Introduction
The increasing prevalence of chronic diseases, such as heart failure, as the population ages is expected to result in a severe shortage in healthcare resources including nurses, physicians, and hospital beds [1], [2], [3]. Telemonitoring is a promising tool that can potentially alleviate some of the burden on the healthcare system by empowering patients to care for themselves and enabling more efficient clinical care, such as through automated alerts at the earliest sign of deteriorating patient health.
Currently, computerized clinical decision support systems exist for heart failure diagnosis and treatment plans. The rule sets for these systems can be based on clinical practice guidelines [4], [5], [6]. Although methods to transform clinical practice guidelines for use in decision support have been developed [7], [8], no guidelines presently exist that are comprehensive enough for direct use in the relatively new area of automated heart failure patient decision support and alerting systems (e.g., to create patient instructions for each possible combination of symptoms, blood pressure, and weight readings). Furthermore, even if such guidelines did exist, clinical input during the development of computerized clinical practice guidelines would be important to bridge the “gap between the guideline text and clinical practice” [9]. That is, rule sets need to take into consideration each healthcare institution's own workflow and policies, as well as the individuality of the patients. The clinical input could be implemented through various user-centered design processes, such as participatory design, usability testing, or contextual design [10], [11], [12].
An expert system is a “knowledge-intensive program that solves a problem by capturing the expertise of a human in limited domains of knowledge and experience” [13]. Expert systems have been used in a variety of fields, including medicine, space, and business [14], [15], [16], [17], [18]. For rule-based expert systems, the knowledge from experts is translated by a knowledge engineer into a set of rules [19].
The purpose of this work was to develop a rule-based expert system for a heart failure mobile phone-based telemonitoring system. The expert system would be used to automatically generate appropriate alerts for clinicians and patients, as well as to generate suitable patient instructions. This expert system would be fundamentally different from existing clinical decision support tools available to healthcare providers for diagnosis and management because it would function independently of the clinician with automated alerts being sent directly to the patient and to the clinician. This article describes the process used to create the heart failure rule set, the resultant rule set, and the lessons learned. To determine the efficacy of the rule set, it was incorporated into a heart failure telemonitoring system that was evaluated in a randomized controlled trial. The effects of the rule set on the trial results are also summarized below. Details of the trial results have been published elsewhere [20], [21].