Content of quality-of-life instruments is affected by item-generation methods

Methods used to generate items for complex measurement scales are heterogeneous and probably produce heterogeneous data, yet nothing is known about the advantages of one method over another.

We aimed to compare methods of generating items for tools designed to measure quality-of-life for patients.

We used five methods to develop a quality-of-life instrument for patients with lower-limb osteoarthritis: individual interviews with patients involving two different techniques (semi-structured and cognitive), individual interviews with health professionals, and focus groups of patients and health professionals. The process generated 80 items, of which 37 were excluded after content and psychometric analysis. With the final 43-item scale used as a 'reference standard', we estimated the contribution of each method.

For health professionals, the focus group and individual interviews produced 35 and 81% of the items, respectively. For patients, the focus groups produced 74% of the items and both interview techniques 100% of the items. Health professionals provided a narrower picture of the effects of the disease on quality-of-life. Focus groups contributed less to social domains than did individual interviews. The two patient interview techniques highlighted different themes.

In developing a complex measurement scale for patients, we found individual interviews with patients the best method for formulating items; other methods such as physician interviews and focus groups contributed no additional information. Reports of instrument generation should include details of the item-generation step, the methods used to develop items and the number of people involved.