ED intervention to decrease opioids

[Will do these interventions as 2 blogs (too long for one); next one on Monday]

Informing emergency dept (ED) clinicians about their misperceptions of their own opioid prescribing led to decreased prescriptions (see opioid ER MD behav change Academermed2018 in dropbox, or doi: 10.1111/acem.13400).

Details:

--109 clinicians in 4 EDs (emergency departments) in Massachusetts, all part of the same system and senior leadership, were enrolled in the study. Providers not included were those not seeing adult patients and those who were no longer in the ED for the intervention (eg residents rotating through for a month)

--eligible clinicians had their prior 12 months of opioid prescriptions reviewed

    --a total of 75,203 prescriptions (20.1% were for opioids) in 119,428 discharged patients

--65 attending physicians/36 resident physicians/8 advanced practice providers (APPs), 33% female, mean of 1124 patients per provider giving 98 opioid prescriptions (10.5% of patients), for median of 21.0 opioid prescriptions per 100 total prescriptions

--median time since professional school graduation (excluding residents)=13 years, though wide range of 3-34 years and interquartile range of 14 yrs

--51 were randomized to the intervention: clinicians stated what they thought was their frequency of opiate prescriptions in comparison to their peers. this was then compared to their actual opiate prescribing pattern for the preceding 12 months compared to their peers

    --incorrect estimation of opioid prescriptions written was defined as: +/- one decile from their actual decile of all clinicians

    --65% of these clinicians underestimated their opioid prescribing (73% of attendings and APPs, by a median of 2 deciles; and 27% of residents); only 5 providers (3 residents and 2 attendings) overestimated it

--residents and APPs used their own DEA numbers, so the initial and subsequent evaluations were specific to a given provider (ie, were not included in the data of the attending, though the attending might have influenced the decision)

--primary outcome: change in proportion of patients discharged with an opioid prescription per each individual provider

--secondary outcome: change in % of prescriptions written for opioids; change in total morphine milligram equivalents (MME) for the opioid prescriptions written

Results:

--overall change in opiate prescribing by individual providers, including the control group, was decreased by 3.5 prescriptions/100 patients at 6 months and 4.3 at 12 months

--at 6 months:

    --in clinicians who underestimated their opiate prescribing, there was a further 2.1/100 patient decrease in opiate prescribing

--at 12 months:

    --in clinicians who underestimated their opiate prescribing, there was a further 2.2/100 patient decrease in opiate prescribing​

--change in MME per total patients discharged with an opiate: significantly lower in the underestimators at 6-months, approximate values of -7 MME, and at 12-months of -7.5 MME vs controls (-4.5MME and -5 MME) vs non-underestimators (-3 MME and -4 MME) [they lumped together the few overestimators with those on target]

    

Commentary:

--there was no real difference between the control group and the intervention group of clinicians who did not underestimate their opiate prescriptions. in fact there was a trend to a higher % of patients discharged with opiates in the control group.  this brings up the potential role of cognitive dissonance: those confronted with their underestimating prescriptions were the most likely to act on it (as has been shown in many other types of studies); those shown that they were on-target in their perceptions were more likely to feel that what they were doing was appropriate and change the least; and those unaware of how they stacked up (the control group) had more uncertainty about how they fit in and had a trend (nonsignficant) to write a bit more opiate prescriptions

--this last point brings up this perhaps intuitive notion: if one is trying to change anyone's behavior, it is better to give specific feedback to them ("you actually only gave flu shots to xx% of your patients", or "only tested xx% for their lipids" etc) vs a broadcast message "we should be giving flu shots or testing lipids on all patients in category yy"). this deals with issues of cognitive dissonance, as above, though should be done in a very supportive and private way ("what can we do as a health center to help you improve your rates of xxx?")

--it was notable that the overall group decreased their opioid prescriptions over the study year, likely reflecting implementing new clinical practice guidelines from Massachusetts as well as the general increasing awareness of the issue and consciousness of the need to decrease opiate prescribing

--this brings to mind the study of older patients going to the ED and seeing high-opiate prescribers (vs lower) for similar clinical indications and finding that there was a 30% increase in these patients being on long-term opiates (see http://gmodestmedblogs.blogspot.com/2017/02/opiate-prescribing-in-elderly-and.html  )​, reinforcing the importance to minimize opiate prescriptions

--one concern about this above type of study is the lack of specific data, just a global assessment. although we clinicians are much more aware of the risks of opiate prescribing (and data show we are doing less), there is no gold standard here. are there certain patients, conditions, differences in physiology/genetics/receptors that identify some individuals who really need opiates to function?? are there even some patients who need particularly high doses of opiates (never really studied well)? if so, maybe the clinicians prescribing fewer opiates are really undertreating patients? and maybe even some of the higher prescribers are underprescribing? same thing applies to some old studies of showing gross statistics of chest xray ordering. just displaying total numbers of which clinicians order more xrays than their peers does lead to fewer xrays ordered. peer pressure works. but are these decisions based on peer pressure or on the real clinical needs of the individual patients? or, insurance companies imposing obstacles to ordering CTs or MRIs  (read: prior approvals through both tortuous and torturous processes) do decrease these exams. but is that clinically appropriate??

--a positive aspect of the current ED study is that opiate prescribing information was supplied privately and only to the specific clinician. all other clinicians in the display were anonymized. so, the issue of direct peer-pressure is minimized

--it would have been useful to have patient outcome data: was there a change in patients satisfied with their interaction with clinicians? did more patients come back to this or another ED or their primary care clinician to get different meds for pain relief or perhaps opiates?

--but, my feeling, as it has evolved over time, is that the risks of opiates is so high that we should do whatever we can to decrease prescribing opiates as safely and patient-centeredly as we can. which may mean undertreating some pain (though the recourse being that the patient comes back and gets another evaluation of the pain).

so,

--this study does reinforce the differences in prescribing styles of clinicians, as well as our own misperceptions of what we really think we are doing. This data-driven display of our objective opiate-prescribing is a potentially useful tool to decrease these prescriptions. 

--Another, more time-intensive but perhaps more targeted/clinically useful approach may be to have case conferences on a regular basis of clinicians who prescribe opiates, to review individual cases and get feedback on perhaps other, non-pharmacologic or pharmacologic approaches. and in a suportive manner.