Original Articles

Application of analytic hierarchy process for measuringand comparing the global performance of intensivecare units

Seetharaman Hariharan MDa,*, Prasanta K. Dey PhDb, Deryk R. Chen FRCAa,Harley S.L. Moseley FFARCSc, Areti Y. Kumar MDc

aAnaesthesia and Intensive Care, The University of the West Indies, St. Augustine, Trinidad and TobagobAston Business School, Aston University, Birmingham, B4 7ET UKcSchool of Clinical Medicine and Research, The University of the West Indies, Barbados

0883-9441/$ – see front matter D 2005

doi:10.1016/j.jcrc.2005.04.002

T Corresponding author. Tel.: +1 868

E-mail address: hseetharaman@fms

Keywords:Analytic hierarchy

process;

Global performance;

Intensive care unit

AbstractPurpose: To develop a model for the global performance measurement of intensive care units (ICUs)

and to apply that model to compare the services for quality improvement.

Materials and Methods: Analytic hierarchy process, a multiple-attribute decision-making technique, is

used in this study to evolve such a model. The steps consisted of identifying the critical success factors

for the best performance of an ICU, identifying subfactors that influence the critical factors, comparing

them pairwise, deriving their relative importance and ratings, and calculating the cumulative

performance according to the attributes of a given ICU. Every step in the model was derived by

group discussions, brainstorming, and consensus among intensivists.

Results: The model was applied to 3 ICUs, 1 each in Barbados, Trinidad, and India in tertiary care

teaching hospitals of similar setting. The cumulative performance rating of the Barbados ICU was 1.17

when compared with that of Trinidad and Indian ICU, which were 0.82 and 0.75, respectively, showing

that the Trinidad and Indian ICUs performed 70% and 64% with respect to Barbados ICU. The model also

enabled identifying specific areas where the ICUs did not perform well, which helped to improvise

those areas.

Conclusions: Analytic hierarchy process is a very useful model to measure the global performance

of an ICU.

D 2005 Elsevier Inc. All rights reserved.

1. Introduction

Performance measurement is not easy in health care

services because the measurement of efficiency of health

Elsevier Inc. All rights reserved.

662 4030; fax: +1 868 662 4030.

.uwi.tt (S. Hariharan).

care delivery is multifactorial, the metrics are highly

variable and difficult to be defined accurately.

Until recently, performance measurement of an inten-

sive care unit (ICU) is done by prognostic scoring sys-

tems such as Acute Physiology and Chronic Health

Evaluation, Simplified Acute Physiology Score, and Mor-

tality Prediction Model [1]. All these systems consider

binomial patient outcomes, namely, survival or death, as the

Journal of Critical Care (2005) 20, 117–125

S. Hariharan et al.118

indicators of measurement. These systems incorporate

logistic regression equations to predict the mortality for a

case-mix in a particular ICU. The ratio of the predicted

mortality to the observed mortality (standardized mortality

ratio [SMR]) is used to compare the performance of differ-

ent ICUs [2].

Although validated by many prospective studies, there

are many inherent problems with these models. A study

that used all the 3 models to compare ICUs from

32 hospitals (Project IMPACT) reported that they

exhibited only fair to moderate agreement in identifying

the quality measures [3]. Most of the ICUs studied were

found to be performing very well according to these

standards. The editorial accompanying this article queried

that, bif most ICUs graduate with honors, is it genuine

quality or grade inflation?Q With the presently available

models of performance measurement, it is difficult to

distinguish between the two [4]. Many studies have

reported poor goodness of fit for the scoring systems,

implying that the prognostic models do not perform

consistently in all intensive care patients [5-10]. Although

patient outcome may be the ultimate goal of any ICU,

there are many other factors that are equally important in

the day-to-day functioning of an ICU, which are remark-

ably missed by these scoring systems when used as

performance measurement models.

Recently, some studies have tried to identify other ICU

quality indicators [11,12]. Other performance measurement

methods of ICU involve data envelopment analysis and

specific functions of ICU such as measurements and

guidelines to improve patient care and resource use in the

ICU to establish and implement best practices [13,14].

These methods will enable to improvise only specific areas

of the functioning of an ICU.

When one considers individual factors such as mortality

rate of an ICU as a measure of performance, an increase in

the mortality rate is interpreted as a reduction in the level of

overall performance. It is imperative that, before arriving at

any conclusion, one has to investigate the reasons for the

increase in the mortality rate, which may be multifactorial.

However, if the method of performance measurement itself

incorporates all the factors, it makes the identification of

deficiencies easier for a manager. In an attempt to consider

all the functions of an ICU globally and measure the

performance of every function individually, we conceptu-

alized and designed the present study. Accordingly, the

objective of the study was to develop a holistic perfor-

mance measurement model for ICU services and practically

apply the model so as to enable improvement of the quality

of services.

2. Design and methodology

Analytic hierarchy process (AHP), a multiple-criteria

decision-making technique, has been applied in this study to

measure the performance of an ICU. Although in medical

field there have been many applications of AHP [15-18], to

our knowledge it has not been applied as a performance

measurement model for an ICU.

The model was practically applied initially to a surgical

ICU of Queen Elizabeth Hospital, Barbados, a 650-bed

tertiary care teaching hospital affiliated to the University of

the West Indies. Subsequently, it was applied to ICUs in

Trinidad (Eric Williams Medical Sciences Complex,

Mount Hope) and India (University General Hospital,

Vijayawada), which are tertiary care teaching hospitals of

similar setting.

Initially, a questionnaire survey was conducted among

the clinical professionals involved in the management of

the ICU to determine what they consider the important

processes of an ICU. This included 15 anesthesiologists,

5 senior nurses (sisters-in-charge), and 10 staff nurses. Five

senior anesthesiologists, including the 3 authors who are

actively involved in the day-to-day management of the

ICU, had multiple brainstorming sessions to discuss every

step in detail. By group consensus, every aspect in one step

was agreed upon; after which, we proceeded toward the

next step.

Our model involved the following steps:

2.1. Step 1: Identifying the critical success factorsand the subfactors

Critical success factors by definition are the character-

istics, conditions, or variables that when properly sustained,

maintained, or managed can have a significant impact on the

successful management of an organization [19]. Following

the most commonly used approach suggested by Donabedian

[20] for performance evaluation in health care, namely, the

structure, process, and outcome of health care delivery, we

identified these 3 parameters as the critical success factors.

Later, we had multiple brainstorming sessions that helped

us to identify subfactors that influence the critical factors,

which in turn would facilitate the overall measurement of

the performance.

1. processes of patient care in ICU;

2. structure (resources) of ICU;

3. outcome—morbidity/mortality of patients in ICU.

2.2. Processes of ICU

We classified the processes of ICU into 2 main areas,

namely, the processes related to patient care and those

related to patient comfort.

Subfactors related to patient care are as follows:

! therapeutic intervention—aggressive interventions

such as invasive lines, early tracheostomy in deserv-

ing patients, and bedside procedures (not waiting for

other units such as operating rooms);

! monitoring—availability of constant monitoring con-

ditions for all the patient parameters;

Table 1 Pairwise comparison and normalized matrix of the

factors

Factors Process Structure Morbidity/

mortality

Overall

importance

Process 1 2 2

(0.5) (0.5) (0.5) 0.5

Structure 1/2 1 1

(0.25) (0.25) (0.25) 0.25

Morbidity/

mortality

1/2 1 1

(0.25) (0.25) (0.25) 0.25

Numbers in parentheses denote the normalized matrix.

Global performance of intensive care units 119

! admission-discharge protocols—defining admission

and discharge criteria and adhering to them;

! daily rounds—organizing rounds to review patient’s

condition and progress;

! case conferences—conducting regular case confer-

ences to avoid morbidity in future.

Providing better patient comfort is an important measure

of the performance of an ICU. This can be achieved by

the following:

! sedation—appropriate use of sedatives and muscle

relaxants;

! audiovisual comforts—providing music and televi-

sion for appropriate patients;

! attitudes of staff—friendly behavior of staff toward

patients and relatives.

2.3. Structure

Performance of an ICU depends upon the availability of

the overall resources and facilities in the ICU:

! capital equipment—ensuring state-of-the-art technol-

ogy equipment;

! staffing—adequate and competent medical, nursing,

paramedical, and support staff;

! maintenance—appropriate and regular maintenance

of equipment by a dedicated biomedical engineering

team;

! supply chain management—ensuring constant supply

of drugs and other materials, establishing good

interdepartmental and intradepartmental communica-

tion, and constant availability of the services of

related departments such as laboratory and radiology.

2.4. Outcome: morbidity/mortality

We considered morbidity of a patient as an adverse

patient occurrence that resulted in prolonged ICU/hospital

stay or disability causing reduced quality of life during the

hospital stay and/or after discharge. Morbidity is a very

important factor in the assessment of the quality of an ICU,

because a poorly performing ICU may show an increased

morbidity of patients, even if mortality may be to the

expected standards. Morbidity and mortality of the patients

are influenced by these factors:

! patient factor—clinical presentation of the patients in

an ICU;

! iatrogenic complications—reducing the iatrogenic

complications to minimum;

! nosocomial infections—ICU-acquired infections be-

ing minimal;

! antimicrobial drugs—avoiding inappropriate and

overuse of antimicrobials by establishing and adher-

ing to prescription protocols;

! patient nutrition—ensuring the availability of enteral

and parenteral nutrition and appropriately using

them.

2.5. Step 2: Pairwise comparison of critical successfactors and subfactors

The next step was determining the importance of each

critical success factor and the subfactors by pairwise

comparison. Extensive brainstorming sessions were held

and the priorities were derived by comparing each set of

elements in a pairwise fashion with respect to each of the

element in higher stratum [21]. A 9-point numerical scale

was used for the comparison. The intensity and the

definitions of the pairwise comparison are as follows:

1 = equal importance, 2 activities contribute equally to

the object;

3 = moderate importance, slightly favors one over

another;

5 = essential or strong importance, strongly favors one

over another;

7 = demonstrated importance, dominance of the demon-

strated importance in practice;

9 = extreme importance, evidence favoring one over

another of highest possible order of affirmation;

2, 4, 6, 8 = intermediate values, when compromise

is needed.

In a common objectives context where all members of

the group have the same objectives, there are 4 ways that

could be used for setting the priorities: (1) consensus,

(2) vote or compromise, (3) geometric mean of the

individuals’ judgments, and (4) separate models or players

[22]. In our situation, although our participants had differing

opinions at the initial stage, the differences were not much

and adequate discussions could settle the differences so as to

enable to arrive at a consensus.

The numerical representations of the level of importance

for each critical success factor in the 9-point scale were

derived (first row of Table 1). For example, the factor process

was considered to have a moderate importance with an

intensity numerically represented by b2Q with respect to the

Therapeutic interventionsAggressive interventions in patientsGood >75% Average 50-75% Poor <50%

MonitoringGood full Average partial Poor minimal

Admission discharge protocolsGood Established protocol

followed Average Established protocol

not followed Poor No protocol

Daily roundsGood Well organized Average Not organized Poor No daily rounds

Case conferencesGood At organized

intervals Average Random Poor None

SedationGood Optimal use of

sedatives and relaxants Average Sub-optimal or

excessive use Poor Poor use of drugs

Audiovisual comfortsGood All deserving patients

receive Average Inadequate Poor None

Attitudes of staffGood Staff interaction with

patients/relatives regular

Average No regular interactions Poor None

Patient care Patient comfort

Fig. 1 Definition of ratings for the factor process.

Capital equipmentGood Fully equipped Average Partially equipped Poor Poorly equipped

StaffingGood Complete/ideal Average Partially deficient Poor Poor staffing pattern

Maintenance of equipmentGood Regular Average Irregular Poor Poor

Supply chain management

Drugs/suppliesGood Timely Average Short deficiencies Poor Long deficiencies

CommunicationGood Organized Average Not well organized Poor Poorly organized

Related departments (laboratory/radiology etc.)Good Fully equipped Average Partially equipped Poor Poorly equipped

Patient factorClinical presentation of patients Low No lead time - <50 %

patients present in a bad shape

Moderate 50-75 % patients present in a bad shape

High >75 % patients present in a bad shape

Iatrogenic complications andNosocomial infections

Low Minimal Moderate Average incidenceHigh High incidence

Antibiotic usage

Optimal Protocol followedExcessive/ Excessive/ Inadequate Indiscriminate use Poor No protocol followed

NutritionOptimal Correct use of enteral

and parenteral feeds Inadequate Indiscriminate use Poor No facilities available

for nutrition

Structure Morbidity/mortality

Fig. 2 Definition of ratings for the factors structure and

outcome.

S. Hariharan et al.120

factors structure as well as outcome. The factors structure and

outcome were considered to have equal importance, numer-

ically represented as b1.Q By this pairwise comparison, it

follows that the importance of structure and outcome with

respect to process will be 1/2. A normalized matrix was

derived from these comparisons (numbers given in paren-

theses in Table 1). This is calculated as the fraction of the

importance of each critical success factor with respect to the

sum of the overall comparison between the factors. Thus, the

sum of all the values of importance for the factor process is 2

(sum of column bProcessQ in Table 1) (1 + 1/2 + 1/2) and the

fraction of process alone will be 1 over 2 (0.5) and that of

structure will be 1/2 over 2 (0.25) and so on. The overall

importance of the factors was calculated as an average for

each factor (last column in Table 1). Thus, for process, it is 0.5

(0.5 + 0.5 + 0.5 divided by 3).

This procedure was done for all the subfactors to derive

their respective importance (local perspectives) by pairwise

comparison. The values for the individual subfactors

derived by pairwise comparison and normalization are

shown in Fig. 3 (numbers in parentheses after each

subfactor). Consistency check was done for all the

normalized matrices and was found to be less than 0.1,

which is within acceptable limits.

2.6. Step 3: Derivation of the attributes andweights for subfactors

For all the subfactors, 3 attributes were identified,

namely, good/adequate/low, average/moderate/inadequate,

and poor/nil/high. This was done by defining the catego-

ries of each subfactor by way of interviewing, brainstorm-

ing sessions, and discussions. For example, for the

subfactor staffing, we had 3 attributes, namely, good,

average, and poor. These were based on the characteristics

of the medical and nursing staffing of an ICU—good, if a

dedicated intensivist was available around the clock and

the nurse-patient ratio was 1:1; average, if an intensivist

was available part time and the nurse-patient ratio was 1:2

or 1:3; and poor, if no intensivist was available and/or

nurse-patient ratio was more than 1:3. For each subfactor,

the best attribute received a weight of 0.6, the moderate

0.3, and the poor 0.1.

Figs. 1 and 2 show the definitions we used to derive the

ratings of the subfactors.

Fig. 3 shows the entire hierarchical model in AHP

framework of the critical success factors, subfactors, their

relative importance, and the weights and ratings.

2.7. Step 4: Allocating the weights of ratings forthe 3 ICUs

The next step was to allocate the ratings of each

subfactor for the ICUs in Barbados, Trinidad, and India.

The clinicians and managers in each unit were interviewed

regarding the ratings of each subfactor. Ratings of each

subfactor were allocated to each ICU according to the

Process (0.5) Structure (0.25) Outcome (0.25)

Patient Care (0.5)

Patient comfort (0.5)

Sedation (0.4) Audiovisual comforts (0.2) Attitude of staff (0.4)

Capital equipment (0.3) Staffing (0.4) Maintenance of equipment (0.05) Supply-chain management (0.25)

Patient factor (0.3) Iatrogenic complications (0.2) Nosocomialinfections (0.2) Antibiotic usage (0.2) Nutrition(0.1)

Sub-factors

Critical success factors

Adequate (0.6) Partial (0.3) Nil (0.1)

Adequate (0.6) Inadequate (0.3) Poor (0.1)

Low (0.6) Moderate (0.3) High (0.1)

Measuring performance of an ICU Goal

Ratings and weights

Therapeutic interventions (0.25) Monitoring (0.3) Admission/ Discharge protocols (0.2) Daily rounds (0.125) Case conferences (0.125)

Good (0.6) Average (0.3) Poor (0.1)

Drugs/ supplies (0.33) Communication (0.33) Related departments (0.33)

Optimal (0.6) Excessive/ Suboptimal (0.3) Poor (0.1)

Fig. 3 Analytic hierarchy process framework with prioritizations.

Global performance of intensive care units 121

clinician’s view of where the particular unit stands

according to the attribute. For example, the subfactor

admission-discharge protocols for the critical factor process

had 3 attributes. The attribute good had a weight of 0.6;

average, 0.3; and poor, 0.1, respectively. Barbados ICU got

the weight 0.1 because we agreed that this ICU did not

have any such protocol. The weights for characteristics

that best described the performance of each ICU at each

subfactor level were thus assigned to the respective ICUs

in the 3 countries by explaining the criteria and character-

istics of each attribute to the clinicians and managers

attached to the respective ICUs.

Table 2 Weights and performance ratings for process factor and its

Subfactors Barbados ICU

Weighta Performanceb

Patient care

Therapeutic interventions Good 0.075

Monitoring Good 0.09

Admission-discharge protocols Poor 0.01

Daily rounds Poor 0.00625

Case conferences Good 0.0375

Patient comfort

Sedation Adequate 0.12

Audiovisual comforts Partial 0.03

Attitude of staff Average 0.06

Cumulative performance 0.42875a Weight of subfactor by its attribute.b The rating calculated by the steps of AHP.

2.8. Step 5: Calculating and comparing cumulativeperformance of the ICUs

By multiplying the importance (Table 1) of each critical

success factor with those of the subfactors (derived by

pairwise comparison) (Fig. 3), the overall importance

(global perspective) (GP) of individual subfactor was

derived. This GP was then multiplied by the weight each

subfactor received by its attribute, which gave the factor

level performance of each subfactor. The sum of the

performance rating of all subfactors gave the cumulative

performance of each critical success factor (Tables 2-4).

subfactors

Trinidad ICU Indian ICU

Weighta Performanceb Weighta Performanceb

Average 0.0375 Average 0.0375

Average 0.045 Average 0.045

Poor 0.01 Average 0.03

Average 0.01875 Good 0.0375

Good 0.0375 Good 0.0375

Partial 0.06 Partial 0.06

Partial 0.03 Nil 0.01

Average 0.06 Average 0.06

0.29875 0.3175

Table 3 Weights and performance ratings for structure factor and its subfactors

Subfactors Barbados ICU Trinidad ICU Indian ICU

Weighta Performanceb Weighta Performanceb Weighta Performanceb

Capital equipment Adequate 0.18 Inadequate 0.09 Inadequate 0.09

Staffing Inadequate 0.12 Poor 0.04 Inadequate 0.12

Maintenance of equipment Inadequate 0.015 Inadequate 0.015 Poor 0.005

Supply chain management

Drugs/supplies Good 0.0495 Inadequate 0.02475 Inadequate 0.02475

Communication Inadequate 0.02475 Inadequate 0.02475 Poor 0.00825

Related departments Inadequate 0.02475 Inadequate 0.02475 Poor 0.00825

Cumulative performance 0.414 0.21925 0.25625a Weight of subfactor by its attribute.b The rating calculated by the steps of AHP.

S. Hariharan et al.122

This method was used to calculate the performance ratings

of all the ICUs. The sum of performance ratings of all the

critical success factors gave the overall performance rating

of the given ICU (Table 5). The cumulative performance

ratings of the ICUs were then compared to know the level

of performance of each ICU with respect to another.

2.9. Summary of derivation of cumulativeperformance

Pairwise comparison of critical success factors and

normalization: importance of critical success factor (1);

Pairwise comparison of subfactors and normalization:

importance of individual subfactor (local perspective) (2);

Product of 1 and 2: overall importance of each subfactor

(GP) (3);

Allocation of weights by attributes of each subfactor

for sample ICUs: weight by the attributes that fit each

ICU (4);

Product of 3 and 4: performance rating of each subfactor

for each ICU (5);

Sum of all the subfactor performance ratings: perfor-

mance rating for each critical success factor (6);

Sum of performance ratings of all the 3 critical success

factors: cumulative performance rating of each ICU (7).

The theory of AHP is outlined elsewhere [23,24].

Although software is available to automatically enable all

Table 4 Weights and performance ratings for outcome factor and its

Subfactors Barbados ICU Trin

Weighta Performanceb We

Patient factor Moderate 0.09 Mo

Iatrogenic complications Moderate 0.06 Mo

Antibiotic usage Excessive 0.06 Exc

Nosocomial infections Moderate 0.06 Mo

Nutrition Optimal 0.06 Sub

Cumulative performance 0.33a Weight of subfactor by its attribute.b The rating calculated by the steps of AHP.

calculations (Expert Choice Inc., Arlington, VA), we did our

calculations byMicrosoft Excel (MS Office 2000, Microsoft)

to arrive at the results.

3. Results

By pairwise comparison of the factors, the present study

considered process of ICU as the most important factor with

equal importance to the ICU infrastructure and outcome of

patients (Table 1).

The cumulative rating of performance of Barbados ICU

was found to be 1.17, for Trinidad ICU, 0.82, and for Indian

ICU, 0.72. This gave us the percentage of performance of

Trinidad ICU and Indian ICU with respect to the Barbados

ICU as 70% and 64% (Table 5).

The ratings of many subfactors were low in Trinidad and

Indian ICUs when compared with Barbados ICU. This

clearly points out that these units need to improve their

efforts to provide a better care. Some factors such as ina-

dequate practice of admission-discharge protocols brought

down the ratings of Barbados ICU. The rating comparison

also showed that there is a need to improve overall staffing,

reduce nosocomial infection rates, optimize the antibiotic

use, and reduce iatrogenic complications. Although many

areas in the management of all the ICUs required attention,

sensitivity analysis showed that improving certain areas in

the ICUs will have much more impact than improving other

areas in the overall rating of the ICUs.

subfactors

idad ICU Indian ICU

ighta Performanceb Weighta Performanceb

derate 0.09 High 0.03

derate 0.06 Moderate 0.06

essive 0.06 Inadequate 0.02

derate 0.06 Moderate 0.06

optimal 0.03 Poor 0.01

0.3 0.18

Table 5 Overall comparison between Barbados, Trinidad,

and Indian ICUs

Factors Barbados

ICU

Trinidad

ICU

Indian

ICU

Process 0.43 0.3 0.32

Structure 0.41 0.22 0.26

Outcome

(morbidity/mortality)

0.33 0.3 0.18

Cumulative performance 1.17 0.82 0.75

Percentage of performance of Trinidad ICU with respect to Barbados

ICU, 70%; Indian ICU with respect to Barbados ICU, 64%; and Indian

ICU with respect to Trinidad ICU, 92%.

Global performance of intensive care units 123

After applying this model, among other measures, in

Barbados ICU, we have now established admission and

discharge criteria, we did research for elucidating the

infections and antibiotic use [25] and have established

protocols for antibiotic use and sedation practices, and we

have improved the patient-nurse ratio to 1:1 from the

existing 2:1.

4. Discussion

Measurement of the quality of health care delivery may

be done by 3 parameters, namely, structure of the unit,

process of care, and outcome of patients as described by

Donabedian [20].

There have been controversies regarding process- and

outcome-based methods of quality assessment. The pro-

cess-based quality measure may not be able to predict the

outcome and the outcome-based quality measure may not

consider the differences in patients receiving the same type

of treatment, having a poor outcome due to factors that are

not under the control of the physicians [26]. Performance

appraisal models do not incorporate all the 3 parameters

and there has been acknowledged difficulty in creating a

link between the process and outcome [27]. Our present

model attempts to achieve this link by having measures for

all the 3 parameters. The present model also attaches more

significance to measuring the process of ICU than the

outcome of patients.

The main aim of critical care after its introduction as a

separate specialty has been prevention of mortality [28].

However, controversy exists whether mortality alone can be

a reliable indicator for the measurement of the performance

of an ICU [29]. Most patients admitted to ICUs may be

already very sick and provision of the best possible care

does not necessarily ensure preventing mortality in many

cases. Hence, even if a well-performing ICU is available,

many other factors may influence the survival of the patient

from the critical illness. This multifactorial aspect has been

affirmed by another recent study [30].

In contrast to the mortality that depends more on the

patient factor, morbidity depends more on the ICU care

factor, and therefore, morbidity reflects the quality of the

ICU service more than mortality. Hence, we included

morbidity as one of the outcome measures.

In our previously published research, we used the Acute

Physiology and Chronic Health Evaluation II scoring

system to compare the performance of Barbados ICU with

the ICUs of developed countries using the SMR as the tool

for comparison [31]. The SMR was 0.96, which meant that

Barbados ICU performed equally well compared with one

in a developed country and qualified with honors. However,

the present study gave a clearer picture of the functioning

of the ICU with respect to many other aspects in addition

to mortality.

In the view of the authors, patient comfort is a factor that

has been most often ignored in evaluating an ICU perfor-

mance. Providing better patient comfort should be considered

as one of the important aspects of ICU performance, which

the mortality-based models do not take into account at all.

The ICU nurses in general are obsessed to provide better

patient comfort, and in our situation, all our nurses gavemuch

importance to the patient comfort factor in the process of ICU

care. The mortality rate of our ICU is around 16%, which

means 84 of 100 patients recover from the critical illness, and

hence, it makes lot of sense when we emphasize providing

better patient comfort for most of the ICU patients. The

importance of the patient factor is being recognized as bwhatis best for the patient is best for the ICUQ [32].

There are many advantages of using AHP in the

performance measurement of ICU:

! Intensive care unit service is multifactorial and the

factors are both objective and subjective in nature and

measurement of the performance of such a system can

be easily modeled using AHP.

! Performance measurement is also a group decision–

making process, and AHP allows the same.

! Analytic hierarchy process has a sound mathematical

basis, and its application is user-friendly.

! Analytic hierarchy process enables to identify the

deficiencies in the specific areas of ICU care.

! Analytic hierarchy process allows to prioritize the

areas requiring improvement.

! Software that many ICUs following our model can

use is available.

The present study has some drawbacks. Although the

weights and ratings for each subfactor were allocated by

detailed discussions and brainstorming sessions, these may

still be considered subjective. When attributes such as

attitude of staff are measured, it may be impossible to

completely eliminate the subjective component. However,

one should ensure that the pairwise comparisons are

consistent, which was the case in the present study. In

addition, the ratings for the attributes are uniform and it is

possible for any ICU manager to allocate the rating for a

specific ICU by following the definitions. The subjective

component may also be minimized by an honest allocation of

S. Hariharan et al.124

the rating by attributes as well as by involving all the

clinicians and managers of an ICU and obtaining a cross-

sectional opinion. It may be argued that some factors such as

antimicrobial use and provision of nutrition are processes in

an ICU. We categorized them under morbidity/mortality

because of their close relationship to patient outcome.

Because we were able to apply the model to 3 different

ICUs, it should be possible for any manager to similarly

apply the model to the respective ICUs. Health care delivery

is highly complex and the patterns are different in every

region in accordance with the sociocultural, economic, and

political settings. When different factors take priority in

different settings and there are trade-offs between decision

criteria, AHP is one of the most useful tools for the

successful application in multiple-criteria decision-making

situations [33].

The present study establishes AHP as a valuable tool to

design amodel for the global performance measurement of an

ICU and compare it with another. Every ICU may apply this

model and have the ratings for the factors and arrive at

individual conclusions.

References

[1] Zimmerman JE. Measuring intensive care performance: a way to

move forward. Crit Care Med 2002;30:2149-50.

[2] Becker RB, Zimmerman JE. ICU scoring systems allow prediction of

patient outcomes or comparison of ICU performance. Crit Care Clin

1996;12:503-14.

[3] Glance LG, Osler TM, Dick A. Rating the quality of intensive care

units: is it a function of the intensive care scoring system? Crit Care

Med 2002;30:1976-82.

[4] Popovich MJ. If most intensive care units are graduating with

honors, is it genuine quality or grade inflation? Crit Care Med 2002;

30:2145-6.

[5] Marik PE, Varon J. Severity scoring and outcome assessment:

computerized predictive models and scoring systems. Crit Care Clin

1999;15:633-46.

[6] Rowan KM, Kerr JH, Major E, et al. Intensive Care Society’s Acute

Physiology and Chronic Health Evaluation (APACHE II) study in

Britain and Ireland: a prospective, multicenter, cohort study compar-

ing two methods for predicting outcome for adult intensive care

patients. Crit Care Med 1994;22:1392-401.

[7] Muckart DJJ, Bhagwanjee S, Neijenhuis PA. Prediction of the risk of

death by APACHE II scoring in critically ill trauma patients without

head injury. Br J Surg 1996;83:1123-7.

[8] Markgraf R, Deutschinoff G, Pientka L, Scholten T. Comparison of

acute physiology and chronic health evaluations II and III and

simplified acute physiology score II: a prospective cohort study

evaluating these methods to predict outcome in a German interdis-

ciplinary intensive care unit. Crit Care Med 2000;28:26 -33.

[9] Katsaragakis S, Papadimitropoulos K, Antonakis P, et al. Comparison

of Acute Physiology and Chronic Health Evaluation II (APACHE II)

and Simplified Acute Physiology Score II (SAPS II) scoring systems

in a single Greek intensive care unit. Crit Care Med 2000;28:426-32.

[10] Sikka P, Jaafar WM, Bozkanat E, El-Solh AA. A comparison of

severity of illness scoring systems for elderly patients with severe

pneumonia. Intensive Care Med 2000;26:1803-10.

[11] Berenholtz SM, Dorman T, Ngo K, Pronovost PJ. Qualitative review

of intensive care unit quality indicators. J Crit Care 2002;17:1 -12.

[12] Pronovost PJ, Berenholtz SM, Ngo K, et al. Developing and pilot

testing quality indicators in the intensive care unit. J Crit Care 2003;

18:145-55.

[13] Field K, Emrouznejad A. Measuring the performance of neonatal care

units in Scotland. J Med Syst 2003;27:315-24.

[14] Dlugacz YD, Stier L, Lustbader D, Jacobs MC, Hussain E,

Greenwood A. Expanding a performance improvement initiative in

critical care from hospital to system. Joint Comm J Qual Improv

2002;28:419-34.

[15] Weingarten MS. A pilot study of the use of analytic hierarchy process

for the selection of surgery residents. Acad Med 1997;72:400 -2.

[16] Hummel JM. Medical technology assessment: the use of analytic

hierarchy process as a tool for multidisciplinary evaluation of medical

devices. Int J Artif Organs 2000;23:782 -7.

[17] Carter KJ. Analysis of three decision-making methods: a breast cancer

patient as a model. Med Decis Making 1999;19:49-57.

[18] Dolan J, Frisina S. Randomized controlled trial of a patient decision

aid for colorectal cancer screening. Med Decis Making 2002;22:

125-39.

[19] Leidecker JK, Bruno AV. Identifying and using critical success

factors. Long Range Plan 1984;17:23 -32.

[20] Donabedian A. The quality of care—how can it be assessed? JAMA

1988;260:1743-8.

[21] Wind Y, Saaty TL. Marketing applications of the analytic hierarchy

process. Manag Sci 1980;26:641-58.

[22] Dyer RF, Forman EH. Group decision support with the analytic

hierarchy process. Decis Support Syst 1992;8:99-124.

[23] Dolan JG, Isselhardt Jr BJ, Cappuccio JD. Analytic hierarchy

process in medical decision making: a tutorial. Med Decis Making

1989;9:40-50.

[24] Saaty T. Axiomatic foundations of the analytic hierarchy process.

Manag Sci 1986;32:841 -55.

[25] Hariharan S, Nanduri SB, Moseley HSL, Kumar AY, Jonnalagadda R.

Spectrum of microbes and antimicrobial resistance in a surgical

intensive care unit, Barbados. Am J Infect Control 2003;31:280-7.

[26] Brook RH, McGlynn EA. Measuring quality of care. N Engl J Med

1996;335:966 -9.

[27] Lurie JD, Merrens EJ, Lee J, Splaine ME. An approach to hospital

quality improvement. Med Clin North Am 2002;86:825 -45.

[28] Nierman DM. A structure for chronically critically ill. Crit Care Clin

2002;18:477 -91.

[29] Sherck JP, Shatney CH. ICU scoring systems do not allow prediction

of patient outcomes or comparison of ICU performance. Crit Care

Clin 1996;12:515 -23.

[30] Rotondi AJ, Sirio CA, Angus DC, Pinsky MR. A new conceptual

framework for ICU performance appraisal and improvement. J Crit

Care 2002;17:16 -28.

[31] Hariharan S, Moseley HSL, Kumar AY. Outcome evaluation in a

surgical intensive care unit in Barbados. Anaesthesia 2002;57:

434-441.

[32] Green TP. What is best for the patients is the best for the intensive care

unit. Crit Care Med 2001;29:2038-9.

[33] Sloane EB, Liberatoire MJ. Medical decision support using the

analytic hierarchy process. J Healthc Inf Manag 2002;16:38 -43.

Commentary

Is what you have or how you use it or whatyou achieve the more important determinant ofquality?

In this issue of the Journal of Critical Care, Hariharan

et al [1] present an interesting decision support method for