The results obtained from validating our forecasting model demonstrated that the ability to accurately forecast the NICU census was largely a function of the length of forecasts, a common feature of nearly all forecasting models.
As noted from Figure 3, the 1, 3, 5, and 7 day forecasts appear to capture the global trends in observed census counts, however as expected, locally the accuracy of our census forecasts diminishes as a function of the length of forecasts.
This feature is a likely candidate to explain the modest overestimation that was observed in the validation of our forecasting model, which became more pronounced for increasing lengths of forecasts.
Comparing the performance of such different methods, which use extrapolation, explanation, and/or expectation to forecast mortality, can be fair as long as concrete assumptions like the base period and the length of the forecast horizon are equal (Stoeldraijer et al. 2013).
It is also indicated that the deviation is approximately a linear function of the length of the forecast period.
We use values from 5 to 30 with an increase proportional to the length of the forecasting horizon.
Extensive Monte Carlo simulations, using dynamic factor models, clarify the relationship between the achieved reduction in forecast variance and various design parameters, such as the observation length, the number of predictors, and the length of the forecast horizon.
If population parameters are known, it is very easy to calculate the probability, P q (T p, T p + ΔT), for the event in the forecast period with the conditional probability, (1) where F N is the cumulative distribution function (CDF) of a normal distribution, ΔT is the length of the forecast period, x p = ln(T p ), and x f = ln(T p + ΔT).
As a result of the aforementioned works, the Health and Social Care Modelling Group (HSCMG) in collaboration with the Care Services Efficiency Delivery (CSED, Department of Health) developed a software implementation of the forecasting framework [ 10], known as FLoSC (Forecasting Length-of-Stay and Cost) (cf. http://www.healthcareinformatics.org.uk/flosc/).uk/flosc/
In the following section, we will apply this approach to the NHS continuing healthcare field using London PCTs data to forecast existing patients' length-of-stay and estimate the number of patients who are likely to survive.
To apply the proposed method, we only need the length of walkway and passenger forecast data, such as peak-hour factor ((varepsilon)), passenger peak-hour volume (q), mean headway between the trains (h) and the (c_{h}^{2}) of the arrival interval between the train.
