Adjustments for non-telephone bias in random-digit-dialling surveys

Telephone surveys are widely used in the U.S.A. for the study of health-related topics. They are subject to 'coverage bias' because they cannot sample households that do not have telephones. Although only around 5 per cent of households do not have a telephone, rates of telephone coverage show substantial variation by geography, demographic factors and socio-economic factors. In particular, lack of telephone service is more common among households that contain ethnic and racial minorities or that have lower socio-economic status with fewer opportunities for access to medical care and poorer health outcomes. Thus, failure to adequately account for households without telephones in health surveys may yield estimates of health outcomes that are misleading, particularly in states with at least moderate telephone non-coverage. The dynamic nature of the population of households without telephones offers a way of accounting for such households in telephone surveys. At any given time the population of telephone households includes households that have had a break or interruption in telephone service. Empirical results strongly suggest that these households are very similar to households that have never had telephone service. Thus, sampled households that report having had an interruption in telephone service may be used also to represent the portion of the population that has never had telephone service. This strategy can lead to a reduction in non-coverage bias in random-digit-dialling surveys. This paper presents two methods of adjusting for non-coverage of non-telephone households. The effectiveness of these methods is examined using data from the National Health Interview Survey. The interruption-in-telephone-service methods reduce non-coverage bias and can also result in a lower mean squared error. The application of the interruption-in-telephone-service methods to the National Immunization Survey is also discussed. This survey produces estimates for the 50 states and 28 urban areas. The interruption-in-telephone-service estimates tend be slightly lower than estimates resulting from poststratification and from another non-coverage adjustment method. The results suggest that the reduction in bias is greatest for variables that are highly correlated with the presence or absence of telephone service.     

Model-assisted estimators for time-to-event data from complex surveys

We develop model-assisted estimators for complex survey data for the proportion of a population that experienced some event by a specified time t. Theory for the new estimators uses time-to-event models as the underlying framework but have both good model-based and design-based properties. The estimators are compared in a simulation to traditional survey estimation methods and are also applied to a study of nurses' health. The new estimators take advantage of covariates predictive of the event and reduce standard errors compared to conventional alternatives.     

A stratified Wilcoxon-type test for trend

We propose an extension of Cuzick's non-parametric Wilcoxon-like trend test for situations when the observations can be grouped into strata which are thought to be related to the outcome. We illustrate its use in two examples from cancer epidemiology.     

Sample size/power calculation for stratified case–cohort design

The case–cohort (CC) study design usually has been used for risk factor assessment in epidemiologic studies or disease prevention trials for rare diseases. The sample size/power calculation for a stratified CC (SCC) design has not been addressed before. This article derives such result based on a stratified test statistic. Simulation studies show that the proposed test for the SCC design utilizing small sub‐cohort sampling fractions is valid and efficient for situations where the disease rate is low. Furthermore, optimization of sampling in the SCC design is discussed and compared with proportional and balanced sampling techniques. An epidemiological study is provided to illustrate the sample size calculation under the SCC design. Copyright © 2014 John Wiley & Sons, Ltd.     

多阶段抽样调查资料的加权估计法

多阶段抽样技术广泛应用于流行病学现况调查中,但针对其所得资料的统计分析方法往往选择不当.文中介绍一种用于多阶段抽样调查资料的统计分析方法--加权估计法,以推广针对此类资料的恰当的分析方法.在介绍加权估计法基本原理的基础上通过两个二阶段分层整群抽样的实际调查资料实现这种算法.加权估计法可以校正由多阶段抽样导致的三种效应:群效应、分层效应、不等概率性,给出无偏点估计和比较客观的误差估计,并作出正确的统计推断.     

Comparison of tests for categorical data from a stratified cluster randomized trial

Two features commonly exhibited by randomized trials of health promotion interventions are cluster randomization and stratification. Ignoring correlations between individuals within clusters can lead to an inflated type I error rate and hence a P-value which overstates the significance of the result. This paper compares several methods for analysing categorical data from a stratified cluster randomized trial. We propose an extension of a method from survey sampling that uses the design effect to reduce the effective sample size. We compare this with three methods from Zhang and Boos that extend the standard Cochran-Mantel-Haenszel (CMH) statistic by using appropriate covariance matrices, and with a bootstrap method. The comparison is based on empirical type I error rates from a simulation study, in which the number of clusters randomized is small, as in most public health intervention studies. The method that performs consistently well is one of the Zhang and Boos extensions of the standard CMH statistic.     

Regression modeling of combined data from multiple sample surveys

Combined data from multiple sample surveys are often used in population‐based epidemiologic studies. Combining data can be beneficial in that sampling errors are reduced and coverage biases are corrected. Also, it is often necessary in order to use information lacking in one survey but available in another. We propose an estimation equations method for generalized linear models from the combined data. The estimation procedures for logistic regression models and Poisson regression models are developed. An example of estimating the relative risk of death by smoking status is used as an illustration and a simulation study is performed to examine the performance of the method. Copyright © 2009 John Wiley & Sons, Ltd.     

A Cochran–Armitage‐type and a score‐free global test for multivariate ordinal data

We propose a Cochran–Armitage‐type and a score‐free global test that can be used to assess the presence of an association between a set of ordinally scaled covariates and an outcome variable within the range of generalized linear models. Both tests are developed within the framework of the well‐established ‘global test’ methodology and as such are feasible in high‐dimensional data situations under any correlation and enable adjustment for covariates. The Cochran–Armitage‐type test, for which an intimate connection with the traditional score‐based Cochran–Armitage test is shown, rests upon explicit assumptions on the distances between the covariates' ordered categories. The score‐free test, in contrast, parametrizes these distances and thus keeps them flexible, rendering it ideally suited for covariates measured on an ordinal scale. As confirmed by means of simulations, the Cochran–Armitage‐type test focuses its power on set‐outcome relationships where the distances between the covariates' categories are equal or close to those assumed, whereas the score‐free test spreads its power over a wide range of possible set‐outcome relationships, putting more emphasis on monotonic than on non‐monotonic ones. Based on the tests' power properties, it is discussed when to favour one or the other, and the practical merits of both of them are illustrated by an application in the field of rehabilitation medicine. Our proposed tests are implemented in the R package globaltest . Copyright © 2016 John Wiley & Sons, Ltd.     

Estimation of the average correlation coefficient for stratified bivariate data

If the relationship between two ordered categorical variables X and Y is influenced by a third categorical variable with K levels, the Cochran-Mantel-Haenszel (CMH) correlation statistic QC is a useful stratum-adjusted summary statistic for testing the null hypothesis of no association between X and Y. Although motivated by and developed for the case of K I x J contingency tables, the correlation statistic QC is also applicable when X and Y are continuous variables. In this paper we derive a corresponding estimator of the average correlation coefficient for K I x J tables. We also study two estimates of the variance of the average correlation coefficient. The first is a restricted variance based on the variances of the observed cell frequencies under the null hypothesis of no association. The second is an unrestricted variance based on an asymptotic variance derived by Brown and Benedetti. The estimator of the average correlation coefficient works well in tables with balanced and unbalanced margins, for equal and unequal stratum-specific sample sizes, when correlation coefficients are constant over strata, and when correlation coefficients vary across strata. When the correlation coefficients are zero, close to zero, or the cell frequencies are small, the confidence intervals based on the restricted variance are preferred. For larger correlations and larger cell frequencies, the unrestricted confidence intervals give superior performance. We also apply the CMH statistic and proposed estimators to continuous non-normal data sampled from bivariate gamma distributions. We compare our methods to statistics for data sampled from normal distributions. The size and power of the CMH and normal theory statistics are comparable. When the stratum-specific sample sizes are small and the distributions are skewed, the proposed estimator is superior to the normal theory estimator. When the correlation coefficient is zero or close to zero, the restricted confidence intervals provide the best performance. None of the confidence intervals studied provides acceptable performances across all correlation coefficients, sample sizes and non-normal distributions.     

A procedure for linking psychosocial job characteristics data to health surveys.

A system is presented for linking information about psychosocial characteristics of job situations to national health surveys. Job information can be imputed to individuals on surveys that contain three-digit US Census occupation codes. Occupational mean scores on psychosocial job characteristics-control over task situation (decision latitude), psychological work load, physical exertion, and other measures-for the linkage system are derived from US national surveys of working conditions (Quality of Employment Surveys 1969, 1972, and 1977). This paper discusses a new method for reducing the biases in multivariate analyses that are likely to arise when utilizing linkage systems based on mean scores. Such biases are reduced by modifying the linkage system to adjust imputed individual scores for demographic factors such as age, education, race, marital status and, implicitly, sex (since men and women have separate linkage data bases). Statistics on the linkage system's efficiency and reliability are reported. All dimensions have high inter-survey reproducibility. Despite their psychosocial nature, decision latitude and physical exertion can be more efficiently imputed with the linkage system than earnings (a non-psychosocial job characteristic). The linkage system presented here is a useful tool for initial epidemiological studies of the consequences of psychosocial job characteristics and constitutes the methodological basis for the subsequent paper.     

Sample size estimation for the van Elteren test--a stratified Wilcoxon-Mann-Whitney test

The van Elteren test is a type of stratified Wilcoxon-Mann-Whitney test for comparing two treatments accounting for strata. In this paper, we study sample size estimation methods for the asymptotic version of the van Elteren test, assuming that the stratum fractions (ratios of each stratum size to the total sample size) and the treatment fractions (ratios of each treatment size to the stratum size) are known in the study design. In particular, we develop three large-sample sample size estimation methods and present a real data example to illustrate the necessary information in the study design phase in order to apply the methods. Simulation studies are conducted to compare the performance of the methods and recommendations are made for method choice. Finally, sample size estimation for the van Elteren test when the stratum fractions are unknown is also discussed.     

Feasibility of using web surveys to collect time-activity data

Time-activity data are traditionally collected by telephone interviews or through paper diaries, which are time consuming and costly. As a potential alternative that may greatly save staff time, a web survey to collect time-activity data was developed and tested in this study. We collected 24-h recall web diaries from 151 parents of young children mostly under 55 years of age (who also answered for their children) and 55 older adults (≥ 55 years of age) both on a weekday and a weekend day every 3 months during an 18-month period. The performance and reliability of the web surveys collected were evaluated, including the survey-completion rate, and the percentage of surveys with unreasonable time being reported as spent sleeping and with missing reports of being in transit between locations. We also compared the web-survey data with time-activity information we collected from the same subjects in telephone interviews and found that these data sources were fairly consistent with each other. However, we observed slightly more compliance issues for the web than the telephone survey, but most of these issues could be addressed and minimized by refining some questions or the survey interface. Our study suggests that it is critical to reduce participants' burden and improve survey interface design for optimal compliance and data quality. In conclusion, web surveys are a promising method to consider for time-activity data collection.     

Analysis of vitamin A data from the health and nutrition examination surveys

An Expert Panel on Vitamin A Nutriture assessed serum vitamin A values and related data collected during the first and second National Health and Nutrition Examination Surveys (NHANES I and NHANES II) and the southwest component of the Hispanic Health and Nutrition Examination Survey (SW HHANES). The appropriateness of the analytical methodologies and adequacy of the quality-control procedures used to obtain the serum vitamin A values were evaluated. Age-specific guidelines for the interpretation of serum vitamin A values in selected low ranges (less than 20, 20-24 and 25-29 micrograms/dl) were developed, and estimates of the prevalence of serum vitamin A values in these ranges were presented for each of the three surveys.     

Standardizing the power of the Hosmer–Lemeshow goodness of fit test in large data sets

The Hosmer–Lemeshow test is a commonly used procedure for assessing goodness of fit in logistic regression. It has, for example, been widely used for evaluation of risk‐scoring models. As with any statistical test, the power increases with sample size; this can be undesirable for goodness of fit tests because in very large data sets, small departures from the proposed model will be considered significant. By considering the dependence of power on the number of groups used in the Hosmer–Lemeshow test, we show how the power may be standardized across different sample sizes in a wide range of models. We provide and confirm mathematical derivations through simulation and analysis of data on 31,713 children from the Collaborative Perinatal Project. We make recommendations on how to choose the number of groups in the Hosmer–Lemeshow test based on sample size and provide example applications of the recommendations. Copyright © 2012 John Wiley & Sons, Ltd.     

Lifestyle and socio-economic status--verification of data reliability from questionnaire surveys

The study of reliability was part of a project funded by the Grant Agency of the Czech Ministry of Health. The repeatability of answers, 180 questionnaires being sent again after a six week interval, was assessed by two methods: by the percentages of agreement and using the Kappa index. We evaluated 61 questions, which were divided into five sections (general questions, questions on employment, lifestyle, health, and personality). The agreement rate varied from 46% to 100% and the value of the Kappa index from -0.01 to 1. The agreement performed by the Kappa index was divided into 4 groups (< or = 0.4--poor; 0.41-0.6--average; 0.61-0.80--good; 0.81-1 almost perfect). The agreement across our questionnaire was poor in 6.6% of the questions, average--in 31.1%, good--in 45.9% and almost perfect--in 16.4% of the questions. Next we divided the questions by their content into two groups--the factual questions (group 1) and questions where answers contain an evaluative or motivational element (group 2). Significant differences (p < 0.001) were found between groups 1 and 2 in the agreement (86.8%; 72.1%) and the Kappa index (0.73; 0.48). In questions where agreement of answers is evaluated as average or even weak we used three types of solution: a) substitution by the analogous questions with a higher agreement, b) aggregation of the answers if possible, or c) the questions were not used for the analyses. These important results will be considered in further analyses of the data and in the interpretation of the study results. The detailed information about the whole project and the results has already been published or presented are available on the following web site: www.zuova.cz/projekty/ses.php.     

Disclosure control using partially synthetic data for large‐scale health surveys,with applications to CanCORS

Statistical agencies have begun to partially synthesize public‐use data for major surveys to protect the confidentiality of respondents’ identities and sensitive attributes by replacing high disclosure risk and sensitive variables with multiple imputations. To date, there are few applications of synthetic data techniques to large‐scale healthcare survey data. Here, we describe partial synthesis of survey data collected by the Cancer Care Outcomes Research and Surveillance (CanCORS) project, a comprehensive observational study of the experiences, treatments, and outcomes of patients with lung or colorectal cancer in the USA. We review inferential methods for partially synthetic data and discuss selection of high disclosure risk variables for synthesis, specification of imputation models, and identification disclosure risk assessment. We evaluate data utility by replicating published analyses and comparing results using original and synthetic data and discuss practical issues in preserving inferential conclusions. We found that important subgroup relationships must be included in the synthetic data imputation model, to preserve the data utility of the observed data for a given analysis procedure. We conclude that synthetic CanCORS data are suited best for preliminary data analyses purposes. These methods address the requirement to share data in clinical research without compromising confidentiality. Copyright © 2013 John Wiley & Sons, Ltd.     

Tests of non-null hypothesis on proportions for stratified data

It is more reasonable to interpret the efficacy of therapies under non-null hypothesis, hence tests of non-null hypothesis have been carried out. Most clinical trials adopt multi-center campaign and yield stratified data. However, the existing non-null hypothesis tests are not suitable for stratified data. This paper proposes the tests of non-null hypothesis on proportions for stratified data. Averaging the treatment-control difference in each stratum yields the mean treatment-control difference. Comparing its expectation with the minimal detectable difference leads to set up a non-null hypothesis. Its variance is used to construct the equation of the basic relationship for stratified designs under the non-null hypothesis. Then follow the derivations for the one- and two-sample tests. Their performance is demonstrated by the Monte Carlo method. As far as the two-sample tests are concerned, they reduce to the Cochran test and the Mantel-Haenszel test, on setting the minimal detectable difference equal to zero, and to the Dunnett-Gent test when there is only one stratum. As for the one-sample test, it also reduces to its classical counterparts in these situations. The observed power coincides with the prescribed power and the relevant operating characteristic curves. The tests can be applied to the active control clinical trials with multi-center or stratified designs for establishing the clinical superiority or non-inferiority of a tested drug versus control. Worked examples illustrate the methodology.