AIMS: Referrals from primary to secondary care may differ regarding motivation and initiative. Previous research on the frequency and variation of referrals has mostly treated referrals as homogeneous. We intended to develop a taxonomy regarding referrals from primary to secondary care in Germany that could support decision making on a macro level. METHODS: We analyzed 3,988 referrals by 29 German general practitioners (GPs). GPs were asked to document all referrals during one week; in subsequent audits they stated the reasons and initiative for any referral. We postulated the following five referral types: clinical problem, shared care, administrative, patient initiated and shared cost. The data were analyzed with k-means cluster analysis. RESULTS: We identified three of our five postulated referral types with cluster analytic techniques: shared care, clinical problem, and patient initiated. This solution accounted for 11.7% of total variance. The majority of referrals in German primary care practices were initiated by the GP, or they were part of a shared decision with patients. CONCLUSIONS: We suggest a taxonomy of referrals that might offer insights regarding the allocation of resources within the German health system. Referrals might be reduced by improved training of primary care physicians and by giving them more competencies in routine care of chronic patients.
AIMS: Referrals from primary to secondary care may differ regarding motivation and initiative. Previous research on the frequency and variation of referrals has mostly treated referrals as homogeneous. We intended to develop a taxonomy regarding referrals from primary to secondary care in Germany that could support decision making on a macro level. METHODS: We analyzed 3,988 referrals by 29 German general practitioners (GPs). GPs were asked to document all referrals during one week; in subsequent audits they stated the reasons and initiative for any referral. We postulated the following five referral types: clinical problem, shared care, administrative, patient initiated and shared cost. The data were analyzed with k-means cluster analysis. RESULTS: We identified three of our five postulated referral types with cluster analytic techniques: shared care, clinical problem, and patient initiated. This solution accounted for 11.7% of total variance. The majority of referrals in German primary care practices were initiated by the GP, or they were part of a shared decision with patients. CONCLUSIONS: We suggest a taxonomy of referrals that might offer insights regarding the allocation of resources within the German health system. Referrals might be reduced by improved training of primary care physicians and by giving them more competencies in routine care of chronic patients.