Our ePRISM® platform gives healthcare providers instantaneous access to the latest clinical prediction models coupled with rich visualization tools. These models may come from national organizations, outcomes researchers or a specific institution. In addition to decision support applications, ePRISM® can be used to rapidly create personalized educational materials, patient letters, informed consent documents and a broad array of other items that can help elevate the quality of healthcare delivery.

- The Challenge

• The past decade of health services research has witnessed an explosion of prognostic models to help physicians understand the risks and benefits of proposed medical therapies. Yet, the delivery of such models to the forefront of patient care has been limited by their sheer complexity.

  Model Coding and Definition Integrity

  The mathematical complexity of modern prognostic models makes their coding difficult and requires reliable mechanisms to ensure that they are properly implemented and tested prior to deployment.

  Model Revisions and Disseminating Updates

  Medical knowledge is constantly advancing and a sustainable mechanism must be available for updating predictive models and disseminating revisions in a timely fashion to clinical end-users.

  Dynamic Data Collection Requirements

  A major challenge in the design of any computational aid is the creation of a data-collection mechanism flexible enough to accommodate changes to the underyling model definitions without requiring costly and time-consuming redesign of the user interface.

  Platform Integration

  A successful system must be able to adapt to a broad range of clinical practice settings and offer secure yet flexible mechanisms for integrating with existing IT infrastructures, disparate clinical information systems, and third-party vendor solutions.

- The Solution

• ePRISM^® from Health Outcomes Sciences provides a powerful and versatile interface for building and deploying the latest evidence-based predictive models as Web-based tools and XML web services. These tools can be updated in real time and can be used for a broad range of clinical applications, including clinical decision support, shared medical decision making, and patient education.

  HL7 interfaces eliminate the need for redundant date entry. A rich set of graphical output options and customizable report templates makes it easy for physicians to interpret results and share them with patients.

  ePRISM^®’s Remote Modeling XML Web Service gives remote clients access to the modeling engine programmaticall without gaining access to the underlying model definitions. This service allows organizations that develop risk models to leverage their intellectual property by making it available to third-party vendors without the risks associated with exposing the models directly to those vendors.

- Model Design

• Visual Model Designer

  An intuitive, yet powerful, patented graphical user interface allows users to implement prediction formulae derived from standard statistical models without having to write a single line of code. A mathematical equation-rendering engine dynamically displays prediction formulae -- including all beta weights and covariance matrix terms -- during construction, minimizing errors by allowing for rapid visual inspection of models.

  Model coefficients can be imported directly from most statistical packages including SAS and SPSS. Advanced validation checks help find errors that are missed by the user and ensure the internal consistency of formulae before their publication.

  
  

  ePRISM^® employs a general regression modeling framework for expressing predictions encompassing all major types of prognostic models including linear, generalized linear, cumulative multinomial, generalized multinomial, proportional hazard and other non-standard model types.

- Clinical Tools

• Clinical User Interface

  ePRISM^® offers healthcare providers and other clinical end-users a robust and easy-to-use interface for running prognostic models and viewing their results from virtually any web-enabled device.

  Clinicians can browse available models and view details on each such as the population from it was derived, the statistical methods employed in its development and a listing of any published references. A mathematical equation displayer allows users to visually inspect model definitions and associated coefficients, thereby providing complete transparency to the risk- modeling process.

  Once the clinician has selected a patient and one or more models to run, ePRISM^® dynamically creates a data input form containing only those data elements required by the requested model(s). The form is pre-populated with any available data values for the specified patient, and any redundant data elements across multiple selected models are automatically eliminated. After entering and/or verifying a patient’s clinical data, the clinician can run the model(s) and view the results in any number of graphical output formats including bar plots, gradient bars, whisker-line plots, pie charts, stacked bar plots and digital LED displays. Outputs from related models can be overlaid to facilitate comparisons (e.g., risk of stroke with PCI versus CABG). Customizable report templates allow such outputs to be readily configured into decision aids, informed consent documents, patient letters and educational materials.

- Web Services

• ePRISM^® offers a number of integrated XML web services that expose many of its core modeling and data access functions to remote web clients through the industry standard Simple Object Access Protocol (SOAP).

  - Data Integration Service
  

  • Offers healthcare organizations a flexible mechanism to rapidly link ePRISM^®’s Data Bank with existing clinical information systems (CIS). Our companion ePRISM^® Connect messaging server, when used in conjunction with a third-party HL7 Gateway such as HL7Connect, provides a turnkey solution for routing incoming HL7-compliant messages from other systems directly to the web service. Additionally, Data Integration Service can optionally receive data via HL7.

  - Library Publishing Service
  

  • Provides functions for publishing all model and supporting parameter definitions as a Predictive Modeling Markup Language (PMML) compliant XML document. Exported definitions can be used in data mining applications that support the PMML standard or for collaborative sharing of model and parameter definitions across different organizations.

  - Remote Modeling Service
  

  • Provides a powerful set of functions for accepting submitted patient data from remote computing clients and executing a set of requested models against that data; processed results are then returned directly to the requesting remote client application. This service offers organizations the option of licensing their models directly to third-party software vendors as an XML web service, allowing revenue to be derived on either a timed or per use basis, without exposing the mathematical structures and associated coefficients of proprietary models to those clients.