Aim \nThis study asks if practices can collaborate to create a shared formulary, \nwhich influences prescribing. \nSecondary questions \nIf so, how long does the effect last? To what degree can such a formulary \ninfluence volume of prescribing, choice of treatment and cost? Is there a limit \nto the number of practices involved before the effect is lost? Are doctors more \ninfluenced by the formulary if they are closely involved with its creation? \nMethod \nControlled trial in two stages. 100 general practitioners from 20 urban and \nsemi-rural practices and 50 community nurses worked with a small team of \nfacilitators to create a formulary. Between 1991 and 1999, prescribing by the \nparticipating doctors was compared with the prescribing by all other general \npractitioners in the county. \nResults \nCollaborative work resulted in a countywide formulary for primary care. The \nuse of information to support prescribing changed from a few practices using \ntheir own formularies, to an evidence-based formulary supported by all five \nPrimary Care Groups in the county. Choice of treatment changed in seven out \nof thirteen therapeutic groups. Volume of prescribing reduced in three groups. \nCost reduced by F-3000 per doctor per annum. \nConclusions \nSharing resources between practices to create a primary care formulary can \nlead to modest changes in prescribing, sustained over three years, and lower \noverall costs. The largest observed changes were a 14% change in the choice \nof drugs for musculoskeletal conditions, and a saving of £5000 per practice \nper year on antibiotics. Such changes, attributed to the development of a \nformulary, also occur in practices that have no direct involvement, but later by \nseveral years. The greatest change in prescribing is seen immediately after a \nformulary is created and in those involved with its development. The funding for the work is estimated to amount to 17% of the saving on prescribing. \nDoctors and nurses from 32 practices can work together on such an \nintervention.