Individual and joint effects of influenza-like illness and vaccinations on stroke in the young: A case-control study

Abstract

Background:

Influenza-like illness (ILI) is an acute trigger for stroke, though joint effects of vaccinations and ILI have not yet been explored.

Methods:

Data for our case-control study was obtained from MarketScan Commercial Claims and Encounters between 2008 and 2014. Patients 18–65 years old who experienced a stroke were matched on age and admission date to a control, defined as patients with head trauma or ankle sprain at an inpatient or emergency department visit. Exposures were ILI in the prior 30 days, and any type of vaccination during the year prior. Our outcome was ischemic and intracerebral hemorrhagic strokes identified using ICD-9 codes. Logistic regression models estimated adjusted odds ratios (aOR) controlling for preventive care visits, diabetes, valvular heart disease, smoking, alcohol abuse, obesity, and hypertension.

Results:

We identified and matched 24,103 cases 18–44 years old and 141,811 45–65 years old. Those aged 18–44 years had increased stroke risk 30 days after ILI [aOR:1.68 (95% CI:1.51–1.86)] and reduced risk with any vaccination in the year prior [aOR:0.92 (95% CI:0.87–0.99)]. Joint effects indicate that ILI was associated with increased stroke risk among those with [aOR:1.41 (95% CI:1.08–1.85)] and without [aOR:1.73 (95% CI:1.55–1.94)] vaccinations in the prior year (P_interaction=0.16). Among those ages 45–65 years, adjusted analyses indicate increased stroke risk for those with ILI [aOR:1.32 (95% CI:1.26–1.38)], though there was no effect of vaccinations [aOR:1.00 (95% CI:0.97–1.02)]. Joint effects indicate that ILI was not associated with stroke among those with any vaccination [aOR:1.07 (95% CI:0.96–1.18) but was associated with increased risk among those without vaccinations [aOR:1.39 (95% CI:1.32–1.47)] (P_interaction<0.001).

Conclusions:

ILI was associated with increased stroke risk in the young and middle-aged, while vaccinations of any type were associated with decreased risk among the young. Joint effects of ILI and vaccinations indicate vaccinations can reduce the effect of ILI on stroke.

Graphical Abstract

Introduction

The incidence and prevalence of stroke among the young are increasing in the US, with approximately 10–14% of ischemic strokes occurring in people age 18–45.¹ Further, this age group demonstrates greater heterogeneity in stroke etiology compared to the older stroke population, suggesting risk reduction efforts that apply to older populations may not be as beneficial in younger populations. As such, young stroke patients present a unique population where identifying novel modifiable risk factors is a high priority.

Infections have been identified as both a potential chronic risk factor and acute trigger for stroke.^2–7 Respiratory tract infections are the most common cause of infection in adults, and influenza-like illness (ILI) accounts for the majority of these infections.⁸ Though not a clinical diagnosis, ILI is used to identify a group of respiratory infection diagnoses regularly used for surveillance by the Department of Defense and Center of Disease Control and Prevention (CDC). Respiratory tract infections and ILI have been associated with short-term stroke risk in several studies, particularly in people aged 18–45 years.^{2, 7, 9–11} Vaccinating against infections, including influenza, reduces the risk of stroke.¹²

We aimed to evaluate the relationship between ILI, any type of vaccination, and the combined effects of ILI and vaccinations on stroke risk in young and middle-aged populations using a case-control study of an insurance-based administrative dataset. We hypothesized that ILI will be associated with an increase in odds of stroke, vaccines will be associated with a decrease in odds of stroke, and people who experience ILI in the absence of vaccines will have the highest odds of stroke, with vaccines attenuating the risk of stroke in those with ILI. Further, we hypothesize these associations will be stronger in the people aged 18–44 years than in those 45 and older.

Methods

Study Population & Study Design

MarketScan Commercial Claims and Encounters is an administrative database containing longitudinal data for patients with specific employee-sponsored insurance programs. MarketScan includes nearly 273 million de-identified patients with data on patient demographics, International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes, and ICD-9 procedure codes for all inpatient and outpatient visits.¹³ Data for each individual is linked by a de-identified patient identifier code allowing for tracking patients over time through multiple inpatient and outpatient admissions. Within MarketScan, claims data for each discharge, inpatient and outpatient, is collected, de-identified, standardized, and then made available to researchers. Data elements include demographic information such as age, sex, and location. For each admission discharge, data available included diagnosis codes, month of discharge, length of stay in hospital, and procedure codes. Due to restrictions in data licensing, we were limited to examining data between the of 2008 and 2014. Data is available through a purchase agreement with MarketScan.

We conducted a matched case-control study within the MarketScan database in which cases were defined as patients who had a stroke between January 1^st, 2008 and December 31^st, 2014 while enrolled in MarketScan and were 18–65 years of age. Healthcare utilization among this stroke population has been previously examined.¹⁴ For patients who had multiple strokes during this study period, only their first stroke was used. Strokes were defined as ischemic or intracerebral hemorrhagic identified using ICD-9 codes (Table S1). Subarachnoid hemorrhagic strokes were excluded. Controls were selected from patients 18–65 years of age presenting to the hospital with an injury (specifically head trauma or an ankle sprain) identified using ICD-9 codes (Table S1) during an inpatient or emergency department visit who had no indication of a stroke at time of injury. Head trauma and ankle sprain patients were selected as controls because these conditions are likely independent of our exposures. Controls were randomly matched to cases 1:1 on admission date (+/− 14 days) and age (+/− 3 years). Though head trauma patients may be at increased risk of stroke after the trauma event, time after trauma is not part of our study period, and, thus, would not affect our study. We categorized patients by their age at first stroke or age at head trauma/ankle sprain visit as 18–44 (young population) or 45–65 (middle-aged population) years.

Exposure Measures and Covariates

Our study assessed two exposures:1) ILI at or prior to the time of stroke or control (index) visit, and 2) any vaccination in the year prior to index visit. The occurrence of ILI was assessed using previously validated ICD-9 codes associated with the stroke hospitalization or control visit (Table S1) or preceding inpatient, outpatient, or emergency department visits.¹⁵ Our main analysis evaluated the exposure of ILI during 1–30 days prior to index visit. In subsequent analyses, additional time frames were examined, including at time of index visit, 1–15 days prior, and 1–60 days prior. Any prior vaccination was assessed using MarketScan procedure group codes associated with outpatient or emergency department visits occurring within one year prior to index visit. Due to limitations in the availability of data on specific vaccination type, we were unable to identify vaccinations for influenza specifically; therefore, our study examined vaccinations for any illness. However, the majority were likely for influenza given they are administered annually. MarketScan procedure group codes are groups of related outpatient procedures, based on Healthcare Common Procedure Coding System, Current Procedural Terminology, 4th Edition, or ICD-9-CM procedure codes. Procedure group codes used for categorization of outpatient visits are presented in Table S1.

We examined demographics and medical history at the time of index visit, as well as healthcare utilization in the year prior including total number of visits, days from last visit, and whether they had a preventive care visit. To assess medical history and preventive care visits, we utilized ICD-9 codes and MarketScan procedure group codes, respectively (Table S1). Those without an ICD-9 code were assumed to not have the specified condition. No patients had a missing value for variables determined by ICD-9 codes. Preventive care refers to physical exams, counseling/guidance/risk factor reduction, and ordering of laboratory/diagnostic procedures. Mean number of preventative care visits was included in the analyses as preventative care visits could be an important confounder in the relationship between vaccination and stroke. MarketScan procedure group codes classify all immunizations as vaccinations and not as preventive care.

Statistical Analysis

We performed all analyses first for the entire study sample, and then stratified by age groups (18–44 and 45–65 years). Those with missing age or sex were not included in the analysis. We examined distributions of demographic characteristics and clinical risk factors assessed at time of index visit as mean [standard deviation (SD)] for continuous variables and proportions as n (%) for categorical variables. Mean number of healthcare visits prior to index visit and, of those with at least one inpatient, outpatient, or emergency department visit, mean days from last visit were calculated. The median [interquartile range (IQR)] for the total number of visits prior and days from last visit were also examined.

We calculated odds ratios (ORs) and 95% confidence intervals (95% CIs) using unadjusted logistic regression to test the crude associations among ILI, vaccinations in the year prior and stroke events. We examined the combined effects of ILI and vaccinations using an interaction term. We then examined these relationships using a multivariable logistic regression model (adjusted model) adjusting for having a preventive care visit in the year prior, diabetes, valvular heart disease, smoking, alcohol abuse, obesity, and hypertension. Prior studies have demonstrated sex differences when examining ILI and stroke, therefore in secondary analyses we stratified by sex to examine potential differences.¹⁶ Due to small cell sizes, some stratifications were limited to our main ILI exposure assessed at 1–30 days prior to index visit. An alpha of p<0.05 was used for the main statistical analyses and an alpha of p<0.1 was used for the interaction analyses. E-values were calculated for the association between ILI and stroke and the association between vaccination and stroke in the overall cohort to assess for the minimum strength necessary for a confounder to explain away observed associations.¹⁷

Results

We identified a total of 169,358 patients with stroke who met the case inclusion criteria for our study, consisting of 24,103 patients ages 18–44 and 145,255 ages 45–65 years. A control pool of 1,071,259 eligible patients was identified, including 743,665 patients ages 18–44 and 327,594 ages 45–65. We were able to match 100% of cases to a control for those ages 18–44, resulting in 24,103 cases and 24,103 controls. Of the 145,255 cases ages 45–65 years, we were able to match 97.6%, resulting in 141,811 cases and 141,811 controls. Patient characteristics for the entire cohort, and by age group are presented in table 1. For those 18–44 years old, the median time from infection within 1–30 days prior to index visit was 15.0 days (IQR: 8.0–23.0) for cases and 17.0 days (IQR: 8.0–23.0) for controls. For those 45–65 years old, the median time from infection within 1–30 days to index visit was 12.0 days (IQR: 5.0–21.0) for cases and 12.0 days (IQR: 5.0–21.0) for controls. For those 18–44 years old, the median time from vaccination to index visit was 147.0 days (IQR: 63.0–244.0) for cases and 163.0 days (IQR: 75.0–260.0) for controls. For those 45–65 years old, the median time from vaccination to index visit was 147.0 days (IQR: 64.0–248.0) for cases and 157.0 days (IQR: 69.0–254.0) for controls.

Table 1.

Demographics and medical history assessed at time of stroke (Case) or trauma/ankle sprain visit (Control) by age group and visit type

	Entire Population (Ages 18–65)				Young Population (Ages 18–44)				Middle-Aged Population (Ages 45–65)
	Cases		Controls		Cases		Controls		Cases		Controls
	(N=165,914)		(N=165,914)		(N=24,103)		(N=24,103)		(N=141,811)		(N=141,811)
Age (Mean, SD)	54.02	9.11	51.90	9.14	36.22	6.98	34.09	6.92	57.04	5.08	54.92	5.15
Sex – Female (n, %)	74,095	44.66%	105,155	63.38%	12,853	53.33%	13,696	56.82%	61,242	43.19%	91,459	64.49%
Total # of Visits Prior
Mean, SD	14.98	21.09	13.17	17.38	12.23	19.30	10.24	15.15	15.44	21.35	13.67	17.69
Median, IQR	8.00	3.0–19.0	8.00	3.0–17.0	6.00	2.0–15.0	5.00	2.0–13.0	9.00	3.0–20.0	8.00	3.0–18.0
Days from Last Visit †
Mean, SD	30.73	57.86	50.92	67.78	35.16	63.66	66.69	79.08	30	56.82	48.38	65.42
Median, IQR	6.00	1.0–28.0	23.00	7.0–67.0	6.00	1.0–35.0	34	10.0–94.0	6	2.0–27.0	21	6.0–63.00
Preventative Care Visit	30,489	18.38%	47,081	28.38%	5,193	21.55%	6,812	28.26%	25,296	17.84%	40,269	28.40%
Vaccination Visit	26,821	16.17%	29,578	17.83%	2,661	11.04%	3,181	13.20%	24,160	17.04	26,397	18.61
Medical History (n, %)‡
Diabetes	29,801	17.96%	4,803	2.89%	1,916	7.95%	131	0.54%	27,885	19.66%	4672	3.29%
Hypertension	87,552	52.77%	13,724	8.27%	7,522	31.21%	413	1.71%	80,030	56.43%	13,311	9.39%
Obesity	4,843	2.92%	467	0.28%	807	3.35%	35	0.15%	4,036	2.85%	432	0.30%
Infections	18,048	10.88%	3,072	1.85%	2,689	11.16%	177	0.73%	15,359	10.83%	2,895	2.04%
Coagulopathy	6,369	3.84%	680	0.41%	1,634	6.78%	31	0.13%	4,735	3.34%	649	0.46%
Hypercoagulable state	2,484	1.50%	85	0.05%	928	3.85%	<10	<0.04%	1,556	1.10%	81	0.06%
Migraine	6,962	4.20%	411	0.25%	2,722	11.29%	49	0.20%	4,240	2.99%	362	0.26%
Valvular heart disease	19,597	11.81%	635	0.38%	2,645	10.97%	24	0.10%	16,952	11.95%	611	0.43%
Congenital Heart Disease	132	0.08%	<10	0.00%	61	0.25%	0	0.00%	71	0.05%	<10	0.00%
Patent foramen ovale	5,612	3.38%	20	0.01%	1,716	7.12%	<10	<0.04%	3,896	2.75%	19	0.01%
Alcohol abuse	24,697	14.89%	4,833	2.91%	2,971	12.33%	476	1.97%	21,726	15.32%	4,357	3.07%
Drug Abuse/Dependence	4,371	1.32%	2,172	0.65%	735	3.05%	175	0.73%	3,636	2.56%	2,025	1.43%
Smoking	22,198	13.38%	3,624	2.18%	2,600	10.79%	409	1.70%	19,598	13.82%	3,215	2.27%
Trauma	2,510	1.51%	54,152	32.64%	776	3.22%	5,985	24.83%	1,734	1.22%	48,167	33.97%

^*including inpatient, outpatient and/or emergency department

^†Calculated only for those who had at least one visit

^‡Collected at time of stroke/trauma visit

In the entire cohort, unadjusted and adjusted analyses indicated increased odds of stroke for those with ILI 1–30 days prior to index visit (Table 2) and reduced odds of stroke for those having a vaccination in the year prior. The E-value for the association between ILI and stroke was 2.1 and was 1.5 for the association between vaccination and stroke. Results were consistent when stratified by sex. In analyses of combined effects of ILI and vaccinations, among those with no vaccinations in the prior year, those with ILI had increased odds of stroke (adjusted OR, 1.46 [1.39–1.53]) while for those with vaccination the risk associated with ILI was lower (adjusted OR, 1.11 [1.01–1.22]); adjusted P_interaction<0.001).

Table 2.

Individual and joint effects of ILI and vaccines on stroke risk among those aged 18–65 years

Individual Effects
	Entire Population (Ages 18–65)		Young Population (Ages 18–44)		Middle-Aged Population (Ages 45–65)
	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)
ILI vs. No ILI	1.30 (1.25–1.35)	1.38 (1.32–1.44)	1.66 (1.52–1.82)	1.68 (1.51–1.86)	1.24 (1.19–1.29)	1.32 (1.26–1.38)
ILI vs. No ILI in males	1.47 (1.39–1.56)	1.54 (1.43–1.65)	2.27 (1.94–2.65)	2.26 (1.89–2.66)	1.37 (1.28–1.46)	1.42 (1.32–1.53)
ILI vs. No ILI in females	1.32 (1.26–1.39)	1.38 (1.31–1.46)	1.40 (1.25–1.58)	1.41 (1.24–1.61)	1.30 (1.24–1.37)	1.36 (1.28–1.45)
Vaccination vs. No Vaccination	0.89 (0.87–0.91)	0.96 (0.94–0.98)	0.82 (0.77–0.86)	0.92 (0.87–0.99)	0.90 (0.88–0.92)	1.00 (0.97–1.02)
Vaccination vs. No Vaccination in Males	0.93 (0.91–0.96)	0.97 (0.93–1.00)	0.88 (0.81–0.96)	0.99 (0.89–1.11)	0.91 (0.89–0.94)	0.97 (0.93–1.00)
Vaccination vs. No Vaccination in Females	0.90 (0.87–0.92)	0.94 (0.91–0.97)	0.79 (0.74–0.85)	0.88 (0.82–0.96)	0.93 (0.90–0.95)	1.00 (0.97–1.03)
Joint Effects
	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)
Overall
ILI vs. No ILI among those with No Vaccination	1.35 (1.30–1.40)	1.46 (1.39–1.53)	1.70 (1.54–1.88)	1.73 (1.55–1.94)	1.29 (1.23–1.35)	1.39 (1.32–1.47)
ILI vs. No ILI among those with a Vaccination	1.15 (1.07–1.25)	1.11 (1.01–1.22)	1.50 (1.18–1.90)	1.41 (1.08–1.85)	1.12 (1.03–1.22)	1.07 (0.96–1.18)
Vaccination vs. No Vaccination among those without ILI	0.89 (0.88–0.91)	0.97 (0.95–0.99)	0.82 (0.77–0.86)	0.93 (0.87–0.99)	0.90 (0.88–0.92)	1.00 (0.98–1.03)
Vaccination vs. No Vaccination among those with ILI	0.76 (0.70–0.83)	0.73 (0.66–0.81)	0.72 (0.56–0.92)	0.76 (0.57–1.01)	0.78 (0.71–0.86)	0.77 (0.69–0.86)
Males
ILI vs. No ILI among those with No Vaccination	1.51 (1.41–2.62)	1.62 (1.50–1.75)	2.27 (1.92–2.67)	2.25 (1.87–2.71)	1.40 (1.30–1.51)	1.50 (1.38–1.63)
ILI vs. No ILI among those with a Vaccination	1.33 (1.16–1.53)	1.23 (1.05–1.44)	2.36 (1.52–3.67)	2.30 (1.41–3.78)	1.25 (1.09–1.45)	1.14 (0.96–1.35)
Vaccination vs. No Vaccination among those without ILI	0.93 (0.91–0.96)	0.97 (0.94–1.01)	0.88 (0.80–0.96)	0.99 (0.89–1.10)	0.91 (0.89–0.94)	0.97 (0.94–1.01)
Vaccination vs. No Vaccination among those with ILI	0.82 (0.71–0.96)	0.74 (0.62–0.88)	0.91 (0.58–1.45)	1.01 (0.60–1.70)	0.82 (0.70–0.96)	0.74 (0.62–0.89)
Females
ILI vs. No ILI among those with No Vaccination	1.38 (1.31–1.45)	1.46 (1.37–1.55)	1.45 (1.28–1.64)	1.47 (1.28–1.70)	1.35 (1.28–1.43)	1.44 (1.34–1.54)
ILI vs. No ILI among those with a Vaccination	1.18 (1.07–1.30)	1.14 (1.01 −1.29)	1.24 (0.93–1.66)	1.15 (0.83 −1.59)	1.17 (1.05–1.30)	1.13 (1.00–1.28)
Vaccination vs. No Vaccination among those without ILI	0.90 (0.88–0.92)	0.95 (0.92–0.98)	0.80 (0.74–0.86)	0.89 (0.82–0.97)	0.93 (0.91–0.96)	1.01 (0.97–1.04)
Vaccination vs. No Vaccination among those with ILI	0.77 (0.68–0.86)	0.74 (0.65–0.84)	0.68 (0.50–0.93)	0.69 (0.49–0.98)	0.81 (0.72–0.91)	0.79 (0.69–0.92)

ILI: influenza-like illness

Note: Vaccinations were assessed during the year prior to stroke/trauma visit and ILI was assessed 1–30 days prior to stroke/trauma visit

^*controlling for having a preventive care visit in the past year, diabetes, valvular heart disease, smoking, alcohol abuse, obesity, and hypertension

P interaction<0.10

Among the young population (18–44 years), unadjusted and adjusted analyses indicated increased odds of stroke for those with ILI 1–30 days prior to index visit (Table 2) and reduced odds of stroke for those having a vaccination in the year prior. Results were consistent when stratified by sex, albeit vaccinations in males were no longer significantly associated with reduced odds of stroke in the adjusted analyses. In analyses of the combined effects of ILI and vaccinations, there was no significant interaction between ILI and vaccination (adjusted P_interaction=0.16). Among those with no vaccinations in the prior year, those with ILI had increased odds of stroke (Model 1 OR, 1.73 [1.55–1.94]) and among those with a vaccination in the prior year, those with ILI also had increased odds of stroke, although to a lesser magnitude (Model 1 OR, 1.41 [1.08–1.85]). Stratified analyses by sex produced similar results, with a greater magnitude of effect in males.

Among the middle-aged population (45–65 years), unadjusted and adjusted analyses indicated increased odds of stroke for those with ILI 1–30 days prior to index visit. Though in an unadjusted model there was a reduced odds of stroke for those having a vaccination in the year prior, this effect was not significant in multivariable analyses. These results were consistent when stratified by sex, with the exception of vaccinations in males remaining statistically significant. In models assessing the combined effects of ILI and vaccinations, there was a significant interaction (adjusted P_interaction<0.001). Among those with no vaccinations in the prior year, those with ILI had increased odds of stroke (Model 1 OR, 1.39 [1.32–1.47]). Among those with a vaccination in the prior year, ILI was not associated with any increase in risk (Model 1 OR, 1.04 [0.96–1.18]). Analyses stratified by sex produced similar results.

Evaluating different time periods, such as ILI at the time of index visit, 1–15 days prior, and 1–60 days prior provided results consistent with the main findings (Table 3). ILI had the strongest measure of effect at the time of index visit for both age groups. The magnitude of effect seen from ILI declined with increased preceding time intervals. When examining combined effects of ILI and vaccinations (Table 4), ILI assessed at index visit, 1–30 days prior, and 1–60 days prior demonstrated the largest increase in odds of stroke among those with no vaccinations in the prior year, with the greatest effects seen in the younger population.

Table 3.

Individual effects of ILI on stroke among those aged 18–65 years at the time of stroke/control visit and the days prior.

	ILI at time of Stroke/Control Visit		ILI at 1–15 Days Prior		ILI at 1–30 Days Prior		ILI at 1–60 Days Prior
	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)
Entire Population (Ages 18–65)
ILI vs. No ILI	6.23 (5.73–6.78)	6.13 (5.59–6.71)	1.62 (1.54–1.70)	1.69 (1.60–1.79)	1.30 (1.25–1.35)	1.38 (1.32–1.44)	1.09 (1.06–1.12)	1.16 (1.12–1.20)
In males	5.04 (4.46–5.70)	5.17 (4.54–5.89)	1.81 (1.67–1.96)	1.86 (1.70–2.04)	1.47 (1.39–1.56)	1.54 (1.43–1.65)	1.23 (1.17–1.29)	1.28 (1.22–1.36)
In females	7.14 (6.35–8.03)	6.88 (6.06–7.81)	1.64 (1.55–1.75)	1.69 (1.58–1.82)	1.32 (1.26–1.39)	1.38 (1.31–1.46)	1.12 (1.09–1.17)	1.16 (1.12–1.22)
Young Population (Ages 18–44)
ILI vs. No ILI	10.71 (8.16–14.07)	10.75 (8.12–14.24)	2.34 (2.06–2.65)	2.36 (2.06–2.72)	1.66 (1.52–1.82)	1.68 (1.51–1.86)	1.28 (1.19–1.37)	1.31 (1.21–1.42)
In males	9.52 (6.57–13.78)	9.97 (6.79–14.65)	3.33 (2.68–4.12)	3.35 (2.64–4.24)	2.27 (1.94–2.65)	2.26 (1.89–2.66)	1.69 (1.50–1.91)	1.67 (1.46–1.91)
In females	11.97 (8.00–17.89)	11.78 (7.79–17.80)	1.91 (1.63–2.23)	1.92 (1.61–2.29)	1.40 (1.25–1.58)	1.41 (1.24–1.61)	1.11 (1.02–1.21)	1.14 (1.03–1.26)
Middle-Aged Population (Ages 45–65)
ILI vs. No ILI	5.79 (5.30–6.33)	5.66 (5.13–6.23)	1.52 (1.44–1.60)	1.57 (1.48–1.67)	1.24 (1.19–1.29)	1.32 (1.26–1.38)	1.06 (1.03–1.09)	1.12 (1.08–1.16)
In males	4.57 (4.01–5.20)	4.64 (4.04–5.33)	1.63 (1.50–1.78)	1.65 (1.49–1.82)	1.37 (1.28–1.46)	1.42 (1.32–1.53)	1.17 (1.11–1.23)	1.22 (1.15–1.29)
In females	6.74 (5.96–7.63)	6.49 (5.66–7.44)	1.60 (1.50–1.71)	1.65 (1.52–1.78)	1.30 (1.24–1.37)	1.36 (1.28–1.45)	1.12 (1.08–1.17)	1.16 (1.11–1.22)

Table 4.

Joint effects of ILI on stroke among those aged 18–65 years at the time of stroke/control visit and the days prior

	ILI at time of Stroke/Trauma		ILI at 1–15 Days Prior		ILI at 1–30 Days Prior		ILI at 1–60 Days Prior
	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)	Unadjusted OR (95% CI)	Adjusted* OR (95% CI)
Entire Population (Ages 18–65)
ILI vs. No ILI among those with No Vaccination	6.28 (5.73–6.90)	6.21 (5.62–6.87)	1.67 (1.58–1.76)	1.77 (1.67–1.89)	1.35 (1.30–1.40)	1.46 (1.39–1.53)	1.11 (1.07–1.14)	1.19 (1.15–1.24)
ILI vs. No ILI among those with a Vaccination	5.98 (4.88–7.31)	5.71 (4.58–7.11)	1.47 (1.33–1.63)	1.41 (1.25–1.60)	1.15 (1.07–1.25)	1.11 (1.01–1.22)	1.05 (0.99–1.11)	1.04 (0.97–1.12)
Vaccination vs. No Vaccination among those without ILI	0.89 (0.87–0.91)	0.96 (0.94–0.98)	0.89 (0.87–0.91)	0.96 (0.94–0.98)	0.89 (0.88–0.91)	0.97 (0.95–0.99)	0.89 (0.87–0.91)	0.97 (0.94–0.99)
Vaccination vs. No Vaccination among those with ILI	0.85 (0.68–1.06)	0.88 (0.69–1.12)	0.78 (0.70–0.88)	0.77 (0.67–0.88)	0.76 (0.70–0.83)	0.73 (0.66–0.81)	0.84 (0.79–0.90)	0.84 (0.78–0.91)
Young Population (Ages 18–44)
ILI vs. No ILI among those with No Vaccination	10.98 (8.19–14.72)	10.99 (8.12–14.86)	2.31 (2.02–2.65)	2.34 (2.01–2.72)	1.70 (1.54–1.88)	1.73 (1.55–1.94)	1.30 (1.21–1.41)	1.34 (1.22–1.46)
ILI vs. No ILI among those with a Vaccination	8.84 (4.21–18.54)	9.27 (4.31–19.95)	2.57 (1.84–3.58)	2.56 (1.78–3.69)	1.50 (1.18–1.90)	1.41 (1.08–1.85)	1.21 (1.01–1.44)	1.18 (0.97–1.45)
Vaccination vs. No Vaccination among those without ILI	0.82 (0.78–0.87)	0.93 (0.87–0.99)	0.81 (0.77–0.86)	0.92 (0.86–0.98)	0.82 (0.77–0.86)	0.93 (0.87–0.99)	0.82 (0.77–0.87)	0.93 (0.87–0.99)
Vaccination vs. No Vaccination among those with ILI	0.66 (0.30–1.46)	0.78 (0.34–1.78)	0.90 (0.63–1.28)	1.01 (0.68–1.49)	0.72 (0.56–0.92)	0.76 (0.57–1.01)	0.76 (0.63–0.91)	0.82 (0.67–1.02)
Middle-Aged Population (Ages 45–65)
ILI vs. No ILI among those with No Vaccination	5.80 (5.25–6.39)	5.71 (5.13–6.35)	1.57 (1.48–1.66)	1.65 (1.54–1.77)	1.29 (1.23–1.35)	1.39 (1.32–1.47)	1.07 (1.04–1.11)	1.15 (1.11–1.20)
ILI vs. No ILI among those with a Vaccination	5.76 (4.67–7.10)	5.43 (4.31–6.84)	1.38 (1.23–1.54)	1.30 (1.13–1.48)	1.12 (1.03–1.22)	1.07 (0.96–1.18)	1.03 (0.97–1.10)	1.01 (0.93–1.09)
Vaccination vs. No Vaccination among those without ILI	0.90 (0.88–0.92)	1.00 (0.97–1.02)	0.90 (0.88–0.92)	1.00 (0.98–1.02)	0.90 (0.88–0.92)	1.00 (0.98–1.03)	0.90 (0.88–0.92)	1.00 (0.98–1.03)
Vaccination vs. No Vaccination among those with ILI	0.89 (0.71–1.12)	0.95 (0.74–1.22)	0.79 (0.70–0.89)	0.78 (0.67–0.91)	0.78 (0.71–0.86)	0.77 (0.69–0.86)	0.86 (0.81–0.93)	0.88 (0.81–0.95)

^*controlling for having a preventive care visit in the past year, diabetes, valvular heart disease, smoking, alcohol abuse, obesity, and hypertension

Vaccinations were assessed during the year prior to stroke/trauma visit and ILI was assessed at varying times

Discussion

Our case-control study uniquely assessed individual and joint effects of ILI and vaccinations on stroke risk in young and middle-aged commercially insured populations. Our results found ILI increased the odds of stroke, while vaccinations decreased the odds of stroke in the general population, particularly among younger adults (18–44 years). For the middle-age group (45–65 years), ILI similarly resulted in an increase in odds of stroke, however to a lesser magnitude, and the effect of vaccinations was not statistically significant. Further, these associations were strongest at the time of index visit and 1–15 days prior, supportive of the potential for ILI to act as a trigger rather than long-term risk factor. In evaluating joint effects of ILI and vaccinations, our findings suggest that risk of stroke following ILI is greatest among younger individuals who had not been vaccinated in the year prior, particularly in males. These results highlight at-risk groups who may benefit from targeted interventions (e.g. vaccination programs) to reduce risk of stroke.

Our findings are consistent with previous studies indicating increased stroke risk after ILI, though previous work was somewhat limited by their utilization of a case-crossover design, which does not account for increased odds of stroke associated with aging and the development of stroke risk factors over time^{16, 18} A participant in a case-crossover study serves as their own control and they may have developed risk factors from the time they served as a control to the time they became a case. However, this difference in risk factors between cases and controls will not be captured in the case-crossover design. Through uses of a case-control design, our study overcame this limitation by matching on age and controlling for risk factors at time of index visit for both those with and without stroke. However, although we matched at time of the index date, we did not account for changes in risk factor profiles or risk factor management over time for either cases or controls.

While prior literature has demonstrated increased risk, the mechanisms by which infections such as ILI increase the risk for stroke are uncertain. Proposed mechanisms behind the relationship between infection and stroke include mediation through a pro-thrombotic state, inflammation-mediated injury of endothelium, infection-related platelet activation and aggregation, inflammation-induced thrombosis, dehydration-induced thrombosis, or through effects on cardiac endothelium.^2–6 In response to the emerging role of infectious disease in stroke risk, vaccinations have been examined as potential interventions. Prior research indicates vaccinations are an important strategy in preventing not only infectious diseases, but also the development of complications resulting from the infectious diseases including stroke, myocardial infarction (MI), and death.^19–22

Earlier studies identified a reduction in stroke risk for people who were vaccinated against influenza.^{12, 23, 24} Our study included a more diverse sample of any type of vaccination, though the vast majority of vaccinations year to year in adults are influenza vaccines.²⁵ Interestingly, when assessing joint effects there was a statistical interaction between ILI and vaccination in the middle-aged population, but not in the young population. Reasons for this may include sample size variations, as strokes are rare in younger populations and possible differences in types of vaccines received by each age group, as younger age groups are less likely to be vaccinated for influenza²⁶. Additionally, biological differences related to sex and age may play a role. When examining the effect by sex within the younger age group, there appears to be no effect in males. This is further demonstrated when examining joint effects of ILI and vaccinations in males, where vaccinations do not appear protective among either males with or without ILI at 1–30 days prior to index visit. However, vaccines appear protective among females, particularly among those with ILI. While our results suggest no individual effect of vaccination among those middle-aged overall, vaccination demonstrated a significant reduction in stroke risk among those middle-aged with an ILI 1–30 days prior to index visit, supporting the hypothesized role vaccines may play in preventing stroke.

Unlike other available vaccines, the influenza vaccine is not similarly protective every year. Since the 2004–05 influenza season, the CDC has provided estimates of the effectiveness of the influenza vaccination. In the years with increased vaccine effectiveness, the overall rates of ILI, stroke, and MI were nominally lower, but the association between ILI and stroke remained the same.²⁷ While the potential to become infected with influenza even after vaccination remains, the symptoms experienced may be less severe if vaccinated, and post-infectious complications like stroke, MI and death are reduced.^{12, 23, 24, 28–30} Our results support these prior findings through demonstrating a reduced effect of ILI among the vaccinated young populations. Consistent influenza vaccination done yearly even further reduces the risk of stroke.³⁰ As such, the American Heart Association recommends influenza vaccination as a secondary stroke prevention strategy in patients with coronary or atherosclerotic vascular disease.²⁹ While the bulk of the literature consistently indicates influenza vaccination reduces stroke risk, some studies have identified little or no benefit to influenza vaccination in stroke risk reduction, and thus, the topic remains unsettled.^{31, 32} This could be due to differences in study populations, such as variations in age or perhaps sex differences. For example, our results support there to be limited to no individual effect of vaccination among older populations or among males. Further, as demonstrated by Lin et al., the frequency of receiving influenza vaccinations may influence the strength of effect, as only 1–2 influenza vaccines over 5 years had no effect on stroke risk while 3–4 and 5 vaccinations were associated with significant reductions in risk.³⁰

In addition to the influenza vaccine, other vaccines have been identified as protective factors against stroke. A prospective cohort study identified a 35% reduction in stroke risk at one year in patients who received the 23-valent pneumococcal vaccine.³³ However, the protective benefits did not last longer than 3 years, and a separate cohort study among men found no reduction in stroke risk from the pneumococcal vaccine.^{34, 35}

Our study had several limitations. While utilizing administrative data allowed us to achieve a large sample size, particularly for the rare outcome of stroke among the young, we were limited to using ICD-9 codes for exposures, outcomes, and covariates. ICD-9 codes have been validated, but there is potential for misclassification. For our study, ILI had to result in an ICD-9 code associated with an outpatient, inpatient, or emergency department visit to be included. Therefore, we likely capture more moderate to severe ILI, limiting our capability to make conclusions about mild ILI. Further, vaccinations and preventive care visits were assessed through ICD-9 procedure codes associated with outpatient visits. Therefore, our study did not capture whether patients sought care or vaccinations through other sources, such as through work programs or a pharmacy. As a result, vaccination rates may also be underestimated; however, we expect misclassification resulting from use of ICD-9 codes to affect cases and controls similarly. We believe influenza vaccinations to make up the majority of vaccinations observed our study given it is recommended annually. However, the proportion of vaccinations being for influenza and types of other vaccinations given likely vary by age group. Vaccinations may also be a surrogate for other healthy behavior, and thus associations may, in part, represent this. However, to reduce this potential bias we also controlled for preventive visits. Additionally, calculated E-values were moderate, suggesting observed associations may be explained by unmeasured confounders. Finally, MarketScan captures insurance claims data, and thus this study may not be representative of uninsured individuals or individuals with other types of insurance that are not included in MarketScan. However, we believe the large sample size representative of a national population achieved through this study design outweighs this limitation, as it allows us to examine stratified analyses among populations where the outcome of stroke is rare.

In summary, our case-control study found ILI associated with increase in risk of stroke, while vaccines appeared to decrease the risk of stroke, particularly among the young. Further, our study uniquely examines joint effects of ILI and vaccinations on stroke with results indicating vaccinations can reduce the effect of ILI on stroke. Importantly, these findings aid in informing at-risk groups who may most benefit from vaccination programs for stroke prevention and may provide additional motivation for younger populations to get their yearly influenza vaccine. Building on our work, future investigations should further explore the potential for causality of the observed relationship between vaccinations and stroke. Elucidating potential mechanisms for increased stroke risk following ILI infection may further aid the identification of targeted and effective interventions.

Non-standard Abbreviations and Acronyms

ILI

Influenza-like illness

ICD-9

International Classification of Diseases, Ninth Revision

SD

standard deviation

IQR

interquartile

OR

odds ratios