Motorcycle Crashes in Ontario

Direct Medical Costs of Motorcycle Crashes in Ontario
Journal: CMAJ
Manuscript ID CMAJ-17-0337.R2
Manuscript Type: Research - Cohort study (retrospective)
Date Submitted by the Author: n/a

Co
Complete List of Authors: Pincus, Daniel; University of Toronto,

Wasserstein, David; Sunnybrook Health Sciences Centre
Nathens, Avery; Sunnybrook Health Sciences Center, ; Institute for Clinical
Evaluative Sciences,
nf
Bai, Yu; Institute for Clinical Evaluative Sciences,

Redelmeier, Donald A.; University of Toronto, Medicine
Wodchis, Walter; University of Toronto, Health Policy, Management, and
Evaluation
id
Keywords: Epidemiology, Health Policy, Health Economics, Public Health

en
More Detailed Keywords: Motorcycles, Traffic Accidents, Direct medical costs
Background. No reliable estimate of costs incurred by motorcycle crashes

(MCs) exists. Our objective was to calculate the direct costs of all publicly
funded medical care provided to individuals following MCs compared to
tia
automobile crashes (ACs).
Methods. We conducted a population based, matched cohort study of

adults in Ontario that presented to hospital because of a MC or AC from
l
2007 through 2013. For each case, we identified one control absent a
motor vehicle crash during the study period. Direct costs for each case and
control were estimated as 2013 $ Canadian from the payer perspective
using methodology that links healthcare use to individuals over time. MC-
and AC-attributable costs within 2 years were then calculated using a
difference-in-differences approach.
Abstract:
Results. We identified 26,831 patients injured from MCs and 281,826
injured from ACs. Mean MC- and AC-attributable costs were $5,825 and
$2,995, respectively (p<0.0001). The rate of injury was triple for MCs
compared to ACs (2,194 injured yearly/100,000 registered motorcycles
versus 718 injured yearly/100,000 registered automobiles; IRR=3.1, 95%
CI=2.8-3.3, p<0.0001). Severe injuries, defined as those with an
Abbreviated Injury Scale >=3, were 10 times greater (125 severe injuries
per yearly/100,000 registered motorcycles versus 12 severe injuries per
yearly/100,000 registered automobiles; IRR=10.4, 95% CI=8.3-13.1,
p<0.0001).
Interpretation. Considering both the attributable cost and higher rate of

injury, we found each registered motorcycle in Ontario costs the public
For Peer Review Only

Page 1 of 64
1
2
3
4 healthcare system 6 times the amount of each registered automobile.
Medical costs may provide an additional incentive to improve motorcycle
5
safety.
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 2 of 64
COMMENTS 1 MC costs
1
2
3
4
Dear Editorial Board of The CMAJ,
5
6 Thank you very much for the opportunity to revise our manuscript "CMAJ-17-0337.R1: Direct
7 Medical Costs of Motorcycle Crashes in Ontario". We sincerely appreciate the thoughtful
8 reviews and believe that in addressing the feedback from reviewers we have improved the
9 paper. Our responses to specific comments from the editorial board and individual reviewers
10
11 are listed below.
12
13 Thank you very much again.
14
15
16 Daniel Pincus on behalf of the authors.
17
18 --
19
20
21 Editorial Board:
Co
22
23 Comment 1. What sort of motor bikes are covered by your codes? For example are electric-
24 assisted bicycles in here? Do you have any estimate as to how valid the coding is?
25
26
nf
27 This is an important point. 'Mopeds, motor scooters, and motorized bicycles were included in
28 our definition of motorcycles as per the International Classification of Diseases and Related
29 Health Problems, 10th Revision published by the Canadian Institute for Health Information (ICD-
id
30
31 10-CA). There is no way to distinguish between these different vehicle subtypes using the ICD-
32 10-CA to our knowledge.
en
33
34 Including these other motorcycle subtypes may have increased the overall rate of injury
35
observed in our study. However, we know from the Ontario Road Safety Annual Report (ORSAR)
tia
36
37 data that does distinguish between motorcycle subtypes that moped crashes result very few
38 health care interactions: for example, 25 emergency room visits, 2 hospital admissions, and
39 zero deaths in 2013.(1) In other words, mopeds result in <1% of the hospital presentations, and
l
40
41 even fewer of the hospital admissions, attributed to motorcycles overall. Including these lower
42 speed vehicles would have also decreased the proportion of overall injuries that were severe
43 and/or resulted in death. For these reasons, it is unlikely our cost estimates per vehicle were
44
affected by this coding limitation.
45
46
47 Although the ICD codes for external causes of injury are considered extremely accurate, those
48 relating to MVCs specifically have not undergone a formal validation to our knowledge.
49
However, their validity is supported by:
50
51 1) Coding of Type 9 diagnoses, such as those used to define motorcycle and automobile
52 injuries in this study, being a mandatory requirement in the databases used in this
53 study.(2)
54
2) The use of Type 9 codes use in several prior studies of traffic crashes, including those
55
56 published in The CMAJ.(3-5)
57
58
59
60
Page 3 of 64
COMMENTS 2 MC costs
1
2
3
4
3) Data abstraction for diagnosing motor vehicle crashes being intuitively and practically
5 more straightforward than using chart records to diagnose medical conditions such as
6 diabetes,(6) hypertension,(7) chronic obstructive pulmonary disease (COPD),(8) or
7 congestive heart failure (CHF);(9) diagnoses which have been formally validated in
8
9
Ontarios administrative data.
10 4) Ontario hospitals receiving additional payments for road injuries. This includes cases
11 related to motorcycle crashes. MVC codes are thus tied to reimbursement and carefully
12 scrutinized accordingly.
13
14
5) Relative comparisons of motorcycle and automobile crashes using different (police
15 report) data are similar to those we found. These include reports by Ontarios Ministry
16 of Transportation (i.e. ORSAR)(1) and the National Highway Traffic Safety Administration
17 (NHTSA) in the United States.(10) According to the NHTSA, MCs in the United States are
18
19 associated with 6x the death rate as ACs per registered vehicle.(11) Similarly, according
20 to the 2013 ORSAR, MCs in the Ontario are associated with 4x the death rate as ACs per
21 registered vehicle.(1) We found 5x the death rate of MCs versus ACs per vehicle.
Co
22
23
24 Including 'mopeds, motor scooters, and motorized bicycles in our definition of motorcycles is
25 discussed in Supplementary Appendix B of our paper. However, due to the page limit of the
26 Interpretation section referenced in Comment 5 below, and the priority given to other
nf
27 Limitations already discussed in our manuscript, no additional changes to the manuscript have
28
29 been made in response to this comment. However, we can add a note in the main text about
id
30 including lower speed motorcycles if the Editors deem it prudent.

31
32
en
Comment 2. Many of your drivers will be young and not seeing doctors at all. What
33
34 proportion of the young Ontario population is that true for? Are they different from the
35 others, that you used for controls?
tia
36
37 The rationale for excluding those without any contact within 5 years in the first place was to
38
39 exclude false health cards, individuals who died and whose death was not registered, as well as
l
40 those who left the province without notifying the Ontario Health Insurance Plan. This is
41 standard practice in costing research using Ontarios administrative data.(12)
42
43
44 In addition, though this point is well-taken, it is not what we observed in the study. Healthcare
45 utilization in the year prior to the index date, as measured by health care costs accrued during
46 the year prior, was similar between MC-cases and controls (please see Table 3 in the R1
47
version of the manuscript or the updated Supplementary Appendix E in updated R2
48
49 manuscript; mean (SD) costs being 1779.76 (7003.45) and 1664.34 (8732.5), standardized
50 difference = 0.02). The difference-in-differences / self-matching approach would have further
51 removed this difference in utilization, had it existed, after propensity-score matching.
52
53
54 Comment 3. It appears from your data that a much higher proportion of motorcycle drivers
55 than automobile drivers will have died on the scene. If this is correct please comment on the
56 effect of this death rate difference on your cost findings.
57
58
59
60
Page 4 of 64
COMMENTS 3 MC costs
1
2
3
4
As indicated in the Limitations section of the manuscript: We identified patient encounters
5 with health care providers. As a result, our data does not include those injured and/or killed at
6 the scene and not transported to hospital. Nonetheless, by comparing our data to police report
7 data from MTO reports (ORSARs) published during the study period, we may infer proportion of
8
9
deaths occurring prehospital.(1) We agree this may create an interesting competing outcome,
10 with rate of early death (without any cost) being higher in motorcycle group. However, this
11 would underestimate costs in the motorcycle group and we argue is not critical to our payer
12 perspective analysis (though would be if the perspective were changed from payer to society).
13
14
Most importantly, when we actually compare our data to the ORSAR data during the study
15 period, we calculate 40% of motorcycle deaths (at 30 days) occurred prehospital, as compared
16 to 50% automobile deaths (at 30 days).
17
18
19 Comment 4. Please use the STROBE checklist for the revision.
20
21 We have uploaded a revised version of the STROBE checklist to reflect the most recent
Co
22 manuscript version.
23
24
25 Comment 5. Interpretation. This section is about twice as long as we have space for. Can you
26 please shorten it to as near 2 pages as possible, particularly the Implications section.
nf
27
28
29 The Interpretation section has been shortened to 800 words.
id
30
31 Comment 6. Visual Instruments. We will have room for a maximum of 4 in the published
32
en
paper. I suggest you join Tables 1a and 1b. Table 3 could feature in the appendices. Figure 2
33
34 could also go to the appendices. Please tell us what you think about these suggestions.
35
tia
36 We agree. Thank you for these suggestions. We have edited the updated manuscript,
37
figures/tables, and supplementary appendix reflecting these changes.
38
39
l
40 Comment 7. Tables. Please remove shading and bolding from all tables. In Table 2 could you
41 please supply the N for each column?
42
43
44 These changes have been made.
45
46 Comment 8. Figures. Please make sure that each is supplied in an editable format.
47
48
-- Figure 1: please confirm it's original to you, and unpublished elsewhere. We will need an
49 editable Word or PowerPoint version
50
51 We confirm the Figure is original and we have provided it in an editable PowerPoint file.
52
53
54 -- Figure 2: please supply as an editable Excel file with a retained link to the data
55
56 An excel file containing this data for this Figure (now Supplementary Appendix E) is included.
57
58
59
60
Page 5 of 64
COMMENTS 4 MC costs
1
2
3
4
--
5
6 Reviewer 1:
7
8 Comment 1. p. 5: it looks like you did both propensity matching for some variables, and
9
10 standard variables matching for others (age, sex). Does Rosenbaum/Rubins theorem hold
11 for such mixed matching? (perhaps you could prove that in the next revision). You need to
12 be more explicit about what the propensity is here, that you are matching on. For MC, it
13
should be the likelihood of a MC accident (give the degree of propensity scores in terms of
14
15 decimal places you matched for, or at least the algorithm you used, and its relative
16 success). For AC, it should be the likelihood of a AC accident. (Can you be anymore sure that
17 your controls had no MC or ACs in the sample period? Or does patient confidentiality
18
preclude matching with local police citations?)
19
20
21 Greedy matching occurred 1:1 on age (+/- 90 days), sex, and the logit of a propensity score with
Co
22 a caliper of 0.2*standard deviation.(13, 14) Propensity scores were calculated based on patient
23
comorbidity, income quintile, and residential location. We have updated the manuscript to
24
25 reflect these details.
26
nf
27 Success of our matching approach in eliminating known confounding was assessed by

28
29
examining the balance of covariates after matching.(15-17) Please see our response to
id
30 Comment 8 below for a more detailed discussion of the method we used to assess whether
31 covariates were balanced after matching (i.e. standardized differences). As we discuss below,
32
en
the use of standardized differences to assess the success of / balance of prognosis after
33
34
matching is standard practice(16) and has been used in prior studies recently published in The
35 CMAJ.(18, 19) Several studies indicate factors related to the outcome, rather than the
tia
36 exposure, may improve the balancing ability of the propensity score.(20, 21) Several studies
37 have also found that the discriminatory power of propensity score models such as those
38
39
measured by c-statistics, provides no information about whether the model has been correctly
l
40 specified, as long as balance of covariates is achieved.(13) Another measure of relative success

41 we discuss in the manuscript is that 98% of cases were matched to controls.
42
43
44 Balance of observed covariates does not necessarily mean unobserved covariates were also
45 balanced. However, the difference-in-differences (or self-matching) approach would have
46 further removed individual variation and unmeasured confounding between cases and controls,
47 if any existed, after propensity-score matching. However, this was unlikely the case as
48
49 healthcare utilization in the year prior to the index date (measured by health care costs accrued
50 during the year prio), was similar between MC-cases and controls (please see Table 3 in the R1
51 version of the manuscript or the updated Supplementary Appendix E in updated R2
52 manuscript; mean (SD) costs being 1779.76 (7003.45) and 1664.34 (8732.5) in the year prior,
53
54 standardized difference = 0.02).
55
56 Regarding the absence of MCs or ACs among controls during the study period, we can be sure
57
controls did not have any health interactions resulting from these collisions because Ontario
58
59
60
Page 6 of 64
COMMENTS 5 MC costs
1
2
3
4
residents have all their medically necessary health care services, provider information, and
5 demographic characteristics recorded in the health administrative databases used in this study.
6 As indicated in the Limitations section of the manuscript, however, we only identified patient
7 encounters with health care providers, and as a result, our data does not include those injured
8
9
who did not present to hospital. We do not have access to police reports as suggested.
10 However, persons involved in MVCs but not presenting to hospital by definition would have not
11 incurred publicly funded medical costs and not have affected our cost estimates.
12
13
14
Comment 2. p. 5-6 you are fond of citing other validated measures used in your study,
15 without reference to validation in your own study (references 23-31, roughly). Could you add
16 an index indicating how your sample was similarly validated, i.e., that there is external
17 validation (between studies) in the use of these indices, in an additional appendix. Thanks.
18
19
20 Since we are not aware of similar studies of health care interactions of MCs and ACs from a
21 population based sample, there may be differences between our findings and those obtained
Co
22 from police report data [such as those published Ontarios Ministry of Transportation (MTO)(1)
23
24 and the National Highway Traffic Safety Administration (NHTSA) in the United States].(10)
25 These differences are due to the different type of data we collected in this study compared to
26 data collected by these organizations. We identified patient encounters with health care
nf
27 providers whereas the MTO / NHTSA uses police reports. Our data does not include those killed
28
29 at the scene and not transported to hospital and MTO / NHTSA data omit patients injured from
id
30 MVCs when police reports are not filed.

31
32
en
Nonetheless, though these reports do not include costs, our findings regarding the additional
33
34 risk of MCs are similar to those reported by the MTO and NHTSA. According to the NHTSA, MCs
35 in the United States are associated with 6x the death rate as ACs per registered vehicle (see
tia
36 Table 2 from the most recent Motorcycle Traffic Safety Fact Sheet). (11) Similarly, according to
37 the 2013 ORSAR, MCs in the Ontario are associated with 4x the death rate as ACs per registered
38
39 vehicle.(1) We found 5x the death rate of MCs versus ACs per vehicle.
l
40
41 Comment 3. p. 6, and study limitations: you indicate that you would like to capture more (p
42 10) ongoing costs associate with these injuries; but couldnt you get a partial handle on this
43
44 by comparing the robustness of your results with and without rehab costs, and LTC costs (p
45 24) that you have data on now. If those werent greatly different, than other long term costs
46 may also not be greatly different between MC and AC; or vice versa.
47
48
49 Although continuing care costs were not significantly different between MCs and ACs from a
50 clinical perspective (see Table 3), inpatient rehab costs (which are reported as part of acute
51 care hospital costs in the manuscript) were more than twice as great for MCs compared to ACs
52
within 1-year [$670 (95% CI, 505 836) versus $295 (95% CI, 149 441), p = <0.001]. The
53
54 majority of brain rehabilitation over the long-term in our province is provided on an outpatient
55 basis and there is still this limitation, which may be differential between MCs and ACs. As we
56 discuss in the manuscript: the most important limitation of our cost calculations was our
57
58
inability to capture on-going care costs of patients requiring outpatient rehabilitation, such as
59
60
Page 7 of 64
COMMENTS 6 MC costs
1
2
3
4
those suffering acquired brain injury (ABI). However, as we also discuss: Since MC victims are
5 significantly more likely to suffer head trauma,(22-24) and severe injuries in general (Table 2),
6 this limitation only underestimated MC-attributable costs compared to those incurred by ACs.
7
8
9
Comment 6. p. 5, two year followup. Does it matter if its longer? Do you have more data to
10 check?
11
12 As above.
13
14
15 Comment 7. p. 7: its not accidents per vehicle, but accidents per mile driven that are
16 relevant. One way to get a robustness check on this, is compare seasonal variation. A lot
17 less M miles driven in the winter PER motorcycle, then A miles driven per auto.
18
19
20 This is an interesting point. Using the most recently published NHTSA data from 2015, the
21 average registered automobile annually travels approximately 5x the distance compared the
Co
22 average registered motorcycle.(11) All rates expressed per registered vehicle from our study
23
24 could thus be multiplied 5x to be expressed per kilometer (or miles) travelled.
25
26 In response to this comment we have added the following point to the discussion section of the
nf
27 updated manuscript: All rates in our study could also be multiplied by 5 and expressed per
28
29 kilometer travelled since the average automobile travels 5 times the distance of the average
id
30 motorcycle (i.e. 15x injuries, 50x severe injuries, 25x deaths, 30x costs per kilometer
31 travelled).(11)
32
en
33
34 Comment 8. P 8 and table 3 are not standard comparisions: you should use the classical
35 Neymann statistics here and use t-statistics to compare the means. If you think that is too
tia
36 easy, you then need to bootstrap the standard errors around your SD measures (I presume
37 this is for standard deviations), and show that those standard deviation measures are not
38
39 different from zero. (they seem large by my experience with standardized betas in health
l
40 care research)
41
42 Please note this table is now included as Supplementary Appendix E in the updated manuscript
43
44 as per the Editors suggestion. As discussed in the Methods section of the manuscript: Baseline
45 characteristics of cases and controls were reported as means and proportions and compared by
46 using standardized differences (greater than 0.1 being considered indicative of imbalance).(33,
47
34) Standardized differences (also known as standardized mean differences or standardized
48
49 differences of the mean) are calculated as the mean difference (mean group 1 mean group 2)
50 divided by the standard deviation of the measurements. Standardized differences have been
51 shown to provide a measure of effect size and are thought to be particularly useful in large
52
cohort studies using administrative data to avoid identifying clinically spurious statistical
53
54 associations because the statistic is less influenced by sample size compared to statistical
55 significance/classical Neymann statistics.(16) With statistical significance/classical Neymann
56 statistics, in contrast, precision increases and standard errors decrease with increases in sample
57
58
size and no measure of effect size is provided.(16) The use of standardized differences to
59
60
Page 8 of 64
COMMENTS 7 MC costs
1
2
3
4
assess balance of prognosis after matching is standard practice,(16) including prior studies
5 recently published in The CMAJ.(18, 19)
6
7 Comment 9. Also, explain to the reader why you have just a one to one match here; many to
8
9
one matches are more efficient for estimation (though 1 to 1 is asymptotically valid, I
10 agree). Note in the quintiles that your controls are poorer for both groups, but more
11 especially for the MC group. Could you speculate why this is, and how it might bias the
12 comparison (since its a differential response between AC and MC)?
13
14
15 We chose 1:1 matching to facilitate comparisons of covariates and easily and intuitively
16 evaluate the success of our matching algorithm.(15-17) Although we agree adding controls may
17 even further increase the precision of our standard errors, our study was already adequately
18
19 powered using a 1:1 match to detect statistically significant differences between groups.
20 Indeed, one reason we used standardized differences for comparisons in our study was to avoid
21 identifying clinically spurious statistical associations in the large dataset.(16)
Co
22
23
24 Income inequality between cases and controls is an important point. We do acknowledge that
25 although most baseline characteristics were balanced between cases and controls after
26 matching, MC cases were more likely to live in higher income neighborhoods. The clinical
nf
27 relevance of 20.5% cases versus 16.4% controls living in the highest income neighborhoods,
28
29 however, is difficult to contextualize. The fact that motorcycle insurance is already rising in our
id
30 province may explain part of this slight imbalance.(25) Low SES has been shown to predict
31 worse outcomes and higher healthcare utilization in our province.(26-28) As such, poorer MC-
32
en
controls likely would have underestimated MC-attributable costs if anything.

33
34
35 Nonetheless, and as we mentioned above, despite statistically significant differences in income
tia
36 between MC-cases and controls, healthcare utilization in the year prior to the index date was
37 similar between MC-cases and controls (please see Table 3 in the R1 version of the manuscript
38
39 or the updated Supplementary Appendix E in updated R2 manuscript; mean (SD) costs being
l
40 1779.76 (7003.45) and 1664.34 (8732.5), standardized difference = 0.02). The difference-in-
41 differences / self-matching approach would have further removed this difference in utilization,
42 had it existed, after propensity-score matching.
43
44
45 Comment 10. Table 2: why abbreviate in the heading when there is plenty of room to spell
46 out ISS and AIS, rather than add those in the footnote here.
47
48
49 These changes were made.
50
51 --
52
53
54
55 Reviewer 2:
56
57
58
59
60
Page 9 of 64
COMMENTS 8 MC costs
1
2
3
4
Comment 1. a) The calculation of cost differences at the top of page 6 and in figure 1 were to
5 take into account any cost inflation differences between the pre year and post-two study
6 years.
7
8
We apologize this was not clear in the initial version of the manuscript but all costs were
9
10 assigned to each patient based on the year when they were incurred and then inflated to 2013
11 Canadian dollars using the health care component of the Ontario consumer price index (CPI,
12 www.statscan.gc.ca). In response to this comment, the following has been added to the
13
14
updated methods section: All costs were expressed as 2013 Canadian dollars using the health
15 care component of the Ontario consumer price index (CPI, www.statscan.gc.ca).
16
17 Comment 2. b) The authors would add a small calculation as to the costs, mortality rates,
18
serious injury rates PER 100,000 MC or AC vehicle kilometers travelled. I am sure data is
19
20 available as to the average annual kilometrage of both modes.
21
Co
22 This is an interesting point. Using the most recently NHTSA data from 2015, the average
23 registered automobile travels approximately 5x the distance annually compared the average
24
25 registered motorcycle.(11) All rates expressed per registered vehicle in our study could thus be
26 multiplied 5x to be expressed per kilometer (or miles) travelled.
nf
27
28
In response to this comment we have added the following point to the discussion section of the
29
id
30 updated manuscript: Since the average automobile annually travels 5 times the distance of the
31 average motorcycle, all rates in our study could also be multiplied by 5 and expressed per
32 kilometer travelled (i.e. 15x injuries, 50x severe injuries, 25x deaths, 30x costs per kilometer
en
33
travelled).(11)
34
35
Comment 3. Middle of page 9: write "were double" instead of "were 2 times greater"
tia
36
37
38
39
This change (as well as 3 times greater -> triple) has been made to the updated manuscript.
l
40
41 Comment 4. Abstract: last time : ADD may provide AN ADDITIONAL incentive to improve
42 motorcycle safety.
43
44
45 This change was made.
46
47 --
48
49
50 Reviewer 3:
51
52 Introduction
53
Comment 1. - (Major) Similar studies and background literature do not appear to be
54
55 introduced and referenced completely. Though studies on motorcycle crash costs may be
56 limited, it would be valuable to discuss what is known about prevalence and cost of both MCs
57 and ACs. A preliminary Medline search resulted in several relevant papers on MCs alone that
58
59
60
Page 10 of 64
COMMENTS 9 MC costs
1
2
3
4
have not been cited. There are similar cost analyses for automobile crashes as well that
5 should be introduced. For the articles cited, there is lack of detail on what these studies tell
6 us.
7
8
9
We have cited in the introduction of our manuscript the systematic review of prior studies
10 examining medical costs of motorcycle crashes conducted by the National Highway Traffic
11 Safety Administration (NHTSA) in 2003. This systematic review concluded that prior calculations
12 of medical costs attributable to MCs are limited to reviews of hospital charges at single centers
13
14
and that estimates derived from these reviews were incomplete, neglecting costs incurred after
15 a patients discharge.(24, 29, 30) In other words, that no reliable estimate of medical costs
16 incurred by MCs exists.(30) We have been unable to identify any research on this topic
17 published since that time. There is also no literature of which we are aware that directly
18
19 compares costs of MCs and ACs.
20
21 In our study, we examined all health care costs accrued during each patients index
Co
22 visit/admission, any subsequent readmissions/visits, costs for continuing care including
23
24 residential long-term care and home care, and costs for rehabilitation and physicians, all being
25 assessed up to 2 years after the injury. Our study was also population based in contrast to the
26 prior studies on this topic which recorded the hospital charges of inpatients at single centers,
nf
27 primarily trauma centers.(30) As a result, we were able to study the vast majority of MC and AC
28
29 cases who are treated as outpatients and do not have severe injuries (>90%) that were
id
30 excluded in prior studies.

31
32
en
In contrast to data on costs, other reports of clinical data regarding the additional risk of MCs
33
34 are publically available. According to the NHTSA, MCs in the United States are associated with
35 6x the death rate as ACs per registered vehicle (see Table 2 from the most recent Motorcycle
tia
36 Traffic Safety Fact Sheet). (11) Similarly, according to the 2013 ORSAR, MCs in the Ontario are
37 associated with 4x the death rate as ACs per registered vehicle.(1) We found 5x the death rate
38
39 of MCs versus ACs per vehicle. However, these other reports do not consider costs or provide
l
40 detailed descriptions of injury types (using the Abbreviated Injury Scale, for example).
41
42 We have tried to frame our study in the updated Introduction / Interpretation sections within
43
44 the existing literature(30) and publically available data on this topic.(1, 10) We do not think
45 including details of this literature (of which we are aware) beyond what is presented above and
46 in the updated manuscript is prudent given the word limit of the manuscript.
47
48
49 Methods
50 Comment 2. - Page 4 lines 31-36: Is there any literature to show that ICD-10 codes related to
51 motorcycle and automobile crashes are assigned accurately? It would be helpful to know if
52
these codes have been validated in existing literature, and if so, what the positive predictive
53
54 value is.
55
56 This is an important point. Please see our response to the Editorial Boards Comment 1 above.
57
58
59
60
Page 11 of 64
COMMENTS 10 MC costs
1
2
3
4
5 Comment 3. - Page 4 line 44: It would be valuable to clarify what you mean by 30% general
6 sample. This may not come across intuitively to readers.
7
8
9
The following change has been made to the manuscript in response to this comment: For each
10 MC and AC case, we identified one control from a representative subset of the Ontario
11 population.
12
13
14
Comment 4. - Page 4 line 51: The control group consisted of patients who had at least some
15 health system contact during the appropriate inclusion years. This control group may not be
16 fully representative of the underlying population at risk. I assume that many otherwise-
17 healthy men in their 30s who ride motorcycles, or average MC patients, do not have
18
19 regular health system contact. How did you justify using this sample? Limitations of this
20 control group should be addressed in the appropriate section.
21
Co
22 This is also an important point. Please see our response to the Editorial Boards Comment 2
23
24 above.
25
26 Comment 5. - (Major) Page 5 Lines 3-9: It is unclear how you used the propensity score to
nf
27 conduct your match. Clarification on this is essential. If using propensity score matching, what
28
29 technique of matching was used ie. nearest matching, greedy nearest neighbour matching
id
30 with calipers +/- replacement, etc.

31
32
en
Greedy matching occurred 1:1 on age (+/- 90 days), sex, and the logit of a propensity score with
33
34 a caliper of 0.2*standard deviation(13, 14) and was calculated based on patient comorbidity,
35 income quintile, and residential location.
tia
36
37 We have updated the manuscript to reflect these details.
38
39
l
40 Comment 6. - Page 5 Line 51: Since this is not a typically-used method, a reference to the
41 difference-in-differences approach would be valuable.
42
43
44 In response to this comment, we have cited a recent article from JAMA that summarizes the
45 difference-in-differences approach.(31) Although the article focuses on the application of this
46 approach to evaluating the implementation of health policy, the explanations and illustration
47
provided also apply to our study of costs.
48
49
50 Comment 7. - Page 6 Line 18: All costs were expressed as 2013 Canadian dollars. Please
51 reference the source used for inflation rates.
52
53
54 We apologize this was not clear in the original manuscript. Costs were assigned to each patient
55 based on the year when they were incurred and then inflated to 2013 Canadian dollars using
56 the health care component of the Ontario consumer price index (CPI, www.statscan.gc.ca). In
57
58
response to this comment, the following has been added to the updated methods section: All
59
60
Page 12 of 64
1
2
3
4
costs were expressed as 2013 Canadian dollars using the health care component of the Ontario
5 consumer price index (CPI, www.statscan.gc.ca).
6
7 Comment 8. - Page 6 Lines 54-57: Please reference source of proposed 0.1 as indicative of
8
9
imbalance. I believe this is from Normand et al.s 2001 paper.
10
11 Thank you. This reference has been added to the updated manuscript.(32)
12
13
14
Comment 9. - It would be helpful to specify the source used to determine the number of
15 registered automobiles and motorcycles (reference 18) in the Methods section.
16
17 This was the number of each vehicle registered each year during the study period with
18
19 Ontarios Ministry of Transportation. As mentioned in the manuscript: Annual rates of injuries
20 and deaths were expressed per 100,000 registered motorcycles or automobiles in the Province
21 by using publicly available data from the Ministry of Transportation (MTO) (see Supplementary
Co
22 Appendix C).(1)
23
24
25 Results
26 Comment 10. - Page 8 lines 26-28: Covariates were balanced between cases and controls,
nf
27 the exception being that MC cases were more likely to reside in rural areas than controls.
28
29 Please expand on this and potential implications, if any, in Limitations section. Were any
id
30 other model diagnostics done?

31
32
en
This is an important point. Although most baseline characteristics were balanced between cases
33
34 and controls after matching, MC cases were significantly more likely to reside in rural areas
35 than controls. However, the clinical relevance of this difference between a rurality index of 15.9
tia
36 versus 11.3 is difficult to contextualize since a rurality index >40 is generally considered
37 indicative of rural residence.(33) This finding may be the result of motorcycles being difficult
38
39 to ride in the city (and therefore) being more common in the country. Nonetheless and as
l
40 discussed above, healthcare utilization in the year prior to the index date, as measured by
41 health care costs accrued during that year, was actually similar between MC-cases and controls
42 (please see Table 3 in the R1 version of the manuscript or the updated Supplementary
43
44 Appendix E in updated R2 manuscript; mean (SD) costs being 1779.76 (7003.45) and 1664.34
45 (8732.5), standardized difference = 0.02). The difference-in-differences / self-matching
46 approach would have further removed this difference in utilization, had it existed, after
47
propensity-score matching.
48
49
50 Regarding model diagnostics, several studies have found that the discriminatory power of
51 propensity score models such as those measured by c-statistics, provides no information about
52
whether the model has been correctly specified, as long as balance of covariates is
53
54 achieved.(13) Therefore, the success of our matching approach was assessed by examining the
55 balance of covariates after matching.(15-17) We also report the proportion of eligible cases
56 matched to controls (approximately 98%).
57
58
59
60
Page 13 of 64
1
2
3
4
Due to the page limit of the Interpretation section referenced in the Editorial Boards Comment
5 5 above and the priority given to other Limitations already discussed, no changes to the
6 manuscript have been made in response to this comment.
7
8
9
Interpretation
10 Comment 11. - (Major) No reference is made to existing literature on MC and AC prevalence
11 and costs. Though I appreciate that high-quality studies on MC costs may be lacking, relevant
12 literature based out of Canada and the United States exists and should be introduced and
13
14
compared to these findings. Furthermore, many studies have been done internationally (Iran,
15 Nigeria, Brazil, Taiwan, Mexico, etc.) that would add to the discussion portion. Discussing
16 how your estimated AC costs relate to findings in current literature would also be valuable.
17
18
19 Please see our response to Comment 1 above.
20
21 Comment 12. - Page 11 lines 21-26: Expressed in other terms, Ontarios healthcare payer
Co
22 (MOHLTC) would expect to save $13.5 million dollars per year if the Provinces registered
23
24 motorcycles were replaced with automobiles. This is assuming the rate of automobile
25 accidents would stay the same. What if higher-risk motorcycle drivers are also higher-risk
26 automobile drivers, and their probability of crashing is higher than the average automobile
nf
27 driver?
28
29
id
30 Table 6.2: Selected Factors Relevant to Fatal Motorcycle Collisions of the 2013 Ontario Road
31 Safety Annual Report indicates that most motorcycle-related deaths in Ontario are unlikely the
32
en
result of modifiable risk factors or risk-taking behaviors, other than choosing to ride a
33
34 motorcycle itself. For example, we know that 97.9% of motorcyclists in 2013 were wearing
35 helmets at the time of their crash.(1)
tia
36
37 Based on this fact, and that province-level cost estimates may provide context for readers and
38
39 policymakers, we have kept this discussion point. However, our assumption of a constant
l
40 accident rate is a good point and we have updated the manuscript to reflect this: Expressed in
41 other terms, Ontarios healthcare payer (MOHLTC) would expect to save $13.5 million dollars
42 per year if collision rates remained unchanged and the Provinces registered motorcycles were
43
44 replaced with automobiles.
45
46 Comment 13. - Page 12 lines 24-29: How do you anticipate that a tax would increase safety? It
47
may deter drivers with a lower income, but I cannot appreciate how it would inherently
48
49 improve motorcycle safety.
50
51 As suggested, an excise tax may deter motorcycle use, similar to the way tobacco prices have
52
been one of the most effective means of reducing tobacco use.(34-36) Anecdotally, raising
53
54 insurance rates in response to private medical costs has already acted as a deterrent to
55 motorcycle use in our jurisdiction.(25) There are other potential policy initiatives we discuss in
56 the manuscript, such as supplementary private health insurance and costs recoverable from
57
58
59
60
Page 14 of 64
1
2
3
4
motorcycle insurers under the Provinces Insurance Act. The broader expression improving
5 motorcycle safety has been used instead of deterrence for this reason.
6
7 Comment 14. - You did not match based on previous number/type of health care encounters.
8
9
You may want to add in your discussion how costs in the previous year compare for cases
10 versus controls. This may help strengthen the argument that you have comparable groups.
11
12 Thank you for this thoughtful point. We have added the following additional analysis to the
13
14
manuscript in response: Baseline healthcare utilization as measured by costs in the year prior
15 was similar between cases and controls, also indicating groups were comparable (mean (SD)
16 prior year costs for MC-cases and controls were 1779.76 (7003.45) and 1664.34 (8732.5),
17 standardized difference = 0.02; ACs-cases and controls 2385.69 (8223.62) and 2314.57
18
19 (9420.75) standardized difference = 0.01).
20
21 As mentioned above, the difference-in-differences / self-matching approach was intended to
Co
22 remove residual differences in health care utilization between cases and controls, had they
23
24 existed, after propensity-score matching.
25
26 Comment 15. Tables
nf
27 - Table 2: Does the severe injuries score take into account injuries noted during the first
28
29 assessment or the entire health care visit/admission? In other words, are initially-missed
id
30 injuries accounted for?

31
32
en
This is an important point. Any injury diagnoses noted during the patients emergency
33
34 department visit (in the NACRS database) and inpatient stay (in the CIHI-DAD database) were
35 accounted for using an algorithm previously validated in Ontarios administrative data.(37)
tia
36 Initially-missed injuries were thus accounted for. Please see Supplementary Appendix A for
37 more information about the different administrative databases used in this study.
38
39
l
40 Comment 16. - Table 3: The standard deviations for average cost appear quite large. Are
41 these accurate? If so, we lose the ability to calculate costs at the lower end of the spectrum.
42 95% confidence intervals would show the range better.
43
44
45 The updated Table 3 now contains 95% confidence intervals in response to this suggestion.
46 We agree the standard deviations for average costs are large, but these are accurate and
47
similar to cost ranges previously published for different patient populations using the same
48
49 methodology.(38) The distribution is likely due to the wide range of pathology (and associated
50 costs) sampled in our study, ranging from those requiring only an ED visits to those admitted to
51 the ICU and requiring inpatient rehabilitation. Similarly, these severe injuries that required
52
admission to hospital and the ICU likely accounted for the additional costs incurred by MCs
53
54 compared to ACs.
55
56
57
58
59
60
Page 15 of 64
1
2
3
4
References for Comments
5
6 1. Ministry of Transportation. Ontario Road Safety. Annual Report 2013.
7 2. Juurlink D, Preyra C, Croxford R, Chong A, Austin P, Tu J, Laupacis A. Canadian Institute
8
for Health Information Discharge Abstract Database: A Validation Study. Toronto: Institute for
9
10 Clinical Evaluative Sciences; 2006.
11 3. Redelmeier DA, Katz D, Lu H, Saposnik G. Roadway crash risks in recent immigrants.
12 Accid Anal Prev. 2011;43(6):2128-33.
13
14
4. Redelmeier DA, May SC, Thiruchelvam D, Barrett JF. Pregnancy and the risk of a traffic
15 crash. CMAJ. 2014;186(10):742-50.
16 5. Bai YQ, Santos G, WP W. Cost of Public Health Services for Ontario Residents. Applied
17 Health Research Question Series. Toronto: Health System Performance Research Network;
18
19
2016.
20 6. Hux JE, Ivis F, Flintoft V, Bica A. Diabetes in Ontario: determination of prevalence and
21 incidence using a validated administrative data algorithm. Diabetes care. 2002;25(3):512.
Co
22 7. Tu K, Campbell NR, Chen ZL, Cauch-Dudek KJ, McAlister FA. Accuracy of administrative
23
24
databases in identifying patients with hypertension. Open medicine : a peer-reviewed,
25 independent, open-access journal. 2007;1(1):e18.
26 8. Gershon AS, Wang C, Guan J, Vasilevska-Ristovska J, Cicutto L, To T. Identifying
nf
27 individuals with physcian diagnosed COPD in health administrative databases. Copd.

28
29
2009;6(5):388.
id
30 9. Ko DT, Mamdani M, Alter DA. Lipid-lowering therapy with statins in high-risk elderly
31 patients: the treatment-risk paradox. Jama. 2004;291(15):1864.
32
en
10. The National Highway Traffic Safety Administration. TRAFFIC SAFETY FACTS 2015.
33
34 Available from https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/812384. Accessed
35 April 27, 2017.
tia
36 11. NHTSAs National Center for Statistics and Analysis (2015), NHTSA: Motorcycles Traffic
37 Safety Fact Sheet , Accessed June 24, 2017
38
39 from https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/812353.
l
40 12. Wodchis WP, Austin PC, Henry DA. A 3-year study of high-cost users of health care.
41 CMAJ. 2016;188(3):182-8.
42 13. Austin PC. An Introduction to Propensity Score Methods for Reducing the Effects of
43
44 Confounding in Observational Studies. Multivariate behavioral research. 2011;46(3):399.
45 14. Austin PC. Optimal caliper widths for propensity-score matching when estimating
46 differences in means and differences in proportions in observational studies. Pharmaceutical
47 statistics. 2011;10(2):150.
48
49 15. Austin PC. Using the standardized difference to compare the prevalence of a binary
50 variable between two groups in observational research. Communications in Statistics:
51 Simulation and Computation. 2009;38(6):1228.
52 16. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates
53
54 between treatment groups in propensity-score matched samples. Statistics in medicine.
55 2009;28(25):3083.
56 17. Fitzmaurice G. Confounding: Propensity score adjustment. Nutrition. 2006;22(11-
57
12):1214.
58
59
60
Page 16 of 64
1
2
3
4
18. Taipale H, Tolppanen AM, Koponen M, Tanskanen A, Lavikainen P, Sund R, et al. Risk of
5 pneumonia associated with incident benzodiazepine use among community-dwelling adults
6 with Alzheimer disease. CMAJ. 2017;189(14):E519-E29.
7 19. Shih CJ, Chen HT, Kuo SC, Li SY, Lai PH, Chen SC, et al. Comparative effectiveness of
8
9
angiotensin-converting-enzyme inhibitors and angiotensin II receptor blockers in patients with
10 type 2 diabetes and retinopathy. CMAJ. 2016;188(8):E148-57.
11 20. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Strmer T. Variable
12 selection for propensity score models. American Journal of Epidemiology. 2006;163(12):1149.
13
14
21. Rubin DB. On principles for modeling propensity scores in medical research.
15 Pharmacoepidemiology and drug safety. 2004;13(12):855.
16 22. Keng SH. Helmet use and motorcycle fatalities in Taiwan. Accident Analysis and
17 Prevention. 2005;37(2):349.
18
19 23. Hotz GA, Cohn SM, Mishkin D, Castelblanco A, Li P, Popkin C, et al. Outcome of
20 Motorcycle Riders at One Year Post-Injury. Traffic Injury Prevention. 2004;5(1):87.
21 24. Monk JP, Buckley R, Dyer D. Motorcycle-related trauma in Alberta: A sad and expensive
Co
22 story. Canadian Journal of Surgery. 2009;52(6):E235.
23
24 25. Harvey I. Motorcycle rates rising with severity, number of claims: Toronto Star; 2016
25 [Available from: https://www.thestar.com/autos/2016/05/28/motorcycle-rates-rising-with-
26 severity-number-of-claims.html.
nf
27 26. Agabiti N, Picciotto S, Cesaroni G, Bisanti L, Forastiere F, Onorati R, et al. The influence
28
29 of socioeconomic status on utilization and outcomes of elective total hip replacement: a
id
30 multicity population-based longitudinal study. International journal for quality in health care :
31 journal of the International Society for Quality in Health Care / ISQua. 2007;19(1):37.
32
en
27. Santaguida PL, Hawker GA, Hudak PL, Glazier R, Mahomed NN, Kreder HJ, et al. Patient
33
34 characteristics affecting the prognosis of total hip and knee joint arthroplasty: a systematic
35 review. Canadian journal of surgeryJournal canadien de chirurgie. 2008;51(6):428.
tia
36 28. Matheson FI, Moineddin R, Dunn JR, Creatore MI, Gozdyra P, Glazier RH. Urban
37 neighborhoods, chronic stress, gender and depression. Social science & medicine (1982).
38
39 2006;63(10):2604.
l
40 29. Dobson JL. THE PUBLIC COST OF MOTORCYCLE TRAUMA. Jama-Journal of the American
41 Medical Association. 1989;261(8):1149-.
42 30. Administration NHTS. Costs of Injuries Resulting from Motorcycle Crashes: A Literature
43
44 Review 2003 [Available from:
45 http://www.nhtsa.gov/people/injury/pedbimot/motorcycle/Motorcycle_HTML/overview.html.
46 31. Dimick JB, Ryan AM. Methods for evaluating changes in health care policy: the
47
difference-in-differences approach. JAMA. 2014;312(22):2401-2.
48
49 32. Normand ST, Landrum MB, Guadagnoli E, Ayanian JZ, Ryan TJ, Cleary PD, et al. Validating
50 recommendations for coronary angiography following acute myocardial infarction in the
51 elderly: a matched analysis using propensity scores. J Clin Epidemiol. 2001;54(4):387-98.
52
33. Kralj B. Measuring Rurality for Purposes of Health Care Planning: An Empirical
53
54 Measure for Ontario. Toronto: Ontario Medical Association. 2005.
55 34. Chaloupka F. Rational Addictive Behavior and Cigarette-Smoking. Journal of Political
56 Economy. 1991;99(4):722-42.
57
58
59
60
Page 17 of 64
1
2
3
4
35. Gilmore AB, Tavakoly B, Taylor G, Reed H. Understanding tobacco industry pricing
5 strategy and whether it undermines tobacco tax policy: The example of the UK cigarette
6 market. Addiction. 2013;108(7):1317-26.
7 36. Jha P, Chaloupka FJ. The economics of global tobacco control. British Medical Journal.
8
9
2000;321(7257):358-61.
10 37. Haas B, Xiong W, Brennan-Barnes M, Gomez D, Nathens AB. Overcoming barriers to
11 population-based injury research: Development and validation of an ICD-10-to-AIS algorithm.
12 Canadian Journal of Surgery. 2012;55(1):21.
13
14
38. Rosella LC, Fitzpatrick T, Wodchis WP, Calzavara A, Manson H, Goel V. High-cost health
15 care users in Ontario, Canada: demographic, socio-economic, and health status characteristics.
16 BMC health services research. 2014;14:532.
17
18
19
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 18 of 64
MANUSCRIPT 1 MC costs
1
2
3
4
Direct Medical Costs of Motorcycle Crashes in Ontario
5
6 Authors:
7 Daniel Pincus1,2,3 MD
8
9
David Wasserstein1,4 MD, MSc, MPH
10 Avery B. Nathens1,2,3,4 MD, MPH, PhD
11 Yu Qing Bai2,3 MSc
12 Donald A. Redelmeier2,3,5,6 MD, MS(HSR)
13
14
Walter P. Wodchis2,3 PhD
15
16 Affiliations:
17 1
Department of Surgery, University of Toronto, Toronto, Canada
18 2
19 Institute for Clinical Evaluative Sciences, Toronto, Canada
3
20 Institute of Health Policy, Management and Evaluation, University of Toronto, Toronto, Canada
4
21 Department of Surgery, Sunnybrook Health Sciences Centre, Toronto, Canada
Co
22 5
Division of General Internal Medicine, Sunnybrook Health Sciences Centre, Toronto, Canada,
23 6
24 Center for Leading Injury Prevention Practice Education & Research, Toronto, Canada
25
26 Correspondence and address for all authors:
nf
27 Daniel Pincus MD
28
29 Department of Surgery, Division of Orthopaedic Surgery, University of Toronto
id
30 149 College Street, Room 508-A

31 Toronto, ON M5T 1P5
32
en
Phone: 416-946-7957 (w), 647-244-3324 (c)

33
34 Fax: 416-978-3928
35 d.pincus@utoronto.ca
tia
36
37 Competing Interests / Disclosures:
38
39 There are no relevant disclosures or competing interests. All authors meet ICMJE criteria.
l
40
41 Source of funding:
42 This research was supported by grants from the Ontario Ministry of Health and Long-Term Care
43
44 (MOHLTC) and the Ontario SPOR Support Unit to the Health System Performance Research
45 Network (HSPRN: fund #06034, recipient WPW), and by the Institute for Clinical Evaluative
46 Sciences (ICES), which is also funded by an annual grant from the MOHLTC. The funders had no
47
role in study design, data collection and analysis, decision to publish, or preparation of the
48
49 manuscript. Parts of the material are based on data and information compiled and provided by
50 the Canadian Institute for Health Information (CIHI). No endorsement by ICES, the MOHLTC or
51 CIHI is intended or should be inferred. No benefits have been received or will be received from
52
a commercial party related directly or indirectly to the subject of this article.
53
54
55 Key Words: Motorcycles, Accidents, Traffic/economics, Traffic/statistics & numerical data,
56 Insurance, Health/economics, Retrospective Studies
57
58
59
60
Page 19 of 64
1
2
3
4
ABSTRACT (Word Count: 246)
5
6 Background. No reliable estimate of costs incurred by motorcycle crashes (MCs) exists. Our
7 objective was to calculate the direct costs of all publicly funded medical care provided to
8
9
individuals following MCs compared to automobile crashes (ACs).
10
11 Methods. We conducted a population based, matched cohort study of adults in Ontario that
12 presented to hospital because of a MC or AC from 2007 through 2013. For each case, we
13
14
identified one control absent a motor vehicle crash during the study period. Direct costs for
15 each case and control were estimated as 2013 $ Canadian from the payer perspective using
16 methodology that links healthcare use to individuals over time. MC- and AC-attributable costs
17 within 2 years were then calculated using a difference-in-differences approach.
18
19
20 Results. We identified 26,831 patients injured from MCs and 281,826 injured from ACs. Mean
21 MC- and AC-attributable costs were $5,825 and $2,995, respectively (p<0.0001). The rate of
Co
22 injury was triple for MCs compared to ACs (2,194 injured yearly/100,000 registered motorcycles
23
24 versus 718 injured yearly/100,000 registered automobiles; IRR=3.1, 95% CI=2.8-3.3, p<0.0001).
25 Severe injuries, defined as those with an Abbreviated Injury Scale >=3, were 10 times greater
26 (125 severe injuries per yearly/100,000 registered motorcycles versus 12 severe injuries per
nf
27 yearly/100,000 registered automobiles; IRR=10.4, 95% CI=8.3-13.1, p<0.0001).

28
29
id
30 Interpretation. Considering both the attributable cost and higher rate of injury, we found each
31 registered motorcycle in Ontario costs the public healthcare system 6 times the amount of each
32
en
registered automobile. Medical costs may provide an additional incentive to improve

33
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 20 of 64
1
2
3
4
INTRODUCTION
5
6 Motor vehicle safety has improved significantly over the past 100 years.(1) Between
7
8
9
2000 and 2010, for example, mortality related to motor vehicle crashes (MVCs) decreased by
10
11 55.1% in 19 developed countries.(1) In contrast, deaths and injuries among a subset of MVC
12
13
14
patients those injured in motorcycle crashes (MCs) remained stable during the same time
15
16 period.(2)
17
18
19 Although medical costs may provide incentive to improve motorcycle safety,(3-5) no
20
21 reliable estimate of medical costs incurred by MCs exists. Prior calculations of medical costs
Co
22
23
24 attributable to MCs are limited to reviews of hospital charges at single centers.(6) Estimates
25
26 derived from these reviews are incomplete, neglecting costs incurred after a patients
nf
27
28
29 discharge, for example.(6-8)
id
30
31 The calculation of patient level medical costs in Ontario, Canada (population 13.6 million
32
en
33
34 in 2014) is now possible with methodology that links publicly funded healthcare use to
35
tia
36 individuals over time.(9) Using this methodology from the payer perspective, our objective was
37
38
39 to calculate the direct costs of all publicly funded medical care provided to individuals
l
40
41 presenting to hospital following MCs in comparison to automobile crashes (ACs). We also
42
43
44 examined the population incidence of injuries resulting from these crashes. Our hypothesis was
45
46 that medical costs and injury rates attributable to MCs, which can occur at high speeds with less
47
48
49 personal protection, are significantly higher than those attributable to ACs.
50
51
52
53
54 METHODS
55
56 Setting
57
58
59
60
Page 21 of 64
1
2
3
4
We conducted a population based, matched cohort study in Ontario, Canada using
5
6 linked health administrative databases at the Institute for Clinical Evaluative Sciences (ICES)
7
8
9
(see Supplementary Appendix A data sources). Ontarians have their medically necessary
10
11 health care interactions, provider information, and demographic characteristics recorded in
12
13
14
these databases. These data have previously been used to estimate medical costs(10-15) and
15
16 study traffic crashes.(16-18) The research protocol was approved by the Research Ethics Board
17
18
19 at Sunnybrook Health Sciences Centre.
20
21
Co
22
23
24 Patients
25
26 The study design is displayed in Figure 1. Ontarians who presented to the emergency
nf
27
28
29 department (ED) or were admitted to hospital following a MC or AC between April 1, 2007 to
id
30
31 March 31, 2013 were eligible for inclusion. International Classification of Diseases, 10th Revision
32
en
33
34 (ICD-10) codes were used to identify injured motor vehicle occupants [MC (V20-V29) and AC
35
tia
36 (V40-V49, V50-V59, V70-V79), respectively].(19) We excluded non-Ontario residents and those

37
38
39 with prior hospital presentation for a motor-vehicle related injury of any type within 2 years
l
40
41 before the index date (see Supplementary Appendix B database codes).
42
43
44 For each MC and AC case, we identified one control from a representative subset of the
45
46 Ontario population. The purpose of these controls was to determine baseline medical costs
47
48
49 absent a motor vehicle crash and thus enable the calculation of incremental costs. Controls had
50
51 health system contact but no documented motor-vehicle related injury during this timeframe
52
53
54 or dating back 2 years prior to the index date. Index dates for controls were randomly assigned
55
56 based on the distribution of index dates for cases in the same fiscal year. Greedy matching
57
58
59
60
Page 22 of 64
1
2
3
4
occurred on age (+/- 90 days), sex, and the logit of a propensity score with a caliper of
5
6 0.2*standard deviation (20-22) and was calculated based on patient comorbidity, income
7
8
9
quintile, and residential location (see Covariates).
10
11
12
13
14
Covariates
15
16 We measured several covariates that have been shown to influence health care
17
18
19 utilization in Ontario.(13) Age and sex were analyzed as continuous and categorical variables,
20
21 respectively. Comorbidities listed on outpatient and hospital discharge abstracts in the two
Co
22
23
24 years before the index date were categorized according to Collapsed Aggregate Diagnosis
25
26 Groups (CADGs).(23) Neighborhood income quintile was used as a validated surrogate measure
nf
27
28
29 for socioeconomic status and social deprivation.(24-26) Patient location of residence was
id
30
31 classified according to Local Health Integrated Network (LHIN) and the Rurality Index of Ontario
32
en
33
34 (RIO).(27)
35
tia
36
37
38
39 Outcomes
l
40
41 The primary outcome of this study was direct medical costs within 2 years, attributable
42
43
44 to MCs and ACs, and paid by Ontarios Ministry of Health and Long-term Care (MOHLTC). First,
45
46 medical costs were calculated for each case and control during each fiscal year of the study
47
48
49 period using established patient level costing methodology (see Supplementary Appendix A).(9)
50
51 MC- and AC-attributable costs within 2 years were then calculated using a difference-in-
52
53
54 differences approach: a baseline cost accrued the year prior to a crash was subtracted from
55
56
57
58
59
60
Page 23 of 64
1
2
3
4
costs in the first and second year following the crash (first difference) and then compared to
5
6 the same difference among controls (second difference).(28)
7
8
9
Costs were categorized as: (a) acute care hospital including the ED, index admission,
10
11 rehabilitation, and any subsequent readmissions, (b) physician, (c) drugs/laboratory costs
12
13
14
outside of hospitals, (d) continuing care including residential long-term care and home care,
15
16 and (e) assistive devices.(18) Care episodes that spanned more than one fiscal year were
17
18
19 divided on a pro rata basis. All costs were expressed as 2013 Canadian dollars using the health
20
Co
22
23
24 Secondary outcomes included clinical data. First, the annual rates of those injured and
25
26 dying (within 30 days) from MCs and ACs were reported, so long as the patient presented to
nf
27
28
29 hospital after the crash. More detailed clinical data were also compared between MC and AC
id
30
31 cases. We identified the highest level of care required for each patient: ED visit only,
32
en
33
34 hospitalization, or intensive care unit (ICU) admission. The Injury Severity Score (ISS) and
35
tia
36 Abbreviated Injury Scale (AIS) were used to assess injury severity. ISS and AIS were obtained
37
38
39 from ICD-10 codes by means of a validated algorithm.(29) ISS was categorized as <9, 9-15, 16-
l
40
41 24, or >=25. Severe injuries were those with an AIS>=3 and were reported overall and for each
42
43
44 anatomical region. The algorithm and operational definitions have been used previously in
45
46 population based research of Ontarios administrative data.(30-32)
47
48
49
50
51 Statistical Analysis
52
53
54 Baseline characteristics of cases and controls were reported as means and proportions
55
56 and compared by using standardized differences (greater than 0.1 being considered indicative
57
58
59
60
Page 24 of 64
1
2
3
4
of imbalance).(33, 34) We also compared characteristics between MC-cases with AC-cases by
5
6 using independent samples t-tests for continuous variables and chi-square tests for categorical
7
8
9
variables. Longer-term costs (up to 5 years) were also compared between matched cases and
10
11 controls enrolled before FY 2011 for whom longer-term follow-up (>2 years) was available.
12
13
14
Annual rates of injuries and deaths were expressed per 100,000 registered motorcycles or
15
16 automobiles in the Province by using publicly available data from the Ministry of Transportation
17
18
19 (MTO) (see Supplementary Appendix C).(2) Incidence rate ratios (IRRs) with 95% confidence
20
21 intervals (CIs) were then used to compare injury and death rates between MC- and AC-cases.
Co
22
23
24 All analyses were performed using SAS software (SAS version 9.3 and SAS Enterprise Guide
25
26 version 6.1; SAS Institute) and the type-I error probability was set to 0.05.
nf
27
28
29
id
30
31 RESULTS
32
en
33
34 Descriptive clinical data; (Table 1, Table 2)
35
tia
36 We identified 26,831 Ontarians during the study period injured during MCs and 281,826
37
38
39 injured during ACs. The annual incidence of injured persons was triple for MCs compared to ACs
l
40
41 (2,194 injured per year/100,000 registered motorcycles versus 718 injured per year/100,000
42
43
44 registered automobiles; IRR=3.1, 95% CI=2.8-3.3, p<0.0001; Table 1). Deaths (at 30d) were
45
46 approximately 5 times greater (14 deaths per year/100,000 registered motorcycles versus 3
47
48
49 deaths per year/100,000 registered automobiles; IRR=4.7, 95% CI=2.9-7.6, p<0.0001; Table 1).
50
51 In terms of the level of care required for each patient, those injured in MCs were
52
53
54 significantly more likely to require hospital and ICU admission than those injured in ACs (11.93%
55
56 versus 4.24%, p < 0.0001 and 2.31% versus 1.09%, p < 0.0001, respectively). MC patients were
57
58
59
60
Page 25 of 64
1
2
3
4
also more likely to have an ISS > 16 (3.67% versus 1.32%, p < 0.0001) and suffer severe injuries
5
6 (5.69% versus 1.67% with AIS>=3, p < 0.0001). Expressed per registered vehicle, severe injuries
7
8
9
were 10 times greater for MCs than ACs (125 severe injuries per year/100,000 registered
10
11 motorcycles versus 12 severe injuries per year/100,000 registered automobiles; IRR=10.4, 95%
12
13
14
CI=8.3-13.1, p<0.0001). Severe head, thorax, abdominal and extremity injuries were all more
15
16 common among patients injured in MCs compared to ACs (Table 2).
17
18
19
20
21 Baseline characteristics after matching; (Supplementary Appendix E)
Co
22
23
24 There were 26,257 MC cases and 276,760 AC cases (98%) matched to controls.
25
26 Covariates were balanced between cases and controls, the exception being that MC cases were
nf
27
28
29 more likely to reside in rural areas than controls. Baseline healthcare utilization as measured by
id
30
31 costs in the year prior was similar between cases and controls, also indicating groups were
32
en
33
34 comparable (mean (SD) prior year costs for MC-cases and controls were 1779.76 (7003.45)
35
tia
36 and 1664.34 (8732.5), standardized difference = 0.02; AC-cases and controls 2385.69 (8223.62)
37
38
39 and 2314.57 (9420.75) standardized difference = 0.01).
l
40
41 The mean age of individuals injured in MCs was younger (mean 36.08) than those
42
43
44 injured in ACs (mean 38.53) (p < 0.0001). A greater proportion of MC patients were also male
45
46 (81% male) in comparison to AC patients (43% male) (p < 0.0001). Other baseline characteristics
47
48
49 and average costs amongst cases and controls after matching are presented in Supplementary
50
51 Appendix E.
52
53
54
55
56 Patient-level healthcare costs; (Table 3, Supplementary Appendix D)
57
58
59
60
Page 26 of 64
1
2
3
4
Average (mean) MC- and AC- attributable costs for medical treatment within 2 years of
5
6 a crash were $5,825 and $2,995, respectively (p < 0.0001). Acute care hospital costs accounted
7
8
9
for over 75% of this amount for both MCs and ACs (Table 3). Furthermore, most costs were
10
11 accrued within one year of the injury date. Attributable costs declined thereafter for both MCs
12
13
14
and ACs, remaining stable and similar up to 5 years following the injury date (Supplementary
15
16 Appendix D).
17
18
19
20
21
Co
22
23
24 INTERPRETATION
25
26 Main findings
nf
27
28
29 In this population-based study of patients in the largest Canadian province, we
id
30
31 measured the cost of health care provided to individuals who were injured in a MC or AC and
32
en
33
34 required treatment at hospital. Average (mean) attributable medical costs within 2 years were
35
tia
36 $5,825 and $2,995, respectively (p < 0.0001). Since the annual number of injured persons per
37
38
39 registered vehicle was triple and costs per injury were double, we estimate the total cost to the
l
40
41 universal healthcare system of injuries sustained in motor vehicles is approximately 6 times
42
43
44 greater per registered motorcycle compared to automobile. Since acute care accounted for
45
46 over 75% of the costs, severe injuries (AIS >=3) that were 10 times greater for motorcycles
47
48
49 compared to automobiles (IRR=10.4, 95% CI=8.3-13.1, p<0.0001) likely accounted for the
50
51 additional costs incurred by MCs. All rates in our study could also be multiplied by 5 and
52
53
54 expressed per kilometer travelled since the average automobile travels 5 times the distance of
55
56
57
58
59
60
Page 27 of 64
1
2
3
4
the average motorcycle (i.e. 15x injuries, 50x severe injuries, 25x deaths, 30x costs per
5
6 kilometer travelled).(24)
7
8
9
10
11 Limitations
12
13
14
Although our methodology captured the vast majority (an estimated 92%) of publicly
15
16 funded health care costs provided to Ontarians injured in MCs and ACs,(18) the most important
17
18
19 limitation of our cost calculations was our inability to capture on-going care costs of patients
20
21 requiring outpatient rehabilitation, such as those suffering acquired brain injury (ABI).
Co
22
23
24 Accordingly, our cost calculations were less comprehensive over time as patients, and their
25
26 care, are moved from the hospital setting and into the community. Since MC victims are
nf
27
28
29 significantly more likely to suffer head trauma,(8, 35, 36) and severe injuries in general, this
id
30
31 limitation only underestimated MC-attributable costs compared to those incurred by ACs (Table
32
en
33
34 2). Similarly, direct costs borne by private insurance as well as indirect or opportunity costs
35
tia
36 incurred by individual patients and society also were not captured. Indirect costs from missed
37
38
39 work days and permanent disability precluding the ability to earn, for example, may also be
l
40
41 greater after MCs than ACs since MC patients were younger and more severely injured.
42
43
44 Our data also does not include those injured and/or killed at the scene and not
45
46 transported to hospital because we identified patient encounters with health care
47
48
49 providers. Since we only identified each patients first event, it is also important to point out
50
51 that injury rates reported in this study do not include subsequent crashes, if a patient had more
52
53
54 than one.
55
56
57
58
59
60
Page 28 of 64
1
2
3
4
Implications
5
6 We found motorcycle injuries cost the public healthcare system 6 times more than
7
8
9
automobile injuries, per vehicle. Expressed in other terms, Ontarios healthcare payer
10
11 (MOHLTC) would expect to save $13.5 million dollars per year if collision rates remained
12
13
14
unchanged and the Provinces registered motorcycles were replaced with automobiles. We
15
16 argue this estimate is conservative for several reasons previously mentioned (see Limitations).
17
18
19 Despite publically available data indicating that the risk associated with driving a
20
21 motorcycle is much greater than that associated with driving an automobile, this knowledge
Co
22
23
24 has not translated to improvements in motorcycle safety.(2, 38) Medical costs may provide a
25
26 novel financial incentive to mitigate the risk of MCs. Anecdotally, insurance companies raising
nf
27
28
29 their rates in response to private medical costs have already acted as a deterrent to motorcycle
id
30
31 use.(39) Health care costs incurred by the Ontarios healthcare payer the Ministry of Health
32
en
33
34 and Long-Term Care in treating MVC (including MC) patients may be recovered, at least in
35
tia
36 part, by the Government of Ontario under the Provinces Insurance Act. Our study is the first to
37
38
39 accurately estimate costs potentially recoverable from motorcycle insurers under this
l
40
41 legislation. The higher costs and greater incidence of severe injury and death we have observed
42
43
44 may also warrant other improved public health prevention measures. For example,
45
46 supplementary private healthcare insurance, an excise tax, or another novel method of health
47
48
49 coverage for motorcyclists may be justified with the goal of improving motorcycle safety.
50
51 Our study was conducted at the level of a healthcare system serving a population of
52
53
54 approximately 13.6 million. Although exact health care costs vary in other healthcare systems,
55
56 we argue the conclusions drawn from the relative comparison of MCs to ACs apply beyond
57
58
59
60
Page 29 of 64
1
2
3
4
Canada to the rest developed world. For example, in a privately funded health care system,
5
6 insurance companies and individual providers may accept a larger share of the direct health
7
8
9
care costs we have estimated in this study. In our Province, we have captured all hospital and
10
11 physician costs provided in the Provinces publicly funded health care system.
12
13
14
15
16 Conclusions
17
18
19 We found each motorcycle incurs 3x the injuries, 6x the medical costs, and 5x the
20
21 deaths of each automobile. An understanding of these consequences may play a key role in
Co
22
23
24 public health strategy aimed at improving motorcycle safety. Future work is required to
25
26 investigate indirect or opportunity costs incurred by MCs.
nf
27
28
29
id
30
31 ACKNOWLEDGEMENTS
32
en
33
34 This research was supported by grants from the Ontario Ministry of Health and Long-
35
tia
36 Term Care (MOHLTC) and the Ontario SPOR Support Unit to the Health System Performance
37
38
39 Research Network (HSPRN: fund #06034, recipient WPW), and by the Institute for Clinical
l
40
41 Evaluative Sciences (ICES), which is also funded by an annual grant from the MOHLTC. The
42
43
44 funders had no role in study design, data collection and analysis, decision to publish, or
45
46 preparation of the manuscript. Parts of the material are based on data and information
47
48
49 compiled and provided by the Canadian Institute for Health Information (CIHI). No
50
51 endorsement by ICES, the MOHLTC or CIHI is intended or should be inferred. No benefits have
52
53
54 been received or will be received from a commercial party related directly or indirectly to the
55
56 subject of this article.
57
58
59
60
Page 30 of 64
1
2
3
4
The authors would also like to acknowledge Allan S. Detsky for providing comments on
5
6 an earlier version of the manuscript.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 31 of 64
1
2
3
4
REFERENCES
5
6 1. Sauber-Schatz E, Euerer D, Dellinger A, GT B. Vital Signs: Motor Vehicle Injury Prevention
7 United States and 19 Comparison Countries. MMWR Morb Mortal Wkly Rep 2016.
8
9
2. Ministry of Transportation. Ontario Road Safety. Annual Report 2013.
11 Economy. 1991;99(4):722-42.
12 4. Gilmore AB, Tavakoly B, Taylor G, Reed H. Understanding tobacco industry pricing
13
14
strategy and whether it undermines tobacco tax policy: The example of the UK cigarette
17 2000;321(7257):358-61.
18
Co
23
nf
27 9. Wodchis W et al. Guidelines on Person-Level Costing Using Administrative Databases in

28
29 Ontario. Working Paper Series. Toronto: Health System Performance Research Network. 2013.
id
30 10. Chen A, Bushmeneva K, Zagorski B, Colantonio A, Parsons D, Wodchis WP. Direct cost
31 associated with acquired brain injury in Ontario. BMC Neurology. 2012;12.
32
en
11. Nikitovic M, Wodchis WP, Krahn MD, Cadarette SM. Direct health-care costs attributed
33
34 to hip fractures among seniors: a matched cohort study. Osteoporos Int. 2013;24(2):659-69.
35 12. Munce SEP, Wodchis WP, Guilcher SJT, Couris CM, Verrier M, Fung K, et al. Direct costs
tia
36 of adult traumatic spinal cord injury in ontario. Spinal Cord. 2013;51(1):64.

37 13. Rosella LC, Fitzpatrick T, Wodchis WP, Calzavara A, Manson H, Goel V. High-cost health
38
l

41 14. Tanuseputro P, Wodchis WP, Fowler R, Walker P, Bai YQ, Bronskill SE, et al. The health
42 care cost of dying: A population-based retrospective cohort study of the last year of life in
43
44 Ontario, Canada. PLoS ONE. 2015;10(3).
45 15. Rosella LC, Lebenbaum M, Fitzpatrick T, O'Reilly D, Wang J, Booth GL, et al. Impact of
46 diabetes on healthcare costs in a population-based cohort: a cost analysis. Diabetic medicine : a
47
journal of the British Diabetic Association. 2016;33(3):395.
48
50 Accid Anal Prev. 2011;43(6):2128-33.
51 17. Redelmeier DA, May SC, Thiruchelvam D, Barrett JF. Pregnancy and the risk of a traffic
52
crash. CMAJ. 2014;186(10):742-50.
53
56 2016.
57
58
59
60
Page 32 of 64
1
2
3
4
19. Juurlink D, Preyra C, Croxford R, Chong A, Austin P, Tu J, Laupacis A. Canadian Institute
5 for Health Information Discharge Abstract Database: A Validation Study. Toronto: Institute for
7 20. Austin PC. Comparing paired vs non-paired statistical methods of analyses when making
8
9
inferences about absolute risk reductions in propensity-score matched samples. Statistics in
10 medicine. 2011;30(11):1292.
13
14
22. Austin PC. Optimal caliper widths for propensity-score matching when estimating
16 statistics. 2011;10(2):150.
17 23. Weiner JP, Abrams C. The Johns Hopkins ACG System: Technical Reference Guide
18
19 Version 10.0. 2011.
Co
23
25 25. Santaguida PL, Hawker GA, Hudak PL, Glazier R, Mahomed NN, Kreder HJ, et al. Patient
nf

28
id
31 2006;63(10):2604.
32
en
27. Kralj B. Measuring Rurality for Purposes of Health Care Planning: An Empirical
33
tia
36 difference-in-differences approach. JAMA. 2014;312(22):2401-2.

38
l

41 30. Haas B, Stukel TA, Gomez D, Zagorski B, De Mestral C, Sharma SV, et al. The mortality
42 benefit of direct trauma center transport in a regional trauma system: a population-based
43
44 analysis. J Trauma Acute Care Surg. 2012;72(6):1510-5; discussion 5-7.
45 31. Gomez D, Haas B, de Mestral C, Sharma S, Hsiao M, Zagorski B, et al. Gender-associated
46 differences in access to trauma center care: A population-based analysis. Surgery.
47
2012;152(2):179-85.
48
49 32. Gomez D, Haas B, de Mestral C, Sharma S, Hsiao M, Zagorski B, et al. Institutional and
50 provider factors impeding access to trauma center care: an analysis of transfer practices in a
51 regional trauma system. J Trauma Acute Care Surg. 2012;73(5):1288-93.
52
33. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates
53
55 2009;28(25):3083.
56
57
58
59
60
Page 33 of 64
1
2
3
4
34. Normand ST, Landrum MB, Guadagnoli E, Ayanian JZ, Ryan TJ, Cleary PD, et al. Validating
8
9
Prevention. 2005;37(2):349.
13
14
CMAJ. 2016;188(3):182-8.
15 38. The National Highway Traffic Safety Administration. TRAFFIC SAFETY FACTS 2015.
17 April 27, 2017.
18
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 34 of 64
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 1st DIFFERENCE
19 ()
20
21
22
23
Case
Year prior Co
INDEX DATE First year Second Year
//
Up to March 31, 2015
24
25
nf
26
27
id
28
29
30 en
31
32
33 tia
Figure 1: Description of the study design. For each case, we identified one control who had health
34
35
36
37
l
system contact during the study period but no motor vehicle accident code. Index dates for controls
were randomly assigned based on the distribution of index dates for cases in the same fiscal year.
Costs attributable to MCs and ACs were then calculated by using a difference-in-differences
38 approach: the baseline cost accrued the year prior to the MC or AC was subtracted from costs in the
39 first and second year following the collision (first difference) and then compared to the same
40 difference among controls (second difference).
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 35 of 64
1
2
3
4
Table 1. Annual incidence of motorcycle- and automobile-related injured persons and deaths
5 per 100,000 registered vehicles* (fiscal year 2007/12)
6
7 2007 2008 2009 2010 2011 2012 Mean
8
9 INJURIES
10 Motorcycle 2,568 2,328 2,212 2,014 2,105 1,938 2,194
11 Automobile 767 712 711 751 681 685 718
12
13
DEATHS
14 Motorcycle 16 14 12 14 10 14 14
15 Automobile 3 3 3 3 2 3 3
16
17
18 *Calculated using motor vehicle registration data made publically available by Ontarios
19 Ministry of Transportation (see Supplementary Appendix C).
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
Page 36 of 64
1
2
3 Table 2: Clinical outcomes among patients with motorcycle- and automobile-related
4
injuries.
5
6
7 Motorcycle injury, N (%) Automobile injury, N (%) p-value*
8
9 N=26,257 N=276,760
10 Level of care
11
12
ED visit only 23,123 (88.07) 265,020 (95.76) < 0.0001
13 Required Hospitalization 3,123 (11.93) 11,740 (4.24) < 0.0001
14 Required ICU admission 606 (2.31) 3,018 (1.09) < 0.0001
15
Injury Severity Score (ISS)
16
17 <9 19,090 (90.46) 156,529 (97.11) < 0.0001
18 9-15 1,539 (5.86) 4,348 (1.57)
19
16-24 668 (2.54) 2,238 (0.81)
20
21 25 + 297 (1.13) 1,409 (0.51)
Co
22 Severe injuries ( Abbreviated Injury Scale >=3), overall and by body region
23
Any severe injury 1,493 (5.69) 4,621 (1.67) < 0.0001
24
25 Head 58 (0.22) 440 (0.16) 0.0179
26 Face NR NR 0.5204
nf
27
Neck NR NR 0.8807
28
29 Thorax 347 (1.32) 1,173 (0.42) < 0.0001
id
30 Abdomen 53 (0.20) 198 (0.07) < 0.0001

31
Spine 36 (0.14) 390 (0.14) 0.8749
32
en
33 Upper Extremity 396 (1.51) 830 (0.30) < 0.0001

34 Lower Extremity 600 (2.29) 1568 (0.57) < 0.0001
35
Missing NR NR -
tia
36 *
37 p-values were calculated using chi-square tests.
38 NR; small cell sizes not reportable according to privacy guidelines at the Institute for Clinical
39 Evaluative Sciences
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
Page 37 of 64
1
2
3 Table 3: Attributable cost (2013$CAN) of suffering a motorcycle- or automobile-related
4
injury and/or death within 2 years
5
6 Healthcare Sector Mean MC-attributable cost (95% CI) Mean AC-attributable cost (95% CI) p-value*
7
8 Acute care hospital 4,409 (4,247 to 4,571) 2,185 (1,883 to 2,340) < 0.0001
9
Physician 1,175 (1,109 to 1,,241) 664 (534 to 731) < 0.0001
10
11 Drug/Laboratory -6 (-54 to 42) 27 (-77 to 81) < 0.0001
12
Continuing care 224 (166 to 282) 104 (-16 to 165) < 0.0001
13
14 Assistive devices 14 (-7 to 25) 13 (-32 to 36) < 0.0001
15 Total cost 5,825 (5,651 to 5,999) 2,995 (2,669 to 3,161) < 0.0001
16
*
17 p-values compared MC- with AC- attributable costs overall, and by sector, within 2 years and
18 were calculated using independent samples t-tests. CI; confidence interval.
19
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
Page 38 of 64
Supplementary Appendix MC costs

1
2
3
4
Table of contents
5
6
7 Supplementary Appendix A. Data sources Page 2
8
9
10 Supplementary Appendix B. International Classification of Diseases, 10th Revision (ICD-10) codes
11 specifying cohort inclusion and exclusion Page 3
12
13
14
Supplementary Appendix C. Registered motorcycles and automobiles in Ontario, by year during
15 the study period (MTO) Page 4
16
17 Supplementary Appendix D. Patient-level direct healthcare costs attributable to MC- and AC-
18
19 related injury or death over time Page 5
20
21 Supplementary Appendix E. Baseline characteristics and average costs of cases and matched
Co
22 controls (fiscal year 2007/12) Page 6
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
Page 39 of 64

1
2
3
4
Supplementary Appendix A. Data sources.
5
6 Data were obtained from several health administrative databases in Ontario, Canada.
7
8
9
A) The Canadian Institute for Health Information Discharge Abstract Database (CIHI-DAD),
10 and the National Ambulatory Care Reporting System (NACRS) contain detailed, validated
11 diagnostic and procedural information about all hospital admissions and emergency
12 department visits in the Province, respectively.(1)
13
14
15 B) Demographic and vital status information were obtained from the OHIP Registered
16 Persons Database (RPDB).
17
18
19 C) Costs were considered by using previously established methods for case-costing in
20 Ontarios administrative databases:(2)
21 acute hospitalizations, emergency department, and same day surgery costs are
Co
22 calculated using the resource intensity weight method,(3)
23
24 complex continuing care (CCC) and long-term care (LTC) costs are based on resource
25 utilization groups and length of stay,(4)
26 physician service costs and prescription drug costs are based on direct payments to
nf
27
28
the physician/pharmacy,(2)
29 rehabilitation costs are based on the rehabilitation patient group case mix
id
30 classification and length of stay,(5-7)

31 costs for home care are based on the number and costs per visit.(8)
32
en
33
34 These databases are held securely in linked, encoded form at the Institute for Clinical Evaluative
35 Sciences (ICES) and have been used in prior research estimating healthcare costs (9-14) and
tia
36 studying traffic collisions.(15-17)

37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
Page 40 of 64

1
2
3
4
Supplementary Appendix B. International Classification of Diseases, 10th Revision (ICD-10)
5 codes specifying cohort inclusion and exclusion.
6
7 Cohort inclusion:
8
9
Motorcycle:
10 V20-V29: Motorcycle rider/passenger injured in transport accident
11 Includes: moped, motor scooter, motorcycle with sidecar, motorized bicycle
12 Includes: driver, rider, unspecified
13
14
Excludes: three-wheeled motor vehicle (V30-V39)
15
16 Automobile:
17 V40-V49: Car occupant injured in transport accident
18
19 V50-V59: Occupant of pick-up truck or van injured in transport accident
20 V70-V79: Bus occupant injured in transport accident
21
Co
22 Cohort exclusion (motor-vehicle associated injury in the prior 2 years):
23
24 V02: Pedestrian injured in collision with two- or three-wheeled motor vehicle
25 V03: Pedestrian injured in collision with car, pick-up truck or van
26 V04: Pedestrian injured in collision with heavy transport vehicle or bus
nf
27 V09: Pedestrian injured in other and unspecified transport accidents

28
29 V12: Pedal cyclist injured in collision with two- or three-wheeled motor vehicle
id
30 V13: Pedal cyclist injured in collision with car, pick-up truck or van
31 V14: Pedal cyclist injured in collision with heavy transport vehicle or bus
32
en
V19: Pedal cyclist injured in other and unspecified transport accidents

33
34 V20-V29: Motorcycle rider injured in transport accident
35 V30-V39: Occupant of three-wheeled motor vehicle injured in transport accident
tia
36 V40-V49: Car occupant injured in transport accident

37 V50-V59: Occupant of pick-up truck or van injured in transport accident
38
39 V60-V69: Occupant of heavy transport vehicle injured in transport accident
l
40 V70-V79: Bus occupant injured in transport accident

41 V80.3: Rider or occupant injured in collision with two- or three-wheeled motor vehicle
42 V80.4: Rider or occupant injured in collision with car, pick-up truck, van, heavy transport vehicle
43
44 or bus
45 V80.5: Rider or occupant injured in collision with other specified motor vehicle
46 V86: Occupant of special all-terrain or other motor vehicle designed primarily for off-road use,
47
injured in transport accident. Excludes: vehicle in stationary use or maintenance (W31.-)
48
49 V87: Traffic accident of specified type but victim's mode of transport unknown. Excludes:
50 collision involving: pedal cyclist (V10-V19), pedestrian (V01-V09), and V87.9
51 V88: Nontraffic accident of specified type but victim's mode of transport unknown. Excludes:
52
collision involving: pedal cyclist (V10-V19), pedestrian (V01-V09) and V88.9
53
54 V89: Motor- or nonmotor-vehicle accident, type of vehicle unspecified
55
56
57
58
59
60
3
Page 41 of 64

1
2
3
4
Supplementary Appendix C. Registered motorcycles and automobiles in Ontario, by year
5 during the study period (MTO)(18)
6
7 2007 2008 2009 2010 2011 2012
8
REGISTERED MOTORCYCLE 173314 191572 200810 211536 220026 228303
9
10 REGISTERED AUTOMOBILE 6339389 6446988 6488233 6547976 6605791 6683622
11
12
13
14
15
16
17
18
19
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
4
Page 42 of 64

1
2
3
4
Supplementary Appendix D. Patient-level direct healthcare costs attributable to MC- and AC-
5 related injury or death over time.
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
5
Page 43 of 64

1
2
3
4
Supplementary Appendix E. Baseline characteristics and average costs of cases and matched
5 controls (fiscal year 2007/12).
6 Motorcycle-related injury or death Automobile-related injury or death
7 Variable Class Case Control SD a
Case Control SDa p-valueb
8 Mean (SD)/% Mean (SD)/% Mean (SD)/% Mean (SD)/%
9 26,257 26,257 - 276,760 276,760 - -
10 Baseline Characteristics
11 Age, mean (SD) 36.08 (16.34) 36.08 (16.34) 0.00 38.53 (19.09) 38.53 (19.09) 0.00 < 0.0001
12 Male, % 81.44 81.44 0.00 43.07 43.07 0.00 < 0.0001
13 CADG, %
14 acute minor 1 62.69 64.80 0.04 73.40 71.97 0.03 < 0.0001
15 acute major 2 52.36 54.43 0.04 64.19 64.03 0.00 < 0.0001
16 likely to recur 3 47.83 49.43 0.03 59.79 58.22 0.03 < 0.0001
17 asthma 4 5.16 5.70 0.02 7.08 6.17 0.04 < 0.0001
18 chronic medical unstable 5 14.03 12.97 0.03 18.66 16.85 0.05 < 0.0001
19 chronic medical stable 6 25.67 27.92 0.05 36.53 33.78 0.06 < 0.0001
20 chronic specialty stable 7 4.09 3.38 0.04 4.06 3.79 0.01 0.7999
21 eye dental 8 3.48 3.82 0.02 6.28 5.40 0.04 < 0.0001
Co
22 chronic specialty unstable 9 4.27 4.48 0.01 6.81 5.90 0.04 < 0.0001
23 psychosocial 10 28.15 29.92 0.04 33.66 34.56 0.02 < 0.0001
24 prevention, administration 11 23.91 25.82 0.04 37.75 36.81 0.02 < 0.0001
25 pregnancy 12 0.56 1.55 0.10 4.61 5.53 0.04 < 0.0001
26 LHIN, % < 0.0001
nf
27 1 5.30 4.87 0.02 4.21 4.80 0.03

28 2 10.64 8.70 0.07 8.75 8.62 0.00
29 3 5.92 5.64 0.01 5.29 5.30 0.00
id
30 4 13.96 11.76 0.07 11.84 12.14 0.01

31 5 4.57 7.30 0.12 7.73 7.42 0.01
32 6 5.62 7.78 0.09 7.93 7.24 0.03
en
33 7 6.22 6.17 0.00 5.00 6.07 0.05

34 8 8.08 11.85 0.13 13.34 12.04 0.04
35 9 11.38 12.66 0.04 13.30 12.98 0.01
10 4.74 4.38 0.02 4.49 4.45 0.00
tia
36
37 11 10.16 8.38 0.06 8.31 8.40 0.00
38 12 5.11 3.59 0.07 3.71 3.63 0.00
39 13 6.46 4.92 0.07 4.26 4.95 0.03
l
40 14 1.84 2.00 0.01 1.85 1.96 0.01

Rurality Index, mean (SD) 15.85 (19.57) 11.32 (18.12) 0.24 11.39 (17.42) 11.24 (18.00) 0.01 < 0.0001
41
Income quintile, % < 0.0001
42
1 16.94 21.88 0.13 21.18 22.17 0.02
43
2 19.92 21.12 0.03 21.15 21.60 0.01
44
3 20.67 20.88 0.01 21.02 20.75 0.01
45
4 21.97 19.69 0.06 20.27 19.53 0.02
46
5 20.50 16.43 0.10 16.39 15.96 0.01
47
Average cost
48
Previous year, mean (SD) 1779.76(7003.45) - 2385.69(8223.62) 2314.57(9420.75) - < 0.0001
49 1664.34(8732.5)
50 Year 1, mean (SD) 7012.96(23969.21) - 4819.68(18451.29) 2276.46(9582.72) - < 0.0001
51 1625.8(7878.04)
52 Year 2, mean (SD) 2380.74(9522.63) - 2881.8(10993.91) 2287.61(9337.95) - < 0.0001
53 1712.3(8556.51)
a b
54 Standardized differences 0.1 represent meaningful differences in covariates between groups. p-values
55 compared MC-cases with AC-cases and were calculated using independent samples t-tests for continuous variables
56 and chi-square tests for categorical variables.
c
57 LHIN; Local Health Integrated Network.
58
59
60
6
Page 44 of 64

1
2
3
4
5 Supplementary Appendix References
6
8
9
for Health Information Discharge Abstract Database: A Validation Study. Toronto: Institute for
11 2. Wodchis W et al. Guidelines on Person- Level Costing Using Administrative Databases in
13
14
3. Pink GH, Bolley HB. Physicians in health care management: 3. Case Mix Groups and
15 Resource Intensity Weights: An overview for physicians. CMAJ. 1994;150(6):889.
16 4. Canadian Institute for Health Information (2010) CCRS technical document Ontario
17 RUG weighted patient day (RWPD) methodology. http://www.cihi.ca/CIHI-ext-
18
19 portal/pdf/internet/CCRS_RWPD_METHO_TECH_EN.
20 5. Sutherland J, Walker J. Technical report - Development of the rehabilitation patient
21 group (RPG) case mix classification methodology and weighting system for adult inpatient
Co
22 rehabilitation. Joint Policy and Planning Committee. 2006.
23
24 6. Sutherland JM, Walker J. Challenges of rehabilitation case mix measurement in Ontario
25 hospitals. Health Policy. 2008;85(3):336.
26 7. Sutherland JM. Technical Report: Evaluation and Revision of the Rehabilitation Patient
nf
27 Group (RPG) Case Mix System. 2010.

28
29 8. Home Care Database CCAC Guidelines. 2009.
id

32
en
10. Nikitovic M, Wodchis WP, Krahn MD, Cadarette SM. Direct health-care costs attributed
33
tia

38
l

43
47
journal of the British Diabetic Association. 2016;33(3):395.
48
50 Accid Anal Prev. 2011;43(6):2128-33.
52
crash. CMAJ. 2014;186(10):742-50.
53
56 2016.
57
58
18. Ministry of Transportation. Ontario Road Safety. Annual Reports 2007 to 2012.
59
60
7
Page 45 of 64

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
8
Page 46 of 64
1 STROBE StatementChecklist of items that should be included in reports of cohort studies

2 Item Location in
3 No manuscript where
4 Recommendation items are reported
5
Title and abstract 1 (a) Indicate the studys design with a commonly used Page 1 (Title)
6
7 term in the title or the abstract
8 (b) Provide in the abstract an informative and balanced Page 2 (Abstract)
9 summary of what was done and what was found
10
11 Introduction
12 Background/rationale 2 Explain the scientific background and rationale for the Page 3 (Introduction)
13 investigation being reported
14 Objectives 3 State specific objectives, including any prespecified Page 3 (Introduction)
15
hypotheses
16
17 Methods
18 Study design 4 Present key elements of study design early in the paper Page 2 (Abstract)
19
Page 4 (Methods)
20
21 Setting 5 Describe the setting, locations, and relevant dates, Page 5 (Methods)
Co
22 including periods of recruitment, exposure, follow-up,
23 and data collection
24 Participants 6 (a) Give the eligibility criteria, and the sources and Page 4-6 (Methods)
25
methods of selection of participants. Describe methods of
26
nf
27 follow-up
28 (b) For matched studies, give matching criteria and Page 5-6 (Methods)
29
id
number of exposed and unexposed

30
Variables 7 Clearly define all outcomes, exposures, predictors, Page 5-6 (Methods)
31
32 potential confounders, and effect modifiers. Give
en
33 diagnostic criteria, if applicable

34 Data sources/ 8* For each variable of interest, give sources of data and Page 5-6 (Methods)
35 measurement details of methods of assessment (measurement).
tia
36
Describe comparability of assessment methods if there is
37
38 more than one group
39 Bias 9 Describe any efforts to address potential sources of bias Page 8-9
l
40 (Discussion)
41
Study size 10 Explain how the study size was arrived at N/A
42
43 Quantitative variables 11 Explain how quantitative variables were handled in the Page 4-6 (Methods)
44 analyses. If applicable, describe which groupings were
45 chosen and why
46 Statistical methods 12 (a) Describe all statistical methods, including those used Page 6 (Methods)
47
48 to control for confounding
49 (b) Describe any methods used to examine subgroups and Page 6 (Methods)
50 interactions
51 (c) Explain how missing data were addressed N/A
52
53 (d) If applicable, explain how loss to follow-up was N/A
54 addressed
55 (e) Describe any sensitivity analyses N/A
56
Results
57
58 Participants 13* (a) Report numbers of individuals at each stage of N/A
59 studyeg numbers potentially eligible, examined for
60 eligibility, confirmed eligible, included in the study,
1
Page 47 of 64
1 completing follow-up, and analysed

2 (b) Give reasons for non-participation at each stage N/A
3
(c) Consider use of a flow diagram N/A.
4
5 Descriptive data 14* (a) Give characteristics of study participants (eg Table 2.
6 demographic, clinical, social) and information on
7 exposures and potential confounders
8 (b) Indicate number of participants with missing data for Table 2.
9
each variable of interest
10
11 (c) Summarise follow-up time (eg, average and total Page 4-6 (Methods)
12 amount)
13 Outcome data 15* Report numbers of outcome events or summary measures Table 3.
14
over time
15
16 Main results 16 (a) Give unadjusted estimates and, if applicable, Tables 1 and 2.
17 confounder-adjusted estimates and their precision (eg,
18 95% confidence interval). Make clear which confounders
19 were adjusted for and why they were included
20
21 (b) Report category boundaries when continuous Page 4-6 (Methods)
variables were categorized
Co
22
23 (c) If relevant, consider translating estimates of relative N/A
24 risk into absolute risk for a meaningful time period
25
Other analyses 17 Report other analyses doneeg analyses of subgroups N/A.
26
nf
27 and interactions, and sensitivity analyses

28 Discussion
29
id
Key results 18 Summarise key results with reference to study objectives Page 8-11
30
31 (Discussion)
32 Limitations 19 Discuss limitations of the study, taking into account Page 8-9
en
33 sources of potential bias or imprecision. Discuss both (Discussion)

34 direction and magnitude of any potential bias
35
Interpretation 20 Give a cautious overall interpretation of results Page 9-11
tia
36
37 considering objectives, limitations, multiplicity of (Discussion)
38 analyses, results from similar studies, and other relevant
39 evidence
l
40
Generalisability 21 Discuss the generalisability (external validity) of the Page 10-11
41
42 study results (Discussion)
43 Other information
44 Funding 22 Give the source of funding and the role of the funders for Page 2 (Abstract)
45
46 the present study and, if applicable, for the original study and Page 11
47 on which the present article is based (Acknowledgements)
48
49 *Give information separately for exposed and unexposed groups.
50
51
52 Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and
53 published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely
54 available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at
55 http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is
56
available at http://www.strobe-statement.org.
57
58
59
60
2
Page 48 of 64
1
2 YEAR AFTER ACCIDENT MOTORCYCLE CAR
3 1 5308 2436
4
5 2 567 507
6 3 412 461
7 4 394 483
8
9 5 129 451
10
11
12
13
14
15
16
17
18
19
20
21
Co
22
23
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 49 of 64
1
2
3
4
5
6 MANUSCRIPT 1 MC costs
7
8
9 Direct Medical Costs of Motorcycle Crashes in Ontario
10
11 Authors:
12 Daniel Pincus1,2,3 MD
13 David Wasserstein1,4 MD, MSc, MPH
14 Avery B. Nathens1,2,3,4 MD, MPH, PhD
15 Yu Qing Bai2,3 MSc
16 Donald A. Redelmeier2,3,5,6 MD, MS(HSR)
17 Walter P. Wodchis2,3 PhD
18
19 Affiliations:
1
20 Department of Surgery, University of Toronto, Toronto, Canada
2
21 Institute for Clinical Evaluative Sciences, Toronto, Canada
3
Co
22 Institute of Health Policy, Management and Evaluation, University of Toronto, Toronto, Canada
4
23 Department of Surgery, Sunnybrook Health Sciences Centre, Toronto, Canada
5
24 Division of General Internal Medicine, Sunnybrook Health Sciences Centre, Toronto, Canada,
6
25 Center for Leading Injury Prevention Practice Education & Research, Toronto, Canada
26
nf
27 Correspondence and address for all authors:

28 Daniel Pincus MD
29 Department of Surgery, Division of Orthopaedic Surgery, University of Toronto
id
30 149 College Street, Room 508-A

31 Toronto, ON M5T 1P5
32 Phone: 416-946-7957 (w), 647-244-3324 (c)
en
33 Fax: 416-978-3928
34 d.pincus@utoronto.ca
35
tia
36 Competing Interests / Disclosures:

37 There are no relevant disclosures or competing interests. All authors meet ICMJE criteria. Formatted: Font: Not Bold
38
39 Source of funding:
l
40 This research was supported by grants from the Ontario Ministry of Health and Long-Term Care
41 (MOHLTC) and the Ontario SPOR Support Unit to the Health System Performance Research
42 Network (HSPRN: fund #06034, recipient WPW), and by the Institute for Clinical Evaluative
43 Sciences (ICES), which is also funded by an annual grant from the MOHLTC. The funders had no
44 role in study design, data collection and analysis, decision to publish, or preparation of the
45 manuscript. Parts of the material are based on data and information compiled and provided by
46 the Canadian Institute for Health Information (CIHI). No endorsement by ICES, the MOHLTC or
47 CIHI is intended or should be inferred. No benefits have been received or will be received from
48 a commercial party related directly or indirectly to the subject of this article.
49
50 Key Words: Motorcycles, Accidents, Traffic/economics, Traffic/statistics & numerical data,
51 Insurance, Health/economics, Retrospective Studies
52
53
54
55
56
57
58
59
60
Page 50 of 64
1
2
3
4
5
7
8
9 ABSTRACT (Word Count: 2467)
10
11 Background. No reliable estimate of costs incurred by motorcycle crashes (MCs) exists. Our
12 objective was to calculate the direct costs of all publicly funded medical care provided to
13 individuals following MCs compared to automobile crashes (ACs).
14
15 Methods. We conducted a population based, matched cohort study of adults in Ontario that
16 presented to hospital because of a MC or AC from 2007 through 2013. For each case, we
17 identified one control absent a motor vehicle crash during the study period. Direct costs for
18 each case and control were estimated as 2013 $ Canadian from the payer perspective using
19 methodology that links healthcare use to individuals over time. MC- and AC-attributable costs
20 within 2 years were then calculated using a difference-in-differences approach.
21
Co
22 Results. We identified 26,831 patients injured from MCs and 281,826 injured from ACs. Mean
23 MC- and AC-attributable costs were $5,825 and $2,995, respectively (p<0.0001). The rate of
24 injury was also 3 times greatertriple for MCs compared to ACs (2,194 injured yearly/100,000
25 registered motorcycles versus 718 injured yearly/100,000 registered automobiles; IRR=3.1, 95%
26 CI=2.8-3.3, p<0.0001). Severe injuries, defined as those with an Abbreviated Injury Scale >=3,
nf
27 were 10 times greater (125 severe injuries per yearly/100,000 registered motorcycles versus 12
28 severe injuries per yearly/100,000 registered automobiles; IRR=10.4, 95% CI=8.3-13.1,
29 p<0.0001).
id
30
31 Interpretation. Considering both the attributable cost and higher rate of injury, we found each
32 registered motorcycle in Ontario costs the public healthcare system 6 times the amount of each
en
33 registered automobile. Medical costs may provide an additional incentive to improve

35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 51 of 64
1
2
3
4
5
7
8
9 INTRODUCTION
10
11 Motor vehicle safety has improved significantly over the past 100 years.(1) Between
12
13 2000 and 2010, for example, mortality related to motor vehicle crashes (MVCs) decreased by
14
15 55.1% in 19 developed countries.(1) In contrast, deaths and injuries among a subset of MVC
16
17 patients those injured in motorcycle crashes (MCs) remained stable during the same time
18
19 period.(2)
20
21 Although medical costs may provide incentive to improve motorcycle safety,(3-5) no
Co
22
23 reliable estimate of medical costs incurred by MCs exists. Prior calculations of medical costs
24
25 attributable to MCs are limited to reviews of hospital charges at single centers.(6) Estimates
26
nf
27 derived from these reviews are incomplete, neglecting costs incurred after a patients
28
29 discharge, for example.(6-8)
id
30
31 The calculation of patient level medical costs in Ontario, Canada (population 13.6 million
32
en
33 in 2014) is now possible with methodology that links publicly funded healthcare use to
34
35 individuals over time.(9) Using this methodology from the payer perspective, our objective was
tia
36
37 to calculate the direct costs of all publicly funded medical care provided to individuals
38
39 presenting to hospital following MCs in comparison to automobile crashes (ACs). We also
l
40
41 examined the population incidence of injuries resulting from these crashes. Our hypothesis was
42
43 that medical costs and injury rates attributable to MCs, which can occur at high speeds with less
44
45 personal protection, are significantly higher than those attributable to ACs.
46
47
48
49 METHODS
50
51 Setting
52
53
54
55
56
57
58
59
60
Page 52 of 64
1
2
3
4
5
7
8
9 We conducted a population based, matched cohort study in Ontario, Canada using
10
11 linked health administrative databases at the Institute for Clinical Evaluative Sciences (ICES)
12
13 (see Supplementary Appendix A data sources). Ontarians have their medically necessary
14
15 health care interactions, provider information, and demographic characteristics recorded in
16
17 these databases. These data have previously been used to estimate medical costs(10-15) and
18
19 study traffic crashes.(16-18) The research protocol was approved by the Research Ethics Board
20
21 at Sunnybrook Health Sciences Centre.
Co
22
23
24
25 Patients
26
nf
27 The study design is displayed in Figure 1. Ontarians who presented to the emergency
28
29 department (ED) or were admitted to hospital following a MC or AC between April 1, 2007 to
id
30
31 March 31, 2013 were eligible for inclusion. International Classification of Diseases, 10th Revision
32
en
33 (ICD-10) codes were used to identify injured motor vehicle occupants [MC (V20-V29) and AC
34
35 (V40-V49, V50-V59, V70-V79), respectively].(19) We excluded non-Ontario residents and those
tia
36
37 with prior hospital presentation for a motor-vehicle related injury of any type within 2 years
38
39 before the index date (see Supplementary Appendix B database codes).
l
40
41 For each MC and AC case, we identified one control from a 30% generalrepresentative
42
43 sample subset of the Ontario population, without replacement.(20) The purpose of these
44
45 controls was to determine baseline medical costs absent a motor vehicle crash and thus enable
46
47 the calculation of incremental costs. Controls had health system contact but no documented
48
49 motor-vehicle related injury during this timeframe or dating back 2 years prior to the index
50
51 date. Index dates for controls were randomly assigned based on the distribution of index dates
52
53
54
55
56
57
58
59
60
Page 53 of 64
1
2
3
4
5
7
8
9 for cases in the same fiscal year. Greedy mMatching occurred on age (+/- 90 days), sex, and the
10
11 logit of a propensity score with a caliper of 0.2*standard deviation (20-22) and was calculated
12
13 based on patient comorbidity, income quintile, and residential location (see Covariates).
14
15
16
17 Covariates
18
19 We measured several covariates that have been shown to influence health care
20
21 utilization in Ontario.(13) Age and sex were analyzed as continuous and categorical variables,
Co
22
23 respectively. Comorbidities listed on outpatient and hospital discharge abstracts in the two
24
25 years before the index date were categorized according to Collapsed Aggregate Diagnosis
26
nf
27 Groups (CADGs).(23) Neighborhood income quintile was used as a validated surrogate measure
28
29 for socioeconomic status and social deprivation.(24-26) Patient location of residence was
id
30
31 classified according to Local Health Integrated Network (LHIN) and the Rurality Index of Ontario
32
en
33 (RIO).(27)
34
35
tia
36
37 Outcomes
38
39 The primary outcome of this study was direct medical costs within 2 years, attributable
l
40
41 to MCs and ACs, and paid by Ontarios Ministry of Health and Long-term Care (MOHLTC). First,
42
43 medical costs were calculated for each case and control during each fiscal year of the study
44
45 period using established patient level costing methodology (see Supplementary Appendix A).(9)
46
47 MC- and AC-attributable costs within 2 years were then calculated using a difference-in-
48
49 differences approach: a baseline cost accrued the year prior to a crash was subtracted from
50
51
52
53
54
55
56
57
58
59
60
Page 54 of 64
1
2
3
4
5
7
8
9 costs in the first and second year following the crash (first difference) and then compared to
10
11 the same difference among controls (second difference).(28)
12
13 Costs were categorized as: (a) acute care hospital including the ED, index admission,
14
15 rehabilitation, and any subsequent readmissions, (b) physician, (c) drugs/laboratory costs
16
17 outside of hospitals, (d) continuing care including residential long-term care and home care,
18
19 and (e) assistive devices.(18) Care episodes that spanned more than one fiscal year were
20
21 divided on a pro rata basis. All costs were expressed as 2013 Canadian dollars using the health
Co
22
24
25 Secondary outcomes included clinical data. First, the annual rates of those injured and
26
nf
27 dying (within 30 days) from MCs and ACs were reported, so long as the patient presented to
28
29 hospital after the crash. More detailed clinical data were also compared between MC and AC
id
30
31 cases. We identified the highest level of care required for each patient: ED visit only,
32
en
33 hospitalization, or intensive care unit (ICU) admission. The Injury Severity Score (ISS) and
34
35 Abbreviated Injury Scale (AIS) were used to assess injury severity. ISS and AIS were obtained
tia
36
37 from ICD-10 codes by means of a validated algorithm.(29) ISS was categorized as <9, 9-15, 16-
38
39 24, or >=25. Severe injuries were those with an AIS>=3 and were reported overall and for each
l
40
41 anatomical region. The algorithm and operational definitions have been used previously in
42
43 population based research of Ontarios administrative data.(30-32)
44
45
46
47 Statistical Analysis
48
49 Baseline characteristics of cases and controls were reported as means and proportions
50
51 and compared by using standardized differences (greater than 0.1 being considered indicative
52
53
54
55
56
57
58
59
60
Page 55 of 64
1
2
3
4
5
7
8
9 of imbalance).(33, 34) We also compared characteristics between MC-cases with AC-cases by
10
11 using independent samples t-tests for continuous variables and chi-square tests for categorical
12
13 variables. Longer-term costs (up to 5 years) were also compared between matched cases and
14
15 controls enrolled before FY 2011 for whom longer-term follow-up (>2 years) was available.
16
17 Annual rates of injuries and deaths were expressed per 100,000 registered motorcycles or
18
19 automobiles in the Province by using publicly available data from the Ministry of Transportation
20
21 (MTO) (see Supplementary Appendix C).(2) Incidence rate ratios (IRRs) with 95% confidence
Co
22
23 intervals (CIs) were then used to compare injury and death rates between MC- and AC-cases.
24
25 All analyses were performed using SAS software (SAS version 9.3 and SAS Enterprise Guide
26
nf
27 version 6.1; SAS Institute) and the type-I error probability was set to 0.05.
28
29
id
30
31 RESULTS
32
en
33 Descriptive clinical data; (Table 1a and 1b, Table 2)

34
35 We identified 26,831 Ontarians during the study period injured during MCs and 281,826
tia
36
37 injured during ACs. The annual incidence of injured persons was 3 times greatertriple for MCs
38
39 compared to ACs (2,194 injured per year/100,000 registered motorcycles versus 718 injured per
l
40
41 year/100,000 registered automobiles; IRR=3.1, 95% CI=2.8-3.3, p<0.0001; Table 1a). Deaths (at
42
43 30d) were approximately 5 times greater (14 deaths per year/100,000 registered motorcycles
44
45 versus 3 deaths per year/100,000 registered automobiles; IRR=4.7, 95% CI=2.9-7.6, p<0.0001;
46
47 Table 1b).
48
49 In terms of the level of care required for each patient, those injured in MCs were
50
51 significantly more likely to require hospital and ICU admission than those injured in ACs (11.93%
52
53
54
55
56
57
58
59
60
Page 56 of 64
1
2
3
4
5
7
8
9 versus 4.24%, p < 0.0001 and 2.31% versus 1.09%, p < 0.0001, respectively). MC patients were
10
11 also more likely to have an ISS > 16 (3.67% versus 1.32%, p < 0.0001) and suffer severe injuries
12
13 (5.69% versus 1.67% with AIS>=3, p < 0.0001). Expressed per registered vehicle, severe injuries
14
15 were 10 times greater for MCs than ACs (125 severe injuries per year/100,000 registered
16
17 motorcycles versus 12 severe injuries per year/100,000 registered automobiles; IRR=10.4, 95%
18
19 CI=8.3-13.1, p<0.0001). Severe head, thorax, abdominal and extremity injuries were all more
20
21 common among patients injured in MCs compared to ACs (Table 2).
Co
22
23
24
25 Baseline characteristics after matching; (Table 3Supplementary Appendix E)
26
nf
27 There were 26,257 MC cases and 276,760 AC cases (98%) matched to controls (Table
28
29 3). Covariates were balanced between cases and controls, the exception being that MC cases
id
30
31 were more likely to reside in rural areas than controls. Baseline healthcare utilization as
32
en
33 measured by costs in the year prior was similar between cases and controls, also indicating
34
35 groups were comparable (mean (SD) prior year costs for MC-cases and controls were 1779.76
tia
36
37 (7003.45) and 1664.34 (8732.5), standardized difference = 0.02; AC-cases and controls 2385.69
38
39 (8223.62) and 2314.57 (9420.75) standardized difference = 0.01).
l
40
41 The mean age of individuals injured in MCs was younger (mean 36.08) than those
42
43 injured in ACs (mean 38.53) (p < 0.0001). A greater proportion of MC patients were also male
44
45 (81% male) in comparison to AC patients (43% male) (p < 0.0001). Other baseline characteristics
46
47 and average costs amongst cases and controls after matching are presented in Supplementary
48
49 Appendix ETable 3.
50
51
52
53
54
55
56
57
58
59
60
Page 57 of 64
1
2
3
4
5
7
8
9 Patient-level healthcare costs; (Table 34, Supplementary Appendix DFigure 2)
10
11 Average (mean) MC- and AC- attributable costs for medical treatment within 2 years of
12
13 a crash were $5,825 and $2,995, respectively (p < 0.0001). Acute care hospital costs accounted
14
15 for over 75% of this amount for both MCs and ACs (Table 34). Furthermore, most costs were
16
17 accrued within one year of the injury date (Figure 2). Attributable costs declined thereafter for
18
19 both MCs and ACs, remaining stable and similar up to 5 years following the injury date
20
21 (Supplementary Appendix DFigure 2).
Co
22
23
24
25
26
nf
27 INTERPRETATION
28
29 Formatted: Level 1
id
30
31 Principal Main findings
32
en
33 In this population-based study of patients in the largest Canadian province, we

34
35 measured the cost of health care provided to individuals who were injured in a MC or AC and
tia
36
37 required treatment at hospital. There were 26,831 Ontarians during the study period injured
38
39 from MCs and 281,826 from ACs. Average (mean) attributable medical costs within 2 years
l
40
41 were $5,825 and $2,995, respectively (p < 0.0001). Since the annual number of injured persons
42
43 per registered vehicle was 3 times greatertriple and costs per injury were 2 times
44
45 greaterdouble, we estimate the total cost to the universal healthcare system of injuries
46
47 sustained in motor vehicles is approximately 6 times greater per registered motorcycle
48
49 compared to automobile.
50
51
52
53
54
55
56
57
58
59
60
Page 58 of 64
1
2
3
4
5
7
8
9 Since acute care accounted for over 75% of the costs, sSevere injuries (AIS >=3) that
10
11 were also 10 times greater for motorcycles compared to automobiles (IRR=10.4, 95% CI=8.3-
12
13 13.1, p<0.0001) and those injured in MCs were also more likely to require hospital and ICU
14
15 admission (11.93% versus 4.24%, p < 0.0001 and 2.31% versus 1.09%, p < 0.0001, respectively).
16
17 Since acute care accounted for over 75% of the costs after MCs and ACs, these severe injuries
18
19 that required admission to hospital and the ICU likely accounted for the additional costs
20
21 incurred by MCs. All rates in our study could also be multiplied by 5 and expressed per
Co
22
23 kilometer travelled since the average automobile travels 5 times the distance of the average
24
25 motorcycle (i.e. 15x injuries, 50x severe injuries, 25x deaths, 30x costs per kilometer
26
nf
27 travelled).(24)
28
29
id
30
31 Limitations
32
en
33 Although our methodology captured the vast majority (an estimated 92%) of publicly
34
35 funded health care costs provided to Ontarians injured in MCs and ACs,(18) the most important
tia
36
37 limitation of our cost calculations was our inability to capture on-going care costs of patients
38
39 requiring outpatient rehabilitation, such as those suffering acquired brain injury (ABI).
l
40
41 Accordingly, our cost calculations were less comprehensive over time as patients, and their
42
43 care, are moved from the hospital setting and into the community. Attributable costs at 2
44
45 years, rather than over the longer-term, comprised our primary outcome for this reason.
46
47 However, Ssince MC victims are significantly more likely to suffer head trauma,(8, 35, 36) and
48
49 severe injuries in general, this limitation only underestimated MC-attributable costs compared
50
51 to those incurred by ACs . Indeed, for several anatomical regions assessed in our study,
52
53
54
55
56
57
58
59
60
Page 59 of 64
1
2
3
4
5
7
8
9 including head trauma, severe injuries (AIS >=3) were more common among patients injured in
10
11 MCs compared to ACs (Table 2). Similarly,
12
13 Our results also indicated drug costs were negligible (Table 4) because we could only
14
15 record prescriptions dispensed to patients with Ontario Drug Benefit coverage (i.e. those >65
16
17 years of age). Drug costs have been shown to account for a small portion (<15%) of health
18
19 expenditures in Ontario however.(37) Care provided to tourists (non-Ontario residents) injured
20
21 during MVCs also could not be captured. Importantly, we did not investigate ddirect costs
Co
22
23 borne by private insurance as well as indirect or opportunity costs incurred by individual
24
25 patients and society also were not captured. Indirect costs from missed work days and
26
nf
27 permanent disability precluding the ability to earn, for example, may also be greater after MCs
28
29 than ACs since MC patients were younger and more severely injured.
id
30
31 We identified patient encounters with health care providers. As a result, Oour data also
32
en
33 does not include those injured and/or killed at the scene and not transported to hospital
34
35 because we identified patient encounters with health care providers. Since we only identified
tia
36
37 each patients first event, it is also important to point out that injury rates reported in this study
38
39 do not include subsequent crashes, if a patient had more than one.
l
40
41
42
43 Implications
44
45 We found injuries related to motorcycle crashes injuries cost the public healthcare
46
47 system in Ontario 6 times more than automobile injuries, per vehicle. Expressed in other terms,
48
49 Ontarios healthcare payer (MOHLTC) would expect to save $13.5 million dollars per year if
50
51 collision rates remained unchanged and the Provinces registered motorcycles were replaced
52
53
54
55
56
57
58
59
60
Page 60 of 64
1
2
3
4
5
7
8
9 with automobiles. We argue this estimate is conservative for several reasons previously
10
11 mentioned (see Limitations).
12
13 Despite publically available data indicating that the risk associated with driving a
14
15 motorcycle is much greater than that associated with driving an automobile, this knowledge
16
17 has not translated to improvements in motorcycle safetymortality and morbidity related to
18
19 motorcycle crashes has remained stable.(2, 38) In other words, knowledge of the additional risk
20
21 associated with driving a motorcycle does not appear to have led to improvements in
Co
22
23 motorcycle safety. We conducted this study because an understanding of medical costs may
24
25 play a novel role in public health strategy aimed at improving motorcycle safety. Medical costs
26
nf
27 may provide a novel financial incentive to mitigate the risk associated withof MCs.;
28
29 Aanecdotally, insurance companies raising their rates in response to private medical costs have
id
30
31 already acted as a deterrent to motorcycle use in our jurisdiction.(39) However, it was
32
en
33 important to examine public health care costs in this study because private medical costs borne
34
35 by individual patients comprise a very small portion of the total direct healthcare expenditure
tia
36
37 in our Provinces health care system. HMoreover, health care costs incurred by the Ontarios
38
39 healthcare payer the Ministry of Health and Long-Term Care (MOHLTC) in treating MVC
l
40
41 (including MC) patients may be recovered, at least in part, by the Government of Ontario under
42
43 the Provinces Insurance Act. OIn this way, our study is the first to accurately estimate costs
44
45 potentially recoverable from motorcycle insurers under this legislation.
46
47 Medical costs, potentially borne by insurance companies under legislation, and
48
49 transferred to insured motorcyclists can shift risk for additional costs to motorcyclists. Indeed,
50
51 Tthe higher costs of injury and greater incidence of severe injury and death we have observed
52
53
54
55
56
57
58
59
60
Page 61 of 64
1
2
3
4
5
7
8
9 may also warrant other improved public health prevention measures. For example, For
10
11 example, ssupplementary private healthcare insurance, an excise tax, or another novel method
12
13 of health coverage for motorcyclists may be justified with the goal of improving motorcycle
14
15 safety.
16
17 Our study was conducted at the level of a healthcare system serving a population of at
18
19 approximately 13.6 million. Although exact health care costs vary in other healthcare systems,
20
21 we argue the conclusions drawn from the relative comparison of MCs to ACs apply beyond
Co
22
23 Canada to the rest developed world. For example, in a privately funded health care system,
24
25 insurance companies and individual providers may accept a larger share of the direct health
26
nf
27 care costs we have estimated in this study. In our Province, we have captured all hospital and
28
29 physician costs provided in the Provinces publicly funded health care system.
id
30
31
32
en
33 Final Conclusions and future directionsConclusions

34
35 We found each registered motorcycle incurs 3x the injuries, 6x the medical costs, and 5x
tia
36
37 the deaths of each automobile. An understanding of these consequences may play a key role in
38
39 public health strategy aimed at improving motorcycle safety and reducing health and cost
l
40
41 burden. Future work is required to investigate indirect or opportunity costs incurred by MCs.
42
43
44
45
46
47
48
49
50
51 ACKNOWLEDGEMENTS
52
53
54
55
56
57
58
59
60
Page 62 of 64
1
2
3
4
5
7
8
9 This research was supported by grants from the Ontario Ministry of Health and Long-
10
11 Term Care (MOHLTC) and the Ontario SPOR Support Unit to the Health System Performance
12
13 Research Network (HSPRN: fund #06034, recipient WPW), and by the Institute for Clinical
14
15 Evaluative Sciences (ICES), which is also funded by an annual grant from the MOHLTC. The
16
17 funders had no role in study design, data collection and analysis, decision to publish, or
18
19 preparation of the manuscript. Parts of the material are based on data and information
20
21 compiled and provided by the Canadian Institute for Health Information (CIHI). No
Co
22
23 endorsement by ICES, the MOHLTC or CIHI is intended or should be inferred. No benefits have
24
25 been received or will be received from a commercial party related directly or indirectly to the
26
nf
27 subject of this article.

28
29 The authors would also like to acknowledge Allan S. Detsky for providing comments on
id
30
31 an earlier version of the manuscript.
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 63 of 64
1
2
3
4
5
7
8
9 REFERENCES
10
11 1. Sauber-Schatz E, Euerer D, Dellinger A, GT B. Vital Signs: Motor Vehicle Injury Prevention
12 United States and 19 Comparison Countries. MMWR Morb Mortal Wkly Rep 2016.
13 2. Ministry of Transportation. Ontario Road Safety. Annual Report 2013.
15 Economy. 1991;99(4):722-42.
16 4. Gilmore AB, Tavakoly B, Taylor G, Reed H. Understanding tobacco industry pricing
17 strategy and whether it undermines tobacco tax policy: The example of the UK cigarette
20 2000;321(7257):358-61.
Co
nf

28 9. Wodchis W et al. Guidelines on Person-Level Costing Using Administrative Databases in
id
32 11. Nikitovic M, Wodchis WP, Krahn MD, Cadarette SM. Direct health-care costs attributed
en
tia
l
44 journal of the British Diabetic Association. 2016;33(3):395.
46 Accid Anal Prev. 2011;43(6):2128-33.
48 crash. CMAJ. 2014;186(10):742-50.
51 2016.
52
53
54
55
56
57
58
59
60
Page 64 of 64
1
2
3
4
5
7
8
10 for Health Information Discharge Abstract Database: A Validation Study. Toronto: Institute for
12 20. Austin PC. Comparing paired vs non-paired statistical methods of analyses when making
13 inferences about absolute risk reductions in propensity-score matched samples. Statistics in
14 medicine. 2011;30(11):1292.
17 22. Austin PC. Optimal caliper widths for propensity-score matching when estimating
19 statistics. 2011;10(2):150.
20 23. Weiner JP, Abrams C. The Johns Hopkins ACG System: Technical Reference Guide
21 Version 10.0. 2011.
Co
26 25. Santaguida PL, Hawker GA, Hudak PL, Glazier R, Mahomed NN, Kreder HJ, et al. Patient
nf
id
31 2006;63(10):2604.
32 27. Kralj B. Measuring Rurality for Purposes of Health Care Planning: An Empirical
en

35 difference-in-differences approach. JAMA. 2014;312(22):2401-2.
tia

39 30. Haas B, Stukel TA, Gomez D, Zagorski B, De Mestral C, Sharma SV, et al. The mortality
l
40 benefit of direct trauma center transport in a regional trauma system: a population-based

41 analysis. J Trauma Acute Care Surg. 2012;72(6):1510-5; discussion 5-7.
42 31. Gomez D, Haas B, de Mestral C, Sharma S, Hsiao M, Zagorski B, et al. Gender-associated
43 differences in access to trauma center care: A population-based analysis. Surgery.
44 2012;152(2):179-85.
45 32. Gomez D, Haas B, de Mestral C, Sharma S, Hsiao M, Zagorski B, et al. Institutional and
46 provider factors impeding access to trauma center care: an analysis of transfer practices in a
47 regional trauma system. J Trauma Acute Care Surg. 2012;73(5):1288-93.
48 33. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates
50 2009;28(25):3083.
51
52
53
54
55
56
57
58
59
60
Page 65 of 64
1
2
3
4
5
7
8
9 34. Normand ST, Landrum MB, Guadagnoli E, Ayanian JZ, Ryan TJ, Cleary PD, et al. Validating
13 Prevention. 2005;37(2):349.
17 CMAJ. 2016;188(3):182-8.
18 38. The National Highway Traffic Safety Administration. TRAFFIC SAFETY FACTS 2015.
20 April 27, 2017.
Co
24
25
26
nf
27
28
29
id
30
31
32
en
33
34
35
tia
36
37
38
39
l
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Motorcycle Crashes in Ontario

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Motorcycle Crashes in Ontario

Uploaded by

Copyright:

Available Formats

Direct Medical Costs of Motorcycle Crashes in Ontario

Manuscript Type: Research - Cohort study (retrospective)

Date Submitted by the Author: n/a

Complete List of Authors: Pincus, Daniel; University of Toronto,

Bai, Yu; Institute for Clinical Evaluative Sciences,

Keywords: Epidemiology, Health Policy, Health Economics, Public Health

More Detailed Keywords: Motorcycles, Traffic Accidents, Direct medical costs

Background. No reliable estimate of costs incurred by motorcycle crashes

automobile crashes (ACs).

Methods. We conducted a population based, matched cohort study of

Interpretation. Considering both the attributable cost and higher rate of

For Peer Review Only

30 including lower speed motorcycles if the Editors deem it prudent.

27 Success of our matching approach in eliminating known confounding was assessed by

40 specified, as long as balance of covariates is achieved.(13) Another measure of relative success

30 MVCs when police reports are not filed.

controls likely would have underestimated MC-attributable costs if anything.

30 excluded in prior studies.

30 with calipers +/- replacement, etc.

30 other model diagnostics done?

30 injuries accounted for?

27 individuals with physcian diagnosed COPD in health administrative databases. Copd.

30 149 College Street, Room 508-A

Phone: 416-946-7957 (w), 647-244-3324 (c)

27 yearly/100,000 registered automobiles; IRR=10.4, 95% CI=8.3-13.1, p<0.0001).

registered automobile. Medical costs may provide an additional incentive to improve

36 (V40-V49, V50-V59, V70-V79), respectively].(19) We excluded non-Ontario residents and those

27 9. Wodchis W et al. Guidelines on Person-Level Costing Using Administrative Databases in

36 of adult traumatic spinal cord injury in ontario. Spinal Cord. 2013;51(1):64.

40 BMC health services research. 2014;14:532.

27 review. Canadian journal of surgeryJournal canadien de chirurgie. 2008;51(6):428.

36 difference-in-differences approach. JAMA. 2014;312(22):2401-2.

40 Canadian Journal of Surgery. 2012;55(1):21.

30 Abdomen 53 (0.20) 198 (0.07) < 0.0001

33 Upper Extremity 396 (1.51) 830 (0.30) < 0.0001

Supplementary Appendix MC costs

Supplementary Appendix MC costs

30 classification and length of stay,(5-7)

36 studying traffic collisions.(15-17)

Supplementary Appendix MC costs

27 V09: Pedestrian injured in other and unspecified transport accidents

V19: Pedal cyclist injured in other and unspecified transport accidents

36 V40-V49: Car occupant injured in transport accident

40 V70-V79: Bus occupant injured in transport accident

Supplementary Appendix MC costs

Supplementary Appendix MC costs

Supplementary Appendix MC costs

27 1 5.30 4.87 0.02 4.21 4.80 0.03

30 4 13.96 11.76 0.07 11.84 12.14 0.01

33 7 6.22 6.17 0.00 5.00 6.07 0.05

40 14 1.84 2.00 0.01 1.85 1.96 0.01

Supplementary Appendix MC costs

27 Group (RPG) Case Mix System. 2010.

30 9. Chen A, Bushmeneva K, Zagorski B, Colantonio A, Parsons D, Wodchis WP. Direct cost

36 of adult traumatic spinal cord injury in ontario. Spinal Cord. 2013;51(1):64.

40 BMC health services research. 2014;14:532.

Supplementary Appendix MC costs

1 STROBE StatementChecklist of items that should be included in reports of cohort studies

number of exposed and unexposed

33 diagnostic criteria, if applicable

1 completing follow-up, and analysed

27 and interactions, and sensitivity analyses

33 sources of potential bias or imprecision. Discuss both (Discussion)

27 Correspondence and address for all authors: