Background.

Each year, Senior Pharmacy Managers in the London NHS Region collect standardised summative data on pharmacy performance indicators. This data is transferred to a Regional spreadsheet, analysed and presented as descriptive statistics. The pharmacy performance indicators are a list of variables, of which some may be designated as dependent, in the sense that their linear relationship with other variables is of interest. In this study, we used simultaneous multiple regression to investigate the association between pharmacy staff establishment (and therefore staff costs) and various performance indicators, to determine the relative contribution of the selected indicators to the overall variation in staffing establishment.

Method.

The summative performance indicators for a single year were collected for 53 NHS trusts in the London Region, which includes university teaching trusts (n="14"), non-teaching trusts (n="29") and community trusts (n="10"). Staff establishment was the dependant variable, expressed as whole time equivalents (WTE). The variable list was examined for colinearity, and eligible variables entered stepwise into a simultaneous multiple regression model (Table 1.)

Table 1. Independent variable list

Result.

Three variables remained significant after entering all variables into the model, with the adjusted coefficient of determination,R² = 0.892 (explaining 89.2% of variance, F="97".1, p<0.001). Colinearity diagnostics were favourable. Staff establishment, WTE, can be described by inpatient drug expenditure (IDE), ward visits (WV) and occupied bed days (OBD), Table 2. For some trusts, the residual plot revealed a large difference between actual and predicted WTE based on this model.

Table 2. Final regression model.

Conclusions.

Regression analysis enables a statistically robust model of WTE to be produced, but it is of equal value to reflect on those variables that were rejected by the procedure, as well as those that were accepted (Table 1). In terms of predicted staff establishment, teaching hospitals stand out, with a particularly wide variation (both positive and negative) between observed and predicted values. Although exploratory at this stage, we anticipate that multivariate techniques will be of value in modelling pharmaceutical services, and hence enable pharmacy managers to argue effectively for resources to provide those services.

Acknowledgement.

The authors would like to thank London Region Senior Pharmacy Managers for permission to analyse the performance indicator data.