Do general practitioners have different "referral thresholds"?

One of the most interesting puzzles in general practice is why doctors differ so greatly in the frequency with which they refer patients to hospital. Published figures range from less than 0-5 to more than 15 per 100 consultations,1 and from only 0-6% to 25-9% of patients on a practice list a year. ? Obvious differences in patient characteristics, such as age, sex, social class, type and severity of problem, and distance from the practice; in doctor characteristics, such as age, years in general practice and in a particular practice, special medical interests, and postgraduate training ; and in practice characteristics, such as size of list, work load, urban or rural setting, partnership or single-handed structure, and distance from hospital1'11 may explain many of these marked variations. These differences in rates may also be explained by what other authors have called a referral threshold' * >*? that is, doctors have a personal level at which the stimulus of a consultation produces a referral. This concept would be supported if doctors were observed to refer at dissimilar rates even when patient, doctor, and practice characteristics were the same. Unfortunately, only a few studies* ' of the process have controlled for more than one of these characteristics at a tame. Since 1971, the general practice partnership of one of us (BJ) has kept extensive records on every consultation. The size and detail of these data allowed us to carry the observations of previous workers one step further?we compared the rates of several doctors while a number of patient and practice variables were controlled. Our purpose was to examine the hypothesis that general practitioners have a personal threshold. Practice characteristics were the same for each