Self-reported Illness Experiences and Psychosocial Outcomes for Reservation-Area American Indian Youth During COVID-19

Key Points Question How did American Indian youth who live on or near reservations experience the COVID-19 pandemic? Findings This cross-sectional study among 2559 American Indian students in grades 6 through 12 found that approximately 14% of the sample reported having had a test result positive for SARS-CoV-2 infection, a higher rate than for all cases nationally and for children nationally, and three-quarters of the sample reported someone close contracting COVID-19, while more than one-quarter reported someone close dying from COVID-19. Regarding perceived psychosocial impacts, COVID-19 was associated with strained friend relationships, lower school engagement, and less social connectedness, although more than 60% of students reported feeling no change or a decrease in negative emotions, such as sadness and anxiety. Meaning These findings suggest that although COVID-19 mortality and morbidity rates were high on American Indian reservations, psychosocial impacts were complex and many students were resilient in the face of the pandemic.

although ages are grouped in the Census and do not map exactly onto the ages of 6 th -12 th grade students. Many tribal and BIE schools do not report enrollments unless taking part in our survey; therefore, we were not able to calculate the AI population by grade from our sampling frame. The Census provides more complete, though still imperfect, data. Because of the small number of schools in the Northeast, Southern Plains, and Southeast, all schools and/or school districts meeting the requirements of the sample are invited to participate in the survey. In other regions, a participation rate of 20% is assumed; thus, we randomly draw approximately 5 times the number of schools needed in the final sample. For the Spring 2021 sample, schools within a region had an equal probability of selection, as enrollments for many schools in the sampling frame were unknown, prior to sampling.
For schools not including high school grades (e.g., middle schools), the high school most likely to be attended by the AI middle school students is determined and added to the drawn sample. Further information about each school included in the drawn sample is gathered, including current enrollments by grade and current contact information. Given the small number of schools within the sampling frame, 30 schools are sought each year to participate in the survey. Schools are asked to participate for 2 consecutive years, as per Monitoring the Future procedures, in order to better identify trends by dampening outliers and/or noise (similar to using moving averages). Fifteen schools were scheduled to repeat from the previous year (2019-2020); however, only 9 agreed to repeat the survey due to the impacts of COVID-19.
With 9 schools surveying for their second year, for Spring 2021 we sought to recruit 21 "new" schools. 318 middle schools were in the sampling frame for new schools, with 163 related high schools. The final draw of schools with a grade 7 was 68 schools. Including the related high school led to a total number of schools of 102, regionally distributed as NE: 6.9%; NP: 36.3%; NW: 11.8%; SE: 4.9%; SW: 26.5%; UGL: 9.8%. Of these, 3 were disqualified due to 100% remote instruction prior to recruitment. (There may have been more schools operating remotely that did not have this information posted on their website or other public area.) Thus, 99 schools were asked to participate.
Recruitment of sampled schools includes a letter of invitation, an email, and a telephone call explaining the project and procedures. Once a school agrees to participate, we obtain all necessary approvals, including school boards and tribal IRBs.

Survey Procedures.
For each participating school, the appropriate tribal and school board approvals were obtained. Approximately 3 weeks prior to the scheduled survey, letters were sent to parents of enrolled students in grades 6 and above describing the survey and providing instructions for opting their child out of the survey. This information was also posted on other local media sites, where parents were likely to see it. On the scheduled date, the OYOF survey was administered online to students using Qualtrics software during school hours. School staff read directions prior to survey administration indicating that students could decline to participate or leave blank any questions they did not wish to answer, and these instructions were repeated in the online survey. Eleven (11) schools were operating concurrently (some students remote; some students in-person); 6 schools used a hybrid approach where students were part-time remote and part-time in-person; 3 schools were in-person only. In order to ensure that a staff member was present to read instructions and answer questions, schools were asked to only administer the survey during times when a teacher or other staff member was present, either remotely or in-person. Several schools opted not to survey remote students.
All survey responses were collected anonymously, and all procedures were approved by the university institutional review board. Participating schools received a full report of their survey findings and compensation for resources used to complete the survey process, with mean payment of $1400. where the numerator is estimated from Census 2010 American Indians of one race or more than one race ages 10-19 for all reservations within a region and the denominator is calculated after obtaining enrollments from participating schools. Non-response weights are calculated as: W i|g|j, =(# of AI students enrolled in grade g in school j)/(# of AI students completing survey in grade g, school j) The final weight for a student i in grade g in school j which is located in region h is:

Proportion and confidence interval estimation
For each item discussed above, frequency of outcomes (95% CI) were computed using survey commands of Stata® statistical software, designating stratification by region and school as the primary sampling unit. The survey commands for proportion with default linearized variance estimation were used to calculate proportions and 95% CIs. An example command is: svy linearized: proportion CVp25_b1r, over(sex). Differences by grade group (6-8 and 9-12) and sex were tested using Pearson chi-square tests adjusted for complex samples while point differences in proportions between sexes and their confidence intervals were calculated using lincom commands in Stata.

Participants.
In total, 3847 students in grades 6-12 were surveyed, with this study using a subsample of students selfreporting as AI (N=2559). Participants represented 60.4% of eligible students in these schools, with 70% of schools surveying 60% or more of their eligible students. Fewer than 1% of students across participating schools were opted out of participation in OYOF by their parents, and the primary reason that students did not take the survey was due to absence on the day of survey administration. During COVID-19, AI schools had high absentee rates among students, and students not attending school may have had limited access to technology needed for remote learning. 5 This may have introduced bias in this study's findings. A question on the survey asked students in participating schools whether they were taking the survey remotely or in-person. A comparison of responses presented in this study found no significant differences in responses by where the survey was taken (remotely or in-person), lending some credence to remote students not being different from in-person students. However, these remote students may have been more likely to have technology access, thus affecting their responses to school questions in Table 2 and potentially other responses. Finally, the exact number of students who refused to take the survey on the survey administration date (that is, those who were present but refused to complete the survey) cannot be determined because there is no record in the database for students who choose not to open the survey link. Because the survey is anonymous, we ask teachers to be in an area of the classroom or lab where they cannot observe student responses.
Sample sizes by sex and mean age, per grade, are presented in eTable 2, in addition to race/ethnicity and region. It is worth noting that Spring 2021 was a challenging time for surveying (no surveying occurred Spring 2020 or Fall 2020 due to COVID-19), and thus, there may be bias in our data due to a variety of factors that are specific to the pandemic. At the same time, there is no other broad-scale survey of reservation-area AI adolescents of this type. Therefore, even with its limitations, the data can make a positive contribution to understanding COVID-19's "impacts" on this unique population.   .9 (-13.8, 5.9) 9.2 (-1.7, 19.6) a Sample sizes for all but the last question in this table are significantly less than 2559 partially due to the planned missingness design and partially due to students not responding to the question. For all questions but the last one, the sample sizes by sex averaged 722 for females, 668 for males, and 39 for another sex. b This question was based on a 5-point scale of much less socially connected, less socially connected, about the same, more socially connected, and much more socially connected. Percentages indicate those who answered they felt much less or less socially connected. The larger N is due to this question not being part of the planned missing design so all students could answer it. Also, it was near the start of the survey and thus had a lower percentage of missingness due to survey attrition. c Estimate is significantly different from zero at p<. 05