Drug use and severe outcomes among adults hospitalized with influenza, 2016–2019

Abstract Background Influenza is a persistent public health problem associated with severe morbidity and mortality. Drug use is related to myriad health complications, but the relationship between drug use and severe influenza outcomes is not well understood. The study objective was to evaluate the relationship between drug use and severe influenza‐associated outcomes. Methods Data were collected by the Influenza Hospitalization Surveillance Network (FluSurv‐NET) from the 2016–2017 through 2018–2019 influenza seasons. Among persons hospitalized with influenza, descriptive statistics and logistic regression models were used to analyze differences in demographic characteristics, risk and behavioral factors, and severe outcomes (intensive care unit [ICU] admission, mechanical ventilation, or death) between people who use drugs (PWUD), defined as having documented drug use within the past year, and non‐PWUD. Results Among 48,430 eligible hospitalized influenza cases, 2019 were PWUD and 46,411 were non‐PWUD. PWUD were younger than non‐PWUD and more likely to be male, non‐Hispanic Black or Hispanic/Latino, smoke tobacco, abuse alcohol, and have chronic conditions including asthma, chronic liver disease, chronic lung disease, or immunosuppressive conditions. PWUD had greater odds of ICU admission and mechanical ventilation, but not death compared with non‐PWUD; however, these findings were not statistically significant after adjustment. Opioid use specifically was associated with increased risk of ICU admission and mechanical ventilation. Conclusion These results support targeted initiatives to prevent influenza in this population, including influenza vaccination, which remains one of the most important tools to prevent influenza infection and associated severe outcomes.

(Grant NU38OT000297-02-00). C.E.P. is funded by the National Institute on Alcohol Abuse and Alcoholism (Grant T32AA025877). models were used to analyze differences in demographic characteristics, risk and behavioral factors, and severe outcomes (intensive care unit [ICU] admission, mechanical ventilation, or death) between people who use drugs (PWUD), defined as having documented drug use within the past year, and non-PWUD.
Results: Among 48,430 eligible hospitalized influenza cases, 2019 were PWUD and 46,411 were non-PWUD. PWUD were younger than non-PWUD and more likely to be male, non-Hispanic Black or Hispanic/Latino, smoke tobacco, abuse alcohol, and have chronic conditions including asthma, chronic liver disease, chronic lung disease, or immunosuppressive conditions. PWUD had greater odds of ICU admission and mechanical ventilation, but not death compared with non-PWUD; however, these findings were not statistically significant after adjustment. Opioid use specifically was associated with increased risk of ICU admission and mechanical ventilation.
Conclusion: These results support targeted initiatives to prevent influenza in this population, including influenza vaccination, which remains one of the most important tools to prevent influenza infection and associated severe outcomes. proportionately such as older adults, persons with chronic underlying conditions, pregnant women, and children. 2 It is important to identify additional risk factors for severe influenza outcomes to help inform public health recommendations and provide guidance for groups at higher risk for severe influenza outcomes.
Drug use is associated with a myriad of health complications and diseases. 3 Use and abuse of drugs as well as overdose deaths have increased in the United States since 2000. 4,5 There is a growing body of evidence that drug use can lead to increased risk for acquiring and having severe outcomes due to respiratory diseases. Several studies have found that opioids, including prescription opioids, can impair respiration and lung function, weaken the immune system, and lead to lung-damaging inflammation. 6,7 Inhaling drugs such as marijuana, methamphetamine, and crack/cocaine can lead to lung damage and a variety of chronic lung conditions. [8][9][10][11] Intravenous (IV) drug use has also been associated with lung disease. 12 However, the epidemiology of influenza among people who use drugs (PWUD), including the association between drug use and severe influenza disease, is not well understood.
Early studies indicate that PWUD have higher risk of acquiring and experiencing severe outcomes from SARS-CoV-2 infection. [13][14][15] PWUD might also be at an increased risk for acquiring and experiencing severe outcomes from influenza. With the ongoing national opioid epidemic, there is an urgent need to better understand the association between drug use and the risk of influenza illness and severe influenza-associated outcomes in this population. The Centers for Disease Control and Prevention (CDC)-funded Influenza Hospitalization Surveillance Network (FluSurv-NET) conducts populationbased surveillance for laboratory-confirmed, influenza-related hospitalizations. FluSurv-NET sites are located around the country and represent approximately 9% of the total US population or 27 million people. 16 FluSurv-NET data can provide insight into the epidemiology of severe influenza-associated outcomes among PWUD hospitalized with influenza.
The objectives of this analysis were to compare demographic, clinical, and behavioral characteristics among PWUD and non-PWUD   cases can include catchment area residents admitted to hospitals outside the catchment area. 16 Cases with a positive influenza test >3 days after hospital admission were excluded from the analysis. Influenza testing is clinician-driven and influenza testing practices vary across the hospitals participating in FluSurv-NET. 17 1. depressants such as benzodiazepines and barbiturates; 2. stimulants which included amphetamines and cocaine; 3. opioids, primarily prescription drugs and heroin; 4. hallucinogens such as mushrooms, PCP, and "club drugs" like MDMA; and

inhalants (excluding e-cigarettes)
Tobacco use and alcohol abuse were classified as behavioral factors. Marijuana use and e-cigarette use were excluded from this analysis due to varying state laws and regulations. More information about the sampling scheme and data collection by season can be found in Tables 1, 2A, and 2B in Appendix S1.

| Data analysis
Descriptive statistics, chi-square tests of independence, odds ratios, and 95% confidence intervals were used to evaluate differences in demographic characteristics, behavioral factors, specific drug use, and severe influenza outcomes among PWUD and non-PWUD. Severe outcomes were defined as ICU admission, mechanical ventilation, and in-hospital death. Univariable and multivariable logistic regression models were created to examine the association between PWUD status and ICU admission or mechanical ventilation, but not for death due to small sample size (56 deaths occurred among the PWUD group). PWUD with non-missing data for the demographic, behavioral, and outcome variables of interest were included in multivariable logistic regression models. Multivariable logistic regression models were formed by initially including all statistically significant (p < 0.05) demographic, behavioral, and underlying health condition variables, then using backwards elimination until only those that remained significant were included. PWUD status was kept in the final model, regardless of significance, as the primary exposure of interest, as well as smoking status due to its known role as a confounder in the relationship between substance use and respiratory disease. 23 After backwards elimination, the ICU admission model controlled for age, chronic lung disease, chronic liver disease, asthma, sex, and vaccination status, and the mechanical ventilation model controlled for age, chronic lung disease, asthma, sex, and vaccination. Additional multivariable logistic regression models examining stimulant and opioid use specifically were created for ICU admission and mechanical ventilation. These models compared PWUD with specific drug use with non-PWUD and controlled for the same variables as the general drug use models. All analyses were conducted using SAS 9.4 24 and accounted for the complex survey design using methods previously described. [17][18][19][20] 3 | RESULTS

| Descriptive analysis
The analytic sample included 48,430 adults. Of these, 2019 (4.2%) were PWUD. Demographic characteristics are presented in Table 1. Factors associated with PWUD in bivariate analyses are described in Table 2. Compared with non-PWUD, PWUD hospitalized with influenza were more likely to be current tobacco smokers (OR: 12. Bivariate associations between PWUD status and severe influenza-associated outcomes are described in Table 3. PWUD had significantly greater odds of having a hospital stay longer than 2 weeks   Note: Ns in each of the categories might not add up to total numbers of PWUD (n = 2019) and/or non-PWUD (n = 46,411) cases due to missing or unknown data. Significant differences between groups (p < 0.05) were detected by chi-square test for categorical variables.

| Logistic regression models
In multivariable analyses shown in

| DISCUSSION
An analysis of a large, population-based, and geographically diverse sample of hospitalized adults with influenza over three seasons found that there were significant differences in demographics, comorbidities, behavioral factors, and severe outcomes experienced among PWUD compared with non-PWUD. PWUD had greater odds of being admitted to the ICU and requiring mechanical ventilation. While these effects were no longer significant in multivariable analysis after adjusting for other factors, the direction of the associations remained unchanged. Exclusive opioid use, but not stimulant use, was associated with increased risk for ICU admission and mechanical ventilation.
Further research using data from a variety of populations is needed to confirm these findings.
There are many strengths to this study, chief among them that the dataset is a large, rich, sample of people hospitalized with influenza in the United States. Evidence has shown that substance use has increased during the COVID-19 pandemic, so it is possible that in the future PWUD will be even more impacted by influenza. 13,14 T A B L E 3 Severe outcomes among adults hospitalized with laboratory-confirmed influenza by PWUD status, Note: The ICU model controlled for significant categorical variables which were age, chronic lung disease, chronic liver disease, asthma, sex, and vaccination. The mechanical ventilation model controlled for significant categorical variables which were age, chronic lung disease, asthma, sex, and vaccination. Both models also were forced to control for smoking, a known confounder that was not significant in either model. The comparison group for these models was non-PWUD.
However, there are some limitations to the study. Some variables known to be associated with PWUD status and health outcomes, like health insurance, were not available in this dataset. Additional analyses can use data not available to this study to assess associations with characteristics that might be impactful in the relationship between PWUD status and influenza outcomes. While it was a strength that data on drug use came from medical record reviews and were collected systematically by the FluSurv-NET team, the information written in the medical records might be incomplete or incorrect, especially that on substance use and abuse. Self-report by patients could lead to substantial undercounting of PWUD and thus misclassification of PWUD as non-PWUD. Estimated illicit drug use among adults in the United States is 11.7%, 25 but our data show 4% of hospitalized influenza cases are PWUD. A potential recommendation from this study is to expand data collection to allow for more complete and consistent collection of data on drug use among hospitalized patients, which could help address this discrepancy. Additionally, hospital providers could screen more often for substance use and/or conduct more thorough interviews with cases where possible to get a full drug use history. Finally, hospitalization is itself a severe influenza outcome, and this dataset might not be representative of the general population who did not experience this severe outcome. Repeating this study with an entire population-level dataset could show the true effect of drug use on severe influenza outcomes. Such a dataset would also have the benefit of not having the potential influence of admission bias or other biases masking true effects.
PWUD were less likely to have received their seasonal influenza vaccine than non-PWUD. Influenza vaccinations remain one of the most effective and most recommended methods to prevent severe outcomes from influenza. Although the study was not designed to assess vaccine effectiveness and does not take into account the propensity for PWUD to be vaccinated, these findings suggest an opportunity to improve vaccine uptake among PWUD, as vaccination remains one of our most important tools to prevent influenza and associated severe outcomes.
Identifying factors that impact the spread and severity of influenza can help to inform recommendations and best practices for public health and medical professionals that can help groups at higher risk for severe influenza outcomes such as PWUD. Future research can build upon this work and inform initiatives targeting PWUD with the goal of preventing severe outcomes among this population.  Note: The ICU model controlled for significant categorical variables which were age, chronic lung disease, chronic liver disease, asthma, sex, and vaccination. The mechanical ventilation model controlled for significant categorical variables which were age, chronic lung disease, asthma, sex, and vaccination. All models also were forced to control for smoking, a known confounder that was not significant in either model. The comparison group for these models was non-PWUD.