Sunday, February 12, 2006

Evidence-based programming

I've been playing around with the training software at home; I think this will work.

I've been thinking about what works in EMR programs. There seems to be a lot of research done in computer-aided clinical decision making. Some programs don't work, for example, a large study showing that the EMR did not help with angina or asthma, found here (for the trial), and here (for an explanation of why it didn't work--basically it was intrusive, did not fit into the flow of work, and the so physicians didn't like it and didn't use it).

It seems to me that the EMR has to present the right evidence at the right time, in a way that is not too intrusive. This is challenging: patients present with multiple problems, and you don't want too much information popping up when you don't need it, or it is going to consistently be bypassed.

I think what will work is to start slowly, by giving us things we are interested in first. In Ontario, we have incentives to provide mammograms, paps, flu shots and children's vaccinations, and soon we will start on Fecal Occult Blood. The EMR will help us to maintain lists of patients that are eligible for these interventions. There is evidence that automatic prompts improve the provision of cancer screening and vaccinations (see Garg et al). The EMR can probably automatically generate a prompt if a patient overdue for a preventive service comes in.

As well, I will start doing electronic audits for these preventive services, this fall. I will assign this to one of my staff members, and we will generate recall letters for patients who are overdue (because recall letters work). I have been talking about this with a colleague from Nova Scotia, who will be using the same software, and who has been thinking along the same lines. We can start using some of the provincial databases that have secure access (for example, Cytobase for pap smears) to make sure our registers are up to date.

I think that for any EMR to work, it has to be used, and used consistently. It has to fit well into the workflow. Prompts and clinical decision support will likely prove their value, but they must be carefully integrated into the system. A good first step is preventive services, because I have incentives to do them, I can maintain registers, there is already evidence of successful implementation in the literature, and I can modify work processes in my office (assigning a staff member to do audits, generating recall letters, having alerts to remind patients) relatively easily.

I have been notified by my hospital that I can now have on-line access to their electronic system, which will help me with patients seen in Emergency, and for those recently discharged from the hospital. Interestingly, there was a recent research article in CMAJ showing that giving family physicians on-line access to Emerg patient data made no difference. I wonder if the reason for that was that they picked very busy physicians (over 4,000 patients in each practice), and also picked people who were not very familiar with computers. The family physicians may not have used the software, and so it did not work; if systems are not well integrated into daily practice, they don't work. You have to think about effectiveness; maybe we should call this evidence-based programming.


No comments: