What 265 families reveal about child versus parent reporting

When a study involving children collects outcome data from both the child directly and a parent as proxy, it's tempting to treat the two sources as interchangeable, or at least as two measurements of the same underlying thing that should mostly agree. A psychometric evaluation of the PROMIS pediatric and parent-proxy anxiety short forms, using data from 265 families with children averaging 11.14 years old enrolled in the Kids FACE FEARS trial, found that assumption doesn't fully hold up.

Both instruments worked well on their own terms

Before getting to the disagreement, it's worth being clear that both versions of the scale performed soundly individually. Confirmatory factor analysis supported adjusted single-factor solutions for both the youth-report and caregiver-report versions, meaning each instrument measured a coherent, internally consistent construct rather than a jumble of unrelated items. High omega coefficients, a measure of internal consistency, were reported for both forms. Both versions also showed medium-to-large associations with anxiety-related impairment and severity, supporting that each is genuinely measuring something meaningful about anxiety, not just producing a number that happens to correlate weakly with the condition it's meant to capture.

In short: neither instrument was the problem. Each one, evaluated in isolation, is a psychometrically sound tool.

The disagreement shows up specifically between the two sources

Where things got more complicated was in comparing the two reports directly against each other. The study found moderate, rather than strong, cross-informant reliability between child self-report and parent-proxy report. That's a specific, quantifiable way of saying: a child's own account of their anxiety and their parent's account of that same child's anxiety don't align as closely as a single, unified measure of "the child's anxiety level" would imply.

This is a well-documented pattern in mental health assessment more broadly, not a flaw specific to this particular scale or trial. Children and caregivers often have genuinely different vantage points on internal experiences like anxiety: a child has direct access to their own subjective distress in a way a parent doesn't, while a parent may be picking up on behavioural signals, or applying their own baseline expectations about what counts as anxious behaviour, that don't map precisely onto what the child would report about their own internal state.

Why "moderate" reliability is a genuinely important number, not just a caveat

For a study relying on either source alone, this finding has direct consequences. If a trial uses only parent-proxy report because it's more practical to collect, for younger children or shorter study visits, it's capturing a related but distinct construct from what the child would report about themselves, not simply a slightly noisier version of the same measurement. The gap isn't measurement error in the usual sense that averages out with a larger sample; it reflects a genuine difference in perspective between the two reporters.

That has implications beyond the specific anxiety scale studied here. Any pediatric or child-inclusive study weighing whether to collect child self-report, parent-proxy report, or both, is making a real methodological choice with consequences for what the resulting data actually represents, not just a practical decision about convenience or participant burden.

What this means for designing pediatric outcome measurement

A few practical implications follow from a finding this specific:

• Where feasible, collect both child self-report and parent-proxy report rather than choosing one as a stand-in for the other. Given only moderate cross-informant reliability, one report is not simply a proxy for the other; each captures something at least partially distinct.
• Be explicit in a study's design about which perspective the primary outcome is actually measuring. A protocol that says "anxiety will be measured using PROMIS" without specifying which reporter, and why, is leaving an important methodological decision implicit rather than deliberate.
• Don't assume disagreement between child and parent reports indicates a data quality problem to be resolved. The moderate correlation found here is a structural feature of measuring internal states through two different reporters, not evidence that one report is simply wrong and the other correct.
• Age and developmental stage plausibly affect how much the two reports diverge, and a study spanning a wide paediatric age range should consider whether cross-informant reliability might vary across that range, since a single aggregate reliability figure, like the one found across this study's 11-year-old average age, may not hold uniformly at every age within the sample.

The broader lesson for any research involving proxy reporting, whether from parents, caregivers, or another informant standing in for a participant who can't reliably self-report, is that proxy data is genuinely useful and often necessary, but it isn't a direct substitute for self-report. Both instruments here were well built. The interesting, useful finding is precisely that even two well-built instruments measuring the same nominal construct from different vantage points don't converge as tightly as a single "true anxiety score" model would predict.