Discordance between respiratory drive and sedation depth in critically ill patients receiving mechanical ventilation

Abstract

Objective:

In mechanically ventilated patients, deep sedation is often assumed to induce respirolysis, i.e. lyse spontaneous respiratory effort, while light sedation is often assumed to preserve spontaneous effort. This study was conducted to determine validity of these common assumptions, evaluating the association of respiratory drive with sedation depth and ventilator-free days in acute respiratory failure.

Design:

Prospective cohort study.

Setting:

Patients were enrolled during two month-long periods in 2016–2017 from five ICUs representing medical, surgical, and cardiac specialties at a U.S. academic hospital.

Patients:

Eligible patients were critically ill adults receiving invasive ventilation initiated ≤ 36 hours before enrollment. Patients with neuromuscular disease compromising respiratory function or expiratory flow limitation were excluded.

Interventions:

Respiratory drive was measured via P_0.1, the change in airway pressure during a 0.1-second airway occlusion at initiation of patient inspiratory effort, every 12 ± 3 hours for 3 days. Sedation depth was evaluated via the Richmond agitation-sedation scale (RASS). Analyses evaluated the association of P_0.1 with RASS (primary outcome) and ventilator-free days.

Measurements and Main Results:

Fifty-six patients undergoing 197 bedside evaluations across five intensive care units were included. P_0.1 ranged between 0–13.3 (median 0.1, interquartile range 0.0–1.3) cm H₂O. P_0.1 was not significantly correlated with RASS (R_Spearman 0.02, 95% CI −0.12 to 0.16; p = 0.80). Considering P_0.1 terciles (range < 0.2, 0.2–1.0, and > 1.0 cm H₂O), patients in the middle tercile had significantly more ventilator-free days than the lowest tercile (incidence rate ratio [IRR] 0.78, 95% CI 0.65–0.93; p < 0.01) or highest tercile (IRR 0.58, 95% CI 0.48–0.70; p < 0.01).

Conclusion:

Sedation depth is not a reliable marker of respiratory drive during critical illness. Respiratory drive can be low, moderate, or high across the range of routinely targeted sedation depth.

INTRODUCTION

During invasive mechanical ventilation with critical illness, clinical practice guidelines recommend titrating analgesics and sedatives according to estimated pain and agitation/sedation, targeting light sedation where possible (1, 2).

Nevertheless, deep sedation sometimes is prescribed in attempt to facilitate lung-protective ventilation or eliminate patient-ventilator dyssynchrony (3–8). Patients with status asthmaticus, for example, may receive deep sedation in attempt to induce hypoventilation with permissive hypercapnia, thereby abating catastrophic dynamic hyperinflation (9). Similarly, in acute respiratory distress syndrome (ARDS), deep sedation may be chosen in attempt to maintain low tidal volumes and prevent certain patient-ventilator dyssynchronies (3, 10). In such cases, analgesics and sedatives are given not just for analgesic, anxiolytic, or sedating purposes, but also to suppress spontaneous respiratory effort, which we term their respirolytic effect. Efficacy of this practice is unknown.

While partial or complete respirolysis might help facilitate lung protection in select patients, in others it is an unintended side-effect of analgesics and sedatives. Prolonged, complete respirolysis may contribute to diaphragm disuse atrophy, impeding ventilator weaning (11–14). Better understanding of the relationship between sedation depth and respiratory drive is important to provide lung- and diaphragm-protective ventilation (15) and to minimize potentially harmful side-effects of sedation (16, 17).

Respiratory drive can be quantified easily at bedside using P_0.1, the change in airway pressure during a brief (0.1 second) airway occlusion at the start of patient inspiratory effort (18). P_0.1 quantifies motor output from the respiratory center at the beginning of inspiration and is a well-validated measure of drive available on most modern intensive care unit (ICU) ventilators (19–21).

Yet, respiratory drive is not monitored routinely in clinical practice. Rather, it is often assumed that deep sedation suppresses respiratory drive while light sedation preserves it (1, 3, 22–24). This common clinical assumption conflicts with our own experience at bedside, during which we have observed the effects of sedation on drive to be highly variable. Therefore, this study was conducted to evaluate the association of respiratory drive with sedation depth in patients with acute respiratory failure requiring invasive mechanical ventilation. Given the potential implications of respiratory drive for lung and diaphragm protection, we also evaluated its association with ventilator-free days.

METHODS

Patients

This prospective cohort included critically ill adults for whom invasive mechanical ventilation was newly initiated within the prior 36 hours. Patients were enrolled during two month-long periods in 2016–2017 from five ICUs (medical, surgical, and cardiac) at an academic hospital. Patients were excluded for neuromuscular disease that compromised respiratory function, obstructive airway disease causing expiratory flow limitation, chest tube with active air leak, tracheostomy, or ventilator lacking capability to measure P_0.1. The UCSD review board approved the study (Project #181061XX) with consent waiver.

Measurement of P_0.1

P_0.1 was measured using a validated, automated maneuver on full-feature ICU ventilators (21, 25). During the maneuver, the ventilator inspiratory valve remains closed during the first 100 milliseconds (0.1 second) of the next patient-triggered inspiratory cycle, a period thought too short for detection, conscious or unconscious, by the patient (3). Change in airway pressure during this brief occlusion is measured by the ventilator as P_0.1. After the 0.1-second occlusion, the inspiratory valve opens and flow is delivered. Although some ventilators estimate P_0.1 during the trigger phase without an occlusion (supplement) (26), for this study, only ventilators that performed the conventional method of P_0.1 measurement with a proper 0.1-second occlusion were used. For each assessment, P_0.1 measurement was repeated three times consecutively and the average value reported (27, 28).

P_0.1 measurement was performed every 12 ± 3 hours for three days (total of up to six assessments), immediately prior to bedside assessment of sedation depth, pain, and delirium. To ensure that data were representative of exposure, measurements were not performed during temporary spontaneous awakening and breathing trials. Periods during which enrolled patients received neuromuscular blockade were excluded from analyses.

Bedside Assessment

Bedside assessments were performed by trained study personnel twice daily immediately after each P_0.1 measurement. Sedation depth was assessed with the Richmond Agitation-Sedation Scale (RASS), pain with the Critical Care Pain Observation Tool (CPOT), and delirium with the Confusion Assessment Method for ICU (CAM-ICU) (supplement). Assessments were performed without knowledge of target RASS or nurse-charted value.

Vital signs, ventilator settings, analgesic/sedative exposure, and the most recent blood gas were recorded. Illness severity was assessed via baseline Acute Physiology and Chronic Health Evaluation-II (APACHE-II) and daily Sequential Organ Failure Assessment (SOFA).

Analgesic and Sedative Exposure

Analgesia and sedation were managed per usual care, consisting of nurse-driven titration to achieve a prescribed target RASS range (e.g. RASS 0 to −2; −1 to −3; −2 to −4; or −3 to −5) (29–32). Daily interruption of sedation and spontaneous breathing trials for extubation readiness screening were performed per hospital protocol, consistent with practice guidelines (2, 33). To quantify analgesic/sedative exposure, three time-intervals were considered: (1) the current dose of continuous infusion sedatives/analgesics at evaluation; (2) the total cumulative dose, including boluses, received during the hour immediately preceding bedside evaluation; and (3) the total dose, including boluses, received during the 12 hours immediately preceding bedside evaluation.

Clinical Outcomes

Patients were followed until discharge for vital status, ventilator-free days through day 28, duration of mechanical ventilation, and ICU/hospital lengths of stay. Ventilator-free days was computed as time between successful liberation from mechanical ventilation and day 28 for survivors, and zero days for non-survivors (34). Patients discharged alive before day 28 were presumed alive at day 28.

Statistical Analysis

For the primary outcome, Pearson and Spearman correlations evaluated the association of P_0.1 with RASS for each of the six timepoints and for all observations in aggregate. Linear mixed models were constructed to evaluate the within-subject association of P_0.1 with sedation depth over time (fixed effects) with random subject-specific intercept. Model marginal r² quantified the proportion of variance explained by fixed effects. Sensitivity analyses are detailed in the supplement.

Linear mixed models with random intercept also were constructed to evaluate the association of P_0.1 with pain, delirium, analgesic and sedative medication exposure, blood gas results, respiratory parameters, and illness severity over time.

To evaluate the association of respiratory drive with ventilator-free days, Poisson regression was conducted entering each patient’s average P_0.1 over the available evaluations as the predictor of interest. P_0.1 was entered as both a linear and quadratic term, with statistical significance for the quadratic term indicative of a nonlinear relationship warranting its retention in models. Multivariable models adjusted for APACHE-II, baseline PaO₂:FiO₂, and average RASS. To aid interpretability of a nonlinear relationship, unadjusted Poisson regression was rerun entering P_0.1 as a categorical variable split into terciles. LOESS regression was performed to visualize data.

To evaluate association of respiratory drive with in-hospital mortality, logistic regression models were used with P_0.1 entered as linear and quadratic terms as above.

Two-sided alpha of 0.05 was considered statistically significant.

RESULTS

Patients

Sixty-one patients met screening criteria, four were excluded for support with a ventilator lacking P_0.1 capability, and one was excluded for neuromuscular disease that compromised respiratory function. Fifty-six patients were included in the final analysis. Accounting for early deaths, extubations, and neuromuscular blockade use, there were 204 eligible bedside observation periods during the 3-day measurement window (supplement). Seven observations (3.4%) were missing for study staff unavailability, leaving 197 bedside observations with P_0.1 measurement for inclusion in the final analysis.

Patients were enrolled median [interquartile range] 19 [10.5–29] hours after intubation. Most patients (77%) were intubated for hypoxemic respiratory failure, and half of patients had ARDS at enrollment (Tables 1, E1). Patients were severely ill at enrollment, indicated by APACHE-II and SOFA scores predicting mortality of 33–50%. Observed in-hospital mortality was 33.9%.

Table 1.

Baseline Characteristics and Outcomes of Study Participants

Characteristic	Value (n=56)
Age, years	62.5 [53–75.5]
Female	18 (32%)
Primary reason(s) for intubationa
Hypoxemia	43 (77%)
Hypercapnia	5 (9%)
Airway protection	11 (20%)
Time from intubation to enrollment, hours	19 [10.5–29]
APACHE-IIb	32 [28–36]
Acute respiratory distress syndrome	28 (50%)
Vasopressor-dependent shock	23 (41%)
Continuous infusion analgesia and sedation at enrollment	50 (89%)
Propofol infusion	40 (71%)
Infusion rate, mcg/kg/min	35 [25–40]
Benzodiazepine infusion, n (%)	5 (9%)
Midazolam-equivalent infusion rate, mg/hc	4 [2–5]
Opioid infusion, n (%)	25 (45%)
Fentanyl-equivalent infusion rate, mcg/hd	50 [50–100]
Dexmedetomidine infusion	1 (2%)
Infusion rate, mcg/kg/h	0.6
No continuous infusion	6 (11%)
Richmond Agitation Sedation Scale (RASS)	−4 [−4 to −3]
Critical Care Pain Observation Tool score (CPOT)	0 [0–0]
Delirium positive Confusion Assessment Method for ICU	3 (5%)
Initial P0.1, cm H2O	0.4 [0.0–1.2]
Initial ventilator management
Volume-targeted pressure assist/control mode	53 (95%)
Tidal volume (mL/kg predicted body weight)	7.3 [6.5–8.3]
Airway driving pressure, cm H2O	11 [9–14]
PEEP, cm H2O	5 [5–9]
PaO2:FiO2	213 [168–312]
Patient outcomes
Duration of mechanical ventilation, days	4.5 [2.5–15.5]
Ventilator-free days	12 [0–25]
ICU length of stay, days	8.5 [5–21]
Hospital length of stay, days	14 [6.5–22.5]
New tracheostomy prior to discharge	7 (13%)
Death before hospital discharge	19 (33.9%)

^aParticipants could have multiple reasons for intubation; categories are not mutually exclusive.

^bAn APACHE score of 30 corresponds to a predicted in-hospital mortality of approximately 50%.

^cMidazolam-equivalent dose calculated as midazolam 1 mg = lorazepam 0.5 mg.

^dFentanyl-equivalent dose calculated as fentanyl 100 mcg = hydromorphone 2 mg = morphine 10 mg.

Sedation Depth, Pain, and Delirium

Patients were deeply sedated (RASS −4 or −5) during approximately half of evaluations (48.2%); most patients (75%) had at least one study assessment indicating deep sedation. The lightest sedation depth observed was RASS +1, corresponding to restless/anxious appearance without aggressive movements. Within-patient range in RASS during study observations was at least 1 level in 67.9% of patients, and at least 2 levels in 41.1% of patients (Table E2). Patients had behavioral findings suggestive of pain (CPOT > 2) during 6.1% of evaluations. Patients screened positive for delirium during 20.6% of eligible evaluations.

Respiratory Drive Assessed by P_0.1

Across 197 separate evaluations, P_0.1 ranged from 0 to 13.3 cm H₂O, with median [IQR] value 0.1 [0.0–1.3] cm H₂O (Figure E1). During only 38 of 197 measurements (19%) was P_0.1 between 0.5 and 1.5 cm H₂O, a range considered normal in healthy individuals at rest (20). P_0.1 exceeded 1.5 cm H₂O during 24.4% (48 of 197) of measurements. By contrast, P_0.1 was zero, indicating no respiratory drive, during half (49.2%) of measurements.

Nearly half of all patients (41.1%) had at least one evaluation with P_0.1 exceeding 1.5 cm H₂O, including 44.4% of patients with ARDS. The average within-patient range of P_0.1 during the study period was 1.3 cm H₂O (minimum range 0 cm H₂O, maximum range 12.2 cm H₂O).

Respiratory Drive and Sedation Depth

Figure 1 presents P_0.1 values according to RASS for all observations. No compelling association is conspicuous graphically.

Figure 1. P_0.1 and Richmond Agitation-Sedation Scale (RASS).

Overlaid box-and-whisker and scatter plot. Boxes represent median with interquartile range; whiskers extend 1.5 times the interquartile range beyond the first and third quartiles per the Tukey method. Diamonds represent mean values. Circles indicate individual observations; total of 197 observations were performed and all displayed. P_0.1 values within each RASS level were randomly jittered to facilitate visualization. During the study period, the highest RASS observed was +1.

Eighteen of the 20 highest P_0.1 values, including all ten highest P_0.1 values, were observed in patients with RASS −3 to −5.

In cross-sectional analysis, RASS was not significantly correlated with P_0.1 at any evaluation time (Figure 2). Correlation coefficients were consistently near zero.

Figure 2. Respiratory drive is poorly correlated with sedation depth.

Pearson and Spearman correlations for cross-sectional relationship between P_0.1 and sedation depth assessed via Richmond Agitation-Sedation Scale (RASS). Higher RASS corresponds to increased arousal (lighter sedation depth).

Considering average within-patient effects over time via linear mixed models, higher RASS was poorly correlated with higher P_0.1 (marginal r² = 2.4%; p = 0.02; Table E3). RASS was not significantly associated with P_0.1 in sensitivity analyses adjusting for pain and delirium (Table E3).

Respiratory Drive and Pain

P_0.1 exceeded 1.5 cm H₂O during 21.6% of observations without pain. Sixteen of the 20 highest P_0.1 values were observed in patients without apparent pain.

In linear mixed models evaluating within-patient associations, higher CPOT was poorly associated with higher P_0.1 (marginal r² = 4.0%) (Table E3). A similar lack of meaningful association was observed when entering CPOT as a dichotomous measure of pain (marginal r² = 1.6%).

Respiratory Drive and Delirium

P_0.1 exceeded 1.5 cm H₂O during 21.8% of observations without delirium. Thirteen of the 20 highest P_0.1 values were observed in patients not meeting criteria for delirium.

Delirium was weakly correlated with higher P_0.1 in linear mixed models evaluating within-patient associations (marginal r² = 5.8%) (Table E3).

Respiratory Drive and Sedative Exposure

Higher propofol dose over the prior hour was weakly associated with a modest decrease in P_0.1 (Table 2). Benzodiazepine and opioid titrations were not significantly associated with P_0.1.

Table 2.

Association of P_0.1 with Analgesic and Sedative Exposure Over Time^a

Medication	Median [IQR] Dose	ß (95% CI)	P	Marginal r2 for fixed effectsb	Conditional r2 for fixed and random effectsb
Infusion rate at time of evaluation
Propofol, mcg/kg/min	20 [0–40]	−0.015 (−0.031, 0.001)	0.07	2.7%	57.4%
Midazolam-equivalent, mg/hc	0 [0–0]	0.014 (−0.068, 0.097)	0.74
Fentanyl-equivalent, mcg/hd	25 [0–50]	0.003 (−0.003, 0.010)	0.31
Total dose over prior 1 h
Propofol, mcg/kg	1200 [0–2400]	−3.3×10−4 (−5.8×10−4, −0.7×10−4)	0.01	3.6%	58.1%
Midazolam-equivalent, mg	0 [0–0]	0.012 (−0.070, 0.094)	0.77
Fentanyl-equivalent, mcg	25 [0–75]	0.002 (−0.004, 0.009)	0.50
Total dose over prior 12 h
Propofol, mcg/kg	14,400 [0–25,200]	−1.8×10−5 (−4.2×10−5, 6.2×10−6)	0.14	1.8%	56.4%
Midazolam-equivalent, mg	0 [0–0]	3.3×10−4 (−6.8×10−3, 7.4×10−3)	0.93
Fentanyl-equivalent, mcg	300 [0–825]	2.2×10−4 (−3.9×10−4, 8.3×10−4)	0.48

^aAnalyses included 197 bedside evaluations among 56 patients. Reported regression coefficient (95% CI) and p-value were calculated from linear mixed effects model with all three classes of medications and time entered as fixed effects and random subject effect (intercept). The resulting regression coefficients report the change in P_0.1 per 1-unit change in sedative dose, adjusting for the effects of the other sedatives and time. Separate models were run for infusion rate at time of evaluation and total dose over prior 1 h. Negative values indicate decrease in P_0.1 with higher medication dose.

^bMarginal and conditional r² were computed per the method of Nakagawa and Schielzeth. Marginal r² quantifies the proportion of variance explained by model fixed effects only, while conditional r² quantifies the proportion of variance explained by fixed and random effects. A low marginal r² and high conditional r² together indicate that most model variance is explained by the random effect (subject intercept).

^cMidazolam-equivalent dose calculated as midazolam 1 mg = lorazepam 0.5 mg.

^dFentanyl-equivalent dose calculated as fentanyl 100 mcg = hydromorphone 2 mg = morphine 10 mg.

Respiratory Drive and Other Factors

Within-patient associations of P_0.1 with other factors were evaluated to reconfirm face-validity of P_0.1 as a measure of respiratory drive. Lower pH, higher respiratory rate, and higher minute-volume were associated with higher P_0.1 (Tables E4, E5). After adjusting for sedation depth, greater illness severity (SOFA) also was associated with higher P_0.1.

Clinical Outcomes

Average P_0.1 exhibited a significant, non-linear association with ventilator-free days (unadjusted Poisson model with coefficients: VFD = −0.0801 P_0.1² + 0.1166 P_0.1 + 2.5686; p < 0.01 for P_0.1² term). This association remained statistically significant and qualitatively unchanged in a multivariable model adjusting for baseline APACHE-II, baseline PaO₂:FiO₂, and average RASS. Rerunning the unadjusted model entering P_0.1 tercile (range 0.0–0.1, 0.2–1.0, and 1.1–8.1) as a class variable with the middle tercile as the reference group demonstrated a similar relationship. Patients in the middle tercile had significantly more ventilator-free days compared to patients in either the lowest tercile (incidence rate ratio [IRR] 0.78, 95% CI 0.65–0.93; p < 0.01) or highest tercile (IRR 0.58, 95% CI 0.48–0.70; p < 0.01). The same relationship was evident graphically with LOESS regression: patients with moderate respiratory drive on average had more ventilator-free days than patients with either no drive or high drive (Figure 3).

Figure 3. Both high and low respiratory drive, compared to moderate drive, are associated with fewer ventilator-free days.

(A) Overlaid box-and whisker and scatter plots of all P_0.1 values for each patient; red diamonds represent mean values, and circles indicate values from individual observations. Boxes represent median with interquartile range; whiskers extend 1.5 times the interquartile range beyond the first and third quartiles per the Tukey method. Shaded vertical bars indicate number of ventilator-free days for each patient (secondary Y-axis). Patients are sorted by number of ventilator free days and average P_0.1 value to aid interpretation. (B) Locally estimated scatterplot smoothing (LOESS) regression line with 95% confidence bands for model associating P_0.1 with ventilator-free days.

P_0.1, entered into logistic models as either a linear or quadratic term, was not significantly associated with mortality in univariable or multivariable models.

DISCUSSION

This study demonstrates that sedation depth is a poor surrogate of respiratory drive, as measured by P_0.1, in critically ill adults with acute respiratory failure. Across the range of routinely targeted sedation depth, respiratory drive could be low, moderate, or high. Deep sedation often did not suppress respiratory drive, light sedation did not ensure respiratory effort was preserved, and many non-sedated awake patients had normal respiratory drive. Patients with moderate respiratory drive experienced more ventilator-free days than patients in whom drive was either high or completely extinguished.

Clinical practice guidelines addressing ICU sedation advocate using validated tools for the identification, treatment and prevention of pain, agitation/anxiety, and delirium (2). While absent from these guidelines, monitoring respiratory drive could be useful to evaluate whether the intended clinical effect is achieved, and to recognize an undesirable side-effect when sedation is prescribed for other indications. If deep sedation is prescribed for respirolysis but fails to decrease drive, it exposes patients to potential harm from excessive sedatives/analgesics (35) with no benefit. Sedation prescribed to alleviate pain or anxiety may cause occult respirolysis, risking disuse atrophy of respiratory muscles (11, 13).

Critically ill patients with acute respiratory failure have a multitude of factors that might contribute to high respiratory drive: pain, anxiety, systemic and cerebral inflammation, infection, pulmonary mechanical perturbations, acidemia, hypercapnia, and hypoxemia, among others (36). Medications that solely alleviate pain or anxiety are unlikely to achieve respirolysis in the setting of other contributors to drive. Common tools for assessing pain and anxiety/sedation, CPOT and RASS, include “fighting the ventilator” as a marker of both pain and anxiety/agitation. Our data indicate that high drive, a proximate cause to patient-ventilator dyssynchronies that might be construed as “fighting the ventilator,” is often not responsive to analgesia or anxiolysis and frequently occurs absent agitation.

Many sedatives and analgesics can modulate respiratory pattern directly. Prior studies in non-critically ill populations have found propofol depresses inspiratory muscle effort and induces a rapid shallow breathing pattern (37–39). A similar effect has been reported with midazolam, which also may blunt ventilatory response to carbon dioxide (40–42). Dexmedetomidine, in contrast, is thought to have little effect on respiratory drive or pattern (43). Opioids may cause bradypnea with less effect on drive, and can produce a unique respiratory pattern characterized by prolonged active exhalation with abdominal muscle contraction (44). In the present study, higher propofol dose was associated with a modest decrease in P_0.1, while no discernible within-patient effect was observed with benzodiazepines or opioids. Clinician rationale for a particular sedative or sedation depth was not recorded in this study, and there are several reasons why a particular sedation depth may be targeted in a given patient. Nevertheless, our data suggest dose-response relationships described in healthy volunteers or surgical patients are inconsistently reproduced during critical illness at routinely prescribed doses.

P_0.1 does have limitations as a measure of respiratory drive. P_0.1 measures motor output from the respiratory center and can be discordant with neural drive due to muscle weakness, fatigue, or other disruption to neuromechanical coupling. This study enrolled patients early in their ventilator course and excluded patients with known neuromuscular weakness to mitigate this concern. Breath-to-breath variability in P_0.1 requires averaging multiple consecutive measures to obtain representative values (20, 28), an approach used in this study. Expiratory flow limitation with intrinsic PEEP can introduce a phase delay between inspiratory muscle effort and pressure change at the airway opening. Patients with overt flow limitation were excluded from this study to negate this potential concern, although prior data suggest P_0.1 remains valid despite intrinsic PEEP (27, 45). Measurement technique differs across ventilators, and resulting imprecision could impede clinical applications. Reassuringly, reported values appear reasonably similar on newer ventilators regardless of technique performed (21, 46).

Alternative measures of respiratory drive, including electrical activity of the diaphragm, esophageal pressure, and transdiaphragmatic pressure, are technically challenging to perform and not widely available. By comparison, P_0.1 is well-validated over decades, near-ubiquitous on modern ICU ventilators, often completely automated requiring no technical expertise, and can be monitored breath to breath continuously over time. P_0.1 has not been studied previously for monitoring sedation effects but has been used to titrate assistance during pressure support ventilation (47–49) and evaluate extubation readiness (50).

The association of P_0.1 extremes (high or low) with fewer ventilator-free days may not be causal. Both high drive and complete pharmacologic respirolysis could be markers of underlying pathology that contributes to morbidity independent of drive. Yet, the association of P_0.1 with ventilator-free days remained significant after adjusting for APACHE-II, PaO₂:FiO₂, and RASS. In at-risk patients, high drive could prolong ventilator weaning by (i) exacerbating lung injury (high tidal volumes, breath stacking, pendelluft, atelectrauma (10, 51)); (ii) exacerbating diaphragm fatigue and injury (under-assisted or dyssynchronous eccentric contractions (52)); and (iii) prompting administration of additional sedatives/analgesics with accompanying adverse effects (3). Low drive might prolong ventilator weaning via respiratory muscle atrophy and excess sedative/analgesic exposure (11, 13).

The optimal level of respiratory drive in critical illness is unknown and likely depends on potentially conflicting risks of lung injury, diaphragm injury, and the side-effect profile of interventions designed to modulate drive. Adjusting the ventilator to optimize patient-ventilator interactions and treating causes underlying high drive should be attempted before considering pharmacologic interventions solely for respirolysis. In less critically ill patients without severe lung injury, acceptance of heightened drive and permissive dyssynchrony (53) is likely preferable to added sedative, analgesic, or paralytic exposure, while the converse might be true in select patients at extremely high risk of lung injury. Whether attempting to use sedatives and analgesics for respirolysis affords benefit to this narrow highest-risk population is unclear, but risks of deep sedation (16, 17, 33) and respiratory muscle inactivity (11–14) are well established.

CONCLUSION

Sedation depth is a poor marker of respiratory drive and should not be used as a surrogate of drive during critical illness. Extremes of respiratory drive (high and low) were observed across the range of routinely targeted sedation depth and were independently associated with fewer ventilator-free days. Whether measuring respiratory drive to guide clinical decision-making can improve outcomes is unknown and warrants evaluation.