Assessing Service Needs: The Sampled Survey Approach

The standard survey-based approach to assessing local service needs.

This is the approach most commonly taken by organisations wishing to estimate service needs in local areas. An appropriately-sized sample of individuals is thus used to estimate local prevalences of socio-economic or lifestyle characteristics linked to service needs. For instance, a survey revealing the local prevalence of smoking or other health-risk behaviour might provide an appropriate basis upon which to target area-based programmes aimed at smoking reduction.

With expertise in questionnaire design, implementation and analysis, RAE Consulting can support clients wishing to undertake simple sample surveys, although our usual advice would be to use modelling techniques to enhance the effectiveness of the survey. This can help address a fundamental problem with surveys aimed at generating robust local area estimates; namely that such surveys must include a sufficient number of respondents from all local areas for which estimates are required. The smaller the geographic resolution the larger, and more costly, the survey.

The implications of this are illustrated by a recent postal survey aimed at obtaining ward-level lifestyle data on adults in a single Primary Care Trust (PCT) in the North West of England.  To obtain satisfactory prevalence estimates at this scale required a 5% sample of adults registered with the PCT.  The total cost of designing, implementing and analysing this survey was well in excess of £100,000. Few public sector organisations could justify such expenditure in order to establish the service needs of an adult population of less than 500,000.

Nevertheless, the sample survey approach can offer a cost-effective approach to needs assessment when the local areas for which estimates are required are relatively large—or indeed when all that is required is an assessment of the level of need in a service provider’s overall client population. In such situations the sampling ratio, and thus the number of respondents who would have to be surveyed, can be kept relatively low. In all other circumstances we recommend clients adopting either a modelled survey or synthetic estimation approach.