Forecasting Healthcare Demand

Why is healthcare forecasting important? Because it prevents deaths!

Hospitals. Does the idea of being in dire need of medical attention freak you out? Does the idea of overwhelmed healthcare staff in this situation freak you out even more? If not, you can probably stop reading and get on with your daredevil life, but if you are like me you will understand the importance of staffing and efficient healthcare forecasting.

Hospitals face demand uncertainty and – as a consequence – a trade-off between (temporary) staff cost and patient care/safety: Overstaffing is costly and unsustainable, i.e., it represents a waste of financial resources that could be utilized to improve patient care by investment in new technology, funding of clinical trials, research, etc. Understaffing could severely jeopardize patient care – several studies in industrialized countries document a surge in deaths on weekends, consistent with a reduction in clinical personnel that does not match the consistent day-to-day burden of disease (e.g., Lamn, 1973; Rogot et al., 1976; Trudeau, 1997; Evans et al., 2000; Bell and Redelmeier, 2001; Aylin et al., 2010). Empirical studies have shown that a lower nurse-to-patient ratio can increase medication errors (Frith et al., 2012) and increases the occurrence of avoidable events such as inpatient falls and hospital-acquired pressure ulcers (Staggs and He, 2013).


Healthcare Forecasting

Extant approaches focus on aggregate forecasting…

You probably already see the point I am trying to make: If healthcare demand forecasting is bad, this will lead to bad staffing, which in turn will lead to your death. Or maybe not your death, but the numbers don’t lie — it most certainly leads to deaths. “Well, that’s not good”, you say? Agreed, that’s not good at all. Extant healthcare forecasting focuses on either the total number of admissions on the aggregate (i.e., hospital) level (e.g., Kim et al., 2015) or focus on specific subsample of total admissions, such as emergency admissions (e.g., Jones et al., 2002, 2008, 2009; Abraham et al., 2009; Schweigler et al., 2009; Sun et al., 2009; Kadri et al., 2014), but there is no application of hierarchical forecasting that uses information at all levels of disaggregation (aggregate, emergency vs inpatient, divisions, etc.).

Such an approach is needed, as there is no clear consensus on health care staffing best practices – while CNOs and other administrators predominantly optimize staffing on the division level, there are advocates of centralized staffing who argue that only a centralized view of staffing needs can ensure an efficient allocation of staff throughout the entire organization – i.e., an efficient matching of full-time nurses and temporary staff with patient care needs (Crist-Grundman and Mulrooney, 2011; AORN, 2013). This indicates the need for a combination of decentralised monitoring and centralised planning in order to achieve effective staffing and consequently a balance between quality, safety, labour costs, and staff satisfaction.

…we exploit the hierarchical organisation structure of hospitals!

In current work in progress, we aim to close this gap in healthcare forecasting by using hierarchical time series (HTS) methods that can be used in a system to optimally model health care demand at both the aggregate level and disaggregated levels (divisions, primary specialities, etc.). Our goal is to decrease the complexity of staffing management using this system forecast, which makes it possible to efficiently plan the total requirement for temporary staff (using the aggregate forecast), as well as efficiently (pre-)allocate the staff to the disaggregated units (using the disaggregated forecasts). In addition to HTS methods, we explore the value of forecast combination for disaggregated healthcare forecasting. Preliminary results of our empirical study using a large sample of disaggregated hospital data are very promising for both HTS and forecast combination and indicate a significant improvement in forecasting accuracy over the aggregate approach.

Stay tuned for more detailed results and leave a comment on your views about the need for disaggregated healthcare forecasting!

Leave a Reply

Your email address will not be published. Required fields are marked *

Protected with IP Blacklist CloudIP Blacklist Cloud