As aforementioned, SMART Flags are applied on the basis of statistical plausibility to ensure better data quality. If the surveyed population follows a normal distribution, SMART assumes a standard deviation of 1 (statistical properties of a normal distribution) to identify any SMART Flags that lie outside of ±3 of the observed mean. If you have 1000 children measured and for example there were 20 SMART flags identified, only 3 of those children are expected versus the remaining 17 are big mistakes.
On a day-to-day basis, you are generally only looking at WHO Flags (±5 of the reference population’s mean). It doesn’t make sense to use SMART Flags throughout the data collection as the observed mean on a day-to-day basis or for a given team will be different from the overall observed mean once the data collection has finished. You may make sense at start looking at the number of Flags after one week of data collection after retrieving all of the data from the teams; however, any flagged data should never be erased or removed at any stage. The reason for Flags is to highlight any big mistakes, whether it is during data collection with WHO or at the end of data collection with SMART. Consequently, the survey manager should double-check each flagged measurement with the paper or electronic copy of the questionnaire to ensure that no data entry errors were made. If the same values appear on the questionnaire, then these flags remain and you will obtain the associated amount of penalty points based on the SMART Plausibility Check.
