Multilevel Logit Models Research Articles

Non-linear associations between zonal built environment attributes and transit commuting mode choice accounting for spatial heterogeneity

Understanding how built environment attributes are associated with transit commuting mode choice is essential for planners to promoting transit through land use and transportation policies. Scholars usually assume that their relationships follow a (generalized) linear pattern and are homogeneous over space. These assumptions may lead to inconsistent estimates. This study develops a semi-parametric multilevel mixed logit model to identify non-linear and spatially heterogeneous relationships between built environment attributes and transit commuting in Nanjing, China. The results show that built environment variables in residential areas have saliently non-linear associations with transit commuting, and the associations vary across traffic analysis zones. Densification facilitates transit use but it has a diminishing return. A medium level of mixed-use is conducive to transit commuting. Transit supply has to exceed a certain threshold to be effective. These findings offer nuanced guidance for transit-oriented neighborhood planning.

Interpersonal and Intrapersonal Variabilities in Daily Activity-Travel Patterns: A Networked Spatiotemporal Analysis

Interpersonal and intrapersonal variabilities are two important perspectives to understand daily travel behaviors, while only a small number of studies incorporate them for understanding human dynamics. This paper employed a network analysis approach to detecting daily activity-travel patterns of 680 Beijing’s residents within a week and then used a multilevel multinomial logit model to analyze the intrapersonal variability in patterns and the socioeconomic linkages behind them. Results suggest that most activity-travel patterns have significant day-to-day intrapersonal and interpersonal variabilities. This suggests that the application of a typical day of activity-travel behaviors to measure and represent a week’s or even longer-term behaviors may be biased, due to the existence of day-to-day intrapersonal variability. This study also provides a hint for the selection of days of a week to conduct a diary survey for activity pattern mining or travel demand modeling.

Determinants of patient-reported outcome trajectories and symptomatic recovery in Improving Access to Psychological Therapies (IAPT) services.

Despite evidence for the general effectiveness of psychological therapies, there exists substantial heterogeneity in patient outcomes. We aimed to identify factors associated with baseline severity of depression and anxiety symptoms, rate of symptomatic change over the course of therapy, and symptomatic recovery in a primary mental health care setting. Using data from a service evaluation involving 35 527 patients in England's psychological and wellbeing [Improving Access to Psychological Therapies (IAPT)] services, we applied latent growth models to explore which routinely-collected sociodemographic, clinical, and therapeutic variables were associated with baseline symptom severity and rate of symptomatic change. We used a multilevel logit model to determine variables associated with symptomatic recovery. Being female, younger, more functionally impaired, and more socioeconomically disadvantaged was associated with higher baseline severity of both depression and anxiety symptoms. Being older, less functionally impaired, and having more severe baseline symptomatology was associated with more rapid improvement of both depression and anxiety symptoms (male gender and greater socioeconomic disadvantage were further associated with rate of change for depression only). Therapy intensity and appointment frequency seemed to have no correlation with rate of symptomatic improvement. Patients with lower baseline symptom severity, less functional impairment, and older age had a greater likelihood of achieving symptomatic recovery (as defined by IAPT criteria). We must continue to investigate how best to tailor psychotherapeutic interventions to fit patients' needs. Patients who begin therapy with more severe depression and/or anxiety symptoms and poorer functioning merit special attention, as these characteristics may negatively impact recovery.

Swing Voters’ Vote Choice Prediction Using Multilevel Logit Model to Improve Election Survey Accuracy

Public opinion surveys are often used to predict an election result. However, the predictions are not always accurate due to many factors. The presence of swing voters at the time of survey is one of the sources of the inaccuracy. On the other hand, election surveys are also often conducted by using multi-stage random sampling method so that ordinary models such as logit model generally do not provide satisfactory results. The data, hence, is complex and may be approached by multilevel models. The study is conducted to assess the extent to which a prediction of swing voters’ vote choice through a multilevel logit model can improve survey accuracy. The data used in this study was the result of a survey conducted using stratified multistage random sampling method two weeks before the 2019 presidential election. The model with 15 predictors and random effects for villages and neighborhood providing 96.3% accuracy and AUC reached 99.1% in the validation process. Based on the final model, the swing voters in this survey were predicted to vote more for Candidate B (10.4%) than Candidate A (7.5%). The direction of the swing voters’ support different from the loyal voters who prefer Candidate A (49.1%) than Candidate B (33.0%). The prediction of swing voters’ vote choice using multilevel logit model significantly improved the survey accuracy. Before the swing voters’ support was predicted the absolute deviation between the survey result and the election result was quite large, around 6.4%-11.5%. After swing voters’ support was predicted the absolute difference shrank to 1.1%.

Heterogeneous relationships between the health-related lifestyle and risk of overweight and obesity in urbanizing China

BackgroundHealth-related lifestyle is widely believed to play an essential role in the prevalence of overweight and obesity. However, it is unclear how urbanization moderates the effect of health-related lifestyle on overweight and obesity. MethodsBased on survey data of 8266 individuals across China, we applied several multilevel logit models to examine whether the effect of health-related lifestyle on overweight and obesity differ between urban and rural areas and how the effect varies with urbanization level. ResultsWe found differences in people's health-related lifestyle between urban and rural areas and variations with urbanization levels. The estimated results showed that most health-related lifestyle variables influence individual’ risks of overweight and obesity in urban areas, whereas only car ownership, exercise time and smoking show effects on being overweight and obese in rural areas. Except for active commuting time, and exercise time, all other health-related lifestyle variables showed different effects on overweight and obesity in areas with different urbanization levels. ConclusionsThese findings provide a better understanding of the role played by urbanization in moderating the effects of health-related lifestyle on overweight and obesity. It suggests that, to mitigate the overweight epidemic, policy makers should pay more attention to local characteristics when promoting related strategies.

Inequality in hospitalization due to non-communicable diseases in Sweden: Age-cohort analysis of the Uppsala Birth Cohort Multigenerational Study.

We aimed to investigate cohort differences in age trajectories of hospitalization due to non-communicable conditions, and if these varied by paternal socioeconomic position. We used the Uppsala Birth Cohort Multigenerational Study—including virtually complete information on medical diagnoses.Our sample constituted 28,448 individuals (103,262 observations). The outcome was five-year prevalence of hospitalization due to major non-communicable conditions in 1989–2008. The exposures were age (19–91), year-of-birth (1915–1929; 1938–1972), gender (man vs woman), and parental socioeconomic position (low, medium, and high). We used multilevel logit models to examine associations between exposures and the hospitalization outcome.Younger cohorts had a higher prevalence of hospitalization at overlapping ages than those born earlier, with inter-cohort differences emerging from early-adulthood and increasing with age. For instance, at age 40 predicted probability of hospitalization increased across birth-cohorts—from 1.2% (born in 1948-52) to 2.0% (born in 1963-67)—whereas at age 50 it was 2.9% for those born in 1938-42 compared with 4.6% among participants born in 1953-57. Those with medium and low socioeconomic position had 13.0% and 20.0% higher odds of experiencing hospitalization during the observation period, respectively—when age, year-of-birth and gender were accounted for.We found that no progress was made in reducing the socioeconomic inequalities in hospitalization across cohorts born between 1915 and 1972. Hence, more effective policies and interventions are needed to reduce the overall burden of morbidity—particularly among the most vulnerable.

Computerized Conjoint Analysis of the Weight Treatment Preferences of Individuals With Schizophrenia.

Innovative approaches are needed for assessing treatment preferences of individuals with schizophrenia. Conjoint analysis methods may help to identify preferences, but the usability and validity of these methods for individuals with schizophrenia remain unclear. This study examined computerized conjoint analysis for persons with schizophrenia and whether preferences for weight management programs predict service use. A computerized, patient-facing conjoint analysis system was developed through iterative consultation with 35 individuals with schizophrenia enrolled at a community mental health clinic. An additional 35 overweight participants with schizophrenia then used the system to choose among psychosocial weight management programs varying in four attributes: location (community or clinic), delivery mode (Internet or in person), leader (clinician or layperson), and training mode (individual or group). A multilevel logit model with partial preference data determined contributions of each attribute to groupwide preferences. Associations were studied between preferences and use of a psychosocial weight management group. Conjoint analysis system usability was rated highly. Groupwide preferences were significantly influenced by location (p<0.001; clinic was preferred), leader (p=0.02; clinician was preferred), and training mode (p<0.001; group was preferred) but not delivery mode (p=0.68). Preferences did not correlate with age, gender, body mass index, illness severity, or subsequent program use. Participants described barriers to program attendance, including transportation, scheduling, privacy, psychiatric illness, and lack of motivation. Computerized conjoint analysis can produce valid assessments of treatment preferences of persons with schizophrenia and inform treatment development and implementation. Although preferences may affect treatment use, they are one of multiple factors.

Entrepreneurship and social capital: some evidence on micro-spatial interactions

We analyze the determinants of three different entrepreneurship patterns namely, the innovative entrepreneur, the socio-cultural entrepreneur, and the regional differentiation model. We assume that these patterns do not have closed boundaries but overlap one another as a result of the entrepreneurs mobilizing different social capital, namely trust, norms and values, and networks. We test this hypothesis using European Social Survey data for Greece and multi-level logit models. Our results verify the importance of social capital, albeit mixed (both positive and negative) effects are reported in the case of all constructs. Furthermore, the regional differentiation of entrepreneurship coincides with the mobilization of different social capital constructs. To the extent that the study uses data for the pre-crisis period in Greece we might refer to the current results as exploratory. Nonetheless, they present a useful comparison basis with future evidence. The scope of entrepreneurship, as exercised in the country, is sketched and thus valuable policy level information is provided.

Infant mortality among native-born children of immigrants in France, 2008-17: results from a socio-demographic panel survey.

BackgroundWithin Europe, France stands out as a major country that lacks recent and reliable evidence on how infant mortality levels vary among the native-born children of immigrants compared with the native-born children of two parents born in France.MethodsWe used a nationally representative socio-demographic panel consisting of 296 400 births and 980 infant deaths for the period 2008–17. Children of immigrants were defined as being born to at least one parent born abroad and their infant mortality was compared with that of children born to two parents born in France. We first calculated infant mortality rates per 1000 live births. Then, using multi-level logit models, we calculated odds ratios of infant mortality in a series of models adjusting progressively for parental origins (M1), core demographic factors (M2), father's socio-professional category (M3) and area-level urbanicity and deprivation score (M4).ResultsWe documented a substantial amount of excess infant mortality among those children born to at least one parent from Eastern Europe, Northern Africa, Western Africa, Other Sub-Saharan Africa and the Americas, with variation among specific origin countries belonging to these groups. In most of these cases, the excess infant mortality levels persisted after adjusting for all individual-level and area-level factors.ConclusionsOur findings, which can directly inform national public health policy, reaffirm the persistence of longstanding inequality in infant mortality according to parental origins in France and add to a growing body of evidence documenting excess infant mortality among the children of immigrants in Europe.

Going Off the Grid: Assessing Sentencing Departures for Individuals Convicted of Sexual Offenses

This research examines factors associated with departures from sentencing guidelines for a sample of individuals convicted of sexual offenses in Pennsylvania between 2004 and 2015 ( N = 26,093). We contribute to the literature on the sentencing of these cases by considering whether the impact of individual and contextual factors is conditioned by the type of sexual offense being sentenced (e.g., assaultive, child pornography, failure to comply with registration). Bayesian multilevel multinomial logit models suggested a number of legal, extralegal, and contextual correlates of departure likelihood, and indicated that the effects of offense severity, prior record, a previous sex offense conviction, multiple concurrent convictions, and court size vary across different subsets of sexual convictions.

Creativity and security as a cultural recipe for entrepreneurship

AbstractWe posit that societal cultural values of creativity and security are associated with the likelihood that a person will engage in a business start-up. Creativity supports the opportunity identification and security the opportunity exploitation aspects of entrepreneurship. In contrast, both emphasis on performance and acceptance of risk-taking may not play the role that is typically assumed. To verify our hypotheses we construct a multilevel dataset, combining Global Entrepreneurship Monitor individual-level data with country-level data from the World Values Survey. We use a multilevel logit model to address the hierarchical structure of our data. We found that odds of start-up engagement are higher if people in a society value security, yet also appreciate thinking up new ideas and being creative. Our results support McCloskey's distinction between aristocratic and bourgeois values, and John and Storr's proposition that different cultural traits support different aspects of entrepreneurship.

Inpatient and Discharge Fluoroquinolone Prescribing in Veterans’ Affairs Hospitals Between 2014 and 2017

Background: Between 2007 and 2015, inpatient fluoroquinolone use declined in US Veterans’ Affairs (VA) hospitals. Whether fluoroquinolone use at discharge has also declined, in particular since antibiotic stewardship programs became mandated at VA hospitals in 2014, is unknown. Methods: In this retrospective cohort study of hospitalizations with infection between January 1, 2014, and December 31, 2017, at 125 VA hospitals, we assessed inpatient and discharge fluoroquinolone (ciprofloxacin, levofloxacin, and moxifloxacin) use as (1) proportion of hospitalizations with a fluoroquinolone prescribed and (2) fluoroquinolone days per 1,000 hospitalizations. After adjusting for illness severity, comorbidities, and age, we used multilevel logit and negative binomial models to assess for hospital-level variation and longitudinal prescribing trends. Results: Of 560,219 hospitalizations meeting inclusion criteria as hospitalizations with infection (Fig. 1), 209,602 of 560,219 (37.4%) had a fluoroquinolone prescribed either during hospitalization (182,337 of 560,219, 32.5%) or at discharge (110,003 of 560,219, 19.6%) (Fig. 1). Hospitals varied appreciably in inpatient, discharge, and total fluoroquinolone use, with 71% of hospitals in the highest prescribing quartile located in the southern United States. Nearly all measures of fluoroquinolone use decreased between 2014 and 2017, with the largest decreases found in inpatient fluoroquinolone and ciprofloxacin use (Fig. 2). In contrast, there was minimal decline in fluoroquinolone use at discharge (Fig. 2), which accounted for 1,433 of 2,339 (61.3%) of hospitalization-related fluoroquinolone days by 2017. Between 2014 and 2017, fluoroquinolone use decreased in VA hospitals, largely driven by a decrease in inpatient fluoroquinolone (especially ciprofloxacin) use. Fluoroquinolone prescribing at discharge, and levofloxacin prescribing overall, remain prime targets for stewardship.Funding: This work was funded by a locally initiated project grant from the Center for Clinical Management Research at the VA Ann Arbor Healthcare System.Disclosures:Valerie M. Vaughn reports contract research for Blue Cross and Blue Shield of Michigan, the Department of Veterans’ Affairs, the NIH, the SHEA, and the APIC. She also reports fees from the Gordon and Betty Moore Foundation Speaker’s Bureau, the CDC, the Pew Research Trust, Sepsis Alliance, and The Hospital and Health System Association of Pennsylvania.Disclaimer The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the US government.

Multilevel modelling of the risk of malaria among children aged under five years in Nigeria.

Malaria is still a major cause of morbidity and mortality among children aged <5 y (U5s). This study assessed individual, household and community risk factors for malaria in Nigerian U5s. Data from the Nigerian Malaria Health Indicator Survey 2015 were pooled for analyses. This comprised a national survey of 329 clusters. Children aged 6-59 mo who were tested for malaria using microscopy were retained. Multilevel logit model accounting for sampling design was used to assess individual, household and community factors associated with malaria parasitaemia. A total of 5742 children were assessed for malaria parasitaemia with an overall prevalence of 27% (95% CI 26 to 28%). Plasmodium falciparum constituted 98% of the Plasmodium species. There was no significant difference in parasitaemia between older children and those aged ≤12 mo. In adjusted analyses, rural living, northwest region, a household size of >7, dependence on river and rainwater as primary water source were associated with higher odds of parasitaemia, while higher wealth index, all U5s who slept under a bed net and dependence on packaged water were associated with lower odds of parasitaemia. Despite sustained investment in malaria control and prevention, a quarter of the overall study population of U5s have malaria. Across the six geopolitical zones, the highest burden was in children living in the poorest rural households.

Some politics is local: the determinants of engagement in local and national politics across Europe

Understanding the roots of political engagement has been one of the critical tasks performed by students of comparative political behaviour. This paper adds to the literature by examining the determinants of political discussion about local and national affairs in Europe. A series of multilevel logit models are fitted to the data (n = 28,563 from 31 European countries) to test the individual and country level determinants of political discussion about local and national matters. At the individual level, we find that gender, the type of community, the type of civil society organisations people are members of, and their level of education affect the type of politics they engage with. At the macro level, citizens from countries with a higher economic development are more likely to engage in discussions about national affairs, while the impact of local government autonomy does not seem to make individuals more likely to engage in discussions about local politics. The findings suggest that if local politics is considered the share of politically disengaged citizens can be smaller than is typically estimated. The full range of democratic practice may thus remain underappreciated if non-national politics is left out of the picture in the study of political engagement.

Inpatient and Discharge Fluoroquinolone Prescribing in Veterans Affairs Hospitals Between 2014 and 2017.

Between 2007 and 2015, inpatient fluoroquinolone use declined in US Veterans Affairs (VA) hospitals. Whether fluoroquinolone use at discharge also declined, in particular since antibiotic stewardship programs became mandated at VA hospitals in 2014, is unknown. In this retrospective cohort study of hospitalizations with infection between January 1, 2014, and December 31, 2017, at 125 VA hospitals, we assessed inpatient and discharge fluoroquinolone (ciprofloxacin, levofloxacin, moxifloxacin) use as (a) proportion of hospitalizations with a fluoroquinolone prescribed and (b) fluoroquinolone-days per 1000 hospitalizations. After adjusting for illness severity, comorbidities, and age, we used multilevel logit and negative binomial models to assess for hospital-level variation and longitudinal prescribing trends. Of 560219 hospitalizations meeting inclusion criteria as hospitalizations with infection, 37.4% (209602/560219) had a fluoroquinolone prescribed either during hospitalization (32.5%, 182337/560219) or at discharge (19.6%, 110003/560219). Hospitals varied appreciably in inpatient, discharge, and total fluoroquinolone use, with 71% of hospitals in the highest prescribing quartile located in the Southern United States. Nearly all measures of fluoroquinolone use decreased between 2014 and 2017, with the largest decreases found in inpatient fluoroquinolone and ciprofloxacin use. In contrast, there was minimal decline in fluoroquinolone use at discharge, which accounted for a growing percentage of hospitalization-related fluoroquinolone-days (52.0% in 2014; 61.3% by 2017). Between 2014 and 2017, fluoroquinolone use decreased in VA hospitals, largely driven by decreased inpatient fluoroquinolone (especially ciprofloxacin) use. Fluoroquinolone prescribing at discharge, as well as levofloxacin prescribing overall, is a growing target for stewardship.

Cultural values and definitions of career success

AbstractThis paper examines how national culture relates to the ways that individuals define career success. Data are drawn from interviews with 269 professional services employees in 15 countries. Interviews are content coded and linked with country‐level Global Leadership and Organizational Behavior Effectiveness measures of cultural values. We test our hypotheses using a multilevel multinomial logit model. The results demonstrate that the ways in which employees define career success vary across countries, due in part to differences in cultural values after controlling for gender, occupation, job level, and national economic development. We find that employees from countries high in future orientation, uncertainty avoidance, and performance orientation are more likely to define career success in terms of interpersonal outcomes, and those from countries high in collectivism (institutional and in‐group), humane orientation, and gender egalitarianism are more likely to prefer intrapersonal outcomes. We find that employees from countries that are high in assertiveness, uncertainty avoidance, and performance orientation are more likely to define career success in terms of achievement‐oriented outcomes. Finally, we find that employees from countries high in power distance report career success definitions in terms of safety and security outcomes. We discuss the implications of these findings for theories of cultural differences in careers across countries.

A predictive model and socioeconomic and demographic determinants of under-five mortality in Sierra Leone

Sierra Leone is among the countries that recorded high under-five child mortality rate in the world. To design and implement policies that can address this public health challenge, the present study developed a predictive model of factors that explained under-five mortality in Sierra Leone using the 2008 and 2013 Sierra Leone Demographic and Health Survey (SDHS) datasets. LASSO regression technique was used to select the predictors to build the under-five predictive single-level logit and multilevel logit models. Statistical analyses were performed in the R freeware version 3.6.1. About 588 (10.4%) and 1320 (11.1%) children under five were reported dead in 2008 and 2013, respectively. The significant predictors of under-five mortality in Sierra Leone were the total number of children ever born, number of children under five in the household, mother's birth in the last five years, mother's number of living children, and number of household members, household wealth, maternal contraceptive use and intention, number of eligible women in the household, type of toilet facility, sex of the child, and weight of the child at birth. The study identified certain predictors that deserve policy attention and interventions to strengthen the efforts of creating child welfare and survival atmosphere in Sierra Leone.

Individual- and Company-Level Predictors of Receiving Vocational Rehabilitation: A Multilevel Study of Finnish Private Sector Workplaces.

Purpose The aim of this study was to examine the magnitude of company-level variation in vocational rehabilitation (VR) and to determine which individual- and company-level characteristics are associated with receiving VR due to mental disorders, musculoskeletal diseases, and other somatic diseases. Methods A 30% random sample of all Finnish private sector companies with more than 10 employees aged 25-62years at the end of 2010 (5567 companies with 300,601 employees) was followed up for the receipt of VR over the next 6 years. Company size and industry, as well as gender, age, education, social class and sickness absence measured both at the individual- and company-level were used as explanatory variables in multilevel logit models. Results After controlling for the individual-level characteristics, 12% of the variance in VR was attributed to the company level. The proportion was largest in VR due to musculoskeletal diseases. Receiving VR was more common among women, older employees (except the oldest age group), those with low education (particularly due to musculoskeletal diseases), low social class, and previous sickness absence. Receiving VR was more common in larger companies, and in construction and in health and social work, and less common in professional, scientific and technical activities. Furthermore, receiving VR was more common in companies with low proportion of highly educated employees and with higher sickness absence rates. Conclusions Company-level variation in receiving VR was substantial. Adopting the practices of the companies with highest participation in VR could help to avoid work disability problems.

Individual- and company level determinants of vocational rehabilitation: a multilevel study

Abstract Backround Vocational rehabilitation (VR) is considered an important means to combat work ability problems and enable people to continue working despite health problems. We examined the magnitude of company level variation in VR and determined which individual- and company level characteristics are associated with access to VR due to mental disorders, musculoskeletal diseases, and other somatic diseases. Methods A 30% random sample of all Finnish private sector companies with more than 10 employees aged 25-62 years at the end of 2010 (5.567 companies with 300.601 employees) was followed up for incident VR for six years. Company size and industry, as well as gender, age, education, social class and sickness absence measured both at the individual- and company level, were used as explanatory variables in multilevel logit models. Results After controlling for the individual level covariates, companies accounted for 12% of the variance in VR. The proportion was largest in VR due to musculoskeletal diseases. VR was more common among women, older employees (except the very oldest age group), those with low education (particularly due to musculoskeletal diseases), low social class, and previous sickness absence. VR was more common in larger companies, and in construction and in health and social work, and less common in professional activities. VR was more common in companies with low proportion of highly educated employees and with higher sickness absence rates. Conclusions Company level variation in VR was substantial after controlling for individual level characteristics of the employees. The employer may have an important role in the prevention of work disability through vocational rehabilitation. Key messages Company level variation in vocational rehabilitation was substantial after controlling for individual level characteristics of the employees. The employer may have an important in the prevention of work disability through vocational rehabilitation.

Route choice modeling with overlapping in super-networks

In modern transportation networks, there is a common phenomenon of overlapping of physical road sections and transport modes. In this study, the overlapping phenomenon in traffic network is defined as the generalized overlapping problem which is related to the failure of many existing models for users' route selection. When modeling the route selection behavior in the super-networks, the available routes are constrained by the overlapping of both the physical segments and the transport modes. In this paper, we propose a multi-level mixed logit model to investigate the correlation of generalized routes. Based on the proposed model, the traffic assignment of super network can be simplified as fixed point problem. By analyzing the error terms of the physical segments and the transport modes, we discuss the overlapping effects on the route choice. The results show that the proposed model can sole the overlapping problem effectively, which has significant effects on the route choice and traffic assignment problems in practice.