predicting outcome from hypoxic-lschemic coma

8/18/2019 Predicting Outcome From Hypoxic-lschemic Coma

1/7

Predicting Outcome From

Hypoxic-lschemic ComaDavid E. Levy, MD; John J. Caronna, MD; Burton H. Singer, PhD;Robert H. Lapinski, PhD; Halina Frydman, PhD; Fred Plum, MD

Outcome from

coma

caused by cerebral hypoxia-ischemia (eg, cardiacarrest) was compared with serial neurological findings in 210 patients.Thirteen percent of patients regained independent function at some pointduring the first postarrest year. Computer application of new multivariatetechniques to the prospectively observed findings generated easily utilizedrules that classified patients by likely outcome. At the time of initialexamination, 52 patients (one fourth of the total population) had absentpupillary light reflexes, and none of these patients ever regained indepen-dent daily function. By contrast, the initial presence of pupillary lightreflexes, the development of spontaneous eye movements that were rovingconjugate or better, and the findings of extensor, flexor, or withdrawalresponses to pain identified a smaller group of 27 patients, 11 (41%) ofwhom regained independence in their daily lives. By 24 hours after onset, 93poor-outcome patients were identified by motor responses that were absent,

extensor, or flexor and by spontaneous eye movements that were neitherorienting nor roving conjugate; only one regained independent function. Thiscontrasts with recovery in 19 (63%) of 30 patients who at that time showedimprovement in their eye-opening responses and obeyed commands or hadmotor responses that were withdrawal or localizing. Similarly simple rulesdistinguished between good- and poor-prognosis patients on postarrest days3, 7, and 14.

(JAMA 1985;253:1420-1426)

MOST patients sustaining cardiacarrest either undergo irreversibleasystole or reawaken quickly andmake a good physical and mental

recovery. A few, however, experiencesufficiently severe brain injury thatthey remain at least transiently inpostarrest coma.

Among this less for¬

tunate group morbidity is high, butsome retain the neurological capacityto do well. Others, however, are leftwith overwhelmingly severe braindamage. This study reports newlyconstructed, empirically derivedguidelines to predict within the firstfew days which of these brain-injuredpatients will do well and which willdo badly following cardiac arrest orsimilar global hypoxic-ischemic in¬sults.

METHODSPatient Evaluation

The analysis was performed on 210patients with cerebral hypoxia-ischemia,drawn from a larger group of 500 subjectswith nontraumatic coma.12 The opera¬tional definition of coma w as that patientsfailed to open their eyes either spontane¬ously or in response to noise, that theyexpressed no comprehensible words, andthat they neither obeyed commands normoved their extremities appropriately tolocalize or resist painful stimuli. At prede¬termined intervals (the initial examina¬tion and 0 to 1, 2 to 3, 4 to 7, and 8 to 14

days after the onset

of coma), patientsunderwent focused neurological examina¬tions." More than half (115, or 55%) werefirst examined by the investigators withinsix hours, 176 (84%) within 12 hours, and200 (95%) within the first day. Wheninformation was available from several

examinations, the best response in a giveninterval was used for analysis. To concen¬trate on patients in whom prognosis wasin doubt, we excluded those with an obvi¬ously good (eg, prolonged syncope ordelayed recovery from anesthesia) or poorprognosis (eg, brain death) and requiredthat patients remain in coma for at least

six hours. Patients, their families, or

health professionals were questioned regu¬larly about a variety of neurological anddaily life functions; these included thepatient's housing, speech, return to theprior level of function, and independencein ambulation, bathing, dressing, foodpreparation, eating, and use of toilet facil¬ities. Their functional state was catego¬rized at 1, 3, 6, and 12 months into one offive grades4: (1) no recovery (continuedcoma until death), (2) the vegetative state(eyes-open wakefulness without evidenceof cognitive awareness), (3) severe disabil-

From the Department of Neurology and ResearchCenter in Cerebrovascular Disease, New York Hos-

pital-Cornell Medical Center (Drs Levy, Caronna,Lapinski, and Plum); Department of Statistics,Columbia University (Dr Singer); and the GraduateSchool of Business, New York University (Dr Fryd-man), New York.

A preliminary report o f this work was read beforethe American Academy of Neurology, Boston, April12, 1984.

Reprint requests to Department of Neurology(A-569), Cornell Medical Center, 1300 York Ave,New York, NY 10021 (Dr Levy).

ownloaded From: http://jama.jamanetwork.com/ by a Vanderbilt University User on 02/20/2016


2/7

Table 1.—Factors Not

Associated With Outcome*

No. (%) WhoAchieved

IndependentVariable No. Function

Age, yr60 121 14 (12)

SexM 128 12 (9)F 82 14 (17)

Site of onset

Out-of-hospital 72 11 (15)Intensive care 53 7 (13)Operating suite 19 2 (11)General ward 65 6 (9)

Cause o f comaCardiac arrest 150 15 (10)Respiratory failure 22 4 (18)Miscellaneous 38 7 (18)

Paroxysmal activityConvulsions 32 4 (13)Isolated myoclonus 21 2 (10)Neither 157 20 (13)

'None of these variables was significantly associ¬ated with outcome from coma by

x2 testing with

appropriate degrees of freedom.

ity (conscious but dependent on others foraspects of daily function), (4) moderatedisability (independent but unable toresume the prior level of activities), and(5) good recovery (able to resume the priorlevel of function). To avoid influencingpatient outcomes, the investigators re¬frained from any involvement in theircare.

Analysis

Data were subjected to univariate and

multivariate analyses in an

effort to relatethe early clinical picture to the ultimatefunctional state. Among the analyticmethods applied was a computerizedrecursive partitioning algorithm' that con¬structed classification trees to predict thebest functional state within the first year.Recursive partitioning first identifies thevariable having greatest predictive powerand, thereby, divides the population intotwo groups, each with relatively homoge¬neous outcomes. It then divides these

subgroups on the basis of the most predic¬tive variables for each; the process contin¬ues until a group either has uniform

outcome or

is too small to split further.The program utilizes categorical andordered data rather than requiring inter¬val scales (equal steps between the values)and thus avoids the artificiality of inter¬preting scores as strict numerical values.Recursive partitioning can be directed tomaximize overall accuracy or, alternative¬

ly, to minimize particularly costly errors.Cost in this sense incorporates personal,social, humane, and economic considera¬tions, and different assessments of costwill often result in different classificationtrees." For th e present study, we arbitrari¬ly decided that the error of predicting a

Table 2.—Clinical Signs That Predict Future IndependentFunction: Percent (No.) of Patients"

NeurologicalResponse

No. of Days After Onset of Coma

45 (14/31)50 (5/10)

20 (5/25) 5 (1/21)1 1 (20/185) 6 (6/104)

Verbal responseOriented

...

...

Confused...

71 (5/7)

Inappropriate...

50 (1/2)Incomprehensible 60 (3/5) 33 (3/9)None 13 (11/82) 8 (5/59)

Eye openingSpontaneousTo noise

To painNone

Pupillary light reflexPresent

Absent

Corneal reflex

Present

Absent

Spontaneous eye movementsOrienting

Roving conjugateRoving dysconjugateOther movement

None

Oculocephalic responseNormal

Full

Minimal

None

Oculovestibular responseNormal

Full or tonic atypicalMinimal

None

Motor response (best limb)Obeying

...

64 (7/11)

tocalizing...

38 (3/8)

Withdrawal 29 (10/35) 38 (13/34)Flexor 14 (5/35) 3 (1/30)Extensor 18 (7/38) 0 (0/22)None 4 (4/102) 3 (2/61)

16 (25/157)0 (0/52)

17 (22/130)4 (3/71)

36 (12/33)17 (2/12)12 (3/25)

6 (8/140)

17 (16/96)13 (4/31)

7 (6/83)

18 (25/138)0 (0/29)

21 (25/119)0 (0/37)

70 (7/10)

29 (14/48)14 (1/7)

0 (0/21)5 (4/80)

62 (10/16)13 (12/92)12 (2/16)

5 (2/42)

0 (0/2) 59 (10/17)18 (21/115) 13 (12/89)

5 (1/21) 0 (0/13)5 (3/56) 6 (2/33)

100 (8/8)

80 (8/10)

25 (1/4)12 (1/8)

5 (2/37)

44 (22/50)17 (1/6)

0 (0/11)

4 (2/57)

21 (24/112)0 (0/11)

24 (24/101)0 (0/ 14)

77 (20/26)

1 1 (4/37)0 (0/5)

0 (0/9)

2 (1/46)

70 (19/27)6 (4/64)

12 (1/8)

4 (1/24)

65 (13/20)5 (3/58)

17 (1/6)6 (1/17)

86 (18/21)12 (1/8)

24 (6/25)0 (0/18)0 (0/20)

0 (0/31)

100 (11/11)75 (9/12)

50 (1/2)0 (0/5)6 (1/17)

46 (23/50)33 (1/3)

0 (0/6)

0 (0/17)

32 (23/71)

0 (0/4)

34 (23/68)0 (0/2)

74 (20/27)

10 (3/30)0 (0/4)

17 (1/6)

0 (0/9)

84 (21/25)5 (2/41)

20 (1/5)

0 (0/3)

50 (10/20)8 (2/25)

50 (1/2)

0 (0/4)

78 (18/23)55 (5/9)

8 (1/13)0 (0/12)0 (0/8)

0 (0/11)

'For each time period, the denominator in parentheses is the number of patients demonstrating eachneurological response. The numerator indicates the number of these patients subsequently achieving a bestone-year functional state of moderate disability or good recovery. Percentages represent the ratios inparentheses. Responses are arranged hierarchically, with the best in each group appearing first, and thepoorest, last. With few exceptions, the relationship between different recovery patterns and the clinicalresponses was always significant by appropriate \ tests with P


3/7

210 Patients at Initial

Examination

Pupillary Reflex:Present?

^I Motor:Extensor or Better?

Motor:Flexor or Better?

Spont Eye Movt:Rov Conj or Better?

Total No.ol

Patients

52

-» 62

47

22

Best 1 -yr Recovery (% of Total)

No Recov Mod Disab

Veg State Sev Disab Good Recov

94 6 0(84-99) (1-16) (0-7)

89 5 6(78-95) (1-14) (2-16)

77 13 11(62-88) (5 -26 ) (4-23)

59 14 27(36-79) (3-35) (11-50)

41 19 41(22-61) (6-38) (22-61)

124 Patients at 3 Days

3D Motor:Withdrawal or Better?

3D Spont Eye Movt:Orienting?

Total N o.of

Patients

28

26


No Recov Mod Disab


93 7 0(84-98) (2-16) (0-5)

61 21 18(41-79) (8-4 1) (6-37)

8 15 77

(1-25) (4-35) (56-91)

Fig 1.—Rules based on neurological examination that classify patientsin hypoxic-ischemic coma in accordance with their best functionalstate within first year. Figures in parentheses represent the 95%confidence intervals for percentages given immediately above. Abbre¬viations used in describing outcomes follow: recov indicates recovery;veg, vegetative; disab, disability; sev, severe; and mod, moderate. The

following abbreviations are used in describing neurological signs:

spont eye movt indicates spontaneous eye movements; and rov conj,roving conjugate. Initial examination (top left) was obtained within sixhours of onset of coma in 55% of patients and within 12 hours i n 84%.Modifiers INIT, 1D, 3D, 1W, and 2W refer to initial, 0- to 1-day, 2- to

3-day, 4- to 7-day, and 8- to 14-day examinations, respectively. If nodata are available, response to relevant question is considered to be"no." Where changes in clinical signs are utilized (top right andbottom), difference is between best response of previous time intervaland best response of current interval. Physicians applying these rulesmust exercise caution when residual anesthetics or depressant drugs(including anticonvulsants) are present, should wait until the end ofeach time interval to make certain that the best clinical response has

been observed, and should not use any rules that require responsesthat are unavailable for a specific patient.

i— 168 Patients at 1 Day

ID Spont Eye Movt:Rov Conj or Better?


1D Eye Opening:Improved at Least2 Grades?_

Total No.of

Patients

93

22

23

30

Best 1 -yr Recovery (% o f Total)

No Recov Mod Disab


95 4(88-98) (1-11)

77(55-92)

7(1-22)

14

(3-35)

70 13(47-87) (3-34)

1(0-6)

9(1-29)

17

(5-39)

3 63(0 -1 7) (4 4-8 0)

76 Patients at 1 wk

3D Eye Opening:Spontaneous?

Init Spont Eye Movt:Rov Conj or Better?

1W Motor:

Obeys Commands?

Total No.of

Patients

26

12

32


No Recov Mod DisabVeg State Sev Disab Good Recov

100 0 0(87-100) (0-13) (0-1 3)

58 42 0(27-85) (15-72) (0-26)

67

(22-96)

17

(0-64)

6 22

(1-21) (9-40)

17

(0-64)

72(53-86)

61 Patients at 2 wk

-Any?-3D Motor:Obeys Commands?3D Ey e Opening:Spontaneous?2W Eye Opening:Improved at Least2 Grades?

Yes

2W Oculocephalic:Normal?

Total No.

of

Patients

17

18

26

8est 1 -yr Recovery (% of Total)

No Recov Mod DisabVeg State Sev Disab Good Recov

100 0(80-100) (0-20)

44 44

(22-69) (22-69)

4

(0-20)15

(4-35)

0

(0-20)

22

(6-48)

81

(61-93)

esthetic accidents being the mostcommon. Roughly equal numbers ofpatients sustained their hypoxic-ischemic insults in an intensive caresetting (53, or 26%), general hospitalward (66, or 31%), or out of hospital(72, or 34%). Nineteen patients (9%)failed to awaken after operative pro¬cedures. The population consisted of108 men and 82 women. The median

age was 61 years, with all but 16 olderthan 30 years of age.

Outcome

Most patients who were destined toawaken and do well did so within ashort time. By three days after theonset of coma, 25 patients regainedconsciousness (with cognitive aware¬ness); 19 of these went on to becomeindependent. After two weeks, thenumber of conscious patients roseonly to 28, of whom 21 recoveredindependence. Many patients reached

a vegetative state rapidly; among 47such patients, within the first day,only 11 became independent subse¬quently, and among 33 patients vege¬tative at one week, only threeregained independence. The numberof comatose patients dropped rapidlyafter the insult, with 106 still coma¬tose after one day, 57 at three days,and 17 at one week, of whom only oneever regained consciousness.

Improvement after one month was



4/7

rare. None of the 15 patients vegeta¬tive at one month ever regained inde¬pendent function, and only three of 16patients severely disabled at onemonth did so. None of these three

patients had early clinical signs thatwould have placed them in the groupwith the poorest prognosis.

The death rate associated

with

hypoxic-ischemic coma of six or morehours was high: 42 (20% ) died withinthe first 24 hours, 86 (41%) by the endof three days, and 134 (64%) by theend of the first week. Only 19 (10%)survived one year after the onset ofcoma. Of the 191 patients who diedwithin a year, most (105, or 55%)succumbed to nonneurological prob¬lems; 67 (35%) died of neurologicalcauses, and in 18 (10%), no exactdetermination of the cause of deathwas possible. Because of the frequen¬cy with which nonneurological factorsinterfered with full recovery, we ana¬lyzed our data to predict the bestfunctional neurological state attainedwithin the first year.

In terms of best outcome, overthree fourths of the patients eitherdied without opening their eyes (121,or 57%) or, although they openedtheir eyes, showed no evidence of

cognitive function and were classifiedas vegetative (43, or 20% ). Among theremaining patients who regained con¬

sciousness, 20 (10%) remained severe¬

ly disabled and dependent on othersfor routine needs. Six patients (3% )achieved a moderate disability, and 20(10% ) achieved a good recovery; these26 patients are grouped together ashaving recovered independent func¬tion. This is a useful distinctionbecause Maas et al" found littleinterobserver variability in differen¬tiating dependent from independentfunction.

Factors Not Related to Recovery

None of the following factors sig¬nificantly (P>.05 by x; testing withappropriate degrees of freedom) in¬fluenced the degree of recovery:patient age or sex, the site of theinitial insult, the cause of coma, orthe presence of postanoxic seizures(Table 1).

Individual NeurologicalSigns Related to Recovery

The absence of certain brainstemreflexes at the initial examination

Table 3.—Rules That Identify PatientsWith Poor or Good Prognosis*

Time After

CardiacArrest Clinical Sign

Patients With Virtually No Chanceof Regaining Independence

Initial No pupillary light reflexexamination

1 day 1-Day motor response nobetter than flexor an d 1-day spontaneous eyemovements neither

orienting nor roving con¬ jugate

3 days 3-Day motor response nobetter than flexor

1 wk 1 -wk motor response not

obeying commands andinitial spontaneous eyemovements neither

orienting nor roving con¬ jugate and 3-day eyeopening not sponta¬neous

2 wk 2-wk oculocephalic re¬sponse not normal and

3-day motor responsenot obeying commandsand 3-day eye openingnot spontaneous and 2-wk eye opening not im¬proved at least twogrades

Patients With Best Chanceof Regaining Independence

Initial Pupillary light reflexesexamination present and motor re¬

sponse flexor or exten¬sor an d spontaneouseye movements rovingconjugate or orienting

1 day 1-Day motor responsewithdrawal or better a nd

1-day eye opening im¬proved at least 2grades

3 days 3-Day motor responsewithdrawal or better a nd

3-day spontaneous eyemovements normal

1 wk 1-wk motor responseobeying commands

2 wk 2-wk oculocephalic re¬sponse normal

"This table summarizes the rules on the t op andbottom lines of Fig 1.

identified patients with little or nolikelihood of meaningful recovery(Table 2). None of 52 patients lackingpupillary reflexes at the initial exam¬

ination ever became independent, andonly three regained consciousness.Three (4%) of 71 patients lackingcorneal reflexes when first examinedattained a good recovery within ayear, but no patient who lacked cor¬neal reflexes on or after the first dayregained consciousness.

Although the pattern of motorresponses eventually correlated withrecovery, neither their initial absencenor the presence of extensor or flexor

posturing ruled out recovery. After

three days, however, absent or pos¬turing motor responses were incom¬patible with future independence.Five patients who had motor re¬sponses poorer than withdrawal atthree days and four patients at sevendays regained consciousness, but allremained severely disabled.

Certain early signs

were associatedwith relatively good chances of recov¬ery. At the initial examination, themost favorable sign was incompre¬hensible speech (moaning), but thiswas rare. At one day, each of thefollowing signs was associated with atleast a 50% chance of regaining inde¬pendence: confused or inappropriatespeech, orienting spontaneous eyemovements, normal oculocephalic oroculovestibular responses, and obe¬dience to commands. Information onskeletal muscle tone and deep-tendonreflexes has been omitted becauseof concern about interobserver varia¬

bility.

Multivariate Analysisof Prognostic Variables

Reliance on individual clinical signscan be potentially misleading. Forexample, the likely outcome of apatient with intact pupillary lightreflexes but absent corneal reflexescannot be deduced from inspection ofTable 2. We applied several multivar¬iate

techniques to the data and found

that recursive partitioning' was mostuseful in segregating patients intodifferent groups on the basis of prog¬nosis.

Figure 1 illustrates the classifica¬tion rules generated by recursive par¬titioning for the 210 patients at dif¬ferent times after the onset of coma.

Figure 1 (top left) shows, for example,that 52 patients at initial examina¬tion could be identified on the basis ofabsent pupillary light reflexes asbeing in the poorest prognosis group

and that none of these ever achievedindependent function; the 95% confi¬dence interval for this result is 0% to7%. By contrast, 27 patients hadpreserved pupillary light reflexes,motor responses that were extensoror better, and roving conjugate ororienting spontaneous eye move¬ments. Eleven (41%) regained inde¬pendent function; the confidence in¬terval for this figure is 22% to 61%.Figure 1 (top right, center left, centerright, and bottom) shows correspond-



5/7

106 ComatosePatients at 1 Day

1D Oculovestibular:Any Response?

Init Motor:Withdrawal or Better?

ID Spont Eye Movt:Rov Conj or Better?

Total No.of

Patients

45

19


No RecovVeg State

Mod DisabSev Disab Good Recov

98 0 2(88-100) (0-8) (0-15)

91 9 0(77-98) (2-23) (0-10)

86 0 14( 42-100) ( 0- 41) (0-56)

63 16 21(38-84) (3-40) (6-46)

Fig 2. —

Rules based on neurological examination that classify patientssurviving one day after hypoxic-ischemic coma with different levels ofconsciousness in accordance with their best functional state withinfirst year. Abbreviations are as described for Fig 1. Operationaldefinitions of levels of consciousness follow: coma (top left)—no eyeopening regardless of stimulus, no comprehensible words, andcommands not obeyed; vegetative state (top right)—eye openingspontaneously, to noise, or to pain, but no comprehensible words, andcommands not obeyed; and consciousness (bottom)—comprehensi¬ble words or commands obeyed. These definitions, which wereadopted on detailed examination of vegetative patients," wereconsidered an improvement over those used at time study was begunand in previous publications'2 because they distinguish coma fromvegetative state.

47 VegetativePatients at 1 Day

1D Spont Eye Movt:Any Rov or Better?


Total No.of

Patients

10

11

26


Mod DisabVeg State Sev Disab Good Recov

100 0 0(69-100) (0-31) (0-31)

82(48-98)

38

(20-59)

18

(2-52)

19(7-39)

0(0-28)

42(23-63)

15 ConsciousPatients at 1 Day

Init Pupillary Reflex:Present?

1D Spont Eye Movt:

Rov Conj or Better?1D Oculovestibular:Normal?

Total No.of

Patients


Sev Disab Mod Disab Good Recov

86

(42-99)

0

(0-37)

0(0-41)

0

(0-37)

14(0-58)

100(63-100)

ing rules for patients who survived 1,3, 7, and 14 days. Validity testing onsplit samples showed the rules to bestable. For example, at the initialexamination, only one patient whodid well was mistakenly classified inthe group with poorest prognosis, anerror rate well within the confidenceinterval of 0% to 7%.

The algorithms of Fig 1 permitseveral conclusions. Inspection of theclinical signs actually utilized in theclassification rules shows that themost predictive variable at the initialexamination was the pupillary lightreflex, the sign shown by both Teas-dale et al8 and van den Berge et al9 tohave the least interobserver variabili¬

ty. Thereafter, the motor responsebecame the variable having the great¬est predictive power. Other powerful

clinical observations that contributedempirically to useful rules were pupil¬lary light reactions, spontaneous eyemovements, eye opening, and oculo¬cephalic response. The verbal re¬sponse, a key component in the Glas¬gow Coma Scale,10 never appeared asan important predictor of outcome.

Recursive partitioning identified ateach time a portion of survivingpatients as having very little chanceof recovering independent function,and only one patient was incorrectly

assigned to these groups with thepoorest prognosis. That misclassifica-tion occurred at day 1 and concerneda young woman with underlying renalfailure who suffered a cardiac arrest.Her early misclassification reflectedpoor clinical responses probablycaused

by renal failure.

Although the

clinician might weigh the mitigatinginfluence of this metabolic insult, thecomputer program did not. Other¬wise, the rules identified at each timea substantial proportion of survivors,none of whom ever recovered (25% atadmission, 55% at one day, 56% atthree days, 34% at seven days, and28% at 14 days). These rules aresummarized in Table 3.

The rules also identified early asizable subpopulation of patients whodid well after hypoxic-ischemic coma

(Table 3). For example, the classifica¬tion rules for all 168 survivors at oneday placed 123 into either the poorestprognosis group (only one of 93patients regaining independence) orthe best prognosis group (with 21 of30 patients regaining independentfunction). The overall proportion ofsurvivors assigned the best prognosisrose gradually from 13% at admis¬sion, to 18% at one day, 21% at threedays, 42% at seven days, and 43% at14 days.

Likely Outcome AmongSurviving Patients With Different

Levels of Consciousness

The ease of applying recursive par¬titioning permitted us to examinespecific subpopulations. For example,

patients who survived various inter¬

vals after the onset of coma were

subcategorized as comatose, vegeta¬tive, or conscious. Figure 2 illustratesthe classification rules for patientswho survived one day; other figures(not shown but available on requestfrom the authors) were generated forpatients with different levels of sen-sorium at later intervals after the

onset of coma. Over one third of

patients vegetative at one day couldbe classified as likely to remain vege¬tative on the basis of absent or pos¬

turing motor responses (Fig 2, topright). Among patients conscious atone day, recursive partitioning differ¬entiated surprisingly well betweenthose who were left with severe dis¬

ability and those who regained inde¬pendent function (Fig 2, bottom). Thepersistence of coma at seven daysprecluded recovery of independence.This was not true, however, for thevegetative state, where even after twoweeks, two of 21 patients subsequent¬ly did well; these two patients were



6/7


7/7

normalities or drugs (eg, anticho-linergics on pupillary reactions, de¬pressants on corneal reflexes, orparalytic agents on motor re¬sponses).

The ability to predict with assur¬ance that even half of a population ofpatients in coma would have good orpoor outcome and to act on these

predictions can offer a major benefitto personal physicians, patients, pa¬tients' families, and health planners.One might ask why, if 134 of 210patients died within a week, predic¬tion even matters. Perhaps the bestanswer is that if one waits until

patients either awaken or die, theirmedical condition will have stabilized.The result can be prolonged survivalfor vegetative patients who mightotherwise have died. For most of us,this is an undesirable state. Many lay

persons fear permanently incapaci¬tating neurological disability: "What¬ever you do, doctor, don't leave me avegetable or a permanent burden onmy family." Even for those whoregain mental competence, being re¬suscitated seems to be a harrowingexperience: 42% in one study24 statedthat they wished not to be similarlyresuscitated in the future. The abilityto predict outcome could also sparefamilies the emotional and financialburden of prolonged care of patientswith

hopeless prognoses. The

impacton allocation of scarce health re¬sources can be appreciated from con¬sideration of the potential effect ofconfidently identifying poor-progno¬sis patients. At one day, we identifiedas having a poor prognosis 93 patients(Fig 1, top right), all b ut one of whomfailed to regain independence and 88of whom failed even to recover con¬sciousness. Continued support ofthese 88 patients involved a total ofover 500 hospital days, frequently inintensive care units. Not only was this

useless terminal care costly (over$250,000) but the intensive care bedscould have been used to treat patientswith greater chances of recovery.These results contrast, incidentally,with those of Liberthson et al21 adecade ago, who concluded that 41%of patients who died did so within thefirst day, and 75% within the firstweek. Among our patients who died,only 23% did so within the first day,and half within the first week. Earlyidentification of patients likely to

remain vegetative after cardiac arrestcould reduce the suffering of familiesand limit encroachment on the medi¬cal commons.

Even for patients who want everypossible effort made, accurate prog¬nosis can be valuable in choosingmanagement. Recent evidence2526 sug¬gests that neuropathological abnor¬malities can continue to evolve forhours or days after global brainischemia. In such desperate cases,delayed treatment might still limitthe eventual extent of ischémie braindamage. Identification of poor- andgood-prognosis patients in the earlyhours or days after an arrest willfacilitate the design and interpreta¬tion of appropriately stratified treat¬ment trials.

The care of patients after cardio-pulmonary arrest poses serious ques¬

tions in terms of allocation ofresources and humane care for both

patients and their families. Implicitestimations of prognosis are involvedin the management of all seriously i llpatients, but in the absence of well-established facts, such predictionsusually amount to little more thaninformed guess work. Judicious appli¬cation of the present data may pro¬vide a more rational approach formanaging patients who sustain hy-poxic-ischemic coma.

This study was supported by contract NS-42328 and grant NS-03346 from the NationalInstitute of Neurological and CommunicativeDisorders and Stroke, Bethesda, Md, as well asgrant 1356 from the Josiah Macy, Jr, Founda¬tion, New York, and grant 6024 from the RobertWood Johnson Foundation, Princeton, NJ. DrLevy was supported by an Established Investiga-torship from the American Heart Association,Dallas, with funds contributed in part by theNew York Heart Association, New York.

The authors thank the following for theirclinical participation: David Shaw, MB; NiallE. F. Cartlidge, MB; and David Bates, MB; ofRoyal Victoria Infirmary, Newcastle-upon-Tyne,England; Seth Finklestein, MD, of San FranciscoGeneral Hospital; and Barrie Hurwitz, MD; R.John Leigh, MD; John H. Dougherty, Jr, MD; andRobert L. Van

Uitert, MD; o f New York

Hospi¬tal-Cornell Medical Center, New York. Specialthanks are due Jerome H. Friedman, PhD, ofStanford University, Palo Alto, Calif, for provid¬ing the recursive partitioning program andRobin P. Knill-Jones, MB, of Ruchill Hospital,Glasgow, Scotland, for helpful advice.

References

1. Bates D, Caronna JJ, Cartlidge NEF, et al:A prospective study of nontraumatic coma:Methods and results in 3 10 patients. Ann Neurol1977;2:211-220.

2. Levy DE, Bates D, Caronna JJ, et al:Prognosis in nontraumatic coma. Ann InternMed 1981;94:293-301.

3. Plum F, Posner JB : The Diagnosis of Stupor

and Coma, ed 3. Philadelphia, FA Davis Co,1980.

4. Jennett B, Bond M: Assessment of outcomeafter severe brain damage: A practical scale.Lancet 1975;1:480-484.

5. Breiman L, Friedman J, Olshen R, et al:Classification and Regression Trees. Belmont,Calif, Wadsworth International Group, 1984.

6. Levy DE: Ethical considerations in the careof unconscious patients, in Pfaff DW (ed): Ethi-cal Questions Related to Brain Behavior. NewYork, Raven Press, 1983, pp 57-72.

7. Maas AIR, Braakman R, Schouten HJA, etal: Agreement between physicians on assess-ment of outcome following severe head injury.J Neurosurg 1983;58:321-325.

8. Teasdale G, Knill-Jones R, van der Sande J:Observer variability in assessing impaired con-sciousness and coma. J Neurol Neurosurg Psy-chiatry 1978;41:603-610.

9. Van den Berge JH, Schouten HJA, Boom-stra S, et al: Interobserver agreement in assess-ment of ocular signs in coma. J Neurol Neuro-surg Psychiatry 1979;42:1163-1168.

10. Teasdale G, Jennett B: Assessment ofcoma and impaired consciousness: A practicalscale. Lancet 1974;2:81-84.

11. Dougherty JH Jr, Rawlinson DG, LevyDE, et al: Hypoxic-ischemic brain injury and thevegetative state: Clinical and neuropathologicalcorrelation.

Neurology 1981;31:991-997.12. Bell JA, Hodgson HJF: Coma after cardiacarrest. Brain 1974;97:361-372.

13. Willoughby JO, Leach BG: Relation ofneurological findings after cardiac arrest tooutcome. Br Med J 1974;3:437-439.

14. Snyder BD, Ramirez-Lassepas M, LippertDM: Neurologic status and prognosis after car-diopulmonary arrest: I. A retrospective study.Neurology 1977;27:807-811.

15. Snyder BD, Loewenson RB, Gumnit RJ, etal: Neurologic prognosis after cardiopulmonaryarrest: II. Level of consciousness. Neurology1980;30:52-58.

16. Snyder BD, Hauser WA, Loewenson RB, etal: Neurologic prognosis after cardiopulmonaryarrest: III. Seizure activity. Neurology 1980;30:1292-1297.

17. Snyder BD, Gumnit RJ, Leppik IE, et al:

Neurologic prognosis after cardiopulmonary ar-rest: IV. Brainstem reflexes. Neurology 1981;31:1092-1097.

18. Earnest MP, Breckinridge JC, Yarnell PR,et al: Quality of survival after out-of-hospitalcardiac arrest: Predictive value of early neuro-logic evaluation. Neurology 1979;29:56-60.

19. Earnest MP, Yarnell PR, Merrill SL, et al:Long-term survival and neurologic status afterresuscitation from out-of-hospital cardiac ar-rest. Neurology 1980;30:1298-1302.

20. Longstreth WT Jr, Diehr P, Inui TS:Prediction of awakening after out-of-hospitalcardiac arrest. N Engl J Med 1983;308:1378\x=req-\1382.

21. Liberthson RR, Nagel EL, Hirschman JC,et al: Prehospital ventricular defibrillation:Prognosis and follow-up course. N Engl J Med

1974;291:317-321.22. Rosenberg GA, Johnson SF, Brenner RP:Recovery of cognition after prolonged vegetativestate. Ann Neurol 1977;2:167-168.

23. Shuttleworth E: Recovery to social andeconomic independence from prolonged post\x=req-\anoxic vegetative state. Neurology 1983;33:372\x=req-\374.

24. Bedell SE, Delbanco TL, Cook EF, et al:Survival after cardiopulmonary resuscitation inthe hospital. N Engl J Med 1983;309:569-576.

25. Pulsinelli WA, Brierley JB, Plum F: Tem-poral profile of neuronal damage in a model oftransient forebrain ischemia. Ann Neurol 1982;11:491-498.

26. Kirino T: Delayed neuronal death in thegerbil hippocampus following ischemia. BrainRes 1982;239:57-69.

predicting outcome from hypoxic-lschemic coma

Documents