Background. No instrument has been developed specifically for assessing the clinician–patient therapeutic relationship (TR) in community psychiatry. This study aimed to develop a measure of the TR with clinician and patient versions using psychometric principles for test construction.

Method. A four-stage prospective study was undertaken, comprising qualitative semi-structured interviews about TRs with clinicians and patients and their assessment of nine established scales for their applicability to community care, administering an amalgamated scale of more than 100 items, followed by Principal Components Analysis (PCA) of these ratings for preliminary scale construction, test–retest reliability of the scale and administering the scale in a new sample to confirm its factorial structure. The sample consisted of patients with severe mental illness and a designated key worker in the care of 17 community mental health teams in England and Sweden.

Results. New items not covered by established scales were identified, including clinician helpfulness in accessing services, patient aggression and family interference. The new patient (STAR-P) and clinician scales (STAR-C) each have 12 items comprising three subscales: positive collaboration and positive clinician input in both versions, non-supportive clinician input in the patient version, and emotional difficulties in the clinician version. Test–retest reliability was r=0·76 for STAR-P and r=0·68 for STAR-C. The factorial structure of the new scale was confirmed with a good fit.

Conclusions. STAR is a specifically developed, brief scale to assess TRs in community psychiatry with good psychometric properties and is suitable for use in research and routine care.