Predicting student outcomes using digital logs of learning behaviors: Review, current standards, and suggestions for future work.

Using traces of behaviors to predict outcomes is useful in varied contexts ranging from buyer behaviors to behaviors collected from smart-home devices. Increasingly, higher education systems have been using Learning Management System (LMS) digital data to capture and understand students' learning and well-being. Researchers in the social sciences are increasingly interested in the potential of using digital log data to predict outcomes and design interventions. Using LMS data for predicting the likelihood of students' success in for-credit college courses provides a useful example of how social scientists can use these techniques on a variety of data types. Here, we provide a primer on how LMS data can be feature-mapped and analyzed to accomplish these goals. We begin with a literature review summarizing current approaches to analyzing LMS data, then discuss ethical issues of privacy when using demographic data and equitable model building. In the second part of the paper, we provide an overview of popular machine learning algorithms and review analytic considerations such as feature generation, assessment of model performance, and sampling techniques. Finally, we conclude with an empirical example demonstrating the ability of LMS data to predict student success, summarizing important features and assessing model performance across different model specifications.

Nonmelanoma Skin Cancer in Patients Older Than Age 85 Years Presenting for Mohs Surgery: A Prospective, Multicenter Cohort Study.

Importance: It has been suggested that Mohs surgery for skin cancer among individuals with limited life expectancy may be associated with needless risk and discomfort, along with increased health care costs. Objective: To investigate patient- and tumor-specific indications considered by clinicians for treatment of nonmelanoma skin cancer in older individuals. Design, Setting, and Participants: This multicenter, prospective cohort study was conducted using data from US private practice and academic centers. Included patients were those older than age 85 years presenting for skin cancer surgery and referred for Mohs surgery, with reference groups of those younger than age 85 years receiving Mohs surgery and those older than age 85 years not receiving Mohs surgery. Data were analyzed from November 2018 through January 2019. Exposures: Mohs surgery for nonmelanoma skin cancer. Main Outcomes and Measures: Reason for treatment selection. Results: Among 1181 patients older than age 85 years referred for Mohs surgery (724 [61.9%] men among 1169 patients with sex data; 681 individuals aged >85 to 88 years [57.9%] among 1176 patients with age data) treated at 22 sites, 1078 patients (91.3%) were treated by Mohs surgery, and 103 patients (8.7%) received alternate treatment. Patients receiving Mohs surgery were more likely to have tumors on the face (738 patients [68.5%] vs 26 patients [25.2%]; P < .001) and nearly 4-fold more likely to have high functional status (614 patients [57.0%] vs 16 patients [15.5%]; P < .001). Of 15 distinct reasons provided by surgeons for opting to proceed with Mohs surgery, the most common were patient desire for treatment with a high cure rate (712 patients [66.0%]), good or excellent patient functional status for age (614 patients [57.0%]), and high risk associated with the tumor based on histology (433 patients [40.2%]). Conclusions and Relevance: This study found that older patients who received Mohs surgery often had high functional status, high-risk tumors, and tumors located on the face. These findings suggest that timely surgical treatment may be appropriate in older patients given that their tumors may be aggressive, painful, disfiguring, and anxiety provoking.

Deep learning: A primer for psychologists.

Deep learning has revolutionized predictive modeling in topics such as computer vision and natural language processing but is not commonly applied to psychological data. In an effort to bring the benefits of deep learning to psychologists, we provide an overview of deep learning for researchers who have a working knowledge of linear regression. We first discuss several benefits of the deep learning approach to predictive modeling. We then present three basic deep learning models that generalize linear regression: the feedforward neural network (FNN), the recurrent neural network (RNN), and the convolutional neural network (CNN). We include concrete toy examples with R code to demonstrate how each model may be applied to answer prediction-focused research questions using common data types collected by psychologists. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

A Deep Learning Algorithm for High-Dimensional Exploratory Item Factor Analysis.

Marginal maximum likelihood (MML) estimation is the preferred approach to fitting item response theory models in psychometrics due to the MML estimator's consistency, normality, and efficiency as the sample size tends to infinity. However, state-of-the-art MML estimation procedures such as the Metropolis-Hastings Robbins-Monro (MH-RM) algorithm as well as approximate MML estimation procedures such as variational inference (VI) are computationally time-consuming when the sample size and the number of latent factors are very large. In this work, we investigate a deep learning-based VI algorithm for exploratory item factor analysis (IFA) that is computationally fast even in large data sets with many latent factors. The proposed approach applies a deep artificial neural network model called an importance-weighted autoencoder (IWAE) for exploratory IFA. The IWAE approximates the MML estimator using an importance sampling technique wherein increasing the number of importance-weighted (IW) samples drawn during fitting improves the approximation, typically at the cost of decreased computational efficiency. We provide a real data application that recovers results aligning with psychological theory across random starts. Via simulation studies, we show that the IWAE yields more accurate estimates as either the sample size or the number of IW samples increases (although factor correlation and intercepts estimates exhibit some bias) and obtains similar results to MH-RM in less time. Our simulations also suggest that the proposed approach performs similarly to and is potentially faster than constrained joint maximum likelihood estimation, a fast procedure that is consistent when the sample size and the number of items simultaneously tend to infinity.

A review of cutaneous toxicities from targeted therapies in the treatment of colorectal cancers.

Currently there are three targeted therapies approved for the treatment of colorectal cancers. These include the epidermal growth factor receptor (EGFR) inhibitors, cetuximab and panitumumab, and the multikinase inhibitor regorafenib. It is important to understand and recognize the common presentations of cutaneous toxicity that result from these agents to effectively manage symptoms and prevent premature discontinuation of anticancer treatment.

Special stains in Mohs surgery.

The excellent cure rates associated with Mohs micrographic surgery depend on accurate interpretation of complete and high-quality microscopic frozen sections. Reliable interpretation of microscopic slides is only possible if the surgeon can distinguish tumor cells from surrounding normal tissue. By highlighting tumor cells with a chromogen that is visible on light microscopy, immunostaining allows the Mohs surgeon to distinguish tumor from normal cells in these challenging scenarios. This article focuses on practical aspects involving the most commonly used immunostains in dermatologic surgery, including MART-1 for melanocytic neoplasms, cytokeratin stains for keratinocytic neoplasms, and CD34 stains for dermatofibrosarcoma protuberans.

Decision analysis of treatment options for vestibular schwannoma.

OBJECT: Widespread use of MR imaging has contributed to the more frequent diagnosis of vestibular schwannomas (VSs). These tumors represent 10% of primary adult intracranial neoplasms, and if they are symptomatic, they usually present with hearing loss and tinnitus. Currently, there are 3 treatment options for quality of life (QOL): wait and scan, microsurgery, and radiosurgery. In this paper, the authors' purpose is to determine which treatment modality yields the highest QOL at 5- and 10-year follow-up, considering the likelihood of recurrence and various complications. METHODS: The MEDLINE, Embase, and Cochrane online databases were searched for English-language articles published between 1990 and June 2008, containing key words relating to VS. Data were pooled to calculate the prevalence of treatment complications, tumor recurrence, and QOL with various complications. For parameters in which incidence varied with time of follow-up, the authors used meta-regression to determine the mean prevalence rates at a specified length of follow-up. A decision-analytical model was constructed to compare 5- and 10-year outcomes for a patient with a unilateral tumor and partially intact hearing. The 3 treatment options, wait and scan, microsurgery, and radiosurgery, were compared. RESULTS: After screening more than 2500 abstracts, the authors ultimately included 113 articles in this analysis. Recurrence, complication rates, and onset of complication varied with the treatment chosen. The relative QOL at the 5-year follow-up was 0.898 of normal for wait and scan, 0.953 for microsurgery, and 0.97 for radiosurgery. These differences are significant (p < 0.0052). Data were too scarce at the 10-year follow-up to calculate significant differences between the microsurgery and radiosurgery strategies. CONCLUSIONS: At 5 years, patients treated with radiosurgery have an overall better QOL than those treated with either microsurgery or those investigated further with serial imaging. The authors found that the complications associated with wait-and-scan and microsurgery treatment strategies negatively impacted patient lives more than the complications from radiosurgery. One limitation of this study is that the 10-year follow-up data were too limited to analyze, and more studies are needed to determine if the authors' results are still consistent at 10 years.

Development of and psychometric testing for the Brief Pain Inventory-Facial in patients with facial pain syndromes.

OBJECT: Outcomes in clinical trials on trigeminal pain therapies require instruments with demonstrated reliability and validity. The authors evaluated the Brief Pain Inventory (BPI) in its existing form plus an additional 7 facial-specific items in patients referred to a single neurosurgeon for a diagnosis of facial pain. The complete 18-item instrument is referred to as the BPI-Facial. METHODS: This study was a cross-sectional analysis of patients who completed the BPI-Facial. The diagnosis of classic versus atypical trigeminal neuralgia (TN) was made before analyzing the questionnaire results. A hypothesis-driven factor analysis was used to determine the principal components of the questionnaire. Item reliability and questionnaire validity were tested for these specific constructs. RESULTS: Data from 156 patients were analyzed, including 114 patients (73%) with classic and 42 (27%) with atypical TN. Using orthomax rotation factor analysis, 3 factors with an eigenvalue > 1.0 were identified-pain intensity, interference with general activities, and facial-specific pain interference-accounting for 97.6% of the observed item variance. Retention of the 3 factors was confirmed via a Cattell scree plot. Internal reliability was demonstrated by calculating Cronbach's alpha: 0.86 for pain intensity, 0.89 for interference with general activities, 0.95 for facial-specific pain interference, and 0.94 for the entire instrument. Initial validity of the BPI-Facial instrument was supported by the detection of statistically significant differences between patients with classic versus atypical pain. Patients with atypical TN rated their facial pain as more intense (atypical 6.24 vs classic 5.03, p = 0.013) and as having greater interference in general activities (atypical 6.94 vs classic 5.43, p = 0.0033). Both groups expressed high levels of facial-specific pain interference (atypical 6.34 vs classic 5.95, p = 0.527). CONCLUSIONS: The BPI-Facial is a rigorous measure of facial pain in patients with TN and appears to have sound psychometric properties and is responsive to differences between classic and atypical TN. Future studies must assess the instrument's test-retest reliability, validity in additional populations, and responsiveness with respect to changes in patient outcomes following neurosurgical interventions and medical therapies.