Mixed-effects Multinomial Logistic Regression Research Articles

Aboriginal women have a higher risk of cervical abnormalities at screening; South Australia, 1993-2016.

Cervical cancer mortality has halved in Australia since the national cervical screening program began in 1991, but elevated mortality rates persist for Aboriginal and Torres Strait Islander women (referred to as Aboriginal women in this report). We investigated differences by Aboriginal status in abnormality rates predicted by cervical cytology and confirmed by histological diagnoses among screened women. Using record linkage between cervical screening registry and public hospital records in South Australia, we obtained Aboriginal status of women aged 20-69 for 1993-2016 (this was not recorded by the registry). Differences in cytological abnormalities were investigated by Aboriginal status, using relative risk ratios from mixed effect multinomial logistic regression modelling. Odds ratios were calculated for histological high grade results for Aboriginal compared with non-Aboriginal women. Of 1,676,141 linkable cytology tests, 5.8% were abnormal. Abnormal results were more common for women who were younger, never married, and living in a major city or socioeconomically disadvantaged area. After adjusting for these factors and numbers of screening episodes, the relative risk of a low grade cytological abnormality compared with a normal test was 14% (95% confidence interval 5-24%) higher, and the relative risk of a high grade cytological abnormality was 61% (95% confidence interval 44-79%) higher, for Aboriginal women. The adjusted odds ratio of a histological high grade was 76% (95% confidence interval 46-113%) higher. Ensuring that screen-detected abnormalities are followed up in a timely way by culturally acceptable services is important for reducing differences in cervical cancer rates between Aboriginal and non-Aboriginal women.

Where beavers (Castor canadensis) build: testing the influence of habitat quality, predation risk, and anthropogenic disturbance on colony occurrence

Species distributions are shaped by numerous factors that vary in importance across spatiotemporal scale. Understanding drivers of the distribution of North American beavers (Castor canadensis Kuhl, 1820) is paramount given their profound influence on ecological communities. Our objectives were to evaluate the influence of habitat quality, risk of gray wolf (Canis lupus Linnaeus, 1758) predation, and anthropogenic disturbance on the occurrence of beaver colonies in northeast British Columbia (BC), Canada. We used mixed-effects multinomial logistic regression to model the occurrence of active and inactive colonies and t tests to compare landscape covariates associated with active versus inactive colonies. We determined that occurrence of beavers was driven by habitat quality. Occurrence increased in areas with higher vegetation-class richness and greater proportions of open water, nutrient-rich fen, and deciduous swamp. We also observed that active colonies were surrounded by greater amounts of deciduous swamps relative to inactive colonies. We found no evidence that predation risk or industrial activities decreased the occurrence of beavers in northeast BC, although numerical changes in abundance might occur without changes in distribution. This research illuminated drivers of beaver distribution while providing a means to predict the occurrence of a keystone species in the boreal ecosystem.

Building walls or opening borders? Global immigration policy attitudes across economic, cultural and human security contexts

Using data from the World Values Survey for 51 countries, we conduct a multi-level analysis with mixed effects multinomial logistic regression models to explore the effects of economic context, cultural context, and national security events on immigration policy attitudes. Analyses of attitudes towards immigration to date have been limited in key respects: the scope has been mostly restricted to Western Europe and the Americas; limited attention has been paid to institutional and sociopolitical features of the macro-context; and national security events have been rarely taken into account. We propose a theoretical framework incorporating three overarching contextual domains. We find support for two of the three contextual domains. Contrary to our expectation, macro-level economic context shows no association immigration policy attitudes, and socio-cultural context evinces unexpected patterns. However, in the national security domain, recent terrorist events demonstrate a consistent association with anti-immigration policy attitudes.

Common risk indicators for oral diseases and obesity in 12-year-olds: a South Pacific cross sectional study

BackgroundDespite the increasing need to prevent obesity and oral diseases in adolescents worldwide, few studies have investigated the link existing between these conditions and their common risk factors. This study aims to evaluate the oral health and weight status of New Caledonian Children (aged 6,9,12 years) and to identify, amongst 12-year-olds, risk indicators that may characterize the groups of children affected by oral diseases, obesity or both diseases.MethodsThis survey evaluated in 2011–2012 the oral health and stature-weight status and related risk indicators in a national representative sample of 6, 9 and 12 years-old children in New Caledonia. Dental status, chewing efficiency, height, weight and waist circumference were clinically recorded at school. The body mass index (BMI) and the waist to height ratio (WtHR) were calculated. For BMI the WHO Cut-offs were used. Twelve years-old participants responded to a questionnaire concerning socio-demographic and behavioural variables. For statistical analysis, the Clinical Oral and Global Health Index (COGHI) was developed and used. Twelve years-old children were categorised into four groups; Oral Diseases (ODG), Obesity (OG), Obesity and Oral Diseases (ODOG) and a Healthy Group (HG). A multivariate analysis was conducted using mixed-effects multinomial logistic regression models.ResultsPrevalence of overweight and obesity was greatly increasing from 6 years (respectively 10.8% [8.8;13.3] and 7.8% [6.0;9.9]) to 12 years (respectively 22.2% [19.9;24.7] and 20.5% [18.2;22.9]) and one third of the 12-yr-olds had an excess of abdominal adiposity. At age 12, 36.6% of the children were healthy (HG), 27.3% had oral diseases (ODG), 19.7% were obese (OG) and 16.5% had both conditions (ODOG). Geographical location, ethnicity, tooth-brushing frequency and masticatory disability were significant risk factors for the OG, ODOG and ODG groups. Ethnicity and masticatory impairment were common risk indicators for the association of oral diseases and obesity.ConclusionsIn NC health promotion programs should be specifically addressed towards Native populations who are particularly exposed to oral diseases and obesity, integrating a multiple risk factors approach, in order to prevent the onset of chronic diseases in adulthood. The impact of masticatory ability on children’s weight status is a major issue for future research.

The Role of Senior Housing in Hospitalizations Among Vulnerable Older Adults With Multiple Chronic Conditions: A Longitudinal Perspective.

This study focuses on vulnerable elders (i.e., those with moderate or low incomes who live alone) and examines to what extent a senior housing environment moderates the effects of multiple chronic conditions (MCCs) on hospitalization over time. Data came from six waves (2002-2012) of the Health Retirement Study (N = 1,401 individuals, 3,705 observations). Mixed-effect multinomial logistic regression modeling estimated the effects of senior housing on changes in hospitalization. Vulnerable elders with MCCs were more likely to be hospitalized at both moderate and heavy levels at baseline. Consistent with the environment docility hypothesis, findings show that older individuals with MCCs who live in a senior housing environment have fewer hospitalizations over time. As one of the first efforts to build empirical knowledge on health care use among vulnerable elders in senior housing, our findings underscore the importance of continued research into these environments as a possible alternative to existing models.

Water, Sanitation and Hygiene (WASH) and environmental risk factors for soil-transmitted helminth intensity of infection in Timor-Leste, using real time PCR.

BackgroundNo investigations have been undertaken of risk factors for intensity of soil-transmitted helminth (STH) infection in Timor-Leste. This study provides the first analysis of risk factors for intensity of STH infection, as determined by quantitative PCR (qPCR), examining a broad range of water, sanitation and hygiene (WASH) and environmental factors, among communities in Manufahi District, Timor-Leste.MethodsA baseline cross-sectional survey of 18 communities was undertaken as part of a cluster randomised controlled trial, with additional identically-collected data from six other communities. qPCR was used to assess STH infection from stool samples, and questionnaires administered to collect WASH, demographic, and socioeconomic data. Environmental information was obtained from open-access sources and linked to infection outcomes. Mixed-effects multinomial logistic regression was undertaken to assess risk factors for intensity of Necator americanus and Ascaris infection.Results2152 participants provided stool and questionnaire information for this analysis. In adjusted models incorporating WASH, demographic and environmental variables, environmental variables were generally associated with infection intensity for both N. americanus and Ascaris spp. Precipitation (in centimetres) was associated with increased risk of moderate-intensity (adjusted relative risk [ARR] 6.1; 95% confidence interval [CI] 1.9–19.3) and heavy-intensity (ARR 6.6; 95% CI 3.1–14.1) N. americanus infection, as was sandy-loam soil around households (moderate-intensity ARR 2.1; 95% CI 1.0–4.3; heavy-intensity ARR 2.7; 95% CI 1.6–4.5; compared to no infection). For Ascaris, alkaline soil around the household was associated with reduced risk of moderate-intensity infection (ARR 0.21; 95% CI 0.09–0.51), and heavy-intensity infection (ARR 0.04; 95% CI 0.01–0.25). Few WASH risk factors were significant.ConclusionIn this high-prevalence setting, strong risk associations with environmental factors indicate that anthelmintic treatment alone will be insufficient to interrupt STH transmission, as conditions are favourable for ongoing environmental transmission. Integrated STH control strategies should be explored as a priority.

Lymph node yield is an independent predictor of survival in rectal cancer regardless of receipt of neoadjuvant therapy

AimsLymph node yield (LNY) is used as a marker of adequate oncological resection. The American Joint Committee on Cancer (AJCC) currently recommends that at least 12 nodes are necessary to...

Daily diary study of adult men's and women's event-level sexual motivations and sexual behaviour.

Adult men (n=156) and women (n=192) completed thrice-daily electronic diaries assessing individual- and partner-specific attributes and non-coital or coital sexual behaviours. Sexual motivations were: interest in sex, feeling in love with partner, wanted to have sex and partner wanted to have sex. The outcome variable was: sexual behaviour type (no sex, one vaginal sex event, one vaginal sex event+any other sex types, multiple vaginal sex events, any other sex types). Mixed-effect multinomial logistic regression modelled the influence of each sexual motivation on sexual behaviour type (Stata; all p<0.05). 'No sex' was the referent in all models; all models controlled for gender. Participants contributed 14856 total partner-associated diary entries. Most (54%; women: 56.5%, men: 51.2%) were associated with no sex; when sex occurred, the most common behaviour type was one vaginal sex event (13.1%) for women and other sex types (16.4%) for men. Wanting to have sex or perceiving a partner wanted to have sex were the strongest predictors of sexual behaviour type, and were associated with a greater number of reported sexual behaviours. Event-specific sexual motivations are associated with the choice to have sex, and with variation in the chosen sexual behaviours.

Women's views and postpartum follow-up in the CHIPS Trial (Control of Hypertension in Pregnancy Study)

ObjectiveTo compare women's views about blood pressure (BP) control in CHIPS (Control of Hypertension In Pregnancy Study) (NCT01192412). DesignQuantitative and qualitative analysis of questionnaire responses. SettingInternational randomised trial (94 sites, 15 countries). Population/sample911 (92.9%) women randomised to ‘tight’ (target diastolic blood pressure, 85mmHg) or ‘less tight’ (target diastolic blood pressure, 100mmHg) who completed questionnaires. MethodsA questionnaire was administered at ∼6–12 weeks postpartum regarding post-discharge morbidity and views about trial participation. Questionnaires were administered by the site co-ordinator, and contact was made by phone, home or clinic visit; rarely, data was collected from medical records. Quantitative analyses were Chi-square or Fisher's exact test for categorical variables, mixed effects multinomial logistic regression to adjust for confounders, and p<0.001 for statistical significance. NVivo software was used for thematic analysis of women's views. Main outcome measuresSatisfaction, measured as willingness to have the same treatment in another pregnancy or recommend that treatment to a friend. ResultsAmong the 533 women in ‘tight’ (N=265) vs. ‘less tight’ (N=268) control who provided comments for qualitative analysis, women in ‘tight’ (vs. ‘less tight’) control made fewer positive comments about the amount of medication taken (5 vs. 28 women, respectively) and intensity of BP monitoring (7 vs. 17, respectively). However, this did not translate into less willingness to either have the same treatment in another pregnancy (434, 95.8% vs. 423, 92.4%, respectively; p=0.14) or recommend that treatment to a friend (435, 96.0% and 428, 93.4%, respectively; p=0.17). Importantly, although satisfaction remained high among women with an adverse outcome, those in ‘tight’ control who suffered an adverse outcome (vs. those who did not) were not consistently less satisfied, whereas this was not the case among women in ‘less tight’ control among whom satisfaction was consistently lower for the CHIPS primary outcome (p<0.001), severe hypertension (p≤0.01), and pre-eclampsia (p<0.001). ConclusionsWomen in ‘tight’ (vs. ‘less tight’) control were equally satisfied with their care, and more so in the face of adverse perinatal or maternal outcomes.

TECHNOLOGICAL COMPLEXITY: A TOOL FOR UNDERSTANDING THE BEHAVIOUR OF CONSUMERS OF HIGH VALUE-ADDED FOODSTUFFS

This article proposes and develops the concept of technological complexity (TC) as a useful and simple tool for grouping key attributes that give added value to a product. In addition, it reports an empirical application of this concept to two different food products (cured ham and cured sausage). The authors used a mixed-effects multi-nomial logistic regression model and show that in the cured pork product agribusiness, a low frequency of consumption favours the acceptance of high TC products. The results also confirm that marketing high TC products in stores with a large assortment decreases the chances of success for agribusiness companies that produce cured pork food products. These finding can be used by the managers for designing complementary attributes that improve their product portfolio. Besides, advertising expenditures associated with introducing new products could be reduced if companies strengthened their presence in specialty stores.

Tree-based approach to evaluate size dependence of residual tree damage caused by selective logging: Case study in tropical semi-evergreen forests of Cambodia

Logging damage to residual trees is one of the fundamental components in evaluating the sustainability of tropical selective logging in terms of timber production, carbon retention and biodiversity conservation. Although many studies have taken an area-based approach to tropical rain forests, we adopted a tree-based approach to quantify the dependence of residual tree damage on the size of residual and felled trees. We used data from 179 plots, each 25m×40m, covering the stump and crown of one felled tree in Cambodian tropical semi-evergreen forests. We used the mixed-effects multinomial logistic regression model to predict the probability of a residual tree sustaining severe, slight or no damage. Increasing size of residual trees decreased the probability of severe damage and increased that of slight damage. The probability of total damage (severe plus slight) was nearly constant, regardless of the size of residual trees. Increasing size of the felled tree caused greater probability of severe damage, but did not influence the probability of slight damage. Interestingly, our prediction of total damage rate from the tree-based modeling approach of the Cambodian semi-evergreen forests is very consistent with the findings of previous studies using the area-based approach in Indonesia. The departure of our study results from those of other studies may be explained by differences in felled tree size and terrain slope. The inclusion of tree-size dependence in logging damage estimation would increase its accuracy and comparability across different types of tropical forests.

Low body condition predisposes cattle to lameness: An 8-year study of one dairy herd

Lameness in dairy cows is a multifactorial and progressive disease with complex interactions between risk factors contributing to its occurrence. Detailed records were obtained from one United Kingdom dairy herd over an 8-yr period. Weekly locomotion scores were used to classify cows as not lame (score 1 to 2), mildly lame (score 3) and severely lame (score 4 to 5). These outcomes were used to investigate the hypothesis that low body condition score (BCS) is associated with an increased risk of lameness in dairy cows. Mixed effect multinomial logistic regression models were used to investigate the association between prior BCS and repeat lameness events during the longitudinal period of the study. Discrete time survival models were used to explore the relationship between prior BCS and first lifetime lameness events. In total, 79,565 cow weeks at risk were obtained for 724 cows. The number of lameness events was 17,114, of which 8,799 were categorized as mildly lame and 8,315 as severely lame. The median BCS was 2.25 (range, 0.75 to 4.25) and the mean body weight (BW) and age at first calving were 619.5kg (range, 355.6 to 956.4kg) and 25.8mo (range, 20.5 to 37.8mo), respectively. Subsets of the data were used in the discrete time survival models: 333 mild and 211 severe first lifetime lameness events in heifers (first lactation cows), and 81 mild and 49 severe first lifetime lameness events in cows second lactation or greater. Low BCS 3wk before a repeated lameness event was associated with a significantly increased risk of lameness. Cows with BCS <2 were at greatest risk of mild or severe lameness, and an increased BCS above 2 was associated with a reduced risk of mild or severe lameness. Low BCS 16 or 8wk before a first mild or severe lifetime lameness event, respectively, also had a positive association with risk of lameness in cows second lactation or greater. This provides evidence to support targeting management toward maintaining BCS to minimize the risk of lameness. Low BW (independent of BCS) and increased age at first calving above 24mo were also associated with increased long-term risk of repeated lameness events. Overall, the model explained 62 and 60% of the variability for mild and severe lameness, respectively, highlighting the importance of these variables as risk factors and hence where management could be targeted to significantly affect reducing the risk of lameness.

Infant risk factors associated with internalizing, externalizing, and co-occurring behavior problems in young children.

The purpose of the current study was to examine the unique and interactive contributions of infant negative emotionality and family risk factors in the development of internalizing-only, externalizing-only, and co-occurring behavior problems in early childhood. The sample included 412 infants and their primary caregivers. Interviews and temperament assessments took place when infants were 5-7 months old, and primary caregivers completed child behavior ratings at ages 2 1/2 and 5 years. Mixed-effects multinomial logistic regression was used to examine associations between infant risk factors and "pure" and co-occurring child behavior problems, and test whether these associations changed over time. The results of this study showed that hostile parenting during infancy increased the likelihood that children would develop internalizing-only problems, whereas infants who were highly distressed in response to novelty were at increased risk of developing externalizing-only problems. Multiple risk factors, including maternal anxious and depressive symptoms, family conflict, and younger maternal age, independently predicted early childhood co-occurring problems. Additionally, there was a significant interaction between infant anger/frustration and hostile parenting: In the context of hostile parenting, infants high in anger were at increased risk of developing early co-occurring problems, though this association faded by age 5. These findings point to the importance of considering the infant's family context, and differentiating between "pure" and co-occurring behaviors when examining the etiology of early childhood behavior problems.

Beyond political claims: women's interest in and emotional response to viewing their ultrasound image in abortion care.

In the United States, abortion opponents have supported legislation requiring that abortion patients be offered the opportunity to view their preprocedure ultrasound. Little research has examined women's interest in and emotional response to such viewing. Data from 702 women who received abortions at 30 facilities throughout the United States between 2008 and 2010 were analyzed. Mixed-effects multinomial logistic regression analysis was used to determine which characteristics were associated with being offered and choosing to view ultrasounds, and with reporting positive or negative emotional responses to viewing. Grounded theory analytic techniques were used to qualitatively describe women's reports of their emotional responses. Forty-eight percent of participants were offered the opportunity to view their ultrasound, and nulliparous women were more likely than others to receive an offer (odds ratio, 2.3). Sixty-five percent of these women (31% overall) chose to view the image; nulliparous women and those living in a state that regulates viewing were more likely than their counterparts to do so (1.7 and 2.5, respectively). Some 213 women reported emotional responses to viewing; neutral emotions (fine, nothing) were the most commonly reported ones, followed by negative emotions (sad, guilty, upset) and then positive emotions (happy, excited). Women who visited clinics with a policy of offering viewing had increased odds of reporting a negative emotion (2.6). Ultrasound viewing appears not to have a singular emotional effect. The presence of state regulation and facility policies matters for women's interest in and responses to viewing.

Measuring online endorsement of prescription opioids abuse: an integrative methodology

The ability to track prescription opioid abusers' endorsement of compounds may be valuable in forecasting abuser's preferences. We developed a metric, referred to as the Endorsement Ratio (ERo), to explore the feasibility of quantifying the sentiment expressed by opioid abusers who post online about prescription opioid products. Internet posts written between April 1, 2010 and March 31, 2011 and related to six prescription opioid compounds (oxycodone ER, hydrocodone, hydromorphone, oxymorphone, morphine ER, and tramadol) were identified and qualitatively coded. A mixed effects multinomial logistic regression was employed to model the probability of observing endorsing, discouraging, mixed, or unclear messages per compound. Endorsement-to-discouragement ratio (ERo) and ratio of endorsement-to-discouragement ratios (RERo) were estimated for each compound. In the study sample of 12 838 Internet posts, a ranking of endorsement could be established, with oxymorphone as most endorsed (ERo = 7.39), followed by hydromorphone (ERo = 5.02), hydrocodone (ERo = 3.53), oxycodone ER (ERo = 3.23), morphine ER (ERo = 2.71), and tramadol (ERo = 1.76). Relative rankings of products required that route of administration be considered. This study expands existing Internet monitoring approaches by developing a metric to estimate the endorsement of specific prescription opioid compounds as viewed through the lens of Internet communities. Route of administration must be taken into account when examining preferences of drug abusers for various products. This study did not assess whether the novel metric is valid with respect to classification of abuse rates of different drugs. Further studies examining external validation studies are warranted.

Using knowledge brokers to facilitate the uptake of pediatric measurement tools into clinical practice: a before-after intervention study

BackgroundThe use of measurement tools is an essential part of good evidence-based practice; however, physiotherapists (PTs) are not always confident when selecting, administering, and interpreting these tools. The purpose of this study was to evaluate the impact of a multifaceted knowledge translation intervention, using PTs as knowledge brokers (KBs) to facilitate the use in clinical practice of four evidence-based measurement tools designed to evaluate and understand motor function in children with cerebral palsy (CP). The KB model evaluated in this study was designed to overcome many of the barriers to research transfer identified in the literature.MethodsA mixed methods before-after study design was used to evaluate the impact of a six-month KB intervention by 25 KBs on 122 practicing PTs' self-reported knowledge and use of the measurement tools in 28 children's rehabilitation organizations in two regions of Canada. The model was that of PT KBs situated in clinical sites supported by a network of KBs and the research team through a broker to the KBs. Modest financial remuneration to the organizations for the KB time (two hours/week for six months), ongoing resource materials, and personal and intranet support was provided to the KBs. Survey data were collected by questionnaire prior to, immediately following the intervention (six months), and at 12 and 18 months. A mixed effects multinomial logistic regression was used to examine the impact of the intervention over time and by region. The impact of organizational factors was also explored.ResultsPTs' self-reported knowledge of all four measurement tools increased significantly over the six-month intervention, and reported use of three of the four measurement tools also increased. Changes were sustained 12 months later. Organizational culture for research and supervisor expectations were significantly associated with uptake of only one of the four measurement tools.ConclusionsKBs positively influenced PTs' self-reported knowledge and self-reported use of the targeted measurement tools. Further research is warranted to investigate whether this is a feasible, cost-effective model that could be used more broadly in a rehabilitation setting to facilitate the uptake of other measurement tools or evidence-based intervention approaches.

Characteristics of eyewitness identification that predict the outcome of real lineups

AbstractData were analysed from 640 attempts by eyewitnesses to identify the alleged culprit in 314 lineups organized by the Metropolitan Police in London. Characteristics of the witness, the suspect, the witness's opportunity to view the culprit, the crime and the lineup were recorded. Data analysis, using mixed effects multinomial logistic regression, revealed that the suspect was more likely to be identified if the witness is younger than 30, the suspect is a white European (rather than African‐Caribbean), the witness gave a detailed description, viewed the culprit for over a minute and made a fast decision at the lineup. None of the explanatory variables were significantly associated with a mistaken identification of a foil. No independent, statistically reliable effects of weapon focus, cross‐race identification, or of the delay before the identification attempt were observed. Copyright © 2003 John Wiley &amp; Sons, Ltd.

A mixed-effects multinomial logistic regression model.

A mixed-effects multinomial logistic regression model is described for analysis of clustered or longitudinal nominal or ordinal response data. The model is parameterized to allow flexibility in the choice of contrasts used to represent comparisons across the response categories. Estimation is achieved using a maximum marginal likelihood (MML) solution that uses quadrature to numerically integrate over the distribution of random effects. An analysis of a psychiatric data set, in which homeless adults with serious mental illness are repeatedly classified in terms of their living arrangement, is used to illustrate features of the model.