2.3 Repeatable

Repeatable data collection means that for a study population, two or more different individuals would collect the same data if using the same approach/method (allowing for small differences due to random error). The data collection process should be designed to minimise the amount of subjectivity and interpretation left to the data collector. Repeatability is in part linked with standardisation, as clear definitions of the data to be collected will allow the same results to be collected by different observers.

For example, the World Health Organization (WHO) provides a standardised methodology called a point prevalence survey (PPS) for measuring AMU in hospital inpatients. Some questions that might arise include:

• Is a patient who was discharged at lunchtime on the day of the survey eligible for inclusion?
• If a patient is admitted half an hour after the survey starts, should they be included?

To address this, the WHO specifies that for the purpose of a PPS, an inpatient is defined as a patient who has been admitted to a hospital ward at or before 8 a.m. on the day of the survey. Applying this definition means that there is less room for interpretation of the methodology by the data collector, and hence the data collection process is more repeatable.

Inevitably, there will always be some inherent variability and uncertainty in measurements of data, which is related to random error or bias. This means that repeatability may not always be achieved. For example, the temperature of a patient taken by a nurse at the time of sample collection may be different from the temperature taken by the same nurse, at the same time, with a different thermometer. Recording additional metadata, such as the method used for measuring temperature, may help to understand the variability when analysing and interpreting the data.