374

Predicting outcome in critical care: the current status of the APACHE prognostic scoring system The APACHE (Acute Physiology and Chronic Health Evaluation) prognostic scoring system was developed in 1981 at the George Washington University Medical Center as a way to measure disease severity. APACHE II, introduced in 1985, was a simplified modification of the original APACHE. The APACHE !1 score consisted of three parts: 12 acute physiological variables, age and chronic health status. Probability of death can be derived by using the disease category and the APACHE H score. The uses of APACHE H include risk stratification to account for case mix in clinical studies, comparison of the quality of care among ICUs, and assessment of group and individual prognoses. APACHE III, a retinement of APACHE H, will be introduced in late 1990. The APACHE I11 data base includes 17,457 patients from a representative sample of 40 American hospitals. Additional potential uses of APACHE HI include the identification offactors in the ICU which contribute to outcome and assistance in individual patient decisionmaking. This article reviews the development, current uses and potential applications of the APACHE ~3,stem. L'APACHE (acute physiology and chronic health evaluation) est un systOme de gradation pronostique qui s'est ddveloppE en 1981 d George Washington University Medical Center atit, de mesurer la sEvEritd de la maladie. APACHE II, introduit en 1985, fut une modification qui a simplifid I'APACHE original. L'APACHE H consiste en trois parties : 12 variables physiologiques aiguEs. 6ge et dtat de santE chronique. La probabilitE de

K e y

w o r d s

INTENSIVE CARE: A P A C H E , assessment. From the Department of ICU Research, The George Washington University Medical Center, 2300 K Street NW, Washington, DC 20037. Address correspondence to: Dr. D. T. Wong, Department of Anaesthesia, Toronto Western Division, The Toronto Hospital, 399 Bathurst Street, Toronto, Ont., Canada, M5T 2S8. Accepted for publication 15th November, 1990.

CAN J ANAESTH 1991 / 38:3 / pp374-83

David T. Wong MD, William A. Knaus MD

la mortalitd peut Etre ddduite en utilisant la catdgorie de maladie et le systdme d'dvaluation APACHE II. Les utilisations de I'APACHE H incluent la stratification du risque atin de tenir compte de l'identification des cas dans des dtudes cliniques mixtes, la comparaison de la qualitd des soins entre les unitds de soins intensifs et l'Evaluation des pronostics individuel et de groupe. APACHE HI, un ratinement de I'APACHE II, sera introduit vers latin des anndes 1990. Les donndes de I'APACHE II! incluent 17,457 patients d'une population representative de 40 h6pitaux amEricains. Les utilisations potentielles additionnelles de I'APA CHE IH incluent l' identification desfacteurs attr soins intensify qui contribuent d l'issue et tJ l'assistance concernant les decisions sur certains patients. Cet article revolt le ddveloppement des utilisations courantes, des applications potentielles du systEme APACHE.

Contents Development of A P A C H E - A P A C H E , 1981 - A P A C H E II, 1985 Current uses of A P A C H E I1 - Control for case mix - Quality assurance in ICUs - Non-ICU patients Resource allocation Computer versus clinicians Individual patient decisions Future Uses - A P A C H E Ill, 1990 - Objectives Major features Prediction over time Automation Conclusion Recent American data indicate that intensive care unit (ICU) beds account for 7% o f all hospital beds, 1 5 - 2 0 % of hospital costs and 1% of the gross national product (GNP).I Improving therapeutic capabilities and increas-

Wong and Knaus: THE APACHE SYSTEM ing use of monitoring technology have escalated ICU demand at a time of economic constraint and resource limitation. Precise disease classification and accurate outcome prediction can optimize ICU bed usage by reducing unnecessary low-risk monitored-only patients and futile care of terminally ill patients. The APACHE (acute physiology and chronic health evaluation) prognostic scoring system was developed in 1981 at the George Washington University Medical Center as a way to measure disease severity. 2 The APACHE score was found to correlate directly with hospital mortality. APACHE 11, introduced in 1985, was a simplified refinement of the original APACHE. 3 Probability of death can be derived by using the disease category, acute physiological score, age and the chronic health status. The uses of APACHE ii include risk stratification, comparison of the quality of care, and prognosis. APACHE I11, a refinement of APACHE 1I will be introduced in late 1990. 4 This article reviews the development, current uses and potential applications of the APACHE system.

Development of the APACHE prognostic scoring system Although few question the need for ICUs, there is a lack of data to document their effectiveness in improving patient outcome. 5 This is because it is difficult to conduct randomized studies in patients at high risk of dying and, until recently, there was no accurate method to account for case mix in an ICU overtime or among different ICUs. Therefore it was difficult to determine if outcome variations were related to therapeutic effort or to a difference in case mix. A severity-of-disease scoring system allows the investigator to risk-stratify patients and compare patient risk from different units. A conceptual model for the APACHE prognostic scoring system was formulated in 1979. 2 The initial model involved the identification of factors that influence outcome from an acute illness (Table I). APACHE incorporated the three patient factors (disease, reserve, severity) which were treatment-independent. Disease was classified using the primary diagnostic reason for ICU admission, lfa specific diagnosis could not be given, then TABLE I

Factors influencing outcome for an acute illness

Patient factors (before treatment)

Treatment factors (post-treatment)

Type of disease Physiologic reserve - age - chronic disease Severity of illness

Type of therapy Amount of therapy Response to therapy

375 the principle organ system dysfunction precipitating ICU admission was used. Physiological reserve was reflected by age and prior health status. Severity of disease was measured by acute physiological abnormalities, represented by clinical and laboratory measurements available at presentation to the ICU. The original APACHE consisted of two parts: the APS (acute physiology score) representing the degree of acute illness and CHE (chronic health evaluation) indicating physiological reserve before the acute illness. The APS variables were developed by a panel of physicians from medicine, surgery and anaesthesia. Thirty-four variables were selected and relative weights (0-4) were assigned according to the clinicians' clinical experience and a review of the literature. The greatest degree of abnormality (worst value) of each variable within the first 32 hr after admission was used. The CHE consisted of a questionnaire inquiring about the number of recent physician visits, activities of daily living and the presence of carcinoma. Patients were then classified into A for excellent health to D for severely failing health. The final score consisted of an APS and a CHE (e.g., 25-D). Probability calculations were not part of the original APACHE system. Results from the initial study group indicated a direct relationship between the APS score and the probability of death. However, with the CHE, only class D was found to be independently associated with mortality. The predictive ability of the APACHE score was subsequently validated in new populations of patients. 6-s Criticisms of the APACHE included the large number of variables and the 32 hr allowed for data collection. Further analysis and modifications led to the development of the APACHE 11. APACHE I13 which was introduced in 1985 incorporated a number of important changes. First, the number of APS variables was reduced from 34 to 12 (Figure I). This was accomplished by eliminating infrequently measured variables (e.g., serum lactate, osmolality) and redundant variables (e.g., BUN). Subsequently, using multivariate analysis, the smallest number of variables that reflected physiological derangement yet maintained statistical precision was found to be 12. 3 In addition, the threshold and weights of variables were modified according to their statistical correlation to hospital mortality. Particularly, the Glasgow Coma Scale was given an increased weight of 12 and acute renal failure was double-weighted with a maximum score of 8. The most abnormal APS values within the first 24 hr of ICU admission were used. Chronic health points were assigned only for severe organ system dysfunction. Nonoperative and emergency surgery were given additional weight and age was incorporated into the APACHE II score. With these modifications, APACHE II then consisted

376

CANADIAN

THE APACHE PHYSIOLOGIC

II S E V E R I T Y O F D I S E A S E C L A S S I F I C A T I O N

VARIABLE

0

TEMPERATURE - - reclal ('C)

--

--

33.5".33 9 ~

,oQ29

0

34 "-35 9"

~0~o9

+2

+3

0

+4

0

32 "-33 9"

30".31 9"

0

"~29.9"

c) .,,;a9

0

0

0

0

0

0

140.179

t I O-139

70-109

55.69

40.54

'~ 39

0

0

,'t.50

35.49

0 ~ S00

@

,

350,490

0

0

0

0

0

12-24

10-11

6"g

.'; 5

(3

200,349

< 200

'-0"~'6,~"~

A"TER'ALp" SERUMSOD'UM,mMoI'L)

,s%g

,so%g

SERUM POTASSIUM (mMoI/L)

~70

6 0-6.9

SERUM CREATININE (molt00 ml) (DouOle po.nt score for IIr renal failure)

0 ~'3 5

0 2.3.4

15 ,g

o 50-69

25-34

O

-- --

,,% ,,0%4 0

5.5"5.9 0 1.5.1 9

0 po, 5170

7,33,7,49

o 139.149 0

3.5-5 4

0

3.3.4

ode,

55.~

7,25,7 32

,,,%4

o 120.129

o 111.119

5 ~ : < ss

0

0

2.5-2.9 '

< 2.5

0 06.1 4

0 < 0.6

O 60

O 50-59.9

CD 46-49.9

(D 30-45.9

O 20-29.9

O < 20

0 2"40

0 20-39.9

0 15.199

0 3.14 9

0 1.29

0