Multilevel analysis of group-randomized trials with binary outcomes

Power considerations for generalized estimating equations analyses of four-level cluster randomized trials

In this article, we develop methods for sample size and power calculations in four-level intervention studies when intervention assignment is carried out at any level, with a particular focus on cluster randomized trials (CRTs). CRTs involving four levels are becoming popular in healthcare research, where the effects are measured, for example, from evaluations (level 1) within participants (level 2) in divisions (level 3) that are nested in clusters (level 4). In such multilevel CRTs, we consider three types of intraclass correlations between different evaluations to account for such clustering: that of the same participant, that of different participants from the same division, and that of different participants from different divisions in the same cluster. Assuming arbitrary link and variance functions, with the proposed correlation structure as the true correlation structure, closed-form sample size formulas for randomization carried out at any level (including individually randomized trials within a four-level clustered structure) are derived based on the generalized estimating equations approach using the model-based variance and using the sandwich variance with an independence working correlation matrix. We demonstrate that empirical power corresponds well with that predicted by the proposed method for as few as eight clusters, when data are analyzed using the matrix-adjusted estimating equations for the correlation parameters with a bias-corrected sandwich variance estimator, under both balanced and unbalanced designs.

Targeted HIV Screening in Eight Emergency Departments: The DICI-VIH Cluster-Randomized Two-Period Crossover Trial

STUDY OBJECTIVE

This study compares the effectiveness and cost-effectiveness of nurse-driven targeted HIV screening alongside physician-directed diagnostic testing (intervention strategy) with diagnostic testing alone (control strategy) in 8 emergency departments.

METHODS

In this cluster-randomized, 2-period, crossover trial, 18- to 64-year-old patients presenting for reasons other than potential exposure to HIV were included. The strategy applied first was randomly assigned. During both periods, diagnostic testing was prescribed by physicians following usual care. During the intervention periods, patients were asked to complete a self-administered questionnaire. According to their answers, the triage nurse suggested performing a rapid test to patients belonging to a high-risk group. The primary outcome was the proportion of new diagnoses among included patients, which further refers to effectiveness. A secondary outcome was the intervention's incremental cost (health care system perspective) per additional diagnosis.

RESULTS

During the intervention periods, 74,161 patients were included, 16,468 completed the questionnaire, 4,341 belonged to high-risk groups, and 2,818 were tested by nurses, yielding 13 new diagnoses. Combined with 9 diagnoses confirmed through 97 diagnostic tests, 22 new diagnoses were established. During the control periods, 74,166 patients were included, 92 were tested, and 6 received a new diagnosis. The proportion of new diagnoses among included patients was higher during the intervention than in the control periods (3.0 per 10,000 versus 0.8 per 10,000; difference 2.2 per 10,000, 95% CI 1.3 to 3.6; relative risk 3.7, 95% CI 1.4 to 9.8). The incremental cost was €1,324 per additional new diagnosis.

CONCLUSION

The combined strategy of targeted screening and diagnostic testing was effective.

Effectiveness of a blended web-based intervention on return to work for sick-listed employees with common mental disorders: results of a cluster randomized controlled trial

Background

Common mental disorders are strongly associated with long-term sickness absence, which has negative consequences for the individual employee’s quality of life and leads to substantial costs for society. It is important to focus on return to work (RTW) during treatment of sick-listed employees with common mental disorders. Factors such as self-efficacy and the intention to resume work despite having symptoms are important in the RTW process. We developed “E-health module embedded in Collaborative Occupational health care” (ECO) as a blended Web-based intervention with 2 parts: an eHealth module (Return@Work) for the employee aimed at changing cognitions of the employee regarding RTW and a decision aid via email supporting the occupational physician with advice regarding treatment and referral options based on monitoring the employee’s progress during treatment.

Objective

This study evaluated the effect of a blended eHealth intervention (ECO) versus care as usual on time to RTW of sick-listed employees with common mental disorders.

Methods

The study was a 2-armed cluster randomized controlled trial. Employees sick-listed between 4 and 26 weeks with common mental disorder symptoms were recruited by their occupational health service or employer. The employees were followed up to 12 months. The primary outcome measures were time to first RTW (partial or full) and time to full RTW. Secondary outcomes were response and remission of the common mental disorder symptoms (self-assessed).

Results

A total of 220 employees were included: 131 participants were randomized to the ECO intervention and 89 to care as usual (CAU). The duration until first RTW differed significantly between the groups. The median duration was 77.0 (IQR 29.0-152.3) days in the CAU group and 50.0 (IQR 20.8-99.0) days in the ECO group (hazard ratio [HR] 1.390, 95% CI 1.034-1.870, P=.03). No significant difference was found for duration until full RTW. Treatment response of common mental disorder symptoms did not differ significantly between the groups, but at 9 months after baseline significantly more participants in the ECO group achieved remission than in the CAU group (OR 2.228, 95% CI 1.115-4.453, P=.02).

Conclusions

The results of this study showed that in a group of sick-listed employees with common mental disorders, applying the blended eHealth ECO intervention led to faster first RTW and more remission of common mental disorder symptoms than CAU.

Trial Registration

Netherlands Trial Register NTR2108; http://www.trialregister.nl/trialreg/admin/rctview.asp?TC=2108. (Archived by WebCite at http://www.webcitation.org/6YBSnNx3P).

Hospital-level associations with 30-day patient mortality after cardiac surgery: a tutorial on the application and interpretation of marginal and multilevel logistic regression

Background

Marginal and multilevel logistic regression methods can estimate associations between hospital-level factors and patient-level 30-day mortality outcomes after cardiac surgery. However, it is not widely understood how the interpretation of hospital-level effects differs between these methods.

Methods

The Australasian Society of Cardiac and Thoracic Surgeons (ASCTS) registry provided data on 32,354 patients undergoing cardiac surgery in 18 hospitals from 2001 to 2009. The logistic regression methods related 30-day mortality after surgery to hospital characteristics with concurrent adjustment for patient characteristics.

Results

Hospital-level mortality rates varied from 1.0% to 4.1% of patients. Ordinary, marginal and multilevel regression methods differed with regard to point estimates and conclusions on statistical significance for hospital-level risk factors; ordinary logistic regression giving inappropriately narrow confidence intervals. The median odds ratio, MOR, from the multilevel model was 1.2 whereas ORs for most patient-level characteristics were of greater magnitude suggesting that unexplained between-hospital variation was not as relevant as patient-level characteristics for understanding mortality rates. For hospital-level characteristics in the multilevel model, 80% interval ORs, IOR-80%, supplemented the usual ORs from the logistic regression. The IOR-80% was (0.8 to 1.8) for academic affiliation and (0.6 to 1.3) for the median annual number of cardiac surgery procedures. The width of these intervals reflected the unexplained variation between hospitals in mortality rates; the inclusion of one in each interval suggested an inability to add meaningfully to explaining variation in mortality rates.

Conclusions

Marginal and multilevel models take different approaches to account for correlation between patients within hospitals and they lead to different interpretations for hospital-level odds ratios.

Clustering of oral and general health risk behaviors in Korean adolescents: a national representative sample

PURPOSES

To investigate the distribution of modifiable oral and general health risk behaviors according to socioeconomic status and to examine the extent of clustering of risk behaviors among Korean adolescents.

METHODS

Self-reported data from 71,404 adolescents were obtained using a stratified cluster sample of students in the 7th to 12th grades who participated in the Korea Youth Risk Behavior Web-based Survey. Multiple logistic regression analysis was used to determine the influence of family affluence on health behaviors and the strength of associations of general health behaviors in models for oral health behaviors.

RESULTS

Adolescents of high family affluence level were more likely to frequently brush their teeth, receive preventive dental care, eat breakfast, exercise regularly and become drunk, and were less likely to frequently smoke compared with those from less affluent families. Those who brush their teeth frequently were 19% less likely to smoke, 1.26 times more likely to eat breakfast, and 1.15 times more likely to do exercise. Those who received preventive dental care during the past year were 1.1 times more likely to smoke frequently and 1.38 times more likely to experience drunkenness once or more during the past year.

CONCLUSIONS

The co-occurrence of oral and general health risk behaviors among Korean adolescents and relatedness to the underlying socioeconomic environment suggest that clustering of health behaviors occurs before adulthood. Integrated efforts toward oral and general health promotion in the socioeconomic construct of the family environment may be needed to effect change in health risk behaviors among adolescents.

Two-year effects of a school-based prevention programme on adolescent cigarette smoking in Guangzhou, China: a cluster randomized trial

BACKGROUND

The prevalence of adolescent smoking has been increasing rapidly in China. Theory-based smoking prevention programmes in schools may be an effective approach in preventing smoking among Chinese adolescents.

METHODS

A school-level cluster randomized controlled trial was conducted among 7th and 8th grade students (N = 2343) in four junior high schools in southern China during 2004-06. The theory-based, multi-level intervention was compared with the standard health curriculum. Outcome measures comprised changes in students' smoking-related knowledge, attitudes and behaviour.

RESULTS

The mean knowledge scores from baseline to the 1- and 2-year follow-ups increased more in the intervention group than in the control group, whereas there was little change in attitude scores. At the 1-year follow-up (the total sample), the interventions reduced the probability of baseline experimental smokers' escalating to regular smoker [7.9 vs 18.3%; adjusted odds ratio (OR) 0.34, 95% confidence interval (CI) 0.12-0.97, P = 0.043], but did not reduce the probability of baseline non-smokers' initiating smoking (7.9 vs 10.6%; adjusted OR 0.86, 95% CI 0.54-1.38, P = 0.538). At the 2-year follow-up (only 7th grade students), similar proportions of baseline non-smokers initiated smoking in the intervention group and the control group (13.5 vs 13.1%), while a possibly lower proportion of baseline experimental smokers escalated to regular smoking in the intervention group than the control group (22.6 vs 40.0%; adjusted OR 0.43, 95% CI 0.12-1.57, P = 0.199).

CONCLUSIONS

This multi-level intervention programme had a moderate effect on inhibiting the escalation from experimental to regular smoking among Chinese adolescents, but had little effect on the initiation of smoking. The programme improved adolescents' smoking-related knowledge, but did not change their attitudes towards smoking.

Alteration of PTGS2 promoter methylation in chronic periodontitis

Levels of prostaglandin E(2) and the prostaglandin-endoperoxide synthase-2 (PTGS2, or COX-2) increase in actively progressing periodontal lesions, but decrease in chronic disease. We hypothesized that chronic inflammation is associated with altered DNA methylation levels within the PTGS2 promoter, with effects on COX-2 mRNA expression. PTGS2 promoter methylation levels from periodontally inflamed gingival biopsies showed a 5.06-fold increase as compared with non-inflamed samples (p = 0.03), and the odds of methylation in a CpG site in the inflamed gingival group is 4.46 times higher than in the same site in the non-inflamed group (p = 0.016). The level of methylation at -458 bp was inversely associated with transcriptional levels of PTGS2 (RT-PCR) (p = 0.01). Analysis of the data suggests that, in chronically inflamed tissues, there is a hypermethylation pattern of the PTGS2 promoter in association with a lower level of PTGS2 transcription, consistent with a dampening of COX-2 expression in chronic periodontitis. These findings suggest that the chronic persistence of the biofilm and inflammation may be associated with epigenetic changes in local tissues at the biofilm-gingival interface.

Two-year clinical effectiveness of adhesives and retention form on resin composite restorations of non-carious cervical lesions

The current study investigated the clinical effectiveness of three adhesives and the use of retention form in Class V resin composite restorations of the non-carious cervical lesion (NCCL) over a two-year period. One-hundred and fifty NCCLs in 39 patients were restored with resin composites according to six experimental protocols combining the presence or absence of retention form and three adhesives: ScotchBond Multi-Purpose (MP, 3M ESPE), an experimental adhesive (EX, Vericom) and Adper Prompt (AP, 3M ESPE). All restorations were evaluated at baseline, 6, 12 and 24 months. Modified United States Public Health Service (USPHS) criteria were used to evaluate the restorations. MP was found to have significantly superior marginal adaptation than AP in cumulative logistic regression analysis (odds ratio, 2.12; 95% confidence interval, 1.05-4.31; p = 0.0397). In analysis using the Pearson's Chi-square or Fisher's Exact Test to compare the clinical performance of restorations with and without retention form, EX with retention form showed a significantly higher retention rate at two years than that without retention form (p = 0.0089). Restorations with retention form also showed significantly less marginal discoloration than those without retention form in all three adhesives (p = 0.0336).

Chewing function impacts oral health-related quality of life among institutionalized and community-dwelling Korean elders

OBJECTIVE

The aim of this study was to assess the association of chewing ability to oral health-related quality of life (OHRQoL) measured by the Oral Health Impact Profile-14 (OHIP-14) controlling for clinical oral health status, self-reported health status, demographic factors, and socioeconomic conditions among community-dwelling and institutionalized Korean elders.

METHODS

This cross-sectional study comprised a sample of 307 community-dwelling and 102 institutionalized people over the age of 60, using a cluster sampling procedure. A questionnaire was implemented and a clinical oral examination was completed for each subject. The outcome variable of interest was the OHIP-14 score, and its associations with chewing ability, objective oral health status, self-reported health status, demographic factors, and socioeconomic conditions were assessed. Because of highly-skewed distribution of the OHIP-14 scores, nonparametric analytic methods were used. The final model was developed using a multivariable two-level logistic regression model for a dichotomized OHIP-14 score to account for the cluster sampling method applied to this study.

RESULTS

The mean age of the participants was 75.4 years, with 67.7% being women. The median OHIP-14 score was 7. Negative oral health impacts were experienced fairly often or very often by a total of 36.4% of elderly. In the final model, elders who could chew none to three and four to six foods among seven indicator foods were 3.4 (P = 0.010) and 2.0 (P = 0.040) times more likely, respectively, to have worse OHRQoL compared with elders who could chew all seven food types. Also significant associations with worse OHRQoL were shown for being concerned about oral health [Odds Ratio (OR) = 4.9, P = 0.002], fair or better self-reported oral health (OR = 0.12, P = 0.002), very good/good self-reported general health (OR = 0.38, P = 0.008), being married (OR = 2.0, P = 0.054), and having a favourable economic status (OR = 0.43, P = 0.042).

CONCLUSIONS

This study showed highly significant association between chewing ability and OHRQoL measured by the OHIP-14 score after controlling for related factors. Amelioration of chewing ability might independently contribute to improving the OHRQoL of elders.

Introduction to multi-level community based culturally situated interventions

This introduction to a special issue of the American Journal of Community Psychiatry is the result of a symposium at the Annual Meeting of the Society for Applied Anthropology, 2006, that brought together anthropologists and psychologists involved in community based collaborative intervention studies to examine critically the assumptions, processes and results of their multilevel interventions in local communities with local partners. The papers were an effort to examine context by offering a theoretical framework for the concept of "level" in intervention science, and advocating for "multi-level" approaches to social/behavioral change. They presented examples of ways in which interventions targeted social "levels" either simultaneously or sequentially by working together with communities across levels, and drawing on and co-constructing elements of local culture as components of the intervention. The papers raised a number of important issues, for example: (1) How are levels defined and how should collaborators be chosen; (2) does it matter at which level multilevel interventions begin; (3) do multilevel interventions have a greater effect on desired outcomes than level-specific interventions; (4) are multilevel interventions more sustainable; (5) are multilevel interventions cost effective to run, and evaluate; (6) how can theories of intervention be generated and adapted to each level of a multilevel intervention; (7) how should intervention activities at each level coordinate to facilitate community resident or target population empowerment? Many of these questions were only partially addressed in the papers presented at that time, and are more fully addressed in the theoretical papers, case studies and approach to evaluation included in this collection.

Identification of risk factors for the prevalence and persistence of Salmonella in Belgian broiler chicken flocks

According to the European Food Safety Authority, salmonellosis is still one of the main causes of infectious foodborne gastroenteritis in humans. Broilers are an important source of salmonellosis after eggs and pork. Between 1987 and 1999 the trend of human salmonellosis incidence in Belgium increased constantly. However, from 2000 until 2005 a decrease in human cases was observed, probably following the sanitary measures implemented in the poultry breeder and laying sector. In order to decrease human infections it is essential to tackle the problem at the farm level to minimize cross-contamination from farm to fork. This paper seeks to answer two questions: (i) given the Salmonella status of the farm at a certain occasion (equal to the sampling time of the flock), what are the risk factors that the farm will be Salmonella positive at a following occasion? And (ii) what are the risk factors for a farm to be persistently positive for two consecutive flocks? We used surveillance data on 6824 broiler flocks studied for Salmonella infectivity from 2005 to 2006 in Belgium. The farms were tested regularly (3 weeks before slaughter of each broiler flock) for the presence of Salmonella based on multiple faecal samples per flock on a farm yielding clustered data. Generalized estimating equations, alternating logistic regression models, and random-intercept logistic regression models were employed to analyse these correlated binary data. Our results indicated that there are many factors that influence Salmonella risk in broiler flocks, and that they interact. Accounting for interactions between risk factors leads to an improved determination of those risk factors that increase infection with Salmonella. For the conditional analysis, the risk factors found to increase the risk of Salmonella infection on a farm at a current occasion given the previous Salmonella status included: Salmonella infection of day-old chicks (of the current flock); a previously infected flock even though the farm was equipped with a hygiene place to change clothes prior to entering the broiler house; having temporary workmen when there was a separation between birds of different species; and separating birds of different species in the Walloon region relative to the Flanders region. Sanitary measures such as a cleaning and disinfecting procedure conducted by an external cleaning firm, applying the all-in all-out procedure, and hand washing decreased the risk despite their interaction with other factors. From the joint analysis, the most important factors identified for increased risk for persistent Salmonella on a farm involved the interaction between having temporary workmen when there were poultry or farmers in contact with foreign poultry or persons, and the interaction between having temporary workmen when there were poultry or farmers in contact with external poultry or persons.

Parental perceptions of children's oral health: the Early Childhood Oral Health Impact Scale (ECOHIS)

BACKGROUND

Dental disease and treatment experience can negatively affect the oral health related quality of life (OHRQL) of preschool aged children and their caregivers. Currently no valid and reliable instrument is available to measure these negative influences in very young children. The objective of this research was to develop the Early Childhood Oral Health Impact Scale (ECOHIS) to measure the OHRQL of preschool children and their families.

METHODS

Twenty-two health professionals evaluated a pool of 45 items that assess the impact of oral health problems on 6-14-year-old children and their families. The health professionals identified 36 items as relevant to preschool children. Thirty parents rated the importance of these 36 items to preschool children; 13 (9 child and 4 family) items were considered important. The 13-item ECOHIS was administered to 295 parents of 5-year-old children to assess construct validity and internal consistency reliability (using Cronbach's alpha). Test-retest reliability was evaluated among another sample of parents (N = 46) using the intraclass correlation coefficient (ICC).

RESULTS

ECOHIS scores on the child and parent sections indicating worse quality of life were significantly associated with fair or poor parental ratings of their child's general and oral health, and the presence of dental disease in the child. Cronbach's alphas for the child and family sections were 0.91 and 0.95 respectively, and the ICC for test-retest reliability was 0.84.

CONCLUSION

The ECOHIS performed well in assessing OHRQL among children and their families. Studies in other populations are needed to further establish the instrument's technical properties.