Stacked survival models for residual lifetime data.

When modelling the survival distribution of a disease for which the symptomatic progression of the associated condition is insidious, it is not always clear how to measure the failure/censoring times from some true date of disease onset. In a prevalent cohort study with follow-up, one approach for removing any potential influence from the uncertainty in the measurement of the true onset dates is through the utilization of only the residual lifetimes. As the residual lifetimes are measured from a well-defined screening date (prevalence day) to failure/censoring, these observed time durations are essentially error free. Using residual lifetime data, the nonparametric maximum likelihood estimator (NPMLE) may be used to estimate the underlying survival function. However, the resulting estimator can yield exceptionally wide confidence intervals. Alternatively, while parametric maximum likelihood estimation can yield narrower confidence intervals, it may not be robust to model misspecification. Using only right-censored residual lifetime data, we propose a stacking procedure to overcome the non-robustness of model misspecification; our proposed estimator comprises a linear combination of individual nonparametric/parametric survival function estimators, with optimal stacking weights obtained by minimizing a Brier Score loss function.

Parametric models for combined failure time data from an incident cohort study and a prevalent cohort study with follow-up.

A classical problem in survival analysis is to estimate the failure time distribution from right-censored observations obtained from an incident cohort study. Frequently, however, failure time data comprise two independent samples, one from an incident cohort study and the other from a prevalent cohort study with follow-up, which is known to produce length-biased observed failure times. There are drawbacks to each of these two types of study when viewed separately. We address two main questions here: (i) Can our statistical inference be enhanced by combining data from an incident cohort study with data from a prevalent cohort study with follow-up? (ii) What statistical methods are appropriate for these combined data? The theory we develop to address these questions is based on a parametrically defined failure time distribution and is supported by simulations. We apply our methods to estimate the duration of hospital stays.

Risk Taking Runners Slow More in the Marathon.

Much research has explored the physiological, energetic, environmental, and psychological factors that influence pacing in endurance events. Although this research has generally neglected the role of psychological variation across individuals, recent studies have hinted at its importance. Here we conducted an online survey of over 1,300 marathon runners, testing whether any of five psychological constructs - competitiveness, goal achievement, risk taking in pace (RTP), domain-specific risk taking, and willingness to suffer in the marathon - predicted slowing in runners' most recent marathons. Analyses revealed that RTP - the extent to which runners agreed that they began the marathon at a pace that was so fast that it would jeopardize their capacity to maintain this pace throughout the event - was a robust predictor of marathon slowing. RTP proved a substantial predictor even in regression models controlling for the other psychological constructs, training, experience, and other known pacing correlates. This result suggests that marathoners consider trade-offs when making pacing decisions, and that individuals vary in their pacing decision making.

Benefits of combining prevalent and incident cohorts: An application to myotonic dystrophy.

It is frequently of interest to estimate the time that individuals survive with a disease, that is, to estimate the time between disease onset and occurrence of a clinical endpoint such as death. Epidemiologic survival data are commonly collected from either an incident cohort, whose members' disease onset occurs after the study baseline date, or from a cohort with prevalent disease that is followed forward in time. Incident cohort survival data are limited by study termination, while prevalent cohort data provide biased (left-truncated) survival data. In this article, we investigate the advantages of a study design featuring simultaneous follow-up of prevalent and incident cohorts to the estimation of the survivor function. Our analyses are supported by simulations and illustrated using data on survival after myotonic dystrophy diagnosis from the United Kingdom Clinical Practice Research Datalink (CPRD). We demonstrate that the NPMLE using combined incident and prevalent cohort data estimates the true survivor function very well, even for moderate sample sizes, and ameliorates the disadvantages of using a purely incident or prevalent cohort.

Methods for Real-Time Prediction of the Mode of Travel Using Smartphone-Based GPS and Accelerometer Data.

We propose and compare combinations of several methods for classifying transportation activity data from smartphone GPS and accelerometer sensors. We have two main objectives. First, we aim to classify our data as accurately as possible. Second, we aim to reduce the dimensionality of the data as much as possible in order to reduce the computational burden of the classification. We combine dimension reduction and classification algorithms and compare them with a metric that balances accuracy and dimensionality. In doing so, we develop a classification algorithm that accurately classifies five different modes of transportation (i.e., walking, biking, car, bus and rail) while being computationally simple enough to run on a typical smartphone. Further, we use data that required no behavioral changes from the smartphone users to collect. Our best classification model uses the random forest algorithm to achieve 96.8% accuracy.

Using Survival Analysis to Identify Risk Factors for Treatment Interruption among New and Retreatment Tuberculosis Patients in Kenya.

Despite high tuberculosis (TB) treatment success rate, treatment adherence is one of the major obstacles to tuberculosis control in Kenya. Our objective was to identify patient-related factors that were associated with time to TB treatment interruption and the geographic distribution of the risk of treatment interruption by county. Data of new and retreatment patients registered in TIBU, a Kenyan national case-based electronic data recording system, between 2013 and 2014 was obtained. Kaplan-Meier curves and log rank tests were used to assess the adherence patterns. Mixed-effects Cox proportional hazards modeling was used for multivariate analysis. Records from 90,170 patients were included in the study. The cumulative incidence of treatment interruption was 4.5% for new patients, and 8.5% for retreatment patients. The risk of treatment interruption was highest during the intensive phase of treatment. Having previously been lost to follow-up was the greatest independent risk factor for treatment interruption (HR: 4.79 [3.99, 5.75]), followed by being HIV-positive not on ART (HR: 1.96 [1.70, 2.26]) and TB relapse (HR: 1.70 [1.44, 2.00]). Male and underweight patients had high risks of treatment interruption (HR: 1.46 [1.35, 1.58]; 1.11 [1.03, 1.20], respectively). High rates of treatment interruption were observed in counties in the central part of Kenya while counties in the northeast had the lowest risk of treatment interruption. A better understanding of treatment interruption risk factors is necessary to improve adherence to treatment. Interventions should focus on patients during the intensive phase, patients who have previously been lost to follow-up, and promotion of integrated TB and HIV services among public and private facilities.

Fast men slow more than fast women in a 10 kilometer road race.

Background. Previous studies have demonstrated that men are more likely than women to slow in the marathon (footrace). This study investigated whether the sex difference in pacing occurs for a shorter race distance. Materials & Methods. Data were acquired from the Bolder Boulder 10 km road race for the years 2008-2013, which encompassed 191,693 performances. There were two pacing measures, percentage change in pace of the first 3 miles relative to the final 3.2 miles and percentage change in pace of the first mile relative to the final 5.2 miles. Pacing was analyzed as a continuous variable and as two categorical variables, as follows: "maintain the pace," defined as slowing <5% and "marked slowing," defined as slowing ≥10%. Results. Among the fastest (men < 48:40; women < 55:27) and second fastest (men < 53:54; women < 60:28) sex-specific finishing time sextiles, men slowed significantly more than women with both pacing measures, but there were no consistently significant sex differences in pacing among the slower four sextiles. For the fastest sextile, the odds for women were 1.96 (first pacing measure) and 1.36 (second measure) times greater than men to maintain the pace. For the fastest sextile, the odds for women were 0.46 (first measure) and 0.65 (second measure) times that of men to exhibit marked slowing. Multiple regression indicated that being older was associated with lesser slowing, but the sex difference among faster runners persisted when age was controlled. Conclusions. There was a sex difference in pacing during a 10 km race where glycogen depletion is not typically relevant. These results support the hypothesis that the sex difference in pacing partly reflects a sex difference in decision making.

U.S. masters track participation reveals a stable sex difference in competitiveness.

It is well established that men are more likely than women to engage in direct competition, but it is unclear if this reflects social structural conditions or evolved predispositions. These hypotheses can be addressed by quantifying competitiveness in sports and testing if the sex difference is decreasing over time in the U.S., a society where social roles have converged. Study 1 assessed participation and the occurrence of relatively fast performances by masters runners (40-74 years old) at recent road races and track meets. Fast performances occurred over 20 times more often at track meets than at road races. Women comprised 55% of finishers at roads races but only 15-28% of finishers at track meets. Thus, the sex difference in masters track participation can serve as a measure of the sex difference in competitiveness. Study 2 used data from national championship meets and yearly rankings lists to test whether the sex difference in masters track participation decreased from 1988-2012. The sex difference decreased overall, but there was no evidence of change since the late 1990s. Therefore, the sex difference in the willingness to engage in direct sports competition appears to reflect both social structural conditions and evolved predispositions.

A behavioural dynamic model of the relative age effect.

The relationship between date of birth and success in a variety of sports, including hockey, is well established. This phenomenon is known as the relative age effect (RAE). We model the RAE in Canadian youth hockey as a positive feedback loop where an initial age advantage is reinforced through additional training and playing opportunities based on perceived skill superiority. The same causal mechanism leads to a higher quit rate for relatively younger players. Our model effectively replicates the birth month distribution of Canadian National Hockey League players (R2 = 86.79%) when driven by Canadian birth distributions. We use this model to evaluate three policies that aim to lessen the RAE. All of the policies reduce the RAE with a significant delay. The most effective policy is a combination of providing additional support to age disadvantaged children and rotating the cut-off date for youth leagues between January 1st and July 1st annually. In equilibrium, this approach leads to a 96% reduction in the RAE compared to the base case.

Testing the assumptions for the analysis of survival data arising from a prevalent cohort study with follow-up.

In a prevalent cohort study with follow-up subjects identified as prevalent cases are followed until failure (defined suitably) or censoring. When the dates of the initiating events of these prevalent cases are ascertainable, each observed datum point consists of a backward recurrence time and a possibly censored forward recurrence time. Their sum is well known to be the left truncated lifetime. It is common to term these left truncated lifetimes "length biased" if the initiating event times of all the incident cases (including those not observed through the prevalent sampling scheme) follow a stationary Poisson process. Statistical inference is then said to be carried out under stationarity. Whether or not stationarity holds, a further assumption needed for estimation of the incident survivor function is the independence of the lifetimes and their accompanying truncation times. That is, it must be assumed that survival does not depend on the calendar date of the initiating event. We show how this assumption may be checked under stationarity, even though only the backward recurrence times and their associated (possibly censored) forward recurrence times are observed. We prove that independence of the lifetimes and truncation times is equivalent to equality in distribution of the backward and forward recurrence times, and exploit this equivalence as a means of testing the former hypothesis. A simulation study is conducted to investigate the power and Type 1 error rate of our proposed tests, which include a bootstrap procedure that takes into account the pairwise dependence between the forward and backward recurrence times, as well as the potential censoring of only one of the members of each pair. We illustrate our methods using data from the Canadian Study of Health and Aging. We also point out an equivalence of the problem presented here to a non-standard changepoint problem.

Association of unmet need with self-rated health in a community dwelling cohort of disabled seniors 75 years of age and over.

Self-rated health (SRH) is a measure of perceived health that has been shown to predict use of community services, functional decline, pain, and mortality. Many factors associated with SRH have been identified, but unmet need for physical assistance with activities of daily living (ADL) has not yet been examined. The objective of this paper is to examine the association between unmet need and SRH while accounting for the effects of other, previously identified, correlates of SRH. We conducted a secondary analysis of a population-based study of 839 residents of Montréal, Québec who were 75 years of age or older, not cognitively impaired, and living in the community. Multivariable logistic regression was used to evaluate the association between met and unmet personal ADL (PADL) and instrumental ADL (IADL) need for physical assistance with SRH. Among 508 disabled community-dwelling elderly, for each additional unmet IADL need, subjects were 1.70 (95% CI: 1.11-2.61) times more likely to report poorer SRH. For each additional unmet PADL need, subjects were 2.26 (95% CI: 1.31-3.91) times more likely to report poorer SRH. Subjects at increased risk of malnutrition, with greater comorbidity and whose income was insufficient to meet their needs were also more likely to report poorer SRH. After adjustment for important correlates, unmet PADL and IADL needs retain a statistically significant association with poorer SRH, with nutritional status, comorbid conditions, and income satisfaction being important confounders of the relationship.

A formal test for the stationarity of the incidence rate using data from a prevalent cohort study with follow-up.

In a prevalent cohort study with follow-up, the incidence process is not directly observed since only the onset times of prevalent cases can be ascertained. Assessing the "stationarity" of the underlying incidence process can be important for at least three reasons, including an improvement in efficiency when estimating the survivor function. We propose, for the first time, a formal test for stationarity using data from a prevalent cohort study with follow-up. The test makes use of a characterization of stationarity, an extension of this characterization developed in this paper, and of a test for matched pairs of right censored data. We report the results from a power study assuming varying degrees of departure from the null hypothesis of stationarity. The test is also applied to data obtained as part of the Canadian Study of Health and Aging (CSHA) to verify whether the incidence rate of dementia amongst the elderly in Canada has remained constant.