Therapy of Critically Ill Postoperative Patients Based on Outcome Prediction and Prospective Clinical Trials

Symposium on Critical Care

Therapy of Critically Ill Postoperative Patients Based on Outcome Prediction and Prospective Clinical Trials William C. Shoemaker, M.D.,* Richard D. Bland, M.D., t and Paul L. Appel, M .P .A .:f:

Traditional concepts of surgical management were centered on technical aspects of the surgical correction of anatomic lesions in the operating room. By contrast, postoperative intensive care unit (ICU) management has traditionally focused on physiologic and biochemical deficiencies that are identified and then corrected. Usually, this one-at-a-time search for defects after they occur results in fragmented, episodic care. Moreover, this traditional approach usually assumes that normal values are the appropriate therapeutic goals and does not take into account the increased metabolic and hemodynamic needs of the critically ill patient. Sophisticated physiologic monitoring is a major function of the ICU that provides important diagnostic as well as hemodynamic and oxygen transport data needed to develop rational criteria for clinical decision making. However, invasive hemodynamic and oxygen transport monitoring is usually started after the patient is in critical condition, with lethal complications. By this time, the patient has had various therapies that make it impossible to differentiate the prior physiologic deficits from the effects of therapy; that is, dehydration, hypoxemia, hypotension, hypovolemia, and oliguria may have been overcorrected, but the patient is still in shock and unresponsive to therapy. In most shock patients, appropriate therapy is eventually given but not necessarily at the right time, in optimal doses, or in the proper order. Marked differences were observed between survivors and nonsurvivors' hemodynamic patterns during the early postoperative course of critically ill surgical patients, but no one physiologic variable was able to predict outcome. However, there are patterns of physiologic events that *Professor of Surgery, Department of Surgery, UCLA School of Medicine, Los Angeles, California tSurgical Resident, Department of Surgery, Los Angeles County Harbor/UCLA Medical Center, Torrance, California tResearch Associate, Department of Surgery, Los Angeles County Harbor/UCLA Medical Center, Torrance, California

Surgical Clinics of North America-Yo!. 65, No.4, August 1985

811

812

WILLIAM

C.

SHOEMAKER, RICHARD

D.

BLAND AND PAULL. APPEL

characterize the survivors and nonsurvivors of life-threatening surgical problems. It is also intuitively obvious that no single factor is responsible for all postoperative deaths; rather, mortality is a multifactorial problem that requires multivariate analysis. This article describes methods for prediction of outcome, definition of therapeutic goals, and results of clinical trials using prospectively the median values of the survivors as therapeutic goals.

PREDICTION OF OUTCOME A number of different severity (or predictive) indexes have been developed for various conditions, including Trauma Severity Scores, 2• 8-10 Glasgow Coma Scale, 31 Therapeutic Intervention Scoring System (TISS),n. 12 arterial lactate levels,28. 32 hemodynamic predictor of Siegel and colleagues, 29 the acute physiology and chronic health evaluation (APACHE) system, 13 the ICU outcome measure, 30 and the Norris 14 predictive index for acute myocardial infarction. Over the past 10 years, we have developed several objective, empiric models based on the distribution of cardiorespiratory values of survivors and nonsurvivors at various arbitrarily defined time periods. 4, s. 20. 22. 24 These and other efforts have stimulated considerable interest for more objective and rational criteria for clinical decision making. Many of these predictive indexes are based on either retrospective correlations or normative standards generated by a panel of experts. More sophisticated approaches to pattern recognition involve cluster analyses and other complex computer programs, whereas others require only hand-held calculators to generate their algorithms or to calculate their prediction for a given patient.

ASSESSMENT OF SEVERITY OF ILLNESS

Both the severely traumatized patient and the critically ill postoperative patient may have sudden unexpected catastrophic circulatory problems of postoperative shock or its sequelae-shock lung, renal failure, sepsis, disseminated intravascular coagulation (DIC), and so on. Obviously, the urgency of these unanticipated problems requires special care to handle specific life-threatening emergencies. Although numerous studies have claimed efficacy in treatment of emergency patients, they are often seriously Hawed because they use historical rather than concurrent controls and because they have not rigorously developed an appropriate experimental design for prospective testing. Comparative studies on the relative efficacy of specific medical problems would be greatly facilitated by the development of an accurate, reproducible index of severity based on outcome measures. This severity index also could be used as a "proxy outcome" measure to evaluate the relative effectiveness of various therapies; that is, improvement or lack of improvement may be an objective physiologic criterion for evaluating alternative therapeutic modalities.

813

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

SELECTION OF VARIABLES FOR OUTCOME PREDICTION

Table 1 summarizes over 50,000 measurements of the most commonly monitored variables in a series of critically ill postoperative survivors and nonsurvivors. These variables include vital signs, mean arterial pressure (MAP), heart rate (HR), central venous pressure (CVP), pulmonary artery wedge pressure (WP), and cardiac output. We were able, with vigorous therapy, to bring these values back into the normal range in 76 per cent of the nonsurvivors; nevertheless, they still died. By comparison, 75 per cent of the survivors also had two or more values in the normal range. Clearly, we were measuring the wrong variables and had the wrong values as therapeutic goals for these variables; in essence, we have relied on custom rather than on objective criteria for the selection of physiologic variables to be measured and for the selection of appropriate goals of therapy. 6• 23 Although a few controlled, prospective studies have been performed, considerable work in this area needs to be done to determine criteria for the following: which variables are most relevant to the most important biologic end point, survival or death; in what circumstances these monitoring variables are appropriate; what combination of variables is most useful for the initial resuscitation as well as subsequent monitoring in critical periods; which variables are valuable for clinical decision making, such as selection of specific therapy and the titration of this therapy to optimal goals; and which variables are useful for surveillance of postoperative patients who are not considered critical or who do not have complications. 23 The percentage of correct outcome predictions for each cardiorespiratory variable is a reasonable indication of its biologic significance as well as its relevance to clinical management. That is, if a variable is unable to differentiate the dying patient from the patient who survives, it is not very useful or relevant. On the other hand, if it is an excellent predictor of outcome, it may be related to an important pathophysiologic problem and, therefore, is a useful criterion for therapeutic decision making. The percentages of correct outcome predictions for each cardiorespiratory variable were evaluated at each stage and over all stages (Tables 2 to 4). 22-24 The most commonly measured variables were found to be the poorest, least relevant predictors. The advantage of the outcome predictors is that they are heuristically or phenomenologically determined and do not depend on

Table 1.

Number and Percentage of Patients with Two or More Values in the Normal Range SURVIVORS

NONSURVIVORS

Mean arterial pressure (MAP) Heart rate (HR) Central venous pressure (CVP) Pulmonary wedge pressure (WP) Cardiac index (CI) Mean of these variables

Number

Percent

Number

Percent

29 30 35

78 81 95 30 95

68 66 72 21 64

89 87 95 28 84

ll

35

76

75

r:¥:J

...... ,j;o..

Table 2.

:;a....

Cardiorespiratory Variables: Abbreviations, Units, Calculations, Normal Values, and Optimal Values

t"'

ABREVIATIONS

UNITS

Volume-Related Variables Mean arterial pressure Central venous pressure Central blood volume

MAP CVP CBV

mmHg em H 20 ml per m2

Stroke index Hemoglobin Mean pulmonary arterial pressure Wedge pressure Blood volume

SI Hgb MPAP WP BV

ml per m2 gm per dl mmHg mm Hg ml per m2

Red cell mass

RCM

ml per m2

MEASUREMENTS OR CALCULATIONS

Direct measurement Direct measurement CBV = mean transit time X CI X 16.7 SI = CI + HR Direct measurement Direct measurement Direct measurement BV = plasma volume + (1 - Hct)* X body surface area RCM = BV- PV

Flow-Related Variables Cardiac index Left ventricular stroke work index Left cardiac work index Right ventricular stroke work index Right cardiac work index

CI LVSW LCW RVSW RCW

L per min·m2 gm·m per m2 kg·m per m2 gm·m per m2 kg·m per mz

Direct measurement LVSW = SI x MAP x 0.0144 LCW = CI x MAP x 0.0144 RVSW = SI x MPAP x 0.0144 RCW = CI x MPAP x 0.0144

Stress-Related Variables Systemic vascular resistance index Pulmonary vascular resistance index Heart rate Rectal temperature

SVRI PVRI HR temp

dyne·sec per cm 5 ·m2 dyne·sec per cm'·mz beat per minute

SVR = 79.92 (MAP - CVP)t + CI PVR = 79.92 (MPAP - WP)t + CI Direct measurement Direct measurement

NORMAL VALUES

OPTIMAL VALUES

82--102 1-9 660-1000

> 84 <5 > 925

30-50 12--16 11-15 0-12 men, 2.74 women, 2.37 men, 1.1 women, 0.95

> 48 > 12 < 19 > 9.5 > 3.0 > 2.7 >1.1 > 0.95

t"' ....

> ;:::: C1 en

:I: 0

trl

;::::

~

trl ~

"

Z? n

:I:

> ~ I:)

v tl::l

t"'

OF

2.8--3.6 44-68 3--4.6 4-8 0.4-0.6 1760-2600 45--225 72--88 97.8--98.6

> 4.5 >55 >5 > 13 > 1.1 < < < >

1450 226 100 100.4

> z I:)

> z I:)

~ et"'

r >

"C "C

trl t"'

Oxygen-Related Variables Hgb saturation Arterial C0 2 tension Arterial pH Mixed venous 0 2 tension Arterial-mixed venous 0 2 content difference 0 2 delivery 0 2 consumption 0 2 extraction rate Perfusion-Related Variables Red cell flow rate Blood flow-volume ratio 0 2 transport/red cell mass Tissue 0 2 extraction Efficiency of tissue 0 2 extraction 0 2 transport/red cell flow

0

Sao' Paco, pH PVo, C(a-v)o,

%

torr ml per dl

Direct measurement Direct measurement Direct measurement Direct measurement C(a-v)o, = Cao, - Cvo2

Do, Vo, 0 2 ext

ml per min· m2 ml per min·m 2 %

Do, = Cao 2 x CI x 10 Vo, = C(a-V)o 2 x CI x 10 0 2 ext = (Cao, - Cvo,) · Cao,

torr

95-99 36-44 7.36-7.44 33-53 4-5.5

> > > >

95 30 7.47 36 < 3.5

<:! ..., C"l

0

a::

1:'1

0

l'l:j

'"'d

520--720 100--180 22-30

> 550 > 167 < 31

1;] 0

n ..., .... 0

z

RCFR BFVR OTRM TOE ETOE OTRF

RCFR = CI x Hct BFVR = CI + BV OTRM = Vo, + RCM TOE = C(a-V)o, + RCFR ETOE = C(a-v)o, + RCM OTRF = Vo2 + RCFR

*Hct corrected for packing fraction and ratio of large-vessel hematocrit to total-body hematocrit. tVenous pressures expressed in mm Hg.

0.6-1.8 0.6-1.8 0.06-0.18 1.8-6.6 0.06-0.18

> > > > >

1.3 1.7 0.25 5.7 1.3 <3

~

0 '"'d ~ 0

"'1:''C1

C"l ..., ....

< 1:'1

(l

I:"'

z

n> I:"'

~

~

I:"'

"'

(;,:J ~

"'

816

WILLIAM

C.

SHOEMAKER, RICHARD

D.

BLAND AND PAULL. APPEL

Table 3. Rank and Percentage of Correct Predictions for Each Cardiorespiratory Variable Average Over all Stages* VARIABLE

RANK

1 2 3 4 5 6 7 8 9 10 11

12 13 14 15 16 17 18 19 20 21 22 23 24 25

26 27 28 29 30 31 32 33 34

ETOE RCM OTR PVR LCW BV MAP Do2 BFVR TOEI pH LVSW STTI RCFR MTT OTRF CI RVSW WP Vo2 Paco2 0 2 ext RCW P(A-a)02 Pv02 MPAP Sao2 SI MSER Hgb temp CVP SVR HR

PERCENT CORRECT

91 85 79 77 76 76 76 76 75 75 74 74 73 72 71 71

70 70 70 69 69 69 69 68 68 68 67 67 66 66 64

62 62 60

*The number of patient-stages is defined as the number of patients in whom the variable was measured during each stage; a variable measured in one patient throughout one stage constitutes one patient-stage. (From Shoemaker, W. C., and Czer, L. S. C.: Evaluation of the biologic importance of various hemodynamic and oxygen transport variables. Crit. Care Med., 7:424, 1979; with permission.)

"clinical opinion" or "party line." Moreover, they are not dependent on a given probability distribution, and in the statistical sense, they are "distribution free." The cardiorespiratory variables with the largest and smallest capability of predicting outcome at each stage were evaluated as the percentage of correct predictions 22 • 23 (Tables 3 and 4). These percentages have pronounced variations from stage to stage; this indicates their stage specificity. For example, 0 2 delivery (Do 2) and pulmonary vascular resistance index (PVRI) are good predictors in the early stage (B and Low Stages), but not

817

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

Table 4. STAGE B

Outcome Predictions for Cardiorespiratory Variables Greater than 75 Per cent and Less Than 67 Per cent in Each Stage STAGE LOW

STAGE C 1

STAGE C2

STAGES D, E, F

ETOE OTR RVSW

ETOE BVD RCM BFVR OTR MAP LCW

Outcome predictions greater than 75 per cent correct:

ETOE PVR Do, RCM TOE WP

ETOE RCM BFVR OTR OTRF TOE TTl pH

ETOE TOE

TTl

Do,

~vsw

0 2 ext

Do,

LCW PVR PVo2

Cl pH

RVSW RCFR Vo 2

TRR Sl

PVo 2 Paco 2

in the late stages; MAP is a poor predictor in the early stages, but a good predictor in the late stage. In the late stage, most variables predict outcome well, but clinical judgment at this time also may be excellent; thus, the clinical usefulness of prediction at this time is minimal. 22

PHYSIOLOGIC BASIS FOR PREDICTORS AND THEIR APPLICATION TO THERAPY

It is crucial to identify, describe, and evaluate physiologic compensations that maintain circulatory integrity, because the failure to compensate adequately may precipitate death or lead to life-threatening complications. For example, physiologic responses, such as increased cardiac output, may be a compensation for reduced hematocrit, reduced Pao 2 , or reduced tissue oxygenation. If increased cardiac output is a compensation with survival value, then it follows that therapy should be directed toward augmenting this response rather than returning circulatory values to the normal range. Thus, it is of great importance to understand the interactions of physiologic variables that are common not only in patients who have suffered accidental and surgical trauma, but also in those with other critical illnesses. Maintenance of the physiologic responses to stress may have great

818

WILLIAM

C.

SHOEMAKER, RICHARD

D.

BLAND AND PAULL. APPEL

relevance to all critically ill patients. Even though particular mechanisms are known to be directly concerned with the origin of a specific disease, it may be necessary to provide general support of the physiologic compensatory responses that are essential for survival. The ICU provides a unique opportunity to measure hemodynamic, physiologic, and metabolic variables and to titrate therapy to achieve optimal values.

OUTCOME PREDICTION FOR POSTOPERATIVE GENERAL SURGICAL PATIENTS

Without knowledge of the basic pathophysiology and underlying regulatory mechanisms of shock and trauma syndromes, the traditional approach to therapy of shock and trauma states may be focused on relatively superficial manifestations of shock, such as MAP, HR, CVP, hematocrit, Pao 2, and so forth. Even when physiologic measurements are made, the conventional approach is usually to correct physiologic deficits after their appearance is detected, rather than to direct early or preventative therapy toward the most likely underlying pathophysiologic mechanisms. The recognition of various patterns is based on the description of the natural history of the effects of surgical trauma from the physiologic viewpoint. There were clearly defined cardiorespiratory patterns in various etiologic categories of shock, if the data were separated in temporal stages by objective criteria. 16· 18· 27 Moreover, the sequential cardiorespiratory patterns of survivors were found to be different from those of nonsurvivors, despite a wide variety of illnesses and an even wider spectrum of operations. We observed that in the early period of shock and trauma syndromes, there was usually a normal or high cardiac output (unless limited by hypovolemia or myocardial functional impairment), together with an inadequate 0 2 transport.18 These data and other experimental findings led us to the conclusion that the basic physiologic defect in shock states is not just low flow, but maldistribution of flow, which results in inadequate 0 2 transport.l 5· 18 In contrast, patients in the late period of shock (terminal and preterminal stages) had hypotension and low cardiac output. We have used 0 2 transport variables as measures of tissue perfusion and changes in 0 2 transport as measures of the therapeutic effectiveness in comparative studies of various agents; in these studies, the patients act as their own controls. 25, 26, 28--30 Our present approach to prediction was taken (a) to describe the pattern of nonsurvivors and to define criteria for early warning of impending disaster to permit immediate maximal therapeutic effort; (b) to describe the pattern of survivors and to define objective criteria for therapeutic goals; and (c) to determine which variables are able to differentiate between survivors and nonsurvivors. Thus, the capacity of a variable to predict outcome was used as a measure of its relevance and usefulness in making clinical decisions (Table 3).19• 23 This allows therapeutic decisions to be made on the basis of empirically derived evidence, rather than on opinion based on anecdotes, philosophy, or a "party line." This analysis showed that the most commonly monitored variables-

819

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

like blood pressure, Pao2, CVP, HR, and hematocrit-were very poor predictors of outcome, and were, therefore, of limited value in therapeutic management.23 Figure 1 illustrates the patterns of survivors and nonsurvivors for a few selected variables. The differences between survivors' and nonsurvivors' MAP and Pao2 patterns were minimal, indicating poor prediction until the preterminal stage. Cardiac index was somewhat better than MAP in predicting outcome; more importantly, survivors had values at the late stage that were 50 per cent in excess of normal. Pulmonary vascular resistance in~ex, Do2 _and 0 2 consumption (Vo 2) were better predictors. In survivors, Do2 and Vo 2 were well maintained in the early stage. In the nonsurvivors, they were generally reduced, but there was a compensatory increase in Vo 2 in the middle stage. These patterns were remarkably consistent, despite the wide variety of illnesses and the even wider variety of operations (Table 2). However, no one variable was satisfactory as a predictor. In essence, determination of the usefulness of the variables is a multifactorial problem that requires multivariate analysis. 20· 27

RATIONALE FOR THE PREDICTIVE INDEX AND ITS APPLICATION TO THERAPY

Our approach to the predictive index is based on the old principle that if one knows everything that is important about a system, one should

Figure 1. Schematic representation of the predictive score of a cardiorespiratory variable showing the frequency distribution of values of a given variable for patients who survived and for those who did not. The point Q maximally separates these two groups, the ranges of survivors and nonsurvivors, the 90 percentile value of nonsurvivors, the 10 percentile value of survivors, and the overlap region. The probability of survival is determined by relative distance of an observed value between Q and the lOth or 90th percentile value. (Adapted from Shoemaker, W. C., Appel, P. L., and Bland, R.: Use of physiologic monitoring to predict outcome and to assist in clinical decisions in critically ill postoperative patients. Am. J. Surg., 146:3, 1983; with permission.

Regton of Overlap

Reg ton of

Lethal tty Range

of

Regton of Survivability

Nonsurvtvors Range

of

Survi.vors

820

WILLIAM

C.

SHOEMAKER, RICHARD

D.

BLAND AND PAULL. APPEL

be able to predict the outcome, and if one can predict outcome, one should be able to modify it. Translated into contemporary scientific terms, this means that adequate description of circulatory events in surviving and nonsurviving patients with medical emergencies will provide the basis for a pathophysiologic understanding of the disease. Physiologic studies also help the physician to interpret mechanistically the sequential physiologic events of acute circulatory failure from inadequate tissue perfusion during or after the time of the operation. Objective evaluation of these physiologic descriptions and the changes produced by specific therapeutic interventions may then be systematically evaluated. Subsequently, organized coherent therapeutic regimens or protocols may be proposed and tested prospectively. The major premise is that death in critically ill postoperative general surgical patients is the end result of well-defined, predictable cardiorespiratory patterns that are independent of the specific clinical diagnosis of accidental or operative traumatic lesions. Irrespective of the illness or injury, there are a number of physiologic events that characterize the pattern of survivors and distinguish it from that of the nonsurvivors. For example, the post-traumatic shock and subsequent lethal renal, cardiac, or respiratory failures after pelvic fracture from blunt trauma are not measurably different from intraoperative or postoperative shock and those same vital organ failures after extensive ablative surgery for carcinoma. This concept is not new; it was Sir William Osler who first pointed out that patients rarely died of the primary disease diagnosed; more often they died of complications. The physiologic derangements of postoperative shock and its sequelae can be identified and described quantitatively in their temporal relationships with the primary etiologic event. The major hypothesis of the clinical studies is that a systematic, coherent, physiologic approach to management of the life-threatening illness based on clinical and physiologic criteria (defined operationally from retrospective analysis of critically ill patients with trauma) will improve morbidity and mortality. This hypothesis has three essential components that may be examined within this context. First, the cardiorespiratory patterns of surviving patients are distinctly different from those of nonsurvivors, despite the wide spectrum of clinical diagnoses and the even wider spectrum of surgical operations, accidental trauma, hemorrhage, sepsis, and so on. Second, the monitored cardiorespiratory pattern of survivors of lifethreatening illness provides objective physiologic criteria that may be used to develop goals of therapy for the critically ill patient. Finally, these operationally defined goals may also be used to develop a coherent systematic protocol for therapy of critically ill postoperative patients. Ideally, a protocol may be expressed in terms of a branched-chain decision tree; the branching chain defines (that is, preselects) clinical groups and sequentially organizes, in a temporal order, therapeutic criteria defined by the survivors' physiologic pattern to provide the physician with information to make the most effective clinical decisions. The underlying assumptions of this hypothesis are as follows. First, most postoperative deaths are due to circulatory problems and the subse-

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

821

quent multiple vital organ failures. Critically ill patients most often die of physiologic derangements rather than of the anatomic lesions of their primary disease or the technical features of the surgical operation. Second, the circultory system, like other fluid systems, can be characterized by measurements of pressure, flow, volume, and physiologic function. These measurements can be taken easily and repetitively by currently available technology and can be used to describe the physiologic patterns of survivors and nonsurvivors. Functional aspects of the circulation are best assessed by measurements of 0 2 transport and metabolism, because 0 2 transport is essential to life, consistently reduced in shock states, and considerably different in survivors and nonsurvivors. This information is particularly relevant from the technologic standpoint because 0 2 has the highest extraction ratio of any blood constituent and, therefore, is the most flow dependent. Moreover, 0 2 uptake by the tissues is very close to 0 2 transport across the alveolar-capillary membrane and the cell membrane, because 0 2 cannot be stored, nor can a sizable 0 2 debt be accumulated for significant periods of time. These and other experimental findings have suggested that Vo 2 is the regulating mechanism in tissue hypoxia (stagnant, anemic, and hypoxic) described in Barcroft's classic studies. Thus, Vo 2, which reflects the sum of all oxidative metabolism as well as the adequacy of the circulatory delivery system, is maintained at normal or slightly elevated values, by hemodynamic compensations, including increased HR, cardiac output, and 0 2 extraction; when these compensations fail, Vo 2 falls and death occurs.l8 Measurements of 0 2 metabolism are the most sensitive and specific of the monitored cardiorespiratory variables for acute circulatory failure. 22, 23 Finally, the longer the patient has circultory deficits, the longer he or she will be critically ill and the more likely he or she will have complications, including multiple vital organ failures (renal, respiratory, hepatic, and central nervous system [CNS] failure), sepsis, DIC, and nutritional problems. Patients rarely die in the operating room of technical problems; usually postoperative deaths occur after prolonged critical illness that begins as circulatory impairment but then leads into multiple organ failure.

FORMULATION OF OUTCOME PREDICTORS

A predictive index based on the probability distributions of each variable was developed with the use of a simple algorithm.22 The relative distance of the value of a given variable from the ninetieth percentile of nonsurvivors to the tenth percentile of survivors was calculated according to an algorithm (Fig. 2). The sum of the weighted scores of each variable gives an overall severity (predictive) index, which serves as a yardstick that tells how far it is to the brink of disaster and how far to safe territory. The sole criterion of this empiric analysis was survival. Recently, a greatly simplified and more useful predictor was based solely on the survival rates of patients whose cardiorespiratory values fell within each of 10 equally spaced divisions in the spectrum of values of a

822

WILLIAM

l

60

SHOEMAKER, RICHARD

D.

BLAND AND PAULL. APPEL Stages

Stages

A

mm Hg

100 80

C.

8

Low

C1

C2

D

Mean arterial pressure

F percent

100~

A

8

I

I

Low C1 I

I

C2

D

E

F

I

I

I

I

Arterial Oxygen saturation

96 92

88

40

84

20 liters/min/M 2

E

j

ml/mm/M 2 Cardiac index

800 600

400

ml/min/M 2 Pulmonary vascular resistance

200 175

150 125 100

Figure 2. Sequential patterns of mean arterial pressure, cardiac index, pulmonary vascular resistance, Pa02 , 0 2 delivery, and 0 2 consumption for postoperative survivors and nonsurvivors of critical illnesses.

given variable. 4· 5 The average score of each variable's prediction is computed to give a total predictive index for the patient for each postoperative period. Figure 3 illustrates the survival rates of patients whose left ventricular stroke work index values ranged from 15 to 90 gm ·m per m 2 in the first 8 hours postoperatively in an initial retrospective series of critically ill postoperative patients. It may be seen from these data that survival rates increased progressively until a plateau was reached; the height of the plateau reflects the capacity of the variable to predict outcome, and the width of this plateau defines the optimal values for this variable. A mathematic model based on the least squares method defines the horizontal plateau, as well as the slope of the line ascending to, and descending from, this plateau (Figs. 3 to 5). Figure 4 illustrates Do2 and yo 2 values in the same series. This figure shows a wide, high plateau for Do 2 , indicating that this variable is an excellent predictor and that optimal values for survival are over 600 ml per min.ute·m 2 ; the Vo 2 values show that survival increases with rising values of Vo 2 . Figure 5 depicts the survival rates of patients for HR and PVRI values. It may be seen that the survival rates were approximately 50 per cent for HR values throughout the observed spectrum. In

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

823

WIDTH OF PLATEAU PS THE RANGE OF OPTIMAL VALUES OF A GIVEN VARIABLE ~

100 'Yo

90

80 70 60

-~

HEIGHT OF PLATEAU PS'Y.OF SURVIVORS

~50

(/)

30

20 10

0 30

60

Figure 3. Survival rates are shown for patients whose left ventricular stroke work index values ranged from 15 to 90 g·m/m2 during the first 8 hours postoperatively in the initial series of postoperative patients.

contrast, PVRI values did not have a high plateau, indicating poor prediction of survival, but the decrease to 0 per cent survival above values of 500 dyne·sec per cms·m 2 signifies that this variable was a good predictor of death.

PROSPECTIVE EVALUATION OF THE PREDICTIVE INDEX

Since this overall predictive index was developed retrospectively in a large series of postoperative patients at Cook County Hospital in Chicago

824

WILLIAM

OXYGEN

C.

SHOEMAKER, RICHARD

DELIVERY

D.

BLAND AND PAULL. APPEL

OXYGEN CONSUMPTION %

100

80

60

40

20

0 115

I~

19!5

235

275

Figure 4. Survival rates versus values for 0 2 delivery (left) and 0 2 consumption (right) during the first 8 hours postoperatively in the initial series.

and at Mount Sinai Hospital in New York, it is necessry to test the validity of the predictor in a fresh series of critically ill postoperative patients. Table 5 shows the results of this predictive index applied prospectively to a new series of 300 critically ill postoperative patients during the first 5 years of the surgical ICU at Harbor-UCLA Medical Center. 20 Only about 2 per cent of the surgical patients were monitored; these were critically ill, highrisk, high-mortality patients. As seen from this standard truth table, the index was 93 per cent correct. All eight incorrectly predicted to survive died of late complications or carcinomatosis; the 13 who were predicted to die but who survived had their catheter removed less than 12 to 18 hours postoperatively. Thus, the predictive index was found to be reasonably accurate and reproducible over a wide range of socioeconomic levels, clinical mixes, and private or public institutions. The most recently developed predictor was also tested in a prospective trial in a fresh series of patients and found to be 94 per cent correct. Prediction of outcome requires a more rigorous statistical approach than does the more commonly used P-values for the Student's t-test. The upper section of Figure 6 illustrates schematically the distribution of survivors' and nonsurvivors' values of a variable when the mean values of the two groups were statistically significant (P value < 0.01); however, prediction of survival could be made only in the 10 per cent of patients whose values fell outside the range of the nonsurvivors, and prediction of death could be made only in a small percentage of those whose values fell below

825

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

PULMONARY VASCULAR

HEART RATE

RESISTANCE

Figure 5. Survival rates versus heart rate (left) and pulmonary vascular resistance (right) during the first 8 hours postoperatively in the initial series.

the range of the survivors (indicated by the cross-hatched areas). The lower section of Figure 6 illustrates distributions of survivors' and nonsurvivors' values for which there was a narrow range of overlap and a large percentage of prediction, shown by the cross-hatched area. Thus, classificationthat is, prediction-requires separation of the two populations by an order of magnitude greater than that required for statistical significance by the conventional Student's t-test.

Table 5.

Results of Predictors after Surgical Operations LAST AVAILABLE PREDICTED VALUE

Actual Outcome

Survived Died Total Per cent Correct

Survival

Death

Total

Per cent Correct

206 8 214 96*

13 73 86 85**

219 81 300

90t 93

94*

*Sensitivity: percentage of survivors with correctly predicted outcome. tSpecificity: percentage of nonsurvivors with correctly predicted outcome. *Predictive accuracy: percentage of survivors among patients predicted to live. **Predictive precision: percentage of nonsurvivors among patients predicted to die.

826

WILLIAM

C.

SHOEMAKER, RICHARD

Number Nonsurvivors

D.

BLAND AND PAULL. APPEL

Survivors

: • ss;f ;ca bon - 1 0 % 0

Xn

X5

100 P<.01

Figure 6. Idealized distribution of survivors and nonsurvivors where there was a wide area of overlap for the two groups illustrates poor prediction but statistical significance P < 0.01 (upper section) and a narrow area of overlap with good prediction and good statistical significance (lower section).

DEFINITION OF OPTIMAL THERAPEUTIC GOALS

BIOLOGIC SIGNIFICANCE OF VARIOUS MONITORED VARIABLES

What should be measured, when should it be measured, and why? Since the majority of cardiovascular diseases are responsible for deaths in this country, since most patients with fatal accidental and surgical trauma die from cardiorespiratory problems, and since all patients who die of acute illness go through a stage of circulatory shock, the circulatory system is a good place to begin. Moreover, hemodynamic and 0 2 transport measurements lend themselves to frequent repetitive monitoring and can provide crucial information on underlying physiologic mechanisms useful for prognosis and therapy. H>-21. 21 The results of invasive cardiorespiratory monitoring have been compared with the standard conventional monitoring from the standpoint of the capacity to anticipate death or cardiopulmonary arrest. For example, electrocardiogram (EKG), MAP, HR, hematocrit, CVP, urine output, and blood gases, which are the conventionally monitored variables, are useful parameters that describe the end stage of circulatory failure in the trauma patient; they have not been found to be sensitive or accurate in early warning of death in acutely ill postoperative and trauma patients. 6• 23 Not only do multivariate predictors provide a system for classification of outcome in the early postoperative period, but also, more importantly,

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

827

they are able to define the goals of therapy. That is, they provide a simple, straightforward, objective, physiologic basis for therapeutic decisions. As a first approximation, therapeutic goals may be defined empirically from the median values of survivors for each cardiorespiratory variable (Table 6).

DEVELOPMENT OF A THERAPEUTIC PLAN

A branched-chain decision tree was developed from data of the predictors (Fig. 7). Criteria for assignment of priorities were based on survival statistics. A more coherent and effective strategy is vigorous volume load without exceeding WP greater than 18 mm Hg. We have found that it is easier to achieve these goals with colloids that expand the plasma volume without undue increase in the interstitial water.25, 26, 2~3° Mter the maximum effect of fluids has been obtained, we then add an inotropic agent such as dobutamine, beginning with 2 flg per kg and increasing in a stepwise fashion; the optimal dose is obtained by titration to achieve the best goals in terms of cardiac index (CI), Do2 , and Vo 2 • If the patient has high MAP and systemic vascular resistance index (SVRI), vasodilation with nitroglycerin or nitroprusside is considered; the optimal dose is obtained by titration to achieve improved cardiac index without producing hypotension (that is, MAP > 80 mm Hg, systolic pressure > 110 mm Hg). If fluids, inotropic agents, and vasodilators fail to achieve optimal goals, vasopressors such as dopamine then are given in the smallest possible dose needed to maintain MAP at 80 mm Hg and systolic pressure at 110 mm Hg. Vasopressors are given last because they raise venous pressure in addition to increasing MAP and thereby may limit optimal fluid administration; no amount of dopamine will make up for blood volume deficits. The increased Vo 2 often found in preoperative patients with severe trauma, stress, sepsis, and hypercatabolic states indicates a greater than normal metabolic rate, but this does not mean that all the patient's metabolic needs have been met. This definition of therapeutic goals and adequacy of therapy cannot be assessed directly; however, tissue demand may be inferred indirectly by an empiric approach, that is, by a trial of therapy. If the therapy raises cardiac output and Vo 2 , it may be assumed that therapy opened up additional microcirculatory channels that perfused relatively hypoxic tissues, which then extracted more 0 2 ; since tissue cannot

Table 6.

Optimal Goals of Therapy

l. Cardiac index 50 per cent in excess of normal, i.e., 4.5 L per minute·m 2 . 2. Blood volume 500 ml greater than normal, provided this could be attained without exceeding WP of 2,0 mm Hg. 3. Oxygen delivery (Do2) greater than 600 ml per minute·m 2 • 4. Oxygen consumption (Yo2) greater than 170 ml per minute·m2 . 5. Normal blood pressure. 6. PVRI less than 250 dyne·sec per cm 5 ·m2 • 7. Metabolic and nutritional support.

828

WILLIAM

C.

SHOEMAKER, RICHARD

D.

BLAND AND PAULL. APPEL

Critically Ill Postop Patient

Figure 7. Decision tree for fluid management of the critically ill postoperative patient. Preliminary evaluation by routine work-up in the intensive care unit includes arterial blood gases, chest radiograph, routine blood chemistries, electrocardiogram, and coagulation studies. These tests should be either performed or in process and the observed defects corrected; that is, ifPao 2 < 70 torr, 7.3 >pH> 7.5, Paco 2 >55 torr, or respiration rate (RR) > 30 breaths per minute, place on the respiratory protocol. If none of the preceding conditions are present, proceed to step.l. (1) Determine if the patient has reached the optima\ goals. Measure (cardiac index) CI, Do 2 , Vo2 , and blood volume (BV). IfCI < 4.5 Umin·m2 , Do2 < 600 ml!min·m 2 , Vo2 < 160 ml!min·m 2 and BV 3.0 Um2 for men or 2. 7 Um 2 for women, take hematocrit (Hct). lf any of the preceding optimal values has not been reached, proceed to step 2. If the goals are reached, the objective of the algorithm has been achieved. Re-evaluate and recycle at intervals to maintain these goals. (2) Take pulmonary wedge pressure (WP). If > 20 mm Hg, proceed to step 3; if < 20, proceed to step 4. (3) If WP > 20, give furosemide (Lasix) intravenously at increasing dose levels (20, 40, 80, 160 mg) if there is clinical or radiographic evidence of salt and water overload or clinical findings of pulmonary congestion. If not, consider vasodilators, nitroprusside, or nitroglycerin if mean arterial pressure (MAP) > 80 and systolic pressure > 100 mm Hg. Recycle up to five times to titrate the dose needed to reduce WP < 20 mm Hg but maintain MAP > 80 mm Hg. If unsuccessful, place on cardiac protocol. (4) If Hct < 32 per cent, give 1 unit of whole blood (WB) or 2 units of packed red blood cells (Prbc). If Hct > 32 per cent, give a fluid load (volume challenge) consisting of one of the following (depending on clinical indications of plasma volume deficit or hydration): 5 per cent PPF, 500 ml; 5 per cent albumin, 500 ml; 25 per cent albumin (25 g), 100 ml; 6 per cent hydroxyethyl starch, 500 ml; 6 per cent dextran 60, 500 ml; D5RL, 1000 mi. (5) If the blood or fluid load improved any of the optimal therapeutic goals defined h1 step 1, proceed to step 6; if none are improved, proceed to step 7. (6) If goals are not reached, recycle steps 2 through 6 until these goals are met or WP > 20 mm Hg. (7) If MAP > 70 mm Hg or systolic arterial pressure (SAP) < 100 mm Hg, give dobutamine (Dobutrex) by constant intravenous infusion in' increasing doses. (8) Titrate dobutamine beginning with 1 to 2 fJ.g/min·kg and gradually increasing doses up to 20 fJ.g/min·kg provided there is improvement in CI, Do2 , or Vo2 without further

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

829

take up more 0 2 than it uses, the increased Vo 2 after fluid challenge means that an 0 2 debt had been present and was at least partially relieved.

PROSPECTIVE CLINICAL TRIALS OF EMPIRICALLY DEFINED GOALS OF THERAPY The therapeutic plan outlined in the preceding section was prospectively tested in clinical trials against a control group, in which normal values were used as the goal of therapy.I9, 21 Both protocol and control groups had the same monitoring, availability of x-ray films and lab tests, nursing care, ancillary facilities, and therapy; patients whose critical illnesses met prearranged criteria (Table 7) were prospectively allocated to both groups, and they were managed by the same pool of Harbor-UCLA-trained residents. The only real difference was that the therapeutic goals for control patients were normal values of cardiorespiratory variables, whereas the therapeutic goals for the protocol group were the median values of the survivors (Table 6), Comparison of the clinical data, clinical diagnosis, operations, and associated medical conditions demonstrated that the patients in the protocol group were at least as ill and probably more at risk than those in the control group. The results of this study showed marked reduction in morbidity and mortality in the protocol patients (Tables 8 and 9). Recently, we have undertaken an additional three-part (CVP catheter, pulmonary artery catheter with normal values as goals, and pulmonary artery catheter with optimal values as goals of therapy), prospective, randomized trial of this concept that was begun preoperatively; this trial was strictly randomized, with the use of sealed envelopes containing cards identifying one of the three therapeutic systems. The results of the first 10 months of this ongoing trial showed no statistically significant difference between the ICU mortality (29 per cent) of patients managed with a CVP catheter and that of patients managed with the pulmonary artery catheter with normal values as therapeutic goals; however, use of the pulmonary artery catheter with optimal goals led to significantly reduced (5 per cent) mortality. lowering of arterial pressure until goals are met. (9) If goals are reached, re-evaluate and recycle. If goals are not reached or if it becomes evident that higher doses of the drug are not more effective or that they produce hypotension and tachycardia, continue dobutamine at its most effective dose range. (10) If MAP > 80 mm Hg and SAP > llO mm Hg, give sodium nitroprusside or nitroglycerin in gradually increasing doses. If the arterial pressures are lower than MAP> .80 mm ~g and SAP> llO mm Hg, give vasopressor. (ll) If there is no improvement in CI, Do2 , or Vo2 with the vasodilator or if hypotension (MAP < 80 mm Hg, SAP < .llO mm Hg) ensues, discontinue the vasodilator. If there is improvement in CI, Do2 , or Vo2 , titrate vasodilator to its maximum effect consistent with satisfactory pressures. (12) If optimal goals are reached, re-evaluate and recycle at intervals. If these goals are not reached and MAP < 80 mm Hg, SAP < llO mm Hg, give vasopressor. (13) Titrate doses of vasopressor (dopamine) in the lowest doses possible to maintain arterial pressures, MAP > 80 mm Hg, SAP > llO mm Hg. If pressures cannot be maintained, the patient is considered to be a protocol failure. (Modified from Shoemaker, W. C.: Pathophysiology of shock syndromes. In Shoemaker, W. C., Thompson, W. L., and Holbrook, P. R. (eds.): Textbook of Critical Care. Philadelphia, W. B. Saunders Company, 1984.)

830

WILLIAM

C.

SHOEMAKER, RICHARD

Table 7.

D.

BLAND AND PAULL. APPEL

Criteria for Critical Illness

Preoperative Patients Who Have One or More of the Following: 1. Previous severe cardiorespiratory illness (acute myocardial infarction (MI), chronic obstructive pulmonary disease (COPD), stroke, and so on). 2. Extensive ablative surgery planned for carcinoma: for example, esophagectomy and total gastrectomy, prolonged surgery (> 6 hours). 3. Severe multiple trauma: for instance, that involving > three organs or > two systems; opening of two body cavities (left side of chest and abdomen); multiple long-bone and pelvic fractures. 4. Massive blood loss (> 8 units): BV < 1.5 L per m2, Hct < 20 ml per dl within 48 hours prior to admission. 5. Age over 70 years and evidence of limited physiologic reserve of one or more vital organs. 6. Shock: MAP < 60 mm Hg; CVP < 15 mm Hg; urinary output (UO) < 20 mllh; cold, clammy skin. 7. Septicemia: positive blood culture, white blood cell count (WBC) > 12,000 per mm 3 , spiking fever to 101 °F for 48 hours, chills. 8. Evidence of sepsis (temp> 101°F, WBC count> 12,000 per mm 3) plus hypotension (MAP< 70 mm Hg). 9. Severe nutritional problems associated with a surgical illness: weight loss > 20 lb, albumin concentration < 3, osmolarity < 280 mOsm per L. 10. Respiratory failure: for example, Pao 2 < 60 or FI02 < 0.4, Q,JQ, > 30 per cent, patient on mechanical ventilation. ll. Acute abdominal catastrophe: for instance, pancreatitis, gangrenous bowel, peritonitis, perforated viscus, internal gastrointestinal (GI) bleeding. 12. CVP > 15 mm Hg after fluid resuscitation. 13. Acute renal failure: blood urea nitrogen ([BUN] > 50 mg per dl, creatinine > 3 mg per dl, CH 20 > - 10 mllh). 14. Acute hepatic failure (bilirubin > 3 mg per dl, albumin concentration < 3 g/dl, lactic dehydrogenase [LDH] > 200 units per ml, alkaline phosphatase > 100 units/ml, ammonia> 120 fLg per ml). 15. Acute agitation, depressed nervous system, semicoma, or coma. Postoperative Patients Who Have One or More of the Following: 1. Acute catastrophic change, suggesting fresh MI, pulmonary embolus, postoperative bleeding. 2. Hypotension: MAP < 70 mm Hg or unstable vital signs. 3. Operative misadventure: for example, use of < 8 units of whole blood (WB) or packed red blood cells (PRBC) or estimated 4000-ml blood loss in the operating room. 4. Severe sepsis, perforated viscus, gangrene of bowel, peritonitis, pneumonia, and positive blood culture, aspiration pneumonia, temperature elevation > 101°F for> 2 days. 5. Any vital organ failure, that is, the same as 9 to 15 of the above list of preoperative conditions. 6. Postoperative fluid-electrolyte problem requiring more than 5000 ml of fluids per day. 7. Failure to respond to adequate volume therapy, which is replacement of blood losses estimated from sponge and lap counts, as judged by clinical criteria, such as arterial pressure, UO, Hct, level of consciousness, and motor responses.

Table 8.

Number of patients Number of nonsurvivors Mortality (per cent)

Outcome Data CONTROL

PROTOCOL

143 50 35

80 10 12.5

831

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

Table 9.

Complications of Control and Protocol Groups CONTROL

COMPLICATION

(NUMBER =

Respiratory failure (requiring ventilation) Sepsis, systemic Cardiogenic problems, including pulmonary edema, cardiac arrest, arrhythmia Renal failure (requiring dialysis) Disseminated intravascular coagulation Pulmonary embolism Hepatic failure, coma Delirium tremens Upper GI hemorrhage Urinary tract infection Postoperative bowel obstruction Decubitus ulcer Carcinomatosis Pneumothorax Bronchopleural fistula Pancreatitis Metabolic alkalosis Allergic reaction Total Average number per patient

66 44

143)

PROTOCOL (NUMBER = 8o)

27 11

9 33

25

5

8

3

2

3 1 7

5 2 1

1 1 1

1 2 1

1 19[

61

1.34

0.76

CONCLUSION An objective physiologic approach, using survival as the criterion for outcome prediction as well as for therapeutic decision making, is feasible and efficacious. The improved mortality in these prospective studies supports the hypothesis that compensatory responses of the survivors are the major determinants of outcome. Therefore, therapy that supports these compensations and produces the survivor pattern improves survival rates. These prospective studies confirm the validity of an organized, coherent physiologic approach, in contrast to the traditional approach, whose objectives are to correct hemodynamic and chemical abnormalities if and when they are discovered. This approach emphasizes aggressive fluid management in tacit acknowledgment that unrecognized hypovolemia, delay in treatment of hypovolemia, of inadequate volume therapy is the primary precipitating event in most cases of postoperative, hemorrhagic, and traumatic shock. Moreover, most surgical complications (such as acute respiratory failure, acute renal failure, and, to some extent, sepsis and DIC) are the sequelae of shock; for example, shock lung is a complication of shock, and without shock~ there is. no shock lung. However, episodes of reduced cardiac index, Do2 , and Vo2 may occur intraoperatively with little or no hypotension or with hypotension that is corrected by administration of ephedrine or other vasopressors.

832

WILLIAM

C.

SHOEMAKER, RICHARD

D.

BLAND AND PAULL. APPEL

We conclude that normal values are appropriate for normal, unstressed, resting subjects, but the empirically determined cardiorespiratory patterns of surviving patients are the appropriate goals of therapy for critically ill postoperative patients. The need for this proposed approach is evident: More than 24 million major surgical operations are performed annually in the United States with an estimated annual mortality of over 400,000. These studies indicate that the mortality for a university-run hospital may be sharply reduced and suggest that it may be possible to achieve comparable reductions in postoperative mortalities nationwide. The use of a branched-chain decision tree helps to achieve expeditiously these therapeutic goals by providing a coherent, organized patient management plan. The essence of this plan is to maintain prophylactically the patient in an optimal hemodynamic state that does not allow him or her to develop tissue hypoxia from blood volume, hemodynamic, and 0 2 transport deficits. It is not necessary to wait for patients to develop cardiorespiratory deficits before initiating therapy. Therapy should be started if key values are not optimal. It must be stated that this approach, which is primarily directed toward fluid management, will not correct diagnostic errors, anatomic problems, misadventures at the time of surgery, transfusion reactions, idiosyncratic drug reactions, iatrogenic ineptitudes, or lack of the patient's desire to live. Furthermore, even if physiologic variables are optimized, success is not guaranteed if the patient's cardiac and pulmonary reserve capacities are not adequate or if the surgical stress is overwhelming.

REFERENCES 1. Appel, P. L., and Shoemaker, W. C.: Evaluation of fluid therapy in adult respiratory failure. Crit. Care Med., 9:862, 1981. 2. Baker, S. P., O'Neill, B., Haddon, W., et al.: The injury severity score: A method for describing patients with multiple injuries and evaluating emergency care. J. Trauma, 14:187, 1974. 3. Bland, R. D., and Shoemaker, W. C.: Common physiologic patterns in general surgical patients: Hemodynamic and oxygen transport changes during and after operation in patients with and without associated medical problems. Surg. Clin. North Am., 65:793, 1985. 4. Bland, R. D., and Shoemaker, W. C.: The severity index score as a method for data reduction in the surgical ICU. Comput. Biomed. Res., 16:395, 1983. 5. Bland, R., Shoemaker, W. C., Abraham, E., et al.: Hemodynamic and oxygen transport patterns in surviving and nonsurviving postoperative patients. Crit. Care Med., 13:85, 1985. 6. Bland, R., Shoemaker, W. C. and Shabot, M. M.: Physiologic monitoring goals for the critically ill patient. Sug. Gynecol. Obstet., 147:833, 1978. 7. Carey, J. S., Brown, R. S., Woodward, N. W., et al.: Comparison of hemodynamic responses to whole blood and plasma expanders in clinical traumatic shock. Surg. Gynecol. Obstet., 121:1059, 1965. 8. Champion, H. R., S~co, W. J., Carnazzo, A. J., et al.: The Trauma Score. Crit. Care Med., 9:672, 1981. 9. Champion, H. R., Sacco, W. J., Hannon, D. S. et al.: Assessment of injury severity: The Triage Index. Crit. Care Med., 8:201, 1980. 10. Cowley, R. A., Sacco, W. J., Gill, W., et al.: A prognostic index for severe trauma. J. Trauma, 14:1029, 1971.

OUTCOME OF PREDICTION AND PROSPECTIVE CLINICAL TRIALS

833

11. Cullen, D. ]., Civetta, J. M., Briggs, B. A., eta!.: Therapeutic intervention scoring system: A method for quantitative comparison of patient care. Crit. Care Med., 2:57, 1974. 12. Cullen, D. ]., Ferrara, L. C., Gilbert, J., eta!.: Indicators of intensive care in critically ill patients. Crit. Care Med., 5:173, 1974. 13. Knaus, W. A., Zimmerman, J. E., Wagner, D. P., eta!.: APACHE: Acute physiology and chronic health evaluation: A physiologically based classification system. Crit. Care Med., 9:591, 1981. 14. Norris, R. M., Brandt, P. W. T., Caughey, D. E., et a!.: A new coronary prognostic index. Lancet, i:274, 1969. 15. Schwartz, S., Frantz, R. A., and Shoemaker, W. C.: Sequential hemodynamic and oxygen transport responses in hypovolemia, anemia and hypoxia. Am. ]. Physiol., 241:H864, 1981. 16. Shoemaker, W. C.: Analysis of physiologic mechanisms in various etiologic types of clinical shock from sequential cardiorespiratory measurements. In Forscher, B. K., Lillehei, R. C., and Stubbs, S. S. (eds.): Shock in Low- and High-Flow States. Amsterdam, Excerpta Medica, 1972, pp. 119--130. 17. Shoemaker, W. C.: Comparison of the relative effectiveness of whole blood transfusions and various types of fluid therapy in resuscitation. Crit. Care Med., 4:71, 1976. 18. Shoemaker, W. C.: Pathophysiology of shock syndromes. In Shoemaker, W. C., Thompson, W. L., and Holbrook, P.R. (eds.): Textbook of Critical Care. Philadelphia, W. B. Saunders Co., 1984, pp. 52-71. 19. Shoemaker, W. C., Appel, P. L., and Bland, R.: Use of physiologic monitoring to predict outcome and to assist in clinical decisions in critically ill postoperative patients. Am. J. Surg., 146:43, 1983. 20. Shoemaker, W. C., Appel, P. L., Bland, R., eta!.: Clinical trial of an algorithm for outcome prediction in acute circulatory failure. Crit. Care Med., 10:390, 1982. 21. Shoemaker, W. C., Appel, P. L., Waxman, K., et a!.: Clinical trial of survivors' cardiorespiratory patterns as therapeutic goals in critically ill postoperative patients. Crit. Care Med., 10:398, 1982. 22. Shoemaker, W. C., Chang, P. C., Czer, L. S.C., eta!.: Cardiorespiratory monitoring in postoperative patients: I. Prediction of outcome and severity of illness. Crit. Care Med., 7:237, 1979. 23. Shoemaker, W. C., and Czer, L. S. C.: Evaluation of the biologic importance of various hemodynamic and oxygen transport variables. Crit. Care Med., 7:424, 1979. 24. Shoemaker, W. C., Elwyn, D. H., Levin, H., eta!.: Early prediction of death and survival in postoperative patients with circulatory shock by non parametric analysis of cardiorespiratory variables. Crit. Care Med., 2:317, 1974. 25. Shoemaker, W. C., and Hauser, C. ]. : Critique of crystalloid versus colloid therapy in shock lung. Crit. Care Med., 7:117, 1979. 26. Shoemaker, W. C., and Monson, D. 0.: Effect of whole blood and plasma expanders on volume-flow relationships in critically ill patients. Surg. Gynecol. Obstet., 137:453, 1973. 27. Shoemaker, W. C., Montgomery, E. S., Kaplan, E., et a!.: Physiologic patterns in surviving and nonsurviving shock patients. Arch. Surg., 106:630, 1973. 28. Shubin, H., Wei!, M. H., Afifi, A. A., et a!.: Selection of hemodynamic, respiratory and metabolic variables for evaluation of patients in shock. Crit. Care Med., 2:326, 1974. 29. Siegel,]., Goldwin, R. M., and Friedman, H. P.: Pattern in process in the evolution of human septic shock. Surgery, 70:232, 1971. 30. Snyder, J. V., McGuirk, M., Grenvik, A., eta!.: Outcome of intensive care: An application of a predictive model. Crit. Care Med., 9:598, 1981. 31. Teasdale, G., and Jannett, B.: Assessment of coma and impaired consciousness: A practical scale. Lancet, i:81, 1974. 32. Wei!, M. H., and Afifi, A. A.: Experimental and clinical studies on lactate and pyruvate as indicators of severity of acute circulatory failure (shock). Circulation, 41:989, 1970. Department of Surgery Los Angeles County Harbor/UCLA Medical Center 1000 West Carson Street Torrance, California 90509