To identify conceptual and methodological concerns with Azola et al.’s evaluation of the RECOVER PASC score as a Long COVID classification tool.
Approach:
Sensitivity assessment: The commentary argues that recruiting cases from a Long COVID clinic may overrepresent people with multiple symptoms, creating spectrum bias. Requiring at least one neuropsychiatric symptom may also bias sensitivity upward because that symptom category contrib…
Specificity assessment: The authors argue that fully recovered participants and healthy controls do not represent patients with overlapping symptoms from other causes, potentially inflating specificity.
Predictive values: The commentary notes that case-control studies do not reflect disease prevalence in a target population and therefore cannot establish positive or negative predictive values.
Key Findings:
The commentary argues that the clinic-based case sample may not represent the broader population of people with Long COVID and may overrepresent those with multiple symptoms.
The requirement that cases have at least one neuropsychiatric symptom may have biased estimated sensitivity upward.
The commentary argues that using fully recovered individuals and healthy controls rather than people with overlapping symptoms may have inflated estimated specificity.
The commentary concludes that the study did not validate the PASC score: it argues that sensitivity and specificity estimates were biased and that a case-control design cannot establish predictive values for a population with a given disease prevalence.
Interpretation:
The commentary characterizes Azola et al.’s results as estimating the proportion of patients in their Long COVID clinic who had at least one neuropsychiatric symptom and met the PASC score threshold, rather than validating the score for diagnostic classification.
Limitations:
The commentary critiques Azola et al.’s evaluation and does not report an independent validation study. It notes that case-control validation cannot determine positive or negative predictive values because it does not reflect disease prevalence in a population of interest.
Conclusion:
The commentary concludes that Azola et al.’s study did not validate the PASC score, arguing that the clinic-based case sample and comparison groups biased sensitivity and specificity estimates, while the case-control design could not establish positive or negative predictive values.
