Caregiver-reported newborn term and preterm motor abilities: psychometrics of the PediaTracTM Motor domain.

BACKGROUND: Approximately 5-10% of children exhibit developmental deviations in motor skills or other domains; however, physicians detect less than one-third of these abnormalities. Systematic tracking and early identification of motor deviations are fundamental for timely intervention. METHODS: Term and preterm neonates were prospectively assessed at the newborn (NB) period in a study of the psychometric properties of the Motor (MOT) domain of PediaTracTM v3.0, a novel caregiver-based development tracking instrument. Item response theory graded response modeling was used to model item parameters and estimate theta, an index of the latent trait, motor ability. Exploratory factor analysis (EFA) was conducted to examine the dimensionality and factor structure. RESULTS: In a cohort of 571 caregiver/infant dyads (331 term, 240 preterm), NB MOT domain reliability was high (rho = 0.94). Item discrimination and item difficulty of each of the 15 items could be reliably modeled across the range of motor ability. EFA confirmed that the items constituted a single dimension with second-order factors, accounting for 43.20% of variance. CONCLUSIONS: The latent trait, motor ability, could be reliably estimated at the NB period. IMPACT: The caregiver-reported Motor domain of PediaTrac provides a reliable estimate of the latent trait of motor ability during the newborn period. This is the first known caregiver-reported instrument that can assess motor ability in the newborn period with high reliability in term and preterm infants. Item response theory methods were employed that will allow for future characterization of developmental subgroups and motor trajectories. The PediaTrac Motor domain can support early identification of at-risk infants. Including caregivers in digital reporting and child-centered monitoring of motor functioning may improve access to care.

Longitudinal caregiver-reported motor development in infants born at term and preterm.

AIM: To examine the extent to which estimates of a latent trait or underlying construct of motor ability differ in infants born at term and preterm, based on caregiver ratings of the motor domain of PediaTrac v3.0. METHOD: The sample consisted of 571 caregiver-infant dyads (331 born at term, 240 born preterm), 48% female, with 51.7% of caregivers identifying as an ethnic minority. Latent trait of motor ability was estimated based on item response theory modeling. Gestational group differences (term and preterm birth) were examined at the newborn/term-equivalent, 2-, 4-, 6-, 9-, and 12-month time points. RESULTS: Caregiver ratings of latent trait of motor ability were reliably modeled across the range of abilities at each time point. While the group born preterm exhibited significantly more advanced motor abilities at the term-equivalent time point, by 6 months the group born at term was more advanced. Biological sex difference main and interaction effects were not significant. INTERPRETATION: Caregivers provided reliable, longitudinal estimates of motor ability in infancy, reflecting important differences in the motor development of infants born at term and preterm. The findings suggest that significant motor development occurs in infants born preterm from birth to the term-equivalent time point and provide a foundation to examine motor growth trajectories as potential predictors in the early identification of neurodevelopmental conditions and needs.

Family functioning within the context of families with adolescent children in urban India.

Research on family functioning within given cultural contexts is needed. This study aims to describe salient dimensions of family functioning in two urban contexts in India and to examine differences in family functioning by sociodemographic groups. We measured differences in family functioning using cross-sectional survey questionnaire data collected from 13 to 15-year-old adolescents and one of their parents/primary caregivers in Mumbai (n = 843) and Kolkata (n = 913) during 2019-2020. We drew a multi-stage sample representative of neighborhoods and households in both cities. We assessed a multi-dimensional family functioning latent factor that included parent-reported measures (parent-adolescent communication, family cohesion, and parent monitoring of peers) and adolescent-reported measures (parent support, family cohesion, and parent supervision). Our results support an overall measure of family functioning manifested by multiple dimensions for parent- and adolescent-reported data. Families with male adolescents had worse adolescent-reported family functioning in Mumbai and parent-reported family functioning in Kolkata. Higher socioeconomic status was associated with better parent-reported family functioning in both cities and better adolescent-reported family functioning in Kolkata. Muslim religious identification in Kolkata and the Hindi native language in both cities were associated with better adolescent-reported family functioning. Our findings indicate heterogeneity in family functioning across demographic and social-cultural groups within the two urban contexts of India. This study may inform the development of culturally congruent prevention interventions for families with adolescents in India.

Neighbourhood tobacco retail access and tobacco use susceptibility in young adolescents in urban India.

BACKGROUND: Neighbourhood tobacco retail access may influence adolescent tobacco use. In India, we examined the association between neighbourhood tobacco retail access and cognitive risks for tobacco use during early adolescence. METHODS: In 2019-2020, a population-based sample (n=1759) of adolescents aged 13-15 years was surveyed from 52 neighbourhoods in Mumbai and Kolkata. Neighbourhood tobacco retail access was measured as the frequency of visits to tobacco retailers, mapped tobacco retailer density and perceived tobacco retailer density. We estimated associations between neighbourhood tobacco retail access and cognitive risks for tobacco use (perceived ease of access to tobacco, perceived peer tobacco use and intention to use tobacco). RESULTS: There was high neighbourhood tobacco retail access. Tobacco retailer density was higher in lower income neighbourhoods (p<0.001). Adolescent frequency of tobacco retailer visits was positively associated with cognitive tobacco use risks. Mapped tobacco retailer density was associated with perceived ease of access in Kolkata but not in Mumbai, and it was not associated with perceived peer tobacco use nor intention. Perceived tobacco retailer density was associated with perceived ease of access and perceived peer use, but not with intention. In Kolkata, higher perceived retailer density and frequency of tobacco retailer visits were negatively associated with perceived ease of access. CONCLUSIONS: Efforts to reduce neighbourhood tobacco retail access in India may reduce cognitive tobacco use risk factors in young adolescents. The frequency of tobacco retailer visits and perceived tobacco retailer density increased cognitive risks, though there were some exceptions in Kolkata that further research may explain.

A Satellite Account for Health in the United States.

This paper develops a satellite account for the US health sector and measures productivity growth in health care for the elderly population between 1999 and 2012. We measure the change in medical spending and health outcomes for a comprehensive set of 80 conditions. Medical care has positive productivity growth over the time period, with aggregate productivity growth of 1.5% per year. However, there is significant heterogeneity in productivity growth. Care for cardiovascular disease has had very high productivity growth. In contrast, care for people with musculoskeletal conditions has been costly but has not led to improved outcomes.

Incomplete Response of Gastroesophageal Reflux Symptoms Poorly Predicts Erosive Esophagitis or Barrett's Esophagus.

BACKGROUND & AIMS: Many patients with symptoms of gastroesophageal reflux disease (GERD) not responding to a proton pump inhibitor (PPI) undergo an upper endoscopy. We hypothesized that an incomplete response to a PPI is not associated with findings of esophageal pathology on endoscopy, and that psychological distress is associated inversely with pathology. METHODS: We enrolled consecutive individuals aged 40 to 79 years with prior heartburn or regurgitation. Logistic regression was used to estimate the effects of incomplete response of GERD symptoms and psychological distress on the odds of finding erosive esophagitis or Barrett's esophagus. RESULTS: A total of 625 patients completed the study. A total of 254 (66.8% of those taking PPI) had GERD symptoms incompletely responsive to a PPI and 352 (56.3%) had severe somatization. An incomplete response to a PPI was associated with psychological distress (P < .001). Erosive esophagitis was found in 148 subjects (23.7%) and Barrett's esophagus in 58 (9.3%). Overall, an incomplete response to a PPI was not found to be associated with these pathologies (odds ratio, 1.17; 95% CI, 0.720-1.91). In contrast, greater psychological distress was associated inversely with erosive esophagitis or Barrett's esophagus (in particular, highest vs lowest tertile somatization: odds ratio, 0.590; 95% CI, 0.365-0.952). CONCLUSIONS: Patients undergoing upper endoscopy frequently have GERD symptoms incompletely responding to a PPI and a high burden of somatization. However, an incomplete response of GERD symptoms is a poor predictor for endoscopic pathology, and should not be relied upon for selecting patients for screening for Barrett's esophagus. Patients with high psychological distress are less likely to have esophageal pathology.

Validation and Comparison of Tools for Selecting Individuals to Screen for Barrett's Esophagus and Early Neoplasia.

BACKGROUND & AIMS: Guidelines suggest endoscopic screening of individuals who are at increased risk for Barrett's esophagus (BE) and esophageal adenocarcinoma. Tools based on clinical factors are available for identifying patients at risk, but only some have been validated. We aimed to compare and validate available tools. METHODS: We performed a prospective study of 1241 patients, ages 40 to 79 years, presenting either for their first esophagogastroduodenoscopy (EGD) or their first endoscopic therapy of early neoplastic BE, from April 2015 through June 2018. We calculated risk scores for 6 previously published tools (the Gerson, Locke, Thrift, Michigan BE pREdiction Tool [M-BERET], Nord-Trøndelag Health Study [HUNT], and Kunzmann tools). We also investigated the accuracy of frequency and duration of gastroesophageal reflux disease (GERD), using data from a randomly selected 50% of patients undergoing their first EGD. We compared the ability of all these tools to discriminate patients with BE and early neoplasia from patients without BE, using findings from endoscopy as the reference standard. RESULTS: BE was detected in 81 of 1152 patients during their first EGD (7.0%). GERD symptoms alone identified patients with BE with an area under the receiver operating characteristic curve (AuROC) of 0.579. All of the tools were more accurate in identifying patients with BE than the frequency and duration of GERD (AuROC for GERD, 0.579 vs range for other tools, 0.660-0.695), and predicted risk correlated well with observed risk (calibration). The AUROCs of the HUNT tool (0.796), the M-BERET (0.773), and the Kunzmann tool (0.763) were comparable in discriminating between patients with early neoplasia (n = 94) vs no BE. Each tool was more accurate in discriminating BE with early neoplasia than GERD frequency and duration alone (AuROC, 0.667; P < .01). CONCLUSIONS: The HUNT, M-BERET, and Kunzmann tools identify patients with BE with AuROC values ranging from 0.665 to 0.695, and discriminate patients with early neoplasia from patients without BE with AuROC values ranging from 0.763 to 0.796. These tools are more accurate than frequency and duration of GERD in identifying individuals at risk for neoplastic BE.

Validation of Tools for Predicting Incident Adenocarcinoma of the Esophagus or Esophagogastric Junction.

INTRODUCTION: Guidelines suggest screening of individuals who are at increased risk of esophageal adenocarcinoma (EAC). Tools for identifying patients at risk of Barrett's esophagus have been validated. Here, we aimed to compare and validate the tools for the primary outcomes of interest: EAC and esophagogastric junction adenocarcinoma (EGJAC). METHODS: Retrospective longitudinal analysis of the Kaiser Permanente Northern California Multiphasic Health Checkup Cohort, a community-based cohort including 206,974 patients enrolled between 1964 and 1973 followed through 2016. Baseline questionnaires and anthropometrics classified predictor variables for each tool and were linked to cancer registry outcomes. Analyses used logistic regression, Cox proportional hazards regression, and Kaplan-Meier survival curves. RESULTS: We identified 168 incident EAC cases and 151 EGJAC cases at a mean of 32 years after enrollment (mean follow-up among controls 26 years). Gastroesophageal reflux disease (GERD) symptoms predicted incident EAC (hazard ratio 2.66; 95% confidence interval 1.01, 7.00), but not EGJAC. The Nord-Trøndelag Health Study tool, Kunzmann tool, and Michigan Barrett's Esophagus pREdiction Tool were more accurate than GERD for predicting EAC, with individuals in the fourth quartile of Kunzmann having 17-fold the risk of those in the 1st quartile (hazard ratio = 16.7, 95% confidence interval = 4.72, 58.8). Each tool also predicted incident EGJAC with smaller magnitudes of effect. DISCUSSION: The study independently validated 4 tools for predicting incident EAC and EGJAC in a large community-based population. The Kunzmann tool appears best calibrated; all appear preferable to using GERD alone for risk stratification. Future studies should determine how best to implement such tools into clinical practice.

Measurement error correction and sensitivity analysis in longitudinal dietary intervention studies using an external validation study.

In lifestyle intervention trials, where the goal is to change a participant's weight or modify their eating behavior, self-reported diet is a longitudinal outcome variable that is subject to measurement error. We propose a statistical framework for correcting for measurement error in longitudinal self-reported dietary data by combining intervention data with auxiliary data from an external biomarker validation study where both self-reported and recovery biomarkers of dietary intake are available. In this setting, dietary intake measured without error in the intervention trial is missing data and multiple imputation is used to fill in the missing measurements. Since most validation studies are cross-sectional, they do not contain information on whether the nature of the measurement error changes over time or differs between treatment and control groups. We use sensitivity analyses to address the influence of these unverifiable assumptions involving the measurement error process and how they affect inferences regarding the effect of treatment. We apply our methods to self-reported sodium intake from the PREMIER study, a multi-component lifestyle intervention trial.

Small Area Estimation of Cancer Risk Factors and Screening Behaviors in US Counties by Combining Two Large National Health Surveys.

BACKGROUND: National health surveys, such as the National Health Interview Survey (NHIS) and the Behavioral Risk Factor Surveillance System (BRFSS), collect data on cancer screening and smoking-related measures in the US noninstitutionalized population. These surveys are designed to produce reliable estimates at the national and state levels. However, county-level data are often needed for cancer surveillance and related research. METHODS: To use the large sample sizes of BRFSS and the high response rates and better coverage of NHIS, we applied multilevel models that combined information from both surveys. We also used relevant sources such as census and administrative records. By using these methods, we generated estimates for several cancer risk factors and screening behaviors that are more precise than design-based estimates. RESULTS: We produced reliable, modeled estimates for 11 outcomes related to smoking and to screening for female breast cancer, cervical cancer, and colorectal cancer. The estimates were produced for 3,112 counties in the United States for the data period from 2008 through 2010. CONCLUSION: The modeled estimates corrected for potential noncoverage bias and nonresponse bias in the BRFSS and reduced the variability in NHIS estimates that is attributable to small sample size. The small area estimates produced in this study can serve as a useful resource to the cancer surveillance community.

FPCA-based method to select optimal sampling schedules that capture between-subject variability in longitudinal studies.

A critical component of longitudinal study design involves determining the sampling schedule. Criteria for optimal design often focus on accurate estimation of the mean profile, although capturing the between-subject variance of the longitudinal process is also important since variance patterns may be associated with covariates of interest or predict future outcomes. Existing design approaches have limited applicability when one wishes to optimize sampling schedules to capture between-individual variability. We propose an approach to derive optimal sampling schedules based on functional principal component analysis (FPCA), which separately characterizes the mean and the variability of longitudinal profiles and leads to a parsimonious representation of the temporal pattern of the variability. Simulation studies show that the new design approach performs equally well compared to an existing approach based on parametric mixed model (PMM) when a PMM is adequate for the data, and outperforms the PMM-based approach otherwise. We use the methods to design studies aiming to characterize daily salivary cortisol profiles and identify the optimal days within the menstrual cycle when urinary progesterone should be measured.

Longitudinal study of adolescent tobacco use and tobacco control policies in India.

BACKGROUND: This project will use a multilevel longitudinal cohort study design to assess whether changes in Community Tobacco Environmental (CTE) factors, measured as community compliance with tobacco control policies and community density of tobacco vendors and tobacco advertisements, are associated with adolescent tobacco use in urban India. India's tobacco control policies regulate secondhand smoke exposure, access to tobacco products and exposure to tobacco marketing. Research data about the association between community level compliance with tobacco control policies and youth tobacco use are largely unavailable, and are needed to inform policy enforcement, implementation and development. METHODS: The geographic scope will include Mumbai and Kolkata, India. The study protocol calls for an annual comprehensive longitudinal population-based tobacco use risk and protective factors survey in a cohort of 1820 adolescents ages 12-14 years (and their parent) from baseline (Wave 1) to 36-month follow-up (Wave 4). Geographic Information Systems data collection will be used to map tobacco vendors, tobacco advertisements, availability of e-cigarettes, COTPA defined public places, and compliance with tobacco sale, point-of-sale and smoke-free laws. Finally, we will estimate the longitudinal associations between CTE factors and adolescent tobacco use, and assess whether the associations are moderated by family level factors, and mediated by individual level factors. DISCUSSION: India experiences a high burden of disease and mortality from tobacco use. To address this burden, significant long-term prevention and control activities need to include the joint impact of policy, community and family factors on adolescent tobacco use onset. The findings from this study can be used to guide the development and implementation of future tobacco control policy designed to minimize adolescent tobacco use.

Graphical and numerical diagnostic tools to assess suitability of multiple imputations and imputation models.

Multiple imputation has become a popular approach for analyzing incomplete data. Many software packages are available to multiply impute the missing values and to analyze the resulting completed data sets. However, diagnostic tools to check the validity of the imputations are limited, and the majority of the currently available methods need considerable knowledge of the imputation model. In many practical settings, however, the imputer and the analyst may be different individuals or from different organizations, and the analyst model may or may not be congenial to the model used by the imputer. This article develops and evaluates a set of graphical and numerical diagnostic tools for two practical purposes: (i) for an analyst to determine whether the imputations are reasonable under his/her model assumptions without actually knowing the imputation model assumptions; and (ii) for an imputer to fine tune the imputation model by checking the key characteristics of the observed and imputed values. The tools are based on the numerical and graphical comparisons of the distributions of the observed and imputed values conditional on the propensity of response. The methodology is illustrated using simulated data sets created under a variety of scenarios. The examples focus on continuous and binary variables, but the principles can be used to extend methods for other types of variables. Copyright © 2016 John Wiley & Sons, Ltd.

Prevalence of groups A and C rotavirus antibodies in infants with biliary atresia and cholestatic controls.

OBJECTIVE: To analyze the prevalence of acute asymptomatic group A and C rotavirus (RV-A and RV-C) infection in neonates with cholestasis. STUDY DESIGN: Participants were infants <180 days of age with cholestasis (serum direct or conjugated bilirubin >20% of total and ≥2 mg/dL) enrolled in the Childhood Liver Disease Research and Education Network during RV season (December-May). Forty infants with biliary atresia (BA), age 62 ± 29 days (range, 4.7-13 weeks) and 38 infants with cholestasis, age 67 ± 44 days (range, 3-15.8 weeks) were enrolled. RESULTS: At enrollment, RV-A IgM positivity rates did not differ between infants with BA (10%) vs those without (18%) (P = .349). RV-C IgM was positive in 0% of infants with BA vs 3% in those without BA (P = .49). RV-A IgG was lower in infants with BA: 51 ± 39 vs 56 ± 44 enzyme-linked immunoassay unit, P = .045 but this difference may lack biological relevance as maternal RV-A IgG titers were similar between groups. Infant RV-A IgM titers at 2-6 months follow-up increased markedly vs at presentation in both infants with BA (50 ± 30 vs 9 ± 9) and those without (43 ± 18 vs 16 ± 20 enzyme-linked immunoassay unit) (P < .0001), without differences between groups. CONCLUSIONS: RV-A infection in the first 6 months of life is common in infants with cholestasis of any cause. RV-A could have different pathogenetic effects by initiating different hepatic immune responses in infants with vs without BA or could lack pathogenetic significance.

Bayesian sensitivity analysis of incomplete data: bridging pattern-mixture and selection models.

Pattern-mixture models (PMM) and selection models (SM) are alternative approaches for statistical analysis when faced with incomplete data and a nonignorable missing-data mechanism. Both models make empirically unverifiable assumptions and need additional constraints to identify the parameters. Here, we first introduce intuitive parameterizations to identify PMM for different types of outcome with distribution in the exponential family; then we translate these to their equivalent SM approach. This provides a unified framework for performing sensitivity analysis under either setting. These new parameterizations are transparent, easy-to-use, and provide dual interpretation from both the PMM and SM perspectives. A Bayesian approach is used to perform sensitivity analysis, deriving inferences using informative prior distributions on the sensitivity parameters. These models can be fitted using software that implements Gibbs sampling.

Total serum bilirubin predicts fat-soluble vitamin deficiency better than serum bile acids in infants with biliary atresia.

OBJECTIVE: Fat-soluble vitamin (FSV) deficiency is a well-recognized consequence of cholestatic liver disease and reduced intestinal intraluminal bile acid. We hypothesized that serum bile acid (SBA) would predict biochemical FSV deficiency better than serum total bilirubin (TB) level in infants with biliary atresia. METHODS: Infants enrolled in the Trial of Corticosteroid Therapy in Infants with Biliary Atresia after hepatoportoenterostomy were the subjects of this investigation. Infants received standardized FSV supplementation and monitoring of TB, SBA, and vitamin levels at 1, 3, and 6 months. A logistic regression model was used with the binary indicator variable insufficient/sufficient as the outcome variable. Linear and nonparametric correlations were made between specific vitamin measurement levels and either TB or SBA. RESULTS: The degree of correlation for any particular vitamin at a specific time point was higher with TB than with SBA (higher for TB in 31 circumstances vs 3 circumstances for SBA). Receiver operating characteristic curve shows that TB performed better than SBA (area under the curve 0.998 vs 0.821). Including both TB and SBA did not perform better than TB alone (area under the curve 0.998). CONCLUSIONS: We found that TB was a better predictor of FSV deficiency than SBA in infants with biliary atresia. The role of SBA as a surrogate marker of FSV deficiency in other cholestatic liver diseases, such as progressive familial intrahepatic cholestasis, α-1-antitrypsin deficiency, and Alagille syndrome in which the pathophysiology is dominated by intrahepatic cholestasis, warrants further study.

Convergent Validity Between the Motor Domain of PediaTracTM and Ages and Stages in Term and Preterm Infants at 2, 4, 6, and 9 Months of Age.

The primary aim of this study was to evaluate the convergent validity of the Motor domain (MOT) of PediaTracTM v3.0, an online developmental tracking instrument based on caregiver reports, with fine and gross motor domains (ASQ-FM and ASQ-GM) of the Ages and Stages Questionnaire (ASQ-3) in infants between 2- and 9 months of age. Participants were caregivers of 571 infants born term or preterm (gestational age <37 weeks) enrolled in a multi-site psychometric study of PediaTracTM. Findings revealed significant correlations between MOT and ASQ-3 scores at 2, 4, 6, and 9 months across time periods, term-preterm status, and biological sex. A significantly higher percentage of infants born preterm, compared with those born at term, was identified as a moderate or high risk on both the ASQ-3 and PediaTrac. Future investigations are warranted to further examine the psychometric properties of the MOT domain, including sensitivity, specificity, and positive and negative predictive value.

Measuring Early Relational Health Using PediaTrac TM in a Diverse Sample of Infant-Caregiver Dyads.

OBJECTIVE: Early relational health (ERH) is a key developmental predictor and outcome in infancy and early childhood that reflects social-emotional well-being and promotes resilience throughout childhood. Currently, there is no gold-standard developmental screening tool for ERH in pediatric care settings. This study examined the psychometric properties of items assessing ERH that are part of a web-based, caregiver-report screening tool called PediaTrac TM . It was hypothesized that ERH could be reliably estimated and that second-order factors would be revealed within the latent construct ERH. METHOD: Participants included 571 caregivers of term (n = 331; ≥37 weeks) and preterm (n = 240; <37 weeks) infants recruited shortly after birth from several academic medical centers and a community health clinic. Caregivers completed PediaTrac modules at birth and 2, 4, 6, 9, 12, 15, and 18 months; data for this study are from the newborn through 12-month periods. RESULTS: Results from Item Response Theory Graded Response Modeling revealed excellent reliability for the PediaTrac ERH domain at all time points, ranging from 0.96 to 0.98. Exploratory factor analyses revealed 4 to 5 second-order factors, representing Parent-Child Relationship, Parent Distress, Parenting Stress, Parenting Efficacy, Sensitivity, and Perceptions of Child, depending on period. CONCLUSION: The caregiver-report developmental screening tool, PediaTrac, reliably measures ERH during the first year of life. The measure has promising clinical utility in pediatric clinic settings for tracking ERH over time to ensure early social-emotional well-being and to identify concerns as early as possible.

A Bayesian model for time-to-event data with informative censoring.

Randomized trials with dropouts or censored data and discrete time-to-event type outcomes are frequently analyzed using the Kaplan-Meier or product limit (PL) estimation method. However, the PL method assumes that the censoring mechanism is noninformative and when this assumption is violated, the inferences may not be valid. We propose an expanded PL method using a Bayesian framework to incorporate informative censoring mechanism and perform sensitivity analysis on estimates of the cumulative incidence curves. The expanded method uses a model, which can be viewed as a pattern mixture model, where odds for having an event during the follow-up interval $$({t}_{k-1},{t}_{k}]$$, conditional on being at risk at $${t}_{k-1}$$, differ across the patterns of missing data. The sensitivity parameters relate the odds of an event, between subjects from a missing-data pattern with the observed subjects for each interval. The large number of the sensitivity parameters is reduced by considering them as random and assumed to follow a log-normal distribution with prespecified mean and variance. Then we vary the mean and variance to explore sensitivity of inferences. The missing at random (MAR) mechanism is a special case of the expanded model, thus allowing exploration of the sensitivity to inferences as departures from the inferences under the MAR assumption. The proposed approach is applied to data from the TRial Of Preventing HYpertension.

Multiple Imputation with Massive Data: An Application to the Panel Study of Income Dynamics.

Multiple imputation (MI) is a popular and well-established method for handling missing data in multivariate data sets, but its practicality for use in massive and complex data sets has been questioned. One such data set is the Panel Study of Income Dynamics (PSID), a longstanding and extensive survey of household income and wealth in the United States. Missing data for this survey are currently handled using traditional hot deck methods because of the simple implementation; however, the univariate hot deck results in large random wealth fluctuations. MI is effective but faced with operational challenges. We use a sequential regression/chained-equation approach, using the software IVEware, to multiply impute cross-sectional wealth data in the 2013 PSID, and compare analyses of the resulting imputed data with those from the current hot deck approach. Practical difficulties, such as non-normally distributed variables, skip patterns, categorical variables with many levels, and multicollinearity, are described together with our approaches to overcoming them. We evaluate the imputation quality and validity with internal diagnostics and external benchmarking data. MI produces improvements over the existing hot deck approach by helping preserve correlation structures, such as the associations between PSID wealth components and the relationships between the household net worth and sociodemographic factors, and facilitates completed data analyses with general purposes. MI incorporates highly predictive covariates into imputation models and increases efficiency. We recommend the practical implementation of MI and expect greater gains when the fraction of missing information is large.