# Clinical decision support systems and how critical care clinicians use them

Dr. Weber’s article is a review of the literature and a discussion of the use of clinical decision support systems in the intensive care unit.

## Contents

## Introduction

Introduced in the 1980s, computer-operated clinical decision support systems were created to improve the accuracy of decision-making and more importantly, improve patient outcomes. These systems have been used in the intensive care unit to decide when patient transfer is appropriate; to decide whether to use invasive medical procedures such as mechanical ventilation, pulmonary artery catheters, and renal replacement therapy; to ascertain whether or not to use high risk or expensive medications such as vasopressors or drotrecogin; and to determine disposition of patients at time of hospital discharge.

## Decision making

Three phases of decision making were described. The intelligence phase depends on collection of data and formulation of a question. This typically requires a fair bit of human input. The subsequent design phase explores alternatives by means of a mathematical or statistical prediction technique. Finally, evaluation of alternatives is the choice phase. This phase is well supported by computer systems as it is typically structured.

### Types

With respect to the types of decisions we make, three categories were described. Structured decisions are typically the easiest to make as intelligence, design, and choice are all well-defined. Clinical decision support would be unlikely to be helpful in this scenario as there is little uncertainty. In unstructured decisions all three phases of decision making are unknown with a resultant high degree of uncertainty. Decision support systems may add some, but little additional help in this situation. The majority of decisions in healthcare are semi-structured where at least one phase of decision making is well defined. This is where decision support systems are likely to have the most impact. Critical care providers are making mostly semi-structured or unstructured decisions.

### Mathematical models

In computer-based decision algorithms simulation, mathematical, or statistical models may be used. Simulation models predict events by modeling the expected interactions of the components of the situation over time. Times of occurrence of events and nature of events are determined by use of a statistical sample distribution. Mathematical models focus on probabilities. Statistical models help with high-complexity situations by use of multivariate analyses. Systems using mathematical relationships to make clinical predictions are typically easily interfaced with existing clinical computerized systems that contain documentation of laboratory values, vital signs, and respiratory care.

The author states that some studies have shown that the reliability and validity of the clinical decision support systems were 95 % (+/- 3 %). Prospective studies with large sample sizes (up to 20,000 patients) were performed to analyze the relationship between hospital survival, acuity of illness, intensive care unit length of stay, need for active treatment and diagnosis, acute physiologic issues, age, co-morbidities, and pre-existing functional issues.

## Conclusion

The question not addressed in this article: Does clinical decision support improve prediction of outcomes versus unassisted decision making? This would require design and implementation of a prospective randomized trial. I suspect we’ll be seeing this in the near future.

## References

Reviewed by Jennifer LeTourneau, DO