Using postal questionnaires to evaluate physical activity and diet behaviour change: case study exploring implications of valid responder characteristics in interpreting intervention outcomes.

BACKGROUND

Patient reported outcome measures (PROMs) are used to evaluate lifestyle interventions but little is known about differences between patients returning valid and invalid responses, or of potential for bias in evaluations. We aimed to examine the characteristics of patients who returned valid responses to lifestyle questionnaires compared to those whose responses were invalid for evaluating lifestyle change.

METHODS

We conducted a secondary data analysis from the SPHERE Study, a trial of an intervention to improve outcomes for patients with coronary heart disease in primary care. Postal questionnaires were used to assess physical activity (Godin) and diet (DINE) among study participants at baseline and 18 month follow-up. Three binary response variables were generated for analysis: (1) valid Godin score; (2) valid DINE Fibre score; and (3) valid DINE Total Fat score. Multivariate analysis comprised generalised estimating equation regression to examine the association of patients' characteristics with their return of valid responses at both timepoints.

RESULTS

Overall, 92.1% of participants (832/903) returned questionnaires at both baseline and 18 months. Relatively fewer valid Godin scores were returned by those who left school aged <15 years (36.5%) than aged 18 and over (50.5%), manual workers (39.5%) than non-manual (49.5%) and those with an elevated cholesterol (>5 mmol) (34.7%) than those with a lower cholesterol (44.4%) but multivariate analysis identified that only school leaving age (p = 0.047) was of statistical significance.Relatively fewer valid DINE scores were returned by manual than non-manual workers (fibre: 80.8% v 86.8%; fat: 71.2% v 80.0%), smokers (fibre: 72.6% v 84.7%; fat: 67.5% v 76.9%), patients with diabetes (fibre: 75.9% v 82.9%; fat: 66.9% v 75.8%) and those with cholesterol >5 mmol (fat: 68.2% v 76.2%) but multivariate analysis showed statistical significance only for smoking (fibre: p = 0.013; fat: p = 0.045), diabetes (fibre: p = 0.039; fat: p = 0.047), and cholesterol (fat: p = 0.039).

CONCLUSIONS

Our findings illustrate the importance of detailed reporting of research methods, with clear information about response rates, respondents and valid outcome data. Outcome measures which are relevant to a study population should be chosen carefully. The impact of methods of outcome measurement and valid response rates in evaluating healthcare requires further study.