A Bayesian Dirichlet-Multinomial Test for Cross-Group Differences

Testing for differences within data sets is an important issue across various applications. Our work is primarily motivated by the analysis of microbiomial composition, which has been increasingly relevant and important with the rise of DNA sequencing. We first review classical frequentist tests that are commonly used in tackling such problems. We then propose a Bayesian Dirichlet-multinomial framework for modeling the metagenomic data and for testing underlying differences between the samples. A parametric Dirichlet-multinomial model uses an intuitive hierarchical structure that allows for flexibility in characterizing both the within-group variation and the cross-group difference and provides very interpretable parameters. A computational method for evaluating the marginal likelihoods under the null and alternative hypotheses is also given. Through simulations, we show that our Bayesian model performs competitively against frequentist counterparts. We illustrate the method through analyzing metagenomic applications using the Human Microbiome Project data.

A cross-site, comparative effectiveness study of an integrated HIV and substance use treatment program.

Co-occurrence of HIV and substance abuse is associated with poor outcomes for HIV-related health and substance use. Integration of substance use and medical care holds promise for HIV patients, yet few integrated treatment models have been reported. Most of the reported models lack data on treatment outcomes in diverse settings. This study examined the substance use outcomes of an integrated treatment model for patients with both HIV and substance use at three different clinics. Sites differed by type and degree of integration, with one integrated academic medical center, one co-located academic medical center, and one co-located community health center. Participants (n=286) received integrated substance use and HIV treatment for 12 months and were interviewed at 6-month intervals. We used linear generalized estimating equation regression analysis to examine changes in Addiction Severity Index (ASI) alcohol and drug severity scores. To test whether our treatment was differentially effective across sites, we compared a full model including site by time point interaction terms to a reduced model including only site fixed effects. Alcohol severity scores decreased significantly at 6 and 12 months. Drug severity scores decreased significantly at 12 months. Once baseline severity variation was incorporated into the model, there was no evidence of variation in alcohol or drug score changes by site. Substance use outcomes did not differ by age, gender, income, or race. This integrated treatment model offers an option for treating diverse patients with HIV and substance use in a variety of clinic settings. Studies with control groups are needed to confirm these findings.

Fluid resuscitation practice patterns in intensive care units of the USA: a cross-sectional survey of critical care physicians.

BACKGROUND: Fluid resuscitation is a cornerstone of intensive care treatment, yet there is a lack of agreement on how various types of fluids should be used in critically ill patients with different disease states. Therefore, our goal was to investigate the practice patterns of fluid utilization for resuscitation of adult patients in intensive care units (ICUs) within the USA. METHODS: We conducted a cross-sectional online survey of 502 physicians practicing in medical and surgical ICUs. Survey questions were designed to assess clinical decision-making processes for 3 types of patients who need volume expansion: (1) not bleeding and not septic, (2) bleeding but not septic, (3) requiring resuscitation for sepsis. First-choice fluid used in fluid boluses for these 3 patient types was requested from the respondents. Descriptive statistics were performed using a Kruskal-Wallis test to evaluate differences among the physician groups. Follow-up tests, including t tests, were conducted to evaluate differences between ICU types, hospital settings, and bolus volume. RESULTS: Fluid resuscitation varied with respect to preferences for the factors to determine volume status and preferences for fluid types. The 3 most frequently preferred volume indicators were blood pressure, urine output, and central venous pressure. Regardless of the patient type, the most preferred fluid type was crystalloid, followed by 5 % albumin and then 6 % hydroxyethyl starches (HES) 450/0.70 and 6 % HES 600/0.75. Surprisingly, up to 10 % of physicians still chose HES as the first choice of fluid for resuscitation in sepsis. The clinical specialty and the practice setting of the treating physicians also influenced fluid choices. CONCLUSIONS: Practice patterns of fluid resuscitation varied in the USA, depending on patient characteristics, clinical specialties, and practice settings of the treating physicians.