-
Methods for stratification of person-time and events
- a prerequisite for Poisson regression and SIR estimation
Many epidemiological methods for analysing follow-up studies require the calculation of rates based on accumulating person-time and events, stratified by various factors. Managing this stratification and accumulation is often the most difficult aspect of this type of analysis.
We provide a tutorial on accumulating person-time and events, stratified by various factors i.e. creating event-time tables. We show how to efficiently generate event-time tables for many different outcomes simultaneously. We also provide a new vocabulary to characterise and differentiate time-varying factors. The tutorial is focused on using a SAS macro to perform most of the common tasks in the creation of event-time tables. All the most common types of time-varying covariates can be generated and categorised by the macro. It can also provide output suitable for other types of survival analysis (e.g. Cox regression). The aim of our methodology is to support the creation of bug-free, readable, efficient, capable and easily modified programs for making event-time tables. We briefly compare analyses based on event-time tables with Cox regression and nested case-control studies for the analysis of follow-up data.
Anyone working with time-varying covariates, particularly from large detailed person-time data sets, would gain from having these methods in their programming toolkit.
-
Case-case analysis of enteric diseases with routine surveillance data: Potential use and example results
Background:
Case-control studies and outbreak investigations are the major epidemiological tools for providing detailed information on enteric disease sources and risk factors, but these investigations can be constrained by cost and logistics.
Methods:
We explored the advantages and disadvantages of comparing risk factors for enteric diseases using the case-case method. The main issues are illustrated with an analysis of routine notification data on enteric diseases for 2006 collected by New Zealand's national surveillance system.
Results:
Our analyses of aggregated New Zealand surveillance data found that the associations (crude odds ratios) for risk factors of enteric disease were fairly consistent with findings from local case-control studies and outbreak investigations, adding support for the use of the case-case analytical approach. Despite various inherent limitations, such an approach has the potential to contribute to the monitoring of risk factor trends for enteric diseases. Nevertheless, using the case-case method for analysis of routine surveillance data may need to be accompanied by: (i) reduction of potential selection and information biases by improving the quality of the surveillance data; and (ii) reduction of confounding by conducting more sophisticated analyses based on individual-level data.
Conclusion:
Case-case analyses of enteric diseases using routine surveillance data might be a useful low-cost means to study trends in enteric disease sources and inform control measures. If used, it should probably supplement rather than replace outbreak investigations and case-control studies. Furthermore, it could be enhanced by utilising high quality individual-level data provided by nationally-representative sentinel sites for enteric disease surveillance.
-
Partitioning the population attributable fraction for a sequential chain of effects
Background:
While the population attributable fraction (PAF) provides potentially valuable information regarding the community-level effect of risk factors, significant limitations exist with current strategies for estimating a PAF in multiple risk factor models. These strategies can result in paradoxical or ambiguous measures of effect, or require unrealistic assumptions regarding variables in the model. A method is proposed in which an overall or total PAF across multiple risk factors is partitioned into components based upon a sequential ordering of effects. This method is applied to several hypothetical data sets in order to demonstrate its application and interpretation in diverse analytic situations.
Results:
The proposed method is demonstrated to provide clear and interpretable measures of effect, even when risk factors are related/correlated and/or when risk factors interact. Furthermore, this strategy not only addresses, but also quantifies issues raised by other researchers who have noted the potential impact of population-shifts on population-level effects in multiple risk factor models.
Conclusion:
Combined with simple, unadjusted PAF estimates and an aggregate PAF based on all risk factors under consideration, the sequentially partitioned PAF provides valuable additional information regarding the
process
through which population rates of a disorder may be impacted. In addition, the approach can also be used to statistically control for confounding by other variables, while avoiding the potential pitfalls of attempting to separately differentiate direct and indirect effects.
-
Flexible Two-Phase studies for rare exposures: Feasibility, planning and efficiency issues of a new variant
The two-phase design consists of an initial (Phase One) study with known disease status and inexpensive covariate information. Within this initial study one selects a subsample on which to collect detailed covariate data. Two-phase studies have been shown to be efficient compared to standard case-control designs. However, potential problems arise if one cannot assure minimum sample sizes in the rarest categories or if recontact of subjects is difficult.In the case of a rare exposure with an inexpensive proxy, the authors propose the flexible two-phase design for which there is a single time of contact, at which a decision about full covariate ascertainment is made based on the proxy. Subjects are screened until the desired numbers of cases and controls have been selected for full data collection. Strategies for optimizing the cost/efficiency of this design and corresponding software are presented. The design is applied to two examples from occupational and genetic epidemiology. By ensuring minimum numbers for the rarest disease-covariate combination(s), we obtain considerable efficiency gains over standard two-phase studies with an improved practical feasibility.The flexible two-phase design may be the design of choice in the case of well targeted studies of the effect of rare exposures with an inexpensive proxy.
-
Factors related to the frequency of citation of epidemiologic publications
Background:
Previous studies have demonstrated that the frequency with which a publication is cited varies greatly. Our objective was to determine whether author, country, journal, or topic were associated with the number of times an epidemiological publication is cited.
Methods:
We used outcome-based sampling and investigated one public health issue – child injury prevention, and one clinical topic – coronary artery disease (CAD) prevention. Using the Institute for Scientific Information's (ISI) Web of Science® databases, we limited searches to full articles involving humans published in English between 1998 and 2004. We calculated the citation rate and, after frequency-matching on year of publication, selected the 36 most frequently cited and 36 least frequently cited articles per year, for a total of 252 highly-cited and 252 infrequently-cited articles per topic area (child injury prevention and CAD prevention).
Results:
Highly-cited articles in both CAD and child injury prevention were more likely to be published in medium or high impact journals or in journals with medium or high circulations. They were also more likely to be published by authors from U.S. institutions. Among articles examining CAD prevention, the highly-cited articles often involved risk factors, and the association between topics and frequency of citation persisted after adjusting for impact factor. Among articles addressing child injury prevention, topic was not statistically associated with citation.
Conclusion:
Journal and country appear to be the factors most strongly associated with frequency of citation. In particular, highly-cited articles are predominantly published in high-impact, high-circulation journals. The factors, however, differ somewhat depending on the area of research the journals represent. Among CAD prevention articles, for example, topic is also an important predictor of citation whereas the same is not true for articles addressing injury prevention.Condensed AbstractOur objective was to determine whether author, country, journal, or topic were associated with the number of times an epidemiological publication is cited. We used outcome-based sampling and investigated one public health issue, child injury prevention, and one clinical topic, coronary artery disease (CAD) prevention. Using the Institute for Scientific Information (ISI) Web of Science® databases, we limited searches to full articles involving humans published in English between 1998 and 2004. We calculated the citation rate and, after frequency-matching on year of publication, selected the 36 most frequently cited and 36 least frequently cited articles per year, for a total of 252 highly-cited and 252 infrequently-cited articles per topic area (child injury prevention and CAD prevention). Highly-cited articles in both CAD and child injury prevention were more likely to be published in medium or high impact journals or in journals with medium or high circulations. They were also more likely to be published by authors from U.S. institutions. Among articles examining CAD prevention, the highly-cited articles often involved risk factors, and the association between topics and frequency of citation persisted after adjusting for impact factor. Among articles addressing child injury prevention, topic was not statistically associated with citation.
-
Should adjustment for covariates be used in prevalence estimations?
Background:
Adjustment for covariates (also called auxiliary variables in survey sampling literature) is commonly applied in health surveys to reduce the variances of the prevalence estimators. In theory, adjusted prevalence estimators are more accurate when variance components are known. In practice, variance components needed to achieve the adjustment are unknown and their sample estimators are used instead. The uncertainty introduced by estimating variance components may overshadow the reduction in the variance of the prevalence estimators due to adjustment. We present empirical guidelines indicating when adjusted prevalence estimators should be considered, using gender adjusted and unadjusted smoking prevalence as an illustration.
Methods:
We compare the accuracy of adjusted and unadjusted prevalence estimators via simulation. We simulate simple random samples from hypothetical populations with the proportion of males ranging from 30% to 70%, the smoking prevalence ranging from 15% to 35%, and the ratio of male to female smoking prevalence ranging from 1 to 4. The ranges of gender proportions and smoking prevalences reflect the conditions in 1999–2003 Behavioral Risk Factors Surveillance System (BRFSS) data for Massachusetts. From each population, 10,000 samples are selected and the ratios of the variance of the adjusted prevalence estimators to the variance of the unadjusted (crude) ones are computed and plotted against the proportion of males by population prevalence, as well as by population and sample sizes. The prevalence ratio thresholds, above which adjusted prevalence estimators have smaller variances, are determined graphically.
Results:
In many practical settings, gender adjustment results in less accuracy. Whether or not there is better accuracy with adjustment depends on sample sizes, gender proportions and ratios between male and female prevalences. In populations with equal number of males and females and smoking prevalence of 20%, the adjusted prevalence estimators are more accurate when the ratios of male to female prevalences are above 2.4, 1.8, 1.6, 1.4 and 1.3 for sample sizes of 25, 50, 100, 150 and 200, respectively.
Conclusion:
Adjustment for covariates will not result in more accurate prevalence estimator when ratio of male to female prevalences is close to one, sample size is small and risk factor prevalence is low. For example, when reporting smoking prevalence based on simple random sampling, gender adjustment is recommended only when sample size is greater than 200.
-
Feasibility of an automated telephone survey to enable prospective monitoring of subjects whose confidentiality is paramount: a four-week cohort study of partner violence recurrence after Emergency Department discharge
ObjectiveA goal in intimate partner violence (IPV) research is to identify victims when they are treated in a hospital Emergency Department (ED) and predict which patients will sustain abuse again after discharge, so interventions can be targeted. Following patients to determine those prognostic factors is difficult, however, especially to study IPV given the risk to be assaulted if their partner learns of their participation. We assessed the feasibility of an automated telephone survey and a wireless incentive delivery system to follow ED patients after discharge, enabling detection of IPV recurrence.
Methods:
A four-week prospective cohort pilot study was conducted at an urban academic medical center ED in the U.S. Thirty patient subjects (24 women, 6 men; 18–54 years) who had sustained IPV in the past six months, 12 of whom presented for an acute IPV-related condition, were interviewed in the ED and were asked to report weekly for four weeks after discharge to a toll-free, password protected telephone survey, and answer recorded questions using the telephone keypad. A $10 convenience store debit card was provided as an incentive, and was electronically recharged with $10 for each weekly report, with a $20 bonus for making all four reports.
Results:
Twenty-two of 30 subjects (73.3%) made at least one report to the telephone survey during the four weeks following discharge; 14 of the 30 subjects (46.7%) made all four weekly reports. Each time the telephone survey was accessed, the subject completed all questions (i.e., no mid-survey break-offs). Eight months after follow-up ended, almost all debit cards (86.7%) had been used to make purchases.
Conclusion:
Approximately three of every four subjects participated in follow-up after ED discharge, and approximately two of every four subjects completed all follow-up reports, suggesting the method of an automated telephone survey and wireless incentive delivery system makes it feasible to study IPV prospectively among discharged patients. That finding, along with evidence that IPV recurrence risk is high, suggests the protocol tested is warranted for use conducting full-scale studies of IPV. The protocol could benefit efforts to study other outcomes, especially when patient confidentiality is paramount for their safety.
-
The effects of neighborhood density and street connectivity on walking behavior: the Twin Cities walking study
A growing body of health and policy research suggests residential neighborhood density and street connectivity affect walking and total physical activity, both of which are important risk factors for obesity and related chronic diseases. The authors report results from their methodologically novel Twin Cities Walking Study; a multilevel study which examined the relationship between built environments, walking behavior and total physical activity. In order to maximize neighborhood-level variation while maintaining the exchangeability of resident-subjects, investigators sampled 716 adult persons nested in 36 randomly selected neighborhoods across four strata defined on density and street-connectivity – a matched sampling design. Outcome measures include two types of self-reported walking (from surveys and diaries) and so-called objective 7-day accelerometry measures. While crude differences are evident across all outcomes, adjusted effects show increased odds of travel walking in higher-density areas and increased odds of leisure walking in low-connectivity areas, but neither density nor street connectivity are meaningfully related to overall mean miles walked per day or increased total physical activity. Contrary to prior research, the authors conclude that the effects of density and block size on total walking and physical activity are modest to non-existent, if not contrapositive to hypotheses. Divergent findings are attributed to this study's sampling design, which tends to mitigate residual confounding by socioeconomic status.
-
Case-cohort design in practice – experiences from the MORGAM Project
When carefully planned and analysed, the case-cohort design is a powerful choice for follow-up studies with multiple event types of interest. While the literature is rich with analysis methods for case-cohort data, little is written about the designing of a case-cohort study. Our experiences in designing, coordinating and analysing the MORGAM case-cohort study are potentially useful for other studies with similar characteristics. The motivation for using the case-cohort design in the MORGAM genetic study is discussed and issues relevant to its planning and analysis are studied. We propose solutions for appending the earlier case-cohort selection after an extension of the follow-up period and for achieving maximum overlap between earlier designs and the case-cohort design. Approaches for statistical analysis are studied in a simulation example based on the MORGAM data.
-
Older adults' perspectives on key domains of childhood social and economic experiences and opportunities: a first step to creating a multidimensional measure
ObjectivesAlthough research has found that childhood socioeconomic status (SES) is associated with physical and mental health in mid- and later life, most of these studies used conventional, single dimension SES measures for the childhood period such as household income or educational attainment of parents. Life course and health disparities research would benefit from identification and measurement of a variety of childhood social and economic experiences and opportunities that might affect health in later life.DesignThis study utilized qualitative research methods to identify key dimensions of childhood experiences related to SES. We conducted in-depth interviews with 25 adults age 55 to 80 years from diverse economic and ethnic backgrounds. Topics included home, neighborhood, school, and work experiences during early childhood and adolescence. Interviews were audio-taped and transcripts were coded to identify thematic domains.
Results:
We identified eight thematic domains, many of which had clear subdomains: home and family circumstances, neighborhood, work and money, potential for advancement through schooling, school quality and content, discrimination, influence and support of adults, and leisure activities. These domains highlight individual characteristics and experiences and also economic and educational opportunities.
Conclusion:
These domains of childhood social and economic circumstances add breadth and depth to conventional conceptualization of childhood SES. When the domains are translated into a measurement tool, it will allow for the possibility of classifying people along multiple dimensions, such as from a low economic circumstance with high levels of adult support.
|
