0
Original Research: CRITICAL CARE MEDICINE |

Prolonged Acute Mechanical Ventilation: Implications for Hospital Benchmarking FREE TO VIEW

Marya D. Zilberberg, MD, MPH, FCCP; Andrew A. Kramer, PhD; Thomas L. Higgins, MD, MBA; Andrew F. Shorr, MD, MPH, FCCP
Author and Funding Information

*From the School of Public Health and Health Sciences (Dr. Zilberberg), University of Massachusetts, Amherst, MA; Cerner Corporation (Dr. Kramer), Kansas City, MO; the Division of Critical Care Medicine (Dr. Higgins), Baystate Medical Center, Springfield, MA; and the Division of Pulmonary and Critical Care (Dr. Shorr), Washington Hospital Center, Washington, DC.

Correspondence to: Marya D. Zilberberg, MD, MPH, FCCP, University of Massachusetts, School of Public Health and Health Sciences, PO Box 303, Goshen, MA 01032; e-mail: Marya@EviMedGroup.org


Dr. Kramer is an employee of Cerner Corporation, which holds the rights to the APACHE system. Drs. Zilberberg, Higgins, and Shorr have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (www.chestjournal.org/site/misc/reprints.xhtml).


© 2009 American College of Chest Physicians


Chest. 2009;135(5):1157-1162. doi:10.1378/chest.08-1928
Text Size: A A A
Published online

Background:  Hospital performance measures rely on aggregate outcomes. For patients receiving mechanical ventilation (MV), outcomes depend on severity of illness, hospital MV volume, and case mix. Patients requiring prolonged acute MV (PAMV) [MV for ≥ 96 h] comprise a resource-intensive group, but the impact of its volume on aggregate outcomes is unknown. We investigated whether observed outcomes differed from those predicted by APACHE (acute physiology and chronic health evaluation) IV risk adjustment and the relationship between hospital MV volume and outcomes among patients receiving PAMV.

Methods:  We conducted a retrospective cohort study using the APACHE IV database between the years 2001 and 2005.

Results:  Of the 94,553 patients receiving MV at 45 hospitals, 24,366 (25.8%) were receiving PAMV. Unadjusted mortality was 32.3% for patients receiving PAMV and 22.9% for patients receiving short-term MV (STMV) [< 96 h]. Although mortality predictions were accurate in both groups, the length-of-stay (LOS) predictions underestimated duration of MV, ICU LOS, and hospital LOS by 5.2, 4.6, and 5.4 days, respectively, in the PAMV group. After stratifying the PAMV group by hospital MV volume, except for quintile 1, the standardized mortality ratio (SMR) was found to be inversely related to the volume quintile. The difference between actual and predicted MV durations, however, exhibited a consistent direct relationship with the MV volume.

Conclusions:  In patients requiring PAMV, the SMR is inversely proportional to hospital MV volume. Conversely, the PAMV group had a disproportionate effect on durations of MV, ICU LOS, and hospital LOS, and these marginal excesses increased with the hospital MV volume quintile. Development of specific predictive equations for patients receiving PAMV is recommended. Benchmarking measures must consider the case mix of patients receiving STMV vs those receiving PAMV.

Hospital performance measures, the most compelling of which is hospital mortality, are aimed at improving quality and efficiency of health-care delivery.1 Although imperfect, several statistical models24 are used to adjust mortality rates for severity of illness. Even in the face of risk adjustment, however, controversy exists about its accuracy and validity as a measure of an individual institution's performance.5,6 Hospital length of stay (LOS) has been used as a surrogate for efficiency conflating quality and cost of care.1 Similar to mortality risk adjustment, LOS risk-adjustment models allow benchmarking of utilization performance, thus aiding in predicting ICU and hospital LOS and time receiving mechanical ventilation (MV).7 Some instruments, such as the diagnosis-related group system,8 are used to determine reimbursement levels given patient and hospital characteristics. The accuracy of these instruments lies in how well they factor in disease, patient, and process-of-care characteristics. This accuracy is particularly important in the face of a changing case mix, especially among populations with disproportionately high and inherently variable utilization intensity.

One such recently identified population is patients who require prolonged acute MV (PAMV) [MV for ≥ 96 h], who comprise one third of all patients receiving MV in US hospitals.9 Although this group's hospital survival may not differ from patients receiving short-term MV (STMV) [MV < 96 h], their hospital LOS is substantially longer.9 At $16 billion in 2003, the PAMV population accounted for approximately two thirds of the cost of care of all patients receiving MV in US hospitals.9 Highlighting the need to optimize efficiency of health-care delivery to this population, these patients' numbers have been projected to more than double by the year 2020.10 The increasing volume of patients requiring PAMV will alter the institutional case mix of the MV population, thereby influencing the aggregate outcomes that determine hospital performance measures.

The present study explored the relationship between APACHE (acute physiology and chronic health evaluation) IV risk-adjusted mortality and resource utilization predictions among patients receiving PAMV. Furthermore, given previous findings11 that hospital MV volume is predictive of mortality but has no relationship to resource utilization among patients receiving MV, we also investigated how well the existing APACHE IV models measure these domains of hospital performance.

No human subjects were enrolled in the study. Thus, the study was exempt from regulations guiding protection of human subjects.

Data for the study were obtained from a database (APACHE IV; Cerner Corporation; Kansas City, MO). Only the sites participating for ≥ 12 months and those contributing data on at least 50 patients receiving MV were eligible for enrollment. Included were patients admitted to ICUs between January 1, 2001, and June 30, 2005, at hospitals that were part of the APACHE system. Patients admitted from another hospital ICU or those who were readmitted to the hospital ICU were excluded. Patients undergoing coronary artery bypass grafting surgery were excluded because their outcomes require specialized predictive models. Information was available on physiologic measures, diagnosis, comorbidities, demographics, hospital admission and discharge locations, and various clinical variables. Detailed information on this database can be found elsewhere.7 Duration of MV was captured as the first episode if occurring on day 1 of the hospital ICU stay. It was collected as an integer (ie, consecutive number of days); thus, patients receiving PAMV were ventilated at some point during their first day in the ICU, and this episode lasted for ≥ 4 days.

Outcomes analyzed included hospital mortality, duration of MV, and ICU and hospital LOS. Predicted values were obtained from the corresponding APACHE IV equation for each of these outcomes, and predicted values were compared with the observed values. The standardized mortality ratio (SMR) was calculated by dividing the observed by the predicted mortality. For the MV duration and ICU and hospital LOS, the mean of the differences between observed and predicted values for each patient was derived. This process was repeated after stratifying the hospitals by volume quintile based on their average annual number of patients receiving MV at APACHE ICUs within each institution.

Results for categorical variables are reported as proportions; for nonparametrically distributed variables, as medians with interquartile ranges (IQRs); and for parametric continuous variables, as means with 95% confidence intervals (CIs). Significance testing for categorical variables was performed using the χ2 test and for parametrically and nonparametrically distributed continuous variables, with the Student t test and the Kruskall-Wallis test, respectively. Statistical significance was deemed to be present at the α level of < 0.05.

Of the 61 hospitals in the APACHE IV database representing 285,874 patients between January 1, 2001, and June 30, 2005, 45 eligible sites contributed 93,075 patients receiving MV on hospital ICU day 1 (STMV, 69,124 patients [74.3%]; PAMV, 23,951 patients [25.7%]). Although similar in age and gender distribution, patients receiving PAMV differed from those receiving STMV in many ways (Table 1). Most importantly, patients receiving PAMV had a significantly greater severity of illness than those receiving STMV, as measured by the median acute physiology score component of the APACHE IV score (PAMV, 59 [IQR, 43 to 79]; STMV, 46 [IQR, 32 to 67]; p < 0.001). Patients receiving PAMV also were more likely than those receiving STMV to be classified as medical rather than surgical and to receive their care in a mixed ICU (Table 1). The top five primary diagnoses given patients in the PAMV and STMV groups differed between the groups (Table 2).

Table Graphic Jump Location
Table 1 Baseline Characteristics of Patients on STMV vs PAMV*

*Values are given as the median (IQR) or %, unless otherwise indicated. APS = acute physiology score.

Table Graphic Jump Location
Table 2 Most Frequent Diagnostic Groups in Descending Order of Frequency

Outcomes were significantly different between the PAMV and STMV groups. Crude hospital mortality was higher in the PAMV than in the STMV group (relative risk, 1.35; 95% CI, 1.33 to 1.38) [Table 3]. However, no statistically significant difference was found in the SMR between the two groups (PAMV, 0.965; STMV, 0.990; p = 0.083). Table 3 shows the observed and predicted outcomes for the two groups. Although the predicted LOS in each of the strata among the patients receiving STMV approximated the observed LOS among the patients receiving PAMV, the predicted LOS value consistently underestimated the observed value by a substantial duration. Thus, the differences between the observed and predicted values were 5.2, 4.6, and 5.4 days for MV duration, ICU LOS, and hospital LOS, respectively (p < 0.001 for each). Put another way, the actual duration receiving MV, ICU LOS, and hospital LOS were longer than the predicted values by 111%, 67%, and 36%, respectively. The differences between predicted and observed LOS values in each stratum of care also were significantly greater among patients requiring PAMV than STMV (Table 3).

Table Graphic Jump Location
Table 3 Actual and Predicted Outcomes by Stratum of Hospital Care Among Patients on STMV vs PAMV

*The p value for mortality based on finding the smallest CIs that do not overlap. The p values for MV duration, ICU LOS, and hospital LOS were calculated using a t test of the difference between paired observations (observed and predicted values).

Results for the 45 hospitals divided into quintiles by annual institutional MV volume are presented in Table 4. Although the lowest SMR occurred among patients receiving PAMV in the lowest volume quintile, the highest volume quintile had an SMR lower than that for quintiles 2 through 4. Although no clear trend was found in the difference between the observed and predicted values for hospital or ICU LOS among the quintiles, the difference in the duration of MV showed a consistent rise, with the smallest difference observed in the lowest volume quintile and the greatest difference in the highest volume quintile (p < 0.001). Thus, although the patients requiring PAMV in the highest volume quintile had 9% lower mortality than predicted, they needed 5 additional days of MV.

Table Graphic Jump Location
Table 4 SMRs and Differences Between Actual and Predicted Duration of MV, ICU LOS, and Hospital LOS Among Patients Receiving PAMV by Quintile of Annual Hospital Volume of MV

*Values are given as the mean (95% CI).

†Values are given as the median (95% CI). Differences in LOS in each stratum of care are calculated as the actual minus the predicted number of days.

We have demonstrated that the observed mortality in the PAMV population is consistent with the APACHE IV predictions, with an SMR that approximates 1.0. Observed durations of MV, ICU LOS, and hospital LOS, however, are significantly longer than those predicted. As previously shown in all patients receiving MV,11 our data confirm an inverse hospital MV volume-to-survival relationship among most patients receiving PAMV, with the exception of those in the lowest volume quintile. Interestingly, with respect to hospital resource utilization, the difference between observed and predicted values for hospital LOS becomes greater as the annual hospital MV volume increases. This finding suggests that LOS models developed in a general population may miss important differences in highly select population subgroups, such as the population requiring a longer duration of ventilation support. This finding is consistent with other studies; for example, Murphy-Filkins et al12 demonstrated in a simulation study that artificially changing the proportion of patient characteristics beyond a critical threshold can generate misleading conclusions regarding hospital ICU performance.

The PAMV population is large and resource intensive, representing > 300,000 hospital discharges in 2003.9 Previous estimates of the median hospital LOS among patients receiving PAMV (17 days) far exceed those for patients receiving STMV (6 days).9 The growth of patients being discharged from the hospital while receiving PAMV is outpacing the rise in US hospitalizations overall, and this group is expected to number > 600,000 by year 2020.10 Given such explosive growth and resource intensity, it is critical to understand how the utilization patterns of such patients may affect quality and efficiency benchmarking. Our data suggest that mortality of patients receiving PAMV can be adequately predicted using day-of-hospital-admission physiologic measurements. However, measures of resource use, such as MV duration, ICU LOS, and hospital LOS, appear to be affected by factors unaccounted for in the APACHE day 1 models,13,14 such as neurologic, cardiac, respiratory, renal, nutritional, musculoskeletal, and psychological dysfunction, which may develop during the hospital ICU stay.

How do we address the imprecision of utilization estimates based only on day 1 physiology? Analogous to our findings in the PAMV population, the developers of APACHE found that patients remaining in the hospital ICU for > 4 days had an LOS that could not be predicted with adequate accuracy using day 1 data. As a result, a special equation15 was developed to estimate with greater precision the remaining LOS for patients who are still in the hospital ICU on day 5. In fact, applying the day 5 models to the PAMV population resulted in a much more precise estimate of the remaining LOS (data not shown), suggesting that a similar strategy may be needed for accurate LOS predictions in this group. However, until such equations are widely incorporated into routine information technology applications, performance benchmarking must be calibrated carefully so as not to penalize institutions with greater volumes of patients receiving PAMV. This admonition likely holds for other predictive models in wide use for performance benchmarking.

Given the near-universal finding that greater experience improves efficiency, the observation of a direct relationship between a hospital's annual MV volume and risk-adjusted MV duration among patients requiring PAMV is puzzling and could serve as a marked disincentive for regionalizing PAMV care. Although the explanation for this novel finding needs further exploration, our observation that the larger the annual hospital MV volume, the greater the discrepancy between their predicted and observed durations of MV carries important implications. Because MV is a disproportionate driver of hospital costs, seemingly small errors in the overall hospital LOS estimates may result in large discrepancies between actual costs and reimbursements for these institutions.16,17

Although mostly in agreement, our study is partially at odds with the study by Kahn et al,6 which indicated an inverse relationship between a hospital's annual volume of MV and mortality outcomes for all patients requiring MV. Contrary to those results, we have found in the PAMV population that although this relationship is the case for volume quintiles 2 through 5, it breaks down in the lowest volume quintile, which appears to have the lowest SMR. Several potential explanations exist for this finding, the most likely of which is that the small number of PAMV cases contributed by the hospitals in the lowest volume quintile has reduced the precision of the estimates. On the other hand, lower volume hospitals have been noted to be more likely to transfer their patients receiving MV to a higher volume facility,18 introducing a potential selection bias into the evaluation of the volume–outcomes relationship among patients receiving MV. Another possibility, though less likely, is that lower volume institutions may have more aggressive tracheostomy practices,19 resulting in their improved ability to transfer patients to a subacute facility sooner, where their outcomes are not captured. Finally, it is possible that lower MV volume institutions represent community centers where hospital discharge-planning efforts are less challenging than those in the larger referral centers. The latter scenarios would result in lower observed mortality because the SMR outcome is highly sensitive to hospital discharge practices.6 These and other possible explanations for this discrepancy need to be explored in future studies.

Our study has a number of limitations. First, as a retrospective cohort design, it is prone to several forms of bias. Second, the APACHE IV database captures only MV episodes that occur within the first 24 h of hospital admission. Although this procedure may have underestimated both the total MV and the PAMV cases, the population that requires MV after the initial 24-h period partly represents patients with nosocomial complications and, as such, can be expected to have different outcomes not based on the primary reason for hospitalization. Third, the PAMV case definition in the present study was slightly different from that used in previous work.9,10 Initial studies9,10 in this population defined PAMV as any patient whose case has been given an International Classification of Diseases, ninth edition, clinical modification, procedure code of 96.72 (MV duration for ≥ 96 h), but because of the way the APACHE data are collected, we classified PAMV as any patient who required MV on day 4 or beyond. Although our definition may have resulted in the misclassification of 4,531 (4.6%) cases as PAMV rather than STMV, this misclassification would carry a slight bias against our hypotheses. The fact that our results are significant even in the face of this possible misclassification implies that the relationships we have uncovered are robust.

We have confirmed that the PAMV population represents a substantial proportion of the total number of patients receiving MV; the hospital mortality rate for the group is 32%. Additionally, we have shown that although mortality predictions in this population are accurate, the predictions of LOS in every stratum of care significantly underestimate the actual LOS. Furthermore, while confirming the inverse volume-outcomes relationship in hospital mortality, we have demonstrated that patients receiving PAMV in hospitals with higher annual MV volumes are likely to have a longer LOS than those in the lower MV volume institutions. This divergence in mortality and hospital resource utilization outcomes should not by any means deter us from studying the potential of regionalizing MV services as a quality improvement measure because there is a substantial survival advantage associated with being cared for in a higher MV volume institution.11,20 It should, on the other hand, serve as further impetus to refine LOS prediction models for the increasingly specialized and complex populations cared for in our hospital ICUs. The implication that better hospital survival in the PAMV population can be achieved while at the same time expending increased hospital resources demands further attention. It underscores the importance of studying such potential interventions as early tracheostomy as an efficiency measure to streamline the care of patients receiving PAMV. It also demands better understanding of patient outcomes and utilities downstream from their acute hospitalization. Further study of what systems can eliminate bottlenecks in hospital throughput is critical in dealing with the welcome challenges created by improved clinical outcomes of patients with critical illness. Finally, as the PAMV population rapidly grows in US institutions, more accurate risk-stratification models that reflect the changing case mix are of paramount importance in order to develop valid performance measures and to align quality with reimbursement.

APACHE

acute physiology and chronic health evaluation

CI

confidence interval

IQR

interquartile range

LOS

length of stay

MV

mechanical ventilation

PAMV

prolonged acute mechanical ventilation

SMR

standardized mortality ratio

STMV

short-term mechanical ventilation

Kroch E, Duan M, Silow-Carroll S, et al. Hospital performance improvement: trends on quality and efficiency; a quantitative analysis of performance improvement in US hospitals. 2007; New York, NY The Commonwealth Fund
 
Zimmerman JE, Kramer AA, McNair DS, et al. Acute physiology and chronic health evaluation (APACHE) IV: hospital mortality assessment for today's critically ill patients. Crit Care Med. 2006;34:1297-1310. [PubMed] [CrossRef]
 
Higgins TL, Teres D, Copes WS, et al. Assessing contemporary intensive care unit outcome: an updated Mortality Probability Admission Model (MPM0-III). Crit Care Med. 2007;35:827-835. [PubMed]
 
Moreno RP, Metnitz PGH, Almeida E, et al. SAPS 3: from evaluation of the patient to evaluation of the intensive care unit: part 2. Development of a prognostic model for hospital mortality at ICU admission. Intensive Care Med. 2005;31:1345-1355. [PubMed]
 
Lee TH, Torchiana DF, Lock JE. Is zero the ideal death rate? N Engl J Med. 2007;357:111-113. [PubMed]
 
Kahn JM, Kramer AA, Rubenfeld GD. Transferring critically ill patients out of hospital improves the standardized mortality ratio: a simulation study. Chest. 2007;131:68-75. [PubMed]
 
Zimmerman JE, Kramer AA, McNair DS, et al. Intensive care unit length of stay: benchmarking based on acute physiology and chronic health evaluation (APACHE) IV. Crit Care Med. 2006;34:2517-2529. [PubMed]
 
Centers for Medicare & Medicaid Services Acute inpatient PPS.Accessed January 21, 2008 Available at:http://www.cms.hhs.gov/AcuteInpatientPPS.
 
Zilberberg MD, Luippold RS, Sulsky S, et al. Prolonged acute mechanical ventilation, hospital resource utilization and mortality in the United States. Crit Care Med. 2008;36:724-730. [PubMed]
 
Zilberberg MD, de Wit M, Pirone JR, et al. Growth in adult prolonged acute mechanical ventilation: implications for healthcare delivery. Crit Care Med. 2008;36:1451-1455. [PubMed]
 
Kahn JM, Goss CH, Heagerty PJ, et al. Hospital volume and the outcomes of mechanical ventilation. N Engl J Med. 2006;355:41-50. [PubMed]
 
Murphy-Filkins R, Teres D, Lemeshow S, et al. Effect of changing patient mix on the performance of an intensive care unit severity-of-illness model: how to distinguish a general from a specialty intensive care unit. Crit Care Med. 1996;24:1968-1973. [PubMed]
 
Kollef MH, Ahrens TS, Shannon W. Clinical predictors and outcomes for patients requiring tracheostomy in the intensive care unit. Crit Care Med. 1999;27:1714-1720. [PubMed]
 
Rady MY, Ryan T. Perioperative predictors of extubation failure and the effect on clinical outcome after cardiac surgery. Crit Care Med. 1999;27:340-347. [PubMed]
 
Kramer A. The APACHE IV equations: benchmarks for mortality and resource use [white paper report].Accessed June 1, 2008 Available at:http://www.cerner.com/public/Cerner_3.asp?id=27300.
 
Dasta JF, McLaughlin TP, Mody SH, et al. Daily cost of an intensive care unit day: the contribution of mechanical ventilation. Crit Care Med. 2005;33:1266-1271. [PubMed]
 
Halpern NA, Pastores SM, Greenstein RJ. Critical care medicine in the United States 1985–2000: an analysis of bed numbers, use, and costs. Crit Care Med. 2004;32:1254-1259. [PubMed]
 
Needham DM, Bronskill SE, Rothwell DM, et al. Hospital volume and mortality for mechanical ventilation of medical and surgical patients: a population-based analysis using administrative data. Crit Care Med. 2006;34:2349-2354. [PubMed]
 
Griffiths J, Barber VS, Morgan L, et al. Systematic review and meta-analysis of studies of the timing of tracheostomy in adult patients undergoing artificial ventilation. BMJ. 2005;330:1243. [PubMed]
 
Kahn JM, Linde-Zwirble WT, Wunsch H, et al. Potential value of regionalized intensive care for mechanically ventilated medical patients. Am J Respir Crit Care Med. 2008;177:285-291. [PubMed]
 

Figures

Tables

Table Graphic Jump Location
Table 1 Baseline Characteristics of Patients on STMV vs PAMV*

*Values are given as the median (IQR) or %, unless otherwise indicated. APS = acute physiology score.

Table Graphic Jump Location
Table 2 Most Frequent Diagnostic Groups in Descending Order of Frequency
Table Graphic Jump Location
Table 3 Actual and Predicted Outcomes by Stratum of Hospital Care Among Patients on STMV vs PAMV

*The p value for mortality based on finding the smallest CIs that do not overlap. The p values for MV duration, ICU LOS, and hospital LOS were calculated using a t test of the difference between paired observations (observed and predicted values).

Table Graphic Jump Location
Table 4 SMRs and Differences Between Actual and Predicted Duration of MV, ICU LOS, and Hospital LOS Among Patients Receiving PAMV by Quintile of Annual Hospital Volume of MV

*Values are given as the mean (95% CI).

†Values are given as the median (95% CI). Differences in LOS in each stratum of care are calculated as the actual minus the predicted number of days.

References

Kroch E, Duan M, Silow-Carroll S, et al. Hospital performance improvement: trends on quality and efficiency; a quantitative analysis of performance improvement in US hospitals. 2007; New York, NY The Commonwealth Fund
 
Zimmerman JE, Kramer AA, McNair DS, et al. Acute physiology and chronic health evaluation (APACHE) IV: hospital mortality assessment for today's critically ill patients. Crit Care Med. 2006;34:1297-1310. [PubMed] [CrossRef]
 
Higgins TL, Teres D, Copes WS, et al. Assessing contemporary intensive care unit outcome: an updated Mortality Probability Admission Model (MPM0-III). Crit Care Med. 2007;35:827-835. [PubMed]
 
Moreno RP, Metnitz PGH, Almeida E, et al. SAPS 3: from evaluation of the patient to evaluation of the intensive care unit: part 2. Development of a prognostic model for hospital mortality at ICU admission. Intensive Care Med. 2005;31:1345-1355. [PubMed]
 
Lee TH, Torchiana DF, Lock JE. Is zero the ideal death rate? N Engl J Med. 2007;357:111-113. [PubMed]
 
Kahn JM, Kramer AA, Rubenfeld GD. Transferring critically ill patients out of hospital improves the standardized mortality ratio: a simulation study. Chest. 2007;131:68-75. [PubMed]
 
Zimmerman JE, Kramer AA, McNair DS, et al. Intensive care unit length of stay: benchmarking based on acute physiology and chronic health evaluation (APACHE) IV. Crit Care Med. 2006;34:2517-2529. [PubMed]
 
Centers for Medicare & Medicaid Services Acute inpatient PPS.Accessed January 21, 2008 Available at:http://www.cms.hhs.gov/AcuteInpatientPPS.
 
Zilberberg MD, Luippold RS, Sulsky S, et al. Prolonged acute mechanical ventilation, hospital resource utilization and mortality in the United States. Crit Care Med. 2008;36:724-730. [PubMed]
 
Zilberberg MD, de Wit M, Pirone JR, et al. Growth in adult prolonged acute mechanical ventilation: implications for healthcare delivery. Crit Care Med. 2008;36:1451-1455. [PubMed]
 
Kahn JM, Goss CH, Heagerty PJ, et al. Hospital volume and the outcomes of mechanical ventilation. N Engl J Med. 2006;355:41-50. [PubMed]
 
Murphy-Filkins R, Teres D, Lemeshow S, et al. Effect of changing patient mix on the performance of an intensive care unit severity-of-illness model: how to distinguish a general from a specialty intensive care unit. Crit Care Med. 1996;24:1968-1973. [PubMed]
 
Kollef MH, Ahrens TS, Shannon W. Clinical predictors and outcomes for patients requiring tracheostomy in the intensive care unit. Crit Care Med. 1999;27:1714-1720. [PubMed]
 
Rady MY, Ryan T. Perioperative predictors of extubation failure and the effect on clinical outcome after cardiac surgery. Crit Care Med. 1999;27:340-347. [PubMed]
 
Kramer A. The APACHE IV equations: benchmarks for mortality and resource use [white paper report].Accessed June 1, 2008 Available at:http://www.cerner.com/public/Cerner_3.asp?id=27300.
 
Dasta JF, McLaughlin TP, Mody SH, et al. Daily cost of an intensive care unit day: the contribution of mechanical ventilation. Crit Care Med. 2005;33:1266-1271. [PubMed]
 
Halpern NA, Pastores SM, Greenstein RJ. Critical care medicine in the United States 1985–2000: an analysis of bed numbers, use, and costs. Crit Care Med. 2004;32:1254-1259. [PubMed]
 
Needham DM, Bronskill SE, Rothwell DM, et al. Hospital volume and mortality for mechanical ventilation of medical and surgical patients: a population-based analysis using administrative data. Crit Care Med. 2006;34:2349-2354. [PubMed]
 
Griffiths J, Barber VS, Morgan L, et al. Systematic review and meta-analysis of studies of the timing of tracheostomy in adult patients undergoing artificial ventilation. BMJ. 2005;330:1243. [PubMed]
 
Kahn JM, Linde-Zwirble WT, Wunsch H, et al. Potential value of regionalized intensive care for mechanically ventilated medical patients. Am J Respir Crit Care Med. 2008;177:285-291. [PubMed]
 
NOTE:
Citing articles are presented as examples only. In non-demo SCM6 implementation, integration with CrossRef’s "Cited By" API will populate this tab (http://www.crossref.org/citedby.html).

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging & repositioning the boxes below.

Find Similar Articles
CHEST Journal Articles
PubMed Articles
  • CHEST Journal
    Print ISSN: 0012-3692
    Online ISSN: 1931-3543