Bias and Precision of the "Multiple Imputation, Then Deletion" Method for Dealing With Missing Outcome Data

Socioeconomic status and sleep duration among a representative, cross-sectional sample of US adults

BACKGROUND

Sleep is a crucial determinant of physical and mental health outcomes, and insufficient sleep is highly prevalent among United States adults. Although some risk factors of poor sleep have been extensively studied, including substance use, age, health behaviors, and others, the associations between socioeconomic status (SES) and sleep remain inconclusive. There is limited evidence on SES and sleep duration among the US adult population. This study analyzed the relationships between three SES indicators (poverty, education, and food security), and sleep duration.

METHODS

We used responses from the 2017-March 2020 National Health and Nutrition Examination Survey (NHANES). Respondents younger than 25 years old were excluded. Sleep duration was classified using self-reported sleep time and stratified by work vs. non-workdays. SES was operationalized using three indicators: poverty-income ratio, educational attainment, and food security status. We imputed missing data for socioeconomic status and outcome variables using multiple imputation. Weighted Poisson regression models with robust standard errors were used to calculate the crude and adjusted prevalence ratios for insufficient sleep duration (< 7 h of self-reported sleep) on workdays and non-workdays separately by each of the three SES indicators.

RESULTS

We included a total of 8,457 individuals. In the adjusted model, participants with lower income, educational status, and food security had significantly higher prevalence of insufficient sleep duration on both workdays and non-workdays. For example, low-income individuals (poverty-income ratio < 1) had 1.22 (95% CI 1.04-1.44) and 2.08 (95% CI 1.61-2.67) higher prevalence of insufficient sleep as compared to high income individuals on workday and non-workdays, respectively. In general, we found larger differences by level of SES indicator for the non-workday than for the workday outcome. There were no major differences in gender-stratified analysis. We also found that lower SES was associated with higher prevalence of excessive sleep (≥ 9 h).

CONCLUSION

Socioeconomic status indicators are significantly associated with sleep duration in the US adult population. Lower SES correlates with increased prevalence of insufficient sleep duration, which has implications for the overall wellbeing of US adults with lower SES. Targeted interventions and further research are needed to reduce this disparity.

Association between blood lipid levels in early pregnancy and urinary organophosphate metabolites in the Japan Environment and Children's Study

BACKGROUND

High low-density lipoprotein cholesterol levels (LDL-C) during pregnancy have been associated with adverse pregnancy and offspring outcomes. While previous studies have suggested a potential link between organophosphate pesticide (OPP) exposure and higher LDL-C in the general population and agricultural workers, the relationship in pregnant women and the effect of body mass index on this relationship remain unclear. We examined the association between the urinary concentrations of OPP metabolites (dialkylphosphates) and blood lipid levels in pregnant women.

METHODS

We used data from the Japan Environment and Children's Study, which included 5,169 pregnant women with urinary dialkylphosphate data. We examined the association between urinary concentrations of six dialkylphosphates (DEP, DETP, DEDTP, DMP, DMTP, DMDTP) and blood lipid levels (LDL-C, total cholesterol, high-density lipoprotein cholesterol, and triglycerides) during the first trimester using multiple linear regression under a Bayesian paradigm. We examined the association between high LDL-C, defined as ≥90th percentile of LDL-C, and urinary dialkylphosphate concentrations, using multiple logistic regression under a Bayesian paradigm. These analyses were repeated in underweight, normal-weight, and overweight participants.

RESULTS

DEP, DMP, and DMTP were detected in >50 % of the participants. Multiple linear regression analyses did not show associations between LDL-C and these dialkylphosphates. Stratified analyses showed a positive association between DEP and LDL-C in overweight women (beta coefficient = 2.13, 95 % credible interval = 0.86-3.38, probability of direction (PD) = 100 %); however, the association was not significant (percentage in region of practical equivalence (% in ROPE) = 84.0). Higher DEP was significantly associated with high LDL-C (odds ratio = 1.32, 95 % credible interval = 1.13-1.55, PD = 100 %, % in ROPE = 0.2).

CONCLUSIONS

Among overweight pregnant women in the first trimester, higher urinary DEP concentrations were associated with high LDL-C. The effects of OPP on blood lipid profiles merit further investigation.

Lower HAGOS subscale scores associated with a longer duration of groin problems in football players in the subsequent season

Introduction

Groin injuries represent a considerable problem in football. Although the Adductor Strengthening Programme reduced groin injury risk, players can still experience groin symptoms throughout the season. This study aimed to determine whether preseason Copenhagen Hip and Groin Outcome Score (HAGOS) and a history of previous injury can identify individuals at risk of having a longer duration of groin problems the subsequent season, using an 'any physical complaint' definition of injury.

Methods

Preseason HAGOS score and weekly groin problems were registered with the Oslo Sports Trauma Research Center Overuse questionnaire during one full season in 632 male semiprofessional adult players.

Results

The prognostic model showed a decreased number of weeks with groin problems for each increase in HAGOS score for 'groin-related quality of life' (QOL) (IRR=0.99, p=0.003). A 10-point higher 'QOL' score predicted 10% fewer weeks of groin problems. Additionally, previous hip/groin injury was associated with a 74% increase in the number of weeks with symptoms (p<0.001).

Conclusion

The HAGOS questionnaire applied preseason can detect players at risk of getting more weeks with groin problems the following season. The 'QOL' subscale seems to be the superior subscale for estimating subsequent groin problem duration. While HAGOS appears promising in identifying players at risk, previous groin injury is the most robust indicator, showing a substantial 74% increase in weeks with symptoms.

The impact of being overweight or obese on 12 month clinical recovery in patients following lumbar microdiscectomy for radiculopathy

BACKGROUND CONTEXT

The proportion of patients who undergo lumbar microdiscectomy due to lumbar radiculopathy who are also overweight or obese is high. However, whether high body mass index (BMI) affects clinical outcomes is not well-studied.

PURPOSE

To investigate the difference in the clinical course between normal weight, overweight, and obese patients with radiculopathy who underwent lumbar microdiscectomy followed by physical therapy and to evaluate whether high BMI is associated with poor recovery.

STUDY DESIGN/SETTING

A prospective cohort study with a 12-month follow-up was conducted in a multidisciplinary clinic.

PATIENT SAMPLE

We included 583 patients (median [IQR] age: 45 [35-52] years; 41% female) with clinical signs and symptoms of lumbar radiculopathy, consistent with magnetic resonance imaging findings, who underwent microdiscectomy followed by postoperative physical therapy.

OUTCOME MEASURES

Outcomes were leg pain and back pain intensity measured with a visual analogue scale, disability measured with the Roland Morris Disability Questionnaire at 3 and 12-month follow-ups, and complications.

METHODS

Patients were classified as being normal weight (46.9%), overweight (38.4%), or obese (14.7%). A linear mixed-effects model was used to assess the difference in the clinical course of pain and disability between the three BMI categories. The association between BMI and outcomes was evaluated using univariable and multivariable logistic regression analyses.

RESULTS

All three patient groups experienced a significant improvement in leg pain, back pain, and disability over 3 and 12-month follow-up. Patients who were overweight, obese, or normal weight experienced comparable leg pain (p=.14) and disability (p=.06) over the clinical course (p=.14); however, obese patients experienced higher back pain (MD=-6.81 [95%CI: -13.50 to -0.14]; p=.03). The difference in back pain scores was not clinically relevant.

CONCLUSIONS

In the first year following lumbar microdiscectomy, patients demonstrated clinical improvements and complications that were unrelated to their preoperative BMI.

Handling missing values in the analysis of between-hospital differences in ordinal and dichotomous outcomes: a simulation study

Missing data are frequently encountered in registries that are used to compare performance across hospitals. The most appropriate method for handling missing data when analysing differences in outcomes between hospitals with a generalised linear mixed model is unclear. We aimed to compare methods for handling missing data when comparing hospitals on ordinal and dichotomous outcomes. We performed a simulation study using data from the Multicentre Randomised Controlled Trial of Endovascular Treatment for Acute Ischaemic Stroke in the Netherlands (MR CLEAN) Registry, a prospective cohort study in 17 hospitals performing endovascular therapy for ischaemic stroke in the Netherlands. The investigated methods for handling missing data, both case-mix adjustment variables and outcomes, were complete case analysis, single imputation, multiple imputation, single imputation with deletion of imputed outcomes and multiple imputation with deletion of imputed outcomes. Data were generated as missing completely at random (MCAR), missing at random and missing not at random (MNAR) in three scenarios: (1) 10% missing data in case-mix and outcome; (2) 40% missing data in case-mix and outcome; and (3) 40% missing data in case-mix and outcome with varying degree of missing data among hospitals. Bias and reliability of the methods were compared on the mean squared error (MSE, a summary measure combining bias and reliability) relative to the hospital effect estimates from the complete reference data set. For both the ordinal outcome (ie, the modified Rankin Scale) and a common dichotomised version thereof, all methods of handling missing data were biased, likely due to shrinkage of the random effects. The MSE of all methods was on average lowest under MCAR and with fewer missing data, and highest with more missing data and under MNAR. The 'multiple imputation, then deletion' method had the lowest MSE for both outcomes under all simulated patterns of missing data. Thus, when estimating hospital effects on ordinal and dichotomous outcomes in the presence of missing data, the least biased and most reliable method to handle these missing data is 'multiple imputation, then deletion'.

Characteristics and health outcomes associated with activation for self-management in patients with non-specific low back pain: A cross-sectional study

BACKGROUND

Research has shown that the course of non-specific low back pain (LBP) is influenced by, among other factors, patients' self-management abilities. Therefore, clinical guidelines recommend stimulation of self-management. Enhancing patients' self-management potentially can improve patients' health outcomes and reduce future healthcare costs for non-specific LBP.

OBJECTIVES

Which characteristics and health outcomes are associated with activation for self-management in patients with non-specific LBP?

DESIGN

Cross-sectional study.

METHOD

Patients with non-specific LBP applying for primary care physiotherapy were asked to participate. Multivariable linear regression analysis was performed to analyze the multivariable relationship between activation for self-management (Patient Activation Measure, range 0-100) and a range of characteristics, e.g., age, gender, and health outcomes, e.g., self-efficacy, pain catastrophizing.

RESULTS

The median activation for self-management score of the patients with non-specific LBP (N = 208) was 63.10 (IQR = 19.30) points. The multivariable linear regression analysis revealed that higher self-efficacy scores (B = 0.54), female gender (B = 3.64), and a middle educational level compared with a high educational level (B = -5.47) were associated with better activation for self-management in patients with non-specific LBP. The goodness-of-fit of the model was 17.24% (R2 = 0.17).

CONCLUSIONS

Patients with better activation for self-management had better self-efficacy, had a higher educational level, and were more often female. However, given the explained variance better understanding of the factors that influence the complex construct of self-management behaviour in patients who are not doing well might be needed to identify possible barriers to engage in self-management.

Development and Validation of a Model to Predict Malignancy Within the First Year After Adult Heart Transplantation

Purpose: Malignancy after heart transplantation is associated with poor outcomes. At present, no prediction model exists for any malignancy within the first year after transplant. Methods: We studied adults who underwent heart transplantation included in the multicenter, national Scientific Registry of Transplant Recipients from January 2000 through April 2021. Possible predictors of malignancy were identified based on their known association with malignancy. Multiple imputations were conducted for missing values using predictive mean matching. A multivariable logistic regression model for predicting malignancy development within the first year after transplant was developed and internally validated via 500 bootstrapped samples to estimate the optimism-corrected measures of model accuracy and performance. Results: Among the 47 212 recipients comprising 16% females, 76% whites, 7% with prior malignancy, and a median age of 56 years; 865 (2.3% of those with non-missing data) developed malignancy within the first year after transplant. Prior malignancy, older age at heart transplantation, white race, and nonischemic heart failure etiology were the strongest predictors of new malignancy. The optimism-corrected model had modest discrimination (C-statistic: 0.70, 95% CI: 0.69-0.72) and good calibration and performance (calibration slope: 0.96; Cox-Snell R²: 0.063), particularly at lower predicted risk. A nomogram for the practicing clinician was developed. Conclusions: Using selection variables previously linked to cutaneous malignancy, our model was modestly predictive of the development of any malignancy in the first year after heart transplantation. Future research could identify factors that may improve malignancy prediction, including incorporation of time-to-event data.

Is there a dose-response relationship? Investigating the functional form between COVID-19 incidence rates and life satisfaction in a multilevel framework

While there is plenty of research linking the effects of the global COVID-19 pandemic to a drastic reduction of life satisfaction in the population, there is little information on the functional form of this relationship. Until now, one could suspect that this association is linear and a higher number of COVID-19 infections in a region leads to a continuous decline of satisfaction. However, there are reasons to assume that this interrelation is indeed more complex and deserves further attention. To resolve this question, high-quality panel data of the first wave of COVID-19 from Germany are analysed in a fixed-effect multilevel framework. With information from more than 6,000 respondents (after imputation) nested in 339 federal districts, we estimate linear models with higher-order terms up to the fifth degree of median COVID-19 incidence rates and random intercepts for districts to describe the functional form. The results indicate that even regions with very low incidences are affected and a linear decline of satisfaction is only apparent for rather low incidence levels, quickly reaching a plateau, which is then quite constant, even for higher incidence levels. These findings indicate that at least in rich and industrialized countries like Germany, assuming a strictly linear relation between incidences and change of satisfaction is not appropriate.

Development and validation of a meta-learner for combining statistical and machine learning prediction models in individuals with depression

BACKGROUND

The debate of whether machine learning models offer advantages over standard statistical methods when making predictions is ongoing. We discuss the use of a meta-learner model combining both approaches as an alternative.

METHODS

To illustrate the development of a meta-learner, we used a dataset of 187,757 people with depression. Using 31 variables, we aimed to predict two outcomes measured 60 days after initiation of antidepressant treatment: severity of depressive symptoms (continuous) and all-cause dropouts (binary). We fitted a ridge regression and a multi-layer perceptron (MLP) deep neural network as two separate prediction models ("base-learners"). We then developed two "meta-learners", combining predictions from the two base-learners. To compare the performance across the different methods, we calculated mean absolute error (MAE, for continuous outcome) and the area under the receiver operating characteristic curve (AUC, for binary outcome) using bootstrapping.

RESULTS

Compared to the best performing base-learner (MLP base-learner, MAE at 4.63, AUC at 0.59), the best performing meta-learner showed a 2.49% decrease in MAE at 4.52 for the continuous outcome and a 6.47% increase in AUC at 0.60 for the binary outcome.

CONCLUSIONS

A meta-learner approach may effectively combine multiple prediction models. Choosing between statistical and machine learning models may not be necessary in practice.

The impact of COVID-19 on the psychological distress of youths in Japan: A latent growth curve analysis

BACKGROUND

This study expands on previous studies that have investigated the impact of the novel coronavirus disease (COVID-19) on mental health in two ways. We first model the change in mental health, then examine the various factors that predict changes in psychological distress.

METHOD

Longitudinal surveys were conducted once each in 2015, 2017, and 2019 on mothers and their children born between April 2000 and March 2001 (n = 1854), and three times in 2020 (February, July, and December) on the children in Japan. A latent growth curve model with four time points from December 2019 to December 2020 was used to depict the changes in the psychological distress of youths and to examine the factor associated with the level and change in psychological distress.

RESULTS

The psychological distress of youths increased from December 2019 to July 2020, especially among female youths, then decreased in December 2020. Initial health status and psychological traits were related to the initial level of psychological distress, but not the change. Gender was not related to the initial level of psychological distress but an increase in distress.

CONCLUSION

Although the effect size was small, gender was related to changes in distress during the COVID-19 pandemic. Other factors, such as health-related characteristics and personality traits, were associated with the level of distress before the pandemic but could not explain the changes in distress during the pandemic.

A population-based study of the intersection of sexual identity and race/ethnicity on physiological risk factors for CVD among U.S. adults (ages 18-59)

Objectives: Sexual minorities face significant psychosocial stressors (such as discrimination and violence) that impact their health. Several studies indicate that sexual minority women (SMW) and bisexual men may be at highest risk for cardiovascular disease (CVD), but limited research has examined physiological CVD risk or racial/ethnic differences. This study sought to examine racial/ethnic differences in physiological risk factors for CVD among sexual minority and heterosexual adults.Design: We analyzed data from the National Health and Nutrition Examination Survey (2001-2016) using sex-stratified multiple linear regression models to estimate differences in physiological CVD risk. We compared sexual minorities (gay/lesbian, bisexual, 'not sure') to heterosexual participants first without regard to race/ethnicity. Then we compared sexual minorities by race/ethnicity to White heterosexual participants.Results: The sample included 22,305 participants (ages 18-59). Lesbian women had higher body mass index (BMI) but lower total cholesterol than heterosexual women. Bisexual women had higher systolic blood pressure (SBP). Gay men had lower BMI and glycosylated hemoglobin (HbA1c) relative to heterosexual men. White and Black lesbian women and bisexual women of all races/ethnicities had higher BMI than White heterosexual women; Black bisexual women had higher SBP and HbA1c. Black sexual minority men had higher HbA1c relative to White heterosexual men. Latino 'not sure' men also had higher SBP, HbA1c, and total cholesterol than White heterosexual men.Conclusions: Given evidence of higher CVD risk in sexual minority people of color relative to White heterosexuals, there is a need for health promotion initiatives to address these disparities. Additional research that incorporates longitudinal designs and examines the influence of psychosocial stressors on CVD risk in sexual minorities is recommended. Findings have implications for clinical and policy efforts to promote the cardiovascular health of sexual minorities.

The association between rheumatoid arthritis and cardiovascular disease among adults in the United States during 1999-2018, and age-related effect modification in relative and absolute scales

PURPOSE

To explore the rheumatoid arthritis (RA)-cardiovascular diseases (CVD) association in relative and absolute risk scales among US adults aged ≥20 years over time and the effect modification of the association by age.

METHODS

We analyzed aggregated data from all ten continuous National Health and Nutrition Examination Survey cycles. A sample of 43,184 complete-case subjects was considered. The design-based regressions were used to investigate the associations in relative and absolute scales.

RESULTS

In relative scale, the CVD odds ratio was 2.32, 2.19, and 1.97 among adults with RA than no arthritis in 1999-2006, 2007-2012, and 2013-2018 cycles, respectively. This time trend was not statistically significant. The absolute risk estimates were 11, 10, and 9 per 100 CVD events. We also observed a significant effect modification by age; the higher relative risk among younger adults (<50 years) with RA and higher absolute risk in older adults (≥80 years) with RA were consistent across survey cycles.

CONCLUSIONS

There is a significant association between RA and CVD among US adults in both relative and absolute risks. Moreover, age is a significant effect modifier for this association; but with opposing age-related trends in relative and absolute scales.

The 3-Month Effectiveness of a Stratified Blended Physiotherapy Intervention in Patients With Nonspecific Low Back Pain: Cluster Randomized Controlled Trial

Background

Patient education, home-based exercise therapy, and advice on returning to normal activities are established physiotherapeutic treatment options for patients with nonspecific low back pain (LBP). However, the effectiveness of physiotherapy interventions on health-related outcomes largely depends on patient self-management and adherence to exercise and physical activity recommendations. e-Exercise LBP is a recently developed stratified blended care intervention comprising a smartphone app integrated with face-to-face physiotherapy treatment. Following the promising effects of web-based applications on patients’ self-management skills and adherence to exercise and physical activity recommendations, it is hypothesized that e-Exercise LBP will improve patients’ physical functioning.

Objective

This study aims to investigate the short-term (3 months) effectiveness of stratified blended physiotherapy (e-Exercise LBP) on physical functioning in comparison with face-to-face physiotherapy in patients with nonspecific LBP.

Methods

The study design was a multicenter cluster randomized controlled trial with intention-to-treat analysis. Patients with nonspecific LBP aged ≥18 years were asked to participate in the study. The patients were treated with either stratified blended physiotherapy or face-to-face physiotherapy. Both interventions were conducted according to the Dutch physiotherapy guidelines for nonspecific LBP. Blended physiotherapy was stratified according to the patients’ risk of developing persistent LBP using the Keele STarT Back Screening Tool. The primary outcome was physical functioning (Oswestry Disability Index, range 0-100). Secondary outcomes included pain intensity, fear-avoidance beliefs, and self-reported adherence. Measurements were taken at baseline and at the 3-month follow-up.

Results

Both the stratified blended physiotherapy group (104/208, 50%) and the face-to-face physiotherapy group (104/208, 50%) had improved clinically relevant and statistically significant physical functioning; however, there was no statistically significant or clinically relevant between-group difference (mean difference −1.96, 95% CI −4.47 to 0.55). For the secondary outcomes, stratified blended physiotherapy showed statistically significant between-group differences in fear-avoidance beliefs and self-reported adherence. In patients with a high risk of developing persistent LBP (13/208, 6.3%), stratified blended physiotherapy showed statistically significant between-group differences in physical functioning (mean difference −16.39, 95% CI −27.98 to −4.79) and several secondary outcomes.

Conclusions

The stratified blended physiotherapy intervention e-Exercise LBP is not more effective than face-to-face physiotherapy in patients with nonspecific LBP in improving physical functioning in the short term. For both stratified blended physiotherapy and face-to-face physiotherapy, within-group improvements were clinically relevant. To be able to decide whether e-Exercise LBP should be implemented in daily physiotherapy practice, future research should focus on the long-term cost-effectiveness and determine which patients benefit most from stratified blended physiotherapy.

Trial Registration

ISRCTN Registry 94074203; https://doi.org/10.1186/ISRCTN94074203

International Registered Report Identifier (IRRID)

RR2-https://doi.org/10.1186/s12891-020-3174-z

How Trust Makes a Difference: The Impact of the First Wave of the COVID-19 Pandemic on Life Satisfaction in Germany

The extraordinary COVID-19 pandemic is one of the most severe disruptions of human life since the end of World War II, even in rich and industrialized countries like Germany. The introduction of a rather comprehensive "lockdown" and the restriction of multiple basic civil rights have affected the population in many areas of life, like employment, economic prosperity, health and trust in public institutions. The question arises how life satisfaction is influenced by these measures in detail and whether there are interactions between institutional trust, life satisfaction and time of crisis. Fixed-effect regression analyses using German National Educational Panel Study (NEPS) data demonstrate that life satisfaction has fallen sharply after the onset of the crisis and that interaction effects with institutional trust are present. Individuals with low levels of pre-crisis trust in institutions like the government, courts or the media report a stronger decrease of satisfaction than individuals with higher levels of trust. We believe that these results are relevant to explain the role of institutions in times of crisis and might serve as foundations for interventions to strengthen trust and increase overall satisfaction.

A New Path to Address Multimorbidity? Longitudinal Analyses of Retirement Sequences and Chronic Diseases in Old Age

Chronic disease and multimorbidity are growing health challenges for aging populations, often coinciding with retirement. We examine late-life predictors of multimorbidity, focusing on the association between retirement sequences and number of chronic diseases. We modeled the number of chronic diseases as a function of six types of previously identified 10-year retirement sequences using Health and Retirement Study (HRS) data for 7,880 Americans observed between ages 60 to 61 and 70 to 71. Our results show that at baseline, the adjusted prevalence of multimorbidity was lowest in sequences characterized by late retirement from full-time work and highest in sequences characterized by early labor-force disengagement. Age increases in multimorbidity varied across retirement sequences, though overall differences in prevalence persisted at age 70 to 71. Earlier life disadvantages did not moderate these associations. Findings suggest further investigation of policies that target health limitations affecting work, promote continued beneficial employment opportunities, and ultimately leverage retirement sequences as a novel path to influence multimorbidity in old age.

A machine learning-based risk stratification model for ventricular tachycardia and heart failure in hypertrophic cardiomyopathy

BACKGROUND

Machine learning (ML) and artificial intelligence are emerging as important components of precision medicine that enhance diagnosis and risk stratification. Risk stratification tools for hypertrophic cardiomyopathy (HCM) exist, but they are based on traditional statistical methods. The aim was to develop a novel machine learning risk stratification tool for the prediction of 5-year risk in HCM. The goal was to determine if its predictive accuracy is higher than the accuracy of the state-of-the-art tools.

METHOD

Data from a total of 2302 patients were used. The data were comprised of demographic characteristics, genetic data, clinical investigations, medications, and disease-related events. Four classification models were applied to model the risk level, and their decisions were explained using the SHAP (SHapley Additive exPlanations) method. Unwanted cardiac events were defined as sustained ventricular tachycardia occurrence (VT), heart failure (HF), ICD activation, sudden cardiac death (SCD), cardiac death, and all-cause death.

RESULTS

The proposed machine learning approach outperformed the similar existing risk-stratification models for SCD, cardiac death, and all-cause death risk-stratification: it achieved higher AUC by 17%, 9%, and 1%, respectively. The boosted trees achieved the best performing AUC of 0.82. The resulting model most accurately predicts VT, HF, and ICD with AUCs of 0.90, 0.88, and 0.87, respectively.

CONCLUSIONS

The proposed risk-stratification model demonstrates high accuracy in predicting events in patients with hypertrophic cardiomyopathy. The use of a machine-learning risk stratification model may improve patient management, clinical practice, and outcomes in general.

Polypharmacy and Frailty among Hemodialysis Patients

BACKGROUND

Both polypharmacy and frailty are highly prevalent among the patients on hemodialysis and associated with adverse outcomes; however, little is known about the association between them.

METHODS

We examined 337 patients enrolled in the ACTIVE/ADIPOSE dialysis cohort study between 2009 and 2011. The number of prescribed medications and frailty were assessed at baseline, 12, and 24 months. Frailty was defined based upon the Fried's frailty phenotype. We used logistic regression with generalized estimating equations to model the association of the number of medications and frailty at baseline and over time. A competing-risk regression analysis was also used to assess the association between the number of medications and incidence of frailty.

RESULTS

The mean number of medications was 10 ± 5, and 94 patients (28%) were frail at baseline. Patients taking >11 medications showed higher odds for frailty than the patients taking fewer than 8 medications (OR 1.54, 95% CI 1.05-2.26). During the 2-year of follow-up, 87 patients developed frailty among those who were nonfrail at baseline. Compared with the patients taking fewer than 8 medications, the incidence of frailty was approximately 2-fold in those taking >11 medications (sub-distribution hazard ratio 2.15, 95% CI 1.32-3.48).

CONCLUSIONS

Using a higher number of medications was associated with frailty and the incidence of frailty among hemodialysis patients. Minimizing polypharmacy may reduce the incidence and prevalence of frailty among dialysis patients.

Missing Data in Prediction Research: A Five-Step Approach for Multiple Imputation, Illustrated in the CENTER-TBI Study

In medical research, missing data is common. In acute diseases, such as traumatic brain injury (TBI), even well-conducted prospective studies may suffer from missing data in baseline characteristics and outcomes. Statistical models may simply drop patients with any missing values, potentially leaving a selected subset of the original cohort. Imputation is widely accepted by methodologists as an appropriate way to deal with missing data. We aim to provide practical guidance on handling missing data for prediction modeling. We hereto propose a five-step approach, centered around single and multiple imputation: 1) explore the missing data patterns; 2) choose a method of imputation; 3) perform imputation; 4) assess diagnostics of the imputation; and 5) analyze the imputed data sets. We illustrate these five steps with the estimation and validation of the IMPACT (International Mission on Prognosis and Analysis of Clinical Trials in Traumatic Brain Injury) prognostic model in 1375 patients from the CENTER-TBI database, included in 53 centers across 17 countries, with moderate or severe TBI in the prospective European CENTER-TBI study. Future prediction modeling studies in acute diseases may benefit from following the suggested five steps for optimal statistical analysis and interpretation, after maximal effort has been made to minimize missing data.

Ultrasound-based risk model for preoperative prediction of lymph-node metastases in women with endometrial cancer: model-development study

OBJECTIVE

To develop a preoperative risk model, using endometrial biopsy results and clinical and ultrasound variables, to predict the individual risk of lymph-node metastases in women with endometrial cancer.

METHODS

A mixed-effects logistic regression model for prediction of lymph-node metastases was developed in 1501 prospectively included women with endometrial cancer undergoing transvaginal ultrasound examination before surgery, from 16 European centers. Missing data, including missing lymph-node status, were imputed. Discrimination, calibration and clinical utility of the model were evaluated using leave-center-out cross validation. The predictive performance of the model was compared with that of risk classification from endometrial biopsy alone (high-risk defined as endometrioid cancer Grade 3/non-endometrioid cancer) or combined endometrial biopsy and ultrasound (high-risk defined as endometrioid cancer Grade 3/non-endometrioid cancer/deep myometrial invasion/cervical stromal invasion/extrauterine spread).

RESULTS

Lymphadenectomy was performed in 691 women, of whom 127 had lymph-node metastases. The model for prediction of lymph-node metastases included the predictors age, duration of abnormal bleeding, endometrial biopsy result, tumor extension and tumor size according to ultrasound and undefined tumor with an unmeasurable endometrium. The model's area under the curve was 0.73 (95% CI, 0.68-0.78), the calibration slope was 1.06 (95% CI, 0.79-1.34) and the calibration intercept was 0.06 (95% CI, -0.15 to 0.27). Using a risk threshold for lymph-node metastases of 5% compared with 20%, the model had, respectively, a sensitivity of 98% vs 48% and specificity of 11% vs 80%. The model had higher sensitivity and specificity than did classification as high-risk, according to endometrial biopsy alone (50% vs 35% and 80% vs 77%, respectively) or combined endometrial biopsy and ultrasound (80% vs 75% and 53% vs 52%, respectively). The model's clinical utility was higher than that of endometrial biopsy alone or combined endometrial biopsy and ultrasound at any given risk threshold.

CONCLUSIONS

Based on endometrial biopsy results and clinical and ultrasound characteristics, the individual risk of lymph-node metastases in women with endometrial cancer can be estimated reliably before surgery. The model is superior to risk classification by endometrial biopsy alone or in combination with ultrasound. Copyright © 2019 ISUOG. Published by John Wiley & Sons Ltd.

Predictors of interleukin-1β and interleukin-1 receptor antagonist in infant saliva

OBJECTIVES

This study assesses the feasibility of measuring interleukin-1β (IL-1β) and interleukin-1 receptor antagonist (IL-1ra) in infant salivary samples as representative of pro- and anti-inflammatory processes, and explores predictors of these biomarkers in a US population.

METHODS

Data were collected from 73 US mother-infant pairs. Salivary samples were collected with an infant swab and analyzed for IL-1β, IL-1ra, and immunoglobulin A (IgA) using ELISA. Household, maternal, infant, and anthropometric predictors were selected using stepwise regression to build final multivariate models.

RESULTS

Both IL-1β and IL-1ra can be feasibly measured in infant saliva. The predictors in the final IL-1β model were IL-1ra and reported infant illness. IL-1β, IgA, infant age, household income, maternal BMI, and infant weight-for-age z-score were significant in the final model for IL-1ra.

CONCLUSIONS

IL-1β and IL-ra are useful biomarkers of immune function for infants. In particular, IL-1ra has the potential to address the relationship between immune function and body composition in the mother-infant dyad.

Sexual Identity, Adverse Life Experiences, and Cardiovascular Health in Women

BACKGROUND

Adverse life experiences (ALE; eg, discrimination and sexual abuse) may contribute to cardiovascular disease (CVD) risk in sexual minority women (SMW), but few studies have tested whether ALE explain the association of sexual identity with cardiovascular health (CVH) markers in women.

OBJECTIVE

The aim of this study was to examine sexual identity differences in CVH among women and the role of ALE.

METHODS

In the Epidemiologic Study of Risk in Women, we used multinomial logistic regression to assess sexual identity differences (SMW vs heterosexual women [reference group]) in CVH markers (ideal vs poor, intermediate vs poor) using the American Heart Association's Life's Simple 7 metric and the total score. Next, we tested whether the association of sexual identity with the total CVH score was attenuated by traditional CVD risk factors or ALE.

RESULTS

The sample consisted of 867 women (395 heterosexual, 472 SMW). Sexual minority women were more likely to have experienced discrimination (P < .001) and lifetime sexual abuse (P < .001) than heterosexual women. Sexual minority women were also less likely to meet ideal CVH criteria for current tobacco use (adjusted odds ratio, 0.43; 95% confidence interval, 0.24-0.73) or intermediate CVH criteria for body mass index (adjusted odds ratio, 0.60; 95% confidence interval, 0.40-0.92). Sexual minority women had a lower cumulative CVH score (B [SE] = -0.35 [0.14], P < .01) than heterosexual women. This difference was not explained by traditional CVD risk factors or ALE.

CONCLUSIONS

Smoking, body mass index, and fasting glucose accounted for much of the CVH disparity due to sexual identity, but those differences were not explained by ALE. Health behavior interventions tailored to SMW should be considered.

Baseline Patient Characteristics Commonly Captured Before Surgery Do Not Accurately Predict Long-Term Outcomes of Lumbar Microdiscectomy Followed by Physiotherapy

STUDY DESIGN

Prospective cohort study.

OBJECTIVE

To develop and internally validate prognostic models based on commonly collected preoperative data for good and poor outcomes of lumbar microdiscectomy followed by physiotherapy.

SUMMARY OF BACKGROUND DATA

Lumbar microdiscectomy followed by physiotherapy is a common intervention for lumbar radiculopathy. Postoperatively, a considerable percentage of people continues to experience pain and disability. Prognostic models for recovery are scarce.

METHODS

We included 298 patients with lumbar radiculopathy who underwent microdiscectomy followed by physiotherapy. Primary outcomes were recovery and secondary outcomes were pain and disability at 12 months follow-up. Potential prognostic factors were selected from sociodemographic and biomedical data commonly captured preoperatively. The association between baseline characteristics and outcomes was evaluated using multivariable logistic regression analyses.

RESULTS

At 12 months follow-up, 75.8% of the participants met the criterion for recovery. Variables in the model for good recovery included: younger age, leg pain greater than back pain, high level of disability, and a disc herniation at another level than L3-L4. The model for poor recovery included: lower educational level, prior back surgery, and disc herniation at L3-L4. Following internal validation, the explained variance (Nagelkerke R) and area under the curve for both models were poor (≤0.02 and ≤0.60, respectively). The discriminative ability of the models for disability and pain were also poor.

CONCLUSION

The outcome of microdiscectomy followed by postoperative physiotherapy cannot be predicted accurately by commonly captured preoperative sociodemographic and biomedical factors. The potential value of other biomedical, personal, and external factors should be further investigated.

LEVEL OF EVIDENCE

3.

Multinomial logistic regression with missing outcome data: An application to cancer subtypes

Many diseases such as cancer and heart diseases are heterogeneous and it is of great interest to study the disease risk specific to the subtypes in relation to genetic and environmental risk factors. However, due to logistic and cost reasons, the subtype information for the disease is missing for some subjects. In this article, we investigate methods for multinomial logistic regression with missing outcome data, including a bootstrap hot deck multiple imputation (BHMI), simple inverse probability weighted (SIPW), augmented inverse probability weighted (AIPW), and expected estimating equation (EEE) estimators. These methods are important approaches for missing data regression. The BHMI modifies the standard hot deck multiple imputation method such that it can provide valid confidence interval estimation. Under the situation when the covariates are discrete, the SIPW, AIPW, and EEE estimators are numerically identical. When the covariates are continuous, nonparametric smoothers can be applied to estimate the selection probabilities and the estimating scores. These methods perform similarly. Extensive simulations show that all of these methods yield unbiased estimators while the complete-case (CC) analysis can be biased if the missingness depends on the observed data. Our simulations also demonstrate that these methods can gain substantial efficiency compared with the CC analysis. The methods are applied to a colorectal cancer study in which cancer subtype data are missing among some study individuals.

Bullying and Peer Victimization of Minority Youth: Intersections of Sexual Identity and Race/Ethnicity

BACKGROUND

Youth with multiple minority identities, such as those who are both sexual minority (eg, lesbian, gay, bisexual) and racial/ethnic minority (eg, Black, Latino) may be at increased risk for bullying and peer victimization.

METHODS

Youth Risk Behavior Surveillance data (2011-2017) were analyzed (N = 114,881; 50.8% girls; mean age = 15.7 years, SD = 0.03). We used chi-square tests and sex-stratified multiple linear regression models to examine sexual identity and racial/ethnic differences and the intersection between sexual identity and race/ethnicity across 3 forms of bullying and peer victimization, co-occurrence of traditional and electronic bullying, and any type of bullying or peer victimization.

RESULTS

Sexual minority youth reported higher odds of bullying and peer victimization than heterosexual youth. White youth reported higher odds of bullying than racial/ethnic minority youth. In intersectional analyses, all sexual minority and racial/ethnic minority boys, and bisexual racial/ethnic minority girls were at higher risk for bullying and peer victimization compared to heterosexual peers of the same race/ethnicity.

CONCLUSIONS

This study of a large diverse sample of youth advances our understanding of vulnerability to bullying and peer victimization among youth with multiple minority identities. This research can inform policy initiatives and interventions to prevent peer victimization of vulnerable youth.

The Associations Between Physical and Psychological Symptoms and Traumatic Military Deployment Exposures

Current paradigms regarding the effects of traumatic exposures on military personnel do not consider physical symptoms unrelated to injury or illness as independent outcomes of trauma exposure, characteristically dealing with these symptoms as comorbidities of psychological disorders. Our objective was to ascertain the proportions of deployed military personnel who experienced predominantly physical symptoms, predominantly psychological symptoms, and comorbidity of the two and to examine the association between traumatic deployment exposures (TDEs) and these symptomatic profiles. Data were taken from a cross-sectional study of Australian Defence Force personnel who were deployed to the Middle East during 2001-2009 (N = 14,032). Four groups were created based on distributional splits of physical and psychological symptom scales: low-symptom, psychological, physical, and comorbid. Multinomial logistic regression models assessed the probability of symptom group membership, compared with low-symptom, as predicted by self-reported TDEs. Group proportions were: low-symptom, 78.3%; physical, 5.0%; psychological, 9.3%; and comorbid, 7.5%. TDEs were significant predictors of all symptom profiles. For subjective, objective, and human death and degradation exposures, respectively, the largest relative risk ratios (RRRs) were for the comorbid profile, RRRs = 1.47, 1.19, 1.48; followed by the physical profile, RRRs = 1.27, 1.15, 1.40; and the psychological profile, RRRs = 1.22, 1.07, 1.22. Almost half of participants with physical symptoms did not have comorbid psychological symptoms, suggesting that physical symptoms can occur as a discrete outcome trauma exposure. The similar dose-response association between TDEs and the physical and psychological profiles suggests trauma is similarly associated with both outcomes.

Preventing bias from selective non-response in population-based survey studies: findings from a Monte Carlo simulation study

Background

Health researchers often use survey studies to examine associations between risk factors at one time point and health outcomes later in life. Previous studies have shown that missing not at random (MNAR) may produce biased estimates in such studies. Medical researchers typically do not employ statistical methods for treating MNAR. Hence, there is a need to increase knowledge about how to prevent occurrence of such bias in the first place.

Methods

Monte Carlo simulations were used to examine the degree to which selective non-response leads to biased estimates of associations between risk factors and health outcomes when persons with the highest levels of health problems are under-represented or totally missing from the sample. This was examined under different response rates and different degrees of dependency between non-response and study variables.

Results

Response rate per se had little effect on bias. When extreme values on the health outcome were completely missing, rather than under-represented, results were heavily biased even at a 70% response rate. In most situations, 50–100% of this bias could be prevented by including some persons with extreme scores on the health outcome in the sample, even when these persons were under-represented. When some extreme scores were present, estimates of associations were unbiased in several situations, only mildly biased in other situations, and became biased only when non-response was related to both risk factor and health outcome to substantial degrees.

Conclusions

The potential for preventing bias by including some extreme scorers in the sample is high (50–100% in many scenarios). Estimates may then be relatively unbiased in many situations, also at low response rates. Hence, researchers should prioritize to spend their resources on recruiting and retaining at least some individuals with extreme levels of health problems, rather than to obtain very high response rates from people who typically respond to survey studies. This may contribute to preventing bias due to selective non-response in longitudinal studies of risk factors and health outcomes.

Examining Sleep Duration and Sleep Health Among Sexual Minority and Heterosexual Adults: Findings From NHANES (2005-2014)

Objectives: This study proposed to examine sexual identity differences in sleep duration and sleep health (use of sleep medications or sedatives, trouble sleeping, and diagnosis of sleeping disorders) among American adults. Methods: Data from the National Health and Nutrition Examination Survey (2005-2014) were used. Sex-stratified multiple logistic regression models were used to compare sleep duration and sleep health between sexual minority (gay/lesbian, bisexual, not-sure) and heterosexual participants, adjusted for predetermined covariates. Heterosexual participants were the reference group. Results: The analytic sample included 16,332 participants. No differences in sleep duration or sleep health were detected when gay and bisexual men were compared to heterosexual men. Not-sure men had significantly higher rates of adequate sleep duration than heterosexual men (aOR 2.35 [1.16-4.79]. Compared to heterosexual women, bisexual women reported higher rates of short sleep duration (aOR 1.29 [95% CI = 1.01-1.65]). Bisexual women were also more likely than heterosexual women to use sleep medication or sedatives (aOR 1.85 [95% CI = 1.19-2.88]), to have ever told a health professional they had trouble sleeping (aOR 1.64 [95% CI = 1.15-2.34), and to have been told by a health professional they had a sleeping disorder (aOR 2.38 [95% CI = 1.50-3.80). Lesbian and not-sure women exhibited no differences in sleep duration or sleep health compared to heterosexual women. Conclusions: Findings suggest there is a need to promote sleep health and further investigate sleeping disorders among bisexual women. Additional research should incorporate objective measures of sleep health and examine whether sleep health is associated with chronic disease in sexual minorities.

Risk factors for childhood enteric infection in urban Maputo, Mozambique: A cross-sectional study

BACKGROUND

Enteric infections are common where public health infrastructure is lacking. This study assesses risk factors for a range of enteric infections among children living in low-income, unplanned communities of urban Maputo, Mozambique.

METHODS & FINDINGS

We conducted a cross-sectional survey in 17 neighborhoods of Maputo to assess the prevalence of reported diarrheal illness and laboratory-confirmed enteric infections in children. We collected stool from children aged 1-48 months, independent of reported symptoms, for molecular detection of 15 common enteric pathogens by multiplex RT-PCR. We also collected survey and observational data related to water, sanitation, and hygiene (WASH) characteristics; other environmental factors; and social, economic, and demographic covariates. We analyzed stool from 759 children living in 425 household clusters (compounds) representing a range of environmental conditions. We detected ≥1 enteric pathogens in stool from most children (86%, 95% confidence interval (CI): 84-89%) though diarrheal symptoms were only reported for 16% (95% CI: 13-19%) of children with enteric infections and 13% (95% CI: 11-15%) of all children. Prevalence of any enteric infection was positively associated with age and ranged from 71% (95% CI: 64-77%) in children 1-11 months to 96% (95% CI: 93-98%) in children 24-48 months. We found poor sanitary conditions, such as presence of feces or soiled diapers around the compound, to be associated with higher risk of protozoan infections. Certain latrine features, including drop-hole covers and latrine walls, and presence of a water tap on the compound grounds were associated with a lower risk of bacterial and protozoan infections. Any breastfeeding was also associated with reduced risk of infection.

CONCLUSIONS

We found a high prevalence of enteric infections, primarily among children without diarrhea, and weak associations between bacterial and protozoan infections and environmental risk factors including WASH. Findings suggest that environmental health interventions to limit infections would need to be transformative given the high prevalence of enteric pathogen shedding and poor sanitary conditions observed.

TRIAL REGISTRATION

ClinicalTrials.gov NCT02362932.

The effects of medical marijuana laws on cannabis-involved driving

This study uses data from the Fatality Analysis Reporting System and a differences-in-differences model to examine the effect of state medical marijuana laws (MMLs) on cannabis-involved driving among U.S. drivers involved in a fatal crash between 1993-2014. Findings indicate that MMLs in general have a null effect on cannabis-positive driving, as do state laws with specific supply provisions including home cultivation and unlicensed or quasi-legal dispensaries. Only in jurisdictions with state-licensed medical marijuana dispensaries did the odds of marijuana-involved driving increase significantly by 14 percent, translating into an additional 87 to 113 drivers testing positive for marijuana per year. Sensitivity analyses reveal these findings to be generally robust to alternate specifications, although an observed spillover effect consistent with elevated drugged driving enforcement in bordering states weakens a causal interpretation. Still, reasonable policy implications are drawn regarding drugged driving prevention/enforcement and regulations governing dispensary delivery services and business siting decisions.

Does pattern mixture modelling reduce bias due to informative attrition compared to fitting a mixed effects model to the available cases or data imputed using multiple imputation?: a simulation study

BACKGROUND

Informative attrition occurs when the reason participants drop out from a study is associated with the study outcome. Analysing data with informative attrition can bias longitudinal study inferences. Approaches exist to reduce bias when analysing longitudinal data with monotone missingness (once participants drop out they do not return). However, findings may differ when using these approaches to analyse longitudinal data with non-monotone missingness.

METHODS

Different approaches to reduce bias due to informative attrition in non-monotone longitudinal data were compared. To achieve this aim, we simulated data from a Whitehall II cohort epidemiological study, which used the slope coefficients from a linear mixed effects model to investigate the association between smoking status at baseline and subsequent decline in cognition scores. Participants with lower cognitive scores were thought to be more likely to drop out. By using a simulation study, a range of scenarios using distributions of variables which exist in real data were compared. Informative attrition that would introduce a known bias to the simulated data was specified and the estimates from a mixed effects model with random intercept and slopes when fitted to: available cases; data imputed using multiple imputation (MI); imputed data adjusted using pattern mixture modelling (PMM) were compared. The two-fold fully conditional specification MI approach, previously validated for non-monotone longitudinal data under ignorable missing data assumption, was used. However, MI may not reduce bias because informative attrition is non-ignorable missing. Therefore, PMM was applied to reduce the bias, usually unknown, by adjusting the values imputed with MI by a fixed value equal to the introduced bias.

RESULTS

With highly correlated repeated outcome measures, the slope coefficients from a mixed effects model were found to have least bias when fitted to available cases. However, for moderately correlated outcome measurements, the slope coefficients from fitting a mixed effects model to data adjusted using PMM were least biased but still underestimated the true coefficients.

CONCLUSIONS

PMM may potentially reduce bias in studies analysing longitudinal data with suspected informative attrition and moderately correlated repeated outcome measurements. Including additional auxiliary variables in the imputation model may also reduce any remaining bias.

Depressive Symptoms, Antidepressant Use, and Hypertension in Young Adulthood

BACKGROUND

Among adults, depressive symptoms are associated with higher rates of cardiovascular disease; however, the evidence is mixed regarding the association between depressive symptoms and hypertension, especially among young adults. The deleterious effects of some antidepressant medications on blood pressure may contribute to mixed findings.

METHODS

Adolescents enrolled in Add Health (N = 11,183) (1994-2008) completed an abbreviated Center for Epidemiologic Studies Depression Scale at three waves (mean ages, 16, 22, and 29). Antidepressant use was measured at age 22 and at age 29. Hypertension at age 29 was defined as measured systolic blood pressure of 140 mm Hg or greater, diastolic blood pressure of 90 mm Hg or greater, or staff-inventoried anti-hypertensive medication use.

RESULTS

The prevalence of hypertension at age 29 was 20%. High depressive symptoms in adolescence or young adulthood were not associated with hypertension in young adulthood. Antidepressant use at age 29 was associated with increased prevalence of hypertension (prevalence ratio [PR], 1.4; 95% CI, 1.2, 1.7) and an interaction with sex was observed (PRMen, 1.6; 95% CI, 1.2, 2.0; PRWomen, 1.2; 95% CI, 0.89, 1.6; pinteraction = 0.0227). Selective serotonin reuptake inhibitor and non-selective serotonin reuptake inhibitor antidepressant use were associated with hypertension (PRSSRI, 1.3; 95% CI, 1.0, 1.6; PRnon-SSRI, 1.6; 95% CI, 1.2, 2.1).

CONCLUSIONS

In this sample, antidepressant use, but not depressive symptoms, was associated with hypertension in young adulthood. Further research is recommended to examine joint and independent relationships between depression and antidepressant use and hypertension among young adults. See video abstract at, http://links.lww.com/EDE/B355.

Diet and anthropometry at 2 years of age following an oral health promotion programme for Australian Aboriginal children and their carers: a randomised controlled trial

There are marked disparities between indigenous and non-indigenous children's diets and oral health. Both diet and oral health are linked to longer-term health problems. We aimed to investigate whether a culturally appropriate multi-faceted oral health promotion intervention reduced Aboriginal children's intake of sugars from discretionary foods at 2 years of age. We conducted a single-blind, parallel-arm randomised controlled trial involving women who were pregnant or had given birth to an Aboriginal child in the previous 6 weeks. The treatment group received anticipatory guidance, Motivational Interviewing, health and dental care for mothers during pregnancy and children at 6, 12 and 18 months. The control group received usual care. The key dietary outcome was the percent energy intake from sugars in discretionary foods (%EI), collected from up to three 24-h dietary recalls by trained research officers who were blind to intervention group. Secondary outcomes included intake of macronutrients, food groups, anthropometric z scores (weight, height, BMI and mid-upper arm circumference) and blood pressure. We enrolled 224 children to the treatment group and 230 to the control group. Intention-to-treat analyses showed that the %EI of sugars in discretionary foods was 1·6 % lower in the treatment group compared with control (95 % CI -3·4, 0·2). This culturally appropriate intervention at four time-points from pregnancy to 18 months resulted in small changes to 2-year-old Aboriginal children's diets, which was insufficient to warrant broader implementation of the intervention. Further consultation with Aboriginal communities is necessary for understanding how to improve the diet and diet-related health outcomes of young Aboriginal children.

Results of Database Studies in Spine Surgery Can Be Influenced by Missing Data

BACKGROUND

National databases are increasingly being used for research in spine surgery; however, one limitation of such databases that has received sparse mention is the frequency of missing data. Studies using these databases often do not emphasize the percentage of missing data for each variable used and do not specify how patients with missing data are incorporated into analyses. This study uses the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database to examine whether different treatments of missing data can influence the results of spine studies.

QUESTIONS/PURPOSES

(1) What is the frequency of missing data fields for demographics, medical comorbidities, preoperative laboratory values, operating room times, and length of stay recorded in ACS-NSQIP? (2) Using three common approaches to handling missing data, how frequently do those approaches agree in terms of finding particular variables to be associated with adverse events? (3) Do different approaches to handling missing data influence the outcomes and effect sizes of an analysis testing for an association with these variables with occurrence of adverse events?

METHODS

Patients who underwent spine surgery between 2005 and 2013 were identified from the ACS-NSQIP database. A total of 88,471 patients undergoing spine surgery were identified. The most common procedures were anterior cervical discectomy and fusion, lumbar decompression, and lumbar fusion. Demographics, comorbidities, and perioperative laboratory values were tabulated for each patient, and the percent of missing data was noted for each variable. These variables were tested for an association with "any adverse event" using three separate multivariate regressions that used the most common treatments for missing data. In the first regression, patients with any missing data were excluded. In the second regression, missing data were treated as a negative or "reference" value; for continuous variables, the mean of each variable's reference range was computed and imputed. In the third regression, any variables with > 10% rate of missing data were removed from the regression; among variables with ≤ 10% missing data, individual cases with missing values were excluded. The results of these regressions were compared to determine how the different treatments of missing data could affect the results of spine studies using the ACS-NSQIP database.

RESULTS

Of the 88,471 patients, as many as 4441 (5%) had missing elements among demographic data, 69,184 (72%) among comorbidities, 70,892 (80%) among preoperative laboratory values, and 56,551 (64%) among operating room times. Considering the three different treatments of missing data, we found different risk factors for adverse events. Of 44 risk factors found to be associated with adverse events in any analysis, only 15 (34%) of these risk factors were common among the three regressions. The second treatment of missing data (assuming "normal" value) found the most risk factors (40) to be associated with any adverse event, whereas the first treatment (deleting patients with missing data) found the fewest associations at 20. Among the risk factors associated with any adverse event, the 10 with the greatest effect size (odds ratio) by each regression were ranked. Of the 15 variables in the top 10 for any regression, six of these were common among all three lists.

CONCLUSIONS

Differing treatments of missing data can influence the results of spine studies using the ACS-NSQIP. The current study highlights the importance of considering how such missing data are handled.

CLINICAL RELEVANCE

Until there are better guidelines on the best approaches to handle missing data, investigators should report how missing data were handled to increase the quality and transparency of orthopaedic database research. Readers of large database studies should note whether handling of missing data was addressed and consider potential bias with high rates or unspecified or weak methods for handling missing data.

Parkinson's Disease Severity at 3 Years Can Be Predicted from Non-Motor Symptoms at Baseline

Objective

The aim of this study is to present a predictive model of Parkinson’s disease (PD) global severity, measured with the Clinical Impression of Severity Index for Parkinson’s Disease (CISI-PD).

Methods

This is an observational, longitudinal study with annual follow-up assessments over 3 years (four time points). A multilevel analysis and multiple imputation techniques were performed to generate a predictive model that estimates changes in the CISI-PD at 1, 2, and 3 years.

Results

The clinical state of patients (CISI-PD) significantly worsened in the 3-year follow-up. However, this change was of small magnitude (effect size: 0.44). The following baseline variables were significant predictors of the global severity change: baseline global severity of disease, levodopa equivalent dose, depression and anxiety symptoms, autonomic dysfunction, and cognitive state. The goodness-of-fit of the model was adequate, and the sensitive analysis showed that the data imputation method applied was suitable.

Conclusion

Disease progression depends more on the individual’s baseline characteristics than on the 3-year time period. Results may contribute to a better understanding of the evolution of PD including the non-motor manifestations of the disease.

Efficacy of an oral health literacy intervention among Indigenous Australian adults

OBJECTIVE

To determine the effect of an oral health literacy intervention on oral health literacy-related outcomes among rural-dwelling Indigenous Australian adults.

METHODS

A total of 400 Indigenous adults (203 intervention and 197 control participants) were recruited into a randomized controlled trial; a functional, context-specific oral health literacy interventions were developed and implemented by Indigenous staff. The intervention comprised five sessions, each lasting 1.5 hours, across a 1-year period. The primary outcome was oral health literacy as assessed by the HeLD-14 instrument, with secondary outcomes including the social impact of oral disease, and psychosocial and knowledge-related factors. Three scenarios were used in data analysis: (I) intention to treat; (II) as treated and; (III) adherence only. Multiple imputation (MI) was used to replace missing data.

RESULTS

The proportion reporting that "water with fluoride" was good increased in the intervention group within both crude and MI data analyses under the three scenarios. Other crude data analysis yielded no significant differences for either primary or secondary outcomes between intervention and control groups under the three scenarios. After MI, oral health literacy improved when assessed under scenario II (mean change=1.3, 95% CI: 1.1, 1.6). Improvements under three scenarios were also observed for the Oral Health Impact Profile (OHIP-14; mean change ranged from -0.7 to -3.8), sense of control (mean change ranged from 0.4 to 1.1), oral health-related fatalism (mean change ranged from -0.7 to -0.4) and perceived stress (mean change ranged from -2.1 to -1.1). The proportion reporting that "cordial was good" decreased in the intervention group from MI analysis under scenarios II and III.

CONCLUSIONS

A context-specific oral health literacy intervention was partially successful in improving oral health literacy and oral health literacy-related outcomes in this vulnerable population, but only after MI.

Multiple imputation for handling missing outcome data when estimating the relative risk

Background

Multiple imputation is a popular approach to handling missing data in medical research, yet little is known about its applicability for estimating the relative risk. Standard methods for imputing incomplete binary outcomes involve logistic regression or an assumption of multivariate normality, whereas relative risks are typically estimated using log binomial models. It is unclear whether misspecification of the imputation model in this setting could lead to biased parameter estimates.

Methods

Using simulated data, we evaluated the performance of multiple imputation for handling missing data prior to estimating adjusted relative risks from a correctly specified multivariable log binomial model. We considered an arbitrary pattern of missing data in both outcome and exposure variables, with missing data induced under missing at random mechanisms. Focusing on standard model-based methods of multiple imputation, missing data were imputed using multivariate normal imputation or fully conditional specification with a logistic imputation model for the outcome.

Results

Multivariate normal imputation performed poorly in the simulation study, consistently producing estimates of the relative risk that were biased towards the null. Despite outperforming multivariate normal imputation, fully conditional specification also produced somewhat biased estimates, with greater bias observed for higher outcome prevalences and larger relative risks. Deleting imputed outcomes from analysis datasets did not improve the performance of fully conditional specification.

Conclusions

Both multivariate normal imputation and fully conditional specification produced biased estimates of the relative risk, presumably since both use a misspecified imputation model. Based on simulation results, we recommend researchers use fully conditional specification rather than multivariate normal imputation and retain imputed outcomes in the analysis when estimating relative risks. However fully conditional specification is not without its shortcomings, and so further research is needed to identify optimal approaches for relative risk estimation within the multiple imputation framework.

Electronic supplementary material

The online version of this article (10.1186/s12874-017-0414-5) contains supplementary material, which is available to authorized users.

Factors associated with low birth weight in Nepal using multiple imputation

Background

Survey data from low income countries on birth weight usually pose a persistent problem. The studies conducted on birth weight have acknowledged missing data on birth weight, but they are not included in the analysis. Furthermore, other missing data presented on determinants of birth weight are not addressed. Thus, this study tries to identify determinants that are associated with low birth weight (LBW) using multiple imputation to handle missing data on birth weight and its determinants.

Methods

The child dataset from Nepal Demographic and Health Survey (NDHS), 2011 was utilized in this study. A total of 5,240 children were born between 2006 and 2011, out of which 87% had at least one measured variable missing and 21% had no recorded birth weight. All the analyses were carried out in R version 3.1.3. Transform-then impute method was applied to check for interaction between explanatory variables and imputed missing data. Survey package was applied to each imputed dataset to account for survey design and sampling method. Survey logistic regression was applied to identify the determinants associated with LBW.

Results

The prevalence of LBW was 15.4% after imputation. Women with the highest autonomy on their own health compared to those with health decisions involving husband or others (adjusted odds ratio (OR) 1.87, 95% confidence interval (95% CI) = 1.31, 2.67), and husband and women together (adjusted OR 1.57, 95% CI = 1.05, 2.35) were less likely to give birth to LBW infants. Mothers using highly polluting cooking fuels (adjusted OR 1.49, 95% CI = 1.03, 2.22) were more likely to give birth to LBW infants than mothers using non-polluting cooking fuels.

Conclusion

The findings of this study suggested that obtaining the prevalence of LBW from only the sample of measured birth weight and ignoring missing data results in underestimation.

Outcome-sensitive multiple imputation: a simulation study

Background

Multiple imputation is frequently used to deal with missing data in healthcare research. Although it is known that the outcome should be included in the imputation model when imputing missing covariate values, it is not known whether it should be imputed. Similarly no clear recommendations exist on: the utility of incorporating a secondary outcome, if available, in the imputation model; the level of protection offered when data are missing not-at-random; the implications of the dataset size and missingness levels.

Methods

We used realistic assumptions to generate thousands of datasets across a broad spectrum of contexts: three mechanisms of missingness (completely at random; at random; not at random); varying extents of missingness (20–80% missing data); and different sample sizes (1,000 or 10,000 cases). For each context we quantified the performance of a complete case analysis and seven multiple imputation methods which deleted cases with missing outcome before imputation, after imputation or not at all; included or did not include the outcome in the imputation models; and included or did not include a secondary outcome in the imputation models. Methods were compared on mean absolute error, bias, coverage and power over 1,000 datasets for each scenario.

Results

Overall, there was very little to separate multiple imputation methods which included the outcome in the imputation model. Even when missingness was quite extensive, all multiple imputation approaches performed well. Incorporating a secondary outcome, moderately correlated with the outcome of interest, made very little difference. The dataset size and the extent of missingness affected performance, as expected. Multiple imputation methods protected less well against missingness not at random, but did offer some protection.

Conclusions

As long as the outcome is included in the imputation model, there are very small performance differences between the possible multiple imputation approaches: no outcome imputation, imputation or imputation and deletion. All informative covariates, even with very high levels of missingness, should be included in the multiple imputation model. Multiple imputation offers some protection against a simple missing not at random mechanism.

Electronic supplementary material

The online version of this article (doi:10.1186/s12874-016-0281-5) contains supplementary material, which is available to authorized users.

External validation, update and development of prediction models for pre-eclampsia using an Individual Participant Data (IPD) meta-analysis: the International Prediction of Pregnancy Complication Network (IPPIC pre-eclampsia) protocol

BACKGROUND

Pre-eclampsia, a condition with raised blood pressure and proteinuria is associated with an increased risk of maternal and offspring mortality and morbidity. Early identification of mothers at risk is needed to target management.

METHODS/DESIGN

We aim to systematically review the existing literature to identify prediction models for pre-eclampsia. We have established the International Prediction of Pregnancy Complication Network (IPPIC), made up of 72 researchers from 21 countries who have carried out relevant primary studies or have access to existing registry databases, and collectively possess data from more than two million patients. We will use the individual participant data (IPD) from these studies to externally validate these existing prediction models and summarise model performance across studies using random-effects meta-analysis for any, late (after 34 weeks) and early (before 34 weeks) onset pre-eclampsia. If none of the models perform well, we will recalibrate (update), or develop and validate new prediction models using the IPD. We will assess the differential accuracy of the models in various settings and subgroups according to the risk status. We will also validate or develop prediction models based on clinical characteristics only; clinical and biochemical markers; clinical and ultrasound parameters; and clinical, biochemical and ultrasound tests.

DISCUSSION

Numerous systematic reviews with aggregate data meta-analysis have evaluated various risk factors separately or in combination for predicting pre-eclampsia, but these are affected by many limitations. Our large-scale collaborative IPD approach encourages consensus towards well developed, and validated prognostic models, rather than a number of competing non-validated ones. The large sample size from our IPD will also allow development and validation of multivariable prediction model for the relatively rare outcome of early onset pre-eclampsia.

TRIAL REGISTRATION

The project was registered on Prospero on the 27 November 2015 with ID: CRD42015029349.

Should multiple imputation be the method of choice for handling missing data in randomized trials?

The use of multiple imputation has increased markedly in recent years, and journal reviewers may expect to see multiple imputation used to handle missing data. However in randomized trials, where treatment group is always observed and independent of baseline covariates, other approaches may be preferable. Using data simulation we evaluated multiple imputation, performed both overall and separately by randomized group, across a range of commonly encountered scenarios. We considered both missing outcome and missing baseline data, with missing outcome data induced under missing at random mechanisms. Provided the analysis model was correctly specified, multiple imputation produced unbiased treatment effect estimates, but alternative unbiased approaches were often more efficient. When the analysis model overlooked an interaction effect involving randomized group, multiple imputation produced biased estimates of the average treatment effect when applied to missing outcome data, unless imputation was performed separately by randomized group. Based on these results, we conclude that multiple imputation should not be seen as the only acceptable way to handle missing data in randomized trials. In settings where multiple imputation is adopted, we recommend that imputation is carried out separately by randomized group.

Leukocyte Telomere Length in Relation to 17 Biomarkers of Cardiovascular Disease Risk: A Cross-Sectional Study of US Adults

BACKGROUND

Leukocyte telomere length (LTL) is a putative biological marker of immune system age, and there are demonstrated associations between LTL and cardiovascular disease. This may be due in part to the relationship of LTL with other biomarkers associated with cardiovascular disease risk. However, the strength of associations between LTL and adiposity, metabolic, proinflammatory, and cardiovascular biomarkers has not been systematically evaluated in a United States nationally representative population.

METHODS AND FINDINGS

We examined associations between LTL and 17 cardiovascular biomarkers, including lipoproteins, blood sugar, circulatory pressure, proinflammatory markers, kidney function, and adiposity measures, in adults ages 20 to 84 from the cross-sectional US nationally representative 1999-2002 National Health and Nutrition Examination Survey (NHANES) (n = 7,252), statistically adjusting for immune cell type distributions. We also examine whether these associations differed systematically by age, race/ethnicity, gender, education, and income. We found that a one unit difference in the following biomarkers were associated with kilobase pair differences in LTL: BMI -0.00478 (95% CI -0.00749--0.00206), waist circumference -0.00211 (95% CI -0.00325--0.000969), percentage of body fat -0.00516 (95% CI -0.00761--0.0027), high density lipoprotein (HDL) cholesterol 0.00179 (95% CI 0.000571-0.00301), triglycerides -0.000285 (95% CI -0.000555--0.0000158), pulse rate -0.00194 (95% CI -0.00317--0.000705), C-reactive protein -0.0363 (95% CI 0.0601--0.0124), cystatin C -0.0391 (95% CI -0.0772--0.00107). When using clinical cut-points we additionally found associations between LTL and insulin resistance -0.0412 (95% CI -0.0685--0.0139), systolic blood pressure 0.0455 (95% CI 0.00137-0.0897), and diastolic blood pressure -0.0674 (95% CI -0.126--0.00889). These associations were 10%-15% greater without controlling for leukocyte cell types. There were very few differences in the associations by age, race/ethnicity, gender, education, or income. Our findings are relevant to the relationships between these cardiovascular biomarkers in the general population but not to cardiovascular disease as a clinical outcome.

CONCLUSIONS

LTL is most strongly associated with adiposity, but is also associated with biomarkers across several physiological systems. LTL may thus be a predictor of cardiovascular disease through its association with multiple risk factors that are physiologically correlated with risk for development of cardiovascular disease. Our results are consistent with LTL being a biomarker of cardiovascular aging through established physiological mechanisms.

Relative efficiency of joint-model and full-conditional-specification multiple imputation when conditional models are compatible: The general location model

Estimating the parameters of a regression model of interest is complicated by missing data on the variables in that model. Multiple imputation is commonly used to handle these missing data. Joint model multiple imputation and full-conditional specification multiple imputation are known to yield imputed data with the same asymptotic distribution when the conditional models of full-conditional specification are compatible with that joint model. We show that this asymptotic equivalence of imputation distributions does not imply that joint model multiple imputation and full-conditional specification multiple imputation will also yield asymptotically equally efficient inference about the parameters of the model of interest, nor that they will be equally robust to misspecification of the joint model. When the conditional models used by full-conditional specification multiple imputation are linear, logistic and multinomial regressions, these are compatible with a restricted general location joint model. We show that multiple imputation using the restricted general location joint model can be substantially more asymptotically efficient than full-conditional specification multiple imputation, but this typically requires very strong associations between variables. When associations are weaker, the efficiency gain is small. Moreover, full-conditional specification multiple imputation is shown to be potentially much more robust than joint model multiple imputation using the restricted general location model to mispecification of that model when there is substantial missingness in the outcome variable.