Automated Alerting and Recommendations for the Management of Patients with Preexisting Hypoxia and Potential Acute Lung Injury: A Pilot Study

ACUTE lung injury (ALI) is a devastating condition with significant mortality. The clinical syndromes of ALI and acute respiratory distress syndrome (ARDS) are defined by descriptive clinical findings, regardless of the specific etiology of acute pulmonary dysfunction. The American–European Consensus Committee in 1994 defined clinical ALI as requiring, (1) respiratory failure of acute onset, with a ratio of partial pressure of arterial oxygen (Pao₂) to inspired fraction of oxygen (Fio2) (Pao₂/Fio2 or P/F) less than 300 mmHg (regardless of the level of positive end-expiratory pressure [PEEP]), (2) bilateral infiltrates on frontal chest radiograph, and (3) a pulmonary capillary wedge pressure less than 18 mmHg (if measured) or no evidence of left atrial hypertension.¹  ARDS was defined identically, except for a lower limiting value of less than 200 mmHg for Pao₂/Fio₂.¹  ALI and ARDS affect a large number of patients and have a poor prognosis. The incidence of ALI/ARDS has been variably reported to be 50,000–190,000 cases per year in the United States.^1–7 

The only ventilator intervention to date that has clearly demonstrated a survival benefit in controlled studies in adults with ARDS has been the adoption of a low tidal volume ventilation strategy (6 ml/kg predicted body weight [PBW]) with plateau pressures less than 30 cm H₂O.⁸  Additionally, the use of recruitment maneuvers, higher levels of PEEP, and judicious administration of fluids have been shown to reduce ventilator days, and increase Pao₂, while avoiding potential hyperoxia-induced lung injury.^9–12  Together, these methods comprise a lung-protective ventilation strategy (LPVS). A number of animal and clinical studies have shown that resorting to a nonprotective strategy, by adoption of high tidal volumes or higher plateau pressures, can result in serious lung injury and in some cases, parallel the course of patients with severe ARDS.^13,14 

Despite dissemination of this management strategy, we have recently found that LPVSs were not specifically undertaken in patients with ALI in the operating room setting.¹⁵  Of 1,286 patients who underwent procedural anesthesia with a Pao₂/Fio2 ratio less than 300, 242 met the criteria for diagnosis of ALI before the procedure. Intraoperative lung-protective ventilator management was not routinely performed on these patients. Patients, on average, received 8.5 cc/kg PBW ventilation and approximately 5 cm H₂O of PEEP with high Fio₂. Typically, these patients had higher Pao₂ and tidal volumes in the operating room compared with the preanesthetic setting.

In the current study, we tested the hypothesis that information to anesthetic providers of recommended ventilation strategy for potential ALI in patients with low preoperative P/F ratios would result in an increased use of lower tidal volumes and higher PEEP consistent with a LPVS.

Institutional Review Board approval was obtained for this randomized, open-label, parallel arm clinical trial from the Institutional Review Boards of the University of Michigan Medical School (Ann Arbor, MI), under waiver of consent without use of a data safety monitoring board, due to the low-risk nature of the study. Figure 1 depicts the flow of patients through the trial. The study was conducted at a single site, the University of Michigan Medical Center. Patients were recruited using a “just-in-time” automated anesthesia information management system (Centricity, Wishahaka, WI) script, which was created to screen all patients undergoing an anesthetic at the time of the start of anesthesia or entry into the operating room, with central laboratory arterial blood gases collected in the past 2 days. From these arterial blood gases, the average P/F ratio was calculated for each patient. If the patient was 18 yr old or more, the average P/F ratio was less than 300 or the last P/F ratio was less than 300, and a valid history and physical, including the patient’s height, was present. Patients were enrolled and randomized, using a real-time electronic pseudorandom number generator to intervention or no intervention (control) in a 1:1 allocation ratio, to enroll two equal groups of 50 patients. No system of stratification was used during randomization. If randomized to intervention, the height of the patient was obtained from the electronic history and physical, and the PBW was calculated. An automated text page was sent to the attending anesthesiologist’s and/or in-room provider’s hospital pager, depending on “signed in” status at either start of anesthesia, or at the time the patient was in the operating room, stating:

“Patient: LAST, FIRST has a P/F ratio consistent with acute lung injury. If this patient has ALI, recommended tidal volume is <XXX>cc. Recommended PEEP and Fio2 are shown on the Web site.”^††

There was no protocol-based care in this study. All clinical decisions were left to the discretion of the anesthetic care team, who could decide whether to follow the recommendations of the automated alert. The anesthesia machine used in this study was the Aisys (General Electric Healthcare, Milwaukee, WI). This machine provides basic volume control ventilation, in addition to advanced pressure control modes, pressure control mode with a volume guarantee, pressure support, and inverse ratio ventilation. Only basic ventilator information was collected by the anesthesia information management system, including delivered tidal volume, PEEP, peak inspiratory pressure, respiratory rate, and Fio2. No recommendation on ventilator mode was included in the alert, and the ventilator mode was not recorded.

The study was intended to demonstrate the superiority of the alert group, with a reduction in the total cc/kg PBW administered. The primary outcome measure was median tidal volume in cc/kg PBW between the two groups, from surgical incision to dressing completed. Secondary outcomes included median total tidal volumes, PEEP, and peak inspiratory pressure. Finally, we calculated what would have been the recommended LPVS tidal volumes for the patients not receiving the alert (control group) and compared these tidal volumes with the actual tidal volumes administered, to determine the mean difference. After the completion of enrollment and initial statistical analysis, the preoperative anesthetic records were reviewed for common conditions, which have been associated with increased morbidity and mortality. Additionally, preoperative chest radiographs were examined, if available, to look for the presence of bilateral infiltrates that would be consistent with a diagnosis of ALI. Radiograph review was completed using a consensus process by the investigators while they were blinded to additional information about the patients. Finally, the hospital course of each patient was examined for postoperative complications, which could be associated with low tidal volume ventilation, including prolonged intubation, stroke/herniation, myocardial infarction, new onset renal failure, and death.

The study was designed with a sample size of 50 patients in each group to have 80% power to detect a difference in means of 1.0 cc/kg PBW, assuming that the common SD was 1.00 cc/kg PBW, using a two-sided independent t test (α = 0.05). Comparison between groups was made with SPSS version 18 (SPSS Software, Chicago, IL) and R version 2.16 (R Foundation for Statistical Research, Vienna, Austria), using the two-tailed Student’s t test and/or Mann–Whitney U test for continuous variables or chi-square analysis for dichotomous variables. P values less than 0.05 were considered to be statistically significant.

We successfully enrolled and collected data on 100 patients (fig. 1). The study enrolled from July 8, 2010 until February 21, 2011. Twenty-two patients were excluded from the dataset after randomization due to lack of ventilator data capture. No other data were missing in the dataset that was collected. There were no statistically significant differences between comorbidities in the control and intervention groups (table 1). For the 50 patients enrolled in the intervention group, the average recommended tidal volume was 390 cc. The mean PBW in both populations was 65 kg. The intervention group received lower total tidal volumes (458 vs. 508 cc; P = 0.033) and lower tidal volumes per kilogram of PBW (7.19 vs. 7.97 cc/kg; P = 0.040; fig. 2). The distribution of the low tidal volume cc/kg PBW was not entirely normal, as such we examined it using both the Mann–Whitney U test and t test, which provided P values of 0.040 and 0.058, respectively.

There was no statistical difference in the PEEP or peak inspiratory pressure between the two groups. These data are summarized in table 2 and Figure 2. Patients in the intervention group received total tidal volumes closer to the recommended tidal volumes than those in the intervention group (mean difference 67 cc [95% CI, 33–101] vs. 117 cc [95% CI, 81–153]; P = 0.046). The absolute value of the difference in patients who received the alert was 109 cc (95% CI, 86–133), which was lower than the absolute value of 147 cc (95% CI, 121–173) in those patients who did not receive the alert (P = 0.034). There were also more patients in the intervention group who received tidal volumes less than 6.5 cc/kg PBW (17 patients vs. 8 patients; P = 0.065), although this did not meet statistical significance. On review of all cases, there was no statistically or clinically increased rate of complications associated with the intervention group. This included a statistically identical number of postoperative ventilator days, strokes, myocardial infarctions, and deaths. These data are further illustrated in table 2.

On review of the initial dataset, three patients were found to have profoundly low tidal volumes of less than 3 cc/kg PBW, and three were found to be of The American Society of Anesthesiologists Physical Status 6. A sensitivity analysis was completed excluding these patients from the dataset (tables 3 and 4). Overall, there were no statistically significant changes in the results.

In this prospective randomized controlled trial, an automated intraoperative notification alerting clinicians to the potential presence of ALI/ARDS led to statistically significant reduction in delivered total median tidal volumes and in volumes measured in cc/kg PBW. There were no differences in the PEEP between the two groups. This is the first trial that we are aware of that changed clinician behaviors in patients with possible ALI in the perioperative environment.

ALI is an established syndrome that has profound effects on numerous patients each year. The incidence of ALI in two recent studies has been estimated at 22–86 cases per 100,000 persons per year,^5,7  with 40–43% of these patients having ARDS.⁷  Survival statistics for patients with ALI/ARDS vary with lung injury etiology and age, but overall mortality rates in both adult and pediatric patients remain substantial at 20–50%, despite sophisticated intensive care.^{2,3,5–7,16–19}  The significance of distinguishing between the two clinical syndromes in a practical sense is uncertain, because a meta-analysis of 102 studies before 1996 showed little or no difference in mortality rates between patients meeting criteria for ALI compared with ARDS.²⁰  This was also the conclusion in the recent NEJM article by Rubenfeld et al.,⁷  which reported mortality rates of 38.5% for ALI and 41% for ARDS, with an estimated 74,500 deaths per year, and an aggregate 3.6 million hospital days of care in the United States.

Over the past 2 decades, a variety of interventions and intensive care strategies have been used in treating patients with ALI/ARDS. Historically, ALI/ARDS was treated with large tidal volumes and high peak pressures, in an attempt to improve oxygenation. However, it has been shown that such efforts actually add to ventilator-induced lung injury through multiple mechanisms, including alteration of the pulmonary cytoskeleton, disturbed alveolar fluid balance, and increased inflammatory response.^21–25  Current standard of care for ARDS includes mechanical ventilation, with adoption of lung protective strategies, judicious fluid management, adjunct nutritional support, and more importantly the diagnosis and treatment of the underlying cause. Despite this care, the only multicenter randomized controlled evidence for reducing ARDS-associated mortality consists of a reduction in tidal volumes from 12 to 6 cc/kg PBW and the administration of neuromuscular blockade for 48 h after the onset of the syndrome.²⁶ 

The application of relatively low tidal volumes during anesthesia is not a new concept and has been advocated in thoracic anesthesia, where ALI is a devastating and common diagnosis. In work by Licker,²⁷  it was determined that increased tidal volumes were associated with the development of ALI. Subsequent work showed that protocol-based care, with reduced tidal volumes in thoracic oncologic surgery during one-lung ventilation, reduced the incidence of ALI from 3.7 to 0.9%.²⁸  However, Licker’s prevention trial was conducted in nonrandomized patients without ALI, undergoing only a single-type procedure and did not address the management of patients with preexisting ALI.

Recently, we have shown that patients meeting criteria for ALI preoperatively are managed in a similar format to those who undergo anesthesia with hypoxia for another reason.¹⁵  This care frequently mirrors the care provided in the intensive care unit before their procedure, with an important exception that patients receiving LPVS with tidal volumes less than 6 cc/kg PBW, received significantly increased tidal volumes intraoperatively. Although the reason for this difference in practice is unknown, the current study shows that an automated alert triggered by P/F ratios can increase compliance with a LPVS for patients who may have ALI. This is especially important for patients undergoing nonthoracic surgeries, where ALI may not be a foremost concern for anesthetic providers and is likely underdiagnosed.

We have demonstrated that a single pager notification that simply suggests the institution of a LPVS can impact provider behavior regards to intraoperative management of patients at risk for ALI. Notifications, such as those in this pilot study, may be an effective way to intervene and reduce the potential for further injury in patients with ALI. However, it is important to allow clinicians the flexibility to selectively implement a set of recommendations rather than protocolize care because situations where it is either impractical, or unsafe to follow a set of predefined recommendations may occur. For example, in a patient at risk for ALI with underlying pulmonary dysfunction as well as concomitant head injury and increased intracranial pressures, low tidal volumes with permissive hypercapnea may be contraindicated. These complex clinical situations may have contributed to the nonadherence to strict LPVS in this study. For example, this may be the reason for lower PEEP values than were recommended by the table on the Web site. As increasing PEEP may cause hemodynamic compromise, providers may have been concerned about applying PEEP above a certain level in these potentially critically ill patients.

Our study has several limitations, which are important to discuss. First, the trial was undertaken at a single center with a relatively small number of patients. These results and the changes in provider behavior may not be generalizable to other settings. In particular, the study relied on the relatively advanced anesthesia information management system architecture in place at the study hospital, although most centers with an anesthesia information management system should be able to replicate the notifications. This is notable, as this study design introduces a new method of studying the impact of low-risk clinical reminders on care in ALI and other conditions, which is consistent with the highest level of medical evidence, the randomized controlled trial. This minimizes the concerns frequently associated with pre- and postanalysis that is common in quality improvement literature.

Second, in this study, we enrolled patients based on only one of the three criteria for ALI; because the study was performed with a “just-in-time,” automated screening and enrollment strategy that would be replicable at other centers, it was not possible to obtain intravascular volume status or chest radiograph information to confirm a diagnosis of ALI before the anesthetic. However, previous work has shown that lower tidal volumes can also help prevent the development of ALI.^27,28  Therefore, interventions that lower tidal volumes may be beneficial to any patient with a low P/F ratio, regardless of whether the other criteria for ALI are met.

Finally, Although there was an attempt made to determine whether the outcomes were different between the two groups, the study was distinctly underpowered to detect a subtle increase in complications from implementation of the reminder intervention. Further studies assessing the impact of these types of interventions are therefore, warranted.

In conclusion, we have demonstrated that anesthesiologists will modify their intraoperative ventilation strategies when notified about the potential risk for ALI underlying preoperative hypoxia. This relatively simple alert system is a potentially useful tool to bring provider behavior more in-line with the most recent standards of care from the literature. It would be reasonable to undertake a multicenter prospective trial in the future to determine the influence this type of notification can have on morbidity and mortality.