Geographical variations of motor-vehicle injuries in Quebec, 1983–1988

Geographical variations of motor-vehicle injuries in Quebec, 1983–1988

She.Sci.Med. Vol.33.No. 4,pp.415-421,1991 Printed in GreatBritain 0277-9536/91 $3.00+0.00 PergamonPressplc GEOGRAPHICAL VARIATIONS OF MOTOR-VEHICLE ...

783KB Sizes 0 Downloads 37 Views

She.Sci.Med. Vol.33.No. 4,pp.415-421,1991 Printed in GreatBritain

0277-9536/91 $3.00+0.00 PergamonPressplc

GEOGRAPHICAL VARIATIONS OF MOTOR-VEHICLE INJURIES IN QUEBEC, 1983-1988 JEAN-PIERRE THOUEZ,’ MARIE-FRANCEJOLY,~ANDRE RANNOU,’ YVES Bu.ssrE~~’ and ROBERTBOURBEAUJ

‘Dipartement de gkographie, Universitt de Montreal, Unite de recherche en tpidkmiologie, HBtel-Dieu de Montrbal, BP 6128, SUCC.A, MontrCal,._Q&bet. Canada H3C 3J7. 2Deoartement de mtdecine sociale . et prtventive, Universitt de Montrtal, Centre de recherche sur les transports, UniversitC de Montreal, )Institut national de la recherche scientifique (INRS)--Urbanisation, Universite du Qutbec and 4Departement de demographic, UniversitC de MontrCal, Centre de recherche sur les transports Universitt de Mont&al Abstract-Data furnished by the R&e de I’Assurance Automobile du QuCbec (RAAQ) were used to describe the geography of motor vehicle accidents in the province of Quebec for the period 1983-1988. These were also used to evaluate the risk factors associated with zones of high risk with regards to accidents for the 97 Municipalids Regionales de CorntO (MRC). The results demonstrate that non-severe accidents are more frequent in the urban context. On the other hand, severe accidents are more frequent in the rural context. The Standard Morbidity Ratio (SMR) highlights those MRC’s with the risk of severe or non-severe accidents, where risks are twice that of Quebec as a whole. The demographic characteristics (age, sex) of the driver and passengers and the place of the accident (region, density) were used in the model LOGIT to evaluate risk factors associated with high risk zones. Results of the model for those severely injured are different from those for non-severely injured people. This holds true for the location of accidents as well as for demographic characteristics. In both models, women and people over 65 years of age are low-risk groups for accidents. The authors indicate certain action to be undertaken in Quebec by the Government to improve this situation. Key words-motor

vehicle accidents, high risk zones, LOGIT model

According to annual statistics compiled by the Regie de I’Assurance Automobile du Quebec (RAAQ), during the period 1983-1988, the incidence of injuries to drivers and passengers of motor vehicles appear to have been stable or to have decreased slightly for severe injury (hospitalization required), while increasing among those less severely injured [l]. There were approximately 45 injured for each death. Road accident injuries are also one of the main causes of shortand long-term disability. Such trauma has an impact on the level of activity and on expectations concerning a healthy life [2-51. A detailed demographic and epidemiological study by Bourbeau [6] showed that between 1926 and 1978, morbidity was in many ways a significant dimension of road accidents in Quebec: there were substantial numbers of injured, a sizeable proportion of whom (around 20%) required hospitalization. Socio-economit costs were high because of the age group of the population most involved: young adults between 15 and 35. In 1982, the average total cost for severely injured individuals showing functional disability was estimated at %400,000 [7]. Despite its importance, the image of morbidity remains relatively vague. According to Bourbeau (61, two reasons explain this situation: mortality, and available data are scarce in many countries [8]. In general, the incidence estimates are believed to be conservative, both in overall magnitude and in distribution by severity [9]. In Quebec, the RAAQ data banks were used in several studies which sought either to present the demographic and epidemiologi415

cal profile of this phenomenon [6, 10, 111, or to analyze its spatial variations in peripheral regions [ 121, in urban areas [ 13-151 or in rural areas [16]. Other researchers have sought to measure the role of explanatory factors using various models [14]. In most cases, geographical and some descriptive studies have sought to identify dangerous sites or areas where a high concentration of road accidents are observed. In this study, we will first analyze the geographical distribution of morbidity in the Municipalit& R&gionales de Comte (MRC) between 1983-1988. The 1979 provincial planning law (Loi sur I’amCnagement et l’urbanisme), modified in 1983, defines an MRC as a corporation with legal rights over a territory for which letters patent were delivered in order, to design a regional plan, among other things [ 181.There are 97 MRC’s in Quebec including two Urban Communities, Montreal (CUM) and Quebec City (CUQ), and the Communautt rCgionale de l’outaouais (CRO). We have excluded region 010, the Kativik regional administration and Northern Territories, which falls under the James Bay Convention. Population density varied between 0.07 inhabitants/km2 in the Caniapiscau MRC and 1798 in the Champlain MRC. It reached 3415 in the CUM, 219 in the CUQ and was about 21 in the CRO. In the second part of the study, a LOGIT regression model was used to single out explanatory factors that are related to the geographical variations of morbidity, while separating those who suffeed severe injuries from those who did not require hospitalization.

416

JEAN-PIERRE THOUEZ et

Table

I. Frequency

and average annual Total

Age

rate (per l00.000)

of non-severely

MRC

injured Urban

Men

Women

Men

PI-O”0

N

Rate

N

Rate

N

Rate

34 3H 45-54 55-64 ~65

1134 6779 5062 2795 1430 779 515

155.20 1329.01 856. I4 574.70 425. I5 269.46 196.08

1277 4086 3330 2215 I250 818 595

184.61 861.60 555.33 449.92 360.81 253.56 154.02

896 5322 4085 2337 1213 649 419

150.59 1301.22 821.93 565.86 425.61 270.18 205.39

18,494

58 I .23

13,571

409.60

14,921

564.12

Total

THE

DATA

Statistics originate from the RAAQ for the period 1983 to 1988. These statistics correspond to maps of localities where accidents have occurred. According to a case study conducted by Morin ef al. [ 161, a certain number of errors relating to localization may be attributed to a pre-established division of space referring to Mercator, a system used by the police. With respect to the larger zone of the MRC this problem disappears or at least was less significant. Comparing the statistics of hospitalized injuries from road accidents to data from the Ministry of Transport, Bourbeau [6] has observed significant differences between the Ministry of Transport data and hospital morbidity. About 22% of these differences have an impact on the indices of gravity (seriousness) of injuries, measured either by the AIS (Abbreviated Injury Scale), or by the ISS (Injury Severity Score). Bourbeau and Messier [15] noted elsewhere that the definitions used to classify accident victims as severely or non-severely injured may be at the root of differences between the data and the reality. According to the RAAQ, an injury is classified as severe when there is hospitalization (one or more days in hospital). Availability of hospital beds may influence the distribution by severity [13, 191. However, when individuals die at the hospital as a result of a traffic accident, those accidents are classified as fatal. If death occurs after this period, the accident is classified as severe rather than fatal. The authors estimate that the gap between statistics and reality is between 10% and 20%. This problem of classification must be considered when interpreting the results of this study and when making direct comparisons to other studies. Since the age effect must be taken into account, we used the indirect method of standardization. The crude rate for the standard population (mean number of cases reported between 1984 and 1988 for age groups of 10 year spans according to sex for the whole of Quebec) was applied to the given population Table 2. Frequency

and average annual Total

to sex and age, Quebec,

MRC

Rural MRC Mea

Women N Rate 1038 3416 2824 1943 1087 700 511 Il.519

19831988

WClmen N Rate

N

Rate

182.69 854.00 553.73 459.34 365.99 254.71 163.26

238 1457 977 458 217 130 96

178.95 1798.77 1039.36 594.81 409.43 245.28 188.24

239 670 506 272 163 II8 84

192.24 892.97 562.23 367.57 307.55 240.82 137.70

414.95

3573

666.60

2052

387.17

RESULTS

The demographic morbidity

composition

of

motor-vehicle

Tables 1 and 2 show the mean number and average annual rate of non-severe and severe injury victims according to sex and age in Quebec from 1983 to 1988. For all MRC’s, there is an average of 1.36 males non-severely injured for one non-severely injured female, while for severe injuries the ratio is 1.74: 1. Drivers and passengers between 16-24 years have the highest morbidity rates for both sexes. Globally, there is a negative relationship between traffic morbidity and age. In these tables we also separated the urban MRC’s, from the rural MRC’s. An MRC is defined as urban when more than half of its population dwells in an urban region: the population of an urban region must be greater than 1000 and the density must be over 400 inhabitants/km2 [18]. For the non-severely injured, morbidity rates are higher in urban MRC’s for women and in rural MRC’s for men. In both cases the 16-24 age group was clearly over-repesented. For severe injuries, the rates are higher in rural MRC’s regardless of sex, and among men in the 16-24 age group have the highest rate. The higher morbidity among men when compared to women is obvious in all age groups. For victims younger than 15 years of age, the average

rate (per 100,000) of severely injured Urban Men

Women

cases according

to obtain the expected number of cases that would exist if the given population was subject to the schedule of rates of the standard population [20]. The standardized morbidity ratio (SMR) is the ratio of the actual to the indirect adjusted rate. The SMR was calculated for severe and non-severe injuries for each of the 97 MRC’s. In order to decide if there was a statistically significant difference between the actual and indirect adjusted rates, a Poisson test following the Bailar and Ederer [21] methods was used.

MRC

Men

al.

Age group 65

N

Rate

N

Rate

N

Rate

146 1033 765 404 210 147 II3

20.02 212.12 129.39 83.07 62.43 50.85 43.02

I25 478 371 256 163 I21 I05

18.07 100.79 61.87 52.00 47.05 37.51 27.18

92 705 547 295 I48 102 79

15.46 172.37 110.06 71.43 51.93 42.15 38.73

Total

2818

88.56

1619

48.86

1968

74.41

cases according

to sex and age, Quebec, 1983-1988 Rural

MRC Women N Rate 89 352 283 I95 II9 82 79 II99

Men

MRC Women Rate N

N

Rate

16.01 88.03 55.49 46. IO 40.07 29.60 25.24

54 328 218 109 62 45 34

178.95 404.94 231.91 141.56 116.98 93.75 66.67

36 126 88 61 44 39 26

29.03 161.54 97.78 82.43 83.02 79.59 42.62

43.19

850

158.58

420

79.25

Geographical variations of motor- vehicle injuries in Quebec, 1983-1988

annual rate is greater among women when nonsevere injuries are considered while the same is true for men in tbe case of severe injuries. Geographic variations of motor-vehicle

morbidity

Combined severe and non -severe injuries. In a recent

article concerning road accident mortality measured in MRC’s in Quebec, three kinds of indicators were used: population, the number of vehicles in circulation and the number of driver’s licenses [22]. The strong Pearson and Spearman correlations between these variables resulted in our retaining population only as a denominator for the remainder of the study. The average number of non-severely and severely injured persons per MRC between 1983-1988 was 429 (SD = 1002). The distribution ranged from 13 (Caniapiscau) to 9498 (CUM). The average rate observed was 730.41 injured per 100,000 population. Some of the MRC’s had a high SMR (Standardized Morbidity Ratio) for non-severely and severely injured person that was statistically significant at the 99% probability level. MRC’s for Mirabel (418.72), Montcalm (230.07), the Pays-den-Haut (275.33), the Laurentians (23 1.87), Matawinie (220.44) and Pontiac (217.97) showed the highest levels. Non-severe injuries. The average number of those non severely injured per MRC was 383.03 (SD = 938.09). The distribution varied between 11 (Caniapiscau) and 8878 (CUM). The average annual rate was 625.66 per 100,000 population. Mirabel has

417

the highest number of occurrences (2055.96). Half of the geographic areas had a high level that was statistically significant at the 95% probability level. Except for Matawinie (44.26) and Pontiac (197.13), the MRC’s listed in the above paragraph had the highest SMR, twofold of what would be expected from province-wide morbidity. Figure 1 illustrates the geographic variation of the SMR while differentiating between urban and rural MRC’s. Severe injuries. The average number of severely injured persons per geographic area was 45.64 (SD = 65.12) in our study. The distribution varied between 3 (Caniapiscau) and 620 (CUM). According to the variation coefficient (SD/mean), the geographic dispersion of the data distribution was stronger for victims of accidents with non-severe injuries (2.44) than for those of accidents involving severe injuries (1.42). The average rate was 104.75 per 100,000 and varied between 19.60 (Laval) and 346.72 (Mirabel). More than one third of the areas had a high level that was statistically significant at the 95% level. Cote de Beauprt (289.84). Haut-Saint-Francois (216.84), Coaticook (212.53) Brome-Missisquoi (203.20), Jardins-de-Napierville (239.05), Mirabel (492.98) Montcalm (297.43), Pays-d’en-Haut (237.40), the Laurentians (249.18), Matawinie (247.13), Papineau (25 1.22), Vallie-de-Gatineau (256.41), Pontiac (333.49) and Antoine-Labelle (247.05), all showed high levels of SMR twice the expected provincial base (Fig. 2). It should be noted that these MRC’s are

URBAN

M.R.C.

RURAL

M.R.C I-.

-1

srandard

deviation

(__)

-0.99 to io.99

;--:

standard

deviation

&--l

deviation

,c --. \._ _.j

standard

Fig. 1. Standard morbidity ratio of motor-vehicle driver and passengers: non severe injuries, Quebec 1983-1988.

I

JEAN-PIERRETHOLEZ et

418

al

URBAN

0

M.R.C.

-1

standard

-0.99 to standard

cl

=?

RURAL

+I

standard

.-.

M.R.C

deviation

(__)

l0.99

:--; L ._:

deviation

deviation

c --. :.__.)

Fig. 2. Standard morbidity ratio of motor-vehicle driver and passengers: severe injuries, Quebec 1983-1988.

mostly rural and that the four MRC’s which showed a high level of risk for non-severe injuries are also part of this list.

The explanatory Xl Age

1 16-24 years 2 25-64 years 3 65 years over

X2 Sex

0 Man 1 Woman

X3 Region

0 Rural 1 Urban

X4 Density

0 b;;.

Risk factors and motor -vehicle morbidity

Recently the Quebec [23] and French [24] records provided detailed information on the user, the vehicle, the infrastructure and the accident itself. In this study, we retained only four parameters which do not completely explain the complexity of the phenomena. Our objective is to single out those parameters that are associated with zones having high levels of risk for motor-vehicle injuries. In order to do this, we have chosen the LOGIT model [25]. Interpretation is identical to that of multiple regression, however, it must be remembered that the dependent variable is dichotomous: its value is 1 when the average annual morbidity rate is higher than average and 0 when it is below the average. Thus, it is a special case of the multiple regression model designed to deal with the situation when we have only the measurement of presence or absence occurrence or non-occurrence of a particular factor. The program we used was the BMD PLR version (or BMDP-83) used by Clark and Hoskins [25] in their study on residential mobility. The program generates design variables for each categorical variable. The latter are used in the model instead of the value or category numbers recorded for the covariate (26,271.

variables are:

than

50 inhabitants/

1 Over 50 inhabitants/km2t *50 people per square kilometre. TThreshold equals one standard deviation above the mean. The design of the model facilitates the evaluation of the risk factors associated with high morbidity as a function of the characteristics of the victims (age and sex) and of geographic area (region and density). The BMD program is based upon the notion that the logistic model is a variation of the log linear model. According to Clark and Hoskins [23], the model (with a maximum likelihood procedure) would take the following form: log e =

;s =

B, + B,X, + . . . + &X4

Geographical variations of motor-vehicle injuries in Quebec, 1983-1988

419

Table 3. Results of the LOGIT analysis: severely injured cases, Quebec, 1983-1988 Independent variables

Estimated coefficient

Standard error

coefficient/ standard error

Transformed coefficient

Sex: Men Women

-0.64

0.72 E 0.01

-8.88

-0.17

Age: 16-24 25-64 65 and over

-0.31 -2.26

0.12 0.17

-2.46 - 12.73

-0.08 -0.07

Region: Rural Urban

-0.19

0.74 E 0.01

-2.61

-0.05

Density: t50 hab/km’ > 50 hab/km’

-0.67

0.10

-6.42

-0.18

Constant

-0.28

0.13

-2.14

pz=,

_(-598.68)=0,26, (-814.52)

where Pi/i is the likelihood (or the odds) that an accident will occur in a high risk geographic unit and PO/i is the likelihood that it will not.

In the first column of Tables 3 and 5, the independent variables are identified; in the second, the LOGIT regression coefficient appears; in the third, we find the standard error and in the fourth column, the coefficient standard error ratio is found. Following Clark and Hoskins [25], the value of the logistic regression coefficient corresponding to a particular category of an independent variable is found by multiplying the value of each design variable for that category by the corresponding logit regression coefficient and summing. In our study, calculations did not differ from the values of the estimated coefficients because of the low number of categories per variable. As a result, in the last column of Tables 3 and 5, we indicated the antilog of the estimated coefficient. In Table 3, the negative values for the parameters of women aged 25 and over, urban regions and densities greater than 50 represent attributes of the low risk zone for severely injured cases, in other words, the likelihood of a low risk zone for severely injured cases being associated with these parameters. For non-severely injured cases, the principle differences from the estimates indicated above are: the density variable is not significant at a 95% confidence level, the 25 to 64 age group (and also 16 to 24 age group) and the urban region are positively associated with the high risk zone for non-severely injured (Table 5). In the two tables, there is a (expected) decline in the likelihood (odds) of an accident occurring among women and age group 65 years and over. To obtain a measure of goodness of fit similar to R’, Clark and Hoskins [25] suggested the calculation ofp*. The estimates p2 = 0.26 for the severely injured and p* = 0.30 for the non-severely injured indicate that the models are an extremely good fit for the data [28]. Tables 4 and 6 illustrate the probability associated with several of the associations according to the formula derived from discrete choice [29]. For example, the highest probability to have a severe Table 4. Prediction associated probabilities according to the LOGIT analysis: severely injured cases, Quebec Severely injured cases Women Men

65 and over

Urban

> 50 hab./km*

16-24

Rural

< 50 hab./kmz

P(s) = 0.01 P(s) = 0.98

vehicle injury accident is associated with the following: men, aged 16-24, rural MRC and low density of population (Table 4). On the other hand, there is little difference in the high probability of having a nonsevere injury vehicle accident between men and women between 16-24 years of age in the urban MRC’s. DISCUSSION

The matter of road accident morbidity is important because as with mortality, its amplitude is considerable and confirms the high indirect burden of motor vehicle accidents to families and society. The results of this study confirm an over-morbidity for males which has already been documented in the literature. On the other hand, we note that the ratio of males with severe injuries is higher than that for men with lesser injuries. The latter can be explained by the degree of exposure to the risks involved: more precisely men may drive more quickly and for longer distances than females. The fact that there are more serious accidents in rural areas than in urban areas can also be explained by the difference in geographical accessibility of health care between these two milieus. The population at risk for accidents is the age group of 16-24 years. The propensity of involvement in an accident for this age group may be seen throughout Quebec and this justifies a province-wide program of accident prevention organized by the R.A.A.Q. We have also remarked that individuals aged 25-64 years are the second group most at risk for non-severe injuries. The results of the LOGIT analysis reveal two explanatory models that are more effective for the analysis of the demographic characteristics of injured persons than of the locational characteristics. The LOGIT model is a significant outlet for the isolation of specific categories of variables related to the hypothesis (mean rate above average vs below average) and to evaluate the probabilities of those categorized responses. The results are consistent with our n priori expectations. Urban MRC’s show higher risk for non-severe injuries while risks of severe injuries are higher in the rural MRC’s. In both models, women 65 years and over are positively associated with low risk vehicle accidents [6]. The geographic division of the MRC’s is determined administratively. There is a strong difference in

420

JEAN-PIERRE THOUEZet al. Table 5. Results of the LOGIT analysis: non-severely injured casts, Quebec, 1983-1988 Independent variables

Estimated coefficient

Standard

Women

-0.44

Age: 16-24 25-64 65 and over

0.18 -4.00

coefficient/ standard error

Transformed coefficient

0.72 E 0.01

-6.14

-0.12

0.24 0.41

0.74 -9.66

0.05 -1.50

2.01

0.40

UT0r

Sex: Men

Region: Rural Urban

0.14

0.79 E 0.01

Density: < 50 hab/km* :, 50 hab/km’ Constant

-0.26

p’ = , - (-563.21) (-814.65)

0.23

Table 6. Prediction associated probabilities according to lhc LOGIT analysis: non-severely injured cases, Quebec

65 and over 65 and over 16-24 16-24

In Collection Demographic Canadienne, Vol. 424. Uni-

versity of Montreal Press, Montreal, 1983. 7. All Industry Research Advisory Council. Automobile

Non-severely Injured Cases Women Men Women Men

-1.11

= 0.30,

Rural Rural Urban Urban

P(s) P(s) P(s) P(s)

= = = =

0.0080 0.0300 0.9512 0.9840

the type of road system and its upkeep (for example, irregular salting etc.). In this study, the geographical variations of traffic morbidity show the high risk regions such as the Eastern Townships (south east of Montreal) and the Laurentian axis (north of Montreal) as areas which deserve more in depth analysis looking at traffic intensity road network generally higher density on weekends and traffic lights in those specific areas. Finally, the data are not free from error of localization. One can argue that the error is uniformly spread throughout the area; thus, there is not an important difference in the overall picture of traffic accidents. Nevertheless, the need to improve the data is obvious if one wants finer measurements or if one wants to make comparisons with other studies. The geography of motor-vehicle accidents constitutes an as yet little developed approach in the area of geography of health. The perspective is not exclusively epidemiological but also social, cultural and behavioral. Such an approach demands that the geographer be open to other disciplines. REFERENCES 1. Use of the new form to report accidens began 1January 1978. On this form, oolice officers must distinauish between serious injured victims and those less set-i&sly injured. Also on 1 March 1978, the Quebec Public Auto Insurance Plan (RAAQ) came into effect. 2. Cowley R. A. Accidental death and disability: The neglected Disease of modern society-Where is the fifth component? Am. Emerg. Med. 11, 582-585, 1982. 3. Dill&d S. Duree ou Qu&P de Vie? Conseil des Affaires Sociale. Collection la Sante des Outbecois. Outbec. 1983. 4. Baker S. P., Whitfield R. A. and O’Neil B. The Injury Facr Book. DC. Heath, Lexington, MA, 1983. 5. Rice D. P., MacKenzie E. J. ef a!. Cosf of Injury in the United States. A report to Congress produced by the Institute for Health and Aging. University of California, San Francisco, 1989. 6. Bourbeau R. Les Accidents de la route au Quebec 19261978, etude dtmographique et ipidimiologique.

Injuries and their compensation in the United States. In Automobile Injury Costs in Michigan, New Jersey and Pennsylvwiu. A.I.R.A.C.. Washington, 1982. 8. EURO.

9.

10. 11. 12.

13. 14. 15.

16.

Les statistique relatives aux accidents de la route. In Rapports et .&u&s, Vol. 19. World Health Organization, Copenhagen, 1981. In the United States, the major data sources used in estimating numbers of injuries are the National Morbidity Detail-File (NMDF) for deaths, the National Hospital Discharae Survev MHDS) for live hosoital discharges, and the National Health Interview S&vey (NHIS) for less severe non-hospitalized injuries. The number of hospitalized injuries is underestimated since admission to the Veterans Administration and other state or federal hospitals are not included in the NHDS sample. Also, the NHIS collects information only on household members who are not institutionalized at the time of the interview. The NHIS is based on self-reports, unlike the NMDF and the NHDS which are based on medical records. Laberge-Nadeau C. and Bourbeau R. Mortalitt et morbidite par accidents de la route au Canada, evolution 1960-1974. Routes Transports 10, 14-19. 1979. Charest S. and Beaudoin R. Approches de Recherche sur les Sites Dangereux. Internal Report, Rimouski Community Health Department, QuCbec, 1987. Pouliot M., Morin D., Audet B., Beaudoin M., Letendre M. and Vandennissen P. Aspects geographiques des accidents routiers en regions piripheriques au Quebec. In Rapport FCARIRAAQ, Vol. 1 86-24-0060. D. 154. Geoaraohv Deoartment. Sherbrooke Ur&rsity, Sherbrooke, ~Qu&c, 1986. Joly M. F. GCographie des Accidents de la Route Cher les Enfants en Milieu Urbain. Doctoral Thesis, p. 246. Geography Department, University of Montreal, 1985. Saint-Laurent G. Importance des accidents de la route a Rouyn-Noranda. Routes Transporfs 15, 17-23, 1985. Bourbeau R. and Messier C. Les sites dangereux sur file de Montreal et les accidents de la circulation routiere, 1984-1986: Principaux resultats. Routes Transports 19, 14-21, 1989. Morin D., Pouliot M. and Vandermissen M. H. La localisation de la delinquance routiere et des accidents automobiles en milieu rural. Roures Transports 19. 33-38,

1989.

17. Gaudry M. Explication des Accidents de la Route au Quebec de 1954 d 1982: Quelques R&hats du Mode/e DRAG. Road Security Colloquium, ATR, Montreal,

June 1986. 18. Commission de la Representation Electorale du Quibet. MRC Socio-Economic Dossier, Vol. l-4. Quebec Government, 1982.

421

Geographical variations of motor-vehicle injuries in Quebec, 1983-1988

category 1 2 3

Publ. Hlth 77, 35833600, 1987. 20. Fleiss J. L. Statistical Methoa!s for Rates and Proportions. Wiley, New York, 1981.-

21. Bailar J-C. N. and Ederer F. Significance factors for the ratio of a Poisson variable to its expectation. Biometrics 20, 639643, 1964. 22. Thouez J-P., Joly M. F., Bussiere Y., Bourbeau R. and

Rannou A. La gtographie de la mortalitt sur la route au Quebec, 1984-l 988. Colloque Inegalites GCograph iques de la Mortalite, Population and Geographical Health Commission, IGU, Lille, France, April, 1990. 23. Vandermissen M. H., Pouliot M. and Morin D. Les traumatismes routiers et leurs caracteristiques geographiques dans les regions p&iphCriques du Quebec.

27. 28.

Routes Transports 18, 20-26, 1988. 24. Fleury D., Yerpez J. and Michel J. E. Lieux des

accidents, profils des accidents et des deplacements. Rdcherche Transports Securith 23, 11-18, 1989. 25. Clark W. A. and Hoskins P. L. Statistical Metho& for Geographers. Wiley, New York, 1986. 26. Generated designed variables either contrast the first category with later categories, or are orthogonal, polynomial components. Assuming three categories, the BMDPLR program generates two design variables, Dl and D2 of one of the following types

(b) Partial Type:

(a) Marginal Type:

19. Pless B., Verrault R., Arsenault L. and Froppier J. The euidemiology of road accidents in childhood. Am. J.

29. 30.

31.

DI -1 1 0

D2 -1 0 1

category

Dl

D2

1

0

0

2 3

1 0

0 1

Using this coding scheme, the two other age groups of our research are compared to the 16-24 years olds, as a reference group. It is worth mentioning that these diverse approaches give identical results with the exception of the size of the coefficients. Dixon W. J. (Ed.) EMDP Statistical Software. University of California Press, Los Angeles, 1983. The analysis of residuals should also be integrated into the reasoning. For example, the residuals of the chosen models (the full model in the case of serious injuries and the full model minus population density in the severe group) should be examined for mis-specification and other troublesome occurrences. Wrigley N. Categorical Data Analysis. Longman, London, 1985. Sergerie-D. Identification et Priorisation des Sites Daneereux en Montereeie. Internal Renort. D. 37. DSC Monteregie Region;1 Committee for Road Security, 1988. Letendre P. La Geographic des Accidents de la Route en Milieu Rural: Le Cas de la MRC de Coaticook en 1984.

Masters Thesis, p. 127. Department of Geography, Sherbrooke University, Sherbrooke, Quebec, 1988.

APPENDIX Municipaliths Regionales de ComtP (M.R.C.)

(Figs I and 2)

Code Name

Code Name

Code Name

100 Les Iles-de-la-Madeleine 110Avignon 120 Bonaventure 130 Pabok 140 Le Cdte-de-Gasp& 150 Denis-Rivarin 160 Matane 170 La Matapedia 180 La Mitis 190 Rimouski-Neigette 210 Le Fjord-du Saguenay 230 Lac-Saint-Jean-Est 300 Temiscouate 310 Les Basques 315 Rivitre-du-Louo 320 Kamouraska _ 325 L’Islet 330 Montmagny 335 Les Etchemins 340 Beauce-Sartigan 345 L’Amiante 350 Robert-Clihe 355 La Nouvelle-Beauce 360 Bellechasse 365 Desjardins 370 Les Chutes-de-la-Chaudiere 375 Lotbinitre 378 Portneuf 380 La Jacques-Cartier 385 L’ae-d’orlians 390 LaCBte-de-Beaupre

395 Charlevoix 398 Charlevoix-Est 405 L’Erable 410 Arthabaska 415 Drummond 420 Nicolet-Yamaska 425 BCcanour 435 Francheville 440 Le Centre-de-la-Mauricie 450 Maskinonge 470 Mtkinac 480 Le Haut-Sant-Maurice 5 10 Le Granit 520 Le Haut-Saint-Fran9ois 530 Coaticook 540 Menphremagog 560 Sherbrooke 570 Le Val-Saint-Fragois 580 L’Or-Blanc 600 Brome-Missiquoi 610 Le Haut-Richelieu 615 Les Jardins-de-Napierville 620 Le Haut-Saint-Laurent 625 Beauhamois-Salaberry 628 Yaudreuil-Soulanges 632 Roussilon 634 Champlain 635 La Vallee-du-Richelieu 640 Rouville 645 Le Haute-Vamaska 650 Acton

655 Les Maskoutains 660 Le Bas-Richelieu 665 Lajemmerais 670 Lava1 671 Deux-Montagnes 672 Mirabel 673 The&e-de-Blainville 674 Les Moulins 676 L’Assomption 678 D’Autray 680 Joliette 682 Montcalm 684 La Riviere-du-Nord 686 Argenteuil 687 Les Pays-d’en-Haut 689 Les Laurentides 690 Matawinie 7 10 Papineau 730 La Vallte-de-Gatineau 740 Pontiac 750 Antoine-Labelle 8 IO Temiscaminque 820 Rouyn-Noranda 830 La Vallee-de-l’Or 850 Abitibi 890 Abitibi-Ouest 910 Minganie 920 Sept-Rivieres 930 Manicouagan 940 Le Haute-C&e-Nord 990 Caniapiscau