My second miniproject was in conjunction with the NHS's Commissioning Support Unit for Birmingham, Black Country and Solihull. In addition to an external NHS supervisor, I had two internal supervisors Dr Colm Connaughton and Dr Tom Nichols.
Patient choice is a central theme in the most recent Government White Paper on health. In particular a pregnant woman in England can choose the maternity unit in which she will give birth. A woman is encouraged to make this decision early in the pregnancy and to make a booking at her unit of choice for her due date. A discrete choice survey was commissioned by the NHS with the aim of using the results to predict maternity unit usage, particularly to see how patient flow will change when specific units are closed or changed.
Utilities were derived from the responses to the survey and a utility maximising predictive model was implemented. However, when this model was applied to predict the location of each birth in the West Midlands in 2008/2009, the results of the model did not provide a good match to the observed numbers. The aim of this project was to investigate causes of the discrepancy between the predictions of the utility maximisation model and the data, and to implement better predictive models.
I investigated three further models; a nearest unit model, a mixed utility maximisation-nearest unit model and a capacity based model. I found that these all gave better results than the utility maximising model when using the chi-squared goodness of fit test to compare with the data. We concluded that incorporating constraints on choice into the model; such as a probability of not making an informed choice and caps on the number of births per unit per day, improved the agreement of the predictions with the data. These results are able to have a direct impact on the way the NHS predict maternity service usage.