Background

The Secondary Care Pharmaceutical Adviser for London collates standardised, summative data on pharmacy performance indicators (PIs) from NHS trusts on a regular basis. In this study, and in earlier work¹, we have used multiple regression to determine the relative contribution of selected PIs to the variation in pharmacy establishments. For predictive models to be useful for service planning, this regression approach needs to be explored and validated.

Method

A previous study¹ demonstrated the use of multivariate regression analysis for predicting establishment in 52 NHS trusts, based on 1999/00 PIs. A second consecutive year of data (2000/01) has been captured to determine the robustness of this modelling technique. Establishment, expressed as whole time equivalents (WTEs), was designated as the dependent variable, and regression equations derived for each year of data using identical PIs. The year-specific regression coefficients were then applied to the data from the alternative year, yielding two predicted WTEs for trusts in each year of data. If modelling is robust, we would not expect to see deviations in these values on a case-wise application.

Results

The figures show a scatter plot of the two predicted WTEs in each year of data. The modelling appears robust, with high agreement between the predicted value for WTEs, suggesting that the regression coefficients are stable over time.

Conclusion

We have demonstrated previously that regression models can be used to predict pharmacy workforce in secondary care.¹ The present sensitivity study indicates that the relationship between WTEs and selected PIs is consistent over 2 years of data capture. The next stage will employ this method to predict WTEs by professional type. Multivariate techniques appear to be of value in modelling pharmaceutical services, and could in future provide evidence for examining an appropriate skill mix. The results presented here provide evidence that the modelling approach is sufficiently robust and sensitive to be developed for this purpose.

Acknowledgement

The authors would like to thank London senior pharmacy managers for permission to analyse the performance indicator data.

References

1. Sharott P, Webb DG, Bates I. Multiple regression as a procedure to interrogate pharmacy performance indicators. 8th Health Services Research and Pharmacy Practice Conference, Leeds University 2002. (ISBN 0 9538505-3-6. The Royal Pharmaceutical Society), p12.