Letters

Author Affiliations: Department of Biomedical and Health Information Sciences, University of Illinois at Chicago, Chicago (Abraham); Department of Family Medicine, College of Medicine, University of Illinois at Chicago, Chicago (Kannampallil); Department of Internal Medicine, University of Texas Health Science Center, Houston (Almoosa). Corresponding Author: Joanna Abraham, PhD, Department of Biomedical and Health Information Sciences, University of Illinois at Chicago, 1919 W Taylor St, Chicago, IL 60618 ([email protected]). Conflict of Interest Disclosures: None reported. 1. Devlin MK, Kozij NK, Kiss A, Richardson L, Wong BM. Morning handover of on-call issues: opportunities for improvement. JAMA Intern Med. 2014;174(9): 1479-1485. 2. Abraham J, Kannampallil T, Patel VL. A systematic review of the literature on the evaluation of handoff tools: implications for research and practice. J Am Med Inform Assoc. 2014;21(1):154-162. 3. Abraham J, Kannampallil TG, Almoosa KF, Patel B, Patel VL. Comparative evaluation of the content and structure of communication using two handoff tools: implications for patient safety. J Crit Care. 2014;29(2):311.e1-311.e7. 4. Abraham J, Kannampallil TG, Patel VL. Bridging gaps in handoffs: a continuity of care based approach. J Biomed Inform. 2012;45(2):240-254. 5. Jones LK, Franklin A, Kannampallil TG, Buchman TG. Effects of structuring clinical rounds on communication and efficiency. In: Patel VL, Kaufman DR, Cohen T, eds. Cognitive Informatics in Health and Biomedicine: Case Studies on Critical Care, Complexity and Errors. London, England: Springer; 2014.

In Reply Abraham and colleagues raise important points regarding the challenge of developing metrics to evaluate the clinical consequences of incomplete handover communication,1 based on an impressive body of research focused on handover communication in the intensive care unit (ICU) setting.2,3 We developed our measurement approach with a specific goal to quantitatively measure how often on-call residents fail to handover on-call issues to the daytime physician team. This approach involved direct observation of morning handover communication to determine the proportion of clinically important issues (identified via real-time medical chart review) handed over by the on-call resident. Unfortunately, our approach did not allow us to determine whether these omissions actually resulted in patient harm or delays in care provision. They assert that our estimation of the clinical importance of handover omissions may have been overstated by citing examples from their research of handover in the ICUs.2,3 The trouble is that the consequence of omitting an on-call issue at morning handover is very much dependent on context. The ICU is a highly centralized and monitored environment where information continuity (eg, ICU flow sheets, one-on-one nursing patient assignments) makes patient information and clinical status more readily available to members of the health care team. Also, structured interprofessional team rounds are the norm. As such, the omission of on-call issues at morning handover in the ICU may either occur less frequently, or even when they do occur, may not delay patient care as often because other redundancies in the system make up for lapses in physician-to-physician communication. However, in the general internal medicine (GIM) ward setting, as typified by our study findings, the timing and format of morning handover, as well as the environment in which handover took place, was highly variable. Resident documentation of on-call issues was the exception, not the rule. Similarly, residents rarely updated the electronic sign-out tool to highlight new issues arising overnight. Other structural difjamainternalmedicine.com

ferences limit information sharing and continuity, which Abraham and colleagues acknowledge when they suggested that our findings might have resulted from the fragmented information systems seen in our GIM environment. However, this very fragmentation of patient information means that on-call issues omitted at morning handover are more likely to slip through the cracks, making consistent and complete morning handover of on-call issues even more critical. While some of the methodological concerns raised may contribute to an overestimation of the prevalence and magnitude of the problem, we believe that the “true” proportion of omissions is still too high. For some of these issues, important delays in following up on on-call issues remain a real concern, especially in the GIM setting. Therefore, in addition to improving actual physician communication practices at morning handover, Abraham and colleagues remind us that we should also learn from other care settings such as the ICU and determine the extent to which changes to other parts of our system both before and after the handover event (eg, pre-handover information organization, dedicated team-based handover rounds, electronic documentation) can result in better continuity of care, and ultimately, better outcomes for our patients. Natalie K. Kozij, MD Megan K. Devlin, MD Brian M. Wong, MD Author Affiliations: Department of Medicine, Queen’s University, Kingston, Ontario, Canada (Kozij); Department of Medicine, University of Toronto, Toronto, Ontario, Canada (Devlin, Wong); Division of General Internal Medicine, Sunnybrook Health Sciences Centre, Centre for Quality Improvement and Patient Safety, Toronto, Ontario, Canada (Wong). Corresponding Author: Brian M. Wong, MD, Division of General Internal Medicine, Sunnybrook Health Sciences Centre, Centre for Quality Improvement and Patient Safety, 2075 Bayview Ave, Room H466, Toronto, ON M4N 3M5, Canada ([email protected]). Conflict of Interest Disclosures: None reported. 1. Abraham J, Kannampallil T, Patel VL. A systematic review of the literature on the evaluation of handoff tools: implications for research and practice. J Am Med Inform Assoc. 2014;21(1):154-162. 2. Abraham J, Kannampallil TG, Almoosa KF, Patel B, Patel VL. Comparative evaluation of the content and structure of communication using two handoff tools: implications for patient safety. J Crit Care. 2014;29(2):311.e1-311.e7. 3. Abraham J, Kannampallil TG, Patel VL. Bridging gaps in handoffs: a continuity of care based approach. J Biomed Inform. 2012;45(2):240-254.

What Ecologic Analyses Cannot Tell Us About Medical Marijuana Legalization and Opioid Pain Medication Mortality To the Editor Bachhuber et al1 found that states with legalized medical marijuana and those that had longer periods of legalized medical marijuana had lower adjusted opioid analgesic mortality rates than those that did not. Although they acknowledged that mechanisms underlying these relationships are speculative, the one receiving the most attention was that medical marijuana may reduce patients’ reliance on opioid pain medications, thereby reducing their mortality risk. The authors noted that one limitation of their analyses was that they were “ecologic” (ie, linking state-level variables), so they “cannot adjust for characteristics of individuals within the (Reprinted) JAMA Internal Medicine April 2015 Volume 175, Number 4

Copyright 2015 American Medical Association. All rights reserved.

Downloaded From: http://archinte.jamanetwork.com/ by a University of California - San Diego User on 08/20/2015

655

Letters

states….”1(p1671) We would suggest that the primary drawback of the ecologic analyses is that they provide no information whether individual pain patients who use marijuana have a lower or higher opioid analgesic mortality risk. Almost 65 years ago, Robinson2 illustrated that relationships between ecologic- and individual-level variables can differ in magnitude and even direction. Thus, the negative statelevel correlations identified by Bachhuber et al1 could be present even if, within each state, individual patients receiving medical marijuana had opioid overdose rates equal to or higher than patients not receiving medical marijuana. Inferring individuallevel relationships from relationships at higher levels of aggregation became known as the “ecological fallacy,” a fallacy well appreciated in sociology3 and epidemiology,4 but less so in health care policy ciricles.5 The authors also observed that the reported state-level associations might be driven by other state-level factors not included in the analyses. However, if additional potential statelevel confounds (eg, having more competent pain care providers or emergency services that respond more quickly to opioid overdoses) were included, the results might indicate that medical marijuana legalization is associated with higher state-level opioid medication mortality (as the unadjusted analyses found). More critically, no state-level analysis, no matter how completely specified, can tell us whether pain patients are more or less likely to die from opioid medication overdoses if they use medical marijuana. Instead, prospective data on individual pain patients’ marijuana, opioid pain medication, and other substance use, as well as higher-level data (eg, rigor of state medical marijuana regulation, county-level availability of legal medical marijuana from dispensaries), need to be analyzed with multilevel or “mixed-effects” models4,5 to partition the possible effects of patient and policy variables on subsequent patient-level pain medication use and mortality risk. Such studies would be much more expensive than additional ecologic analyses with administrative data but also substantially more informative.

4. Greenland S. A review of multilevel theory for ecologic analyses. Stat Med. 2002;21(3):389-395.

In Reply We agree with Finney et al that large, prospective studies of the effects of medical cannabis in individuals with chronic pain are warranted. However, we believe that results from studies of state-level data can also be informative. Robinson1 identified a fundamental concept, that associations between variables at the individual level may be different than associations between variables at the group level. However, group-level studies can be of great value when the unit of inference is the group (in the case of our report, the state) and individual-level data are unavailable.2 Furthermore, we are not simply comparing states with and without the policy at a single point of time, an inferential method often leading to flawed ecological correlations as Robinson1 illustrated. Rather, we are comparing changes before and after a statelevel policy change relative to states where there was no policy change. This over-time variation allows us to statistically adjust for many of the idiosyncratic state-level differences that might otherwise lead to biased conclusions.3 A study examining the effects of medical cannabis use among individuals with chronic pain, as Finney et al propose, will elucidate a critically important aspect of the larger question of what happens in states that legalize medical cannabis. The use of state-level outcomes, such as state-level opioid overdose mortality rates in our report, offers the advantage of being able to estimate an average state-level association spanning multiple groups (ie, not limited to individuals with chronic pain). We chose to study this state-level association because it is of broad interest to many, including lawmakers, health officials, and voters potentially deciding on whether to support or oppose medical cannabis ballot initiatives. Studies of individuals with chronic pain who use medical cannabis would help further characterize several observations from previous research. Surveys of people being evaluated for medical cannabis or attending dispensaries have found that up to two-thirds (51%-66%) of respondents reported substituting cannabis for prescription drugs; however, the proportion who substituted cannabis for opioids is unclear.4-6 Furthermore, over one-quarter (26%-30%) of patients reported substituting cannabis for an illicit drug. Studies of people with chronic pain could further elucidate whether cannabis is a substitute for opioids and whether people who misuse opioids might reduce their use and potentially switch to cannabis. Such studies would also allow us to see if, and how much, average opioid doses change and whether the effect is concentrated among individuals with the highest opioid doses. Only individual-level data can help us sort through these potential mechanisms. Examining the impact of medical marijuana laws on a state level and among individuals with chronic pain are both important goals, each with their own strengths and limitations. The effect of medical marijuana laws on individual behavior remains unclear. In addition, important questions remain with regard to the different regulatory schemes states use in legal-

JAMA Internal Medicine April 2015 Volume 175, Number 4 (Reprinted)

jamainternalmedicine.com

John W. Finney, PhD Keith Humphreys, PhD Alex H. S. Harris, PhD Author Affiliations: Center for Innovation to Implementation, VA Palo Alto Health Care System, Menlo Park, California (Finney, Humphreys, Harris); Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, California (Finney, Humphreys); Department of Orthopedic Surgery, Stanford University School of Medicine, Stanford, California (Harris). Corresponding Author: John W. Finney, PhD, VA Palo Alto HCS (152-MPD), Center for Innovation to Implementation, 795 Willow Rd, Menlo Park, CA 94025 ([email protected]). Conflict of Interest Disclosures: None reported. 1. Bachhuber MA, Saloner B, Cunningham CO, Barry CL. Medical cannabis laws and opioid analgesic overdose mortality in the United States, 1999-2010. JAMA Intern Med. 2014;174(10):1668-1673. 2. Robinson WS. Ecological correlations and the behavior of individuals. Am Sociol Rev. 1950;15:351-357. 3. van Poppel F, Day LH. A test of Drukeim’s theory of suicide—without committing the “ecological fallacy.” Am Sociol Rev. 1996;61:500-507.

656

5. Finney JW, Humphreys K, Kivlahan DR, Harris AHS. Why health care process performance measures can have different relationships to outcomes for patients and hospitals: understanding the ecological fallacy. Am J Public Health. 2011;101(9):1635-1642.

Copyright 2015 American Medical Association. All rights reserved.

Downloaded From: http://archinte.jamanetwork.com/ by a University of California - San Diego User on 08/20/2015

Letters

izing medical marijuana, and analyses of state-level data are well-poised to inform policy. Marcus A. Bachhuber, MD Brendan Saloner, PhD Colleen L. Barry, PhD, MPP

Corresponding Author: Marcus A. Bachhuber, MD, Center for Health Equity Research and Promotion, Philadelphia Veterans Affairs Medical Center, 423 Guardian Dr, 1303-A Blockley Hall, Philadelphia, PA 19104 (marcus.bachhuber @gmail.com). Conflict of Interest Disclosures: None reported. 1. Robinson WS. Ecological correlations and the behavior of individuals. Am Sociol Rev. 1950;15:351-357. 2. Schwartz S. The fallacy of the ecological fallacy: the potential misuse of a concept and the consequences. Am J Public Health. 1994;84(5):819-824. 3. Angrist J, Pischke J. Mostly Harmless Econometrics: An Empiricist's Companion. Princeton, NJ: Princeton University Press; 2009. 4. Reiman A. Cannabis as a substitute for alcohol and other drugs. Harm Reduct J. 2009;6:35. 5. Nunberg H, Kilmer B, Pacula RL, Burgdorf J. An analysis of applicants presenting to a medical marijuana specialty practice in California. J Drug Policy Anal. 2011;4(1):1. 6. Grella CE, Rodriguez L, Kim T. Patterns of medical marijuana use among individuals sampled from medical marijuana dispensaries in Los Angeles. J Psychoactive Drugs. 2014;46(4):263-272.

β-Blockers in Diabetic Patients With Heart Failure To the Editor Pasternak et al1 report that the effectiveness of metoprolol succinate and carvedilol in patients with heart failure is similar. Counteracting neurohormonal hyperactivation through β-blockers represents an established cornerstone for the treatment of heart failure. However, owing to their glycometabolic effects there are some issues concerning the use of β-blockers in patients with diabetes mellitus, which account for approximately one-third of patients with heart failure.2,3 The large and accurate population study by Pasternak et al1 could be very helpful in unveiling potentially diverse outcomes in diabetic patients receiving a selective β1-adrenergic receptor blocker (metoprolol succinate) vs a nonselective β1β2-α1–blocker (carvedilol), providing meaningful clinical and pathophysiological insights. Indeed, the different adrenergic receptors play a key role in the regulation of glucose homeostasis and insulin release.3,4 Unfortunately, the authors did not report any analysis on the group of diabetic patients, albeit clearly indicating the presence of patients receiving insulin or oral hypoglycemics.1 Similarly, given the importance of β-adrenergic receptors in the regulation of vascular tone,3,5 an evaluation in hypertensive patients could be interesting to readers.

jamainternalmedicine.com

Corresponding Author: Gaetano Santulli, MD, PhD, Columbia University Medical Center, New York, NY 10032 ([email protected]; [email protected]). Conflict of Interest Disclosures: None reported.

Author Affiliations: Center for Health Equity Research and Promotion, Philadelphia Veterans Affairs Medical Center, Philadelphia, Pennsylvania (Bachhuber); Robert Wood Johnson Foundation Clinical Scholars Program, University of Pennsylvania, Philadelphia (Bachhuber); Leonard Davis Institute of Health Economics, University of Pennsylvania, Philadelphia (Bachhuber, Barry); Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland (Saloner, Barry); Department of Mental Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland (Saloner, Barry).

Gaetano Santulli, MD, PhD

Author Affiliation: Columbia University Medical Center, New York, New York.

1. Pasternak B, Svanström H, Melbye M, Hviid A. Association of treatment with carvedilol vs metoprolol succinate and mortality in patients with heart failure. JAMA Intern Med. 2014;174(10):1597-1604. 2. Arnold SV, Spertus JA, Lipska KJ, et al. Type of β-blocker use among patients with versus without diabetes after myocardial infarction. Am Heart J. 2014;168 (3):273-279.e1. 3. Kveiborg B, Hermann TS, Major-Pedersen A, et al. Metoprolol compared to carvedilol deteriorates insulin-stimulated endothelial function in patients with type 2 diabetes—a randomized study. Cardiovasc Diabetol. 2010;9:21. 4. Santulli G, Lombardi A, Sorriento D, et al. Age-related impairment in insulin release: the essential role of β(2)-adrenergic receptor. Diabetes. 2012;61(3): 692-701. 5. Ciccarelli M, Cipolletta E, Santulli G, et al. Endothelial beta2 adrenergic signaling to AKT: role of Gi and SRC. Cell Signal. 2007;19(9):1949-1955.

In Reply We thank Dr Santulli for his suggestions for additional analyses. We have now conducted a subgroup analysis according to history of diabetes, defined as treatment with any oral antidiabetic drug or insulin within the last year. Among patients with diabetes, the crude all-cause mortality rate was 11.0 deaths per 100 person-years among carvedilol users and 11.9 deaths per 100 person-years among metoprolol succinate users; the propensity score-adjusted hazard ratio (aHR) for all-cause mortality was 0.94 (95% CI, 0.76-1.17) comparing carvedilol vs metoprolol succinate. Among patients without diabetes, the mortality rates were 6.0 and 6.1 deaths per 100 person-years among carvedilol and metoprolol users, respectively, and the aHR was 1.00 (95% CI, 0.87-1.15). Data on history of hypertension are not available in this data set. Björn Pasternak, MD, PhD Henrik Svanström, MSc Anders Hviid, DrMedSci Author Affiliations: Department of Epidemiology Research, Statens Serum Insitut, Copenhagen, Denmark. Corresponding Author: Björn Pasternak, MD, PhD, Department of Epidemiology Research, Statens Serum Insitut, Artillerivej 5, 2300 Copenhagen S, Denmark ([email protected]). Conflict of Interest Disclosures: None reported.

Concern About the Use of Venlafaxine to Treat Vasomotor Symptoms To the Editor In a recent issue of JAMA Internal Medicine, Joffe et al1 report the results of a 3-arm, double-blind trial randomizing healthy perimenopausal and postmenopausal women with bothersome vasomotor symptoms (VMS) to low-dose estradiol and venlafaxine. Their findings contribute to a better understanding of the differences between estrogen therapy and venlafaxine therapy. However, there is a concern about the use of venlafaxine to treat VMS. It is well known that menopause is a risk factor for cardiovascular disease, the leading cause of death in women in the United States.2 Thus, the blood pressure– increasing effect of venlafaxine, observed by Joffe et al1 and others,3 may potentiate the menopausal risk for cardiovascu(Reprinted) JAMA Internal Medicine April 2015 Volume 175, Number 4

Copyright 2015 American Medical Association. All rights reserved.

Downloaded From: http://archinte.jamanetwork.com/ by a University of California - San Diego User on 08/20/2015

657

What ecologic analyses cannot tell us about medical marijuana legalization and opioid pain medication mortality.

What ecologic analyses cannot tell us about medical marijuana legalization and opioid pain medication mortality. - PDF Download Free
80KB Sizes 12 Downloads 36 Views