Article MDM Policy & Practice 1–12 Ó The Author(s) 2019 Traffic Lights Intervention Reduces Article reuse guidelines: sagepub.com/journals-permissions Therapeutic Inertia: A Randomized DOI: 10.1177/2381468319855642 journals.sagepub.com/home/mdm Controlled Trial in Multiple Sclerosis Care Gustavo Saposnik , Muhammad Mamdani, Xavier Montalban, Maria Terzaghi, Berenice Silva, Maria Laura Saladino, Philippe N. Tobler*, and Fernando Caceres* Abstract Background: Therapeutic inertia (TI) is a common phenomenon among physicians who care for patients with chronic conditions. We evaluated the efficacy of the traffic light system (TLS) educational intervention to reduce TI among neurologists with MS expertise. Methods: In this randomised, controlled trial, 90 neurologists who provide care to MS patients were randomly assigned to the TLS intervention (n = 45) or to the control group (n = 45). The educa- tional intervention employed the TLS, a behavioral strategy that facilitates therapeutic choices by facilitating reflec- tive decisions. The TLS consisted in a short, structured, single session intervention of 5-7 min duration. Participants made therapeutic choices of 10 simulated case-scenarios. The primary outcome was a reduction in TI based on a published TI score (case-scenarios in which a participant showed TI divided by the total number of scenarios where TI was possible ranging from 0 to 8). Results: All participants completed the study and were included in the primary analysis. TI was lower in the TLS group (1.47, 95% CI 1.32-1.61) compared to controls (1.93; 95% CI 1.79-2.08). The TLS group had a lower prevalence of TI compared to controls (0.67, 95% CI 0.62-0.71 vs. 0.82, 95% CI 0.78- 0.86; p = 0.001). The multivariate analysis, adjusted for age, specialty, years of practice, and risk preference showed a 70% reduction in TI for the TLS intervention compared to controls (OR 0.30; 95% CI 0.10-0.89). Conclusions: In this randomized trial, the TLS strategy decreases the incidence of TI in MS care irrespective of age, expertise, years for training, and risk preference of participants, which would lead to better patient outcomes. Keywords decision making, disease-modifying therapy, educational intervention, multiple sclerosis, randomized clinical trial Date received: February 8, 2019; accepted: May 3, 2019 The management of multiple sclerosis (MS) evolves with caring for patients with chronic conditions.1,4–6 Insufficient the availability of new disease-modifying agents, varying knowledge integration and knowledge-to-action gaps are dosage forms (oral, injectable, infusion), and different among the most common explanations for suboptimal safety and efficacy profiles. As a result, physicians who care therapeutic decisions. However, TI is a complex process for MS patients face more complex therapeutic decisions when considering individual patients, number of relapses, activity on brain imaging, and the need to escalate therapy. Corresponding Author Moreover, many MS patients remain undertreated.1–3 Gustavo Saposnik, Neuroeconomics Program, Department Economics, University of Zurich, Switzerland. (gustavo.saposnik@econ.uzh.ch) Therapeutic inertia (TI) corresponds to the absence of Department of Medicine (Neurology), St. Michael’s Hospital, treatment initiation or intensification when treatment University of Toronto, 55 Queen St E, Toronto, Ontario M5C 1R6, goals are unmet. It affects more than 50% of clinicians Canada; Telephone: (416) 864-5155 (saposnikg@smh.ca). This Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 License (http://www.creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage). 2 MDM Policy & Practice 00(0) also related to other characteristics of the providers such labels enhance the coupling between brain regions associ- as lack of knowledge about appropriate goals, high patient ated with valuation (i.e., ventromedial prefrontal cortex) volume, and time constraints. Some physicians fail to inte- and self-control.17 Evidence from the literature suggests grate the available information (e.g., severity of the condi- that medical decisions leading to TI are likely related to tion, risk of progression, imaging findings affecting knowledge-to-action gaps.7 The design and application of outcomes) with best practice recommendations for a given the TLS as an educational intervention provides a unique risk-scenario.7–10 Furthermore, physicians have limited opportunity to overcome knowledge-to-action gaps in training in risk management and formal learning in medi- MS care.18 cal decision making.11–13 All of these factors may contrib- In a previous pilot study, we assessed the feasibility ute to TI and undertreatment. and potential efficacy of a TLS educational intervention Educational interventions have been designed to in 25 neurologists from Spain.19 TI was present in optimize knowledge integration and bridge knowledge- 72.0% of participants in at least one case scenario. The to-action gaps for complex medical decisions (e.g., diag- primary feasibility outcome, the completion rate of the nostic challenges, varying risk categories, availability of study, was 100% (25/25 participants). While not pow- multiple agents with a broad range of safety/efficacy ered to detect a significant difference between groups, ratio).11 One such intervention is the traffic lights system our pilot study demonstrated a nonsignificant reduc- (TLS). The TLS emerged as risk categorization and tion in TI for the targeted intervention group relative warning strategy to reduce human errors that relies on to the control group (22.6% v. 33.9% postintervention; a relatively automatic, well-established, and cross- odds ratio [OR] = 0.57; 95% confidence interval [CI] cultural concept to increase the chance of an optimal = 0.26 to 1.22).19 The TLS also showed a high usabil- course of action.14,15 The TLS facilitates the decision- ity score (74.7; 95% CI = 70.1 to 79.2) when tested in making process using traffic light terminology, which a larger study comprising neurologists from Argentina, creates a link between a color, representing a risk level, Chile, and Canada.20 and an action: red light (‘‘high risk’’/‘‘stop and think’’), In the present study, we evaluated the efficacy of our yellow light (‘‘intermediate risk’’/‘‘reassess soon’’), and simple and pilot-tested educational TLS intervention for green light (‘‘low risk’’/‘‘continue the same strategy’’). reducing TI among neurologists who routinely provide For example, studies showed that the TLS facilitates care of MS patients. healthier food choices by interrupting automatic behavior and triggering a reevaluation processes.16 A functional magnetic resonance imaging study showed that TLS Methods Study Design and Participants Division of Neurology, Department of Medicine (GS) and Decision Neuroscience Unit, Li Ka Shing Knowledge Institute (GS, MT), St. The overarching goal of the TLS intervention was to Michael’s Hospital, University of Toronto, Toronto, Canada; facilitate the integration of gaps between risk stratifica- Laboratory for Social and Neural Systems Research, Department of tion and treatment decisions (initiation or escalation) in Economics, University of Zurich, Zurich, Switzerland (GS, PNT); Li Ka Shing Centre for Healthcare Analytics Research and Training MS care. Specifically, our randomized parallel trial tested (LKS-CHART), Toronto, Toronto, Canada (MM); Department of the efficacy of an educational TLS intervention (active Neurology-Neuroimmunology, Centre d’Esclerosi Múltiple de group) against usual care (control group) for reducing TI Catalunya (CEMCAT), Hospital Universitari Vall d’Hebron, in the management of MS (Figure 1). Participants were Barcelona, Spain (XM); Institute of Neuroscience Buenos Aires (INEBA), Buenos Aires, Argentina (BS, MLS, FC). Prof. Philippe randomly assigned (1:1 ratio) to the TLS or the control Tobler was funded by the Swiss National Science Foundation group by Qualtrics (Qualtrics.com; Online Appendix, (PP00P1_150739 and 100014_165884). Dr. Gustavo Saposnik is a Figure e1). Allocation concealment was implemented in neurologist educated in Argentina, currently practicing neurology in Qualtrics, so that participants did not know which inter- Canada supported by the HSF Scientist Award. Drs. Mamdani, vention they were allocated to after completing the initial Caceres, Saladino, Silva, Terzaghi, and Montalban have no conflicts to disclose. The study was sponsored by the Institute of Neuroscience demographic information. Investigators were also Buenos Aires (INEBA) with endorsement by the Sociedad Argentina blinded to the treatment allocation. Participants were de Neurologia (SNA) and NeuroEconSolutions, and funded by an recruited by automatic e-mail invitation from the study operating grant from Roche. Neither the SNA nor Roche were platform. The Institute of Neuroscience Buenos Aires involved in the design, execution, analysis, and interpretation or and the Argentinian Neurological Society facilitated the reporting of the results. mailing of information to potential participants who met *Drs. Caceres and Tobler contributed equally. the inclusion criteria. Saposnik et al. 3 Figure 1 Consort flow diagram. Of 117 eligible participants, 90 participants were randomized to the educational intervention (n = 45) and control (n = 45). All participants completed the intervention and contributed a complete set of data to the analysis. Data Collection unfavorable) over one where this probability is unknown (Figure 2A and B). Risk experiments involved determin- Participants answered questions as follows: 1) demo- ing the subjective value of a risky (50/50) option in terms graphic and practice-based information, 2) behavioral of a safe (100%) option. By asking participants to indi- experiments, and 3) 10 case scenarios that assessed TI. cate the magnitude of the safe option at which they Behavioral experiments used previously established would be indifferent between the two options, we were in designs to assess participants’ risk preferences and toler- a position to determine the point of subjective equiva- ance to ambiguity in the health and financial lence of the risky and the safe option. The higher this domains.21,22 In brief, ambiguity aversion is defined as point, the higher the propensity to take risk. Further dislike for events with unknown probability compared to details of these experiments were published in previous events with known probability.21 For example, in the studies (and appendix).3 medical domain, an ambiguity-averse individual would Participants were also asked to select those MS drugs rather choose a treatment where the probability of bene- that they use and then rank them from a list including all fits or side effects is known (even if these are somewhat available agents approved by the local regulatory body in 4 MDM Policy & Practice 00(0) Figure 2 Experiments to assess ambiguity in the financial and health domains. Participants were told to imagine two different types or urns. For urn type A, they knew that 50% of the balls were red and the other 50% were blue. For urn type B, they did not know the exact proportion of blue to red balls, with the grey bar representing the unknown proportion of balls. For the financial domain (Panel A), participants knew that if they drew a blue ball, they would win the full amount of $400. If they drew a red ball, they would win $ 0. For the health domain (Panel B), participants decided between two treatments for a patient. With ‘‘Treatment A,’’ the patient had a 50% probability of survival. With ‘‘Treatment B,’’ the exact probability of survival was unknown, with the grey bar representing the unknown probability. In our tasks, participants were asked to choose between one option (presented as two-colored bar) with known 50/50 probability of winning 400 or 0 American dollars (urn A) versus an option with unknown probability of the same outcomes (urn B). Participants who chose the 50/50 options were classified as averse to ambiguity; the remaining participants were classified as tolerant to ambiguity.1 A similar approach was used to determine aversion to ambiguity in the health domain (Panel B). Saposnik et al. 5 Argentina by March 30, 2018. The purpose of this strat- present analysis, we used the aforementioned scheme egy was to compare how TI derived from case scenarios according to the current clinical practice.24–26,33 relates with agents used for treatment escalation (e.g., fin- golimod, natalizumab, alemtuzumab) in routine clinical practice. At the time of conducting this trial, cladribine The Traffic Light System (Figure 3) and ocrelizumab were not available and therefore not In our study, the TLS was applied to help participants included in case scenarios. identify scenarios with poor prognosis (high risk of dis- In line with the learning and education literature, ease progression; Figure 3B). Consequently, participants vignettes, clinical case scenarios, or ‘‘real-world’’ encoun- would be able to identify the ‘‘red’’ traffic light as a ters are regarded as the best simple strategy to evaluate warning sign for a high-risk situation, and subsequently potential cognitive biases and medical decisions.23 Case escalate treatment. The ‘‘yellow’’ traffic light represents scenarios were designed by our research team and MS caution for scenarios with either a clinical relapse or experts derived from the most common situations in clin- some degree of activity on brain imaging (but not both), ical practice as previously reported in our pilot study.19 which requires a reassessment within 6 to 12 months. Eight case scenarios were designed to assess appropriate The control group made therapeutic decisions without treatment initiation or escalation, whereas the remaining being exposed to the TLS intervention as part of the cur- two cases were designed to assess overtreatment (defined rent standard practice. Further details of the TLS inter- as treatment escalation when there was low risk of dis- vention are described elsewhere.19 An example of the ease progression and no evidence of disease activity).24–26 presented case scenario is represented in Figure 3C. Participants from each randomized group were exposed to the same case scenarios. Inclusion criteria comprised Outcome Measures neurologists who were actively involved in managing MS patients. Physicians whose practice was primarily in car- Therapeutic inertia was the primary outcome of interest, ing for MS patients or who obtained a subspecialty measured as a continuous variable (TI score) and as bin- degree were classified as ‘‘MS specialists.’’ Physicians ary. The TI score was defined as the number of case sce- who were not practicing neurology or seeing less than narios in which a participant showed TI (ranging from 0 one MS patient per month were excluded from the study. to 8). The TI score was reported in our previous studies to reflect the magnitude of TI.19,34 A low TI score repre- sents low TI, whereas a higher score represents higher. A Definitions 0.5 point difference in the TI score was deemed as clini- cally meaningful given the impact in clinical practice. In For the primary analysis, bad prognosis was defined as our study, detecting a difference equal to or greater than the combination of a clinical relapse plus the presence of 0.50 between groups in the TI score would represent a new brain lesions in follow-up magnetic resonance ima- clinically meaningful improvement. The TI score was ging (MRI) scans or at least one gadolinium-enhancing derived from case scenarios, which were aligned with the lesion.27,28 All high-risk cases included a description of current Argentinian, North American, and European an MRI with more than five new T2 lesions or at least practice recommendations.24–26 one enhancing lesion.29 The use of these definitions com- A reduction in the TI score reflects that participants bining a clinical relapse and MRI activity is consistent appropriately switched from a first-line agent (e.g., gla- with recent evidence regarding the risk of treatment fail- tiramer, interferons) to a high-efficacy treatment (e.g., ure among patients receiving interferon-b.30 Disease pro- fingolimod as a second-line therapy or monoclonal anti- gression was defined as at least one point worsening bodies as third-line therapies) when clinical and radiolo- from baseline to 1-year follow-up in the Expanded gical evidence of disease progression. Disability Status Scale (EDSS) score.31 Recent meta- TI as a binary outcome was defined as the proportion analysis confirmed that alemtuzumab, natalizumab, and of participants demonstrating TI in at least one of the fingolimod are the best available choices for preventing eight scenarios (prevalence of TI). clinical relapses in patients with relapsing-remitting MS (RRMS).32 The current treatment option for RRMS include first-line (b-interferons, glatiramer acetate), Statistical Analysis second-line (fingolimod, cladribine), and third-line (nata- We used parametric tests (t test and Fisher’s exact test) to lizumab, alemtuzumab, ocrelizumab) therapies. For the compare continuous and categorical variables between 6 MDM Policy & Practice 00(0) Figure 3 The TLS intervention. Panel A illustrates background information on the TLS and application to therapeutic decisions. Panel B illustrates how the TLS facilitates the decision-making process using traffic light terminology, which creates a link between a color, representing a risk level, and an action: red light (‘‘high risk’’/‘‘stop and think’’), yellow light (‘‘intermediate risk’’/‘‘reassess soon’’), and green light (‘‘low risk’’/‘‘continue the same strategy’’). Panel C provides a case scenario as an example of those given the participants. Saposnik et al. 7 Table 1 Baseline Characteristics of Participants Characteristics Control (N = 45) Educational Intervention (N = 45) P Value Age in years (mean 6 SD) 46.8 6 10.3 46.0 6 10.4 0.72 Sex, n (%) 0.67 Female 20 (44.4) 22 (48.9) Male 25 (55.6) 23 (51.0) Specialty, n (%) 0.82 General neurologists who care for MS patients 29 (64.4) 30 (66.7) MS specialists 16 (35.6) 15 (33.3) Practice setting, n (%) 0.37 Public 17 (37.8) 13 (28.9) Private 28 (62.2) 32 (71.1) % time in clinical practice, n (%) .75% 19 (42.2) 20 (44.4) 0.83 Years in practice (mean 6 SD) 21.1 6 10.5 19.5 6 11 0.48 MS patients seen per week (mean 6 SD) 21.8 6 4.5 23.0 6 8.3 0.37 Author of a peer-reviewed publication in the last 1 year, n (%) 20 (44.4) 23 (51.1) 0.53 Restriction to prescribe MS drugs, n (%) 0.83 No restrictions 28 (62.2) 29 (64.4) Treatment escalation, mean drugs (6SD) 2.91 (1.24) 3.09 (2.05) 0.62 MS, multiple sclerosis groups. Linear regression analysis was used to determine completed the survey (response rate 76.9%) between the efficacy of the TLS for reducing TI scores in the inter- April and September 2018. There was representation vention group versus the control group. Similarly, logistic from all provincial territories. The mean (SD) age was regression analysis was used to determine the efficacy of 46.4 (610.3) years; 48 (53%) were male neurologists. the TLS with respect to the proportion of participants Thirty-one (34.4%) participants primarily focused their with TI in at least one case scenario. We also evaluated practice on MS care. They had 20.3 (610.9) years of the association between restrictions for prescribing DMTs practice and were assessing 22 (66.6) MS patients per and the number of second- and third-line agents com- week. Table 1 compares baseline characteristics between monly used in clinical practice with the TI score. groups. Groups did not differ in demographics or in risk For multivariate analysis of individual responses, we preferences as measured by the behavioral risk tasks (P constructed multilevel mixed-effects models where parti- = 0.40 for risk preferences and P = 0.63 for aversion to cipants (n = 90) and individual responses (n = 720; 90 ambiguity). There were no differences in treatment esca- participants each completing 8 case scenarios) entered as lation at baseline between groups. On average, partici- random effects. The aim of this analysis was to evaluate pants in the TLS group used 3.09 agents for treatment the contribution of individual-specific variables to the escalation versus 2.91 in the control group (P = 0.62). variation of TI. Given the findings from our previous There was no association between participants’ restric- studies, we included the following a priori variables: par- tions to prescribe MS drugs and TI score (P = 0.44) or ticipant age, expertise (MS specialist v. general neurolo- the prevalence of TI (P = 0.78). gist), years of experience, and aversion to ambiguity.3,19 Table 2 summarizes the primary and secondary out- All tests were 2-tailed, and P values \0.05 were consid- come measures at the participant and individual response ered significant. We used STATA 13 (StataCorp LP, levels. TI scores were significantly lower in the TLS inter- College Station, TX) to conduct all analyses. vention group than in the control group (1.36, 95% Further details of the protocol are published in CI = 1.23 to 1.50, v. 2.04, 95% CI = 1.90 to 2.17) after ClinicalTrials.gov (# NCT03134794). adjustment for the prespecified variables (age, specialty, years of practice, and risk preferences). The observed 0.68 difference between groups in the adjusted TI scores Results was greater than the minimal clinically meaningful mea- Out of the 117 neurologists with expertise in MS care sure of 0.5 to detect an improvement. Similarly, partici- who were invited to participate in the study, 90 pants in the TLS intervention group had a lower 8 MDM Policy & Practice 00(0) Table 2 Multivariate Analysis for the Primary and Secondary Outcome Measuresa Difference Between Multivariate Regression Outcome Measures Control Intervention Groups Analysis (95% CI) Primary outcome Participant-level analysis n = 45 n = 45 TI score, mean (SD) 1.93 (1.42) 1.47 (1.42) (0.46) 20.68 (21.24 to 20.11)b Secondary outcome measures TI (present v. absent) in at least one 37 (82.2) 30 (66.7) (15.5) 0.30 (0.10 to 0.89)c case scenario, n (%) Individual responses n = 360 n = 360 TI score, mean (SD) 1.93 (1.41) 1.47 (1.41) (0.46) 20.68 (20.87 to 20.48)d TI present versus absent, n of 87/360 (24.2) 66/360 (18.3) (5.9) 0.60 (0.41, 0.88)d individual responses/total CI, confidence interval; OR, odds ratio; TI, therapeutic inertia. a All models adjusted for age, specialty, years of practice, risk preference, and group (intervention v. control). b Derived from linear regression models and expressed in b coefficients (95% CI) with TI score as dependent variable. c Derived from multivariate logistic regression analysis with TI (present v. absent) as dependent variable. d Derived from multilevel mixed effects models expressed as OR (95% CI) for binary outcomes (TI present v. absent) and b coefficients (95% CI) for the TI score. prevalence of TI compared to controls (63.7%, 95% activity), there will be over 24 patients who will remain CI = 58.9% to 68.6%, v. 83.9%, 95% CI = 80.4% to with the same treatment if managed by neurologists 88.4%) (Figure 4). without educational intervention (control group). That The multivariate analysis also revealed a significant number would be decreased to 10 patients if treated by reduction in the TI score for the TLS intervention com- neurologists who received the TLS educational pared to the control group (b = 20.68; 95% CI = intervention. 21.24 to 20.11). The multivariate logistic regression analysis showed 70% reduction in the odds of TI for the TLS group (OR = 0.30, 95% CI = 0.10 to 0.89). Discussion Specialist status (P = 0.002), higher years of experience In the present randomized controlled trial (RCT), we (P = 0.007), and tolerance to ambiguity (P = 0.043) evaluated the efficacy of a newly designed pilot-tested19 were associated with lower TI. The adjusted models educational intervention to overcome TI among practi- showed good discrimination (c-statistic = 0.74) and cali- cing neurologist care for MS patients. We found TI in 7 bration (goodness-of-fit test P = 0.52). out of 10 participating neurologists in at least one case Results were consistent for the analyses of individual scenario. The TLS educational intervention was associ- responses (Table 2). There were no significant differences ated with a 68% reduction in the TI score or 70% reduc- between fixed and random effect models (data not tion in the odds of TI. In other words, participants shown). Figure 3 represents the relationship between the appropriately choose a higher efficacy treatment (e.g., observed and predicted TI scores (Figure 5A) and strati- monoclonal antibodies) instead of continuing with the fies the data by group (Figure 5B), revealing that TI was same agents (e.g., glatiramer, interferon) when clinical consistently lower in the intervention group. and radiological evidence of disease progression. The Participants who commonly used agents for treatment effect of the educational intervention was similar for all escalation in their daily practice had lower TI scores categories of TI scores. Specialist status, years of experi- (1.48 v. 1.8; P \ 0.01). Accordingly, participants who ence, and tolerance to ambiguity were associated with do not commonly use agents for treatment escalation in lower TI. Moreover, selection of common agents used their daily practice benefited from the intervention (for for treatment escalation in participants’ routine practice TI score: b = 20.78, 95% CI = 21.03 to 20.52; for TI was associated with lower TI scores. More interestingly, prevalence: OR = 0.56, 95% CI = 0.36-0.88). the TLS intervention was effective among participants The analysis of individual responses also revealed that who do not commonly use agents for treatment escala- for every 100 MS patients with a bad prognosis (e.g., tion in their daily practice by showing a significant both clinical and radiological evidence of disease reduction in TI. Our results were consistent for the Saposnik et al. 9 Figure 4 TLS intervention decreased therapeutic inertia (TI). (A) Comparison of adjusted TI scores in the intervention and control groups. This graph was derived from the multivariate linear regression analysis adjusted for age, years of practice, participants risk preference, and specialty (general neurologist v. MS expert). TI scores were significantly higher in the control group compared to the intervention group (*P value \0.001). (B) Comparison of adjusted prevalence of TI between the intervention and control groups. This graph was derived from the multivariate logistic regression analysis adjusted for age, years of practice, participants risk preference, and specialty (general neurologist v. MS expert). The prevalence of TI was significantly higher in the control group compared with the intervention group (*P value \0.001). analysis at the participant level and individual responses participants. Although the authors evaluated outcomes in for both categories of TI and TI scores. multiple domains (e.g., self-efficacy, knowledge, behavior There are not many proven effective interventions in change), none of the studies assessed patients’ benefits.35 In medical education associated with improvements in clini- MS care, TI may lead to undertreatment. By extension, the cally meaningful outcomes.35 A recent systematic review TLS intervention used here may eventually have patient evaluated 302 controlled studies that had investigated the benefits if it reduces TI in MS care. effect of evidence-based educational interventions. Of 85 We used TLS to reduce TI in MS care. Previous articles that met the inclusion criteria, 46 (54%) studies were authors proposed the TLS to monitor treatment response randomized trials, 51 (60%) included postgraduate-level in patients with RRMS. They included a more 10 MDM Policy & Practice 00(0) Figure 5 Adjusted probability of therapeutic inertia (TI). (A) Adjusted probability of TI as a function of TI scores. (B) Adjusted TI score categories stratified by intervention assignment group. The x-axis represents categories of the TI to evaluate whether the intervention had a different effect among participants with low, medium, and high TI scores. The y-axis represents the TI scores to be able to show the lack of overlap of 95% CI between TLS and controls for each TI category (P value TLS v. controls = 0.02). Data derived from multivariate linear regression with TI score as the dependent variable. ‘‘I’’ represent 95% CI error bars. sophisticated scoring system (0 to 3) for different cate- Our results have limitations that deserve comment. gories (clinical relapses, evidence of disease progression, First, our sample size is relatively small. However, our cognitive status, and MRI findings) making practical use RCT was designed and powered to detect differences in in daily practice more difficult. This scoring system leads TI following our pilot study.19 Second, case scenarios to a decision model that uses the TLS to facilitate thera- may not necessarily reflect participants’ daily practice. peutic choices.36 However, this strategy has not been pre- It is also possible that general neurologists apply the viously tested in a RCT. The findings of our RCT ‘‘first do not harm rule’’ when not escalating treatment suggest that the TLS may be a useful medical educational (commonly associated with more severe side effects). intervention, in keeping with research on the manage- Following these arguments, our study underestimates the ment of obesity, fever in children, blood pressure control, prevalence of TI in real-life practice given a tightly con- and healthy food choices.14,15 trolled environment and applicability of treatment Saposnik et al. 11 recommendations in our RCT design. Third, we cannot Supplemental Material rule out the possibility that unmeasured confounders Supplementary material for this article is available on (e.g., health policy, restrictive prescription rules) may the Medical Decision Making Policy & Practice website at play a role for the studied outcome measures. We con- https://journals.sagepub.com/home/mpp. trolled for this issue by measuring the prevalence of pre- scription restrictions in the workplace for each ORCID iD participant. No association was found between prescrip- Gustavo Saposnik https://orcid.org/0000-0002-5950-9886 tion restrictions and the outcomes of interest. Fourth, we have no information regarding the sustainability of the References TLS effect on TI given the design of our study. Future 1. O’Connor PJ, Sperl-Hillen JAM, Johnson PE, Rush WA, research will be needed to investigate this question. Biltz G. Clinical inertia and outpatient medical errors. In: Finally, the definition of TI applied to MS care may not Henriksen K, Battles JB, Marks ES, Lewin DI, eds. be widely used. Nevertheless, we used a practical and Advances in Patient Safety: From Research to Implementa- conservative definition of TI (absence of escalation with tion (Volume 2: Concepts and Methodology). Rockville the concomitant presence of a clinical relapse and evi- (MD): Agency for Healthcare Research and Quality; 2005. dence of imaging activity) consistent with our previous 2. Ziemssen T, Derfuss T, de Stefano N, et al. Optimizing studies, which is supported by guidelines showing treatment success in multiple sclerosis. J Neurol. 2016;263(6):1053–65. doi:10.1007/s00415-015-7986-y improvements in clinical outcomes when escalating 3. Saposnik G, Sempere AP, Prefasi D, et al. Decision-mak- therapies (i.e., blood pressure and diabetes).6,24,26 ing in multiple sclerosis: the role of aversion to ambiguity In the evolving landscape of MS treatment, new and for therapeutic inertia among neurologists (DIScUTIR MS). more effective agents with improvements in safety pro- Front Neurol. 2017;8:65. doi:10.3389/fneur.2017.00065 files are becoming available.33 Despite such advance- 4. Mohan AV, Phillips LS. Clinical inertia and uncertainty in ments, many MS patients remained undertreated.25,26 medicine. JAMA. 2011;306(4):383. Several conditions affect the risk of TI, but physicians’ 5. Phillips LS, Branch WT, Cook CB, et al. Clinical inertia. factors are regarded as the most influential.33 Our results Ann Intern Med. 2001;135(9):825–34. revealed that an innovative and highly usable educa- 6. Okonofua EC, Simpson KN, Jesri A, Rehman SU, Dur- tional intervention may revert the incident risk of TI kalski VL, Egan BM. Therapeutic inertia is an impediment to achieving the Healthy People 2010 blood pressure con- among neurologists who care for MS patients. This is trol goals. Hypertension. 2006;47(3):345–51. doi:10.1161/01 also supported by the following facts: 1) participants .HYP.0000200702 who commonly use agents for treatment escalation had 7. Gongora-Ortega J, Segovia-Bernal Y, Valdivia-Martinez lower TI, 2) the significant reduction in TI for the TLS Jde J, Galaviz-DeAnda JM, Prado-Aguilar CA. Educa- intervention group among neurologist who do not com- tional interventions to improve the effectiveness in clinical monly use agents for treatment escalation, and 3) the competence of general practitioners: problem-based versus relevance of the role of MS specialist in making thera- critical reading-based learning. BMC Med Educ. 2012;12: peutic decisions given the lower prevalence of TI. 53. doi:10.1186/1472-6920-12-53 Our findings have practical clinical and health policy 8. Kennedy T, Regehr G, Rosenfield J, Roberts SW, Lingard implications, which may not only lead to improving out- L. Exploring the gap between knowledge and behavior: a comes for patients but also to the implementation of qualitative study of clinician action following an educa- tional intervention. Acad Med. 2004;79(5):386–93. educational interventions in physicians managing high- 9. Djulbegovic B, Beckstead JW, Elqayam S, et al. Evalua- risk and complex patients. Our intervention has the tion of physicians’ cognitive styles. Med Decis Making. potential to be translated to other highly prevalent medi- 2014;34(5):627–37. doi:10.1177/0272989X14525855 cal conditions, including the management of hyperten- 10. Kostopoulou O, Sirota M, Round T, Samaranayaka S, sion, diabetes, and dyslipidemia commonly affecting Delaney BC. The role of physicians’ first impressions in individuals at high risk of cardiovascular and cerebro- the diagnosis of possible cancers without alarm symptoms. vascular diseases. Med Decis Making. 2017;37(1):9–16. doi:10.1177/02729 89X16644563 11. Dijkstra IS, Pols J, Remmelts P, Brand PL. Preparedness Acknowledgments for practice: a systematic cross-specialty evaluation of the The authors are most grateful to all physicians participating in alignment between postgraduate medical education and the study. We appreciate the support from the Department of independent practice. Med Teach. 2015;37(2):153–61. doi: Economics at the University of Zurich. 10.3109/0142159X.2014.929646 12 MDM Policy & Practice 00(0) 12. Monrouxe LV, Grundy L, Mann M, et al. How prepared in relapsing remitting multiple sclerosis patients. J Neurol are UK medical graduates for practice? A rapid review of Sci. 2018;385:217–24. doi:10.1016/j.jns.2018.01.004 the literature 2009–2014. BMJ Open. 2017;7(1):e013656. 25. Montalban X, Gold R, Thompson AJ, et al. ECTRIMS/ doi:10.1136/bmjopen-2016-013656 EAN guideline on the pharmacological treatment of people 13. Kostopoulou O, Russo JE, Keenan G, Delaney BC, Douiri with multiple sclerosis. Mult Scler. 2018;24(2):96–120. doi: A. Information distortion in physicians’ diagnostic judg- 10.1177/1352458517751049 ments. Med Decis Making. 2012;32:831–9. doi:10.1177/ 26. Rae-Grant A, Day GS, Marrie RA, et al. Practice guideline 0272989X12447241 recommendations summary: disease-modifying therapies 14. Sonnenberg L, Gelsomin E, Levy DE, Riis J, Barraclough for adults with multiple sclerosis: report of the guideline S, Thorndike AN. A traffic light food labeling intervention development, dissemination, and implementation Subcom- increases consumer awareness of health and healthy choices mittee of the American Academy of Neurology. Neurology. at the point-of-purchase. Prev Med. 2013;57(4):253–7. doi: 2018;90(17):777–88. doi:10.1212/WNL.0000000000005347 10.1016/j.ypmed.2013.07.001 27. Prosperini L, Mancinelli CR, De Giglio L, De Angelis F, 15. Murphy MS, Baker M. When used appropriately, NICE’s Barletta V, Pozzilli C. Interferon beta failure predicted by traffic light system for febrile children helps clinicians make EMA criteria or isolated MRI activity in multiple sclerosis. safe and appropriate decisions. BMJ. 2014;348:g2518. doi: Mult Scler. 2014;20(5):566–76. doi:10.1177/1352458513502399 10.1136/bmj.g2518 28. Bermel RA, You X, Foulds P, et al. Predictors of long- 16. Enax L, Krajbich I, Weber B. Salient nutrition labels term outcome in multiple sclerosis patients treated with increase the integration of health attributes in food deci- interferon b. Ann Neurol. 2013;73(1):95–103. doi:10.1002/ sion-making. Judgment Decis Making. 2016;11(5):460–71. ana.23758 17. Enax L, Krapp V, Piehl A, Weber B. Effects of social sus- 29. Sormani MP, Rio J, Tintore M, et al. Scoring treatment tainability signaling on neural valuation signals and taste- response in patients with relapsing multiple sclerosis. Mult experience of food products. Front Behav Neurosci. 2015;9: Scler. 2013;19(5):605–12. doi:10.1177/1352458512460605 247. doi:10.3389/fnbeh.2015.00247 30. Sormani MP, Gasperini C, Romeo M, et al. Assessing 18. Phillips AC, Lewis LK, McEvoy MP, et al. Development response to interferon-b in a multicenter dataset of patients and validation of the guideline for reporting evidence- with MS. Neurology. 2016;87(2):134–40. doi:10.1212/WNL based practice educational interventions and teaching .0000000000002830 (GREET). BMC Med Educ. 2016;16(1):237. doi:10.1186/ 31. Sormani MP, De Stefano N. Defining and scoring response s12909-016-0759-1 to IFN-b in multiple sclerosis. Nat Rev Neurol. 2013;9(9): 19. Saposnik G, Maurino J, Sempere AP, et al. Overcoming 504–12. doi:10.1038/nrneurol.2013.146 therapeutic inertia in multiple sclerosis care: a pilot rando- 32. Tramacere I, Del Giovane C, Salanti G, D’Amico R, Filip- mized trial applying the traffic light system in medical edu- pini G. Immunomodulators and immunosuppressants for cation. Front Neurol. 2017;8:430. doi:10.3389/fneur.2017 relapsing-remitting multiple sclerosis: a network meta-anal- .00430 ysis. Cochrane Database Syst Rev. 2015;(9):CD011381. 20. Saposnik G, Tobler PN, Caceres F, et al. Usability of an doi:10.1002/14651858.CD011381.pub2 educational intervention to overcome therapeutic inertia in 33. Saposnik G, Montalban X. Therapeutic inertia in the new multiple sclerosis care. Front Neurol. 2018;9:522. doi: landscape of multiple sclerosis care. Front Neurol. 2018;9: 10.3389/fneur.2018.00522 174. doi:10.3389/fneur.2018.00174 21. Levy I, Snell J, Nelson AJ, Rustichini A, Glimcher PW. 34. Saposnik G, Montalban X, Selchen D, et al. Therapeutic Neural representation of subjective value under risk and inertia in multiple sclerosis care: a study of Canadian neurolo- ambiguity. J Neurophysiol. 2010;103(2):1036–47. doi: gists. Front Neurol. 2018;9:781. doi:10.3389/fneur.2018.00781 10.1152/jn.00853.2009 35. Albarqouni L, Hoffmann T, Glasziou P. Evidence-based 22. Anderson LR, Mellor JM. Predicting health behaviors with practice educational intervention studies: a systematic an experimental measure of risk preference. J Health Econ. review of what is taught and how it is measured. BMC 2008;27(5):1260–74. doi:10.1016/j.jhealeco.2008.05.011 Med Educ. 2018;18(1):177. doi:10.1186/s12909-018-1284-1 23. Peabody JW, Luck J, Glassman P, Dresselhaus TR, Lee 36. Stangel M, Penner IK, Kallmann BA, Lukas C, Kieseier M. Comparison of vignettes, standardized patients, and BC, Gold R. Multiple Sclerosis Decision Model (MSDM): chart abstraction: a prospective validation study of 3 meth- Entwicklung eines Mehrfaktorenmodells zur Beurteilung ods for measuring quality. JAMA. 2000;283(13):1715–22. des Therapie- und Krankheitsverlaufs bei schubförmiger 24. Cristiano E, Alonso R, Pinheiro AA, et al. Argentinean Multipler Sklerose. Akt Neurol. 2013;40(9):486–93. doi: recommendations on the identification of treatment failure 10.1055/s-0033-1349643