UHM 2013, Vol. 40, No.

1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA

An analysis of the causes of compressed-gas diving fatalities

in Australia from 1972-2005
John Lippmann 1,2, Adrian Baddeley 3,4, Richard Vann 5, Douglas Walker 1

1 Divers Alert Network Asia-Pacific (AP), Ashburton, Victoria, Australia
2
James Cook University, Townsville, Ph.D. candidature
3 CSIRO Mathematics, Informatics and Statistics, Floreat, Western Australia

4 School of Mathematics & Statistics, University of Western Australia

5 Divers Alert Network, America; Anesthesiology Dept., Duke University, Durham, NC, USA

CORRESPONDING AUTHOR: John Lippmann– johnl@danasiapacific.org

_____________________________________________________________________________________________
ABSTRACT

In order to investigate causative factors, root cause with rough water, buoyancy trouble, equipment
analysis (RCA) was applied to 351 Australian com- trouble, and gas supply trouble. CAGE was asso-
pressed-gas diving fatalities from 1972-2005. Each ciated with gas supply trouble and ascent trouble,
case was described by four sequential events (trigger, while cardiac cases were associated with exertion,
disabling agent, disabling injury, cause of death) that cardiovascular disease, and greater age. Exertion was
were assessed for frequency, trends, and dive and diver more common in younger cardiac deaths than in
characteristics. The average age increased by 16 years, older deaths. Asphyxia became less common with
with women three years younger than men annually. increasing age. Equipment-related problems were
For the entire 34-year period, the principal disabling most common during the late 1980s and less so in
injuries were asphyxia (49%), cerebral arterial gas 2005. Buoyancy-related deaths usually involved loss
embolism (CAGE; 25%), and cardiac (19%). There of buoyancy on the surface but decreased when buoy-
was evidence of a long-term decline in the rate of ancy control devices were used. Countermeasures
asphyxia and a long-term increase in CAGE and to reduce fatalities based on these observations will
cardiac disabling injuries. Asphyxia was associated require validation by active surveillance.
____________________________________________________________________________________________

INTRODUCTION Stickybeak data for 1972-2005 in order to better identify

The publication of diving fatality reporting began in
Australia in 1972 with the introduction of “Project Sticky-
beak” by Walker [1], who painstakingly collected and re-
ported information on snorkeling and compressed-air div-
ing fatalities for more than 40 years. Since its inception,
Project Stickybeak reports have been regularly published
in the journals of the South Pacific Underwater Medicine
Society (SPUMS) and, later, Diving and Hyperbaric
Medicine, (joint journal of SPUMS and EUBS) and re-
printed with permission in several compilations pro-
duced by Divers Alert Network Asia-Pacific (DAN AP)
[2-4].
In 2008, Denoble et al. [5] introduced sequential
analysis (“root cause analysis,” or “RCA”) for the in-
vestigation of diving fatalities, dividing the course of
each accident into four sequential events: the trigger,
disabling agent, disabling injury, and cause of death.
Our aim was to apply this methodology to the Project
causes and progressions of diving accidents and, con-
sequently, try to identify appropriate preventative
strategies.
Methods
Data collection
By arrangement with various state and territory coro-
nial offices, Walker obtained relevant coronial, police,
witness and autopsy reports, and these formed the ba-
sis of his analyses. However, during the last decade,
access to such documents became more difficult, with
increased privacy protection and requirements for ethics
approvals. As a result of these difficulties, and of its
interest in investigating diving-related accidents as part
of its mission to improve diving safety, DAN AP (Asia-
Pacific) formed a dive fatality review committee (which
includes Walker) in order ensure this important project
continues.

Copyright © 2013 Undersea & Hyperbaric Medical Society, Inc. 49

 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA
In 2005, DAN AP began incorporating Project Sticky-
beak data into a broader dive fatality database, which
includes data from elsewhere in the Asia-Pacific, and
now produces annual Australian diving fatality reports
[6-9]. Following ethics approvals from various state
coronial and heath ethics review boards in Australia,
DAN AP gained access to diving-related deaths
(including compressed-gas, snorkel and breath-hold
divers) recorded in the National Coronial Information
System (NCIS) as well as more detailed information
from various state coronial offices. This has overcome
some of the obstacles to data collection.
The DAN AP database was used to identify relevant
cases and the likely sequence of events involved in the
accidents. This study was conducted under general ethics
approval from The Human Research Ethics Committee,
Department of Justice, Government of Victoria, Australia.
Sequential analysis
In accordance with the methods used by Denoble et al.
[5], accident descriptions for all Australian dive-related
fatalities between 2003-2006 inclusive (n=81) were
studied by a panel of experts and classified into a
sequence of four events defined below. These experts
included one of the authors (JL), three dive physicians,
a retired judge and a diving instructor/researcher. Agree-
ment on the classifications was reached by consensus.
Two of the above group, including one of the authors
(JL) then retrospectively analyzed all reported cases
from 1972-2002 inclusive. This was done individually
and differences were discussed and resolved.
In general, information available in the case descrip-
tions, and autopsy reports improved with time. In part,
this seemed the result of a growing awareness of the
special requirements of a diving autopsy by coro-
nial pathologists.
Events in the accident sequence were defined as [5]:
1. Trigger. The earliest identifiable event that appeared
to transform an unremarkable dive into an emergency.
2. Disabling agent. A hazardous behaviour or circum- R code used in our analysis is available on request.
stance that was temporally or logically associated
with the trigger and perhaps caused the disabling
injury.
3. Disabling injury. An injury directly responsible for
death or incapacitation followed by death due to
drowning.
4. Cause of death (COD). COD was specified by the
medical examiner and could be the same as the
disabling injury, or could be drowning secondary
to injury.
50
The disabling injury was determined using criteria
similar to those described by Denoble [5] as follows:
(a) Asphyxia: asphyxia with or without aspiration of
water and with no indication of a prior disabling
injury. (In this study, Asphyxia was defined as a
condition arising when the body was deprived
of oxygen, causing unconsciousness or death [10]).
(b) Cerebral arterial gas embolism (CAGE): gas in the
cerebral arteries with or without evidence of lung
rupture.
(c) Cardiac: acute episode of chest discomfort signaled
by the diver, distress without obvious cause, history
of cardiac disease, or autopsy findings.
(d) Trauma: witnessed trauma, traumatic findings at
autopsy.
(e) Decompression sickness (DCS): signs or symptoms,
autopsy findings.
(f) Other: stroke, cerebral hemorrhage, etc.
(g) Unknown: body not recovered, no autopsy available,
no indications of disabling injury at autopsy.
In addition to the accident sequence, we investigated a
variety of other information from the reports, including
the experience of the victims, weight management,
BCD management, remaining gas, buddy situation,
the depth at which the accident occurred, the gender
and age of the victim and whether or not high levels of
anxiety or stress appeared to have been a major trigger.
Determination of the presence of anxiety or stress
was generally made on the basis of witness reports,
the victim’s lack of experience and/or other factors such
as the witnessed presence of water in the mask early in
an accident sequence.
Statistical analysis
Software
Statistical analyses were performed using the open source
statistical package R [11,12]. Pareto charts were plotted
using the additional library qcc [13]. This software is
available from the R archive network [14]. Additional
Trends in accident counts
Since diving accidents are rare events, it is reasonable
to make the working assumption that annual accident
counts follow a Poisson distribution. Trends in the
number of accidents over time were investigated ini-
tially using loglinear Poisson regression. This assumes
that the predicted mean number of accidents in a par-
ticular year, µ, depends on the calendar year, x, through
J. Lippmann, A. Baddeley, R. Vann, D. Walker
 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA
log µ = β0 + β1 x [1]
so that the logarithm of the mean number of accidents
in a particular year is a linear function of the year. For
more flexible and realistic tracking of trends, the linear
function on the right-hand side of equation (1) can be
replaced by a quadratic or higher-order function, making
a Poisson generalized linear model (GLM, [15]) or
be replaced by a smooth function, making a Poisson
generalized additive model (GAM, [16]). The likeli-
hood ratio test was used to assess whether a trend was
significant.
Trends in relative frequencies
Trends over time in the relative frequency of accidents
of a particular type (relative to all accidents) were pre-
dicted by fitting a Poisson GLM or GAM to the event
counts as above, and normalizing the predicted mean
counts to obtain predicted proportions for a given year.
This is effectively equivalent to prediction based on
logistic regression [15, pgs 209-214]. The same tech-
nique was used to track trends in buoyancy control
device (BCD) usage and other characteristics.
Trends in age
Trends in the age of victims over time were tracked
using linear models (e.g., linear regression) or GAM
with normally distributed errors, and the likelihood
ratio test was used to assess significance.
Influence of trigger
The endpoint for an accident was taken to be the dis-
abling injury, rather than cause of death, for reasons
articulated by Denoble et al. [5, pg 394].
To investigate the influence of accident triggers on
the outcome, we followed Denoble et al. [5] in applying
a logistic regression technique. For each disabling
injury (e.g., asphyxia), we defined a binary outcome
variable y which was equal to 1 for this injury and equal
to 0 for all other injuries. To assess whether a particular
injury was associated with a particular trigger (e.g.,
exertion), we also defined a binary indicator variable
x equal to 1 for this trigger and 0 for all other triggers.
The association between this disabling injury and trigger
was evaluated by fitting a multiple logistic regression [5]
p defined by the disabling injury, as above. For a parti-
log ____ = αx + β0 + β1 z1 + β2 z2 . . . [2]
1– p
where p is the probability that the outcome variable
y is 1 (e.g., the probability that the disabling injury is
J. Lippmann, A. Baddeley, R. Vann, D. Walker
asphyxia), α and β0, β1, β2 . . . are parameters to be
estimated by the regression procedure, and z1, z2 . . . are
other relevant co-variates recorded for the accident.
Co-variates used in our analysis included the age
of victim, gender, BCD status, experience, weight
belt status and depth of accident.
To determine whether the injury and the trigger
were associated after allowing for the co-variates z1, z2
. . . , we applied the likelihood ratio test of α = 0
against α ≠ 0. If the test result was significant, then
the strength of the effect was measured by the esti-
mated odds ratio OR=e α, where α is the estimate
obtained from the fitted model.
Interpretation of odds ratio
In this context, the odds ratio is formally defined by
taking the odds p/(1–p) of the injury when the trigger
is present, and dividing by the odds of the same injury
when the trigger is absent. The OR for trigger:
exertion and disabling injury: asphyxia is
P(asphyxia\exertion,co-variates) P(not asphyxia\not exertion,co-variates)
OR = _______________________ x _____________________________ [3]
P(not asphyxia\exertion,co-variates) P(asphyxia\not exertion,co-variates)
where P(A\B) denotes the probability of A given B.
An odds ratio of 1 indicates a lack of association
since the presence or absence of the trigger does not af-
fect the odds of the particular injury occurring.
The precise interpretation of odds ratios in this context
is complicated, because the probabilities and odds refer
to the population of fatal accidents, rather than the popu-
lation of all recreational dives. A small odds ratio
(OR<<1) for the link between trigger = exertion and
disabling injury = asphyxia (for example) does not in-
dicate that exertion makes diving safer by reducing
the chance of asphyxia. Rather, it indicates that the
proportion of accidents that end in asphyxia is lower
among those triggered by exertion than among those
with another trigger. Exertion lowers the relative
risk of asphyxia because it raises the relative risk
of another endpoint (e.g., cardiac incident).
Influence of disabling agent
The logistic regression technique was also used to in-
vestigate the influence of the disabling agent on the
outcome. The accident outcome variable y was again
cular disabling agent (e.g., cardiovascular disease), the
independent variable x was now defined to equal 1 in
the presence of this Disabling Agent, and 0 for any other
Disabling Agent. Explanatory co-variates in the logistic
51
 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA

_____________________________________________
logistic regression model, equation [2], was then fitted
Figure 1. Annual compressed gas fatalities
in Australia 1972-2005 and subsequent calculations performed as above.

Gender differences
The frequency of each type of accident was broken
15

down by gender of the victim. Evidence for gender

differences was assessed using standard contingency
table methods [17] and the logistic regression strategy
of equation [2].
10
Deaths

Results
From 1972 to 2005 inclusive, there were records of
5

351 deaths in divers who were breathing compressed

gas underwater. These included 288 scuba divers (in-
cluding five using rebreathers), 62 using surface-
supply and one with an unknown breathing apparatus.
0

1975 1980 1985 1990 1995 2000 2005

Annual incidence
Year Figure 1 shows the annual total number of deaths, a
fitted trend based on Poisson loglinear regression, and
regression were the disabling agents and again in- two-standard-error confidence limits for the fitted trend.
cluded gender, age of victim, BCD status, experience, There is a slight suggestion of an upward trend in the
weight belt status and depth of accident. The test of sig- annual number of deaths, although this is not significant
nificance and the odds ratio were calculated as above. (p = 0.167) using the likelihood ratio test for non-zero
slope.
Influence of buoyancy-related problems
Buoyancy-related problems could be present either as Triggers
a trigger, as a disabling agent, or both. To answer the The triggers identified were classified as equipment-
question “Does buoyancy trouble promote a particular related, gas supply-related, rough water, anxiety/stress,
type of accident?” we modified the logistic regression exertion, other, and unknown (the latter two being
technique by defining the variable x to equal 1 if either combined for the analysis). The various trigger catego-
the trigger was buoyancy-related, or the disabling agent ries and a breakdown within some of these categories
was buoyancy-related, or both. If no buoyancy-related are shown in Table 1. Figure 2 shows the relative
problems were present then x was equal to 0. The frequencies of the triggers as a Pareto chart.
________________________________________________________________________________
Table 1. Constituents identified within each trigger category
(n = 292; 59 cases with unidentified triggers were omitted)
21% other, mean age 38, 62% male 18% equipment trouble, mean age 33, 92% male
Hit by boat BCD (e.g. sticky inflator)
Silting Broken fin/mask strap
Spearfishing / fish collecting Drysuit blowup
Suicidal intentions Hose entanglement
Vomiting Tank slippage
Water in snorkel Weight belt detachment
18% gas supply trouble, mean age 35, 82% male 16% rough water, mean age 36, 80% male
Gas contamination Current
Gas supply interruption Surface conditions
Inappropriate gas Suction
Out of breathing gas Surge
11% exertion, mean age 47, 91% male 11% anxiety/stress, mean age 38, 62% male
5% buoyancy trouble, mean age 34, 80% male
________________________________________________________________________________

52 J. Lippmann, A. Baddeley, R. Vann, D. Walker

 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA

_____________________________________________ _____________________________________________
Figure 2. Triggers identified Figure 3. Disabling agents identified
in 292 diving fatalities in 275 diving fatalities

100%
100%
250 300

50 100 150 200 250 300

Cumulative percentage
Cumulative percentage

75%
75%
100 150 200

Frequency
Frequency

50%
50%

25%
25%
50

0%
0%

0
0

gas supply-related

ascent-related

CVD

buoyancy-related

entrapment

other

shark

other medical
other

equipment-related

gas supply-related

rough water

exertion

anxiety-stress

buoyancy-related

Disabling agents The frequencies of the various groups of disabling

The Disabling Agents identified were classified as gas agents are shown in Table 2 and Figure 3. The most
supply-related, cardiovascular disease (CVD), ascent- common was gas supply-related, predominantly being
related, buoyancy-related, and other, shown in Table 2. “out of gas.” Almost 50% of cases are covered by gas
supply-related and ascent-related disabling agents.

________________________________________________________________________________
Table 2. Constituents identified within each disabling agent category
(n = 313; 38 cases with unidentified triggers were omitted)
26% gas supply-related, mean age 33, 85% male 21% ascent-related, mean age 36, 80% male
Gas contamination Breath-holding on ascent
Gas supply interruption Gas sharing on ascent
Inappropriate gas Rapid ascent
Out of gas Tank slippage
Weight belt detachment
16% CVD, mean age 50, 92% male 12% buoyancy-related, mean age 31, 75% male
Evidence at autopsy Inadequate buoyancy control underwater (- or +)
Medical history Lack/loss of buoyancy on surface
Self/witness description of indications

11% entrapment, mean age 30, 88% male 8% other, mean age 34, 87% male
Blow to head
Crocodile attack
Hypothermia
Inadequate decompression
Laryngospasm
Narcosis / CO2
Propeller
Vomiting underwater
3% medical condition, mean age 44, 100% male 3% shark, mean age 30, 90% male
Aortic aneurysm
Cardiomyopathy
Cerebral aneurysm
Duodenal ulcer
Epilepsy
________________________________________________________________________________

J. Lippmann, A. Baddeley, R. Vann, D. Walker 53

 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA

__________________________________________________ _____________________________________________
TABLE 3. Relative occurrence of disabling injuries Figure 4. Disabling injury identified
(n = 311; 40 cases w/unidentified disabling injuries omitted) in 311 diving fatalities

100%
Disabling Frequency Male Female Mean age

200 250 300

Injury (%) (%) (%)
__________________________________________________(yrs)
Asphyxia 49 82 18 31

75%
Cumulative percentage
CAGE 25 84 16 36
Cardiac 18 91 9 48

Frequency
Trauma 5 93 7 31

50%
100 150
DCS 1 100 0 35
__________________________________________________
Other 2 100 0 47

25%
Disabling injuries

50
The disabling injuries identified were asphyxia, cerebral

0%
0
arterial gas embolism (CAGE), cardiac, trauma, decom-

asphyxia

CAGE-PBT

cardiac

trauma

other

DCS
pression sickness (DCS), and other, which included
stroke and gastrointestinal hemorrhage. The frequencies
of the various groups of disabling injuries are shown in
Table 3 and Figure 4. The predominant disabling injury
(49%) was asphyxia from the inhalation of water while _____________________________________________
diving. CAGE was thought to have contributed to one- Figure 5. Causes of death identified
quarter of the deaths and cardiac involvement to 18%. in 315 diving fatalities
__________________________________________________

100%
200 250 300

TABLE 4. Relative occurrence of causes of death

(n = 315; 36 cases w/unidentified disabling injuries omitted)

75%

Cumulative percentage
Cause of Frequency Male Female Mean age
__________________________________________________
death (%) (%) (%) (yrs)
Frequency

50%
Drowning 56 82 18 33
100 150

CAGE / PBT 21 86 14 36
Cardiac 16 88 12 48

25%
Trauma 5 93 7 31
50

DCS 1 100 0 35
__________________________________________________
Other 2 100 0 47

0%
0

drowning

CAGE-PBT

cardiac

trauma

other

DCS

Cause of death (COD)

CODs identified were drowning, CAGE, cardiac, DCS,
trauma, and other. The frequencies of these causes are
shown in Table 4 and Figure 5. The predominant COD
was drowning, which was reported in 56% of the cases,
followed by CAGE (21%) and cardiac (16%). Drown-
ing was often listed by default when a lifeless diver was
recovered from the water, and no other COD was reported
on autopsy. The difference between 56% drowning as
COD and 49% asphyxia as disabling injury may reflect
cases where drowning was secondary to an injury such
as CAGE.
Association of triggers and disabling agents
with disabling injuries
Drowning was the most frequently assigned COD, but
disabling injuries are more useful than COD for under-
standing the sequence of events and for designing
countermeasures to improve safety. The characteris-
tic triggers and disabling agents associated with the
three principal disabling injuries are shown in Tables
5 a, b and c. Note that the odds ratios (OR) refer to the
odds of having one specific disabling injury rather
than any other injury (equation 3). Odds ratios greater
than 30 or less than 1/30 have been truncated.

54 J. Lippmann, A. Baddeley, R. Vann, D. Walker

 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA
__________________________________________________
TABLE 5a. Triggers and disabling agents
associated with asphyxia (n = 153)
__________________________________________________
Trigger p-value odds ratio (95% CI)
Rough water < 0.001 11.2 (3.5, 36)
Buoyancy trouble < 0.01 7 (1.6, 31)
Equipment trouble < 0.05 2.93 (1.2, 6.9)
Exertion < 0.001 0.0127 (0.0012, 0.13)
__________________________________________________
Disabling agent p-value odds ratio (95% CI)
Gas supply trouble < 0.001 15.4 (6.4, 37)
Buoyancy trouble < 0.001 9.5 (2.7, 33)
__________________________________________________
CVD < 0.001 0.0129 (0.0014, 0.12)
Rough water, buoyancy trouble and equipment trouble were the
most significant triggers for asphyxia, while gas supply trouble
and buoyancy trouble were the most significant disabling agents.
Exertion and CVD had odds ratios (OR) of less than 1, indicating
these factors were associated with other disabling injuries rather
than asphyxia.
_____________________________________________
Figure 6. Age and calendar year over the observation
period for males and females. Fitted trends in age
for males (solid line) and females (dashed line).
45
40
Age
35
30
25
1975 1980 1985 1990 1995 2000 2005
Year
Association with diver and dive characteristics
The mean age of victims was 35.8 years, with a range of
13 to 71 years. The median age was 34, with quartiles 26
and 44 years. Divers aged older than 60 made up 4% of
fatalities of each gender. Overall, 88% of victims were
male and 12% were female, but the proportion of fe-
male deaths increased from 7% in 1972 to 22% in 2005.
Males were an average of one year older than females.
Figure 6 shows the relationship between age of victim
and calendar year of accident, for male and female vic-
tims, over the 34-year observation period. Mean age
J. Lippmann, A. Baddeley, R. Vann, D. Walker
__________________________________________________
TABLE 5b. Triggers and disabling agents
associated with CAGE (n = 76)
__________________________________________________
Trigger p-value odds ratio (95% CI)
Gas supply trouble < 0.01 3.11 (1.34, 7.2)
__________________________________________________
Disabling agent p-value odds ratio (95% CI)
Ascent trouble < 0.001 >30
Gas supply trouble < 0.05 0.298 (0.115, 0.774)
__________________________________________________
CVD < 0.05 0.049 (0.0014, 0.118)
For CAGE, gas supply trouble was the only significant trigger,
while ascent trouble was the most important disabling agent.
For example, running out of air as a trigger followed by rapid
ascent as a disabling agent was a common occurrence.
However, in cases where the disabling agent was a gas supply-
problem, the trigger was rarely (if ever) gas supply-related,
and CAGE was not implicated as the disabling injury.
__________________________________________________
TABLE 5c. Triggers and disabling agents
associated with cardiac events (n = 57)
__________________________________________________
Trigger p-value odds ratio (95% CI)
Exertion < 0.001 >30
Gas supply trouble < 0.05 0.058 (0.0065, 0.52)
Rough water < 0.05 0.0656 (0.0068, 0.63)
__________________________________________________
Disabling agent p-value odds ratio (95% CI)
CVD < 0.001 >30
__________________________________________________
Gas supply trouble < 0.01 0.059 (0.0065, 0.52)
For cardiac disabling injuries, exertion was the only trigger,
while CVD was the only disabling agent. Gas supply trouble or
rough water without exertion were rarely associated with
cardiac injuries and were predictive of other disabling injuries,
as discussed in Methods.
for both male and female victims increased signifi-
cantly at a rate of about 0.5 years per calendar year
(p < 0.001). In a given calendar year, male victims were
on average three years older than female victims. The
overall average male-female age difference was only
one year, because males were over-represented in the
earlier decades when divers were, on average, younger.
The χ2 test of independence between trigger and
gender was rejected (p = 0.0272), and there was a
positive association between female gender and the
anxiety/stress-related trigger (odds ratio 3.86).There
were no significant associations between gender and
disabling agent, disabling injury or COD.
There was a highly significant association between
dive purpose and gender (p < 0.001). Females were under-
represented among accident victims whose dive purpose
was “collecting seafood” and over-represented among
55
 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA

_____________________________________________
Figure 7. Estimated trend in the prevalence of CVD or
other medical condition as the disabling agent against
diver age. Shaded band is 2-standard-error confidence
interval. Dots show averages over 3-year age groups.
% of disabling agents that are CVD or other medical
100
_____________________________________________
Figure 8. Proportion of cases where exertion was the
trigger, as a function of age of victim. Shaded band
is 2-standard-error confidence interval. Dots show
averages over 3-year age groups.

100
80

80
% of triggers that are exertion
60

60
40

40
20

20
0

0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age

those whose dive purpose was “research.” This was 40 implies that the only route to cardiac injury in this
probably attributable to an association between gender age group is exertion, while the dip around ages 50-
and professional occupation. Experience, depth of ac- 60 is consistent with a less important role for exer-
cident, weight belt status, and buddy status were also tion in older divers, who are at greater natural risk of
evaluated, but no significant associations were found. cardiac injury. The involvement of exertion in younger
Cardiovascular disease (CVD) was believed to be divers may be indicative of undiagnosed or occult
the disabling agent in 16% of the cases. The preva- CVD, as suggested by Mitchell and Bove [23].
lence of CVD and other medical conditions as the dis-
abling agent increased significantly with age (p < 0.001; _____________________________________________
Figure 7). Denoble et al. (18; Figure 2) found younger Figure 9. Proportion of cardiac injury cases where
males had greater CVD involvement than younger the trigger was exertion plotted against diver age.
females, but the difference disappeared at about age Shaded band is 2-standard-error confidence interval.
60. In our data, the difference between genders was Dots show 3-year averages.
not significant, probably due to small annual numbers.
100

Figure 8 shows the proportion of cases where the

trigger was exertion plotted against diver age. The solid
% of triggers (cardio) that are exertion
80

curve was estimated by smooth GAM logistic regression

– the light shading shows the 95% confidence intervals,
and the dots are three-year averages. Age was strongly
60

associated with exertion as the trigger in cardiac events

(p<0.001). Only 4% of the victims were older than 60,
40

and gender had no influence as exertion was reported for

only four females. This agrees with the very large odds
ratio reported for exertion as a trigger of cardiac injuries.
20

A different picture emerges if we restrict the analysis

to cases where the disabling injury was cardiac. Figure
0

9 shows the proportion of cardiac injuries in which the

20 30 40 50 60 70
trigger was exertion, plotted against diver age. The trend
is significant (p < 0.05). The peak at 100% for ages 30- Age

56 J. Lippmann, A. Baddeley, R. Vann, D. Walker

 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA

_________________________________________________________________________________________
Figure 10. Prevalence of the three main disabling injuries against diver age.
0 20 40 60 80 100 Left to right: cardiac, asphyxia and CAGE. Dots show averages over 3-year age groups.

0 20 40 60 80 100

20 40 60 80 100
% of injuries that are asphyxia
% of injuries that are cardiac

% of injuries that are CAGE

0
20 30 40 50 60 70 20 30 40 50 60 70 20 30 40 50 60 70
Age Age Age

Figure 10 shows the prevalence of the three main The most common disabling agent was gas supply
disabling injuries plotted against diver age. Dots show trouble (26%), with running out of gas most frequent.
averages over three-year age groups, and fitted trends Ascent trouble (21%) was next, with rapid ascent the
were obtained using smooth GAM logistic regression. greatest problem. Cardiovascular disease (16%) was
The proportion of cardiac injuries increased with age, prominent in older males. Buoyancy trouble (13%) usu-
and the proportion of asphyxia decreased with age, ally involved loss of buoyancy on the surface.
both highly significant (p < 0.001), while injuries due to Entrapment (11%) was less common as a trigger and
CAGE were not associated with age. Gender was not more common as a disabling agent secondary to triggers
significant for cardiac injury, which affected only six such as silting or narcosis.
females.
Trends in diving fatalities
Discussion The results described above represent averages over
Overall fatality characteristics the 34-year observation period and were generally
The average age of diving fatalities in Australia increased similar to those of Denoble et al. [18], Cumming et al.
by 16 years from 1972-2005. Women who died while [19] and Richardson [20]. Conclusions based on many
diving were three years younger than men in the same years of pooled data, however, do not necessarily
calendar year (Figure 6) and increased in proportion reflect the present, which is of greatest interest to diving
from 7% to 22% of total fatalities – probably reflecting safety, and was most closely approximated in our
greater involvement of women in recreational diving data by the year 2005. This is apparent in Figure 11,
rather than greater hazard. Overall, the most common which illustrates statistical trends from 1972 to 2005.
disabling injuries were asphyxia (49%), CAGE (25%) Triggers showing trends over time included equip-
and cardiac events (18%). Asphyxia was commonly ment-related problems that were greatest during the
associated with rough water, buoyancy trouble, equip- late-1980s (Figure 11a) but have become less frequent
ment trouble, and gas supply trouble. CAGE was com- since then. This could reflect improved design or bet-
monly associated with gas supply trouble and ascent ter diver training and familiarity. A surprising finding
trouble, while cardiac injuries were associated with (Figure 11b) was that anxiety or stress seemed to have
exertion, cardiovascular disease, and greater age. increased significantly and particularly so in women.
Equipment trouble and gas supply trouble were the Determining whether anxiety or stress was present
most common triggers, each accounting for 18% of could be difficult, however, and this result should be
accidents. Rough water was also prominent, being iden- interpreted cautiously.
tified as the likely trigger in 15% of the accidents. Figure 11c shows that buoyancy-related disabling
Exertion (11%) occurred primarily in older males, while agents decreased over the observation period, and Fig-
anxiety/stress (11%) was disproportionally represented ure 12 suggests this decrease was related to BCD use.
in females. Buoyancy trouble (5%) was also a rele- The smooth estimates of proportions in Figure 12 were
vant trigger. derived by GAM logistic regression with normalization

J. Lippmann, A. Baddeley, R. Vann, D. Walker 57

 UHM 2013, Vol. 40, No. 1 – CoMPRESSED-GAS DIVE FATAlITIES IN AUSTRAlIA

_________________________________________________________________________________________
Figure 11. Trends over time for triggers, disabling sgents, disabling injuries and cause of death.
Confidence limits for the fitted trends are 2-standard error. Dots are 3-year averages.
Triggers
a. equipment-related (p=0.015) b. Anxiety/Stress (p<0.001)

% of trigers that are anxiety/stress

% of trigers that are equip-related

80
80

60
40 60

40
20

20
0

0
1975 1980 1985 1990 1995 2000 2005 1975 1980 1985 1990 1995 2000 2005
Year Year

Disabling agents
c. buoyancy-related (p<0.001) d. Ascent-related (p=0.002) e. CVD-related (p=0.026)
% of injuries that are ascent-related
80

80

80
% of injuries that are buoyancy

% of injuries that are CVD

60

60

60
40

40

40
20

20

20
0

1975 1980 1985 1990 1995 2000 2005 1975 1980 1985 1990 1995 2000 2005 0 1975 1980 1985 1990 1995 2000 2005
Year Year Year

Disabling injuries
f. Asphyxia (p<0.001) g. CAge (p=0.003) h. Cardiac injury (p=0.0039)
% of injuries that are cardiovascular
% of injuries that are CAGE-related
80

80

80
% of injuries that are asphyxia
60

60

60
40

40

40
20

20

20
0

1975 1980 1985 1990 1995 2000 2005 1975 1980 1985 1990 1995 2000 2005 1975 1980 1985 1990 1995 2000 2005
Year Year Year

Cause of death
i. Drowning (p<0.001) j. CAge (p<0.001) k. Cardiac (p<0.001)
% of deaths that are cardiovascular
% of deaths that are CAGE-related
80

80

80
% of deaths that are drowning
60

60

60
40

40

40
20

20

20
0

1975 1980 1985 1990 1995 2000 2005 1975 1980 1985 1990 1995 2000 2005 1975 1980 1985 1990 1995 2000 2005
Year Year Year

58 J. Lippmann, A. Baddeley, R. Vann, D. Walker

 UHM 2013, Vol. 40, No. 1 – CoMPRESSED-GAS DIVE FATAlITIES IN AUSTRAlIA
_____________________________________________
Figure 12. bCD use among diving fatalities.
100
80
Percentage
60
40
20
0
1975 1980 1985 1990 1995 2000
Year
to add up to 100 percent. For a particular BCD use
(“none,” “inflated,” etc.), the proportion of that use
among all accidents is represented by the vertical
separation between pairs of lines or between a line
and the upper or lower boundary.
Figure 12 shows trends over time in the pattern
of BCD usage. The percentage breakdown of BCD
usage (into the categories “none,” “not inflated,” “faulty,”
etc.) is represented by the vertical spacing between
curves. For example, for the year 1990, if we draw a
vertical line through 1990 on the graph, it can be seen
that BCD usage was “none” for approximately 25% of
fatalities; it was “not inflated” for approximately 50%
of fatalities (the space between 25% and 75%); and
“inflated” for about 10% of fatalities. The percentages
in Figure 12 are smoothed estimates, derived by GAM
logistic regression, normalized to add up to 100 percent.
The proportion of victims not wearing BCDs declined
from 80% in 1972 to 10% in 2005 (p < 0.001), although
50% of the victims in 2005 had not inflated their BCD.
Inflation might have improved the chances of surviv-
al, but 25% of the BCDs were partly or completely in-
flated, indicating that BCD inflation did not guarantee
survival. Less than 5% of BCDs were reported as faulty
during the observation period, suggesting good BCD
reliability. Overall, diving with no BCD was less com-
mon among women (4/52 = 8%) than men (109/299 =
36%), possibly reflecting minimal BCD use in the early
1970s when there were few female fatalities. The
decrease in buoyancy-related disabling agents among
fatalities over the study period coincided with an
J. Lippmann, A. Baddeley, R. Vann, D. Walker
increase in BCD use, both in the diving population and
in the recorded fatalities. A causal relationship may be
conjectured but cannot be explored using these data.
Disabling agents related to ascent trouble (Figure
11d) increased in relative frequency consistent with
an accident scenario in which inattention to the gas
supply leads to insufficient gas and emergency ascent.
Disabling agents related to cardiovascular disease
(Figure 11e) increased in relative frequency. This is
probably attributable to the increasing age of victims
(Figure 6), as the diving population is aging [21], age
and CVD were very strongly associated (Table 5c), and
age-related CVD is common in the general population
[22].
Disabling injury and cause of death showed similar
trends (Figures 11f-11k), with decreased percentages of
2005
asphyxia and drowning probably reflecting improved
understanding by pathologists of the special require-
ments for diving autopsies. These decreases were
accompanied by increases in CAGE and cardiac inju-
ries. The upward trend in CAGE might reflect, in part,
an increased availability of pre-autopsy CT scans.
It may also reflect an over-reporting of gas in the cir-
culation as CAGE, rather than as artifacts from decom-
position, resuscitation attempts and decompression.
Interventions and countermeasures
for improving diving safety
A useful strategy for decreasing diving fatalities might
be to develop interventions and countermeasures that
focused on reducing the triggers and disabling agents
characteristic of each disabling injury. Fatalities due to
asphyxia, for example, became less common with greater
BCD use, which would be expected to help distressed
divers return to the surface and to keep the heads of
unconscious divers above water, although many modern
BCDs will not float an unconscious diver in a face-up
position. Additional BCD training and routine main-
tenance might lead to further reductions in asphyxia.
As fatalities due to CAGE were associated with run-
ning out of gas and rapid ascent, pre-dive briefings
might place stronger emphasis on gas supply monitoring,
and air-sharing might be practiced before each dive.
The increase in disabling injuries associated with
cardiac events is worrisome because of the natural
development of cardiac risk that occurs with age [24].
Many older individuals appear to be taking up diving
due to its increased accessibility. In addition, after
certification at a younger age, an older diver might not
have had medical evaluation over the ensuing years
59
 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA
despite accumulating occult cardiac disease. The recent
diving fatality workshop addressed this problem and
provided recommendations for diving physicians and a
self-assessment questionnaire to assist older divers in
deciding whether to seek medical advice [24]. The
observation that exertion is strongly associated with
cardiac events argues for careful selection of dive
sites to avoid environmental conditions that might lead
to heavy exercise (Figure 8).
The suggestion that anxiety or stress is an impor-
tant trigger, especially in women, is new and warrants
further investigation for confirmation due to potential
for diagnostic bias, as noted above. Nonetheless, it is
reasonable that careful selection of dive sites to avoid
rough water and exposures beyond a diver’s training
and experience might help reduce anxiety or stress.
The interventions and countermeasures mentioned
above seem plausible enough, but whether they would
be effective cannot be known without active surveil-
lance. Active surveillance is widely used in public health
but requires continuous, systematic, and timely obser-
vation, with rapid dissemination and communication to
the public [25]. Continued surveillance of this nature
is essential if recreational diving safety is to improve.
Limitations
There were inevitable limitations and uncertainties
associated with our investigations. These included:
• Incomplete case data. Fatalities were sometimes not
witnessed, various people collected information over
34 years, and autopsy methods have improved.
Future efforts should standardize all phases of fatality
investigation to improve data accuracy and
completeness.
• Classification of cases into a sequence of four events
(trigger, disabling agent, disabling injury, cause of
death). This requires a single choice for each event
which may omit important factors in some cases.
A longer sequence of events might be useful.
• Difficulties in establishing unambiguous cause-
and-effect relationships in survey data. Some of our
main findings relate to correlations between triggers,
disabling agents and disabling injuries. These
conclusions are necessarily provisional and require
continuing active surveillance. Correlations may be
60
spurious due to confounding (e.g., apparent correla-
tions due to parallel changes in the diver population
and diving practice), diagnostic bias (e.g., a greater
propensity for investigators to impute anxiety to
inexperienced rather than experienced divers),
or missing data effects (e.g., some types of accidents
are intrinsically more likely to result in incomplete
data).
• Limited annual case data. An average of 10 fatalities
per year is a small number, making it difficult to
investigate complex relationships between multiple
variables in the presence of trends over time. This
is reflected in the very wide confidence intervals
obtained for some of the odds ratios. However, the
main associations and trends were statistically well
established over the 34-year period. Low numbers of
female deaths meant that gender effects were difficult
to investigate. Some of the p-values are inaccurate,
since they were based on the usual chi-squared
approximation to the likelihood ratio test, which
is a poor approximation in small samples.
• Regional conclusions. The report describes the
Australian experience, and some findings may not
extrapolate to other locations. Comparisons with
other regions may eventually require regional
variables such as water temperature, marine life,
sea state, reporting and investigation protocols
and expertise, etc.
• Absence of information on divers at risk. Without
a denominator, our analysis of factors affecting
disabling injuries had to be made for a given injury
relative to all other injuries. This problem arises
in many epidemiological and safety studies.
• Lack of currency. The most recent year investigated
was 2005, some seven years ago. An ideal surveil-
lance system would publish results annually within
not more than two years of the current year.
Acknowledgments
We acknowledge Monash University National Centre for
Coronial Information for providing access to the National
Coronial Information System (NCIS), State and Territory
Coronial Offices, various police officers, dive operators
and divers who provided information on these fatalities.
We also thank Dr Harri Kiiveri (CSIRO) for comments on the
statistical analysis and Scott Jamieson for organizing the data.
n
J. Lippmann, A. Baddeley, R. Vann, D. Walker
 UHM 2013, Vol. 40, No. 1 – COMPRESSED-GAS DIVE FATALITIES IN AUSTRALIA
____________________________________________________________________________________________________
REFERENCES
1. Walker D. Provisional report on Australian diving-related
deaths in 1972. SPUMS J. 1972; 2(4):37-42.
2. Walker D. Report on Australian diving deaths 1972-1993.
Melbourne: JL Publications; 1998.
3. Walker D. Report on Australian diving deaths 1994-1998.
Melbourne: Divers Alert Network Asia-Pacific; 2002.
4. Walker D. Report on Australian diving deaths 1999-2002.
Melbourne: Divers Alert Network Asia-Pacific; 2009.
5. Denoble PJ, Caruso JL, de L Dear G, Vann RD. Common
causes of open-circuit recreational diving fatalities. Undersea
Hyperb Med 2008;35(6):393-406.
6. Walker D, Lippmann J. Provisional report on diving-related
fatalities in Australian waters 2003. Diving and Hyperbaric
Medicine. 2009;39(1):4-19.
7. Walker D, Lippmann J, Lawrence C, Houston J, Fock A.
Provisional report on diving-related fatalities in Australian
waters 2004. Diving and Hyperbaric Medicine. 2009;
39(3):138-61.
8. Walker D, Lippmann J, Lawrence C, Houston J, Fock A.
Provisional report on diving-related fatalities in Australian wa-
ters 2005. Diving and Hyperbaric Medicine. 2010; 40(3):131-
49.
9. Lippmann J, Walker D, Lawrence C, Houston J, Fock
A. Provisional report on diving-related fatalities in Australian
waters 2006. Diving and Hyperbaric Medicine. 2011; in press.
10. http://oxforddictionaries.com/definition/asphyxia?region=us.
(Accessed 2012 March 1).
11. Ihaka R, Gentleman R. R: A language for data analysis
and graphics. Journal of Computational and Graphical Statis-
tics. 1996; 5(3):299-314.
12. R Development Core Team. R: A language and environment
for statistical computing. Vienna: R Foundation for Statistical
Computing, 2011.
13. Scrucca L. qcc: An R package for quality control charting
and statistical process control. R News. 2004;1:11-17.
14. The comprehensive R archive network. http://cran.r-proj-
ect.org.
15. McCullagh P, Nelder JA. Generalized Linear Models.
2nd ed. Boca Raton, Florida: Chapman & Hall/CRC;
1989:209-214.
16. Hastie TJ, Tibshirani RJ. Generalized Additive Models.
Chapman & Hall /CRC. Boca Raton, Florida.
J. Lippmann, A. Baddeley, R. Vann, D. Walker
17. Rudas T. Odds ratios in the analysis of contingency tables.
Series: Quantitive Applications in the Social Sciences.
Thousand Oaks, CA:Sage; 1998.
18. Denoble PJ, Pollock NW, Vaithiyanathan P, Caruso JL,
Dovenbarger JA, Vann RD. Scuba injury death rate among
insured DAN members. Diving Hyperbaric Med. 2008;
38(4):182-188.
19. Cumming B, Peddie C, Watson J. A review of the nature
of diving in the United Kingdom and of diving fatalities Jan. 1
1998 – Dec. 31 2009. In: Vann RD, Lang MA, eds. Recreational
Diving Fatalities. Proceedings of the Divers Alert Network
2010 April 8-10 Workshop. Durham, N.C.: Divers Alert
Network, 2011. ISBN #978-0-615-54812-8. Pp 99-118.
20. Richardson D. Training scuba divers: A fatality and risk
analysis. In: Vann RD, Lang MA, eds. Recreational Diving
Fatalities. Proceedings of the Divers Alert Network 2010 April
8-10 Workshop. Durham, N.C.: Divers Alert Network, 2011.
ISBN #978-0-615-54812-8. Pp 119-64.
21. Denoble PJ, Marroni A, Vann RD. Annual fatality rates
and associated risk factors for recreational scuba diving. In:
Vann RD, Lang MA, eds. Recreational Diving Fatalities.
Proceedings of the Divers Alert Network 2010 April 8-10
Workshop. Durham, N.C.: Divers Alert Network, 2011. ISBN
#978-0-615-54812-8. Pp 73-85.
22. Australian Bureau of Statistics. 4821.0-55.001 –
Cardiovascular disease in Australia: A Snapshot, 2004-5.
http://www.abs.gov.au/ausstats/abs@.nsf/mf/4821.0.55.001.
Accessed 3 May, 2011.
23. Vann RD, Lang MA, eds. Recreational Diving Fatalities.
Proceedings of the Divers Alert Network 2010 April 8-10
Workshop. Durham, N.C.: Divers Alert Network, 2011. ISBN
#978-0-615-54812-8.
24. Mitchell S, Bove A. Medical screening of recreational
divers for cardiovascular disease: Consensus discussion at the
Divers Alert Network Fatality Workshop. Undersea and Hyperb
Med 2011; 38(4):289-96.
25. Kucera KL, Marshall SW. Developing and evaluating
interventions using surveillance data. In: Vann RD, Lang MA,
eds. Recreational Diving Fatalities. Proceedings of the Divers
Alert Network 2010 April 8-10 Workshop. Durham, N.C.:
Divers Alert Network, 2011. ISBN #978-0-615-54812-8.
Pp 63-71.
u
61