Effect of non-response bias in pressure ulcer prevalence studies
Article first published online: 15 MAY 2006
Journal of Advanced Nursing
Volume 55, Issue 2, pages 230–236, July 2006
How to Cite
Lahmann, N., Halfens, R. J.G. and Dassen, T. (2006), Effect of non-response bias in pressure ulcer prevalence studies. Journal of Advanced Nursing, 55: 230–236. doi: 10.1111/j.1365-2648.2006.03906.x
- Issue published online: 15 JUN 2006
- Article first published online: 15 MAY 2006
- Accepted for publication 20 December 2005
- non-response bias;
- point prevalence survey;
- pressure ulcers;
Aim. This paper reports a study to determine the prevalence of pressure ulcers in German hospitals and nursing homes for national and international comparison, and analyses the influence of non-response bias.
Background. Outcome rates are often used to evaluate provider performance. The prevalence of pressure ulcers is seen as a possible parameter of outcome healthcare quality. However, the results from different pressure ulcer prevalence studies cannot be compared, because there is no standardized methodology and terminology. Observed and published prevalence rates may reflect variations in quality of care, but differences could also relate to differences in case-mix or to random variation.
Methods. A point prevalence survey was carried out for 2002 and 2003 using data from 21,574 patients and residents in 147 different kinds of institutions throughout Germany. Participation rates and reasons for not participating in the study were documented. Non-responders were considered in different calculations to show the range of possible prevalence rate for a hypothetic 100% participation.
Results. In 2002 and 2003, the calculated prevalence rate (among participating persons at risk) in hospitals was 25·1% and 24·2% respectively, while in nursing homes it was 17·3% and 12·5% respectively. Non-response varied from 15·1% to 25·1%. The majority of non-responders in hospitals and nursing homes had not been willing to participate in the study. Based on different assumptions about the characteristics of the non-responders, we calculated minimum and maximum prevalence rates as if 100% participation was achieved.
Conclusions. Calculating the non-response bias of prevalence rates is an inconvenient but necessary thing to do because its influence on calculated prevalence rates was high in this study. High participation rates in clinical studies will minimize non-response bias. If non-response cannot be avoided, the formula provided will help researchers calculate possible minimum and maximum prevalence rates for the total sample of both the responding and non-responding groups.