Introduction This study was designed to examine the geographic distribution of prescription clients around community pharmacies. We wished to test the assumption made in previous analyses of prescribing, that pharmacy location was a good proxy for patient location.
In addition, we wished to explore the application of geographic information system (GIS) technology to pharmacy practice research. GIS technology has been used to study the distribution of community pharmacies in the north-west of England¹ and in a number of health related epidemiology studies.^2-4 However, there have been no reports of this technology being used to analyse pharmacy prescription databases.

Method Twelve New Zealand pharmacies were selected. Four were central city, four suburban and four rural pharmacies. Addresses of clients who had had prescriptions dispensed were downloaded from each pharmacy's database. For each pharmacy, a random sample of 1,067 addresses was selected.
All of the addresses were checked and corrected manually. These were then electronically matched (geocoded) against a reference database containing a list of all the streets and suburbs in the country. The matching was represented in two visual forms: first, as a spreadsheet table consisting of a unique identifier, meshblock number, area unit number and X and Y co-ordinates for each of the addresses; secondly, as a series of maps for each pharmacy, indicating the plotted position of each address in relation to the pharmacy and coloured-coded meshblocks illustrating customer density.

Focal points

• Assumptions about where pharmacies draw their customers from need testing
• Combining geographic information systems (GIS) technology with pharmacy computer prescription databases through geocoding offers a method of obtaining information about the geographical spread of pharmacy customers
• Different types of pharmacies have different patterns of customer patronage
• Further uses for other information stored in pharmacy databases need exploration

Results Figure 1 is a representation of the type of information produced by the geocoding process. It is based on a model map generated from the address database of a simulated suburban pharmacy, with the pharmacy clearly marked and colour-coded meshblocks showing a high density of customers in close proximity to the pharmacy. There are a few outlying high-density meshblocks which can be variously explained as containing rest homes or hospitals serviced by the pharmacy or customers belonging to a prescription-charge reimbursement scheme.
Central city pharmacies had a broad scattering of low-density clientele throughout the city with few customers living in the immediate vicinity.
Suburban pharmacies drew their clients in large numbers from the surrounding residential suburbs.
Rural pharmacies' customers came from the town (high density) and surrounding countryside (low density) with some travelling substantial distances.

Discussion It cannot be assumed that pharmacies draw their customers from the immediate vicinity; different types of pharmacies have different patterns of customer patronage. Mapping customer distribution enables pharmacists to target their marketing activities appropriately. The results obtained are only as accurate as the pharmacy's database entries, with some addresses being unclear or uncodable. In this study the sampling procedure gave no indication of the date or frequency of visits.
A further study is being carried out this year which attempts to address these inadequacies. This is investigating the proportion of customers who visit one, two or more pharmacies. Further uses for other information stored in pharmacy databases, such as the geographic distribution of prescribed medications, will also be explored.

Figure 1. Map based on a model generated by a geographic information system (GIS) from the address data base of a simulated suburban pharmacy