Welcome to the API – Your entry to the HDAI predictive platform

The HDAI API offers privileged access to HDAI’s full suite of validated, accurate, and calibrated analytic models. Upon approval, users may use the API to apply all or some of the models to patient records of interest. Users can build on the HDAI baseline clinical and cost predictors by applying other patient data.

Who benefits from using the HDAI API?

The HDAI API is designed to support researchers, practitioners, administrators, and data scientists throughout healthcare.

Public health, clinical and academic researchers can use the API (at no cost for academic work) to augment their data sets with a broad array of predictors, helping them risk adjust and understand baseline hazard so they can more effectively and efficiently isolate treatment and pandemic effects.

Health systems, payers, and contracting entities such as ACO can use the predictions as a foundational risk layer, augmenting their existing clinical and cost analyses. Data scientists who are building models to predict mortality, complication risks, readmissions, length of stay, and other key business and clinical metrics can access highly calibrated models that can speed up their work and improve understanding of the actual quality and cost performance relative to expected for a given population.

When the API is integrated into EHRs and care management systems, clinicians at the point of care can improve visibility to patient- and population-level predictive models in real-time.

API users are invited to participate in the HDAI Learning Network, designed to provide open discussion for continuous improvement and discovery.

What does the API provide?

The API enables users to access a comprehensive suite of risk indicators using widely available, structured inputs (demographics, billing codes). Categories include adverse events, utilization, cost, interventions, and chronic conditions and can be run for inpatient and community settings. These models are updated and augmented regularly throughout the year.

How does HDAI measure the models’ performance and accuracy?

HDAI’s predictive models are built with a published methodology with diagnostic, prescription, demographic, and other data from 100 million Americans over more than 20 years. For each predictor, we provide a range of performance characteristics, including measures of accuracy and calibration and information about populations and subpopulations to which the models apply. We combine established statistical techniques and proprietary machine-learning approaches to deliver high-resolution and stable models.

Apply to access the API