Measuring the reliability of proxy respondents in behavioural assessments: an open question

In behavioural assessment, information can be gathered from internally referenced self-reports or from proxy informants.

This study aimed to fine-tune a brief but reliable method for evaluating the proxy accuracy in cases where responses obtained from adult and older adults' patient cannot be considered reliable.

We generated a set of items reflecting both overt and covert behaviours related to the basic instrumental activities of daily living. The psychometric properties of the content, factorial, and criterium validity of these items were then checked. The Proxy Reliability Questionnaire-ProRe was created. We tested the frequency of "I don't know" responses as a measure of proxy reliability in a sample of healthy older adults and their proxies, and in a second sample of proxy respondents who answered questions about their parents.

As expected, response precision was lower for items characterizing covert behaviours; items about covert compared to overt behaviours generated more "I don't know" answers. Proxies provided less "I don't know" responses when evaluating the parent, they claimed they knew better. Moreover, we tried to validate our approach using response confidence. Encouragingly, these results also showed differences in the expected direction in confidence between overt and covert behaviours.

The present study encourages clinicians/researchers to how well the proxy the patient know each other, the tendency of proxies to exhibit, for example, response bias when responding to questions about patients' covert behaviours, and more importantly, the reliability of informants in providing a clinical assessment of neurocognitive diseases associated with aging.