We thank Desai & Chaturvedi for their interest in our paper. We agree that additional dimensions could be included as possible predictors of health status and healthcare use and that the latter would be influenced by the nature of local healthcare facilities. It was impossible to include such additional measures in our study because we were restricted to those measures that had been used in the original studies.

Desai & Chaturvedi mention duration and severity of symptoms as possible predictors of outcome. Duration is important but may not predict number of subsequent doctor visits. ^{Reference Fiddler, Jackson, Kapur, Wells and Creed1} Severity is important and five of our studies used questionnaires (including the Patient Health Questionnaire-15) which assessed the degree of bothersomeness of each somatic symptom, a subjective measure of severity. The distinction between intensity and severity is complex, but one study noted that severity of pain did not explain the association between number of somatic symptoms and subsequent health status. ^{Reference Creed, Tomenson, Chew-Graham, Macfarlane, Davies and Jackson2}

The point raised by Desai & Chaturvedi regarding the co-occurrence of medically explained and unexplained symptoms is very important and forms one of the main points of the paper. Such co-occurrence of symptoms is common and constitutes one of the main difficulties of trying to make a diagnosis purely on the presence of medically unexplained symptoms. In the four sites where data were available, we found that the association of somatic symptoms with health status, after adjustment for confounders, was stronger for total somatic symptom score than for number of medically unexplained symptoms. We could not test this in relation to healthcare use but the association between number of somatic symptoms with healthcare use appears similar for medically explained and unexplained symptoms. ^{Reference Jackson, Fiddler, Kapur, Wells, Tomenson and Creed3}

Assessing abnormal illness behaviour is difficult in population-based studies using self-administered questionnaires, as most measures include items about how often the respondent visits doctors, which would overlap with our outcome measure of healthcare use. A better dimension might be a person’s general tendency to visit doctors even for minor reasons; this is a predictor of healthcare use independent of number of bothersome somatic symptoms. ^{Reference Mewes, Rief, Brähler, Martin and Glaesmer4}

The other dimension mentioned by Desai & Chaturvedi, health anxiety, is very important. In two studies a high number of somatic symptoms and pronounced health anxiety were both independent predictors of primary healthcare contacts (see Tomenson et al ^{Reference Tomenson, McBeth, Chew-Graham, MacFarlane, Davies and Jackson5} ). Two other studies have shown a complex interaction between these dimensions, with health anxiety being a predictor of subsequent healthcare use only in respondents without a high number of somatic symptoms or who also have serious medical illnesses. ^{Reference Jackson, Fiddler, Kapur, Wells, Tomenson and Creed3,Reference Tomenson, McBeth, Chew-Graham, MacFarlane, Davies and Jackson5}

This field of research suffers from lack of prospective studies. The correlates, or predictors, of healthcare use are somewhat different for past use and future use. ^{Reference Tomenson, McBeth, Chew-Graham, MacFarlane, Davies and Jackson5,Reference Barsky, Ettner, Horsky and Bates6} One paper made the intriguing, but plausible, suggestion that frequent visits to the physician could increase health anxiety and precipitate more somatic symptoms rather than the other way round. ^{Reference Barsky, Ettner, Horsky and Bates6} Further prospective studies using well-validated questionnaires are needed. ^{Reference Zijlema, Stolk, Löwe, Rief, White and Rosmalen7}