DSM criterion. I vastly prefer DSM-IV-TR conceptualizations to the more flawed DSM-V. Thankfully, i rarely if ever run into criterion differences that would impact the diagnosis. I code everything via ICD-10 nomenclature.
My typical referral is for neuropsych eval for TBI and/or polytrauma, so I almost always have the advantage of patient report, clinical judgement, and objective data to inform my dx’s and recommendations.
I make provisional diagnoses with rule outs during the initial intake and explain in the assessment report (neuro, pain, pre-surg, etc) if any of the provisional diagnoses changed as a result of the data and/or outside records I didn’t have at the initial intake.