Change in settings for early-season influenza vaccination among US adults, 2012 to 2013.

Vaccination in non-medical settings is recommended as a strategy to increase access to seasonal influenza vaccine. To evaluate change in early-season influenza vaccination setting, we analyzed data from the National Internet Flu Survey. Bivariate comparison of respondent characteristics by location of vaccination was assessed using chi-square tests. Multinomial logistic regression was performed to compare the predicted probability of being vaccinated in medical, retail, and mobile settings in 2012 vs 2013. In both 2012 and 2013, vaccination in medical settings was more likely among elderly adults, those with chronic conditions, and adults with a high school education or less. Adults 18-64 without a chronic condition had a lower probability of vaccination in the medical setting, and higher probability of vaccination in a retail or mobile setting, in 2013 compared to 2012. Adults 18-64 with a chronic condition had no change in their location of flu vaccination. Elderly adults had a lower probability of vaccination in the medical setting, and higher probability of vaccination in a retail setting, in 2013 compared to 2012. Non-medical settings continue to play an increasing role in influenza vaccination of adults, particularly for adults without a chronic condition and elderly adults. Retail and mobile settings should continue to be viewed as important mechanisms to ensure broad access to influenza vaccination.

Introduction

Vaccination in non-medical settings is recommended as a strategy to increase access to seasonal influenza vaccine. (Fiore et al., 2010) National data from the 2011–2012 influenza season showed that 57% of adults ≥ 18 years received influenza vaccine in medical settings, such as a doctor's office, while the remaining 43% were vaccinated in non-medical settings. (Lu et al., 2014) Prior research has shown a downward trend in influenza vaccination in medical settings, with a corresponding upward trend in vaccination in non-medical settings. (Lu et al., 2014) In addition, the setting for influenza vaccination has been shown to vary by characteristics such as age, race/ethnicity, education level, and chronic condition status. (Lu et al., 2014; Centers for Disease Control and Prevention (CDC), 2011).

To build upon these findings using more recent national data, our objective was to explore changes in place of early season influenza vaccination among adults from 2012 to 2013, by age and chronic condition status. We hypothesized that for elderly persons and non-elderly adults with chronic diseases, medical settings would predominate, with minimal change across the two years studied. Conversely, we hypothesized that non-elderly adults without chronic conditions would have less vaccination in medical settings, and would demonstrate a shift toward increasing use of non-medical settings across the two years studied. Studying changes in place of vaccination can support public health officials' efforts to increase flu vaccine coverage rates among adults by offering guidance on whether educational campaigns, public health funding opportunities, and immunization policy initiatives are appropriately targeted.

Methods

Study design

In November 2012 and November 2013, we conducted the National Internet Flu Survey, a nationally representative, cross-sectional survey of US adults using the internet-enabled KnowledgePanel® (GfK Custom Research, LLC). The purpose of the National Internet Flu Survey was to provide information on early-season influenza vaccination results (Santibanez et al., 2012, Srivastav et al., 2013) for use by vaccination campaigns during National Influenza Vaccination Week. The study was approved by the University of Michigan Medical School Institutional Review Board.

Survey sample and administration

KnowledgePanel® members have been selected by GfK using address-based probability sampling and cover approximately 97% of US households. (GfK, 2013) Households without internet access at the time of panel enrollment are given a web-enabled computer and free internet service. For these surveys, a random sample of KnowledgePanel® members was invited to participate. Panel members in racial/ethnic minority groups were oversampled to ensure adequate representation in the results. Surveys were fielded by GfK during the first 2 weeks of November, in both English and Spanish. The completion rate was 63% (5057 of 8039) in 2012 and 59% (5333 of 9039) in 2013.

Survey questions

In the 29-question survey, respondents reported their influenza vaccination status for that year's influenza season, as of the time of survey completion; whether they had visited a health professional since July of the survey year; and whether they had any of 10 health conditions that would place them at increased risk for influenza complications. Those who had received influenza vaccine indicated the setting for vaccination, selecting from 13 options, including other.

Data analyses

GfK provided a data file with survey responses, respondent demographic information (age, gender, race/ethnicity, and education attainment) from KnowledgePanel® member profile data, and U.S. Census-based post-stratification weights to match the U.S. population distribution on respondent sex, age, race/ethnicity, education, and U.S. Census region. The weights include adjustments for survey non-response and are applied to the survey data to produce nationally representative response estimates. (GfK, 2013, Dennis, 2010).

For each survey year, among respondents who had received influenza vaccine by the time of survey completion, we explored differences in the setting for vaccination by respondent age, gender, race/ethnicity, education level, chronic condition status, and attendance at a health care visit since July of that year. For analysis purposes, the 13 response options for place of vaccination were classified into 3 groups: medical (doctor's office, clinic or health center, hospital, health department), retail (pharmacy or drugstore, supermarket or grocery store), and mobile (workplace, school, senior center, nursing home, military, home). These 3 groups represent the outcome of interest (location of vaccination). Bivariate comparison of respondent characteristics by location of vaccination was assessed for each year using chi-square tests. A multinomial logistic regression model was performed to compare the predicted probability of being vaccinated in each location, in 2012 vs 2013, for the three groups traditionally used in adult flu vaccination recommendations and coverage assessments: adults 18–64 without a chronic condition, adults 18–64 with a chronic condition, and adults ≥ 65 years. The model included year, age group/chronic condition and the interaction of year by age group/chronic condition variable.

Analyses were conducted using Stata 12 (Stata Corp, College Station, TX). Weighted proportions and predicted probabilities are reported.

Results

Among respondents to the National Internet Flu Survey, 35% in 2012 and 39% in 2013 had received influenza vaccine at the time of survey completion. Table 1 presents the setting for vaccination, overall and by respondent characteristics. Overall, medical settings were the most commonly reported place of vaccination, but the proportion declined from 2012 to 2013, with a concurrent increase for both retail and mobile settings.

Table 1

Place of early season influenza vaccination by respondent demographics.

Characteristic	2012 place of vaccination, %(n = 1982)				2013 place of vaccination, %(n = 2250)
Medical	Retail	Mobile	P-value	Medical	Retail	Mobile	P-value
Overall	57.2	18.2	24.6	50.0	21.9	28.1
Age
18–49 years	51.7	15.0	33.3	≤ 0.0001	44.2	16.9	38.9	≤ 0.0001
50–64 years	57.0	19.3	23.7		49.9	20.9	29.2
≥ 65 years	67.1	22.7	10.2		59.2	30.4	10.4
Gender
Female	58.3	20.4	21.3	0.01	52.8	22.4	24.8	0.03
Male	55.7	15.5	28.8		46.9	21.3	31.8
Race/ethnicity
White, non-Hispanic	53.7	19.6	26.7	0.50	48.0	22.7	29.3	0.08
Black, non-Hispanic	58.4	16.1	25.5		60.5	15.0	24.5
Other, non-Hispanic	63.5	15.9	20.6		44.9	25.6	29.5
Hispanic	59.3	18.9	21.8		55.9	20.5	23.6
Education level
Less than high school	78.8	11.3	9.9	≤ 0.0001	59.7	22.8	17.5	≤ 0.0001
High school	61.7	18.4	19.9		57.0	23.9	19.1
Some college	52.1	19.0	28.9		50.5	19.4	30.1
College degree or higher	49.5	19.6	30.9		40.3	21.6	38.1
Chronic condition
Yes	61.3	19.5	19.2	0.0005	58.6	20.1	21.3	≤ 0.0001
No	53.1	16.9	30.0		42.3	23.5	34.2
Health care visit since July 1
Yes	62.6	17.3	20.1	≤ 0.0001	58.3	19.8	21.9	≤ 0.0001
No	41.5	21.1	37.4		30.1	26.6	43.3

In both 2012 and 2013, setting for vaccination varied significantly by age, gender, education level, chronic condition status, and attendance at a health care visit since July (Table 1). In both years, a greater proportion of elderly (≥ 65 years) were vaccinated in medical settings, compared to non-elderly adults, and adults 18–49 years were more likely than older adults to be vaccinated in mobile settings. Adults with a high school education or less, with a chronic condition, and with a health care visit after July 1 of the calendar year all had higher proportions of vaccination in a medical setting in both years, compared to those with more education, no chronic condition, and no health care visit, respectively.

Table 2 presents the comparison of the predicted probability of flu vaccination generated from the multinomial logistic regression model in different settings (medical, retail, and mobile) for 2012 vs 2013 survey respondents. Adults 18–64 without a chronic condition had a lower probability of vaccination in a medical setting in 2013 compared to 2012; though not significant, adults 18–64 without a chronic condition had a higher probability of vaccination in retail and mobile settings in 2013 compared to 2012. Adults 18–64 with a chronic condition had no change in their location of flu vaccination. Elderly adults had a lower probability of vaccination in a medical setting, and higher probability of vaccination in a retail setting, in 2013 compared to 2012.

Table 2

Predicted probability of receiving influenza vaccination in medical, retail and mobile settings, by age and chronic disease status, 2012 vs 2013.

Predicted probability of receiving influenza vaccination in a medical setting
Group	2012	95% CI		2013	95% CI		P-value
18–64 without chronic disease	48.9%	43.5%	54.3%	37.8%	33.3%	42.4%	0.0021
18–64 with chronic disease	58.8%	53.3%	64.4%	58.0%	52.8%	63.1%	0.8285
65 and above	67.1%	62.6%	71.6%	59.2%	54.3%	64.2%	0.0209

CI = confidence interval; P-value compares predicted probability in 2012 vs 2013.

Discussion

The initial recommendation for universal flu vaccination of all persons ≥ 6 months of age recognized that utilization of non-medical settings would be important for expanding access to influenza vaccine, (Fiore et al., 2010) especially to persons who do not regularly access the health care system. (Fiore et al., 2010, Uscher-Pines et al., 2010; National Vaccine Advisory Committee (NVAC), 2012) Findings from this study indicate that non-medical settings continue to play an increasing role in influenza vaccination of adults; adults without a chronic condition and elderly adults demonstrated a decreased proportion of vaccination in medical settings from 2012 to 2013. In contrast, adults with a chronic condition did not experience a change in their use of medical settings for flu vaccination. Between 2012 and 2013, early-season flu vaccination rates showed a larger increase (35% to 39%) than comparable end-of-season flu vaccination rates (41.5% to 42.2%), (Centers for Disease Control and Prevention, 2013b, Centers for Disease Control and Prevention, 2014a) which suggests that non-medical settings may facilitate earlier vaccination. To provide a more nuanced look at non-medical settings, this study divided non-medical settings into two different categories: retail and mobile. Retail settings, such as pharmacies, were integral to 2009–2010 H1N1 vaccination efforts, (Association of State and Territorial Health Officials, 2009) and have continued to expand their immunization capacity. (American Pharmacists Association, 2013) Retail settings are advantageous because they are very accessible to the general public – almost 95% of the US population lives within 5 miles of a community pharmacy (National Vaccine Program Office, 2016) – and most have existing expertise and infrastructure for administering immunizations and promoting vaccines through mass media campaigns. (Rothholz, 2013) The use of retail settings for flu vaccination increased from 2012 to 2013 among both the elderly population and younger adults without a chronic condition.

For mobile settings, only younger adults without chronic conditions demonstrated an increase from 2012 to 2013, which was predominantly driven by workplace vaccination. Factors contributing to this increase could include the employee wellness program incentive policies created by the Affordable Care Act (US Department of Labor, 2014) and implementation of strategies to increase coverage rates among health care personnel. (National Vaccine Advisory Committee, 2012) The promotion of flu vaccination among large employers across many industry sectors is promising. (Graves et al., 2014).

To fully realize the benefits of vaccination in non-medical settings, as well as continue to support vaccination in medical settings, public health officials must continue to mitigate barriers to vaccination. For example, while the Affordable Care Act mandates coverage of vaccines, including flu vaccine, some health plans limit their coverage to certain locations (e.g., a doctor's office). Trends in the location of flu vaccination call into question whether such policies may deter flu vaccination for certain plan members. Other barriers may be related to the inexperience of non-medical immunization providers in the logistics of delivering vaccines. Trends in flu vaccination in non-medical settings suggest that public health training and educational efforts, as well as funding for demonstrated projects, should include a focus on retail clinics and mobile settings. (National Vaccine Program Office, 2016, Rothholz, 2013) Those efforts should include strategies to ensure that non-medical immunization providers participate in state immunization information systems, which would allow medical providers to verify which patients were vaccinated in other settings, and allow public health officials to accurately track flu vaccination receipt. (National Vaccine Program Office, 2016).

Limitations

Data on respondents' flu vaccination status and setting for vaccination are based on self-report and are not validated with clinical records. Prior studies comparing self-reported flu vaccination to medical records have found self-report to have high sensitivity and moderate to high specificity. (Zimmerman et al., 2003, Mangtani et al., 2007, Rolnick et al., 2013) Discordance between self-report and medical records may be due to recall, but also may indicate incomplete documentation. (Mangtani et al., 2007, Greene et al., 2009) One study found that over half of self-reported vaccinations not captured in electronic medical records were administered in non-medical settings. (Greene et al., 2009).

Although response bias is possible, the survey's probability-based sampling and post-stratification weighting help to minimize the risk; moreover, the response rates for this study were favorable compared to the surveys used to generate the CDC's end-of-season flu vaccination rates (Centers for Disease Control and Prevention, 2013b, Centers for Disease Control and Prevention, 2014a).

Data were collected in November of each year and thus reflect early-season settings for vaccination; data for vaccinations administered later in each flu season were not available. Finally, our data are not directly comparable to setting for flu vaccination data from prior years due to different categorizations and use of early-season vs. end-of-season data. However, overall trends were similar, with adults more likely to be vaccinated in medical settings if they were older, non-White, had a chronic condition, and lower educational attainment. (Lu et al., 2014; Centers for Disease Control and Prevention (CDC), 2011).

Though differences across flu seasons in vaccine supply and influenza activity could impact our findings, these two flu seasons were similar in early season vaccine availability (both in terms of number of doses distributed and the proportion of total doses distributed by the survey date) (Centers for Disease Control and Prevention, 2013c, Centers for Disease Control and Prevention., 2014b) and influenza activity. (Centers for Disease Control and Prevention, 2013a, Epperson et al., 2014).

Conclusion

This analysis of consecutive years of national early-season flu vaccination surveys demonstrates increased use of non-medical settings by adults. Retail and mobile settings offer convenient and accessible locations for vaccination across age groups, particularly for those who do not have visits with a medical provider during the flu season. Future education, policy, and funding to promote flu vaccination among adults should include non-medical settings.

Conflict of interest

The authors declare that there are no conflicts of interest.

Transparency document

Transparency document