solution, and the company names the tradeoff its category is known for: high sensitivity in sepsis detection is easily bought at the cost of specificity and alert fatigue. Its research page lists papers on triage accuracy, including one on older adults with heart attack and heart failure, without results by race, ethnicity, language, sex or age for the KATE models themselves.

Emergency triage acuity is a documented site of disparity across race, language and presentation, and a model trained on past triage decisions learns from those assignments, so results by patient group are the missing evaluation.

Mednition publishes error rates for its sepsis model with counts. Its March 2025 paper on detecting Sepsis-3 in the emergency department, listed on the company's research page with company authors, reports an area under the curve of 0.99 with 95 percent sensitivity and 96 percent specificity on a retrospective cohort of 540,884 patients containing 14,676 sepsis cases across 16 hospital sites, against the Sepsis-3 definition; an earlier June 2023 preprint, which the company marks as not peer reviewed, reported 0.94. The figures give a buyer an operating point and the positive case count, which is what makes a rare event claim testable.

The evaluation is retrospective and run by the company, and no warranty, indemnity or remediation commitment attaches to the output. The customer data privacy policy routes patient privacy complaints to a compliance officer, which covers data handling rather than what the model says.

Named record systems, one of them through a marketplace. The partners page says KATE integrates natively with Epic, is highlighted in the Connection Hub on Epic Showroom and deploys in two to three months, and integrates natively with Oracle Health (Cerner) systems. The customer application data privacy policy says KATE ingests record data by bulk transfer or stream messaging and sends messages back to clinicians. Named sites include Adventist Health White Memorial and Adventist Health Glendale.

A Connection Hub entry is a directory listing rather than integration documentation, nothing published describes data written back into the record, and no site is tied to a named record system.

A single hosted option, with the host named and no location committed. The customer application data privacy policy says KATE is a cloud based managed service hosted on Google Cloud Platform, with record data ingested from each customer by bulk transfer or stream messaging into the company's cloud storage, and that it may later use other cloud providers with substantially the same protections. No region, tenancy model or residency commitment is published.

Latency shapes the design, since a recommendation that arrives after the nurse assigns acuity has no value, and nothing says where triage text is processed.

No price is published. The ROI calculator estimates savings from a hospital's own inputs (its sample shows about $3.0 million) and offers a value report on request, and the partners page says deployment takes two to three months on Epic; neither states a unit of charge, a fee, or whether KATE Triage, KATE Sepsis and the Clinical Data Engine are licensed together. The customer application data privacy policy describes KATE as a subscription managed service, which says how it is delivered, not what drives its cost.

Nothing on cost can be established before a sales conversation.

Coverage is narrow and named, with validation behind part of it. KATE works at emergency department triage, with KATE Sepsis for sepsis at triage and a solution for maternal risk at triage, and the company sells it to hospital leaders on emergency department throughput, length of stay and capacity as well as safety. The company's research reports a retrospective sepsis evaluation across 16 hospital sites, and named deployments include Adventist Health White Memorial and Adventist Health Glendale.

Validation covers the sepsis model in one retrospective cohort, the throughput and length of stay claims carry no study, and nothing is published on performance by type of site or outside the United States.