OVERVIEW OF METHODS
The study population consisted of Asian American (AA) and Native Hawaiians and Pacific Islander (NHPI) adults, 18 years and older, residing in the United States. Eligible respondents included people who reside in the U.S. in permanent or temporary quarters (e.g., dormitories, apartments, hotels), but who still consider their permanent residence within the U.S. A primary objective of this study was to assess emerging needs because of COVID-19 and to have diverse representation of different AA and PI ethnic groups. To meet this end, the sample was stratified by five AA ethnic groups (Chinese, Filipino, South Asian, Vietnamese, and Korean) and five PI ethnic groups (Native Hawaiian, Samoan, CHamoru, and Marshallese). AA groups included the five largest in most census estimates. The South Asian ethnic group included people with roots in India, Pakistan, Bangladesh, Sri Lanka, Nepal, among others. The NHPI ethnic groups included the largest in the three major categories: Polynesian (Native Hawaiian, Samoan, Tongan); Micronesian (CHamoru, Marshallese); and Melanesian (Fijian). A systematic screening process verified eligibility (i.e., US residency; AA/PI subgroup membership) for the study and ask eligible respondents to participate in the study. The study relied on self-report to measure the ethnicity of the respondent.
Children under the age of 18 years old at the time of the survey were excluded from this survey. The exclusion of children is a function of cost and time considerations since it would take considerably more effort to secure parental consent to recruit children into the sample. To obtain some information about how children are faring in the pandemic environment, the survey includes a few questions asking parents about this issue. It may also be possible to obtain information on children from other types of data collection activities. We also exclude people who do not reside in the U.S. but are here as tourists or as a student who intends to return to their home country. Another major exclusion is an indirect function of the use of the online survey method to obtain responses. About 84% of the U.S. population is considered computer-literate which is above the international average, 77% (Mamedova & Pawlowski, 2018). Despite this ranking, 16% are not computer-literate with lower education, English-speaking abilities, urbanicity, income, as some of the demographic factors associated with this group (Perrin & Maeve, 2015). Accordingly, the mode of study may inadvertently exclude older, non-English speakers, low-income and rural residents.
This study consists of two complementary components with the first composed of the reliance on a secondary data, the PULSE survey conducted by the U.S. Census and other organizations, and original data collection of eligible respondents obtained through community programs and social media platforms. The first component, the PULSE survey,
The design used a dual frame to recruit eligible respondents for the survey. The intent of the survey was to secure a broad swath of NHPIs who live in different states and who represented different NHPI ethnic groups. The first frame recruited respondents from a Qualtrix panel that provided an overall national dataset about how different AAPI ethnic groups are doing on certain dimensions during the pandemic. Since the Qualtrix panel would likely have be biased toward highly educated and middle to high income respondents, we supplemented this sample with a frame derived from recruiting residents from community organizations and social media platforms. Convenience samples, like the used here, are relatively efficient and less costly means to recruit samples especially from relatively rare population. The non-probability characteristic of the convenience sample is its most serious disadvantage. It is, by nature, difficult to generalize to a specific population since we do not have sufficient information about the types of people among AA and PI populations who are do not use social media or participate in the selected organizations which are critical elements of our recruitment strategy. Accordingly, it will not be possible to make precise prevalence estimates of the outcome variables. Despite this disadvantage, the convenience sample can be enhanced to increase its value. The dual sampling strategy provided the opportunity to recruit people who different profiles which will increase the coverage and inclusion of a heterogenous final sample. In subsequent data analyses, we plan to consider recent statistical innovations that can complement convenience sample which may make it amenable to use powerful inferential statistical tools in our analyses (Hedt & Pagano, 2014). For this current report, we used the pool unweighted samples from the dual frames.
A systematic screening process verified eligibility (i.e., US residency; NHPI subgroup membership) for the study. The study relied on self-reports to measure the ethnicity of the respondent. Respondents were provided a $10 honorarium for their participation in the survey. Eligible respondents completed a web-based survey. If eligible participants did not have access to a computer or preferred a different mode of responding to the survey, they were given a self-administered questionnaire or completed the survey through a phone or virtual interview. When this occurred, the questionnaire responses were coded and researchers entered the responses into the survey database.
One of the goals of this survey was to make it accessible to a broad range of respondents. Since language is a key facet in the AAPI communities, the survey was translated from English into the following languages corresponding to the 10 AA and PI ethnic groups: Chinese (traditional and simplified); Bangla, Hindi, Urdu; Vietnamese; Korean; Samoan; Tongan; Chamorro; and Marshallese. We also worked extensively with different national and community AAI organizations to gather input on the survey design and content, to insure that the data can be useful and useable for policy and programmatic purposes, and to facilitate the recruitment of eligible respondents into the survey.
Three principles guided the selection of the measures. First, the core measures agreed upon by the Alliance members were given priority for inclusion in the AA and PI survey. Second, some survey measures are taken from the Census Bureau’s PULSE project conducted from April to June 2020 about the public’s response to the COVID-19 pandemic. The inclusion of the PULSE measures provided a means to compare the findings from this survey with a national probability sample. Finally, the members of the AA and PI group agreed on the remaining measures for the survey that were derived from other local and national COVID surveys. Measures were generally taken from established scales. At the outset, the AA and PI group agreed that the survey was to take no longer than 30 minutes to complete to minimize respondent burden.
DATA CLEANING AND VERIFICATION
Extensive time was spent verifying the survey responses and cleaning the resulting survey data. Since bots are a major problem with online survey, we checked for inconsistent responses between variables such as age and birthdays, the inordinate use of the same internet protocol addresses. Since we provided a $10 stipend for participation in the survey, we also checked for inconsistency in the address and the geographic location recorded in the survey.
This needs assessment study has several limitations. First, findings are based on cross-sectional surveys and it is not possible to make causal attributions. Second, we do not have data prior to the start of the pandemic. While we do ask respondents about their perception of changes before and during COVID-19, the absence of survey data on our samples do not allow us to make precise comparison between these two time periods. Third, despite our intent to provide data on specific NHPI ethnic groups, it was not possible to do so in all cases. For example, the South Asian sample includes a number of ethnic groups and for some ethnic groups, like the Bangledeshis, the resulting sample is too small to make a statistically accurate conclusion.
Despite these limitations, the surveys produced rich datasets on AA and NHPI residents and how they are faring during the current pandemic. This report can only touch on some of these important findings and we plan to produce more data briefs for community audiences and policy makers as well as papers for scholarly and academic outlets. It also should be noted that even when some sample sizes are too small to make definitive statistical conclusions about some ethnic groups, having representation from these groups does allow us to identify patterns that may be useful to examine in future studies.
The Asian American Pacific Community Health Organizations (AAPCHO) reviewed and approved the human subjects protocol for the surveys. The AAPI COVID-19 Needs Assessment investigators completed their individual human subjects training and are certified by their local institutions.
Hedt, B. & Pagano, M. (2014). Health indicators: Eliminating bias from convenience sampling estimators. Statistical Methods, 30(5_:560-568.
Mamedova, S. & Pawlowski, E. (2018). A description of U.S. adults who are not digitally literate. Stats in Brief. Washington, D.C.: National Center for Educational Statistics.
Matsueda, R. & Bielby, W.T. (19860. Statistical power in covariance structure models. Sociological Methodology, 16:120-158.
Perrin, A. & Maeve, D. (2015). Americans internet access: 2000-2015. Washington, D.C.: Pew Foundation.