Executive Summary

Integrated Delivery Network System for Ambulatory Electronic Medical Records

Company/Business Overview

The intention of this company is to provide an integrated delivery network for electronic ambulatory medical records using a federated virtual database methodology. This system runs on an artificial intelligence system originally developed for the department of defense by Dr. Eric Mettala. It is currently in use in the state of Connecticut for aggregating medical records form 410 disparate databases in hospitals all over the state. This technology is a result of an eleven man-year research project in the area of artificial intelligence.

Dr. Eric Mettala leads a group of scientists that have developed a suite of tools known as ACT Ontology that can be used for the aggregation of information across disparate data silos and apply business intelligence functionality such as diction support systems and management reporting. The unique value proposition of this technology is an enterprise integration solution with business intelligence that does not use a data warehouse.

This new business model will exploit this technology to develop and sell an off the shelf enterprise solution for hospitals. This solution will aggregate medical records across disparate data sources or silos, including unstructured data, in order to make patient information available in real time to clinicians in the examining room. It will also apply that information to clinical decision support system. The major factor that makes this solution different is the combination of the following attributes:

· Aggregation of medical records across disparate data sources in real time.

· Business intelligence functionality.

· No data warehouse.

Basic functionality:

Aggregation of clinical, and demographic patient information in accordance with HIPAA standards from:

Pharmacy
Lab
Radiology
Accounting and administration

Simple intranet based user interface for doctors and nurses.
Clinical decision support system including:

Suggested diagnosis based on patients who exhibit similar characteristics from aggregated patient information.
Suggested treatment options based on historical methodology form other doctors also from aggregated patient information.
Detailed analysis of success rates for treatment options.

Ability for doctors and nurses to input new information in examining room such as doctors notes and ordering lab test.

An object-oriented “metadata” representation can be used to support queries, browsing, and many kinds of analyses. For example, the artificial intelligence capabilities will also allow doctors, administrators, and researchers to run ad hock queries against patient information. This capability will allow users to develop corollary and causal relationships between different factors effecting patient health.

By enabling doctors to have access to patient information in the examining room, this can increase the number of patients that a doctor can visit by as much as 30% and at the same time increase the amount of time a doctor can spend with each patient. Because the system is intelligent it will be able to assist in the diagnosis and treatment process. ACT*Ontology will allow the doctors to access information about associated corollary factors effecting illness as well as the effectiveness of different treatment methods.

Medical advantages:

o 30% increase in patient visits.

o Increase in amount of time spent with each patient.

o Assist doctors in the process of diagnosis.

o Correlate multiple independent patient attributes and environmental factors with probable illnesses.

o Provide doctors with information concerning the effectiveness of treatment methods.

o Provide hospital administrators and researchers with the cost and effectiveness of differing treatment methods.

Technical advantages:

o Middleware independent of a data warehouse.

o Reduced legal issues concerning data ownership.

o Implementation through existing integrated development environment utilized by non-technical analysts.

o Java based platform independent system.

o Utilization of intranet networking capability through a web portal.

o Intelligent translators to integrate existing databases using built in drivers.

o Greater processing speed of any existing business intelligence or middleware system.

o The model is non-evasive, since neither the source applications nor their data are modified by the system.

o Provides dynamic browsing among related objects, expert system analysis, or flexible growth of the model over time.

o Seamless integration will all known data sources.

Existing Hospital Environment

Commitment by staff and administration to high quality of patient care.
Large complex organization with many disparate processes and organizational structures.
Complex Processes reliant of the efficient flow of information through the organization.
Multiple departmental organizations acting as islands of information with their own set of processes and procedures dependant on external sources of information.

Specialists are dependant on referrals form primary care givers as well as information for lab reports and historical patient records.
Nursing staff dependant on clear instructions from doctors for administering care to the patients.

Medication
Diet
Monitoring

Lab relies on clear instructions from specialists and primary health care provider for the type of test required for proper diagnosis.
Doctors require complete and accurate historical patient records for proper diagnosis and treatment options as well as availability specialists within a geographical area.

Commitment to the community to provide accurate and reliable information about patient care and knowledge of the specific needs of the surrounding community.
Requirement to track patient safety with regard to ordering and administering medication, duplication of testing, and the timely scheduling of specialist appointments.
Ongoing projects dedicated to the continuous improvement of patient care.
Budgetary concerns balancing the need for quality patient care through the equitable allocation of limited capital and human resources.

Complex Hospital Systems Simplified

The diagram above depicts many different departments or islands of data within an organization. Each departments needs to share information with many other entitles. The processes that define the relationships of these disparate entities often rely on physical documentation that must be transported from one location to another.

As the number of entities increase, the number of connections required to link these entities increase exponentially. Likewise the processes that govern the flow of information will grow in complexity. Even a seemingly simple process such as a doctor visit can be become very complex taking into consideration the possibility of lab tests, follow up visits, and scheduling specialists.

All of these factors can result in delays in the process of diagnosis and treatment. Not only will patient care suffer as a result of these inefficiencies, but costs will increase do to reduces patient turnover. The longer a patient stays within the system the greater the opportunity for degradation in patient care likewise every service associated with the direct cost of patient care will increase in direct proportion.

Aggregating information through a single portal or gateway will reduce the complexity of the system and the number of connections or processes needed gather information and share that information with other members in the system.

As information flows more freely through the system patients will be treated more accurately, safely, and with a greater standard of care. Patients will experience a faster response times and the hospital will in turn process a greater number a patients at a lower cost.

Overall effectiveness:

o Reduce processing time

o Reduce costs

o Increase quality of patient care and safety.

o Increase time staff can spend with each patient.

The existing networked environment in a hospital can be described by Metcalfe’s law as follows :

· The number of possible network connections = {[n (n+1)]/2} - n

In a networked environment utilizing a portal there is a 1 to 1 relationship between the number of entities and the number of networked connections. In this simple example one can realize a 33% increase efficiency through the use of an enterprise web portal to aggregate patient information.

Many of the connections as defined in the existing hospital network are not currently automated. Physical processes define many of the network connections, to disparate user groups in the current system. A patient record or results from a lab test being couriered to a specialist for review and diagnosis can illustrate one example of this process.

At Kiser Permanente this process resulted in a 10 day delay in doctors receiving the results of a lab test. After automating this process doctors could get lab results in a matter of hours.

With the use of ACT*Ontology results for a lab test can be made available to doctors in real time. This system will reduce the complexity of the existing system and automate the process thereby greatly reducing inefficiencies and administrative costs. The current system may require staff to place a phone call, make a fax, currier a document, or perform multiple data entries. Through the implementation of ACT*Ontology these processes can be automated and information can flow freely and instantaneously between user groups.

Competition and alternatives

· Data Aggregation:

Currently there are a variety of solutions available that can aggregate patient information and allow doctors real time access in the examining room. This is particularly true with regard to lab records and radiology reports.

Generally these solutions represent 80% to 90% services and 10% software. The reason for this is due to variety of disparate data sources and procedures that are unique to every hospital or clinic. A consulting firm such as IBM or E&Y will build a system that will conform to whatever existing processes, procedures, and data that are persistent in a hospital.

Development tools such as Optimal J by Compuware utilize an easy graphical interface and all of the necessary drivers to drill down into existing databases. Objects that can define abstractions of patient information can be quickly and easily built. These objects exist on a web server and can be accessed remotely through a secure intranet. In this was disparate data can be aggregated from many sources and made available to doctors in an easy to use graphical interface.

With this system information can be accessed timely and accurately, however intelligent analysis combined with real time data is not possible with other existing solutions.

· Business Intelligence:

The industry standard in business intelligence is IBM’s business intelligence group. Again the majority of their work consists mostly of business services where systems are built to comply with an existing process.

The implementation of an IBM business intelligence system requires the implementation of a new data warehouse. Data is mined from existing disparate sourced or extracted, transformed, and loaded into a data warehouse. The intelligent system then runs off of the historical data in the data warehouse. The major issue with this type of system is that it relies on historical data. Real time access to transactional data is not possible. For this reason information such as lab reports can not made available in a timely fashion. Most importantly, the lack of explicit representation of semantic relations in relational databases also severely restricts data warehousing solutions, which makes it difficult for them to provide capabilities such as dynamic browsing among related objects.

The ACT*Ontology intelligent system combines the advantages of these two solutions. Data can be accessed in real time, transformed into usable information, and made available to users over a secure intranet.

Through the use of intelligent capabilities data can be analyzed and transformed into usable information. Business rules can be applied to prompt doctors as to a potential diagnosis as well as different treatment options including appropriate medication based on patient demographics such as age and weight.

Unlike other aggregation tools the ACT*Ontology solution will allow complex data analysis without the implementation of a data warehouse. The following chart compares these attributes:

Functional Requirements	ACT*Ontology Business Intelligence	IBM Business Intelligence (Data Warehouse)	Java based integration tool
Real time data access	a Yes	NO	a Yes
Transactional data access and real time aggregation	a Yes	NO	a Yes
Data Analysis Tools	a Yes	a Yes	NO
Statistical Analysis tools	a Yes	a Yes	NO
Data Discovery and Mining	a Yes	a Yes	NO
Decision support	a Yes	a Yes	NO
Management Reporting	a Yes	a Yes	NO

“ACT*Ontology uses an object-oriented knowledge representation as its federation model. Using ACT*Epilog Transformer, developers construct a common object-oriented enterprise model that represents data from legacy data sources, such as relational databases, object databases, documents, CORBA applications, spreadsheets, XML and so forth.

In our approach each data source is accessed by a data access agent called a “ACT*Data Access Agent” or DAA. A DAA is a software module that knows how to use the data/information source’s Application Programmer’s Interface (APIs) to retrieve specific information as required by the Enterprise model. The DAA also knows how to translate source data into the enterprise model’s object representation. Each data source, together with its DAA, makes up one “federate” in the federation model. The federates, together with the common enterprise model, make up the federation. The benefits are:

The federation approach, at any level, is non-invasive, since neither the source applications nor their data are modified by the system, unless data-updating is desired by the business rules or policies.

The federation approach is more scalable and maintainable than point-to-point solutions, since integration is performed only once per data source. Each federate knows how to translate its data into the enterprise model representation, as well as how to translate the enterprise model representation into source database formats. How that data is used by other applications, including the web, is not of concern to a federate. In other words, any number of applications can use data produced by another application without ever knowing anything about the underlying system or data representation.

The knowledge federation approach overcomes the problems inherent to the solutions discussed above by providing a flexible, evolvable integration framework that serves as a dynamic “virtual” warehouse. Data (structured or unstructured) is dynamically retrieved from its source (database, object store or text) as required by the application, and translated into an object representation in the enterprise model.

New data sources can be dynamically added by extending the enterprise model with new object classes and developing a new translator, or federate. Business rules may be applied to the translation process, to monitoring information events in the enterprise, or to ensure consistency and completeness among information providers and producers.

A further value of this approach is that data is truly integrated as semantically linked objects in the enterprise model. In other words, a “part” object from the inventory database can be linked to a “purchase order” object from the back office database through a semantic relation “ordered-by”, as illustrated in the figure below. The inverse relation, “orders,” provides a means of getting from the purchase order object to the part object. These semantic linkages provide the means of integrating information from different data sources into a meaningful enterprise context.

Representation of semantic links among objects in enterprise model.

This integrated enterprise model provides an object-oriented “metadata” representation that can be used to support queries, browsing, and many kinds of analyses. It also serves as an agreement between all stakeholders (federation participants) about how they will communicate and share information.

This is not possible in point-to-point or EAI solutions, since there is no common model of the enterprise to support analysis, data management or general information sharing. The lack of an explicit representation of semantic relations in relational databases also severely restricts data warehousing solutions, which makes it difficult for them to provide capabilities like dynamic browsing among related objects, expert systems analysis, or flexible growth of the model over time.

Once a federation has been established, clients can search the enterprise model based on their own criteria, as though it were a single virtual knowledge source available over the web. Using web presentation methodologies, such as Java, HTML or XML, this semantically linked network of objects can also be navigated using typical web-enabled browsers. This is important both for navigating from one related object to another and for discovering new implicit relationships in the enterprise knowledge base.

Advantages of the federation approach over data warehousing include:

Evolvable framework: the object model is easy to update when a new data source, application, or process needs to be incorporated, unlike the data warehouse schema.

Fresh source data: data is dynamically retrieved from source databases on demand; it is not duplicated in a warehouse.

Unified, semantically integrated enterprise views: data and other information from various sources are viewed as a unified whole with semantic context.”

Source: The ACT Product Suite White Paper, see appendix

The following diagrams illustrate differences between the ACT*Ontology system and traditional business intelligence solutions.