The digital audio recording was password-protected and encrypted, per IRB instruction, and the transcriber was required to sign a confidentiality agreement. After the interview
transcription was verified by the interviewer, the recordings were destroyed within six months. As the principal student investigator, I have been responsible for data-cleaning and preparing the final datasets for analysis. I developed a respondent matrix and utilized the software package NVivo in conjunction with manual coding. NVivo is a computer-assisted qualitative data analysis software (CAQDAS). Grounded theory methodology in this instance will allow for the incorporation of the perceptions and perspectives of those most affected by the intersection of multiple collateral consequences and stigmas in relation to the criminal justice system.
Analysis began early in data collection, along with a continuing review of the literature, researcher experience, and the interviews which all informed the analytic process. For the purposes of this study, the terms, “concept,” “category,” and “theme” are not interchangeable. Category was descriptive, and a priori categories were deployed as well as new ones developed. The use of semi-structured interview guides provided the initial descriptive framework for analysis. Therefore, the themes identified were strongly linked to the data, however, themes did not necessarily directly reflect questions used in interviews (Braun and Clarke 2006; Creswell 2014; Patton 2015). Through this identification of categories and subcategories, I began to identify candidate themes (Bazeley 2009; Braun and Clarke 2006).
Due to the process of ongoing analysis coupled with simultaneous data collection, findings have been refined, challenging generalizations through a more in-depth analysis of the data. In some instances, I discovered that initial findings continued to be supported, or if support weakened, new themes were developed, combining existing ones or creating subthemes. I have not restricted analysis to individual cases, but have been accessing information from across all data, interview transcripts, field notes, and basic institutional numbers, such as number of attorneys/social workers per office and number of cases per office and per social worker (Miles and Huberman 1994; Henderson and Balfour 2015).
As noted, the client interview guide for the ACLNID is based upon general civil legal service checklists. Additionally, the client interview guide consisted of 70 questions (Appendix F) which were designed to gather background characteristics and to elicit individual experiences regarding housing, safety and security, financial, benefits and subsidies, health, mental health, military, education, employment, driver’s license, voting, and general questions regarding people’s experiences with civil legal needs agencies. Following the suggestion of Steinar (1996) and Patton (2002), all the interview guides (Appendices F, G, H) were designed according to methodological principles, as were the follow-up questions and probes used to clarify and expand responses. Participants were also repeatedly reminded throughout the interview that they were not required to answer any of my questions per IRB requirements.
A participant matrix was created concurrently during the ongoing data collection. Along with basic demographic information such as age, race, and sex, additional categories were also identified and tracked. One example was whether or not a person had graduated from high school or obtained a GED. Another asked whether the participants were homeless at the time of their arrest, or if not, if they had experienced homelessness previously. Also included were questions
about health insurance, mental health issues, driving license status, type of state identification, and ability to vote.
The words “stigma,” “stigmatizing,” or “labelling” were not included in the semi- structured interview guides. These questions sought to identify how issues of stigma could be impacting a client’s willingness to self-disclose. I included the entirety of the transcripts for coding about stigma and mental health diagnosis because although highly structured,
interviewees introduced issues of mental health throughout the interview and did not necessarily remain specific to the mental health section. Participants were not restricted to respondents who met any psychiatric diagnostic criteria for mental illness.
These questions included specific inquiries Q57, Q58, and Q59. For example, Q57 “Have you ever been diagnosed with a mental illness? If yes, where? When?” Or question 59, “have you ever been hospitalized for mental illness? If yes, When? Where?” Also included are general questions such as 62 “Is there any reason you have not or would not seek help for a mental health issue, either your own or a family member?” (Appendix F).
Each interview was recorded using a digital recorder, and these were subsequently transcribed by a professional transcriptionist. I listened to the recordings and verified the transcripts, as did a graduate research assistant from the College of Law. The data was cleaned, missing words identified, and each version saved in a separate password-protected file. When the interview was formatted and verified, I entered it into NVivo. Client, attorney, and social worker interviews were identified as subjects and cases, and interview questions were treated as a priori codes and entered as nodes. After entering the transcribed interviews into NVivo, I and a law school graduate assistant provided additional coding to identify civil legal needs, as well as expand mental health coding.
Initial coding indicated expectations of prejudice and discriminatory treatment among those who disclosed that they have been diagnosed with a mental illness. From these, I then went on to code for perceived and anticipated experiences of stigma. Literature points to confusion and vagueness of terms in stigma research (Pescosolido and Martin 2015). Therefore, so that this research could be situated within the existing literature, I sought to use accepted terms described by previous researchers.
Participants employed a variety of stigma-resisting strategies, and I coded these according to the deflection strategies conceptualized by Thoits (2011, 2016; Thoits and Link 2015). The strategies themselves will be explained in greater detail in subsequent sections. I also developed new conceptualizations as information emerged from the data.
Similarities and differences within and between client, attorney, and social worker interviews emerged around areas of consensus and diverging views related to the identification of barriers to help-seeking and disclosure of mental health issues. Differences and similarities between varying demographics also became apparent; most notably, differences in education attainment correlated with different perceptions of mental illness stigma pertaining to multiple areas of diagnosis and etiology. The interaction of text, codes, and themes in this study involved several iterations before analysis proceeded to the interpretative phase. At this point, connections were drawn with existing or newly developed theoretical constructs.