Computer-based execution of clinical guidelines: A review

Abstract

Purpose

Clinical guidelines are useful tools to standardize and improve health care. The automation of the guideline execution process is a basic step towards its widespread use in medical centres. This paper presents an analysis and a comparison of eight systems that allow the enactment of clinical guidelines in a (semi) automatic fashion.

Methods

This paper presents a review of the literature (2000–2007) collected from medical databases as well as international conferences in the medical informatics area.

Results

Eight systems containing a guideline execution engine were selected. The language used to represent the guidelines as well as the architecture of these systems were compared. Different aspects have been assessed for each system, such as the integration with external elements or the coordination mechanisms used in the execution of clinical guidelines. Security and terminology issues complement the above study.

Conclusions

Although these systems could be beneficial for clinicians and patients, it is an ongoing research area, and they are not yet fully implemented and integrated into existing careflow management systems and hence used in daily practice in health care institutions.

Introduction

Nowadays, the execution of clinical guidelines (CGs) is one of the most interesting topics of study within clinical informatics. CGs contain a set of directions or principles to assist the health care practitioner with patient care decisions about appropriate diagnostic, therapeutic, or other clinical procedures for specific clinical circumstances [1]. They are intended to ensure consistent high quality clinical practice and provide some benefits to both patients and health care managers [2], [3], [4], [5], [6], [7], [8], such as:

• For health care professionals, the use of CGs can improve the quality of clinical decisions and activities and, in consequence, the patient outcomes are also improved (e.g., a clinician will not forget an important aspect to be checked before ordering a certain treatment).

• CGs facilitate reuse of knowledge, because a guideline can be adapted, tailored and applied to different clinical situations.

• Guidelines support rapid dissemination of updates and changes. CGs promote interventions of proved benefits and discourage those that are ineffective.

• CGs help doctors to use the clinical knowledge about the patient at the appropriate point of his care.

• Guideline authors are encouraged to employ rigorous formal techniques, which help to ensure syntactic, logical and medical validity of CGs.

Although several international organisations create and maintain repositories¹ with guidelines in different domains (e.g., oncology, cardiology, paediatrics), they are not being widely used in daily practice. Several factors limiting or restricting complete physicians adherence to clinical guidelines were identified in [9]. Such factors include lack of awareness with the guideline’s existence, lack of agreement, lack of physician self-efficacy, lack of outcome expectancy, or the inherent difficulty to change habits in daily behaviour. These factors could be tackled (in most cases) with an automation and computerisation of the daily management of both clinical guidelines and patient data [10].

Several steps must be considered in the management of clinical guidelines: representation, acquisition, verification and execution. The first three tasks concern the authors of the guideline, whereas the later is related to practitioners. Briefly, these steps can be described as follows:

• Choice of a representation language. A CG contains several elements to be modelled, such as actions, required patient data, decisions to be made, constraints between tasks, temporal constraints in a global plan, etc. Different researchers have defined formal languages to model computer-interpretable clinical guidelines, such as PROforma, EON, GLIF, GUIDE or Asbru [11], [12], [13], [14].

• Acquisition of CGs. Medical guidelines are based on the evidence collected from clinical trials and existing literature [15], [16]. Some authors are also currently working in the semi-automatic construction of guidelines, by applying Machine Learning techniques from a corpus of clinical data collected in a medical centre [17] or directly from textual documents [18].

• Verification of CGs. Verification includes two aspects: is a medical guideline well formed?, and, which of these two available medical guidelines is the best? The first question seeks to verify the formal correctness of the guideline [7]. The second question is more difficult to answer since it is necessary to quantify how appropriate is a medical guideline. To tackle this problem, some authors proposed an evaluation procedure called AGREE which calculates a set of parameters for a given medical guideline to evaluate its quality [19]. In addition a methodology to facilitate the whole development and evaluation of clinical guidelines can be found in [20].

• Execution aspects. As mentioned above, a medical guideline contains a great amount of information to be considered (decisions to be made, constraints between tasks, temporal restrictions). All these data have to be collected and monitored when enacting the guideline.

Concretely, while representation, acquisition and verification stages are currently active research areas [21], [22], [23], the execution of guidelines is a less developed field. This paper is focused on the analysis of systems that allow the automatic (or semi-automatic) execution of guidelines. Several systems are currently being developed for this purpose. In the next section we describe the basic characteristics to be analysed in a guideline execution engine (e.g., coordination issues, security techniques, use of standard medical vocabularies). Section 3 provides a table that summarizes the main information of each system and comments the results obtained for each of the analysed attributes. Section 4 states some comments derived from the comparison of the systems. Finally, the last section briefly summarises a list of concluding remarks.