Regression Models For Clustered Data Research Articles

Catheter size and risk of short-term peripheral venous catheter-associated bloodstream infections: an observational study

ObjectivesShort-term peripheral venous catheter-associated bloodstream infections (PVC-associated BSI) are disregarded in the literature because of their relatively low incidence. No data are available on the association between PVC diameter size and the risk of PVC-associated BSI. MethodsUsing a prospective database, we performed an observational study at the University of Geneva Hospitals from 1 January 2020 to 31 December 2021, including all patients with a PVC. We used univariable and multivariable marginal Cox regression models for clustered data to investigate the association between catheter size and PVC-associated BSI. The main variable of interest ‘catheter size’ was forced into our multivariable models. Confounders, which are thought to influence the risk of PVC-associated BSI, were used as adjustment factors. ResultsA total of 206 804 PVCs were included. In all, 10 806 of 201 413 (5.4%), 80 274 of 201 413 (39.9%), 93 047 of 201 413 (46.2%) and 17 286 of 201 413 (8.6%) PVCs measured ≤16G, 18G, 20G and ≥22G, respectively. The univariable analysis showed that diameters of ≤16G were significantly associated with a higher risk of PVC-associated BSI (hazard ratio [HR] 4.52, 95% CI, 1.14–18.00). Multivariable models confirmed these results (HR 4.65, 95% CI, 1.19–18.20). Sensitivity analyses including PVC inserted only in 2021 (HR 4.80, 95% CI, 1.21–19.10), for dwell time >2 days (HR 3.67, 95% CI, 0.92–14.65) and only in adults (HR 3.97, 95% CI, 0.97–15.39) showed similar results. DiscussionLarger PVC size may increase the risk of PVC-associated BSI. Diameter size should be considered when selecting PVCs to reduce the burden of PVC-associated BSI.

Application of Frailty Quantile Regression Model to Investigate of Factors Survival Time in Breast Cancer: A Multi-Center Study.

The prognostic factors of survival can be accurately identified using data from different health centers, but the structure of multi-center data is heterogeneous due to the treatment of patients in different centers or similar reasons. In survival analysis, the shared frailty model is a common way to analyze multi-center data that assumes all covariates have homogenous effects. We used a censored quantile regression model for clustered survival data to study the impact of prognostic factors on survival time. This multi-center historical cohort study included 1785 participants with breast cancer from four different medical centers. A censored quantile regression model with a gamma distribution for the frailty term was used, and p-value less than 0.05 considered significant. The 10th and 50th percentiles (95% confidence interval) of survival time were 26.22 (23-28.77) and 235.07 (130-236.55) months, respectively. The effect of metastasis on the 10th and 50th percentiles of survival time was 20.67 and 69.73 months, respectively (all p-value < 0.05). In the examination of the tumor grade, the effect of grades 2 and 3 tumors compare with the grade 1 tumor on the 50th percentile of survival time were 22.84 and 35.89 months, respectively (all p-value < 0.05). The frailty variance was significant, which confirmed that, there was significant variability between the centers. This study confirmed the usefulness of a censored quantile regression model for cluster data in studying the impact of prognostic factors on survival time and the control effect of heterogeneity due to the treatment of patients in different centers.

Increasing cluster size asymptotics for nested error regression models

This paper establishes asymptotic results for the maximum likelihood and restricted maximum likelihood (REML) estimators of the parameters in the nested error regression model for clustered data when both of the number of independent clusters and the cluster sizes (the number of observations in each cluster) go to infinity. Under very mild conditions, the estimators are shown to be asymptotically normal with an elegantly structured covariance matrix. There are no restrictions on the rate at which the cluster size tends to infinity but it turns out that we need to treat within cluster parameters (i.e. coefficients of unit-level covariates that vary within clusters and the within cluster variance) differently from between cluster parameters (i.e. coefficients of cluster-level covariates that are constant within clusters and the between cluster variance) because they require different normalisations and are asymptotically independent.

A random‐effects model for clustered circular data

AbstractThis article considers a circular regression model for clustered data, where both the cluster effects and the regression errors have von Mises distributions. It involvesβ, a vector of parameters for the fixed effects, and two concentration parameters for the error distribution. A measure of intra‐cluster circular correlation and a predictor for an unobserved cluster random effect are studied. Preliminary estimators for the vectorβand the two concentration parameters are proposed, and their performance is compared with that of the maximum likelihood estimators in a simulation study. A numerical example investigating the factors impacting the orientation taken by a sand hopper when released is presented.The Canadian Journal of Statistics47: 712–728; 2019 © 2019 Statistical Society of Canada

Physical activity during adolescence and risk of colorectal adenoma later in life: results from the Nurses\u2019 Health Study II

BackgroundPhysical activity during adulthood has been consistently associated with lower risk of colorectal cancers, but whether physical activity during adolescence may also play a role in colorectal carcinogenesis is unclear.MethodsWe included 28,250 women in the Nurses’ Health Study II who provided data on physical activity during adolescence (ages 12–22 years) in 1997 and underwent lower bowel endoscopy (1998–2011). We used logistic regression models for clustered data to examine the association between physical activity during adolescence and risk of adenoma later in life.ResultsPhysical activity during adolescence was inversely associated with risk of colorectal adenoma (2373 cases), independent of physical activity during adulthood. The multivariable-adjusted odds ratio (OR) of adenoma was 0.89 (95% CI 0.77–1.02; Ptrend = 0.03) comparing women with ≥ 72 metabolic equivalent of tasks-hours/week (MET-h/week) to < 21 MET-h/week. Women with high physical activity during both adolescence (≥53.3 MET-h/week) and adulthood (≥23.1 MET-h/week) had significantly lower risk of adenoma (all adenomas: OR 0.76; 95% CI 0.66–0.88; advanced adenoma: OR 0.61; 95% CI 0.45–0.82) compared to women with low physical activity during both stages of life.ConclusionsOur findings suggest that physical activity during adolescence may lower the risk of colorectal adenoma later in life.

Image-based teleconsultation using smartphones or tablets: qualitative assessment of medical experts

BackgroundMobile health has promising potential in improving healthcare delivery by facilitating access to expert advice. Enabling experts to review images on their smartphone or tablet may save valuable time. This...

A prospective study of oral contraceptive use and colorectal adenomas.

The influence of reproductive factors on colorectal cancer, including oral contraceptive (OC) use, has been examined, but less research is available on OC use and adenomas. Participants of the Nurses' Health Study who had a lower bowel endoscopy between 1986 (when endoscopies were first assessed) and 2008 were included in this study. Multivariable logistic regression models for clustered data were used to estimate odds ratios and 95% confidence intervals [OR (95% CIs)]. Among 73,058 participants, 51% (n=37,382) reported ever using OCs. Ever OC use was associated with a slight increase in non-advanced adenomas [OR 1.11, 95% CI (1.02, 1.21)] but not with any other endpoints. Duration of OC use was not associated with adenomas, but longer times since last OC use were associated with increased odds of adenomas [e.g., compared to never use, 15+ years since last use: OR 1.17 (1.07, 1.27)]. Shorter times since last OC use were inversely associated [e.g., ≤4years since last use: OR 0.74 (0.65, 0.84)]. We observed a modest borderline increase in risk of colorectal adenomas with any prior OC use. Additionally, more recent OC use may decrease risk, while exposure in the distant past may modestly increase risk of adenomas.

PS-161 Severe Intraventricular Haemorrhage And Periventricular Leukomalacia Rates In Very Low Gestational Age Infants Admitted To Euroneonet Participant Units

Background/Aim To describe rates and time trends of severe (Grade III/IV) intraventricular haemorrhage (IVH) and cystic periventricular leukomalacia (PVL) and to investigate associated perinatal factors in EuroNeoNet Neonatal Intensive Care Units. Methods Data from 36,607 VLBWI admitted to 194 NICUs from 19 countries (2006–2012). IVH and PVL rates were separately analysed using logistic regression models for clustered data, with a derivation sample to look for associated factors and a validation sample to verify these (70%/30%). Calibration (Hosmer-Lemeshow test) and discrimination (AUC analysis) were assessed. Significance level was 0.05. Results There was a marked, significant difference in severe IVH and PVL rates between Northern, Western and Eastern European regions (IVH rates 5%, 9.3% and 14.5% respectively; PVL: 2%, 3.7% 8.8%). Severe IVH and/or PVL had common risk factors: lower GA, 5’ Apgar score, antenatal steroid use, male gender and need for surfactant. PVL was associated with major birth defects, and IVH with vaginal delivery, multiple birth, lower 1’ Apgar score and need for DR resuscitation manoeuvres. A significant time trend to decreasing IVH rates was found in Western countries (from 10.5% in 2006 to 7.3% in 2012) (H-L; ROC (95% CI): Severe IVH: 0.870; 0.805 (0.797 to 0.813); PVL. 0.198; 0.664 (0.649 to 0.679). Conclusions There is wide variability in IVH and PVL rates across European regions. IVH rates may be decreasing in Western countries. Although both scores could be used to assess quality improvement initiatives, the perinatal model has better predictive value for IVH. Funded by EC DG-SANCO (Contract number:530078).

Abstract 2166: A prospective study on oral contraceptive use and risk of colorectal adenomas

Abstract Background: Most colorectal cancers (CRC) arise from colorectal adenomas. The influence of reproductive factors on CRC, including oral contraceptive (OC) use, has been examined for a number of years, but less research is available on the role of OC use on adenomas. A better understanding of this association will provide insight into the mechanism through which hormones impact colorectal carcinogenesis. Objective: To examine the association of OC use and colorectal adenomas using detailed data from a large prospective cohort study with 32 yrs of follow-up data. Design: The Nurses' Health Study was established in 1976 when 121,701 female registered nurses 30-55 yrs of age completed a mailed questionnaire on medical history and lifestyle factors. We assessed OC use every 2 yrs between 1976-1982 and categorized this use according to ever (≥2 mo) and never use (reference), duration of use (never, ≤1, &amp;gt;1-&amp;lt;2, ≥2-&amp;lt;5, ≥5-&amp;lt;10, 10+ yrs), and time since last use (never, ≤4, &amp;gt;4-&amp;lt;10, ≥10-&amp;lt;15, 15+ yrs). Participants had to undergo a lower bowel endoscopy during the study period to be eligible for the current analyses. All colorectal adenoma cases identified at endoscopy were included from 1986 (when endoscopies were first accessed) through 2010; cases were stratified by subsite (proximal, distal, rectal), stage/size (small/early, large/advanced), and number of adenomas (1, 2+). Methods: Multivariable logistic regression models for clustered data were used to estimate odds ratios [OR (95% CIs)]. Models were adjusted for age, body mass index, height, physical activity, smoking, processed and red meat, folate, calcium, total energy, alcohol, aspirin use, age at first birth, parity, hormone therapy use/duration, family history of CRC, time period, as well as endoscopy information including reason, number, and yr since most recent. Results: Among 70,164 participants who had a lower bowel endoscopy, 49% (N=34,329) reported never using OCs and 51% (N=35,835) reported ≥2 mo of use. We recorded 2,950 never-users with an adenoma compared to 3,075 ever-users. Ever OC use was associated with a slight increase of small/early stage adenomas [OR=1.12 95% CI (1.03-1.22)] but not with large/advanced stage adenomas or any other category (e.g., number of adenomas). Duration of OC use was not associated with adenomas but longer times since last OC use were associated with increased odds for all of the adenoma outcomes [e.g., compared to never use, 15+ yrs since last use for all adenomas: OR=1.18 (1.08, 1.29)]. Shorter times since last OC use were inversely associated [e.g., ≤4 yrs since last use for all adenomas: OR=0.76 (0.67, 0.86)]. Conclusions: The null association between OC use and colorectal adenomas may be more nuanced than initially described in the literature. We found that OC use was associated with certain stages/sizes of colorectal adenomas (e.g., small/early) and further exploration of the time since last OC use associations are warranted. Support: P01CA87969, T32CA09001, T32HD060454 Citation Format: Brittany M. Charlton, Ed Giovannucci, Charles S. Fuchs, Stacey A. Missmer, Bernard A. Rosner, Kana Wu, Karin B. Michels. A prospective study on oral contraceptive use and risk of colorectal adenomas. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 2166. doi:10.1158/1538-7445.AM2014-2166

Dating violence among urban, minority, middle school youth and associated sexual risk behaviors and substance use.

Whereas dating violence among high school students has been linked with sexual risk-taking and substance use, this association has been understudied among early adolescents. We estimated the prevalence of physical and nonphysical dating violence in a sample of middle school students and examined associations between dating violence, sexual, and substance use behaviors. Logistic regression models for clustered data from 7th grade students attending 10 Texas urban middle schools were used to examine cross-sectional associations between dating violence victimization and risk behaviors. The sample (N = 950) was 48.5% African American, 36.0% Hispanic, 55.7% female, mean age 13.1 years (SD 0.64). About 1 in 5 reported physical dating violence victimization, 48.1% reported nonphysical victimization, and 52.6% reported any victimization. Adjusted logistic regression analyses indicated that physical, nonphysical, and any victimization was associated with ever having sex, ever using alcohol, and ever using drugs. Over 50% of sampled middle school students had experienced dating violence, which may be associated with early sexual initiation and substance use. Middle school interventions that prevent dating violence are needed.

Trends and Correlates of Child Passenger Restraint Use in 6 Northwest Tribes: The Native Children Always Ride Safe (Native CARS) Project

We compared proportions of children properly restrained in vehicles in 6 Northwest American Indian tribes in 2003 and 2009, and evaluated risks for improper restraint. During spring 2009 we conducted a vehicle observation survey in Oregon, Washington, and Idaho tribal communities. We estimated the proportions of children riding properly restrained and evaluated correlates of improper restraint via log-binomial regression models for clustered data. We observed 1853 children aged 12 years and younger in 1207 vehicles; 49% rode properly restrained. More children aged 8 years and younger rode properly restrained in 2009 than 2003 (51% vs 29%; P < .001). Older booster seat-eligible children were least likely to ride properly restrained in 2009 (25%). American Indian children were more likely to ride improperly restrained than nonnative children in the same communities. Other risk factors included riding with an unrestrained or nonparent driver, riding where child passenger restraint laws were weaker than national guidelines, and taking a short trip. Although proper restraint has increased, it remains low. Tribe-initiated interventions to improve child passenger restraint use are under way.

Occurrence and clinical correlates of REM sleep behaviour disorder in patients with Parkinson’s disease over time

Objective:To examine the occurrence and clinical and demographic correlates of REM sleep behaviour disorder (RBD) in patients with Parkinson’s disease (PD) in a community-based cohort over 8 years.Methods:231 patients with...

Poisson Regression for Clustered Data

Summary We compare five methods for parameter estimation of a Poisson regression model for clustered data: (1) ordinary (naive) Poisson regression (OP), which ignores intracluster correlation, (2) Poisson regression with fixed cluster‐specific intercepts (FI), (3) a generalized estimating equations (GEE) approach with an equi‐correlation matrix, (4) an exact generalized estimating equations (EGEE) approach with an exact covariance matrix, and (5) maximum likelihood (ML). Special attention is given to the simplest case of the Poisson regression with a cluster‐specific intercept random when the asymptotic covariance matrix is obtained in closed form. We prove that methods 1–5, except GEE, produce the same estimates of slope coefficients for balanced data (an equal number of observations in each cluster and the same vectors of covariates). All five methods lead to consistent estimates of slopes but have different efficiency for unbalanced data design. It is shown that the FI approach can be derived as a limiting case of maximum likelihood when the cluster variance increases to infinity. Exact asymptotic covariance matrices are derived for each method. In terms of asymptotic efficiency, the methods split into two groups: OP &amp; GEE and EGEE &amp; FI &amp; ML. Thus, contrary to the existing practice, there is no advantage in using GEE because it is substantially outperformed by EGEE and FI. In particular, EGEE does not require integration and is easy to compute with the asymptotic variances of the slope estimates close to those of the ML.

Pobreza, jóvenes y consumo de tabaco en México

To characterize tobacco use according to level of poverty in a random, nationally representative sample of adolescents (10 to 21 years old), living in urban areas with less than 50,000 inhabitants. The study was done in 2001 as part of the baseline assessment of the evaluation of the governmental program, Oportunidades. A questionnaire was applied to 29,548 adolescents living in 30 000 selected households and it included specific questions on individual tobacco use among other questions. The prevalence of smokers was 3.5% (95% CI: 3.3%-3.7%) and experimenters 9.9% (95% CI: 9.6%-10.2%). A logistic regression model for clustered data was constructed in order to evaluate the associated factors that distinguish a smoker from an experimenter. After adjusting for level of poverty of the household and use of alcohol and drugs, a significant association (OR = 1.5, p <0.01) was found with having a paid job and a differential association was found between gender and age group. The results of this study suggest that the additional availability of money that an adolescent has, could increase the prevalence of tobacco smoking and that the program Oportunidades should include prevention campaigns directed specifically at this population group.

Random-effects regression models for clustered data with an example from smoking prevention research.

Random-effects regression models for clustered data with an example from smoking prevention research.

The Effect of Retrospective Sampling on Binary Regression Models for Clustered Data

Recently a great deal of attention has been given to binary regression models for clustered or correlated observations. The data of interest are of the form of a binary dependent or response variable, together with independent variables X1,...., Xk, where sets of observations are grouped together into clusters. A number of models and methods of analysis have been suggested to study such data. Many of these are extensions in some way of the familiar logistic regression model for binary data that are not grouped (i.e., each cluster is of size 1). In general, the analyses of these clustered data models proceed by assuming that the observed clusters are a simple random sample of clusters selected from a population of clusters. In this paper, we consider the application of these procedures to the case where the clusters are selected randomly in a manner that depends on the pattern of responses in the cluster. For example, we show that ignoring the retrospective nature of the sample design, by fitting standard logistic regression models for clustered binary data, may result in misleading estimates of the effects of covariates and the precision of estimated regression coefficients.