International Journal of Aquatic Research and Education, 2007, 1, 221-230
© 2007 Human Kinetics, Inc.
The 4W Model of Drowning
Stathis Avramidis, Ronald Butterly, and David J. Llewellyn
The aim of the study was to develop a conceptual model of drowning incidents.
The authors conducted qualitative content analysis of drowning-incident videos
(N = 41) and semistructured interviews of those involved in drowning incidents
(N = 34), followed by the measurement of frequencies and Boolean search with
matrix intersection. Results confirmed that when there is human activity in,
above, or around an aquatic environment, a drowning incident might occur to
whomever, wherever, and under whatever circumstances. Factors that determined
drowning outcome were, in order of importance, rescuer characteristics (Who 1),
casualty characteristics (Who 2), location (Wherever), and general circumstances
(Whatever). The interaction of the rescuer with the casualty largely determines
the outcome of drowning. The 4W model is a promising tool in lifesaving and
lifeguarding.
Key Words: lifeguarding, lifesaving, water safety, swimming, rescue
Participation in aquatic activities increases the likelihood of drowning-related
incidents (Lifesaving Society, 2000). Petridou and Klimentopoulou (2006) have
observed that drowning often occurs to healthy people during leisure time. In
the year 2000, about 450,000 people drowned worldwide, making drowning the
second leading cause of unintentional injury-related mortality globally, after traffic
accidents (World Health Organization, 2003). This figure includes only reported
accidental drownings and submersions, which means that the incidence of actual
drownings from all causes might be much higher. Computer modeling also suggests
that the effects of global warming might lead to an increase in drowning-related
mortality (Epstein, 2000). Taken together these findings demonstrate that drowning
is likely to remain a leading cause of death in the future and a key public health
concern.
In the late 20th century it was thought that lifeguards were the key to drowning
prevention and water safety (Griffiths, 2001). Previous improvements in lifeguard
awareness have resulted from the development of domain-specific theoretical
models and specific interventional suggestions. For example, the 5-minute scanning strategy (Griffiths, 2000) was developed based on the observation that after 15
minutes of performing a simple task, the quality of people’s performance on that
task tends to deteriorate. Based on Griffiths’s strategy, lifeguards should therefore
Avramidis and Butterly are with the Carnegie Faculty of Sport and Education, Leeds Metropolitan University, Headingley Leeds LS63QS UK. Avramidis is also with the European Lifeguard Academy. Llewellyn
is with the Institute of Public Health, University of Cambridge, Cambridge, CB2 2SR, UK.
221
222
Avramidis, Butterly, and Llewellyn
change posture and scanning technique every 5 minutes or so in an attempt to maintain high levels of effectiveness. The C zones (Connolly, 2004) are used to explain
how a drowning casualty tends to progress through a number of worsening stages
or zones (i.e., comfort, concern, crises, critical, and cardiopulmonary resuscitation). Similarly, Pia (1984) revolutionized thinking in lifeguarding by establishing
the R.I.D. model to help explain why drowning can occur even when a lifeguard
is present. He hypothesized that drowning can occur as a result of the failure to
recognize the symptoms of drowning (R), intrusion (I), or distraction (D).
A broad-ranging conceptual model of drowning incidents occurring during
engagement in aquatic and nonaquatic activities has not previously been available
and would give greater insight into the drowning phenomenon. A conceptual model
of drowning incidents should prove to be a useful tool for teaching lifeguarding
skills, targeting those most vulnerable to drowning, and reducing the number of
drowning-related deaths. Whether someone drowns deliberately or accidentally is
of little concern to lifeguards, whose responsibility, simply put, is to prevent injury,
drowning, and death regardless of cause (Isaacs, 2003). Of course drownings do
not only occur when people are supervised by lifeguards. Improved understanding
of the dynamics of drowning by the general public through water safety education
might also lead to a reduction of drownings in alternative circumstances when
lifeguards are not present.
Although the number of people dying each year confirms drowning as a major
problem worldwide, most studies relate to preventive and forensic aspects without
integrating them within an overarching theoretical model (Bierens, Knape, & Gelissen, 2002). Although the studies that have been conducted afford us some insight
into the nature of drownings, a comprehensive conceptual model of drowning
incidents is clearly needed. Questions remain about the characteristics of people
who drown, the location of drowning incidents, the causes of drowning incidents,
and characteristics of rescuers (if present) that might make drowning more or
less likely (for example, see Table 1). The interaction of these factors can only be
addressed by a model that integrates all these factors.
The purpose of this study was to develop a model of drowning incidents and
to classify and rank risk factors (Petridou & Klimentopoulou, 2006) to establish
a platform for effective education and intervention. An extensive review of the
drowning-incident literature (Avramidis, 2004) revealed that drowning incidents
can occur as a result of several factors that can be categorized into four discrete
categories (Table 1). It was hypothesized that these factors are interrelated (see
Figure 1), and given human activity in, on, and around an aquatic environment,
a drowning incident might happen to whomever, wherever, and under whatever
circumstances. We therefore undertook two studies to investigate the importance
of these factors and their relationships.
Study 1
Participants
A criterion-sampling method (Patton, 1990) was used to obtain drowning-incident
videos (N = 41) from a wide variety of sources (BBC1, 2000, 2001, 2002; ITV,
2001; Mega Channel, 2001, 2002a, 2002b; Pia, 1970; Royal National Lifeboat
Table 1 The Four Factors That Contribute to Drowning and
Practical Examples
Rescuer characteristics
Training
Current level of experience,
expertise, physical strength,
vision, health, swimming
speed
Casualty
characteristics
Physical water
fitness
Disability or
medical problems
Age
Knowledge of the particular
dangers of the aquatic area
Sex
Professionalism (adequate
number of lifeguards, visible
appearance, clothing, written
operating procedures)
Socioeconomic
background
Ability to do risk assessment,
work as educator, recognize
the instinctive drowning
response, remain alert, and
react ignoring the bystanders’
lack of response
Ethnicity
Area of residence
Number and
type of family
members
Occupation
Casualty
behaviors
Location
General
circumstances
The specific
geographical
characteristics
of the country
Relationship
between
casualty and
equipment
Size and the
shape of the
working area
Risk as
physical
demand of
aquatic activity
Lighting,
heating, air
quality, water
clarity
Weather and
environmental
conditions
Social and
emotional
environment
Presence of
others
Time, day,
season of
occurrence
Rescue type
Aquatic
activity
Ethical issues
Note. Adapted from Avramidis (2004).
Figure 1 — The 4W model of drowning.
223
224
Avramidis, Butterly, and Llewellyn
Institution, 1994; Twenty First Century Films Production, 1998; Waga News, 2001).
This method facilitated the identification of variables and their relations that would
not otherwise be available for fatal or traumatizing events. These visual narratives
ranged in length from 30 to 720 s (M = 345.0, SD = 2.8).
Apparatus and Procedures
Videos were watched using a JVC television (CM31720-003) and a Panasonic
videocassette recorder (AG-MD830). Videos were watched and data input into
NVIVO software (QSR, 2002) for qualitative analysis. One of the first things to be
established was the length of each video narrative, because the aquatic emergencies were usually on videocassettes that contained other audiovisual narratives.
Therefore it was not always clear when each narrative started and ended. This had
to be defined to guarantee reliable measurements during the test–retest. The reset
time button of the VCR was pressed while the first visual or audio message that was
related to the aquatic emergency was on the screen. For example, in some videos
the audio narrative started before the actual visual portion, and in other cases, the
visual video started before the audio narrative. In both cases the actual start point
of the video was either the very first visual scene or the first audio narrative of the
video. In cases in which the video was connected with transition effects with the
next or preceding video segment on the tape, however (e.g., fade in), the starting and
finishing points were when the whole scene could cover the whole TV screen.
As soon as the start and end points of the video were established, the videotapes
were watched in real time like a movie, so that the researcher could get overall
impressions of the aquatic emergency and take rough notes. The objective and
subjective content of the video were then noted. Objective content was defined as
the observations of audio or visual information on which every person watching
the video would agree (e.g., what type of rescue a lifeguard did, in which aquatic
environment the casualty was drowning, what they said). Unsupported assumptions
and editorial comments were avoided. An example of an objective description is
“Mr. L.H. was immersed in the river. Mrs. L.McD. was shouting to him to hold
on to the collar of the dog that approached to rescue him.” For objective description one would avoid reporting that Mr. L.H. could not hear Mrs. L.McD. because
he was deaf or saying that the dog that rescued him was a Newfoundland until
those things were confirmed as such in the video. Subjective content of records
was defined as the responsive interpretations of the psychosocial dynamics of the
scene. Subjective content should be the same regardless of the number of times that
the same researcher observes the same video. An example of subjective content
is the different interpretation of the same example: “After being saved by the dog,
Mr. L.H. was very scared” or “After being saved by the dog, Mr. L.H. was very
depressed.” A researcher’s observations would therefore be influenced by his or her
own age, gender, personal experience, and so on. The researcher could therefore
misunderstand what he or she is observing, resulting in mistaken interpretations
of the situation (Gratton & Jones, 2004).
To minimize observer bias the videos were observed twice within a period of
3 months, and then information that was common to both observations was finally
recorded and saved as the final narrative. During the test and retest each story
had to be transcribed twice, first to analyze all the audio messages in the video
The 4W Model of Drowning
225
and second to describe what could actually be seen. This ensured that enough
information about each story would be available for analysis instead of relying
completely on the narrator’s comments. Finally, each aquatic-emergency narrative
was divided into manageable sections (30 s long) so specific observations could be
precisely located in the transcribed text. NVIVO also allowed the notation of the
exact location of the coded text in the transcribed narrative by document number,
paragraph, and line (Rich & Chalfen, 1999). Because the subjects of the current
study were the casualties shown in the videos, they had no direct communication
with the researcher, and therefore threats to reliability such as subject error and
subject bias were not present (Gratton & Jones, 2004).
Observational content analysis of the video-recorded incidents enabled us to
examine the drowning experience “from the inside out.” The objective data that
the videos captured were rich in information that other forms of data could not
duplicate; an example of this was the audiovisual record of a girl being submerged
under the water for 4 min, wheezing and gasping for breath with her hair caught
in the water-suction valve of a whirlpool. Content analysis was dependent on the
careful observation of the videos, the categorization of the frequency and nature
of the verbal interactions, the data analysis, and the writing of a brief report with
recommendations for future practice (Booth, 1998). Coding stripes and the node
browser enabled the visual identification of differently coded sections and the
internal comparison of all data that were similarly coded as tree nodes. Frequencies
were measured, and Boolean search with matrix intersection identified the dominant linkages between the four variable factors thought to contribute to drowning
incidents (Table 2).
Results
The analysis of the 115 tree nodes of the four factors resulted in 7,560 theoretically possible variable combinations. From this range of possible relationships,
1,425 (18.84%) were actually observed (see Table 2). A relationship was defined
as every combination of one characteristic of one factor with another characteristic
of another factor. For example, one relationship might be the ability to recognize
(rescuer characteristic) the casualty’s instinctive drowning response of a nonswimmer (casualty characteristic) or the presence of risk during an activity (circumstances
of occurrence) in the deep water of the sea (place of occurrence). The casualty
Table 2 Frequency of Observed Relationships and Possible Matrix
Relationships Between 115 Tree Nodes (in Parentheses) of the Four
Key Drowning-Incident Factors in Drowning-Incident Videos (N = 41)
Who 1 (casualty)
Who 2 (rescuer)
Whatever
circumstances
Wherever
Who 1
(casualty)
Who 2
(rescuer)
Whatever
circumstances
Wherever
—
303 (1,584)
—
104 (864)
359 (1,056)
237 (1,404)
230 (1,716)
—
192 (936)
—
226
Avramidis, Butterly, and Llewellyn
(Who 1) related with the rescuer (Who 2) 303 times, with the place of occurrence
(Wherever) 237 times, and with the circumstances under which the incident took
place (Whatever circumstances) 104 times. The rescuer related with the place of
occurrence 230 times and with the circumstances of the drowning 359 times. The
place of occurrence and the circumstances under which the drowning occurred
were related 192 times. Table 2 also shows that the rescuer and the circumstances
under which each aquatic emergency took place played the biggest role in the
outcome of the examined sample (359 relationships). The second most important
combination of factors was the relationship between the casualty and the rescuer
(303 relationships). The third important combination was between the casualty
and the place where the incident occurred (237 relationships). The fourth-ranked
combination was between the rescuer and the place of occurrence of the aquatic
emergency (230 relationships). The fifth rank was the relationship between the place
of occurrence and the circumstances of the incident (192 relationships). Finally,
the sixth most frequent occurrence was the relationship between the casualty and
the circumstances under which the incident occurred (104 relationships).
Study 2
Participants
With a snowball sampling method (Patton, 1990), we conducted semistructured
interviews with 30 male (age 16–65 years, M = 28.4, SD = 11.3) and 4 female participants (age 19–65 years, M = 37.5, SD = 19.5) who were water safety or aquatic
professionals (e.g., lifeguards, lifesavers, scuba divers, and athletes of aquatic sports)
from Greece (n = 25, 71.4%), the United Kingdom (n = 2, 5.7%), the United States
(n = 1, 2.8%), and Cyprus (n = 6, 17.1%). Participants were selected if they had
witnessed and could describe a drowning-related incident above the surface of the
sea (n = 23, 67.6%), under the surface of the sea (n = 5, 14.7%), in a lake (n = 2,
5.9%), or in swimming pool or water park (n = 4, 11.8%).
Apparatus and Procedures
A Sanyo M-1110C audiotape recorder and 2-hr Maxell cassettes were used to
record the interviews. Institutional ethical approval was first obtained to conduct
semistructured interviews investigating the factors involved in drowning incidents.
The people interviewed were involved in the drowning incidents they described.
A snowball or chain sampling method was used to locate information-rich key
informants and critical cases who either viewed the footage or discussed their
own experience (Patton, 1990). An information sheet was distributed to potential
participants before the interviews explaining the nature and objectives of the study,
and voluntary informed consent was obtained (Gratton & Jones, 2004).
Interviews were conducted using a semistructured interview outline, which
included points relating to each of the four factors of interest (i.e., the rescuer,
the casualty, the location, and the circumstances). Confidentiality and anonymity
were maintained throughout, and individuals were not identifiable from the raw
data (Patton, 1990). Data were transcribed and entered into NVIVO for indexing
and qualitative content analysis (Wengraf, 2001). The procedures adopted were
The 4W Model of Drowning
227
consistent with those used in Study 1. All hard copies were kept in a locked cabinet,
and electronic data were password protected.
Results
The analysis of the four factors revealed 2,910 possible random factorial combinations (see Table 3). From this range of possible relationships 206 (7.8%) were
observed. The interaction of the rescuer and the general circumstances appeared
to play the biggest role in determining the outcome of aquatic emergencies (58
relationships), along with the way the casualty related with the rescuer (58 relationships). The third most important combination was between the casualty and
the circumstances under which the incident occurred (36 relationships). The fourth
combination was between the casualty and the place of occurrence of the aquatic
emergency (27 relationships). The fifth was the relationship between the places of
occurrence and the casualty (16 relationships). Finally, the sixth was the relationship between the place and the circumstances under which the incident occurred
(11 relationships). When we compare the results from Studies 1 and 2 a similar
pattern of relationships between the 4W factors is observed (Table 4).
Discussion
The results of these two studies showed that the rescuer, the casualty, the place,
and the circumstances of occurrence determine the outcome of drowning incidents.
The factors with the highest frequency and therefore potentially having the most
importance for the outcome of the rescue are, in order of significance, the rescuer
characteristics (Who 1), the casualty characteristics (Who 2), the drowning location
(Wherever), and general circumstances (Whatever). It is now clear that although the
rescuer is the most important factor, as was thought before, it is not the key to the
prevention of drowning incidents; rather, the interaction between the rescuer and
the casualty is. Thus, previous conceptual models appear to have neglected the role
of important factors and oversimplified the complex interactions between them.
Our study represents the first attempt to provide a comprehensive model of
drowning-related incidents and has a broad range of applications. By incorporating
Table 3 Frequency of Observed Relationships and Possible Matrix
Relationships Between 115 Tree Nodes (in Parentheses) of the Four
Factors Related to Drowning Incidents in Semistructured Interviews
of Those Involved in Drowning Incidents (N = 34)
Who (casualty)
Who (rescuer)
Whatever
circumstances
Wherever
Who
(casualty)
Who
(rescuer)
Whatever
circumstances
Wherever
—
58 (420)
—
16 (320)
36 (532)
27 (416)
58 (728)
—
11 (494)
—
228
Avramidis, Butterly, and Llewellyn
Table 4 Single and Contributing Factors That Determine the
Outcome of Drowning
Observed
Ranked factor
relationships
combinations
(n)
Rescuer
Videos (n = 41)
rescuer—whatever
359
6
rescuer—casualty
303
5
casualty—wherever
237
—
rescuer—wherever
230
3
wherever—whatever
192
—
casualty—whatever
104
—
Subtotal
1,425
14
Interviews (n = 34)
rescuer—wherever
58
6
rescuer—casualty
58
5
rescuer—whatever
36
4
casualty—wherever
27
—
casualty—whatever
16
—
wherever—whatever
11
—
Subtotal
206
15
Grand total
1,631
29
Casualty
Wherever
Whatever
—
5
4
—
—
1
10
—
—
4
3
2
—
9
6
—
—
—
2
1
9
—
5
—
3
2
—
10
20
6
—
—
3
—
1
10
19
—
—
4
—
2
1
7
16
Note. The first column contains the ranked factor combinations, the second column the observed relationships, and the other four the number of times these factors were actually related. The more times
two factors were interrelated, the higher their rank. The pair with the most relationships ranks 6, and
the pair with the least ranks 1.
both observational and interview data, effectively a form of methodological
triangulation, our findings and the 4W model itself are more likely to have construct
validity (Gratton & Jones, 2004). This maximizes the likelihood that the 4W model
will be generalizable, although further research is required to replicate the current
findings and examine the utility of the 4W model in other drowning contexts.
A number of methodological points also need to be considered. The current
findings are subject to a number of sources of potential bias. For example, those
interviewed might struggle to remember specific details about the drowning incident,
and their perceptions are likely to have been influenced by the stressful nature of the
situation. Considerable time might also have passed between the drowning incident
and the interview, although the in-depth interviews were designed to maximize
recall. The degree to which the 4W model is generalizable to situations in which
a lifeguard is not present is also a topic for future research.
The R.I.D. factor (Pia, 1984) and the 5-minute scanning strategy (Griffiths,
2000) are proposed to be useful to lifeguards, although they have limited application to nonlifeguard settings. The C-zones approach (Connolly, 2004) helps both
lifeguards and the general public understand the drowning process, although in a
relatively limited capacity. The 4W model, on the other hand, offers a far broader
and more detailed model for describing and understanding drowning incidents. For
example, the 4W factors provide insight into the antecedents of drowning incidents
and the influences determining their outcomes. The 4W model applies to a wide
range of human activities, including those that occur in, above, and around aquatic
environments. Those involved in water safety education should therefore consider
The 4W Model of Drowning
229
using the 4W model to educate others about the factors that lead to drowning
incidents and determine their outcome.
These findings have important implications for the general public, because
anyone engaged in activity in and around water should consider the risks that they
are exposed to. The presence of lifeguards and their interaction with other factors is
clearly important, and people should engage in aquatic activities that are supervised
by lifeguards wherever and whenever possible. Our findings suggest the importance of professionalism in lifeguarding, structured training, and high certification
standards. Local authorities have a responsibility to make funding available when
possible for additional professional lifeguards to minimize the number of drowning
incidents. When drowning incidents occur, the 4W model also provides a useful lens
through which to examine the antecedent and situational factors that led to them.
This process might help identify possible cases of negligence and teach valuable
lessons to minimize the number of future drowning incidents.
Several questions were left unanswered, such as the way that the rescuer
characteristics relate with the place and with the circumstances of occurrence of
the drowning incidents. In contrast to the factor relationship “rescuer–casualty,”
these secondary factors and relationships could not be established as important
contributing factors to the outcome of drowning incidents because they appeared
to be important in one study but less important in the other. Given the fact that
the rescuer plays the dominant role in all the linkages of the 4W model, a further
examination of how the rescuer relates with the rest of the factors will enhance
the understanding and the effectiveness of the rescue procedure for the benefit of
both casualty and rescuer.
Conclusions
When there is human activity in, above, or around an aquatic environment a drowning incident might happen to whomever and under whatever circumstances. The
factors that determine the outcome of drowning incidents are, in order of significance, rescuer characteristics (Who 1), casualty characteristics (Who 2), location
(Wherever), and general circumstances (Whatever). Rescuer characteristics, and
their interaction with other factors (mainly with casualty characteristics), appear to
largely determine the outcome of drowning incidents. As a result of the two studies
reported in this article, we conclude that the 4W model of drowning is a promising
tool in lifesaving and lifeguarding training, as well as in understanding the dynamics
of drowning for risk assessment, accident prevention, and safety promotion from
water safety organizations, local authorities, and the general public.
References
Avramidis, S. (2004). The Model 4W of drowning & lifeguarding. In B.B. Barrosso, L.M.P.
Gomez, M.A.G. Marinas, & P.S. Velaso (Eds.), VI Jornadas 2004, VI ESS Congress
of Lifesaving book of abstracts (pp. 221-223). Segovia: Spain: Escuela Segovia de
Socorisimo.
BBC1. (Producer). (2000, January 17). 999 television series [Television series]. London:
BBC1.
230
Avramidis, Butterly, and Llewellyn
BBC1. (Producer). (2001, February 16) Children near drown in a frozen lake [Television
broadcast]. London: BBC1.
BBC1. (Producer). (2002, January 4). Near drowning of family in Indianapolis [Television
broadcast]. London: BBC1.
Bierens, J.J.L.M., Knape, J.T.A., & Gelissen, H.P.M.M. (2002). Drowning. Current Opinion
in Critical Care, 8, 578-586.
Booth, M.B. (1998). Teaching and leading of junior hospital doctors on ward rounds: The use
of video recording and analysis by an educationist. Medical Teacher, 20, 470-472.
Connolly, J. (2004). C-zones: The Connolly framework. In S. Avramidis, E. Avramidou,
& J. Tritaki (Eds.), 1st International Lifeguard Congress (p. 12). Piraeus, Greece:
European Lifeguard Academy.
Epstein, P.R. (2000). Is global warming harmful to health? Scientific American, 283, 5057.
Gratton, C., & Jones, I. (2004). Research methods for sport studies. New York: Taylor &
Francis.
Griffiths, T. (2001, March) Swim at your own risk. Aquatics International, 14.
Griffiths, T. (2000, March). Five-minute scan. Aquatics International, 12.
Isaacs, D. (2003). To kill or to let die? Journal of Paediatric Child Health, 39, 135-136.
ITV. (Producer). (2001, February 20). When stands go wrong [Television broadcast].
London: ITV.
Lifesaving Society. (2000). Boating fatalities in Canada, 2000 special report. Ottawa, ON,
Canada: Author.
Lightfoot, C. (1997). The culture of adolescent risk taking. New York: The Guilford
Press.
Mega Channel. (Producer). (2001, December 5). Drowning in Patra [Television broadcast].
Athens: Mega Channel.
Mega Channel. (Producer). (2002a, March 11). Stowaways drown in river Rio Bravo [Television broadcast]. Athens: Mega Channel.
Mega Channel. (Producer). (2002b, December 3). Batavanos nearly drowns [Television
broadcast]. Athens: Mega Channel.
Patton, M.Q. (1990). Qualitative evaluation and research methods (2nd ed.). Newbury
Park, CA: Sage.
Petridou, E., & Klimentopoulou, A. (2006). Risk factors for drowning. In J.J.L.M. Bierens
(Ed.), The handbook of drowning (pp. 63-69). Berlin: Springer.
Pia, F. (1984). The RID factor as a cause of drowning. Parks and Recreation, 19, 52-55,
67.
Pia, F. (Director). (1970). On drowning: A training aid [Motion picture]. Larchmont, NY:
Water Safety Films.
QSR. (2002). NVIVO: Getting started in NVIVO. Melbourne, Australia: QSR International.
Rich, M., & Chalfen, R. (1999). Showing and telling asthma: Children teaching physicians
with visual narrative. Visual Sociology, 14, 51-71.
Royal National Lifeboat Institution. (Producer). (1994). The history of RNLI. Heroes of the
sea [Motion picture]. Poole, UK: Channel Television Ltd for Castle Communications
PLC.
Twenty First Century Films Production. (Producer). (1998). Great survivors: Witness events
of the 20th century [Motion picture]. London: Author.
Waga News. (Producer). (2001). Animal heroes: Real life rescues [Television broadcast].
Atlanta, GA: New World Communications of Atlanta.
Wengraf, T. (2001). Qualitative research interviewing. London: Sage.
World Health Organization. (2003). The injury fact book. A graphical overview of the global
burden of injuries. Geneva: Author.