I'm more familiar with the BSI (the 53 item version derived from the SCL-90), having been involved in the development of screening for prisoners. The manuals will usually suggest a T-score of 63 (I believe) as a cut-off, but it won't diagnose as far as I can see. When we looked at cut-offs in a prison population against an independent assessment by a mental health nurse of whether an inmate needed services, we found that lower cut-offs were often needed to achieve the best balance of sensitivity and specificity (you'll find the reference to the publication on my profile - "A comparison of scoring models for computerised mental health screening for federal prison inmates" - and I can send you a copy if you don't have access).
I'm not sure if this is helpful for your purposes, but I'm hoping it will steer you in the right direction. I should also qualify this response with the fact that I am not a clinician, but have been involved in a number of research projects using these tools.
Comparison with the norms based on the "normal" population are not helpful because those scores have a very skewed distribution toward the low end of the scale. As a consequence, even a mediocre raw score will look high against such a norm. I advise to search for norms based on people seeking milder psychological help and based on psychiatric patients.
Even such a comparison needs to be supplemented by a look at the client's average raw score per scale. Such for the following reason: if certain complaints are quite common among the norm groups advised above, for instance depressive complaints, then even a rather high raw score on depression may yet receive a moderate norm score because the norms are high. For that reason I have written my own scoring program which calculates the average raw scores on the scales as well. Moreover, it provides an account of the individual items, arranged according to their raw scores from high to low. That kind of output is much more useful for a therapist than just norm scores.
I have been working with the Dutch version of the SCL-90R, so the norm scores I use and the program I have written will be of little service to you.
I added the German manual of the SCL-90(r)-S on Researchgate. You could detect psychological distress, if the T-score is T>=60. You could prove the "Case Definition", 2 T-scores and/or T(GSI) >=63. It is usual to compare the individual score with normative data. Sincerly yours, GHF
Dear colleague valuable data in 3 levels which are useful in interpretation of the respondent's psychological status. In discrete symptom level, % score of items can be compared to that of norm group A,B,C . A table example of such a comparison is in p 24 of SCL-90-R manual. It is useful to get an insight to the importance level of each item. But if you calculate and specify those items with Quite a bit and severely, you can access to the symptom of note and can interpret dimensional level finding more precisely. In dimensional level psychological status of the respondents can be elucidated. The information of this level can put many clues in hand about each dimension. You can convert the scores of each dimension to area T-score and compare them with favorite norm group.According to the nature of your research normative group should be chosen.In pages 22 and 23 of the manual you can find a pattern for such a comparison.a T score of 60, regardless of the symptom dimension, will place an individual in the 84th centile of the normative sample; a T score of 70 will place the respondent in approximately the 98th centile. Obviously, the same raw score will fall at differing points in various normative distributions (e.g., adolescent vs. adult nonpatients) and therefore may be interpreted differently depending on the norm chosen as a referent.
In global level, global indices especially GSI, provide a valuable single score information regarding the overall psychological status of the respondent.