AJPH Video Abstract: County-Level Factors in Influenza & COVID-19 Vaccination in Indiana, 2020-2022

Hello, I'm Monica Kasting from Purdue  University, and I'm going to talk about   our article that was recently published in the  American Journal of Public Health titled County   Level Factors Associated with Influenza  and COVID -19 Vaccination in Indiana,  2020 through 2022. As a brief background,  the COVID -19 pandemic highlighted rural   urban disparities, including access to  health care, differences in health beliefs,  higher vaccination hesitancy and lower  vaccination confidence. While initial pandemic   hotspots centered in densely populated  urban areas, infections in rural areas   began to rise as the pandemic progressed. Nationally, there are significant rural urban   vaccination coverage disparities across multiple  vaccines. Addressing these disparities must begin   with accounting for the triad of local culture, geographic location, and economics of the region.   Therefore, this study aimed to assess existing  differences among adults in both COVID -19 and   influenza vaccination by county in Indiana  and determine what county -level factors are   associated with those vaccination rates. We chose to focus on these two vaccines   because they both involve respiratory viruses,  both have generated greater vaccine hesitancy   than other adult vaccines, and are both  routinely recommended for all adults.  We use several sources of data for  our analyses. We examine COVID -19   and influenza vaccination rates across Indiana's  92 counties among adults aged 18 years and older   from December 2020 through March 2022. For COVID -19 data, we use the Indiana   State Department of Health's Immunization  Registry, which is called the Children and   Hoosier Immunization Registry Program,  or CHIR. CHRP had mandatory reporting   for all COVID -19 vaccinations. These data are publicly available   by request. CHRP also includes information on  county -level vaccination rates by age group,   race -ethnicity, and sex. All data were aggregate  and no individual -level data were available.  However, there's not mandatory reporting in CHRP  for adult influenza vaccinations. Therefore,   for flu data, we use the CDC's U .S. Influencers'  Surveillance Dashboard or FluVax View.  FluVax View reports weekly national influenza  vaccination data by county. From the database,   we extracted influenza vaccination rates using  the federal information processing codes,  which are unique codes assigned to each county  in the U .S. and we filtered down to only Indiana   counties. For our other measures, we derive  population -level estimates from the U .S. Census.  We also included county -level socio -demographic  and health data from the County Health Rankings   Report, which is supported by the Robert  Wood Johnson Foundation and the University   of Wisconsin Population Health Institute. From county health rankings, we extracted data   on socio -demographic distribution of each  county, including median household income,   percent of the county that's rural and percent  of the county that is Hispanic, among others.  We extracted data on health indicators like  the number of primary care providers per capita   and the percentage of the population who smoke.  vaccination rates using linear regression weighted   by the total adult population in each county, and then conducted model selection using forward,   backward, and stepwise selection to determine  the best multiple linear regression model   for each vaccination rate. We selected  the best model based on the adjusted R   -squared value and examined tolerance values for  variables in each final model to ensure issues   with multi -colonarity were negligible, and we assessed residuals for normality.   The selected reduced model for each vaccination  rate is reported, and the squared semi -partial   correlation coefficient for each variable  is reported as a measure of effect size.  You can see the mean percentage of counties that  were rural was 54 .5%. When looking at socio   -demographic characteristics, the mean percentage  of residents who were African -American were 3   % with a wide range from 0 .2 to almost 30%. There were high rates of smoking with a mean of 22   % and obesity with a mean of 36%. The mean COVID  -19 vaccination rate across the 92 counties in   Indiana was 58 % and ranged from 31 .2 to 87 .6%. The mean influenza vaccination rate was 42 .9 %   and ranged from 33 .7 to 53 .1%. This figure  displays the variability in COVID -19 influenza   vaccination rates across Indiana counties, with darker color indicating the higher   vaccination rate. COVID -19 vaccination  rates are shown in panel A on the left,   and flu vaccination rates are shown in  panel B on the right. As you can see,  there is more variability with COVID -19  vaccination than was flu vaccination,   and overall flu rates were lower than COVID -19  vaccination rates. The final selected model for   COVID -19 contained seven variables and  had an adjusted R -squared of 0 .867.  Based on this model, an increase in primary care  providers per capita, median household income,   percentage of Medicare enrollees  who had a mammography screening,   and percentage of African -American residents  were all associated with increases in the   percentage with COVID -19 vaccination. In addition, an increase in uninsured   residents, percentage of female residents,  and percentage of people who smoke were all   associated with decreases in the percentage  with COVID -19 vaccination. The final model   for flu contained five variables and  had an adjusted R -square to 0 .702.  Based on this model, an increase in the  percentage of residents who were uninsured   and those who completed high school were  associated with increases in the percentage   with flu vaccination, and an increase in the  percentage of African -American residents,  Hispanic residents, and percentages of adults  who smoke were all associated with decreases   in the percentage with flu vaccination.  While both COVID -19 and flu vaccinations   are routinely recommended for all adults, county level factors associated with each   very greatly between the two vaccines.  The reduced model for COVID explained   a slightly higher percentage of the  variance, 86 .7 % than the model for flu,  70 .2%. This may reflect greater polarization  surrounding COVID -19 vaccination, leading to   a greater influence in socio -demographic  factors with COVID -19 compared to flu.  Of note were variables reflecting access to care,  like the number of primary care providers per   capita, which were significant for COVID, but not  for flu. One possible explanation is the unique   way in which the COVID -19 vaccine was rolled out, which involved different dates which people were   eligible based on age or occupation,  for instance. The need to sign up for a   vaccine online and difficulties particularly  early on in finding available appointments,  all of which may have increased logistical  barriers and resulted in the differences we   found in our data. Another finding of interest  was the association between each vaccine and   county -level racial ethnic distribution. Their percentage of Hispanic residents in   the county was associated with influence  of vaccination but not COVID -19, and the   percentage of African -American residents in a  county was negatively associated with flu vaccine,  but positively associated with COVID -19 vaccine.  This could be due to targeted outreach programs,   at medical discrimination and  distrust in the medical community,   or a broader reflection of access to health care. It's possible differing findings by the Hispanic   population may be better explained if  we controlled for individual patient   characteristics. That was not possible with our  aggregate data. Future research should explore   these racial ethnic disparities while taking  individual patient characteristics into account.  Importantly, smoking was strongly  negatively associated with vaccination   rates for both vaccines. This may seem  counterintuitive because smoking is a risk   factor for severe respiratory diseases. However, research shows preventive health behaviors cluster   together and people who engage in one healthy  or unhealthy behavior are more likely to engage   in another. Likewise, people who smoke are less  likely to receive routine preventive services.  In addition, while the percentage of the county  that was rural was significantly associated with   both COVID -19 and flu vaccinations in  univariable regression analyses. These   associations were no longer significant when  other variables were included in the model.  It's possible that the rural urban differences  were accounted for with other county -level   factors, including income or smoking status. But  the reasons for the rural urban disparities across   a multitude of health indicators are multifaceted  and are likely a complex elimination of access,  infrastructure, attitudes, and beliefs.  Our findings should be interpreted in the   light of some limitations. First, the  data are cross -sectional and causal   relationships cannot be established. Second, the data are aggregate and   are subject to ecological fallacy.  In addition, our data focused only   on Indiana and findings may not be applicable  to other states or jurisdictions. And lastly,  because CHRP does not track adult influence of  vaccination, we use two different data sources   to examine the two vaccination rates.  It's possible that the two data sources   differ in their accuracy and reporting. Therefore, the results of this study   should be interpreted with caution and further  studies are needed to understand the complex   associations between systems -level factors we  examined and vaccination rates. In conclusion,  while both COVID -19 and flu vaccines  protect against respiratory viruses   and are recommended for all adults. The  factors associated with uptake of each are   varied. Variables reflecting access to care, like the number of primary care providers   per capita and median household income, were  significant for COVID -19, but not flu vaccination   rates. The percentage of uninsured residents in  the county was significant for both vaccines,  but in opposite directions, so that the rate was  negatively associated with COVID and positively   associated with flu. The polarization surrounding  COVID -19 vaccination may have led to a greater   influence of socio -demographic factors with COVID  -19 vaccination as compared with flu vaccination.  Further research, including individual  -level data, is needed to better understand   these associations and develop effective  interventions to address county -level   factors and improve vaccine uptake. In closing, I'd like to thank the rest of our research team   and acknowledge our research support and  funding for this work. Thank you for your   time. For more information, please view  our full article by scanning the QR code.