Purpose: The development of effective computerized Clinical Decision Support Systems (CDSS) requires an understanding of physicians' responses to such systems. This study evaluates physicians' tendency to adjust clinical decisions, based on information from a decision-support system, and estimates the effects of different factors on these responses.
Methods: The Computerized Community Cholesterol Control (4C) project is a large-scale clinical decision support system. It is active since 2001 and focuses on Dyslipidemia management, based on the National Cholesterol Education Program (NCEP-III) guidelines. Periodic advisory letters are mailed to primary medical teams about patients with Atherosclerosis diagnosis, indicating the need for lipid screening, lipid-lowering drugs or metabolic consultation. The advisory letters are based on a wide range of data, including demographics, diagnosis, hospitalizations, laboratory tests and pharmacotherapy. We analyzed data of 14,018 patients and 297 physicians from 112 demographically matched clinics, 56 in an intervention group, and 56 in a control group. Physicians' compliance was measured by the probability of physicians to refer patients to lipid screening within a time period of 3 weeks to 3 months after an advisory letter was generated.We used a logistic regression to predict the physicians' compliance with the decision-support tool as it is affected by clinics', physicians' and patients' characteristics.
Results: The decision-support tool, implemented by the advisory letters, led to greater compliance with clinical guidelines. Compliance was significantly higher in the intervention group than in the control group (p<0.05). Physicians' compliance in the intervention group was higher, the more patients a physician treated, and the more letters he or she received, yet compliance was lower when the physician treated relatively older population. Physicians' compliance in the control group was unaffected by these measures.
Conclusions: Physicians' adherence to computerized Clinical Decision Support System depends on a combination of properties, of the decision-support tool, the clinic's population, the physicians' work conditions and the individual patient. The physicians tended to be more compliant with the decision-support tool when they experienced higher workload. Thus, the decision-support tool assisted physicians in providing adequate treatment even under increasing workload conditions. Systematic response patterns, similar to those characterizing operators in other domains, are also evident in physicians' responses, and can be used to predict physicians' responses as a function of system design and other relevant factors.