Faculty Bibliography

Latent variable network models that accommodate edge correlations implicitly, by assuming an underlying latent factor, are increasing in popularity. Although, these models are examples of what is a growing body of research, much of the research is focused on proposing new models or extending others. There has been very little work on unifying the models in a single framework. In this paper, we present a complete framework that organizes existing latent variable network models within an integrative generalized additive model. Our framework is called Conditionally Independent Dyad (CID) models, and includes existing network models that assume dyad (or edge) independence conditional on latent variables and other components in the model. We further discuss practical aspects of model fitting such as posterior parameter estimation via MCMC, identifiability of parameters, approaches to handle missing data and model selection via cross-validation, for the proposed additive CID models. Finally, by presenting several data examples, we illustrate the utility of the proposed framework and provide advice on selecting components for building new CID models.

BACKGROUND:There is concern about the potential of an increased risk related to medications that act on the renin-angiotensin-aldosterone system in patients exposed to coronavirus disease 2019 (Covid-19), because the viral receptor is angiotensin-converting enzyme 2 (ACE2). METHODS:We assessed the relation between previous treatment with ACE inhibitors, angiotensin-receptor blockers, beta-blockers, calcium-channel blockers, or thiazide diuretics and the likelihood of a positive or negative result on Covid-19 testing as well as the likelihood of severe illness (defined as intensive care, mechanical ventilation, or death) among patients who tested positive. Using Bayesian methods, we compared outcomes in patients who had been treated with these medications and in untreated patients, overall and in those with hypertension, after propensity-score matching for receipt of each medication class. A difference of at least 10 percentage points was prespecified as a substantial difference. RESULTS:Among 12,594 patients who were tested for Covid-19, a total of 5894 (46.8%) were positive; 1002 of these patients (17.0%) had severe illness. A history of hypertension was present in 4357 patients (34.6%), among whom 2573 (59.1%) had a positive test; 634 of these patients (24.6%) had severe illness. There was no association between any single medication class and an increased likelihood of a positive test. None of the medications examined was associated with a substantial increase in the risk of severe illness among patients who tested positive. CONCLUSIONS:We found no substantial increase in the likelihood of a positive test for Covid-19 or in the risk of severe Covid-19 among patients who tested positive in association with five common classes of antihypertensive medications.

Social network data represent interactions and relationships among groups of individuals. One aspect of social interaction is social influence, the idea that beliefs or behaviors change as a result of one's social network. The purpose of this article is to introduce a new model for social influence, the latent space model for influence, which employs latent space positions so that individuals are affected most by those who are "closest" to them in the latent space. We describe this model along with some of the contexts in which it can be used and explore the operating characteristics using a series of simulation studies. We conclude with an example of teacher advice-seeking networks to show that changes in beliefs about teaching mathematics may be attributed to network influence.

Background/Purpose : Psoriatic arthritis (PsA) is a heterogenous inflammatory disease affecting skin, joints, and other domains. While psychiatric diseases (i.e., depression and anxiety) are known comorbidities, little is known about their impact on disease severity and patient reported outcomes (PROs). The objective of this study was to characterize the prevalence of psychiatric comorbidities in an academic combined psoriasis-psoriatic arthritis center and determine their impact on PsA clinical and patient derived outcomes. Methods : Consecutive adult patients meeting CASPAR criteria for PsA (n=436) were prospectively recruited at the NYU Psoriatic Arthritis Center. All data was collected from clinical visits utilizing a standardized EPIC template. Depression was defined by established diagnosis and/or use of anti-depressant medications. Objective measures of disease severity included swollen and tender joint counts (SJC/TJC) and PROs including RAPID3 scores. Data was analyzed using statistical software R. Results : Our cohort was comprised of 436 patients: 54% male, mean age of 47 years, and mostly Caucasians (74.1%). Within our population, 19.5% had depression, 15.6% had anxiety, and 4.8% had ADHD (Table 1). Of those with depression, 71% were on anti-depressive medication. At the initial visit, patients with PsA and depression were more likely to be on medication(s) for PsA (80% vs 65%, p=.01) and had a trend towards higher rates of biologic use (47.5% vs 40.4%, p=.126). Those with depression had a similar TJC to their non-depressed counterparts, but had a trend towards fewer swollen joints and concomitant higher RAPID3 scores (Table 2). When analyzing repeated outcome measures over subsequent visits, individuals with depression were similarly more likely to have a higher TJC, a lower SJC, and a higher RAPID3 score (although only RAPID3 was found to be statistically significant, p=.004). Importantly, these findings persisted when analyzing participants that were matched with propensity scores to adjust for age, sex, comorbidities, and medication use. In addition to joint activity, psoriasis activity measured by body surface area (BSA) was lower in those who were depressed (1.4% vs 3.03%, p=.001) and these differences were maintained over subsequent visits. Conclusion : Our results expand on prior reports of significantly elevated rates of depression in PsA. Notably, individuals with depression were more likely to be on medication(s) for their PsA, had fewer swollen joints, and a lower BSA but, paradoxically reported higher RAPID3 scores. This discrepancy is likely a manifestation of how depression could affect the way patients experience their PsA despite apparent improvement in skin and joint symptoms. Depression should, therefore, be considered a critical comorbidity when addressing PsA care in routine visits. Further work is needed to understand whether modulation of psychiatric comorbidities can lead to improved PsA outcomes

Background/Purpose : About 30% of patients with skin psoriasis (PsO) develop psoriatic arthritis (PsA). The reasons for why only some progress to synovio-enthesial disease from skin involvement remains unknown. Genetic, environmental and clinical-demographic factors have been implicated, but are yet to be characterized in specialized, combined care centers. We aim to describe clinical phenotypes differentiating patients with PsO from those with PsA at a large, urban tertiary care PsO-PsA clinic. Methods : Consecutive adult patients meeting CASPAR criteria for PsA (n= 448) or with dermatologist diagnosed skin psoriasis only (n=161) were prospectively recruited at the NYU Psoriatic Arthritis Center and the NYU Psoriasis and Psoriatic Arthritis Clinic. All data was collected utilizing clinical visit notes and additional on-site questionnaires. Type of psoriasis and body surface area (BSA) was determined by dermatologists or rheumatologists specializing in psoriatic disease. Data was analyzed using statistical software SPSS using chi squared test with Yates Continuity Correction for dichotomous/categorical variables and t-test for continuous variables. Results : Patients with PsO were more likely to be older (52.7 vs. 48.9, p=.032) and have hypertension, obesity, diabetes, and history of myocardial infarction (Figure 1). Patients with PsO had a statistically higher BSA than those with PsA (5.8% vs 3.1%, p=.003). While the type of psoriasis was similar, the site of psoriasis involvement (specifically the scalp and nail) differentiated the populations (Table 1). In PsA compared to PsO, the odds ratio of scalp involvement was 2.96 (95% Confidence Interval [CI] 2.02, 4.34) and that of nail involvement was 14.66 (95% CI 8.21, 26.16). Inverse psoriasis was not different between groups. Additionally, those with PsA were much more likely to have a first degree relative (FDR) with psoriasis compared to those with cutaneous disease alone (31.9% vs. 12.0%, p=.007) (Figure 1). Conclusion : We report for the first time the comorbidities and psoriasis features of PsA and PsO populations in a large, combined center. We found that scalp involvement and any nail involvement was more prevalent in the PsA as compared to PsO. Only one previous study has identified scalp psoriasis[1] as a possible risk factor for progression, while previous studies looking at nail psoriasis reported much lower odds ratios[1,2]. Patients with PsA also demonstrated a higher number of FDRs with skin psoriasis, reinforcing the notion of strong heritability in PsA. The identification of risk factors for progression is of critical importance to study natural history of psoriatic disease and to inform the adequate design of prevention trials in psoriasis patients who have enriched features associated with future transition to synovio-enthesial disease

Background: Complicated grief (CG) is hypothesized to include both attachment and traumatic distress symptoms, and a preliminary diagnosis has been placed in the trauma and stressor related DSM-5 category (APA, 2013). Posttraumatic stress disorder (PTSD) and CG often present comorbidly, and both result from a major stressor (Simon et al., 2007; Marques et al., 2013; Lenferink et al., 2018). Preliminary data suggest posttraumatic stress symptoms (PTSS) may be present across patients with CG, and not vary by whether the loss is violent or accidental in nature such as required for PTSD diagnoses (Simon et al., 2013; Kersting et al., 2011). Much less is known about how PTSS changes with CG targeted treatment, whether this change is impacted by the nature of the death, or whether it may be necessary to target PTSS separately from grief to improve functional outcomes.
Method(s): Participants were 395 individuals (mean age +/- SD = 53.0 +/- 14.5 years; 78.0% women) with a primary diagnosis of CG based on structured clinical interviews and an Inventory of Complicated Grief (ICG) score>=30. Data were derived from the previously published 20-week multi-center RCT of complicated grief therapy plus pill placebo (CGT + PLA), CGT plus citalopram (CGT + CIT), citalopram (CIT), or placebo (PLA) (Shear et al., 2016). DSM-IV PTSS were assessed using the 17-item self-report Davidson Trauma Scale (DTS). DTS total score of 40 was proposed by the developers of the scale as a cut-off for a diagnosis of PTSD, and has been frequently used as a threshold in previous studies (Davidson et al., 1997; Kastello et al., 2016; Khitab et al., 2013). Our primary analysis examined the adjusted mean difference from baseline in the DTS total and subscale scores (i.e., intrusion, avoidance-numbing, hyperarousal) over three follow-up periods (week 12, 16, and 20) by treatment arm using longitudinal mixed effects regression with participant specific random intercepts. In follow-up analyses, we investigated whether cause of death (violent vs. nonviolent) moderated the relationship between treatments and DTS total score by introducing interaction terms between cause of death and treatment arms in the mixed effects regression model.
Result(s): In the full sample, the mean DTS total score at baseline was 63.2 +/- 27.2, and 77.7% (n = 307) had DTS>=40. There was a general decreasing trend of mean DTS total scores over the 20-week period with a mean adjusted reduction of 27.4 points (d = 0.6) from baseline to week 12 (p < 0.001), and a reduction of 30.7 points (d = 0.7) from baseline to week 20 (p < 0.001). There was no significant difference in change in DTS total score at week 12 by treatment group. However, at weeks 16 and 20, CGT + PLA and CGT + CIT were each associated with a significant DTS reduction compared to placebo alone, while CIT was not. For CGT + PLA vs. PLA, there was 8.8 point (d = 0.14) greater reduction in DTS total score from baseline to week 16 (p = 0.01), and 12.5 point (d = 0.19) greater reduction from baseline to week 20 (p < 0.001). For CGT + CIT vs PLA, there was a 10.0 point (d = 0.15) greater reduction in adjusted DTS total score from baseline to week 16 (p < 0.001), and 10.7 point (d = 0.16) greater decrease from baseline to week 20 (p < 0.001). Similar trends were observed for DTS subscales-CGT + PLA and CGT + CIT demonstrated consistent reduction compared to PLA. In the model with interaction terms between treatments and cause of death, the decrease in DTS score for CGT + CIT compared to PLA was 9.5 points (d = 0.12) greater for those who had violent death compared to those who did not experience violent death (p = 0.04). For CGT + CIT vs CGT + PLA, however, the reduction in DTS total score was 4.2 points (d = 0.04) greater in those who experienced violent death compared to those who did not, but the difference was not statistically significant (p = 0.53).
Conclusion(s): Adults with primary CG assigned to CGT with or without medication demonstrated a significantly larger reduction in PTSS compared to pill placebo, whereas citalopram alone did not. These data parallel findings from the primary study findings for grief (Shear et al., 2016), and demonstrate that CGT may be an effective intervention for PTSS in those with CG. A high level of PTSS were present in this primary CG sample, and PTSS were comparable at baseline for those with violent and non-violent losses. For those who lost someone to violent death, while these data found initial support for greater PTSS reduction for combination therapy with CGT and citalopram compared to placebo, we did not find evidence for a significant benefit of combined therapy over CGT alone for CG due to violent loss. More research is needed to fully understand the role of traumatic distress and its optimal treatment in CG

Progressive muscle relaxation (PMR) is an under-utilized Level A evidence-based treatment for migraine prevention. We studied the feasibility and acceptability of smartphone application (app)-based PMR for migraine in a neurology setting, explored whether app-based PMR might reduce headache (HA) days, and examined potential predictors of app and/or PMR use. In this single-arm pilot study, adults with ICHD3 migraine, 4+ HA days/month, a smartphone, and no prior behavioral migraine therapy in the past year were asked to complete a daily HA diary and do PMR for 20 min/day for 90 days. Outcomes were: adherence to PMR (no. and duration of audio plays) and frequency of diary use. Predictors in the models were baseline demographics, HA-specific variables, baseline PROMIS (patient-reported outcomes measurement information system) depression and anxiety scores, presence of overlapping pain conditions studied and app satisfaction scores at time of enrollment. Fifty-one patients enrolled (94% female). Mean age was 39 ± 13 years. The majority (63%) had severe migraine disability at baseline (MIDAS). PMR was played 22 ± 21 days on average. Mean/session duration was 11 ± 7 min. About half (47%) of uses were 1+ time/week and 35% of uses were 2+ times/week. There was a decline in use/week. On average, high users (PMR 2+ days/week in the first month) had 4 fewer days of reported HAs in month 2 vs. month 1, whereas low PMR users (PMR < 2 days/week in the first month) had only 2 fewer HA days in month 2. PROMIS depression score was negatively associated with the log odds of using the diary at least once (vs. no activity) in a week (OR = 0.70, 95% CI = [0.55, 0.85]) and of doing the PMR at least once in a week (OR = 0.77, 95% CI = [0.68, 0.91]). PROMIS anxiety was positively associated with using the diary at least once every week (OR = 1.33, 95% CI = [1.09, 1.73]) and with doing the PMR at least once every week (OR = 1.14 [95% CI = [1.02, 1.31]). In conclusion, about half of participants used smartphone-based PMR intervention based upon a brief, initial introduction to the app. App use was associated with reduction in HA days. Higher depression scores were negatively associated with diary and PMR use, whereas higher anxiety scores were positively associated.

We study regularized estimation in high-dimensional longitudinal classification problems, using the lasso and fused lasso regularizers. The constructed coefficient estimates are piecewise constant across the time dimension in the longitudinal problem, with adaptively selected change points (break points). We present an efficient algorithm for computing such estimates, based on proximal gradient descent. We apply our proposed technique to a longitudinal data set on Alzheimer's disease from the Cardiovascular Health Study Cognition Study. Using data analysis and a simulation study, we motivate and demonstrate several practical considerations such as the selection of tuning parameters and the assessment of model stability. While race, gender, vascular and heart disease, lack of caregivers, and deterioration of learning and memory are all important predictors of dementia, we also find that these risk factors become more relevant in the later stages of life.