An Expert System on Detection, Evaluation and Treatment of Hypertension - Artificial Intelligence and Symbolic Computation - page 252

1 Introduction

A considerable amount of work, including [1] (probably the best known, although
not the only example) has been carried out in the field of medical rule-based ex-
pert systems. For instance, a formal method for creating appropriateness criteria
[2] has been developed at the RAND Health Service (Santa Monica, California):
it is based on combining literature review and ratings from expert panels. One
problem with the current RAND methodology is to ensure logical consistency:
RAND researchers manually develop algorithms from ratings and re-meet with
experts to settle inconsistencies [3]. Alternative approaches to this problem have
been proposed e.g. in [4-6].
Many health organizations are now developing and using evidence-based poli-
cies for medical care. They are created using systematic methodologies. In the
field of major cardiovascular disease (CVD) risk and hypertension (high blood
pressure), we can underline: the International Society of Hypertension of the
World Health Organization (ISH-WHO), the Joint National Committee on Pre-
vention, Detection, Evaluation and Treatment of High Blood Pressure (JNC) of
the National Heart, Lung, and Blood Institute ( U.S. Dept. of Health and Human
Services )... [7-11]. They are not designed for retrospective evaluation of medi-
cal practice, but as a decision-making aid. As far as we know, expert systems
directly based on these reports in the field of hypertension haven't been spread.
A new proposal of expert system on the topic is detailed below.

2 General Description

This article extracts the information and knowledge from different sources [7-
11] and merges it with the knowledge of the three coauthors that are MDs. Let
us underline that one of the authors is the chief of the Hypertension Unit of
one of the main Spanish hospitals and other of the authors is finishing his PhD
in Medicine on this topic. The accurate knowledge representation needed to
implement these topics has required a refinement in the details of the tables and
algorithms provided. Finally the logic and computational processes are simple
but sound.

2.1 Data Acquisition
First, different data from the patient have to be acquired. They are:
- Blood Pressure (BP): Systolic Blood Pressure (SBP) and Diastolic Blood
Pressure (DBP) figures.
- Major cardiovascular disease (CVD) risk (Boolean) factors, apart from hy-
pertension (*): obesity, dyslipidemia, diabetes melitus (DM) (

), cigarette
smoking, physical inactivity, microalbuminuria (estimated filtration rate <
60mL/min)(

†

), age ( > 55 for men, > 65 for women), family history of pre-
mature CVD (men age < 55, women age < 65). Note that hypertension is
another CVD risk factor, but it is not a datum, as it is deduced from the
SBD and DBP.

‡