Sample size considerations and predictive performance of multinomial logistic prediction models

What would affect drivers' stop-and-go decisions at yellow dilemma zones? A driving simulator study in Hong Kong

Yellow dilemma, at which a driver can neither stop nor go safely after the onset of yellow signals, is one of the major crash contributory factors at the signal junctions. Studies have visited the yellow dilemma problem using observation surveys. Factors including road environment, traffic conditions, and driver characteristics that affect the driver behaviours are revealed. However, it is rare that the joint effects of situational and attitudinal factors on the driver behaviours at the yellow dilemma zone are considered. In this study, drivers' propensity to stop after the onset of yellow signals is examined using the driving simulator approach. For instances, the association between driver propensity, socio-demographics, safety perception, traffic signals, and traffic and weather conditions are measured using a binary logit model. Additionally, variations in the effect of influencing factors on driver behaviours are accommodated by adding the interaction terms for driver characteristics, traffic flow characteristics, traffic signals, and weather conditions. Results indicate that weather conditions, traffic volume, position of yellow dilemma in the sequence, driver age and safety perception significantly affect the drivers' propensity to stop after the onset of yellow signals. Furthermore, there are remarkable interactions for the effects of driver gender and location of yellow dilemma.

How to develop, validate, and update clinical prediction models using multinomial logistic regression

OBJECTIVES

Multicategory prediction models (MPMs) can be used in health care when the primary outcome of interest has more than two categories. The application of MPMs is scarce, possibly due to added methodological complexities compared to binary outcome models. We provide a guide of how to develop, validate, and update clinical prediction models based on multinomial logistic regression.

STUDY DESIGN AND SETTING

We present guidance and recommendations based on recent methodological literature, illustrated by a previously developed and validated MPM for treatment outcomes in rheumatoid arthritis. Prediction models using multinomial logistic regression can be developed for nominal outcomes, but also for ordinal outcomes. This article is intended to supplement existing general guidance on prediction model research.

RESULTS

This guide is split into three parts: 1) outcome definition and variable selection, 2) model development, and 3) model evaluation (including performance assessment, internal and external validation, and model recalibration). We outline how to evaluate and interpret the predictive performance of MPMs. R code is provided.

CONCLUSION

We recommend the application of MPMs in clinical settings where the prediction of a multicategory outcome is of interest. Future methodological research could focus on MPM-specific considerations for variable selection and sample size criteria for external validation.

Classification of autonomous vehicle crash severity: Solving the problems of imbalanced datasets and small sample size

Only a few researchers have shown how environmental factors and road features relate to Autonomous Vehicle (AV) crash severity levels, and none have focused on the data limitation problems, such as small sample sizes, imbalanced datasets, and high dimensional features. To address these problems, we analyzed an AV crash dataset (2019 to 2021) from the California Department of Motor Vehicles (CA DMV), which included 266 collision reports (51 of those causing injuries). We included external environmental variables by collecting various points of interest (POIs) and roadway features from Open Street Map (OSM) and Data San Francisco (SF). Random Over-Sampling Examples (ROSE) and the Synthetic Minority Over-Sampling Technique (SMOTE) methods were used to balance the dataset and increase the sample size. These two balancing methods were used to expand the dataset and solve the small sample size problem simultaneously. Mutual information, random forest, and XGboost were utilized to address the high dimensional feature and the selection problem caused by including a variety of types of POIs as predictive variables. Because existing studies do not use consistent procedures, we compared the effectiveness of using the feature-selection preprocessing method as the first process to employing the data-balance technique as the first process. Our results showed that AV crash severity levels are related to vehicle manufacturers, vehicle damage level, collision type, vehicle movement, the parties involved in the crash, speed limit, and some types of POIs (areas near transportation, entertainment venues, public places, schools, and medical facilities). Both resampling methods and three data preprocessing methods improved model performance, and the model that used SMOTE and data-balancing first was the best. The results suggest that over-sampling and the feature selection method can improve model prediction performance and define new factors related to AV crash severity levels.

Stage of Change of Transtheoretical Model for Nine Health-Related Behaviors Among Hypertensive Patients: Cross-Sectional Study

Aim

This study aimed to determine the current stage of change (SOG) toward seven healthy eating behaviors and two healthy lifestyle behaviors related to blood pressure (BP) control. The lifestyle behaviors included smoking Behavior and practicing regular exercise, while the dietary behaviors included the DASH diet guidelines.

Methods

A total of 1109 outpatients participated in this cross-sectional study that was conducted between 2021 and 2022 in Jordan. A staging algorithm assessed SOG for several BP control-related behaviors for diagnosed hypertension patients. Data were collected by a structured interview-based questionnaire.

Results

There was a high degree of maintenance toward consuming diets with high grains, fruit, vegetables, meat, and poultry, less saturated fat, and more low-fat dairy products. More than half of the participants were in the pre-action stage for quitting smoking, practicing physical exercise, and consuming sweets and added sugars. Significant associations were observed between the degree of maintenance for several behaviors (p < 0.01). Age, income, education level, disease duration, and nutrition consultation availability were the most related factors to the SOG of the studied behaviors (p < 0.01).

Conclusion

Patients with hypertension in Jordan are still in the pre-action stages for quitting smoking, practicing physical exercise, and consuming 5 servings of refined sweets and added sugars weekly. The current outcome suggests a need for nutritional counseling, education, and interventions to raise awareness of lifestyle factors influencing BP among hypertension patients.

Analysis of injury severity of work zone crashes on rural and urban work zones: Accounting for out-of-sample prediction and temporal instability

This research utilizes data collected in Florida to examine the differentials in injury severities among single-vehicle drivers involved in work zone-related incidents, specifically focusing on the distinctions between rural and urban areas. The study encompasses a four-year period (2016-2019) of crash dataset. A likelihood ratio test was performed to examine model estimate's temporal consistency in datasets from rural and urban areas across several time periods throughout the year. Separate statistical models were estimated for both rural and urban datasets to understand different driver injury severity outcomes (no injury, minor injury, and severe injury) using a mixed logit approach with possible heterogeneity in mean and variance of random parameters. Out-of-sample simulations were conducted to see the effect of different parameter changes on injury severity probabilities in rural and urban work zone crashes. Over multiple years, various years in both rural and urban models have generated statistically significant random factors that effectively capture the presence of heterogeneity in means, accounting for unobservable variations within the data. Clear evidence of factors such as speed limits, work zone type, and traffic volume affecting the work zone injury severities were found to vary significantly between rural and urban work zone areas. However, despite this difference, rural and urban work zones share common safety problems and countermeasures such as driver education, improved signage, and appropriate traffic controls; combining ITS technologies and enhanced law enforcement can help mitigate crash severity in urban and rural work zone areas.

Suicidal Ideation and Attempts Among Youth With Physical-Mental Comorbidity in Canada: Proposal for an Epidemiological Study

BACKGROUND

Evidence suggests that having a chronic physical illness (CPI; eg, asthma, diabetes, and epilepsy) is an independent risk factor for suicidality (ie, suicidal ideation or attempts) among youth. Less is known about the mechanisms linking CPI and suicidality. Some evidence suggests that mental illness (eg, depression and anxiety) or neurodevelopmental disorder (eg, attention-deficit/hyperactivity disorder) mediates or moderates the CPI-suicidality association. Missing from the knowledge base is information on the association between having co-occurring CPI and mental illness or neurodevelopmental disorder (MIND) on youth suicidality.

OBJECTIVE

This study uses epidemiological data from the 2019 Canadian Health Survey of Children and Youth (CHSCY) to study the intersection of CPI, MIND, and suicidality in youth. We will estimate prevalence, identify predictors, and investigate psychosocial and service use outcomes for youth with CPI-MIND comorbidity versus other morbidity groups (ie, healthy, CPI only, and MIND only).

METHODS

Conducted by Statistics Canada, the CHSCY collected data from 47,850 children (aged 1-17 years) and their primary caregiving parent. Measures of youth CPI, MIND, family environment, and sociodemographics are available using youth and parent informants. Information on psychiatric services use is available via parent report and linkage to national administrative health data found in the National Ambulatory Care Reporting System and the Discharge Abstract Database, which allow the investigation of hospital-based mental health services (eg, emergency department visits, hospitalizations, and length of stay in hospital). Questions about suicidality were restricted to youths aged 15-17 years (n=6950), which form our analytic sample. Weighted regression-based analyses will account for the complex survey design.

RESULTS

Our study began in November 2023, funded by the American Foundation for Suicide Prevention (SRG-0-008-22). Access to the linked CHSCY microdata file was granted in May 2024. Initial examination of CHSCY data shows that approximately 20% (1390/6950) of youth have CPI, 7% (490/6950) have MIND, 7% (490/6950) seriously considered suicide in the past year, and 3% (210/6950) had attempted suicide anytime during their life.

CONCLUSIONS

Findings will provide estimates of suicidality among youth with CPI-MIND comorbidity, which will inform intervention planning to prevent loss of life in this vulnerable population. Modeling correlates of suicidality will advance understanding of the relative and joint effects of factors at multiple levels-information needed to target prevention efforts and services. Understanding patterns of psychiatric service use is vital to understanding access and barriers to services. This will inform whether use matches need, identifying opportunities to advise policy makers about upstream resources to prevent suicidality. Importantly, findings will provide robust baseline of information on the link between CPI-MIND comorbidity and suicidality in youth, which can be used by future studies to address questions related to the impact of the COVID-19 pandemic and associated countermeasures in this vulnerable population of youth.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

DERR1-10.2196/57103.

Bayesian Multivariate Logistic Regression for Superiority and Inferiority Decision-Making under Observable Treatment Heterogeneity

The effects of treatments may differ between persons with different characteristics. Addressing such treatment heterogeneity is crucial to investigate whether patients with specific characteristics are likely to benefit from a new treatment. The current paper presents a novel Bayesian method for superiority decision-making in the context of randomized controlled trials with multivariate binary responses and heterogeneous treatment effects. The framework is based on three elements: a) Bayesian multivariate logistic regression analysis with a Pólya-Gamma expansion; b) a transformation procedure to transfer obtained regression coefficients to a more intuitive multivariate probability scale (i.e., success probabilities and the differences between them); and c) a compatible decision procedure for treatment comparison with prespecified decision error rates. Procedures for a priori sample size estimation under a non-informative prior distribution are included. A numerical evaluation demonstrated that decisions based on a priori sample size estimation resulted in anticipated error rates among the trial population as well as subpopulations. Further, average and conditional treatment effect parameters could be estimated unbiasedly when the sample was large enough. Illustration with the International Stroke Trial dataset revealed a trend toward heterogeneous effects among stroke patients: Something that would have remained undetected when analyses were limited to average treatment effects.

The trajectory of smoking cessation after treatment and its related factors in Taiwan

Smoking has multiple negative effects on health; therefore, the Taiwanese government provides smoking cessation clinics to smokers. This study aimed to explore the trajectory of smoking cessation after smokers received treatment and the variables related to different trajectories. A retrospective longitudinal study was conducted, in which 735 adult smokers who received smoking cessation medications were recruited. The participants' demographic characteristics, chronic diseases, smoking characteristics, and cigarette dependence were collected from chart review. The amount of smoking was collected at baseline, and at 1 week, 1 month, 3 months, and 6 months after treatment. The Proc Traj procedure for group-based modeling and multinomial logistic regression were used for statistical analysis. Three trajectories were identified: early quitters (28.03%), late quitters (11.43%) and reducers (60.54%). Compared with early quitters, reducers were younger and had a higher probability of severe cigarette dependence. Compared with early quitters, late quitters had a higher number of taking smoking cessation medications. The findings revealed that approximately 60% of participants who received smoking cessation treatment could not completely quit smoking, and that age, number of medications taken, and cigarette dependence were significant predictors of different trajectories.

Identification of Distinct Risk Factors for Antepartum and Postpartum Preeclampsia in a High-Risk Safety-Net Hospital

OBJECTIVE

Postpartum preeclampsia (PE), defined as de novo PE that develops at least 48 hours following delivery, can be particularly dangerous as many patients are already discharged at that point. The goal of our study was to identify risk factors uniquely associated with the development of late postpartum preeclampsia (PPPE).

STUDY DESIGN

In a retrospective cohort study of deliveries between July 1, 2016 and June 30, 2018 at a safety-net hospital in Atlanta, Georgia, we used multinomial logistic regression models to estimate adjusted odds ratios (aORs) and 95% confidence intervals (CIs) for associations between demographic, medical, and obstetric factors and development of PE, categorized as a three-level outcome: no PE, antepartum/intrapartum preeclampsia (APE) (diagnosed prior to or < 48 hours of delivery), and late PPPE (diagnosed ≥ 48-hour postpartum).

RESULTS

Among 3,681 deliveries, women were primarily of ages 20 to 35 years (76.4%), identified as non-Hispanic Black (68.5%), and covered by public health insurance (88.6%). PE was diagnosed prior to delivery or within 48-hour postpartum in 12% (n = 477) of the study population, and 1.5% (57) developed PE greater than 48-hour postpartum. In the adjusted models, maternal age ≥ 35, race/ethnicity, nulliparity, a diagnosis of pregestational or gestational diabetes, and chronic hypertension were associated with increased odds of APE only, while maternal obesity (OR: 1.9; 95% CI: 1.0-3.5) and gestational hypertension (OR: 2.7; 95% CI: 1.5-4.8) were uniquely associated with PPPE. Multifetal gestations and cesarean delivery predicted both PPPE and APE; however, the association was stronger for PPPE.

CONCLUSION

Patients with obesity, gestational hypertension, multifetal gestations, or cesarean delivery may benefit from additional follow-up in the early postpartum period to detect PPPE.

KEY POINTS

· Late postpartum preeclampsia may go undetected, particularly in low-income patients.. · In a delivery cohort in Georgia, 1.5% of patients developed late postpartum preeclampsia.. · Maternal obesity and gestational hypertension were strongly associated only with late postpartum preeclampsia..

Tutorial: dos and don'ts in clinical prediction research for venous thromboembolism

Clinical prediction modeling has become an increasingly popular domain of venous thromboembolism research in recent years. Prediction models can help healthcare providers make decisions regarding starting or withholding therapeutic interventions, or referrals for further diagnostic workup, and can form a basis for risk stratification in clinical trials. The aim of the current guide is to assist in the practical application of complicated methodological requirements for well-performed prediction research by presenting key dos and don'ts while expanding the understanding of predictive research in general for (clinical) researchers who are not specifically trained in the topic; throughout we will use prognostic venous thromboembolism scores as an exemplar.

Examination of protective factors that promote prosocial skill development among children exposed to intimate partner violence

This retrospective cohort study examined prosocial skills development in child welfare-involved children, how intimate partner violence (IPV) exposure explained heterogeneity in children's trajectories of prosocial skill development, and the degree to which protective factors across children's ecologies promoted prosocial skill development. Data were from 1,678 children from the National Survey of Child and Adolescent Well-being I, collected between 1999 and 2007. Cohort-sequential growth mixture models were estimated to identify patterns of prosocial skill development between the ages of 3 to 10 years. Four diverse pathways were identified, including two groups that started high (high subtle-decreasing; high decreasing-to-increasing) and two groups that started low (low stable; low increasing-to-decreasing). Children with prior history of child welfare involvement, preschool-age IPV exposure, school-age IPV exposure, or family income below the federal poverty level had higher odds of being in the high decreasing-to-increasing group compared with the high subtle-decreasing group. Children with a mother with greater than high school education or higher maternal responsiveness had higher odds of being in the low increasing-to-decreasing group compared with the low stable group. The importance of maternal responsiveness in fostering prosocial skill development underlines the need for further assessment and intervention. Recommendations for clinical assessment and parenting programs are provided.

Digital biomarkers of mobility and quality of life in Parkinson's disease based on a wearable motion analysis LAB

BACKGROUND AND OBJECTIVE

Functional mobility, an indicator of the quality of life (QoL), requires fast and flexible changes during motion, which are limited in Parkinson's disease (PD). Recent body-worn sensors have emerged in the last decades as potential solutions to produce digital biomarkers able to quantify mobility outside routine consultations and during real-life scenarios for multiple days at a time. The proposed research aims to study the ability of a wearable motion analysis lab, developed by our team, to produce digital biomarkers of mobility and QoL levels in patients with PD.

METHODS

A cross-sectional study was followed, including 40 patients stratified into three subgroups according to a clinic motor examination and a QoL questionnaire.

RESULTS

The achieved outcomes demonstrate the ability of the proposed high-tech solution to measure prototypical gait impairments and discriminate motor condition (AUC=0,890) and patients' QoL levels (AUC=0,950). Also, from the measured multiple gait-associated parameters, we identified the variables with the most potential to be applied as digital biomarkers of mobility (67 % of the metrics) and QoL (72 % of the metrics) in PD.

CONCLUSIONS

Overall, we confirmed our hypothesis of using our body-worn sensor-based solution for passive or active monitoring of mobility and QoL in PD to produce objective, feasible, and continuous digital biomarkers.

Postoperative and postdischarge nausea and vomiting following ambulatory eye, head, and neck surgeries: a retrospective cohort study comparing incidence and associated factors

BACKGROUND

Ambulatory surgery is often followed by the development of nausea and/or vomiting (N/V). Although risk factors for postoperative nausea and vomiting (PONV) are frequently discussed, the distinction between PONV and postdischarge nausea and vomiting (PDNV) is unclear. This is especially troublesome given the potential consequences of postdischarge nausea and vomiting (PDNV), which include major discomfort and hospital readmission.

METHODS

In this retrospective cohort study, data from 10,231 adult patients undergoing ambulatory ophthalmology or otolaryngology procedures with general anesthesia were collected and analyzed. Binary and multinomial logistic regression was used to assess the association between patient and anesthetic characteristics (including age, body mass index (BMI), American Society of Anesthesiologists Physical Status (ASA P/S) classification, current smoker status, and intra- and postoperative opioid usage) and the odds ratios of experiencing only PDNV, only PONV, or both PONV and PDNV, as compared to not experiencing N/V at all.

RESULTS

We found that 17.8% of all patients developed N/V (PONV and/or PDNV). Patients who experienced PONV had a 2.79 (95% confidence interval 2.24-3.46) times greater risk of reporting PDNV. Binary logistic regression found that younger age, opioid use, and female sex were associated with an increased likelihood of experiencing any N/V. Increased use of nitrous oxide and a higher ASA P/S class was associated with elevated likelihood of PONV, but not PDNV or PONV plus PDNV.

CONCLUSIONS

Patients experiencing N/V in the PACU are observed to develop PDNV disproportionately by a factor of 2.79. The patients have distinct predictors, indicating important opportunities for care improvements beyond current guidelines.

Latent Class Groups of Concurrent Substance Use Among Adolescents in an Urban Community: Correlates With Mental Health, Access to Drugs and Alcohol, and Risk Perception

BACKGROUND

Concurrent substance use among adolescents has been associated with an increase in physical and mental health problems. These outcomes tend to be exacerbated among adolescents of color in underserved urban settings. The purpose of this study was to understand alcohol and concurrent drug use patterns among adolescents in an underserved urban community to provide targeted prevention and treatment recommendations.

METHOD

This study examined data among adolescents in an underserved urban community (N = 1789; 56.90% female; 70.86% Hispanic/Latino/a; meanage = 15.96 ± 1.56). Using latent class analysis (LCA) and multinomial logistic regression modeling, analyses identified independent correlates of latent class membership.

RESULTS

Five latent classes (LC) were identified including LC group 1: Predominant alcohol use and limited to no concurrent-drug use (n = 213; 11.9%); LC group 2: Concurrent drug and alcohol use including methamphetamine, marijuana and synthetic marijuana use, and alcohol use (n = 74; 4.2%); LC group 3: Concurrent drug and alcohol use, with no marijuana use (n = 204; 11.39%); LC group 4: High Concurrent drug use and alcohol use (n = 204; 11.40%); and LC group 5: Concurrent drug use without alcohol use (n = 1101; 61.52%). Significant between group differences were noted between latent class groups and sociodemographic characteristics. Multinomial logistic regression models identified the associations between sociodemographic characteristics and corollary clinical features of substance use on latent class groupings of alcohol and concurrent drug use.

CONCLUSION

Understanding concurrent substance use LC groups among adolescents is essential to providing targeted interventions and treatment programs, as well as early intervention programs that may help reduce substance use during adolescence.

Bayesian multilevel multivariate logistic regression for superiority decision-making under observable treatment heterogeneity

BACKGROUND

In medical, social, and behavioral research we often encounter datasets with a multilevel structure and multiple correlated dependent variables. These data are frequently collected from a study population that distinguishes several subpopulations with different (i.e., heterogeneous) effects of an intervention. Despite the frequent occurrence of such data, methods to analyze them are less common and researchers often resort to either ignoring the multilevel and/or heterogeneous structure, analyzing only a single dependent variable, or a combination of these. These analysis strategies are suboptimal: Ignoring multilevel structures inflates Type I error rates, while neglecting the multivariate or heterogeneous structure masks detailed insights.

METHODS

To analyze such data comprehensively, the current paper presents a novel Bayesian multilevel multivariate logistic regression model. The clustered structure of multilevel data is taken into account, such that posterior inferences can be made with accurate error rates. Further, the model shares information between different subpopulations in the estimation of average and conditional average multivariate treatment effects. To facilitate interpretation, multivariate logistic regression parameters are transformed to posterior success probabilities and differences between them.

RESULTS

A numerical evaluation compared our framework to less comprehensive alternatives and highlighted the need to model the multilevel structure: Treatment comparisons based on the multilevel model had targeted Type I error rates, while single-level alternatives resulted in inflated Type I errors. Further, the multilevel model was more powerful than a single-level model when the number of clusters was higher. A re-analysis of the Third International Stroke Trial data illustrated how incorporating a multilevel structure, assessing treatment heterogeneity, and combining dependent variables contributed to an in-depth understanding of treatment effects. Further, we demonstrated how Bayes factors can aid in the selection of a suitable model.

CONCLUSION

The method is useful in prediction of treatment effects and decision-making within subpopulations from multiple clusters, while taking advantage of the size of the entire study sample and while properly incorporating the uncertainty in a principled probabilistic manner using the full posterior distribution.

Propensity-based standardization to enhance the validation and interpretation of prediction model discrimination for a target population

External validation of the discriminative ability of prediction models is of key importance. However, the interpretation of such evaluations is challenging, as the ability to discriminate depends on both the sample characteristics (ie, case-mix) and the generalizability of predictor coefficients, but most discrimination indices do not provide any insight into their respective contributions. To disentangle differences in discriminative ability across external validation samples due to a lack of model generalizability from differences in sample characteristics, we propose propensity-weighted measures of discrimination. These weighted metrics, which are derived from propensity scores for sample membership, are standardized for case-mix differences between the model development and validation samples, allowing for a fair comparison of discriminative ability in terms of model characteristics in a target population of interest. We illustrate our methods with the validation of eight prediction models for deep vein thrombosis in 12 external validation data sets and assess our methods in a simulation study. In the illustrative example, propensity score standardization reduced between-study heterogeneity of discrimination, indicating that between-study variability was partially attributable to case-mix. The simulation study showed that only flexible propensity-score methods (allowing for non-linear effects) produced unbiased estimates of model discrimination in the target population, and only when the positivity assumption was met. Propensity score-based standardization may facilitate the interpretation of (heterogeneity in) discriminative ability of a prediction model as observed across multiple studies, and may guide model updating strategies for a particular target population. Careful propensity score modeling with attention for non-linear relations is recommended.

How to explore within-person and between-person measurement model differences in intensive longitudinal data with the R package lmfa

Intensive longitudinal data (ILD) have become popular for studying within-person dynamics in psychological constructs (or between-person differences therein). Before investigating the dynamics, it is crucial to examine whether the measurement model (MM) is the same across subjects and time and, thus, whether the measured constructs have the same meaning. If the MM differs (e.g., because of changes in item interpretation or response styles), observations cannot be validly compared. Exploring differences in the MM for ILD can be done with latent Markov factor analysis (LMFA), which classifies observations based on the underlying MM (for many subjects and time points simultaneously) and thus shows which observations are comparable. However, the complexity of the method or the fact that no open-source software for LMFA existed until now may have hindered researchers from applying the method in practice. In this article, we provide a step-by-step tutorial for the new user-friendly software package lmfa, which allows researchers to easily perform the analysis LMFA in the freely available software R to investigate MM differences in their own ILD.

Development and internal validation of a non-invasive clinical tool to predict sufficient omega-3 levels in early pregnancy

BACKGROUND

Complications from preterm birth (PTB) are the leading cause of death and disability in those under five years. Whilst the role of omega-3 (n-3) supplementation in reducing PTB is well-established, growing evidence suggests supplementation use in those replete may increase the risk of early PTB.

AIM

To develop a non-invasive tool to identify individuals with total n-3 serum levels above 4.3% of total fatty acids in early pregnancy.

METHODS

We conducted a prospective observational study recruiting 331 participants from three clinical sites in Newcastle, Australia. Eligible participants (n = 307) had a singleton pregnancy between 8 and 20 weeks' gestation at recruitment. Data on factors associated with n-3 serum levels were collected using an electronic questionnaire; these included estimated intake of n-3 (including food type, portion size, frequency of consumption), n-3 supplementation, and sociodemographic factors. The optimal cut-point of estimated n-3 intake that predicted mothers with total serum n-3 levels likely above 4.3% was developed using multivariate logistic regression, adjusting for maternal age, body mass index, socioeconomic status, and n-3 supplementation use. Total serum n-3 levels above 4.3% was selected as previous research has demonstrated that mothers with these levels are at increased risk of early PTB if they take additional n-3 supplementation during pregnancy. Models were evaluated using various performance metrics including sensitivity, specificity, area under receiver operator characteristic (AUROC) curve, true positive rate (TPR) at 10% false positive rate (FPR), Youden Index, Closest to (0,1) Criteria, Concordance Probability, and Index of Union. Internal validation was performed using 1000-bootstraps to generate 95% confidence intervals for performance metrics generated.

RESULTS

Of 307 eligible participants included for analysis, 58.6% had total n-3 serum levels above 4.3%. The optimal model had a moderate discriminative ability (AUROC 0.744, 95% CI 0.742-0.746) with 84.7% sensitivity, 54.7% specificity and 37.6% TPR at 10% FPR.

CONCLUSIONS

Our non-invasive tool was a moderate predictor of pregnant women with total serum n-3 levels above 4.3%; however, its performance is not yet adequate for clinical use.

TRIAL REGISTRATION

This trial was approved by the Hunter New England Human Research Ethics Committee of the Hunter New England Local Health District (Reference 2020/ETH00498 on 07/05/2020 and 2020/ETH02881 on 08/12/2020).

Development of a Machine Learning Model to Identify Colorectal Cancer Stage in Medicare Claims

PURPOSE

Staging information is essential for colorectal cancer research. Medicare claims are an important source of population-level data but currently lack oncologic stage. We aimed to develop a claims-based model to identify stage at diagnosis in patients with colorectal cancer.

METHODS

We included patients age 66 years or older with colorectal cancer in the SEER-Medicare registry. Using patients diagnosed from 2014 to 2016, we developed models (multinomial logistic regression, elastic net regression, and random forest) to classify patients into stage I-II, III, or IV on the basis of demographics, diagnoses, and treatment utilization identified in Medicare claims. Models developed in a training cohort (2014-2016) were applied to a testing cohort (2017), and performance was evaluated using cancer stage listed in the SEER registry as the reference standard.

RESULTS

The cohort of patients with 30,543 colorectal cancer included 14,935 (48.9%) patients with stage I-II, 9,203 (30.1%) with stage III, and 6,405 (21%) with stage IV disease. A claims-based model using elastic net regression had a scaled Brier score (SBS) of 0.45 (95% CI, 0.43 to 0.46). Performance was strongest for classifying stage IV (SBS, 0.62; 95% CI, 0.59 to 0.64; sensitivity, 93%; 95% CI, 91 to 94) followed by stage I-II (SBS, 0.45; 95% CI, 0.44 to 0.47; sensitivity, 86%; 95% CI, 85 to 76) and stage III (SBS, 0.32; 95% CI, 0.30 to 0.33; sensitivity, 62%; 95% CI, 61 to 64).

CONCLUSION

Machine learning models effectively classified colorectal cancer stage using Medicare claims. These models extend the ability of claims-based research to risk-adjust and stratify by stage.

Minimum sample size for developing a multivariable prediction model using multinomial logistic regression

AIMS

Multinomial logistic regression models allow one to predict the risk of a categorical outcome with > 2 categories. When developing such a model, researchers should ensure the number of participants (n ) is appropriate relative to the number of events (E k ) and the number of predictor parameters (p k ) for each category k. We propose three criteria to determine the minimum n required in light of existing criteria developed for binary outcomes.

PROPOSED CRITERIA

The first criterion aims to minimise the model overfitting. The second aims to minimise the difference between the observed and adjusted R 2 Nagelkerke. The third criterion aims to ensure the overall risk is estimated precisely. For criterion (i), we show the sample size must be based on the anticipated Cox-snell R 2 of distinct 'one-to-one' logistic regression models corresponding to the sub-models of the multinomial logistic regression, rather than on the overall Cox-snell R 2 of the multinomial logistic regression.

EVALUATION OF CRITERIA

We tested the performance of the proposed criteria (i) through a simulation study and found that it resulted in the desired level of overfitting. Criterion (ii) and (iii) were natural extensions from previously proposed criteria for binary outcomes and did not require evaluation through simulation.

SUMMARY

We illustrated how to implement the sample size criteria through a worked example considering the development of a multinomial risk prediction model for tumour type when presented with an ovarian mass. Code is provided for the simulation and worked example. We will embed our proposed criteria within the pmsampsize R library and Stata modules.

Immigration Experience and Cognitive Function Trajectories Among Older Chinese Immigrants

OBJECTIVES

Although a number of studies have documented cognitive health among older immigrants in the United States, little is known about how the life-course immigration experiences are associated with cognitive trajectories among older Chinese immigrants. We assess patterns of cognitive functioning and change over time and examine whether age at migration, reasons for migration, acculturation, perceived discrimination, and preferred dialects are related to cognitive trajectories.

METHODS

The sample comprised 2,075 participants from the Population Study of Chinese Elderly (PINE), who completed a battery of cognitive tests at four time points (2011-2019). Latent class growth analysis and multinomial logistic regression were utilized.

RESULTS

Three latent classes of cognitive trajectories were identified: the low functioning with the fastest decline (LCF, 12%), the moderate functioning with a medium decline rate (MCF, 39%), and the high functioning with the slowest decline (HCF, 48%). Perceiving more discrimination reduced, whereas speaking Taishanese increased the odds of being in the LCF and MCF. High acculturation only distinguished MCF from HCF after controlling for the known factors of cognitive health such as age, education, and social engagement.

DISCUSSION

This study identifies a group of older Chinese immigrants who are especially vulnerable to cognitive impairment and indicates that the risk of cognitive decline appears to be elevated with lower levels of acculturation and unidentified racial discrimination. More research is needed to fully understand the underlying mechanisms that link the life-course immigration experiences to cognitive health outcomes in later life.

Changing trends in ethnicity and academic performance: observational cohort data from a UK medical school

OBJECTIVE

Quantify differential attainment by ethnicity in undergraduate medical assessments and evaluate whether institutional efforts to reduce the attainment gap have had impact.

DESIGN

Observational cohort study.

SETTING

A single UK MBBS medical programme.

PARTICIPANTS

Pseudonymised data of adults aged ≥18 years enrolled in one of the UK MBBS medical programmes between 2012 and 2018. Ethnicity was self-declared during enrolment as White, Asian, Black, mixed and other.

MAIN OUTCOME MEASURE

Module mark (distinction, merit, pass, fail) graded according to a variety of assessments, including single best answer examinations, objective structured clinical examinations and coursework submissions. All modular assessments are graded as a percentage. Logistic regression models were used to calculate relative risk ratio to study the association between ethnicity and attainment gap over a calendar and scholastic year. Models were adjusted for age, gender, social deprivation and scholastic year of study.

RESULTS

3714 student records were included. In the sample, 2134 students (57%) were non-white. The proportion of non-white students increased from 2007 (49%) to 2018 (70%). Mean age was 18 (IQR 18-21) and 56.6% were females. Higher proportion of non-white students 218 (24.8%) were from more deprived backgrounds versus white 76 (14.8%). Compared with non-white, there were no significant differences in the proportion of students failing assessments. However, white students were more likely to achieve merit (relative risk ratio 1.29 (95% CI 1.08 to 1.45)) or distinction (1.69 (95% CI 1.37 to 2.08)). Differences in attainment gap have remained unchanged over time, and for black students, attainment gap grew between their first and final year of study.

CONCLUSION

A similar proportion (97%) of non-white and white students had a passing score, but attainment gap for higher grades persists over years despite widespread efforts in medical schools to diminish the attainment gap linked to ethnicity. Our findings are from a single institution, thus affecting generalisability.

Effects of fun-seeking and external locus of control on smoking behaviour: a cross-sectional analysis on a cohort of working men in Singapore

OBJECTIVES

We examined the combined effects of behavioural inhibition and behavioural activation, on one hand, and locus of control, on the other hand, on different categories of smoking behaviour (non-smoking, ex-smoking, occasional smoking, daily smoking).

DESIGN

This study adopted a cross-sectional design. Participants completed questionnaires regarding demographics, smoking patterns, behavioural inhibition/behavioural activation systems and locus of control.

SETTING

The study was conducted across four companies from the transportation, cooling plant and education sectors in Singapore.

PARTICIPANTS

Three hundred sixty-nine male working adults were included in the final sample.

RESULTS

Corroborating previous research, a logistic regression model examining behavioural inhibition/behavioural activation systems revealed that the fun-seeking aspect of behavioural activation was a unique predictor in distinguishing non-smokers from daily smokers (OR=1.24, p=0.012). By contrast, in a separate model examining locus of control, external locus of control was found to be a unique predictor in distinguishing non-smokers from daily smokers (OR=1.13, p<0.001). In addition, a third model combining both behavioural inhibition/behavioural activation systems and locus of control found that only external locus of control remained a significant predictor (OR=1.12, p<0.001). Further analyses revealed a mediating effect of external locus of control on the relationship between fun-seeking and smoking behaviour. That is, the increase in the odds of daily smoking due to fun-seeking was explained by external locus of control (direct pathway OR=1.20, p=0.058; indirect pathway OR=1.04, p<0.050).

CONCLUSIONS

Overall, fun-seeking through its influence on external locus of control indirectly affects daily smoking behaviour, suggesting a more complex relationship than shown in previous research.

Mental health emergencies attended by ambulances in the United Kingdom and the implications for health service delivery: A cross-sectional study

OBJECTIVE

In the context of increasing demand for ambulance services, emergency mental health cases are among the most difficult for ambulance clinicians to attend, partly because the cases often involve referring patients to other services. We describe the characteristics of mental health emergencies in the East Midlands region of the United Kingdom. We explore the association between 999 (i.e. emergency) call records, the clinical impressions of ambulance clinicians attending emergencies and the outcomes of ambulance attendance. We consider the implications of our results for optimizing patient care and ambulance service delivery.

METHODS

We conducted a retrospective observational study of records of all patients experiencing mental health emergencies attended by ambulances between 1 January 2018 and 31 July 2020. The records comprised details of 103,801 '999' calls (Dispatch), the preliminary diagnoses by ambulance clinicians on-scene (Primary Clinical Impression) and the outcomes of ambulance attendance for patients (Outcome).

RESULTS

A multinomial regression analysis found that model fit with Outcome data was improved with the addition of Dispatch and Primary Clinical Impression categories compared to the fit for the model containing only the intercept and Outcome categories (Chi-square = 18,357.56, df = 180, p < 0.01). Dispatch was a poor predictor of Primary Clinical impression. The most common predictors of Outcome care pathways other than 'Treated and transported' were records of respiratory conditions at Dispatch and anxiety reported by clinicians on-scene.

CONCLUSIONS

Drawing on the expertise of mental health specialists may help '999' dispatchers distinguish between physical and mental health emergencies and refer patients to appropriate services earlier in the response cycle. Further investigation is needed to determine if training Dispatch operatives for early triage and referral can be appropriately managed without compromising patient safety.

Musculoskeletal pain trajectories of employees working from home during the COVID-19 pandemic

Objectives

In March 2020, the COVID-19 pandemic necessitated a rapid public health response which included mandatory working from home (WFH) for many employees. This study aimed to identify different trajectories of multisite musculoskeletal pain (MSP) amongst employees WFH during the COVID-19 pandemic and examined the influence of work and non-work factors.

Methods

Data from 488 participants (113 males, 372 females and 3 other) involved in the Employees Working from Home (EWFH) study, collected in October 2020, April and November 2021 were analysed. Age was categorised as 18–35 years (n = 121), 36–55 years (n = 289) and 56 years and over (n = 78). Growth Mixture Modelling (GMM) was used to identify latent classes with different growth trajectories of MSP. Age, gender, working hours, domestic living arrangements, workstation comfort and location, and psychosocial working conditions were considered predictors of MSP. Multivariate multinomial logistic regression was used to identify work and non-work variables associated with group membership.

Results

Four trajectories of MSP emerged: high stable (36.5%), mid-decrease (29.7%), low stable (22.3%) and rapid increase (11.5%). Decreased workstation comfort (OR 1.98, CI 1.02, 3.85), quantitative demands (OR 1.68, CI 1.09, 2.58), and influence over work (OR 0.78, CI 0.54, 0.98) was associated with being in the high stable trajectory group compared to low stable. Workstation location (OR 3.86, CI 1.19, 12.52) and quantitative work demands (OR 1.44, CI 1.01, 2.47) was associated with the rapid increase group.

Conclusions

Findings from this study offer insights into considerations for reducing MSP in employees WFH. Key considerations include the need for a dedicated workstation, attention to workstation comfort, quantitative work demands, and ensuring employees have influence over their work.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00420-022-01885-1.

Differences and Similarities in Breast and Colorectal Cancer Screening Uptake among Municipalities in Flanders, Belgium

Despite the recognized benefits of fecal occult blood test (FOBT) and mammography screenings, participation in breast (BC) and colorectal cancer (CRC) screening programs is still suboptimal. This study investigates municipal characteristics associated with their BC/CRC screening uptake profiles among women aged 55–69 years. Using data from 308 municipalities of Flanders from 2014 to 2017, a profile for each municipality based on its BC/CRC screening uptake compared with the median screening uptake was created. Logistic regression with generalized estimating equations was used to assess the associations between municipal characteristics and BC/CRC screening uptake profiles. The overall median uptake of cancer screening was higher for CRC (57.4%) than for BC (54.6%). The following municipal characteristics were associated with worse performance in terms of only CRC, only BC, or both CRC and BC screening uptake, respectively: foreign nationality, self-employment rate, (early) retirement rate, diabetes, disabilities; (early) retirement rate; age group 65–69, foreign nationality, self-employment rate, (early) retirement rate, wage-earners, diabetes. The following municipal characteristics were associated with better performance in terms of only CRC, only BC, or both CRC and BC screening uptake respectively: residential stability, having a partner, having children, jobseeker rate, GP visits, preventive dental visits; having children, GP visits; age group 55–59, residential stability, having a partner, having children, jobseeker rate, higher education, GP visits, preventive dental visits. This study’s results regarding the interrelation between the BC and CRC screening could be used to tailor interventions to improve the participation of the target population in both programs.

Tenets of Good Practice in Regression Analysis. A Brief Tutorial

BACKGROUND

Regression analysis quantifies the relationships between one or more independent variables and a dependent variable and is one of the most frequently used types of analysis in medical research. The aim of this article is to provide a brief theoretical and practical tutorial for neurosurgeons wishing to conduct or interpret regression analyses.

METHODS AND RESULTS

Data preparation, univariable and multivariable analysis, choice of model, model requirements and assumptions are discussed, as essential prerequisites to any regression analysis. Four main types of regression techniques are presented: linear, logistic, multinomial logistic, and proportional odds logistic. To illustrate the applications of regression to real-world data and exemplify the concepts introduced, we used a previously reported data set of patients with intracranial aneurysms treated by microsurgical clip reconstruction at the Department of Neurosurgery of Erasmus MC University Medical Center Rotterdam, between January 2000 and January 2019.

CONCLUSIONS

Regression analysis is a powerful and versatile instrument in data analysis. This material is intended as a starter for those wishing to critically interpret or perform regression analysis and we recommend multidisciplinary collaborations with trained methodologists, statisticians, or epidemiologists.

Demographic discrepancies between administrative-prioritisation and self-prioritisation of multiple ethnic identifications

Ethnic classification is an inherently subjective process, especially when multiple ethnic identifications are involved. There are two methods commonly used to classify multiple ethnicities into single categories: administrative-prioritisation (assignment via a predetermined hierarchy) and self-prioritisation (where individuals select their "main" ethnicity). Currently, little is known about whether the demographic composition of outputted ethnic groups differs by prioritisation method. This study utilised large-scale data of multi-ethnic children (N = 1,860), adolescents (N = 2,413), and adults (N = 1,056) from Aotearoa New Zealand to examine individual and contextual demographic characteristics associated with discrepancies between administratively-prioritised and self-prioritised ethnicity. Results showed that discrepancy rates, which exceeded 50%, were systematically associated with neighbourhood ethnic composition and socioeconomic deprivation, but largely not associated with gender, age, and birthplace. The contextual nature of self-prioritisation highlights the importance of researchers' choice of ethnic classification method. Implications are discussed in the context of increasing multi-ethnic prevalence.

Patient's Perspective on the Impact of COVID-19 on Cancer Treatment in Nigeria

Because of the global COVID-19 pandemic, health care organizations introduced guidelines for modifications to health and cancer medical care delivery to mitigate transmission and ensure quality health outcomes. To examine the extent and impact of these modifications on oncology service disruptions in Nigeria, we surveyed oncology patients across selected public and private cancer treatment centers.

Service interruptions because of the COVID-19 pandemic—Nigerian cancer patients' experience.

A Latent Class Analysis of Mental Health Severity and Alcohol Consumption: Associations with COVID-19-Related Quarantining, Isolation, Suicidal Ideations, and Physical Activity

The present study examined latent class cluster group patterns based on measures of depression and anxiety symptom severity and alcohol consumption during the COVID-19 pandemic. Hypothesized correlates with latent class cluster groups including quarantining, self-isolation, suicidal ideations, sitting hours per day, and physical activity (vigorous intensity exercise in minutes per week) were examined. The delimited participant sample consisted of 606 university young adults 18 to 25 years of age (M = 21.24 ± 1.62). Latent cluster analysis (LCA) modeled patterns of depression and anxiety symptom severity and alcohol consumption during the COVID-19 pandemic. Between group analysis and multinomial logistic regression analysis were used to examine relationships between latent class clusters and correlates including quarantining, self-isolation, suicidal ideations, sitting hours per day, and physical activity (vigorous intensity exercise in minutes per week). LCA results showed that six latent cluster groups provided optimal model-to-date fit based on mental health symptom severity and alcohol consumption (L 2  = 56.31, BIC = 5012.79, AIC = 4849.74, and the bootstrap L 2 p-value = .88; Entropy R 2  = .89). Identified latent class clusters were as follows: cluster one = moderate anxiety and depression severity and moderate alcohol consumption (n = 156; 25.7%); cluster two = high mental health severity and alcohol consumption (n = 133; 21.9%); cluster three = low mental health symptoms and moderate alcohol consumption (n = 105; 17.3%); cluster four = lowest mental health severity and alcohol consumption (n = 95; 15.7%); cluster five = moderate depression severity, low anxiety severity, and low alcohol consumptions (n = 74; 12.2%); and cluster six = moderate anxiety severity, low depression severity, and low alcohol consumption (n = 43; 7.1%). Multinomial logistic regression analysis results found that quarantining, self-isolation, suicidal ideations, sedentary behavior, and physical activity were differentially associated with cluster group membership. Findings from this study demonstrate associations between COVID-19 public health restrictions, suicidal ideations, and declines in mental health and increases in alcohol consumption among young adult university students.

Incorporating respondent-driven sampling into web-based discrete choice experiments: preferences for COVID-19 mitigation measures

To slow the spread of COVID-19, most countries implemented stay-at-home orders, social distancing, and other nonpharmaceutical mitigation strategies. To understand individual preferences for mitigation strategies, we piloted a web-based Respondent Driven Sampling (RDS) approach to recruit participants from four universities in three countries to complete a computer-based Discrete Choice Experiment (DCE). Use of these methods, in combination, can serve to increase the external validity of a study by enabling recruitment of populations underrepresented in sampling frames, thus allowing preference results to be more generalizable to targeted subpopulations. A total of 99 students or staff members were invited to complete the survey, of which 72% started the survey (n = 71). Sixty-three participants (89% of starters) completed all tasks in the DCE. A rank-ordered mixed logit model was used to estimate preferences for COVID-19 nonpharmaceutical mitigation strategies. The model estimates indicated that participants preferred mitigation strategies that resulted in lower COVID-19 risk (i.e. sheltering-in-place more days a week), financial compensation from the government, fewer health (mental and physical) problems, and fewer financial problems. The high response rate and survey engagement provide proof of concept that RDS and DCE can be implemented as web-based applications, with the potential for scale up to produce nationally-representative preference estimates.

Association between rural workers’ sociodemographic and reproductive characteristics and their reproductive autonomy

ABSTRACT Objectives: to verify the association between sociodemographic and reproductive characteristics with rural workers’ reproductive autonomy. Methods: a cross-sectional study, with a sample of 346 women and application of the Reproductive Autonomy Scale. Multinomial regression was performed to analyze associations between independent variables and outcomes. Results: in the analysis of subscales “Decision-making”, “My sexual partner or someone else such as a parent”, “Both me and my partner” and “Me”, women experienced greater reproductive autonomy in relation to their partners. For outcomes “Decision about which method to use”, “When to have a baby” or “About unplanned pregnancy”, the highest prevalence was for category “Me”, with statistically significant associations. Conclusions: the sociodemographic and reproductive characteristics among the most vulnerable women, in terms of the social, economic and cultural context in which they are inserted, may be associated with greater difficulties in exercising reproductive autonomy.

Associação entre as características sociodemográficas e reprodutivas com a autonomia reprodutiva das trabalhadoras rurais

RESUMO Objetivos: verificar a associação entre as características sociodemográficas e reprodutivas com a autonomia reprodutiva das trabalhadoras rurais. Métodos: estudo transversal, com amostra de 346 mulheres e aplicação da Escala de Autonomia Reprodutiva. Foi realizada regressão multinomial para análises de associações entre as variáveis independentes e desfechos. Resultados: na análise das subescalas “Tomada de decisão”, “Meu parceiro sexual ou alguém da família tem mais a dizer”, “Eu e meu parceiro sexual” e “Eu decido”, as mulheres experimentaram maior autonomia reprodutiva em relação aos parceiros. Para os desfechos “Decisão sobre qual método utilizar”, “Quando ter um bebê” ou “Sobre gravidez não planejada”, as maiores prevalências foram para a categoria “Eu decido”, com associações estatisticamente significante. Conclusões: as características sociodemográficas e reprodutivas entre mulheres mais vulneráveis, tratando-se do contexto social, econômico e cultural que estão inseridas, podem estar associadas a maiores dificuldades para exercerem a autonomia reprodutiva.

Why don't all nurse practitioners precept? A comparative study

BACKGROUND

Preceptors are integral in nurse practitioner (NP) education. A shortage of willing preceptors limits graduations from NP programs. However, little is known about why NPs decide not to precept.

PURPOSE

To identify the factors significantly associated with NPs' status as currently, previously, or never precepting, using the Integrated Behavioral Model as the conceptual framework.

METHODOLOGY

This was a cross-sectional, comparative, descriptive study of NPs using survey methodology. Our survey was based on published surveys with items added and adapted to reflect our framework. Subscales included personal factors, primary determinants of intent to precept, and external factors. We mailed recruitment postcards, with an online survey link, to all NPs in Arkansas. Data were analyzed using bivariate and stepwise multinomial logistic regression for each subscale.

RESULTS

We obtained 261 responses. Participants who had previously and/or never precepted differed from current preceptors on multiple variables on bivariate analysis. Predictive personal factors included experience and hours worked per week. Predictive primary determinants included knowing NPs that precept, support for precepting, recognition of preceptors, and clinical expertise. Predictive external factors included space, liability, having a "gatekeeper," NP program, importance of precepting, and number of requests.

CONCLUSIONS

Different factors predict NPs who currently, previously, and have never precepted. However, frequency of requests predicted both nonprecepting groups. Various supports in the clinical setting and program factors predicted one or other nonprecepting group.

IMPLICATIONS

Strategies should be developed to ensure all potential preceptors are recruited, increase support for precepting, and ensure educational programs meet preceptors' needs.

Risk prediction models for discrete ordinal outcomes: Calibration and the impact of the proportional odds assumption

Calibration is a vital aspect of the performance of risk prediction models, but research in the context of ordinal outcomes is scarce. This study compared calibration measures for risk models predicting a discrete ordinal outcome, and investigated the impact of the proportional odds assumption on calibration and overfitting. We studied the multinomial, cumulative, adjacent category, continuation ratio, and stereotype logit/logistic models. To assess calibration, we investigated calibration intercepts and slopes, calibration plots, and the estimated calibration index. Using large sample simulations, we studied the performance of models for risk estimation under various conditions, assuming that the true model has either a multinomial logistic form or a cumulative logit proportional odds form. Small sample simulations were used to compare the tendency for overfitting between models. As a case study, we developed models to diagnose the degree of coronary artery disease (five categories) in symptomatic patients. When the true model was multinomial logistic, proportional odds models often yielded poor risk estimates, with calibration slopes deviating considerably from unity even on large model development datasets. The stereotype logistic model improved the calibration slope, but still provided biased risk estimates for individual patients. When the true model had a cumulative logit proportional odds form, multinomial logistic regression provided biased risk estimates, although these biases were modest. Nonproportional odds models require more parameters to be estimated from the data, and hence suffered more from overfitting. Despite larger sample size requirements, we generally recommend multinomial logistic regression for risk prediction modeling of discrete ordinal outcomes.

Evaluation of competing risks prediction models using polytomous discrimination index

For competing risks data, it is often important to predict a patient's outcome status at a clinically meaningful time point after incorporating the informative censoring due to competing risks. This can be done by adopting a regression model that relates the cumulative incidence probabilities to a set of covariates. To assess the performance of the resulting prediction tool, we propose an estimator of the polytomous discrimination index applicable to competing risks data, which can quantify a prognostic model's ability to discriminate among subjects from different outcome groups. The proposed estimator allows the prediction model to be subject to model misspecification and enjoys desirable asymptotic properties. We also develop an efficient computation algorithm that features a computational complexity of O(n log n). A perturbation resampling scheme is developed to achieve consistent variance estimation. Numerical results suggest that the estimator performs well under realistic sample sizes. We apply the proposed methods to a study of monoclonal gammopathy of undetermined significance.

Long working hours, poor sleep quality, and work-family conflict: determinant factors of fatigue among Indonesian tugboat crewmembers

BACKGROUND

Tugboat crewmembers are susceptible to fatigue during their 24-h work shifts, despite the availability of rest time. The fatigue experienced by seafarers contributes to marine accidents and metabolic and cardiovascular diseases, which have long-term effects. This study aimed to analyse the association between working hours and fatigue and other possibly related factors in tugboat crewmembers.

METHOD

This comparative cross-sectional study included 127 tugboat crew members from 15 randomly chosen tugboats in Samarinda Harbor, Indonesia. Their fatigue levels while at work were measured using a reaction timer and standardised questionnaire. Personal and occupational data of crewmembers, including age, marital status, rating (job ranking), duration on board, length of seafaring experience, watch system, smoking status, coffee and alcohol consumption, and working hours, were collected. Moreover, sleep quality and stress levels related to work-family conflict were measured and analysed using the Pittsburgh Sleep Quality Index (PSQI) and Work-Family Conflict Scale (WCFS), respectively.

RESULTS

The study found that 40.2% of the subjects were classified as having fatigue. The determinant factors were long working hours (> 72 h/week), poor sleep quality, and work-family conflict [adj. OR = 13.32; 95%-CI (4.78-31.23)] and p < 0.001, [adj. OR = 4.49 (1.39-14.52)] and p = 0.012, [adj. OR = 2.87 (1.12-7.33)] and p = 0.028, respectively. However, personal and occupational factors, including age, marital status, duration on board, length of seafaring experience, smoking status, and coffee and alcohol consumption, were not significantly associated with crewmember fatigue.

CONCLUSION

The incidence of fatigue among Indonesian tugboat crewmembers operating on the Mahakam River was considerably high. Working hours, sleep quality, and work-family conflict were strongly associated with fatigue in tugboat crewmembers; therefore, the working hours of tugboat crewmembers need to be improved. Crewmember lifestyle variables need to be studied further.

Predictive markers of nonalcoholic fatty liver disease in lean patients. A multinomial regression model and a 2k factorial analysis

OBJECTIVE

Nonalcoholic fatty liver disease (NAFLD) is associated with obesity and insulin resistance; however, there is a group of non-obese patients with NAFLD that need to be characterized. Our aim was to evaluate the factors associated with NAFLD in non-obese subjects in a third-level hospital.

METHODS

A comparative cross-sectional study was performed. Participants were divided into four groups: non-obese without NAFLD (group 1), non-obese with NAFLD (group 2), obese without NAFLD (group 3), and obese with NAFLD (group 4). We evaluated the effect of clinical and biochemical characteristics with the disease by groups using a multinomial regression model and a 2K factorial analysis.

RESULTS

We included 278 participants. Low platelet-lymphocyte ratio (PLR) as a novel parameter associated with NAFLD in non-obese subjects. Age, uric acid, alanine transaminase (ALT), high-density lipoprotein (HDL)-cholesterol, and neutrophil-lymphocyte ratio (NLR) were other related parameters (akaike information criterion = 557). NLR had the larger OR in groups with NAFLD (lean with NAFLD 7.12, obese with NAFLD 13.02). The 2k factorial design found inverse effect on PLR by NAFLD (effect -21.89, P < 0.001), which was higher than BMI (effect -1.33, P < 0.045).

CONCLUSION

Our study found that PLR is a novel parameter with inverse correlation with NAFLD in non-obese patients. Other related parameters are age, hyperuricemia, elevation of ALT and NLR, and low HDL-cholesterol.

Predicting correlated outcomes from molecular data

MOTIVATION

Multivariate (multi-target) regression has the potential to outperform univariate (single-target) regression at predicting correlated outcomes, which frequently occur in biomedical and clinical research. Here we implement multivariate lasso and ridge regression using stacked generalisation.

RESULTS

Our flexible approach leads to predictive and interpretable models in high-dimensional settings, with a single estimate for each input-output effect. In the simulation, we compare the predictive performance of several state-of-the-art methods for multivariate regression. In the application, we use clinical and genomic data to predict multiple motor and non-motor symptoms in Parkinson's disease patients. We conclude that stacked multivariate regression, with our adaptations, is a competitive method for predicting correlated outcomes.

AVAILABILITY AND IMPLEMENTATION

The R package joinet is available on GitHub (https://github.com/rauschenberger/joinet) and cran (https://cran.r-project.org/package=joinet).

SUPPLEMENTARY INFORMATION

Supplementary tables and figures are available at Bioinformatics online.

Patterns of online health information seeking after visiting a physician: perceptions of adolescents from high schools in central Belgrade, Serbia

BACKGROUND

Browsing webpages that offer health information allows visitors to remain anonymous, particularly when researching sensitive topics. Uncertainty about confidentiality may be a barrier for adolescents to discuss their health in-person with a physician and seek further health information on the Internet after seeing a physician.

OBJECTIVE

To explore factors contributing to perceived online health information seeking after visiting a physician in a sample of high school students.

METHODS

A cross-sectional study was conducted from December 2016 to January 2017. The study included 702 high school students. Socio-demographic and behavioral questionnaire as well as the electronic health (e-health) literacy scale (eHEALS) were used to collect data. Students described their perceptions of what they commonly do after visiting a physician, which was not connected to a particular clinic at any point in time.

RESULTS

A total of 347 students (49.4%) perceived that they search for online health information after visiting a physician. Attending humanities-languages school program, lower education level of parents, being older at first Internet use, stronger influence of online health information on students' behaviour, better e-health literacy, use of smartphones, YouTube, social networks and heath forums were associated with perceived online health information seeking after visiting a physician.

CONCLUSION

One-half of high school students in this study perceived that they search for online health information after having visited a physician. Our results suggest that many adolescents might seek additional information about health.

Developing more generalizable prediction models from pooled studies and large clustered data sets

Prediction models often yield inaccurate predictions for new individuals. Large data sets from pooled studies or electronic healthcare records may alleviate this with an increased sample size and variability in sample characteristics. However, existing strategies for prediction model development generally do not account for heterogeneity in predictor‐outcome associations between different settings and populations. This limits the generalizability of developed models (even from large, combined, clustered data sets) and necessitates local revisions. We aim to develop methodology for producing prediction models that require less tailoring to different settings and populations. We adopt internal‐external cross‐validation to assess and reduce heterogeneity in models' predictive performance during the development. We propose a predictor selection algorithm that optimizes the (weighted) average performance while minimizing its variability across the hold‐out clusters (or studies). Predictors are added iteratively until the estimated generalizability is optimized. We illustrate this by developing a model for predicting the risk of atrial fibrillation and updating an existing one for diagnosing deep vein thrombosis, using individual participant data from 20 cohorts (N = 10 873) and 11 diagnostic studies (N = 10 014), respectively. Meta‐analysis of calibration and discrimination performance in each hold‐out cluster shows that trade‐offs between average and heterogeneity of performance occurred. Our methodology enables the assessment of heterogeneity of prediction model performance during model development in multiple or clustered data sets, thereby informing researchers on predictor selection to improve the generalizability to different settings and populations, and reduce the need for model tailoring. Our methodology has been implemented in the R package metamisc.

Predicting onset of early- and late-treatment resistance in first-episode schizophrenia patients using advanced shrinkage statistical methods in a small sample

Evidence suggests there are two treatment-resistant schizophrenia subtypes (i.e. early treatment resistant (E-TR) and late-treatment resistant (L-TR)). We aimed to develop prediction models for estimating individual risk for these outcomes by employing advanced statistical shrinkage methods. 239 first-episode schizophrenia (FES) patients were followed-up for approximately 5 years after first presentation to psychiatric services; of these, n=56 (25.2%) were defined as E-TR and n=24 (12.6%) were defined as L-TR. Using known risk factors for poor schizophrenia outcomes, we developed prediction models for E-TR and L-TR using LASSO and RIDGE logistic regression models. Models' internal validation was performed employing Harrell's optimism-correction with repeated cross-validation; their predictive accuracy was assessed through discrimination and calibration. Both LASSO and RIDGE models had high discrimination, good calibration. While LASSO had moderate sensitivity for estimating an individual risk for E-TR and L-TR, sensitivity estimated for RIDGE model for these outcomes was extremely low, which was due to having a very large estimated optimism. Although it was possible to discriminate with sufficient accuracy who would meet criteria for E-TR and L-TR during the 5-year follow-up after first contact with mental health services for schizophrenia, further work is necessary to improve sensitivity for these models.

Depression, Anxiety, and Lifestyle Among Essential Workers: A Web Survey From Brazil and Spain During the COVID-19 Pandemic

BACKGROUND

Essential workers have been shown to present a higher prevalence of positive screenings for anxiety and depression during the COVID-19 pandemic. Individuals from countries with socioeconomic inequalities may be at increased risk for mental health disorders.

OBJECTIVE

We aimed to assess the prevalence and predictors of depression, anxiety, and their comorbidity among essential workers in Brazil and Spain during the COVID-19 pandemic.

METHODS

A web survey was conducted between April and May 2020 in both countries. The main outcome was a positive screening for depression only, anxiety only, or both. Lifestyle was measured using a lifestyle multidimensional scale adapted for the COVID-19 pandemic (Short Multidimensional Inventory Lifestyle Evaluation-Confinement). A multinomial logistic regression model was performed to evaluate the factors associated with depression, anxiety, and the presence of both conditions.

RESULTS

From the 22,786 individuals included in the web survey, 3745 self-reported to be essential workers. Overall, 8.3% (n=311), 11.6% (n=434), and 27.4% (n=1027) presented positive screenings for depression, anxiety, and both, respectively. After adjusting for confounding factors, the multinomial model showed that an unhealthy lifestyle increased the likelihood of depression (adjusted odds ratio [AOR] 4.00, 95% CI 2.72-5.87), anxiety (AOR 2.39, 95% CI 1.80-3.20), and both anxiety and depression (AOR 8.30, 95% CI 5.90-11.7). Living in Brazil was associated with increased odds of depression (AOR 2.89, 95% CI 2.07-4.06), anxiety (AOR 2.81, 95%CI 2.11-3.74), and both conditions (AOR 5.99, 95% CI 4.53-7.91).

CONCLUSIONS

Interventions addressing lifestyle may be useful in dealing with symptoms of common mental disorders during the strain imposed among essential workers by the COVID-19 pandemic. Essential workers who live in middle-income countries with higher rates of inequality may face additional challenges. Ensuring equitable treatment and support may be an important challenge ahead, considering the possible syndemic effect of the social determinants of health.

Clinical prediction models to predict the risk of multiple binary outcomes: a comparison of approaches

Clinical prediction models (CPMs) can predict clinically relevant outcomes or events. Typically, prognostic CPMs are derived to predict the risk of a single future outcome. However, there are many medical applications where two or more outcomes are of interest, meaning this should be more widely reflected in CPMs so they can accurately estimate the joint risk of multiple outcomes simultaneously. A potentially naïve approach to multi-outcome risk prediction is to derive a CPM for each outcome separately, then multiply the predicted risks. This approach is only valid if the outcomes are conditionally independent given the covariates, and it fails to exploit the potential relationships between the outcomes. This paper outlines several approaches that could be used to develop CPMs for multiple binary outcomes. We consider four methods, ranging in complexity and conditional independence assumptions: namely, probabilistic classifier chain, multinomial logistic regression, multivariate logistic regression, and a Bayesian probit model. These are compared with methods that rely on conditional independence: separate univariate CPMs and stacked regression. Employing a simulation study and real-world example, we illustrate that CPMs for joint risk prediction of multiple outcomes should only be derived using methods that model the residual correlation between outcomes. In such a situation, our results suggest that probabilistic classification chains, multinomial logistic regression or the Bayesian probit model are all appropriate choices. We call into question the development of CPMs for each outcome in isolation when multiple correlated or structurally related outcomes are of interest and recommend more multivariate approaches to risk prediction.

Data sharing policies of journals in life, health, and physical sciences indexed in Journal Citation Reports

Background

Many scholarly journals have established their own data-related policies, which specify their enforcement of data sharing, the types of data to be submitted, and their procedures for making data available. However, except for the journal impact factor and the subject area, the factors associated with the overall strength of the data sharing policies of scholarly journals remain unknown. This study examines how factors, including impact factor, subject area, type of journal publisher, and geographical location of the publisher are related to the strength of the data sharing policy.

Methods

From each of the 178 categories of the Web of Science’s 2017 edition of Journal Citation Reports, the top journals in each quartile (Q1, Q2, Q3, and Q4) were selected in December 2018. Of the resulting 709 journals (5%), 700 in the fields of life, health, and physical sciences were selected for analysis. Four of the authors independently reviewed the results of the journal website searches, categorized the journals’ data sharing policies, and extracted the characteristics of individual journals. Univariable multinomial logistic regression analyses were initially conducted to determine whether there was a relationship between each factor and the strength of the data sharing policy. Based on the univariable analyses, a multivariable model was performed to further investigate the factors related to the presence and/or strength of the policy.

Results

Of the 700 journals, 308 (44.0%) had no data sharing policy, 125 (17.9%) had a weak policy, and 267 (38.1%) had a strong policy (expecting or mandating data sharing). The impact factor quartile was positively associated with the strength of the data sharing policies. Physical science journals were less likely to have a strong policy relative to a weak policy than Life science journals (relative risk ratio [RRR], 0.36; 95% CI [0.17–0.78]). Life science journals had a greater probability of having a weak policy relative to no policy than health science journals (RRR, 2.73; 95% CI [1.05–7.14]). Commercial publishers were more likely to have a weak policy relative to no policy than non-commercial publishers (RRR, 7.87; 95% CI, [3.98–15.57]). Journals by publishers in Europe, including the majority of those located in the United Kingdom and the Netherlands, were more likely to have a strong data sharing policy than a weak policy (RRR, 2.99; 95% CI [1.85–4.81]).

Conclusions

These findings may account for the increase in commercial publishers’ engagement in data sharing and indicate that European national initiatives that encourage and mandate data sharing may influence the presence of a strong policy in the associated journals. Future research needs to explore the factors associated with varied degrees in the strength of a data sharing policy as well as more diverse characteristics of journals related to the policy strength.

Dysbiosis in the Dead: Human Postmortem Microbiome Beta-Dispersion as an Indicator of Manner and Cause of Death

The postmortem microbiome plays an important functional role in host decomposition after death. Postmortem microbiome community successional patterns are specific to body site, with a significant shift in composition 48 h after death. While the postmortem microbiome has important forensic applications for postmortem interval estimation, it also has the potential to aid in manner of death (MOD) and cause of death (COD) determination as a reflection of antemortem health status. To further explore this association, we tested beta-dispersion, or the variability of microbiomes within the context of the “Anna Karenina Principle” (AKP). The foundational principle of AKP is that stressors affect microbiomes in unpredictable ways, which increases community beta-dispersion. We hypothesized that cases with identified M/CODs would have differential community beta-dispersion that reflected antemortem conditions, specifically that cardiovascular disease and/or natural deaths would have higher beta-dispersion compared to other deaths (e.g., accidents, drug-related deaths). Using a published microbiome data set of 188 postmortem cases (five body sites per case) collected during routine autopsy in Wayne County (Detroit), MI, we modeled beta-dispersion to test for M/COD associations a priori. Logistic regression models of beta-dispersion and case demographic data were used to classify M/COD. We demonstrated that beta-dispersion, along with case demographic data, could distinguish among M/COD – especially cardiovascular disease and drug related deaths, which were correctly classified in 79% of cases. Binary logistic regression models had higher correct classifications than multinomial logistic regression models, but changing the defined microbial community (e.g., full vs. non-core communities) used to calculate beta-dispersion overall did not improve model classification or M/COD. Furthermore, we tested our analytic approach on a case study that predicted suicides from other deaths, as well as distinguishing MOD (e.g., homicides vs. suicides) within COD (e.g., gunshot wound). We propose an analytical workflow that combines postmortem microbiome indicator taxa, beta-dispersion, and case demographic data for predicting MOD and COD classifications. Overall, we provide further evidence the postmortem microbiome is linked to the host’s antemortem health condition(s), while also demonstrating the potential utility of including beta-dispersion (a non-taxon dependent approach) coupled with case demographic data for death determination.

Understanding consumer behavior patterns in antibiotic usage for upper respiratory tract infections: A study protocol based on the COM-B framework

BACKGROUND

Irrational use of antibiotics is prevalent worldwide. But our understanding on consumer behaviors in the use of antibiotics is very limited. This study aims to identify consumer behavior patterns in the use of antibiotics for upper respiratory tract infections (URTIs).

METHODS

The study will employ a mixed methods approach based on the "Capacity & Opportunity & Motivation - Behavior" (COM-B) framework. The COM-B attributes of consumers in relation to the use of antibiotics will be extracted from a systematic literature review. Semi-structured in-depth interviews will be conducted on 20-25 community residents with URTI symptoms over the past three months to illustrate the meaning and implications of the thematic categories of COM-B attributes for the purpose of measurement development. The measurement instruments will be modified and validated through Delphi consultations with 15 experts and a survey of 300 adult residents in Wuhan. A cross-sectional survey using the finalised measurement instruments will be conducted on 2700 adult residents randomly selected from 18 residential communities across 9 municipalities in 3 provinces in China. Multi-level latent class analyses will be performed to categeorise the respondents based on the indicators measuring the behavioral features (need recognition, information searching, alternative assessment, purchase, use, and post-use evaluation) of consumers in purchasing, consuming and disposing antibiotics for URTIs. Multi-nominal regression analyses will be performed to determine the predictors of different behavior patterns.

DISCUSSION

This study aims to classify consumers into distinguished categories of behavior patterns toward the use of antibiotics for URTIs. Such a classification system categories the consumers with similar behavior features into the same group so that better targeted interventions can be developed. The COM-B model adopted in this study can also help us better understand the underlying mechanisms of different behavior patterns of consumers.