Dataset Information

Which Propensity Score Method Best Reduces Confounder Imbalance? An Example From a Retrospective Evaluation of a Childhood Obesity Intervention.

ABSTRACT: BACKGROUND:Propensity score (PS) methods are increasingly being employed by researchers to reduce bias arising from confounder imbalance when using observational data to examine intervention effects. OBJECTIVE:The purpose of this study was to examine PS theory and methodology and compare application of three PS methods (matching, stratification, weighting) to determine which best improves confounder balance. METHODS:Baseline characteristics of a sample of 20,518 school-aged children with severe obesity (of whom 1,054 received an obesity intervention) were assessed prior to PS application. Three PS methods were then applied to the data to determine which showed the greatest improvement in confounder balance between the intervention and control group. The effect of each PS method on the outcome variable-body mass index percentile change at one year-was also examined. SAS 9.4 and Comprehensive Meta-analysis statistical software were used for analyses. RESULTS:Prior to PS adjustment, the intervention and control groups differed significantly on seven of 11 potential confounders. PS matching removed all differences. PS stratification and weighting both removed one difference but created two new differences. Sensitivity analyses did not change these results. Body mass index percentile at 1 year decreased in both groups. The size of the decrease was smaller in the intervention group, and the estimate of the decrease varied by PS method. DISCUSSION:Selection of a PS method should be guided by insight from statistical theory and simulation experiments, in addition to observed improvement in confounder balance. For this data set, PS matching worked best to correct confounder imbalance. Because each method varied in correcting confounder imbalance, we recommend that multiple PS methods be compared for ability to improve confounder balance before implementation in evaluating treatment effects in observational data.

SUBMITTER: Schroeder K

PROVIDER: S-EPMC5098456 | biostudies-literature | 2016 Nov/Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Which Propensity Score Method Best Reduces Confounder Imbalance? An Example From a Retrospective Evaluation of a Childhood Obesity Intervention.

Schroeder Krista K Jia Haomiao H Smaldone Arlene A

Nursing research 20161101 6

<h4>Background</h4>Propensity score (PS) methods are increasingly being employed by researchers to reduce bias arising from confounder imbalance when using observational data to examine intervention effects.<h4>Objective</h4>The purpose of this study was to examine PS theory and methodology and compare application of three PS methods (matching, stratification, weighting) to determine which best improves confounder balance.<h4>Methods</h4>Baseline characteristics of a sample of 20,518 school-aged ...[more]

PMID: 27801717

Similar Datasets

Project description:Obesity is increasing worldwide in prepubertal children, reducing the age of onset of associated comorbidities, including type 2 diabetes. Sulfur-containing amino acids, methionine, cysteine, and their derivatives play important roles in the transmethylation and transsulfuration pathways. Dysregulation of these pathways leads to alterations in the cellular methylation patterns and an imbalanced redox state. Therefore, we tested the hypothesis that one-carbon metabolism is already dysregulated in prepubertal children with obesity. Peripheral blood was collected from 64 children, and the plasma metabolites from transmethylation and transsulfuration pathways were quantified by HPLC. The cohort was stratified by BMI z-scores and HOMA-IR indices into healthy lean (HL), healthy obese (HO), and unhealthy obese (UHO). Fasting insulin levels were higher in the HO group compared to the HL, while the UHO had the highest. All groups presented normal fasting glycemia. Furthermore, high-density lipoprotein (HDL) was lower while triglycerides and lactate levels were higher in the UHO compared to HO subjects. S-adenosylhomocysteine (SAH) and total homocysteine levels were increased in the HO group compared to HL. Additionally, glutathione metabolism was also altered. Free cystine and oxidized glutathione (GSSG) were increased in the HO as compared to HL subjects. Importantly, the adipocyte secretory function was already compromised at this young age. Elevated circulating leptin and decreased adiponectin levels were observed in the UHO as compared to the HO subjects. Some of these alterations were concomitant with alterations in the DNA methylation patterns in the obese group, independent of the impaired insulin levels. In conclusion, our study informs on novel and important metabolic alterations in the transmethylation and the transsulfuration pathways in the early stages of obesity. Moreover, the altered secretory function of the adipocyte very early in life may be relevant in identifying early metabolic markers of disease that may inform on the increased risk for specific future comorbidities in this population.

Project description:Previous evidence revealed that central obesity played a vital role in the development of diabetes mellitus (DM). However, because of imbalanced confounding variables, some studies have not wholly established the association between central obesity and diabetes. Propensity score matching (PSM) analysis can minimize the impact of potential confounding variables. Therefore, the aim of the present study was to explore the relationship between central obesity and diabetes in the Japanese population by using PSM analysis. This retrospective cohort study included 15,453 Japanese adults who were free of diabetes at baseline between 2004 and 2015, which provided all medical records for individuals participating in the physical exam. Central obesity at baseline was an independent variable, and incident diabetes during follow-up was an outcome variable. Using a 1:1 PSM analysis, the present retrospective cohort study included 1639 adults with and without central obesity. Additionally, we employed a doubly robust estimation method to identify the association between central obesity and diabetes. Subjects with central obesity were 92% more likely to develop DM (HR = 1.65, 95%CI 1.12, 2.41). After adjusting for covariates, subjects with central obesity had a 72% increased risk of developing DM compared with subjects with non-central obesity in the PSM cohort (HR = 1.72, 95% CI 1.16, 2.56). Central obesity individuals had a 91% higher risk of DM than non-central obesity individuals, after adjustment for propensity score (HR = 1.91, 95% CI 1.29, 2.81). In sensitivity analysis, the central obesity group had a 44% (HR = 1.44, 95% CI 1.09, 1.90) and 59% (HR = 1.59, 95% CI1.35, 1.88) higher risk of DM than the non-central obesity group in the original and weighted cohorts after adjusting for confounding variables, respectively. Central obesity was independently associated with an increased risk of developing diabetes. After adjustment for confounding covariates, central obesity participants had a 72% higher risk of development of diabetes than non-central obesity individuals in the PSM cohort.

Project description:BackgroundChildhood obesity is high on the policy agenda of wealthier nations, and many interventions have been developed to address it. This work describes an overview of schemes for obese and overweight children and young people in England, and the 'mapping' approach we used.MethodsOur search strategy, inclusion criteria and coding frame had to be suitable for describing a potentially large number of schemes within a short timeframe. Data were collected from key informants, scheme publicity and reports, and via a web-survey. To be included, schemes had to be based in England, follow a structured programme lasting at least two weeks, promote healthy weight, and be delivered exclusively to overweight and/or obese children and young people (age range 4-18). Data were entered into a coding frame recording similar information for each scheme, including any underpinning research evidence, evaluation or monitoring reports. Priority questions were identified in consultation with colleagues from the Department of Health and the Cross Government Obesity Unit.ResultsFifty-one schemes were identified. Some operated in multiple areas, and by using estimates of the number of schemes provided by multi-site scheme leads, we found that between 314 and 375 local programmes were running at any time. Uncertainty is largely due to the largest scheme provider undergoing rapid expansion at the time of the mapping exercise and therefore able to provide only an estimate of the number of programmes running. Many schemes were similar in their approach, had been recently established and were following NICE guidelines on interventions to promote healthy weight. Rigorous evaluation was rare.ConclusionsOur methods enabled us to produce a rapid overview of service activity across a wide geographic area and a range of organisations and sectors. In order to develop the evidence base for childhood obesity interventions, rigorous evaluation of these schemes is required. This overview can serve as a starting point for evaluations of interventions to address obesity. More generally, a rapid and systematic approach of this type is transferable to other types of service activity in health and social care, and may be a tool to inform public health planning.

Dataset Information

Which Propensity Score Method Best Reduces Confounder Imbalance? An Example From a Retrospective Evaluation of a Childhood Obesity Intervention.

Publications

Which Propensity Score Method Best Reduces Confounder Imbalance? An Example From a Retrospective Evaluation of a Childhood Obesity Intervention.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets