Intensive Care Medicine

, Volume 29, Issue 12, pp 2330–2335

Do different mattresses affect the quality of cardiopulmonary resuscitation?


    • Department of Intensive Care MedicineBirmingham Heartlands Hospital
  • Robert Benny
    • Leicester Warwick Medical SchoolUniversity of Warwick
  • Simon Giles
    • Department of Intensive Care MedicineBirmingham Heartlands Hospital
  • Fang Gao
    • Department of Intensive Care MedicineBirmingham Heartlands Hospital
  • Michael J. Tweed
    • Leicester Warwick Medical SchoolUniversity of Warwick

DOI: 10.1007/s00134-003-2014-6

Cite this article as:
Perkins, G.D., Benny, R., Giles, S. et al. Intensive Care Med (2003) 29: 2330. doi:10.1007/s00134-003-2014-6



To determine the effect of different mattresses on cardiopulmonary resuscitation performance and establish whether emergency deflation of an inflatable mattress improves the quality of resuscitation.

Design and setting

Randomised controlled cross-over trial performed in a general ICU


Critical care staff from a general ICU.


Cardiopulmonary resuscitation on a manikin on the floor or on a bed with a standard foam mattress and inflated and deflated pressure redistributing mattresses. Maximal compression force was measured at different bed heights.

Measurements and results

Compression depth, duty cycle and rate and percentage correct expired air ventilation were recorded on a manikin. Compression depth was significantly lower on the foam (35.2 mm), inflated (37.2 mm) and deflated mattress (39.1 mm) than the floor (44.2 mm). There were no clinically important differences in duty cycle or compression rate. The quality of ventilation was poor on all surfaces. Maximal compression force declined as bed height increased.


Resuscitation performance is adversely affected when performed on a bed (irrespective of mattress type) compared to the floor. There were no differences between the inflated and deflated mattresses, although the deflation process did not adversely affect performance. This study does not support the routine deflation of an inflated mattress during resuscitation and questions the potential benefits from using a backboard. The finding that bed height affects maximal compression forces, challenges the recommendation that cardiopulmonary resuscitation be performed with the bed at middle-thigh level and requires further investigation.


Cardiopulmonary resuscitationPressure ulcersSupport surfaceBasic life supportExternal chest compression


In hospital cardiac arrest is rarely an unheralded event. Many patients show signs of physiological deterioration prior to cardiac arrest [1]. These critically ill patients are also known to be at risk of developing pressure ulcers and their associated complications [2]. In an attempt to minimise this risk a high proportion are nursed on some form of pressure redistributing mattress. These surfaces commonly use air movement and pressure alterations in order reduce the risk of developing pressure-related ulceration. Current guidelines recommend that during a cardiac arrest, external chest compression (ECC) is performed on a firm surface [3]. In order to facilitate this most air-filled mattresses now incorporate a fast-deflate mechanism for use in cardiac arrest. However, the need for and effects of emergency deflation for cardiopulmonary resuscitation (CPR) have not previously been studied in detail.

The aim of this study was to examine the effect of foam and air-filled mattresses on CPR performance and to establish whether the emergency deflation mechanism provides any improvement or deterioration in the quality of CPR.


The study was conducted in the Department of Intensive Care Medicine at Birmingham Heartlands Hospital, Birmingham, United Kingdom, during June 2002. Twenty critical care physicians/nurses were recruited, all of whom had received in-house training and assessment in basic life support in the preceding 6 months according to current guidelines [3]. The study was considered by the local research and ethics committee.


Prior to starting the study participants were allowed a period of familiarisation with the resuscitation manikin and equipment with feedback monitors visible. During this time they were asked to demonstrate 15 correct consecutive chest compressions and two successful ventilations with the manikin placed on the floor and on a foam mattress. Participants were randomised to perform 75 s of dual operator CPR whilst kneeling perpendicular to a manikin placed on the floor and standing perpendicular to a manikin placed on a bed covered with the following surfaces: foam mattress, air-filled mattress whilst inflated, deflating and deflated. Ventilation was performed by one operator using a bag valve and mask whilst the other performed the ECC at a ratio of two ventilations to every 15 compressions in accordance with current guidelines [3]. Participants were blinded as far as possible by placing a sheet over the top of the mattress. Each participant performed ventilations and chest compressions on all surfaces and were allowed 5 min rest between each test and alternated between providing chest compressions and ventilations. We compared CPR variables for each mattress with the floor (gold standard) and the foam with the inflated mattress. The inflated and deflated mattresses and effects of deflation were also investigated.

In a second series of experiments we investigated the effect of body position on the maximum force that a CPR provider can exert during simulated chest compression. Participants were instructed to simulate a chest compression and exert the maximum force that they could on a pair of scales placed at different heights and body positions. These positions were: kneeling on the floor; standing with scales placed in the centre of a bed height adjusted to heights of 45 cm (lowest level for bed), 65 cm (average middle-thigh level of participants) and 85 cm (maximum bed height). The maximum compression force (kg) for each position, height, sex and weight were recorded for each participant. The coefficient of variation for this test model was less than 2%.


CPR was performed on a Rescusi Annie Modular System Skill Reporter (Laerdal Medical, Orpington, UK). Data were collected using the voice advisory messaging (VAM) software system onto a laptop computer (Dell) and downloaded to Microsoft Excel. The VAM system records ECC depth (mm), rate (min) and duty cycle (ratio of compression to relaxation phase, %) [4]. The coefficient of variation for this system is less than 3% for ventilations and less than 1% for chest compressions. Individual compression and ventilation are measured and recorded by the VAM system as an overall average for the 75-s test. In addition, a breakdown of the averages for the following time intervals were recorded: 0–30 s, 31–45 s, 46–60 s, 61–75 s. According to current guidelines chest compression depth should be 40–50 mm, duty cycle 50% and rate 100/min [3].

The percentage of correct ventilations was derived from the number of ventilations that achieved a tidal volume of 400–600 ml, with an inflation rate less than 1000 ml/s. Incorrect ventilations were categorised as inadequate (<400 ml), excessive (>600 ml) and/or too fast (inflation rate >1000 ml/s) in accordance with current international guidelines for ventilation with supplemental oxygen in cardiac arrest [3].

The manikin was placed on the floor or on a bed with a foam mattress (Softform, Medical Support Systems, Cardiff, UK) or inflatable pressure redistributing mattress (Nimbus III, Huntleigh Healthcare, Luton, UK). The manikin was weighted to 70 kg using sandbags along side the chest and abdomen to increase its weight to within the operating range for the mattresses. Care was taken to ensure that the sandbags did not interfere with CPR performance. The bed height was adjusted to the middle-thigh level of the participant providing compressions [5]. A standard 500-ml bag valve and mask (Laerdal Medical, Norway) was used for ventilation.

Statistical analysis

From our pilot study [6] we calculated that 20 measurements would be required to detect a 10% difference in the depth of ECC compression between inflated and deflated mattresses with a p value of 0.05 and 80% power. The data were stored in Microsoft Excel and analysed using SPSS 10 for Windows. The data were tested for normality using the Sharpiro-Wilk test. Normally distributed data are presented as mean ±SD and were analysed by repeated-measure analysis of variance using Bonferroni's correction for multiple comparisons. Non-normally distributed data are presented as median (interquartile range) and were analysed using Friedman's repeated-measure analysis of variance on ranks with Dunn's test for between subject comparisons. The χ2 test and Fisher's exact test were used for comparing categorical data. A p value less than 0.05 was considered statistically significant.


Floor vs. foam and inflated/deflated mattresses

Mean chest compression depth for the floor and mattresses are shown in Table 1. Chest compression depth was significantly less on all mattresses than on the floor (p<0.001). There was no difference between compression rate or duty cycle between the floor and mattresses. The number of attempted ventilations was similar although most were of inadequate tidal volume (<400 ml). There was no substantial difference in the proportion of "correct" ventilations.
Table 1.

Compression and ventilation parameters for different surfaces. Data are presented as mean (standard deviation) or median (interquartile range) for the total time for each test





Test duration (s)

76 (75–76)

76 (75–76)

76 (75–76)

76 (75–77)

Compression depth (mm)

44.2 (5.2)

35.2 (5.6)*

37.2 (6.3)*

39.1 (5.6)*

Compression rate (min)

121.8 (22.5)

124.7 (16.8)

126.5 (15.9)

125.7 (15.9)

Duty cycle (%)

45.5 (6.2)

47.6 (3.2)

44.6 (2.8)**

44.2 (4.1)

Number of ventilations during test

14 (12–16)

14 (14–15.5)

14 (12–14)

14 (12.5–16)

Number of shallow ventilations (<400 ml)

11 (7.3–12.8)

12 (4.5–14)

12 (6.3–14)

12 (10.3–14)

Number of deep ventilations (>600 ml)





Percentage of correct ventilations

12.1 (0–36.5)

0 (0–18.8)*

4.17 (0–25.6)

0 (0–26.56)

*p<0.05 vs. floor. **p<0.05 vs. foam mattress

Foam vs. inflated mattress

There was no difference in ECC rate or depth or percentage correct ventilations between these surfaces. The duty cycle was marginally superior on the foam mattress than on the inflated mattress (Table 1).

Inflated vs. deflated and deflating mattresses

There was no difference in ECC rate, depth, and duty cycle or percentage correct ventilations between the inflated and deflated mattress. The inflated mattress took on average 30 s (range 29–30) to deflate totally. During active deflation there was no significant difference in compression depth, rate, duty cycle or ventilation parameters (data not shown).

Compression parameters over time (75 s)

ECC depth declined over time on all surfaces (Fig. 1). There was no significant difference in the amount of deterioration in compression depth between surfaces over time compared to baseline (0–30 s). There were very small (although statistically significant) changes in duty cycle on the inflated (44.1±3.7% to 45.2±3.0%, p=0.001) and deflated mattress (42.7±0.9% to 44.4±0.9%, p=0.001) over time. There was no difference in chest compression rate.
Fig. 1.

ECC depth over time on the four surfaces

Effect of body position on maximum compression force

Simulated compression force was greatest when performed kneeling on the floor. It fell when participants stood to perform simulated compressions and as the height of the bed was increased (Fig. 2). We found moderate association between height and maximal force on the floor (r=0.60, p=0.005), at 45 cm (r=0.59, p=0.002), 65 cm (r=0.65, p=0.002) and 85 cm (0.67, p=0.001). Weight was significantly associated with compression force only with the bed at 65 cm (r=0.51, p=0.022) and 85 cm (r=0.60, p=0.005).
Fig. 2.

Simulated compression force with participant kneeling on the floor and standing with scales placed in the centre of a bed height adjusted the lowest level (45 cm), average middle-thigh level (65 cm) and maximum bed height (85 cm)


The principal findings of this study are that the performance of basic life support is reduced on a bed with a foam mattress, inflated or deflated pressure redistributing mattress when compared to the gold standard of the floor. There were no important differences in the measures of CPR efficacy between CPR performed on a foam mattress and on the inflated mattress. We found no evidence to support using a fast deflate mechanism to deflate an inflated mattress when performing CPR, although the deflation process did not interfere adversely with the provision of CPR.

Basic life support (ventilation and chest compression) is an integral part of CPR. ECC generates at best one-third of the normal cardiac output, coronary and cerebral blood flow [7]. Despite this the early initiation and quality of CPR have been shown to be important determinants of the success rate of resuscitation from cardiac arrest [8]. ECC rate [9], duty cycle (proportion of time spent in compression: relaxation) [10] and depth have all been shown to affect cardiac output during CPR. Following a comprehensive review of published studies the International Liaison Committee for Resuscitation [3] have recently produced recommendations for "optimal" chest compression techniques. These guidelines recommend that the depth of chest compressions should be 4–5 cm. Babbs et al. [11] studied the effect of chest compression depth delivered by a mechanical device in dogs. They found a linear relationship between compressions depth (in the range of 2.5–6 cm), cardiac output and mean arterial blood pressure. In this model a 1-cm reduction in compression depth was associated with a 50% relative reduction in cardiac output and 30% relative reduction in mean arterial blood pressure. In adult pigs Bellamy et al. [12] demonstrated improved global and coronary blood flow with increasing depth of chest compression in the range of 3.75–6.25 cm. In contrast to these findings Wik et al. [13] found that coronary blood flow fell when compression depth was increased from 4–5 cm. As flow appears to be dependent on compression depth until a certain limit is reached [14], one potential explanation for these conflicting finding were that Wik et al. used smaller pigs in which the 5-cm compression depth exceeded this "upper" threshold. In adult victims of cardiac arrest Ornato et al. [15] also found a linear relationship between compression force delivered by a CPR Tumper, end-tidal CO2 and systolic blood pressure. End-tidal CO2 accurately reflects cardiac output during CPR and is a useful predictor of survival [16]. The 10–20% (0.5–1 cm) reduction in chest compression depth between the floor and mattresses described in the present study would seem likely to cause a reduction in the cardiac output associated with CPR which could potentially have an adverse effect on outcome.

The reduced compression depth found with the mattresses compared to the floor could be explained by differences in the rigidity of the surface or the differences in body position adopted by the CPR provider (standing rather than kneeling on the floor). When the chest is compressed on a firm surface the distance the sternum is depressed is directly related to compression of the chest. However, when the chest is compressed on a soft surface such as a mattress, the distance that the sternum is depressed is also influenced by the amount of compression of the underlying mattress. A recent simulation study using mathematical models of CPR investigated the effects of mattress stiffness on sternal-spine displacement during simulated ECC [17]. The study compared two mathematical models of ECC: a constant force model (where a constant force is applied to the sternum) and a constant displacement model (where the sternum is displaced a constant distance). The constant force model allowed adequate compression on all but the softest of surfaces, at the expense of greatly increased work to the simulated rescuer on soft surfaces. In contrast, the constant displacement model led to substantial "under compression" on soft surfaces. The presence of a simulated backboard improved sternal-spine displacement with both models and reduced rescuer work with the constant force model.

Whilst we did not formally measure mattress stiffness, users in our pilot study perceived the inflated mattress as more unstable than the foam mattress [6]. However, we did not detect any difference in the present study in mean compression depth between the foam and inflated mattress or the deflated mattress (in effect the bed frame which is a rigid surface). This suggests that the presence of the manikin on a bed per se, compared to the floor, may itself impair the application of compressions. These findings are consistent with a recent study examining patterns of chest compressions which found an average reduction of 4 mm in compression depth in between CPR performed on the floor compared to a table [18]. This hypothesis is supported by our findings that maximal compression force declines when participants simulated CPR on a bed from a standing position compared to kneeling and simulating compressions on the floor. Compression force reduced further when the height of the bed was increased. This challenges the current recommendations that CPR is performed with the bed at middle-thigh level [5]. Our study suggests that maximal compression force is generated when the bed is as low as possible. Our findings also question the benefit that may be derived from the insertion of a backboard behind a patient that has arrested in a bed. Apart from the logistical difficulties of placing large patients or those attached to invasive monitors/pumps on such a device, we found no benefit on CPR performance from the presence of a rigid surface (deflated mattress), suggesting that there is little to gain from a backboard. We suggest that the effect of the CPR provider's body position is more important and requires further evaluation. In particular the use of a stool to raise a CPR provider's body position relative to the position of the simulated patient or the practice sometimes observed in which the CPR provider kneels on the bed next to the patient should be explored.

Overall the deleterious effects of performing CPR on a soft surface were less marked than would be expected from the constant displacement model described above. This suggests that the CPR providers compensated for the softer surface by increasing compression force with a corresponding increase in workload. The potentially adverse effects of this increased workload leading to fatigue and decay in performance [19] were not, however, evident in this study as compression depth declined uniformly whether CPR was performed on the floor, inflated or deflated surface. However, we studied CPR over only a relatively short period (75 s), and it is possible that performance would have decayed to a greater extent on the foam and inflated mattress had the test been continued for longer.

Recent evidence from studies in humans has shown that in the presence of supplemental oxygen tidal volumes of 400–600 ml should be sufficient to ventilate a patient in cardiac arrest [20]. When the inflated mattress is first deflated, it deflates in the patient's thoracic/abdominal region prior to complete deflation. We hypothesised that this differential deflation process promotes neck flexion and may impair ventilation. However, the quality of ventilation in the study was disappointingly poor and precluded us from undertaking any meaningful comparisons between mattresses. The finding of poor quality single operator bag-valve mask ventilation in similar studies using manikin models and humans has been previously reported [21, 22]. This raises question of the optimal modality for ventilating a patient in cardiac arrest prior to endotracheal intubation. Other studies have suggested that the laryngeal mask, cuffed oropharyngeal airway or pocket mask are superior alternatives [22, 23].

This study used a resuscitation manikin to simulate the application of CPR. Whilst this manikin does not provide the degree of variation in the chest wall compliance and elasticity found in the human thorax, it has the advantage of providing consistent conditions during experimental study [20]. It is increasingly difficult to undertake this kind of study in patients during CPR. Several other investigators [18, 24] have used this or similar systems for evaluating CPR and we believe it provides a reasonable surrogate to real CPR. It was not possible to blind participants when performing CPR on the floor and difficult when on a bed. We did attempt to blind participants from identifying the surface upon which they were undertaking CPR by placing a sheet over the mattress. Whilst we did not test the success of this blinding technique, participants in our pilot study were able to perceive differences in mattress stiffness, and therefore it is unlikely that our attempt at blinding was wholly successful. Whilst we cannot be certain that the probable failure of blinding did not introduce some bias, we consider it unlikely that knowledge of surface type influenced our participants CPR performance. Automatic electronic data collection ensured that knowledge of surface type could not bias the data collection. Finally, this study investigated CPR performed on a Nimbus III system mattress and Softform foam mattress. As the mechanical properties such as stiffness may vary between different mattresses, care is needed in extrapolating these findings to other mattress.


The quality of CPR is adversely affected when performed on a foam, inflated and deflated mattress compared to the floor through a reduction in compression depth. There was no difference in CPR performance between the inflated and deflated mattresses, although the deflation process did not adversely affect CPR quality. There appears to be little evidence to support the routine deflation of an inflated mattress in order to perform CPR. This finding also questions the potential benefit of inserting a backboard when performing CPR on a mattress. We found that maximal compression force declines with increasing bed height. These data suggest that compression depth is improved by positioning the bed as low as possible relative to the CPR provider. This challenges the current recommendation that the bed should be adjusted to the CPR provider's middle-thigh level.


We thank the Resuscitation Council UK for providing financial support for this study, Huntleigh Healthcare Ltd. for providing the pressure redistributing mattress and the staff at Birmingham Heartlands Hospital for their participation. We also acknowledge the technical advice and support provided by Dr. K.G. Morallee, Laerdal Medical (UK) in relation to the operation of the VAM system.

Copyright information

© Springer-Verlag 2003