GE HealthCare Command Center

Prediction is what the newer product is for. CareIntellect for Operations is built around two models: Pressure Forecast, which reads historical operations data and live feeds from the record and resource systems to predict constraints up to 72 hours ahead, and Estimated Day of Discharge, which predicts each patient's likely discharge day from longitudinal data and updates every hour. The recommendations and reprioritized work queues come out of those predictions, and a digital twin lets teams test what if scenarios on a simulated copy of the health system before changing staffing or capacity.

The installed Command Center is a mix. Many tiles are real time visibility tools for beds, transfers and care progression that work without a model, while others carry predictive census and capacity forecasts; GE HealthCare describes Command Center as using AI and machine learning to optimize capacity at Alfred Health. A hospital buying tiles for visibility alone gets workflow software, and one buying CareIntellect for Operations gets a forecasting product.

The software advises and people decide. Pressure Forecast flags where constraints are building, Estimated Day of Discharge gives each patient a predicted discharge day, and the application recommends actions and reorders operational work queues for capacity, discharge, imaging and transfer teams. GE HealthCare's chief science and technology officer describes the design goal as predictions teams can understand and recommendations tied to their workflow. No step acts on a patient, bed or order without a person.

What staff see about why a forecast changed, and how often recommendations are overridden, is not described.

Pressure Forecast combines historical operations data with live record and resource system data to predict census, staffing, boarding and waits across the emergency department, therapies, procedural recovery and inbound transfers up to 72 hours ahead, at unit, department and enterprise level. Estimated Day of Discharge uses longitudinal patient data to predict each patient's discharge day, with estimates refreshed hourly. A digital twin simulates the health system so teams can test what if scenarios before acting. All of it is proprietary to GE HealthCare and runs in the CareIntellect cloud infrastructure on AWS.

The model type, training data, whether models are fitted per hospital, update cadence and any accuracy measure are not given.

Both models are GE HealthCare's own, so no outside model provider sits in the forecasting path. CareIntellect for Operations runs on Amazon Web Services and is sold through AWS Marketplace as well as directly, which puts AWS in the chain as the host.

No subprocessor list is published, and whether data is shared with AWS services beyond hosting, or used to train models across customers, is not stated.

The evidence is customer case studies with numbers and a stated method. At The Queen's Health Systems in Honolulu, ten months after its command center opened, length of stay fell 1.07 days, emergency department length of stay for admitted patients fell 41.2 percent (248 minutes), boarding fell 63.9 percent and accepted transfers rose 22.2 percent, about 100 more a month. The case study says the metrics came from Queen's, covers January to October 2024, and bases the $20 million first year saving on year over year length of stay at an assumed $1,200 per patient day. Providence Swedish reports capacity for 19,000 more patients a year from a regional command center.

Children's Mercy Kansas City opened its Patient Progression Hub on Command Center tiles in May 2023. Comparing June to December 2023 with the same months of 2022, it reports avoidable days down 24 percent, room for 300 more patients on its medical and surgical floors, transfers from pediatric intensive care to the floor 20 percent faster, deferrals down 94 percent and admissions up 2.8 percent, with census and surge days about the same.

These are before and after results for whole programs, which at Queen's also included surge plans and daily discharge rounds, so they cannot be read as the effect of the software or its models alone. GE HealthCare's own footnote says results vary with the record system and level of adoption. No study of the Pressure Forecast or Estimated Day of Discharge models has been published.

The models draw on hundreds of patient and operational data points, including bed availability, patient delays, staffing, wait times and ancillary services, combined with historical operations data and live feeds from the record system. CareIntellect applications share one data layer on a fully managed AWS infrastructure with single sign on, so a hospital's data feeds several applications from one place rather than being copied into each.

How long that data is kept, whether one hospital's data is used to train or tune the models offered to others, and whether any of it is de identified are not stated.

The product reads identifiable patient data from the record system, including bed placement, delays and longitudinal patient information for discharge prediction, so it sits inside a US customer's HIPAA program. The Command Center and CareIntellect for Operations pages, the AWS Marketplace listing and the launch announcement say nothing about business associate status, and no agreement, template or tier is published.

Outside the United States, at Alfred Health or NHS sites, the equivalent obligations come from Australian privacy law and UK data protection law, and no data processing position is published for those either.

GE HealthCare runs a product security portal and a cybersecurity program across its equipment and software, linked from its site navigation. For Command Center and CareIntellect for Operations, though, no SOC 2 report, HITRUST certification or ISO 27001 certificate is named, and the product pages describe security only as secure onboarding of new applications. Running on AWS brings AWS's own certifications for the infrastructure, which do not cover GE HealthCare's controls over the application and its data.

Command Center and CareIntellect for Operations are presented as operational software for capacity, staffing and patient flow, and nothing on the product pages claims a clinical diagnosis or treatment function. GE HealthCare states no regulatory position for either product. The CareIntellect for Operations page says only that it is commercially available in the United States and may not be available elsewhere, and the Command Center pages carry the general notice that not all products or features are available in all markets.

Estimated Day of Discharge predicts a patient level event from clinical data, which is the part of the product closest to clinical decision making.

The Queen's Health Systems and Duke Health are the first clinical evaluation sites for CareIntellect for Operations, and frontline feedback from both shapes later releases. GE HealthCare frames the aim as predictions teams can understand and trust, which describes how the models are developed rather than how they are checked.

No validation results, monitoring practice or subgroup analysis is published for either model. A discharge prediction can run consistently late for patients who wait on post acute placement or social care, and those waits differ by payer, age and community; whether the model's errors are tracked by group is not stated.

Each forecast and recommendation lands with operational staff, who decide whether to act on it. Pressure Forecast and Estimated Day of Discharge carry no published accuracy, error rate or calibration figure, so a hospital has no stated level of performance to hold them to, and no warranty, service commitment on forecasts or correction route appears in public material.

The case study footnote says results are not prescriptive and vary with the user, the record software and the level of adoption, which places outcome risk with the hospital.

Command Center integrates streaming data from electronic medical records and other source systems, and CareIntellect connects to the medical record and to the customer's single sign on. Alfred Health and close to 500 other facilities run it on top of their own record systems, which it reads rather than replaces.

The record systems behind named deployments, the integration standards used, and any documentation or listing in a record vendor's app marketplace are not published, and GE HealthCare's case study note says results vary with the record software in use.

CareIntellect for Operations is software as a service on a fully managed AWS infrastructure, part of a cloud first family meant to let customers add applications without integrating each one separately, and it is listed on AWS Marketplace. It is commercially available in the United States only. Command Center itself is delivered as tiles activated in phases, as at Alfred Health.

The AWS regions used, the tenancy model and any residency commitment are not stated. For Command Center sites in Australia, Canada and the United Kingdom, where patient data is processed is not addressed either.

CareIntellect for Operations is sold by GE HealthCare directly and through AWS Marketplace, where the listing offers custom pricing through a private offer rather than a public rate. Command Center is bought as a set of tiles that can be added over time, from Capacity Expediter and Patient Manager to Hospital Pulse and the ED Expediter, and Alfred Health is activating its tiles in phases.

No price, unit of charge, subscription term or implementation fee appears on the product pages, the Marketplace listing or the launch announcement, so the cost of a starting set of tiles, or of moving from Command Center to CareIntellect for Operations, is not shown.

Command Center runs across hospital types and health systems in several countries. GE HealthCare cites close to 500 hospitals and facilities, with named users including The Queen's Health Systems (a 575 bed trauma center and a 120 bed community hospital on Oahu), Duke Health (a user since 2019), Providence Swedish in the Puget Sound region, Children's Mercy Kansas City, Halifax Infirmary in Nova Scotia, Bradford Royal Infirmary in England, and Alfred Health's three Melbourne hospitals. Tiles cover beds, transfers, intake, discharge, procedural growth and the emergency department.

The CareIntellect for Operations forecasts cover inpatient units, the emergency department, imaging, therapies, procedural recovery and inbound transfers, but the product is commercially available in the United States only, so sites such as Alfred Health and Bradford run Command Center without it. The published results come from United States sites, and how the models would perform in public systems with different admission patterns and discharge rules is not reported.