Abstract
Background:
Traumatic injuries are a leading cause of death and morbidity. Despite their comprising a small (<5%) segment of all hospitalizations, the length of stay (LOS) is above average; and the cost of care for the more than 20 million trauma inpatients nears $30 billion per year. Adding insult to injury, risk factors for health-care–associated infections (HAI), including invasive devices and comprised integrity, may be particularly pronounced in this population, potentially exacerbating the clinical and economic burden. Our aim was to determine the distribution, determinants, and burden of HAI after traumatic injury using LOS as a surrogate for health-care–related expense.
Patients and Methods:
This retrospective cohort study used the Trauma Quality Improvement Project (TQIP) database (2013–2016). Patients 16 to 89 years of age were included. Those who developed at least one of the following were counted as cases: Catheter-related central blood stream infection, catheter-related urinary tract infection, ventilator-associated pneumonia, surgical site infection, osteomyelitis, and severe sepsis. Outcomes included the hospital LOS, intensive care unit (ICU) days, and ventilator days. Uni-variable and propensity-matched analyses were conducted to determine differences among patients with and without an HAI.
Results:
Of 806,066 patients, 5.6% (n = 44,844) developed an HAI. A higher proportion of patients with HAI had co-morbid risk factors of chronicity and history of blood transfusion and rated higher on the Abbreviated Injury Scale than those without HAI. After matching, those with HAI also had significantly longer (3 × ) overall LOS, ICU LOS, and prolonged mechanical ventilation (p < 0.05).
Conclusions:
This updated epidemiology study of trauma patients showed the HAI burden to be higher than past estimates and disproportionate of all patient estimates. The associated economic burden of a longer ICU stay with a tripling of the LOS and longer mechanical ventilation demands responsible administrative policies and support for infection prevention programs and interventions.
A leading cause of death in the United States is trauma; which is estimated to total 79,000 deaths annually in persons under 45 years of age [1]. Trauma also has been identified as the most important cause of potential years of life lost for those under 65 years of age [1]. This is more striking, as trauma patients comprise only a small segment of all hospitalizations: Of a total of 465,342,651 hospital discharges between 2000 and 2011, 4.4% (20,659,684) were inpatient trauma discharges. In 2016, DiMaggio et al. provided a national estimate of the average length of hospital stay (LOS) of 5.1 days in all trauma patients seen between 2001 and 2011. The total cost of that care was $240.7 billion, reflecting increases year over year from $12 billion in 2001 to $29.1 billion in 2011 [1].
Those who survive to receive treatment in hospitals face an additional threat of health-care–associated infections (HAIs), a largely preventable adverse consequence of the delivery of healthcare [2]. Despite the most recent (2015) prevalence data from the Centers for Disease Control and Prevention (CDC) showing a reduction over prior years, the continued burden of HAI is staggering; approximately 1 in 31, or a little over 3%, of all hospitalized patients have an HAI on any given day, approximately 10% of whom die during their hospitalization [3]. The annual cost of the five major categories of HAIs (surgical site infections [SSIs], central catheter-associated blood stream infection, catheter-associated urinary tract infection, ventilator-associated pneumonia, and Clostridioides difficile) was US$9.8 billion (2012 dollars) [2]. A major factor in the high cost is the long hospitalization associated with HAI.
Risk factors for HAI are numerous and include factors such as patient age, co-morbid conditions, and the number and types of invasive procedures and devices [4,5]. The epidemiology of trauma patients has shifted such that they are increasingly of advanced age. Also, by definition, traumatic injury care typically necessitates invasive procedures and devices and therefore potentially a greater risk for HAI [1,6,7]. However, we do not know if trauma patients are disproportionately found to have an HAI. The consequences suffered and impact of both trauma patients and patients with an HAI thus is potentially amplified by trauma patients who have an HAI.
A few studies have examined the clinical and economic impact of various categories of HAI in trauma patients using both single center examinations and publicly available national administrative datasets [8-11]. Unfortunately, many of these findings are dated and thus may not reflect current trends in either trauma patients or HAI prevention and control groups. Moreover, the use of administrative definitions or markers of HAIs may not capture accurately the standardized clinical epidemiology definitions set forth by the CDC National Healthcare Safety Network (NHSN), which is used for benchmarking and public reporting across the United States [3].
Therefore, in the context of the considerable clinical and economic impacts of trauma and HAIs, the aim of this study was to address gaps in our knowledge of patients with a dual diagnosis (trauma and HAI) and provide evidence to update the state of the science. The specific purpose of the study was to evaluate the burden of HAIs after traumatic injury using hospital LOS as a surrogate for health-care–related expense.
Patients and Methods
Patient selection and characterization
This was a retrospective cohort study. Information on all adult patients who sustained injury was accessed from the Trauma Quality Improvement Project (TQIP) database from the calendar years 2013–2016. The TQIP is a quality improvement program maintained by the American College of Surgeons (ACS). The database is a subset of the National Trauma Data Bank (NTDB); approximately 65% of all trauma patients' data deposited in the NTDB are qualified for the inclusion in the TQIP. More than 825 institutions throughout the United States currently participate in the TQIP data sharing program [12]. The TQIP also provides feedback to the participating centers on certain quality measures. All patients, age 16 to 89 years old, who sustained either blunt or penetrating mechanisms of injury and were brought to the hospital were included in the study. The patient characteristics captured were race, sex, initial systolic blood pressure (SBP mm Hg), hypotension (SBP <90 mm Hg) at the time of initial hospital presentation, heart rate, Injury Severity Score (ISS), Glasgow Coma Scale (GCS), packed red blood cell (PRBC) transfusion within four hours and 24 hours of hospital arrival, and patient co-morbidities (chronic alcoholism, on chemotherapy for cancer, history of disseminated malignant disease, current smoker, chronic renal failure on dialysis, diabetes mellitus (DM), respiratory disease including chronic obstructive pulmonary disease (COPD), history of steroid use). Patients younger than 16 years or who presented to the hospital with no signs of life were excluded from the study.
Infections
For this analyses, we assigned the designation “HAI positive” to a patient who developed at least one of the following infection outcomes during the initial hospitalization: Central catheter-associated blood stream infection (CCABSI), urinary tract infection including catheter-associated urinary tract infection (CAUTI), pneumonia including ventilator-associated pneumonia (VAP), SSI, osteomyelitis, and severe sepsis (SS). Classification of these individual outcomes occurs by trained infection prevention and control staff as part of routine hospital surveillance and follows strict CDC NHSN guidelines per TQIP data requirements [1,2]. An “HAI negative” value was assigned to patients who developed none of these outcomes.
The primary outcome of this study was the hospital LOS. Secondary outcomes were intensive care unit (ICU) days and ventilator days. Because the data were de-identified and available to researchers, the study was exempted from the Institutional Review Board review as per institutional policy.
Statistical methods
Patient characteristics and outcomes were summarized using median with interquartile range (IQR) (first quartile–third quartile) for continuous variables and frequency and percentage for categorical variables. First uni-variable analyses were performed to compare the group in which the patient contracted an HAI (HAI positive) with patients in the group where no HAI was found (HAI negative). The two groups, HAI positive and HAI negative, were compared using the Wilcoxon rank sum test for continuous variables, and the χ2 test for the categorical variables. The propensity score for the HAI-positive group was calculated for each subject using age, race (white versus nonwhite), sex, presence of hypotension, ISS, GCS, injury type, mechanism of injury, PRBC given in 24 hours, and co-morbidities (alcoholism, chemotherapy for cancer, history of smoking, renal failure, diabetes mellitus, malignant disease, and COPD). Then the one-to-one matching was performed using the nearest neighbor to pair a subject who was HAI positive with a subject who was HAI negative. The propensity score matching was performed using the R package “MatchIt” [13]. The numeric and graphic diagnostics were used to evaluate the improvement. After propensity matching, data were again summarized using summary statistics as described above.
The Wilcoxon signed rank test was used to compare the continuous variables between matched groups. The McNemar test was used to compare the categorical variables between matched groups if the level of a categorical variable was two. If the level of a categorical variable was greater than two, the Stuart-Maxwell test was used [13]. For the length of total hospital stay, ICU days, and ventilator days, the Kaplan-Meier procedure was used to estimate the median time, and the standard error was estimated using the Greenwood formula [14]. Kaplan-Meier curves were generated for the hospital LOS. The log rank test was used to compare the time (Kaplan-Meier curves) between groups. The two-sided p value was reported for each test. A p value <0.05 was considered statistically significant. Statistical analysis was performed using the R language [14].
Results
Uni-variable analysis
Of a total of 806,066 patients, 44,844 (5.56%) developed an HAI while in the hospital. The median age was approximately 53 years. They were predominantly male, and the majority (∼75%) were Caucasian. Blunt trauma was the most common mechanism of injury, found in more than 90% of the patients.
On uni-variable analysis, significant differences were found between the two groups regarding ISS, GCS, and the presence of hypotension at initial evaluation. A higher proportion of patients with a history of diabetes mellitus, COPD, and respiratory disease developed an HAI. A higher proportion of patients in the HAI-positive group received PRBC within 24 hours (Table 1).
Characteristics Associated with Health-care–Associated Infection (HAI) in Trauma Patients
Continuous variables are presented as median with interquartile range (first quartile–third quartile). Categorical variables are expressed as count (percentage) frequency. Health-care–associated groups were compared using Wilcoxon rank sum test for continuous variables and χ2 test for the categorical variables.
PRBC = packed red blood cells; SW = stab wound.
Propensity-matched analysis
After propensity score matching, there was an improvement in the standardized mean difference for the majority of patient characteristics. However, some differences remained, although most were 1 point or less (Table 2). When the two groups were compared on the Abbreviated Injury Scale (AIS) of different body regions, the HAI-positive group was found to have a higher frequency of torso injury than the HAI-negative group, whereas the HAI-negative group sustained brain injury at a much higher frequency than the HAI-positive group (Table 3).
Characteristics of Patients with Healthcare-Associated Infection (HAI) after Propensity Matching
Notes: Continuous variables are presented as median with interquartile range (first quartile–third quartile). Categorical variables are expressed as count (percentage) frequency. Health-care-associated groups were compared using the Wilcoxon rank sum test for continuous variables and the χ2 test for the categorical variables.
PRBC = packed red blood cells; SW = stab wound.
Abbreviated Injury Scale (≥2 Scores) by Body Region after Propensity Matching
Value 1 = present, 0 = absent.
AIS = Abbreviated Injury Scale; HAI = health-care–associated infection.
The patients in the HAI-positive group had a significantly longer LOS—almost three times that of the patients who did not develop an HAI. Similarly, the patients in the HAI-positive group had a significantly longer ICU stay and were on mechanical ventilation longer (Table 4).
Log-Rank Test of Outcomes Using Matched Health-Care–Associated Infection (HAI) Data (n = 44,844)
Kaplan-Meier procedure was used to estimate the median time.
CI = confidence interval.
Further analysis showed that pneumonia, including VAP, was the most common HAI (55.4%), followed by CAUTI (35.8%) among all patients who suffered from any HAI. Ventilator-associated pneumonia was the most common infection in patients who had just one infection and had the longest ventilator days (median [95% confidence interval] 11 [11,12]).
Discussion
In this study, we found that trauma patients with and without HAI were similar in terms of age, race, gender, malignant disease, and smoking status but differed in terms of COPD/respiratory disease, diabetes mellitus, alcoholism, and receipt of PRBCs. This differs from what is documented in the literature with regard to gender and adds to what is known regarding other risk factors [9,15–17]. In this sample, the incidence of HAI was greater (> 5%) in trauma patients than in recent national all-patient estimates (3%) [3].
Similar to the findings of Glance et al. (2011) [9], we found a substantially greater in-hospital LOS in trauma patients with an HAI. These patients had a LOS three times longer, and during that hospitalization spent four times longer in the ICU and seven days on a ventilator compared with zero for patients without an HAI. The finding of a greater number of ventilator days in trauma patients with an HAI is concerning. Regardless of the factor necessitating the mechanical ventilation, such care is the main predictor of VAP. Of all HAI categories, the case fatality rate related to VAP is highest at 14.4% [18]. Although case-specific factors and recommendations for mechanical ventilation are beyond the scope of this paper, clearly, efforts to reduce the number of ventilator day are needed. These might include patient-specific interventions or organizational initiatives such as active rounding, VAP prevention bundles, and aggressive weaning protocols.
There are numerous estimates of cost per hospital day and unit type. Glance et al. conducted an analysis using the Healthcare Cost and Utilization Project Nationwide Inpatient Sample and found that trauma patients with an HAI had inpatient costs approximately double those of patients without an HAI [9]. Given the most recent report available from the Kaiser Family Foundation [19] (2017), which shows the average hospital-adjusted expenses per inpatient day is $2,424 and our finding of a three times longer LOS, at a minimum, one could expect care of a trauma patient with an HAI to cost $7,272 more than that of a patient without an HAI. The per-case estimated cost of VAP (in 2012 dollars) was $40,144 [18]. This does not take into account the severity of illness and ICU treatments and expenses.
Although this paper is not an economic analysis, from using LOS as a surrogate marker of economic impact, it is evident that the burden of HAIs for trauma patients is costly. Whereas the hospital investment in an infection prevention program can be costly, given the number of HAI in trauma patients (44,844 in this series) and the potential impact of LOS ($7,272 per day), the cost of HAIs in this four-year sample would equate to more than $326 million, or $395,279 per participating hospital. Although the excess cost of $100,000 per year for trauma-related HAIs may seem an absorbable expense, of note, trauma patients comprise approximately 4.4% of all inpatient hospital discharges. The investment in a robust infection prevention program, including personnel to provide guidance, recommendations, and interventions for all hospitalized patients, could result in further cost savings related to HAI prevention.
Limitations
The study was performed from the national trauma data repository, and the retrospective nature of the study carries some inherent limitations, including absence of detailed information on the timing of the event of HAI and practice guidelines for the prevention of HAI at the treating facility and selection bias. We were unable to determine if the HAI was related to being ventilated or part of another category or if ventilation preceded or followed the HAI. In order to remove the selection bias, we performed propensity score matching and pair matched analyses. However, propensity score matching does not take into account unobserved variables. Furthermore, the TQIP data set does not provide detailed information on the sterility of the procedures performed at the initial evaluation of the trauma patients, which may have impacted the incidences of some of the HAIs. Finally, despite the use of strict definitions for HAIs in TQIP using data reported to the CDC NHSN system by trained infection prevention and control staff, these data are subject to variability in the application of the classification system.
Conclusions
Health-care–associated infections affected approximately 6% of trauma patients over a four-year period in a national sample; and pneumonia, including VAP, was found to be the most common HAI. This updated epidemiology of trauma patients shows that the clinical HAI burden is higher than past estimates and disproportionate to all patient estimates. The associated economic burden of a significantly longer ICU stay, the need for mechanical ventilation, and tripling the LOS demands responsible administrative policies and support for infection prevention programs and interventions.
Footnotes
Author Disclosure Statement
Dr. Hessels is supported by the Department of Health and Human Services Centers for Disease Control and Prevention National Institute for Occupational Safety and Health Career Development Award 1K01OH011186.
The authors have no conflicts of interest to report.