Multiple Imputation Research Articles

Association of socioeconomic status with unintended pregnancy and induced abortion in married couples: An analysis of the 2021 national survey data in Japan

BackgroundAn association of the major indicators of socioeconomic status with unintended pregnancy and induced abortion in married couples has not been investigated in Japan, and we investigated it using recent nationwide survey data in Japan. MethodsData from the National Fertility Survey in 2021 in Japan, which evaluated the outcomes and intentions of previous pregnancies in married couples, were used. Number of unintended pregnancies, induced abortions, and unintended pregnancies ending in induced abortion were used as outcomes, and the number of unintended pregnancies ending in induced abortion was analyzed among couples who experienced an unintended pregnancy. Associations with current educational attainment, employment status, and household income were investigated using Poisson regression analysis. A complete-case analysis was conducted for dealing with missing data, as well as an analysis using multiple imputation. ResultsAverage number of unintended pregnancy and induced abortion per pregnancy of junior high school graduates was the highest among educational attainments of wives and husbands. The adjusted regression analysis showed that junior high school graduate in husbands was significantly associated with higher number of unintended pregnancies and induced abortions, regardless of the missing data analysis methods. In addition, high school or training school graduate in husbands was significantly associated with higher number of induced abortions. A significant association between socioeconomic status and the unintended pregnancy ending in induced abortion was not observed. ConclusionThis study showed that a lower educational attainment in husband was associated with higher number of unintended pregnancies and induced abortions in Japan.

Methods to Handle Missing depression (PHQ-8 Score) Values among Adults in the United Arab Emirates

Abstract Background The UAE Healthy Future Study (UAEHFS) is one of the first large prospective cohort studies in the Gulf region which examines causes and risk factors for chronic diseases among adult UAE nationals. Missing values are often unavoidable in empirical research and can in many cases, lead to bias. The aim of this study is to estimate the percentage of depression in the UAEHFS pilot data using the eight-item Patient Health Questionnaire (PHQ-8) variables, using different statistical methods. Methods Five common statistical machine learning methods of handling missing values were included in this analysis. These are mode imputation, k-nearest neighbor (KNN) imputation, classification, and regression trees (CART), random forest (RF) imputations, and random sample from observed values (Sample). 100 multiple imputations were used. Results 487 (94.2 %) eligible participants were included in the analysis. 231 (44.7%) were included in the complete case analysis. The median age was 30 years (Interquartile-Range: 23 - 38). More males (67.8%) than females included in the analysis. The estimated percentage of depression was 8.4%, 8.9%, 9.9%, 12.5%, 15.4% and 17.9% by the mode, complete case, sample, RF, CART, and KNN respectively. In additional analyses, the estimated proportions of depression were 11.5% by the Complete Case, 11.9% by KNN, 13.2% by K-means clustering, and 13.2% by Random Forest. Conclusions The estimated percentage of depression in the UAEHFS pilot data varies between the applied methods of handling missing values. This shows that the problem of missing values in the variables is not negligible. Further research is needed using multiple imputations in the main UAEHFS dataset after completing recruitment. Key messages • For the depression missing values, we recommend using multiple imputations not to generate data but to prevent the exclusion of observed data. • To have a better estimate of the percentage of depression, is recommended to use different machine learning methods.

Assessment of body composition in young adulthood and its associations to early changes in cardiovascular phenotypes: a cross-sectional study

Abstract Introduction Obesity in young adulthood is associated with various cardiovascular structural and functional changes, commonly used as markers of early subclinical cardiovascular disease. Whether these changes occur predominantly in relation to increased fat mass (FM) or are instead more closely related to simultaneous increases in underlying fat-free mass (FFM) that are known to occur in obesity, is incompletely understood. Purpose To investigate associations between both FM and FFM and a wide range of cardiac (left ventricular mass, LVM; end-diastolic volume, EDV; cardiac output, CO; stroke volume, SV), vascular (systolic blood pressure, SBP; diastolic blood pressure, DBP; mean arterial pressure, MAP; systemic vascular resistance, SVR) and autonomic (heart rate variability triangular index, HRVTI; baroreceptor reflex sensitivity, BRS) phenotypes commonly used to monitor the early evolution of subclinical CVD. Methods The study used data from extensive adipose and cardiovascular phenotyping of 408 healthy young adults (mean age 21 years) participating in a sub-study of the long-running ALSPAC birth cohort. We performed cross-sectional analysis using multiple imputation, multivariable linear regression and interaction testing to assess the association between both FM and FFM and cardiovascular phenotypes. Results Our study revealed that FFM resulted in a much greater increase in CO (difference in outcome per 1-SD difference in exposure β=0.58L/min, p<0.01, 95% CI=0.38, 0.77) than FM (β=0.20L/min, p=0.01, 95% CI=0.05, 0.36), following adjustment for both tissue types. FFM was also associated with an increase in LVM (β=19.1g, p<0.01, 95% CI=16.61, 21.58), EDV (β=22ml, p<0.01, 95% CI=18.18, 25.13) and SV (β=13ml, p<0.01, 95% CI=10.96, 15.75). FFM was also associated with improved autonomic and peripheral vascular function, including lower HR (β=-4bpm, p<0.01, 95% CI=-6.12, -1.95), higher HRVTI (β=2.1, p<0.01, 95% CI=0.72, 3.48) and lower SVR (β=-1.1, p<0.01, 95% CI=-2.12, -1.07), resulting in unchanged MAP despite elevations in CO. On the contrary, FM was not associated with LVM (β=-0.91g, p=0.35, 95% CI=-2.85, 1.02), and was associated with lower EDV (β=-4.62ml, p<0.01, 95% CI=-7.34, -1.88) and reduced autonomic function, as seen with HRVTI (β=-1.83, p<0.01, 95% CI=-3.02, -0.64) and BRS (β=-2.47, p=0.01, 95% CI=-4.27, -0.67). FM was also associated with increased HR (β=3.26bpm, p<0.01, 95% CI=1.57, 4.94) and increased SBP (β=2.14mmHg, p<0.01, 95% CI=1.01, 3.27) DBP (β=2.68mmHg, p<0.01, 95% CI=1.60, 3.77), and MAP (β=2.34mmHg, p<0.01, 95% CI=1.34, 3.34). Conclusion In young adults, changes in surrogate clinical markers of cardiac, vascular, and autonomic function commonly attributed solely to increases in fat mass should be interpreted with caution as they may be associated with underlying changes in FFM. Future work should focus on longitudinal studies of this cohort to confirm remodelling patterns based on these clinical markers.

Neighborhood environment and falls: a longitudinal Japan gerontological evaluation study

Abstract Background Older adults who experience falls are at a high risk of injury and death and require long-term care. Notably, the risk of falls varies among different neighborhood environments. No prior longitudinal studies have examined the association between falls and neighborhood environment at the community level. Identifying such associations will help prevent people falling by improving their neighborhood environment. Herein, we aimed to identify neighborhood environmental factors associated with falls among older adults in Japan. Methods We analyzed the data of 27,346 older adults living in 416 communities who participated in the 2010, 2013, and 2016 Japan Gerontological Evaluation Study surveys. The dependent variable was the occurrence of falls in 2016, and the explanatory variables were the mean scores of each of eight community-level neighborhood environment variables, categorized into tertiles (high, middle, and low groups) in 2013. Our model was adjusted for covariates (e.g., sex, age, and socioeconomic status) using the variables from 2010. A multilevel Poisson regression analysis was performed to obtain the relative risk (RR) and 95% confidence interval (CI) between the neighborhood environment and falls. Missing variable values were supplemented by multiple imputations. Results Older adults living in neighborhood environments with more parks or foot paths suitable for exercise or walking had a lower risk of falling after 3 years compared to those living in neighborhood environments with fewer parks or foot paths (RR: 0.89; CI: 0.80-0.99). For the variable ‘locations difficult for walking, such as hills or steps’, older adults in the middle group had a lower risk of falling than those in the low group (RR: 0.94; CI: 0.88-0.99). Conclusions Parks or foot paths are important environmental improvements that may prevent falls. Having hills or steps in the neighborhood could be beneficial when properly integrated into the environment. Key messages • Older adults with access to parks, footpaths, and challenging terrains suitable for exercise and walking, had a lower risk of falling. • These findings could be applied in urban designs to prevent falls among older adults.

BMI Trajectories and the Influence of Missing Data

Abstract Introduction Body Mass Index (BMI) trajectories have been estimated in various ways. These estimates are important to understand how BMI develops over time and for use in cost-effectiveness analysis. However, missing data is often stated as a limitation in studies that analyse BMI over time and there is little research into how missing data can influence these BMI trajectories. The aim of this study is to determine how much influence missing data can have when estimating BMI trajectories and to explore the effects this has on subsequent analysis. Methods This study uses data from the English Longitudinal Study of Aging. First, a growth mixture model is used to estimate distinct BMI trajectories in adults over the age of 50. Next, methods that assume data is missing at random (MAR) are used: complete case analysis and multiple imputation. Finally, Diggle Kenward and Roy methods that assume data is missing not at random (MNAR) are implemented. Estimated trajectories from each method are then used to predict the risk of developing type 2 diabetes (T2DM) using discrete-time survival analysis. Results Four distinct trajectories are identified using each of the methods to account for missing data: stable overweight, elevated BMI, increasing BMI, and decreasing BMI. However, the likelihoods of individuals following the different trajectories differs between the different methods. Results show that the influence of BMI trajectory on T2DM is reduced after accounting for missing data. More work is needed to understand which methods for missing data are most appropriate and give the most reliable results. Conclusions Missing data can significantly influence estimations of BMI trajectories. When using BMI trajectories to inform cost-effectiveness analysis or policymaking, missing data should be considered. More research is needed to examine the extent to which accounting for missing data might influence the cost-effectiveness of policies, e.g. weight management interventions. Key messages • Missing data is important when modelling BMI trajectories. • More research is needed to examine the extent to which accounting for missing data might influence the cost-effectiveness of policies, e.g. weight management interventions.

Right ventricular dysfunction parameters at echocardiography in patients with acute pulmonary embolism: meta-analysis of individual participant data

Abstract Background & Aim Echocardiography has a crucial role for decision making on management of patient with acute pulmonary embolism (PE) according to current international guidelines. However, the definition of right ventricle dysfunction (RVD) at echocardiography requires standardization. We performed an Individual patient data meta-analysis (IPDMA) to assess the role of individual parameters of RVD and of combinations of parameters for the prediction of short-term all-cause mortality and PE-related death. Methods A systematic search was run to identify clinical studies reporting on RVD at echocardiography and on short-term clinical outcome in patients with acute PE. Individual patient data from the original studies were pooled to perform a meta-analysis (IPDMA). The study outcomes were short-term death defined as death occurring within 30 days from PE diagnosis and PE-related death. Multiple imputation was used for the management of missing data. The odds ratio for all cause and PE related mortality were calculated using generalized linear mixed effects model. Results Individual data from 9,233 patients with acute PE from 16 studies were pooled in a unique database. All cause short-term mortality occurred in 9% of patients (95% CI 0.06-0.13, I2 93%). The IPDMA was feasible for the following parameters: tricuspid anulus systolic excursion (TAPSE), estimated pulmonary artery pressure (PAP) >30mmHg, right to left ventricle diameter ratio (RV/LV)>1, hypokinesia, paradoxical septal motion, RV diameter>30 mm and McConnel sign. All these parameters, except for PAP>30mmHg and RV diameter>30 mm, were predictors of short-term all cause death in the overall population and in hemodynamically patients (table). All the assessed RV findings, except for the McConnell’s sign, were predictors of PE-related death (Table). The presence of one RVD parameter only was not associated with increased risk of all cause death (OR 1.22, 95% CI 0.97-1-54), while the presence of two parameters or ≥ three parameters were (OR 1.63, 95% CI 1.24-2.15; OR 2.07 95% CI 1.59-2.69, respectively). Conclusion In patients with acute PE, the majority of RVD parameter at echocardiography are predictors of short-term death and of PE-related death. The higher the number of RVD parameters the worst the prognosis. These results claim for standardization in RVD definition to guide decision making for patient management.

Past trends and future projections of psychological distress among general population in Finland

Abstract Introduction The aim was to examine past and future trends of psychological distress in Finland from 2013 to 2040. Methods Six representative cross-sectional surveys from the years 2013, 2014, 2015, 2018, 2020, and 2022 (N = 172,547, aged 20+ y) were used. Multiple imputation was used in projecting psychological distress (MHI-5, with cut-off value ≤60). As the COVID-19 pandemic had a strong effect on psychological distress which persisted after the pandemic, projections were made with and without data from the years 2020 and 2022. Results The prevalence of psychological distress was slightly declining during the period 2013-2018 for both men and women, ending up around 18% for men and 19% for women. When including the years 2020 and 2022, however, trends showed a steep increase in the prevalence of psychological distress during 2018-2022. Among 20-39-year-olds, the prevalence of psychological distress increased 9.8%-points over the four-year period. Similarly, among 40-54-year-olds, an 8.2%-points increase was observed. In older age groups (55-64, 65-74 and 75 years and older) psychological distress first increased from 2018 to 2020 but then plateaued especially among the oldest. Projections based on the observed past trends during the period 2013-2018 showed a steady decline in psychological distress, the prevalence of psychological distress ending up around 11% for men and around 15% for women in 2040. When including the years 2020 and 2022, however, the projections suggested an increase in psychological distress in the general population and the prevalence of psychological distress is projected to end up around 37 % for both men and women in 2030. Conclusions The prevalence of psychological distress has increased substantially in the 2020s particularly in young adults. Mental health needs to be given a priority in all policies in order to change the future projections. Key messages • The upward trend of psychological distress did not reverse or plateau after COVID-19 pandemic in Finland. • Increased efforts to protect and promote mental health in all policies seem indicated especially among young adults.

GenePHIT phase 2 study design: a double blind, placebo-controlled trial to assess efficacy, safety, and tolerability of AB-1002 gene therapy in adults with heart failure

Abstract Introduction Heart failure (HF) is associated with substantial morbidity and mortality, highlighting a critical need for novel therapies. AB-1002, an engineered adeno-associated virus (AAV) vector gene therapy candidate, delivers a constitutively active form of protein phosphatase inhibitor (I-1c) to restore cardiomyocyte intracellular calcium homeostasis. Preliminary results from an ongoing Phase (Ph) 1 trial (NCT04179643) of AB-1002 showed favourable safety and tolerability. Here we describe the design of AB-1002 Ph 2 trial (GenePHIT; NCT05598333) to evaluate efficacy, safety, and tolerability. Methods This phase 2 adaptive, double-blind, placebo-controlled, randomised, multi-centre trial (currently United States; European sites planned) includes a 52-week observation and 4-year follow-up period. Eligible patients are ≥18 years of age with non-ischaemic cardiomyopathy and New York Health Association (NYHA) Class III heart failure symptoms (Figure). Patients will be randomised 1:1:1 (n=30–50/treatment arm), including a single dose of AB-1002 administered via antegrade intracoronary infusion at 1 of 2 different doses, or placebo, all in addition to standard of care. The primary endpoint will evaluate efficacy at 52 weeks based on the composite endpoint by modified win ratio, which includes cardiovascular-related death, change in NYHA classification, and for the responders, improvements in left ventricular ejection fraction (LVEF) (≥5%), peak oxygen uptake (pVO2) ≥1.5 mL/kg/min, and 6-min walk test (6MWT) >30 metres. Key secondary efficacy endpoints include functional status and hospitalisations at 24 and 52 weeks, by assessing changes over time in NYHA classification, LVEF, pVO2, 6MWT, biomarkers (N-terminal pro-hormone b-type natriuretic peptide and troponin), quality of life questionnaires (Minnesota Living with Heart Failure, Kansas City Cardiomyopathy), and physiological assessments. Key safety endpoints will be assessed through 52 weeks and include treatment-emergent adverse events (AEs) and serious AEs (SAEs). Exploratory endpoints will evaluate AAV2i8 antibody titres, T-cell response to AAV2i8 capsid and I-1c, and optional right ventricular biopsy testing. Long-term follow-up will assess treatment-emergent AEs and SAEs, and number of HF hospitalisations. Data safety monitoring board assessment will occur every 3 months. Primary efficacy endpoint will compare both active dose groups combined versus placebo using the joint rank test (1). Key secondary endpoints will be analysed using Analysis of Covariance with baseline as a continuous covariate. Multiple imputations will be used to address missing data. Interim efficacy analysis will be performed using a Bayesian adaptive sample size algorithm. Conclusions GenePHIT is currently recruiting and will evaluate the efficacy, safety, and tolerability of intracoronary infusion of AB-1002, an innovative gene therapy targeting ventricular dysfunction in HF.

Timing of inopressor initiation in patients with acute myocardial infarction complicated by cardiogenic shock

Abstract Introduction Inotropes and vasopressors (inopressors) are commonly used in the treatment of acute myocardial infarction complicated by cardiogenic shock (AMICS). The characteristics of patients subjected to early or late initiation of inopressors are not yet described, and the impact of timing on outcomes remains uncertain. Purpose This retrospective real-world data study aims to identify patient-, treatment-, and outcome characteristics of AMICS patients treated with early- or late initiated inopressors. Methods The Netherlands Heart Registration (NHR) prospectively records baseline, procedural, and outcome data for PCI patients. Between 2017-2021, additional retrospective data on CS patients were collected by fourteen Dutch hospitals. Patients were categorized based on the timing of inopressor initiation, either early (pre PCI) or solely late (at the cathlab or within 24 hours after PCI). Patients presenting with an out of hospital cardiac arrest were excluded. Missing data were imputed through multiple imputation (30x). Multivariable logistic regression analysis was performed to evaluate the association between the timing of inopressor initiation and 30-day mortality, adjusting for univariate significant confounders among variables known at admission. Results A total of 956 consecutive CS patients were included of whom 418 (43.7%) and 538 (57.3%) patients were subjected to early and late initiation of inopressors respectively. The most frequently used drug in both groups was noradrenaline. At baseline, patients in the early initiation group were younger (67.4 vs. 70.0 years, p <0.001), presented with a lower mean arterial pressure (MAP) (72.5 vs. 76.3, p = 0.048) and higher lactate (4.2 vs. 3.6, p = 0.012), and presented with a NSTEMI more often (36.0% vs. 14.2%, p <0.001). They also had a higher prevalence of previous myocardial infarction (33.9% vs. 22.9%, p <0.001) and previous CABG (14.3% vs. 6.2%, p <0.001). Moreover, 30-day mortality rates were higher in the early initiated group (48.6% vs. 37.6%, pLogrank <0.001). After multivariate adjustment for age, admission MAP and lactate, early inopressor initiation remained independently associated with 30-day mortality (OR 1.44, 95%-CI 1.08-1.98) compared to late initiation. Conclusion In this Dutch cohort of CS patients undergoing PCI, baseline- treatment- and outcome characteristics differed between patient categories receiving either early or late initiated inopressors. Patients who were subjected to early initiated inopressors were younger, presented with worse hemodynamics, had a more extensive cardiovascular medical history and higher 30-day mortality rates. The results of this study reemphasize the importance of identifying patients who benefit from inopressors.30-day mortality late vs. earlyTable patient characteristics

Respective impact of bleeding and ischemic complications of mortality after percutaneous coronary interventions: substudy from the FRANCE-PCI registry

Abstract Background Patients undergoing percutaneous coronary interventions (PCI) remain at high risk of both bleeding and ischemic events, both burdened with high morbimortality. The respective impact of both type of complications in contemporary PCI, however, remains unclear. Purpose To evaluate and compare the respective impact of ischemic and bleeding complications on mortality after PCI Methods The FRANCE PCI is an ongoing registry of consecutive French patients undergoing PCI and monitored for one year thereafter. Patients presenting with complications after the index PCI were categorized as ischemic complications in case of stent thrombosis, myocardial infarction, stroke or unplanned revascularization, as hemorrhagic complications in case of bleeding BARC ≥3 or as ischemic and hemorrhagic when both types of events occurred in the same patients. The impact of ischemic and hemorrhagic complications on mortality was evaluated separately for the index hospitalization and the 12-months follow-up after hospital discharge, using multivariate logistic model and Cox proportional hazards model, respectively. The multivariate Cox model was developed using complications as time-dependent variables and estimating the hazard ratios (HRs) over 3 periods of time (0 to 8 weeks, 8 to 24 weeks, and 24 to 52 weeks). Consequently, the proportional hazard assumption, tested using the Schoenfeld test, was met for each period. Multiple imputations were performed for missing data on covariates. Results A total of 54,599 patients, with a median age of 69 [60.0-78.0] years, 75.0% male, underwent PCI in 30 centers between 2014 and 2020, in an elective setting for 47% of the cases. During the index hospitalization, an ischemic and hemorrhagic complications occurred in 610 (1.1% 95%CI 1.0-1.2) and 531 (1.0% 95%CI 0.9-1.1) patients, respectively, while the mortality rate was 2.1% 95%CI 2.0-2.2. The association between ischemic, hemorrhagic and ischemic and hemorrhagic events with mortality was aOR 8.4 95%CI 6.4-10.9, aOR 10.3 95%CI 8.1-13.2 and aOR 13.2 95%CI 7.0-24.7, respectively. Among the 51,738 patients alive and not lost-to-follow-up after discharge, an ischemic and hemorrhagic complications occurred in 2,727 (5.3% 95%CI 5.1-5.5) and 1,022 (2.0% 95%CI 1.9-2.1) patients, respectively, while the mortality incidence was 3.8% 95%CI 3.6-4.0. The associations between ischemic, hemorrhagic and ischemic and hemorrhagic events with mortality according to the periods of time are detailed in Table 1 and Figure 1. Conclusion Both ischemic and hemorrhagic complications after PCI are associated with increased risk of mortality. The risk is particularly increased during the early period after the index PCI and diminishes over time, albeit faster after ischemic complications alone, compared to hemorrhagic events alone or the association of ischemic and hemorrhagic complications.

Long-term outcomes of percutaneous coronary intervention versus coronary artery bypass grafting in women with severe coronary artery disease

Abstract Background The optimal revascularization strategy for severe coronary artery disease (CAD) in women remains controversial. Despite differences between women and men in the development of CAD, there remains a paucity of comparative real-world observational evidence comparing percutaneous coronary intervention (PCI) to coronary artery bypass grafting (CABG) in women. Purpose We evaluated short and long-term outcomes of PCI and CABG in women with severe CAD. Methods We conducted a propensity score-matched retrospective cohort study using linked clinical and administrative databases in Ontario, Canada to obtain records of women with left main, 3-vessel disease, or 2-vessel disease with proximal left anterior descending artery involvement undergoing PCI or CABG from April 1st, 2012 to December 31st, 2021. Patients who were >80 years of age and patients who underwent urgent or emergent procedures were excluded for the primary analysis. The primary outcome was a composite of myocardial infarction, stroke, repeat revascularization, and death (MACCE). Secondary outcomes included the individual components of MACCE and a composite of readmission for MI, heart failure, and stroke. Following multiple imputation and propensity score-matching, the Cox proportional hazards model and Fine-Gray sub-distribution hazard model with robust variance estimators accounting for death as a competing risk were used to evaluate long-term outcomes. Results In the primary cohort, 2469 patients underwent PCI and 3,721 patients underwent CABG. After propensity score matching, including by coronary disease territory, 2,033 well-balanced pairs were identified, with standardized mean differences <0.10 for all covariates. In-hospital death and stroke were significantly lower with PCI than with CABG (0.7% vs 1.5%, p=0.036) and (<0.2% vs 1.4%, p<0.001), respectively, and there was no difference in perioperative MI between PCI and CABG (1.1% vs 0.6%, p=0.121). At a median (IQR) of 5.1 (2.9-7.5) years, major adverse cardiac and cerebrovascular events (MACCE) was higher with PCI compared with CABG (37.7% vs 23.3%, HR 1.79, [95% CI: 1.66-1.91], p<0.001) (Figure 1). All-cause mortality was higher with PCI compared with CABG (17.8% vs 13.4%, HR 1.33, [95% CI: 1.21-1.45], p<0.001) (Figure 2). The composite readmission outcome (MI, heart failure, stroke) was higher with PCI compared with CABG (16.2% vs 11.2%, HR 1.42, [95% CI: 1.37-1.48], p<0.001). Conclusion In women with severe CAD, CABG appears to be associated with a long-term reduction in MACCE and mortality compared with PCI. Our findings provide evidence to suggest women with severe CAD may derive greater long-term benefit from CABG compared to PCI and suggest that the treatment recommendation should shift towards CABG in appropriately selected women.Figure 1.MACCE free survivalFigure 2.Overall survival

Validation of digital cardiovascular risk score (DiCAVA) in the United States All of Us dataset

Abstract Introduction The cardiovascular disease (CVD) risk score DiCAVA was developed in 2021 using UK Biobank data with the aim to be used for population health screening by deriving its predictions only from variables readily available to every individual, omitting difficult-to-obtain measurements such as blood pressure and invasive blood tests for metrics like cholesterol or HbA1c, which are commonly used in clinical CVD risk assessment tools. The original set of variables was reduced to 30 in the DiCAVA Lite score to further improve usability in a digital solution accessible to the general population, while maintaining its discrimination ability. Purpose Like many countries, the US has a large burden of CVD and would benefit from easily accessible screening tools. However, to ensure generalisability of the UK-developed score, it is necessary to evaluate its performance in a population where the score is intended to be used. Methods Data from a large real-world modern observational study in the US, All of Us, was used to evaluate discrimination and calibration metrics of the risk score. Participants with pre-existing CVD and without available electronic health records were excluded from the study. Variables from the All of Us dataset were matched to the original variables and any missing values were imputed using multiple imputation. Results 216,985 participants with a median follow-up time of 3.44 years were included in the study. The median age was 51 years and 63% of the population were female. Discrimination ability in the All of Us dataset (C-index 0.76) slightly surpassed the discrimination achieved at training with UK Biobank (C-index 0.73), and that of Framingham risk score calculated for the All of Us population (C-index 0.72), demonstrating the score’s usability in the US. The score showed very good discrimination in both males (C-index 0.74) and females (C-index 0.77), all ethnic groups and age groups, with the exception of the population over 65 years of age, where the C-index dropped to 0.65 as expected (similar to that observed in the UK Biobank). Calibration of the original model was assessed at 3 years and showed under-estimation of the predicted risk (calibration-in-the-large of 2.37%). Following re-calibration, the algorithm demonstrated excellent results with the average predicted 3-year risk showing 0.59% difference from the observed CVD rate (calibration plot in Figure). Conclusions Overall, the DiCAVA Lite score has shown very good discrimination in the US, showing its generalisability in different demographic populations. Its re-calibrated form is now validated for use in the US population and because of its ability to be utilised outside clinical settings, it can be deployed widely in the general population as an initial screening tool and potentially boost awareness of cardiovascular health in an effort to drive preventative care of CVD.Re-calibration of the DiCAVA Lite score

Longitudinal association between nonstandard work schedule and depressive symptoms

Abstract Background Nonstandard work schedules (NSWSs) have been linked to increased risk of depression, but the evidence is mostly from cross-sectional studies and data on the risk distribution across various groups is lacking. We examined the impact of different NSWSs on depressive symptoms and assessed effect modification by sex, ethnicity, and immigrant status. Methods We used five waves of longitudinal data from the Understanding Society, the UK Household Longitudinal Study, which were collected biennially from 2010/11 to 2018/19. The study sample comprised 11371 employed workers aged 16 to 58 years in 2010/11. Generalised estimating equation with logistic regression was used to assess crude and adjusted effects of NSWS types (work time arrangement, work hours, and weekend working) on depressive symptoms. Independent correlation was specified in all the models. Complete-case and multiple imputation analyses were performed. Results After adjusting for confounders, only weekend work was associated with depressive symptoms. Workers who worked most/all (OR 1.24; 95%CI 1.13 - 1.37) or some (OR 1.10; 95%CI 1.02 - 1.18) weekends had a higher risk of depressive symptoms than workers who did not work on weekends. Ethnicity modified the association between work hours and depressive symptoms, indicating a higher risk among White workers working > 40 hours/week (OR 1.10; 95%CI 1.02 - 1.20) and Asian workers working <35 hours/week (OR 1.72; 95%CI 1.32 - 2.25) compared to White workers working 35-40 hours/week. No effect modification by sex or immigrant status was observed. The results were the same as those obtained from the imputation model. Conclusions Weekend work, long work hours (for White workers), or working < 35 hours/week (for Asian workers) were associated with increased risk of depressive symptoms. These findings indicate the need to review the working conditions of workers exposed to NSWSs, especially those working on weekends and Asian workers who work part-time. Key messages • Weekend work schedule is associated with an increased risk of depressive symptoms among employed workers in the UK. • The effect of work hours on depressive symptoms varies by ethnicity.

Impact of household income on acute decompensated heart failure outcomes: national representative sample

Abstract Background The clinical outcomes of ADHF have been extensively studied; however, the impact of household income still needs to be better understood. This study aimed to elucidate the relationship between household income and outcomes in patients with ADHF. Methods This national retrospective cohort study used data from the 2016-2020 National Inpatient Database, focusing on patients admitted with ADHF. We stratified the cohort into two groups based on the median household income: below $50k and above $86k. The selection criteria included patients aged ≥ 18 years diagnosed with ADHF, excluding those with incomplete income data. The primary outcome was all-cause inpatient mortality, with a length of stay (LOS) and total hospital charges as secondary outcomes. Statistical analysis employed logistic and linear regressions adjusted for confounders, with the significance level set at p<0.05. We accounted for missing data using multiple imputation analyses. Results 1,123,024 patients admitted with ADHF met the inclusion criteria. We divided the groups into two cohorts: low-income (64%, N=718,735) and high-income (36%, N=404,289). No increase in mortality was identified (adjusted aOR: 1.05; p = 0.212). The study revealed that high-income patients had marginally longer average LOS (5.62 days vs. 5.35 days, p < 0.0001) and higher hospital charges ($62,512.46 vs. $49,440.14, p < 0.0001) compared to their low-income counterparts. Low-income patients exhibited a higher prevalence of hypertension and obesity, whereas high-income individuals were more affected by atrial fibrillation and hyperlipidemia (hypertension and obesity, p < 0.0001; atrial fibrillation, p < 0.0001). Conclusion In this national retrospective cohort of patients admitted for ADHF over five years, income did not affect mortality. However, LOS and total charges were affected by the median household income.

Evaluating the median p-value method for assessing the statistical significance of tests when using multiple imputation

Rubin’s Rules are commonly used to pool the results of statistical analyses across imputed samples when using multiple imputation. Rubin’s Rules cannot be used when the result of an analysis in an imputed dataset is not a statistic and its associated standard error, but a test statistic (e.g. Student’s t-test). While complex methods have been proposed for pooling test statistics across imputed samples, these methods have not been implemented in many popular statistical software packages. The median p-value method has been proposed for pooling test statistics. The statistical significance level of the pooled test statistic is the median of the associated p-values across the imputed samples. We evaluated the performance of this method with nine statistical tests: Student’s t-test, Wilcoxon Rank Sum test, Analysis of Variance, Kruskal-Wallis test, the test of significance for Pearson’s and Spearman’s correlation coefficient, the Chi-squared test, the test of significance for a regression coefficient from a linear regression and from a logistic regression. For each test, the empirical type I error rate was higher than the advertised rate. The magnitude of inflation increased as the prevalence of missing data increased. The median p-value method should not be used to assess statistical significance across imputed datasets.

Multiple imputation integrated to machine learning: predicting post-stroke recovery of ambulation after intensive inpatient rehabilitation

Good data quality is vital for personalising plans in rehabilitation. Machine learning (ML) improves prognostics but integrating it with Multiple Imputation (MImp) for dealing missingness is an unexplored field. This work aims to provide post-stroke ambulation prognosis, integrating MImp with ML, and identify the prognostic influential factors. Stroke survivors in intensive rehabilitation were enrolled. Data on demographics, events, clinical, physiotherapy, and psycho-social assessment were collected. An independent ambulation at discharge, using the Functional Ambulation Category scale, was the outcome. After handling missingness using MImp, ML models were optimised, cross-validated, and tested. Interpretability techniques analysed predictor contributions. Pre-MImp, the dataset included 54.1% women, 79.2% ischaemic patients, median age 80.0 (interquartile range: 15.0). Post-MImp, 368 non-ambulatory patients on 10 imputed datasets were used for training, 80 for testing. The random forest (the validation best-performing algorithm) obtained 75.5% aggregated balanced accuracy on the test set. The main predictors included modified Barthel index, Fugl-Meyer assessment/motricity index, short physical performance battery, age, Charlson comorbidity index/cumulative illness rating scale, and trunk control test. This is among the first studies applying ML, together with MImp, to predict ambulation recovery in post-stroke rehabilitation. This pipeline reliably exploits the potential of incomplete datasets for healthcare prognosis, identifying relevant predictors.

Body mass index is associated with disability at discharge as indicated by the modified Rankin Scale in patients with ischemic stroke: a JROAD-DPC study

ABSTRACT Background Body mass index (BMI) of patients with ischemic stroke (IS) has been associated with prognosis and disability in studies in the United States. Although the Asian population is leaner, the optimal BMI for stroke-related disability remains unknown. Objectives To clarify the association between BMI and disability in patients with IS from a national database in Japan. Methods The present study included 522,421 patients with IS identified in the JROAD-DPC database from April 2016 to March 2020. We used the WHO classification of BMI, which divides Asia-Pacific patients into five groups, to categorize BMI and the modified Rankin Scale (mRS) to assess stroke-related disability at admission and discharge. After multiple imputation for missing values, we conducted a multiple mixed-effect logistic regression analysis for poor mRS score (>2) in September 2023. In addition, we created a restricted cubic spline model between the odds ratio (OR) for poor mRS and BMI. Results The mRS score worsened during hospitalization in 60.1% of the patients with IS, and 45.7% had a poor mRS score at discharge. Overweight patients had the lowest OR of having a poor mRS score (OR: 0.898, 95% confidence interval: 0.895–0.902). The spline curve for the OR for poor mRS score was U-shaped with a BMI of 24.7 kg/m2as the apex value. Conclusion The present study revealed a U-shaped relationship between BMI and stroke-related disability, with overweight patients having the lowest OR for disability at discharge.

Using data fusion with multiple imputation to correct for misclassification in self-reported exposure: a case-control study of cannabis use and homicide victimization

BackgroundCannabis use has been causally linked to violent behaviors in experimental and case studies, but its association with homicide victimization has not been rigorously assessed through epidemiologic research.MethodsWe performed a case-control analysis using two national data systems. Cases were homicide victims from the National Violent Death Reporting System (NVDRS), and controls were participants from the National Survey on Drug Use and Health (NSDUH). While the NVDRS contained toxicological testing data on cannabis use, the NSDUH only collected self-reported data, and thus the potential misclassification in the self-reported data needed to be corrected. We took a data fusion approach by concatenating the NSDUH with a third data system, the National Roadside Survey of Alcohol and Drug Use by Drivers (NRS), which collected toxicological testing and self-reported data on cannabis use for drivers. The data fusion approach provided multiple imputations (MIs) of toxicological testing results on cannabis use for the participants in the NSDUH, which were then used in the case-control analysis. Bootstrap was used to obtain valid statistical inference.ResultsThe analyses revealed that cannabis use was associated with 3.55-fold (95% CI: 2.75–4.35) increased odds of homicide victimization. Alcohol use, being Black, male, aged 21–34 years, and having less than a high school education were also significantly associated with increased odds of homicide victimization.ConclusionsCannabis use is a major risk factor for homicide victimization. The data fusion with MI method is useful in integrative data analysis for harmonizing measures between different data sources.

Imputation methods for mixed datasets in bioarchaeology

Missing data is a prevalent problem in bioarchaeological research and imputation could provide a promising solution. This work simulated missingness on a control dataset (481 samples × 41 variables) in order to explore imputation methods for mixed data (qualitative and quantitative data). The tested methods included Random Forest (RF), PCA/MCA, factorial analysis for mixed data (FAMD), hotdeck, predictive mean matching (PMM), random samples from observed values (RSOV), and a multi-method (MM) approach for the three missingness mechanisms (MCAR, MAR, and MNAR) at levels of 5%, 10%, 20%, 30%, and 40% missingness. This study also compared single imputation with an adapted multiple imputation method derived from the R package “mice”. The results showed that the adapted multiple imputation technique always outperformed single imputation for the same method. The best performing methods were most often RF and MM, and other commonly successful methods were PCA/MCA and PMM multiple imputation. Across all criteria, the amount of missingness was the most important parameter for imputation accuracy. While this study found that some imputation methods performed better than others for the control dataset, each imputation method has advantages and disadvantages. Imputation remains a promising solution for datasets containing missingness; however when making a decision it is essential to consider dataset structure and research goals.

Direct Estimation for Commonly Used Pattern-Mixture Models in Clinical Trials

Pattern-mixture models have received increasing attention as they are commonly used to assess treatment effects in primary or sensitivity analyses for clinical trials with nonignorable missing data. Pattern-mixture models have traditionally been implemented using multiple imputation, where the variance estimation may be a challenge because the Rubin’s approach of combining between- and within-imputation variance may not provide consistent variance estimation while bootstrap methods may be time-consuming. Direct likelihood-based approaches have been proposed in the literature and implemented for some pattern-mixture models, but the assumptions are sometimes restrictive and the theoretical framework is fragile. In this article, we propose an analytical framework for an efficient direct likelihood estimation method for commonly used pattern-mixture models corresponding to return-to-baseline, jump-to-reference, placebo washout, and retrieved dropout imputations. A parsimonious tipping point analysis is also discussed and implemented. Results from simulation studies demonstrate that the proposed methods provide consistent estimators. We further illustrate the utility of the proposed methods using data from a clinical trial evaluating a treatment for type 2 diabetes.