Background: To address three methodological challenges when attempting to measure patients' experiences and views of a system of inter-related health services rather than a single service: the feasibility of a population survey for identifying system users, the optimal recall period for system use, and the mode of administration which is most feasible and representative in the context of routine measurement of system performance.
Methods: Postal survey of a random sample of 900 members of the general population and market research telephone survey of quota sample of 1000 members of the general population.
Results: Response rates to the postal and market research telephone population surveys were 51% (457 out of 893 receiving the questionnaire) and 9% (1014 out of 11924 contactable telephone numbers) respectively. Both surveys were able to identify users of the system in the previous three months: 22% (99/457) of postal and 15% (151/1000) of telephone survey respondents. For both surveys, recall of event occurrence reduced by a half after four weeks. The telephone survey more accurately estimated use of individual services within the system than the postal survey. Experiences and views of events remained reasonably stable over the three month recall time period for both modes of administration. Even though the response rate was lower, the telephone survey was more representative of the population, was faster and cheaper to undertake, and had fewer missing values.
Conclusions: It is possible to identify users of a health care system using a population survey. A recall period of three months can be used to estimate experiences and views but one month is more accurate for estimating use of the system. A quota sample market research telephone survey gives a low response rate yet is more representative and accurate than a postal survey of a random sample of the population.