Indoor Air 2010; 20: 284–297  2010 John Wiley & Sons A/S

www.blackwellpublishing.com/ina
Printed in Singapore. All rights reserved INDOOR AIR
doi:10.1111/j.1600-0668.2010.00653.x

Removal of exhaled particles by ventilation and deposition in a

multibed airborne infection isolation room

Abstract Removal of airborne particles in airborne infection isolation rooms is H. Qian1,2, Y. Li,1
important for infection control of airborne diseases. Previous studies showed
1
that the downward ventilation recommended by Centers for Disease Control and Department of Mechanical Engineering, The University
Prevention (CDC) could not produce the expected ÔlaminarÕ flow for pushing of Hong Kong, Pokfulam, Hong Kong SAR, China,
2
School of Energy and Environment, Southeast
down respiratory gaseous contaminants and removing them via floor-level
University, Nanjing, China
exhausts. Instead, upper-level exhausts were more efficient in removing gaseous
contaminants because of upward body plumes. The conventional wisdom in the
current CDC-recommended design is that floor-level exhausts may efficiently Key words: Airborne infection; Isolation room; Large
remove large droplets/particles, but such a hypothesis has not been proven. We droplets; Ventilation.
investigated the fate of respiratory particles in a full-scale six-bed isolation room Dr H. Qian
with exhausts at different locations by both experimental and computational School of Energy and Environment
studies. Breathing thermal manikins were used to simulate patients, and both Southeast University
gaseous and large particles were used to simulate the expelled fine droplet nuclei 2 Sipailou Road, Nanjing
and large droplets. Gaseous and fine particles were found to be removed more Jiangsu 210096
China
efficiently by ceiling-level exhausts than by floor-level exhausts. Large particles
Tel.: +86 13645186001
were mainly removed by deposition rather than by ventilation. Our results show Fax: +86 2583792722
that the existing isolation room ventilation design is not effective in removing e-mail: keenwa@gmail.com
both fine and large respiratory particles. An improved ventilation design is hence
recommended. Received for review 14 August 2009. Accepted for
publication 4 March 2010.

Practical Implications
Our findings of the relatively poor performance of fine-particle removal by the existing CDC design of isolation room
ventilation suggests a need for improvement, and the findings of the removal of large particles by deposition, not by
ventilation, suggest that floor-level exhausts are unnecessary, and that regular surface cleaning and disinfection is
necessary, thus providing evidence for maintaining isolation room surface hygiene.

airflow patterns, the air supply and exhaust should be

Introduction
Ventilation in isolation rooms is an important envi-
ronmental control method for infection control.
Airborne infection isolation (AII) rooms are recom-
mended for infected patients with airborne diseases
such as TB, measles and chickenpox (AIA, 2001;
ASHRAE, 2003; CDC, 1994, 2005) and opportunistic
airborne diseases such as SARS and influenza (WHO,
2007).
The most recommended airflow pattern design is
shown in Figure 1. The principle of the current AII
room ventilation is to control the direction of airflow
to prevent contamination of air in areas adjacent to the
infectious source (CDC, 1994, 2005). For ÔoptimalÕ
located such that clean air first moves to parts of the
room where health care workers (HCWs) are likely to
work, and then flows across the infectious source and
finally into the exhaust. The conventional wisdom in
the current CDC-recommended design is that floor-
level exhausts may efficiently remove large droplets/
particles exhaled from patients or produced by clinical
procedures. To the best of our knowledge, the recom-
mendation of such an airflow pattern remains just a
hypothesis, and it has not been proved. There has not
been a systematic study of the effectiveness of this
CDC-recommended design.
Reported nosocomial SARS outbreaks in hospitals
worldwide have reminded us of the need to create a

284

To aus
ex
ile t
h
t
Flo
or
lev
el
ex
hau
st
Fig. 1 An illustration of the currently recommended ventilation design for airborne infection isolation rooms (WHO, 2007)
safe and healthy environment in hospitals. HCWs can
be at a high risk of infection during an outbreak of
airborne or droplet-borne diseases. During the 2003
SARS epidemic, over 20% of the infected individuals
were HCWs (WHO, 2003). The question of developing
an efficient air distribution system for isolation rooms
is equivalent to asking how to remove simultaneously
the fine droplet nuclei and large particles. Fine droplet
nuclei are believed to be the transmission vector for
airborne diseases, while large droplets are responsible
in droplet-borne diseases. There have been many
studies on the performance of ventilation systems in
isolation rooms since the SARS outbreak, such as
evaluation of the ventilation performance in isolation
rooms by field measurement (Li et al., 2007; Saravia
et al., 2007), numerical investigation of ventilation
performance for removing contaminants (Kao and
Yang, 2006; Lai and Cheng, 2007; Shih et al., 2007)
and laboratory experimental studies (Qian et al., 2006;
Tung et al., 2009). Our previous studies showed that
the so-called downward ventilation could not produce
a ÔlaminarÕ flow as expected to push down respiratory
gaseous contaminants to floor level and remove them
from floor-level exhausts (Qian et al., 2006). The
interaction between upward body plumes and down-
ward supply air streams creates a mixing flow pattern
in the room. The upper-level exhausts were shown to
be efficient in removing gaseous contaminants when the
upward body plumes were considered (Qian et al.,
2008).
The purpose of this study is to investigate in detail
the fate of respiratory particles in a full-scale six-bed
isolation room using both experimental and computa-
Airborne infection isolation room ventilation
Toilet
Isolation room
Anteroom
Transfer grille
Corridor
tional methods. Three systems of exhaust location are
examined, i.e. floor exhausts only (the same as in the
existing CDC design), ceiling exhausts only, and a
combination of both floor and ceiling exhausts. In the
latter system, we only consider an exhaust air ratio of
50:50%. The effectiveness of such a combined system is
clearly demonstrated later without investigating other
flow rate ratios. The conclusion from our multibed
study should also apply to single-bed isolation rooms.
A number of excellent related studies exist. These
include the study of dispersion of droplets in a room by
Chao and Wan (2006), Chao et al. (2008), understand-
ing of body plumes by Zhu et al. (2006) and Gao and
Niu (2007), as well as the movement and removal of
particles in buildings (Bouilly et al., 2005; Lai, 2002;
Nazaroff, 2004).
Methdologies
Experimental design
Experiments were conducted in a full-scale AII test
room in the University of Hong Kong; see Figure 2.
There were six beds, numbered from 1 to 6. Nine
supply diffusers were installed in the ceiling, also
numbered from 1 to 9. Exhausts were available at the
ceiling and floor level. The airflow rate for each exhaust
could be adjusted. A typical downward ventilation
system (CDC, 1994) could be obtained when only the
floor-level exhausts were open. To investigate the
ventilation efficiencies of exhausts at different heights,
the ratio of ventilation rate of exhausts between ceiling
and floor level were set to be 100%:0, 50:50%, and
285
Qian & Li

(a)

(b)

1 4

1 7 4

Column 2

2 5
2 8 5

Column 1

3 6

VSD
3 9 6
VSD

Fig. 2 (a) Illustration of the setup of the six-bed isolation room. The two levels of exhausts are shown. When the floor-level exhaust is
used, the system is referred to as the Centers for Disease Control and Prevention-recommended ceiling supply and floor return design.
(b) The floor plan showing Column 1 and Column 2 for placing the slides. All beds are numbered from 1 to 6 at the pillow location,
while all supply diffusers are also numbered from 1 to 9

0:100%, respectively. Three beds were placed along
both the left- and right-hand sides of the room. The
distance between beds at one side was 1 m. Two
vertical columns were located in the room for placing
particle and temperature sensors. The supply ventila-
tion rate was 12 air changes per hour (ACH), while the
exhaust was 13.2 ACH to maintain a negative pressure
in the room.
In order to reach the steady state for each test, we
waited for at least 4 h after switching on the ventilation
system, and/or changing the heat sources including the
lights and manikins. Six thermal manikins were used to
mimic patients, whose heat generation was set to be
76W, corresponding to a person at rest. During our
experiments, the tracer gas SF6 and particles were
released through the exhaled air of the source manikin
using a specially designed artificial lung with an airflow
rate of 6 l/min, and exhaled air temperature was
controlled at 32 ± 1.5C (Brohus, 1997). The Arizona
Test Dust (A4 Coarse) was used to simulate exhaled
particles. The mode particle size was 80 lm. Six
samplers for the tracer gas were available. One sampler
was located in the exhaust tube and another five
samplers were located in the mouth of each receiving
thermal manikin.
The particles were measured by two methods. The
first was a 16-channel dust monitor (Grimm model
1.108; Grimm Aerosol Technik GmbH & Co. KG,
Ainring, Germany), which could provide real-time size
measurements of particles from 0.5 to 20 lm. The
sampling flow rate was 1.2 l/min. The second method
was to count and measure under a microscope the
particles deposited by impaction and settling on glass
slides placed in the air. Each slide was scanned under a
microscope (Leica DM1000; Leica Microsystems,
Wetzlar, Germany) and photographed with a high
resolution CCD camera (Leica DFC320; Leica Micro-
systems) connected to the microscope. The sizes of
deposited particles were analyzed using image process-
ing software developed in our laboratory (Sun et al.,
2005). The Grimm dust monitor was calibrated by the
manufacturer before the experiments, and the slide and
microscope method was evaluated in our laboratory
with particles of known sizes.
For the slide and microscope method, the airborne
particle concentration may not be fully revealed by the

286

number of deposited particles, because of the variation
in deposition rates of different-sized particles and the
difficulty in completely removing the ultra-fine particles
by cleaning the slide. This may be shown by two photos
of deposited particles taken by the microscope in the
room with and without the particle source (not shown
here). Other possible sources of errors include the
particles overlying each other or small particles joining
up to form a larger one. We counted only the particles
larger than 3 lm.
Numerical simulation
Computational fluid dynamics (CFD) was used to
investigate the airflow pattern and particle dispersion.
The turbulence effect was considered by using the
Re-Normalization Group k- model (Ferziger and
Peric, 2002). The second-order upwind scheme was
used for discretizing the convection terms.
The air velocity from the supply diffuser was set the
same as the average velocity in the experiment,
obtained by dividing the airflow rate by the diffuser
area. The geometry of the simulated isolation room
was the same as the experimental ward. All walls were
well insulated in the experiment and assumed to be
adiabatic in CFD simulations. Non-slip boundary
conditions are applied to all walls. We assumed that
half the heat from the manikins was transferred
through convection (Brohus, 1997). The particle tra-
jectory was calculated by a Lagrangian method. To
simplify the calculation, the following assumptions
were made, i.e. the heat and mass exchange between air
and particles were neglected, and the influence of
particles on airflow was also neglected.
The Lagrangian method calculates the trajectory of
each particle by solving the following equation:
! ! ! !
du p
¼ Fdrag þ Fg þ Fa ð1Þ
dt
! !
!Fdrag is the drag force, Fg is the gravity force,
where
and Fa is the additional force, which includes the
pressure force, virtual mass force, Basset force, Brown-
ian force, and SaffmanÕs lift force. The pressure force
and virtual mass force are sufficiently small and so they
are ignored (Zhao et al., 2004).
The turbulent dispersion of particles was predicted
by integrating the trajectory equations for individual
particles using instantaneous fluid velocity,
u ¼ u þ u0 ðtÞ. The random effects of turbulence on
particle dispersion are modeled by the discrete random
walk method. The values of u¢, v¢, w¢ are assumed to
obey a Gaussian probability distribution, and the
fluctuating components are assumed to be isotropic.
The fluctuating velocity components can be obtained
as follows:
Airborne infection isolation room ventilation
r
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
0 0 0
pffiffiffiffiffiffiffiffiffiffi
u ¼v ¼w ¼1 u02 þ v02 þ w02 =3 ¼ 1 2k=3 ð2Þ
where k is the kinetic turbulent energy and V is the
normally distributed number.
Different boundary conditions of particles are con-
sidered including escape at exhaust outlets, reflection
on vertical walls and ceiling, and entrapment on the
floor, skin of inpatient and beds.
A commercial CFD software package, Fluent 6.2,
with user-defined functions was used for numerical
simulation. CFD simulations were carried out as
follows. First, we obtained the steady solutions for the
airflow without particles. The particles were then
injected into the ward through the mouth of the source
manikin at a rate of 192 particles per second for each
particle size. Particles were injected for a period of 800 s,
and a total of 153,600 particles were injected into the
isolation room during this period for each particle size.
Susceptible exposure index
We define an exposure index for the susceptible
receptor to quantify the relative intake of substances
exhaled by the source patient or the infectious host
(Nazaroff et al., 1998). The susceptible exposure index
is defined as e ¼ CCri C
Cs , where Cr, Ci, and Cs are
s
concentrations at the exhaust (return), in the inhaled
air of the susceptible (i.e. receiving individual), and at
supply, respectively. When steady state conditions are
considered in our experiments, the average concentra-
tions can be used. The tracer gas (SF6) concentration of
supply air was zero (0), thus e ¼ CCri . Note that our
defined susceptible exposure index  is the reciprocal of
the personal exposure index defined in Brohus (1997).
A high susceptible exposure index  also means a high
exposure by the susceptible to the airborne substances
exhaled by the infected.
Results and discussion
Airflow patterns in the isolation room
Figure 3 shows the predicted air velocity field in a
plane through patient 2 and patient 5 in the test room
(see Figure 2). The main flow characteristics are
dominated by the downward negative buoyancy flow
through the nine diffusers, agreeing with smoke visu-
alization results (not shown). The supply air velocity is
sufficiently low so that negative buoyancy dominates
the supply air stream. The supply air stream becomes a
negative thermal plume, instead of a negatively buoy-
ant jet. The two-side diffuser air streams tend to bend
toward the heads of the manikins. This is possibly
because of the rising manikin thermal plumes. The
plumes rise to the ceiling level, bend at the wall corner,
and then join the supply air stream. When the exhausts
287
Qian & Li

(a)
1 m/s

(b)

1 m/s

(c)

Fig. 3 Predicted airflow pattern in the isolation room in the plane of patient 2 (left) and patient 5 (right). (a) Floor-level exhausts;
(b) ceiling-level exhausts, and (c) both levels of exhausts with airflow rate ratio of 50:50%

are located at the lower level (Figure 3a), we have the
traditional CDC isolation room design. The supplied
diffuser air streams fall vertically to floor level and
leave the room through the floor-level exhausts.
Although there is an overall downward flow pattern
in the corridor area, the airflow pattern in the vicinity
of the beds is generally mixed. This is because of the
interaction between the plume generated by the ther-
mal manikin and the air stream from the supply
diffuser.
The plume generated by the thermal manikin is
above the torso and head. When the exhausts are
located at the upper level (Figure 3b), part of the rising
plumes go directly to the exhausts and part is entrained
into the supply air stream. When the exhausts are
located at both upper and lower levels (Figure 3c), part
of the rising plumes go directly to the exhausts at the
upper level, but the airflow rate through the upper
exhaust is lower than that in Figure 3b. The downward
middle supply air stream bends towards the left in
Figure 3a,c, but towards the right in Figure 3b. The
boundary conditions are symmetrical, and the resulting
asymmetrical flows may be because of the nature of the
complex flows. The airflow is mixed, but not fully
mixed, as there is no room-scale recirculation. The
upward plume will help to bring gaseous pollutants to
ceiling level and a ceiling exhaust will improve the
removal efficiency.

288

Measured susceptible exposure index for removing gaseous pollutants
Table 1 summarizes the measured susceptible expo-
sure index of the receiving manikins for the gaseous
pollutants. In a fully mixed room, the pollutant
concentration everywhere in the room should be the
same as in the exhaust. Thus, the measured suscep-
tible exposure index in a fully mixed room should be
unity (1). The measured susceptible exposure index is
between 0.24 and 4.35. The air in the six-bed ward
here is not as well mixed as in the two-bed
experimental ward (Qian et al., 2008). The position
of the source patient significantly influences the
susceptible exposure index. Generally, the susceptible
exposure index is lower when the distance between
the receiving and source patient is greater. For the
last row of results in Table 1, bed 4 is directly facing
the source patient; hence, the susceptible exposure
index is 4.35 at bed 4, suggesting a very high
exposure level.
Comparatively, the overall susceptible exposure
index for the gaseous pollutants is the lowest when
the exhausts are at ceiling level. For the two ceiling
exhaust results shown in Table 1, the measured
susceptible exposure index is mostly much lower
than unity (1), except for those beds close to the
index (source) patient bed. This result suggests
that the ventilation removal effectiveness in this
ventilation design is better than the fully mixed
design using the momentum jet principle. The result
may exclude the need to study other types of
ventilation systems based on the mixing principle
for isolation rooms.
The results are also consistent with the conclusions
of our previous study (Qian et al., 2008). The upward
flow plume appears to play a significant role in
transporting the exhaled pollutants upwards so that
they can be removed through the ceiling-level exhausts.
When the airflow rate is 50% through the ceiling-level
exhausts and 50% through the floor-level exhausts, the
susceptible exposure index is generally less than that
with only floor-level exhausts and greater than that
with only ceiling-level exhausts.
Table 1 Measured susceptible exposure index of each patient of the tracer gas in a full-
scale 6-bed test isolation room when the index patient (the source) is in bed 5 or 6. The
index patient (the source) is facing upward if not stated otherwise
Exhaust location Bed 1 Bed 2 Bed 3
Floor level only 0.90 0.86 0.89
Ceiling level only 0.24 0.28 0.52
Both floor and ceiling levels 0.79 0.68 0.61
Floor level only 1.01 0.85 0.65
Ceiling level only 0.76 0.60 0.45
Both floor and ceiling levels 0.97 0.81 0.63
Both floor and ceiling levels* 1.28 0.91 0.70
*Index patient facing bed 4.
Airborne infection isolation room ventilation
Measured particle distribution
The preceding results were for gaseous pollutants,
which can be a model for fine droplet nuclei with a
diameter of <10 lm. Our further work focuses on
particles larger than 10 lm. The concept of D90 is
used here to show the particle-size distributions,
measured by the slide and microscopy method. In a
particular size distribution, 90% of the total number of
particles has a diameter less than the value of D90.
Figure 4 shows the measured D90 (only particles
larger than 3 lm are considered in the calculation as
discussed earlier) and the measured number of parti-
cles at different locations in the ward with different
exhaust locations. The number of particles at each bed
is the measured value at the manikin head location.
Each setting is referred to as (exhaust location; source
patient bed number) where Ôexhaust locationÕ refers to
floor, ceiling or both levels of exhausts of ratio
50:50%, and Ôsource patient bed numberÕ refers to
bed 5 or 6.
The distribution of particles is found to be influenced
by the distance from the particle source. The particles
are found to be larger at locations close to the source.
Smaller particles are easily influenced by airflow and
can travel a long distance, while large particles tend to
deposit on surfaces close to the source.
The difference in D90 and the total number of
particles along each vertical column is not as obvious
as that at two different horizontal locations, as the two
columns are mostly further apart in Figure 4 for each
setting. The results measured by the particle counter
(Grimm model 1.108) also indicate that particle sizes
do not differ at different heights of 0.5 and 2 m from
the floor for different settings (data not shown).
The trends in the results measured by Grimm and
those measured by microscope are similar, but the
values are different. The D90 measured by Grimm is
between 7.5 and 10 lm, while the D90 measured by
slides and microscope is between 10 and 23 lm. There
are two possible explanations. The first is that in the
slide and microscope method, only particles larger
than 3 lm are counted. The second is that large
particles with a large settling velocity are more easily
deposited on the slide, while fine particles with a low
settling velocity easily follow the airflow and bypass
the slide, suggesting that the slide and microscope
method may not measure all the fine particles. The
Bed 4 Bed 5 Bed 6 large particles deposited on the slide may overlie
smaller particles. Some small particles may join
0.83 2.08 Source
0.26 1.22 Source together and become a single larger one, although
1.16 2.00 Source efforts were made to identify this situation in our
2.00 Source 0.92 imaging processing. Thus, the number of large parti-
1.92 Source 0.57
cles may have been overestimated, and the number of
2.78 Source 0.79
4.35 Source 0.42 small particles may have been underestimated. How-
ever, the measurements by both the Grimm and
microscope methods suggest a greater influence by
289
Qian & Li

2.5 2.5
Column 1 Column 1
Column 2 Column 2
2.0 2.0
4521 5508

Height (m)

Height (m)
1.5 1.5 C2
4785 9472
C1
13 14 4482 Source
1.0 1.0 VSD

C2
VSD

13 23
0.5 C1 0.5
13 Source
VSD
VSD

0.0 0.0
11 12 13 14 15 16 17 18 19 20 21 22 23 0 2000 4000 6000 8000 10,000
Diameter of particles D90 Particle numbers
(a) [Floor exhaust; Source bed 6] (b) [Floor exhaust; Source bed 6]

2.5 2.5
Column 1
Column 2 Column 1
2.0 2.0 Column 2

Height (m)
Height (m)

1.5 1.5
1854 2922
C2
10 13
1.0 1.0 1440 4316
C2
16 C1
13 Source
C1 3197
0.5 0.5
VSD
VSD

18 Source
VSD
VSD

0.0 0.0
8 10 12 14 16 18 20 22 0 2000 4000 6000 8000 10,000
Diameter of particles D90 Particle numbers
(c) [Ceiling exhaust; Source bed 6] (d) [Ceiling exhaust; Source bed 6]

2.5 2.5
Column 1
Column 1 Column 2
2.0 Column 2 2.0
Height (m)
Height (m)

1.5 1.5

4106 6991
1.0 15 17 1.0 C2
C2 4985 11827
15 19 C1
0.5 0.5 7821 Source
C1 VSD
VSD

18 Source
VSD
VSD

0.0 0.0
8 10 12 14 16 18 20 22 0 5000 10,000 15,000 20,000 25,000
Diameter of particles D90 Particle numbers
(e) [Both levels exhaust; Source bed 6] (f) [Both levels exhaust; Source bed 6]

Fig. 4 The vertical distribution of the measured D90 of particles (a, c, and e) and the vertical distribution of the measured particle
number (b, d, and f) along two columns shown in Figure 2(b) at different settings. The numbers shown in a bed are the corresponding
values in a position close to the manikin head in the bed. Each setting is referred to as (exhaust location; source patient bed number)
where Ôexhaust locationÕ refers to floor, ceiling, or both levels of exhausts of ratio 50:50%, and Ôsource patient bed numberÕ refers to bed
5 or 6

the horizontal distance from the source and lesser the source patient was facing upwards. These were
influence by the difference in height. calculated by adding up the total particles removed or
It should be noted that the absolute values (D90 or suspended by each mechanism after the particles were
the particle number) shown in Figure 4 cannot be released continuously for 800 s.
compared between two different experimental settings Looking at the last column, which shows the
because of possible differences in the particle source percentage of particles suspended in the air, 31.5% of
strength. It was not possible to maintain a constant the 1-lm particles remain in suspension in the air in the
particle release rate during our experiments. system with only floor-level exhausts when the source
patient is in bed 5, compared with only 5.2% for the
system with only ceiling-level exhausts. The ceiling-
Predicted fate of particles using CFD
level exhausts are clearly shown to remove fine particles
Table 2 summarizes the percentage of particles sus- (<20 lm) more efficiently than the floor-level
pended in the air and those removed by two major exhausts. This is because of the upward plume as
mechanisms, i.e. by ventilation through the exhaust previously discussed. For particles larger than 30 lm,
outlet(s) and by deposition on different surfaces, when both the floor- and ceiling-level exhausts performed

290
 Airborne infection isolation room ventilation

Table 2 The fate of particles released by a source patient in bed 5 or 6 (measured by the percentage of the total released particles), including deposition on skin, beds, and the floor, and
ventilation removal by exhausts at the ceiling and/or floor level, and suspended in the room air. The source patient is facing upward

Receiving Source
Source Diameter manikins manikin All Ceiling- level Floor-level Suspended
patient Exhaust (lm) skin (%) skin (%) beds (%) Floor (%) exhausts (%) exhausts (%) in air (%)

Bed 5 Floor-level exhausts 1 0.8 2.0 10.1 26.3 – 29.4 31.5

5 0.8 1.9 9.8 27.6 – 28.9 31.0
10 0.6 1.8 8.8 33.0 – 19.4 36.4
20 0.2 2.9 7.3 51.8 – 19.4 18.4
30 0.1 11.6 14.7 50.5 – 11.9 11.2
40 0.0 32.7 27.5 28.3 – 4.7 6.8
50 0.0 75.4 15.4 4.9 – 0.5 3.8
Ceiling-level exhausts 1 0.1 0.8 1.3 2.0 90.4 – 5.2
5 0.1 0.8 1.4 2.2 90.4 – 5.1
10 0.2 0.8 1.2 2.5 90.4 – 4.9
20 0.3 0.9 1.3 4.6 88.9 – 4.0
30 0.8 2.4 2.3 4.5 86.0 – 4.0
40 1.0 9.6 5.0 2.3 77.4 – 4.7
50 0.3 38.7 11.4 1.9 41.6 – 6.1
Both-level exhausts 1 1.2 2.3 8.8 20.7 27.8 11.4 27.8
5 1.2 2.2 8.7 21.9 27.5 11.4 27.1
10 1.1 2.2 8.0 26.5 26.5 10.9 24.8
20 0.8 2.5 8.2 44.5 20.1 7.3 16.5
30 1.9 5.4 14.1 45.0 17.7 3.9 12.1
40 2.5 13.7 19.6 39.4 14.2 2.6 8.0
50 0.5 47.5 17.5 24.1 4.5 0.9 5.0
Bed 6 Floor-level exhausts 1 0.5 4.3 9.7 17.7 – 28.0 39.9
5 0.5 4.3 9.4 18.9 – 27.4 39.5
10 0.6 4.0 8.7 21.6 – 27.1 38.1
20 0.5 7.1 10.9 31.0 – 23.5 27.0
30 0.2 13.4 13.6 40.8 – 15.9 16.2
40 0.0 29.5 17.2 33.4 – 5.4 14.5
50 0.0 51.1 7.5 24.3 – 0.9 16.1
Ceiling-level exhausts 1 0.3 1.8 0.8 11.6 65.8 – 19.7
5 0.3 2.0 4.1 11.8 65.7 – 16.1
10 0.2 2.0 3.7 13.0 66.3 – 14.8
20 0.1 3.2 3.4 12.6 67.0 – 13.7
30 0.0 6.0 4.9 16.9 61.2 – 11.0
40 0.0 15.3 6.1 14.1 53.3 – 11.2
50 0.0 44.4 3.7 4.5 33.4 – 14.1
Both-level exhausts 1 0.5 3.6 5.6 9.7 49.5 7.8 23.5
5 0.5 3.5 5.4 10.2 49.4 7.3 23.8
10 0.5 3.9 5.2 11.7 49.0 6.7 23.0
20 0.7 6.1 7.9 13.9 46.8 5.2 19.4
30 0.9 11.4 12.4 13.0 43.6 3.0 15.6
40 0.6 22.3 14.0 8.5 38.1 1.4 15.1
50 0.1 49.5 9.7 2.4 20.5 1.1 16.8

similarly, with about 5–10% remaining in the air. It
can be seen later that most of these large airborne
particles were recently released into the air. The
performance of the system with exhausts at both levels
was surprisingly close to that of the system with floor-
level exhausts only.
It is interesting to examine the amount removed by
the exhaust air, i.e. by comparing the values in the
columns for ceiling-level exhausts and floor-level
exhausts in Table 2. Of the 1-lm particles, 90.4% are
removed by exhaust in the ceiling-level system, while
only 29.4% are removed by exhaust in the floor-level
system when the source patient is in bed 5.
For large particles, the number of particles removed
by deposition is much greater than by ventilation
through exhausts. Only 0.5% of the 50-lm particles are
removed by the exhaust in the floor-level system when
the source patient is in bed 5. In contrast, under the
same conditions, 41.6% of the 50-lm particles are
removed by the exhaust in the ceiling-level system.
More than 50% of the 50-lm particles are removed
by deposition on skin, beds, and floors in the ceiling-
level system. However, more than 90% of the 50-lm
particles are removed by deposition in the floor-level
system. The floor-level exhausts are not more efficient
in removing large particles than the ceiling-level
exhausts. The results strongly suggest that in the
existing CDC-recommended system, the large parti-
cles are removed mainly by deposition, not by venti-
lation. This disproves the common belief that the

291
Qian & Li

CDC-recommended system pushes down the large particles in the air is much lower than the number of
particles to floor level, so that they are removed by fine particles in each setting. A greater number of large
the floor-level exhaust. particles can be found at a height of around 1.0 m, as
Based on our discussion with the HCWs and the height of the exhaling mouth is 1.06 m when the
infection control experts, floor-level exhausts are also source patient faces upwards and 1.1 m when the
inconvenient because they are usually located close to source patient faces sidewards. Most of the suspended
the bed heads, where the space is valuable for clinical large particles are close to the exhaling mouth; see
purposes. Ceiling-level exhausts would free up this Figure 6.
space for clinical use.
Figure 5 shows the vertical distribution of the
Predicted dispersion of particles using CFD
predicted number of suspended particles in the entire
room. For each setting, exhausts at ceiling level seem to The above-observed pattern of the fate of particles is
be the most efficient in removing fine particles, while closely related to the particle dispersion pattern. In
exhausts at floor level seem to be the least efficient in Figure 6, we show the suspended particles after they
removing them. The number of suspended large are continuously released for 800 s, for both the ceiling

2.5 2.5

2.0 2.0
Height (m)

Height (m)
1.5 1.5

1.0
[Floor exhausts; Source bed 5]
[Ceiling exhausts; Source bed 5]
0.5 [Both exhausts; Source bed 5]
1.0
[Floor exhausts; Source bed 6]
0.5 [Ceiling exhausts; Source bed 6]
[Both exhausts; Source bed 6]

0.0 0.0
0 1000 2000 3000 4000 5000 6000 7000 0 1000 2000 3000 4000 5000 6000 7000
Particle number Particle number
(a) (b)

2.5 2.5

2.0 2.0
Height (m)

Height (m)

1.5 1.5

1.0
[Floor exhausts; Source bed 5]
[Ceiling exhausts; Source bed 5]
0.5 [Both exhausts; Source bed 5]
1.0
[Floor exhausts; Source bed 6]
0.5 [Ceiling exhausts; Source bed 6]
[Both exhausts; Source bed 6]

0.0 0.0
0 1000 2000 3000 4000 5000 6000 7000 0 1000 2000 3000 4000 5000 6000 7000
Particle number Particle number
(c) (d)

2.5 2.5

2.0 2.0
Height (m)

Height (m)

1.5 1.5

1.0 1.0

0.5 [Floor exhausts; Source bed 5] [Floor exhausts; Source bed 6]

[Ceiling exhausts; Source bed 5] 0.5
[Ceiling exhausts; Source bed 6]
[Both exhausts; Source bed 5] [Both exhausts; Source bed 6]
0.0 0.0
0 1000 2000 3000 4000 5000 6000 7000 0 1000 2000 3000 4000 5000 6000 7000
Particle number Particle number
(e) (f)

Fig. 5 The vertical distribution of the predicted number of particles of different sizes in the isolation room. The source patient is facing
upward. (a) Particle size is 1 lm and source patient in bed 5; (b) particle size is 1 lm and source patient in bed 6; (c) particle size is
10 lm and source patient in bed 5; (d) particle size is 10 lm and source patient in bed 6; (e) particle size is 50 lm and source patient in
bed 5; (f) particle size is 50 lm and source patient in bed 6. Each setting in the figure is referred to as (exhaust location; source patient
bed number) where Ôexhaust locationÕ refers to floor, ceiling, or both levels of exhausts of ratio 50:50%, and Ôsource patient bed
numberÕ refers to bed 5 or 6

292
 Airborne infection isolation room ventilation

(a1) dp = 1 mm (b1) dp = 1 mm

(a2) dp = 10 mm (b2) dp = 10 mm

(a3) dp = 20 mm (b3) dp = 20 mm

(a4) dp = 50 mm (b4) dp = 50 mm

Fig. 6 The birth time of suspended particles in the air after 800 s since the first particle was released. A birth time of 0 (zero) second
means that the particle was released 800 s ago. A birth time of 800 s means that the particle was just released. Hence, the exhalation jet
is seen as a ÔredÕ air stream from the source patient. The particles removed by either deposition on surfaces or ventilation through
exhausts are not shown. (a) Floor exhaust; source patient bed 5, facing upward; exhalation velocity 0.12 m/s. (b) Ceiling exhaust;
source patient bed 5, facing upward; exhalation velocity 0.12 m/s

and the floor exhausts. The source patient is in bed 5,
facing upward with an exhalation velocity of 0.12 m/s.
A low exhalation velocity suggests that the exhalation
air stream behaves as a thermal plume, which joins the
body plume. The particles were colored by the birth
time; that is the time at which the particles are released
to the ward. A birth time of 800 s means that the
particle has just been released; hence, the exhalation jet
can be seen as a ÔredÕ air stream from the source
patient. There are fewer suspended particles in the air
for the ceiling-level system (left-hand figures) than for
the floor-level system (right-hand figures). Small par-
ticles of 1–20 lm disperse throughout the entire room,
with smaller particles dispersing more evenly. Large
particles (e.g. 50 lm) stay closer to the source. Only the
very young large particles remain in the air in Figure 6,
suggesting that large particles are removed by surface
deposition.

293
Qian & Li

(a1) dp = 10 mm (a2) dp = 20 mm

(b1) dp = 10 mm (b2) dp = 20 mm

(c1) dp = 10 mm (c2) dp = 20 mm

(d1) dp = 10 mm (d2) dp = 20 mm

Fig. 7 The birth time of suspended particles in the air after 800 s since the first particle was released. (a) Both-levels exhaust; source
patient bed 5, facing upward; exhalation velocity 0.12 m/s. (b) Both-levels exhaust; source patient bed 6, facing upward; exhalation
velocity 0.12 m/s. (c) Floor-level exhaust; source patient bed 6, facing upward; exhalation velocity 0.12 m/s. (d) Ceiling-level exhaust;
source patient bed 6, facing upward; exhalation velocity 0.12 m/s

Figure 6 clearly demonstrates that the ceiling-level
exhausts are efficient in removing fine particles, while
the removal of large particles is relatively less depen-
dent on the ventilation system, as its major removal
mechanism is by surface deposition. Fine particles can
remain suspended in the air for a long time and be
transported a long distance, whereas large particles
cannot.
In Figure 7, the dispersion patterns for the 10- and
20-lm particles are compared for four other condi-
tions, i.e. with both levels of exhausts in Figure 7a,b,
and with the source patient in bed 6 for Figure 7b–d.
In all cases, the source patient is facing upward with a
low exhalation velocity of 0.12 m/s. The 20-lm parti-
cles have a tendency to stay close to the corner bed
source in Figure 7b–d.
The mode of exhalation and the position of the
source patient are also important. In Table 3, we
include two other situations when the source patient
is located in bed 6 and also faces sideways to bed 5.

294
 Airborne infection isolation room ventilation

Table 3 The fate of particles released by a source patient in bed 6 (measured by the percentage of the total released particles) at two exhalation velocities 0.12 or 2.00 m/s, including
deposition on skin, beds, and the floor, and ventilation removal by exhausts at the ceiling and/or floor level, and suspended in the room air. The source patient is facing bed 5

Exhaled Diameter Receiving Source manikin Ceiling-level Floor-level Suspended

Exhaust velocity (m/s) (lm) manikins skin (%) skin (%) All beds (%) Floor (%) exhausts (%) exhausts (%) in air (%)

Both-level 0.12 1 1.4 2.6 3.2 4.0 65.8 3.7 19.4

exhausts 5 1.5 2.7 3.5 4.5 64.0 4.1 19.8
10 1.7 4.7 4.7 5.7 56.6 5.2 21.4
20 1.7 19.8 11.4 8.4 40.5 4.8 13.4
30 0.3 57.7 13.2 7.7 14.8 1.5 4.8
40 0.0 71.7 18.3 2.7 0.2 3.2 4.0
50 0.0 26.4 68.1 1.4 0.0 0.2 3.9
2.00 1 6.1 1.4 7.5 11.9 33.8 10.4 28.9
5 6.0 1.1 7.7 12.9 34.7 10.9 26.7
10 2.2 0.8 4.8 18.1 34.7 10.6 28.7
20 7.2 0.3 12.0 31.2 18.9 11.4 18.9
30 9.8 0.3 19.6 35.7 15.0 6.3 13.3
40 13.3 1.1 30.6 32.4 8.9 3.8 9.8
50 9.6 1.3 56.5 21.7 3.6 4.4 2.8
Ceiling-level 0.12 1 0.6 2.0 2.8 2.1 79.6 – 13.0
exhausts 5 0.7 2.1 3.1 2.4 76.2 – 15.4
10 0.9 3.0 4.1 3.2 75.6 – 13.2
20 0.6 8.9 8.0 7.1 66.1 – 9.3
30 0.0 30.6 8.0 10.7 41.1 – 9.6
40 0.0 49.4 18.8 15.9 12.3 – 3.6
50 0.0 28.3 57.9 10.4 0.3 – 3.1
2.00 1 1.8 1.9 5.7 7.3 60.1 – 23.2
5 1.7 1.6 5.3 7.9 59.0 – 24.5
10 1.9 0.9 5.0 11.1 57.9 – 23.2
20 2.1 0.7 6.1 25.9 48.9 – 16.3
30 2.5 1.9 11.0 25.6 48.6 – 10.3
40 4.1 1.9 19.4 18.8 47.5 – 8.3
50 3.5 1.3 36.9 15.6 36.2 – 6.4
Floor-level 0.12 1 2.3 2.4 9.9 14.0 – 20.8 50.6
exhausts 5 2.2 2.4 9.0 14.8 – 20.5 51.1
10 2.3 2.9 8.9 21.4 – 22.7 41.7
20 1.9 6.8 14.0 33.8 – 22.3 21.2
30 0.2 18.4 23.0 36.7 – 12.5 9.1
40 0.0 33.6 41.6 13.3 – 8.4 3.0
50 0.0 19.6 77.3 0.4 – 1.7 1.0
2.00 1 2.8 2.7 9.9 15.6 – 22.3 46.7
5 2.7 2.6 9.5 16.6 – 22.4 46.2
10 3.1 2.4 9.0 21.2 – 23.9 40.5
20 4.8 4.4 16.6 24.6 – 23.6 26.1
30 6.6 10.2 28.5 21.3 – 17.1 16.2
40 6.4 15.3 36.3 12.8 – 17.7 11.5
50 1.9 5.7 43.2 7.7 – 36.8 4.7

As the source patient faces sideways, the exhalation
jet does not follow the upward body plume; see
Figure 8. When the exhaled air velocity is 0.12 m/s,
the exhaled jet can be captured by the body plume.
However, when the exhaled velocity is 2.0 m/s, the
exhaled jet can escape the plume capture. This
phenomenon results in a greater percentage of
particles suspended in the air for a high exhaled
velocity of 2.0 m/s than that for a low exhaled
velocity of 0.12 m/s.
Conclusions
We have presented a detailed experimental and numer-
ical study of the distribution of the exhaled particles
and gaseous pollutants in a full-scale experimental
isolation room with a downward supply and three
different exhaust designs. The existing CDC-recom-
mended ventilation design is not found to be effective
in removing both fine and large respiratory particles
for the purpose of infection control. For gaseous
pollutants and fine particles, the system with only
ceiling-level exhausts in the isolation room can remove
the pollutants and fine particles more efficiently than
the system with only floor-level exhausts and the
system with exhausts at both levels. For large particles,
deposition played a more important role than ventila-
tion in removing these. The floor-level exhausts did not
appear to be much more efficient in removing large
particles than the ceiling-level exhausts.
Comparatively, the ceiling-level exhausts are found
to be the most efficient in removing exhalation

295
Qian & Li

(a1) dp = 1 mm (a2) dp = 50 mm (d1) dp = 1 mm (d2) dp = 50 mm

(b1) dp = 1 mm (b2) dp = 50 mm (e1) dp = 1 mm (e2) dp = 50 mm

(c1) dp = 1 mm (c2) dp = 50 mm (f1) dp = 1 mm (f2) dp = 50 mm

Fig. 8 The birth time of suspended particles in the air after 800 s since the first particle was released. (a) Both-levels exhaust; source
patient bed 6, facing towards bed 5; exhalation velocity 0.12 m/s. (b) Both-levels exhaust; source patient bed 6, facing towards bed 5;
exhalation velocity 2.00 m/s. (c) Floor-level exhausts; source patient bed 6, facing towards bed 5; exhalation velocity 0.12 m/s.
(d) Floor-level exhausts; source patient bed 6, facing towards bed 5; exhalation velocity 2.00 m/s. (e) Ceiling-level exhausts; source
patient bed 6, facing towards bed 5; exhalation velocity 0.12 m/s. (f) Ceiling-level exhausts; source patient bed 6, facing towards bed 5;
exhalation velocity 2.00 m/s

pollutants. Hence, we recommend the use of a down-

ward ventilation supply and ceiling-level exhaust for
future isolation room design, subject to a possible field
study with epidemiological data. In a separate study,
the same conclusion is reached when the respiratory
droplets are studied, including the effect of evapora-
tion. Additionally, surface deposition as the major
removal mechanism of particles reveals the significance
of regular surface cleaning and disinfection, and
provides good evidence for maintaining isolation room
surface hygiene.
Acknowledgements
The work described was supported by the Research
Grants Council of the Hong Kong SAR Government
(Project No. HKU 7150/06E), Research Grants from
National Nature and Science Foundation of China
(NSFC) (Project No. 50808038) and Research Grants
from Jiangsu Fundamental Research Project (Project
No. BK2009289). The authors thank Dr Sun Hequan
for providing guidance in using the particle image
processing software.

References
AIA (2001) Guidelines for Design and Con-
struction of Hospital and Health Care
Facilities, Washington, D.C., American
Institute of Architects.
ASHRAE (2003) HVAC Design Manual for
Hospitals and Clinics, Atlanta, Ga.,
American Society of Heating Refrigerat-
ing and Air-Conditioning Engineers Inc.
Bouilly, J., Limam, K., Béghein, C. and
Allard, F. (2005) Effect of ventilation
strategies on particle decay rates indoors:
an experimental and modelling study,
Atmos. Environ., 39, 4885–4892.
Brohus, H. (1997) Personal Exposure to
Contaminant Sources in Ventilated Rooms,
PhD Thesis, Aalborg University,
Denmark.
CDC (1994) Guidelines for preventing trans-
mission of Mycobacterium tuberculosis in
health-care settings, Morb. Mortal. Wkly
Rep., 43 (RR-13):1–132, Centers for
Disease Control and Prevention.
CDC (2005) Guidelines for preventing the
transmission of Mycobacterium tubercu-
losis in health-care facilities, Morb.
Mortal. Wkly Rep., 54 (RR-17):1–142,
Centers for Disease Control and
Prevention.
Chao, C.Y.H. and Wan, M.P. (2006) A
study of the dispersion of expiratory
aerosols in unidirectional downward and
ceiling-return type airflows using a
multiphase approach, Indoor Air, 16,
296–312.
Chao, C.Y.H., Wan, M.P. and To, G.N.S.
(2008) Transport and removal of
expiratory droplets in hospital ward
environment, Aerosol Sci. Technol., 42,
377–394.

296

Ferziger, J.H. and Peric, M. (2002) Compu-
tational Methods for Fluid Dynamics,
Berlin, New York, Springer.
Gao, N.P. and Niu, J.L. (2007) Modeling
particle dispersion and deposition in
indoor environments, Atmos. Environ.,
41, 3862–3876.
Kao, P.H. and Yang, R.J. (2006) Virus dif-
fusion in isolation rooms, J. Hosp. Infect.,
62, 338–345.
Lai, A.C. (2002) Particle deposition indoors:
a review, Indoor Air, 12, 211–214.
Lai, A.C.K. and Cheng, Y.C. (2007) Study of
expiratory droplet dispersion and trans-
port using a new Eulerian modeling
approach, Atmos. Environ., 41, 7473–7484.
Li, Y., Ching, W.H., Qian, H., Yuen, P.L.,
Seto, W.H., Kwan, J.K., Leung, J.K.C.,
Leung, M. and Yu, S.C.T. (2007) An
evaluation of the ventilation performance
of new SARS isolation wards in nine
hospitals in Hong Kong, Indoor Built
Environ., 16, 400–410.
Nazaroff, W.W. (2004) Indoor particle
dynamics, Indoor Air, 14(Suppl. 7), 175–
183.
Airborne infection isolation room ventilation
Nazaroff, W.W., Nicas, M. and Miller, S.L.
(1998) Framework for evaluating mea-
sures to control nosocomial tuberculosis
transmission, Indoor Air, 8, 205–218.
Qian, H., Li, Y., Nielsen, P.V., Hyldgaard,
C.E., Wong, T.W. and Chwang, A.T.Y.
(2006) Dispersion of exhaled droplet
nuclei in a two-bed hospital ward with
three different ventilation systems, Indoor
Air, 16, 111–128(Correction, 16 (3): 256-
256; 2006).
Qian, H., Li, Y., Nielsen, P.V., Hyldgaard,
C.E. and Wong, T.W. (2008) Dispersion
of exhalation pollutants in a two-bed
hospital ward with a downward ventila-
tion system, Build. Environ., 43, 344–354.
Saravia, S.A., Raynor, P.C. and Streifel, A.J.
(2007) A performance assessment of air-
borne infection isolation rooms, Am. J.
Infect. Control, 35, 324–331.
Shih, Y.C., Chiu, C.C. and Wang, O. (2007)
Dynamic airflow simulation within an
isolation room, Build. Environ., 42, 3194–
3209.
Sun, H., Yu, L., Shen, Y., Yu, D. and Zhu,
G. (2005) Application of CDF(2,2) to
analysis of suspended sediment image,
China Particuo., 3, 204–207.
Tung, Y.C., Hu, S.C., Tsai, T.I. and Chang,
I.L. (2009) An experimental study on
ventilation efficiency of isolation room,
Build. Environ., 44, 271–279.
WHO (2003) Summary of Probable SARS
Cases With Onset of Illness From 1
November 2002 to 31 July 2003, Geneva,
World Health Organization.
WHO (2007) Infection Prevention and Control
of Epidemic- and Pandemic-prone Acute
Respiratory Diseases in Health Care –
WHO Interim Guidelines, Geneva, World
Health Organization.
Zhao, B., Zhang, Y., Li, X.T., Yang, X.D.
and Huang, D.T. (2004) Comparison of
indoor aerosol particle concentration and
deposition in different ventilated rooms
by numerical method, Build. Environ., 39,
1–8.
Zhu, S.W., Kato, S. and Yang, J.H. (2006)
Study on transport characteristics of sal-
iva droplets produced by coughing in a
calm indoor environment, Build. Environ.,
41, 1691–1702.
297