Identifying commonalities and differences between EHR representations of PASC and ME/CFS in the RECOVER EHR cohort.

Observed Findings

• Approximately 50% of patients identified by one condition phenotype were also identified by the other phenotype.
• Dyspnea, fatigue, and cognitive impairment were the most predictive conditions shared by both PASC and ME/CFS machine learning models.
• Cardiopulmonary conditions showed partial specificity for PASC diagnosis, appearing more frequently in PASC than ME/CFS records.
• Pain and sleep disturbances were more commonly recorded with ME/CFS diagnoses than PASC diagnoses.
• Recorded conditions clustered into three major groups: cardiopulmonary, neurological, and comorbidity-related patterns.

Inferred Conclusions

• PASC and ME/CFS show substantial overlap in their clinical presentations, supporting the possibility of shared pathophysiological mechanisms.
• Cardiopulmonary involvement appears somewhat more characteristic of PASC, while neuropsychiatric and pain-related symptoms are more typical of ME/CFS.
• The substantial overlap in core symptoms (fatigue, dyspnea, cognitive impairment) suggests that symptom management strategies could be effectively shared across both conditions.
• Data-driven approaches can help distinguish PASC and ME/CFS while identifying their commonalities, potentially guiding coordinated treatment approaches.

Remaining Questions