Hello HBN community,

I am currently working with the HBN diagnostic data and defining clinical groups for my analysis.

I have two main questions:

1. Is it the standard and accepted practice to rely on the Diagnosis_ClinicianConsensus variables rather than the raw/computational KSADS variables (such as Diagnosis_KSADS_D/P/T or KSADS_C/P) when assigning final diagnoses for analysis?

2. Assuming the Clinician Consensus is the preferred variable, what is the accepted practice for handling the diagnostic confidence levels (Confirmed, Presumptive, Requires Confirmation, and Rule-out)?

a. Should we filter diagnoses based on these levels to ensure diagnostic accuracy?

b. I noticed a significant number of missing values (NAs) in the Confirmed and Presumptive columns for valid diagnoses. Given this, is it best practice to simply filter out diagnoses explicitly marked as RC or Rule-out (excluding them from the patient group)?

c. As a reference point, how was this handled in the foundational Alexander et al. (2017) paper? Did the reported sample sizes include all consensus diagnoses regardless of their confidence levels? Or was a specific filtering logic applied?

Thank you very much in advance,

Best regards,

Tal