The staff in the neonatal intensive care units is required to have highly specialized training and the using equipment in this unit is so expensive. The random number of arrivals, the rejections or transfers due to lack of capacity and the random length of stays, make the advance knowledge of the optimal staff; equipments and materials requirement for levels of the unit behaves as a stochastic process. In this paper, the number of arrivals, the rejections or transfers due to lack of capacity and the random length of stays in a neonatal intensive care unit of a university hospital has been statistically analyzed.

1. J Med Syst
DOI 10.1007/s10916-009-9259-8
ORIGINAL PAPER
Statistical Analysis of Patients’ Characteristics in Neonatal
Intensive Care Units
Ali Kokangul & Ayfer Ozkan & Serap Akcan &
Kenan Ozcan & Mufide Narli
Received: 2 December 2008 / Accepted: 26 January 2009
# Springer Science + Business Media, LLC 2009
Abstract The staff in the neonatal intensive care units is Introduction
required to have highly specialized training and the using
equipment in this unit is so expensive. The random number The child/baby mortality is one of the most important sub-
of arrivals, the rejections or transfers due to lack of capacity indexes of national life standards, determined by Human
and the random length of stays, make the advance Developing Index [1]. A baby might require neonatal
knowledge of the optimal staff; equipments and materials intensive care due to premature birth, low birth weight or
requirement for levels of the unit behaves as a stochastic a respiratory disorder. The staff and physicians in the
process. In this paper, the number of arrivals, the rejections or neonatal intensive care units (NICU) are the most experi-
transfers due to lack of capacity and the random length of enced and specialized, and they are required to have highly
stays in a neonatal intensive care unit of a university hospital specialized training to work in the NICU.
has been statistically analyzed. The arrival patients are The patients admitted to NICU are placed in one of four
classified according to the levels based on the required nurse: levels based on the required nurse: patient ratio. This ratio is
patient ratio and gestation age. Important knowledge such as calculated by using patient level or dependency scales, which
arrivals, transfers, gender and length of stays are analyzed. operate on the assumption that the more critically ill the patient,
Finally, distribution functions for patients’ arrivals, rejections the more nurse time is needed to care for the patient [2]. Most
and length of stays are obtained for each level in the unit. National Health Service hospital trusts in England began
using the EURICU-1 level system, which calculates nurse:
Keywords Newborn . Neonatal intensive care units . patient ratio [3, 4]. A nurse to patient ratio of 1:2 and 1:4 for
Patients’ characteristics . Statistical analysis non-ventilation beds are maintained round the clock [5].
In this study, the NICU of a university hospital is
considered. It is capable of invasive and non-invasive
A. Kokangul : A. Ozkan : S. Akcan (*) : M. Narli monitoring of the neonate’s cardio respiratory systems and
Department of Industrial Engineering, can provide oxygen, and age appropriate thermoregulation.
Faculty of Engineering and Architecture, Cukurova University, In addition, heart rate, respiratory rate, blood pressure and
01330, Adana, Turkey oxygen saturation are monitored continuously. There are
e-mail: sakcan@cu.edu.tr
three levels in the unit. The patients in these levels may
A. Kokangul become stabilized or worse and re-categorized into one of
e-mail: kokangul@cu.edu.tr
the other levels. The arrival patients are placed in one of
A. Ozkan three levels as described below.
e-mail: ayferdursoy@yahoo.com
M. Narli & Level I patients are those who require close observation,
e-mail: znarli@yahoo.com but not necessarily the continuous presence of a nurse at
the bedside.
K. Ozcan
& Level II patients are those who require close observa-
Neonatal Intensive Care Unit, Cukurova University,
01330, Adana, Turkey tion, but not necessarily the continuous presence of a
e-mail: kenanozcan73@gmail.com nurse at the bedside.

2. J Med Syst
Materials and methods
It is crucial that the staff or equipment requirement should
be appropriate to the dependency of the patients in the unit.
To present the statistical analysis we consider the patients
flow in the NICU. Patients come from outside to one of the
levels in the NICU. When the patients arrive to the NICU:
If there is unoccupied staff and equipments they will be
admitted and accepted, if there is not any unoccupied staff
or equipment in the level, they will be rejected or
transmitted to elsewhere. The patients in the levels may
become stabilized or worse and re-categorized into one of
Fig. 1 Yearly amount of unclassified patient admissions the other levels or they may die during the shift. Patients
come from a level to another level have higher priority then
patient come from outside the NICU. When a patient
transmitting from a level to another one, if there is no
& Level III patients are those who require a nurse at the
unoccupied staff and equipments the transmitting patient
bedside continuously for 24 h/day.
will stay in the same level for any unoccupied staff and
The number of arrivals, transmitting rate among levels equipments.
and length of stays (LOS) are random and this randomness The considered NICU is able to serve patients with
make the number of patients in each level behave as a different level of medical treatment. The NICU observed
stochastic process. In practice, the variance of the differ- through this study has dedicated equipments to serve at
ence of the arrival and departure rates can deplete patients three different levels of medical treatments. The distribu-
in the levels causing unoccupied staff and equipment or fill tions of the equipments with respect to the levels of the unit
the unit causing rejection or transfer. Naturally, the are listed as follows: Level I: two open beds; level II: three
management of the unit tries to avoid these problems. open beds and eight incubators; level III: three open beds,
Several methods such as ratio based method [6–8], discrete ten incubators and ten ventilators. To perform the statistical
event simulation [9, 10], stochastic simulation and queuing analysis of the NICU the patients flow in the unit is
models [11, 12], a combination of simulation, queuing theory observed for 5 years (2000–2004), then for each level and
and statistical analysis [13, 14] and integrated methodologies gestation age obtain distributions for the following param-
that combined stochastic and deterministic approaches [15– eters: Daily arrivals, LOS and rejected or transmitted
17] have been suggested to solve the equipment planning arrivals. In addition; the gestation age, gender, birth weight,
problem of hospital in the literature. diagnosis, arrival date, leave date and leave reasons data of
All of these methods are based on the statistical analysis of 3,330 patients treated in 2000–2004 period is collected.
patients’ arrivals and LOS. In most of these studies, the This study aims to investigate the characteristics of the
variation of requested admissions over time has not been taken patients and to form basis for other studies. For this
into account. The variations in demand of staff, equipments or proposes, the registration data of the patients were analyzed
materials arise due to unpredictable nature of NICU admission and interpreted statistically and graphically.
rates and LOS. Using the average number of patients expected
in a year, average LOS and target occupancy level to calculate
the capacity needed is mathematically incorrect because of
non-linearity and variability in the factors that control LOS
[18]. Also, the level of the economic use of any units’
capacity may be changed from a unit to another unit in the
same hospital. Therefore, the statistical analysis for each unit
in the hospital is useful in practice.
In this study, the probability distributions for accepted,
transferred or rejected arrivals and LOS are determined for
each gestation age and level in the NICU. Also, arrival
sources, transfer reasons, gender of arrivals and transmitting
rate among levels are statistically analyzed. To the best of the
authors’ knowledge these kinds of statistical analysis have
not been performed for a NICU of a university hospital. Fig. 2 Gender based distribution of accepted patients

3. J Med Syst
Table 1 Patient source
Source Year Total %
patients
2000 2001 2002 2003 2004
The number of % The number of % The number of % The number of % The number of %
patients patients patients patients patients
Considered 379 61 461 69 452 70 536 75 453 73 2281 70
Hospital
Other 119 19 100 15 114 18 101 14 84 14 518 16
hospitals
Other 126 20 108 16 83 13 75 11 82 13 474 14
provinces
Statistical analysis was done in two different cases. In arrivals are in agreement with the Poisson distribution.
the first case, no special features of the patients were taken Cochran and Bharti [15] performed statistical analysis
into consideration and no classifying has been done. In the related to arrival, stay distributions, and unit usage rate of
second case, patients are classified according to the gestation patients that covers the whole units in the hospital.
age and levels of the unit. In both cases the analyses were In this study, patients applying to NICU were statisti-
done considering patients’ arrivals frequency, LOS and cally investigated without classifying. As seen in Fig. 1,
discharge probability at monthly and yearly basis. Further- 3,330 admitted patients are considered and more than 600
more, to obtain the LOS and rejection rate, patients were patients have been accepted per year. It is also observed that
classified in terms of the level base. the highest admission occurred in the year 2003.
To analyze the patient characteristics statistically, hypoth- Gender based distribution of accepted patients with
eses were developed and tested with a statistical software respect to years is shown in Fig. 2. It can be concluded
package such as Statistica 6.0 [19] and MatLab 7.0 [20]. that newborn boy babies need more treatment compared to
girl babies. Although the total number of accepted patients
may change, mostly the distribution of the patients at the
Research and finding gender base remains unaltered.
The arrival characteristics for 57 admitted patients could
The analysis of non-classified patients not be collected. Therefore, arrival characteristics of 3,273
patients are considered. To determine the arrival character-
Gronescu et al. [21] investigated all patients’ arrivals to the istics of the patients, their admission data are classified into
geriatric unit in terms of the arrival frequencies, LOS and three groups. These are the patients that come from the
rejection rate to determine the bed fill rate and cost. other units of the same, the patients that come from the
Groothius et al. [22] did not classify the patients applied other hospitals that are located in the same province, and
to the cardiology units. They performed the statistical patients who come from other provinces.
analyses to determine the demands and the characteristics It can be seen in Table 1, that 70% of patients have come
of patients. Analysis results showed that the patients’ from the considered, 16% of patients have come from the
Table 2 Departure reasons of patients per year
Reasons Year Total %
Patients
2000 2001 2002 2003 2004
The number of % The number of % The number of % The number of % The number of %
patients patients patients patients patients
Getting Well 508 85 497 75 510 79 623 88 348 77 2486 80,9
Transmitting 28 4,7 58 9 67 10 13 1,8 11 2 177 5,8
Death 60 10 111 17 71 11 73 10 95 21 410 13,3

4. J Med Syst
Fig. 3 Monthly patients’
arrivals
other hospitals that are in the same province and 14% of H1 = Patients’ arrivals are not in agreement with the
patients have come from the other provinces. The number of Poisson distribution
patients originated from other hospitals and provinces Hypotheses were analyzed with chi-square test, which is
decreased during 5 years period. Transmitting a sick baby used to determine the differences between the variables and
from a hospital to another one may increase the mortality operates with the frequency distributions. Statistical soft-
rate; therefore, most of patients preferred the closed hospital. ware analyzed the patients’ arrivals using the chi-square test
Patients are accepted to be departure from the hospital in with 95% confidence. The test parameters, namely, the chi-
the event of getting well, death, or when they are sent to square test statistics and the level of significance (p) were
another hospital while the illness is continued. The departure found to be 3.57 and 0.46 respectively. As the obtained chi-
characteristics for 257 admitted patients could not be square test statistic is lower than the chi-square table value,
collected. Therefore, during this study the departure character- the difference between the expected and the real values are
istics of 3,073 patients are considered. As it can be seen from not significantly low.
the Table 2, while 80.9% of the patients are departure from As the test, significance level (p) is larger than of 0.05;
the NICU due to recovery of health, 13.3% departure due to H0 hypothesis cannot be rejected [24]. Thus, the patients’
death and 5.8% sent to another hospital or another unit. arrival per year found to be in agreement with the Poisson
When the number of patients’ arrivals is investigated at distribution. Similarly chi-square test was applied for the
the month basis, it can be seen in Fig. 3 that, there exist an other 4 years and the patients’ arrivals and distribution
increase by January, which is decreased around April and parameters were obtained as seen in Table 3.
re-increased around June. The same hypotheses test was applied on the monthly
patients’ arrivals and they also found to be in agreement
Arrival distribution for the non-classified patients with the Poisson distribution.
To introduce the demand on the NICU and inspect the
arrived patients’ profile, the number of patients applied in a
5 years period has been obtained. The distribution of daily
patients’ arrivals in 2000 is given in Fig. 4 as an example.
Although there were no arrivals for 73 days either due to the
lack of capacity in the equipments used for the treatment or
due to the existence of no application, the number of arrivals
in some days increased up to six patients.
In most of studies, the arrival pattern of patients is
considered as Poisson process [21–23]. It can be concluded
from Fig. 4 that the patients’ arrivals are in agreement with
Poisson process. To check the suitability with the Poisson
process a zero and an alternative hypothesis were formed.
In that respect,
H0 = Patients’ arrivals are in agreement with the Poisson
distribution Fig. 4 Patients’ arrival distribution in 2000

5. J Med Syst
Table 3 The statistical distributions of the admitted patients
Year Chi-Square Significance (p) Distribution Parameter
2000 3,57 0,46 Poisson λ=1,6612
2001 2,03 0,72 Poisson λ=1,83562
2002 4,34 0,36 Poisson λ=1,80274
2003 3,65 0,6 Poisson λ=1,94795
2004 4,32 0,36 Poisson λ=1,74590
Analyses of the classified patients
Studies performed on the basis of whole hospital and on the
basis of different units’ show that the classifying of patients
Fig. 5 Gender distribution of groups (%)
varies with respect to the units. Ridge et al. [25] and Utley
et al. [26] classify the patients into two groups after the
studies carried at the intensive care units, as urgent patients
and patients with appointments. Darzi et al. [27] and high probability of premature birth and they need treatment
Navarro et al. [28] classified the patients in geriatrics unit after birth. On the year based examination, however, it has
in to three groups with respect to the treatment period as been found that while the numbers of group IV patients
short term, mid-term and long term treated. Utley et al. [29] were decreasing, numbers of group II patients were
on the other hand, classified the patients’ arrivals and LOS increasing.
as acute and non-acute. In this study the arrival patients are It can be seen from the Fig. 5 that the number of boy
classified according to the gestation age and treatment babies applied to newborn care units are higher than the
level. number of girl babies. Group IV has the highest ratio of boy
babies with 59.9%.
Patient analysis with respect to gestation age Table 5 presents the departure reasons of the patients
based on their gestation age’s groups. Group 0 patients have
In a study carried by the Turkish Neonatology Association relatively low probability of survival and group III patients
in 2003 [30], patients were classified with respect to the have relatively high probability of survival. It can be
gestation period while analyzing the death ratio of newborn depicted that patients of the last four groups (Group II,
infants. In this study, similarly, patients were classified Group III, Group IV and Group V) have higher probability
according to their gestation period in to six groups. of getting well compared to the patients of the first two
Classification with respect to gestation age is given in groups.
Table 4. When the distribution of the patients applied to the The collected data for the group 0 is not sufficient for
NICU is investigated, it can be seen that while the majority statistical analysis. Therefore, group based patients’ arrivals
of the patients belong to group III, minority belong to group distributions are analyzed for the last five groups. The
0. Thus, it can be concluded that the group III patients has a group-based patients’ arrivals distributions, done with chi-
square test statistical analysis, are in an agreement with the
Poisson and Geometrical distributions as seen in Table 6.
Table 4 The number of admitted patients with respect to gestation Patients’ analysis with respect to treatment level
age
Group Gestation Year The Total of Patients are directed by the doctors to the levels (Level I,
age (week) Groups Level II and Level III) according to the necessary
2000 2001 2002 2003 2004
equipment and medical staff needed for their treatments.
0 21–25 7 9 8 9 10 43 The LOS of the emergency patients and planned patients
1 26–28 52 60 46 60 57 275 were described using either the negative exponential curve
2 29–32 122 116 129 136 167 670 or a Weibull curve fitting routine [31]. Tu et al. [32] have
3 33–37 161 220 200 246 209 1036 used discriminate analysis for post-operative prediction of
4 38–42 236 178 122 120 56 712
LOS. A monogram is described for predicting the LOS of
5 43– 38 73 140 139 119 509
neonatal patients [33]. Utley et al. [29] assumed that the
Total 616 656 645 710 618 3245
time a patient stays in a particular pool of beds before either

6. J Med Syst
Table 5 Departure reasons of the patients based on their gestation age
Reasons Group (Gestation Age (week))
0 (21–25) 1 (26–28) 2 (29–32) 3 (33–37) 4 (38–42) 5 (43– )
The number of % The number of % The number of % The number of % The number of % The number of %
patients patients patients patients patients patients
Getting well 10 24 139 54 480 79 847 88 570 82 371 79
Transmitting 0 0 14 6 41 7 60 6,2 48 7 42 9
Death 32 76 103 40 84 14 56 5,8 74 11 58 12
Table 6 Group based patient arrivals and statistical distribution parameters
Group Gestation Chi-Square Significance Distribution Parameter
age (week) (p)
0 21–25 - - - -
1 26–28 0,05 0,81 Geometric p=0,86959
2 29–32 5,24 0,07 Poisson λ=0,36344
3 33–37 3,8 0,15 Poisson λ=0,56103
4 38–42 3,7 0,054 Geometric p=0,72385
5 43– 1,62 0,44 Poisson λ=0,27805
Table 7 Level based patient arrivals’ and LOS distributions and parameters
Arrival LOS
Distribution Test Distribution Test
Level I Poisson (λ=0,24725) K-S (0.0144 < 0.56) Lognormal (μ=1.670 σ=0.998) K-S ( 0.073 < 0.074)
Level II Poisson (λ=0,76438) K-S ( 0.0068 <0.56) Lognormal (μ=2.16 σ=0.689) K-S ( 0.087 < 0.098)
Level III Poisson (λ=0,63288) K-S ( 0.012 < 0.56) Lognormal (μ=1.54 σ=1.139) Chi-Square (9.01 < 14.06)
Table 8 Level based rejected patients’ distributions and parameters
Rejected patients Arrival
Distribution Test
Level I Poisson (λ=0,87097) K-S (0.0422 < 0.565)
Level II Poisson (λ=0,97561) K-S (0.0175 < 0.624)
Level III Poisson (λ=0,96078) K-S (0.04926 < 0.624)

7. J Med Syst
discharge or transfer to the other pool can be treated as distributions will make possible applying methods such as
identically and independently distributed. Each level of the simulation, queuing models, and integrated methodologies
NICU required different specialized equipments and qual- that combined stochastic and deterministic approaches to
ified staff. This requirement level varies from a level to determine the optimal necessary staff or equipments for any
another level in the unit. All of these equipment and staff levels of a NICU. This study can further be enriched by
requirement for each level based on the patients’ arrivals adding the population-increasing rate.
and LOS of the levels. Therefore, in this section the
patients’ arrivals and LOS for each level have been
obtained by employing the chi-square and kolmogorov- References
simirnov (K-S) tests.
Considering the levels of the unit, as seen in Table 7 the
1. Turkey State Planning Organization: (http://ekutup.dpt.gov.tr/
patients’ arrival profile is in agreement with Poisson ekonomi/gosterge).
distribution and LOS distribution is in agreement with the 2. Adomat, R., and Hicks, C., Measuring nursing workload in
Lognormal distribution. Each NICU in practice have intensive care: an observational study using closed circuit video
cameras. J. Adv. Nurs. 42(4):402–412, 2003. doi:10.1046/j.1365-
limited equipment or staff capacities. These limited capac-
2648.2003.02632.x.
ities can cause to rejections. The number of rejection 3. DoH., Circular HSG(95)8, NHS responsibilities for meeting
patients when increased, the management of the unit may continuing health care needs. London: DoH, 1996.
prefer to increase their limited capacities. To determine the 4. Miranda, D. R., Moreno, R., and Iapichino, G., Nine equivalents
of nursing manpower use score (NEMS). Intensive Care Med.
real demand for any equipment and staff of levels, the 23:760–765, 1997. doi:10.1007/s001340050406.
rejected patients should be analyzed for each level. 5. Narang, A., Kiran, P. S., and Kumar, P., Cost of neonatal intensive
Therefore in this study the rejected patients are also care in a tertiary care center. Indian Pediatrics. 42-October 17;
considered beside the accepted patients and the obtained 42:989–997, 2005.
6. Jung, A. L., and Streeter, N. S., Total population estimate of
statistical analysis parameters were summarized in Table 8.
newborn special-care bed needs. Pediatrics. 75:993–996, 1985.
It was found that rejected patients’ arrivals are also follows 7. Parmanum, J., Field, D., Rennie, J., and Ster, P., National census
a Poisson distribution. of availability of neonatal intensive care. BMJ. 321:727–729,
2000. doi:10.1136/bmj.321.7263.727.
8. Gorunescu, F., McClean, S. I., and Millard, P. H., A queuing
model for bed-occupancy management and planning of hospitals.
Results & discussions J. Oper. Res. Soc. 53:19–24, 2002. doi:10.1057/palgrave/jors/
2601244.
When compared on the year and gestation age basis the 9. Davies, R., Simulation for planning services for patients with
coronary artery disease. Eur. J. Oper. Res. 71:323–332, 1994.
arrival numbers of girl to the NICU are lower than the
doi:10.1016/0377-2217(94)90313-1.
arrival number of the boy. It can be concluded that the girl 10. Hudson, P. E., Peel, V., and Rayner, A., Using computer
babies needs less treatment after the birth. The arrivals simulation to plan accident & emergency services. Br. J. Health
show seasonal variations. While the number of patients is Care Manage. 3:160–162, 1997.
11. Jacobson, S., Hall, S., and Swisher, J., Discrete event simulation
increasing between April and July, it is decreasing between
of health care systems. In: Delay Management in Healthcare
January and April and between July and December. Systems. Hall R. Springer in Health Care Management, 2006.
Although most of the patients are discharged, patients of 12. Milne, E., and Whitty, P., Calculation of the need for pediatric
group 0 and 1 have higher percentage of death. intensive care beds. Arch. Dis. Child. 73:505–507, 1995.
doi:10.1136/adc.73.6.505.
In the recent year simulation, queuing models, statistical
13. Romanin-Jacur, G., and Facchin, P., Optimal planning of a
analysis and integrated methodologies that combined pediatric semi-intensive care unit via simulation. Eur. J. Oper.
stochastic and deterministic approaches [15–17] have been Res. 29:192–198, 1987. doi:10.1016/0377-2217(87)90109-3.
suggested to solve the equipment-planning problem of 14. Hershey, J. C., Weiss, E. N., and Cohen, M. A., A stochastic
service network model with application to hospital facilities. Oper.
hospitals. All of these methods are based on the statistical
Res. 29:1–22, 1981. doi:10.1287/opre.29.1.1.
analysis of patients’ arrivals and LOS. Using the average 15. Cochran, J. K., and Bharti, A., A multi-stage stochastic method-
number of patients expected in a year and average LOS to ology for whole hospital bed planning under peak loading. Int. J.
calculate the capacity needed is mathematically incorrect Ind. Syst. Eng. 1:8–36, 2006. doi:10.1504/IJISE.2006.009048.
16. Cochran, J. K., and Bharti, A., Stochastic bed balancing of an
because of the variation due to unpredictable nature of
obstetrics hospital. Health Care Manage. Syst. 9:31–45, 2006.
NICU admission rates and LOS. Therefore, one way of doi:10.1007/s10729-006-6278-6.
determining the number of patients in the NICU over time 17. Kokangul, A., A combination of deterministic and stochastic
is construct a simulation model, using the obtained approaches to optimize bed capacity in a hospital unit. Comput.
Methods Programs Biomed. 90:56–65, 2008. doi:10.1016/j.
distributions for arrivals and LOS. In this study the arrivals
cmpb.2008.01.001.
and LOS distribution is obtained not only for each level of 18. Costa, A. X., Ridley, S. A., Shahani, A. K., Harper, P. R., and De
the NICU, but also for gestation ages. These obtained Senna, V., Mathematical modeling and simulation for planning

8. J Med Syst
critical care capacity. Anaesthesia. 58:320–327, 2003. 27. Darzi, E. E., Vasilakis, C., Chaussalet, T., and Millard, P. H., A
doi:10.1046/j.1365-2044.2003.03042.x. simulation modeling approach to evaluating length of stay, occupan-
19. http://www.statsoft.com cy, emptiness and bed blocking in a hospital geriatric department.
20. http://www.mathworks.com Health Care Manage. Sci. 1:143–149, 1998. doi:10.1023/
21. Gorunescu, F., Mclean, S. I., and Millard, P. H., Using a queuing A:1019054921219.
model to help plan bed allocation in a department of geriatric 28. Navarro, A., and Thompson, W. A., Analysis of bed usage and
medicine. Health Care Manage. Sci. 5:307–312, 2002. occupancy following the introduction of geriatric rehabilitative
doi:10.1023/A:1020342509099. care in a hospital in Huesca, Spain. Health Care Manage. Sci.
22. Groothuis, S., Hasman, A., Pol, P. E. J., Lencer, N. H. M. K., 4:63–66, 2001. doi:10.1023/A:1009657817365.
Janssen, J. J. J., Jans, J., et al., Predicting capacities required in 29. Utley, M., Gallivan, S., Davis, K., Daniel, P., Reeves, P., and
cardiology units for heart failure patients via simulation. Comput. Worrall, J., Estimating bed requirement for an intermediate care
Methods Programs Biomed. 74:129–141, 2004. doi:10.1016/ facility. Eur. J. Oper. Res. 150:92–100, 2003. doi:10.1016/S0377-
S0169-2607(03)00080-4. 2217(02)00788-9.
23. Millard, P. H., Mackay, M., Vasilakis, C., and Christodoulou, G., 30. Turkish Neonatology Association., Türkiye’de Yenidoğan Bakım
Measuring and modeling surgical bed usage. Ann. R. Coll. Surg. Ünitelerinde Mortalite. Turk Neonatoloji Dernegi Bulteni. 7:7–12,
Engl. 82:75–82, 2000. 2003.
24. Montgomery, D. C., Runger, G. C., and Hubele, N. F., Engineering 31. Shahani, A. K., Korve, N., Jones, K. P., and Paynton, D. J.,
statistics, 3rd edition, p. 461, Wiley, New York, 2004. Towards an operational model for preventing and treatment of
25. Ridge, J. C., Jones, S. K., Nielsen, M. S., and Shahani, A. K., asthma attacks. J. Oper. Res. Soc. 45:916–926, 1994.
Capacity planning for intensive care units. Eur. J. Oper. Res. 32. Tu, J. V., Mazer, C. D., Levinton, C., Armstrong, P. V., and Naylor,
105:346–355, 1998. doi:10.1016/S0377-2217(97)00240-3. C. D., A predictive index for length of stay in the intensive care unit
26. Utley, M., Gallivan, S., Treasure, T., and Valencia, O., Analytical following cardiac surgery. CMAJ. 151 (2)177–185, 1994.
methods for calculating the capacity required to operate an 33. Rawlings, J. S., Smith, F. R., and Garcia, J., Expected duration of
effective booked admissions policy for elective inpatient service. hospital stay of low birth weight infants: Graphic depiction in
Health Care Manage. Sci. 6:97–104, 2003. doi:10.1023/ relation to birth eight and gestational age. J. Pediatr. 123(2):307–
A:1023333002675. 309, 1993. doi:10.1016/S0022-3476(05)81708-1.