- Kenny Lin, MD
The October 1st issue features the third article in AFP's new series "Implementing Effective Health Care Reviews," a summary of the Agency for Healthcare Research and Quality's comparative effectiveness report on treatments for gastroesophageal reflux disease. Notably, the report found no differences in efficacy between proton pump inhibitors; better symptom relief from continuous daily compared with on-demand dosing; and limited data on endoscopic treatments. What are the chances that results from this and other high-quality comparative effectiveness studies will quickly change your practice? Not very good, unfortunately. As I wrote in an editorial that introduced the series:
To date, the track record of translating comparative effectiveness research findings into clinical practice has been mixed, at best. For example, several years after a landmark randomized controlled trial demonstrated the superiority of thiazide diuretics compared with other first-line medications for hypertension, prescribing of thiazide diuretics had increased only modestly. An evaluation of diabetes practice guidelines produced after the publication of an Effective Health Care review of oral treatments found numerous inconsistencies between guideline recommendations and evidence-based conclusions. Despite extensive evidence that initial coronary stenting provides no advantages over optimal medical therapy for stable coronary artery disease, more than one-half of patients who undergo stenting in the United States have not had a prior trial of medical therapy.
In the October issue of Health Affairs, Justin Timbie and colleagues propose five reasons that scientific evidence is slow to change how physicians practice:
1) Misalignment of financial incentives - e.g., fee-for-service payment systems tend to reward invasive therapies, such as surgery for back pain, that may be no better than conservative management.
2) Ambiguity of results - "Without consensus on evidentiary standards prior to the release of comparative effectiveness results, ambiguous results become fuel for competing interpretations, making it difficult for providers, insurers, and policy makers to act on the evidence."
3) Cognitive biases in interpreting new information - e.g., a tendency to reject evidence that contradicts previous strongly held beliefs, such as the superiority of atypical to conventional antipsychotics.
4) Failure to address the needs of end users - e.g., designing a study to compare the benefits of two therapeutic strategies, but not the harms.
5) Limited use of decision support - e.g., poorly designed electronic or paper patient decision aids that do not fit into the workflow of primary care practices.
Do these reasons sound about right to you? How do you think these obstacles could be overcome in order for front-line family physicians to rapidly incorporate the best scientific evidence into their practices?
Wednesday, October 10, 2012
Subscribe to:
Post Comments (Atom)
The 5 reasons proposed DO sound "about right". I think it can be simplified: i) It is all too often "about the money", rather than about what is truly best for the patient (MANY reasons why this is so - but much in medicine unfortunately has become about business and profit motive ....); and ii) it is often in how "the answers" are phrased.
ReplyDeleteTo take the example provided by Dr. Jerome Hoffman in a talk about NNT (Number-Needed-to-Treat) - Say a study shows that a medication (or intervention) reduces mortality in a group of patients by 20% from 2% to 1.6% over a 5-year period. On the other hand - say that side effects from that medication increase from 1% to 3%:
i) IF results of this study are reported as, "a 20% reduction in mortality - with just a 3% incidence of side effects" - it would sound like a great study and a worthwhile pill/intervention.
ii) But IF results of this study are reported as a reduction in mortality from 2% down to 1.6% - with a 200% percent increase in side effects - a totally different impression might be gotten.
iii) IF the patient was told instead of a 20% reduction in mortality - that 250 people would have to take this medication for 5 years in order for ONE patient to benefit - but that the other 249 people would get absolutely NO BENEFIT (yet they'd still have to pay for the medication, be monitored for it, and might get side effects from it) - then it might be an entirely different story.
iv) IF the benefit was GREAT (your life will be saved) - and the cost of medication, follow-up monitoring, and severity of side effects were small - then to some people it may be worth the reality that 249 out of every 250 people who take the pill will get absolutely NO BENEFIT from that pill. But to others - they probably would not want to take the medication if those were their chances to benefit ....
Obviously the above example is simplified - but it is NOT far from the truth on many matters of benefit from screening, and commonly used medications and interventions. The reality is that all-too-many studies in the literature report benefit from medication/interventions in RELATIVE terms (ie, 20% reduction in mortality) - yet report adverse effects in ABSOLUTE terms (3% chance of side effects) - and fail to present the patient with NNT (and NNH = Number-Needed-to-Harm) data that is needed to make a truly informed decison about whether or not to be treated. One has to ask why this may be so. Perhaps my first premise in my first paragraph has something to do with those who sponsored the study ...
Ken Grauer, MD (ekgpress@mac.com)