"No sufro, estoy bien/I am not suffering, so I am doing OK": A mixed method exploration of individual and network-level factors and Type 2 Diabetes Mellitus (T2DM) among Mexican American adults in New York City.

INTRODUCTION: The prevalence of type 2 Diabetes Mellitus (T2DM) is 2-3 times greater among Mexican Americans than non-Latino whites, and Mexican Americans are more likely to develop T2DM at younger ages and experience higher rates of complications. Social networks might play a crucial role in both T2DM etiology and management through social support, access to resources, social engagement, and health behavioral norms. OBJECTIVE: To quantitatively identify the social network features associated with glycated hemoglobin (HbA1c) in a community sample of Mexican immigrants residing in New York City, and to explore the extent to which these quantitative findings converge with qualitative narratives of their lived experiences. METHODS: This study used a convergent mixed methods design. To collect personal network data, we used EgoWeb, and obtained 1,400 personal network ties from 81 participants. HbA1c readings were collected using dried blood spots and categorized according to the laboratory criteria of the American Diabetes Association. Additional survey data were collected using Qualtrics software. To investigate the significance of the network-level factors after accounting for the socioeconomic and demographic individual-level factors that the literature indicates to be associated with T2DM, we used a multiple regression model on quantitative data sources. For the qualitative portion of the study, we selected a subset of individuals who participated in the quantitative portion, which represented 500 personal network ties from 25 participants. We conducted in-depth interviews guided by the visualization of these ties to explore who was helpful or difficult in managing their health and health behaviors. RESULTS: Individual-level indicators associated with lower HbA1c scores were body mass index (ß = -0.07, p<0.05), and healthy eating index scores (ß = -0.03, p<0.02). The network-level predictor associated with higher HbA1c levels was the percentage of diabetic alters in the network (ß = 0.08, p <0.001, with a 25% increase in the percentages associated 2.0 change in HbA1c levels. The qualitative data highlighted that most of the diabetes-related information diffused through the social networks of our participants was related to dietary practices, such as reducing sugar and red meat consumption, eating out less, and reducing portion sizes. Notably, even among those with elevated levels and diabetes-related health complications, HbA1c was not considered a part of the lay descriptions of good health since they were not "suffering." Participants regarded doctors as the ultimate authority in diabetes care, even if they had supportive members in their personal networks. CONCLUSION: Our study provides quantitative evidence for the significant role of diabetic network members in the etiology and management of T2DM among Mexican Americans. Our qualitative findings suggest important ley terms for T2DM management and the importance of physicians, which could be included in in future social networks studies seeking to diffuse diabetes-related health information for T2DM prevention and management efforts in this population.

A new look at state-space models for neural data.

State space methods have proven indispensable in neural data analysis. However, common methods for performing inference in state-space models with non-Gaussian observations rely on certain approximations which are not always accurate. Here we review direct optimization methods that avoid these approximations, but that nonetheless retain the computational efficiency of the approximate methods. We discuss a variety of examples, applying these direct optimization techniques to problems in spike train smoothing, stimulus decoding, parameter estimation, and inference of synaptic properties. Along the way, we point out connections to some related standard statistical methods, including spline smoothing and isotonic regression. Finally, we note that the computational methods reviewed here do not in fact depend on the state-space setting at all; instead, the key property we are exploiting involves the bandedness of certain matrices. We close by discussing some applications of this more general point of view, including Markov chain Monte Carlo methods for neural decoding and efficient estimation of spatially-varying firing rates.