*Analysis*

A preliminary codebook for this study was developed prior to data collection with template codes based on the study's research questions and relevant technical assistance literature similar to the code manual development described by Fereday and Muir Cochrane [35]. Four broad code categories were included in this a priori codebook: perception of TA role, intensity of TA, impact of TA, and importance of TA. These broad categories and sub-codes within each category were included in the a priori codebook with a definition and description of each code. Emergent codes were added during the preliminary analysis steps, and are described below. The codebook used for this study was separate to that used for the overall evaluation of the HWC process.

The in-person focus group with UIC TA providers and all phone interviews with labor expert TA providers and TA recipients were audio recorded and professionally transcribed. The transcripts were analyzed using a hybrid approach that involved both inductive and deductive coding and theme development, similar to the approach described by Fereday and Muir-Cochrane [35]. Dedoose software (Dedoose Version 7.0.23, web application for managing, analyzing, and presenting qualitative and mixed method research data, SocioCultural Research Consultants LLC, Los Angeles, CA, USA) was used for all qualitative analyses in this study.


**Table 2.** HWC Participants. TA: technical assistance, UIC: University of Illinois at Chicago.

\* Note: Two of the worker center representatives served in both labor expert presenter roles and participant team roles in the HWC initiative.


#### **Table 3.** Data Collection Instruments.

Immediate post-HWC interviews with labor expert TA providers and the focus group with UIC TA providers were coded by a single coder. For each transcribed interview, the following analysis protocol was used:

(1) Each full interview and the focus group transcript was read and key points were summarized in a memo. At this point, additional codes were added to the codebook based on new categories that emerged from the textual data.

(2) Then, a priori codes and emergent codes were applied to the interview text where text segments were considered representative of and matched the definition of an individual code.

(3) As segments of text were coded, each new excerpt was compared with segments that had previously been assigned the same code. In the event that a code did not seem to fit for both segments, a new code was added to the codebook, and relevant sections of the transcribed interview were recoded.

(4) After all interviews were coded and additions to and refinements of the codebook were complete, a new cycle of coding began. Each interview was re-coded using the updated codebook.

(5) After the second coding cycle, final coded segments were read and subjected to a process of clustering around similar patterns. Themes were identified when all data supporting a given pattern were clustered and saturation was reached. At this stage, differences in themes across interviewees were examined.

Three-month follow-up interviews were coded by two separate coders, and a slightly different analysis protocol was used. Steps 1 and 2 from the baseline and immediate post-HWC interview protocol were followed, as described above, with both coders reading and summarizing transcribed interviews and collaboratively making additions to the codebook. Then, the following steps were completed by the two coders in lieu of steps 3–5 from the baseline and immediate post-HWC analysis protocol: After each transcribed interview was coded by both coders, coded segments were compared for agreement. In the event that the two coders did not agree on coding for a particular segment, they discussed the segment and attempted to come to agreement as to which code(s) should be applied. In the event that the two coders could not come to agreement, a third coder was asked to code the interview and discuss applied codes with the original two coders. Additionally, in the event that no codes seemed to fit a given segment, a new code was added to the codebook, and relevant sections of the transcribed interview were recoded. Both coders reviewed already coded interviews for a comparison of applied codes and recoded those interviews as needed to reflect codebook updates. Final coded segments were read and a process of clustering around similar patterns and themes began. At this stage, differences in themes across participant type were examined.
