Imputation strategies for missing continuous outcomes in cluster randomized trials

Assessing treatment effect heterogeneity in the presence of missing effect modifier data in cluster-randomized trials

Understanding whether and how treatment effects vary across subgroups is crucial to inform clinical practice and recommendations. Accordingly, the assessment of heterogeneous treatment effects based on pre-specified potential effect modifiers has become a common goal in modern randomized trials. However, when one or more potential effect modifiers are missing, complete-case analysis may lead to bias and under-coverage. While statistical methods for handling missing data have been proposed and compared for individually randomized trials with missing effect modifier data, few guidelines exist for the cluster-randomized setting, where intracluster correlations in the effect modifiers, outcomes, or even missingness mechanisms may introduce further threats to accurate assessment of heterogeneous treatment effect. In this article, the performance of several missing data methods are compared through a simulation study of cluster-randomized trials with continuous outcome and missing binary effect modifier data, and further illustrated using real data from the Work, Family, and Health Study. Our results suggest that multilevel multiple imputation and Bayesian multilevel multiple imputation have better performance than other available methods, and that Bayesian multilevel multiple imputation has lower bias and closer to nominal coverage than standard multilevel multiple imputation when there are model specification or compatibility issues.

End-of-life care costs and place of death across health and social care sectors

OBJECTIVES

This study explores the relationship between end-of-life care costs and place of death across different health and social care sectors.

METHODS

We used a linked local government and health data of East London residents (n=4661) aged 50 or over, deceased between 2016 and 2020. Individuals who died in hospital were matched to those who died elsewhere according to a wide range of demographic, socioeconomic and health factors. We reported mean healthcare costs and 95% CIs by care sectors over the 12-month period before death. Subgroup analyses were conducted to investigate if the role of place of death differs according to long-term conditions and age.

RESULTS

We found that mean difference in total cost between hospital and non-hospital decedents was £4565 (95% CI £3132 to £6046). Hospital decedents were associated with higher hospital cost (£5196, £4499 to £5905), higher mental healthcare cost (£283, £78 to £892) and lower social care cost (-£838, -£1,209 to -£472), compared with individuals who died elsewhere. Subgroup analysis shows that the association between place of death and healthcare costs differs by age and long-term conditions, including cancer, mental health and cardiovascular diseases.

CONCLUSION

This study suggests that trajectories of end-of-life healthcare costs vary by place of death in a differential way across health and social care sectors. High hospital burden for cancer patients may be alleviated by strengthening healthcare provision in less cost-intensive settings, such as community and social care.

Collaborating to Improve Neonatal Care: ParentAl Participation on the NEonatal Ward-Study Protocol of the neoPARTNER Study

Parents are often appointed a passive role in the care for their hospitalised child. In the family-integrated care (FICare) model, parental involvement in neonatal care is emulated. Parental participation in medical rounds, or family-centred rounds (FCR), forms a key element. A paucity remains of randomised trials assessing the outcomes of FCR (embedded in FICare) in families and neonates, and outcomes on an organisational level are relatively unexplored. Likewise, biological mechanisms through which a potential effect may be exerted are lacking robust evidence. Ten level two Dutch neonatal wards are involved in this stepped-wedge cluster-randomised trial FCR (embedded in FICare) by one common implementation strategy. Parents of infants hospitalised for at least 7 days are eligible for inclusion. The primary outcome is parental stress (PSS:NICU) at discharge. Secondary outcomes include parental, neonatal, healthcare professional and organisational outcomes. Biomarkers of stress will be analysed in parent-infant dyads. With a practical approach and broad outcome set, this study aims to obtain evidence on the possible (mechanistic) effect of FCR (as part of FICare) on parents, infants, healthcare professionals and organisations. The practical approach provides (experiences of) FICare material adjusted to the Dutch setting, available for other hospitals after the study.

Avoidance Self-Efficacy: Personal Indicators of Risky Sex and Substance Use among At-Risk Youth

Adolescent substance use (SU) is associated with risky sex behavior and sexually transmitted infections and is a risk factor for subsequent risky sex decisions. Based on a sample of 1,580 youth in residential SU treatment, this study investigated how a static factor (race) and two dynamic personal factors (risk-taking, assertiveness) contributed to adolescents' perceived ability to avoid high-risk SU and sex behavior (avoidance self-efficacy). Results showed that race correlated with risk-taking and assertiveness, with White youth reporting higher ratings of assertiveness and risk-taking. Self-reported assertiveness and risk-taking also predicted SU and risky sex avoidance. This study underscores the importance of race and personal factors in relation to adolescents' confidence in avoiding high-risk situations.

Masitinib for mild-to-moderate Alzheimer's disease: results from a randomized, placebo-controlled, phase 3, clinical trial

BACKGROUND

Masitinib is an orally administered tyrosine kinase inhibitor that targets activated cells of the neuroimmune system (mast cells and microglia). Study AB09004 evaluated masitinib as an adjunct to cholinesterase inhibitor and/or memantine in patients with mild-to-moderate dementia due to probable Alzheimer's disease (AD).

METHODS

Study AB09004 was a randomized, double-blind, two parallel-group (four-arm), placebo-controlled trial. Patients aged ≥50 years, with clinical diagnosis of mild-to-moderate probable AD and a Mini-Mental State Examination (MMSE) score of 12-25 were randomized (1:1) to receive masitinib 4.5 mg/kg/day (administered orally as two intakes) or placebo. A second, independent parallel group (distinct for statistical analysis and control arm), randomized patients (2:1) to masitinib at an initial dose of 4.5 mg/kg/day for 12 weeks that was then titrated to 6.0 mg/kg/day, or equivalent placebo. Multiple primary outcomes (each tested at a significance level of 2.5%) were least-squares mean change from baseline to week 24 in the Alzheimer's Disease Assessment Scale - cognitive subscale (ADAS-cog), or the Alzheimer's Disease Cooperative Study Activities of Daily Living Inventory scale (ADCS-ADL). Safety for each masitinib dose level was compared against a pooled placebo population.

RESULTS

Masitinib (4.5 mg/kg/day) (n=182) showed significant benefit over placebo (n=176) according to the primary endpoint of ADAS-cog, -1.46 (95% CI [-2.46, -0.45]) (representing an overall improvement in cognition) versus 0.69 (95% CI [-0.36, 1.75]) (representing increased cognitive deterioration), respectively, with a significant between-group difference of -2.15 (97.5% CI [-3.48, -0.81]); p<0.001. For the ADCS-ADL primary endpoint, the between-group difference was 1.82 (97.5% CI [-0.15, 3.79]); p=0.038 (i.e., 1.01 (95% CI [-0.48, 2.50]) (representing an overall functional improvement) versus -0.81 (95% CI [-2.36, 0.74]) (representing increased functional deterioration), respectively). Safety was consistent with masitinib's known profile (maculo-papular rash, neutropenia, hypoalbuminemia). Efficacy results from the independent parallel group of titrated masitinib 6.0 mg/kg/day versus placebo (n=186 and 91 patients, respectively) were inconclusive and no new safety signal was observed.

CONCLUSIONS

Masitinib (4.5 mg/kg/day) may benefit people with mild-to-moderate AD. A confirmatory study has been initiated to substantiate these data.

TRIAL REGISTRATION

EudraCT: 2010-021218-50.

CLINICALTRIALS

gov : NCT01872598.

Multiple imputation for handling missing outcome data in randomized trials involving a mixture of independent and paired data

Randomized trials involving independent and paired observations occur in many areas of health research, for example in paediatrics, where studies can include infants from both single and twin births. Multiple imputation (MI) is often used to address missing outcome data in randomized trials, yet its performance in trials with independent and paired observations, where design effects can be less than or greater than one, remains to be explored. Using simulated data and through application to a trial dataset, we investigated the performance of different methods of MI for a continuous or binary outcome when followed by analysis using generalized estimating equations to account for clustering due to the pairs. We found that imputing data separately for independent and paired data, with paired data imputed in wide format, was the best performing MI method, producing unbiased point and standard error estimates for the treatment effect throughout. Ignoring clustering in the imputation model performed well in settings where the design effect due to the inclusion of paired data was close to one, but otherwise led to moderately biased variance estimates. Including a random cluster effect in the imputation model led to slightly biased point estimates for binary outcome data and variance estimates that were too small in some settings. Based on these results, we recommend researchers impute independent and paired data separately where feasible to do so. The exception is if the design effect due to the inclusion of paired data is close to one, where ignoring clustering may be appropriate.

Interaction of Neighborhood and Genetic Risk on Waist Circumference in African-American Adults: A Longitudinal Study

BACKGROUND

Understanding determinants of metabolic risk has become a national priority given the increasingly high prevalence rate of this condition among U.S. adults.

PURPOSE

This study's aim was to assess the impact of gene-by-neighborhood social environment interactions on waist circumference (WC) as a primary marker of metabolic risk in underserved African-American adults. Based on a dual-risk model, it was hypothesized that those with the highest genetic risk and who experienced negative neighborhood environment conditions would demonstrate higher WC than those with fewer risk factors.

METHODS

This study utilized a subsample of participants from the Positive Action for Today's Health environmental intervention to improve access and safety for walking in higher-crime neighborhoods, who were willing to provide buccal swab samples for genotyping stress-related genetic pathways. Assessments were conducted with 228 African-American adults at baseline, 12, 18, and 24 months.

RESULTS

Analyses indicated three significant gene-by-environment interactions on WC outcomes within the sympathetic nervous system (SNS) genetic pathway. Two interactions supported the dual-risk hypotheses, including the SNS genetic risk-by-neighborhood social life interaction (b = -0.11, t(618) = -2.02, p = .04), and SNS genetic risk-by-informal social control interaction (b = -0.51, t(618) = -1.95, p = .05) on WC outcomes. These interactions indicated that higher genetic risk and lower social-environmental supports were associated with higher WC. There was also one significant SNS genetic risk-by-neighborhood satisfaction interaction (b = 1.48, t(618) = 2.23, p = .02) on WC that was inconsistent with the dual-risk pattern.

CONCLUSIONS

Findings indicate that neighborhood and genetic factors dually influence metabolic risk and that these relations may be complex and warrant further study.

TRIAL REGISTRATION

NCT01025726.

Design and sample characteristics of COordinated Oral health Promotion (CO-OP) Chicago: A cluster-randomized controlled trial

COordinated Oral health Promotion (CO-OP) Chicago is a two-arm cluster-randomized trial with a wait-list control. The primary aim is to evaluate the efficacy of an oral health community health worker (CHW) intervention to improve oral health behaviors in low-income, urban children under the age of three years. Exploratory aims will determine cost-effectiveness, and if any CHW intervention impact on child tooth brushing behaviors varies when CHWs are based out of a medical clinic compared to a community setting. This paper describes progress toward achieving these aims. Participating families were recruited from community social service centers and pediatric primary care medical clinics in Cook County, Illinois. Sites were cluster-randomized to CHW intervention or usual services (a wait-list control). The intervention is oral health support from CHWs delivered in four visits to individual families over one year. The trial sample consists of 420 child/caregiver dyads enrolled at the 20 participating sites over 11 months. Participant demographics varied across the sites, but primary outcomes values at baseline did not. Data on brushing frequency, plaque, and other oral health behaviors are collected at three timepoints: baseline, 6-, and 12-months. The primary analysis will assess differences in caregiver-reported child brushing frequency and observed plaque score between the two arms at 12-months. The trial is currently in the active intervention phase. The trial's cluster-randomized controlled design takes a real-world approach by integrating into existing health and social service agencies and collecting data in participant homes. Results will address an important child health disparity. ClinicalTrials.gov identifier: NCT03397589. CLINICAL TRIAL REGISTRATION: University of Illinois at Chicago Protocol Record 2017-1090. National Institutes of Dental & Craniofacial Research of the National Institutes of Health (NIDCR) Protocol Number: 17-074-E. NCT03397589.

Can autonomy support have an effect on type 2 diabetes glycemic control? Results of a cluster randomized controlled trial

OBJECTIVES

To assess whether social support or autonomy support intervention for patients with type 2 diabetes can achieve glycemic control at the end of intervention, and to test whether the glycemic control effect can be maintained for a long time.

RESEARCH DESIGN AND METHODS

In this cluster randomized controlled trial, 18 community healthcare stations (CHSs) were randomized to the following: (1) usual care group (UCG) offering regular public health management services, (2) social support group (SSG) providing 3-month social support intervention based on problem solving principles, and (3) autonomy support group (ASG) offering 3-month autonomy support intervention based on self-determination theory. A total of 364 patients registered in the CHSs were enrolled into either of the three groups. The primary outcome was hemoglobin A1c (HbA1c), and secondary outcomes were diabetes self-management (DSM) behaviors. Assessment was conducted at baseline and at 3 and 6 months.

RESULTS

Patients in ASG achieved better HbA1c reduction at the end of intervention (0.53% or 7.23 mmol/mol, p<0.001) than those in the UCG and successfully maintained it up to 6 months (0.42% or 5.41 mmol/mol, p<0.001). However, patients in SSG did not experience significant change in HbA1c at 3 or 6 months when compared with patients in UCG. Besides, patients in both the SSG (0.12, p<0.05) and ASG (0.22, p<0.001) experienced improvement in exercise at 3 months. Patients in ASG sustained improvement in exercise up to 6 months (0.21, p<0.001), but those in the SSG did not.

CONCLUSIONS

Autonomy support for patients with type 2 diabetes could help achieve glycemic control at the end of intervention and successfully maintain it up to 6 months. These findings indicate that autonomy support has positive long-term effects on DSM behaviors and glycemic control and can be recommended in future diabetes intervention programs.

TRIAL REGISTRATION NUMBER

ChiCTR1900024354.

Multiple imputation by predictive mean matching in cluster-randomized trials

Background

Random effects regression imputation has been recommended for multiple imputation (MI) in cluster randomized trials (CRTs) because it is congenial to analyses that use random effects regression. This method relies heavily on model assumptions and may not be robust to misspecification of the imputation model. MI by predictive mean matching (PMM) is a semiparametric alternative, but current software for multilevel data relies on imputation models that ignore clustering or use fixed effects for clusters. When used directly for imputation, these two models result in underestimation (ignoring clustering) or overestimation (fixed effects for clusters) of variance estimates.

Methods

We develop MI procedures based on PMM that leverage these opposing estimated biases in the variance estimates in one of three ways: weighting the distance metric (PMM-dist), weighting the average of the final imputed values from two PMM procedures (PMM-avg), or performing a weighted draw from the final imputed values from the two PMM procedures (PMM-draw). We use Monte-Carlo simulations to evaluate our newly proposed methods relative to established MI procedures, focusing on estimation of treatment group means and their variances after MI.

Results

The proposed PMM procedures reduce the bias in the MI variance estimator relative to established methods when the imputation model is correctly specified, and are generally more robust to model misspecification than even the random effects imputation methods.

Conclusions

The PMM-draw procedure in particular is a promising method for multiply imputing missing data from CRTs that can be readily implemented in existing statistical software.

A Comparison of Missing-Data Imputation Techniques in Exploratory Factor Analysis

BACKGROUND AND PURPOSE

To compare the effects of missing-data imputation techniques, mean imputation, group mean imputation, regression imputation, and multiple imputation (MI), on the results of exploratory factor analysis under different missing assumptions.

METHODS

Missing data with different missing assumptions were generated from true data. The quality of imputed data was examined by correlation coefficients. Factor structures were compared indirectly by coefficients of congruence and directly by factor structures.

RESULTS

MI had the best quality and matching factor structure to the true data for all missing assumptions with different missing rates. Mean imputation had the least favorable results in factor analysis. The imputation techniques revealed no important differences with 10% of data missing.

CONCLUSION

MI showed the best results, especially with larger proportions of missing data.

Properties and pitfalls of weighting as an alternative to multilevel multiple imputation in cluster randomized trials with missing binary outcomes under covariate-dependent missingness

The generalized estimating equation (GEE) approach can be used to analyze cluster randomized trial data to obtain population-averaged intervention effects. However, most cluster randomized trials have some missing outcome data and a GEE analysis of available data may be biased when outcome data are not missing completely at random. Although multilevel multiple imputation for GEE (MMI-GEE) has been widely used, alternative approaches such as weighted GEE are less common in practice. Using both simulations and a real data example, we evaluate the performance of inverse probability weighted GEE vs. MMI-GEE for binary outcomes. Simulated data are generated assuming a covariate-dependent missing data pattern across a range of missingness clustering (from none to high), where all covariates are measured at baseline and are fully observed (i.e. a type of missing-at-random mechanism). Two types of weights are estimated and used in the weighted GEE: (1) assuming no clustering of missingness (W-GEE) and (2) accounting for such clustering (CW-GEE). Results show that, even in settings with high missingness clustering, CW-GEE can lead to more bias and lower coverage than W-GEE, whereas W-GEE and MMI-GEE provide comparable results. W-GEE should be considered a viable strategy to account for missing outcomes in cluster randomized trials.

Bias through selective inclusion and attrition: Representativeness when comparing provider performance with routine outcome monitoring data

Background

Observational research based on routine outcome monitoring is prone to missing data, and outcomes can be biased due to selective inclusion at baseline or selective attrition at posttest. As patients with complete data may not be representative of all patients of a provider, missing data may bias results, especially when missingness is not random but systematic.

Methods

The present study establishes clinical and demographic patient variables relevant for representativeness of the outcome information. It applies strategies to estimate sample selection bias (weighting by inclusion propensity) and selective attrition bias (multiple imputation based on multilevel regression analysis) and estimates the extent of their impact on an index of provider performance. The association between estimated bias and response rate is also investigated.

Results

Provider‐based analyses showed that in current practice, the effect of selective inclusion was minimal, but attrition had a more substantial effect, biasing results in both directions: overstating and understating performance. For 22% of the providers, attrition bias was estimated to be in excess of 0.05 ES. Bias was associated with overall response rate (r = .50). When selective inclusion and attrition bring providers' response below 50%, it is more likely that selection bias increased beyond a critical level, and conclusions on the comparative performance of such providers may be misleading.

Conclusions

Estimates of provider performance were biased by selection, especially by missing data at posttest. Results on the extent and direction of bias and minimal requirements for response rates to arrive at unbiased performance indicators are discussed.

Classroom-based physical activity improves children's math achievement - A randomized controlled trial

This RCT investigated the effect on children of integrating physical activity (PA) into math lessons. The primary outcome was math achievement and the secondary outcomes were executive functions, fitness and body mass index. Twelve Danish schools were randomized to either an intervention group or a control group. A total of 505 children with mean age 7.2 ± 0.3 years were enrolled in the study. Change in math achievement was measured by a 45-minute standardized math test, change in executive function by a modified Eriksen flanker task, aerobic fitness by the Andersen intermittent shuttle-run test, and body mass index by standard procedures. PA during the math lessons and total PA (including time spent outside school) were assessed using accelerometry (ActiGraph, GT3X and GT3X+). Children in the intervention group improved their math score by 1.2 (95% CI 0.3; 2.1) more than the control group (p = 0.011) and had a tendency towards a higher change in physical activity level during math lessons of 120,4 counts/min (95% CI -9.0;249.8.2, p = 0.067). However, the intervention did not affect executive functions, fitness or body mass index. Participation in a 9-month PA intervention (from 2012-2013) improved math achievement among elementary school children. If replicated, these findings would suggest that implementation of physical activity in school settings could lead to higher academic achievement.

A Comparison of Multilevel Imputation Schemes for Random Coefficient Models: Fully Conditional Specification and Joint Model Imputation with Random Covariance Matrices

Literature addressing missing data handling for random coefficient models is particularly scant, and the few studies to date have focused on the fully conditional specification framework and "reverse random coefficient" imputation. Although it has not received much attention in the literature, a joint modeling strategy that uses random within-cluster covariance matrices to preserve cluster-specific associations is a promising alternative for random coefficient analyses. This study is apparently the first to directly compare these procedures. Analytic results suggest that both imputation procedures can introduce bias-inducing incompatibilities with a random coefficient analysis model. Problems with fully conditional specification result from an incorrect distributional assumption, whereas joint imputation uses an underparameterized model that assumes uncorrelated intercepts and slopes. Monte Carlo simulations suggest that biases from these issues are tolerable if the missing data rate is 10% or lower and the sample is composed of at least 30 clusters with 15 observations per group. Furthermore, fully conditional specification tends to be superior with intraclass correlations that are typical of crosssectional data (e.g., ICC = .10), whereas the joint model is preferable with high values typical of longitudinal designs (e.g., ICC = .50).

A pattern-mixture model approach for handling missing continuous outcome data in longitudinal cluster randomized trials

We extend the pattern-mixture approach to handle missing continuous outcome data in longitudinal cluster randomized trials, which randomize groups of individuals to treatment arms, rather than the individuals themselves. Individuals who drop out at the same time point are grouped into the same dropout pattern. We approach extrapolation of the pattern-mixture model by applying multilevel multiple imputation, which imputes missing values while appropriately accounting for the hierarchical data structure found in cluster randomized trials. To assess parameters of interest under various missing data assumptions, imputed values are multiplied by a sensitivity parameter, k, which increases or decreases imputed values. Using simulated data, we show that estimates of parameters of interest can vary widely under differing missing data assumptions. We conduct a sensitivity analysis using real data from a cluster randomized trial by increasing k until the treatment effect inference changes. By performing a sensitivity analysis for missing data, researchers can assess whether certain missing data assumptions are reasonable for their cluster randomized trial.

Hierarchical imputation of systematically and sporadically missing data: An approximate Bayesian approach using chained equations

In health and medical sciences, multiple imputation (MI) is now becoming popular to obtain valid inferences in the presence of missing data. However, MI of clustered data such as multicenter studies and individual participant data meta-analysis requires advanced imputation routines that preserve the hierarchical structure of data. In clustered data, a specific challenge is the presence of systematically missing data, when a variable is completely missing in some clusters, and sporadically missing data, when it is partly missing in some clusters. Unfortunately, little is known about how to perform MI when both types of missing data occur simultaneously. We develop a new class of hierarchical imputation approach based on chained equations methodology that simultaneously imputes systematically and sporadically missing data while allowing for arbitrary patterns of missingness among them. Here, we use a random effect imputation model and adopt a simplification over fully Bayesian techniques such as Gibbs sampler to directly obtain draws of parameters within each step of the chained equations. We justify through theoretical arguments and extensive simulation studies that the proposed imputation methodology has good statistical properties in terms of bias and coverage rates of parameter estimates. An illustration is given in a case study with eight individual participant datasets.

An information, education and communication module to reduce dietary salt intake and blood pressure among tea garden workers of Assam

OBJECTIVE

High salt diet increases blood pressure. Tea garden workers (TGW) of Assam, India have high (60.8%) prevalence of hypertension (HTN), which may be due to consumption of extra salt (salt as side dish) and salted tea at work place and home. The present study evaluated an information, education and communication (IEC) module to reduce salt intake and blood pressure among TGW.

METHODS

Two tea gardens (usual care and intervention) were selected at random covering a total population of 13,458. The IEC module consisting of poster display, leaflets, health rally, documentary show, individual and group discussion was introduced in the intervention garden targeting study participants, health care providers, key stake holders, school children and teachers. IEC intervention was continued for one year. Participants from usual care and intervention were followed at three monthly intervals and BP and other information were compared after one year.

RESULTS

A total of 393 study participants (Non intervention: 194; intervention: 199) were included. After one year of follow up, consumption of extra salt was reduced significantly in the intervention participants (66.3 vs. 45.5%, p=0.000). Intention to treat analysis revealed significant reduction in systolic [-6.4 (-8.6 to -4.2)] and diastolic [-6.9 (-8.1 to -5.7)] blood pressure after one year. Prevalence of HTN was reduced significantly (52.5 vs. 40.0%, p=0.02) among them.

CONCLUSIONS

Our IEC module created awareness about risk of hypertension associated with high salt intake and could reduce dietary salt intake and BP.

Community-based participatory research to design a faith-enhanced diabetes prevention program: The Better Me Within randomized trial

Reducing obesity positively impacts diabetes and cardiovascular risk; however, evidence-based lifestyle programs, such as the diabetes prevention program (DPP), show reduced effectiveness in African American (AA) women. In addition to an attenuated response to lifestyle programs, AA women also demonstrate high rates of obesity, diabetes, and cardiovascular disease. To address these disparities, enhancements to evidence-based lifestyle programs for AA women need to be developed and evaluated with culturally relevant and rigorous study designs. This study describes a community-based participatory research (CBPR) approach to design a novel faith-enhancement to the DPP for AA women. A long-standing CBPR partnership designed the faith-enhancement from focus group data (N=64 AA adults) integrating five components: a brief pastor led sermon, memory verse, in class or take-home faith activity, promises to remember, and scripture and prayer integrated into participant curriculum and facilitator materials. The faith components were specifically linked to weekly DPP learning objectives to strategically emphasize behavioral skills with religious principles. Using a CBPR approach, the Better Me Within trial was able to enroll 12 churches, screen 333 AA women, and randomize 221 (M_age=48.8±11.2; M_BMI=36.7±8.4; 52% technical or high school) after collection of objective eligibility measures. A prospective, randomized, nested by church, design will be used to evaluate the faith-enhanced DPP as compared to a standard DPP on weight, diabetes and cardiovascular risk, over a 16-week intervention and 10-month follow up. This study will provide essential data to guide enhancements to evidence-based lifestyle programs for AA women who are at high risk for chronic disease.

Review of Recent Methodological Developments in Group-Randomized Trials: Part 2-Analysis

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.

Multiple imputation of missing covariate values in multilevel models with random slopes: a cautionary note

Multiple imputation (MI) has become one of the main procedures used to treat missing data, but the guidelines from the methodological literature are not easily transferred to multilevel research. For models including random slopes, proper MI can be difficult, especially when the covariate values are partially missing. In the present article, we discuss applications of MI in multilevel random-coefficient models, theoretical challenges posed by slope variation, and the current limitations of standard MI software. Our findings from three simulation studies suggest that (a) MI is able to recover most parameters, but is currently not well suited to capture slope variation entirely when covariate values are missing; (b) MI offers reasonable estimates for most parameters, even in smaller samples or when its assumptions are not met; and

Improving Care for Patients With or at Risk for Chronic Kidney Disease Using Electronic Medical Record Interventions: A Pragmatic Cluster-Randomized Trial Protocol

BACKGROUND

Many patients with or at risk for chronic kidney disease (CKD) in the primary care setting are not receiving recommended care.

OBJECTIVE

The objective of this study is to determine whether a multifaceted, low-cost intervention compared with usual care improves the care of patients with or at risk for CKD in the primary care setting.

DESIGN

A pragmatic cluster-randomized trial, with an embedded qualitative process evaluation, will be conducted.

SETTING

The study population comes from the Electronic Medical Record Administrative data Linked Database®, which includes clinical data for more than 140 000 rostered adults cared for by 194 family physicians in 34 clinics across Ontario, Canada. The 34 primary care clinics will be randomized to the intervention or control group.

INTERVENTION

The intervention group will receive resources from the "CKD toolkit" to help improve care including practice audit and feedback, printed educational materials for physicians and patients, electronic decision support and reminders, and implementation support.

MEASUREMENTS

Patients with or at risk for CKD within participating clinics will be identified using laboratory data in the electronic medical records. Outcomes will be assessed after dissemination of the CKD tools and after 2 rounds of feedback on performance on quality indicators have been sent to the physicians using information from the electronic medical records. The primary outcome is the proportion of patients aged 50 to 80 years with nondialysis-dependent CKD who are on a statin. Secondary outcomes include process of care measures such as screening tests, CKD recognition, monitoring tests, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker prescriptions, blood pressure targets met, and nephrologist referral. Hierarchical analytic modeling will be performed to account for clustering. Semistructured interviews will be conducted with a random purposeful sample of physicians in the intervention group to understand why the intervention achieved the observed effects.

CONCLUSIONS

If our intervention improves care, then the CKD toolkit can be adapted and scaled for use in other primary care clinics which use electronic medical records.

TRIAL REGISTRATION

ClinicalTrials.gov Identifier: NCT02274298.

Multiple imputation by chained equations for systematically and sporadically missing multilevel data

In multilevel settings such as individual participant data meta-analysis, a variable is 'systematically missing' if it is wholly missing in some clusters and 'sporadically missing' if it is partly missing in some clusters. Previously proposed methods to impute incomplete multilevel data handle either systematically or sporadically missing data, but frequently both patterns are observed. We describe a new multiple imputation by chained equations (MICE) algorithm for multilevel data with arbitrary patterns of systematically and sporadically missing variables. The algorithm is described for multilevel normal data but can easily be extended for other variable types. We first propose two methods for imputing a single incomplete variable: an extension of an existing method and a new two-stage method which conveniently allows for heteroscedastic data. We then discuss the difficulties of imputing missing values in several variables in multilevel data using MICE, and show that even the simplest joint multilevel model implies conditional models which involve cluster means and heteroscedasticity. However, a simulation study finds that the proposed methods can be successfully combined in a multilevel MICE procedure, even when cluster means are not included in the imputation models.

A comparison of imputation strategies in cluster randomized trials with missing binary outcomes

In cluster randomized trials, clusters of subjects are randomized rather than subjects themselves, and missing outcomes are a concern as in individual randomized trials. We assessed strategies for handling missing data when analysing cluster randomized trials with a binary outcome; strategies included complete case, adjusted complete case, and simple and multiple imputation approaches. We performed a simulation study to assess bias and coverage rate of the population-averaged intervention-effect estimate. Both multiple imputation with a random-effects logistic regression model or classical logistic regression provided unbiased estimates of the intervention effect. Both strategies also showed good coverage properties, even slightly better for multiple imputation with a random-effects logistic regression approach. Finally, this latter approach led to a slightly negatively biased intracluster correlation coefficient estimate but less than that with a classical logistic regression model strategy. We applied these strategies to a real trial randomizing households and comparing ivermectin and malathion to treat head lice.

Missing continuous outcomes under covariate dependent missingness in cluster randomised trials

Attrition is a common occurrence in cluster randomised trials which leads to missing outcome data. Two approaches for analysing such trials are cluster-level analysis and individual-level analysis. This paper compares the performance of unadjusted cluster-level analysis, baseline covariate adjusted cluster-level analysis and linear mixed model analysis, under baseline covariate dependent missingness in continuous outcomes, in terms of bias, average estimated standard error and coverage probability. The methods of complete records analysis and multiple imputation are used to handle the missing outcome data. We considered four scenarios, with the missingness mechanism and baseline covariate effect on outcome either the same or different between intervention groups. We show that both unadjusted cluster-level analysis and baseline covariate adjusted cluster-level analysis give unbiased estimates of the intervention effect only if both intervention groups have the same missingness mechanisms and there is no interaction between baseline covariate and intervention group. Linear mixed model and multiple imputation give unbiased estimates under all four considered scenarios, provided that an interaction of intervention and baseline covariate is included in the model when appropriate. Cluster mean imputation has been proposed as a valid approach for handling missing outcomes in cluster randomised trials. We show that cluster mean imputation only gives unbiased estimates when missingness mechanism is the same between the intervention groups and there is no interaction between baseline covariate and intervention group. Multiple imputation shows overcoverage for small number of clusters in each intervention group.

Multiple imputation methods for bivariate outcomes in cluster randomised trials

Missing observations are common in cluster randomised trials. The problem is exacerbated when modelling bivariate outcomes jointly, as the proportion of complete cases is often considerably smaller than the proportion having either of the outcomes fully observed. Approaches taken to handling such missing data include the following: complete case analysis, single‐level multiple imputation that ignores the clustering, multiple imputation with a fixed effect for each cluster and multilevel multiple imputation.

We contrasted the alternative approaches to handling missing data in a cost‐effectiveness analysis that uses data from a cluster randomised trial to evaluate an exercise intervention for care home residents.

We then conducted a simulation study to assess the performance of these approaches on bivariate continuous outcomes, in terms of confidence interval coverage and empirical bias in the estimated treatment effects. Missing‐at‐random clustered data scenarios were simulated following a full‐factorial design.

Across all the missing data mechanisms considered, the multiple imputation methods provided estimators with negligible bias, while complete case analysis resulted in biased treatment effect estimates in scenarios where the randomised treatment arm was associated with missingness. Confidence interval coverage was generally in excess of nominal levels (up to 99.8%) following fixed‐effects multiple imputation and too low following single‐level multiple imputation. Multilevel multiple imputation led to coverage levels of approximately 95% throughout. © 2016 The Authors. Statistics in Medicine Published by John Wiley & Sons Ltd.

Statistical analysis and handling of missing data in cluster randomized trials: a systematic review

Background

Cluster randomized trials (CRTs) randomize participants in groups, rather than as individuals and are key tools used to assess interventions in health research where treatment contamination is likely or if individual randomization is not feasible. Two potential major pitfalls exist regarding CRTs, namely handling missing data and not accounting for clustering in the primary analysis. The aim of this review was to evaluate approaches for handling missing data and statistical analysis with respect to the primary outcome in CRTs.

Methods

We systematically searched for CRTs published between August 2013 and July 2014 using PubMed, Web of Science, and PsycINFO. For each trial, two independent reviewers assessed the extent of the missing data and method(s) used for handling missing data in the primary and sensitivity analyses. We evaluated the primary analysis and determined whether it was at the cluster or individual level.

Results

Of the 86 included CRTs, 80 (93 %) trials reported some missing outcome data. Of those reporting missing data, the median percent of individuals with a missing outcome was 19 % (range 0.5 to 90 %). The most common way to handle missing data in the primary analysis was complete case analysis (44, 55 %), whereas 18 (22 %) used mixed models, six (8 %) used single imputation, four (5 %) used unweighted generalized estimating equations, and two (2 %) used multiple imputation. Fourteen (16 %) trials reported a sensitivity analysis for missing data, but most assumed the same missing data mechanism as in the primary analysis. Overall, 67 (78 %) trials accounted for clustering in the primary analysis.

Conclusions

High rates of missing outcome data are present in the majority of CRTs, yet handling missing data in practice remains suboptimal. Researchers and applied statisticians should carry out appropriate missing data methods, which are valid under plausible assumptions in order to increase statistical power in trials and reduce the possibility of bias. Sensitivity analysis should be performed, with weakened assumptions regarding the missing data mechanism to explore the robustness of results reported in the primary analysis.

Electronic supplementary material

The online version of this article (doi:10.1186/s13063-016-1201-z) contains supplementary material, which is available to authorized users.

The Results of the "Positive Action for Today's Health" (PATH) Trial for Increasing Walking and Physical Activity in Underserved African-American Communities

BACKGROUND

The "Positive Action for Today's Health" (PATH) trial tested an environmental intervention to increase walking in underserved communities.

METHODS

Three matched communities were randomized to a police-patrolled walking plus social marketing, a police-patrolled walking-only, or a no-walking intervention. The 24-month intervention addressed safety and access for physical activity (PA) and utilized social marketing to enhance environmental supports for PA. African-Americans (N=434; 62% females; aged 51±16 years) provided accelerometry and psychosocial measures at baseline and 12, 18, and 24 months. Walking attendance and trail use were obtained over 24 months.

RESULTS

There were no significant differences across communities over 24 months for moderate-to-vigorous PA. Walking attendance in the social marketing community showed an increase from 40 to 400 walkers per month at 9 months and sustained ~200 walkers per month through 24 months. No change in attendance was observed in the walking-only community.

CONCLUSIONS

Findings support integrating social marketing strategies to increase walking in underserved African-Americans (ClinicalTrials.gov #NCT01025726).

Evaluation of the Family Integrated Care model of neonatal intensive care: a cluster randomized controlled trial in Canada and Australia

Background

Admission to the neonatal intensive care unit (NICU) may disrupt parent-infant interaction with adverse consequences for infants and their families. Several family-centered care programs promote parent-infant interaction in the NICU; however, all of these retain the premise that health-care professionals should provide most of the infant’s care. Parents play a mainly supportive role in the NICU and continue to feel anxious and unprepared to care for their infant after discharge. In the Family Integrated Care (FICare) model, parents provide all except the most advanced medical care for their infants with support from the medical team. Our hypothesis is that infants whose families complete the FICare program will have greater weight gain and better clinical and parental outcomes compared with infants provided with standard NICU care.

Methods/Design

FICare is being evaluated in a cluster randomized controlled trial among infants born at ≤ 33 weeks’ gestation admitted to 19 Canadian, 6 Australian, and 1 New Zealand tertiary-level NICU. Trial enrollment began in April, 2013, with a target sample size of 675 infants in each arm, to be completed by August, 2015. Participating sites were stratified by country, and by NICU size within Canada, for randomization to either the FICare intervention or control arm. In intervention sites, parents are taught how to provide most of their infant’s care and supported by nursing staff, veteran parents, a program coordinator, and education sessions. In control sites standard NICU care is provided. The primary outcome is infants’ weight gain at 21 days after enrollment, which will be compared between the FICare and control groups using Student’s t-test adjusted for site-level clustering, and multi-level hierarchical models accounting for both clustering and potential confounders. Similar analyses will examine secondary outcomes including breastfeeding, clinical outcomes, safety, parental stress and anxiety, and resource use. The trial was designed, is being conducted, and will be reported according to the CONSORT 2010 guidelines for cluster randomized controlled trials.

Discussion

By evaluating the impact of integrating parents into the care of their infant in the NICU, this trial may transform the delivery of neonatal care.

Trial registration

NCT01852695, registered December 19, 2012

Neighborhood Social Predictors of Weight-related Measures in Underserved African Americans in the PATH Trial

African Americans have the highest rate of obesity in the United States relative to other ethnic minority groups. Bioecological factors including neighborhood social and physical environmental variables may be important predictors of weight-related measures specifically body mass index (BMI) in African American adults. Baseline data from the Positive Action for Today's Health (PATH) trial were collected from 417 African American adults. Overall a multiple regression model for BMI was significant, showing positive associations with average daily moderate-to-vigorous physical activity (MVPA) (B =-.21, P<.01) and neighborhood social interaction (B =-.13, P<.01). Consistent with previous literature, results show that neighborhood social interaction was associated with healthier BMI, highlighting it as a potential critical factor for future interventions in underserved, African American communities.

A longitudinal study of the effects of instrumental and emotional social support on physical activity in underserved adolescents in the ACT trial

BACKGROUND

Few previous studies have examined the influence of instrumental and emotional social support on physical activity (PA) longitudinally in underserved adolescents.

PURPOSE

This longitudinal study was a secondary analysis of the Active by Choice Today (ACT) trial examining whether instrumental social support predicts increases in PA in underserved adolescents, above and beyond emotional social support provided by family or peers.

METHODS

Students in the sixth grade (N = 1,422, 73 % African American, 54 % female, M age = 11 years) in the ACT trial participated. At baseline and 19 weeks, previously validated measures of social support (family instrumental, family emotional, and peer emotional) were completed and moderate-to-vigorous PA (MVPA) was assessed using 7-day accelerometry estimates.

RESULTS

A mixed ANCOVA demonstrated that baseline (p = 0.02) and change in family instrumental support (p = 0.01), but not emotional support from family or peers, predicted increases in MVPA across a 19-week period.

CONCLUSIONS

Future interventions in underserved adolescents should enhance opportunities for instrumental support for PA.

An overview of the Families Improving Together (FIT) for weight loss randomized controlled trial in African American families

BACKGROUND

The Families Improving Together (FIT) randomized controlled trial tests the efficacy of integrating cultural tailoring, positive parenting, and motivational strategies into a comprehensive curriculum for weight loss in African American adolescents. The overall goal of the FIT trial is to test the effects of an integrated intervention curriculum and the added effects of a tailored web-based intervention on reducing z-BMI in overweight African American adolescents.

DESIGN AND SETTING

The FIT trial is a randomized group cohort design the will involve 520 African American families with an overweight adolescent between the ages of 11-16 years. The trial tests the efficacy of an 8-week face-to-face group randomized program comparing M + FWL (Motivational Plus Family Weight Loss) to a comprehensive health education program (CHE) and re-randomizes participants to either an 8-week on-line tailored intervention or control on-line program resulting in a 2 (M + FWL vs. CHE group) × 2 (on-line intervention vs. control on-line program) factorial design to test the effects of the intervention on reducing z-BMI at post-treatment and at 6-month follow-up.

INTERVENTION

The interventions for this trial are based on a theoretical framework that is novel and integrates elements from cultural tailoring, Family Systems Theory, Self-Determination Theory and Social Cognitive Theory. The intervention targets positive parenting skills (parenting style, monitoring, communication); cultural values; teaching parents to increase youth motivation by encouraging youth to have input and choice (autonomy-support); and provides a framework for building skills and self-efficacy through developing weight loss action plans that target goal setting, monitoring, and positive feedback.

Challenges of cluster randomized trials

Cluster randomized trials are trials that randomize clusters of people, rather than individuals. They are becoming increasingly common. A number of innovations have been developed recently, particularly in the calculation of the required size of a cluster trial, the handling of missing data, designs to minimize recruitment bias, the ethics of cluster randomized trials and the stepped wedge design. This article will highlight and illustrate these developments. It will also discuss issues with regards to the reporting of cluster randomized trials.

Identifying Continence OptioNs after Stroke (ICONS): an evidence synthesis, case study and exploratory cluster randomised controlled trial of the introduction of a systematic voiding programme for patients with urinary incontinence after stroke in secondary care

Background

Urinary incontinence (UI) following acute stroke is common, affecting between 40% and 60% of people in hospital, but is often poorly managed.

Aim

To develop, implement and evaluate the preliminary effectiveness and potential cost-effectiveness of a systematic voiding programme (SVP), with or without supported implementation, for the management of UI after stroke in secondary care.

Design

Structured in line with the Medical Research Council framework for the evaluation of complex interventions, the programme comprised two phases: Phase I, evidence synthesis of combined approaches to manage UI post stroke, case study of the introduction of the SVP in one stroke service; Phase II, cluster randomised controlled exploratory trial incorporating a process evaluation and testing of health economic data collection methods.

Setting

One English stroke service (case study) and 12 stroke services in England and Wales (randomised trial).

Participants

Case study, 43 patients; randomised trial, 413 patients admitted to hospital with stroke and UI.

Interventions

A SVP comprising assessment, individualised conservative interventions and weekly review. In the supported implementation trial arm, facilitation was used as an implementation strategy to support and enable people to change their practice.

Main outcome measures

Participant incontinence (presence/absence) at 12 weeks post stroke. Secondary outcomes were quality of life, frequency and severity of incontinence, urinary symptoms, activities of daily living and death, at discharge, 6, 12 and 52 weeks post stroke.

Results

There was no suggestion of a beneficial effect on outcome at 12 weeks post stroke [intervention vs. usual care: odds ratio (OR) 1.02, 95% confidence interval (CI) 0.54 to 1.93; supported implementation vs. usual care: OR 1.06, 95% CI 0.54 to 2.09]. There was weak evidence of better outcomes on the Incontinence Impact Questionnaire in supported implementation (OR 1.22, 95% CI 0.72 to 2.08) but the CI is wide and includes both clinically relevant benefit and harm. Both intervention arms had a higher estimated odds of continence for patients with urge incontinence than usual care (intervention: OR 1.58, 95% CI 0.83 to 2.99; supported implementation: OR 1.73, 95% CI 0.88 to 3.43). The process evaluation showed that the SVP increased the visibility of continence management through greater evaluation of patients’ trajectories and outcomes, and closer attention to workload. In-hospital resource use had to be based on estimates provided by staff. The response rates for the postal questionnaires were 73% and 56% of eligible patients at 12 and 52 weeks respectively. Completion of individual data items varied between 67% and 100%.

Conclusions

The trial was exploratory and did not set out to establish effectiveness; however, there are indications the intervention may be effective in patients with urge and stress incontinence. A definitive trial is now warranted.

Study registration

This study is registered as ISRCTN08609907.

Funding details

The National Institute for Health Research Programme Grants for Applied Research programme. Excess treatment costs and research support costs were funded by participating NHS trusts and health boards, Lancashire and Cumbria and East Anglia Comprehensive Local Research Networks and the Welsh National Institute for Social Care and Health Research.

Accounting for multiple births in randomised trials: a systematic review

OBJECTIVES

Multiple births are an important subgroup to consider in trials aimed at reducing preterm birth or its consequences. Including multiples results in a unique mixture of independent and clustered data, which has implications for the design, analysis and reporting of the trial. We aimed to determine how multiple births were taken into account in the design and analysis of recent trials involving preterm infants, and whether key information relevant to multiple births was reported.

DESIGN

We conducted a systematic review of multicentre randomised trials involving preterm infants published between 2008 and 2013. Information relevant to multiple births was extracted.

RESULTS

Of the 56 trials included in the review, 6 (11%) excluded multiples and 24 (43%) failed to indicate whether multiples were included. Among the 26 trials that reported multiples were included, only one (4%) accounted for clustering in the sample size calculations and eight (31%) took the clustering into account in the analysis of the primary outcome. Of the 20 trials that randomised infants, 12 (60%) failed to report how infants from the same birth were randomised.

CONCLUSIONS

Information on multiple births is often poorly reported in trials involving preterm infants, and clustering due to multiple births is rarely taken into account. Since ignoring clustering could result in inappropriate recommendations for clinical practice, clustering should be taken into account in the design and analysis of future neonatal and perinatal trials including infants from a multiple birth.

Identifying continence options after stroke (ICONS): a cluster randomised controlled feasibility trial

Background

Urinary incontinence (UI) affects half of patients hospitalised after stroke and is often poorly managed. Cochrane systematic reviews have shown some positive impact of conservative interventions (such as bladder training) in reducing UI, but their effectiveness has not been demonstrated with stroke patients.

Methods

We conducted a cluster randomised controlled feasibility trial of a systematic voiding programme (SVP) for the management of UI after stroke. Stroke services were randomised to receive SVP (n = 4), SVP plus supported implementation (SVP+, n = 4), or usual care (UC, n = 4).

Feasibility outcomes were participant recruitment and retention. The main effectiveness outcome was presence or absence of UI at six and 12 weeks post-stroke. Additional effectiveness outcomes included were the effect of the intervention on different types of UI, continence status at discharge, UI severity, functional ability, quality of life, and death.

Results

It was possible to recruit patients (413; 164 SVP, 125 SVP+, and 124 UC) and participant retention was acceptable (85% and 88% at six and 12 weeks, respectively). There was no suggestion of a beneficial effect on the main outcome at six (SVP versus UC: odds ratio (OR) 0.94, 95% CI: 0.46 to 1.94; SVP+ versus UC: OR: 0.62, 95% CI: 0.28 to 1.37) or 12 weeks (SVP versus UC: OR: 1.02, 95% CI: 0.54 to 1.93; SVP+ versus UC: OR: 1.06, 95% CI: 0.54 to 2.09).

No secondary outcomes showed a strong suggestion of clinically meaningful improvement in SVP and/or SVP+ arms relative to UC at six or 12 weeks. However, at 12 weeks both intervention arms had higher estimated odds of continence than UC for patients with urge incontinence.

Conclusions

The trial has met feasibility outcomes of participant recruitment and retention. It was not powered to demonstrate effectiveness, but there is some evidence of a potential reduction in the odds of specific types of incontinence. A full trial should now be considered.

Trial registration

ISRCTN Registry, ISRCTN08609907, date of registration: 7 July 2010.

Electronic supplementary material

The online version of this article (doi:10.1186/1745-6215-15-509) contains supplementary material, which is available to authorized users.

Are missing data adequately handled in cluster randomised trials? A systematic review and guidelines

BACKGROUND

Missing data are a potential source of bias, and their handling in the statistical analysis can have an important impact on both the likelihood and degree of such bias. Inadequate handling of the missing data may also result in invalid variance estimation. The handling of missing values is more complex in cluster randomised trials, but there are no reviews of practice in this field.

OBJECTIVES

A systematic review of published trials was conducted to examine how missing data are reported and handled in cluster randomised trials.

METHODS

We systematically identified cluster randomised trials, published in English in 2011, using the National Library of Medicine (MEDLINE) via PubMed. Non-randomised and pilot/feasibility trials were excluded, as were reports of secondary analyses, interim analyses, and economic evaluations and those where no data were at the individual level. We extracted information on missing data and the statistical methods used to deal with them from a random sample of the identified studies.

RESULTS

We included 132 trials. There was evidence of missing data in 95 (72%). Only 32 trials reported handling missing data, 22 of them using a variety of single imputation techniques, 8 using multiple imputation without accommodating the clustering and 2 stating that their likelihood-based complete case analysis accounted for missing values because the data were assumed Missing-at-Random.

LIMITATIONS

The results presented in this study are based on a large random sample of cluster randomised trials published in 2011, identified in electronic searches and therefore possibly missing some trials, most likely of poorer quality. Also, our results are based on information in the main publication for each trial. These reports may omit some important information on the presence of, and reasons for, missing data and on the statistical methods used to handle them. Our extraction methods, based on published reports, could not distinguish between missing data in outcomes and missing data in covariates. This distinction may be important in determining the assumptions about the missing data mechanism necessary for complete case analyses to be valid.

CONCLUSIONS

Missing data are present in the majority of cluster randomised trials. However, they are poorly reported, and most authors give little consideration to the assumptions under which their analysis will be valid. The majority of the methods currently used are valid under very strong assumptions about the missing data, whose plausibility is rarely discussed in the corresponding reports. This may have important consequences for the validity of inferences in some trials. Methods which result in valid inferences under general Missing-at-Random assumptions are available and should be made more accessible.

Use of text messages to communicate clinical recommendations to health workers in rural China: a cluster-randomized trial

OBJECTIVE

To compare the effectiveness of mobile phone text messaging and that of traditional health worker training in communicating clinical recommendations to health workers in China.

METHODS

A cluster-randomized controlled trial (Chinese Clinical Trial Register: ChiCTR-TRC-09000488) was conducted in 100 township health centres in north-western China between 17 October and 25 December 2011. Health workers were allocated either to receive 16 text messages with recommendations on the management of viral infections affecting the upper respiratory tract and otitis media (intervention group, n = 490) or to receive the same recommendations through the existing continuing medical education programme - a one-day training workshop (control group, n = 487). Health workers' knowledge of the recommendations was assessed before and after messaging and traditional training through a multiple choice questionnaire. The percentage change in score in the control group was compared with that in the intervention group. Changes in prescribing practices were also compared.

FINDINGS

Health workers' knowledge of the recommendations increased significantly in the intervention group, both individually (0.17 points; 95% confidence interval, CI: 0.168-0.172) and at the cluster level (0.16 points; 95% CI: 0.157-0.163), but not in the control group. In the intervention group steroid prescriptions decreased by 5 percentage points but antibiotic prescriptions remained unchanged. In the control group, however, antibiotic and steroid prescriptions increased by 17 and 11 percentage points, respectively.

CONCLUSION

Text messages can be effective for transmitting medical information and changing health workers' behaviour, particularly in resource-limited settings.

Nonlinear multiple imputation for continuous covariate within semiparametric Cox model: application to HIV data in Senegal

Multiple imputation is commonly used to impute missing covariate in Cox semiparametric regression setting. It is to fill each missing data with more plausible values, via a Gibbs sampling procedure, specifying an imputation model for each missing variable. This imputation method is implemented in several softwares that offer imputation models steered by the shape of the variable to be imputed, but all these imputation models make an assumption of linearity on covariates effect. However, this assumption is not often verified in practice as the covariates can have a nonlinear effect. Such a linear assumption can lead to a misleading conclusion because imputation model should be constructed to reflect the true distributional relationship between the missing values and the observed values. To estimate nonlinear effects of continuous time invariant covariates in imputation model, we propose a method based on B-splines function. To assess the performance of this method, we conducted a simulation study, where we compared the multiple imputation method using Bayesian splines imputation model with multiple imputation using Bayesian linear imputation model in survival analysis setting. We evaluated the proposed method on the motivated data set collected in HIV-infected patients enrolled in an observational cohort study in Senegal, which contains several incomplete variables. We found that our method performs well to estimate hazard ratio compared with the linear imputation methods, when data are missing completely at random, or missing at random.

Multiple imputation methods for handling missing data in cost-effectiveness analyses that use data from hierarchical studies: an application to cluster randomized trials

PURPOSE

Multiple imputation (MI) has been proposed for handling missing data in cost-effectiveness analyses (CEAs). In CEAs that use cluster randomized trials (CRTs), the imputation model, like the analysis model, should recognize the hierarchical structure of the data. This paper contrasts a multilevel MI approach that recognizes clustering, with single-level MI and complete case analysis (CCA) in CEAs that use CRTs.

METHODS

We consider a multilevel MI approach compatible with multilevel analytical models for CEAs that use CRTs. We took fully observed data from a CEA that evaluated an intervention to improve diagnosis of active labor in primiparous women using a CRT (2078 patients, 14 clusters). We generated scenarios with missing costs and outcomes that differed, for example, according to the proportion with missing data (10%-50%), the covariates that predicted missing data (individual, cluster-level), and the missingness mechanism: missing completely at random (MCAR), missing at random (MAR), or missing not at random (MNAR). We estimated incremental net benefits (INBs) for each approach and compared them with the estimates from the fully observed data, the "true" INBs.

RESULTS

When costs and outcomes were assumed to be MCAR, the INBs for each approach were similar to the true estimates. When data were MAR, the point estimates from the CCA differed from the true estimates. Multilevel MI provided point estimates and standard errors closer to the true values than did single-level MI across all settings, including those in which a high proportion of observations had cost and outcome data MAR and when data were MNAR.

CONCLUSIONS

Multilevel MI accommodates the multilevel structure of the data in CEAs that use cluster trials and provides accurate cost-effectiveness estimates across the range of circumstances considered.

Comparison of population-averaged and cluster-specific models for the analysis of cluster randomized trials with missing binary outcomes: a simulation study

Abstracts

Person mobility in the design and analysis of cluster-randomized cohort prevention trials

Person mobility is an inescapable fact of life for most cluster-randomized (e.g., schools, hospitals, clinic, cities, state) cohort prevention trials. Mobility rates are an important substantive consideration in estimating the effects of an intervention. In cluster-randomized trials, mobility rates are often correlated with ethnicity, poverty and other variables associated with disparity. This raises the possibility that estimated intervention effects may generalize to only the least mobile segments of a population and, thus, create a threat to external validity. Such mobility can also create threats to the internal validity of conclusions from randomized trials. Researchers must decide how to deal with persons who leave study clusters during a trial (dropouts), persons and clusters that do not comply with an assigned intervention, and persons who enter clusters during a trial (late entrants), in addition to the persons who remain for the duration of a trial (stayers). Statistical techniques alone cannot solve the key issues of internal and external validity raised by the phenomenon of person mobility. This commentary presents a systematic, Campbellian-type analysis of person mobility in cluster-randomized cohort prevention trials. It describes four approaches for dealing with dropouts, late entrants and stayers with respect to data collection, analysis and generalizability. The questions at issue are: 1) From whom should data be collected at each wave of data collection? 2) Which cases should be included in the analyses of an intervention effect? and 3) To what populations can trial results be generalized? The conclusions lead to recommendations for the design and analysis of future cluster-randomized cohort prevention trials.

Arriba-lib: association of an evidence-based electronic library of decision aids with communication and decision-making in patients and primary care physicians

AIM

In shared decision-making, patients are empowered to actively ask questions and participate in decisions about their healthcare based on their preferences and values. Decision aids should help patients make informed choices among diagnostic or treatment options by delivering evidence-based information on options and outcomes; however, they have rarely been field tested, especially in the primary care context. We therefore evaluated associations between the use of an interactive, transactional and evidence-based library of decision aids (arriba-lib) and communication and decision-making in patients and physicians in the primary care context.

METHODS

Our electronic library of decision aids ('arriba-lib') includes evidence-based modules for cardiovascular prevention, diabetes, coronary heart disease, atrial fibrillation and depression. Twenty-nine primary care physicians recruited 192 patients. We used questionnaires to ask patients and physicians about their experiences with and attitudes towards the programme. Patients were interviewed via telephone 2 months after the consultation. Data were analysed by general estimation equations, cross tab analyses and by using effect sizes.

RESULTS

Only a minority (8.9%) of the consultations were felt to be too long because physicians said consultations were unacceptably extended by arriba-lib. We found a negative association between the detailedness of the discussion of the clinical problem's definition and the age of the patients. Physicians discuss therapeutic options in less detail with patients who have a formal education of less than 8 years. Patients who were counselled by a physician with no experience in using a decision aid more often reported that they do not remember being counselled with the help of a decision aid or do not wish to be counselled again with a decision aid.

CONCLUSIONS

Arriba-lib has positive associations to the decision-making process in patients and physicians. It can also be used with older age groups and patients with less formal education. Physicians seem to adapt their counselling strategy to different patient groups. Prior experience with the use of decision aids has an influence on the acceptance of arriba-lib in patients but not on their decision-making or decision implementation.

Arriba-lib: evaluation of an electronic library of decision aids in primary care physicians

Background

The successful implementation of decision aids in clinical practice initially depends on how clinicians perceive them. Relatively little is known about the acceptance of decision aids by physicians and factors influencing the implementation of decision aids from their point of view. Our electronic library of decision aids (arriba-lib) is to be used within the encounter and has a modular structure containing evidence-based decision aids for the following topics: cardiovascular prevention, atrial fibrillation, coronary heart disease, oral antidiabetics, conventional and intensified insulin therapy, and unipolar depression. The aim of our study was to evaluate the acceptance of arriba-lib in primary care physicians.

Methods

We conducted an evaluation study in which 29 primary care physicians included 192 patients. The physician questionnaire contained information on which module was used, how extensive steps of the shared decision making process were discussed, who made the decision, and a subjective appraisal of consultation length. We used generalised estimation equations to measure associations within patient variables and traditional crosstab analyses.

Results

Only a minority of consultations (8.9%) was considered to be unacceptably extended. In 90.6% of consultations, physicians said that a decision could be made. A shared decision was perceived by physicians in 57.1% of consultations. Physicians said that a decision was more likely to be made when therapeutic options were discussed “detailed”. Prior experience with decision aids was not a critical variable for implementation within our sample of primary care physicians.

Conclusions

Our study showed that it might be feasible to apply our electronic library of decision aids (arriba-lib) in the primary care context. Evidence-based decision aids offer support for physicians in the management of medical information. Future studies should monitor the long-term adoption of arriba-lib in primary care physicians.

Acceptance of shared decision making with reference to an electronic library of decision aids (arriba-lib) and its association to decision making in patients: an evaluation study

Background

Decision aids based on the philosophy of shared decision making are designed to help patients make informed choices among diagnostic or treatment options by delivering evidence-based information on options and outcomes. A patient decision aid can be regarded as a complex intervention because it consists of several presumably relevant components. Decision aids have rarely been field tested to assess patients' and physicians' attitudes towards them. It is also unclear what effect decision aids have on the adherence to chosen options.

Methods

The electronic library of decision aids (arriba-lib) to be used within the clinical encounter has a modular structure and contains evidence-based decision aids for the following topics: cardiovascular prevention, atrial fibrillation, coronary heart disease, oral antidiabetics, conventional and intensified insulin therapy, and unipolar depression. We conducted an evaluation study in which 29 primary care physicians included 192 patients. After the consultation, patients filled in questionnaires and were interviewed via telephone two months later. We used generalised estimation equations to measure associations within patient variables and traditional crosstab analyses.

Results

Patients were highly satisfied with arriba-lib and the process of shared decision making. Two-thirds of patients reached in the telephone interview wanted to be counselled again with arriba-lib. There was a high congruence between preferred and perceived decision making. Of those patients reached in the telephone interview, 80.7% said that they implemented the decision, independent of gender and education. Elderly patients were more likely to say that they implemented the decision.

Conclusions

Shared decision making with our multi-modular electronic library of decision aids (arriba-lib) was accepted by a high number of patients. It has positive associations to general aspects of decision making in patients. It can be used for patient groups with a wide range of individual characteristics.

Results of the "Active by Choice Today" (ACT) randomized trial for increasing physical activity in low-income and minority adolescents

OBJECTIVE

This study reports the results of the "Active by Choice Today" (ACT) trial for increasing moderate-to-vigorous physical activity (MVPA) in low-income and minority adolescents.

DESIGN

The ACT program was a randomized controlled school-based trial testing the efficacy of a motivational plus behavioral skills intervention on increasing MVPA in underserved adolescents. Twenty-four middle schools were matched on school size, percentage minorities, percentage free or reduce lunch, and urban or rural setting before randomization. A total of 1,563 6th grade students (mean age, 11.3 years, 73% African American, 71% free or reduced lunch, 55% female) participated in either a 17-week (over one academic year) intervention or comparison after-school program.

MAIN OUTCOME MEASURE

The primary outcome measure was MVPA based on 7-day accelerometry estimates at 2-weeks postintervention and an intermediate outcome was MVPA at midintervention.

RESULTS

At midintervention students in the intervention condition engaged in 4.87 greater minutes of MVPA per day (95% CI: 1.18 to 8.57) than control students. Students in intervention schools engaged in 9.11 min (95% CI: 5.73 to 12.48) more of MVPA per day than those in control schools during the program time periods; indicating a 27 min per week increase in MVPA. No significant effect of the ACT intervention was found outside of school times or for MVPA at 2-weeks postintervention.

CONCLUSIONS

Motivational and behavioral skills programs are effective at increasing MVPA in low-income and minority adolescents during program hours, but further research is needed to address home barriers to youth MVPA.

Study protocol: ICONS: identifying continence options after stroke: a randomised trial

BACKGROUND

Urinary incontinence following acute stroke is common, affecting between 40%-60% of people in hospital after a stroke. Despite the availability of clinical guidelines for urinary incontinence and urinary incontinence after stroke, national audit data suggest incontinence is often poorly managed. Conservative interventions (e.g. bladder training, pelvic floor muscle training and prompted voiding) have been shown to have some effect with participants in Cochrane systematic reviews, but have not had their effectiveness demonstrated with stroke patients.

METHODS/DESIGN

A cluster randomised controlled pilot trial designed to assess the feasibility of a full-scale cluster randomised trial and to provide preliminary evidence of the effectiveness and cost-effectiveness of a systematic voiding programme for the management of continence after stroke. Stroke services will be randomised to receive the systematic voiding programme, the systematic voiding programme plus supported implementation, or usual care. The trial aims to recruit at least 780 participants in 12 stroke services (4 per arm). The primary outcome is presence/absence of incontinence at six weeks post-stroke. Secondary outcomes include frequency and severity of incontinence, quality of life and cost-utility. Outcomes will be measured at six weeks, three months and (for participants recruited in the first three months) twelve months after stroke. Process data will include rates of recruitment and retention and fidelity of intervention delivery. An integrated qualitative evaluation will be conducted in order to describe implementation and assist in explaining the potential mediators and modifiers of the process.

TRIAL REGISTRATION

ISRCTN: ISRCTN08609907

High normal levels of albuminuria and risk of hypertension in Indo-Asian population

BACKGROUND

Urine albumin excretion in the high normal range [urine albumin to creatinine ratio (UACR) 10-29 mg/g)] predicts hypertension in European-origin populations. However, the prognostic significance of UACR in the high normal range for incident hypertension is unclear in Indo-Asians. The objective of this study was to examine the relationship of normal to high normal levels of UACR and incident hypertension.

METHODS

We conducted a nested cohort study within a cluster randomized controlled trial in Pakistan on 1272 normotensive non-diabetic adults aged ≥ 40 years with UACR <30 mg/g. Incident hypertension was defined as new onset of systolic blood pressure (SBP) ≥ 140 mmHg or diastolic ≥ 90 mmHg or initiation of antihypertensive therapy.

RESULTS

A total of 920 (72.3%) participants completed the 2-year final follow-up. During this time, 105 (11.4%) developed incident hypertension. In the multivariable model, the odds [95% confidence interval (CI)] for incident hypertension were 2.45(1.21-4.98) for those in the fourth (top) quartile (≥ 6.1 mg/g) and 2.12 (1.04-4.35) in the third quartile (3.8-6.1 mg/g) compared to those in the lowest quartile (<2.8 mg/g). In addition, a significant interaction between UACR and baseline SBP was observed suggesting that the odds (95% CI) of incident hypertension with UACR were greater at lower baseline SBP (interaction P = 0.044).

CONCLUSIONS

High normal levels of albuminuria as measured by UACR predict hypertension in non-diabetic Indo-Asians, and this relationship may be enhanced in individuals with low baseline SBP. Further research is needed to assess the clinical applicability of these findings.