Hospital Variation in Mortality and Ventilator Management among Mechanically Ventilated Patients with ARDS

Abstract

Rationale

Acute Respiratory Distress Syndrome (ARDS) is associated with significant mortality. Despite the mortality benefits of lung protective ventilation, adherence rates to evidence-based ventilator practice have remained low and ARDS mortality has remained high.

Objective

Determine variation in ARDS mortality and adherence to low tidal volume ventilation (LTV) across US hospitals.

Materials and Methods

We identified mechanically ventilated patients with ARDS using data from Philips eICU (2014-2015). We then used multi-variable hierarchical logistic regression models with hospital site as the random effect and patient and hospital level factors as fixed effects to assess the hospital risk adjusted mortality rate and median odds ratio for the association between mortality and hospital site. We then assessed associations between adherence to LTV (defined as 4-8 mL/kg PBW) and hospital risk adjusted mortality rates using Spearman correlation.

Results

Among 4441 patients admitted at 110 hospitals with ARDS, the hospital risk-adjusted mortality rate ranged from 19% to 39%, and the MOR for hospital of admission was 1.33 (95% CI 1.25-1.41). Among 3070 patients at 72 hospitals with available ventilator data, 73% of patients had a median set Vt between 4 to 8 mL/kg PBW; hospital adherence rates to LTV ranged from 13% to 95%. There was no association between hospital adherence to LTV and risk-adjusted mortality rate (spearman correlation coefficient −0.01, p = .93). Similarly, among 956 patients who started with a Vt > 8 mL/kg PBW, there was no association between the percent of patients at each hospital whose Vt was decreased to ≤ 8 mL/kg PBW and risk adjusted mortality rate (spearman correlation coefficient .05, p = .73).

Conclusion

Risk adjusted mortality and use of LTV for patients with ARDS varied widely across hospitals. However, hospital adherence to LTV was not associated with ARDS mortality rates. Further evaluation of hospital practices associated with lower ARDS mortality are warranted.

Keywords

ARDS mechanical ventilation hospital variation mortality low tidal volume ventilation

Introduction

Acute Respiratory Distress Syndrome (ARDS) is characterized by inflammatory pulmonary edema leading to acute hypoxemic respiratory failure, and is associated with significant morbidity and mortality.¹ The primary treatment for patients with ARDS is lung productive ventilation (LPV) to reduce ventilator-associated lung injury, based on the results of the 2000 ARMA trial.² However, adherence to LPV has remained relatively low,^3,5 and ARDS mortality has remained high over the last two decades.³ Thus, professional societies have recently proposed quality measures for patients with ARDS to improve outcomes as part of guideline recommendations.⁶ As part of these quality initiatives, the American Thoracic Society developed performance measures for patients with ARDS who are mechanically ventilated, which includes examining the proportion of ventilator settings that meet criteria for LPV (defined as 4-8 mL/kg PBW AND plateau pressure or peak inspiratory pressure < 30 cm H₂O) over the first 72 h of mechanical ventilation.⁶ In this study, we sought to assess variation in adherence to proposed performance measures and their association with mortality rates across US hospitals among patients with ARDS in order to provide benchmarks for future efforts to improve ARDS processes and outcomes.

Materials and Methods

Study Design & Population

We performed a multicenter cohort study of mechanically ventilated patients with ARDS using data from Philips eICU Collaborative database from 2014–2015.⁷ Philips eICU is a large multi-center database, that includes ICU admissions to 208 hospitals across the United States that participate in the Philips telehealth program. We included ICU patients ages 18 to 89, with ARDS, who received invasive mechanical ventilation for greater than 1 h and had documented hospital mortality data. ARDS was defined by a clinical problem list charted diagnosis or an Acute Physiology and Chronic Health Evaluation (APACHE)⁸ admission diagnosis of ARDS in the eICU database, and use of invasive mechanical ventilation was extracted from nurse and respiratory bedside charting tables. For patients with multiple eligible ICU admissions, we selected a random ICU admission. We excluded any admission to a hospital with fewer than 10 patients that met eligibility criteria to increase the stability of hospital-level estimates.

Outcomes

The primary outcome of interest was the median risk-adjusted hospital mortality, a measure of hospital mortality after accounting for differences in patient- and hospital- level factors, was calculated by taking the median of the predicted individual mortality rates during hospitalization from hierarchical regression models accounting for patient- and hospital-level factors (see statistical analysis section) per hospital. We then explored the association between hospital adherence to LTV and risk adjusted mortality rate. Adherence to LTV was calculated as the percentage of patients at each hospital site who received a median Vt between 4 to 8 mL/kg PWB as further discussed below (see ventilator data parameters section).

Patient- and Hospital Level-Factors

Patient and hospital factors for inclusion in models were decided a priori based on potential to confound associations between hospital and mortality. Patient factors were age, gender, race/ethnicity (white, black, Hispanic, other), body mass index (BMI), Charlson Comorbidity Index (CCI),⁹ Sequential organ failure assessment (SOFA) score,¹⁰ APACHE-IV score,⁸ duration of admission prior to ICU transfer, managing provider specialty, location of care prior to ICU admission (emergency department, surgical area including operating room and recovery areas, and transfer from floor or outside hospital), and ARDS severity at time of mechanical ventilation, and admission diagnosis. Hospital factors were US region, hospital bed number, teaching status, and ICU type. Further details regarding patient and hospital level factors are included in the supplement.

Ventilator Data Parameters

In order to examine associations between the risk-adjusted hospital mortality and adherence to evidence-based ventilator management, we extracted several ventilator parameters. We obtained tidal volumes (Vt) for each patient in units of mL/kg predicted body weight (PBW) at the time of initiation of mechanical ventilation and then twice daily for the first 72 h as proposed by the ATS workshop group,⁶ where available. We used documented height and gender to calculate predicted body weight using the following formula: [height(cm) – 152.4]*0.91 + 50 for males, and; [height(cm) – 152.4]*0.91 + 45.5 for females.¹¹ We defined low tidal volume (LTV) ventilation as 4 to 8 mL/kg PBW based upon current guideline recommendations.¹² We only included patients who were ventilated with a volume control mode based upon unclear accuracy of tidal volumes that were not entered as “set Vt” in the database for other ventilator modes. We did not include plateau pressure (or peak inspiratory pressure), despite its inclusion in the performance measures from the ATS workshop group,⁶ as this was missing in large numbers from eICU data, with only 65% of patients having at least one recorded value (either plateau pressure or peak inspiratory pressure) in the initial 72 h of mechanical ventilation. We excluded patients with a Vt that was less than 4 mL/kg PBW, and greater than 14 mL/kg PBW, based upon unclear accuracy of data.

For each individual patient, we examined both the lowest Vt within 24 h of mechanical ventilation, and the median Vt over the first 72 h of mechanical ventilation. Because we sought to evaluate clinician practice patterns, we examined the lowest set Vt to evaluate attempts at LTV, acknowledging that occasionally Vts require adjustment to clinical circumstances and that patient achieved Vt may differ from set Vt. We also evaluated median Vt over 72 h to capture the central tendency of early ventilator management. We were unable to record Vt every 12 h as proposed by the ATS performance measures due to lack of twice-daily Vt recorded in the dataset.

Statistical Analysis

We summarized patient and hospital factors using mean with standard deviation or median with IQR, as appropriate, for continuous variables, and counts with proportions for categorical variables. For unadjusted data, we calculated the overall average hospital mortality, and then stratified by ARDS severity according to the Berlin definition.¹¹ We used complete cases in our primary analyses of the data. We used a multi-variable hierarchical logistic regression models with hospital site as the random effect (random intercept) and included each patient and hospital level factor as fixed effects in the models. From these models, we determined the risk adjusted mortality rate for each hospital site and the Median Odds Ratio (MOR). The MOR is the median change in risk for an individual if managed at a randomly chosen higher risk versus a lower risk hospital,¹³ which allows evaluation of hospital variation on the same odds ratio scale as fixed effects. We calculated the e-value for the MOR, which is the minimum strength of association that an unmeasured confounder would need to have with both the exposure and the outcome to fully explain a specific exposure-outcome association.^14,15

To examine associations between risk-adjusted mortality and compliance with evidence-based ventilator management, we calculated the percentage of patients who received LTV (defined as a median Vt between 4-8 mL/kg PBW) at each hospital site within the initial 72 h of mechanical ventilation, and then used a spearman rank correlation coefficients¹⁶ to calculate the correlation between hospital risk-adjusted mortality rate and hospital adherence to LTV. We evaluated the association between hospital-level adherence to LTV (as measured by the percentage of patients who received LTV at 72 and 24 h at each hospital) and individual patient mortality adjusted for patient- and hospital-level factors. Lastly, we examined the proportion of patients who received LTV by patient and hospital characteristics in order to identify variations in LTV use based upon these variables.

In order to examine ventilator changes over time, we examined trends in Vt over the initial 72 h of mechanical ventilation among patients with at least 2 documented Vts. We calculated (1) the percentage of patients with any change in Vt, (2) the percentage of patients with a decrease in Vt and (3) the median decrease in Vt over the initial 72 h of mechanical ventilation. Among patients with an initial Vt higher than 8 ml/kg PBW, we calculated (1) the percentage of patients with any decrease in Vt and (2) the percentage of patients with a decrease in Vt to ≤ 8 mL/kg PBW within the first 72 h of mechanical ventilation. We assessed the association between the proportion of patients with a decrease in Vt to ≤ 8 mL/kg PBW at each hospital site and hospital risk adjusted mortality rate using spearman rank correlation coefficients.

Sensitivity Analyses

We performed multiple sensitivity analyses in order to evaluate the robustness of our findings to missing data, different definitions of LTV, different ARDS severity, different cohorts with possible ARDS, and to compare our results to findings from ARDS literature: (1) Using multiple imputation with chained equations (MICE)¹⁷ to impute missing values prior to model creation, (2) defining LTV as Vt ≤ 6.5 mL/kg PWB, as Vt ≤ 6.5 mL/kg PBW has been used to define LTV frequently throughout ARDS literature,^4,5,18,20 based upon initial ARMA trial,² (3) Defining hospital median Vt within the initial 24 h and 72 h of mechanical ventilation, (4) using a less restrictive inclusion definition (diagnosis of pneumonia and a P:F ≤ 300) because 40% of ARDS may go unrecognized³ and (5) limiting our inclusion criteria to those who received mechanical ventilation for ≥ 48 h. In order to better approximate the ATS process measure and evaluate results in the context of a recent study of ARDS mortality¹⁸ we performed an additional sensitivity analysis among patients with moderate-severe ARDS (P:F ≤ 150) who had recorded plateau or peak pressures.

Deidentified data was declared non-human subjects research and exempt from review by the Boston University Medical Center institutional review board. All data management, graphs and statistical analyses were performed using R version 4.0.2.²¹ All tests were two-sided, and a P < .05 was considered statistically significant.

Results

There were 4441 patients admitted to 110 hospitals who met inclusion criteria (Figure 1). Median patient age was 63 (IQR 52-73), pneumonia was the most common admission diagnosis (23%), and 50% of patients met criteria for severe ARDS. Approximately half of hospitals had ≥ 500 beds and 33% were teaching hospitals (Table 1). The unadjusted in hospital ARDS mortality rate ranged from 0 to 50%, with an average hospital mortality of 28% (95% CI 27-29%). The average mortality rate was 19% (95% CI 15-23%) for mild ARDS, 24% (95% CI 21-26%) for moderate, and 34% (95% CI 32-36%) for severe ARDS. The median duration of mechanical ventilation was 2.9 days (IQR 2.8-3.1), and the median hospital length of stay was 11.8 days (IQR 11.3-12.2).

Table 1.

Baseline Patient and Hospital Characteristics.

	n = 4441
Region (%)
Midwest	1324 (29.8)
Northeast	388 (8.7)
South	1611 (36.3)
West	864 (19.5)
Other/Unknown	254 (5.7)
Hospital Size, # of beds (%)
≥ 500	2112 (47.6)
250 to 499	1083 (24.4)
< 250	860 (19.4)
Other/Unknown	386 (8.7)
Teaching hospital (%)	1483 (33.4)
Unit Type (%)
Medical ICU	579 (13.0)
Cardiac/Cardiothoracic ICU	869 (19.6)
Medical-Surgical ICU	2432 (54.8)
Neuro ICU	217 (4.9)
Surgical ICU	344 (7.7)
ICU Admission Location (%)
Emergency Dept or Direct Admit	1945 (43.8)
Surgical (OR/Recovery/Pacu)	301 (6.8)
Transfer (Floor/Stepdown/OSH)	924 (20.8)
Other/Unknown	1271 (28.6)
Provider Type (%)
Critical Care	1184 (26.7)
Generalist	1386 (31.2)
Non-critical care subspecialist	514 (11.6)
Surgeon	443 (10.0)
Other/Unknown	914 (20.6)
Age (median [IQR])	63.00 [52.00, 73.00]
Male Gender (%)	2418 (54.4)
Race/ethnicity (%)
White	3243 (73.0)
Black	571 (12.9)
Hispanic	276 (6.2)
Unknown/Other	351 (7.9)
Body Mass Index (median [IQR])	28.28 [23.76, 34.87]
Charlson Comorbidity Index (median [IQR])	3.00 [1.00, 5.00]
Admission Diagnosis (%)
ARDS	412 (9.3)
Cardiac	404 (9.1)
Cardiac/Respiratory Arrest	460 (10.4)
Neurologic	274 (6.2)
Pneumonia	1009 (22.7)
Pulmonary, other than pneumonia	597 (13.4)
Sepsis, other than pneumonia	354 (8.0)
Surgery/Trauma	501 (11.3)
Other	430 (9.7)
APACHE IV score (median [IQR])	74.00 [56.00, 96.00]
SOFA score (median [IQR])	8.00 [5.00, 10.00]
Admission duration prior to ICU, hours (median [IQR])	5.68 [1.57, 56.20]
ARDS Severity (%)
Initial P:F > 300	318 (7.5)
Mild ARDS	468 (11.0)
Moderate ARDS	1319 (31.1)
Severe ARDS	2139 (50.4)

Missingness: BMI 3.2%, Charlson Comorbity Index 0.83%, APACHE IV Score 12.74%, ARDS Severity 4.44%.

Abbreviations: ARDS, acute respiratory distress syndrome; APACHE, acute physiology and chronic health evaluation; SOFA, sequential organ failure assessment.

There were 3554 admissions at 97 hospitals included in the primary complete case analysis. After adjusting for both patient and hospital level factors, the risk-adjusted mortality rate by hospital site ranged from 19% to 39% (Figure 2), with an average risk-adjusted hospital mortality of 28%. The MOR for the hospital of admission was 1.33 (95% CI 1.25-1.41), with an e-value of 1.57. Few categorical covariates (geographical location, severe ARDS, cardiac arrest) were more strongly associated with mortality than the MOR for hospital site (EFigure 1).

Ventilator Data

Patient-level tidal volume data

Set Vt was documented within 24 h of mechanical ventilation for 2918 patients at 70 hospitals, and within 72 h of mechanical ventilation for 3070 patients at 72 hospitals. The lowest set Vt documented for patients in the first 24 h ranged from 4.2 to 13.4 mL/kg PBW, and the median set Vt over the initial 72 h of mechanical ventilation ranged from 4.1 to 13.4 mL/kg PBW. Overall, 75% of patients had at least one set Vt between 4 to 8 mL/kg PBW within the first 24 h, and 73% had a median set Vt between 4 to 8 mL/kg PBW over the initial 72 h. However, when using a lower cut off for LTV of ≤ 6.5 mL/kg PBW, 28% of patients met LTV criteria at 24 h, and 25% had a median set Vt ≤ 6.5 mL/kg PBW over the initial 72 h. Of those patients who were excluded for presumed inaccurate tidal volumes (Vt < 4 mL/kg PBW or > 14 mL/kg PBW), there were no trends in patient or hospital characteristics identified.

Hospital adherence to LTV and outcomes

When examining adherence to LTV (defined as 4-8 mL/kg PBW) by hospital, 6 to 95% of patients at each hospital site received LTV in the first 24 h, and 13 to 95% received LTV over the initial 72 h of mechanical ventilation. We found no association between risk adjusted mortality rate and hospital adherence to LTV in the initial 72 h of mechanical ventilation (spearman correlation coefficient −.01, p = .93). In a patient-level model, hospital rate of LTV was not associated with mortality (adjusted OR 1.00, 95% CI 0.99-1.01). Similarly, there was no association between hospital adherence to LTV in the initial 24 h of mechanical ventilation and risk adjusted mortality rate (spearman correlation coefficient −0.04, p = 0.74) (Figure 3). When evaluating trends in the use of LTV based upon hospital characteristics, patients more frequently received LTV at large hospitals (≥ 500 beds), teaching institutions, hospitals in the northeast region, with no significant difference based upon managing provider type or ICU type (ETable 1).

Trends in tidal volume over 72 h

When examining practice patterns surrounding changes in Vt made over the initial 72 h of mechanical ventilation, 2880 patients had at least 2 documented Vt and were included for comparison. Of these, half (53%) had a change in Vt, with 28% having a decrease in Vt. The median decrease in Vt in the initial 72 h of mechanical ventilation was 1.07 mL/kg PBW.

Throughout the initial 72 h of mechanical ventilation, 68% of patients received LTV on initiation of mechanical ventilation, 76% received LTV in the initial 48 h, 78% received LTV in the initial 72 h, with approximately 65% of patients receiving a Vt between 4 to 8 mL/kg PWB throughout the entire initial 72 h of mechanical ventilation. Approximately 20% of patients received a Vt > 8 mL/kg PWB throughout the initial 72 h of mechanical ventilation. For hospitals, 76% had a median Vt between 4 to 8 mL/kg PBW over the initial 72 h of mechanical ventilation, and all but 10% achieved LTV by 72 h.

Among 956 patients who started with an initial set Vt > 8 mL/kg PBW, half (47%) had a decrease in Vt over the first 72 h, and a third (35%) had a decrease in Vt to < 8 mL/kg PBW. In this cohort, there was no correlation between the percentage of patients with a Vt decrease at each hospital and risk adjusted mortality rate (Spearman .05, p-value = .73), or the percentage of patients with a of Vt decrease to ≤ 8 mL/kg PBW at each hospital and risk adjusted mortality rate (Spearman .02, p-value = .90) (EFigure 2).

Sensitivity analyses

Results from the primary and sensitivity analyses are included in Table 2, ETable 2 (baseline characteristics of moderate-severe ARDS cohort), ETable 3 (baseline characteristics of pneumonia + P:F ≤ 300 cohort), and EFigures 3–5 (ventilator data for moderate-severe ARDS cohort, pneumonia + P:F ≤ 300 cohort, and alternative definitions of Vt respectively). Overall, our sensitivity analyses showed similar results to our primary analysis, including wide ranges in hospital risk adjusted mortality rates, a strong association between mortality and hospital of admission, and no association between hospital adherence to LTV and risk adjusted mortality rate. In the sensitivity analysis restricted to cases of moderate-severe ARDS (P:F ≤ 150) with plateau or peak pressures recorded, the mean Vt within the initial 24 h of mechanical ventilation overall was 7.3 mL/kg PBW (ranging from 5.9-9.5 mL/kg PBW by hospital of admission), with 69% of patients receiving a mean Vt > 6.5 mL/kg PBW and 27% receiving a mean Vt > 8 mL/kg PWB over the initial 24 h of mechanical ventilation. The overall adherence to LPV (defined as Vt ≤ 6.5 mL/kg PBW and plateau or peak inspiratory pressure ≤ 30 cm H₂O) was 28.6%. Among those in the group adherent to LPV, the mean Vt was 6.0 mL/kg PBW, and mean plateau pressure was 23 cm H₂O, compared to those in the non-adherent group, where the mean Vt was 8.0 mL/kg PBW and the mean plateau pressure was 24 cm H₂O. Similar to our primary analysis, there was no association between non-adherence rates to LPV and hospital risk adjusted mortality rate (Spearman .14, p-value = .41). When limiting our inclusion criteria to those that received mechanical ventilation for ≥ 48 h, the median duration of invasive mechanical ventilation was 5.5 days (IQR 5.3-5.75), which is consistent with findings from ARDS literature.²²

Table 2.

Primary and Sensitivity Analyses.

Analysis	Patients/Hospitals	Hospital Risk Adjusted Mortality Rate Range	MOR for Hospital of Admission (95% CI)	Spearman Correlation Coefficient between Hospital Rate of Adherence to LTV and Risk Adjusted Mortality Rate (p-value)
Primary Analysis	3554/97	19% to 39%	1.33 (1.25-1.41)	72 h: −0.01 (.93) 24 h: −0.04 (.74)
Alternative Cohort Definitions:
Moderate-Severe ARDS (P:F ≤ 150)	1245/35	25% to 38%	1.41 (1.32-1.52)	0.14 (.41)
Unrecognized ARDS (Pneumonia and P:F ≤ 300)	6267/119	13% to 40%	1.45 (1.35-1.56)	0.03 (.76)–0.04 (.70)*
Alternative Vt Definitions:
LTV ≤ 6.5 mL/kg PBW	3554/97	19% to 39%	1.33 (1.25-1.41)	72 h: 0.02 (.84) 24 h: 0.08 (.52)
Median Vt by hospital	3554/97	19% to 39%	1.33 (1.25-1.41)	72 h: −0.01 (.37) 24 h: −0.02 (.84)
MICE Imputation	4441/110	17% to 40%	1.39 (1.30-1.48)	N/A**
Limited inclusion criteria: Mechanical Ventilation ≥ 48 h	2758/75	22% to 33%	1.22 (1.16-1.28)	72 h: −0.01 (.90) 24 h: −0.08 (.49)

*Range based upon Hospital Rate of Adherence to LTV defined as ≤ 8 mL/kg PBW and alternative Vt definitions of LTV defined as ≤ 6.5 mL/kg PBW and Hospital Median Vt at 72 and 24 h.

**MICE imputation did not calculate spearman correlation.

Abbreviations: MOR, median odds ratio; LTV, low tidal volume; Vt, tidal volume; ARDS, acute respiratory distress syndrome.

Discussion

In this multicenter cohort study of patients with ARDS, mortality varied widely across hospitals and hospital of admission was strongly associated with hospital mortality. Although most patients received evidence-based tidal volume strategies of 4 to 8 ml/kg PBW, few received a tidal volume of ≤ 6.5 ml/kg PBW. Use of evidence-based tidal volumes varied widely between hospitals but hospital use of evidence-based tidal volumes was not associated with ARDS mortality rates.

Our findings inform efforts to improve ARDS care. First, we provide novel data to identify outcome benchmarking targets, and ranges that identify hospital mortality outliers, that allow hospitals to evaluate ARDS outcomes in relation to others. Second, we identified generally higher rates of LTV during ARDS as compared with other studies, with approximately 70% of patients receiving Vt in ranges generally considered to be evidence-based.^3,5,19 The use of LTV varied widely between hospitals, and multiple hospitals continued to show low rates of LTV. However, low hospital use of LTV was not associated with increased mortality. While surprising, we hypothesize this could be due to many factors. Randomized control trials that have shown mortality benefit of LTV have used control groups that received Vt ≥ 10 mL/kg PBW,²³ and no hospitals included in this study had a median Vt ≥ 10 mL/kg PBW in the initial 24 or 72 h of mechanical ventilation. Thus, while we observed large variations in adherence to LTV between hospitals, it is possible that the observed variation in median Vt between centers is not strongly clinically impactful. In addition, we observed trends in the use of LTV based upon hospital characteristics, which could potentially explain the lack of association between mortality and LTV adherence in our overall cohort. Trials have also shown that driving pressure is a risk factor for death in ARDS even in the setting of LTV,²⁴ but we were unable to further explore how variations in driving pressure impacts outcomes based upon missingness from our dataset. Finally, the management of ARDS is complex and highly variable between centers,¹⁸ thus variations in other practice patterns could contribute to this observed range in mortality.

Our study also provides a real-world evaluation of proposed ARDS quality measures.⁶ Our findings suggest that implementation of electronic health record data collection for tidal volume every 12 h for the first 72 h of mechanical ventilation will be limited by data missingness, and likely would require manual data collection. Additionally, plateau pressures needed to evaluate adherence to pressure-limited ventilation goals were rarely available in electronic record data, highlighting additional areas for improvement in both measurement and documentation. We found few adjustments to Vt during the early management of ARDS, with the majority of patients who started with a Vt > 8 mL/kg PBW never reaching LTV during the initial 72 h of mechanical ventilation, suggesting that clinical inertia plays a strong role in poor adherence rates to LTV. Thus, initial ventilator setting may be a more feasible – and potentially more effective – quality measure than repeated measures.¹⁹ This finding is supported by data from Sjoding et al, who observed patterns of Vt administration at a single center in patients with ARDS and found that initial Vt settings predict exposures to Vt ≥ 8 mL/kg PBW and initial exposure to Vt ≥ 8 mL/kg PBW was associated with increased mortality.²⁵

Previously published data for hospital variation in ARDS mortality is limited. Qadir et al examined associations of LPV with mortality in patients with moderate-severe ARDS,¹⁸ and showed a similar wide range in ARDS mortality between centers, along with wide variation in adherence to lung protective ventilation. However, results showing an association between hospital use of LPV and lower mortality risk differed from our findings. Reasons for the different associations between LPV and mortality identified between the two studies are unclear, although multiple differences between the two studies may have contributed to different results, including different ARDS mortality rates, inclusion of different modes of mechanical ventilation, use of different severity of illness risk adjustment variables (SOFA vs. APACHE IV), and inclusion of different types and sizes of hospitals.

This study had several limitations. The use of Philips eICU limits to inclusions of hospitals who participate in Philips telehealth program, which may decrease generalizability to non-telehealth hospitals. Based upon the deidentified nature of Philips eICU, we were unable to use chart review to validate study variables. We did not include plateau pressure (or peak inspiratory pressure) in our primary definition of LPV ventilation due to missingness and instead measured LTV. We were also limited to the inclusion of recognized ARDS cases in the primary cohort, as we did not have imaging studies available in our dataset. However, a sensitivity analysis using a diagnosis of pneumonia plus P:F ≤ 300 to include potentially unrecognized ARDS showed similar results. As with all observational study, unmeasured confounding may affect results. However, the high e-value calculated for our MOR shows that any unmeasured cofounder would need to have a significant strength of association with both the outcome and exposure in order to explain the association between ARDS mortality and hospital of admission.

In conclusion, this study identified large variation in ARDS mortality rates and use of LTV across hospitals. These findings allow hospitals to benchmark their own practice patterns and management strategies, in order to target process and outcome improvements. Additionally, the lack of correlation between the use of LTV at each hospital and hospital mortality suggests other practices may drive large variation across hospitals in ARDS mortality, which warrants further evaluation. However, LTV remains one of the few management strategies with known mortality benefit in ARDS, and adherence to LTV represents persistent area for practice improvement. Future studies should examine drivers of hospital mortality variation in ARDS, as well as methods to improve data collection for further development of quality measures in order to identify practice patterns and management strategies that could lead to improved practices and outcomes.

Supplemental Material

sj-docx-1-jic-10.1177_08850666221111748 - Supplemental material for Hospital Variation in Mortality and Ventilator Management among Mechanically Ventilated Patients with ARDS

Supplemental material, sj-docx-1-jic-10.1177_08850666221111748 for Hospital Variation in Mortality and Ventilator Management among Mechanically Ventilated Patients with ARDS by Mallory N. LeSieur, Nicholas A. Bosch and Allan J. Walkey in Journal of Intensive Care Medicine

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Ethical Approval

Not applicable, because this article does not contain any studies with human or animal subjects.

ORCID iD

Mallory N. LeSieur

Supplemental Material

Supplemental material for this article is available online.

References

Ware

Matthay

. The acute respiratory distress syndrome. N Engl J Med. 2000;342(18):1334–1349. (In eng). DOI: 10.1056/nejm200005043421806.

Brower

Matthay

Morris

Schoenfeld

Thompson

Wheeler

. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med. 2000;342(18):1301–1308. (In eng). DOI: 10.1056/nejm200005043421801.

Bellani

Laffey

Pham

, et al. Epidemiology, patterns of care, and mortality for patients with acute respiratory distress syndrome in intensive care units in 50 countries. Jama. 2016;315(8):788–800. (In eng). DOI: 10.1001/jama.2016.0291.

Spece

Mitchell

Caldwell

Gundel

Jolley

Hough

. Low tidal volume ventilation use remains low in patients with acute respiratory distress syndrome at a single center. J Crit Care. 2018;44:72–76. (In eng). DOI: 10.1016/j.jcrc.2017.10.021.

Weiss

Baker

Weiner

, et al. Low tidal volume ventilation use in acute respiratory distress syndrome. Crit Care Med. 2016;44(8):1515–1522. (In eng). DOI: 10.1097/ccm.0000000000001710.

Artis

Dweik

Patel

, et al. Performance measure development, use, and measurement of effectiveness using the guideline on mechanical ventilation in acute respiratory distress syndrome. An official American thoracic society workshop report. Ann Am Thorac Soc. 2019;16(12):1463–1472. (In eng). DOI: 10.1513/AnnalsATS.201909-665ST.

Pollard

Johnson

AEW

Raffa

Celi

Mark

Badawi

. The eICU collaborative research database, a freely available multi-center database for critical care research. Sci Data. 2018;5:180178. (In eng). DOI: 10.1038/sdata.2018.178.

Zimmerman

Kramer

McNair

Malila

. Acute physiology and chronic health evaluation (APACHE) IV: hospital mortality assessment for today’s critically ill patients. Crit Care Med. 2006;34(5):1297–1310. (In eng). DOI: 10.1097/01.Ccm.0000215112.84523.F0.

Charlson

Pompei

Ales

MacKenzie

. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373–383. (In eng). DOI: 10.1016/0021-9681(87)90171-8.

10.

Vincent

Moreno

Takala

, et al. The SOFA (sepsis-related organ failure assessment) score to describe organ dysfunction/failure. On behalf of the working group on sepsis-related problems of the European society of intensive care medicine. Intensive Care Med. 1996;22(7):707–710. (In eng). DOI: 10.1007/bf01709751.

11.

Ranieri

Rubenfeld

Thompson

, et al. Acute respiratory distress syndrome: the Berlin definition. Jama. 2012;307(23):2526–2533. (In eng). DOI: 10.1001/jama.2012.5669.

12.

Fan

Del Sorbo

Goligher

, et al. An official American thoracic society/European society of intensive care medicine/society of critical care medicine clinical practice guideline: mechanical ventilation in adult patients with acute respiratory distress syndrome. Am J Respir Crit Care Med. 2017;195(9):1253–1263. (In eng). DOI: 10.1164/rccm.201703-0548ST.

13.

Merlo

Chaix

Ohlsson

, et al. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health. 2006;60(4):290–297. (In eng). DOI: 10.1136/jech.2004.029454.

14.

Mathur

Ding

Riddell

VanderWeele

. Web site and R package for computing E-values. Epidemiology. 2018;29(5):e45–e47. (In eng). DOI: 10.1097/ede.0000000000000864.

15.

VanderWeele

Ding

. Sensitivity analysis in observational research: introducing the E-value. Ann Intern Med. 2017;167(4):268–274. (In eng). DOI: 10.7326/m16-2607.

16.

Schober

Boer

Schwarte

. Correlation coefficients: appropriate use and interpretation. Anesth Analg. 2018;126(5):1763–1768. (In eng). DOI: 10.1213/ane.0000000000002864.

17.

S. van Buuren

KG-O

. Mice: multivariate imputation by chained equations in R [internet]. J Stat Softw. 2011;45:1–67. (Available from: https://www.jstatsoft.org/v45/i03/).

18.

Qadir

Bartz

Cooter

, et al. Variation in early management practices in moderate-to-severe ARDS in the United States: the severe ARDS – generating evidence study. Chest. 2021;160(4):1304–1315. DOI: 10.1016/j.chest.2021.05.047

19.

Walkey

Wiener

. Risk factors for underuse of lung-protective ventilation in acute lung injury. J Crit Care. 2012;27(3):323.e1–9. (In eng). DOI: 10.1016/j.jcrc.2011.06.015.

20.

Umoh

Fan

Mendez-Tellez

, et al. Patient and intensive care unit organizational factors associated with low tidal volume ventilation in acute lung injury. Crit Care Med. 2008;36(5):1463–1468. (In eng). DOI: 10.1097/CCM.0b013e31816fc3d0.

21.

Team RC. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria 2020 (URL https://www.R-project.org/).

22.

Sayed

Riaño

Villar

. Predicting duration of mechanical ventilation in acute respiratory distress syndrome using supervised machine learning. J Clin Med. 2021;10(17):3824. (In eng). DOI: 10.3390/jcm10173824.

23.

Walkey

Goligher

Del Sorbo

, et al. Low tidal volume versus non-volume-limited strategies for patients with acute respiratory distress syndrome. A systematic review and meta-analysis. Ann Am Thorac Soc. 2017;14(Supplement_4):S271–s279. (In eng). DOI: 10.1513/AnnalsATS.201704-337OT.

24.

Guérin

Papazian

Reignier

Ayzac

Loundou

Forel

. Effect of driving pressure on mortality in ARDS patients during lung protective mechanical ventilation in two randomized controlled trials. Crit Care. 2016;20(1):384. (In eng). DOI: 10.1186/s13054-016-1556-2.

25.

Sjoding

Gong

Haas

Iwashyna

. Evaluating delivery of low tidal volume ventilation in six ICUs using electronic health record data. Crit Care Med. 2019;47(1):56–61. (In eng). DOI: 10.1097/ccm.0000000000003469.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.54 MB