University of Groningen
An economic assessment of high-dose influenza vaccine
van Aalst, Robertus
DOI:
10.33612/diss.127973664
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
van Aalst, R. (2020). An economic assessment of high-dose influenza vaccine: Estimating the vaccine-preventable burden of disease in the United States using real-world data. University of Groningen. https://doi.org/10.33612/diss.127973664
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
2
Relative Vaccine Effectiveness of High-Dose versus Standard-Dose Influenza Vaccines among Veterans Health Administration Patients
Yinong Young-Xu a,b, Robertus van Aalst a, Salaheddin M.Mahmud c,d, Kenneth J. Rothman e,f,
Julia Thornton Snider g, Daniel Westreich h, Vincent Mor i,j, Stefan Gravenstein j,k,l,m,
Jason K.H. Lee n,o, Edward W. Thommes p,q, Michael D. Decker p,r, and Ayman Chit n,p
a Clinical Epidemiology Program, Veterans Affairs Medical Center, White River Junction, VT, USA b Department of Psychiatry, Geisel School of Medicine at Dartmouth, Hanover, NH, USA
c Department of Community Health Sciences, College of Medicine, University of Manitoba, Winnipeg, MB, Canada
d George & Fay Yee Center for Healthcare Innovation, University of Manitoba/Winnipeg Regional Health Authority, Winnipeg, MB, Canada
e RTI Health Solutions, Research Triangle Institute, Research Triangle Park, NC, USA f Department of Epidemiology, Boston University School of Public Health, Boston, MA, USA g Precision Health Economics, Oakland, CA, USA
h Department of Epidemiology, Gillings School of Global Public Health, UNC-Chapel Hill, Chapel Hill, NC, USA
i Health Services, Policy, and Practice, Center for Gerontology and Health Care Research, Brown University School of Public Health, Providence, RI, USA
j Veterans Administration Medical Center, Providence, RI, USA
k Center for Geriatrics and Palliative Care, University Hospitals Cleveland Medical Center and Case Western Reserve University, Cleveland, OH, USA
l Center for Gerontology and Health Care Research, School of Public Health, Brown University, Providence, RI, USA
m Department of Health Services, Policy and Practice, School of Public Health, Brown University, Providence, RI, USA
n Leslie Dan School of Pharmacy, University of Toronto, Toronto, ON, Canada o Sanofi Pasteur, Toronto, ON, Canada
p Sanofi Pasteur, Swiftwater, PA, USA
q Department of Mathematics & Statistics, University of Guelph, Guelph, ON, Canada r Department of Health Policy, Vanderbilt University School of Medicine, Nashville, TN, USA
Published on 5 May, 2018
Chapter 2
ABSTRACT
Background
We examined whether a high-dose inactivated influenza vaccine was more efficacious in preventing hospitalizations than a standard-dose vaccine in the Veterans Health Administration (VHA) senior population.
Methods
This study estimated the relative vaccine effectiveness (rVE) of high-dose versus standard-dose using a retrospective cohort of VHA patients 65 years of age or older in the 2015-16 influenza season. To adjust for measured confounders, we matched each high-dose recipient with up to four standard-dose recipients vaccinated at the same location within a two-week period and having two or more pre-existing medical co-morbidities. We used the previous event rate ratio method (PERR), a type of difference-in-differences analysis, to adjust for unmeasured confounders.
Results
We evaluated 104,965 standard-dose and 125,776 high-dose recipients; matching decreased the population to 49,091 standard-dose and 24,682 high-dose recipients; matching decreased the population to 49,091 stadard-dose and 24,682 high-dose recipients. The matched, PERR-adjusted rVE was 25% (95% confidence interval [CI], 2-43%) against influenza- or pneumonia-associated hospitalization, 7% (95% CI: -2 to 14%) against all-cause hospitalization, 14% (95% CI, -8 to 32%) against influenza- or pneumonia-associated outpatient visit, 5% (95% CI, 2-8%) against all-cause outpatient visit, and 38% (95% CI, -5 to 65%) against laboratory-confirmed influenza.
Conclusions
In protecting senior VHA patients against influenza- or pneumonia-associated hospitalization, a high-dose influenza vaccine is more effective than a standard-dose vaccine.
INTRODUCTION
Seasonal influenza epidemics have a substantial public health and economic burden in the United States. On average, more than 200,000 people are hospitalized and upwards of 49,000 people are estimated to die each year from respiratory and circulatory complications associated with seasonal influenza virus infections [1-7]. Medical complications leading to hospitalizations and deaths are greatest among persons aged 65 years and older (hereinafter referred to as seniors) [4]. The heightened susceptibility to complications is largely due to the progressive age-related weakening of the immune system known as immunosenescence that also renders seniors less responsive to vaccines. In 2009, the US Food and Drug Administration licensed an injectable high-dose inactivated trivalent influenza vaccine (Fluzone® High-Dose, Sanofi Pasteur, PA, USA), hereinafter referred to as the high-dose vaccine (HD). HD contains four times more influenza hemagglutinin antigen than standard-dose influenza (SD) vaccines (60 μg vs. 15 μg per strain); designed to improve immune response and therefore protection in seniors.
Analyses comparing the HD and SD vaccines in preventing influenza-related hospitalizations have generated inconsistent results. A randomized clinical trial (RCT) comparing HD and SD in seniors during the 2011-12 and 2012-13 respiratory seasons found a relative vaccine effectiveness (rVE) of 24.2% (95% confidence interval [CI], 9.7%-36.5%) against laboratory-confirmed influenza [8]. In addition, the rate of all-cause hospitalization was calculated to be 6.9% (95% CI, 0.5%-12.8%) lower in the HD group than in the SD group; the rate of serious pneumonia was 39.8% (95% CI, 19.3%-55.1%) lower in the HD group; and the rate of serious cardiorespiratory medical events possibly related to influenza was 17.7% (95% CI, 6. 6%-27.4%) lower in the HD group [9]. A subsequent observational study further supported the benefit of HD versus SD vaccination when comparing two outcomes in US Medicare beneficiaries during the 2012-13 respiratory season [10]. The authors found a relative benefit for HD over SD against rapid response influenza test followed by oseltamivir treatment (22%; 95% CI, 15%-29%), and against influenza attributed hospitalization or emergency department visits (22%, 95% CI, 16%-27%). In contrast, a study among community-dwelling senior US Veterans during the 2010-11 respiratory season found HD no more effective than SD in preventing influenza-attributed hospitalizations, with a calculated relative efficacy of only 2% (95% CI, -40% to 32%) [11].
Chapter 2
Previous observational studies have differed by study design details such as inclusion/ exclusion criteria, timeframe, and outcomes, warranting additional examination of the effectiveness of the HD vaccine as necessary guidance for clinical practice and public policy, especially for Veterans.
In observational studies of treatment effectiveness, confounding by indication is well recognized but challenging to rectify when confounders are difficult to quantify due to a lack of clear definition and verifiable measures. One such confounder, in this case, is frailty – a geriatric condition characterized by an increased risk of catastrophic declines in health and function among older adults [12, 13]. It is possible that two individuals of the same age, sex, and with the same number of documented co-morbid conditions could have considerable differences in their frailty that would be obvious to an examining physician, but elusive to a researcher considering only claims and electronic medical records.
With these considerations, we conducted a retrospective cohort study to estimate rVE of HD versus SD influenza vaccine in senior US Veterans, using a methodology particularly designed to adjust for confounding by indication induced by unmeasured differences in frailty.
METHODS
Design and Data Sources
We conducted a retrospective cohort study among senior Veterans Health Administration (VHA) patients during the 2015-2016 influenza season. We studied three categories of outcomes (1) influenza- or pneumonia-associated with a hospital or outpatient visit, (2) all-cause hospital or outpatient visit, and (3) any medical encounter with a positive influenza test.
The VHA is the single largest integrated health care system in the United States and provides care to military veterans at 144 medical centers, 1,203 community-based outpatient clinics, and 300 Veteran Centers. As of 2015, there were over 6 million patients with healthcare service records in the VHA. The VHA has an integrated and unified electronic medical record (EMR) system that stores information about all
inpatient, emergency, and outpatient visits as well as laboratory tests and medication prescriptions. Each patient is assigned a unique identification number that allows for longitudinal follow-up.
Study Population and Influenza Vaccination
The study population included all VHA patients 65 years and older with at least one inpatient or outpatient encounter in prior respiratory season of 2014-2015. This restriction improves the probability of capturing comorbidities and reduces the chance that participants were incidental VHA users. Influenza vaccination was identified using Current Procedural Terminology (CPT) codes (SD vaccine CPT codes: 90655-90659 and Q2034-Q2039; HD vaccine CPT code: 90662). We applied a number of exclusion criteria to reduce the chances of misclassification and to ensure valid comparison. First, Veterans who self-reported vaccination were excluded, because HD and SD vaccines are similar in presentation and administration making them hard to distinguish by recipients. Second, we excluded all Veterans who had received more than one influenza vaccine in the respiratory season of inquiry (2015-16). Third, we excluded patients who had received an influenza vaccine in any region other than their typical region of care (i.e. “snowbirds” or patients with dual residencies) to simplify matching. Fourth we excluded facilities with HD-vaccination rates lower than 5% (in other words, at least 5% of all patients vaccinated at the facility received HD) to ensure adequate supply of HD at that facility. Finally, recipients of either influenza vaccination during the six months before the start of baseline period (began at the end of the respiratory season in week 27 [beginning of July] of 2015 and ran until the vaccination date) were excluded so as to minimize any potential impacts from previous season vaccination. The study received institutional review board approval from the Veteran’s Institutional Review Board of Northern New England at the White River Junction VA Medical Center and the Geisel School of Medicine at Dartmouth.
Baseline and Observation Period
The baseline period began at the end of the respiratory season in week 27 (beginning of July) of 2015 and ran until the vaccination date [3, 5, 6]. The observation period spanned from two weeks after vaccination to the end of the respiratory season in week 26 (end of June) of 2016 (Figure 1).
Chapter 2
Baseline Characteristics
In addition to demographic characteristics, subjects were classified by the presence of specific health conditions during the period from the beginning of the previous respiratory season to the date of vaccination. Medical conditions considered in this study included chronic cardiac, pulmonary, renal, metabolic, liver, and neurological diseases; diabetes mellitus; hemoglobinopathies; immunosuppressive conditions; and malignancy. These were identified based on primary diagnosis codes (ICD-9/10) from outpatient visits and hospital admissions.
Figure 1. Definition of the study periods for a sample patient. National, weekly positive influenza test
percentages for the 2015–2016 respiratory season were obtained from the Centers for Disease Control and Prevention (CDC).
National, weekly positive influenza test percentages for the 2015-16 respiratory season were obtained from the Centers for Disease Control and Prevention (CDC).
We selected the Care Assessment Need (CAN) score as a proxy measure for frailty. The CAN score, much like the commonly used Charlson and Elixhauser comorbidity scores, is used to predict hospitalization within one year; however CAN was developed specifically for VHA population, and was validated using 4.6 million VHA patients. In addition to incorporating the medical conditions essential to the Charlson and Elixhauser comorbidity scores, CAN includes predictors such as sociodemographic
characteristics, prior year health care utilization, medications, and laboratory tests (some of which are unique to Veterans Administration [VA] EMR data) [14]. We used the maximum CAN score in the four weeks before vaccination.
Outcomes
The primary outcome of the study was any hospitalization where the patient record indicated either pneumonia or influenza (ICD-9: 480-488) as a principal or secondary diagnosis. Four additional outcomes were explored: (1) outpatient visits (defined as a visit to a primary care physician, emergency department, or urgent care provider) where the patient record indicated pneumonia or influenza as a principal or secondary diagnosis; (2) patient record documenting laboratory-confirmed influenza; (3) all-cause outpatient visits; and (4) all-cause hospitalizations.
Statistical Analysis
Outcome rates were calculated for both HD and SD cohorts during each period (baseline and observation). We calculated a person-time denominator by summing the number of weeks that patients were enrolled for each period. The numerator was the total number of outcomes (including multiple outcomes for a single patient) by period. Our primary analysis adjusted for confounding by matching HD recipients to SD recipients on factors described above under Baseline Characteristics. After matching (described below), we adjusted for residual confounding using the prior event rate ratio (PERR) approach [15-19]. This method, similar to the difference-in-differences method used in econometrics, involves the definition of two periods, one before the interventions (e.g. HD and SD) and another after [20]. For each intervention group, the rates of an outcome can be calculated and compared before and after the intervention. To assess the impact of the intervention, the relative rate of the post-intervention period was divided by the relative rate of the pre-intervention period.
Chapter 2
To apply the PERR method, we first computed the incidence rate ratio (RR) for each
study outcome in the observation period (RRo) and then again in the baseline period
(RRb). The RR is the rate of the outcome in HD recipients divided by the rate of the
outcome in SD recipients in any given period. Next, we computed the PERR per the following formula:
73
Statistical Analysis
Outcome rates were calculated for both HD and SD cohorts during each period (baseline and observation). We calculated a person-time denominator by summing the number of weeks that patients were enrolled for each period. The numerator was the total number of outcomes (including multiple outcomes for a single patient) by period.
Our primary analysis adjusted for confounding by matching HD recipients to SD recipients on factors described above under Baseline Characteristics. After matching (described below), we adjusted for residual confounding using the prior event rate ratio (PERR) approach [15-19]. This method, similar to the difference-in-differences method used in econometrics, involves the definition of two periods, one before the interventions (e.g. HD and SD) and another after [20]. For each intervention group, the rates of an outcome can be calculated and compared before and after the intervention. To assess the impact of the intervention, the relative rate of the post-intervention period was divided by the relative rate of the pre-post-intervention period.
To apply the PERR method, we first computed the incidence rate ratio (RR) for each study
outcome in the observation period (RRo) and then again in the baseline period (RRb). The RR is
the rate of the outcome in HD recipients divided by the rate of the outcome in SD recipients in any given period. Next, we computed the PERR per the following formula:
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 =𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑜𝑜𝑜𝑜
𝑏𝑏𝑏𝑏 and, finally, the rVE of HD to SD as:
𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑃𝑃𝑃𝑃 = (1 − 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃) × 100%
and, finally, the rVE of HD to SD as:
73
Statistical Analysis
Outcome rates were calculated for both HD and SD cohorts during each period (baseline and observation). We calculated a person-time denominator by summing the number of weeks that patients were enrolled for each period. The numerator was the total number of outcomes (including multiple outcomes for a single patient) by period.
Our primary analysis adjusted for confounding by matching HD recipients to SD recipients on factors described above under Baseline Characteristics. After matching (described below), we adjusted for residual confounding using the prior event rate ratio (PERR) approach [15-19]. This method, similar to the difference-in-differences method used in econometrics, involves the definition of two periods, one before the interventions (e.g. HD and SD) and another after [20]. For each intervention group, the rates of an outcome can be calculated and compared before and after the intervention. To assess the impact of the intervention, the relative rate of the post-intervention period was divided by the relative rate of the pre-post-intervention period.
To apply the PERR method, we first computed the incidence rate ratio (RR) for each study
outcome in the observation period (RRo) and then again in the baseline period (RRb). The RR is
the rate of the outcome in HD recipients divided by the rate of the outcome in SD recipients in any given period. Next, we computed the PERR per the following formula:
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 =𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑜𝑜𝑜𝑜
𝑏𝑏𝑏𝑏 and, finally, the rVE of HD to SD as:
𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑃𝑃𝑃𝑃 = (1 − 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃) × 100%
Matching
For every HD vaccine recipient at one of the 1,347 VA facilities, we selected between 1 and 4 patient(s) at the same facility who received an SD vaccine within one week from the HD vaccination date, either before or after. This process addressed temporal and geographic factors possibly associated with access to HD (e.g., supply) and influenza disease exposure (i.e., viral activities). To adjust for confounding by indication, we also matched patients on the presence of two or more pre-existing medical comorbidities. We assessed the effectiveness of matching in balancing patient characteristics by comparing the standardized differences in means for baseline characteristics between HD and SD vaccination groups. Both unmatched and matched populations were analyzed with the PERR method and results were shown side by side.
Sensitivity Analysis
A sensitivity analysis using Generalized Estimating Equations (GEE) was performed [21]. GEE was selected in order to account for clustering at both the patient and facility levels and to adjust explicitly for individual characteristics and facility-level factors such as HD adoption rates. The interaction term between intervention (HD vs. SD) and period (baseline and observation) was used to estimate the rVE. To be more representative of the VHA patient population, we applied this sensitivity analysis to an unmatched cohort that included subjects with self-reported vaccinations and those sites with HD adoption rates lower than 5%. Relaxing the exclusion criteria quadrupled our cohort size from one quarter of a million in our main analysis to one million in the sensitivity analysis. An autoregressive of first-order correlation structure was applied.
Relative risk estimates were obtained from the GEE model, and CIs were calculated using a robust covariance estimator for the estimated effects.
RESULTS
We identified 1.3 million Veterans with documented CPT-coded influenza vaccinations during the 2015-2016 influenza season. After applying exclusion criteria, the study population before matching consisted of 104,965 and 125,776 Veterans in the SD and HD cohorts, respectively (Figure 2). The HD cohort was slightly older than the SD cohort, respectively averaging 73.9 and 72.7 years of age (Table 1), whereas sex and race were equally distributed between the cohorts. The SD recipients were vaccinated in facilities with an average HD proportion of 27%, whereas vaccinations of HD recipients occurred in facilities with an average HD proportion of 79%. Because facilities with high and low HD proportions are not evenly distributed across the country, we found substantial differences in the prevalence of HD versus SD by region. In VA Region 9 (AZ, CA, GU, HI, NV), for instance, the majority of Veterans received the HD vaccine (24,495 HD versus 6,860 SD recipients); in Region 4 (southeastern United States), the majority received the SD (3,696 HD versus 11,916 SD recipients). However, this imbalance was ameliorated through matching.
During the baseline period we observed a higher rate of hospitalization where influenza or pneumonia was present in the patient record for those eventually receiving HD than for those eventually receiving SD (RR of 1.29 [95% CI, 1.12-1.48]; Table 2). In contrast, during the observation period, rates of hospitalization for influenza or pneumonia were nearly identical between the two groups, resulting in a RR of 0.99 (95% CI, 0.95-1.03). The PERR-adjusted rVE estimate of HD against influenza- or pneumonia-associated hospitalization in the unmatched cohort was 23% (95% CI, 9%-35%).
After matching, the study population comprised 49,091 and 24,682 Veterans for the SD and HD cohorts, respectively, which as expected were more similar to one another than were the unmatched cohorts for the majority of baseline characteristics (Table 1). We subsequently found a marginally higher rVE estimate against influenza- or pneumonia-associated hospitalization of 25% (95% CI, 2%-43%; Table 2).
Chapter 2
e 1 . B as el in e c ha ra ct er ist ic s b ef or e a nd a ft er m at ch in g Bef or e M at ch ing A fte r M at ch in g II V 3-SD a II V 3-H D b SM D c II V 3-SD II V 3-H D SM D c No . (%) No . (%) No . (%) No . (%) ud y p op ul at io n 10 4, 96 5 12 5, 776 49 ,0 91 24 ,6 82 m al e 2, 255 (2 .1% ) 2, 531 (2 .0% ) -1 .0 1,10 7 (2 .2 %) 463 (1 .8 % ) -2 .7 al e 10 2, 710 (9 7%) 12 3, 24 5 (9 7%) 1. 0 47, 98 4 (9 7%) 24 ,21 9 (9 8%) 2.7 e now n R ac e 7,5 78 (7. 2% ) 10 ,10 0 (8 .0% ) 3.1 3, 46 4 (7 .0% ) 1, 76 3 (7. 1% ) 0. 3 hi te 79 ,17 6 (7 5%) 95, 55 9 (7 5%) 1. 3 37, 10 9 (7 5%) 19, 26 8 (7 8%) 5. 9 ic an -A m er ic an 14 ,7 24 (1 4% ) 14 ,4 76 (1 1% ) -7. 6 6,8 46 (1 3% ) 2, 851 (1 1% ) -7. 2 isp an ic 3, 48 7 (3 .3 %) 5, 64 1 (4 .4 %) 6.0 1, 672 (3 .4 %) 80 0 (3 .2 %) -0 .9 ge g ro up -74 72 ,6 14 (6 9%) 78 ,8 21 (6 2%) -14 32 ,6 65 (6 6%) 13 ,5 97 (5 5%) -2 4 84 23 ,52 5 (2 2%) 30 ,5 11 (2 4%) 4.4 11 ,4 61 (2 3%) 6, 210 (2 5%) 4.2 + 8, 82 6 (8 .4 %) 16 ,444 (1 3% ) 15 4, 965 (1 0% ) 4, 87 5 (1 9% ) 27 H S r eg ion gio n 1 : C T, M E, M A , N H , R I, V T 5, 74 6 (5 .4 %) 2, 567 (2 .0% ) -1 8 3, 85 0 (7. 8% ) 1, 83 8 (7. 4% ) -1 gio n 2 : N J, N Y, P R, V I 12 ,444 (1 1% ) 14 ,14 4 (1 1% ) -2 8, 33 0 (1 6% ) 3, 833 (1 5% ) -4 gio n 3 : D E, D C , M D , P A , V A , W V 15 ,3 72 (1 4% ) 18 ,1 50 (1 4% ) -1 8, 71 2 (1 7% ) 4, 353 (1 7% ) 0 gio n 4 : A L, F L, G A , K Y, M S, N C , S C , T N 11 ,9 16 (1 1% ) 3, 69 6 (2 .9 %) -3 3 6, 999 (1 4% ) 2, 602 (1 0% ) -1 1 gio n 5 : I L, I N , M I, M N , O H , W I 21 ,74 7 (2 0%) 16 ,8 21 (1 3% ) -2 0 7,5 44 (1 5% ) 4,4 58 (1 8% ) 7 gio n 6 : A R, L A , N M , O K , T X 9,11 6 (8 .6 %) 23 ,23 8 (1 8% ) 29 3, 89 9 (7. 9% ) 2, 457 (9 .9 %) 7
2
Chapter 2 Ta bl e 1 (c ont inu ed ). B as el in e c ha ra ct er ist ic s b ef or e a nd a ft er m at ch in g Be fo re Ma tc hi ng Aft er Ma tc hi ng II V 3-SD a II V 3-H D b SM D c II V 3-SD II V 3-H D SM D c No . (%) No . (%) No . (%) No . (%) Re gio n 7 : I A , K S, M O , N E 14 ,6 19 (1 3% ) 12 ,2 28 (9 .7 %) -1 3 4, 78 0 (9 .7 %) 2,3 09 (9 .3 %) -1 Re gio n 8 : C O , M T, N D , S D , U T, W Y 2, 553 (2 .4 %) 10 ,0 98 (8 .0% ) 25 2, 02 3 (4 .1% ) 1, 614 (6 .5 %) 11 Re gio n 9 : A Z, C A , G U , H I, N V 6, 86 0 (6 .5 %) 24 ,4 95 (1 9% ) 39 1, 67 5 (3 .4 %) 879 (3 .5 %) 1 Re gio n 1 0: A K , I D , O R, W A 4, 59 2 (4 .3 %) 33 9 (0 .2 %) -2 8 1, 27 9 (2 .6 %) 33 9 (1 .3 % ) -9 H ig h-ri sk d is or der s C hr on ic c ar di ac d ise as e 27, 91 1 (2 6%) 34 ,533 (2 7%) 1. 9 13 ,0 04 (2 6%) 6, 853 (2 7%) 2. 9 Im m uno sup pre ss iv e d iso rd er s 4, 52 9 (4 .3 %) 5, 83 1 (4 .6 %) 1. 6 2, 09 3 (4.2 % ) 1,16 7 (4 .7 %) 2.2 D ia be te s m el lit us 35 ,0 87 (3 3%) 41, 00 8 (3 2%) -1 .8 16 ,2 42 (3 3%) 7,5 06 (3 0%) -5.7 N eu rol og ic al /m us cu lo sk el et al 4, 772 (4 .5 %) 6,0 48 (4 .8 %) 1. 2 2,3 48 (4 .7 %) 1, 57 7 (6 .3 %) 7.0 C hr on ic re na l d ise as e 6, 58 6 (6 .2 %) 7,7 27 (6 .1% ) -0. 5 3, 14 5 (6 .4 %) 1, 601 (6 .4 %) 0. 3 M al ig na nc ie s 14 ,7 29 (1 4% ) 17, 39 8 (1 3% ) -0. 6 6, 865 (1 3% ) 3, 419 (1 3% ) -0 .4 C hr on ic pu lm on ar y 13 ,4 55 (12% ) 16 ,9 45 (1 3% ) 1. 9 6, 411 (1 3% ) 3, 36 5 (1 3% ) 1. 7 Li ve r d ise as es 1, 46 0 (1 .3 % ) 1, 52 9 (1 .2 % ) -1 .5 687 (1 .3 % ) 265 (1 .0% ) -2 .9 O th er m et ab ol ic a nd i m m un ity d iso rd er s 53 5 (0 .5 %) 72 4 (0 .5 %) 0. 9 267 (0 .5 %) 12 9 (0 .5 %) -0. 3 H em og lo bi nop at hi es 28 0 (0 .2 %) 381 (0 .3 %) 0.7 131 (0 .2 %) 78 (0 .3 %) 0. 9 A t le as t t w o h ig h-ri sk d iso rd er s 28 ,5 04 (2 7%) 34 ,38 4 (2 7%) 0.4 13 ,3 04 (2 7%) 6, 686 (2 7%) 0. 0 Fr ai lty i nd ic ato rs / n on -I CD f ac to rs N ur sin g h om e o r re ha b f ac ili ty re sid en t at tim e o f v ac ci nat io n 56 9 (0 .5 %) 1, 06 2 (0 .8 %) 3. 6 33 9 (0 .6 %) 562 (2 .2 %) 13 .1
e 1 (c ont inu ed ). B as el in e c ha ra ct er ist ic s b ef or e a nd a ft er m at ch in g Be fo re Ma tc hi ng Aft er Ma tc hi ng II V 3-SD a II V 3-H D b SM D c II V 3-SD II V 3-H D SM D c No . (%) No . (%) No . (%) No . (%) itt ed at le as t o nc e t o n ur sin g h om e o r ha b f ac ili ty 1, 52 8 (1 .4 % ) 2, 091 (1 .6 % ) 1. 7 770 (1. 5% ) 716 (2 .9 %) 9. 0 actu re 2,3 81 (2 .2 %) 2, 76 7 (2 .1% ) -0. 5 1,1 28 (2 .2 %) 59 3 (2 .4 %) 0.7 cc in at io n re co rd f ou nd i n p re vio us sp ir at or y s ea so n 81 ,8 58 (7 7%) 96 ,82 2 (7 6%) -2 .4 38 ,17 8 (7 7%) 18 ,3 24 (74 % ) -8. 3 FR e u nd er 4 5 6, 39 2 (6 .0% ) 8, 111 (6 .4 %) 1. 5 2, 88 0 (5 .8 %) 1, 64 8 (6 .6 %) 3. 3 em og lo bi n A1 c f ove r 9 2, 02 0 (1 .9 % ) 2, 16 5 (1 .7 % ) -1 .5 843 (1 .7 % ) 444 (1 .7 % ) 0. 6 nti nu ou s v ar iab le s d 72 .7 (7. 11 ) 73. 9 (7. 82 ) 17 73 .1 (7. 44 ) 75 .5 (8 .5 6) 29 um be r o f p ne um on ia h os pi ta liz at io ns 0. 01 (0 .13 ) 0. 02 (0 .14 ) 1. 5 0. 02 (0 .14 ) 0. 02 (0 .16 ) 2.5 FR e 69. 6 (2 3. 1) 70 .4 (2 2. 6) 3. 6 68. 8 (2 3. 2) 69. 0 (2 3. 5) 0. 5 em og lo bi n A1 c f 6. 63 (1 .3 0) 6. 69 (1 .3 8) 4. 3 6. 62 (1 .3 0) 6. 65 (1. 33 ) 2.4 H A C A N g sc or e 0. 16 (0 .16 ) 0. 17 (0 .17 ) 1. 9 0. 17 (0 .17 ) 0. 17 (0 .17 ) 1. 9 V 3-SD : I na ct iv at ed i nfl ue nz a v ac ci ne , t riv al en t, s ta nd ar d-do se V 3-H D : I na ct iv at ed i nfl ue nz a v ac ci ne , t riv al en t, h ig h-do se ta nd ar di ze d m ea n d iffe re nc e ( SM D ) l es s t ha n 1 0 i n a bs ol ut e v al ue s ug ge st s n o i m po rt an t d iffe re nc e b et w ee n t he t wo c oh or ts [ 27 ] tin uou s v ar ia bl es r ep or te d a s m ea n ( st an da rd d ev ia tio n) R : e st im at ed g lo m er ul ar fi ltr at io n r at e ( ki dn ey f un ct io n) di ca to r o f d ia be te s N : C ar e A ss es sm en t N ee d
2
Chapter 2 Ta bl e 2 . R el at iv e v ac ci ne e ffe ct iv en es s ( rV E ) o f I IV 3-H D f or h os pi ta liz at io ns a nd out pa tie nt v isi ts c w ith u nd er ly in g i nfl ue nz a o r p ne um on ia O ut co m e Ba se line P er io d O bs er va tio n P er io d rV E-H D (9 5% C I) II V3 -S D a II V3 -H D b Re la tive Ri sk HD II V3 -S D a II V3 -H D b Re la tive Ri sk H D No . Pe rs on -we ek s Rat e d No . Pe rs on -we ek s Rat e No . Pe rs on -we ek s Rat e No . Pe rs on -we ek s Rat e Bef or e m at ch ing H osp ita liz at io ns , Infl ue nza / Pn eu m on ia 32 8 1,7 85, 14 8 1. 84 52 2 2,2 10 ,2 53 2. 36 1. 29 981 3, 42 7,1 80 2. 86 1,14 1 4, 02 5,9 35 2. 83 0. 99 23% (9% - 3 5% ) H osp ita liz at io ns , A ll C au se 4, 76 0 1,74 8, 65 9 27. 2 5, 863 2, 16 5, 54 2 27. 1 0. 99 8, 78 3 3, 28 7,5 36 26 .7 10 ,2 88 3, 86 6,9 07 26 .6 1 0% (-5% - 5 % ) PC P/ED /U rg en t C are V isi ts c, Infl ue nza / Pn eu m on ia 53 4 1,7 82 ,81 8 3 721 2, 20 8,0 60 3. 3 1. 09 1,4 64 3, 41 7,7 09 4. 3 1,7 52 4, 01 3, 86 1 4.4 1. 02 7% (-7% - 1 8% ) PC P/ED /U rg en t C are V isi ts c, A ll C au se 49, 65 4 1, 21 3, 67 9 409 59 ,017 1, 50 7,86 3 391 0. 96 81 ,9 38 1,7 11, 46 7 47 9 94 ,15 5 2,0 99 ,01 1 449 0. 94 2% (1% - 4 % ) A fte r m at ch in g H osp ita liz at io ns , Infl ue nza / Pn eu m on ia 17 3 82 4, 251 2.1 13 6 42 0, 29 2 3. 24 1. 54 480 1, 61 2, 73 3 2. 98 27 5 79 7,4 75 3. 45 1.16 25 % (2 % - 4 3% ) H osp ita liz at io ns , A ll C au se 2,3 24 80 6,9 07 28 .8 1, 32 8 410 ,9 75 32 .3 1.1 2 4, 237 1, 54 5, 70 5 27. 4 2, 19 5 76 3, 55 4 28 .7 1. 05 7% (-2% -1 4% )
e 2 ( co nt in ue d) . R el at iv e v ac ci ne e ffe ct iv en es s ( rV E ) o f I IV 3-H D f or h os pi ta liz at io ns a nd out pa tie nt v isi ts c w ith u nd er ly in g i nfl ue nz a o r p ne um on ia ut co m e Ba se line P er io d O bs er va tio n P er io d rV E-H D (9 5% C I) II V3 -S D a II V3 -H D b Re la tive Ri sk HD II V3 -S D a II V3 -H D b Re la tive Ri sk H D No . Pe rs on -we ek s Rat e d No . Pe rs on -we ek s Rat e No . Pe rs on -we ek s Rat e No . Pe rs on -we ek s Rat e P/ED /U rg en t isi ts c, ue nza / eu m on ia 26 0 823 ,23 2 3. 2 16 1 42 0, 05 0 3. 8 1. 21 735 1, 60 7,7 35 4.6 37 9 79 5, 62 7 4.8 1. 04 14 % (-8 % - 3 2% ) P/ED /U rg en t isi ts c, A ll se 23 ,2 09 56 2, 291 413 11 ,12 2 293, 44 9 37 9 0. 92 38 ,4 27 802 ,9 08 47 9 17, 97 2 43 0, 558 417 0. 87 5% (2% - 8 % ) V 3-SD : I na ct iv at ed i nfl ue nz a v ac ci ne , t riv al en t, s ta nd ar d-do se V 3-H D : I na ct iv at ed i nfl ue nz a v ac ci ne , t riv al en t, h ig h-do se pa tie nt V isi ts i nc lu de p rim ar y c ar e ph ys ic ia n ( PC P) , e m er ge nc y d ep ar tm en t ( E D ) a nd u rg en t c ar e v isi ts ut co m es p er 1 0, 00 0 p er so n-w ee ks
2
Chapter 2
For our secondary outcomes, the PERR-adjusted rVE estimate of HD against influenza- or pneumonia-associated outpatient visits in the unmatched cohort was 7% (95% CI, -7% to 18%); matching resulted in an adjusted estimate of 14% (95% CI, -8% to 32%). In the unmatched cohort, the PERR-adjusted rVE for all-cause hospitalizations and for all-cause outpatient visits was 0% (95% CI, -5% to 5%) and 2% (95% CI, 1%-4%), respectively. The PERR-adjusted rVEs for all-cause outcomes in the matched cohort were slightly higher at 7% (95% CI, -2% to 14%) against hospitalizations and 5% (95% CI, 2%-8%) against outpatient visits.
During the observation period, the influenza diagnostic test rate in SD recipients was 3.92 per 10,000 person-weeks (120 were positive and 1,221 negative), and 3.54 per 10,000 person-weeks in the HD recipients (99 positive and 1,322 negative; Table 3). Laboratory-confirmed influenza rates were 0.35 versus 0.24 per 10,000 person-weeks in the SD and HD cohort, respectively, resulting in an rVE of 30% (95% CI, 8%-47%). Matching resulted in a higher test rate in HD than in SD recipients (4.39 vs 3.95 per 100,000 person-weeks, respectively) and in an rVE of 38% (95% CI, -5% to 65%). Laboratory-confirmed influenza rate was nil in the baseline period of the study and as such we did not apply PERR adjustments to the rVE estimates against this outcome.
Sensitivity Analysis Results
Using the GEE method, we found an overall rVE of 29% (95% CI: 14-41%) against influenza- or pneumonia-associated hospitalization. For age groups 65 to74, 75 to 84, and 85 or older, the respective rVEs were as follows: 23% (95% CI: 10-33%), 26% (95% CI: 8-40%), and 30% (95% CI: 15-42%) against influenza- or pneumonia-associated hospitalization.
Table 3. Relative vaccine effectiveness (rVE) of IIV3-HD for laboratory-confirmed influenza Outcome Observation Period IIV3-SDa IIV3-HDb rVE-HD (95% CI) No. Person-weeks Ratec No.
Person-weeks Rate
Before matching
Laboratory Test Rate for
Influenza 1,341 3,419,881 3.92 1,421 4,018,513 3.54 Laboratory-confirmed
Influenza 120 3,440,741 0.35 99 4,041,063 0.24 30%,(8% - 47%)
After matching
Laboratory Test Rate for
Influenza 635 1,619,368 3.95 349 795,727 4.39 Laboratory-confirmed
Influenza 62 1,619,368 0.38 19 801,439 0.24 38%, (-5% - 65%)
a IIV3-SD: Inactivated influenza vaccine, trivalent, standard-dose b IIV3-HD: Inactivated influenza vaccine, trivalent, high-dose c Outcomes per 10,000 person weeks
DISCUSSION
Our main result, a matched and PERR-adjusted HD vaccine rVE of 25% (95% CI, 2%-43%) against influenza- or pneumonia-associated hospitalization, is consistent with the rVE of 40% (95% CI, 20%-45%) against pneumonia admission found from the RCT reported by DiazGranados et al [8]. In addition, we note consistency between our results and those observed when comparing rVE estimates against both our broadest outcome category and our most specific outcome: for all-cause admission, our estimate of 7% (95% CI, -2% to 14%) is consistent with the RCT rVE of 8% (95% CI, 2%-12%), and our rVE of 38% (95% CI, -5 to 65%) against laboratory-confirmed influenza is consistent with the RCT estimate of 24.2% (95% CI, 9.7%-36.5%) against laboratory-confirmed influenza-like illness. Our rVE against laboratory-confirmed influenza also aligns with the result reported by Izurieta et al [10] for two comparable outcomes: rapid response influenza test followed by oseltamivir treatment (22%; 95% CI, 15%-29%) and influenza-attributed hospitalizations or emergency department visits (22%; 95% CI, 16%-27%). Although they were not the focus of Shay et al’s [22] evaluation of the comparative effectiveness of high-dose vaccine in preventing post-influenza deaths during the 2012–13 and 2013–14 seasons, rVE estimates for
Chapter 2
hospitalized influenza and influenza-related illness were reported to be 18.8% (95% CI, 14.4%–23%) and 14.2% (95% CI, 8.5%–19.4%), respectively, which are in accord with our findings. Reporting from the first cluster-randomized, prospective study to investigate the relative effectiveness of HD vaccine in an elderly nursing home population, Gravenstein et al [23] found HD vaccine to be more effective than SD vaccine in reducing both respiratory illness (12.7%, 95% CI, 1.8%-22.4%) and all-cause hospital admissions (6.7%, 95% CI, 1.5%-11.6%) in the 2013-14 season, also consistent with our findings.
Richardson et al [11] studied a cohort of senior Veterans, similar to ours. In contrast, for influenza-related and pneumonia-attributed hospitalizations, they reported a relative efficacy of only 2% (95% CI, -40% to 32%). Several methodological differences may explain our divergent findings. First and foremost is our use of the PERR method: we specifically selected this instrument in order to adjust for confounding by indication induced by unmeasured variables, a factor Richardson and colleagues did not fully address by including influenza or pneumonia hospitalization before the influenza season as an adjustment variable in their multivariate Poisson regression model. Furthermore, Richardson et al’s [11] use of stratification by propensity score quintile to adjust for measured confounders is problematic given the low frequency of HD vaccine usage in the population across the study period [24]. Stratifying in this way may have little impact on adjusting for measured confounders when the exposure to study treatment is rare and the quintiles are based on the entire cohort rather than the exposed subcohort, as demonstrated recently by Desai et al [25]. Moreover, Richardson et al [11] did not address bias introduced due to temporal and geographic factors associated with both vaccine access and influenza exposure.
Our analysis has several limitations. First, we could not apply PERR adjustment to our most specific outcome (laboratory-confirmed influenza) because the outcome was not observed in the baseline period. This lack of adjustment could result in an underestimate of rVE of HD for two reasons: (1) the US Centers for Disease Control and Prevention reported a close match between the circulating influenza strains and the strains included in the vaccine for the 2015-2016 season; and (2), DiazGranados et al [8] reported in their RCT an rVE of 51% against laboratory-confirmed influenza strains similar to vaccine strains [15]. Taken together, these facts imply that without a PERR adjustment to this highly specific outcome, an rVE of 38% is likely an underestimate.
Because we were interested in the impact on the overall health service utilization (e.g., all-cause hospitalizations and all-cause outpatient visits) during the entire influenza season, we did not restrict our analyses to periods of peak influenza activities as some past studies have done. One consequence of using the broader period is bias towards the null arising from misclassification of study outcomes [26]. The differences among our estimates of rVE based on the more specific outcome, i.e. laboratory-confirmed influenza, and our less specific outcomes such as all-cause outpatient visits, suggest such a bias.
Second, the matched PERR analysis is appealing because PERR is a straightforward adaptation of the difference-in-differences method and matching is effective in control for confounding. However, this primary analysis neglects both a potential clustering effect within facilities and correlations between repeated observations for the same patient. Thus, a sensitivity analysis using the GEE method was also performed. Nonetheless, because treatment allocation was not randomized in our study, the potential for confounding by indication and by unmeasured variables cannot be entirely eliminated. Third, VHA patients also use healthcare services outside the VHA, and our study captured and utilized data only from their VHA encounters. Because the observation period immediately followed the baseline period and we excluded “snowbirds,” it seems unlikely that many patients could have significantly changed their healthcare behavior, in terms of seeking care outside the VA, from one period to the next. And the similarity between the matched SD and the HD recipients, as demonstrated in Table 2, also suggested that a meaningful difference between the two groups in terms of seeking care outside the VA was unlikely. Consequently, it is plausible that the PERR method produces low-bias estimates because the missing data regarding healthcare outside the VHA is unlikely to be differential between periods or between groups. Finally, the generalizability of our findings should be interpreted in the context of our source population, which is comprised of patients treated at VHA facilities, and, therefore, may have limited applicability to seniors in the U.S. general population. Moreover, as demonstrated by Shay et al [22], different results (e.g. strength of association) might be observed in years when the relatedness of vaccine and circulating strains differed materially. Of the 5 seasons that have been studied (2010-11 to 2013-14, and ours, 2015-2016), HD’s higher relative effectiveness was observed in all but one – the 2010-2011 season.
Chapter 2
CONCLUSION
In protecting senior VHA patients against influenza- or pneumonia-associated hospitalization, HD vaccine was more effective than SD vaccine. This relative effectiveness was demonstrated using EMRs that included hospitalizations and laboratory results for a vulnerable population receiving healthcare in a real-world setting. Because our study was conducted in a different influenza season than those studies and trials mentioned above, our results may have been impacted by the severity of influenza seasons, virus type and subtype, and the match between the vaccine and circulating influenza strains. We plan to investigate the effectiveness of HD vaccine in a more timely fashion (e.g., within three months from the end of a season) as we automate our data extraction processes and standardize our methodology.
ACKNOWLEDGEMENTS
We thank the Centers for Disease Control and Prevention for providing us with influenza surveillance data and the Public Health Surveillance and Research Group, VHA Office of Quality, Safety and Value for providing us with VA influenza laboratory test data. We also thank Carlos A. DiazGranados of Sanofi Pasteur for his guidance and advice.
REFERENCES
1. Centers for Disease Control and Prevention. Prevention and control of seasonal influenza with vaccines. Recommendations of the Advisory Committee on Immunization Practices--United States, 2013-2014. MMWR Recomm Rep. 2013;62(RR-07):1-43.
2. Zhou H, Thompson WW, Viboud CG, Ringholz CM, Cheng PY, Steiner C, Abedi GR, Anderson LJ, Brammer L, Shay DK. Hospitalizations associated with influenza and respiratory syncytial virus in the United States, 1993-2008. Clin Infect Dis. 2012;54(10):1427-36.
3. Mullooly JP, Bridges CB, Thompson WW, Chen J, Weintraub E, Jackson LA, Black S, Shay DK, Vaccine Safety Datalink Adult Working G. Influenza- and RSV-associated hospitalizations among adults. Vaccine. 2007;25(5):846-55.
4. Thompson WW, Moore MR, Weintraub E, Cheng PY, Jin X, Bridges CB, Bresee JS, Shay DK. Estimating influenza-associated deaths in the United States. Am J Public Health. 2009;99 Suppl 2:S225-30.
5. Thompson WW, Shay DK, Weintraub E, Brammer L, Bridges CB, Cox NJ, Fukuda K. Influenza-associated hospitalizations in the United States. Jama. 2004;292(11):1333-40.
6. Thompson WW, Shay DK, Weintraub E, Brammer L, Cox N, Anderson LJ, Fukuda K. Mortality associated with influenza and respiratory syncytial virus in the United States. JAMA. 2003;289(2):179-86.
7. Thompson WW, Weintraub E, Dhankhar P, Cheng PY, Brammer L, Meltzer MI, Bresee JS, Shay DK. Estimates of US influenza-associated deaths made using four different methods. Influenza Other Respir Viruses. 2009;3(1):37-49.
8. DiazGranados CA, Dunning AJ, Kimmel M, Kirby D, Treanor J, Collins A, Pollak R, Christoff J, Earl J, Landolfi V, et al. Efficacy of high-dose versus standard-dose influenza vaccine in older adults. N Engl J Med. 2014;371(7):635-45.
9. DiazGranados CA, Robertson CA, Talbot HK, Landolfi V, Dunning AJ, Greenberg DP. Prevention of serious events in adults 65 years of age or older: A comparison between high-dose and standard-dose inactivated influenza vaccines. Vaccine. 2015;33(38):4988-93.
10. Izurieta HS, Thadani N, Shay DK, Lu Y, Maurer A, Foppa IM, Franks R, Pratt D, Forshee RA, MaCurdy T, et al. Comparative effectiveness of high-dose versus standard-dose influenza vaccines in US residents aged 65 years and older from 2012 to 2013 using Medicare data: a retrospective cohort analysis. Lancet Infect Dis. 2015;15(3):293-300.
11. Richardson DM, Medvedeva EL, Roberts CB, Linkin DR, Centers for Disease C, Prevention Epicenter P. Comparative effectiveness of high-dose versus standard-dose influenza vaccination in community-dwelling veterans. Clin Infect Dis. 2015;61(2):171-6.
12. Collard RM, Boter H, Schoevers RA, Oude Voshaar RC. Prevalence of frailty in community‐dwelling older persons: A systematic review. Journal of the American Geriatrics Society. 2012;60(8):1487-92. 13. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, Seeman T, Tracy R, Kop WJ,
Burke G. Frailty in older adults evidence for a phenotype. The Journals of Gerontolog y Series A: Biological Sciences and Medical Sciences. 2001;56(3):M146-M57.
14. Wang L, Porter B, Maynard C, Evans G, Bryson C, Sun H, Gupta I, Lowy E, McDonell M, Frisbee K, et al. Predicting risk of hospitalization or death among patients receiving primary care in the
Chapter 2
15. Lin NX, Henley WE. Prior event rate ratio adjustment for hidden confounding in observational studies of treatment effectiveness: a pairwise Cox likelihood approach. Stat Med. 2016;35(28):5149-69. 16. Tannen RL, Weiner MG, Xie D. Replicated studies of two randomized trials of angiotensin‐converting
enzyme inhibitors: further empiric validation of the ‘prior event rate ratio’to adjust for unmeasured confounding by indication. Pharmacoepidemiolog y and Drug Safety. 2008;17(7):671-85.
17. Tannen RL, Weiner MG, Xie D. Use of primary care electronic medical record database in drug efficacy research on cardiovascular outcomes: comparison of database and randomised controlled trial findings. BMJ. 2009;338:b81.
18. Weiner MG, Xie D, Tannen RL. Replication of the Scandinavian Simvastatin Survival Study using a primary care medical record database prompted exploration of a new method to address unmeasured confounding. Pharmacoepidemiol Drug Saf. 2008;17(7):661-70.
19. Yu M, Xie D, Wang X, Weiner MG, Tannen RL. Prior event rate ratio adjustment: numerical studies of a statistical method to address unrecognized confounding in observational studies. Pharmacoepidemiol Drug Saf. 2012;21 Suppl 2(S2):60-8.
20. Angrist JD, Pischke J-S. Mastering’metrics: The path from cause to effect: Princeton University Press; 2014. 21. Diggle P. Analysis of longitudinal data: Oxford University Press; 2002.
22. Shay DK, Chillarige Y, Kelman J, Forshee RA, Foppa IM, Wernecke M, Lu Y, Ferdinands JM, Iyengar A, Fry AM, et al. Comparative Effectiveness of High-Dose Versus Standard-Dose Influenza Vaccines Among US Medicare Beneficiaries in Preventing Postinfluenza Deaths During 2012-2013 and 2013-2014. J Infect Dis. 2017;215(4):510-7.
23. Gravenstein S, Davidson HE, Taljaard M, Ogarek J, Gozalo P, Han L, Mor V. Comparative effectiveness of high-dose versus standard-dose influenza vaccination on numbers of US nursing home residents admitted to hospital: a cluster-randomised trial. The Lancet Respiratory Medicine. 2017;5(9):738-46.
24. Decker MD, DiazGranados CA, Chit A, Hosbach P, Robertson CA, Greenberg DP. Regarding primary care patients who received influenza vaccine at veteran health administration medical centers. Clinical Infectious Diseases. 2015;61(8):1344-5.
25. Desai RJ, Rothman KJ, Bateman BT, Hernandez-Diaz S, Huybrechts KF. A Propensity-score-based Fine Stratification Approach for Confounding Adjustment When Exposure Is Infrequent. Epidemiolog y. 2017;28(2):249-57.
26. Lachenbruch PA. Sensitivity, specificity, and vaccine efficacy. Control Clin Trials. 1998;19(6):569-74. 27. Borenstein M, Hedges LV, Higgins JP, Rothstein HR. Introduction to meta-analysis: John Wiley & Sons;
