Forecasting methods in healthcare planning

However, capacity and financial planning forecasts for healthcare services can be frustrated by a lack of means to allow for such ‘health initiatives/deflections’ and ‘medical developments/disease pressures’. Data are not necessarily available from which to draw deductions, and the challenge is complicated by the various inter-relationships between initiatives: someone improving their lifestyle will probably stop smoking, improve their diet, reduce their alcohol intake and take more exercise, all at the same time.

Given these circumstances, one practical and cost-effective means of forecasting is to apply Delphi methods, which are used in healthcare to obtain accurate professional consensus and information. They have been used for priority setting and guideline development in health-related research¹, and offer a structured, transparent and replicable way of synthesising individual judgments.

The principal aim of Delphi methods is to achieve and measure consensus.Delphi methods were originally devised to be used to obtain consensus from experts where little or no data were available. Delphi studies are undertaken by repeated surveys/interviews over two, three or more rounds, and require responses from individuals who have some knowledge about the subject in question. Accordingly, this can include GPs, specialists, commissioners, community staff, and patients and carers.  

The summarised findings of the group as a whole are presented back to each individual, and their responses to the same questions sought again. The process is repeated until individuals no longer change their responses, at which point the consensus of the group is summarised. Typically little change is seen after three phases.

Delphi methods can either return to the same individuals within a group, particularly when that group of experts is small, or seek responses in different phases from new individuals with similar expertise who have not expressed a response to an earlier phase.

Options for undertaking Delphi methods include: distribution by post/email; via internet; structured events linked to conferences and seminars; and classical one-to-one interviews.

Delphi is based on well-researched principles and provides forecasts that are more accurate than those from unstructured groups^2,3. They are particularly useful where there is little or limited knowledge^4,5,6,7 - a situation common in healthcare planning.

There is no single way of conducting a Delphi study; hence the concept refers to general principles and processes of arriving at solutions to problems through several cycles of revision based on each individual’s feedback.  The end result is considered a better solution than just one individual could have reached⁸.

Pharmaceutical companies have used Delphi methods to forecast changes in prescribing practices  which have been validated by observation of eventual outcomes. In 2007 a change in remuneration structure relating to the use of lipid lowering drugs was introduced. A Delphi study forecast an 8.4 per cent increase in the use of a particular medication in the following six months. The increase was found to be 8.0 per cent. Delphi methods have also been used to formulate treatment pathways.

A 2008 Delphi survey of over 300 GPs in Northern England⁹ projected a 2.2 per cent reduction in smoking prevalence over 10 years. This took account of a 30 per cent reduction in the existing cohort of smokers with new/returning smokers equivalent to 21 per cent of the cohort. This suggested that approximately one in six smokers need to set a quit date each and every year to achieve even a 2.2 per cent reduction, compared to the current one in 16.

A 2009 Delphi survey of the Cheshire & Merseyside Cardiac and Stroke Networks predicted mean trends for heart-disease related hospital cases over the period 2008–2025. They showed a net increase of 3.2 per cent for emergency care, and a net increase of 8.8 per cent for elective/planned care (i.e. 0.2 per cent & 0.5 per cent per annum respectively). The latter would be countered by a switch of 9.3 per cent of elective/planned care from hospitals to primary/community care.

The NHS has mountains of data on what has happened in the past, but little about the future, other than nationally published population projections and some specific research, for example on trends in cancer incidence. As is now automatically stated for financial products: “past performance is not necessarily indicative of future results”.

Planned changes for commissioning NHS services¹⁰ will not mean that the requirement for making forecasts will go away. Indeed it is likely that there will be a greater requirement for planning forecasts when changes to models of healthcare provision and hoped-for innovations are evaluated.