1 Introduction

Positive end expiratory pressure (PEEP), applied to provide therapy to the lungs by restoring expiratory lung volume and improving alveolar ventilation-to-perfusion relationships, is combined with fractional inhaled oxygen concentration (FIO2) to maintain appropriate arterial oxygenation in patients with respiratory failure [1–4]. In addition to the salutary effects of PEEP and FIO2, there are potentially deleterious effects as well. Applying too much PEEP adversely affects hemodynamic function, i.e., acute decreases in ventricular filling pressures, cardiac output, and mean arterial blood pressure (MAP). FIO2 should be applied in a non-toxic range; >0.50 has been associated with the formation of oxygen free radicals, which are cytotoxic [5, 6]. Also, high FIO2 predisposes to absorption atelectasis [7]. Because of the aforementioned reasons, expert and experienced clinicians should be at bedside directing the application of PEEP and FIO2 to provide appropriate ventilatory support.

In some venues, however, expert clinical personnel may not always be available to make timely, proper bedside clinical assessments for setting PEEP and FIO2. For these situations we describe an arterial oxygenation advisor to provide guidance for setting PEEP and FIO2. It is a decision support, rule-based, advisory system that monitors ventilator and physiologic parameters and provides automatic, real time recommendations for increasing, maintaining, or decreasing PEEP and FIO2 for patients with respiratory failure. Decision support ventilator systems have been developed to assist clinicians in providing appropriate care to patients in setting ventilator parameters in a timely manner [8, 9]. The primary goal of the oxygenation advisor is to achieve and maintain a target arterial oxygenation goal, as reflected by pulse oximeter oxygen saturation (SpO2), by providing recommendations for setting PEEP and FIO2 without applying potentially toxic oxygen concentrations and compromising hemodynamic function.

The purpose of this clinical study was to validate recommendations of the oxygen advisor to those recommendations of attending physicians for setting PEEP and FIO2 for patients with respiratory failure. We hypothesized the advisor would supply recommendations comparable to those chosen by experienced clinicians.

2 Methods

With Institutional Review Board approval, 117 adults were studied; all were intubated (endotracheal tube size range 6.5–8.5 mm internal diameter), diagnosed with respiratory failure from various etiologies (Table 1), and received ventilatory support (Model 840, Puritan-Bennett Pleasonton, CA) with PEEP and FIO2 combined with pressure support ventilation (PSV). At enrollment, patients were hemodynamically stable and provided with appropriate sedation and analgesia as needed to maintain Riker Sedation-Agitation Scale (SAS) score of 4 [10]. Excluded were pregnant patients and those with hemodynamic instability (MAP < 50 mmHg).

Table 1 Patient demographic data
Full size table

The oxygenation advisor consists of a commercially available respiratory monitor (NM3, Respironics, Wallingford, CT) and laptop computer. Data from a combined pressure/flow sensor, positioned between the endotracheal tube and Y-piece of the ventilator breathing circuit, were directed to the respiratory monitor for measurements of pressure, flow, tidal volume (VT), breathing frequency (f), PEEP, and PSV (Fig. 1). These data and ventilator settings were directed to the laptop computer via a cable. SpO2, measured with the aforementioned respiratory monitor, and MAP data from a hemodynamic monitor are entered into the laptop computer. MAP data may be entered manually or in real time from the transduced signal of an arterial catheter (Fig. 1). Computer software (Convergent Engineering, Gainesville, FL) employed a rule-based algorithm using two ventilator variables, PEEP and FIO2 settings, and two physiologic variables, SpO2 and MAP, for formulating recommendations for setting PEEP and FIO2. The oxygenation advisor’s target oxygenation goal is to achieve and maintain SpO2 ≥88 and ≤95 % [11, 12]. The advisor’s operating algorithm, developed by our research team, was based on a rule set agreed upon by an independent group of experienced critical care clinicians (n = 10) (Table 2). To prevent potential bias, conflicts of interest, etc., none of these clinicians, as well as members of the developmental team/authors were involved in the general care or making recommendations for setting PEEP and FIO2 for any patient in this study. The oxygenation advisor was used with clinicians who had no knowledge of its development and methods of operation and were blinded to its recommendations for setting PEEP and FIO2.

Fig. 1
figure 1

The oxygenation advisor functions as a decision support system to provide recommendations for setting PEEP and FIO2. Data from a combined pressure and flow sensor, positioned between the endotracheal tube and breathing circuit, and a pulse oximeter attached to a finger to measure pulse oximeter hemoglobin oxygen saturation (SpO2), are directed to a respiratory monitor (NM3, Respironics) and, in turn, to a laptop computer containing the oxygenation advisor software. Additionally, MAP data, obtained from an arterial catheter via a hemodynamic monitor or manually entered, are directed into the advisor

Full size image
Table 2 Operating rule set for oxygenation advisor is shown
Full size table

Regarding operation of the rule set, initially, the advisor assesses SpO2 input data to determine one of three possible rule subsets to be used, i.e., (A) “Low oxygenation,” (B) “On target oxygenation,” or (C) “High oxygenation” (Table 2). For example, if SpO2 is 85 %, then rule subset A is employed. Next, the advisor assesses the patient’s PEEP, FIO2, and MAP to determine the most appropriate rule in the subset to be applied.

Ranges used for PEEP, FIO2, SpO2, and MAP (Table 2) by the advisor are based, in part, on conventional clinical practice methods [13, 14] and patient safety. The range of PEEP is 5–24 cm H2O and in accordance with the Acute Respiratory Distress Syndrome Network (ARDS Net) [11, 12, 15], a reasonable and not excessively high range for treating adults [7]. Most authorities agree a desirable range for FIO2 is ≤0.50 so that pulmonary parenchyma are not damaged from high cytotoxic concentrations of oxygen [6, 7]. For SpO2, the range 88–95 %, as recommended by ARDS Net is used [11, 12, 15]. Monitoring MAP (minimum acceptable value 65 mmHg) is necessary for patient safety reasons, i. e., to determine the onset of hemodynamic instability if PEEP is increased to inappropriately high levels, compromising cardiac filling pressures and output.

Another patient safety algorithm of the advisor is to alert clinicians when inspiratory plateau pressure or static elastic recoil pressure of the respiratory system (Pplt) >30 cm H2O. Pplt should be maintained ≤30 cm H2O to protect the lungs from ventilator induced lung injury [11, 12, 15]. Routine care of our patients is to apply a volume-controlled (6–8 mL/kg IBW) synchronized intermittent mandatory ventilation (SIMV) breath with a 0.50 s end inspiratory pause, while maintaining prescribed PEEP, FIO2, and PSV settings, at appropriate times to determine Pplt. ARDS Net recommends this should be done every four hours and after each change in PEEP or VT [15].

Attending physicians (n = 8) ordered changes for all ventilator settings for all patients at all times during the study (1 day). They determined PEEP and FIO2 settings based on routinely monitored parameters including current ventilator settings, arterial blood gases, SpO2, breathing pattern, chest X-ray, patient tolerance, MAP, and when available, cardiac output and filling pressure data. At no time were any ventilator settings based on recommendations from the oxygenation advisor.

Recommendations from the advisor for increasing, maintaining, or decreasing PEEP and FIO2 were compared to recommendations/decisions of attending physicians. Attending physicians were first asked to evaluate the patient’s PEEP and FIO2 settings using the aforementioned routine approach, their recommendations were recorded. Simultaneously, recommendations from the advisor were recorded.

It was essential to determine a correlation between PEEP and FIO2 settings recommended by physicians to PEEP and FIO2 settings recommended by the advisor. When physicians ordered an increase or decrease in PEEP, generally, this was an increase by 2 cm H2O or decrease by the same amount from the current setting. When the advisor recommended “increase PEEP”, this was interpreted to mean the same as that generally ordered by physicians, namely, increase PEEP by 2 cm H2O from the current level. For example, if PEEP was 10 cm H2O and the advisor recommended “Increase PEEP,” then this meant increase PEEP to 12 cm H2O. Likewise, when “decrease PEEP” was recommended, this meant decrease PEEP by 2 cm H2O from the current setting. When physicians ordered an increase or decrease in FIO2, generally, this was an increase in FIO2 by 0.10 or decrease by the same amount from the current setting. When the advisor recommended “increase FIO2,” this was interpreted to mean increase FIO2 by 0.10 from the current level. For example, if FIO2 was 0.60, then this meant increase FIO2 to 0.70. Likewise, when “decrease FIO2” was recommended, then this meant decrease FIO2 by 0.10 from the current setting. With this approach, all physician ordered PEEP and FIO2 settings were regressed with recommended PEEP and FIO2 changes from the oxygenation advisor.

Data were analyzed using a Fisher’s exact test, Kappa statistic (K) [16–18], and regression analysis with Bland–Altman analysis [19]. Alpha was set at 0.05 for statistical significance.

3 Results

PEEP settings ranged from 2 to 22 cm H2O; FIO2 settings ranged from 0.30 to 0.65. Most commonly used PEEP and FIO2 settings were 5 cm H2O and 0.40, respectively. Pressure support ventilation settings on the ventilators, also determined by attending physicians during the study, were as low as 2 cm H2O and as high as 25 cm H2O. Pplt ranged from 12 to 44 cm H2O. Two patients had Pplt >30 cm H2O; one was 36 cm H2O and one was 44 cm H2O. MAP ranged from 56 to 143 mmHg; SpO2 ranged from 86 to 99 %.

A total 326 recommendations were made by the oxygenation advisor and attending physicians to increase, maintain, or decrease PEEP and FIO2. Attending physicians agreed with 300 of 326 recommendations, 92 % agreement rate (Table 3). The K statistic, a test of the strength of agreement of recommendations between the advisor and attending physicians, was 0.82 (p < 0.0001), indicating “almost perfect agreement” [16–18]. There was a very significant relationship (p < 0.0001) between recommendations of the oxygenation advisor attending physicians for setting PEEP and FIO2.

Table 3 Recommendations made by the oxygenation advisor compared to recommendations made by attending physicians are shown
Full size table

Regarding recommendations from the advisor and attending physicians as shown in Table 3, for the “maintain PEEP and FIO2” recommendations, 70.55 % were by the advisor and 73.92 % were by physicians. Combining the “increase FIO2” and “increase PEEP” recommendations, 2.15 % were by the advisor, and 2.15 % were by physicians. Combining the “decrease FIO2” and “decrease PEEP” recommendations, 27.3 % were by the advisor and 23.93 % were by physicians.

Relationships for recommendations made by attending physicians versus the advisor for setting PEEP and FIO2 were positive and excellent, measurement bias and precision were negligible, i.e., PEEP: r = 0.98 (p < 0.01), r2 = 0.96, bias = −0.14 cm H2O, precision = 0.87 cm H2O (Fig. 2a, b); FIO2: r = 0.91 (p < 0.05), r2 = 0.83, bias = −0.13 %, precision = 2.31 % (Fig. 3a, b).

Fig. 2
figure 2

a Relationship for recommendations by oxygenation advisor for setting positive end expiratory pressure (PEEP) compared to recommendations by attending physicians for setting PEEP were excellent and highly significant. A positive correlation between the advisor and attending physicians was demonstrated (r = 0.98, r2 = 0.96, p < 0.01). b Corresponding Bland–Altman plot for oxygenation advisor and attending physician recommendations for setting PEEP is shown, Bias: −0.14 cm H2O, Precision (1 SD): ± 0.87 cm H2O, and limits of agreement (2 SD): ± 1.74 cm H2O

Full size image
Fig. 3
figure 3

a Relationship for recommendations by oxygenation advisor for setting fractional inhaled oxygen concentration (FIO2) compared to recommendations by attending physicians for setting FIO2 were excellent and highly significant. A positive correlation between the advisor and attending physicians was demonstrated (r = 0.91, r2 = 0.83, p < 0.05). b Corresponding Bland–Altman data for oxygenation advisor and attending physician recommendations for setting FIO2 is shown, Bias: −0.13 %, Precision (1 SD): ± 2.31 %, and limits of agreement (2 SD): ± 4.62 %

Full size image

4 Discussion

PEEP and FIO2 settings recommended by the oxygenation advisor system reflected the treatment philosophy of attending physicians. Valid, clinically appropriate recommendations for setting PEEP and FIO2 were provided by the advisor, i.e., strength of agreement between the advisor and attending physicians was strong and significant; there were no significant differences in recommendations by the advisor compared to those of attending physicians; and there were excellent, highly significant correlations between PEEP and FIO2 settings ordered by attending physicians versus settings recommended by the advisor. The advisor explained or predicted 96 % of the variance in physician’s recommendations for setting PEEP and 83 % of the variance in physician’s recommendations for setting FIO2.

Nearly three fourths of all recommendations by the advisor and recommendations of attending physicians were to maintain PEEP and FIO2. These recommendations characterized a relatively stable group of patients receiving ventilatory support, not ready to be weaned. Nearly a fourth of all recommendations by the advisor and recommendations of attending physicians were to decrease PEEP and FIO2. These recommendations reflected a group of patients who were in the weaning phases of ventilatory support. The remaining small percentage of recommendations by the advisor and attending physicians was to increase PEEP and FIO2. These recommendations reflected a group of patients requiring increased ventilatory support to promote/restore arterial oxygenation. The advisor’s combined recommendations were congruent with recommendations of attending physicians for these three groups of patients.

Although attending physicians disagreed with 8 % of recommendations, this was not considered clinically significant because K was large, meaning “almost perfect agreement,” and correlations between the advisor and attending physicians for setting PEEP and FIO2 were excellent. About this very low percent of disagreement, we observed no individual situations or recurring patterns when recommendations from the oxygenation advisor were so different from recommendations of attending physicians that it may have predisposed to adverse clinical consequences in any study patient. Additionally, no overall discernible or significant clinical pattern of disagreement was observed. At times there may have been philosophical differences between the thinking of some attending physicians and the advisor’s rule set. Or, some physicians may have misinterpreted a patient’s physiologic data/symptoms; for example, inferring FIO2 should be maintained as opposed to decreasing it as recommended by the advisor.

Two “safety nets” integrated into the oxygenation advisor are for hemodynamic safety and lung protection safety. Concerning hemodynamic safety, for example, for one patient receiving PEEP at 15 cm H2O, MAP acutely decreased from 80 to 50 mmHg. In this situation, the oxygenation advisor is programmed to alert the clinician of this potentially unsafe hemodynamic condition prompting a patient assessment so that appropriate interventions can be made. Regarding lung protection safety, two patients had Pplt >30 cm H2O. Following an increase in PEEP to improve arterial oxygenation Pplt increased from 28 to 36 cm H2O for one of these patients. Under this condition the advisor is programmed to alert with the message “Although increasing PEEP may improve SpO 2 , Pplt is >30 cm H 2 O.” Accordingly, this prompts a patient assessment, as well as intervention to decrease Pplt, i.e., decrease VT by 1 mL/kg steps (minimum VT 4 mL/kg) until Pplt ≤30 cm H2O [15].

The SpO2 range for the arterial oxygenation goal (88–95 %), Pplt limit for the lung protection goal (≤30 cm H2O), and MAP range for the minimal hemodynamic function goal (65 mmHg) used in this study may be considered as default settings for the oxygenation advisor. We employed the above SpO2 and Pplt values because these values were used in large clinical trials and deemed appropriate by ARDS Net [11, 12]. Because not all patients requiring ventilatory support have advanced forms of ARDS, clinician users would have options for adjusting these ranges upward or downward in accordance with their clinical practice standards for treating a variety of patients with varying degrees of respiratory failure. For example, rather than a minimum SpO2 oxygenation goal of 88 %, some may prefer 90 or 92 % as a safer minimal goal. Similarly, depending on local practice standards of attending physicians and their clinical experience, some may consider a safe minimal goal for MAP to be 70–75 mmHg.

Our oxygenation advisor is a rule-based system that automates the heuristics of the clinician. It functions as an advisory decision support system by making recommendations for applying PEEP and FIO2 that augment or support patient care management decisions. Ventilator decision support systems similar in operation to the oxygenation advisor have been described. Tehrani [8, 9] described a decision support advisory system to provide clinical guidance. It employs an algorithm for computing optimal ventilator parameters including PEEP, FIO2, minute ventilation, peak inflation pressure (PIP), and inhalation-to-exhalation time ratio. Belel et al. [20] employed an advisory system for ventilating neonates. SpO2, respiratory waveforms, heart rate, transcutaneous PaO2 and PaCO2, MAP, body temperature, ventilator inspiratory time, expiratory time, PIP, PEEP, mean airway pressure, and FIO2 were feedback variables used for formulating recommendations for setting the ventilator to treat ventilation and oxygenation abnormalities. Clinicians agreed with the ventilation recommendations 91 % of the time and with the oxygenation recommendations 94 % of the time. Rees et al. [21] presented a decision support system for optimizing mechanical ventilation by employing mathematical models of oxygen transport, carbon dioxide transport, and lung mechanics combined with penalty functions describing clinical preferences toward the goals and side-effects of mechanical ventilation. This approach was determined to be appropriate for describing patient physiologic data and suggesting optimal ventilatory strategies. Rutledge et al. [22] described a qualitative and quantitative ventilator management advisor (VentPlan) to provide recommendations for setting a ventilator based on a mathematical model of cardiopulmonary physiology that were in agreement with physician preferences for setting the ventilator. Kwok et al. [23] described an adaptive neuro-fuzzy inference system (ANFIS) FIO2 ventilator advisor that estimated intrapulmonary right-to-left shunt by employing a respiratory index, derived in part from calculating the alveolar air equation. In turn, clinical advice is given on an FIO2 needed to attain a target PaO2 for patients with compromised pulmonary function. Fuzzy logic has also been used to create advisor-type algorithms. Computer-based decision support tools employing fuzzy logic are suggested to automate aspects of physician decision making in the ICU, for example recommendations and decisions for patient care [24].

Potential clinical advantages of the oxygenation advisor are that it is automatic and continuously operational; important in clinical situations where expert clinicians may not always be available for patient assessments and timely decisions for setting PEEP and FIO2. Examples of this might include large multi-patient ventilatory care facilities/nursing homes, battlefield military hospitals, mass casualty treatment centers, and large ICUs. In such venues with limited manpower, the advisor may be able to assist in setting PEEP and FIO2 appropriately.

A potential limitation of the study is that recommendations for setting PEEP and FIO2 from the oxygenation advisor were compared to recommendations from attending physicians in a surgical intensive care unit. Additional clinical use of the oxygenation advisor from a broader population of attending physicians and patients, including patients with chronic forms of respiratory failure, may offer additional evidence to assess its use.

In summary, the oxygenation advisor provided valid recommendations for applying PEEP and FIO2 to patients with acute respiratory failure and receiving ventilator support that were in accordance with the treatment philosophy of experienced clinicians.