Estimating Multilevel Logistic Regression Models When the Number of Clusters is Low: A Comparison of Different Statistical Software Procedures

In this paper, multilevel logistic regression models are developed for examining the hierarchical effects of contraceptive use and its selected determinants in Oman using the 2008 Oman National Reproductive Health Survey (ONRHS). Comparison between single level and multilevel logistic regression models has been made to examine the plausibility of multilevel effects of contraceptive use. From the multilevel logistic regression model analysis, it was found that there is real multilevel variation among contraceptive users in Oman. The results indicate that a multilevel logistic regression model is the best fit over ordinary multiple logistic regression models. Generally, this study revealed that women’s age, education, number of living children and region of residence are important factors that affect contraceptive use in Oman. The effect of regional variation for age of women, education of women and number of living children further implies that there exists considerable differences in modern contraceptive use among regions, and a model with a random coefficient or slope is more appropriate to explain the regional variation than a model with fixed coefficients or without random effects. The study suggests that researchers should use multilevel models rather than traditional regression methods when their data structure is hierarchal.

Download Full-text

The Association of Sleep Duration With Vision Impairment in Middle-Aged and Elderly Adults: Evidence From the China Health and Retirement Longitudinal Study

Frontiers in Medicine ◽

10.3389/fmed.2021.778117 ◽

2021 ◽

Vol 8 ◽

Author(s):

Mengsha Sun ◽

Qiyu Bo ◽

Bing Lu ◽

Xiaodong Sun ◽

Minwen Zhou

Keyword(s):

Logistic Regression ◽

Longitudinal Study ◽

Sleep Duration ◽

Regression Models ◽

Vision Impairment ◽

Multilevel Logistic Regression ◽

Elderly Adults ◽

Middle Aged ◽

Logistic Regression Models ◽

Long Sleep

Objective: This study aims to investigate the association of sleep duration with vision impairment (VI) in middle-aged and elderly adults.Methods: This cross-sectional study used the data from the baseline survey of the China Health and Retirement Longitudinal Study (CHARLS) 2011–2012, a national survey of adults aged 45 years or older. Weighted multilevel logistic regression models were used to evaluate the association between self-reported sleep duration and VI.Results: Of the 13,959 survey respondents, a total of 4,776 (34.2%) reported VI. The prevalence of short (≤6 h/night) and long (>8 h/night) sleep durations was higher among respondents with VI than those without VI (P < 0.001). Multilevel logistic regression models showed that compared with a sleep duration of 6–8 h/night, a sleep duration of ≤6 h/night was associated with a 1.45-fold [95% confidence interval (CI) = 1.34–1.56] higher VI risk, and a sleep duration of >8 h/night was associated with a 1.18-fold (95% CI = 1.03–1.34) higher VI risk, after adjusting for sociodemographic data, lifestyle factors, and health conditions. Vision impairment was associated with short sleep duration in respondents from all age or gender categories. However, VI was associated with long sleep duration in respondents from the elderly or female categories. The association between VI and long sleep duration disappeared in respondents of middle-aged or male categories.Conclusions: The potential impact of sleep on the risk of visual functions requires further attention. A more comprehensive and integrated health care and rehabilitation system covering vision and sleep is also needed.

Download Full-text

Do Deviations From Historical Precipitation Trends Influence Child Nutrition? An Analysis From Uganda

American Journal of Epidemiology ◽

10.1093/aje/kwz179 ◽

2019 ◽

Vol 188 (11) ◽

pp. 1953-1960 ◽

Cited By ~ 2

Author(s):

Adrienne Epstein ◽

Jacqueline M Torres ◽

M Maria Glymour ◽

David López-Carr ◽

Sheri D Weiser

Keyword(s):

Logistic Regression ◽

Regression Models ◽

Child Nutrition ◽

Multilevel Logistic Regression ◽

Panel Survey ◽

Logistic Regression Models ◽

Child Nutritional Status ◽

Marginal Risk ◽

Precipitation Trends

Abstract Changes in precipitation patterns might have deleterious effects on population health. We used data from the Uganda National Panel Survey from 2009 to 2012 (n = 3,223 children contributing 5,013 assessments) to evaluate the link between rainfall and undernutrition in children under age 5 years. We considered 3 outcomes (underweight, wasting, and stunting) and measured precipitation using household-reported drought and deviations from long-term precipitation trends measured by satellite. We specified multilevel logistic regression models with random effects for the community, village, and individual. Underweight (13%), wasting (4%), and stunting (33%) were common. Reported drought was associated with underweight (marginal risk ratio (RR) = 1.18, 95% confidence interval (CI): 1.04, 1.35) in adjusted analyses. Positive annual deviations (greater rainfall) from long-term precipitation trends were protective against underweight (marginal RR per 50-mm increase = 0.94, 95% CI: 0.92, 0.97) and wasting (marginal RR per 50-mm increase = 0.93, 95% CI: 0.87, 0.98) but not stunting (marginal RR per 50-mm increase = 1.00, 95% CI: 0.98, 1.01). Precipitation was associated with measures of acute but not chronic malnutrition using both objective and subjective measures of exposure. Sudden reductions in rainfall are likely to have acute adverse effects on child nutritional status.

Download Full-text

The Association of Sleep Duration with Visual Impairment in Middle-aged and Elderly Adults: The China Health and Retirement Longitudinal Baseline

10.21203/rs.3.rs-606887/v1 ◽

2021 ◽

Author(s):

Mengsha Sun ◽

Qiyu Bo ◽

Bing Lu ◽

Xiaodong Sun ◽

minwen zhou

Keyword(s):

Logistic Regression ◽

Visual Impairment ◽

Sleep Disorders ◽

Sleep Duration ◽

Regression Models ◽

Multilevel Logistic Regression ◽

Short Sleep Duration ◽

Logistic Regression Models ◽

Short Sleep ◽

Long Sleep

Abstract Background Sleep disorders may heighten the risk of visual impairment to further impact health outcomes. Little is known regarding the association of visual impairment with sleep disorders in China. Our objective was to examine the association of visual impairment with sleep disorders. Methods This cross-sectional study used the data from 13264 respondents to the 2011 survey of the China Health and Retirement Longitudinal Study, a nationally representative survey of adults aged 45 years or older. Visual impairment (VI) and sleep duration were examined using self-reported questionnaires. Respondents were identified as having VI if they reported blindness or partial blindness. With regards to sleep duration, participants were categorized into three groups: 1) those reporting short sleep duration (≤ 6 hours/night), 2) those reporting long sleep duration (> 8 hours/night), and 3) those reporting 6 to 8 hours of sleep per night (used as the reference group). Weighted multilevel logistic regression models, adjusting for sociodemographic characteristics, health behaviors, and medical history, were used. Results Of 13,264 respondents, 6,880 (51.9%) were women. The mean, standard deviation (SD) age was 59.39 (9.71) years. A total of 842 (6.3%) of respondents reported VI. The prevalence of short and long sleep duration was significantly higher among respondents with VI than those without VI (P < 0.001). The associations also persisted after stratifying the sample by age or sex. Multilevel logistic regression models showed that compared with 6–8 h/night of sleep, sleep duration of ≤ 6h/night was associated with a 1.19-fold (95% confidence interval (CI) = 1.02–1.40) higher VI risk, and sleep duration of > 8 h/night was associated with a 1.36-fold (95% CI = 1.05–1.75) higher VI risk. Higher risk of VI was associated with short (odds ratio [OR] = 1.34, 95% CI: 1.04–1.73) and long (OR = 1.60, 95% CI: 1.04–2.44) sleep durations in middle-aged respondents, as well as short sleep duration (OR = 1.27, 95% CI: 1.05–1.55) in elderly respondents. However, the association between VI and long sleep duration (OR = 1.34, 95% CI = 0.97–1.84) was absent in elderly respondents. Conclusion In this study, both short and long sleep durations were associated with VI. More comprehensive and integrated health care and rehabilitation systems covering vision and sleep are needed to address age-related VI.

Download Full-text

Risk of late cervical cancer screening according to prosperity and medical density in daily frequented neighborhoods in Greater Paris.

10.21203/rs.2.18637/v1 ◽

2019 ◽

Author(s):

Médicoulé TRAORE ◽

Julie Vallée ◽

Pierre CHAUVIN

Keyword(s):

Cervical Cancer ◽

Logistic Regression ◽

Cancer Screening ◽

Cervical Cancer Screening ◽

Regression Models ◽

Individual Characteristics ◽

Multilevel Logistic Regression ◽

Logistic Regression Models ◽

Daily Mobility ◽

Paris Area

Abstract Background: The consideration of multiple spaces frequented daily by individuals is attracting interest for the analysis of socioterritorial health and healthcare inequalities in light of the high daily mobility in urban settings and the increasing availability of mobility data. Our objective was to estimate the associations between attributes of daily frequented neighborhoods and delayed cervical smear tests in the Greater Paris area. Methods: Data were extracted from the 2010 SIRS cohort survey. Participants could report three neighborhoods (residence, work, and the next most regularly frequented). All multivariate analyses were conducted: simple multilevel logistic regression models, cross-classified multilevel logistic regression models were used to simultaneously consider the three types of neighborhoods studied (residential, work or study, visit) with active and mobile women. Finally, associations with socioeconomic and medical diversity scores (adjusted for the five individual characteristics) were estimated by logistic regression models that took sampling design into consideration. Results: One-quarter of the women reported that they had not had a smear test in the previous three years. After adjusting for individual characteristics, there was a significant association between the socioeconomic and medical diversity scores for the multiple neighborhoods frequented and the risk of a delayed smear test. Women who reside and work in poor neighborhoods and whose next most regularly frequented neighborhood was also poor had a significantly higher risk of late cervical cancer screening. Conclusions: In the characterization of social and territorial inequalities in healthcare, social epidemiology and health geography show a growing interest in considering multiple spaces frequented daily by individuals. A cumulative exposure score, such as the one presented here, may be a relevant approach for analyzing their effects. Keywords: Multilevel analysis, neighborhood, daily mobility, cancer prevention, cervical cancer, social inequalities, Paris area

Download Full-text

Bootstrapped inference for variance parameters, measures of heterogeneity and random effects in multilevel logistic regression models

Journal of Statistical Computation and Simulation ◽

10.1080/00949655.2020.1797738 ◽

2020 ◽

Vol 90 (17) ◽

pp. 3175-3199 ◽

Cited By ~ 1

Author(s):

Peter C. Austin ◽

George Leckie

Keyword(s):

Logistic Regression ◽

Random Effects ◽

Regression Models ◽

Multilevel Logistic Regression ◽

Logistic Regression Models ◽

Variance Parameters

Download Full-text

Analyzing the Impact of Practice Environment on Nurse Burnout Using Conventional and Multilevel Logistic Regression Models

Workplace Health & Safety ◽

10.1177/2165079919900796 ◽

2020 ◽

Vol 68 (7) ◽

pp. 325-336

Author(s):

Li Yuan ◽

Chen Yumeng ◽

Zhou Chunfen ◽

Fang Jinbo

Keyword(s):

Logistic Regression ◽

Regression Models ◽

Nursing Practice ◽

Multilevel Logistic Regression ◽

Practice Environment ◽

Logistic Regression Models ◽

Nurse Burnout ◽

Nursing Practice Environment ◽

The Impact ◽

Nursing Practice Environments

Background: Most of the previous studies on nursing practice environment and job burnout employed conventional analyses ignoring the impact of unit-level data clusters. This study addressed this gap by examining the effects of the nursing practice environments on dimensions of occupational burnout among a sample of Chinese nurses using multilevel logistic regression models and demonstrating the superiority of employing multilevel models over conventional models within this context. Methods: A proportionate stratified sampling method was applied in this cross-sectional study that invited 1,300 registered nurses (RNs) from nine clinical units of a large, academic hospital in urban China to complete the questionnaire. Nurse-reported information was obtained using the Practice Environment Scale of the Nursing Work Index (PES-NWI) and the Maslach Burnout Inventory (MBI). Findings: A total of 1,178 valid questionnaires were returned for a response rate of 90.62%. RNs generally perceived their nursing practice environment as favorable as measured by the PES-NWI. Approximately 40% of the respondents reported experiencing emotional exhaustion and depersonalization. The multivariate models indicated that nurse burnout was significantly associated with nurse participation in hospital affairs, nursing foundations for quality of care, and adequate staffing. In addition, our results illustrated the advantage of multilevel modeling over the conventional modeling for handling hierarchical data in terms of the accuracy of the estimates and the goodness-of-fit of the model. Conclusions/Application to Practice: These findings underscore the importance of measures aimed at enhancing nursing practice environments to prevent RNs from experiencing feelings of burnout and of considering multilevel analysis in future nursing research.

Download Full-text