Abstract
BACKGROUND: Several markers of oxygenation are used as prognostic markers in acute hypoxemic respiratory failure. Real-world use is limited by the need for invasive measurements and unreliable availability in the electronic health record. A pragmatic, reliable, and accurate marker of acute hypoxemic respiratory failure is needed to facilitate epidemiologic studies, clinical trials, and shared decision-making with patients. FIO2 is easily obtained at the bedside and from the electronic health record. The FIO2 trajectory may be a valuable marker of recovery in patients with acute hypoxemic respiratory failure.
METHODS: This was a historical cohort study of adult subjects admitted to an ICU with acute hypoxemic respiratory failure secondary to community-acquired pneumonia and/or ARDS.
RESULTS: Our study included 2,670 subjects. FIO2 and SpO2 were consistently more available than was PaO2 in the electronic health record: (FIO2 vs SpO2 vs PaO2 : 100 vs 100 vs 72.8% on day 1, and 100 vs 99 vs 21% on day 5). A worsening FIO2 trajectory was associated with reduced ventilator-free days. From days 2 to 5, every increase in FIO2 by 10% from the previous day was associated with fewer ventilator-free days (on day 2: adjusted mean –1.25 [95% CI –1.45 to –1.05] d, P < .001). The SpO2/FIO2 trajectory also provided prognostic information. On days 3 – 5, an increase in SpO2/FIO2 from the previous day was associated with increased ventilator-free days (on day 3: adjusted mean 2.09 (95% CI 1.44–2.74) d; P < .001). SpO2/FIO2 models did not add predictive information compared with models with FIO2 alone (on day 2: adjusted FIO2 vs SpO2/FIO2 R2 0.122 vs 0.119; and on day 3: 0.153 vs 0.163).
CONCLUSIONS: FIO2 and SpO2/FIO2 are pragmatic and readily available intermediate prognostic markers in acute hypoxic respiratory failure. The FIO2 trajectory in the first 5 d of ICU admission provided important prognostic information (ventilator-free days). Although the SpO2/FIO2 trajectory was also associated with ventilator-free days, it did not provide more information than the FIO2 trajectory alone.
- acute hypoxic respiratory failure
- electronic health record
- hypoxia
- pneumonia
- acute respiratory distress syndrome
- prognosis
Introduction
Acute hypoxic respiratory failure is a common diagnosis that leads to ICU admission.1,2 In patients with acute hypoxic respiratory failure, the disease course is variable and unpredictable, ranging from rapid improvement or decline to slow improvement or decline, or to persistent disease that becomes chronic.3 Our ability to objectively identify the changing disease trajectory day to day in clinical practice and for research purposes is limited in part by a lack of an accurate, reliable, and pragmatic intermediate pulmonary physiologic marker. Intermediate markers are biologic measurements, signs, or symptoms that outwardly reflect the underlying disease process.4-6
An accurate, reliable, pragmatic pulmonary physiologic marker would improve early identification of the clinical course of patients with acute hypoxic respiratory failure. This will provide clinicians with timely information to guide shared decision-making and assist with evolving patient and family counseling. Intermediate markers can also be used as surrogate end points in clinical trials. Compared with traditional end points, for example, mortality, surrogate end points occur earlier in the disease course and thus clinical trials powered for surrogate end points require fewer participants, shorter study duration, and hence cost less.5,7 In addition, as we continue to embark on the era of “big data,” there is a need to identify an intermediate marker for acute hypoxic respiratory failure that can be reliably and accurately extracted from the electronic health record. This marker should be applicable to all patients with acute hypoxic respiratory failure, whether they require mechanical ventilation or another modality of supplemental oxygen delivery.
Several pulmonary physiologic markers have been studied in acute hypoxic respiratory failure.8 and /FIO2 are possibly the most widely used measures. It has been shown to correlate with disease severity in patients with acute hypoxic respiratory failure secondary to ARDS and is often used in clinical research to assess treatment response.9 It is often used in clinical trials to stratify subjects based on disease severity, as an intermediate pulmonary physiologic marker that defines the disease course, or as a surrogate end point, which indicates response to treatment. /FIO2 is intended as an intermediate pulmonary physiologic marker of oxygenation but can be unreliable for this purpose. Studies have demonstrated variability in / with a change in PEEP, , oxygen consumption, and other factors.10-13
Moreover, intrinsic measurement bias exists: many patients do not undergo arterial blood gas analysis, especially if they are not on mechanical ventilation or are outside of the initial hours of ICU admission. The oxygenation index attempts to resolve some limitations of the / measurement by incorporating the mean airway pressure but, again, with varying results.8,14,15 The oxygenation index is also only applicable to patients who are intubated and requires documentation of the mean airway pressure in the electronic health record for pragmatic use. Other physiologic parameters that have been evaluated include the oxygen saturation index, dead space fraction, and lung injury score.16,17 As with / and the oxygenation index, the utility of these markers is limited by their requirement for invasive measures, radiation-generating procedures, complex calculations, and unreliable availability in the electronic health record, and/or only apply to a small subpopulation of patients with acute hypoxic respiratory failure.
Oxygen saturation and / have been shown to correlate with / in predicting and diagnosing ARDS.18-22 It has been validated as a replacement marker for / when determining disease severity in adult and pediatric patients with a critical illness and acute hypoxic respiratory failure.23-25 Clinical characteristics and outcomes such as mechanical ventilation duration and ICU and hospital length of stay are similar between those diagnosed with ARDS when using / as opposed to /.16 Both and are easily accessible at the bedside (regardless of the oxygen delivery device) and in the electronic health record. In hospitalized patients treated with supplemental oxygen, is relatively constant. The patient care team tightly regulates within a narrow range of usually >90%. This is particularly true in the ICU, where monitoring is continuous and nursing and respiratory therapy staff support is readily available. Variations within this range of would not significantly alter the /; therefore, the more illustrative variable within / is alone. In this study, we hypothesized that the change in (ie, the trajectory) alone is associated with the clinically relevant patient outcome of ventilator-free days. We also hypothesized that the trajectory would be at least as reliable as / as an intermediate physiologic marker in acute hypoxic respiratory failure.
Quick Look
Current Knowledge
Although several physiologic markers of oxygenation have been studied, their utility is limited by cost, the risk related to invasive technique or radiation, complex calculations, and/or a lack of accessibility, particularly from the electronic heath record. / is one example of a commonly used marker used clinically to monitor the disease course; however, it has not been validated for this purpose and is unreliable as a prognostic marker in patients with higher needs.
What This Paper Contributes to Our Knowledge
The trajectory (the change in concentration from the previous day) was easily accessible at the bedside and from the electronic health record. In addition, on days 2 – 5, it was associated with the patient outcome of ventilator-free days and can be used as an early (or intermediate) marker to predict patient outcomes.
Methods
Design, Subjects, and Definitions
We performed a historical cohort study by using a convenience sample of 2,771 adult subjects (>18 y old) admitted to the Mayo Clinic–Rochester ICU with community-acquired pneumonia and/or ARDS between January 1, 2009, and June 30, 2014. Development of this cohort was previously described.26 We included all patients with acute hypoxic respiratory failure. Acute hypoxic respiratory failure was defined as requiring supplemental oxygen by any modality at any point within the first 5 d after ICU admission. Community-acquired pneumonia was defined as per the 2007 Infectious Disease Society of America/American Thoracic Society guidelines27 on community acquired pneumonia and identified by the International Classification of Disease, Ninth Edition codes 481–486. ARDS was defined by the 2012 Berlin criteria9 and identified through manual chart review by two independent reviewers. The Mayo Clinic Institutional Review Board reviewed and approved this study, and a waiver of consent was granted. Per Mayo Clinic policy, subjects who had previously declined the use of their medical record data were excluded from this study.
Data Collection
Data were collected by using the METRIC Data Mart System which is an electronic health record based data warehouse with near real-time data input. The development, structure, and data collection methods of this system were previous described.28 The automated system was supplemented by manual review as needed. Data were collected from the first patient admission. Subsequent admissions were omitted. We collected baseline data, which included demographics, comorbid conditions (Charlson comorbidity index), admission diagnoses, illness severity as measured by APACHE (Acute Physiology and Chronic Health Evaluation) III and Sequential Organ Failure Assessment (SOFA) score, and the respiratory failure severity as measured by respiratory SOFA. For those subjects who met the inclusion criteria, all available measures of , , and available within the 5-d measurement period were collected. Oxygen delivery devices (nasal cannula, simple face mask, and high flow nasal cannula) were documented in the electronic health record in L/min. The widely accepted conversion of 4% per liter of flow was used to convert from L/min to percent .18,29
Statistical Analysis
Data were summarized by using counts and percentages for categorical variables, and medians (interquartile ranges) for continuous variables. Differences across levels at ICU admission, were categorized into tertiles of <0.35, 0.35–0.49, 0.50 and were compared by using chi-square tests for categorical variables (sex, race, ethnicity, presence of community-acquired pneumonia, ARDS, COPD, asthma, respiratory SOFA, invasive ventilator and noninvasive ventilator use, hospital mortality, 30-d mortality, and 1-year mortality) and the Kruskal-Wallis test for continuous variables (age, APACHE III, SOFA, invasive ventilator days, noninvasive ventilator days, hospital length of stay, and ICU length of stay).
The primary outcome for this study was ventilator-free days. Ventilator-free days were defined as the number of days that the subject was alive and not requiring invasive mechanical ventilation within 28 d of the study day. Associations between the daily measures and ventilator-free days were analyzed on ICU day 1 through day 5 by using linear regression models. For day 1, the value was the average over the first 6 h of day 1 of the ICU admission. For subsequent days, the 6-h average of the last proceeding day was used. For example, on day 2, the average from hours 18 to 24 were used. For each model, we performed unadjusted univariate analyses as well as adjusted for the a priori clinically relevant adjustment variables of age, sex, Charlson comorbidity index, respiratory SOFA, and day 1 (for days 2 – 5). For days 2 through 5, we analyzed the association of on the current day and from the previous day with ventilator-free days. The same adjustment terms were used for these models. These analyses were repeated by using daily values of / for comparative purposes. The adjusted R2 was reported for each linear regression model to compare the performance of models with measures alone to measures with /. All analyses were performed in SAS version 9.4 (SAS Institute, Cary, North Carolina). Because we examined daily on 5 separate days, we used a Bonferroni correction to account for multiple comparisons. A 2-sided P < .01 was considered statistically significant.
Results
A total of 2,670 subjects met our inclusion criteria. One hundred one patients were excluded due to lack of supplemental oxygen requirement. Subject demographics are presented in Table 1 and are categorized by baseline (day 1) . The subjects in the 0.35–0.49 or >0.50 category of baseline requirement tended to be younger, have higher incidence of ARDS, and a lower incidence of community-acquired pneumonia, lower Charlson comorbidity index score, lower presence of COPD and asthma, and higher SOFA, APACHE III, and respiratory SOFA scores. Compared with the subjects in the <35% baseline category, a higher percentage of the subjects in the 0.35-0.49 and >0.50 categories required invasive ventilation (37.4, 73.7, and 78.9%, respectively; P = .001) and noninvasive ventilation (26.3, 48.5, and 42.7%, respectively; P = .001) during their ICU stay. The subjects in the 0.35-0.49 and >0.50 baseline categories also had worse clinical outcomes, including fewer ventilator-free days (28.0 vs 26.9 and 24 d; P = .001); higher hospital (40.7% vs 43.6% and 59.3%; P = .001), 30-day (43.3% vs 44.1% and 54.8%; P = .001), and 1-year mortality (70.1% vs 66.0% and 74.8%, P = .001); and longer hospital (6.2 vs 7.0 and 8.5 d, P = .001) and ICU (1.2 vs 2.2 and 3.8 d, P = .001) stays.
The daily percentages of the subject population by category on day 1 – 5 are displayed in Figure 1. The percentage of the subjects in the high and mild-moderate requirement categories decreased on each consecutive day. On day 1, 34.1% of the subjects had a high requirement and 65.1% had a mild-moderate requirement. On day 5, 13.2% of the subjects had a high requirement and 24.3% had a mild-moderate requirement. Conversely, the percentage of the subjects on room air increased exponentially from day 1 to 5, with 0.8% on day 1 and 62.5% on day 5. The mean daily value and the number of subjects with available data for the variables of interest (, , , and /) are presented in Table 2. The mean decreased and the mean / increased from day 1 to 5, whereas the mean and remained unchanged. In addition, the number of subjects with available and was considerably higher compared with those with , particularly later in the illness course.
The association between the change in daily and / and ventilator-free days are shown in Tables 3 and 4. In Table 3, the unadjusted increase in on days 2 – 5 compared with the previous day showed a significant decrease in ventilator-free days: on day 2: –1.25 (–1.45 to –1.05) d, P = .001; on day 3: –1.39 (–1.61 to –1.17) d, P = .001; on day 4: –1.55 (–1.81 to –1.28) d, P = .001; on day 5: –1.77 (–2.08 to –1.47) d, P = .001 for each 10% increase in from the previous day. When adjusted for the confounding variables of age, the Charlson comorbidity index, and respiratory SOFA but not of sex, the association remained significant. Similar results were seen when analyzed by the category, as depicted in Figure 2. In Table 4, the unadjusted increase in / on days 3 – 5 compared with the previous day showed a significant increase in ventilator-free days: on day 3: 2.09 (1.44–2.74) d, P = .001; on day 4: 1.38 (0.61–2.16) d, P = .001; on day 5: 1.46 (0.56–2.36) d, P = .001 for each 100-point increase from the previous day. The association between ventilator-free days and the change in / on day 2 compared with day 1 was not significant (on day 2: 0.10 (–0.27 to –0.47) d, P = .60). Again, when adjusted for the confounding variables, the findings on days 3–5 remained significant. As seen in Supplementary Table 1 (see the supplementary materials at http://www.rcjournal.com), the increasing baseline concentration was also associated with significantly fewer ventilator-free days with unadjusted increase, for each 10% increase in , of –1.06 (–1.23 to 0.88) d, P = .001. The predicted change in ventilator-free day for each 10% change in compared with the subjects on >50% on the day of interest and the day before are illustrated in Supplementary Tables 2–5 (see the supplementary materials at http://www.rcjournal.com).
As a sensitivity analysis, we evaluated the association among , /, and ventilator-free days only in those subjects who required invasive mechanical ventilation. A similar association was seen: trajectory: –1.89 (–2.24 to –1.54) d, P = .001 on day 2; –1.16 (–1.54 to –0.78) d, P = .001 on day 3; –1.79 (–2.29 to –1.30) d, P = .001 on day 4; and –1.81 (–2.34 to –1.28) d, P = .001 on day 5; and /: 2.71 (1.11 to 4.31) d, P = .001 on day 2; 1.46 (0.21-2.71) d, P = .02 on day 3; 0.59 (–1.11 to 2.29) d, P = 0.49 on day 4; and 0.32 (–1.54 to 2.18) d, P = .74 on day 5 (Supplementary Tables 6 and 7 [see the supplementary materials at http://www.rcjournal.com]).
Discussion
Our results showed that the trajectory was associated with ventilator-free days and performed as well as / as a prognostic marker. Although / was previously evaluated as a marker of oxygenation, to our knowledge, our study was the first to assess trajectory alone.30,31 Our study also demonstrated the following key findings: (1) was more readily available than ; (2) a higher baseline was associated with increased illness severity and worse outcomes; (3) was the illustrative variable within the /, and the trajectory correlated with the disease course. Unlike other pulmonary physiologic markers, the trajectory is readily determined at the bedside and through the electronic health record, and can be used to facilitate pragmatic clinical trials in subjects with acute hypoxic respiratory failure.
Results from our study showed that, in a population of subjects with acute hypoxic respiratory failure secondary to community-acquired pneumonia or ARDS, higher baseline alone was associated with increased severity of illness and worse outcomes, including increased invasive and noninvasive ventilator use, mortality, hospital and ICU length of stay, and fewer ventilator-free days. This was consistent with a previous study, which demonstrated that baseline is an independent predictor of mortality.32 Our results showed that daily mean concentration decreased and / increased over time, whereas the daily mean and remained largely unchanged. This highlighted as the illustrative variable, which changed over the course of the disease as opposed to and , which remained relatively constant. The more novel result from our study was the association of the trajectory with the patient important outcome of ventilator-free days. In our study, an increase in the concentration by ≥ 10% from the previous day was significantly associated with fewer ventilator-free days. / demonstrated a similar trend in that an increase in / was associated with more ventilator-free days. Interestingly, the R2 value, which depicts the strength of the prognostic relationship between the change in and the change in / were similar, which indicated that the addition of did not provide a substantial benefit.
Our study identified the trajectory as a reliable and pragmatic marker, easily and reliably obtainable at the bedside and from the electronic health record, and associated with the patient important outcome, ventilator-free days. The strengths of our study included the novel assessment of a pragmatic marker, the trajectory, the large cohort size, and the broad applicability to patients with acute hypoxic respiratory failure due to community-acquired pneumonia and/or ARDS. Our study had limitations. The single-center nature of the study limited overall generalizability, and these findings should be investigated further in a multi-center setting. Determining the trajectory also required that bedside teams (nursing, physicians, and respiratory therapists) reduced to the minimum required to meet and/or goals. A potential limitation in doing this study retrospectively was that we were not able to determine this.
However, during the period of our study, all medical ICU subjects followed a respiratory therapy-driven automated protocol (introduced in 2008) to regularly wean supplemental oxygen to the minimum required at multiple times during the day. As such, we can be reasonably confident that the administered at any one time had been appropriately titrated. Any future prospective validation of these study findings should include, for example, automated titration protocol. Also, in the subjects who were not receiving precise (mechanical ventilation, BPAP, CPAP, Venturi device), we used a conversion table to approximate . To our knowledge, a validated conversion does not exist, and this method has been used for similar purposes in several previous studies.18,29 Although this introduced imprecision, it was also a pragmatic necessity because we sought to determine an intermediate marker that can be readily used in a variety of clinical and research settings. Moreover, in our sensitivity analysis of the subjects on invasive mechanical ventilation (in which the exact is known), the association between the trajectory and ventilator-free days was unchanged (Supplementary Tables 6 and 7 [see the supplementary materials at http://www.rcjournal.com]).
Conclusions
In the subjects who were critically ill and with acute hypoxic respiratory failure secondary to community-acquired pneumonia or ARDS, and were readily available in the electronic health record, substantially more so than . A higher baseline and increase in compared with the previous day was associated with fewer ventilator-free days. The trajectory was at least as reliable as the / trajectory and, therefore, the trajectory alone may be sufficient to predict patient important outcomes.
Footnotes
- Correspondence: Sarah J Chalmers MD, Division of Pulmonary and Critical Care, Department of Medicine, Mayo Clinic, 200 First Street SW, Rochester, MN 55905. E-mail: chalmers.sarah{at}mayo.edu
See the Related Editorial on Page 1636
Supplementary material related to this paper is available at http://www.rcjournal.com.
This study was supported by a grant from the Mayo Clinic Critical Care and Research Committee, which did not play a role in the study design, implementation, or analysis of this study or in production of this manuscript. This project was supported by Clinical and Translational Science Award grant UL1 TR002377 from the National Center for Advancing Translational Science. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.
The authors have disclosed no conflicts of interest.
- Copyright © 2021 by Daedalus Enterprises