Automated Clinical Decision Support

Health information technology is supposed to help improve quality, partly by assisting human decision-making through automated analysis of data and application of algorithms.  A number of these efforts fall under the rubric of clinical decision support and one subset helps to check whether drug prescribing and use is appropriate and alerts clinicians to potential mistakes.   A study reported in Journal of the American Association Network Open looks at use of such a system in a hospital setting.   (JAMA Article)   A clinical decision support system is designed to take real-time information in an EHR and other systems and apply evidence-based guidelines to give care recommendations in a particular situation.  The evidence is inconsistent on their value and downsides have been noticed, including cost, usability, alert fatigue and clinician autonomy loss.  This study was based in one hospital in Italy.  The hospital had an advanced EHR and clinical decision support system, with a focus on medication use.  Patients were randomized into groups on which the clinical decision support system’s real-time alert function either was or wasn’t used.  For the control group, the reminder was available, but the physician was not alerted to it in real-time.  There were about 3200 patients in each group.  About 3 reminders were generated by the system for each patient during a hospitalization.  60% of the reminders were clinical in nature and 30% were for potential drug interactions.  In about 4 out of every 100 patients these reminders led to a change in practice.  Resolution of reminders occurred faster in the intervention group.  Use of the system did not appear to have an effect on mortality or length of stay.  Overall, generating real-time reminders of clinical suggestions appeared to slightly improve rates of resolution of recommendations and time to resolution but not to affect actual patient outcomes.  A large number of the reminders were simply clinically insignificant or irrelevant.  This suggests some algorithmic construction issues.  It also has the potential to lead clinicians to ignore recommendations.  It has become clear that while health IT has the potential to be helpful, it isn’t going to provide dramatic cost savings or quality improvements, as some people have suggested, and it can make clinicians’ work lives both better and worse.