key: cord-0007577-8va21fcc
authors: King, James C.; Ajao, Adebola; Lichenstein, Richard; Magder, Laurence S.
title: Surge in Hospitalizations Associated With Mechanical Ventilator Use During Influenza Outbreaks
date: 2014-04-01
journal: Disaster Med Public Health Prep
DOI: 10.1017/dmp.2014.18
sha: d06e67b93762bed89433ce3b50ba733a5b74f9a9
doc_id: 7577
cord_uid: 8va21fcc

OBJECTIVE: Information on surges in critical care services including mechanical ventilator use during seasonal influenza outbreaks is limited. To potentially facilitate preparedness plans for future pandemics, we retrospectively quantitated surges in all-cause mechanical ventilator use during peak influenza for 12 consecutive years in all certified hospitals in Maryland. METHODS: Influenza testing data obtained for the Centers for Disease Control and Protection, Health and Human Services region 3, included defined peak influenza outbreak periods (PIOP), non-influenza time periods (non-ITP), and proportions of circulating influenza types for all study years. Procedure codes for mechanical ventilator use and diagnostic codes for medically attended acute respiratory illness (MAARI) were reviewed for every Maryland hospitalization. Daily counts of hospitalizations associated with ventilator use or MAARI during PIOP compared to non-ITP were analyzed using Poisson regression adjusted for month and year. RESULTS: Ventilator use increased during PIOP by 7% (95% CI, 5-10) over non-ITP (P < .0001) for all study years. These annual surges correlated with influenza season intensity, as measured by MAARI-related hospitalizations (correlation coefficient = 0.91, P < .0001). CONCLUSIONS: Surges in ventilator use were temporally associated with PIOP and were positively correlated with influenza season intensity, as measured by hospitalizations associated with acute respiratory illness. This information may assist resource planning for future pandemics. (Disaster Med Public Health Preparedness. 2014;x:1-7)

I n the United States, federal, state, and local authorities have been challenged to develop and execute preparedness plans to address the potential surge of critically ill patients that could occur in future influenza pandemics. 1 This task may be especially important if the outbreaks are caused by avian influenza viruses, which may be more virulent than seasonal strains and for which no effective vaccines or antivirals may be available. 2, 3 To respond to a surge in critically ill patients, 4 operational capabilities are postulated 4,5 : space, staff, systems, and equipment and supplies. Space includes appropriate medical facilities to provide safe critical care. Staff includes appropriately trained medical professionals to care for critically ill patients. Systems include pre-event planning activities such as forming coalitions; developing and exercising communication protocols; and providing logistics, support, and training before and during an event. Finally, equipment and supplies include access to appropriate durable and disposable critical medical equipment and supplies. 5 Many hospitals have minimal reserves of durable medical equipment such as mechanical ventilators onsite at any one time; additional units are usually rented from medical supply vendors. 5, 6 The US Department of Health and Human Services (HHS) and some states have or are developing stockpiles of mechanical ventilators and associated supplies such as ventilator circuits and drugs such as neuromuscularblocking agents to address surges in critically ill patients during future influenza pandemics. 5, 7 Reliable estimates of such critical care equipment are needed to determine government stockpiling goals for future events. Previous predictions of projected national ventilation needs are based on historical pandemics and limited by the uncertainties in reconstructing these events. 8, 9 Examining increases in the use of mechanical ventilation during multiple seasonal influenza outbreak periods may help to identify the quantitative range of peak surge in the use of mechanical ventilators and associated operational capabilities during seasonal influenza outbreaks. Once a methodology is developed to compare the severity of seasonal to pandemic influenza outbreaks using historical information, these data may be useful to better predict future influenza pandemic preparedness activities, including determining federal ventilator stockpiling goals.

Seasonal influenza outbreaks have been associated with substantial medical utilization. Excess outpatient medical visits, hospitalizations, and admissions to intensive care units have been described with influenza outbreaks. [10] [11] [12] [13] In addition, epidemiologic studies have revealed other serious medical conditions that are not always recognized as associated with influenza infections and that may require hospitalizations and mechanical ventilator support. These conditions include, but are not limited to, secondary infections such as bacterial pneumonia, acute ischemic heart disease, and neurologic complications such as stroke, Guillain-Barré syndrome, and encephalitis. 14-17 Therefore, it is important to consider a broad range of symptoms and diagnostic or procedure codes to capture retrospectively the full impact of seasonal influenza outbreaks on the utilization of the health care system, including medical facilities and equipment.

The primary objective of this investigation was to use retrospective inpatient and viral surveillance data to quantify the surge in all-cause hospitalizations associated with ventilator use during peak influenza outbreak periods (PIOP) compared to non-influenza time periods (non-ITP) throughout multiple years. A secondary objective was to examine the association between the degrees of annual surges in ventilator use with the intensity of the influenza season, as measured by hospitalizations related to medically-attended acute respiratory illnesses (MAARI). An additional secondary objective was to examine the correlation between annual surges in ventilator use during peak periods of influenza activity with the proportion of individual circulating influenza types or subtypes for each season.

The study population included all patients who were admitted to Maryland Anonymous data extracted from the HSCRC database included International Classification of Diseases, Ninth Revision (ICD-9) codes of interest, patient's admission date and age on admission in years (or months in children younger than 2 years of age). Other data extracted from the HSCRC database included all-cause hospital admissions for all ages with at least 1 procedure code for mechanical ventilation. ICD-9 procedural codes for continuous mechanical ventilation included 96.70, 96.71, and 96.72. All-cause mechanical ventilator use results were further broken down into 5 age groups: 1 to 23 months, 2 to 17 years, 18 to 49 years, 50 to 64 years, and older than 64 years.

Data on hospital admissions associated with MAARI diagnoses were defined and used to estimate the intensity of each influenza season in a manner previously published. [17] [18] [19] Diagnoses related to MAARI included 1 or more primary or subsequent ICD-9 discharge codes for upper respiratory tract infections (460.x-465.x), which incorporated acute otitis media, sinusitis, mastoiditis, acute respiratory infection, common cold, pharyngitis, and tonsillitis. Also included were codes for middle respiratory tract diagnoses or symptoms (464.x) such as laryngitis, tracheitis, or croup. Lower respiratory tract illnesses or symptoms included either symptoms related to breathing difficulty, such as stridor (786. to here as the influenza outbreak period) included the fewest number of consecutive weeks, with 85% or more of that season's positive influenza test results reported for HHS region 3. A 2-week buffer period at the beginning and end of this influenza season interval was included to assure that this time period captured the majority of influenza activity in Maryland for each study year. Finally, the non-ITP included the remaining weeks on either side of the influenza season, beginning with the start of each study year to the beginning of the defined influenza season and from the end of the influenza season to the end of the study year ( Figure 1 ).

In addition, the CDC website data were used to identify the types and subtypes of influenza viruses circulating in HHS region 3. These data were used to estimate the proportion of influenza A/H3N2, A/H1N1, and B viruses that circulated in the Maryland area each study year.

Analyses were based on daily counts of key endpoints including all-cause hospital admissions associated with ventilator use as well as MAARI-related hospitalizations for all Maryland hospitals. These counts were analyzed using a Poisson regression model, allowing for overdispersion. The model included terms to allow daily expectation to vary by year and month. It also included terms to allow for higher rates during the PIOP than the non-ITP of each study year. The periods of time between the PIOP and the non-ITP were used in the regression model to adjust for times of the year that likely involved a mix of factors including influenza, other respiratory pathogens, and seasonal or environmental issues that may have influenced the outcome endpoints of interest. The Poisson regression model was used to calculate the rate ratio (RR) of each outcome of interest for each study year's PIOP compared to the corresponding non-ITP. The RR was considered a quantitative indicator of excess rate or surge of these medical outcomes during the peak PIOP for each of the 12 study years. RRs for ventilator use and MAARI-related hospitalizations were calculated for all age groups combined and also for 5 age groups: 1 to 23 months, 2 to 17 years, 18 to 49 years, 50 to 64 years, and older than 64 years.

The correlation coefficient (CC) was calculated by comparing each year's RR for MAARI-related hospitalizations, as a proxy for influenza season intensity, to that of the same year's hospitalizations associated with ventilator use. In addition, the RRs for individual patients requiring mechanical ventilation that had or did not have concomitant MAARI-related diagnoses during the PIOP were compared to those during the non-ITP.

CCs were also calculated comparing the RR of ventilator use during the PIOP with that of the non-ITP to the proportion of individual influenza virus types and subtypes that circulated in the Maryland area for each study influenza season.

All data regarding influenza surveillance and hospitalization were examined and analyzed without personal identifiers. This study was reviewed and approved by the Institutional Review Board of the University of Maryland at Baltimore.

A total of 8 729 857 patients were admitted to all 47 Maryland hospitals from June 1, 2001, through June 30, 2013 (Table 1) . Of these admissions, 255 549 (2.9%) were associated with ventilator use, while 1 018 432 (11.6%) were associated with MAARI diagnostic codes. Table 1 also shows the age breakdown for these admissions and surges (RRs) in ventilator or MAARI-related hospitalizations. For all ages and years combined, there was a 7% increase in the rate of ventilator use during the PIOP compared to the non-ITP (RR 5 1.07, P , .0001) and a 20% increase (RR 5 1.20, P , .0001) in MAARI-related hospitalizations during the PIOP compared to the non-ITP. Table 1 also shows the impact of the influenza outbreaks on ventilator use and MAARI-related hospitalizations by pooling all 12 study years by each age group. The largest increase in ventilator use during the PIOP for all years combined was observed in children aged 1 to 23 months. However the variability of this estimate was high due to the low number of ventilators used in this age group. Of all the admissions associated with ventilator use, 68% were by patients aged 50 years and older and 44% by patients 65 years and older.

In 

The findings in this study indicate an overall statistically significant surge in hospitalizations associated with mechanical ventilator use in Maryland residents during the peak 4 weeks of influenza outbreaks compared to the non-ITP. This finding is important because influenza illnesses likely represent only a fraction of the medical conditions that require the use of mechanical ventilation. 22 The association of influenza infection with surges in ventilator use was suggested by 4 key findings in this study. First, an important finding was a statistically significant increase in ventilator use temporally related to peak influenza activity. Second, a statistically significant correlation was observed between increases in ventilator use and the intensity of the influenza season, as measured by increases in acute respiratory illness-related hospitalizations during peak influenza activity. Third, the surge in ventilator use during the PIOP was predominantly driven by patients who required mechanical ventilation and who also had acute respiratory illness, as defined by MAARI diagnostic codes. Fourth, a statistically significant correlation was observed between the proportions of circulating A/H3N2 viruses and ventilator use in individuals older than 64 years of age, which is the population most likely to need ventilators. This last finding supports previous reports in the literature, which demonstrate that infection with seasonal influenza A/H3N2 causes more severe illness than other influenza virus types or subtypes, especially in the elderly. 11, 12, 17, 19 The data on ventilator use related in this study involved all-cause use. Trauma units, neurological or surgical intensive The overall surge in the use of mechanical ventilators during peak seasonal influenza periods may not appear profound but, in fact, may be very important during pandemic influenza outbreaks, when pandemic strains are likely to induce more severe illness than occur with seasonal strains. It has been speculated that hospitals may not have available on-site all of the mechanical ventilators needed and may have to rent equipment from medical supply companies in times of high demand. 5, 6 To our knowledge, this study is the first to use a large data set to directly quantify the excess all-cause hospitalizations associated with ventilator use during multiple seasonal influenza peak outbreaks. Data from this study may be useful baseline data that could be incorporated into historical pandemic data to model and estimate the surge in hospitalizations associated with ventilator use for future pandemics.

However, future influenza pandemic outbreaks may involve novel influenza strains that may induce more severe illnesses than are observed with annual seasonal strains. 3 In addition, the timeliness and effectiveness of influenza vaccines, the effectiveness of antiviral agents, and attempts to reduce transmission via personal protective devices may be more limited during future pandemics than during seasonal influenza outbreaks. These possibilities make decisions on stockpiling goals for critical care equipment based on seasonal outbreak information challenging. 9 Finally, it is important to note that stockpiling durable and disposable critical medical equipment and supplies is a necessary but not sufficient capability for pandemic influenza preparedness. Enhancing the other operational capabilities such as space, staff, and systems is an imperative part of pandemic preparedness. Experts in the field have indicated that systems are particularly important to effectively utilize stockpiled medical countermeasures such as mechanical ventilators. Systems involve comprehensive planning and training, developing and exercising crisis standards of care protocols, and perfecting logistics and communication before and during pandemics. 23, 24 Study Strengths and Limitations

One strength of this study is having a robust-sized data set on millions of patients of all ages admitted to all certified hospitals in the State of Maryland. Another is having data on 12 consecutive influenza seasons, including 1 pandemic H1N1 outbreak, which allowed an examination of outcomes for a range of seasonal intensities ranging from mild to moderately severe. The Poisson model used in this study had been adjusted for month and year to minimize the influence of other factors such as weather or other circulating respiratory pathogens that could cause a surge in outcomes during the winter months.

The study has 3 main limitations. First, HHS region 3 influenza virus surveillance data may not exactly match the quantity and type of influenza viruses circulating specifically in Maryland. Second, data from Maryland may not precisely reflect other states or regions because of geographic and demographic differences. Third, the small number of ventilators used in children led decreased the precision of data in this age group.

The findings of this study indicate that seasonal influenza infections are associated with small to moderate surges in allcause hospitalizations associated with the use of mechanical ventilators. Pandemic influenza preparedness activities include stockpiling vital medical equipment, such as ventilators and associated supplies, to respond to future pandemic influenza outbreaks. The data from this study may help the HHS to develop increasingly refined models to estimate the need for mechanical ventilators to help prepare for future moderate to severe pandemic influenza outbreaks. 25, 26 About the Authors 

National strategy for pandemic influenza

Preparing for the next pandemic

Pandemic influenza threat and preparedness

Essential equipment, pharmaceuticals and supplies

Definitive care for the critically ill during a disaster: current capabilities and limitations

Survey of mechanical ventilators in US acute care hospitals: a baseline for critical care surge capacity planning

The strategic national stockpile: roles and responsibilities of health care professionals for receiving the stockpile assets

FluSurge: a tool to estimate demand for hospital services during the next pandemic influenza

Definitive care for the critically ill during a disaster: a framework for optimizing critical care surge capacity

Prevention and control of influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP)

Influenza-associated hospitalizations in the United States

Mortality associated with influenza and respiratory syncytial virus in the United States

Critical care services and 2009 H1N1 influenza in Australia and New Zealand

Improving the estimation of influenza-related mortality over a seasonal baseline

Neurologic complications of influenza

Bacterial and viral infections associated with influenza. Influenza Other Respir Viruses

The relationship between influenza outbreaks and acute ischemic heart disease in Maryland residents over a 7-year period

Herd immunity in adults against influenza-related illnesses with the use of trivalent-live attenuated influenza vaccine (CAIV-T) in children

Relationship of influenza vaccine match and use rate to medically attended acute respiratory illnesses in older residents of Maryland

National and regional level outpatient illness and viral surveillance: influenza positive tests reported to CDC, national summary

US Census Bureau

The epidemiology of ventilator use in the United States

Modeling hospital response to mild and severe influenza pandemic scenarios under normal and expanded capacities

Definitive care for the critically ill during a disaster: medical resources for surge capacity

Assistant Secretary for Preparedness and Response, US Department of Health and Human Services