Effect of an Educational Intervention on Therapeutic Inertia in Neurologists With Expertise in Multiple Sclerosis: A Randomized Clinical Trial | Medical Education and Training | JAMA Network Open | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  CONSORT Flow Diagram
CONSORT Flow Diagram

TLS indicates traffic light system (use of established associations between traffic light colors and actions to stop or proceed).

Figure 2.  Study Design and Time Period Illustration
Study Design and Time Period Illustration

A, Participants answered demographic and practice-based questions and provided risk and ambiguity preferences. Next, they listened to simulated case scenarios. Each scenario was followed by 6 therapeutic choices, which remained on the screen until the participant selected 1 of them. After the first 10 simulated case scenarios (pre-intervention), participants were randomized to the intervention or the control group. All participants performed another 10 simulated case scenarios. B, The black dots represent the peak pupil size within each period used to compute pupil responses (pupil peak for each period minus mean baseline at period 0 (T0). Peaks were determined similarly for both groups across all periods and case scenarios. T1 indicates first period (critical clinical information); T2, second period (neurological status of the patient); T3, third period (critical brain imaging information); and T4, fourth period (standardized questions were asked).

Figure 3.  Effects of Intervention and Group Randomization on Pupil Responses
Effects of Intervention and Group Randomization on Pupil Responses

Pupil-linked autonomic arousal responses (peak minus mean baseline) are shown separately for intervention and control groups, stratified by the intervention period. Lower responses in the intervention group extend to T4, in which no critical information was provided, which may suggest that the protective effect of the intervention extends into the period when participants made decisions in the context of uncertainty. T1 indicates first period (critical clinical information); T2, second period (neurological status of the patient); T3, third period (critical brain imaging information); and T4, fourth period (standardized questions were asked).

aP < .01 for the comparison of pupil responses between control and intervention groups.

Table 1.  Baseline Characteristics of Participants
Baseline Characteristics of Participants
Table 2.  Relationship Between Pupil Dilation by Critical Periods and Therapeutic Inertia
Relationship Between Pupil Dilation by Critical Periods and Therapeutic Inertia
1.
Blumenthal-Barby  JS, Krieger  H.  Cognitive biases and heuristics in medical decision making: a critical review using a systematic search strategy.   Med Decis Making. 2015;35(4):539-557. doi:10.1177/0272989X14547740 PubMedGoogle ScholarCrossref
2.
Brück  W, Gold  R, Lund  BT,  et al.  Therapeutic decisions in multiple sclerosis: moving beyond efficacy.   JAMA Neurol. 2013;70(10):1315-1324. doi:10.1001/jamaneurol.2013.3510 PubMedGoogle Scholar
3.
O'Connor  PJ, Sperl-Hillen  JAM, Johnson  PE, Rush  WA, Biltz  G. Clinical inertia and outpatient medical errors. In: Henriksen  K, Battles  JB, Marks  ES, Lewin  DI, eds.  Advances in Patient Safety: From Research to Implementation  (Volume 2: Concepts and Methodology). Agency for Healthcare Research and Quality;2005. doi:10.1037/e448252006-001
4.
Mohan  AV, Phillips  LS.  Clinical inertia and uncertainty in medicine.   JAMA. 2011;306(4):383. doi:10.1001/jama.2011.1044 PubMedGoogle ScholarCrossref
5.
Okonofua  EC, Simpson  KN, Jesri  A, Rehman  SU, Durkalski  VL, Egan  BM.  Therapeutic inertia is an impediment to achieving the Healthy People 2010 blood pressure control goals.   Hypertension. 2006;47(3):345-351. doi:10.1161/01.HYP.0000200702.76436.4b PubMedGoogle ScholarCrossref
6.
Saposnik  G, Sempere  AP, Prefasi  D,  et al.  Decision-making in multiple sclerosis: the role of aversion to ambiguity for therapeutic inertia among neurologists (DIScUTIR MS).   Front Neurol. 2017;8:65. doi:10.3389/fneur.2017.00065 PubMedGoogle Scholar
7.
Almusalam  N, Oh  J, Terzaghi  M,  et al.  Comparison of physician therapeutic inertia for management of patients with multiple sclerosis in Canada, Argentina, Chile, and Spain.   JAMA Netw Open. 2019;2(7):e197093. doi:10.1001/jamanetworkopen.2019.7093 PubMedGoogle Scholar
8.
Burks  J, Marshall  TS, Ye  X.  Adherence to disease-modifying therapies and its impact on relapse, health resource utilization, and costs among patients with multiple sclerosis.   Clinicoecon Outcomes Res. 2017;9:251-260. doi:10.2147/CEOR.S130334 PubMedGoogle ScholarCrossref
9.
Saposnik  G, Mamdani  M, Montalban  X,  et al.  Traffic lights intervention reduces therapeutic inertia: a randomized controlled trial in multiple sclerosis care.   MDM Policy Pract. 2019;4(1):2381468319855642. doi:10.1177/2381468319855642 PubMedGoogle Scholar
10.
Freedman  MS, Selchen  D, Prat  A, Giacomini  PS.  Managing multiple sclerosis: treatment initiation, modification, and sequencing.   Can J Neurol Sci. 2018;45(5):489-503. doi:10.1017/cjn.2018.17PubMedGoogle ScholarCrossref
11.
Rae-Grant  A, Day  GS, Marrie  RA,  et al.  Practice guideline recommendations summary: disease-modifying therapies for adults with multiple sclerosis: report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology.   Neurology. 2018;90(17):777-788. doi:10.1212/WNL.0000000000005347 PubMedGoogle ScholarCrossref
12.
Ontaneda  D, Cohn  S, Fox  R.  Risk stratification and mitigation in multiple sclerosis.   Mult Scler Relat Disord. 2014;3(5):639-649. doi:10.1016/j.msard.2014.05.003 PubMedGoogle ScholarCrossref
13.
Ontaneda  D, Tallantyre  E, Kalincik  T, Planchon  SM, Evangelou  N.  Early highly effective versus escalation treatment approaches in relapsing multiple sclerosis.   Lancet Neurol. 2019;18(10):973-980. doi:10.1016/S1474-4422(19)30151-6 PubMedGoogle ScholarCrossref
14.
Urai  AE, Braun  A, Donner  TH.  Pupil-linked arousal is driven by decision uncertainty and alters serial choice bias.   Nat Commun. 2017;8:14637. doi:10.1038/ncomms14637 PubMedGoogle ScholarCrossref
15.
Mathôt  S.  Pupillometry: psychology, physiology, and function.   J Cogn. 2018;1(1):16. doi:10.5334/joc.18 PubMedGoogle ScholarCrossref
16.
Maier  ME, Ernst  B, Steinhauser  M.  Error-related pupil dilation is sensitive to the evaluation of different error types.   Biol Psychol. 2019;141:25-34. doi:10.1016/j.biopsycho.2018.12.013 PubMedGoogle ScholarCrossref
17.
Wessel  JR, Dolan  KA, Hollingworth  A.  A blunted phasic autonomic response to errors indexes age-related deficits in error awareness.   Neurobiol Aging. 2018;71:13-20. doi:10.1016/j.neurobiolaging.2018.06.019 PubMedGoogle ScholarCrossref
18.
Enax  L, Hu  Y, Trautner  P, Weber  B.  Nutrition labels influence value computation of food products in the ventromedial prefrontal cortex.   Obesity (Silver Spring). 2015;23(4):786-792. doi:10.1002/oby.21027 PubMedGoogle ScholarCrossref
19.
Murphy  MS, Baker  M.  When used appropriately, NICE’s traffic light system for febrile children helps clinicians make safe and appropriate decisions.   BMJ. 2014;348:g2518. doi:10.1136/bmj.g2518 PubMedGoogle ScholarCrossref
20.
Zhang  X, Liu  Y, Gu  Y, Wang  S, Chen  H.  Red for “stop”: “traffic-light” nutrition labels decrease unhealthy food choices by increasing activity and connectivity in the frontal lobe.   Nutrients. 2020;12(1):E128. doi:10.3390/nu12010128 PubMedGoogle Scholar
21.
Enax  L, Krajbich  I, Weber  B.  Salient nutrition labels increase the integration of health attributes in food decision-making.   Judgm Decis Mak. 2016;11(5):460-471.Google Scholar
22.
Levy  I, Snell  J, Nelson  AJ, Rustichini  A, Glimcher  PW.  Neural representation of subjective value under risk and ambiguity.   J Neurophysiol. 2010;103(2):1036-1047. doi:10.1152/jn.00853.2009 PubMedGoogle ScholarCrossref
23.
Anderson  LR, Mellor  JM.  Predicting health behaviors with an experimental measure of risk preference.   J Health Econ. 2008;27(5):1260-1274. doi:10.1016/j.jhealeco.2008.05.011 PubMedGoogle ScholarCrossref
24.
Sormani  MP, Gasperini  C, Romeo  M,  et al; MAGNIMS study group.  Assessing response to interferon-β in a multicenter dataset of patients with MS.   Neurology. 2016;87(2):134-140. doi:10.1212/WNL.0000000000002830 PubMedGoogle ScholarCrossref
25.
Murphy  PR, Vandekerckhove  J, Nieuwenhuis  S.  Pupil-linked arousal determines variability in perceptual decision making.   PLoS Comput Biol. 2014;10(9):e1003854. doi:10.1371/journal.pcbi.1003854 PubMedGoogle Scholar
26.
Privitera  CM, Renninger  LW, Carney  T, Klein  S, Aguilar  M.  Pupil dilation during visual target detection.   J Vis. 2010;10(10):3-3. doi:10.1167/10.10.3 PubMedGoogle ScholarCrossref
27.
Antonakis  J, Bendahan  S, Jacquart  P, Lalive  R.  On making causal claims: a review and recommendations.   The Leadership Quarterly. 2010;21(6):1086-1120. doi:10.1016/j.leaqua.2010.10.010Google ScholarCrossref
28.
MacKinnon  DP, Fairchild  AJ, Fritz  MS.  Mediation analysis.   Annu Rev Psychol. 2007;58(1):593-614. doi:10.1146/annurev.psych.58.110405.085542 PubMedGoogle ScholarCrossref
29.
VanderWeele  TJ.  Mediation analysis: a practitioner’s guide.   Annu Rev Public Health. 2016;37(1):17-32. doi:10.1146/annurev-publhealth-032315-021402 PubMedGoogle ScholarCrossref
30.
de Gee  JW, Knapen  T, Donner  TH.  Decision-related pupil dilation reflects upcoming choice and individual bias.   Proc Natl Acad Sci U S A. 2014;111(5):E618-E625. doi:10.1073/pnas.1317557111 PubMedGoogle ScholarCrossref
31.
Khunti  K, Gomes  MB, Pocock  S,  et al.  Therapeutic inertia in the treatment of hyperglycaemia in patients with type 2 diabetes: a systematic review.   Diabetes Obes Metab. 2017.PubMedGoogle Scholar
32.
Ogura  M, Harada-Shiba  M.  Clinical inertia in the management of hypercholesterolemia: what clinicians need to do.   J Atheroscler Thromb. 2016;23(5):552-553. doi:10.5551/jat.ED037 PubMedGoogle ScholarCrossref
33.
Cooke  CE, Sidel  M, Belletti  DA, Fuhlbrigge  AL.  Review: clinical inertia in the management of chronic obstructive pulmonary disease.   COPD. 2012;9(1):73-80. doi:10.3109/15412555.2011.631957 PubMedGoogle ScholarCrossref
34.
Saposnik  G, Sempere  AP, Raptis  R, Prefasi  D, Selchen  D, Maurino  J.  Decision making under uncertainty, therapeutic inertia, and physicians’ risk preferences in the management of multiple sclerosis (DIScUTIR MS).   BMC Neurol. 2016;16:58. doi:10.1186/s12883-016-0577-4 PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    Medical Education
    December 16, 2020

    Effect of an Educational Intervention on Therapeutic Inertia in Neurologists With Expertise in Multiple Sclerosis: A Randomized Clinical Trial

    Author Affiliations
    • 1Division of Neurology, Department of Medicine, St Michael’s Hospital, University of Toronto, Toronto, Ontario, Canada
    • 2Zurich Center for Neuroeconomics, Department of Economics, University of Zurich, Zurich, Switzerland
    • 3Clinical Outcomes and Decision Neuroscience Unit, St Michael’s Hospital, University of Toronto, Toronto, Ontario, Canada
    • 4Li Ka Shing Knowledge Institute, St Michael’s Hospital, University of Toronto, Toronto, Ontario, Canada
    • 5Applied Health Research Centre, Centre for Urban Health Solutions, St Michael’s Hospital, Toronto, Toronto, Ontario, Canada
    JAMA Netw Open. 2020;3(12):e2022227. doi:10.1001/jamanetworkopen.2020.22227
    Key Points

    Question  How do physicians handle uncertainty when making live therapeutic decisions?

    Findings  In this randomized clinical trial of 34 neurologists from Canada, an educational intervention showed a significant 31% reduction in therapeutic inertia compared with the control group. Pupil dilation, a marker of autonomic arousal, mediated the effect of the educational intervention on therapeutic inertia.

    Meaning  In this study, the educational intervention helped facilitate optimal treatment choices by decreasing therapeutic inertia in neurologists by reducing autonomic arousal responses; further study in other specialties is warranted.

    Abstract

    Importance  Therapeutic inertia (TI) is the failure to escalate therapy when treatment goals are unmet and is associated with low tolerance to uncertainty and aversion to ambiguity in physician decision-making. Limited information is available on how physicians handle therapeutic decision-making in the context of uncertainty.

    Objective  To evaluate whether an educational intervention decreases TI by reducing autonomic arousal response (pupil dilation), a proxy measure of how physicians respond to uncertainty during treatment decisions.

    Design, Setting, and Participants  In this randomized clinical trial, 34 neurologists with expertise in multiple sclerosis (MS) practicing at 15 outpatient MS clinics in academic and community institutions from across Canada were enrolled. Participants were randomly assigned to receive an educational intervention that facilitates treatment decisions (active group) or to receive no exposure to the intervention (usual care [control group]) from December 2017 to March 2018. Participants listened to 20 audio-recorded simulated case scenarios as pupil responses were assessed by eye trackers. Autonomic arousal was assessed as pupil dilation in periods in which critical information was provided (first period [T1]: clinical data, second period [T2]: neurologic status, and third period [T3]: magnetic resonance imaging data). Data were analyzed from September 2018 to March 2020.

    Interventions  The traffic light system (TLS)–based educational intervention vs usual care (unexposed). The TLS (use of established associations between traffic light colors and actions to stop or proceed) assists participants in identifying factors associated with worse prognosis in MS care, thereby facilitating the treatment decision-making process by use of established associations between red, green, and yellow colors and risk levels, and actions (treatment decisions).

    Main Outcomes and Measures  Pupil assessment was the primary autonomic outcome. To test the treatment effect of the educational intervention (TLS), difference-in-differences models (also called untreated control group design with pretest and posttest) were used.

    Results  Of 38 eligible participants, 34 (89.4%) neurologists completed the study. The mean (SD) age was 44.6 (11.6) years; 38.3% were female and 20 (58.8%) were MS specialists. Therapeutic inertia was present in 50.0% (17 of 34) of all participants and was associated with greater pupil dilation. For every additional SD of pupil dilation, the odds of TI increased by 51% for T1 (odds ratio, 1.51; 95% CI, 1.12-2.03), by 31% for T2 (odds ratio, 1.31; 95% CI, 1.08-1.59), and by 49% for T3 (odds ratio, 1.49; 95% CI, 1.13-1.97). The intervention significantly reduced TI (risk reduction, 31.5%; 95% CI, 16.1%-47.0%). Autonomic arousal responses mediated 29.0% of the effect of the educational intervention on TI.

    Conclusions and Relevance  In this randomized clinical trial, the TLS intervention decreased TI as measured by pupil dilation, which suggests that individual autonomic arousal is an indicator of how physicians handle uncertainty when making live therapeutic decisions. Pupil response, a biomarker of TI, may eventually be useful in medical education.

    Trial Registration  ClinicalTrials.gov Identifier: NCT03134794

    Introduction

    Therapeutic decision-making requires an individualized balance of the safety and efficacy profiles of different agents with either imperfect information or uncertain response of that choice.1,2 One outcome in decision-making in the context of uncertainty is therapeutic inertia (TI). Therapeutic inertia is characterized by suboptimal decision-making not to initiate or intensify treatment when treatment goals are unmet.3-5 Therapeutic inertia affects 60% to 90% of physicians caring for patients with chronic conditions (eg, hypertension, diabetes, or multiple sclerosis [MS]).4-7 Suboptimal decision-making is associated with worse clinical outcomes and higher health care costs.3-5,8 A randomized clinical trial on MS care reported a 70% reduction of TI in neurologists, using a short (<5 minutes) and simple (application of the traffic light system [use of established associations between traffic light colors and actions to stop or proceed]) educational intervention.9 For MS care, overcoming TI corresponds to appropriately switching from a first-line agent (eg, glatiramer or interferon) to a high-efficacy treatment (eg, fingolimod or monoclonal antibodies) when given both clinical and radiologic evidence of disease progression.9-13

    Recent studies reported that decision-making in the context of uncertainty is associated with autonomic arousal, as measured by pupil dilation.14,15 In particular, phasic pupil size increases are associated with suboptimal or erroneous decision-making involving high uncertainty.16,17 However, the association between autonomic arousal and therapeutic decision-making in the context of uncertainty is unknown. A better understanding of pupil dilation as a marker of an autonomic arousal response and TI may facilitate the development of decision-aid tools or other educational interventions to overcome suboptimal or erroneous decision-making in the context of other chronic medical conditions (ie, hypertension, diabetes, and dyslipidemia).

    In this randomized clinical trial, we investigated (1) the relation between autonomic arousal responses and TI, (2) how a previously tested and effective educational intervention9 affects autonomic arousal responses and TI, and (3) whether autonomic arousal responses mediate the association between the educational intervention and TI. We used MS care as an appropriate model for complex therapeutic decision-making arising in the management of chronic medical conditions using a previously tested and effective educational intervention to reduce TI.9 Consistent with the details in the protocol for the Canadian study, we hypothesized that our educational intervention would decrease practitioner uncertainty about therapeutic choices, as reflected in decreased autonomic arousal responses, and thereby lead to an improvement in therapeutic decision-making.14,16,17

    Methods
    Study Design

    In the initial study conducted in Argentina, 90 neurologists who provided care to patients with MS were randomly assigned to the TLS intervention (n = 45) or to the control group (n = 45). A 70% reduction in TI after the TLS intervention was found compared with controls (odds ratio, 0.30; 95% CI, 0.10-0.89). In the present study using a similar design, we added the assessment of pupil responses to determine how the TLS would decrease uncertainty and TI. We conducted a randomized clinical trial of MS experts and general neurologists who care for patients with MS practicing at 15 outpatient MS clinics in academic and community institutions from across Canada. We randomly assigned participants to an educational intervention group or to a usual care control group from December 2017 to March 2018. The educational intervention used the traffic light system (TLS) to reduce TI in the management of MS. The control group made therapeutic decision-making without being exposed to the TLS intervention, in line with current standard practice. Randomized group assignment and allocation concealment were controlled by Qualtrics. Participants were not aware to which group they were randomized. Investigators were also blinded to the treatment allocation. The mean (SD) time of study completion was 44.9 (6.7) minutes and participants received CAD$450 (equivalent to $350). The trial protocol is available in Supplement 1. The protocol was amended on October 25, 2017, to add pupil measures as a primary outcome. This amendment was approved by the Research Ethics Board of St Michael’s (which became Unity Health Toronto in May 2018). Data were analyzed from September 2018 to March 2020. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline for randomized clinical trials. Details are available at ClinicalTrials.gov (NCT03134794) and elsewhere.9

    Inclusion Criteria and Participants

    Neurologists actively caring for patients with MS across Canada were invited to participate by email sent from the Canadian Network of MS Clinics and NeuroSens (Lind Publishing Inc) (Figure 1). These networks comprise most MS neurologists in Canada. Participants were recruited from December 13, 2017, to March 2, 2018. Participants who completed a postresidency MS fellowship were classified as MS specialists. Physicians seeing less than 1 patient with MS per month were excluded from the study. Participants provided written informed consent. The study was approved by the Research Ethics Board of St Michael’s Hospital, University of Toronto.

    Educational Intervention
    The Traffic Light System

    We applied a previously proven effective TLS intervention9 to facilitate the identification of patients at high risk of disease progression (based on clinical and imaging evidence of disease progression) introduced by case scenarios. The TLS supports the decision-making process by use of established associations between traffic light colors and actions to stop or proceed.9,18-20 For example, in the context of this intervention, a red light represents high risk and triggers a stop and think action, whereas a green light represents low risk and a triggers a continue the same strategy action. A previous study reported that the TLS can interrupt automatic behavior and lead to more optimal decision-making.21 In our MS care model, the TLS aims for the red traffic light to indicate a warning sign of disease progression and a switch from a low-efficacy agent (eg, interferon or glatiramer) to a more effective disease modifying treatment (eg, monoclonal antibodies).10-12 The TLS aims for the green traffic light to indicate stability in a patient and following a good clinical course (eg, no relapse and stable activity on brain imaging), therefore requiring no immediate therapeutic changes.

    Data Collection and Study Flow

    The study progressed as follows: (1) collection of demographic and practice-based information from participants, (2) participant completion of behavioral experimental procedures, and (3) participant completion of 20 simulated and standardized case scenarios (10 before and 10 after the intervention). We used simulated case scenarios reflecting common situations in clinical practice that were previously designed and validated by our research team (G.S. and J.O.) and MS experts, 16 of whom directly assessed the presence of TI.9 All simulated case scenarios were presented auditorily (via headphones connected to the computer) to avoid interference of visual stimulation and automatic eye movements with pupil responses. The mean (SD) duration of case scenarios was 35.4 (7.1) seconds (range, 27-50 seconds). For each simulated case scenario, we identified 3 periods during which critical information was provided (eTable 1 in Supplement 2). The first period (T1) provided critical clinical information (present and previous clinical relapses, type of relapse, and/or symptoms). The second period (T2) informed about the neurological status of the patient (Expanded Disability Status Scale). The third period (T3) provided critical brain imaging information (number and nature of new lesions, gadolinium-enhancing lesions). The period before the start of case scenarios (T0) served as baseline, while the fourth period (T4) represented the final segment in which standardized questions (eg, “What would you do? Please select one of the options”) were asked in preparation to the treatment options (Figure 2A). Compared with the baseline T0, we expected an autonomic arousal response during critical information periods T1 through T3 and little response during T4.

    Based on previously reported associations of risk and ambiguity aversion with TI,6,7 we considered also their relation to autonomic arousal responses and TI, using established measures in the financial domain.22,23 Briefly, ambiguity aversion is defined as a dislike for events with unknown probability compared with events with known probability.22Ambiguity aversion was assessed asking participants to choose between a known 50/50 option (an urn with equal number of blue and red balls) providing $400 or $0 and an option with unknown probability of the same outcomes. Risk aversion was assessed by asking participants to indicate the minimal certain payoff they would prefer over a gamble with a 50/50 chance of winning $400 or $06 (eAppendix in Supplement 2).

    Outcome Measures

    Therapeutic inertia is defined as lack of treatment escalation despite evidence of disease progression. We defined disease progression as the combination of a clinical relapse plus the presence of 5 or more new lesions (T2 or fluid-attenuated inversion recovery sequences) or at least 1 gadolinium-enhancing lesion in follow-up magnetic resonance imaging scans.10,11 Using the combination of clinical relapse and magnetic resonance imaging activity is consistent with evidence regarding the risk of treatment failure in patients receiving interferon beta.24 For each of our case scenarios, we determined TI as a binary variable (present vs absent; primary outcome). A secondary outcome included TI greater than or equal to 25% of responses, meaning that participants did not escalate treatment when recommended in at least 1 of 4 simulated case scenarios.9-12

    Experimental Procedures

    The study was conducted in an ambulatory clinic–type setting to increase ecological validity. Room temperature, light conditions (100 lumens), and participant sitting positions were held constant. We calculated z scores for pupil time-series scored within each participant to allow comparison of pupil dilation between and within simulated case scenarios, critical periods, and participants. The mean pupil size (measured at T0, ie, 1500 ms - 500 ms before scenario onset) was taken as pupil baseline.25 For each simulated case scenario, we estimated autonomic arousal responses by subtracting the mean baseline pupil diameter from the peak pupil dilation during each critical period (T1-T3) (Figure 2B).15,26 Further details are available in the eAppendix in Supplement 2.

    Sample Size

    Our study was underpowered to evaluate differences in TI between groups.9 The observed 15.5% absolute difference between groups (unadjusted results)9 would require a sample size of a minimum of 78 participants per group with an α = .05 and β = .2. The power to determine differences in phasic pupil response in relation to TI was 99%.

    Statistical Analysis

    We applied 3 analytical approaches: (1) comparison of autonomic arousal responses across critical periods, (2) treatment-effect analysis evaluating the association between autonomic arousal responses and TI, and (3) a mediation analysis to assess how the association between individual participant characteristics and TI may be mediated by autonomic arousal responses. For the first approach, we used nonparametric tests (Wilcoxon rank sum test for continuous variables and Mann-Whitney test for categorical variables). For the second approach, we compared high vs low arousal between groups stratified by intervention period (preintervention vs postintervention). We used generalized estimating equations to assess relationships between the variables of interest with TI accounting for clustering (repeated observations on participants) across all TI scenarios for each time period. This analysis controlled for the predefined explanatory variables age, specialist status (MS expert vs general neurologist), years of practice, risk preferences, and ambiguity aversion as identified in previous research.6 To test the treatment effect of the educational intervention (TLS), we used difference-in-differences models (also called untreated control group design with pretest and posttest).27 This model allowed us to measure the treatment effect of our intervention by comparing the change over time (posttest minus pretest performance) between the intervention and control group. For the third approach, pupil dilation for each participant and case scenario was tested as a mediator. Thus, the mediation analysis evaluated whether individual autonomic arousal responsivity accounted for the benefits of the educational intervention on TI. Mediation analysis is a technique commonly used in the social sciences to explain a relationship between an independent variable (eg, demographic variables) and an outcome via a third variable (called mediator).28,29 We measured whether the effect of our intervention on TI was mediated by pupil-indexed autonomic arousal.

    All tests were 2-tailed, and P < .05 wase considered significant. We used Stata version 13 (StataCorp LLP) and SAS version 9.4 (SAS Institute Inc) to conduct all analyses.

    Results
    Participant Characteristics

    Of 38 eligible participants, 34 (89.4%) neurologists completed the study. The mean (SD) age was 44.6 (11.6) years; 38.3% were female and 20 (58.8%) were MS specialists. Participants had a mean (SD) of 12.5 (12) years of experience and assessed a mean (SD) of 23.1 (16) patients with MS per week. Of the 34 participants who completed the study, pupillary data were available in 30 (88.2%) participants. Two participants in each group had incomplete or missing pupil data (Figure 1). Table 1 presents baseline characteristics of the study population.

    Participants showed risk-neutrality in our measures of risk attitudes and the minimal safe amounts participants preferred over the 50/50 gamble did not differ between groups (Mann-Whitney P = .14; mean [SD] control: $175.7 [$45.2]; mean [SD] intervention: 211 [78.8]). Nineteen (55.9%) participants showed aversion to ambiguity in the financial domain. There was no difference in baseline pupil data between groups for each simulated case scenario (2.82 mm vs 2.96 mm; Mann-Whitney P = .57).

    TI was present in 50.0% (17 of 34) of participants in at least 1 case scenario, representing 7.7% (42 of 544) of all individual responses that assessed TI. Non-MS experts had higher prevalence of TI compared with MS experts (11.5% vs 5.2%; P = .01).

    Arousal Responses and TI

    Overall, pupil size increased for each period relative to baseline (F test of overall significance indicating whether a linear regression model provides a better fit to the data than a model that contains no independent variables, all P < .001). The results remained robust after adjustment for the prespecified covariates (P < .001).

    The multivariate analysis adjusted for age, sex, MS expertise, risk preferences, ambiguity aversion, preintervention vs postintervention period, and intervention group by period (Table 2) showed that pupil dilation was positively related to TI for all critical periods (T1 to T3). For every additional SD of pupil dilation, the odds of TI increased by 51% for T1 (odds ratio, 1.51; 95% CI, 1.12-2.03), by 31% for T2 (odds ratio, 1.31; 95% CI, 1.08-1.59) and by 49% for T3 (odds ratio, 1.49; 95% CI, 1.13-1.97). As expected, there was no association between pupil dilation and TI for T4 (odds ratio, 1.07; 95% CI, 0.86-1.34) as no critical information was provided, which may suggest a relation between TI and the arousal elicited by critical clinical information.

    Our results were robust to using secondary outcome measures (Table 2). Together, stronger autonomic arousal responses at critical periods were associated with stronger TI. For TI equal to or greater than 25%: for T1 (OR, 1.53; 95% CI, 1.11-2.12), for T2 (OR, 1.33; 95% CI, 1.07-1.63), for T3 (OR, 1.51; 95% CI, 1.13-2.00).

    Group Allocation and Autonomic Arousal Responses

    Pupil size did not differ significantly between the TLS and control groups before the intervention but did so after the intervention for T2, T3, and T4 (Figure 3). Overall, the multivariate analysis showed that in the postintervention period, participants in the control group had significantly enlarged pupils (as a continuous variable) compared with the intervention group for T2 (β, −0.46; 95% CI, −0.90 to −0.01; P = .049), T3 (β, −0.63; 95% CI, −1.04 to −0.22; P = .004), and T4 (β, −0.75; 95% CI, −1.12 to −0.37; P < .001) (eTables 2 and 3 in Supplement 2). No difference was observed for T1 (β, −0.23; 95% CI, −0.64 to 0.19; P = .47). The adjusted analysis for the dichotomized pupil response (maximum-peak minus mean-baseline ≥0.1 difference in z score as a high autonomic arousal vs <0.1 difference in z score low autonomic arousal response) showed similar results (hazard ratio control [reference category], 10.1 [95% CI, 8.2-12.0] vs hazard ratio TLS, 7.2 [95% CI, 6.2-8.1] vs LR control, 4.5 [95% CI, 1.6-7.5] vs LR TLS, 3.2 [95% CI, 0.8-5.5]; P < .01]) (eFigure 1 in Supplement 2). Together, these data suggest that the educational intervention has a protective effect, which extends into the period when participants made decisions in the context of uncertainty.

    Educational Intervention and TI

    In a previous study,9 the educational intervention showed a significant reduction in TI. To assess the presence of a treatment effect on individual TI, we used the difference-in-differences analytical strategy (eTable 2 in Supplement 2). We found that participants in the educational intervention group had a significant reduction in TI (risk reduction, 31.5%; 95% CI, 16.1%-47.0%) compared with the control group (eFigure 2 in Supplement 2). Linear regression analysis adjusted for participant age, sex, expertise, risk preference, and pupil dilation showed that for every therapeutic decision, there was a significant TI decrease of 5.0% (−5.0%; 95% CI, −0.8% to −9.3%) in the intervention group. There was no interaction between the education intervention and MS expertise (β, −0.518; 95% CI, −3.936 to −2.901; P = .23). The difference-in-differences analysis revealed no evidence for confounding endogenous effects (eTable 4 in Supplement 2). Together, these data replicate the previous findings that the TLS intervention reduces TI.9

    Autonomic Arousal Response, Mediation, Educational Intervention, and TI

    Autonomic arousal responses mediated 29.0% of the total effect of the educational intervention on TI (eFigure 3 in Supplement 2). The direct effect of the educational intervention on TI was −3.5% (95% CI, −9.2% to 3.5%). Other factors (eg, age, sex, risk preference) had a nonsignificant or a negligible effect. Further details are in the eAppendix in Supplement 2.

    Discussion

    To date, the role of autonomic arousal for therapeutic decision making remains unexplored. In the present study, we addressed this gap in the framework of therapeutic MS care decision-making, with a focus on decisions not to escalate treatment when recommended by best practice guidelines (ie, TI). We analyzed pupil dilation as a marker of autonomic arousal14,15 and found that both continuously measured pupil dilation and dichotomized high vs low pupil responses were associated with TI. For every additional SD of pupil dilation, the odds of TI increased by 31% to 51% depending on the clinical information being provided. Even though our study had low statistical power and Canadian neurologists show comparatively little TI,7 we estimated that participants in the control group would have reduced TI by almost a third if they would have been randomized to the intervention group. Our data suggest that the intervention may ameliorate TI by reducing autonomic arousal responses to critical information. Pupil dilation mediated the effects of the educational intervention on TI (explaining 29% of the total mediated effect).

    Physician uncertainty may be a factor in rapid pupil-linked autonomic arousal responses, affecting behavioral choices14 and the updating of beliefs with presented evidence.30 Our educational intervention may reduce autonomic arousal by reducing uncertainty and thereby facilitating alternative behavioral strategies. Specifically, the warning function of a red traffic light may help emphasize the need for switching to a more effective agent 9 and may concurrently boost the physician’s confidence in the therapeutic decision.

    Relevance for Clinical Practice

    Therapeutic inertia commonly affects physicians caring for patients with chronic medical conditions such as MS, diabetes, and hypertension.31-33 This study may help increase the understanding of how physicians make therapeutic decisions in the context of uncertainty. Critical clinical information increases autonomic arousal and stronger autonomic arousal responses are associated with suboptimal therapeutic decision-making (ie, TI). Moreover, the inertia-reducing effects of our educational intervention appear to be mediated by reduced autonomic arousal responses. We may now be able to identify physicians who are making frequent suboptimal decisions (>25%) and to estimate potential reductions in TI from novel educational interventions. Furthermore, autonomic arousal responses may serve as a marker of the effectiveness of those education interventions. This marker is unaffected by demand effects or cognitive biases. Consequently, our findings may help create other avenues to tailor educational interventions and formal risk-assessment training to decision-makers (medical students, family doctors, and specialists). This approach may help optimize treatment decisions for other more prevalent chronic diseases and lead to improved medical education and better patient outcomes.

    Limitations

    This study has limitations. First, pupil size is not a standard measure in clinical practice. Second, autonomic arousal responses may have different triggers and effects than the ones we tested. Third, given our small sample size, imbalances between groups may lead to residual confounding. We used a statistical approach (eg, generalized estimating equations and mixed models) to address these imbalances. Fourth, simulated case scenarios may not truly reflect therapeutic decision-making in clinical practice. Fifth, the TLS we used is just 1 example of an intervention suitable for MS care. Other interventions may be needed and more effective for the management of other prevalent acute and chronic conditions.

    Despite these limitations, our conclusions are strengthened by a randomized intervention design showing an autonomic arousal–mediated link between an effective therapeutic intervention and therapeutic decision-making made by physicians who care for patients with MS (eFigure 3D in Supplement 2).

    Conclusions

    Therapeutic inertia is a common phenomenon in clinical care, associated with worse outcomes and higher health care costs.6,8,34 The TLS education intervention appeared to help facilitate optimal therapeutic choices by ameliorating the uncertainty associated with clinical information and decreasing TI. Autonomic arousal responses represent therapeutic uncertainty, which was significantly decreased among physicians randomized to the TLS intervention. Our findings may have practical implications for medical education, therapeutic decision-making, and patient outcomes.

    Back to top
    Article Information

    Accepted for Publication: August 17, 2020.

    Published: December 16, 2020. doi:10.1001/jamanetworkopen.2020.22227

    Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License. © 2020 Saposnik G et al. JAMA Network Open.

    Corresponding Author: Gustavo Saposnik, MD, MPH, PhD, Division of Neurology, Department of Medicine, St Michael’s Hospital, University of Toronto, 55 Queen St E, Toronto, ON M5C 1R6, Canada (gustavo.saposnik@unityhealth.to).

    Author Contributions: Drs Saposnik and Nisenbaum had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Tobler and Ruff contributed equally.

    Concept and design: Saposnik, Ruff, Tobler.

    Acquisition, analysis, or interpretation of data: All authors.

    Drafting of the manuscript: Saposnik, Grueschow, Terzaghi, Kostyrko, Nisenbaum, Ruff, Tobler.

    Critical revision of the manuscript for important intellectual content: Saposnik, Grueschow, Oh, Vaidyanathan, Nisenbaum, Ruff, Tobler.

    Statistical analysis: Saposnik, Grueschow, Vaidyanathan, Nisenbaum, Ruff.

    Obtained funding: Saposnik, Tobler.

    Administrative, technical, or material support: Saposnik, Grueschow, Terzaghi, Kostyrko, Ruff, Tobler.

    Supervision: Grueschow, Ruff, Tobler.

    Conflict of Interest Disclosures: Dr Saposnik is supported by the Heart and Stroke Foundation Career Scientist Awards following an open peer-reviewed competition. Dr Saposnik has received personal compensation for consulting or CME activities from Hoffman LaRoche, Celgene, and Servier Canada. Dr Oh has received research funding from the MS Society of Canada, the National MS Society, Brain Canada, Sanofi-Genzyme, Roche, and Biogen. Dr Oh has received personal compensation for consulting or speaking from EMD-Serono, Genzyme, Biogen, Novartis, Celgene, and Roche. Dr Ruff received research funding from the Swiss National Science Foundation (grant 100019L_173248). Dr Tobler received research funding from the Swiss National Science Foundation (grants PP00P1_150739 and 100014_165884) and Pfizer. No other disclosures were reported.

    Funding/Support: The study was sponsored by the Division of Neurology at the University of Toronto with the collaboration of NeuroSens and funded through an unrestricted grant from Roche Canada. Ocrelizumab (manufactured by Roche) was not included as a therapeutic option in any of the simulated case scenarios.

    Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    Data Sharing Statement: See Supplement 3.

    Additional Contributions: The authors thank all colleague neurologists participating in the study and to Kevin Thorpe, MMath (University of Toronto), for suggestions regarding the statistical analysis and to Roula Raptis (St Michael’s Hospital) for the administrative support. We appreciate the support from the Department of Economics at the University of Zurich by facilitating access to the Qualtrics platform.

    References
    1.
    Blumenthal-Barby  JS, Krieger  H.  Cognitive biases and heuristics in medical decision making: a critical review using a systematic search strategy.   Med Decis Making. 2015;35(4):539-557. doi:10.1177/0272989X14547740 PubMedGoogle ScholarCrossref
    2.
    Brück  W, Gold  R, Lund  BT,  et al.  Therapeutic decisions in multiple sclerosis: moving beyond efficacy.   JAMA Neurol. 2013;70(10):1315-1324. doi:10.1001/jamaneurol.2013.3510 PubMedGoogle Scholar
    3.
    O'Connor  PJ, Sperl-Hillen  JAM, Johnson  PE, Rush  WA, Biltz  G. Clinical inertia and outpatient medical errors. In: Henriksen  K, Battles  JB, Marks  ES, Lewin  DI, eds.  Advances in Patient Safety: From Research to Implementation  (Volume 2: Concepts and Methodology). Agency for Healthcare Research and Quality;2005. doi:10.1037/e448252006-001
    4.
    Mohan  AV, Phillips  LS.  Clinical inertia and uncertainty in medicine.   JAMA. 2011;306(4):383. doi:10.1001/jama.2011.1044 PubMedGoogle ScholarCrossref
    5.
    Okonofua  EC, Simpson  KN, Jesri  A, Rehman  SU, Durkalski  VL, Egan  BM.  Therapeutic inertia is an impediment to achieving the Healthy People 2010 blood pressure control goals.   Hypertension. 2006;47(3):345-351. doi:10.1161/01.HYP.0000200702.76436.4b PubMedGoogle ScholarCrossref
    6.
    Saposnik  G, Sempere  AP, Prefasi  D,  et al.  Decision-making in multiple sclerosis: the role of aversion to ambiguity for therapeutic inertia among neurologists (DIScUTIR MS).   Front Neurol. 2017;8:65. doi:10.3389/fneur.2017.00065 PubMedGoogle Scholar
    7.
    Almusalam  N, Oh  J, Terzaghi  M,  et al.  Comparison of physician therapeutic inertia for management of patients with multiple sclerosis in Canada, Argentina, Chile, and Spain.   JAMA Netw Open. 2019;2(7):e197093. doi:10.1001/jamanetworkopen.2019.7093 PubMedGoogle Scholar
    8.
    Burks  J, Marshall  TS, Ye  X.  Adherence to disease-modifying therapies and its impact on relapse, health resource utilization, and costs among patients with multiple sclerosis.   Clinicoecon Outcomes Res. 2017;9:251-260. doi:10.2147/CEOR.S130334 PubMedGoogle ScholarCrossref
    9.
    Saposnik  G, Mamdani  M, Montalban  X,  et al.  Traffic lights intervention reduces therapeutic inertia: a randomized controlled trial in multiple sclerosis care.   MDM Policy Pract. 2019;4(1):2381468319855642. doi:10.1177/2381468319855642 PubMedGoogle Scholar
    10.
    Freedman  MS, Selchen  D, Prat  A, Giacomini  PS.  Managing multiple sclerosis: treatment initiation, modification, and sequencing.   Can J Neurol Sci. 2018;45(5):489-503. doi:10.1017/cjn.2018.17PubMedGoogle ScholarCrossref
    11.
    Rae-Grant  A, Day  GS, Marrie  RA,  et al.  Practice guideline recommendations summary: disease-modifying therapies for adults with multiple sclerosis: report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology.   Neurology. 2018;90(17):777-788. doi:10.1212/WNL.0000000000005347 PubMedGoogle ScholarCrossref
    12.
    Ontaneda  D, Cohn  S, Fox  R.  Risk stratification and mitigation in multiple sclerosis.   Mult Scler Relat Disord. 2014;3(5):639-649. doi:10.1016/j.msard.2014.05.003 PubMedGoogle ScholarCrossref
    13.
    Ontaneda  D, Tallantyre  E, Kalincik  T, Planchon  SM, Evangelou  N.  Early highly effective versus escalation treatment approaches in relapsing multiple sclerosis.   Lancet Neurol. 2019;18(10):973-980. doi:10.1016/S1474-4422(19)30151-6 PubMedGoogle ScholarCrossref
    14.
    Urai  AE, Braun  A, Donner  TH.  Pupil-linked arousal is driven by decision uncertainty and alters serial choice bias.   Nat Commun. 2017;8:14637. doi:10.1038/ncomms14637 PubMedGoogle ScholarCrossref
    15.
    Mathôt  S.  Pupillometry: psychology, physiology, and function.   J Cogn. 2018;1(1):16. doi:10.5334/joc.18 PubMedGoogle ScholarCrossref
    16.
    Maier  ME, Ernst  B, Steinhauser  M.  Error-related pupil dilation is sensitive to the evaluation of different error types.   Biol Psychol. 2019;141:25-34. doi:10.1016/j.biopsycho.2018.12.013 PubMedGoogle ScholarCrossref
    17.
    Wessel  JR, Dolan  KA, Hollingworth  A.  A blunted phasic autonomic response to errors indexes age-related deficits in error awareness.   Neurobiol Aging. 2018;71:13-20. doi:10.1016/j.neurobiolaging.2018.06.019 PubMedGoogle ScholarCrossref
    18.
    Enax  L, Hu  Y, Trautner  P, Weber  B.  Nutrition labels influence value computation of food products in the ventromedial prefrontal cortex.   Obesity (Silver Spring). 2015;23(4):786-792. doi:10.1002/oby.21027 PubMedGoogle ScholarCrossref
    19.
    Murphy  MS, Baker  M.  When used appropriately, NICE’s traffic light system for febrile children helps clinicians make safe and appropriate decisions.   BMJ. 2014;348:g2518. doi:10.1136/bmj.g2518 PubMedGoogle ScholarCrossref
    20.
    Zhang  X, Liu  Y, Gu  Y, Wang  S, Chen  H.  Red for “stop”: “traffic-light” nutrition labels decrease unhealthy food choices by increasing activity and connectivity in the frontal lobe.   Nutrients. 2020;12(1):E128. doi:10.3390/nu12010128 PubMedGoogle Scholar
    21.
    Enax  L, Krajbich  I, Weber  B.  Salient nutrition labels increase the integration of health attributes in food decision-making.   Judgm Decis Mak. 2016;11(5):460-471.Google Scholar
    22.
    Levy  I, Snell  J, Nelson  AJ, Rustichini  A, Glimcher  PW.  Neural representation of subjective value under risk and ambiguity.   J Neurophysiol. 2010;103(2):1036-1047. doi:10.1152/jn.00853.2009 PubMedGoogle ScholarCrossref
    23.
    Anderson  LR, Mellor  JM.  Predicting health behaviors with an experimental measure of risk preference.   J Health Econ. 2008;27(5):1260-1274. doi:10.1016/j.jhealeco.2008.05.011 PubMedGoogle ScholarCrossref
    24.
    Sormani  MP, Gasperini  C, Romeo  M,  et al; MAGNIMS study group.  Assessing response to interferon-β in a multicenter dataset of patients with MS.   Neurology. 2016;87(2):134-140. doi:10.1212/WNL.0000000000002830 PubMedGoogle ScholarCrossref
    25.
    Murphy  PR, Vandekerckhove  J, Nieuwenhuis  S.  Pupil-linked arousal determines variability in perceptual decision making.   PLoS Comput Biol. 2014;10(9):e1003854. doi:10.1371/journal.pcbi.1003854 PubMedGoogle Scholar
    26.
    Privitera  CM, Renninger  LW, Carney  T, Klein  S, Aguilar  M.  Pupil dilation during visual target detection.   J Vis. 2010;10(10):3-3. doi:10.1167/10.10.3 PubMedGoogle ScholarCrossref
    27.
    Antonakis  J, Bendahan  S, Jacquart  P, Lalive  R.  On making causal claims: a review and recommendations.   The Leadership Quarterly. 2010;21(6):1086-1120. doi:10.1016/j.leaqua.2010.10.010Google ScholarCrossref
    28.
    MacKinnon  DP, Fairchild  AJ, Fritz  MS.  Mediation analysis.   Annu Rev Psychol. 2007;58(1):593-614. doi:10.1146/annurev.psych.58.110405.085542 PubMedGoogle ScholarCrossref
    29.
    VanderWeele  TJ.  Mediation analysis: a practitioner’s guide.   Annu Rev Public Health. 2016;37(1):17-32. doi:10.1146/annurev-publhealth-032315-021402 PubMedGoogle ScholarCrossref
    30.
    de Gee  JW, Knapen  T, Donner  TH.  Decision-related pupil dilation reflects upcoming choice and individual bias.   Proc Natl Acad Sci U S A. 2014;111(5):E618-E625. doi:10.1073/pnas.1317557111 PubMedGoogle ScholarCrossref
    31.
    Khunti  K, Gomes  MB, Pocock  S,  et al.  Therapeutic inertia in the treatment of hyperglycaemia in patients with type 2 diabetes: a systematic review.   Diabetes Obes Metab. 2017.PubMedGoogle Scholar
    32.
    Ogura  M, Harada-Shiba  M.  Clinical inertia in the management of hypercholesterolemia: what clinicians need to do.   J Atheroscler Thromb. 2016;23(5):552-553. doi:10.5551/jat.ED037 PubMedGoogle ScholarCrossref
    33.
    Cooke  CE, Sidel  M, Belletti  DA, Fuhlbrigge  AL.  Review: clinical inertia in the management of chronic obstructive pulmonary disease.   COPD. 2012;9(1):73-80. doi:10.3109/15412555.2011.631957 PubMedGoogle ScholarCrossref
    34.
    Saposnik  G, Sempere  AP, Raptis  R, Prefasi  D, Selchen  D, Maurino  J.  Decision making under uncertainty, therapeutic inertia, and physicians’ risk preferences in the management of multiple sclerosis (DIScUTIR MS).   BMC Neurol. 2016;16:58. doi:10.1186/s12883-016-0577-4 PubMedGoogle ScholarCrossref
    ×