Talk:Evidence-based assessment/Step 3: Evaluate risk and protective factors and moderators

From Wikiversity
Jump to navigation Jump to search

Discussion Points[edit source]

Interesting to think about how to organize these. There are certain generic risk factors for diagnosis (nonspecific risk factors, like poverty, low educational attainment, malnutrition), others that are transdiagnostic (affecting clusters of diagnoses, but not all pathology -- short allele of HTTPLR, trauma, sleep disruption...); and a different category could be things that change treatment selection.

These are bundled into one group right now because they are poorly developed and articulated in the field. A more mature approach might have a set of short checklists:

  • General risk factors (ones that are common and cut across most forms of pathology). In a quick and dirty approach, these might have a general DLR+ of 2 or 3, or research might help identify how much these change for specific targets.
  • Intermediate Transdiagnostic Risk Factors (ones that affect sets of diagnoses). These might have DLRs that apply to a group of conditions, but get ignored for others.
  • Target specific risk factors: Currently a very short list for each condition, but likely to change rapidly if the "personalized medicine" approach gains traction. These might not be routinely asked in initial assessment, but could get unlocked and used if other assessment findings move a target probability into the yellow zone.
  • Treatment specific checklists: These could be brief lists that remind us to ask about treatment moderators. A first pass could be a simple box score, keeping track of how many treatment enhancing factors or treatment reducing factors are present, and then discussing with the client and seeing if the combination of moderators and patient preferences results in a change in treatment. Further in the future, studies might use these to come up with predicted treatment response for a variety of conditions. Imagine a dashboard showing the list of top EBPs, with probabilities of successful outcome for each (or projected average effect size in symptom reduction) -- customized for the patient's moderators. If that were online, then it could be linked with Wikipedia pages about treatment, and find-a-therapist directories and online tools.

Where do we find these risk factors? Morrison's book lists many of the general ones. Handbook chapters are being encouraged to include a section about risk factors, and another about treatment moderators. We could skim those to build a version zero of the lists. It would be pretty easy to do some initial validation and estimates of effect sizes from secondary analyses. (Could also play with dropping these into machine learning models if there were appropriate targets for supervised learning, or good indicators to use in unsupervised learning).

One technical reminder: When evaluating the psychometrics of the checklists, remember NOT to use internal consistency measures. These are causal indicators, not something that we would expect all the items to reflect a single underlying factor. Could talk with Ken Bollen about whether there are fancy statistical ways to approach this, but clinically the goal is a brief "greatest hits" of key risk factors -- DIFFERENT risk factors, and the proof is in the validity research.