20 Years of National Life Jacket Wear Observation Study
1. Presenting 20 Years of Data: National Life
Jacket Wear Rate Observational Study
Natalie Spitzer
Research Associate
Thomas W. Mangione
Senior Research Scientist
JSI Research & Training Institute, Inc.
Life Jacket Association 2019 Annual Conference
St. Pete Beach, Florida
May 14-15, 2019
2. National Life Jacket Wear Rate Observation
• 124 Sites in 30 states;
• 675,000 adult recreational boaters and nearly 140,000 boats observed
• Comparison of life jacket wearing behaviors among 8 different boat types
3. Dynamic Risk Factors:
Water temperature
Air temperature
Wind speed
Wave height
Current
Weather
Visibility
Boat movement
Boater position
Boater activity
Boater Impairment
Static Risk Factors:
Boat size
Life jacket regulations
Fishing tournament rules
Boater experience
Boater training
Perceived
Risk of
Capsizing or
Falling
Overboard
Perceived
Negative
Consequences
of Entry into
the Water
Decision
to Wear
Life Jacket
Beliefs in
Benefits/Drawbacks
of wearing Life
Jackets
# of passengers
Presence of children
Boater swimming ability
Boat type
4. Condition “RISKY” vs. “NON-RISKY” Justification
Visibility POOR vs. GOOD/FAIR
Difficulty navigating and operating boats, responding to obstacles or
other boats, finding boaters who have fallen overboard
General Weather
RAINING/STORMY
vs. SUNNY/CLOUDY
Risks related to visibility and wave height.
Strength of
Current
STRONG vs. WEAK/MODERATE
Increased chances of loss of control, capsizing, falling overboard,
boater/swimming fatigue
Water and Air
Temperature*
COLD WATER (<65°F)
and/or COLD AIR (<70°F)
vs. WARM WATER (≥65°F)
and WARM AIR (≥70°F)
Increased chances of hyperventilation, swimming fatigue, loss of
function, hypothermia
Wave Height and
Wind Speed†*
CHOPPY/ROUGH WAVES
and/or HIGH WIND
vs. CALM WAVES and LOW WIND
Increased chances of capsizing, falling overboard, swimming fatigue
Children on Board CHILD PRESENT vs. NO CHILD
Increased chances of entering water (to rescue child), boater
distraction, unpredictable movements that contribute to capsizes or
falls overboard
Size of Boat† SMALL vs. LARGE
Reduced stability, increased chances of capsizing and falling
overboard
Boater Activity†
FISHING/RACING/WHITE WATER
vs. OTHER (pleasure)
Increased chances of standing, loss of balance, entering water,
capsizing, falling overboard
Boat Movement†
MOTORING/PADDLING/SAILING
vs. OTHER (drifting/anchored)
Increased chances of loss of control, capsizing, falling overboard
Number of
Boaters†
SINGLE vs. 2+ BOATERS
Less likely to be rescued if falling overboard (no one to throw
flotation, search and rescue, report accident)
Boater Position PASSENGER vs. OPERATOR Passengers less aware of boating hazards
*Combined variables for analysis due to high correlation; †Risk classification varies by boat type
Categorizing Binary Risks
5. For each boat type: which risky conditions are significantly
related to higher rates of adult life jacket use?
– Chi-square test for equality of proportion
Risky condition
present?
Life Jacket Use?
No Yes
No a% A%
Yes b% B%
6. Runabout
Cabin
Cruiser
Pontoon
Skiff/
Utility
Canoe Kayak
Day
Sailor
Cabin
Sailboat
Cold Water
and/or
Cold Air
Poor Visibility o o o o
Choppy Waves
and/or
High Wind
o o o
Raining o o
Strong Current o
Bivariate analysis:
Environmental conditions associated with increased adult life jacket wear rates
: chi square test p<0.05 where presence of risk -> higher wear rate
o: risk not recognized - chi square test p≥0.05 or presence of risk -> lower wear rate
7. Runabout
Cabin
Cruiser
Pontoon
Skiff/
Utility
Canoe Kayak
Day
Sailor
Cabin
Sailboat
Small Boat
Size *
Dangerous
Activity o
Dangerous
Movement o o o
Children
on Board o o
Single
Boater o * o
Passenger
Position o o o o
Bivariate analysis:
Boat/Boater conditions associated with increased adult life jacket wear rates
* Opposite condition considered risky based on literature
: chi square test p<0.05 where presence of risk -> higher wear rate
o: risk not recognized - chi square test p≥0.05 or presence of risk -> lower wear rate
8. For each boat type: As the number of risky conditions
present increases, what happens to the adult life
jacket wear rate?
– Cochran-Armitage linear trend test (one-sided)
9. 0
20
40
60
80
100
0 risks 1 risk 2 risks 3+ risks
LifeJacketWear(%)
Number of Boating Risks
Skiff / Utility* Runabout* Cabin Cruiser* Pontoon*
Canoes* Kayaks* Day Sailors* Cabin Sailboats*
Adult life jacket wear rate, by cumulative risk count
* One-sided Cochran-Armitage linear trend test significant at p<0.0001
10. For each boat type: What are the top 4 risky
conditions that (in combination with one another) are
most significantly related to higher rates of adult life
jacket use?
– Automated stepwise logistic regression model selection
11. Boat type First OR
P-value
Second OR
p-value
Third OR
P-value
Fourth OR
P-value
Runabout
Cold
temperatures
(water and/or air)
1.9
****
Dangerous activity
1.9
****
Child present
1.7
****
Small boat
2.0
****
Cabin
Cruiser
Cold
temperatures
(water and/or air)
3.7
****
Small boat
1.5
****
Child present
1.7
****
Poor visibility
2.5
****
Pontoon Child present
2.1
****
Small boat
1.5
****
Cold temperatures
(water and/or air)
1.5
****
Raining or stormy
weather
1.8
****
Skiff
/ Utility
Cold
temperatures
(water and/or air)
2.1
****
Boating alone
1.6
****
Dangerous activity
1.3
****
Child present
1.4
****
Canoe
Choppy waves
and/or high wind
2.7
****
Child present
3.2
****
Small boat
1.9
****
Dangerous
activity
2.8
****
Kayak
Cold
temperatures
(water and/or air)
2.3
****
Strong current
3.1
****
More than one
boater
1.7
****
Choppy waves
and/or high wind
1.4
****
Day Sailor Small boat
2.9
****
Boating alone
2.2
****
Dangerous activity
38.8
****
Choppy waves
and/or high wind
1.4
****
Cabin
Sailboat
Cold
temperatures
(water and/or air)
3.0
****
Small boat
1.9
****
Dangerous activity
2.5
****
Sailing or
motoring
1.6
***
Multivariate Analysis: Automated stepwise model selection
“Top 4 priority” risky conditions significantly associated with adult life jacket use
Wald Chi-Square for MLE estimate: ****p<0.0001; *** p<0.001
12. First Second Third Fourth
Runabout
Cold temperatures
(water and/or air)
Dangerous
activity
Child present Small boat
Cabin Cruiser
Cold temperatures
(water and/or air)
Small boat Child present Poor visibility
Pontoon Child present Small boat
Cold
temperatures
(water and/or air)
Raining or stormy
weather
Skiff / Utility
Cold temperatures
(water and/or air)
Boating alone Dangerous activity Child present
Canoe
Choppy waves
and/or high wind
Child present Small boat Dangerous activity
Kayak
Cold temperatures
(water and/or air)
Strong current
More than one
boater
Choppy waves
and/or high wind
Day Sailor Small boat Boating alone Dangerous activity
Choppy waves
and/or high wind
Cabin Sailboat
Cold temperatures
(water and/or air)
Small boat Dangerous activity Sailing or motoring
Multivariate Analysis: Automated stepwise model selection
“Top 4 priority” risky conditions significantly associated with adult life jacket use
Wald Chi-Square for MLE estimate: ****p<0.0001; *** p<0.001
13. Boat type First Second Third Fourth
Runabout
Cold temperatures
(water and/or air)
Dangerous activity Child present Small boat
Cabin Cruiser
Cold temperatures
(water and/or air)
Small boat Child present Poor visibility
Pontoon Child present Small boat
Cold temperatures
(water and/or air)
Raining or stormy
weather
Skiff / Utility
Cold temperatures
(water and/or air)
Boating alone Dangerous activity Child present
Canoe
Choppy waves
and/or high wind
Child present Small boat Dangerous activity
Kayak
Cold temperatures
(water and/or air)
Strong current
More than one
boater
Choppy waves
and/or high wind
Day Sailor Small boat Boating alone Dangerous activity
Choppy waves
and/or high wind
Cabin Sailboat
Cold temperatures
(water and/or air)
Small boat Dangerous activity Sailing or motoring
Multivariate Analysis: Automated stepwise model selection
“Top 4 priority” risky conditions significantly associated with adult life jacket use
Wald Chi-Square for MLE estimate: ****p<0.0001; *** p<0.001
14. Boat type First Second Third Fourth
Runabout
Cold temperatures
(water and/or air)
Dangerous activity Child present Small boat
Cabin Cruiser
Cold temperatures
(water and/or air)
Small boat Child present Poor visibility
Pontoon Child present Small boat
Cold temperatures
(water and/or air)
Raining or stormy
weather
Skiff / Utility
Cold temperatures
(water and/or air)
Boating alone Dangerous activity Child present
Canoe
Choppy waves
and/or high wind
Child present Small boat Dangerous activity
Kayak
Cold temperatures
(water and/or air)
Strong current
More than one
boater
Choppy waves
and/or high wind
Day Sailor Small boat Boating alone Dangerous activity
Choppy waves
and/or high wind
Cabin Sailboat
Cold temperatures
(water and/or air)
Small boat Dangerous activity Sailing or motoring
Multivariate Analysis: Automated stepwise model selection
“Top 4 priority” risky conditions significantly associated with adult life jacket use
Wald Chi-Square for MLE estimate: ****p<0.0001; *** p<0.001
15. Boat type First Second Third Fourth
Runabout
Cold temperatures
(water and/or air)
Dangerous activity Child present Small boat
Cabin Cruiser
Cold temperatures
(water and/or air)
Small boat Child present Poor visibility
Pontoon Child present Small boat
Cold temperatures
(water and/or air)
Raining or stormy
weather
Skiff / Utility
Cold temperatures
(water and/or air)
Boating alone Dangerous activity Child present
Canoe
Choppy waves
and/or high wind
Child present Small boat Dangerous activity
Kayak
Cold temperatures
(water and/or air)
Strong current
More than one
boater
Choppy waves
and/or high wind
Day Sailor Small boat Boating alone Dangerous activity
Choppy waves
and/or high wind
Cabin Sailboat
Cold temperatures
(water and/or air)
Small boat Dangerous activity Sailing or motoring
Multivariate Analysis: Automated stepwise model selection
“Top 4 priority” risky conditions significantly associated with adult life jacket use
Wald Chi-Square for MLE estimate: ****p<0.0001; *** p<0.001
16. Situational Awareness Discussion
• Risky conditions associated with higher rates of adult life jacket
use varies by boat type
– First step of Situational Awareness: boaters evaluate/respond to risks
based on the type of boat they are on
• The bivariate and cumulative risk analysis shows that despite
differences by boat type, boaters are more likely to wear their
life jacket when risky conditions are present
• These associations strongly suggest that boaters understand the
connection between:
– Awareness of perceived risk Unexpected water entry
– Unexpected water entry drowning
– Wearing a life jacket drowning prevention
17. • Multivariate analysis: Lower priority risks are still important!
– Many also associated with higher rates of life jacket use
– Some risks not recognized: not significant; related to other “higher priority”
risks; opposite association
– Worth trying to understand why these risks are not in the top 4 in order to
tailor educational campaigns / interventions that prevent drowning
• Framing is important: drownings occur in all situations.
– How to teach adult boaters how to become more situationally aware and to
respond to risk by wearing a life jacket? While ALSO understanding that:
– No matter what the perceived risks are, boaters should ALWAYS wear a life
jacket while boating because risks can change?
Situational Awareness Discussion, continued
Editor's Notes
Natalie Spitzer
Research associate at JSI
Will be presenting findings from 20 years of data collection that will demonstrate with real data key components of the theoretical model you’ve just learned about
We’ve been working with the US Coast Guard and Coast Guard Auxiliary to conduct boating observations in the summer across the US for the past 20 years
This map shows in blue the states where we have been collecting data
Overall we are at 124 different sites in 30 states (2 in California)
The findings I’m presenting today draw from a sample of about 675,000 adult recreational boaters observed in nearly 140,00 boats
This is the framework Tom presented earlier
The stars indicate which risk factors JSI is able to capture during data collection
Even if we can’t measure boaters’ perception of risk or negative consequences of unexpected water entry, or beliefs in the benefits of wearing life jackets, we can get a sense of how different boat types and risk conditions influence adult life jacket use
The first thing we did was categorize each indicator into a binary form based on risk of unexpected entry into the water and/or drowning
We conducted a literature review to guide this categorization process
You can see in this table how each variable was categorized into either risky or non-risky conditions, and our justification
There are a few cases where the risky condition is different by boat type:
Dangerous wind varies by boat type: sail and power = 6+ knots; paddle = 4+ knots
Size of boat: large kayaks riskier than small; exact boat size dependent on distribution of boats most likely in that specific boat category
Number of boaters: more than one boater in kayak riskier than single boating
Dangerous boat movement varies by boat type: sail = sailing or motoring only; power=motoring only; paddle = paddling or motoring only
Boater activity varies by boat type: sail = fishing or racing only; power=fishing, racing or white water if any; paddling = fishing, racing or white water if any
The first thing we did was figure out, for each boat type which risky conditions were significantly related to higher rates of adult life jacket use
We did this using chi-square tests for equal proportions
For each boat type and each binary risky condition, we compared the proportion of adults wearing life jackets when the risky condition was present and not present
Here’s what we found.
In this table you’ll see that 8 boats types of interest are along the columns
The binary risky conditions are along the rows.
Wherever you see a checkmark is where we found that adult life jacket wear rates were significantly higher in the riskier condition than in the non-risky condition
Wherever you see a red circle is where this was not the case – we’ll call that an unrecognized risk
As you can see for most of our environmental binary risks, there are very few non-significant findings
Here is the same thing for boat and boater risk factors.
Again we are seeing very few red circles, and lots of checkmarks that indicate that in riskier situations adult boaters are wearing their life jackets at higher rates than in lower risk situations
* Opposite condition considered risky based on literature. Therefore opposite risk tested for significance. Ex. larger kayaks are more likely to tip than smaller, therefore larger kayaks were classified as risky. Checkmark indicates that larger kayaks are associated with larger proportion of boaters wearing life jackets than smaller.
The next step in our analysis was to figure out how adult life jacket wearing behaviors change as the number of risky conditions changed.
Based off of our situational awareness model, we would expect to see higher rates of life jacket use in riskier situations, and also expect to see slight differences by boat type
And that is what we see.
Here are the results for all boats, all of which had statistically significant linear trends based off of the Cochran-Armitage linear trend test that show positive increases
Things to notice here are the starting and end life jacket wear rates for each boat as well as the slope to get a sense of how life jacket use varies by boat type as the number of risks changes
Now the last thing we did was try to sort through all of these significant risky conditions to determine which were the most important for predicting adult life jacket use.
To make things simpler I’m just going to show you the results from a stratified automated stepwise logistic regression model selection
We’ll be comparing the first 4 risky conditions selected by this procedure by boat type
Here are the results of the stepwise selection
Along the rows we have each boat type and in the columns we have the top 4 priority risks and odds ratios chosen in the automated model building procedure
You can see by the odds ration that all are greater than one, so adults are more likely to be observed wearing life jackets when the listed risky condition is present compared to the non-risky alternative
To make this more transparent I’m going to switch between slides to pull out some key patterns
Example interpretation: On runabouts, the odds of wearing life jacket when a child is present is 1.7 times that of the odds of wearing a life jacket without a child on board, controlling for boater activity, water/air temperature, and boat size. This association is statistically significant at p<0.0001
Here in this first slide I’ve highlighted how cold air and water temperatures appears as a priority risk in all but 2 boat types
In 5 cases we see that it was the first variable selected out of all 11 risk factors
In this slide you can see that small boat size is a top 4 significant predictor for all boats except skiffs and kayaks (for kayaks small boats isn’t a risk factor)
Here we can see that child presence is a significant predictor for all powerboats and canoes
Boats known to tip: choppy/rough water and/or high winds significant predictor