Skip to content Skip to navigation

WHEN THE COMPUTER OFFERS ADVICE

October 1, 2006
by DENNIS P. MORRISON, PhD
| Reprints
Behavioral health stakeholders have mixed feelings about clinical decision support systems

You are probably familiar with the benefits of electronic health records (EHRs)—legible data, ubiquitous access, increased security, reduced long-term costs, etc. One of the most challenging benefits to implement is a clinical decision support system (CDSS). Our staffs need these systems because, as OpenClinical.org points out, “It is now humanly impossible for unaided healthcare professionals to deliver patient care with the efficacy, consistency and safety that the full range of current knowledge could support.”

Before we had EHRs, we couldn't even consider implementing a CDSS. Paper records are static and unidirectional. What is written on a page can only be read on that page (assuming it's legible). On the other hand, EHRs are legible, dynamic, and interactive. EHRs are not electronic recapitulations of paper records. EHRs can and should do things that paper records cannot, including offer clinical decision support.

Helpful “Advice”?

A CDSS uses client data to generate case-specific advice. This reliance on data is a particular challenge for behavioral healthcare. Other medical specialties’ records rely on lab values and other measures, but our records are text-based and do not lend themselves to quantitative analysis. We do capture some quantifiable data, but we could collect more to help advise clinicians about treatment.

Such “advice” must be programmed into EHRs and, to date, few clinical decision support systems have made their way into behavioral health EHRs. One reason might be that treatment “advice” is not universally known or, perhaps more accurately, does not enjoy universal agreement. Keep in mind that the half-life of psychology literature is estimated to be six years, meaning that 50% of the information we have today will be obsolete six years from now. In addition, the National Institute of Mental Health reports that it takes approximately 17 years for information to evolve from research to practice.

Determining who decides what advice to program in a CDSS also raises serious questions. Is this the responsibility of a facility's medical or clinical director? Are there national norms that could be applied? What is the role of IT vendors? Should managed care companies determine the “advice” programmed into a CDSS?

Types of Decision Support

There are six types of clinical decision support:

  • Alerts and reminders

  • Diagnostic assistance

  • Therapy critiquing and planning

  • Prescribing decision support systems

  • Information retrieval

  • Image recognition and interpretation

Alerts and reminders along with diagnostic assistance are the most immediately relevant for the nonmedical behavioral health industry. Some forms of simple alerts already are commercially available. For example, many EHRs alert clinicians to patients about to run out of authorized sessions. Although somewhat administrative, such alerts do assist the clinical process by assuring continuity of care. A more clinically relevant alert would notify a clinician when a client has endorsed thoughts of suicide on an objective assessment.

More complex alerts and reminders might take a diagnosis code and a functional assessment score to suggest one treatment over another, or use an objective assessment score and a therapist rating of impairment to suggest an evidence-based treatment. Prescribing decision support systems alert clinicians to potentially harmful drug interactions and are available in most prescription management products. This type of clinical decision support reminds prescribers of potential problems but does not prevent them from taking action.

Some basic forms of diagnostic assistance are also available. For example, it would not be difficult to provide a reminder that in the event the number of symptoms endorsed was insufficient to meet the criteria for major depression, the clinician might consider a diagnosis of dysphoria or adjustment disorder with depressed mood.

Barriers

Even when a fully functional EHR is in place, a CDSS still can fail. The system might have a poor human interface design, alerts might not be readily recognizable, or alerts might come too often for trivial matters, leading clinicians to ignore them. Some systems do not fit well with clinical work flows and therefore are irrelevant to providers' work.

And a CDSS won't work if clinicians do not want to interact with computers. They might prefer to dictate their notes instead of being “high-paid typists.” Superficially, this argument makes a modicum of economic sense, but it misses the point entirely. If clinicians are going to take advantage of these systems, they have to interact with computers. Otherwise, the alerts, warnings, and reminders will be useless.

Guiding clinicians to provide “preferred treatments” may be viewed as an unwelcome intrusion into their clinical autonomy. As noted earlier, who decides which treatments are “preferred”? Furthermore, once those choices are made, so are the decisions of what not to do. For example, what will happen if a clinician provides a treatment that is not “preferred,” and how will this be tracked in an EHR?

Some barriers to successful CDSS implementation are outside providers’ control. Some state and federal regulations are at odds with the move toward EHRs generally and preclude the use of a CDSS completely. A Medicaid site reviewer told a provider I know that because information was not documented on paper, it did not exist. Therefore, all the transactions billed to Medicaid and documented in the EHR were invalid. This was the reviewer's position even after the provider offered to print all of the electronically documented transactions.

Pages

Topics