They are screening tools, built to over-refer rather than to be precise. That produces false positives, where someone is flagged who does not have the condition, and false negatives, where someone who has it scores under the cut-off. Treat a score as a prompt, not a finding.