Sunday, June 22, 2008

Research and Quality Improvement

I have been interested in research for several years, and have a couple of projects under my belt (Using a PDA for the diagnosis of angina; Cognitive Behavioural Therapy in primary care; effect of incentives on preventive services). I think EMRs have enormous value for research that is relevant to family medicine and useful for patient care; beyond that, the data in EMRs can certainly contribute information to run our health care system more effectively and efficiently.

The issue will be how to collect data from our practices; you want to fully protect the privacy of both patients and physicians, yet ensure that you still collect enough useful data for analysis.

I am a co-investigator in two studies that have recently received funding:
  1. Canadian Primary Care Sentinel Surveillance Network (CPCSSN), and
  2. Development of an Electronic Medical Record primary care research database, at ICES
These represent two different way of collecting EMR data for research. The CPCSSN sites will collect data on some chronic conditions locally from many different EMR systems in Canada; the data will not contain patient identification. All this data will then be amalgamated at the national site.

The benefit is that you can collect data from several different EMR systems (as we have a tower of Babel of EMR systems in family medicine). The drawback is that you depend on local investigators to try to collect data from their EMRs, and that you are pretty limited in terms of what you collect. You have to say what you collect at the beginning; if you forgot or just couldn't get something important, you don't have it in your database.

For the ICES study, we will explore anonymization software; this will allow, in the future, collection of the entire EMR database while protecting privacy. All names and identifiers are scrambled; after this is done, no one has access to the scrambled data, except for an ICES analyst. Researchers only have access to data after it is analyzed (once they obtain permission), and only for larger groups of patients, never for a single record. In other words, what you get is things like x% of patients in this geographic area have diabetes, y% were prescribed a cholesterol drug, and z% have a cholesterol level that is at goal.

The benefit is that you have a fairly complete description of care; if the database is mirrored (sent automatically to a second server), data collection is on-going and is pretty much live. The drawback is that the anonymizers will have to be programmed for each EMR software, since they all work differently. You will only get data from a select group of physicians, and there will have to be some extrapolation and assumptions if you say that the data represents an entire population.

Because the practices in my group are run remotely out of a single server, we can collect data without having to personally go to each office and access the server. In fact, the amount of bothering physicians, staff or patients is pretty much zero. Ideally, that is the way it should be--we're busy enough during the transition without imposing more work related to research. We will have posters informing patients, and giving them the option of opting out.

I do not know which way is best, and I think these two projects will test both approaches. We will start small, and go slowly and carefully; it may well be that we need a variety of approaches for data collection.

It is very interesting to me that some of the ground work that makes this possible started out of our group Quality improvement initiatives. We wanted to use our common EMR to improve our preventive services, and our diabetes care. Out of that came the realization that data could be collected remotely, and we also learned how to do it.

Research is difficult and time consuming. You also have to go through a Research Ethics Board, and there are lots of forms to fill out. I am not saying this is a bad thing, although sometimes it is very bureaucratic and it doesn't always make sense. Quality Improvement does not require all this overhead; in fact it works better when you have a goal, try small things, see if they work, then fix problems and keep going. This is called a Plan-Do-Study-Act cycle (PDSA); much of what I learned about collecting data in my group's EMR was through that process. For Research, you have to plan thoroughly, apply for your grant, hope it gets accepted, then implement and hope it works. This is not a great way to start something with as many unknowns as EMR data collection. We probably need to re-think the interface between Research and QI.

As for me, I think I probably have something to contribute to this nascent EMR research enterprise. I have now finished the coursework for my Masters, and am writing my thesis (on the Effect of EMRs on preventive services in a Pay-for-Performance environment). I have asked my University department for a second day of research, and will take Wednesdays for this, as of September. I'll practice three days a week, and devote two days (and some evenings and weekends) to research. I have asked for permission to join ICES; we'll see how things go.



Ian Furst said...

good luck with ICES. A co-author of mine and statistician joined back in 1999 and he's a senior scientist there now. Before I started blogging I went to U or T and UW about QI research with them (esp with wait times). Both were interested and even offered an assoc prof position but the grant/research/publish timeline is long and cumbersome. Finally I decided to dedicate my time to improvin my own practice and publish the findings and methods. It's not nearly as rigorous but I think the impact is just as powerful. I think more people in the industry read the blog every day than would ever see an article published in a QI journal. Don't underestimate the power of what you're doing for the community right now by chronicaling your journey with EMR. BTW even google asks me if I want to share my data in an anonymous manner - how do you obtain consent from the patient?

Michelle Greiver said...

Thanks for your comments, Ian.

There are two things here: generation of new knowledge (research), and diffusion of knowledge.

Research is (or should be) rigorous; the process itself, as well as peer review help to ensure this. However, the diffusion of research findings is sometimes less than ideal, especially if published in a less well known journal.

Rogers alludes to the effect of the Internet in the latest edition of "Diffusion of Innovations"; the next edition will no doubt have more about blogs and social networking sites such as Facebook.

The availability and speed of these new diffusion channels does not mean that there is less need for knowledge generation, on the contrary. The flip side is that the diffusion speed of nonsense is greatly increased as well.

As far as patient consent, the rules are the same as for paper chart audits: if you collect data in an anonymized manner, individual patient consent is not needed. None of the large observational studies using chart data require individual patient consent, because that makes those studies unfeasible. REB reviews your audit process to make sure that it is consistent with good privacy practices. Consent is always needed for interventional studies, such as randomized controlled trials, or for studies where a patient is personally contacted to collect data.

Yes, the research process is very difficult (it is one of the most difficult things you can do in medicine); it is also one of the best possible ways of making a contribution to the common good of our society.