1. Presented at CARSP Conference
Analyses of casualty collisions and casualties during statutory holidays in Saskatchewan
Presented at CARSP Conference
Rajib Sahaji
George Eguakun
June 7, 2016

2. Background
Statutory holidays are commonly viewed as times of elevated risk with respect to motor vehicle collisions and severity.
In Saskatchewan, statutory holidays in 2014 constituted 16% all traffic fatalities; however, fatalities per day were 71% higher than fatalities per day during rest of the days of 2014.
Elevated collision risk during statutory holidays triggers enhanced traffic safety enforcements and publicity campaigns around major holidays – E.g. Victoria Day, Thanksgiving and Christmas (VTC)
Do enhanced safety initiatives around the major holidays significantly contribute to increased public safety?
So, Sask stat holidays in 2013 accounted for 463 casualty collisions, more than 30% increase from that in 2009.
In addition, when we consider per day traffic fatalities, stat holidays were associated with 10.5% higher fatalities per day than that during non holidays.
Elevated collision risk during statutory holidays triggered enhanced traffic safety enforcements and publicity campaigns around major holidays, like in Sask, we put additional safety initiatives around Victoria Day, Thanksgiving and Christmas (VTC) holidays.
However, limited research conducted to investigate the impacts of the enhanced safety initiatives during statutory holidays.

3. Objectives
To investigate:
The association of statutory holidays with casualty collisions and casualties
The extent of alcohol involvement in casualty collisions and resulting casualties during statutory holidays
The safety benefits of enhanced enforcement during Victoria Day, Thanksgiving and Christmas (VTC) holidays relative to other holidays.
So, what we would like to achieve in this study is to investigate a few things…
First, we want to know the association of …
Second, we would like to examine the relative risk of …. And
Our last objective is to determine the safety benefits….

4. Study Data Statutory holidays observed in Saskatchewan (2008-2013)
Period
New Year
Dec 29th – Jan 2nd (weekday varies each year)
Easter
Dates vary each year (Friday – Monday)
Family Day
3rd Monday of February (Friday – Monday)
Victoria Day
3rd Monday of May (Friday – Monday)
Canada Day
Jun 29th – July 2nd (weekday varies each year)
Civic Holiday
1st Monday of August (Friday – Monday)
Labour Day
1st Monday of September (Friday – Monday)
Thanksgiving
2nd Monday of October (Friday – Monday)
Remembrance Day
Nov 9th – Nov 12th (weekday varies each year)
Christmas
Dec 22nd – Dec 26th (weekday varies each year)
In this study we looked for stat holidays observed in Sask between 2008 and 2013, and there were 10 major statutory holidays…
Statutory holidays observed in Saskatchewan ( )

5. Study Data Frequency of casualty collisions and casualties
Holiday
Casualty Collisions
AR* Casualty Collisions
% of AR Casualty Collisions
Casualties
AR Casualties
% of AR Casualties
Canada Day
402 60
15% 22 11
50%
Christmas
376 27 7% 21 5
24%
Civic Holiday
336 71
21%
52%
Easter
234 36 6
55%
Family Day
308 42
14% 8 3
38%
Labour Day
341 56
16% 29 12
41%
New Year 9% 15
53%
Remembrance Day
394 28
Thanksgiving
363 25
44%
Victoria Day 54
18% 26
58%
Total Holiday
3,464
466
13%
199 87
Total Non-Holiday**
28,296
2,510
1193
446
37%
By looking at the descriptive collision data, we can see that during statutory holidays for the given study period, there were 3,464 casualty collisions. We used non-holidays as a reference group, and note that non-holidays include regular weekends other than stat holidays as well as all weekdays. So, there were 28, 296 casualty collisions during non-holidays. When it comes to the alcohol involvement in casualty collisions and resulting casualties, holidays were riskier than non-holidays. However, what we cannot tell right from the descriptive data is whether the increase in AR casualty collisions/casualties have been significantly influenced by other factors such as temporal factors, weather effects etc.
*AR – Alcohol Related
**Non-holiday includes regular weekends
Frequency of casualty collisions and casualties
by holiday and non-holiday period ( )

6. Model Development Two modeling techniques were employed:
- Negative binomial regression
- Logistic regression
Controlled for temporal factors (i.e., year, month, day of week) as well as holiday confounding factors
Temporal factors were used as surrogates for the effect of weather, legislative changes and enforcements
The holiday effect considered the aggregated statutory holiday and non-holiday periods
In the model development, we used negative binomial (NB) modeling procedure to model the casualty collisions and resulting casualties.
The model controlled for temporal factors, such as ….. and assumed that the control for temporal factors would give a surrogate control for the effect of factors such as…
The model also controlled for holiday confounding factors, and the holiday effect referred to ….

7. Model Development Two modeling techniques were employed:
- Negative binomial regression
- Logistic regression
Controlled for temporal factors (i.e., year, month, day of week) as well as holiday confounding factors
Temporal factors were used as surrogates for the effect of weather, legislative changes and enforcements
The holiday effect considered the aggregated statutory holiday and non-holiday periods
In the model development, we used negative binomial (NB) modeling procedure to model the casualty collisions and resulting casualties.
The model controlled for temporal factors, such as ….. and assumed that the control for temporal factors would give a surrogate control for the effect of factors such as…
The model also controlled for holiday confounding factors, and the holiday effect referred to ….

8. Results Results NB Regression Results: Risk of Casualty Collisions
(Aggregated Holiday vs. Non-holiday)
Parameter
Estimate
Standard Error
95% C.I.
Pr > |Z|
Relative Risk (RR)
Lower
Upper
Overall Casualty Collisions
Intercept
2.68
0.05
2.58
2.79
<.0001 -
Holiday
0.14
0.04
0.07
0.21
1.15
**Non-holiday
0.00 NA
AR Casualty Collisions
-2.13
0.08
-2.29
-1.97
0.47
0.39
0.54
1.60
Our primary variable of interest was holiday, so, I am showing the modeling results for the predictor holiday. Note that Type 3 GEE statistics indicated that Holiday was strongly related to the rate of overall casualty collisions as well as AR casualty collisions, while the year, month, and day of week had no significant relationship with collision rate.
As we can see, the parameter estimates for both overall and AR casualty collisions during holidays were positive, which means holidays were associated with greater risk of casualty collisions (overall and in terms of alcohol influenced) than non-holidays. The relative risk, which is the exponential of parameter estimate indicates perceived risk of collisions during holidays relative to non-holidays
**Reference level

9. Results Results NB Regression Results: Risk of Casualties
(Aggregated Holiday vs. Non-holiday)
Parameter
Estimate
Standard Error
95% C.I.
Pr > |Z|
Relative Risk (RR)
Lower
Upper
Overall Casualties
Intercept
0.55
0.10
0.35
0.74
<.0001 -
Holiday
0.60
0.07
0.47
0.73
1.82
**Non-holiday
0.00 NA
AR Casualties
-0.23
0.25
-0.74
0.37
0.94
1.14
2.55
In terms of actual casualties associated with collisions, they increase by over 80% during statutory holidays compared to normal days. The likelihood of a casualty resulting from alcohol use during a statutory long-weekend is about 2 and half times that which would occur on a non-holiday period.
**Reference level

10. Results Results NB Regression Results: Risk of Casualty Collisions
(VTC vs Less-targeted Holidays)
Holiday
95% C.I.
Estimate
Lower
Upper
Standard Error
Collision Rate/Day
Pr > |Z|
Relative Risk (RR)
Intercept
2.52
2.36
2.69
0.08
<.0001
Canada Day
0.04
-0.19
0.28
0.12
13.03
0.714
1.04
Civic Holiday
0.10
-0.01
0.20
0.05
13.75
0.070
1.10
Easter
-0.30
-0.08
0.06
10.32
0.001
0.83
Family Day
-0.17
0.29
13.25
0.606
1.06
Labour Day
0.14
-0.05
0.33
14.36
0.154
1.15
New Year
0.01
-0.34
0.37
0.18
12.62
0.950
1.01
Remembrance Day
0.27
0.52
0.13
16.30
0.041
1.31
**VTC
0.00
12.43 -
1.00
When we investigate the risk of casualty collisions for specific statutory holidays, we find that majority of the less-targeted holidays, except Easter and Remembrance day, were not significantly different from VTC holidays regarding casualty collisions.
The Easter holiday had a 17% lower risk of casualty collisions than VTC, while the risk of casualty collisions was 31% greater during the remembrance day holiday than the VTC.
**Reference level

11. Results Results NB Regression Results: Risk of Casualties
(VTC vs Less-targeted Holidays)
Holiday
95% C.I.
Estimate
Lower
Upper
Standard Error
Casualty Rate/Day
Pr > |Z|
Relative Risk (RR)
Intercept
1.20
0.92
1.48
0.14
<.0001
Canada Day
-0.31
0.59
0.23
3.82
0.5321
1.15
Civic Holiday
-0.20
-0.42
0.02
0.11
2.72
0.0752
0.82
Easter
-0.21
-0.74
0.33
0.27
2.69
0.4476
0.81
Family Day
-0.33
-0.57
-0.10
0.12
2.37
0.006
0.72
Labour Day
0.05
-0.34
0.43
0.20
3.47
0.8082
1.05
New Year
-0.30
-0.63
0.03
0.17
2.46
0.0793
0.74
Remembrance Day
0.46
0.22
3.38
0.9294
1.02
**VTC
0.00
3.32 -
1.00
Similar trends can be found when the less-targeted holidays are compared with VTC in regards to actual casualties. Again, No less-targeted holidays, except family day, showed statistically significant difference from VTC in terms of casualties.
The family day exhibits a 30% lower risk of casualties than the VTC.
**Reference level

12. Results Logistic Regression Results: Risk of AR Casualty Collisions
(VTC vs Less-targeted Holidays)
Holiday
Overall
AR Casualty Collisions
Nighttime
(6:01p.m. - 5:59 a.m.)
Odds Ratio Estimates
(C.I.)
Pr >
Chi-square
Canada Day
1.08
( )
0.674
1.26
( )
0.256
Civic Holiday
1.31
( )
0.101
1.40
( )
0.086
Easter
1.17
( )
0.458
1.62
( )
0.055
Family Day
1.03
( )
0.868
1.29
( )
0.292
Labour Day
( )
0.382
( )
0.204
New Year
0.97
( )
0.869
( )
0.919
Remembrance Day
0.66
( )
0.059
0.72
( )
0.226
**VTC
1.00 -
The results of logistic regression indicate that none of the less targeted holidays is significantly different from VTC holidays, when it comes to the likelihood of overall and nighttime casualty collisions .
**Reference level

13. Conclusion
The study supports the common perception of elevated collision risk during statutory holiday periods.
The risk of casualty collisions and resulting casualties was significantly higher during statutory holidays.
The risk of alcohol involvement in casualties during statutory holidays was substantially greater than that for non-holidays.
The risk of casualty collisions during less targeted statutory holidays did not differ significantly from the heavily enforced VTC holidays (except the Remembrance Day)
Easter, Remembrance Day, Family Day, and New Year holidays had higher overall alcohol related casualty collisions than VTC.
The study outcome supports the common perception of elevated collision risk during all holiday periods and warrants additional efforts to improving and promoting traffic safety around statutory holidays.
Safety promotional programs through radio/television advertisement and poster messages
are recommended on an on-going basis around holiday periods

14. Recommendations
Safety promotional programs are recommended on an on-going basis around all holiday periods.
Promotional messages and enhanced enforcements against drinking and driving are highly recommended during all holiday periods, especially Easter, Remembrance Day, Family Day, and New Year’s Day holidays to improve public safety.
This study did not include exposure count (traffic volume/vehicle miles traveled) in the modeling procedure due to data limitations
Future research should include exposure count as well as other factors such as, economic indicators (unemployment rate), population, and driver characteristics (age, gender).
Since the less targeted holidays were not significantly different from the heavily targeted holidays such as on Victoria, Thanksgiving, and Christmas holidays, it would be more beneficial to promote safety consistently during all holiday weekends.
This should be the case with respect to promoting and enforcing drinking and driving messages.

15. Questions?
Thank You!