modelling decision for NBS

Modelling parental decisions for newborn bloodspot screening. : 

Modelling parental decisions for newborn bloodspot screening. Stuart Nicholls Dept. Mathematics and Statistics, Lancaster University 14 April 2010 33rd RSC in Probability and Statistics

Overview : 

Overview Background & motivation Methodological overview Assessing the model Model development and the final model Discussion

Newborn bloodspot screening: motivation : 

Newborn bloodspot screening: motivation 5-8 days post-birth National program since 1969‏ Informed choice (UK Newborn Screening Programme Centre, 2008) but limited evidence (Stewart and Oliver 2003)‏ Aim: To develop and assess a model of parental decisions regarding consent for newborn bloodspot screening

Structural equation modelling: an overview : 

Structural equation modelling: an overview Analysis of covariances Incorporates latent variables Both measurement and structural aspects

Structural equation modelling: an overview : 

Structural equation modelling: an overview COV (X1, Y1) = COV (X1, aX1+e1)‏ = aVAR(X1)‏ a=COV(X1,Y1)/VAR(X1)‏ η=Βη+γξ ML estimation Overidentified Fit indices: χ2, RMSEA, X1 Y1 X2 a b c e1 e2

Defining the model: Questionnaire : 

Defining the model: Questionnaire Parents of children born in 2008 Random sample from Merseyside & Cheshire screening laboratory (n=500)‏ Postal questionnaire 32% response rate Final sample n=148

Defining the model: CFA : 

Defining the model: CFA Multivariate non-normal: bootstrap ML and Bollen-Stine χ2 Goodness-of-fit p=0.152 (Bollen-Stine), RMSEA = 0.048, All parameter estimates significant at the 5% level Adequate model fit.

Initial model assessment : 

Initial model assessment Bollen-Stine, p=0.002, RMSEA=0.082 Modification indices suggest: Midapp→CHOICE eben↔erisk emid↔etrst

Slide 9: 

Bollen-Stine, p=0.32 RMSEA = 0.04 (0.00, 0.07)‏ Good model fit

Model parameters : 

Model parameters 95% confidence intervals (bias-corrected percentile method)‏

Squared multiple correlations : 

Squared multiple correlations 95% confidence intervals (bias-corrected percentile method)‏

Discussion : 

Discussion Trust a significant item in attitude development BUT small SMC. Positive attitudes to screening and perceived choice both impact to improve perceived decisional quality (they decrease uncertainty and ineffectiveness)‏ Prior knowledge is a significant in its effect on perceived choice Suggests current focus on rational psychological process or risk/benefit analysis are limited and sociological factors such as perceived choice and trust incorporated

Acknowledgements : 

Acknowledgements Supervisors Dr Mairi Levitt & Prof Paul Fearnhead (Lancaster University)‏ Dr Kevin Southern & Mrs Elaine Hanmer (Alder Hey)‏ All the parents who took part ESRC