Loading Tree…

DQI-2002

Definition

The proportion of eligible observational units for which no information could be obtained.

Explanation

The “Nonresponse rate” takes only eligible subjects into account and definitions are guided by the American Association for Public Opinion Research (AAPOR) Standard Definitions. Depending on the type of study uncertainty exists as to who is eligible. This may be the case, if the available information about the target sample does not comprise all necessary information about eligibility. For example, the available address might not be up to date. Therefore, in case of a nonresponse to an invitation it may not always be decided with certainty whether the recipient of the invitation was truly eligible.

Nonresponse rates may be computed at all levels of a study, e.g. nonparticipation in an entire study, a follow-up study, a study segment or selected items. All levels other than nonparticipation in a study resemble different degrees of partial responses to a study.

Example

In a cohort study, the target sample comprises 1000 adults with their primary place of residence in a defined region. All subjects have been selected based on population registries. All members of the target sample receive postal invitations. In case of no response contact attempts are made via phone or house visits. Because of the applied mailing method, it is reliably known, that mailings are returned, should the recipient have moved. In the latter case, a new address is provided, if known.

Oriented at American Association for Public Opinion Research survey definitions (AAPOR, 2016 edition), a typical computation is:

Participants

  • I, P: Complete or partially complete examination (N=600)

  • R: Refusal (e.g. target participant declines to take part) and break-offs (an initiated examination is ended at an early stage) (N=100)

  • NC: Non-contacts (e.g. target participant with known eligibility unavailable) (N=40)

  • O: Other eligible no-responder (e.g. is unavailable at projected examination dates, examination data projected but examination not yet conducted) (N=110)

Unknown eligibility, no examination

  • NE: Not eligible (e.g. Persons deceased before the sampling date and persons living outside the sampling region do not belong to the target sample, duplicate listings) (N=150)

In total, 600 out of 850 eligible adults participated, leading to a nonresponse-rate of 29%.

In contrast, a crude missing value indicator at the level missing units would have indicated a proportion of 40% based on 400 out of 1000 individuals who did not participate.

Guidance

Nonresponse-rate related indicators should be computed whenever a full available coding of missing data values is available.

In case of uncertainties about eligibility, AAPOR suggests to assume that a certain proportion of all observational units with unknown eligibility is not-eligible and should be removed from the denominator.

Nonresponse-rates are always equal or lower compared to corresponding missing value percentages.

Interpretation

The higher the nonresponse-rate, the lower the data quality.

Implementations

Literature