Charles R. Hadlock

Clinician-Graded Electronic Facial Paralysis Assessment: The eFACE.

Background: The subjective nature of facial aesthetics and the difficulties associated with quantifying facial function have made outcomes analysis in facial paralysis challenging. Clinicians rely on photographs, subjective descriptions, and scales, limiting assessment, communication among providers, and communication between providers and patients. The authors describe the development and validation of a comprehensive, electronic, clinician-graded facial function scale (eFACE), which generates an overall disfigurement score and offers simple graphic output for clinician communication, assessment of various interventions, and patient understanding. The eFACE application may be used in a variety of electronic devices, including smartphones, tablets, and computers. Methods: An instrument consisting of 16 items in a visual analogue scale format was developed to assess facial function and symmetry (the eFACE). Video recordings of subjects performing facial expressions were viewed, and the eFACE instrument was applied, along with an overall facial disfigurement score. A multiple regression analysis was performed to determine the best linear relationship between overall expert-determined disfigurement and the eFACE items. The resulting equation was tested by three independent facial nerve clinicians, using an additional series of patients, to determine both interrater and intrarater reliability of the instrument. Results: Multiple regression analysis produced good fit of eFACE parameters to overall expert-rated global facial disfigurement when dynamic parameters were weighted twice as heavily as static and synkinesis parameters. eFACE scores demonstrated very high interrater and intrarater reliability. Conclusion: The eFACE is a reliable, reproducible, and straightforward digital clinical measure with which to assess facial function and disfigurement in patients with facial paralysis.

Weighting of Facial Grading Variables to Disfigurement in Facial Palsy

IMPORTANCE A universal, health care professional-graded scale for facial assessment would be a useful tool for reporting, comparing, and assessing facial function among patients with facial paralysis. OBJECTIVES To correlate scores of an assessment tool, the eFACE scale, with expert-rated facial disfigurement and to determine the relative contributions of facial features to facial palsy-related disfigurement. DESIGN, SETTING, AND PARTICIPANTS The eFACE scale yields 15 individual variable scores, in addition to subscores for static, dynamic, and synkinesis elements, and a total score that is based on 100-point scales. Two hundred patients with varying degrees of unilateral facial palsy underwent independent eFACE assessment and assignment of a disfigurement score by 2 facial nerve surgeons. The mean scores were determined, and multivariate regression analysis was performed to fit eFACE subset scores (static, dynamic, and synkinesis) to disfigurement ratings. A hybrid regression model was then used to weight each of the 15 eFACE variables, using stepwise regression to control for the effect of the other variables. Scoring was performed during an 8-week period from March 16 to May 8, 2015. MAIN OUTCOME AND MEASURE Use of the 100-point eFACE variables, together with a 100-point visual analog scale of disfigurement, with 0 representing the most extreme disfigurement possible from a facial nerve disorder and 100 representing no discernible facial disfigurement. RESULTS In the 200 patients included in analysis (126 [63.0% female]; mean [SD] age, 46.5 [16.4] years]), predicted disfigurement scores based on eFACE subset scores demonstrated excellent agreement with surgeon-graded disfigurement severity (r2 = 0.79). Variable weighting demonstrated that the 6 key contributors to overall disfigurement were (in order of importance) nasolabial fold depth at rest (normalized coefficient [NC], 0.18; P < .001), oral commissure position at rest (NC, 0.15; P < .001), lower lip asymmetry while pronouncing the long /ē/ (NC, 0.09; P < .001), palpebral fissure width at rest (NC, 0.09; P < .001), nasolabial fold orientation with smiling (NC, 0.08; P = .001), and palpebral fissure width during attempts at full eye closure (NC, 0.06; P = .03). CONCLUSIONS AND RELEVANCE A mathematical association between eFACE-measured facial features and overall expert-graded disfigurement in facial paralysis has been established. For those using the eFACE grading scale, predictions of the specific effects of various interventions on expert-rated disfigurement are now possible and may guide therapy. LEVEL OF EVIDENCE NA.

VODYS: An Agent-Based Model for Exploring Campaign Dynamics

The literature on campaigns has considered a number of factors that affect whether and how someone votes, including demographics, campaign strategies, and social milieu. Understanding the dynamics of campaigns, however, is complicated by the fact that researchers cannot observe much of what happens during an election cycle. Typically, studies rely on voter recollections of conversations, contacts, and media exposure. In addition, because data are collected at discrete points in time, most models of voter turnout cannot capture the dynamic nature of an individual’s interactions during a campaign cycle. Agent-based models offer a way to overcome these data limitations by allowing us to model the dynamics of voter turnout over the course of many weeks as individuals move back and forth between home and work environments, interacting with neighbors and colleagues. In this article, the authors present an agent-based model of campaign dynamics, VODYS, and conduct three simulations to demonstrate the utility of agent-based models for exploring the effects of contact and context on political behavior.

Proactive Encouragement of Interdisciplinary Research Teams in a Business School Environment: Strategy and Results.

This case study describes efforts to promote collaborative research across traditional boundaries in a business‐oriented university as part of an institutional transformation. We model this activity within the framework of social network analysis and use quantitative tools from that field to characterize resulting impacts.

Generating random networks from a given distribution

Several variations are given for an algorithm that generates random networks approximately respecting the probabilities given by any likelihood function, such as from a p^* social network model. A novel use of the genetic algorithm is incorporated in these methods, which improves its applicability to the degenerate distributions that can arise with p^* models. Our approach includes a convenient way to find the high-probability items of an arbitrary network distribution function.

Asymptotic performance of near-optimum controls obtained by regular and singular perturbations

The dependence of optimal control problems on parameters and the approximate solutions of such problems by perturbation techniques have attracted considerable interest in recent years. Generally speaking, in problems amenable to “regular” perturbation techniques, one hopes to obtain an nth-order Taylor expansion in the parameter as an approximate solution to the optimal control; in problems amenable to “singular” perturbation techniques, one hopes to obtain an nth-order asymptotic expansion in the parameter, but this expansion is usually not a Taylor-type expansion. (For high-order practical problems n is typically <2.) If such approximations are applied to the system, it is natural to ask for a comparison of their performance indexes. In a sense to be made more precise later, Cruz and Werner [l] showed in 1968 that for a certain class of regular problems, the performance of an nth-order approximation to the optimal control approximates the optimal performance index to order 2n + 1; and this result has since been extended to other regular problems [2, 31. However, the proofs have always relied, in a critical way, on the Taylor expansion nature of the approximations, and as well on a lengthy computation of the first 272 + 1 partial derivatives of the performance index with respect to the parameter. It is natural to ask whether the “quadratic” behavior of the Hamiltonian as a function of u, in the neighborhood of an optimal solution, might be used more directly to ohtain a simpler and more transparent proof. The purpose of the present paper is to carry through such a proof and to show that by the same method it is possible to treat a broad class of singular problems. With respect to notation, for a vector or matrix o, i ZJ j denotes the norm given by the sum of the absolute values of the components. The Landau 0 notation will be used: if h and z are variables, h : P( z means that there is a constant K ) such that it is always the case that / 11 / ,( Kz. If h depends on other variables, it is implied that K may be chosen so that this inequality holds uniformly in these other variables. It will be convenient to make frequent use of the following

Service-Learning in the Mathematical Sciences

Abstract Service-learning can be a valuable educational enterprise with both course-specific and general education benefits, as well as contributions to the community. These potential merits and key considerations in achieving them are discussed with special attention to the mathematical sciences.