How to ask about patient satisfaction? The visual analogue scale is less vulnerable to confounding factors and ceiling effect than a symmetric Likert scale

Abstract Aims To study the effects of scale type (visual analogue scale vs. Likert), item order (systematic vs. random), item non‐response and patient‐related characteristics (age, gender, subjective health, need for assistance with filling out the questionnaire and length of stay) on the results of patient satisfaction surveys. Background Although patient satisfaction is one of the most intensely studied issues in the health sciences, research information about the effects of possible instrument‐related confounding factors on patient satisfaction surveys is scant. Design A quasi‐experimental design was employed. A non‐randomized sample of 150 surgical patients was gathered to minimize possible alterations in care quality. Methods Data were collected in May–September 2014 from one tertiary hospital in Finland using the Revised Humane Caring Scale instrument. New versions of the instrument were created for the present purposes. In these versions, items were either in a visual analogue format or Likert‐scaled, in systematic or random order. The data were analysed using an analysis of covariance and a paired samples t ‐test. Results The visual analogue scale items were less vulnerable to bias from confounding factors than were the Likert‐scaled items. The visual analogue scale also avoided the ceiling effect better than Likert and the time needed to complete the visual analogue scale questionnaire was 28% shorter than that needed to complete the Likert‐scaled questionnaire. Conclusion The present results supported the use of visual analogue scale rather than Likert scaling in patient satisfaction surveys and stressed the need to account for as many potential confounding factors as possible.