Trent L. Lalonde

Modeling Correlated Binary Outcomes with Time-Dependent Covariates

We group approaches to modeling correlated binary data according to data recorded cross-sectionally as opposed to data recorded longitudinally; according to models that are population-averaged as opposed to subject-specific; and according to data with time-dependent covariates as opposed to time-independent covariates. Standard logistic regression models are appropriate for cross-sectional data. However, for longitudinal data, methods such as generalized estimating equations (GEE) and generalized method of moments (GMM) are commonly used to fit population-averaged models, while random-effects models such as generalized linear mixed models (GLMM) are used to fit subject-specific models. Some of these methods account for time-dependence in covariates while others do not. This paper addressed these approaches with an illustration using a Medicare dataset as it relates to rehospitalization. In particular, we compared results from standard logistic models, GEE models, GMM models, and random-effects models by analyzing a binary outcome for four successive hospitalizations. We found that these procedures address differently the correlation among responses and the feedback from response to covariate. We found marginal GMM logistic regression models to be more appropriate when covariates are classified as time-dependent in comparison to GEE models. We also found conditional random-intercept models with time-dependent covariates decomposed into components to be more appropriate when time-dependent covariates are present in comparison to ordinary random-effects models. We used the SAS procedures GLIMMIX, NLMIXED, IML, GENMOD, and LOGISTIC to analyze the illustrative dataset, as well as unique programs written using the R language.

Marijuana use, craving, and academic motivation and performance among college students: An in-the-moment study.

INTRODUCTION Marijuana is the most commonly used illicit substance in the U.S., with high rates among young adults in the state of Colorado. Chronic, heavy marijuana use can impact cognitive functioning, which has the potential to influence academic performance of college students. It is possible that craving for marijuana may further contribute to diminished cognitive and affective functioning, thus leading to poor outcomes for students. METHODS College student marijuana users (n=57) were recruited based on heavy use and completed ecological momentary assessment (EMA) via text-messaging. The association between marijuana use and craving in a college setting was explored, as well as how these variables might relate to academic motivation, effort and success. The participants were sent text messages for two weeks, three times per day at random times. RESULTS A temporal association between craving and marijuana use was found, where momentary craving positively predicted greater marijuana use. Similarly, as craving levels increased, the number of minutes spent studying decreased at the next assessment point. A negative association between momentary craving for marijuana and academic motivation was found in the same moment. Greater academic self-efficacy positively predicted cumulative GPA, while average minutes spent smoking marijuana was negatively related. CONCLUSIONS Using EMA, marijuana craving and use were significantly related. These findings provide further evidence that heavy marijuana use is negatively associated with academic outcomes.

GMM logistic regression models for longitudinal data with time-dependent covariates and extended classifications

When analyzing longitudinal data, it is essential to account both for the correlation inherent from the repeated measures of the responses as well as the correlation realized on account of the feedback created between the responses at a particular time and the predictors at other times. As such one can analyze these data using generalized estimating equation with the independent working correlation. However, because it is essential to include all the appropriate moment conditions as you solve for the regression coefficients, we explore an alternative approach using a generalized method of moments for estimating the coefficients in such data. We develop an approach that makes use of all the valid moment conditions necessary with each time-dependent and time-independent covariate. This approach does not assume that feedback is always present over time, or if present occur at the same degree. Further, we make use of continuously updating generalized method of moments in obtaining estimates. We fit the generalized method of moments logistic regression model with time-dependent covariates using SAS PROC IML and also in R. We used p-values adjusted for multiple correlated tests to determine the appropriate moment conditions for determining the regression coefficients. We examined two datasets for illustrative purposes. We looked at re-hospitalization taken from a Medicare database. We also revisited data regarding the relationship between the body mass index and future morbidity among children in the Philippines. We conducted a simulated study to compare the performances of extended classifications.

Cancer Type Does Not Affect Exercise-Mediated Improvements in Cardiorespiratory Function and Fatigue

Purpose. Despite mounting evidence indicating that exercise training has a positive effect on cancer recovery, the influence of cancer type on the response to exercise training remains uncharacterized. Therefore, the adaptations to exercise training were compared between groups composed of 7 different forms of cancer. Methods. A total of 319 cancer survivors completed fatigue inventories and participated in assessments of cardiorespiratory function, which encompassed aerobic capacity (VO2peak), pulmonary function (forced vital capacity [FVC] and forced expiratory volume in 1 second [FEV1]), and resting blood pressure and heart rate. Participants were divided into 7 groups based on cancer type, including breast cancer (BC, n = 170), prostate cancer and other male urogenital neoplasia (PC, n = 38), hematological malignancies (HM, n = 34), colorectal cancer (CC, n = 25), gynecological cancers (GC, n = 20), glandular and epithelial neoplasms (GEN, n = 20), and lung cancer (LC, n = 12). All participants completed an individualized, multimodal exercise intervention consisting of cardiorespiratory, flexibility, balance, and muscular strength training 3 days per week for 3 months. Following the intervention, all subjects were reassessed. Generalized Estimating Equations with exchangeable working correlation structure was used to model each response; the group by time interaction effect represented the effect of cancer type on exercise-associated improvements. Results. No significant (P > .05) group by time interaction effects were observed between different types of cancer for any parameter. Pre- to postexercise contrasts revealed significant improvements in VO2peak in BC, PC, HM, and GEN at the Bonferroni adjusted significance level (.00714). Heart rate was significantly lowered in the BC and CC groups. Mean fatigue indices decreased by at least 17% in all groups, but these changes were only significant in the BC, HM, CC, and GC groups. Systolic blood pressure decreased significantly in BC and GC, and diastolic blood pressure decreased significantly only in the BC group while pulmonary function remained unchanged in all cancer types. Conclusion. Although trends toward improved cardiorespiratory and fatigue parameters only reached significance in some groups, there were no significant differences between cancer types. This suggests that cardiorespiratory and fatigue improvements following rehabilitative exercise are not dependent on cancer type. Further research investigating alternative physiological parameters are needed to confirm the relationship between cancer type and exercise-mediated rehabilitation.

Early Mobility in the Intensive Care Unit: Standard Equipment vs a Mobility Platform

BACKGROUND Despite the general belief that mobility and exercise play an important role in the recovery of functional status, mobility is difficult to implement in patients in intensive care units. OBJECTIVES To compare a mobility platform with standard equipment, assessing efficiency (decreased time and staff required to prepare patient), effectiveness (increased activity time), and safety (no falls, unplanned tube removals, or emergency situations) for intensive care patients. METHODS This observational study was approved by the institutional review board, and informed consent was obtained from the patient or the medical decision maker. Intensive care patients were assigned to a room in the usual manner, with platforms in odd-numbered rooms and standard equipment in even-numbered rooms. Standardized data collection tools were designed to collect data for 24 hours for each patient. The nurses caring for the patients completed the data collection tools in real time during the activity. The stages of activity and the physiological states that would preclude mobility were very specifically defined for the research study. RESULTS Data were collected for a total of 71 patients and 238 activities. Important (although not significant) descriptive statistics regarding early mobility in the intensive care unit were discovered. The unintended result of the research study was a change in the culture and practice regarding early mobility in the intensive care unit. CONCLUSIONS Early mobility can be implemented in intensive care units. Standard equipment can be used to mobilize such patients safely; however, for patients who ambulate, a platform may increase efficiency and effectiveness.

Does social context matter? An ecological momentary assessment study of marijuana use among college students

INTRODUCTION Past research has shown that marijuana use occurs commonly in social situations for young adults, though few studies have examined the association between immediate social context and marijuana use patterns and associated problems. The current study examined the impact of demographics, marijuana use and problem use, alcohol use, craving, and social context on the likelihood of using marijuana with others via ecological momentary assessment (EMA). METHODS College-student marijuana users (N=56) were recruited and completed a baseline assessment and training on the two-week signal-contingent EMA protocol. Participants were sent text messages three times per day randomly for two weeks. RESULTS Of the 1131 EMA instances during which participants reported using marijuana, 862 (76.22%) were labeled as being with others. Forty-five participants (80.36%) reported marijuana use with others present during at least half of the times they used marijuana. Findings from a multilevel logistic regression model showed a significant positive association between the probability of using with others and minutes spent using marijuana (b=0.047, p<0.001), social facilitation (b=0.138, p<0.001), and DSM-IV diagnosis (dependence versus no diagnosis, b=1.350, p=0.047). CONCLUSIONS Cannabis dependence, more time using marijuana in the moment, and using for social facilitation purposes were positively associated with using marijuana in the context of being with others. Daily users had more variability in terms of the social context of their use. This study illustrates the complex relationship between social context and marijuana use.

EXACT LOGISTIC MODELS FOR NESTED BINARY DATA

The use of logistic models for independent binary data has relied first on asymptotic theory and later on exact distributions for small samples. However, the use of logistic models for dependent analysis based on exact analysis is not as common. Moreover, attention is usually given to one-stage clustering. In this paper, we extend the exact techniques to address hypothesis testing (estimation is not addressed) for data with second-stage and probably higher levels of clustering. The methods are demonstrated through a somewhat generic example using C+ + program.

Cancer Rehabilitation: Impact of Physical Activity on Initial Clinical Assessments

ABSTRACT Background: Preconditioning and prehabilitation have been reported to ameliorate a host of health- and cancer-related issues, yet few studies have examined implications of past physical ac...

Monte-Carlo Simulation of Correlated Binary Responses

Simulation studies can provide powerful conclusions for correlated or longitudinal response data, particularly for relatively small samples for which asymptotic theory does not apply. For the case of logistic modeling, it is necessary to have appropriate methods for simulating correlated binary data along with associated predictors. This chapter presents a discussion of existing methods for simulating correlated binary response data, including comparisons of various methods for different data types, such as longitudinal versus clustered binary data generation. The purposes and issues associated with generating binary responses are discussed. Simulation methods are divided into four main approaches: using a marginally specified joint probability distribution, using mixture distributions, dichotomizing non-binary random variables, and using a conditionally specified distribution. Approaches using a completely specified joint probability distribution tend to be more computationally intensive and require determination of distributional properties. Mixture methods can involve mixtures of discrete variables only, mixtures of continuous variables only, and mixtures involving both continuous and discrete variables. Methods that involve discretizing non-binary variables most commonly use normal or uniform variables, but some use count variables such as Poisson random variables. Approaches using a conditional specification of the response distribution are the most general, and allow for the greatest range of autocorrelation to be simulated. The chapter concludes with a discussion of implementations available using R software.

Handbook for Applied Modeling: Non-Gaussian and Correlated Data

Designed for the applied practitioner, this book is a compact, entry-level guide to modeling and analyzing non-Gaussian and correlated data. Many practitioners work with data that fail the assumptions of the common linear regression models, necessitating more advanced modeling techniques. This Handbook presents clearly explained modeling options for such situations, along with extensive example data analyses. The book explains core models such as logistic regression, count regression, longitudinal regression, survival analysis, and structural equation modelling without relying on mathematical derivations. All data analyses are performed on real and publicly available data sets, which are revisited multiple times to show differing results using various modeling options. Common pitfalls, data issues, and interpretation of model results are also addressed. Programs in both R and SAS are made available for all results presented in the text so that readers can emulate and adapt analyses for their own data analysis needs. Data, R, and SAS scripts can be found online at http://www.spesi.org.