Applications: Haematological Screening

A knowledge-based decision support system to screen patients for bleeding disorders.

This application demonstrates the importance of choosing the right application for knowledge-based technology. This particular knowledge base is small and took very little time to build, but the payoff is significant.

The Information Problem

Bleeding disorders and bleeding risk fall on a continuum of severity. People suffering from severe bleeding disorders are likely to have these diagnosed early in life, as the disorders present themselves such that early medical attention is required. However, many bleeding disorders may not be diagnosed. Such undetected disorders could be life-threatening during surgery or childbirth.

People who have bleeding disorders face great danger when they become pregnant or require surgery. Consequently, city hospitals have haematological laboratories where complete diagnostic blood workups are done on people with known bleeding disorders, so that the operating room staff will be equipped to control and stop the bleeding.

A full haematological workup is time consuming. One wants the haematological laboratory staff to get to the patient before they are in trouble, i.e., before they are on the operating table or delivering their baby. It is difficult to find out which of the patients have a bleeding disorder before they end up in surgery or labour. In these situations, routine surgical or birthing procedures become life threatening and some of these patients bleed to death.

Existing Practice

Dr. Jon Gerrard of the Manitoba Institute of Cell Biology (currently Canadian Secretary of State for Science, Research and Development) had been conducting research for many years on isolating readily observable indicators of potential bleeding risk - such as: does the patient have blonde hair; do they bruise easily; do they bruise frequently; etc. These indicators had been discussed in the scientific literature but had not been put to practical use.

Dr. Gerrard felt that these indicators could form a bleeding risk screening test. His idea was to screen patients at admissions or at the time of their referral for hospital admission so that a full haematological workup could be performed prior to surgery or childbirth.

A New Approach

Knowledge-based technology, and Acquire^® in particular, is very well suited to diagnostic screening. It was thought that a knowledge-based system, as opposed to a paper and pencil test, would facilitate screening at hospital admissions as well as ensuring consistency in interpretation.

The Screening System

Dr. Gerrard found the Acquire^® knowledge acquisition system very easy to use. The application prototype was built in one and a half days. The tabular format of the Acquire^® knowledge representation scheme was readily understood, fit the knowledge perfectly and facilitated testing, enhancement and maintenance.

The resultant system involved about 25 patient observations. These observations were interpreted through a variety of intermediate conclusions that ultimately converged on a single conclusion regarding the presence and degree of bleeding risk.

Conclusion

Focusing on a screening system as opposed to a system for complete haematological diagnosis was a good decision. The screening system was built very quickly with an immediate, operational payoff. In addition, the system left the specialists with the more interesting task of isolating the particular haematological problem that the patient faced.

This application indicates that you do not necessarily have to build a large system to get a large benefit.