Indian J Med Res 119, June 2004, pp 273-282

Performance evaluation of APACHE II score for an Indian patient

with respiratory problems

Rajnish Gupta & V.K. Arora

LRS Institute of Tuberculosis & Respiratory Diseases, New Delhi, India

Received July 25, 2003

Background & objectives: Realising the utility of scoring systems in mortality prediction of critically ill
patients admitted to intensive care units (ICUs), studies worldwide have expressed a need to validate
the Acute Physiology and Chronic Health Evaluation (APACHE) II score for databases of respective
countries. Literature available in this area in the Indian context is scanty. The present study was
undertaken to evaluate the performance of APACHE II score in prediction of mortality risk, as well as
in determination of model validity in critically ill Indian patients with respiratory problems.
Methods: The study was prospectively carried out over 18 months at respiratory ICU of a tertiary
Institute in New Delhi, which admitted consecutive medical (with lung ailments) and surgical (who
had undergone any elective thoracic surgical procedure under general anaesthesia) patients. Based
on chief indication of ICU admission, the medical patients were further divided into sub-groups I
(respiratory) and II (non-respiratory). APACHE II points were assigned to all patients for calculating
their individual predicted risks of mortality. Standard mortality ratio (SMR) was computed with 95
per cent confidence intervals (CI). Calibration of model was analysed by calculating Lemeshow and
Hosmer goodness of fit c2 statistic and by plotting calibration curve, whereas discrimination was
evaluated by calculating area under a receiver operating characteristic (ROC) curve.
Results: Of the 393 consecutive patients admitted to respiratory ICU during the study period, 63 were
left out on account of exclusion criteria. Mean APACHE II score of the remaining 330 patients was
12.87±8.25 and range from 1 to 47. There were 287 (87%) survivors and 43 (13%) non-survivors, whose
mean APACHE II scores, being respectively 11.34±6.75 (range 1-37) and 23.09±10.01 (range 5-47), were
significantly different (P<0.01). The study had a predicted mortality of 7.9 per cent and an SMR value
of 1.65 (95% CI from 0.4 to 3.0). Mean APACHE II score of those having medical ailments was significantly
higher (P<0.01) than surgical patients. The non-respiratory sub-group had a significantly higher
(P<0.01) mean APACHE II score than respiratory sub-group. 59 per cent of patients did not get APACHE
II points owing to being <45 yr of age. In addition, against 10 immunocompromised patients, 77 others
did not get APACHE II points despite having apparently compromised immunity due to co-existence of
tuberculosis (TB), diabetes mellitus, dual pathologies or past history of anti-TB treatment. Observed
and predicted mortality rose with 5-point APACHE II score, but did not correlate for patients of any
comparable group. Average ICU stay of 16 days for those with medical disease was significantly longer
(P<0.01) than 9.5 days for surgical patients. APACHE II scoring system showed a poor calibration and
discrimination ability for Indian respiratory patients.
Interpretation & conclusion: Despite the rise in observed and predicted mortality with 5-point APACHE
II score, predicted mortality did not correlate with observed mortality for critically ill patients admitted
to an Indian respiratory ICU. The scoring system also showed a poor calibration as well as discrimination.
The model may be more useful for Indian patients by lowering down the cut-off value in allotment of
age points and by awarding the weightage to factor like co-existing immunocompromised state.
Key words Acute physiology and chronic health evaluation (APACHE) II score - intensive care unit - mortality -
standard mortality ratio (SMR) - respiratory problems
273
274 INDIAN J MED RES, JUNE 2004
Respiratory diseases have been responsible for a
considerable morbidity and mortality in intensive care
units (ICU's). Over the past few decades, several scoring
models like Acute Physiology and Chronic Health
Evaluation (APACHE), Simplified Acute Physiology
Score (SAPS), Mortality Prediction Model (MPM) etc.,
have been developed for predicting outcome of admitted
patients. These have undergone global trials for
determining validity on databases different from originally
studied populations and variable results have been
reported.
APACHE II scoring system1 has been often used to
predict risk of ICU mortality. Using the system, a good
correlation was found between predicted and observed
outcomes in studies from US1, Canada2, New Zealand3
and Singapore4,5. On other hand, APACHE II equation
did not fit British and Irish data6,7. A disproportion in
observed and predicted hospital mortality in individual
ICUs was also reported from Brazil8, Denmark9, Saudi
Arabia10, Cuba11 and India12. Despite non-correlation
with mortality, studies have acknowledged utility of
APACHE II score in selection of triage patients for ICU
admission13, and in provision of optimal management of
small bowel perforations14. Utilisation of score has been
suggested to result in management decisions that could
salvage costly ICU resources scantily available in
developing world. Thus, there is a need, not only to
provide patients with quality care but also utilise available
resources optimally, for example, through usage of
prediction model like APACHE II. However, Indian
literature on this subject is scarcely available. The present
study was, therefore, designed to evaluate performance
of APACHE II score in prediction of mortality risk, as
well as in determination of model validity in critically ill
patients with respiratory problems, in order to develop
or suggest a development methodology for appropriate
prediction model effecting an optimal utilisation of ICU
resources.
Material & Methods
The study was prospectively carried out over 18
months between September 2001 to February 2003 at
respiratory ICU of L.R.S. Institute of Tuberculosis and
Respiratory Diseases, a tertiary Institute based in New
Delhi. The study unit was supposed to admit those
consecutive patients with lung ailments, who either came
from the Institute's chest clinic and chest ward or were
directly shifted from an outside hospital. These were
divided into sub-group I, comprising of those respiratory
or non-respiratory patients whose chief indication of ICU
admission was respiratory for either category, and sub-
group II, comprising of those respiratory patients who
were admitted with a chief non-respiratory indication.
Presence of any past history of chronic illness was
recorded for all admitted patients. The unit also received
consecutive post-operative patients who had undergone
any thoracic surgical procedure under general
anaesthesia (e.g., pneumonectomy, lobectomy, de-
cortication etc.) after having been found pre-operatively
fit for the same on the basis of a pulmonary function test
(PFT) and other investigations deemed necessary.
Patients found unfit on account of low PFT values for a
particular type of surgery were rejected for operative
procedure. The study protocol was approved by the
ethics committee of the Institute.
Enrolled patients under study were screened with
respect to their demographic profile (age and sex),
presence of chronic disease, past history of hospitalisation
and ICU admission, surgical status (elective or
emergency surgery), major reason for ICU admission
(i.e., predominant diagnostic category) and severity of
illness (acute physiologic state)1. Initial and worst values
were taken during the patient's first 24 h of ICU
admission in respect of 12 variables constituting the acute
physiology score (A). However, points were allocated
to the worst values as per protocol of APACHE II
scoring system1. Age (B) and chronic disease (C) were
also assigned points in similar manner. Sum of A, B and
C constituted APACHE II score for a patient, derivation
of which facilitated the subsequent calculation of
predicted risk of mortality and standard mortality ratio
(SMR)1. Mortality comparisons were made between age
categories, males and females, surgical and non-surgical
patients, those having and not having a chronic disease/
history of ICU admission, and survivors and non-
survivors. Results were subjected to analysis for
determining a statistical significance. All mean values
were compared by 't' test and proportions by 'z' test.
The overall goodness of fit of APACHE II scoring system
for patients with respiratory problems was determined
from calibration curve, classification tables, analysis of
receiver operating characteristic (ROC) curve15 and a
c2 statistic proposed by Lemeshow and Hosmer16. 95
per cent confidence intervals (CI) were calculated for
the true positive (predicted to die who died), false positive
 GUPTA & ARORA : APACHE II IN INDIAN PATIENTS
(predicted to die who lived) and overall correct
classification (predicted to live who lived plus predicted
to die who died) rate, in addition to SMR.
Patients <16 yr of age, having ICU stay of <8 h,
admitted due to administrative reason or re-transferred
to ICU during the same hospital stay were excluded from
the study. Every recurrent admission with a defined
outcome of last hospitalisation (e.g., discharge) was
recorded as a separate ICU entry.
Results
Overall 393 consecutive patients were admitted to
the respiratory ICU during the 18 month study period.
Of these, 63 were excluded on account of their age being
<16 yr (18), ICU stay of <8 h (14), admission due to
administrative reason (21) and a re-transfer to ICU
during the same hospital stay (10). Demographic profile
(age and sex) and past history of chronic illness,
hospitalisation or ICU admission of the remaining 330
patients was recorded for each. Mean age of enrolled
patients was 43.32 yr with a standard deviation (SD) of
±16.22 yr. Number of males was significantly more
(P<0.01) than females (Table I). A past history of chronic
illness was present in half the number of patients.
Three fourth of the medical patients had a respiratory
disease (sub-group I) as chief indication of ICU
admission, whereas remaining had a non-respiratory
indication (sub-group II) (Table II). Surgery was
electively performed in all operated thoracic surgical
patients.
Tables I and II provide patient data in relation to their
APACHE II scores, observed deaths, predicted mortality
and standard mortality ratios. The overall mean
APACHE II score was 12.87±8.25 with a range from 1
to 47; 282 (86%) patients had scores <20. There were
287 (87%) survivors and 43 (13%) non-survivors, whose
mean APACHE II scores were respectively 11.34±6.75
(range 1-37) and 23.09±10.01 (range 5-47) with a
significant difference (P<0.01) between them. Observed
mortality (13%) was not significantly higher than
predicted (7.9%) and both were generally seen to
increase with every 5-point rise of APACHE II score
(Fig.1) Overall SMR of study was 1.65 with 95 per cent
CI ranging from 0.4 to 3.0. No survivor was noticed
above the score of 40.
275
Mean APACHE II score was found to increase
progressively with age (Table I). Patients above 65 yr
of age not only had a significantly higher (P<0.01) mean
APACHE II score (19.90±8.13) than of younger age
(12.48±7.00), but also had a higher observed (P<0.01)
and predicted (P>0.05) mortality. APACHE II under-
estimated the mortality risk for the former as compared
to an under or over-estimation for the latter. No
correlation existed between the observed and predicted
mortality for either sex, as well as for those with/without
a past history of chronic illness, hospitalisation or ICU
admission.
Sepsis and post-cardiac arrest (Table II) were
individual disease categories (non-respiratory),
associated with high mean APACHE II scores and
hospital mortality. While a correlation between the
observed and predicted mortality existed for these
categories, it lacked for others within medical or
surgical group.
The medical patient group comprised 10 immuno-
compromised cases, who got weightage of APACHE
II. Of these, 2 fell under 'chronic illness' category1 and
got 5 chronic health points, whereas other 8, having
lung neoplasm, got a coefficient for admitting 'diagnostic
category'. There were 77 (36%) others, who apparently
had a compromised immunity but were not awarded
any extra weightage apart from that permissible for
their admission. Of these, 24 (11.2%) had pulmonary
(10) or extra-pulmonary (14) TB, 39 (18.2%) had an
anti-TB treatment history, 9 (4.2%) had diabetes
mellitus and 5 (2.3%) had dual association of two
pathologies.
Fig.1. Relationship of a 5-point rise of APACHE II score with
12
mortality ( for observed and12 for predicted) and number of patients
 276
Table I. Case distribution according to demography and past history

Parameters No. (%) Observed deaths APACHE II Predicted deaths Standard mortality ratio
of cases and mortality (%) and mortality (%) (95% CI)
(n=330) (n=43) Mean score±SD (Range) (n=26)
Age-group (yr):
16-25 51 (15.5) 1 (2.0) 6.84±6.30 (1-37) 2 (3.9) 0.50 (0.4-0.6)
26-35 80 (24.2) 8 (10.0) 9.06±6.49 (2-36) 5 (6.3) 1.60 (0.6-3.6)
36-45 63 (19.1) 7 (11.1) 12.8±7.36 (3-37) 4 (6.4) 1.75 (1.2-4.8)
46-55 54 (16.4) 15 (27.8) 15.8±8.84 (3-42) 6 (11.1) 2.50 (1.5-6.5)
56-65 57 (17.3) 2 (3.5) 17.9±5.99 (6-31) 6 (10.5) 0.33 (0.2-0.5)
66-75 15 (4.5) 6 (40.0) 18.7±6.32 (10-37) 1 (6.7) 6.00 (5.0-18.0)
76-85 10 (3.0) 4 (40.0) 21.1±9.94 (11-47) 2 (20) 2.00 (0.1-11.0)
Sex:
Males 229 (69.4) 35 (15.3)* 13.47±8.3 (1-42) 20 (8.7) 1.75 (0.1-3.5)
Females 101 (30.6) 8 (7.9) 11.53±8.1 (3-47) 6 (5.9) 1.33 (0.6-3.2)
History of chronic disease:
Present 168 (50.9) 30 (17.9)† 17.82±7.28† (6-47) 22 (13.1)† 1.36 (0.4-3.2)
Absent 162 (49.1) 13 (8.0) 7.74±5.68 (1-37) 4 (2.5) 3.25 (0.6-6.0)
Past hospitalisation:
Present 51 (15.5) 7 (13.7) 15.24±6.99 (3-31) 5 (9.8) 1.4 (0.6-2.8)
Absent 279 (84.5) 36 (12.9) 12.44±8.41 (1-47) 21 (7.5) 1.71 (0.2-3.2)
Past ICU admission:
INDIAN J MED RES, JUNE 2004

Present 41 (12.4) 7 (17.1) 16.03±6.9 (4-31) 4 (9.8) 1.75 (1.2-4.8)

Absent 289 (87.6) 36 (12.5) 12.42±8.34 (1-47) 22 (7.6) 1.64 (0.2-3.0)

* Significantly higher (P<0.05) than females;

†
Significantly higher (P<0.05) than patients lacking a history of chronic disease
APACHE II, acute physiology and chronic health evaluation II; ICU, Intensive care unit; CI, confidence interval
 Table II. Case distribution according to category of disease

Chief indication of admission No. (%) of Observed deaths APACHE II Predicted deaths Standard mortality
cases and mortality (%) and mortality (%) ratio (95% CI)
(n=330) (n=43) Mean score±SD (Range) (n=26)

Medical 214 (64.8)* 40 (18.7)* 16.61±7.85* (2-44) 26 (12.2) 1.54 (0-3.0)

Sub-group I (respiratory) 161 (75.2) 21 (13) 15.37±6.26 (2-42) 11 (6.8) 1.91 (0-3.8)
COPD 81 (37.9) 12 (14.8) 17.83±5.48 (8-42) 5 (6.2) 2.40 (0.4-4.4)
Bronchial asthma 11 (5.1) 0 10.64±2.29 (8-15) 0 0
Infection 19 (8.9) 2 (10.5) 13.37±6.27 (4-27) 1 (5.3) 2.00 (1.0-5.0)
Neoplasm 8 (3.7) 2 (25.0) 18.88±6.71 (10-31) 3 (37.5) 0.67 (0.3-1.0)
Embolism 2 (0.9) 0 5.5±2.12 (4-7) 0 0
ARDS 2 (0.9) 1 (50.0) 13.5±4.95 (10-17) 0 ¥
Others 38 (17.8) 4 (10.5) 12.4±5.93 (2-27) 2 (5.3) 2.00 (1.0-5.0)
Sub-group II (non-respiratory) 53 (24.8) 19 (35.9)† 20.36±10.64† (3-47) 15 (28.3)† 1.27 (0.7-3.3)
Congestive heart failure 9 (4.2) 1 (11.1) 13.11±4.81 (6-23) 0 ¥
Sepsis 32 (15.0) 15 (46.9) 24.09±8.4 (11-38) 13 (40.6) 1.15 (0.1-3.2)
Post cardiac arrest 3 (1.4) 2 (66.7) 34.33±14.19 (19-47) 2 (66.7) 1.00 (0.9-1.1)
Other cardiovascular diseases 2 (0.9) 0 4.0±1.41 (3-5) 0 0
Seizure disorder 1 (0.5) 0 5.0±0 (5) 0 0
Other neurological diseases 3 (1.4) 1 (33.3) 7.67±8.08 (3-17) 0 ¥
Diabetic keto-acidosis 1 (0.5) 0 10.0±0 (10) 0 0
Metabolic disorders 2 (0.9) 0 20.5±10.61 (13-28) 0 0
Surgical (Thoracic) 116 (35.2) 3 (2.6) 5.98±2.62 (1-17) 0 ¥
Pneumonectomy 14 (12.1) 0 7.93±3.2 (3-14) 0 0
Lobectomy 16 (13.8) 0 7.19±3.08 (4-17) 0 0
GUPTA & ARORA : APACHE II IN INDIAN PATIENTS

Resection of mass 1 (0.9) 0 6.0±0 (6) 0 0

Decortication 19 (16.3) 0 5.84±2.39 (3-13) 0 ¥
Thoracoplasty 37 (31.9) 2 (5.4) 5.27±1.71 (3-11) 0 0
Bullectomy 2 (1.7) 1 (50) 5.0±1.41 (4-6) 0 0
Others 27 (23.3) 0 5.33±2.89 (1-15) 0 0

* Significantly higher (P<0.01) than surgical group

†
Significantly higher than Respiratory sub-group (P<0.01) for observed mortality and (P<.05) for mean APACHE II Score and Predicted mortality
APACHE II, Acute physiology and chronic health evaluation II; COPD, chronic obstructive pulmonary disease;
ARDS, adult respiratory distress syndrome; CI, confidence interval
277
278 INDIAN J MED RES, JUNE 2004

Calibration (ability to predict mortality) of APACHE
II system was assessed using Lemeshow Hosmer
goodness of fit test (Table III, Fig.2). Except for ends,
calibration was poor for major part of spectrum. Chi
square value was 10.34, which was not significant
(P>0.05).
Observed and predicted outcomes were compared
at 5 per cent intervals of decision criteria ranging from 5
to 95 per cent. Classification tables were prepared for
10, 25, 50, 75 and 90 percentages of decision criteria
(Table IV). Fig. 3 shows an ROC curve obtained by
plotting the true positive (sensitivity) and false positive
rates (1-specificity) at each decision criterion. The best
overall correct classification rate of 89.7 per cent was
obtained at 70 per cent of decision criterion. The area
under curve (AUC) was .63 indicating a poor
discrimination (ability to distinguish survivors from non-
survivors) of APACHE II scoring system when compared
to the value of >0.8 required for acceptability.
Whereas average ICU stay for all patients was 11.3
days, patients in medical group stayed for 16 days that
was significantly longer (P<0.01) than 9.5 days for
surgical patients group. Patients having a pre-existing
chronic disease or history of past ICU admission also
spent a significantly longer (P<0.01) time in ICU (average
of 12.9 and 14.9 days respectively) as compared to those
not having the either (9.6 and 10.7 days respectively).
Overall, mean hospital stay was significantly higher
(P<0.01) for patients with surgical than medical problems
(92.1 vs 31.9 days respectively). Of the 330 patients, 43
(13%) died, 227 (84%) patients were discharged from
the hospital, while 6 (2%) left against medical advice
(LAMA) and remaining 4 (1%) got referred to a general
medical hospital for seeking services related to other
specialities. Vital status of both LAMA and referred out
patients favoured survival.
Discussion
Our study showed a mean APACHE II score of
12.87±8.25, which was comparable to that reported from
Israel13, Singapore17 and India18. A significantly higher
mean APACHE II study score observed in non-survivors
compared to survivors was in accordance with reports
from Saudi Arabia10,19, Singapore4 and Israel20. Since
every 5-point rise of score led to the increase in both
observed and predicted mortality, association of
APACHE II and risk of dying in hospital, as reported
earlier1,6,7,21 got re-affirmed from the present study. Non-
availability of survivor above score of 40 was also in
consistence with the past observations1,4.
Though under-prediction of mortality for age group
above 76 yr has been reported with scoring system by

Table III. Lemeshow-Hosmer goodness of fit of APACHE II for

Indian patients with respiratory problems

Predicted risk of No. of No. of deaths No. of survivors

mortality (%) cases Observed Expected Observed Expected
Predicted risk of death (%)

0- 172 4 6.28 168 165.72

10- 58 7 8.58 51 49.42
20- 40 8 10.19 32 29.81
30- 20 4 7.18 16 12.82
40- 14 4 6.41 10 7.59
50- 4 1 2.25 3 1.75
60- 7 3 4.58 4 2.42
70- 4 2 2.98 2 1.02
80- 8 7 7.01 1 0.99

Predicted risk of death (%)

90-100 3 3 2.80 0 0.20

Fig.2. Relationship between observed and APACHE II predicted Sum of c2=10.34, df=8
mortality for Indian respiratory patients.
 GUPTA & ARORA : APACHE II IN INDIAN PATIENTS 279

Table IV. Classification Tables for APACHE II applied to Indian patients with respiratory problems

No. predicted to live No. predicted to die Total

Decision criterion of 10% :

No. of observed survivors 168 119 287
No. of observed dead 4 39 43
Total 172 158 330
True positive rate (95% CI) 90.7 (82.1 to 99.3)
False positive rate (95% CI) (%) 41.5 (35.8 to 47.2)
Overall correct classification (95% CI) (%) 62.7 (57.5 to 67.9)
Decision criterion of 25% :
No. of observed survivors 233 54 287
No. of observed dead 17 26 43
Total 250 80 330
True positive rate (95% CI) (%) 60.5 (45.8 to 75.2)
False positive rate (95% CI) 18.8 (14.3 to 23.3)
Overall correct classification (95% CI) (%) 78.5 (74.1 to 82.9)
Decision criterion of 50% :
No. of observed survivors 277 10 287
No. of observed dead 27 16 43
Total 304 26 330
True positive rate (95% CI) (%) 37.2 (22.7 to 51.7)
False positive rate (95% CI) 3.5 (1.3 to 5.7)
Overall correct classification (95% CI) (%) 88.8 (85.4 to 92.2)
Decision criterion of 75% :
No. of observed survivors 285 2 287
No. of observed dead 33 10 43
Total 318 12 330
True positive rate (95% CI) (%) 23.3 (10.8 to 35.8)
False positive rate (95% CI) (%) 0.7 (0 to 1.4)
Overall correct classification (95% CI) (%) 89.4 (86.0 to 92.8)
Decision criterion of 90 % :
No. of observed survivors 287 0 287
No. of observed dead 40 3 43
Total 327 3 330
True positive rate (95% CI) (%) 7.0 (0 to 14.0)
False positive rate (95% CI) (%) 0.0 (0)
Overall correct classification (95% CI) (%) 87.9 (84.2 to 91.6)

several workers7,22 and was also noticed in the present
study, the observed and predicted mortality had no
correlation with any age group, sex and presence of past
histories of chronic disease, hospitalisation or ICU
admission. However, patients with chronic disease, as
reported earlier6,10,12 in comparison to those without, had
a significantly higher risk of hospital mortality. Such a
risk was not seen for those with a past history of ICU
admission.
The significantly low mean APACHE II score found
in thoracic surgical patients was associated with low
hospital mortality observed within group as compared to
medical patients. Available reports also indicate a lower
hospital mortality for surgical than medical
patients2,6,7,18,23, and within group for those undergoing a
thoracic surgical procedure than on any other organ
system6,7, and for those undergoing an elective than
emergency operative intervention6. Since type of surgery
280 INDIAN J MED RES, JUNE 2004
performed in the present study was entirely thoracic,
whose nature was not only elective but required a
mandatory pre-operative procedural fitness, risk of post-
operative mortality was greatly reduced for surgical
patients. Provision of immediate post-operative care may
also be a contributory factor. A lower hospital mortality
observed for respiratory system, as compared to most
other organ systems, was in agreement with earlier
reports6,7,23.
The chi square value for calibration of APACHE II
was not statistically significant suggesting a good
calibration of APACHE II for Indian patients with
respiratory problems. However, good fits have been
reported with small sample sizes24. Such possibility
appears to exist in study as calibration curve showed a
poor calibration of model. Similarly, discrimination of
system was also poor. The results were in consonance
with recent observations from sub-continent18.
We observed an SMR value of 1.65, which was higher
than that reported from many developed nations2,7,23 but
was lower than that observed in most developing
countries8,12,17. Since SMR is a ratio of observed to
predicted mortality, a higher value could result either from
increased observation or from decreased prediction of
mortality. Several factors prevalent in our country need
consideration. Firstly, greater number of people above
Fig.3. ROC curve showing APACHE II application on Indian
respiratory patients. Points A, B, C, D and E correspond to decision
criteria of 10, 25, 50, 75 and 90% respectively for predicted risk of
mortality. The diagonal line indicates an index that operates no better
than chance.
65 yr of age use ICU services in West as compared to
India, where life expectancy at birth is shorter and
families are reluctant to spend money on elderly12. Since
APACHE II awards points to only those ³45 yr of age1,
great majority (59%) of our ICU patients below cut-off
value were left out that would have resulted in an under-
prediction of mortality. Secondly, various communicable
diseases like TB, having a greater prevalence within
country and remaining unaccounted in original APACHE
II, comprise a different case-mix that would contribute
to higher SMR value observed in this part of world. In
addition, as compared to ICUs in developed countries,
overall quality of intensive care in India suffers from
financial constraints, a shortage of trained nursing staff
and a higher workload per nurse12 that potentially
increases observed mortality. Moreover, higher lead time
bias between onset of illness and ICU admission may
also be contributory either due to requisite patient
transportation from distant places to ICUs located in
urban set ups or due to weighing of expenses in minds
of care takers or due to attempted medical interventions
by skilled/unskilled treatment providers or small nursing
homes25 that only delays the delivery of intensive care
inadvertently. A high lead time is likely to be associated
with greater hospital mortality. Further, no factorial
adjustment exists in original APACHE II system, which
would tend to result in the under-estimation of hospital
mortality. The present study showed the evidence of
association with most of these factors. Our nurse-to-
patient ratio of 1:4.7 was more than double of the
American recommendation of 1:2 or less26 that would
suggest a higher work load for nurses, and a near total
bed occupancy of ICU patentially affecting the output
quality of delivered intensive care.
Overall mean ICU stay of 11.3 days for our patients
was higher than that reported in other studies from Brazil
(9.4 days)8 and USA (5 days)27, presumably due to the
fact that majority of enrolled incumbents had a medical
disease of advanced nature. Further, thoracic surgical
patients stayed significantly longer in hospital as compared
to their medical counterparts because a large number,
especially those needing thoracoplasty or de-cortication
had pre-operative admission period varying from few
months to a year, in addition, to post-operative hospital
stay. Pre-operative therapeutic measures, comprising
inter-costal tube drainage procedures, antibiotics, anti-
tubercular treatment etc., were tried for patients of
empyemas or pyo-pneumothoraces to achieve lung
 GUPTA & ARORA : APACHE II IN INDIAN PATIENTS
expansion, symptom relief and surgery avoidance.
More studies need to be carried out with a larger
patient intake in order to evaluate American APACHE
II scoring system for the mortality prediction of critically
ill Indian patients admitted to ICUs with either respiratory
or non-respiratory ailments. Performance of model could
be possibly affected by lowering down cut-off value for
allotment of points for age from 45 yr (suggested in
original scoring system) to an approximately acceptable
level suitable for patient population of our country. It
could also involve giving weightage of points on basis of
a co-existent immunocompromised state like TB, a nurse-
patient ratio of ICU or a lead time bias. Perhaps,
combination of all is likely to yield a suitable modification
of scoring model for predicting mortality of patients
admitted to ICUs in our country. Following a validation,
treatment providers could use it for taking management
decisions resulting in optimal utilization of ICU resources
so crucial for a developing country like India.
In summary, despite a rise in observed and predicted
hospital mortality with 5-point APACHE II score,
predicted mortality did not correlate with observed
mortality for critically ill patients admitted to an Indian
respiratory ICU. The scoring system also showed a poor
calibration as well as discrimination. However, the model
may become more useful for Indian conditions by
lowering down the cut-off value in allotment of age points
and by awarding the weightage to factor like co-existing
immunocompromised state.
Acknowledgment
The authors acknowledge the Council of Scientific and Industrial
Research, New Delhi for financial support, and thank Dr Zaffar
Iqbal for invaluable contribution in preparation of the APACHE II
performas, and Shri Anup Kumar Singh for statistical analysis.
References
1. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE
II : A severity of disease classification system. Crit Care Med
1985; 13 : 818-29.
2. Wong DT, Crofts SL, Gomez M, McGuire GP, Byrick RJ.
Evaluation of predictive ability of APACHE II system and
hospital outcome in Canadian intensive care unit patients.
Crit Care Med 1995; 23 : 1177-83.
3. Zimmerman JE, Knaus WA, Judson JA, Havill JH, Trubuhovich
281
RV, Draper EA, et al. Patient selection for intensive care : A
comparison of New Zealand and United States hospitals.
Crit Care Med 1988; 16 : 318-26.
4. Chen FG, Koh KF, Goh MH. Validation of APACHE II score in
a surgical intensive care unit. Singapore Med J 1993; 34 : 322-4.
5. Lee KH, Hui KP, Lim TK, Tan WC. Acute physiology and
chronic health evaluation (APACHE II) scoring in the Medical
Intensive Care Unit, National University Hospital, Singapore.
Singapore Med J 1993; 34 : 41-4.
6. Rowan KM, Kerr JH, Major E, McPherson K, Short A, Vessey
MP. Intensive Care Society's APACHE II study in Britain and
Ireland-I : Variations in case mix of adult admissions to general
intensive care unit and impact on outcome. BMJ 1993; 307 :
972-7.
7. Rowan KM, Kerr JH, Major E, McPherson K, Short A,
Vessey MP. Intensive Care Society's APACHE II study in Britain
and Ireland-II : Outcome comparisons of intensive care units
after adjustment for case mix by the American APACHE II
method. BMJ 1993; 307 : 977-81.
8. Bastos PG, Sun X, Wagner DP, Knaus WA, Zimmerman JE.
Application of the APACHE III prognostic system in Brazilian
intensive care units : A prospective multi-center study. Intensive
Care Med 1996; 22 : 564-70.
9. Dragsted L, Jorgensen J, Jensen NH, Bonsing E, Jacobsen E,
Knaus WA, et al. Interhospital comparisons of patient outcome
from intensive care : Importance of lead-time bias.
Crit Care Med 1989; 17 : 418-22.
10. Jacobs S, Chang RW, Lee B. Audit of intensive care: A 30 month
experience using the APACHE II severity of disease classification
system. Intensive Care Med 1988; 14 : 567-74.
11. Hernandez JFR, Brito AE, Galsende MLQ, Prendes PN,
Pena MS, Correa NP. Dies anos de atencion al paciente critico.
I. Comparacion de 2 grupos de pacietes ingresados al principio
y final de la ultima decada. Rev Bras Terap Intensiva 1990;
2 : 7-9.
12. Parikh CR, Karnad DR. Quality, cost and outcome of intensive
care in a public hospital in Bombay, India. Crit Care Med 1999;
27 : 1754-9.
13. Sprung CL, Geber D, Eidelman LA, Baras M, Pizov R,
Nimrod A, et al. Evaluation of triage decisions for intensive care
admission. Crit Care Med 1999; 27 : 1073-9.
14. Mani C, Borah C, Krishna SV, Anuj M, Bhatnagar D.
The role of APACHE-II triaging in optimum management
of small bowel perforations. Trop Doct 2001; 31 :
198-201.
15. Hanley JA, McNeil BJ. The meaning and use of the area under
a receiver operating characteristic (ROC) curve. Radiology 1982;
143 : 29-36.
16. Lemeshow S, Hosmer DW Jr. A review of goodness of fit
statistics for use in the development of logistic regression models.
Am J Epidemiol 1982; 115 : 92-106.
282 INDIAN J MED RES, JUNE 2004
17. Lim BL, Chan YW. Audit of 2431 admissions to the surgical
care unit, Singapore General Hospital. Ann Acad Med Singapore
1998; 27 : 314-7.
18. Arunkumar AS, Rajaram MU, Kamat V. Validation of APACHE
II and SAPS II in a multi-disciplinary ICU. Indian J Crit Care
Med 2002; 6 : 12-8.
19. Jacobs S, Chang RW, Lee B. One year's experience with
the APACHE II severity of disease classification system
in a general intensive care unit. Anaesthesia 1987; 42 :
738-44.
20. Porath A, Eldar N, Harman-Bohem I, Gurman G. Evaluation of
the APACHE II scoring system in an Israeli intensive care unit.
Isr J Med Sci 1994; 30 : 514-20.
21. Teoh GS, Mah KK, abd Majid S, Streram, Yee MK. APACHE
II : Preliminary report on 100 intensive care unit cases in
University Hospital, Kuala Lampur. Med J Malaysia 1991;
46 : 72-81.
22. Agarwal N, Sankari M, Cayten CG, Brown BB, Murphy J,
Byrne D, et al. Influence of age on mortality of surgical
intensive care unit (SICU) patients. Crit Care Med 1989;
17 (Suppl) : S85.
Reprint requests : Prof. V.K. Arora, Director, LRS Institute of Tuberculosis & Respiratory Diseases
Sri Aurobindo Marg, New Delhi 110030, India
e-mail: lrsinstitute@eth.net & vk_raksha@yahoo.com
23. Beck DH, Taylor BL, Millar B, Smith GB. Prediction of
outcome from intensive care : A prospective cohort
study comparing acute physiology and chronic health
evaluation II and III prognostic systems in a United
Kingdom intensive care unit. Crit Care Med 1997;
25 : 9-15.
24. Zhu BP, Lemeshow S, Hosmer DW, Klar J, Avrunin J,
Teres D. Factors affecting the performance of the models in
the mortality probability model II system and strategies of
customization : A stimulation study. Crit Care Med 1996;
24 : 57-63.
25. Pandya SK. Mushrooming nursing homes. Natl Med J India
1995; 8 : 239-40.
26. Kahn JM. Recruitment and staffing. In : Adler DC,
Shoemaker NJ, editors. AACN organisation and
management of critical care facilities. St Louis: CV Mosby;
1979 p. 115-28.
27. Knaus WA, Wagner DP, Zimmerman JE, Draper EA. Variations
in mortality and length of stay in intensive care units.
Ann Intern Med 1993; 118 : 753-61.