Chest ACCP Education Calendar
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Guest Access | Sign In via User Name/Password
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Shoemaker, W. C.
Right arrow Articles by Jelliffe, R. W.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Shoemaker, W. C.
Right arrow Articles by Jelliffe, R. W.
(Chest. 2005;128:2739-2748.)
© 2005 American College of Chest Physicians

Outcome Prediction in Chest Injury by a Mathematical Search and Display Program*

William C. Shoemaker, MD; David S. Bayard, PhD; Charles C. J. Wo, BS; Linda S. Chan, PhD; Li-Chien Chien, MD; Kevin Lu, MD and Roger W. Jelliffe, MD

* From the Laboratory of Applied Pharmcokinetics (Dr. Jelliffe), and the Department of Surgery, LAC+USC Medical Center (Drs. Shoemaker, Chan, Chien, Lu, Bayard and Mr. Wo), Keck School of Medicine, University of Southern California, Los Angeles CA; and the Jet Propulsion Laboratory (Dr. Bayard), Pasadena, CA.

Correspondence to: William Shoemaker, MD, LAC+USC Medical Center, Room 9900, 1200 North State St, Los Angeles, CA 90033; e-mail: wcshoemaker00{at}hotmail.com


   Abstract
TOP
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Objective: This study applies a stochastic or probability search and display model to prospectively predict outcome and to evaluate therapeutic effects in a consecutively monitored series of 396 patients with severe thoracic and thoracoabdominal injuries.

Study design: Prospective observational study of outcome prediction using noninvasive hemodynamic monitoring by previously designed protocols and tested against actual outcome at hospital discharge in a level 1 trauma service of a university-run, inner-city public hospital.

Methods: Cardiac index (CI), heart rate (HR), mean arterial pressure (MAP), arterial oxygen saturation measured by pulse oximetry (SpO2), transcutaneous oxygen tension (PtcO2), and transcutaneous carbon dioxide tension (PtcCO2) were monitored beginning shortly after admission to the emergency department. The stochastic search and display model with a decision support program based on noninvasive hemodynamic monitoring was applied to 396 severely ill patients with major thoracic and thoracoabdominal trauma. The survival probability (SP) was calculated during initial resuscitation continuously until patients were stabilized, and compared with the actual outcome when the patient was discharged from the hospital usually a week or more later.

Results: The CI, SpO2, PtcO2, and MAP were appreciably higher in survivors than in nonsurvivors. HR and PtcCO2 were higher in the nonsurvivors. The calculated SP in the first 24-h observation period of survivors of chest wounds averaged 83 ± 18% (± SD) and 62 ± 19% for nonsurvivors. Misclassifications were 9.6%. The relative effects of alternative therapies were evaluated before and after therapy, using hemodynamic monitoring and SP as criteria.

Conclusions: Noninvasive hemodynamic monitoring with an information system provided a feasible approach to early outcome predictions and therapeutic decision support.

Key Words: cardiac index • chest injury • noninvasive hemodynamic monitoring • outcome prediction • stochastic analysis and control program • tissue perfusion by transcutaneous O2


   Introduction
TOP
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
The advantage of noninvasive hemodynamic monitoring systems in acute life-threatening circulatory disorders is that temporal circulatory patterns can be continuously displayed on-line in real-time.123 Although invasive pulmonary artery (PA) thermodilution catheters provide the maximum circulatory data obtained at intervals, they require ICU conditions and, therefore, goal-directed therapy may be delayed until the patient is admitted to an ICU bed.45678

The major assumptions in the present approach are, first, that circulatory deficiencies that ultimately lead to shock, organ failure, and death may be identified by noninvasive monitoring; and, second, that the survival probability (SP) may be predicted early by an information system involving serial hemodynamic patterns in patients with the same diagnosis and covariants.678 Further, the severity of illness may be quantitatively assessed from the SP: the patient with 10% SP is very ill, but the patient with 90% SP is not very ill.

When invasive monitoring was started late in the course of illness or after onset of organ failures, there was no outcome improvement with PA catheters and goal-directed therapy.678910111213141516171819 By contrast, early noninvasive monitoring with an outcome predictor may be a useful approach to identify and correct hemodynamic deficiencies as early as possible.12320 Time may be an important factor in resuscitation and expeditious management of acute emergency patients because delays in correcting circulatory deficiencies of shock and trauma patients lead to organ failures and death.47 Similar to early diagnosis and therapy for cancer, early diagnosis and therapy for circulatory problems may be more cost-effective than therapy delayed until late stages.

Previous studies2 have demonstrated early outcome prediction using discriminant function analysis together with noninvasive hemodynamic monitoring, and a preliminary report3 has introduced the concept a mathematical search and display model to further improve outcome prediction. The present study applies this approach to prediction and therapeutic decision making in chest injuries.


   Materials and Methods
TOP
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Clinical Series
We studied 396 consecutively noninvasively monitored emergency patients with major blunt or penetrating injuries to the chest or chest and abdomen with significant risk of mortality or morbidity within 24-h after emergency department (ED) admission. Patients with major head injury with Glasgow coma scale scores < 6 were excluded as they have been shown to have different hemodynamic patterns.21 All patients who fit the selection criteria were studied. There were 340 survivors and 56 patients who died during their current hospitalization. The hospital mortality was 14%. Noninvasive monitoring was usually begun in the ED shortly after admission, and the patients were followed to the radiology suite when indicated, to the operating room, and then to the ICU (Fig 1 ). PA catheterization was performed when clinically indicated in the ICU. The time of monitoring, time of operations, and times of ICU admission and discharge or death were recorded relative to the time elapsed after ED admission. We also included in the database the following: age, gender, presence of sepsis or the systemic immune response system, APACHE (acute physiology and chronic health evaluation) score,22 Glasgow coma scale,23 injury severity score,24 primary bodily injuries, covariates, hemodynamic patterns by invasive and noninvasive methods, organ failures, other complications, hospital days, ICU days, and hospital outcome. Table 1 lists the salient clinical features. The Institutional Review Board approved the protocol.



View larger version (43K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1.. Hemodynamic patterns and the effects of resuscitation therapy on hemodynamic patterns and survival probability of a 33-year-old man who sustained gunshot wounds to the right chest and abdomen with lacerations of the colon, small bowel, stomach, and spleen with estimated blood loss of 5,000 mL. This was replaced with the rapid transfusions of 10 U of packed RBCs, 5 U of FFP, 500 mL hydroxyethyl starch (Hes), and another 2 U of RBCs along with 1,500 mL Ringer’s lactate solution per hour. After ICU admission, the patient received 2 more units of blood, 1,500 mL 5% albumin, and 2 U of FFP. Note the initial reductions in hemodynamic values and survival probability in the first hour after hospital admission and recovery with control of bleeding and surgical repair of injuries. Time in hours from ED admission is noted below the bottom horizontal line. Therapies are outlined in boxes. DOB = dobutamine; ALB = albumin. Lactated Ringer’s solution was administered at the rate of 500 mL/h during the operation, and 150 mL/h postoperatively. Time in the operating room (OR) and ICU is indicated at the lowest line. The patient survived.

 

View this table:
[in this window]
[in a new window]

  		Table 1.. Clinical Features of the Series*

 
Noninvasive Hemodynamic Monitoring
Hemodynamic values were evaluated by continuous display of noninvasive monitoring of cardiac, respiratory, and tissue perfusion functions. The data were downloaded every 30 s, averaged over 5-min intervals, and entered into the database. When consistent hemodynamic patterns were demonstrated, they were averaged over 4-h periods for presentation (Fig 2 ).



View larger version (17K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2.. Survivor (solid line) and nonsurvivor (dashed line) temporal patterns are shown for the first 24 h after ED admission. Mean values ± SEM are shown for CI, HR, MAP, SpO2, PtcO2/FIO2, and SP. All values are keyed to the time of admission to the ED. Note the survivors’ CI, MAP, SpO2, PtcO2/FIO2, DO2, and SP values were generally higher than those of the nonsurvivors, while HR and PtcCO2 were higher in the nonsurvivors. The mean survivor SP values were significantly higher than the mean nonsurvivor SP values in this initial resuscitation period.

 
Cardiac Output
A thoracic bioelectric impedance device (IQ 101; Noninvasive Medical Technology; Auburn Hills, MI) was applied shortly after arrival in the ED. The noninvasive, disposable, prewired, hydrogel electrodes were positioned on the skin, and three ECG leads were placed on the precordium and right shoulder.2526 Previous studies12 have documented satisfactory correlations between thermodilution and bioimpedance cardiac output values for trauma patients. In the present series, the correlation of cardiac output estimated by impedance vs thermodilution in 907 simultaneous measurements obtained in the operating room and ICU was r = 0.88 (r2 = 0.77, p < 0.01). The precision and bias of cardiac index (CI) estimated by impedance vs thermodilution methods were – 0.17 ± 0.77 L/min/m2. This was similar to a previously reported series.1

Pulse Oximetry
Arterial oxygen saturation measured using pulse oximetry (SpO2) [N-200; Nellcor; Pleasanton, CA] was performed continuously. Values were recorded at the time of the CI measurements. Sudden changes in these values were confirmed by in vitro arterial oxygen saturation obtained by standard blood gas analysis.2

Transcutaneous Oxygen Tension
Conventional transcutaneous oxygen tension (PtcO2) indexed to the fraction of inspired oxygen (FIO2) [PtcO2/FIO2] was calculated because changes of the inspired oxygen produce marked PtcO2 changes throughout the observation period. This technology uses the Clark polarographic oxygen electrode routinely employed in standard blood gas measurements.2127282930 PtcO2 and transcutaneous carbon dioxide tension (PtcCO2) were measured in a representative area of the skin surface heated to 44°C to increase diffusion of oxygen across the stratum corneum and to avoid vasoconstriction in the local area of the skin being measured.29 PtcO2 does not necessarily reflect the state of oxygenation of all tissues, but skin vasoconstriction is an early stress response of hypovolemia and shock.2127282930 Limitations of the transcutaneous methods are that the thermal environment must be reasonably constant. Marked changes in room temperature from drafts or open windows must be avoided. The electrode must be changed to a nearby site and be recalibrated at 4-h intervals to avoid skin erythema.

Patient Database
A database for emergency and acutely injured patients was acquired using software (Excel; Microsoft; Redmond, WA) to describe thoracic injuries, selected covariates, hemodynamic patterns by invasive and noninvasive methods, and outcomes, including hospital survival or death during current hospitalization, organ failures, other complications, hospital days, and ICU days. Covariates included age, gender, the presence of sepsis, systemic immune response system, history of cardiac conditions, prior respiratory conditions, prior renal insufficiency or failure, diabetes, delays in seeking medical care, delays in immediately replacing blood, and fluid losses.

Mathematical Search and Display Program
Bayard et al20 developed a stochastic (probability) search and display program to determine individual SPs from a database of patients with similar clinical/hemodynamic "states," which are defined in terms of the primary diagnosis, covariates, and hemodynamic patterns of changes. Similar is defined as a group of patients, referred to as "nearest neighbors," with the same diagnosis, who share the same or similar covariates, and were found to have the closest hemodynamic patterns to the patient under study. The term similar is meant to express how close the index patient’s data are to the survivors’ patterns and how far from the nonsurvivors’ patterns in comparison with other the database patients.

Mathematically, the stochastic analysis is defined as policy iteration with respect to conventional therapeutic policy used for each patient as the database was developed. It is motivated by methods of machine learning,3132 and methods of dynamic programming for stochastic control.3334 The SPs are estimated from a database of patients with very similar clinical/hemodynamic states, defined by the primary diagnosis, covariates, and hemodynamic variables. The outcome prediction program searches the entire database for patients with identical diagnoses and covariates and with the closest hemodynamic patterns. These patients with the closest clinical/hemodynamic states are statistically referred to as nearest neighbors, and they are used as surrogates for the newly admitted study patient. That is, the survival percentage of these nearest neighbors, whose outcome has been recorded in the database, provides a quantitative estimate of the likelihood of survival of the newly admitted or "index" patient. For example, if 80% of the nearest neighbors survive, it is estimated that the index patient will have an 80% likelihood of survival. The severity of illness may also be estimated from the SP. If the SP is only 20%, the patient is gravely ill; but if the SP is 90%, he is not very ill. Similarly, the nearest neighbors may be used as surrogates to anticipate therapeutic responses of the index patient. For example, if 10 of the nearest neighbors receiving a transfusion of packed RBCs had average CI increase of 0.5 L/min/m2, this response may be compared with the responses of other agents, including crystalloids, colloids, vasopressors, vasodilators, inotropic drugs, and blocking agents. Then the attending physician, armed with the predicted likely responses of his index patient to alternative therapies, may select any one of the agents, or he may administer another agent of his own choosing. In any case, he may observe the actual result of his choice and compare this with the predicted values. The mathematical details of this stochastic analysis have been described.2025

SP
A patient’s SP for a given state x is denoted by S(x), which is calculated by first extracting 40 or more nearest-neighbor states of patients having the same diagnosis and covariates as well as hemodynamic values that are closest to the index patients’ values. The database contains > 1,000 high-risk trauma patients with > 300,000 time lines, each of which represents a patient’s clinical/hemodynamic state. The closest set of values may be selected as a nearest neighbor. The SP is then calculated as the fraction of these nearest neighbors who survived. The SP may serve as an outcome predictor and by the same token as a measure of severity of illness. The SP after initial resuscitation in the first 24 h was compared with in-hospital mortality at the time of hospital discharge, which were usually ≥ 1 weeks after hospital admission.

Statistical Analyses
The survivor and nonsurvivor deficits of mean arterial pressure (MAP), cardiac output, SpO2, and PtcO2 were calculated for the periods of monitoring. For categorical variables, differences in proportions between survivors and nonsurvivors were determined using Excel, GraphPad Prism (GraphPad Software; San Diego, CA), and SPSS (11.5 for Windows; SPSS; Chicago, IL) for all statistical computations.


   Results
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
Hemodynamic and SP Values From the Time of Hospital Admission
Table 2 lists the mean values (± SD) for survivors and nonsurvivors for thoracic and thoracoabdominal wounds during the first 24 h after hospital admission for CI, heart rate (HR), MAP, SpO2, PtcO2/FIO2, oxygen delivery (DO2), and SP. The CI, MAP, Spo2, PtcO2/FIO2, DO2, and SP values of the survivors were significantly higher than the corresponding values of those who died, while the HR was higher in the nonsurvivors during the first 24 h after ED admission. Figure 2 illustrates the time course of survivor and nonsurvivor hemodynamic patterns calculated over intervals during the first 24-h period. The mean SP of survivors during the first 24 h was 83 ± 18%. The SP for nonsurvivors was 62 ± 19%. Since the differences between the thoracic and thoracoabdominal injuries were very similar and not statistically significant, the two groups were combined for additional analysis.


View this table:
[in this window]
[in a new window]

  		Table 2.. Survival Probability and Hemodynamic Values in the First 24 h*

 
SP Calculation Compared With Actual Hospital Outcome
The SPs were continuously calculated until the patient was stable. At discharge from the hospital, the predicted values were compared with the actual hospital outcome. We used 72% as the cut point half way between the survivor and nonsurvivor mean values for the first 24 h. Of 396 chest trauma patients, there were 38 misclassifications, or 9.6% (Table 3 ). Slightly > 90% of the patients were correctly classified after the initial resuscitation within 12 h after admission to the ED. There is an appreciable bias, however, with different sized cells: 86% survivors but only 14% deaths. However, most trauma series have similarly inequality for this analysis.


View this table:
[in this window]
[in a new window]

  		Table 3.. Classifications of Chest Injury Patients (n = 396)*

 
Therapeutic Decision Support System
Preliminarily, the therapeutic decision support system was used in 698 interventions in these acutely ill patients during and shortly after their initial resuscitation. Figure 3 is an example of the use of the therapeutic decision support system for a patient 28 min after ED admission. Figure 3 shows the nearest neighbors’ responses to various therapeutic interventions measured before and after each therapeutic intervention. The prospectively estimated therapeutic responses of each study patient’s nearest neighbors could be compared with the patient’s actual responses. Table 4 summarizes the effects of 698 therapeutic interventions including 1 U packed RBCs, 1 L of crystalloids, usually lactated Ringer’s solution, or 500 mL of colloids, fresh frozen plasma (FFP), 5% albumin, or 6% hydroxyethyl starch, Since there were no significant differences between the effects of each of these colloids, the data of the three colloids were combined. From these preliminary results, packed RBCs and colloids improved CI and PtcO2/FIO2 in survivors but not in late nonsurvivors. Further controlled studies are necessary to objectively evaluate the role of this program in the clinical setting.



View larger version (114K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3.. Data of a patient obtained 28 min after ED admission. The upper half of the window, left side, shows the calculated SP (69%) and most recent therapy received. The center section has two columns: the current hemodynamic values (CI, HR, MAP, SpO2, PtcO2/FIO2, PtcCO2, and hematocrit value in the first column, and the net cumulative excess (+) or deficit (–) of each variable up to this point in time in the second column). In the lower half of the window, the first column shows the number of nearest neighbors receiving each therapy. The second column shows the average SP of these nearest neighbors before therapy. Column 3 shows the number of nearest neighbors of the nearest neighbors who received each of the therapies specified in column 6. Column 4 shows the SP of these nearest neighbors after each therapy. Column 6 is an abbreviated list of a few of the therapies that may be administered.

 

View this table:
[in this window]
[in a new window]

  		Table 4.. Effects of Fluid Therapy After Chest Injury*

 
Receiver Operating Characteristic Curves
Receiver operating characteristic (ROC) curves of data were collected over the first 4 h after ED admission (Fig 4 ), showing the survival predictor together with the other hemodynamic values. The size of the area under the curves represents the sensitivity and specificity for each variable: 1.00 represents 100% correct, and 0.50 represents no correlation. These areas were for 0.83 for SP, 0.80 for PtcO2/FIO2, 0.64 for SpO2, 0.61 for MAP, and 0.53 for CI.



View larger version (18K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4.. ROC curves for MAP, HR, CI, SpO2, PtcO2/FIO2 and SP values of survivors and nonsurvivors during the first 4 h after ED admission. The calculated areas under the curves were for SP (0.83), PtcO2/FIO2 (0.80), SpO2 (0.64), MAP (0.61), and CI (0.53).

 

   Discussion
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
The conventional approach to evaluate controversial therapeutic decisions is to generate prospective randomized control studies. When the series from various hospitals differ, larger numbers of patients or multicenter series may overcome these differences. This approach addresses the problem of generalizability of the results. However, it has the limitation that wide disparities of clinical conditions in large series may not give clear unambiguous solutions to the specific problems of the recently admitted study patient. Because large series of patients may lose specificity with respect to the individual patient at hand, the key question becomes, "What is the appropriate comparison group for this individual patient?" The present study was designed to answer this question by providing a large database with specifically designated diagnoses, covariants, and hemodynamic patterns for each entry, so that subsets of very similar patients would be available to compare with data from the study patient.

The proposed mathematical representation of the circulatory status defines the patient’s clinical/circulatory state by specific diagnostic categories, clinical covariates, hemodynamic variables, their first and second derivatives, and their integrals. Simply stated, the program selects patients in the database with identical diagnosis and covariates, who have hemodynamic patterns that are the closest to the newly admitted study patient, ie, nearest neighbors. These nearest neighbors then serve as surrogates for the study patient.

The accuracy and reliability of this approach depends on the size and comparability of the database needed to provide an adequate group of nearest neighbors. The present database contains > 1,000 high-risk patients with > 300,000 time lines, each of which represents a patient’s clinical/hemodynamic state that potentially may be used as a nearest neighbor. The average difference between a given patient’s variables and the nearest neighbors was < 0.3 of the SD for the variables in the database.

The proposed approach is similar to that of experienced clinicians who on seeing an unusual patient, recall similar patients who responded well to specific therapies. The proposed information system attempts to emulate the processes of good clinical judgment by using a computerized program that obviates memory deficiencies and a large database of similar patients for use as surrogates for the newly admitted study patient. The program searches the database in an analogous manner to find very similar patients and to quantitatively evaluate the relative effectiveness of therapy given to each of them. This approach was tested in the extenuating circumstances of severely traumatized emergency patients in a large, inner-city, public hospital ED. Notwithstanding, the survival probability was found to be a feasible way to track changes throughout the initial observation period. Moreover, during the initially monitored resuscitation, the program correctly predicted hospital outcome in slightly > 90% of the patients. As in cancer, early diagnosis, prognosis, and assessment of physiologic alterations are essential, because this allows therapy to be initiated sooner in the hope that earlier therapy may improve outcome in emergency conditions when time is crucial.67891011

The aim of the therapeutic decision part of the study was to gain experience with the use of this tool in the initial resuscitation when the patients were not completely stabilized and continuing blood and fluid losses could not be entirely ruled out. When blood and fluid losses do occur, the real effects of the infusions may be expected to be underestimated. Since there is no way to know if bleeding is continuing, or how much bleeding is occurring, these data cannot be taken without reservation. If rigorous attention to this problem is required, the data may be ignored. However, if the needs of these severely ill paints are to be addressed, it might be appropriate to consider, albeit tentatively, that the survivors had greater responses to fluid administration compared with the nonsurvivors’ responses, and that packed RBC transfusions and colloids had somewhat greater responses than did Ringer’s lactate solution.

In conclusion, the proposed stochastic search and display program provides an independent mathematical tool to objectively predict outcome, to track circulatory changes throughout the acute illness, and to evaluate relative effectiveness of various therapeutic responses. If confirmed by further studies, this type of information system may provide an objective method to manage chest trauma and other acute critically illnesses.


   Footnotes
 
Abbreviations: APACHE = acute physiology and chronic health evaluation; CI = cardiac index; DO2 = oxygen delivery; ED = emergency department; FFP = fresh frozen plasma; FIO2 = fraction of inspired oxygen; HR = heart rate; MAP = mean arterial pressure; PA = pulmonary artery; PtcCO2 = transcutaneous carbon dioxide tension; PtcO2 = transcutaneous oxygen tension; PtcO2/FIO2 = transcutaneous oxygen tension indexed to the fraction of inspired oxygen; ROC = receiver operating characteristic; SP = survival probability; SpO2 = arterial oxygen saturation measured by pulse oximetry

Supported in part by National Institutes of Health grants RR-11526, and GM-65619, and DOD BAA99–1, Award Number DAMD 17–01-2–0070 by the US Army Medical Research Acquisition Activity, Fort Detrick, MD. The content of the information does not necessarily reflect the position or the policy of the US Government, and no official endorsement should be inferred.

Received for publication March 29, 2004. Accepted for publication April 12, 2005.


   References
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Shoemaker, WC, Belzberg, H, Wo, CCJ, et al (1998) Multicenter study of noninvasive monitoring systems as alternatives to invasive monitoring of acutely ill emergency patients. Chest 114,1643-1652[Abstract/Free Full Text]
  2. Shoemaker, WC, Wo, CCJ, Chan, L, et al Outcome prediction of emergency patients by noninvasive hemodynamic monitoring. Chest 2001;120,528-537[Abstract/Free Full Text]
  3. Shoemaker, WC, Wo, CCJ, Bishop, MH, et al Multicenter trial of a new thoracic electrical bioimpedance device for cardiac output estimation. Crit Care Med 1994;22,1907-1912[ISI][Medline]
  4. Bishop, MH, Shoemaker, WC, Appel, PL, et al Relationship between supranormal values, time delays and outcome in severely traumatized patients. Crit Care Med 1993;21,56-63[ISI][Medline]
  5. Bishop, MH, Shoemaker, WC, Kram, HB, et al Prospective randomized trial of survivor values of cardiac index, oxygen delivery, and oxygen consumption as resuscitation endpoints in severe trauma. J Trauma 1995;38,780-787[ISI][Medline]
  6. Boyd, O, Bennett, D Enhancement of perioperative tissue perfusion as a therapeutic strategy for major surgery. New Horiz 1996;4,453-465[Medline]
  7. Kern, JW, Shoemaker, WC Meta-analysis of hemodynamic optimization in high-risk patients. Crit Care Med 2002;30,1686-1692[CrossRef][ISI][Medline]
  8. Boyd, O, Hayes, M The oxygen trail: the goal. Brit Med Bull 1999;55,125-139[Abstract/Free Full Text]
  9. Shoemaker, WC, Appel, PL, Kram, HB, et al Prospective trial of supranormal values of survivors as therapeutic goals in high-risk surgical patients. Chest 1988;94,1176-1186[Abstract/Free Full Text]
  10. Boyd, O, Grounds, RM, Bennett, ED Preoperative increase of oxygen delivery reduces mortality in high-risk surgical patients. JAMA 1993;270,2699-2707[Abstract]
  11. Wilson, J, Woods, I, Faucett, J, et al Reducing the risk of major elective surgery: randomized control trial of preoperative optimization of oxygen delivery. BMJ 1999;318,1099-1103[Abstract/Free Full Text]
  12. Connors, AF, Jr, Speroff, T, Dawson, NV, et al The effectiveness of right heart catheterization in the initial care of critically ill patients. JAMA 1996;276,889-897[Abstract]
  13. Guyatt, G A randomized control trial of right heart catheterization in critically ill patients. Ontario Intensive Care Group. J Intensive Care Med 1991;6,91-95[Medline]
  14. Zion, MM, Balkin, J, Rosenmann, D, et al Use of pulmonary artery catheters in patients with acute myocardial infarction: analysis of experience in 5,841 patients in the SPRINT registry. Chest 1990;98,1331-1335[Abstract/Free Full Text]
  15. Hayes, MA, Timmins, AC, Yau, EHS, et al Elevation of systemic oxygen delivery in the treatment of critically ill patients. N Engl J Med 1994;330,1717-1722[Abstract/Free Full Text]
  16. Gattinoni, L, Brazzi, L, Pelosi, P, et al A trial of goal-oriented hemodynamic therapy in critically ill patients. N Engl J Med 1995;333,1025-1032[Abstract/Free Full Text]
  17. Taylor, RW Pulmonary Artery Catheter Consensus Conference. Crit Care Med 1997;25,910-925[CrossRef][ISI][Medline]
  18. Bernard, GR, Artegas, A, Brigham, KL, et al The American-European Consensus Conference on ARDS: definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994;149,818-824[Abstract]
  19. Artegas, A, Bernard, GR, Carlet, J, et al The American-European Consensus Conference on ARDS, part 2: Ventilatory, pharmacologic, supportive therapy, study design strategies, and issues related to recovery and remodeling; acute respiratory distress syndrome. Am J Respir Crit Care Med 1998;157,1332-1347[Abstract/Free Full Text]
  20. Bayard, DS, Botnen, A, Shoemaker, WC, et al Stochastic analysis of therapeutic modalities using a database of patient responses. IEEE Symposium on Computer Based Medical Systems (CBMS) 2001;11,439-444
  21. Shoemaker, WC, Bayard, DS, Wo, CCJ, et al A mathematical program for outcome prediction and therapy for trauma beginning within 1-hour after admission. Crit Care Med 2005;33,1499-1506[CrossRef][ISI][Medline]
  22. Seneff, M, Knaus, WA APACHE: a prognostic scoring system. Probl Crit Care ;3,563-577
  23. Teasdale, G, Jennett, B Assessment of coma and impaired consciousness: a practical scale. Lancet 1974;2,81-84[CrossRef][ISI][Medline]
  24. Demetriades, D, Chan, LS, Velmahos, G, et al TRISS methodology in trauma. Br J Surg 1998;85,379-384[CrossRef][ISI][Medline]
  25. Tremper, KK, Waxman, K, Shoemaker, WC Effects of hypoxia and shock on transcutaneous PO2 values in dogs. Crit Care Med 1979;7,526-531[ISI][Medline]
  26. Wang, X, Van De Water, JM, Sun, H, et al Hemodynamic monitoring by impedance cardiography with an improved signal processing technique. Proc IEEE Eng Med Biol 1993;15,699-700
  27. Tremper, KK, Shoemaker, WC Transcutaneous oxygen monitoring of critically ill adults with and without low flow shock. Crit Care Med l981;9,706-709
  28. Tatevossian, RG, Shoemaker, WC, Wo, CCJ, et al Noninvasive hemodynamic monitoring for early warning of adult respiratory distress syndrome in trauma patients. J Crit Care 2000;15,151-159[CrossRef][ISI][Medline]
  29. Tremper, KK, Huxtable, RF Dermal heat transport analysis for transcutaneous O2 measurements. Acta Anesth Scand 1978;68(suppl),4-9
  30. Tremper, KK, Waxman, K, Bowman, R, et al Continuous transcutaneous oxygen monitoring during respiratory failure, cardiac decompensation, cardiac arrest, and CPR. Crit Care Med 1980;8,237-238
  31. Sutton, RS, Barto, AG Reinforced learning. 1998 MIT Press. Cambridge, MA:
  32. Bertsekas, DP, Tsitsiklis, JN Neuro-dynamic programming. 1996 Athena Scientific. Belmont, MA:
  33. Bayard, DS A forward method for optimal stochastic nonlinear and adaptive control. IEEE Trans Auto Control 1991;36,1046-1053[CrossRef]
  34. Bayard, DS Reduced complexity dynamic programming based on policy iteration. J Math Anal Appl 1992;170,75-103[CrossRef]




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Shoemaker, W. C.
Right arrow Articles by Jelliffe, R. W.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Shoemaker, W. C.
Right arrow Articles by Jelliffe, R. W.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS