Clinical Decision Support Systems
Clinical decision support systems (CDSS) represent applications that are human-usable, but use the synthesis of the distinct but related data streams concerning a patient or a population of patients under-the-hood. CDSS can be computerized or non-computerized. Additionally, they can range from paper checklist to a computerized alert triggered if criteria are met. The focus is the use of existing data to improve upon clinical decision-making.
The amount of data surrounding individual patients gathered from modern technology involved in medical care is growing expansively. This is paired with the rapid growth of medical and scientific knowledge. The utilization of various computer-based tools can improve the quality, efficiency, and delivery of health care.
The development of robust CDSS necessitates the understanding of the process of medical decision-making, especially in the setting of probabilistic uncertainty. For example, making a decision to order a test to diagnose acute appendicitis when I am 50% sure it is appendicitis, but 50% unsure as well, is different than ordering a test if I am 90% sure the diagnosis is appendicitis. Understanding the decision analysis techniques such as the PROactive approach formalizes the process of decision-making (Hunink et al., 2001). This method spans defining the problem to be addressed and identifying the objectives to be addressed by a decision support tool, followed by the modeling of decision-making consequences with estimates on probabilities in the structure of decision trees. Comprehensive modeling of the decision making process encountered in medical practice forms the basis of strong tools to assist physicians.
Example of clinical decision tree on whether to vaccinate for Lyme Disease
LD: Lyme Disease
Image: Meltzer, M. (1999). The Cost Effectiveness of Vaccinating against Lyme Disease. Emerging Infectious Diseases, 5(3), 321–328. doi:10.3201/eid0503.990302
My education in decision support systems has lead to a detailed proposal for a CDSS tool assisting in the detection of acute pulmonary embolism (PE) in post-surgical patients. The tool architects the unification of machine-learned variables (vital signs, ECG waveforms, free-text in documentation) with validated scoring systems for PE diagnosis and clinical practice guidelines. A traditional knowledge base is supplemented with non-knowledge base systems such as unsupervised learning systems and artificial neural networks. These systems can generate cluster characteristics of a patient with a PE along with classifiers, respectively. These results can be tested experimentally and findings can be incorporated into the existing knowledge base. This makes the acute PE detection tool in post-surgical patients a continually evolving CDSS.
Hunink, M., Glasziou, P., Siegel, J., Weeks, J., Pliskin, J., Elstein, A., & Weinstein, M. (2001). Decision Making in Health and Medicine Integrating Evidence and Values. Cambridge University Press.
A Novel Acute Pulmonary Embolism Clinical Decision Support Tool for Post-Surgical Patients