Indian Journal of Anaesthesia 2008;52:Suppl (5):663-675

Severity Scoring Systems in Paediatric Intensive Care Units Poonam Bhadoria1, Amit G Bhagwat2

Summary The main purpose of the paediatric intensive care unit (PICU) is to prevent mortality by intensively monitoring and treating critically ill children who are considered at high risk of mortality. The capability to estimate patient risk of death is extremely important because such estimate would be useful in achieving many different goals such as assessing patient’s prognosis, ICU performance, ICU resource utilization and also evaluating therapies, controlling and matching severity of illness in clinical studies. The lack of consistency, reliability, and accuracy in physician’s subjective opinions concerning patient status necessitates quantitative clinical scores. Physicians are in general poor prognosticators. In fact more accurate predictions result from actuarial methods than from clinical methods. Besides scoring systems have been developed in response to increasing emphasis on the evaluation and monitoring of health services. This review focuses on the various scoring system available for predicting mortality risk in paediatric critical care setting, their present, past and future and their uses in ICU management and administration. Key words

Severity scoring system, Paediatric intensive care, PRISM, Risk assessment, Outcome-based quality assurance

assessments are critical for wide range of ICU management and administration. 2, 3, 4

Introduction Paediatric anaesthesiology today is a recognized specialty in anaesthesiology. The current status of paediatric anesthesiology is the product of a history that spans more than 150 years. The emergence of paediatric surgery as a specialty provided some of the impetus for the rapid evolution of paediatric anaesthesia practice, as have advances in paediatric medicine, particularly critical care paediatrics.

In context of intensive care, a rational and objective way to define and quantify severity of illness is through the development of probability models predicting mortality risks. Such predictive models or scoring system have been developed for all age groups including paediatric. 5, 6

The practice of paediatric critical care is dynamic and evolving. Paediatric population is a vulnerable group necessitating standard care for medically and surgically ill children. However, standard care is not well defined for paediatric critical care, as most of the protocols and practices in paediatric intensive care unit (PICU) are actually extrapolation of adult critical care. 1

Scoring systems are arrived at evaluation of the patient’s mortality risk in the ICU by assigning a score to patient and predicting the outcome. However, patient’s mortality is not only affected by ICU performance but also depends on many other factors such as demographic and clinical characteristic of population, infrastructure and non medical factors (management and organization), case mix and admission practice. 7

Whether adult or paediatric, severity of illness,

Therefore there is need for field testing of these

1. Professor, 2. Senior Resident, Department of Anesthesiology, Intensive care & Perioperative Medicine, Maulana Azad Medical College & Associated,New Delhi, Correspondence to: Poonam Bhadoria, Lok Nayak Hospital, Gobind Ballabh Pant Hospital & Guru Nanak Eye Centre, 4 - LF Todarmal Square, Todarmal Lane, Barakhamba road, New Delhi. 110001, Email: [email protected] 663

Indian Journal of Anaesthesia, October 2008(P.G.Issue)

scoring system in setting different from the one in which they were originally developed. The ideal probability model / scoring system would be institution independent and population independent.

Compared with other predictors, it cannot quantify mortality risk. Efforts to evolve the TISS score by combining physiological dysfunction with therapies has been relatively unsuccessful.10

Scoring system in paediatric intensive care units: Past, present & future

The Acute Physiology and Chronic Health Evaluation (APACHE) system

Initially scoring systems were developed for trauma patients and were either specific anatomical methods (abbreviated injury scale 1969, burn score 1971, injury severity score 1974) or specific physiological methods (trauma index 1971, Glasgow coma scale 1974, trauma score 1981 and sepsis score 1983).

The acute physiology and chronic health evaluation (APACHE) system was introduced (for adult patients) in 1981 by an expert panel of physicians who selected and weighed 34 laboratory and clinical measurements based on perceived impact on mortality. It consisted of 2 parts: An acute physiology score that reflected the degree of physiologic derangements and a chronic health evaluation that reflected patient’s status before the acute illness.

Therapeutic Intervention Scoring System (TISS) In 1974 Therapeutic intervention scoring system (TISS) was introduced by Cullen D J et al to quantitate severity of illness according to the therapeutic interventions received by the patients. 8 Each intervention had a value of 1-4 points based upon the complexity and invasiveness of intervention with a total score of 70 interventions. TISS has been utilized for many purposes which include:-

There are now three APACHE system i.e. I, II, III. An increasing APACHE II score reflects increased severity of disease and a higher risk of hospital death. But the system was neither designed nor intended to predict for individual patients and it has an error rate of approximately 15% for the prediction of hospital mortality using 0.50 decision point. APACHE III was introduced in 1991 to expand and improve the prognostic estimate provided by APACHE II. 11

1. Determining the severity of illness

APACHE III system consists of points for physiologic abnormalities, age and chronic health status. Scoring is based on a degree of abnormality in 17 physiologic variables (APS), which reflects value for vital signs, laboratory tests and neurological status. In addition, points are added based on age and 7 co-morbid conditions shown to have a significant impact on short term mortality. 12 It can be used to measure the severity of disease and to risk stratify patients within a single diagnostic category or independently define patient group. It can also be used to compare patient outcomes but only for ICU admissions meeting diagnostic and selection criteria similar to those used in APACHE study.

2. Establishing nurse patient ratio in ICU. 3. Assessing current utilization of hospital intensive care beds. 4. Establishing future need and numbers of ICU beds particularly in response to request for certification of need. TISS was found to be a useful tool for obtaining comparable data which could be utilized for administrative, management and clinical purposes, within and between hospital settings.8 Unfortunately, the TISS score is heavily influenced by diagnosis, indicating the TISS score depends on the monitoring and therapeutic philosophies of the physicians and institutions using it.9

The APACHE system is appropriate for adult ICUs. However the changing physiology with growth 664

Poonam Bhadoria. Severity scoring systems in paediatric ICU

and development within the wide spectrum of ages of paediatric patients prevents its direct application to PICUs.13 The limited number of patients and diverse conditions make diagnostic subgroups difficult to study.13 However in the recent past some scoring systems have been developed for PICU mortality prediction like for example Physiologic stability index ( PSI ).

as TISS only indirectly reflects the severity of illness by assessing therapeutic needs. PSI assesses the severity of acute illness in the total population of paediatric intensive care unit patients by quantitating the degree of derangement in 34 variables from 7 major physiologic systems. Each variable was assigned a score of 1(abnormality worth concern but not to change therapy), 3 (need to change therapy), and 5 (life threatening). (Table 1) This reflected the clinical importance of derangements but not necessarily the amount of deviation from the normal value. The most abnormal value of a variable recorded within 24 hours was used.4

Physiologic Stability Index (PSI). Physiologic Stability index (PSI) was developed by a group of paediatric intensivists in 1984 from TISS Table 1 Physiologic stability Index: Physiologic Systems (7) and Variables (34)

(1) Cardiovascular: systolic blood pressure, diastolic blood pressure, heart rate, cardiac index, C(a-v)O2, CVP, PCWP (2) Respiratory: respiratory rate, PaO2, PaO2/FIO2, PaCO2 (3) Neurologic: Glasgow coma score, intracranial pressure, seizures, pupils (4) Hematologic: hemoglobin, WBC count, platelet count, PT/PTT, FSP (5) Renal: BUN, creatinine, urine output (6) Gastrointestinal: AST/ALT, amylase, total bilirubin, albumin (7) Metabolic: sodium, potassium, calcium, glucose, osmolality, pH, HCO3 Points for each variable: • 0, 1, 3, 5 • reflect clinical importance of derangement, with more abnormal having higher point value • not intended to reflect magnitude of deviation from the normal value Variable

0 points

1 points

3 points

5 points

• Infants

• 66-129

• 55-65, or 130-160

• 40-54, or > 160

• < 40

• Children

• 66-149

• 65-75, or 150-200

• 50-74, or > 200

• < 50

Diastolic blood pressure, in mm Hg

< 90

90-110

> 110

• Infants

• 91-159

• 75-90, or 160-180

• 50-74, or 181-220

• Children

• 81-149

Systolic blood pressure in mm Hg

Heart rate, in beats per minute • < 50, or > 220

• 60-80, or 150-170

• 40-59, or 171-200

• < 40 or > 200

Cardiac index, in L per min per square meter > 3.0

2.0-3.0

1.0-1.9

< 1.0

Arterial to mixed venous oxygen

3.0-5.4

< 3.0, or 5.5-6.5

> 6.5

CVP, in mm Hg

0-15

< 0, or > 15

Wedge pressure, or left atrial

5-14

< 5, or 15-25

content difference, C(a-v)O2, in ml O2 per dL (vol%) > 25

pressure, in mm Hg Respiratory rate, in breaths per minute

Contd......... 665

Indian Journal of Anaesthesia, October 2008(P.G.Issue) Variable

0 points

1 points

3 points

5 points

• Infants

• < 50

• < 30

• 61-90

• > 90, apnea

• Children

• 50-60

• 51-70

• 51-70

• > 70

PaO2, in mm Hg

> 50

50-60

40-49

< 40

PaO2/FIO2

> 300

200-300

< 200

PaCO2 in mm Hg

30-44

< 30, or 45-50

51-65

< 65

pH

7.31-7.54

7.20-7.30, or 7.55-7.65

7.10-7.19, or > 7.65

< 7.10

Glasgow Coma Score

> 11

8-11

5-7

40

focal

grand mal or status

Seizures

epilepticus Pupils

equal and

equal and sluggish

unequal and sluggish fixed and dilated

responsive Hemoglobin, in g/dL

7.1-17.9

5.0-7.0, or 18.0-22.0

3.0-5.0, or 22.1-25.0

WBC count, per µL

5,001 - 19,999

3,000-5,000, or

< 3,000, or > 40,000

< 3.0

20,000 - 40,000 Platelet count, per µL

51,000 - 999,999 20,000-50,000, or > 1 M < 20,000

PT/PTT ratio, relative to normal

1.5

Fibrin split products in µg/mL

40

BUN, in mg/dL

< 40

40-100

> 100

Creatinine, in mg/dL

< 2.0

2.0-10.0

> 10.0

Urine output, in mL per kg per hour

> 1.0

0.5-1.0

< 0.5

AST / ALT, in IU/L

100

Amylase, in U/L

500

Total bilirubin, in mg/dL

3.5

Serum albumin, in g/dL

> 2.0

1.2-2.0

< 1.2

Sodium, in mEq/L

126-149

115-125, or 150-160

< 115, or > 160

Potassium, in mEq/L

3.6-6.4

3.0-3.5, or 6.5-7.5

2.5-2.9, or 7.6-8.0

< 2.5, or > 8.0

Calcium, in mg/dL

8.1-11.9

7.0-8.0, or 12.0-15.0

5.0=6.9, or > 15.0

< 5.0

Glucose, in mg/dL

61-249

40-60, or 250-400

20-39, or > 400

< 20

Osmolality, in mOsm/L

< 320

320-350

> 350

Bicarbonate, in mEq/L

16-32

< 16 or > 32

control PT/PTT

where: • infants are all those under 1 year of age; children are all those older than 1 year of age • AST/ALT is taken to be the ratio of the transaminases • hypoosmolality does not seem be included for evaluation physiologic stability index =SUM (points for each physiologic variable) Contd......... 666

Poonam Bhadoria. Severity scoring systems in paediatric ICU Interpretation Index scores: • minimum score 0 • maximum score 119 • higher scores indicate more severe disease Scores compared: • on day of admission • maximum score • 4-day average • trend over hospital course Trends over hospital course: • decreasing indicates improvement • increasing indicates worsening • unchanging probability of mortality = (EXP((0.277 * (4 day average PSI)) - 5.241)) / (1 + (EXP((0.277 * ( 4 day average PSI)) - 5.241)))

PSI however, is time consuming; requiring the use of 34 variables from 7 physiologic systems and also it is a subjective score. A total of 294 clinical classification system (CCS) class III and IV patients in a PICU were evaluated by using PSI / TISS ratio. Non survivors had significantly higher (p < 0.01) PSI and TISS scores than survivors. Medical patients had the highest PSI / TISS ratio scores while, cardiovascular patients had lowest PSI / TISS ratio scores. 13

Table 2 PRISM

To reduce the number of physiologic variables required for severity of illness assessment and to obtain an objective weighting of remaining variables, a second generation score called pediatric risk of mortality (PRISM) has been devised by Pollack MM et al in 1988. 14

(8) Pupillary reactions to light

Parameters: (1) Systolic blood pressure and age (2) Diastolic blood pressure (3) Heart rate (4) Respiratory rate (5) PaO2 to FIO2 ratio (6) PaCO2 (7) Glasgow coma score (9) PT and PTT (10) Total serum bilirubin (11) Serum potassium (12) Serum total calcium (13) Glucose (14) Bicarbonate Parameter

Pediatric Risk Of Mortality (PRISM)

Systolic blood pressure in mm Hg

Pediatric risk of mortality (PRISM) score allows for mortality risk assessment in the paediatric ICU. PRISM was developed from PSI to reduce the number of variables from 34 to 14 and number of ranges from 75 to 23 without losing the predictive power (Table 2). It is institution independent and can be used within limits to compare different intensive care units. 15

Age Limit

Ranges

Points

infants

130-160 55-65 > 160 40-54 < 40 150-200 65-75 > 200 50-64 < 50

2 2 6 6 7 2 2 6 6 7

children

In 1996 physiological variables and their ranges as well as diagnostic and other risk variables reflective of mortality risk were re evaluated by Pollack MM et

Contd......... 667

Indian Journal of Anaesthesia, October 2008(P.G.Issue) Parameter

Age Limit

Ranges

Diastolic blood pressure in mm Hg

all ages

> 110 mm Hg 6

Heart rate in beats

infants

> 160

4

< 90

4

per minute children Respiratory rate in

infants

breaths per minute children

PaO2/FiO2 PaCO2 in torr

all ages all ages

Points

Where: • Infants: 0-1 years of age • The first implementation of the score had PT and PTT handled as a ratio of PT to PTT. On review it is obvious that the intent is to compare each to the normal control. Unfortunately it is not stated if points are assigned if each is abnormal or if only one.

> 150

4

< 80

4

61-90

1

> 90

5

apnea

5

51-70

1

PRISM score = (systolic blood pressure points) + (diastolic blood pressure points) + (heart rate points) + (respiratory rate points) + (oxygenation points) + (glasgow coma score points) + (pupillary reaction points) + (coagulation points) + (bilirubin points) + (potassium points) + (calcium points) + (glucose points) + (bicarbonate points)

> 70

5

Interpretation:

apnea

5

• minimum score 0, which has an excellent prognosis

200-300

2

< 200

3

• maximum score 76, which is almost invariably associated with death

51-65

1

Prediction of Mortality in ICU

> 65

5

Glasgow coma score

all ages

1 month

> 3.5

6

all ages

3.0-3.5

1

6.5-7.5

1

< 3.0

5

> 7.5

5

7.0-8.0

2

12.0-15.0

2

< 7.0

6

> 15.0

6

40-60

4

Assessment: • sensitivity: correct prediction of nonsurvival

mg/dL Potassium in mEq/L

Calcium in mg/dL

Glucose in mg/dL

all ages

all ages

Bicarbonate in mEq/L all ages

250-400

4

< 40

8

> 400

8

< 16

3

> 32

3

• specificity: correct prediction of survival Limitations: • The PRISM score underpredicts deaths after cardiac surgery.

al to update and improve the performance of second generation PRISM score. Thus PRISM III was developed. 16

PRISM III This was based upon a sample of 11,165 consecutive admissions to 32 paediatric ICUs (10% of PICUs of USA) representing a wide diversity of organizational and structural characteristics. 16 The variables that were most predictive of mortality as indicated by the highest PRISM scores were minimum systolic BP, abnormal pupillary reflexes and stupor/coma were re-

Contd.......... 668

Poonam Bhadoria. Severity scoring systems in paediatric ICU

tained from PRISM score. Variables in the original PRISM that were not included in PRISM III are diastolic BP, respiratory rate, PaCO2/F1O2, serum bilirubin and calcium concentration.

PRISM III has 17 physiologic variables subdivided into 26 ranges and is population independent (Table 3). 16

Table 3 PRISM III The PRISM III score is an improved version of the PRISM score developed at the Children’s National Medical Center in Washington, DC based on data collected at 32 pediatric intensive care units using 11,165 admissions. Age Group

Age Range

neonate

0 to < 1 month

infant

1to 12 months

child

> 12 to 144 months (12 years)

adolescent

> 144 months (> 12 years)

Subscores: (1) cardiovascular and neurologic vital signs: 5 measures (2) acid-base and blood gas: 5 measures (3) chemistry tests: 4 measures (4) hematology tests: 3 measures (with PT and PTT counted as one) Grading variables: • Use the highest and/or lowest values for scoring. Cardiovascular and Neurologic Vital Signs

Findings

Points

Systolic blood pressure

neonate AND > 55 mm Hg

0

neonate AND 40 -55 mm Hg

3

neonate AND < 40 mm Hg

7

infant AND > 65 mm Hg

0

infant AND 45 -65 mm Hg

3

Heart rate

infant AND < 45 mm Hg

7

child AND > 75 mm Hg

0

child AND 55 -75 mm Hg

3

child AND < 55 mm Hg

7

adolescent AND > 85 mm Hg

0

adolescent AND 65 -85 mm Hg

3

adolescent AND < 65 mm Hg

7

neonate AND < 215 beats/minute

0

neonate AND 215 - 225 bpm

3

neonate AND > 225 beats/minute

4

infant AND < 215 beats/minute

0

infant AND 215 - 225 bpm

3

infant AND > 225 beats/minute

4

child AND < 185 beats/minute

0

child AND 185 - 205 bpm

3 Contd.....

669

Indian Journal of Anaesthesia, October 2008(P.G.Issue) Cardiovascular and Neurologic Vital Signs

temperature

mental status pupillary response

Findings

Points

child AND > 205 beats/minute

4

adolescent AND < 145 beats/minute

0

adolescent AND 145 - 155 bpm

3

adolescent AND > 155 beats/minute

4

< 33°C

3

33 - 40°C

0

> 40°C

3

Glasgow coma score >= 8

0

Glasgow coma score < 8

5

both reactive

0

1 reactive AND (1 fixed AND > 3 mm)

7

both fixed AND both > 3 mm 11 where: • The heart rate should not be monitored during crying or iatrogenic agitation. • Pupillary size should not be assessed after iatrogenic dilatation. • Body temperature may be rectal, oral, axillary or blood. • Mental status should not be scored within 2 hours of sedation, paralysis or anesthesia. If sedation, paralysis or anesthesia is continuous, score based status prior to sedation, paralysis or anesthesia. Acid-Base and Blood Gases

Findings

Points

acidosis

pH > 7.28 AND total CO2 >= 17 mEq/L (pH 7.0 - 7.28) OR (total CO2 5 - 16.9 mEq/L) pH < 7.0 OR total CO2 < 5 < 7.48 7.48 - 7.55 > 7.55 < 50 mm Hg 50 - 75 mm Hg > 75 mm Hg 34 mEq/L >= 50 mm Hg 42.0 - 49.9 mm Hg < 42 mm Hg

0 2 6 0 2 3 0 1 3 0 4 0 3 6

pH

PCO2

total CO2 PaO2

where: • PaO2 requires arterial blood. • PCO2 can be measured from arterial, venous or capillary specimens. Chemistry Tests Findings glucose potassium creatinine

200 mg/dL 6.9 mEq/L neonate AND 0.85 mg/dL infant AND 0.90 mg/dL child AND 0.90 mg/dL adolescent AND 1.30 mg/dL neonate AND 11.9 mg/dL not neonate AND 14.9 mg/dL

2 0 2 0 2 0 3 0 3

where: • Whole blood measurements for glucose are increased 10% over serum; for potassium 0.4 mEq/L. Hematologic Tests

Findings

Points

white blood cell count

>= 3,000 per µL

0

< 3,000 per µL > 200,000 per µL 100,000 - 200,000 per µL 50,000 - 99,999 per µL < 50,000 per µL neonate AND PT 85 seconds) not neonate AND PT 57 seconds)

4 0 2 4 5 0

platelet count

PT and PTT

3 0 3

where: • The upper limit of the normal reference ranges for PT and PTT are not given. Other factors to document: (1) nonoperative cardiovascular disease (2) chromosomal anomaly (3) cancer (4) previous ICU admission during current admission (5) pre-ICU CPR during current admission (6) post-operative (not including catheterizations) during past 24 hours (7) acute diabetes with ketoacidosis or other severe complication (8) admission from inpatient unit (do not count if in ICU for < 2 hours or if transferred from surgical recovery room) cardiovascular and neurologic subscore = = (points for systolic pressure) + (points for temperature) + (points for mental status)+ (points for heart rate) + (points for pupillary reflex) acid-base and blood gas subscore = (points for acidosis) + (points for pH) + (points for PaCO2)+ (points for total CO2) + (points for PaO2) chemistry subscore = (points for glucose) + (points for potassium) + (points for creatinine)+ (points for blood urea nitrogen) hematology subscore = (points for WBC count) + (points for platelet count) + (points for PT and PTT testing) Total PRISM III score = (cardiovascular and neurologic subscore) + (acid base and blood gas subscore) + (chemistry subscore)+ (hematology subscore) Interpretation: • minimum subscore and total score: 0 • maximum cardiovascular and neurologic subscore: 30 • maximum acid-base and blood gas subscore: 22 • maximum chemistry subscore: 10 • maximum hematology subscore: 12 • maximum total PRISM III score: 74 Contd..... 671

Indian Journal of Anaesthesia, October 2008(P.G.Issue) • The higher the total score, the worse the prognosis. • A rising score indicates deterioration. • If performed during the first 12 hours in the ICU, the score is designated PRISM-12. • If performed during the first 24 hours in the ICU, it is designated PRISM-24. Predictive equations: • Predictive equations for prognosis are available for the 12 hour and 24 hour scores

PRISM III is a widely accepted and is a standard against which other scores are compared. However there some problems with the use of PRISM III: - A lot of information is needed to calculate it and many units do not calculate it routinely. Worst reading of 12/ 24 hours is used and a lot of deaths occur (in one study over 40%) with in first 24 hours, so the score may be diagnosing death rather predicting it. There may be blurring of differences of 2 units as patient in a good unit may recover rapidly and score may be lower and the same patient in a bad unit might have had higher score due to poor management and high mortality of bad unit may be interpreted as due to sicker patients. The time spent in the hospital before coming to ICU could improve the PRISM score and predict lower than actual mortality ( lead time bias).17 Uses of models of mortality prediction including PRISM III: - These models including PRISM III are most applicable to groups of patients (e.g. to assess institutional performance). These models help us to investigate best ways of organizing PICU by comparing different units.18 They also help us to monitor effect of change in practice by observing trends within the unit over a time.17 They can also be used for controlling severity of illness for various clinical trials.16 They can be applied for resource utilization (rationing intensive care). PRISM III takes 24 hours to complete and can’t be used in regulating admission to PICU. 19 They have been used to assess relation between severity of illness and length of stay or cost. In 1997 MM Pollack et at developed a physiology based measure of physiologic instability for use in pediatric patients that has an expanded scale compared with the prism III score and called as the Pediatric risk of mortality Ill-acute physiology score (prism

Ill-APS) . It has 59 ranges of 21 physiologic variables . It was designed to have a broad severity scale from 0356, with the higher values indicating higher instability. Data were collected from consecutive admissions to 32 Pediatric ICU’s (11165 admissions, 543 deaths). 20 Most patients who had PRISM Ill-APS score of greater than 80 had mortality greater than 97%. It concluded that the PRISM Ill-APS score is an expanded measure of physiologic instability that has been validated against mortality. Compared with prism III, prism Ill-APS should be more sensitive to small changes in physiologic status even those that may not contribute significantly to mortality risk. Patient assessment for future studies for issues e.g. effectiveness of drugs or for other purposes might be more concerned with the physiologic status. However even this should not be used for quality assessments or calculating risk of individual patients. 20

Intensive care uses of scoring systems Mortality risk assessment Mortality risk assessment for the general population of intensive care patients is based on the observation that the mortality risk is proportionate to the number of failed organ system. 21, 22 The mortality risks for PICU patients with one, two, three, and four or more organ failures are approximately 1 %, 10 %, 50%, and 75 %, respectively.23 The major pediatric mortality risk assessment method applicable to the wide variety of PICU patients is the PRISM score which is actually a revision of the PSI. Physiological variables and their ranges were objectively evaluated, resulting in the PRISM score. The resulting PRISM score had only 14 variables and 23 ranges of these variables. 14 The database used for this effort in672

Poonam Bhadoria. Severity scoring systems in paediatric ICU

resources in terms of bed days and diagnostic tests than survivors. In one study, deaths stayed approximately twice as long as survivors, used more than twice as many diagnostic tests per day and their diagnostic tests cost more than twice as much as survivors.26

cluded over l0000 PICU patients from over 32 PICUs and is already described above.

Outcome-based quality assurance Before these scores were introduced, performance evaluations of ICUs using crude mortality rates were extremely difficult because severity of illness and diagnostic distributions varied among ICUs. The rigorous multicentered statistical validation used for the validation of the scores can be readily used in single ICUs or regions to investigate quality issues of local or national interest. The most important performance criteria of ICUs—outcome in terms of survival and death— can be assessed.

Currently, a PICU methodology is available. 26 The method using PRISM to adjust for severity of illness and diagnoses is able to compare expected length of stay (LOS) to actual length of stay (LOS ratio). It is found 20 % of PICUs had LOS ratios lower than expected and 20% had LOS ratios higher than expected.

Predicting death and disability: new frontiers

The constancy of the relationship between physiological profiles and outcome is the backbone of the use of these predictors for quality assurance purposes. If the observed number and distribution of outcomes are similar to the predicted number and distribution of outcomes, then the performance of the institution is equivalent to those institutions validating the predictor in the multi- institutional studies. As with any test, physicians using this quality assurance methodology need to understand its strengths and limitation, when falsepositive and false-negative results can occur, confounding variables, and peculiarities of the method.

A more controversial aspect of outcome prediction is the identification of patients “too sick to benefit” from PICU care. Advances are being made that may in the future result in clinically relevant data. Problems must be recognized and overcome before this effort is suitable for clinical use. The most important problem in identifying patients “too sick to benefit” is the relative inability to achieve sufficient certainty in the prediction of death. Hence more innovative approaches must be sought to improve the prognostic accuracy of assessing mortality risk or assessing the risks of morbidity. One effort to improve prognostic accuracy has been to use the dynamic nature of disease and recovery to improve the PRISM mortality prediction model. Recently, Ruttimann et al expanded this to include a time-series approach to outcome prediction using PRISM scores. 27 Time-series analysis uses past physiological events measured by daily PRISM scores to predict the next day’s PRISM score. Deviations of the observed from the predicted PRISM scores can be used to predict outcome. The usefulness of time-series analysis is the specificity of prediction models at acceptable sensitivities. Unfortunately, there are significant obstacles to be overcome in reliably predicting disability in infants and children First, the measurement of functional status for normally dependent people (e.g. infants and young children) is difficult. When functional dependency is a normal state, differentiating normal from abnormal depen-

However, this method is not designed to replace other quality assurance tasks such as evaluations of nosocomial infections, spontaneous extubation, and so forth. However, it provides a more objective and rigorous review than traditional morbidity and mortality conferences.

Cost containment Clinical scores can be used to conduct economic evaluations and cost-containment interventions in PICUs. These evaluations are of tremendous value because PICUs are both costly and over-utilized. ICU patients use laboratory tests four times more frequently than routine care patients. 24 Twenty percent of hospital charges are generated by the ICU. 25 Deaths use more 673

Indian Journal of Anaesthesia, October 2008(P.G.Issue)

dency is sometimes difficult without time-consuming methods. Second, there are currently no easy, quick and most important, reliable classification criteria for children. Third, long term outcomes may be different from short term outcomes. For example, children who appear dysfunctional on sophisticated neuropsychological testing after discharge from NICUs often achieve much greater function than anticipated.

Use in other areas of research A clear and valuable use for clinical scores is to control for severity of illness in control and experimental populations. When used in this manner, especially if both samples are from the same institution, many of the problems of precisely defining outcome risks are less significant. Investigators should be sure, however, that the clinical score is applicable to the situation. For example, using the TISS score to control for physiological status would be a misuse of the TISS score.

Disease - specific scores There are many disease-specific or condition-specific clinical scores to aid in the evaluation of severity of illness, prognosis, pathophysiological understanding, or therapeutic needs. These are too numerous to evaluate individually. Special concerns must always be raised concerning the applicability of a disease-specific or condition-specific score to a different institution because the relationship between the variables and the outcome may be biased by local conditions. Most importantly they must be validated prospectively on different populations.

Investigators have also used clinical scores as an outcome variable. For example, the Croup score has been used to quantify the clinical symptoms of airway disease and assess therapy. While this is a legitimate use of the score, investigators choose the clinical score that reliably measures the appropriate outcome. Therefore, the score should be appropriately designed and validated. Investigators have occasionally attempted to compare the performance of a population to an expected performance predicted by a clinical score. For example, the outcome of respiratory failure using a new ventilator may be compared with the expected outcome of computed with a clinical score. In this way, the clinical score is being used as a historic control. It is, however, subject to the same problems as historic controls. The relationship between the variables and the outcomes may have changed or may be distorted by institutional factors or population bias. When possible, control and experimental populations ensure a more powerful study.

Tracking ICU therapies Perhaps the most well-known and most commonly used score applicable to the general critical care population is TISS. 8 Initially proposed in 1974, TISS was the first modern-day, quantitative effort directed at measuring severity of illness. TISS has also been applied to pediatric intensive care. The TISS score quantifies 76 therapeutic and monitoring interventions by assigning points from 1 to 4 based on complexity and invasiveness. TISS points are associated with outcome because most diseases follow common pathophysiological pathways, which are symptomatically treated. For example, as multiple organ failure develops, increased monitoring and therapeutic modalities are instituted, increasing the TISS score. Therefore TISS scores correlate with mortality risk. Unfortunately, compared with other predictors, it cannot quantify mortality risk. Efforts to evolve the TISS score by combining physiological dysfunction with therapies has been relatively unsuccessful.10

So presently PRISM III and PIM are the best and latest scores available for PICU mortality prediction but in the years to come surely better scores predicting mortality prediction more accurately for use in individual patients as well as for many other uses discussed already, will be derived.

References 1.

674

Gomke RJBJ, Bonsel GJ, Johannes Van Vught A. Effectiveness and efficiency of a Dutch Pediatriac intensive Care unit: Validity and application of the Pediatric Risk of mortality score. Critical Care Med 1994 ; 2:1477-1484.

Poonam Bhadoria. Severity scoring systems in paediatric ICU 2.

Pollack MM, Ruttiman UE, Getson PR, et al. Accurate Prediction of the outcome of Pediatric intensive care. The New England Journal of Med 1987 ;316: 134-139.

16: 1110-1116. 15.

Balakrishnan G, Tom Aitchison, David Hallworth, Morton NS. Prospective evaluation of Pediatric Risk of Mortality (PRISM) score. Arch Dis Child 1992; 67: 196200.

3.

Pollack MM, Ruttimann UE, Glass NL. Monitoring patients in pediatric intensive care. Pediatric 1985; 76:71924.

16.

4.

Yeh TS, Pollack MM, Ruttimann UE, et al. Validation of a physiologic stability index for use in critically ill infants and children. Pediatric Research 1984; 18:445-452.

Pollack MM, Kantilal M Patel, Urs E Ruttiman. PRISM III: An updated pediatric risk of mortality Score. Crit Care Med 1996; 24: 743-752.

17.

5.

Lemeshow S, Teres D, Kiar J, et al. Mortality probability models (MPM II) based on international cohort of intensive care unit patients. JAMA 1993; 270: 2478-2480.

Shann F, Pearson G, et al. Paediatric index of mortality (PIM). Intensive care Med 1997; 23:201-207.

18.

Knaus WA, Wagner DP, Draper EA, et al. The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults. Chest 1992; 102:1919-20.

Pollock MM, Kantilal M. Patel, et al. Frequency of variable measurement in 16 pediatric intensive care unit. Crit Care Med 1996; 24:74-77.

19.

Teres D, Lemeshow S. Why severity models should be used with caution. Crit Care Clin 1994; 10:93-110.

20.

Pollack MM, Patel and Ruttiman. The PRISM III- Acute Physiology score (PRISM III- APS): A Method of assessing physiologic instability for PICU patients. The Journal of Pediatrics 1997; 131: 575-581.

6.

7.

Bertolini G., Donata RAC, Apolone G. PRISM- An assessment of its performance in a sample of 26 italian ICU’S. Crit Care Med 1988; 26:1427-1432.

8.

Keene AR, Culten Di. Therapeutic intervention scoring system. Update. Crit Care Med 1983; 11:1-3.

21.

Baue AE, Chaudry IH: Prevention of multiple system failure. Surg Clin North Am 1980;60:1167-1178.

9.

Pollack MM: Prognostics - state of art. In Cerra FB. Shoemaker WC, editors: Critical care: state of art, Fullerton,CA, 1987, Society of Critical Care Medicine.

22.

Eiseman B, Beart R, Norton L: Multiple organ failure. Surg Gynecol Obstet 1977; 144:323-326.

10.

Cullen DJ, Keene R. Waternaux C, et al. Objective. Quantitative measurement of severity of illness in critically ill patients. Crit Care Med 1984;12:155.

23.

Wilkinson JD, Pollack MM, Glass NL, Kanter RK, Katz RW, Steinhart CM. Mortality associated with multiple organ system failure and sepsis in pediatric intensive care unit. J Pediatr 1987; 111:324–328.

11.

Knaus WA, Wagner DP, Draper EA, et al. The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults. Chest 1991; 100:1619-1636.

24.

Winkel P, Staland BE. Assessing cost savings when unnecessary utilization of laboratory tests can be abolished. Am J Clin Pathol 1984 ; 82:418-23.

12.

Knaus WA, Wagner DP, Zimmerman JE, et al. Variations in mortality and length of stay in intensive care units. Ann Intern Med 1993; 118:753-761.

25.

The Role of Technology Assessment in Cost Control, in Barbara J. McNeil and Ernest G. Cravalho, eds., Critical Issues in Medical Technology, Auburn House Publishing Co., 1981, 129-138.

13.

Pollack MM, Timothy S. Yeh, Ruttiman UE, P.R. Holbrook, Alan I. fields. Evaluation of Pediatric intensive Care. Crit Care Med 1984 ; 12: 376-383.

26.

Klem SA, Pollack MM, Getson PR. Cost, resource utilization, and severity of illness in intensive care. J Pediatr 1990; 116:231-237.

14.

Pollack MM, Ruttiman UE, Getson P R. The Pediatric Risk of mortality (PRISM) SCORE. Crit Care Med 1988;

27.

Ruttimann UE, Pollack MM. A time-series approach to predict outcome from pediatric intensive care. Comput. Biomed Res 1993;26:353–372.

675