Effects of Mass Media Coverage on Timing and Annual Receipt of Influenza Vaccination among Medicare Elderly

Abstract

Objective

To measure the association between mass media coverage on flu-related topics and influenza vaccination, regarding timing and annual vaccination rates, among the nationally representative community-dwelling elderly.

Data Source

Years 1999, 2000, and 2001 Medicare Current Beneficiary Survey.

Study Design

Cross-sectional survival analyses during each of three influenza vaccination seasons between September 1999 and December 2001. The outcome variable was daily vaccine receipt. We measured daily media coverage by counting the number of television program transcripts and newspaper/wire service articles, including keywords of influenza/flu and vaccine/shot shortage/delay. All models' covariates included three types of media, vaccine supply, and regional/individual factors.

Principal Findings

Influenza-related reports in all three media sources had a positive association with earlier vaccination timing and annual vaccination rate. Four television networks' reports had most consistent positive effects in all models, for example, shifting the mean vaccination timing earlier by 1.8–4.1 days (p<.001) or increasing the annual vaccination rate by 2.3–7.9 percentage points (p<.001). These effects tended to be greater when reported in a headline rather than in text only and if including additional keywords, for example, vaccine shortage/delay.

Conclusions

Timing and annual receipt of influenza vaccination appear to be influenced by media coverage, particularly by headlines and specific reports on shortage/delay.

Influenza is associated with 36,000 deaths and 226,000 hospitalizations per year in the United States (Thompson et al. 2003, 2004). The elderly account for approximately 90 percent (33,000) of these deaths and 63 percent (142,000) of these hospitalizations (Thompson et al. 2003, 2004). Annual vaccination is the most effective strategy to reduce the morbidity from influenza (Centers for Disease Control and Prevention [CDC] 2007a). However, influenza vaccination levels among the noninstitutionalized elderly increased by only 2.4 percent from 1998 to 2002 (Lu et al. 2005), after reaching a “plateau” of about 70 percent during the late 1990s (CDC 2001a). New approaches may be needed to achieve the Healthy People 2010 goal of 90 percent vaccination coverage among the elderly population (CDC 2001a).

Mass media is reported to substantially affect the knowledge of health and the use of health services (Grilli, Ramsay, and Minozzi 2002; Brodie et al. 2003;). For example, a celebrity campaign on colon cancer screening on NBC's Today's Show appeared to increase the colonoscopy rate by 38 percent during 9 months after the show (Cram et al. 2003).

Influenza was ranked fourth among all health news stories followed by the American public during 1996–2002 (Brodie et al. 2003). However, only a limited number of studies (and no nationally representative studies) have attempted to quantify the effect of mass media on influenza vaccination. A long-term radio and television educational campaign was positively but weakly (p<.10) associated with an increase in the vaccination rate among West Virginia Medicare beneficiaries with diabetes (Schade and McCombs 2005). That study lacked a quantitative measure of media exposure and examined annual vaccination rate only, rather than assessing vaccination receipt closer in time to media exposure.

Studies in Massachusetts (Gnanasekaran et al. 2006) and Colorado (Daley et al. 2006) have noted an indirect association between the annual vaccination rate in the pediatric population and media exposure. Neither study measured media exposure quantitatively. Another pediatric study quantified the media effect by counting the total number of media placements by week with either “influenza” or “flu” in the title (Ma et al. 2006). That study found a statistically significant relationship between weekly media coverage and weekly vaccination, but it did not control for other potential confounders (Ma et al. 2006).

One factor that potentially contributes to delayed vaccination is vaccine supply shortage. Such shortages have occurred in four influenza seasons since 2000 (CDC 2000b, 2001b, 2007a). Lack of vaccine availability was the reason for nonvaccination according to 12.7 percent of unvaccinated Medicare beneficiaries aged 65 or older in 2000–2001, and 7.5 percent in 2001–2002 (CDC 2004).

Vaccine supply not only affects the receipt of influenza vaccination but also the timing of vaccination. Vaccination earlier in the season is beneficial because of the 2-week lag before vaccination is fully effective (Kunzel et al. 1996; Chodick et al. 2006;) and because influenza disease outbreaks can occur as early as October, as observed in 2003 (CDC 2008b). In fact, supply shortage and delay led to the guideline changes in recommended timing for different populations, for example, vaccinating higher-risk populations earlier in the season than lower-risk populations (CDC 2000b, 2001b, 2007b). Also, some patients lose incentive to seek vaccination after a certain time in a season. A large number of vaccines delivered to clinics after November were unused in the 2000–2001 season (Iwane et al. 2007), even though an influenza epidemic could start in January and end in April as observed in 2002 (CDC 2008b).

The purpose of our study was to measure the association between mass media coverage on flu-related topics and influenza vaccination regarding vaccination timing and annual vaccination rates among a nationally representative Medicare elderly population. Focusing on short time frames enabled us to distinguish, at the individual level, between media exposure before vaccination and unrelated media exposure after vaccination. It also enabled us to control for vaccine shortages at short intervals.

METHODS

Study Design

Using the Medicare Current Beneficiary Survey (MCBS) and associated claims data (Centers for Medicare & Medicaid Services 2006), we conducted cross-sectional analyses to assess the relationship between media coverage and receipt of influenza vaccination in the subsequent weeks and months in the U.S. elderly population. We focused on three vaccination seasons between September 1999 and December 2001. Because a serious nationwide vaccine supply shortage and delay occurred in 2000, 1999 was chosen as a control year and the starting year of our study period. Also, there was a similar vaccine supply problem in 2001 (CDC 2000b, 2001b). The use of consecutive years of this dataset helps maintain the homogeneity of the subjects throughout the study period, and hence improves the validity of our data analyses. Because of the rotating panel design of MCBS, one-third of the total subjects included in our analyses were continuously enrolled in the study for three consecutive years (and an additional one-third were enrolled for two consecutive years).

Study Population

Our study population was the community-dwelling Medicare elderly population aged 65 or older, continuously enrolled in Medicare Part B from September 1 to December 31, including those who were alive on September 1 but died between September 2 and December 31 (N=7,208, 7,071, 7,136 for 1999–2001, respectively). Medicare managed care enrollees were excluded because exact dates of vaccination were not available. We excluded respondents whose living situation for the year included a skilled nursing facility, because they were likely to be vaccinated within the facility and they have limited control over vaccination receipt or timing. The weighted study population sizes were 22 million, 22 million, and 23 million for 1999–2001, respectively. All analyses used Stata version 10 (StataCorp. 2007).

Statistical Analyses

Our primary outcome of interest was the number of days from the beginning of the influenza vaccination season (September 1) until vaccination each year. To measure this outcome, we used Medicare physician supplier and outpatient claims (ICD-9-CM code V048; CPT/HCPC codes G0008, 90658, 90659, and 90724), censoring the data on December 31. We used survival models to analyze this outcome. We measured mass media coverage by counting the number of newspaper/wire service articles and television program transcripts that included keywords of interest for each day in the LexisNexis data archive (LexisNexis Academic 2008). We used two sets of keywords in our archive searches: (i) influenza OR flu and (ii) (influenza OR flu) AND (delay OR shortage OR late) AND (vaccine OR shot). We recorded media references based on where the keyword was located headline or text only.

Each survival model included three types of media variables defined as follows: (1) a wire service news agency—Associated Press (AP), the number of articles including flu-related keywords, (2) a nationwide newspaper—USA Today, a product of “the number of related articles” and “the circulation number in a resident state [million] (normalized by the number of total population in the state),” distinguishing articles published in weekday and weekend versions (Audit Bureau of Circulations 2009), and (3) four television networks (4TV)—sum of ABC, CBS, FOX, and NBC, each is a product of “the number of related programs” and “the household number tuned to a program in a resident state [million] (normalized by the number of total population in the state)” (The Nielsen Company [U.S.] Inc. 2009).

We included the following time-varying covariates in the model: the number of media reports during the past 1 week with and without 1-week lag, the number of influenza-related press releases by the CDC (2008a) during the past 1 week, the number of vaccine doses distributed in the past 4 weeks at nine census region levels, assuming equal distribution in proportion to the regional total population (CDC, unpublished data 2006), the percentage of specimens testing positive for each week for influenza at nine census region levels (CDC 2008b; CDC National Center for Immunization and Respiratory Diseases [NCIRD], unpublished data 2008) and for respiratory syncytial virus at four regional levels (CDC 2000a, 2002), and the biweekly average temperature at the 50 state levels (National Oceanic and Atmospheric Administration's National Weather Service 2008). CDC press releases were counted exactly the same as AP, distinguishing two sets of keywords and two types of keyword locations, headline or text only. The inclusion of CDC press releases is expected to control for their potential effects on both media reports and physicians seeing Medicare elderly, which could confound the association between media reports and vaccination receipts. We examined media report variables with 1-week lag, since Medicare elderly waited 10 days on average to see their primary care provider for a checkup in 1997 (Trude and Ginsburg 2005).

As others have done (Schneider et al. 2001; Landon et al. 2004; Yoo and Frick 2005; O'Malley and Forrest 2006;), we included among the covariates demographic and socioeconomic information, chronic diseases indicators (classified into two risk levels for influenza (CDC 2007a), general health care-seeking attitudes, preventive behavior indicators (prior season's influenza vaccination), presence of usual source of care, and others. Table 1 provides summary statistics on outcome and explanatory variables.

Table 1.

Summary of Outcome and Explanatory Variables in Survival Analysis (Adjusted by Sampling Weights)

	1999	2000	2001
Sample size (unweighted/weighted)	(7,208/22,232,551)	(7,071/22,182,617)	(7,136/22,999,217)
Outcome variable
Receipt of influenza vaccine (DV)	0.490	0.443	0.486
Explanatory variables
Vaccine dose distributed in past 4 weeks (million)*	1.51	1.88	2.26
	(4.56E−2)	(5.83E−2)	(7.07E−2)
Influenza epidemic level(Percentage of specimens testing positive for flu)*	7.00	2.10	1.59
	(.143)	(.113)	(8.43E−2)
RSV epidemic level (Percentage of specimens testingpositive for flu)†	13.18	10.674	10.05
	(0.112)	(0.102)	(7.49E−2)
Weather temperature (°F)‡	52.59	49.21	53.56
	(0.463)	(0.520)	(0.495)
Female (DV)	0.581	0.588	0.587
Age: ≥75 (DV)	0.522	0.529	0.533
Race: white (DV)	0.885	0.882	0.875
Education: 13 years or more (DV)	0.663	0.662	0.674
Income: U.S.$20,000 or more (DV)	0.501	0.512	0.523
General health: fair or poor (DV)	0.234	0.235	0.235
Chronic disease(s): high-risk level (DV)	0.483	0.501	0.512
Medigap (supplemental private) insurance (DV)	0.746	0.735	0.728
Medicaid enrollment (DV)	0.112	0.115	0.114
Household member: 2 or more (DV)	0.668	0.673	0.676
Residence in metropolitan area (DV)	0.715	0.712	0.717
Influenza vaccination during the prior season (DV)	0.696	0.701	0.688
Pneumonia shot during past 5 years (DV)	0.650	0.661	0.677
Current smoking (DV)	0.111	0.104	0.102
Attitude: avoid going to a physician (DV)	0.275	0.272	0.257
Attitude: visit a physician as soon as feel bad (DV)	0.363	0.365	0.341
Usual source of care (DV)	0.942	0.942	0.950
Number of media reports on “Influenza” per week§
USA Today (text)¶	4.62	4.47	4.43
	(9.50E−02)	(9.44E−02)	(9.88E−02)
USA Today (headline)¶	0.564	0.520	0.811
	(1.16E−02)	(1.09E−02)	(1.82E−02)
4TV (text)¶	36.81	28.61	42.18
	(.104)	(8.68E−2)	(.108)
4TV(headline)¶	8.30	7.57	4.68
	(2.76E−2)	(2.35E−2)	(1.47E−2)
AP (text)∥	18.19	18.46	20.17
	(3.07E−17)	(1.66E−17)	(3.44E−17)
AP (headline)∥	1.48	0.902	1.28
	(1.38E−17)	(4.63E−18)	(7.62E−18)
CDC press releases (text)∥	0.574	0.426	0.459
	(9.70 E−18)	(4.05E−18)	(8.95E−18)
CDC press releases (headline)∥	0.172	0.230	0.230
	(5.75E−18)	(4.04E−18)	(5.09E−18)
Number of media reports on “Influenza (+) Vaccine Delay/Shortage” per week§
USA Today (text)¶	N/A	1.09	1.04
		(2.30E−02)	(2.34E−02)
USA Today (headline)¶	N/A	0.227	N/A
		(4.76E−03)
4TV (text)¶	N/A	9.82	4.70
		(2.93E−2)	(1.30E−2)
4TV (headline)¶	N/A	1.96	N/A
		(7.82E−3)
AP (text)∥	N/A	2.71	2.57
		(1.09E−17)	(1.29E−17)
AP (headline)∥	N/A	0.172	0.057
		(4.14E−18)	(1.50E−18)
CDC press releases (text)∥	N/A	0.311	0.287
		(4.62E−18)	(7.51E−18)
CDC press releases (headline)∥	N/A	N/A	N/A

Note. Mean (SD); chronic disease(s) are defined in main text.

^*Weekly varying at the nine census region level.

^†Weekly varying at the four region level defined by the CDC (Centers for Disease Control and Prevention 2000a, 2002).

^‡Weekly varying at the 50 state levels.

^§Units for media reports (per week): (i) USA Today: a product of “the number of flu-related articles” and “the circulation number in a resident state [million] (normalized by the number of total population in the state),” distinguishing articles published in weekday and weekend versions; (ii) 4TV—sum of ABC, CBS, FOX, and NBC, each is a product of “the number of related programs” and “the household number tuned to a program in a resident state [million] (normalized by the number of total population in the state)”; (iii) AP: the number of articles including flu-related keywords; (iv) CDC press release: the number of releases including flu-related keywords.

^¶Daily varying at the 50 state levels.

^∥Daily varying at the national level.

AP, Associated Press; DV, dichotomous variable; N/A, not available; RSV: respiratory syncytial virus; 4TV, four television (ABC, CBS, FOX, and NBC) networks.

Our primary analysis used a generalized-gamma (gamma) model to analyze time to vaccination. We used this model rather than other models for four reasons. First, Cox models are unable to estimate the effect of key covariates without geographic variations, such as CDC press release and AP articles, while gamma models enable us to estimate their effects on vaccination. Second, the gamma model assumes a very flexible underlying survival curve like Cox models. Third, Wald tests rejected exponential, Weibull, and log-normal models, but they did not reject gamma models among parametric survival models. Fourth, in our sensitivity analyses, the nonconvergence problem was observed only in some Cox models. That is, estimates of Cox models were less stable and less robust to sensitivity analyses (i.e., did not fit well with the data), compared with parametric models implemented.

Hypothesized mass media effects on both the mean vaccination timing among the vaccinated (measured by days earlier) and the annual vaccination rate were predicted based on the estimated coefficients of gamma models. A problem of survival models, including gamma models, is that they can only distinguish the media effect on vaccination timing from that on the annual vaccination rate under two restrictive conditions. Under the first condition where a subset of 50 states has the same total vaccination rate, gamma models can measure the media effect on timing applicable only among this subset. Similarly, in the second condition where a subset of 50 states has the same mean vaccination timing, gamma models can estimate the media effect on the annual total vaccination rate only among this subset. Instead of showing the estimated media effects under such restrictive conditions (which was feasible in a subset of our data), our paper presents two hypothetical settings where media affects either vaccination timing or annual rate only. This is because media is expected to have both effects but smaller in magnitude than those estimated in these hypothetical settings in reality. These estimates are generalizable to the entire U.S. Medicare population and are useful for policy.

As a secondary analysis to directly test the mass media effect on annual vaccine receipt, logit models were used where the outcome variable was vaccine receipt within various time windows.

RESULTS

Demographics

Based on the demographic covariates included in Table 1, the study population was similar in all 3 years and reflected the Medicare population.

Association between Media Reports and Vaccination Timing

Table 2 shows the multivariate associations between vaccination timing and the mass media reports. Columns 1, 2, and 4 in Table 2 show the associations when a media report contained keywords of “influenza” or “flu.” Other columns included additional keywords of “vaccine” or “shot” and of “delay” or “shortage” or “late.”

Table 2.

Effects of Mass Media Reports of “Influenza” and “Influenza Vaccine Delay/Shortage” on Influenza Vaccination Timing among Medicare Elderly (Years 1999, 2000, 2001) Estimated by Generalized Gamma Model (Coefficient [SE])^†

	1999 Keyword: Influenza/Flu (Headline)	2000 Keyword: Influenza/Flu (Headline)	2000 Keyword: Influenza and Vaccine/Shot and Delay/Shortage (Headline)	2001 Keyword: Influenza/Flu (Headline)	2001 Keyword: Influenza and Vaccine/Shot and Delay/Shortage (Full Text)
USA Today‡§	0.997	0.995	0.965	0.970	0.981
	(5.63E−03)	(6.15E−03)	(1.17E−02)***	(6.91E−03)***	(5.13E−03)***
4TV‡§	0.997	0.997	0.993	0.997	0.993
	(1.06E−03)***	(6.70E−04)***	(1.71E−03)***	(1.27E−03)**	(1.83E−03)***
AP‡¶	0.973	0.976	0.985	0.983	0.990
	(8.50E−03)***	(7.91E−03)***	(1.95E−02)	(1.05E−02)	(4.33E−03)**
CDC press releases‡¶	1.062	0.944	N/A	1.001	0.977
	(2.08E−02)***	(1.68E−02)***		(2.35E−02)	(1.07E−02)**
Vaccine supply (million)∥	0.948	0.963	0.962	0.944	0.931
	(1.04E−02)***	(1.83E−02)*	(1.83E−02)**	(1.25E−02)***	(1.20E−02)***
Influenza epidemic level (% positive)∥	1.045	1.018	1.019	1.032	1.032
	(4.33E−03)***	(4.63E−03)***	(4.69E−03)***	(8.29E−03)***	(8.00E−03)***
RSV epidemic level (% positive)††	1.009	1.005	1.005	1.015	1.012
	(2.22E−03)***	(1.77E−03)***	(1.76E−03)***	(2.08E−03)***	(2.09E−03)***
State-level weather temperature (°F)‡‡	0.998	0.997	0.997	1.000	1.001
	(1.63E−03)	(1.05E−03)**	(1.00E−03)***	(1.52E−03)	(1.50E−03)
Female	0.957	0.977	0.977	0.949	0.952
	(1.47E−02)***	(1.44E−02)	(1.42E−02)	(1.51E−02)***	(1.40E−02)***
Age ≥75	0.975	0.967	0.968	0.986	0.986
	(1.28E−02)*	(1.42E−02)**	(1.40E−02)**	(1.57E−02)	(1.47E−02)
Race: white	0.886	0.885	0.886	0.870	0.876
	(2.12E−02)***	(2.49E−02)***	(2.45E−02)***	(2.53E−02)***	(2.36E−02)***
Education: 13 years or more	0.993	0.989	0.989	0.984	0.984
	(1.77E−02)	(1.47E−02)	(1.45E−02)	(2.05E−02)	(1.91E−02)
Income: U.S.$20,000 or more	0.970	0.957	0.957	0.997	0.996
	(1.50E−02)*	(1.56E−02)***	(1.54E−02)***	(1.87E−02)	(1.77E−02)
General health: fair or poor	1.019	1.027	1.027	1.041	1.039
	(1.69E−02)	(2.00E−02)	(1.97E−02)	(1.58E−02)***	(1.44E−02)***
Chronic disease: high-risk level	0.948	0.929	0.930	0.912	0.921
	(1.22E−02)***	(1.49E−02)***	(1.47E−02)***	(1.70E−02)***	(1.57E−02)***
Medigap insurance	0.876	0.903	0.905	0.898	0.903
	(2.04E−02)***	(2.41E−02)***	(2.37E−02)***	(2.13E−02)***	(2.04E−02)***
Medicaid enrollment	0.994	1.003	1.003	1.022	1.022
	(3.17E−02)	(2.93E−02)	(2.90E−02)	(3.55E−02)	(3.27E−02)
Household member: 2 or more	0.986	0.959	0.960	0.965	0.969
	(1.37E−02)	(1.32E−02)***	(1.31E−02)***	(1.55E−02)**	(1.42E−02)**
Residence in metropolitan area	1.047	1.013	1.014	0.969	0.971
	(3.76E−02)	(3.08E−02)	(3.03E−02)	(3.23E−02)	(3.05E−02)
Flu vaccination during prior season	0.552	0.531	0.538	0.548	0.568
	(2.20E−02)***	(2.64E−02)***	(2.54E−02)***	(1.67E−02)***	(1.61E−02)***
Pneumonia shot during past 5 years	0.911	0.905	0.906	0.825	0.840
	(1.64E−02)***	(2.12E−02)***	(2.06E−02)***	(1.88E−02)***	(1.76E−02)***
Current smoking	1.032	0.996	0.996	1.051	1.043
	(2.22E−02)	(2.98E−02)	(2.94E−02)	(3.04E−02)*	(2.75E−02)
Attitude: avoid going to a physician	1.012	1.010	1.009	0.974	0.976
	(1.70E−02)	(1.48E−02)	(1.46E−02)	(1.72E−02)	(1.59E−02)
Attitude: visit a physician as soon as feel bad	1.008	1.014	1.013	0.988	0.989
	(1.46E−02)	(1.42E−02)	(1.39E−02)	(1.58E−02)	(1.49E−02)
Usual source of care	0.847	1.007	1.007	0.900	0.910
	(3.44E−02)***	(4.06E−02)	(3.98E−02)	(3.39E−02)***	(3.18E−02)***

^†Proportion of time spent unvaccinated if exposed to factor, versus if not exposed to the factor (see text).

^‡Units for media reports (per week): (i) USA Today: a product of “the number of flu-related articles” and “the circulation number in a resident state [million] (normalized by the number of total population in the state),” distinguishing articles published in weekday and weekend versions; (ii) 4TV—sum of ABC, CBS, FOX, and NBC, each is a product of “the number of related programs” and “the household number tuned to a program in a resident state [million] (normalized by the number of total population in the state)”; (iii) AP: the number of articles including flu-related keywords; (iv) CDC press release: the number of releases including flu-related keywords.

^§Daily varying at the 50 state levels.

^¶Daily varying at the national level.

^∥Weekly varying at the nine census region level.

^††Weekly varying at the four region level defined by the CDC (Centers for Disease Control and Prevention 2000a, 2002).

^‡‡Weekly varying at the 50 state levels.

^*p-value <0.1

^**p-value <0.05

^***p-value <0.01.

AP, Associated Press; N/A, not available; RSV, respiratory syncytial virus; 4TV, four television (ABC, CBS, FOX, and NBC) networks.

Each model included three types of media variables measuring the common set of keywords appearing in the headline, except a model containing keywords of “delay/shortage” in 2001 where keywords appeared in the text only. Because the headline tended to have a stronger association with vaccination than keyword matches in the text only, models in Table 2 included the media variables of headlines unless headlines were unavailable. These models used the periods of measuring media reports with a stronger association with vaccination, unique to each year. Namely, the models in 2001 used the media reports measured during an immediate past 1 week (1–7 days), and the models in 1999 and 2000 did the media reports during a past 1 week with a 1-week lag (8–14 days).

It should be noted that the coefficients of gamma models in Table 2 indicate the proportion of “time spent in the unvaccinated state over a period of 4 months conditional on being exposed to an influenza-related report with a circulation/audience households (AHH) of one million” divided by “the same time conditional on no report exposure.” That is, these coefficients are normalized to measure the effect of changing circulation/AHH by 1 million. The actual circulation/AHH numbers differ from that. Also, a covariate's coefficient being smaller than one implies the positive association between the covariate and earlier vaccination timing.

For example, the coefficient of 0.981 for USA Today in 2001, in the second row in the far right column in Table 2, indicates that unvaccinated time of a Medicare beneficiary with the exposure of one USA Today article containing “influenza vaccine delay/shortage” in the past week is estimated to be 98.1 percent of unvaccinated time without such exposure. Because the actual circulation of USA Today in 2001 was 2.36 million, the actual effect on vaccination timing or the proportion of unvaccinated time was not 0.981 (per 1 million circulation) but 0.956 (=0.981^2.36), which is statistically significant (p<.01). Similarly, because on the average 4.25 million households tuned to CBS network in 2001, one CBS network program reporting “influenza vaccine delay/shortage” reduced the proportion of unvaccinated time to 0.971 (=0.993^4.25), based on the 4TV coefficient of 0.993 (third row, far right column in Table 2).

The association between receipt of influenza vaccine and number of any type of media reports was multiplicative, so that reports from multiple sources during the same week were associated with greater likelihood of vaccination. Extending two examples described in the previous paragraph, when “influenza vaccine delay/shortage” was reported by one USA Today article and one CBS network program during the same week in 2001, the proportion of unvaccinated time was 0.928 (=0.956 (USA Today) × 0.971 (CBS) shown in the previous paragraph), that is, the vaccination timing became much earlier than a case exposed to single media source.

Among all models in Table 2, all three media variables were positively associated with earlier vaccination timing. These associations were statistically significant (p<.05) for 4TV in all five models and for USA Today and AP in three out of five models.

Effects of Additional Keywords

Adding the keywords “vaccine” or “shot” and “delay” or “shortage” or “late” tended to increase the association between a media report and vaccination timing in terms of magnitude or statistical significance level, regardless of the type of media. Coefficients in columns 3 and 5 are smaller than or equal to corresponding coefficients in columns 2 and 4 in Table 2 (p<.05).

Effects of Controlling Variables

CDC press releases had inconsistent associations with the vaccination timing. The association between vaccine availability and vaccination timing was always positive and statistically significant (p<.01) in all models. For example, the model in 1999 in Table 2 (row 6, column 1) implied that an additional 1 million vaccine doses distributed in a resident region over the past 4 weeks was associated with an unvaccinated period reduced by 5.2 percent (=1−0.948), relative to a region in which no additional vaccines were distributed.

The results of other control variables were consistent with the literature (Schneider et al. 2001; Landon et al. 2004; Yoo and Frick 2005; O'Malley and Forrest 2006;). For example, vaccination rates were positively associated with having received influenza vaccination in the prior season, having ever received pneumococcal vaccination, being white, and having Medigap insurance.

As discussed earlier, we present two hypothetical extreme settings where media affected either vaccination timing only (Table 3) or annual-rate only (Table 4). In reality, media is expected to have both effects but in a smaller magnitude than those presented in these tables.

Table 3.

Effects of Mass Media Reports of “Influenza” and “Influenza Vaccine Delay/Shortage” on Mean Influenza Vaccination Timing^† among the Vaccinated Medicare Elderly (Year 1999, 2000, 2001) Estimated by Generalized Gamma Model (Days Received Earlier [SE]) (Based on Generalized Gamma Model in Table 2)

	1999		2000		2001
	Change in Mean Vaccination Timing (Day)		Change in Mean Vaccination Timing (Day)		Change in Mean Vaccination Timing (Day)
	1 Unit Increase‡	Observed All Reports	1 Unit Increase‡	Observed All Reports	1 Unit Increase‡	Observed All Reports
USA Today§
Influenza¶	0.208	0.145	0.504	0.258	3.18	2.05
	(2.08E−04)***	(7.28E−04)***	(2.88E−04)***	(1.16E−03)***	(2.22E−03)***	(1.03E−02)***
Influenza vaccine delay/shortage∥			3.79	0.97	2.01	2.26
	N/A	N/A	(2.11E−03)***	(5.50E−03)***	(1.32E−03)***	(8.25E−03)***
4TV§
Influenza¶	0.253	1.96	0.324	2.41	0.342	1.77
	(2.51E−04)***	(8.34E−03)***	(1.84E−04)***	(6.61E−03)***	(2.37E−04)***	(5.86E−03)***
Influenza vaccine delay/shortage∥			0.81	1.79	0.73	4.08
	N/A	N/A	(4.53E−04)***	(7.00E−03)***	(4.73E−04)***	(1.01E−02)***
AP§
Influenza¶	2.15	3.21	2.64	2.16	1.81	2.39
	(2.10E−03)***	(1.32E−02)***	(1.51E−03)***	(5.20E−03)***	(1.27E−03)***	(7.15E−03)***
Influenza vaccine delay/shortage∥			1.65	0.223	1.06	2.81
	N/A	N/A	(9.23E−04)***	(1.25E−03)***	(6.89E−04)***	(6.53E−03)***

^†These estimated effects assume a hypothetical extreme setting where mass media reports affect vaccination timing only, not annual vaccine receipt.

^‡1 unit is 1 media report in every week in (i) articles in USA Today with 1 million circulation, (ii) programs in 4TV that 1 million households tuned, and (iii) articles in AP.

^§Units for media reports (per week): (i) USA Today: a product of “the number of flu-related articles” and “the circulation number in a resident state [million] (normalized by the number of total population in the state),” distinguishing articles published in weekday and weekend versions; (ii) 4TV—sum of ABC, CBS, FOX, and NBC, each is a product of “the number of related programs” and “the household number tuned to a program in a resident state [million] (normalized by the number of total population in the state)”; (iii) AP: the number of articles including flu-related keywords.

^¶Headline report in each year.

^∥Headline report (2000), Full text report (2001).

^***p-value <0.01.

AP, Associated Press; N/A, Not available; 4TV, four television (ABC, CBS, FOX, and NBC) networks.

Table 4.

Effects of Mass Media Influenza-Related Reports on Annual Influenza Vaccine Receipt^† among Medicare Elderly (Year 1999, 2000, 2001) (% change in annual vaccination rate) (based on generalized-gamma model in Table 2)

	1999		2000		2001
	Change in Annual Vaccination Rate		Change in Annual Vaccination Rate		Change in Annual Vaccination Rate
	1 Unit Increase‡ (SE)	Observed All Reports (SE)	1 Unit Increase‡ (SE)	Observed All Reports (SE)	1 Unit Increase‡ (SE)	Observed All Reports (SE)
USA Today§
Influenza¶	0.432%	0.529%	0.646%	0.378%	3.83%	5.00%
	(1.02E−04)***	(1.80E−04)***	(1.30E−04)***	(1.16E−04)***	(6.33E−04)***	(1.67E−03)***
Influenza vaccine delay/shortage∥			5.18%	1.29%	2.73%	3.15%
	N/A	N/A	(1.05E−03)***	(4.08E−04)***	(4.98E−04)***	(1.10E−03)***
4TV§
Influenza¶	0.457%	7.94%	0.384%	3.40%	0.407%	3.02%
	(1.09E−04)***	(2.11E−03)***	(7.74E−05)***	(7.34E−04)***	(6.92E−05)***	(5.30E−04)***
Influenza vaccine delay/shortage∥			1.03%	2.27%	0.913%	5.77%
	N/A	N/A	(2.10E−04)***	(5.00E−04)***	(1.72E−04)***	(1.19E−03)***
AP§
Influenza¶	3.81%	10.40%	3.24%	3.56%	2.17%	3.71%
	(9.21E−04)***	(2.69E−03)***	(6.55E−04)***	(7.22E−04)***	(3.69E−04)***	(6.05E−04)***
Influenza vaccine delay/shortage∥			2.22%	0.431%	1.37%	4.95%
	N/A	N/A	(4.53E−04)***	(9.44E−05)***	(2.58E−04)***	(9.52E−04)***

^†These estimated effects assume a hypothetical extreme setting where mass media reports affect annual vaccine receipt only, not vaccination timing.

^‡1 unit is 1 media report in every week in (i) articles in USA Today with 1 million circulation, (ii) programs in 4TV that 1 million households tuned, and (iii) articles in AP.

^§Units for media reports (per week): (i) USA Today: a product of “the number of flu-related articles” and “the circulation number in a resident state [million] (normalized by the number of total population in the state),” distinguishing articles published in weekday and weekend versions; (ii) 4TV—sum of ABC, CBS, FOX, and NBC, each is a product of “the number of related programs” and “the household number tuned to a program in a resident state [million] (normalized by the number of total population in the state)”; (iii) AP: the number of articles including flu-related keywords.

^¶Headline report in each year.

^∥Headline report (2000), Full text report (2001).

^***p-value <0.01.

AP, Associated Press; N/A, not available; 4TV, four television (ABC, CBS, FOX, and NBC) networks.

Association between Media Reports and Vaccination Timing Measured by Days Received Earlier

Table 3 presents the shift of mean vaccination timing among the vaccinated associated with media reports, corresponding to Table 2, which presented gamma model coefficients. Table 3 presents the media effects in a hypothetical scenario (columns 1, 3, and 5) and those in an observed scenario (columns 2, 4, and 6). In one of the hypothetical scenarios (row 1, column 5 in 2001), if USA Today reports one headline article on “influenza” every week from September 1 to December 31, the mean vaccination timing is expected to shift earlier by 15.3 days (15.3=(3.18)^2.36, assuming 2.36 million circulation) compared with a hypothetical season without such headline at all. In fact, such headline articles (seven in total) appeared in 5 weeks in 2001 and were associated with the shift in the mean vaccination timing by 2.05 days earlier (row 1, far right column in 2001).

Comparing the magnitude of 1 unit increase of a certain media variable across 3 years, the magnitude in 2000 tended to be greater. However, the magnitude of observed reports tended to be smaller in 2000 than in 2001, probably due to the most serious vaccine supply limitation in 2000. It should be noted that the mean vaccination timing would be delayed, if, for example, media reports in late December increased the annual vaccination rate but also increased the proportion of late vaccination timing among the all vaccinated.

Association between Media Reports and Annual Vaccination

Table 4 presents the multivariate associations between annual vaccine receipt and mass media reports, corresponding to Table 2. Influenza-related mass media reports were positively associated with the annual vaccination rate. Like Table 3, Table 4 also presents the media effects in a hypothetical scenario and those in an observed scenario. In one of the hypothetical scenarios (row 1, column 5 in 2001), if USA Today reports one headline article on “influenza” every week from September 1 to December 31, the annual vaccination rate is expected to increase by 9.3 percentage points (1.093=(1.0383)^2.36, assuming 2.36 million circulation) compared with a hypothetical season without such headline at all. Actual headline articles (seven in total) in 2001 were associated with the increase in the annual vaccination by 5 percentage points (row 1, far right column in 2001), in comparison with a scenario where those headlines did not occur.

As observed in Table 2, the positive association between annual vaccination and the media reports tended to be greater, particularly in 4TV and USA Today, when a report included additional keywords like shortage/delay. Comparing the magnitude of a 1 unit increase of a certain media variable across 3 years, the magnitude in 2000 tended to be greater. However, the magnitude of observed reports tended to be smaller in 2000 among 3 years, probably due to the most serious vaccine supply limitation in 2000.

In secondary analysis, logit models yielded unstable and implausibly large estimates compared with those of gamma models.

DISCUSSION

The literature to date has found little evidence that community-wide education-only interventions (e.g., newspapers and television campaign) improve the delivery of vaccinations (Briss et al. 2000). To our knowledge ours is the first study to note a positive association between flu-related media reports and influenza vaccination rates in the weeks following the reports in a nationally representative population of elderly individuals.

We found that reports about influenza in both print and television media had a statistically significant positive association with the influenza vaccine receipt in terms of earlier timing as well as the annual vaccination rate, at least among a subset of 50 states with the same level of mean vaccination timing or annual rate. Tables 3 and 4 presented these associations that are hypothetical but generalizable to the entire Medicare population.

The magnitude of this positive association had two general tendencies. First, reports with a headline that included flu-related keywords tended to have a greater association with influenza vaccination than did reports including flu-related keywords but without such a headline. Second, additional specific information, including articles about impending vaccine shortages, tended to increase the likelihood of earlier timing and annual vaccination even further. These findings are plausible—key terms in the headline are more likely to be recognized by readers and viewers than those only in text, and a shortage is a more urgent cause for vaccination than a general warning about the dangers of the flu.

In deriving policy implications from these empirical results, three points should be considered. First, media reports are often not independent events. For instance, our media variables may contain the effect of other types of media not included in our model, for example, the Internet and radio, since newspaper and television stories are often published in those media as well. Television and newspaper media sometimes explicitly copy reports from other media outlets. However, our models are expected to capture major media effects during our study period for two reasons. The first reason is that 53 percent of the general public identified television as the main information source about current news according to a national survey in 2001 (National Science Foundation Division of Science Resources Statistics 2002). The corresponding statistics for newspapers, Internet, radio, and magazines were 29, 7, 5, and 3 percent, respectively. These statistics are consistent with our results where 4TV had more consistent positive associations with vaccination than other media sources. The second reason is that approximately 45 percent of the stories in an average newspaper are attributed to wire services, for example, AP included in our analyses (The Readership Institute 2004).

Second, it is important to control for vaccine supply in evaluating the overall media effect on both the vaccination timing and the annual rates. Out of the 3 years of this study, 2000 had the lowest annual vaccination rate despite the largest number of flu-related headline reports including shortage/delay that were estimated to have a larger association (per 1 unit of media report) with vaccination. Also, the vaccine supply appears to affect the time lag between media reports and vaccine receipts in our results. In 1999, a year without any vaccine supply shortage or delay, media reports with a 1-week lag had stronger associations than those without a 1-week lag. This is consistent with the reported average waiting time (i.e., time lag between appointment making and service receipt) for a checkup being 10 days among Medicare beneficiaries (Trude and Ginsburg 2005). On the other hand, if media reports on vaccine delay or shortage increased the incentive to be vaccinated as soon as possible, time lag between media reports and vaccine receipt would be shorter. As a result, media reports during the immediate past 1 week, without a time lag, are expected to have a larger association than media reports with a 1-week lag in 2000 and 2001. However, such expected association was robust only in 2001 when there was vaccine supply delay but no serious shortage. The absence of such robust associations in 2000 could be explained partly by the more severe vaccine shortage/delay in 2000, compared with 2001.

The public release of regional and local vaccine supply data could improve the evaluation of the media effect on the vaccination rate. Expected stabilized influenza vaccine supply at the national level (CDC 2007b) might mask vaccine supply shortages and/or delays that occur at local and regional levels. Also, vaccine supply problems could still occur even at a national level due to uncertainties in production and regulatory processes (CDC 2007a). Vaccine supply issues and their relation to coverage may continue to be important concerns if influenza vaccination recommendation is expanded to include additional target populations (Schwartz et al. 2006).

The final point to be considered for policy is that it may not be optimal to encourage vaccination early in the season for lower-risk populations (CDC 2000b, 2001b, 2007a). Intensive media coverage on flu, combined with limited supplies, could lead to suboptimal vaccine allocation in which the highest-risk target populations might delay vaccination if lower-risk populations receive vaccination earlier than necessary. Additionally, because influenza could spread in February or later, vaccination after January is still recommended (CDC 2007a).

Our empirical results may impact policy by helping the CDC or other public institutions set priorities among various forms of vaccination campaigns. Specific strategies to use mass media (newspapers or television) effectively include (i) using a headline, as well as text, containing specific key words like shortage/delay in addition to influenza in general, (ii) implementing repeated media use rather than one-time use, due to the short-term media effect and the cumulative effects during a short term, and (iii) assuring sufficient vaccine supply before a media release on supply delay/shortage because of individuals' quick response to such information.

Limitations

One limitation is that generally increased public attention to the flu could lead to both media coverage and vaccination. This potential mechanism would overestimate the mass media effects on vaccination. On the other hand, the measurement errors of mass media reports, partly due to some simplifying assumptions, tend to attenuate their effects on vaccination. These simplifying assumptions did not account for article/program specific circulation/AHH, but assigned the year-specific average circulation/AHH for all articles/programs. Similar attenuated effects are possible for vaccine supply assumed to be distributed in proportion to the regional total population.

Another limitation is the potential measurement error in vaccination receipt. In our data, the vaccination rates were 49.0, 44.3, and 48.6 percent (claims data), 70.0, 68.7, and 69.8 percent (self-reported data) for 1999–2001, respectively. The agreement rates between these two data were 75, 73, and 74 percent (κ coefficient: 0.50, 0.48, and 0.49) for 1999–2001, respectively. These discrepancies could be partly explained by vaccinations received outside Medicare billing institutions, absence of bills despite the actual receipt, and recall bias in survey data (CDC 1994; Zimmerman et al. 2003;).

The generalizability of our findings may be limited by the exclusion of Medicare managed care enrollees (around 20 percent of community-dwelling Medicare beneficiaries), who have been shown to be more likely to receive influenza vaccination than those enrolled in fee-for-service Medicare plans in past work (Schneider et al. 2001; Landon et al. 2004; O'Malley and Forrest 2006;).

CONCLUSION

Media reports in both television and newspaper mass media appear to be associated with earlier and increased influenza vaccination among elderly Americans. To further improve vaccination coverage among the elderly population, new integrated approaches should be explored including the use of mass media such as television and newspapers.

Supporting Information

Additional supporting information may be found in the online version of this article:

Appendix SA1: Author Matrix.

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.