Meds often stopped after ICU stays: deaths result? (JAMA) - PulmCCM
Dec 262011

Medications are often stopped during transfers of care. Bell et al analyzed administrative-level data for almost 400,000 hospitalizations in Ontario, Canada, as well as 90-day follow-up outpatient prescription data. They conclude that medications were likely to be discontinued after discharge from the hospital or ICU, “potentially unintentionally.” Five medication classes were analyzed (inhalers, anticoagulant/antiplatelets, acid-suppressors, thyroxine replacement, and statins). The “discontinuation” is everywhere referred to or implied as being done by the discharging physicians — and probably was — but this is never demonstrated; only whether a patient refilled a Rx within 90 days was measured.

Patients who had medications “potentially unintentionally” discontinued had a higher rate of the composite of “death, ED visit, or emergent hospitalization” at one year — with a just barely significant odds ratio of 1.07 if statins were stopped and 1.1 if antithrombotics were stopped.

Not surprisingly, pro-bleeding medications were particularly likely to be stopped (in about 20-23% of patients after a hospitalization/ICU admission). Inhalers were stopped only about 5% of the time.

There was no chart-level review to determine whether medication discontinuation was intentional/rational — which, with 400,000 records, surely would have been prohibitive. (They did exclude patients with clear reasons for discontinuation in their diagnoses in the database — bleeding ulcers on warfarin, for example.) They attempted to model patient nonadherence factors and intensity of outpatient physician attention after discharge — which sounds pretty tricky.

These were all adherent patients who had taken the drugs for a year straight prior to hospitalization. Unfortunately, we don’t know what people were readmitted with. Did those whose statins or antithrombotics were stopped have heart attacks or strokes — or GI bleeds?

I’m sure their methodology was impressive, since it’s in JAMA. Medication reconciliation at discharge is surely beneficial, and a source of major errors, and we need continual vigilance & systems improvement. It’s certainly possible that inappropriate discontinuation of meds led to worse outcomes. But there are a few leaps here, and it’s hard to know what to conclude from a 10% relative difference (at most) in a composite endpoint in a complex, ill patient population with conceivably unmeasured confounders on the outcome.

On that note, it’s intriguing that their finding was stronger in those who died within 90 days (they were excluded from the final analysis to avoid bias). You could argue, that makes sense — they died because their meds were stopped, or care was worse, etc. But what if in fact the confounder here is patient frailty, polypharmacy, and/or other unmeasurable (in a database) markers of sicker people? In other words, maybe the docs were just more likely to stop a medicine on someone who was going to be readmitted or die soon inevitably — and the med being prescribed or not was immaterial. (For example, stopping warfarin in newly frail patients for “fall risk” is common in the U.S.)  The fact that these patients’ primary physicians didn’t restart the medicines within 90 days adds weight to this possibility, I think. Anticoagulant drugs are rarely stopped accidentally (I hope); there’s usually a reason. Whether or not it’s a good one is a separate question, but that can’t be answered from this data.

Bell CM et al. Association of ICU or Hospital Admission With Unintentional Discontinuation of Medications for Chronic Diseases. JAMA 2011;306:840-847. FREE FULL TEXT


Liked this post? Get a weekly email update (no spam, ever), and explore our library of clinical guidelines, practice updatesreview articles. and board review questions.

PulmCCM is an independent publication, not affiliated with or endorsed by any other organization, society or journal referenced on the website.

Authors: contribute your work to PulmCCM Journal or in a guest blog post.

  2 Responses to “Meds often stopped after ICU stays: deaths result? (JAMA)”

  1. As a practicing intensivist/hospitalist I have to think there is a potential positive aspect to this. I frequently see people admitted w/ 10+ meds, many of which have unanticipated and unmonitored potential drug interactions. Accordingly I frequently purposely delete some of the meds that I think are marginally effective and potentially catastrophic. Often I see patients on warfarin or antiplatelet therapy for no clear reason. I see 90+ y/o patients w/ demonstrable muscle weakness who remain on statins w/ no history of coronary or cerebrovascular disease. Should I not be thinning their polypharmacy?

  2. I am a big fan of this kind of observational, real-world-based research. But this article also shows its limitations, namely in trying to isolate an effect in a complex system. A hospitalization/ICU stay, followed by a changing of some medicines, followed by a yes-or-no event in that composite outcome months later — are incredibly complex and “multifactorial,” to use that tired word. So many events happen in space, time, in the patients’ lives and bodies, in the care systems, and in the doctors’ brains during that period. Some effects are intentional, some “potentially unintentional,” but surely they are diverse and unpredictable.

    They authors told one thread of that story here, but I suspect there are many others that could be told with the same data. E.G., I can easily imagine someone else parsing related data to demonstrate the prevention of GI and brain bleeds by stopping anticoagulation. Or multiple other positive effects of rationally thinning polypharmacy.

    If this effect size were larger I would be less skeptical. But a less than 10% relative change in a composite endpoint, in an observational study (no matter how well-devised)? Just for comparison, a just-published study shows the spouses of people with familial venous thromboembolism have a statistically significant 1.07 odds ratio of having VTE. (Circulation: Is this proof that VTE is contagious? Of course not — the authors use this number as a control, to contrast it with the affected spouses’ odds ratio of 53.

    That’s just one example from this week. There are many others, and leading epidemiology experts have long cautioned against making these kinds of conclusions from small effect sizes (Ioaniddis PLoS 2005:

    I know we can’t have it both ways (reductionistic rat studies = high confidence in conclusion for clinically irrelevant finding; observational epidemiologic/database studies = low confidence in a conclusion that’s highly relevant), but JAMA should be one place we can expect not to be “sold” findings like this, IMHO.

Leave a Comment