How do I identify skip patterns to create measures for hospitalization for mental health problem and hospitalization for substance abuse problems?
The Interactive Codebook section on the CPES Web site provides the universe for each question, that is, the path for each question. Having said that, however, this is a complex set of questions made even more so by the fact that it is impacted by the long/short version of the NCS-R (see ?About CPES? for a description).
The following logic is all online but varies by study.
- NCS-R: All respondents who reach the Services (SR) section (i.e. are in Long Group) are asked SR2, regardless of their response to the "ever hospitalized" item in each of the earlier disorder sections. So, the missing cases at SR2 are only Short Group respondents (good news here is that if a respondent got as far as the "ever hospitalized" question in an earlier section, there is a very high likelihood that they would have been in the Long Group -- so you don't have to worry much about missing respondents who never reached the Services section).
- NSAL: There is an extra checkpoint (SR1) that evaluates the responses to all of the earlier "every hospitalized" items. If at least one YES response to the disorder-specific items, SR2 is not asked. So, for the missing SR2 cases, you will need to cross-reference all of the earlier items to see which one was endorsed (please also note that White respondents were not asked the Services section, so all Whites contribute to the missing values). The entire list of disorder-specific hospitalization items is included in the on-line documentation for SR1.
- NLAAS: Same structure as NSAL -- extra SR1 checkpoint that evaluates earlier disorder-specific hospitalization items. If at least one YES, SR2 is not asked. Again, the entire list of hospitalization items is listed at the SR1 online documentation.
You also indicate an interest in knowing whether the respondent was hospitalized for either a mental/emotional problem or a substance problem. There is no way to know this for sure, since SR2 lumps these all together. However, you could go back to each of the section-specific items and determine which were endorsed. The pitfall here is that respondent may give a positive answer at SR2 without having revealed a hospitalization in earlier sections -- in this case, you would never know the cause(s).