Understanding the transitions between disease states is often the goal in studying chronic disease. These studies, however, are typically subject to a large amount of missingness either due to patient dropout or intermittent missed visits. The missing data is often informative since missingness and dropout are usually related to either an individual's underlying disease process or the actual value of the missed observation. Our motivating example is a study of opiate addiction that examined the effect of a new treatment on thrice‐weekly binary urine tests to assess opiate use over follow‐up. The interest in this opiate addiction clinical trial was to characterize the transition pattern of opiate use (in each treatment arm) as well as to compare both the marginal probability of a positive urine test over follow‐up and the time until the first positive urine test between the treatment arms. We develop a shared random effects model that links together the propensity of transition between states and the probability of either an intermittent missed observation or dropout. This approach allows for heterogeneous transition and missing data patterns between individuals as well as incorporating informative intermittent missing data and dropout. We compare this new approach with other approaches proposed for the analysis of longitudinal binary data with informative missingness.
Objectives Daily driving of diesel-powered tractors has been linked to increased lung cancer risk in farmers, yet few studies have quantified exposure levels to diesel exhaust during tractor driving or during other farm activities. We expanded an earlier task-based descriptive investigation of factors associated with real-time exposure levels to black carbon (BC, a surrogate of diesel exhaust) in Iowa farmers by increasing the sample size, collecting repeated measurements, and applying statistical models adapted to continuous measurements.
Methods The expanded study added 43 days of sampling, for a total of 63 sample days conducted in 2015 and 2016 on 31 Iowa farmers. Real-time, continuous monitoring (30-s intervals) of personal BC concentrations was performed using a MicroAeth AE51 microaethelometer affixed with a micro-cyclone. A field researcher recorded information on tasks, fuel type, farmer location, and proximity to burning biomass. We evaluated the influence of these variables on log-transformed BC concentrations using a linear mixed-effect model with random effects for farmer and day and a first-order autoregressive structure for within-day correlation.
Results Proximity to diesel-powered equipment was observed for 42.5% of the overall sampling time and on 61 of the 63 sample days. Predicted geometric mean BC concentrations were highest during grain bin work, loading, and harvesting, and lower for soil preparation and planting. A 68% increase in BC concentrations was predicted for close proximity to a diesel-powered vehicle, relative to far proximity, while BC concentrations were 44% higher in diesel vehicles with open cabins compared with closed cabins. Task, farmer location, fuel type, and proximity to burning biomass explained 8% of within-day variance in BC concentrations, 2% of between-day variance, and no between-farmer variance.
Conclusion Our findings showed that farmers worked frequently near diesel equipment and that BC concentrations varied between tasks and by fuel type, farmer location, and proximity to burning biomass. These results could support the development of exposure models applicable to investigations of health effects in farmers associated with exposure to diesel engine exhaust.
We described physical activity measures and hourly patterns in patients with chronic obstructive pulmonary disease (COPD) after stratification for generic and COPD-specific characteristics and, based on multiple physical activity measures, we identified clusters of patients. In total, 1001 patients with COPD (65% men; age, 67 years; forced expiratory volume in the first second [FEV1], 49% predicted) were studied cross-sectionally. Demographics, anthropometrics, lung function and clinical data were assessed. Daily physical activity measures and hourly patterns were analysed based on data from a multisensor armband. Principal component analysis (PCA) and cluster analysis were applied to physical activity measures to identify clusters. Age, body mass index (BMI), dyspnoea grade and ADO index (including age, dyspnoea and airflow obstruction) were associated with physical activity measures and hourly patterns. Five clusters were identified based on three PCA components, which accounted for 60% of variance of the data. Importantly, couch potatoes (i.e. the most inactive cluster) were characterised by higher BMI, lower FEV1, worse dyspnoea and higher ADO index compared to other clusters (p < 0.05 for all). Daily physical activity measures and hourly patterns are heterogeneous in COPD. Clusters of patients were identified solely based on physical activity data. These findings may be useful to develop interventions aiming to promote physical activity in COPD. ; The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: RM is supported by the National Council of Scientific and Technological Development (CNPq), Brazil (246704/2012-8). DB holds a Canada Research Chair, Canada. SSCK was funded by the Medical Research Council, UK. WD-CM was funded by the Medical Research Council, UK, and the National Institute for Health Research, UK. MSP was supported by an unrestricted research grant from Astra Zeneca. KCF is supported by the Coordination for the Improvement of Higher Education Personnel (CAPES), Brazil. SZ, DM, SD and JDL were supported by the following foundations: 'Gottfried und Julia Bangerter-Rhyner-Stiftung', 'Freiwillige Akademische Gesellschaft Basel' and 'Forschungsfonds der Universitat Basel', Switzerland. DS was supported by GSK and by the Medical Research Council, UK (G0701628). FP is supported by CNPq, Brazil. PRE was supported by an NHMRC Research Fellowship, Australia (1042341). MIP's contribution to this manuscript was funded by the NIHR Respiratory Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College, UK. EFMW was supported by Point-One funding from AgentschapNL, Dutch Ministry of Economic affairs, the Netherlands. AWV was supported by 'Stichting de Weijerhorst' and Point-One funding from AgentschapNL, Dutch Ministry of Economic affairs, Netherlands. MAS was supported by Point-One funding from AgentschapNL, Dutch Ministry of Economic affairs, the Netherlands. Part of the data was sponsored by GlaxoSmithKline (data from the ECLIPSE cohort sub-study). Data from Ireland was supported by Beaumont Foundation, Ireland and SwordMedical Ltd, Ireland. The Australian sites were supported by a National Health and Medical Research Grant, Australia (grant no.: 570814). Part of the data collection in the UK (data from Leicester) was supported by the National Institute for Health Research (NIHR) Leicestershire, Northamptonshire and Rutland Collaboration for Leadership in Applied Health Research and Care and took place at University Hospitals of Leicester NHS Trust, UK, and by the NIHR Leicester Respiratory Biomedical Research Unit, UK. Data from the PAC-COPD study was funded by grants from the following Spanish institutions: Fondo de Investigacion Sanitaria, Ministry of Health (FIS PI020541); Agencia d'Avaluacio de Tecnologia i Recerca Mediques, Catalonia Government (AATRM 035/20/02); Spanish Society of Pneumology and Thoracic Surgery (SEPAR 2002/137); Catalan Foundation of Pneumology (FUCAP 2003 Beca Maria Rava); Red RESPIRA (RTIC C03/11); Red RCESP (RTIC C03/09), Fondo de Investigacion Sanitaria (PI052486); Fondo de Investigacion Sanitaria (PI052302); Fundacio La Marato de TV3 (No. 041110); and DURSI (2005SGR00392); and by unrestricted educational grants from Novartis Farmaceutica and AstraZeneca Farmaceutica.