Extension of the modified Poisson regression model to prospective studies with correlated binary data
TL;DR: The Poisson regression model using a sandwich variance estimator has become a viable alternative to the logistic regression model for the analysis of prospective studies with independent binary outcomes and is extended to studies with correlated binary outcomes as arise in longitudinal or cluster randomization studies.
Abstract: The Poisson regression model using a sandwich variance estimator has become a viable alternative to the logistic regression model for the analysis of prospective studies with independent binary outcomes. The primary advantage of this approach is that it readily provides covariate-adjusted risk ratios and associated standard errors. In this article, the model is extended to studies with correlated binary outcomes as arise in longitudinal or cluster randomization studies. The key step involves a cluster-level grouping strategy for the computation of the middle term in the sandwich estimator. For a single binary exposure variable without covariate adjustment, this approach results in risk ratio estimates and standard errors that are identical to those found in the survey sampling literature. Simulation results suggest that it is reliable for studies with correlated binary data, provided the total number of clusters is at least 50. Data from observational and cluster randomized studies are used to illustrate the methods.
Citations
More filters
TL;DR: Offering self-collection of samples for HPV testing by community health workers during home visits resulted in a four-fold increase in screening uptake, showing that this strategy is effective to improve cervical screening coverage.
175 citations
TL;DR: The association between intubation and decreased survival was observed in the majority of the sensitivity and subgroup analyses, including when accounting for missing data and in a subgroup of patients with a pulse at the beginning of the event.
Abstract: Importance Tracheal intubation is common during pediatric in-hospital cardiac arrest, although the relationship between intubation during cardiac arrest and outcomes is unknown. Objective To determine if intubation during pediatric in-hospital cardiac arrest is associated with improved outcomes. Design, Setting, and Participants Observational study of data from United States hospitals in the Get With The Guidelines–Resuscitation registry. Pediatric patients ( Exposures Tracheal intubation during cardiac arrest . Main Outcomes and Measures The primary outcome was survival to hospital discharge. Secondary outcomes included return of spontaneous circulation and neurologic outcome. A favorable neurologic outcome was defined as a score of 1 to 2 on the pediatric cerebral performance category score. Patients being intubated at any given minute were matched with patients at risk of being intubated within the same minute (ie, still receiving resuscitation) based on a time-dependent propensity score calculated from multiple patient, event, and hospital characteristics. Results The study included 2294 patients; 1308 (57%) were male, and all age groups were represented (median age, 7 months [25th-75th percentiles, 21 days, 4 years]). Of the 2294 included patients, 1555 (68%) were intubated during the cardiac arrest. In the propensity score–matched cohort (n = 2270), survival was lower in those intubated compared with those not intubated (411/1135 [36%] vs 460/1135 [41%]; risk ratio [RR], 0.89 [95% CI, 0.81-0.99]; P = .03). There was no significant difference in return of spontaneous circulation (770/1135 [68%] vs 771/1135 [68%]; RR, 1.00 [95% CI, 0.95-1.06]; P = .96) or favorable neurologic outcome (185/987 [19%] vs 211/983 [21%]; RR, 0.87 [95% CI, 0.75-1.02]; P = .08) between those intubated and not intubated. The association between intubation and decreased survival was observed in the majority of the sensitivity and subgroup analyses, including when accounting for missing data and in a subgroup of patients with a pulse at the beginning of the event. Conclusions and Relevance Among pediatric patients with in-hospital cardiac arrest, tracheal intubation during cardiac arrest compared with no intubation was associated with decreased survival to hospital discharge. Although the study design does not eliminate the potential for confounding, these findings do not support the current emphasis on early tracheal intubation for pediatric in-hospital cardiac arrest.
160 citations
Cites background from "Extension of the modified Poisson r..."
...Never 321 (43) 391 (25) Initial pulseless rhythmf...
[...]
TL;DR: Among children with in-hospital cardiac arrest with an initial nonshockable rhythm who received epinephrine, delay in administration of epine dopamine was associated with decreased chance of survival to hospital discharge, ROSC, 24-hour survival, and survival toospital discharge with a favorable neurological outcome.
Abstract: Importance Delay in administration of the first epinephrine dose is associated with decreased survival among adults after in-hospital, nonshockable cardiac arrest. Whether this association is true in the pediatric in-hospital cardiac arrest population remains unknown. Objective To determine whether time to first epinephrine dose is associated with outcomes in pediatric in-hospital cardiac arrest. Design, Setting. and Participants We performed an analysis of data from the Get With the Guidelines–Resuscitation registry. We included US pediatric patients (age Exposure Time to epinephrine, defined as time in minutes from recognition of loss of pulse to the first dose of epinephrine. Main Outcomes and Measures The primary outcome was survival to hospital discharge. Secondary outcomes included return of spontaneous circulation (ROSC), survival at 24 hours, and neurological outcome. A favorable neurological outcome was defined as a score of 1 to 2 on the Pediatric Cerebral Performance Category scale. Results Among the 1558 patients, 487 (31.3%) survived to hospital discharge. The median time to first epinephrine dose was 1 minute (IQR, 0-4; range, 0-20; mean [SD], 2.6 [3.4] minutes). Longer time to epinephrine administration was associated with lower risk of survival to discharge in multivariable analysis (multivariable-adjusted risk ratio [RR] per minute delay, 0.95 [95% CI, 0.93-0.98]). Longer time to epinephrine administration was also associated with decreased risk of ROSC (multivariable-adjusted RR per minute delay, 0.97 [95% CI, 0.96-0.99]), decreased risk of survival at 24 hours (multivariable-adjusted RR per minute delay, 0.97 [95% CI, 0.95-0.99]), and decreased risk of survival with favorable neurological outcome (multivariable-adjusted RR per minute delay, 0.95 [95% CI, 0.91-0.99]). Patients with time to epinephrine administration of longer than 5 minutes (233/1558) compared with those with time to epinephrine of 5 minutes or less (1325/1558) had lower risk of in-hospital survival to discharge (21.0% [49/233] vs 33.1% [438/1325]; multivariable-adjusted RR, 0.75 [95% CI, 0.60-0.93]; P = .01). Conclusions and Relevance Among children with in-hospital cardiac arrest with an initial nonshockable rhythm who received epinephrine, delay in administration of epinephrine was associated with decreased chance of survival to hospital discharge, ROSC, 24-hour survival, and survival to hospital discharge with a favorable neurological outcome.
152 citations
Additional excerpts
...Time of day of arrest Nighttimed 457 (29) 118 (24) 339 (32) ....
[...]
...Neonate, <1 mo 421 (27) 117 (24) 304 (28)...
[...]
...001 Infant, 1 mo-<1 y 406 (26) 147 (30) 259 (24)...
[...]
TL;DR: The ideal cardiovascular health concept is beneficially associated with vascular health already in adolescence, supporting the relevance of targeting these metrics as part of primordial prevention.
Abstract: Background—In the Special Turku Coronary Risk Factor Intervention Project for Children (STRIP) study, repeated dietary counseling introduced in infancy and maintained until 20 years of age has led ...
150 citations
TL;DR: The developments of the past 13 years in design with a companion article to focus on developments in analysis are highlighted, including constrained randomization and a range of randomized designs that are alternatives to the standard parallel-arm GRT.
Abstract: In 2004, Murray et al. reviewed methodological developments in the design and analysis of group-randomized trials (GRTs). We have updated that review with developments in analysis of the past 13 years, with a companion article to focus on developments in design.We discuss developments in the topics of the earlier review (e.g., methods for parallel-arm GRTs, individually randomized group-treatment trials, and missing data) and in new topics, including methods to account for multiple-level clustering and alternative estimation methods (e.g., augmented generalized estimating equations, targeted maximum likelihood, and quadratic inference functions).In addition, we describe developments in analysis of alternative group designs (including stepped-wedge GRTs, network-randomized trials, and pseudocluster randomized trials), which require clustering to be accounted for in their design and analysis.
145 citations
References
More filters
TL;DR: In this article, an extension of generalized linear models to the analysis of longitudinal data is proposed, which gives consistent estimates of the regression parameters and of their variance under mild assumptions about the time dependence.
Abstract: SUMMARY This paper proposes an extension of generalized linear models to the analysis of longitudinal data. We introduce a class of estimating equations that give consistent estimates of the regression parameters and of their variance under mild assumptions about the time dependence. The estimating equations are derived without specifying the joint distribution of a subject's observations yet they reduce to the score equations for multivariate Gaussian outcomes. Asymptotic theory is presented for the general class of estimators. Specific cases in which we assume independence, m-dependence and exchangeable correlation structures from each subject are discussed. Efficiency of the proposed estimators in two simple situations is considered. The approach is closely related to quasi-likelih ood. Some key ironh: Estimating equation; Generalized linear model; Longitudinal data; Quasi-likelihood; Repeated measures.
17,111 citations
Additional excerpts
...Note further that it is also possible to improve the statistical efficiency of bb using a working correlation matrix other than the independence.(17) However, the modified Poisson modelling approach discussed here represents not only the use of a working correlation structure as commonly seen in practice,(22,24) but also the use of a working marginal Poisson distribution for binary data....
[...]
TL;DR: Results from a limited simulation study indicate that this approach is very reliable even with total sample sizes as small as 100, and the method is illustrated with two data sets.
Abstract: Relative risk is usually the parameter of interest in epidemiologic and medical studies. In this paper, the author proposes a modified Poisson regression approach (i.e., Poisson regression with a robust error variance) to estimate this effect measure directly. A simple 2-by-2 table is used to justify the validity of this approach. Results from a limited simulation study indicate that this approach is very reliable even with total sample sizes as small as 100. The method is illustrated with two data sets.
7,045 citations
"Extension of the modified Poisson r..." refers background in this paper
...This approach has become a popular alternative to the logistic regression model.(2) The chief advantage of the modified Poisson model is that it directly estimates the relative risk (i....
[...]
01 Jan 1967
TL;DR: In this paper, the authors prove consistency and asymptotic normality of maximum likelihood estimators under weaker conditions than usual, such that the true distribution underlying the observations belongs to the parametric family defining the estimator, and the regularity conditions do not involve the second and higher derivatives of the likelihood function.
Abstract: This paper proves consistency and asymptotic normality of maximum likelihood (ML) estimators under weaker conditions than usual. In particular, (i) it is not assumed that the true distribution underlying the observations belongs to the parametric family defining the ML estimator, and (ii) the regularity conditions do not involve the second and higher derivatives of the likelihood function. The need for theorems on asymptotic normality of ML estimators subject to (i) and (ii) becomes apparent in connection with robust estimation problems; for instance, if one tries to extend the author's results on robust estimation of a location parameter [4] to multivariate and other more general estimation problems. Wald's classical consistency proof [6] satisfies (ii) and can easily be modified to show that the ML estimator is consistent also in case (i), that is, it converges to the 0o characterized by the property E(logf(x, 0) log f(x, O0)) < 0 for 0 . Oo, where the expectation is taken with respect to the true underlying distribution. Asymptotic normality is more troublesome. Daniels [1] proved asymptotic normality subject to (ii), but unfortunately he overlooked that a crucial step in his proof (the use of the central limit theorem in (4.4)) is incorrect without condition (2.2) of Linnik [5]; this condition seems to be too restrictive for many purposes. In section 4 we shall prove asymptotic normality, assuming that the ML estimator is consistent. For the sake of completeness, sections 2 and 3 contain, therefore, two different sets of sufficient conditions for consistency. Otherwise, these sections are independent of each other. Section 5 presents two examples.
5,339 citations
"Extension of the modified Poisson r..." refers methods in this paper
...As previously shown, 2 the use of the Poisson model for binary data represents a model mis-specification which can be accounted for by applying the sandwich variance estimator approach.(1,20) Under mild regularity conditions, the estimated model coefficients are still asymptotically normally distributed....
[...]
TL;DR: Small worked examples and one real data set are used to help end-users appreciate the essence of the GEE method and allow nonstatisticians to imagine the calculations involved when the Gee method is applied to more complex multivariate data.
Abstract: The method of generalized estimating equations (GEE) is often used to analyze longitudinal and other correlated response data, particularly if responses are binary. However, few descriptions of the method are accessible to epidemiologists. In this paper, the authors use small worked examples and one real data set, involving both binary and quantitative response data, to help end-users appreciate the essence of the method. The examples are simple enough to see the behind-the-scenes calculations and the essential role of weighted observations, and they allow nonstatisticians to imagine the calculations involved when the GEE method is applied to more complex multivariate data.
1,999 citations
"Extension of the modified Poisson r..." refers methods in this paper
...Empirical results from the limited simulation study described below also suggest that this approach is as reliable as that obtained using binomial regression with a log link for clustered data.(22)...
[...]
...As a first example, we consider data consisting of age and sex-standardized height, expressed as number of standard deviations above US age- and sex-specific norms, of 144 children from 54 randomly selected families in Mexico.22 It is of interest here to study the relationship between family socioeconomic status and whether the child height standardized score is less than 1....
[...]
...As a first example, we consider data consisting of age and sex-standardized height, expressed as number of standard deviations above US age- and sex-specific norms, of 144 children from 54 randomly selected families in Mexico.(22) It is of interest here to study the relationship between family socioeconomic status and whether the child height standardized score is less than 1....
[...]