Survey development

We designed a 40-item structured survey to evaluate symptom reporting behavior among NH staff. Survey constructs and questions were developed based upon confidential conversations with NH staff through our ongoing role as the Orange County NH COVID-19 Prevention Team.¹⁵ Although we did not use formal “grounded theory” methods,^{Reference Glaser16} we performed a review of the relevant literature and used a modified Delphi approach to formulate item content through iterative discussion, comment, and revision. Many NH staff disclosed to us that they had not reported ill household contacts or COVID-19–like symptoms to their supervisors. Staff often conveyed longstanding concerns with finances, stigma of missing work, reprisal from supervisors, as well as fears of known and unknown consequences of COVID-19. These conversations led us to examine 8 discrete constructs related to symptom reporting, including 4 barriers (ie, fear of COVID-19, lack of knowledge, monetary barriers, logistic barriers) and 4 potential solutions (ie, encouragement from supervisors/coworkers, adequate staffing, access to a confidential helpline, lack of stigma from coworkers for having COVID-19). Construct-specific items were developed to be scored on 5-point Likert scales. We also assessed participant demographics, health status, course of COVID-19 illness, frequency of participation in COVID-19–exposing activities, likelihood of seeking care when ill, and trust in COVID-19 vaccines. This work was approved by the University of California–Irvine Institutional Review Board.

Initial testing

Prior to launch, we asked 10 NH staff to complete and comment on the survey using cognitive interviewing techniques.^{Reference Willis17} We asked respondents about the appropriateness of survey language, word choice, format, time required to completion, and understandability of questions. Based on these responses, we modified survey language, format, and ordering of response options. The survey instrument is provided (Supplementary Material online).

Survey recruitment and administration

We conducted telephone-based surveys of NH staff in Orange County, California, from December 2020 through February 2022, with the goal of recruiting 120 participants. The survey was rolled out in 2 phases: wave 1 during the first pandemic wave and wave 2 during the subsequent SARS-CoV-2 δ (delta) and (omicron) waves. During this time, NHs were conducting routine and symptom based testing of staff. Recruitment of SARS-CoV-2–positive staff was facilitated via posters shared with all 70 NHs in the county, weekly outreach to NHs with known COVID-19 staff cases based upon publicly reported data,¹⁸ and referrals from NH leadership. We applied the following inclusion criteria: a positive SARS-CoV-2 test within the prior 8 weeks, current employment in 1 of the 70 NHs in the county, at least 18 years of age, and willingness to participate.

Each participant received a $50 electronic gift card for completing the 20–30-minute phone-based survey. Surveys were administered in English and Spanish. Participants were reassured that their participation and responses would remain confidential and would not be shared with public health, coworkers, or supervisors. Data were recorded by trained study staff using a standardized instrument in the Research Electronic Data Capture (REDCap) system.^{Reference Harris, Taylor and Minor19}

Hypotheses

Due to expanding information and access to COVID-19 vaccines over time, we expected that staff would be more likely to disclose symptoms earlier in the pandemic versus later in the pandemic and that barriers and facilitators of symptom reporting would change over time. Additionally, given differences in job duties and level of training among staff, we expected that symptom reporting behaviors would differ between CNAs, nurses (LVNs and RNs), and nonfrontline staff (eg, environmental services, dietary, administrative). We expected that frontline workers would be more likely to disclose symptoms given their greater level of interaction with NH residents and that monetary barriers would be less important for higher paid staff (eg, RNs/LVNs) compared to CNAs and nonfrontline staff.

Analysis

Characteristics of the survey participants and responses were summarized using descriptive statistics. The Cronbach α was computed to assess internal consistency and reliability for responses within each construct; α = 0.7 was the benchmark for acceptable reliability.^{Reference Cronbach20}

Multiple-item scales were developed within the 8 constructs using simple algebraic sums of the Likert-scaled items for each construct. For example, 4 questions were related to the construct of “fear of known or unknown consequences of COVID-19.” To create a composite scale, 5-point Likert-scaled responses to the 4 questions were summed together, with a minimum score of 4 and a maximum score of 20. Composite scales for each construct were then transformed to range from 0 to 100 by subtracting scale means from theoretical-scale minimums and dividing the result by the difference between the theoretical-scale maximum and minimum and multiplying that result by 100. Higher scale scores indicated that the barrier or facilitator was more important when deciding whether to report possible symptoms.

To assess construct validity, we calculated single-item summary scores within each of the 8 constructs. To assess discriminant validity, we used t tests and analysis of variance to compare responses among subgroups (1) by time of positive SARS-CoV-2 test, comparing surveys conducted earlier versus later in the pandemic (eg, wave 1 vs wave 2), and (2) by NH staff job role (frontline vs nonfrontline). All data were analyzed using SAS version 9.4 software (SAS Institute, Cary, NC).