Journal of Science and Medicine in Sport (2006) 9, 243—248

ORIGINAL PAPER

Feasibility study of a water safety data collectionfor beaches

Ann Williamson ∗

NSW Injury Risk Management Research Centre, University of New South Wales, Building G2,Western Campus, UNSW Sydney, NSW 2052, Australia

Accepted 22 March 2006

KEYWORDS

I

Sp1

1d

Water safety;Rescues;Injury surveillance

Summary This study examined the feasibility of a methodology for collectinginformation on water safety on beaches. Previous work showed that lifeguards andlifesavers were unreliable data collectors as they often did not have time to devoteto it. The aim of this study was to trial a data collection using dedicated data collec-tors. The results showed the water safety-related items that could be consistentlyreported, and those items that were not. Items relating to beach conditions thatcould be reported were: wave type, tide times, sea conditions, rips and weatherand wind conditions. Items associated with rescues that could be reported were:sex, age group, activity before rescue, who performed the rescue, water depth,safety flag location, nearest rescue and rescue equipment. Data collectors couldcollect some data on an hourly basis especially: exact or best estimates of beachattendance, exact or best estimates of major rescues and rescues, wind directionand first aid used. Some items could not be collected consistently and their inclusionin data collections should be reviewed. These include: exact age, suburb, indigenousstatus and swimming ability. While the validity of the collected information was notable to be established, the feasibility study showed which items could be practicallyincluded in a water safety data collection, but that dedicated data collectors maybe needed to maintain a safety data collection on busy or dangerous beaches. Thisapproach together with data collection by water safety professionals at less busytimes would be cost-effective.© 2006 Sports Medicine Australia. Published by Elsevier Ltd. All rights reserved.

ntroduction

wimming is one of the main forms of sport andhysical activity for Australians over the age of7 years and has the highest participation rate

∗ Tel.: +61 2 93854599; fax: +61 2 93856040.E-mail address: a.williamson@unsw.edu.au.

in sporting activities for children in the 5—14age group.1 Nevertheless, drowning is a seriousinjury problem in many communities, not the leastbecause it tends to affect the most vulnerablegroups, especially under 5-year olds and olderpersons.2 Management of water safety requiresinformation on the size and nature of the watersafety problem.

440-2440/$ — see front matter © 2006 Sports Medicine Australia. Published by Elsevier Ltd. All rights reserved.oi:10.1016/j.jsams.2006.03.023

244 A. Williamson

In Australia, ocean and beach drowning accountfor around one-third of all non-boating drownings.2

New South Wales (NSW), like many states of Aus-tralia, has no common data collection of watersafety-related incidents occurring on beaches. Anumber of independent groups collect informa-tion on drowning, but there is no information col-lected on the circumstances in which drowningoccurs or on the number and circumstances of near-drownings. Better information on why and howdrowning and near-drowning occurs on beaches willprovide clearer direction for beach safety strate-gies. Considerable effort is already put into beachsafety in Australia through the important work ofbeach safety groups (e.g., professional lifeguardsand volunteer surf lifesavers) and there is evidenceof their success. For example, data collected byQueensland Surf Lifesaving on incidents involvingits members where some resuscitation was requiredshowed that only around 17% occurred in the areaswhere the beach was patrolled by lifesavers.3 Fur-thermore, 95% of cases occurring within patrolledareas were resuscitated successfully, whereas only62% cases occurring outside the patrolled area weresuccessfully resuscitated. Better information on

istics of people who are rescued, and to comparethe quality of the results collected by water safetyprofessionals with that of dedicated data collectors(Trial 1).

An acknowledged problem with surveillance forinjury is defining the type of data to be collected.Clearly, this depends on the aim of the surveillance,whether it is to estimate the burden of the injury tothe community, to assess trends or to look for injuryrisk factors. Unfortunately, the sensitivity of a datacollection for identifying injury will vary with theseverity of the injury indicator used.5 For example,collections of injury fatalities such as drowning aremore likely to be reliable than collections of near-drowning incidents or near-miss incidents such asnumber of rescues. This problem can be overcometo a large extent by employing effective criteria forclassifying less severe injury indicators. The crite-ria need to include mutually exclusive categoriesand be exhaustive so that all categories can beclassified.6 The current study attempted to takethese points into account. Rescues were used as thewater safety indicator with a comprehensive inci-dent inclusion definition developed that includedmajor rescues and rescues as well as preventivea

M

Sc

how drowning and near-drowning incidents occurwill assist these groups further in directing theiractivities; including the deployment of rescue ser-vices, the development of new interventions andevaluation of the effectiveness of water safetyactivities as well as the development of watersafety policy by government.

Establishment of a NSW Water Safety Taskforceby the NSW government provided the impetus todevelop a comprehensive data collection on drown-ing, near-drowning and water-related injuries. Thefirst stage of the development of the data collectionmethodology involved establishing a standard set ofvariables or minimum dataset for water rescues tobe used to collect rescue information at beachesand public swimming pools.4 A small feasibility trialof a minimum dataset was conducted using life-guards and lifesavers as data collectors on eightbeaches along the NSW coast. The results demon-strated that, while some information could be col-lected, there were problems with obtaining com-prehensive information on busy, popular beachesas well as problems with the definition of a watersafety incident. A different approach was clearlyneeded.

This paper describes the results of a feasi-bility study of a data collection methodologythat involved observers undertaking comprehensivedata collection at specific times in the year. The aimwas to test the feasibility of collecting water safetydata on beaches, including details of the character-

ctions by water safety staff.

ethods and procedures

ite selection and characteristics of dataollectors

Trial 1—–Water safety professionals: The first trialinvolved 4 weeks of data collection in eight loca-tions chosen in collaboration with the Surf LifeSaving Association (SLSA) and the Australian Pro-fessional Lifeguards Association (APOLA) to reflecthigh and low pressure rescue locations. The trialoccurred in late summer/early autumn and datacollection occurred when water safety personnelpatrolled the beach and safe swimming areas weredefined by flags. The data collectors were stafffrom the two beach safety organisations (SLSAand APOLA). The beaches were typically patrolledfrom around 08:00 to 17:00 h or 18:00 h by pro-fessional lifeguards during the week and volunteerlifesavers during the weekend. The primary role ofboth lifeguards and lifesavers is visually to patrolor watch the beach, especially the area judged tobe the safest at the time which is marked by flagsto make it obvious to all water users. The life-guards and lifesavers were given basic training onthe data collection and the items to be collected.Trial 2—–Dedicated data collectors: The seconddata collection trail involved data collected over a

Feasibility study of a water safety data collection for beaches 245

1-week period on five beaches at the end of sum-mer and during an Easter holiday period on twobeaches. Each data collection again occurred atall times when water safety personnel patrolledthe beach. The SLSA and APOLA aided selection ofsafe and least safe beaches in the Sydney regionbased on ratings of beach safety by Short.7 Thebeaches chosen were not the same as for the firsttrial.

Data collectors were recruited by word of mouththrough surf clubs and through the university stu-dent body. They were not required to have anyparticular experience in water safety, althoughsome were volunteer lifesavers during the summerperiod. They were trained in sessions of 2—3 h inlength where the data collection methodology wasexplained and potential problems discussed. Theywere paid for the training session and for the datacollection time which was typically around 8 h oneach of the data collection days.

Information collected

The information collected for each beach was thesame at each data collection and included an hourlyoofcTf

1

2

sand, in the water in and outside the flagged orpatrolled areas*.

For items marked with an asterisk (*), data col-lectors were asked to make their best estimates if itwas not possible to make exact recordings. Whetherexact or best estimates were made was indicatedon the collection form.

Reported incidents were specific water safetyactions taken by water safety professionals, definedas follows:

Rescue-major: any rescue involving more than onelifeguard/lifesaver or a member of the public ren-dering assistance OR where the rescued personshad to be resuscitated OR where another agencyhad to render assistance (e.g., ambulance and res-cue helicopter);Rescue: any rescue where a person was physicallyassisted or supported to return to shore or otherplace of safety (e.g., retrieving a person in diffi-culty);Preventive action: any water safety advice pro-vided to the public (e.g., asking people to swimbetween the flags and use of public address sys-tem and/or whistle).

clqs

R

Tans2dE31ilcptiao

flo

bservation sheet which was completed at the endf each hour that the beach was patrolled and aorm designed to collect information about spe-ific reported incidents or rescues as they occurred.he information required for each form was asollows:

. Major rescue and rescue sheet: Age or agegroup*, gender, main language spoken at home,suburb of usual residence or overseas, indige-nous status, swimming ability of person res-cued*, activities of person just before res-cue, who performed the rescue and how theyknew the person was in difficulties, locationof the person at rescue point including depthof water and with respect to the flags, useof rescue equipment, use of first aid equip-ment, need for ambulance, involvement of alco-hol and drugs and narrative description of whathappened.

. Hourly observation sheet: Number of life-guards/lifesavers on the beach over the hour,tide times (first hour only), location of theflags over the past hour, number of major res-cues*, number of rescues*, number of preven-tive actions in total and inside and outsidethe flagged or patrolled area*, details of theweather, including temperature, sea conditions,wind conditions and direction, wave type andwave height, presence of rips or currents, num-ber of people in the location including on the

The analysis focused on the quality of the dataollected and compared the results of this data col-ection using dedicated data collectors with theuality of data collected by professional rescuetaff.

esults

rial 1 involved a total of 3199 hourly observationscross the 4-week period. One very busy beach didot collect any information during the study periodo was not included in any further analysis. For Trial, a total of 440 hourly observations were madeuring the 7-day summer collection and the 4-dayaster collection. For Trial 1, 19 major rescues and5 non-major rescues were recorded. For Trial 2,4 people were involved in major rescues and 52n rescues. Comparison with data collected by theocal water safety professionals showed that all res-ues were captured. Most rescues occurred in non-atrolled areas. In Trial 2, there were 1147 preven-ive actions counted across all of the beaches, vary-ng from 1.3 to 38.4% of reported incidents. Almostll related to events and circumstances occurringutside the flagged or patrolled area (90.2%).

As shown in Table 1, there were less missing datarom the hourly records when dedicated data col-ectors were used. The data collected for hourlybservation sheets by the dedicated data collectors

246 A. Williamson

Table 1 Percentage of missing data from hourlyrecords and rescues sheets for Trial 1 using watersafety professionals to collect data compared to Trial2 using dedicated data collectors

Variable Trial 1(n = 3199)

Trial 2(n = 440)

Hourly record sheetsExact count or best estimate

for attendance14.8 4.1

On sand n/a 1.5Within flags n/a 2.5Outside flags n/a 2.8

Exact count or best estimatefor major rescues

13.6 0.5

Exact count or best estimatefor rescues

12.6 0.5

Exact count or best estimatefor preventive actions

n/a 1.5

Wind direction 8.8 0Onshore or offshore 7.6 5.6

Air/water temperature 6.0Air temperature n/a 14.0Water temperature n/a 1.3

Wave height 4.7 5.1Rips (strong currents) 4.0 0.3Wave type 3.0 0.5Sea conditions 2.9 0.3Weather 2.8 0.5Wind conditions 2.8 0.5

Rescue sheetsYear of age 53 82Age group 27 27Postcode 88 86Language 24 33Swimming ability 22 21First aid used 82 35Drug and alcohol involvement 40 33

were almost complete for nearly all variables, withthe exception of air temperature, direction of thewind (onshore or offshore) and wave height. Most ofthe variables on the major rescue and rescue datacollection sheet could be collected, although somepresented problems, especially age and postcode(see Table 1). The same problems were seen in bothdata collection trials. The dedicated data collectorswere more successful in collecting data on first aiduse as this was missing for most cases when the datacollection was by water safety professionals.

Most of the counts of beach attendance wereby dedicated data collectors making best estimatesrather than exact counts of numbers (Fig. 1). Asmight be expected, there was a clear correla-tion between increasing numbers on the beachand the need for data collectors to estimate the

Figure 1 Number of observations made with exactcounts and best estimates for each section of the beachfor Trial 2.

numbers (point-biserial correlations for: on thesand, r(387) = 0.25, p < 0001; in water-within flaggedarea, r(381) = 0.6, p < 0.0001; in water-outside flags,r(382) = 0.57, p < 0.001).

Most importantly, the collectors were able to usethe type of rescue variables and to differentiatebetween major rescues and rescues and preventiveactions with only a very small amount of missingdata for these classifications (0.5% each for majorrescues and rescues and 1.5% for preventive actionsfor Trial 2). For Trial 2, however the rescue detailssheet was not completed for just over one-thirdof major rescues (35.7%) and around one-quarterof other rescues (23.1%). The missed informationfor both types of rescues was from one beach onlyand occurred mainly on occasions where a num-ber of people were rescued at the same time. Thiswas almost certainly because of the logistical prob-lems of obtaining information from rescued peo-ple when others were in danger, although informa-tion was collected in some rescues involving mul-tiple persons. Most of the missing data occurredearly in the data collection period, suggesting thatthe data collectors on this beach may have beenbecoming used to the requirements of the datac

D

Tasit

ollection.

iscussion

here are very good reasons for establishing reli-ble and valid data collections on the circum-tances of injury, especially for establishing trendsn injury occurrence and for evaluating the effec-iveness of injury prevention strategies. However,

Feasibility study of a water safety data collection for beaches 247

data collections for water safety, usually in the formof numbers of drownings, present particular prob-lems. Drowning is a comparatively rarer injury out-come compared to other injury mechanisms suchas road traffic injury so that it can be difficult tocollect sufficiently large amounts of informationto establish reliable trends and to determine riskfactors for the development of appropriate inter-ventions.

This study was an attempt to evaluate the fea-sibility of a water safety data collection method-ology. The results demonstrated that dedicatednon-professional data collectors could collect arange of information on the circumstances of watersafety incidents with comparatively little trainingand collect more complete information than watersafety professionals. While it would seem sensibleto establish a collection using lifesaving profes-sionals who are already working on beaches, theearlier attempt to develop such a data collectionwas limited by the problems of balancing watersafety work with requirements of the data collec-tion. This conflict between the main activities andresponsibilities of health and safety professionalsand the demands of on-going surveillance is likelytTob

bcnspmtlorMmunmrtta

ecTbwt

estimates were taken into account. While furtherwork on the validity of these estimates is needed,it seems that the method is a practical solutionto the problem of obtaining information on beachattendance.

The prime objective of this study was to col-lect broader information on water safety beyondthe number of drownings. Using incidents ratherthan actual injury can present problems of caseinclusion criteria, however. Lack of clarity of cri-teria will threaten the validity of the indicatorbeing collected. The definitions of major rescueand rescue used in this study were found to bereadily usable and the distinction between the twotypes of rescues was found to be unambiguous.The definitions also have the advantage of defin-ing the severity of the incident, since more severeor potentially life-threatening incidents are likelyto require physical intervention by rescuers andhence be distinguished as major rescues rather thanother rescues. Further work is needed, to demon-strate the reliability of these definitions betweencoders.

This study also included counts of the numberof actions that lifeguards and lifesavers made toppsoslbtdtsbthrnso

tseAtctlowtd

o be a common problem for many data collections.his study demonstrated an alternative methodf collecting good quality data, even on busyeaches.

A range of items relating to beach conditions,each attendance and the circumstances of res-ues could be reported successfully but some couldot. Personal information of people rescued pre-ented the most difficulty for both water safetyrofessionals and non-professional data collectors,ainly because it was often not possible to locate

he person to interview after the rescue. This prob-em may be difficult to overcome. Similarly, mostf the data collected on the swimming ability ofescued persons were based on rescuer estimates.ost rescued persons were judged to be weak swim-ers, though it is possible that rescuers tended to

se this classification on the basis that a rescue waseeded. The person was judged to be a strong swim-er in only two cases, both self-estimates by the

escued person. Defining swimming ability is a con-entious issue,8 so further work is needed to explorehe validity of collecting information on swimmingbility.

Problems were also encountered in obtainingxact counts of beach attendance, especially forounts of the number of people in the water.his type of information potentially provides aackground for interpreting the circumstances ofater safety incidents. The number of observa-

ions in this study increased markedly when best

revent water safety incidents during the studyeriod because the preventive activities of profes-ional water safety staff on beaches are likely toffset the number of rescues needed. Where waterafety staff actively interact with beach-goers toimit risky behaviour, fewer full rescues are likely toe required. This also means that judgements abouthe riskiness of beaches based on the number ofrownings or the nature of beach conditions shouldake into account the activities of water safetytaff. For example, based on the ratings developedy Short,7 the highest and lowest (safest) ratings ofhe beaches corresponded to the beaches with theighest and lowest numbers of preventive actionsespectively. A larger study that includes a largerumber of beaches is needed to look at the relation-hips between preventive actions and major andther rescues.

The results indicate that dedicated data collec-ors are an effective method for collecting waterafety information, although this may not be a cost-ffective approach for all beaches and at all times.combination of collections by dedicated collec-

ors on very busy or more dangerous beaches andollections by lifeguard and lifesavers at less busyimes are likely to be the most cost-effective col-ections of rescue data. The next stage in the devel-pment of an effective beach safety data collectionill be to evaluate the reliability or consistency of

he variables included in the data collected by auplicate data collection. As much of the material

248 A. Williamson

collected is factual, issues of the validity of the dataare less likely to be a problem. A few variables suchas the type of rescue and swimming ability haveproblems of definition that may affect their valid-ity and would benefit from further work to clarifytheir meaning.

Conclusion

Overall, this study suggests that information can becollected that is of value for water safety as it pro-vides more depth of understanding of the wider cir-cumstances in which water safety incidents occur.This relatively brief trial indicates that a larger,in-depth data collection would be invaluable for awide range of water safety purposes. These couldinclude:

• analysis of the relationships between beach con-ditions and water safety incidents;

• the timing and allocation of water safety staff tobeaches;

• development and evaluation of new water safetyprograms;

• development of water safety policy.

• A water safety data collection would also bevaluable for evaluating the effectiveness ofstrategies that we think might reduce drown-ing and near-drowning.

Acknowledgement

The author is very grateful to the Australian Profes-sional Ocean Lifeguards Association, the Surf LifeSaving Australia, the Local councils and lifeguardsand lifesavers who participated in the data collec-tion. The project was funded by NSW Health andsupported by the NSW Water Safety Taskforce.

References

1. Australian Bureau of Statistics. Children’s participation insports and leisure activities. Year Book Australia. Cul-ture and recreation. Australian Bureau of Statistics Report1301.0. Canberra; 2004.

2. Mackie IJ. Patterns of drowning in Australia, 1992—1997. MedJ Aust 1999;171:587—9.

3. Fenner PJ, Harrison SL, Williamson JA, et al. Success of surflifesaving resuscitations in Queensland, 1973—1992. Med J

Practical implications

• A water safety data collection for NSW beacheswould be feasible.

• Collection of information about safety-relatedincidents would be a valuable resource forunderstanding the common causal factors fordrowning and near-drowning incidents andfor developing better targeted approaches todrowning prevention.

Aust 1995;163:580—3.4. NSW Water Safety Taskforce. Minimum water safety dataset.

www.safewaters.nsw.gov.au.5. Robertson LS. Injury epidemiology. New York: Oxford Univer-

sity Press; 1992.6. Macaskill P, Driscoll T. National occupational injury statistics:

what can the data tell us? In: Feyer A, Williamson AM, edi-tors. Occupational injury: risk prevention and intervention.London: Taylor & Francis; 1998.

7. Short A. Beaches of the New South Wales coast. A guideto their nature, characteristics, surf and safety. Australianbeach safety and management program. University of Syd-ney; 2000.

8. Gilchrist J, Sacks JJ, Branche CM. Self-reported swim-ming ability in U.S. adults, 1994. Public Health Rep2000;115(2—3):110—1.