Unintended consequences of reducing QT-alert overload in a computerized physician order entry system

From Clinfowiki
Revision as of 02:43, 28 October 2015 by Eyao (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

This is a qualitative study by van der Sijs H, Kowlesar R, Aarts J, Berg M, Vulto A, van Gelder T. (2009) in the Journal of American Medical Informatics Association, entitled “Unintended consequences of reducing QT-alert overload in a computerized physician order entry system.” [1]


A prolonged “Q-T interval”, a period from the beginning of the Q wave to the end of the T wave on an electrocardiogram represents the full electrical depolarization (QRS) and repolarization (S to T) of the heart. Mechanically, this period covers contraction and ejection of blood by the heart, and part of the relaxation of the heart. Prolongation of this interval, normally less than 420 milliseconds, predisposes to lethal arrhythmias, the most extreme being “torsade de pointes.” Although there are many inherited forms of prolonged Q-T intervals (LQTS), acquired forms vastly predominate in adults (Cubeddu, 2003). About 60% of cardiovascular deaths occur outside the hospital, nearly 85% due to sudden cardiac death, and most patients cannot be resuscitated. Both inside and outside the hospital, prolongation of the Q-T interval is a potent risk factor for sudden death (Go et al, 2014). The QT interval may be prolonged with age, female gender, disease (diabetes types I and II, rheumatoid arthritis, heart disease, kidney failure, hypothyroidism, low potassium or high calcium levels, among others), genetic predisposition, and as an adverse drug reaction, particularly combinations of two drugs, each of which prolong the QT interval (van Noord, Eijgelsheim, & Stricker, 2010). On the short list of preventable causes, drug combinations lead. Without a recent electrocardiogram (ECG), and in busy clinical practice, the practitioner may not be fully aware of the dangers of prescribing such drugs, especially when the patients are using many pharmaceuticals (Al-Khatib et al, 2003). Because of the relative high priority given to avoidance, clinical decision support (CDS) alerts are common to avoid dangerous QT prolongation.


Because of the complaints about an excess of low-specificity alerts, in 2007 QT-prolonging drug alerts via a computerized physician order entry system (CPOE) were limited to important QT-prolonging drugs in the Dutch database, rather than those with little clinical significance. The purpose of this study (van der Sijs et al, 2009) was to assess the effects of the new alert policy and positive predictive value. The authors used improvements in at-risk patient identification in developing a highly associated deadly rhythm, Torsades de Pointes (TdP), as a surrogate in patients receiving drugs with this adverse effect.


During the period between 1 February 2006 and 31 July 2006, inpatient entry criteria included a) overridden QT-related DDI alerts b) an ECG within 1 month of the override, and c) without a pacemaker and not using a low-QT-risk combination of cotrimoxazole and tacrolimus. QT-interval prolongation and the risk of developing TdP were calculated for all participants. This risk was then compared to the number of patients for whom a QT-alert would be issued using the post-2007 alert policy.


Of 49 patients (13%) meeting the study criteria, alerts restricted to important QT-prolonging drugs (termed “the adjusted knowledge base”) resulted in a 53% reduction in the number of alerts. Unfortunately, the new alert system only identified 47% of participants at risk for developing TdP. Moreover, there was no change in the positive predictive value of QT alerts.


Even though the number of alerts fell, the specificity of the alerts generated did not increase sufficiently to raise the predictive value of the newer alert policies, ie, the adjusted knowledge base. As such, the newer system missed over half of the patients at significant risk. The large number of unknown risk factors, including drug class and dose that may influence the QT interval, improving the predictive value of alert policies cannot be accomplished. The authors suggest that if these data were available, QT alerts with high sensitivity, specificity, and positive predictive value could be developed. In addition, they suggest an ECG be done before and after a QT-related alert override to assist in the identification of at-risk patients and adverse events.


Prompting the “unintended consequences” used in the title, the restricted use of alerts regarding drugs that prolong the QT interval identified fewer at-risk patients than the older paradigm. Postulating that many non-drug risk factors confounded the results, the authors proposed that these non-drug factors be identified and their contributions better characterized in order to improve predictive value of alert policies.


Faced with a practical problem, the authors sought to verify whether the assumption of the revised policy of filtering QT-interval alerts would be beneficial. Instead, they documented “negative” results which reflected a far more vexing and complex situation than anticipated. While the authors offered “other risk factors” as a means of “understanding” their results, neither their study design nor the data actually demonstrated this was the case. Hence the basis of their hypothesis was actually elimination combined with their view of the possible confounding variables that might influence the QT interval. In their limitations section, they mention other variables, such as diurnal variation of the QT interval, nonstandardized ECGs, and variability in QT intervals to this list, and the inability to generalize their results. In addition to their interpretations, this report can be further explained by (1) the use of surrogates such as QT interval-related alerts, TdP, and poor-quality (or absent) electrocardiograms to reflect rhythm deterioration in complex, ill patients, (2) individual variability to QT effects of drugs which cannot be quantitated, and (3) the use of an oversimplified model of the QT interval and its determinants. The QT interval is actually the sum of all the action potentials (electrical events) that exist in all heart muscle cells in the ventricle at that time. This number is dependent upon opening and closing numerous channels for ion transport, all in constant flux, with a much larger number of independent variables than assumed in this paper (Kallergis et al, 2012). In a sense, predicting total risk from partial underpinnings of a surrogate—the QT-interval—was far more complicated than the study design envisioned.

Related Articles


  1. van der Sijs, et.al. (2009). Unintended consequences of reducing QT-alert overload in a computerized physician order entry system.http://www.ncbi.nlm.nih.gov/pubmed/19415251.