Relative household wealth and non-fatal road crashes: analysis of population-representative data of Kenyan adults

Relative household wealth and non-fatal road crashes: analysis of population-representative data... Abstract Background This study aims to examine potential road crash disparities across relative wealth and location of residence in Kenya by analyzing population-representative Demographic and Health Survey data. Methods Relative wealth was measured by household assets, converted into an index by polychoric principal components analysis. Location and sex-stratified associations between wealth quantiles and crashes were flexibly estimated using fractional polynomial models. Structural equation models were fit to examine whether observed differences may operate through previously identified determinants. Results In rural areas, crashes were least common for both the poorest men (−5.2 percentage points, 95% CI: −7.3 to −3.2) and women (−1.6 percentage points, 95% CI: −2.9 to −0.4). In urban areas, male crashes were lowest (−3.0 percentage points, 95% CI: −5.2 to −0.8) among the wealthiest, while they peaked in the middle of the female wealth distribution (2.0 percentage points, 95% CI: 0.3–3.8). Male differences operate partially though occupational driving and vehicle ownership. Urban female differences operate partially through household vehicle ownership, but differences for rural women were not explained by modeled determinants. Conclusions Relative wealth and road crash have opposite associations in rural and urban areas. Especially in rural areas, it is important to mitigate potential unintended effects of economic development. accidents, public health, transport Introduction Road crashes are estimated to kill ~150 000 people in sub-Saharan Africa (SSA) each year.1 Unlike most other regions, SSA has not experienced marked reductions in road crash injuries over the past 25 years, and the region’s rapid economic growth and urbanization will likely drive increased road crash risk in the coming years.2,3 Nonetheless, road cash in SSA remains understudied.4,5 One particular area with relatively little research is how wealth and poverty affect with road injury risks in SSA. There is some evidence that risk tends to be greatest among those with economic disadvantages, because they are more likely to use less safe transport options and to more time vulnerably exposed to vehicular traffic as pedestrians;6,7 however, this has not been a consistent finding.8,9 One Nigerian study found that higher education—a correlate of wealth—was associated with higher road crash risk.10 Studies of rural–urban disparities—often a partial proxy for wealth—have found differing associations, suggesting context-specific relationships.11–13 Globally, studies often find that increasingly motorized transport results in increased road crashes as developing countries become wealthier.14–16 Simultaneously, cross-country analyses suggest that road crash fatalities increase as countries reach middle-income status, likely because of greater access to motorcycles and the greater risk that motorized transport will strike vulnerable road users (pedestrians, cyclists and motorcyclists), though risk may remain concentrated in populations’ lower income segments.17–19 The enactment of Kenya’s 2010 constitution, which devolved authority to 47 counties, created an opportunity to examine factors associated with road crash in more detail. The 2014 Kenyan Demographic and Health Survey, which includes a module on injury, was increased greatly in size to be able to produce county-specific indicators for key development priorities. This resulted in the largest-ever African population-based survey with road crash information. The Kenyan government’s bivariate results were tantalizing, finding a general trend toward greater road crash frequency among wealthier than the poorest women, but the opposite trend among men. It also found differences by rural and urban residence, with urban women but rural men having higher risk.20 Because location of residence tends to be highly correlated with wealth, these findings may be entangled. This paper aims to extend existing analyses of the 2014 Kenyan DHS survey. First, it seeks to understand the relationship between relative wealth and road crashes with more granularity and stratified by sex and residence. Second, to the extent associations with relative wealth are identified, it seeks to understand the extent to which they operate through well-established causes of road crash risk, such as vehicle ownership and certain occupations. Methods Participants, sampling and survey administration This study makes secondary use of cross-sectional data collected through Kenya’s 2014 Demographic and Health Survey (DHS), which has been fully described previously.20 Its essential features are briefly recapped here. The survey employed a stratified, two-stage cluster sample design. For each of Kenya’s 47 counties, rural and urban strata were developed (except for the entirely urban Nairobi and Mombasa counties). Within each stratum, clusters were selected probability proportionate to size and then 25 households were randomly selected per cluster.20,21 Half the selected households were asked a longer questionnaire that included an unintentional injury module. Enumerators invited all women aged 15–49 years and men aged 15–54 years in selected households to participate. Among households selected for the injury module, 98.8% participated and, within them, 96.2% of eligible women and 90.2% of eligible men participated.20 Because risk factors for youth are likely different, respondents under age 18 were excluded from this analysis. Data were collected via interviewer-administered surveys between May and October 2014. Interviewers received approximately three weeks of classroom and field training. Surveys were written in English and translated to 16 additional languages. Accuracy of translation was assessed during pretesting and altered if necessary. All data were double-entered for quality assurance.20 Data used for this article were downloaded from a public repository maintained by ICF International at http://www.dhsprogram.com/Data/. Measures The main outcome variable was self-reported involvement ‘in a road traffic accident as a driver, passenger, pedestrian or cyclist’ within 12 months preceding the survey. The main predictor variable was relative household wealth, as measured through an asset index. DHS datasets include a standard wealth index score derived via a principal component analysis (PCA) technique developed by Filmer and Pritchett.22–24 However, Kolenikov and Angeles recently identified potential bias from this approach’s handling of discrete variables (such as water source) as a series of mutually exclusive indicator variables.25 This article, therefore, uses two modifications they recommend to the standard DHS approach. First, some asset index items, such as water source and toilet type, are recategorized as ordinal variables. Second, polychoric correlations, which correctly handle dichotomous and ordinal variables, were used for the PCA. The wealth index score was calculated as the score corresponding to the first principal component. This score was converted to three separate ventile (1/20th quantile) rankings: one for the overall Kenyan population and separate rankings for rural and urban areas. The residence-specific rankings were used to maximize the ability to distinguish between ranks in each area. See the Methods appendix provided in Supplement 1 for more detail on the wealth index construction and quality assurance. Additional variables were included in standard DHS form as described previously.26,27 Respondents’ location of residence was considered a potential moderator of associations between wealth and road crash risk and dichotomized as urban or rural, with urban defined by the Kenyan government as settlements with at least 2000 residents.28,29 Variables through which associations between wealth and crash risk were expected to operate included self-reported alcohol use (dichotomized as reporting any alcohol use in the last 14 days or not);30–33 household bicycle, motorbike, automobile ownership (dichotomous);32,34 and occupations hypothesized to alter exposure to road crash risk. Occupations were classified according to the International Standard Classification of Occupations,35 and dichotomous variables were included for drivers7,36–40 and sales and services elementary occupations, for which a large fraction are street and informal vendors frequently exposed to crash risks.41–45 Occupational drivers were almost exclusively men and therefore not included in female analyses. Unemployment, which was hypothesized to reduce transport for work, was also included.46 Age, which may correlate with wealth and with a variety of risk-taking behaviors,8,20 was included as a categorical control variable in 5-year increments. Statistical methods Descriptive analyses were conducted to characterize the relationship between relative wealth and reported road crashes, stratified by gender and rural/urban location. With no strong a priori expectation that associations would take a particular form, the flexible fractional polynomial approach recommended by Royston and colleagues was followed.47,48 Logistic regression models with all possible first and second-degree fractional polynomials with power transformations from −2 to 3 were fit and the best fitting, parsimonious model was chosen using an adaptation of Bayesian Information Criterion for survey data.49,50 Subsequently, modeled probabilities of road crash and confidence intervals were calculated and plotted. The Methods appendix in Supplement 1 has more details. The resulting plots were then examined, and portions of the wealth distribution were identified that had markedly different crash rates than the rest. For males, these were the lowest four rural ventiles and highest three urban ventiles. For females, it was the lowest three rural ventiles and the highest and lowest four urban ventiles. Indicator variables were constructed for these portions of the wealth distribution and significant difference was determined using a sampling design-corrected test for difference in proportions. If found to be significant, structural equation models were fit to determine whether associations between road crash and wealth were attributable to differences in factors expected a priori to increase crash risk. In basic models, the wealth category was modeled to predict road crash risk as well as vehicle ownership, occupation and alcohol consumption, each of which was also modeled to predict road crash risk. In adjusted models, age was also included as a control variable. Relationships between variables were modeled with the assumption they are additive on a probability scale. All analyses incorporated sampling weights and used Taylor linearization to adjust standard errors for clustering. Analyses used Stata 15.0. Statistical code to fully replicate the paper is provided as Supplement 2. Sensitivity analyses Two analyses were conducted to check robustness to modeling assumptions. In the first, the analysis is replicated using the ventiles of the original DHS wealth index. In the second, the structural equation models’ variance estimation was based on 10 000 bootstrap replications. Both are described in more detail in Supplement 1. Because this study made secondary use of anonymized data, Georgetown University’s institutional review board did not require review. Results A total of 24 011 respondents met inclusion criteria and completed the survey version that included injury items. Of these, 12 940 were female and 14 602 lived in rural areas. Participant characteristics are provided in Table 1. Table 1 Participant characteristics Urban (n = 9409) Rural (n = 16 402) Total (n = 24 011) n Weighted % (95% CI) n Weighted % (95% CI) n Weighted % (95% CI) Sex  Male 4429 48.6 (47.1–50.1) 6642 45.0 (44.2–45.8) 11 071 46.6 (45.8–47.4)  Female 4980 51.4 (49.9–52.9) 7960 55.0 (54.2–55.8) 12 940 53.4 (52.6–54.2) Wealth index score, mean (CI) 9409 −0.2 (−0.3 to −0.1) 16 402 1.7 (1.5–1.8) 24 011 0.6 (0.5–0.7) Vehicle ownership  Automobile 708 7.6 (6.1–9.3) 421 3.4 (2.8–4.0) 1129 5.2 (4.5–6.0)  Motorcycle 762 6.3 (5.4–7.2) 1468 10.3 (9.3–11.4) 2230 8.5 (7.9–9.3)  Bicycle 1785 17.9 (16.1–19.9) 3824 28.5 (26.8–30.3) 5609 23.8 (22.6–25.2) Drinks alcohol 1732 22.0 (20.7–23.4) 2261 16.0 (15.2–16.9) 3993 18.7 (17.9–19.4) Relevant occupations  Driver 390 4.4 (3.9–5.1) 334 2.5 (2.2–2.9) 724 3.3 (3.0–3.7)  Sales and Services Elem. Occ. 2127 25.8 (23.9–27.8) 2106 15.0 (13.9–16.2) 4233 19.7 (18.7–20.8)  Not working 1935 17.6 (16.4–19.0) 3222 19.4 (18.4–20.4) 5157 18.6 (17.8–19.5) Age  25 and under 3091 33.8 (32.0–35.5) 4567 32.0 (30.9–33.0) 7658 32.8 (31.8–33.7)  26–30 2129 23.6 (22.2–25.1) 2716 18.2 (17.4–19.0) 4845 20.6 (19.8–21.4)  31–35 1473 16.5 (15.3–17.8) 2083 14.3 (13.5–15.1) 3556 15.3 (14.6–16.0)  36–40 1174 11.8 (10.8–12.9) 2094 13.7 (13.1–14.4) 3268 12.9 (12.3–13.5)  40–45 801 8.0 (7.2–8.8) 1585 11.1 (10.5–11.8) 2386 9.8 (9.3–10.2)  46 and over 741 6.3 (5.6–7.2) 1557 10.6 (10.0–11.4) 2298 8.8 (8.2–9.3) Urban (n = 9409) Rural (n = 16 402) Total (n = 24 011) n Weighted % (95% CI) n Weighted % (95% CI) n Weighted % (95% CI) Sex  Male 4429 48.6 (47.1–50.1) 6642 45.0 (44.2–45.8) 11 071 46.6 (45.8–47.4)  Female 4980 51.4 (49.9–52.9) 7960 55.0 (54.2–55.8) 12 940 53.4 (52.6–54.2) Wealth index score, mean (CI) 9409 −0.2 (−0.3 to −0.1) 16 402 1.7 (1.5–1.8) 24 011 0.6 (0.5–0.7) Vehicle ownership  Automobile 708 7.6 (6.1–9.3) 421 3.4 (2.8–4.0) 1129 5.2 (4.5–6.0)  Motorcycle 762 6.3 (5.4–7.2) 1468 10.3 (9.3–11.4) 2230 8.5 (7.9–9.3)  Bicycle 1785 17.9 (16.1–19.9) 3824 28.5 (26.8–30.3) 5609 23.8 (22.6–25.2) Drinks alcohol 1732 22.0 (20.7–23.4) 2261 16.0 (15.2–16.9) 3993 18.7 (17.9–19.4) Relevant occupations  Driver 390 4.4 (3.9–5.1) 334 2.5 (2.2–2.9) 724 3.3 (3.0–3.7)  Sales and Services Elem. Occ. 2127 25.8 (23.9–27.8) 2106 15.0 (13.9–16.2) 4233 19.7 (18.7–20.8)  Not working 1935 17.6 (16.4–19.0) 3222 19.4 (18.4–20.4) 5157 18.6 (17.8–19.5) Age  25 and under 3091 33.8 (32.0–35.5) 4567 32.0 (30.9–33.0) 7658 32.8 (31.8–33.7)  26–30 2129 23.6 (22.2–25.1) 2716 18.2 (17.4–19.0) 4845 20.6 (19.8–21.4)  31–35 1473 16.5 (15.3–17.8) 2083 14.3 (13.5–15.1) 3556 15.3 (14.6–16.0)  36–40 1174 11.8 (10.8–12.9) 2094 13.7 (13.1–14.4) 3268 12.9 (12.3–13.5)  40–45 801 8.0 (7.2–8.8) 1585 11.1 (10.5–11.8) 2386 9.8 (9.3–10.2)  46 and over 741 6.3 (5.6–7.2) 1557 10.6 (10.0–11.4) 2298 8.8 (8.2–9.3) N.B. Sample sizes (n) are unweighted observations, while proportions are weighted. Wealth index score is the component score for the first principal component calculated via polychoric PCA. Table 1 Participant characteristics Urban (n = 9409) Rural (n = 16 402) Total (n = 24 011) n Weighted % (95% CI) n Weighted % (95% CI) n Weighted % (95% CI) Sex  Male 4429 48.6 (47.1–50.1) 6642 45.0 (44.2–45.8) 11 071 46.6 (45.8–47.4)  Female 4980 51.4 (49.9–52.9) 7960 55.0 (54.2–55.8) 12 940 53.4 (52.6–54.2) Wealth index score, mean (CI) 9409 −0.2 (−0.3 to −0.1) 16 402 1.7 (1.5–1.8) 24 011 0.6 (0.5–0.7) Vehicle ownership  Automobile 708 7.6 (6.1–9.3) 421 3.4 (2.8–4.0) 1129 5.2 (4.5–6.0)  Motorcycle 762 6.3 (5.4–7.2) 1468 10.3 (9.3–11.4) 2230 8.5 (7.9–9.3)  Bicycle 1785 17.9 (16.1–19.9) 3824 28.5 (26.8–30.3) 5609 23.8 (22.6–25.2) Drinks alcohol 1732 22.0 (20.7–23.4) 2261 16.0 (15.2–16.9) 3993 18.7 (17.9–19.4) Relevant occupations  Driver 390 4.4 (3.9–5.1) 334 2.5 (2.2–2.9) 724 3.3 (3.0–3.7)  Sales and Services Elem. Occ. 2127 25.8 (23.9–27.8) 2106 15.0 (13.9–16.2) 4233 19.7 (18.7–20.8)  Not working 1935 17.6 (16.4–19.0) 3222 19.4 (18.4–20.4) 5157 18.6 (17.8–19.5) Age  25 and under 3091 33.8 (32.0–35.5) 4567 32.0 (30.9–33.0) 7658 32.8 (31.8–33.7)  26–30 2129 23.6 (22.2–25.1) 2716 18.2 (17.4–19.0) 4845 20.6 (19.8–21.4)  31–35 1473 16.5 (15.3–17.8) 2083 14.3 (13.5–15.1) 3556 15.3 (14.6–16.0)  36–40 1174 11.8 (10.8–12.9) 2094 13.7 (13.1–14.4) 3268 12.9 (12.3–13.5)  40–45 801 8.0 (7.2–8.8) 1585 11.1 (10.5–11.8) 2386 9.8 (9.3–10.2)  46 and over 741 6.3 (5.6–7.2) 1557 10.6 (10.0–11.4) 2298 8.8 (8.2–9.3) Urban (n = 9409) Rural (n = 16 402) Total (n = 24 011) n Weighted % (95% CI) n Weighted % (95% CI) n Weighted % (95% CI) Sex  Male 4429 48.6 (47.1–50.1) 6642 45.0 (44.2–45.8) 11 071 46.6 (45.8–47.4)  Female 4980 51.4 (49.9–52.9) 7960 55.0 (54.2–55.8) 12 940 53.4 (52.6–54.2) Wealth index score, mean (CI) 9409 −0.2 (−0.3 to −0.1) 16 402 1.7 (1.5–1.8) 24 011 0.6 (0.5–0.7) Vehicle ownership  Automobile 708 7.6 (6.1–9.3) 421 3.4 (2.8–4.0) 1129 5.2 (4.5–6.0)  Motorcycle 762 6.3 (5.4–7.2) 1468 10.3 (9.3–11.4) 2230 8.5 (7.9–9.3)  Bicycle 1785 17.9 (16.1–19.9) 3824 28.5 (26.8–30.3) 5609 23.8 (22.6–25.2) Drinks alcohol 1732 22.0 (20.7–23.4) 2261 16.0 (15.2–16.9) 3993 18.7 (17.9–19.4) Relevant occupations  Driver 390 4.4 (3.9–5.1) 334 2.5 (2.2–2.9) 724 3.3 (3.0–3.7)  Sales and Services Elem. Occ. 2127 25.8 (23.9–27.8) 2106 15.0 (13.9–16.2) 4233 19.7 (18.7–20.8)  Not working 1935 17.6 (16.4–19.0) 3222 19.4 (18.4–20.4) 5157 18.6 (17.8–19.5) Age  25 and under 3091 33.8 (32.0–35.5) 4567 32.0 (30.9–33.0) 7658 32.8 (31.8–33.7)  26–30 2129 23.6 (22.2–25.1) 2716 18.2 (17.4–19.0) 4845 20.6 (19.8–21.4)  31–35 1473 16.5 (15.3–17.8) 2083 14.3 (13.5–15.1) 3556 15.3 (14.6–16.0)  36–40 1174 11.8 (10.8–12.9) 2094 13.7 (13.1–14.4) 3268 12.9 (12.3–13.5)  40–45 801 8.0 (7.2–8.8) 1585 11.1 (10.5–11.8) 2386 9.8 (9.3–10.2)  46 and over 741 6.3 (5.6–7.2) 1557 10.6 (10.0–11.4) 2298 8.8 (8.2–9.3) N.B. Sample sizes (n) are unweighted observations, while proportions are weighted. Wealth index score is the component score for the first principal component calculated via polychoric PCA. The association between relative household wealth and road crash incidence is shown in Fig. 1. When rural and urban areas are combined, associations had an inverted-U shape for both men and women. However, when disaggregated, rural and urban areas had markedly different associations. In rural areas, crashes were least common among the poorest respondents for both men (5.2 percentage points lower, 95% CI: −7.3 to −3.2) and women (1.6 percentage points lower, 95% CI: −2.9 to −0.4). In urban areas, risk was lowest among the wealthiest men (3.0 percentage points lower, 95% CI: −5.2 to −0.8), whereas it peaked for women in the middle of the wealth distribution (2.0 percentage points, 95% CI: 0.3–3.8). Fig. 1 View largeDownload slide One-year cumulative road crash incidence, Kenyan adults. Fig. 1 View largeDownload slide One-year cumulative road crash incidence, Kenyan adults. Across all populations, results were similar whether adjusted for age or not (Tables 2 and 3). In age-adjusted models, the poorest rural men were 5.4 percentage points (95% CI: −7.5 to −3.3) less likely to report a crash in the last year. Of this, 1.0 percentage point (95% CI: −1.6 to −0.3) of the reduction was explainable through other variables. Poorer men were less likely to be occupational drivers (−3.5 percentage points, 95% CI: −5.2 to −1.8), and drivers were 7.0 (95% CI: 1.9–12.2) percentage points more likely to be in a crash. They were also 10.9 (95% CI: −12.5 to −9.2) percentage points less likely to own a motorcycle, and motorcycle owners were 8.2 (95% CI: 4.7–11.6) percentage points more likely to be in a crash. Table 2 Structural equation model estimates for men Urban males Rural males Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Direct association with road crash  Wealth   High −3.3 (−6.0 to −0.6) 0.017 −3.2 (−5.8 to −0.6) 0.017   Low −4.3 (−6.5 to −2.2) <0.001 −4.5 (−6.6 to −2.3) <0.001  Alcohol use 1.7 (−0.6 to 4.0) 0.138 1.6 (−0.7 to 3.9) 0.161 2.6 (0.7–4.6) 0.009 3.5 (1.4–5.5) 0.001  Occupation   Driver 9.8 (4.7–14.9) <0.001 9.9 (4.8–15.0) <0.001 7.8 (2.5–13.0) 0.004 7.0 (1.9–12.2) 0.008   Sales and Services Elem. Occ. −2.3 (−4.9 to 0.3) 0.088 −2.2 (−4.8 to 0.4) 0.094 0.5 (−1.9 to 3.0) 0.659 0.2 (−2.3 to 2.6) 0.903   Not working −3.5 (−7.0 to −0.1) 0.017 −4.0 (−7.4 to −0.7) 0.018 −1.5 (−4.1 to 1.1) 0.250 −3.5 (−6.5 to −0.6) 0.019  Household owns   Automobile −0.2 (−4.8 to 4.5) 0.938 0.2 (−4.5 to 4.9) 0.929 −3.2 (−8.0 to 1.5) 0.182 −2.8 (−7.5 to 1.9) 0.244   Motorcycle 8.3 (3.4–13.3) 0.001 8.3 (3.3–13.2) 0.001 8.3 (4.8–11.8) <0.001 8.2 (4.7–11.6) <0.001   Bicycle 3.8 (−0.5 to 8.1) 0.086 4.0 (−0.2 to 8.2) 0.059 −0.8 (−2.7 to 1.2) 0.451 −0.6 (−2.5 to 1.4) 0.567 Direct association of wealth with risk factors  Alcohol use 11.2 (3.6–18.8) 0.004 10.3 (2.6–17.9) 0.009 0.4 (−4.2 to 5.1) 0.851 −0.4 (−5.0 to 4.2) 0.863  Occupation   Driver −2.8 (−6.5 to 0.8) 0.129 −3.2 (−7.0 to 0.6) 0.099 −3.2 (−4.8 to −1.6) <0.001 −3.5 (−5.2 to −1.8) <0.001   Sales and Services Elem. Occ. −6.6 (−11.7 to −1.6) 0.010 −7.1 (−12.1 to −2.1) 0.006 3.2 (−0.7 to 7.0) 0.111 2.8 (−1.0 to 6.7) 0.149   Not working 2.2 (−1.7 to 6.0) 0.276 4.4 (1.0–7.7) 0.011 0.1 (−2.7 to 3.0) 0.920 1.6 (−1.0 to 4.3) 0.225  Household owns   Automobile 28.1 (22.2–34.0) <0.001 28.0 (22.1–33.9) <0.001 −4.0 (−4.8 to −3.2) <0.001 −4.0 (−4.8 to −3.2) <0.001   Motorcycle 0.4 (−2.4 to 3.1) 0.786 0.4 (−2.4 to 3.1) 0.799 −10.6 (−12.3 to −8.9) <0.001 −10.9 (−12.5 to −9.2) <0.001   Bicycle 10.1 (4.1–16.0) 0.001 9.4 (3.5–15.3) 0.002 −24.2 (−27.3 to −21.1) <0.001 −23.8 (−27.0 to −20.7) <0.001 Total indirect association between wealth and crashes 0.4 (−1.0 to 1.7) 0.615 0.3 (−1.1 to 1.7) 0.677 −0.8 (−1.4 to −0.1) 0.017 −1.0 (−1.6 to −0.3) 0.005 Total difference in crashes by wealth −2.9 (−5.1 to −0.7) 0.010 −2.9 (−5.0 to −0.7) 0.009 −5.1 (−7.2 to −3.0) <0.001 −5.4 (−7.5 to −3.3) <0.001 Urban males Rural males Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Direct association with road crash  Wealth   High −3.3 (−6.0 to −0.6) 0.017 −3.2 (−5.8 to −0.6) 0.017   Low −4.3 (−6.5 to −2.2) <0.001 −4.5 (−6.6 to −2.3) <0.001  Alcohol use 1.7 (−0.6 to 4.0) 0.138 1.6 (−0.7 to 3.9) 0.161 2.6 (0.7–4.6) 0.009 3.5 (1.4–5.5) 0.001  Occupation   Driver 9.8 (4.7–14.9) <0.001 9.9 (4.8–15.0) <0.001 7.8 (2.5–13.0) 0.004 7.0 (1.9–12.2) 0.008   Sales and Services Elem. Occ. −2.3 (−4.9 to 0.3) 0.088 −2.2 (−4.8 to 0.4) 0.094 0.5 (−1.9 to 3.0) 0.659 0.2 (−2.3 to 2.6) 0.903   Not working −3.5 (−7.0 to −0.1) 0.017 −4.0 (−7.4 to −0.7) 0.018 −1.5 (−4.1 to 1.1) 0.250 −3.5 (−6.5 to −0.6) 0.019  Household owns   Automobile −0.2 (−4.8 to 4.5) 0.938 0.2 (−4.5 to 4.9) 0.929 −3.2 (−8.0 to 1.5) 0.182 −2.8 (−7.5 to 1.9) 0.244   Motorcycle 8.3 (3.4–13.3) 0.001 8.3 (3.3–13.2) 0.001 8.3 (4.8–11.8) <0.001 8.2 (4.7–11.6) <0.001   Bicycle 3.8 (−0.5 to 8.1) 0.086 4.0 (−0.2 to 8.2) 0.059 −0.8 (−2.7 to 1.2) 0.451 −0.6 (−2.5 to 1.4) 0.567 Direct association of wealth with risk factors  Alcohol use 11.2 (3.6–18.8) 0.004 10.3 (2.6–17.9) 0.009 0.4 (−4.2 to 5.1) 0.851 −0.4 (−5.0 to 4.2) 0.863  Occupation   Driver −2.8 (−6.5 to 0.8) 0.129 −3.2 (−7.0 to 0.6) 0.099 −3.2 (−4.8 to −1.6) <0.001 −3.5 (−5.2 to −1.8) <0.001   Sales and Services Elem. Occ. −6.6 (−11.7 to −1.6) 0.010 −7.1 (−12.1 to −2.1) 0.006 3.2 (−0.7 to 7.0) 0.111 2.8 (−1.0 to 6.7) 0.149   Not working 2.2 (−1.7 to 6.0) 0.276 4.4 (1.0–7.7) 0.011 0.1 (−2.7 to 3.0) 0.920 1.6 (−1.0 to 4.3) 0.225  Household owns   Automobile 28.1 (22.2–34.0) <0.001 28.0 (22.1–33.9) <0.001 −4.0 (−4.8 to −3.2) <0.001 −4.0 (−4.8 to −3.2) <0.001   Motorcycle 0.4 (−2.4 to 3.1) 0.786 0.4 (−2.4 to 3.1) 0.799 −10.6 (−12.3 to −8.9) <0.001 −10.9 (−12.5 to −9.2) <0.001   Bicycle 10.1 (4.1–16.0) 0.001 9.4 (3.5–15.3) 0.002 −24.2 (−27.3 to −21.1) <0.001 −23.8 (−27.0 to −20.7) <0.001 Total indirect association between wealth and crashes 0.4 (−1.0 to 1.7) 0.615 0.3 (−1.1 to 1.7) 0.677 −0.8 (−1.4 to −0.1) 0.017 −1.0 (−1.6 to −0.3) 0.005 Total difference in crashes by wealth −2.9 (−5.1 to −0.7) 0.010 −2.9 (−5.0 to −0.7) 0.009 −5.1 (−7.2 to −3.0) <0.001 −5.4 (−7.5 to −3.3) <0.001 N.B. Direct association with road crash represents the change in road crash likelihood, controlling for the other factors. Direct association of wealth with risk factors represents the change in each risk factor associated with the difference in wealth. The total indirect association is the total difference in road crash likelihood comparing wealth levels that operates through the change in risk factors. The total difference in crashes by wealth is the overall difference in crash likelihood comparing wealth levels. Adjusted models also control for age. The difference in crash risk by wealth category varies inconsequentially from the values in the text because of listwise deletion of observations with missing data in the SEMs. Table 2 Structural equation model estimates for men Urban males Rural males Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Direct association with road crash  Wealth   High −3.3 (−6.0 to −0.6) 0.017 −3.2 (−5.8 to −0.6) 0.017   Low −4.3 (−6.5 to −2.2) <0.001 −4.5 (−6.6 to −2.3) <0.001  Alcohol use 1.7 (−0.6 to 4.0) 0.138 1.6 (−0.7 to 3.9) 0.161 2.6 (0.7–4.6) 0.009 3.5 (1.4–5.5) 0.001  Occupation   Driver 9.8 (4.7–14.9) <0.001 9.9 (4.8–15.0) <0.001 7.8 (2.5–13.0) 0.004 7.0 (1.9–12.2) 0.008   Sales and Services Elem. Occ. −2.3 (−4.9 to 0.3) 0.088 −2.2 (−4.8 to 0.4) 0.094 0.5 (−1.9 to 3.0) 0.659 0.2 (−2.3 to 2.6) 0.903   Not working −3.5 (−7.0 to −0.1) 0.017 −4.0 (−7.4 to −0.7) 0.018 −1.5 (−4.1 to 1.1) 0.250 −3.5 (−6.5 to −0.6) 0.019  Household owns   Automobile −0.2 (−4.8 to 4.5) 0.938 0.2 (−4.5 to 4.9) 0.929 −3.2 (−8.0 to 1.5) 0.182 −2.8 (−7.5 to 1.9) 0.244   Motorcycle 8.3 (3.4–13.3) 0.001 8.3 (3.3–13.2) 0.001 8.3 (4.8–11.8) <0.001 8.2 (4.7–11.6) <0.001   Bicycle 3.8 (−0.5 to 8.1) 0.086 4.0 (−0.2 to 8.2) 0.059 −0.8 (−2.7 to 1.2) 0.451 −0.6 (−2.5 to 1.4) 0.567 Direct association of wealth with risk factors  Alcohol use 11.2 (3.6–18.8) 0.004 10.3 (2.6–17.9) 0.009 0.4 (−4.2 to 5.1) 0.851 −0.4 (−5.0 to 4.2) 0.863  Occupation   Driver −2.8 (−6.5 to 0.8) 0.129 −3.2 (−7.0 to 0.6) 0.099 −3.2 (−4.8 to −1.6) <0.001 −3.5 (−5.2 to −1.8) <0.001   Sales and Services Elem. Occ. −6.6 (−11.7 to −1.6) 0.010 −7.1 (−12.1 to −2.1) 0.006 3.2 (−0.7 to 7.0) 0.111 2.8 (−1.0 to 6.7) 0.149   Not working 2.2 (−1.7 to 6.0) 0.276 4.4 (1.0–7.7) 0.011 0.1 (−2.7 to 3.0) 0.920 1.6 (−1.0 to 4.3) 0.225  Household owns   Automobile 28.1 (22.2–34.0) <0.001 28.0 (22.1–33.9) <0.001 −4.0 (−4.8 to −3.2) <0.001 −4.0 (−4.8 to −3.2) <0.001   Motorcycle 0.4 (−2.4 to 3.1) 0.786 0.4 (−2.4 to 3.1) 0.799 −10.6 (−12.3 to −8.9) <0.001 −10.9 (−12.5 to −9.2) <0.001   Bicycle 10.1 (4.1–16.0) 0.001 9.4 (3.5–15.3) 0.002 −24.2 (−27.3 to −21.1) <0.001 −23.8 (−27.0 to −20.7) <0.001 Total indirect association between wealth and crashes 0.4 (−1.0 to 1.7) 0.615 0.3 (−1.1 to 1.7) 0.677 −0.8 (−1.4 to −0.1) 0.017 −1.0 (−1.6 to −0.3) 0.005 Total difference in crashes by wealth −2.9 (−5.1 to −0.7) 0.010 −2.9 (−5.0 to −0.7) 0.009 −5.1 (−7.2 to −3.0) <0.001 −5.4 (−7.5 to −3.3) <0.001 Urban males Rural males Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Direct association with road crash  Wealth   High −3.3 (−6.0 to −0.6) 0.017 −3.2 (−5.8 to −0.6) 0.017   Low −4.3 (−6.5 to −2.2) <0.001 −4.5 (−6.6 to −2.3) <0.001  Alcohol use 1.7 (−0.6 to 4.0) 0.138 1.6 (−0.7 to 3.9) 0.161 2.6 (0.7–4.6) 0.009 3.5 (1.4–5.5) 0.001  Occupation   Driver 9.8 (4.7–14.9) <0.001 9.9 (4.8–15.0) <0.001 7.8 (2.5–13.0) 0.004 7.0 (1.9–12.2) 0.008   Sales and Services Elem. Occ. −2.3 (−4.9 to 0.3) 0.088 −2.2 (−4.8 to 0.4) 0.094 0.5 (−1.9 to 3.0) 0.659 0.2 (−2.3 to 2.6) 0.903   Not working −3.5 (−7.0 to −0.1) 0.017 −4.0 (−7.4 to −0.7) 0.018 −1.5 (−4.1 to 1.1) 0.250 −3.5 (−6.5 to −0.6) 0.019  Household owns   Automobile −0.2 (−4.8 to 4.5) 0.938 0.2 (−4.5 to 4.9) 0.929 −3.2 (−8.0 to 1.5) 0.182 −2.8 (−7.5 to 1.9) 0.244   Motorcycle 8.3 (3.4–13.3) 0.001 8.3 (3.3–13.2) 0.001 8.3 (4.8–11.8) <0.001 8.2 (4.7–11.6) <0.001   Bicycle 3.8 (−0.5 to 8.1) 0.086 4.0 (−0.2 to 8.2) 0.059 −0.8 (−2.7 to 1.2) 0.451 −0.6 (−2.5 to 1.4) 0.567 Direct association of wealth with risk factors  Alcohol use 11.2 (3.6–18.8) 0.004 10.3 (2.6–17.9) 0.009 0.4 (−4.2 to 5.1) 0.851 −0.4 (−5.0 to 4.2) 0.863  Occupation   Driver −2.8 (−6.5 to 0.8) 0.129 −3.2 (−7.0 to 0.6) 0.099 −3.2 (−4.8 to −1.6) <0.001 −3.5 (−5.2 to −1.8) <0.001   Sales and Services Elem. Occ. −6.6 (−11.7 to −1.6) 0.010 −7.1 (−12.1 to −2.1) 0.006 3.2 (−0.7 to 7.0) 0.111 2.8 (−1.0 to 6.7) 0.149   Not working 2.2 (−1.7 to 6.0) 0.276 4.4 (1.0–7.7) 0.011 0.1 (−2.7 to 3.0) 0.920 1.6 (−1.0 to 4.3) 0.225  Household owns   Automobile 28.1 (22.2–34.0) <0.001 28.0 (22.1–33.9) <0.001 −4.0 (−4.8 to −3.2) <0.001 −4.0 (−4.8 to −3.2) <0.001   Motorcycle 0.4 (−2.4 to 3.1) 0.786 0.4 (−2.4 to 3.1) 0.799 −10.6 (−12.3 to −8.9) <0.001 −10.9 (−12.5 to −9.2) <0.001   Bicycle 10.1 (4.1–16.0) 0.001 9.4 (3.5–15.3) 0.002 −24.2 (−27.3 to −21.1) <0.001 −23.8 (−27.0 to −20.7) <0.001 Total indirect association between wealth and crashes 0.4 (−1.0 to 1.7) 0.615 0.3 (−1.1 to 1.7) 0.677 −0.8 (−1.4 to −0.1) 0.017 −1.0 (−1.6 to −0.3) 0.005 Total difference in crashes by wealth −2.9 (−5.1 to −0.7) 0.010 −2.9 (−5.0 to −0.7) 0.009 −5.1 (−7.2 to −3.0) <0.001 −5.4 (−7.5 to −3.3) <0.001 N.B. Direct association with road crash represents the change in road crash likelihood, controlling for the other factors. Direct association of wealth with risk factors represents the change in each risk factor associated with the difference in wealth. The total indirect association is the total difference in road crash likelihood comparing wealth levels that operates through the change in risk factors. The total difference in crashes by wealth is the overall difference in crash likelihood comparing wealth levels. Adjusted models also control for age. The difference in crash risk by wealth category varies inconsequentially from the values in the text because of listwise deletion of observations with missing data in the SEMs. Table 3 Structural equation model estimates for women Urban females Rural females Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Direct association with road crash  Wealth   High −1.3 (−3.7 to 1.1) 0.297 −1.3 (−3.7 to 1.1) 0.301   Low −2.7 (−4.5 to −0.8) 0.009 −2.7 (−4.5 to −0.9) 0.004 −1.3 (−2.6 to 0.1) 0.061 −1.3 (−2.6 to 0.0) 0.058  Alcohol use 1.1 (−3.0 to 5.2) 0.602 1.0 (−3.1 to 5.0) 0.631 3.4 (0.1–6.8) 0.046 3.3 (−0.1 to 6.6) 0.054  Occupation   Sales and Services, Elem. Occ. −0.2 (−2.3 to 2.0) 0.891 −0.2 (−2.4 to 2.0) 0.884 0.8 (−0.7 to 2.2) 0.289 0.8 (−0.7 to 2.2) 0.285   Not working −1.9 (−4.0 to 0.2) 0.079 −2.1 (−4.4 to 0.1) 0.060 −0.8 (−1.9 to 0.2) 0.122 −0.8 (−2.0 to 0.3) 0.150  Household owns   Automobile −2.7 (−4.8 to −0.7) 0.010 −2.8 (−4.8 to −0.8) 0.006 1.5 (−2.4 to 5.3) 0.451 1.4 (−2.4 to 5.3) 0.468   Motorcycle 1.7 (−2.7 to 6.1) 0.453 1.8 (−2.7 to 6.2) 0.431 1.0 (−0.8 to 2.8) 0.278 1.1 (−0.8 to 2.9) 0.257   Bicycle −0.9 (−2.5 to 0.8) 0.303 −1.0 (−2.6 to 0.7) 0.239 0.6 (−0.6 to 1.8) 0.352 0.5 (−0.7 to 1.7) 0.402 Direct association of high wealth with risk factors  Alcohol use 5.6 (2.0–9.2) 0.002 5.6 (2.0–9.1) 0.002  Occupation   Sales and Services Elem. Occ. −2.0 (−6.9 to 2.8) 0.409 −2.2 (−7.0 to 2.6) 0.369   Not working −5.4 (−10.5 to −0.2) 0.043 −4.7 (−9.7 to 0.3) 0.068  Household Owns:   Automobile 23.9 (18.4–29.5) <0.001 23.8 (18.2–29.4) <0.001   Motorcycle −0.3 (−2.5 to 1.8) 0.765 −0.4 (−2.6 to 1.7) 0.714   Bicycle 8.0 (3.4–12.6) 0.001 7.7 (3.2–12.2) 0.001 Direct association of low wealth with risk factors  Alcohol use 1.3 (−1.8 to 4.3) 0.405 1.3 (−1.7 to 4.3) 0.400 6.1 (3.2–9.1) <0.001 6.2 (3.3–9.2) <0.001  Occupation   Sales and Services, Elem. Occ. −4.4 (−10.0 to 1.1) 0.118 −4.8 (−10.1 to 0.5) 0.073 −1.3 (−4.7 to 2.0) 0.438 −1.0 (−4.4 to 2.3) 0.544   Not working −1.8 (−6.9 to 3.3) 0.479 −0.6 (−5.5 to 4.2) 0.799 35.7 (30.7–40.8) <0.001 34.5 (29.7–39.3) <0.001  Household owns   Automobile −1.2 (−1.6 to −0.7) <0.001 −1.4 (−2.0 to −0.9) <0.001 −3.4 (−4.0 to −2.7) <0.001 −3.3 (−4.0 to −2.7) <0.001   Motorcycle −3.8 (−5.6 to −2.0) <0.001 −3.8 (−5.6 to −2.0) <0.001 −8.2 (−10.1 to −6.3) <0.001 −8.2 (−10.1 to −6.3) <0.001   Bicycle −6.4 (−9.6 to −3.2) <0.001 −7.1 (−10.4 to −3.9) <0.001 −26.2 (−28.3 to −24.0) <0.001 −26.2 (−28.4 to −24.0) <0.001 Total indirect association between wealth and crashes  High −0.6 (−1.2 to 0.0) 0.072 −0.6 (−1.2 to 0.0) 0.053  Low 0.1 (−0.2 to 0.3) 0.544 0.1 (−0.2 to 0.3) 0.543 −0.4 (−1.0 to 0.2) 0.217 −0.4 (−1.0 to 0.3) 0.256 Total difference in crashes by wealth  High −1.8 (−3.8 to 0.2) 0.073 −1.9 (−3.9 to 0.2) 0.075  Low −2.6 (−4.4 to −0.8) 0.005 −2.6 (−4.4 to −0.8) 0.005 −1.6 (−2.9 to −0.4) 0.012 −1.7 (−3.0 to −0.4) 0.011 Urban females Rural females Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Direct association with road crash  Wealth   High −1.3 (−3.7 to 1.1) 0.297 −1.3 (−3.7 to 1.1) 0.301   Low −2.7 (−4.5 to −0.8) 0.009 −2.7 (−4.5 to −0.9) 0.004 −1.3 (−2.6 to 0.1) 0.061 −1.3 (−2.6 to 0.0) 0.058  Alcohol use 1.1 (−3.0 to 5.2) 0.602 1.0 (−3.1 to 5.0) 0.631 3.4 (0.1–6.8) 0.046 3.3 (−0.1 to 6.6) 0.054  Occupation   Sales and Services, Elem. Occ. −0.2 (−2.3 to 2.0) 0.891 −0.2 (−2.4 to 2.0) 0.884 0.8 (−0.7 to 2.2) 0.289 0.8 (−0.7 to 2.2) 0.285   Not working −1.9 (−4.0 to 0.2) 0.079 −2.1 (−4.4 to 0.1) 0.060 −0.8 (−1.9 to 0.2) 0.122 −0.8 (−2.0 to 0.3) 0.150  Household owns   Automobile −2.7 (−4.8 to −0.7) 0.010 −2.8 (−4.8 to −0.8) 0.006 1.5 (−2.4 to 5.3) 0.451 1.4 (−2.4 to 5.3) 0.468   Motorcycle 1.7 (−2.7 to 6.1) 0.453 1.8 (−2.7 to 6.2) 0.431 1.0 (−0.8 to 2.8) 0.278 1.1 (−0.8 to 2.9) 0.257   Bicycle −0.9 (−2.5 to 0.8) 0.303 −1.0 (−2.6 to 0.7) 0.239 0.6 (−0.6 to 1.8) 0.352 0.5 (−0.7 to 1.7) 0.402 Direct association of high wealth with risk factors  Alcohol use 5.6 (2.0–9.2) 0.002 5.6 (2.0–9.1) 0.002  Occupation   Sales and Services Elem. Occ. −2.0 (−6.9 to 2.8) 0.409 −2.2 (−7.0 to 2.6) 0.369   Not working −5.4 (−10.5 to −0.2) 0.043 −4.7 (−9.7 to 0.3) 0.068  Household Owns:   Automobile 23.9 (18.4–29.5) <0.001 23.8 (18.2–29.4) <0.001   Motorcycle −0.3 (−2.5 to 1.8) 0.765 −0.4 (−2.6 to 1.7) 0.714   Bicycle 8.0 (3.4–12.6) 0.001 7.7 (3.2–12.2) 0.001 Direct association of low wealth with risk factors  Alcohol use 1.3 (−1.8 to 4.3) 0.405 1.3 (−1.7 to 4.3) 0.400 6.1 (3.2–9.1) <0.001 6.2 (3.3–9.2) <0.001  Occupation   Sales and Services, Elem. Occ. −4.4 (−10.0 to 1.1) 0.118 −4.8 (−10.1 to 0.5) 0.073 −1.3 (−4.7 to 2.0) 0.438 −1.0 (−4.4 to 2.3) 0.544   Not working −1.8 (−6.9 to 3.3) 0.479 −0.6 (−5.5 to 4.2) 0.799 35.7 (30.7–40.8) <0.001 34.5 (29.7–39.3) <0.001  Household owns   Automobile −1.2 (−1.6 to −0.7) <0.001 −1.4 (−2.0 to −0.9) <0.001 −3.4 (−4.0 to −2.7) <0.001 −3.3 (−4.0 to −2.7) <0.001   Motorcycle −3.8 (−5.6 to −2.0) <0.001 −3.8 (−5.6 to −2.0) <0.001 −8.2 (−10.1 to −6.3) <0.001 −8.2 (−10.1 to −6.3) <0.001   Bicycle −6.4 (−9.6 to −3.2) <0.001 −7.1 (−10.4 to −3.9) <0.001 −26.2 (−28.3 to −24.0) <0.001 −26.2 (−28.4 to −24.0) <0.001 Total indirect association between wealth and crashes  High −0.6 (−1.2 to 0.0) 0.072 −0.6 (−1.2 to 0.0) 0.053  Low 0.1 (−0.2 to 0.3) 0.544 0.1 (−0.2 to 0.3) 0.543 −0.4 (−1.0 to 0.2) 0.217 −0.4 (−1.0 to 0.3) 0.256 Total difference in crashes by wealth  High −1.8 (−3.8 to 0.2) 0.073 −1.9 (−3.9 to 0.2) 0.075  Low −2.6 (−4.4 to −0.8) 0.005 −2.6 (−4.4 to −0.8) 0.005 −1.6 (−2.9 to −0.4) 0.012 −1.7 (−3.0 to −0.4) 0.011 N.B. Direct association with road crash represents the change in road crash likelihood, controlling for the other factors. Direct association of wealth with risk factors represents the change in each risk factor associated with the difference in wealth. The total indirect association is the total difference in road crash likelihood comparing wealth levels that operates through the change in risk factors. The total difference in crashes by wealth is the overall difference in crash likelihood comparing wealth levels. Adjusted models also control for age. The difference in crash risk by wealth category varies inconsequentially from the values in the text because of listwise deletion of observations with missing data in the SEMs. Table 3 Structural equation model estimates for women Urban females Rural females Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Direct association with road crash  Wealth   High −1.3 (−3.7 to 1.1) 0.297 −1.3 (−3.7 to 1.1) 0.301   Low −2.7 (−4.5 to −0.8) 0.009 −2.7 (−4.5 to −0.9) 0.004 −1.3 (−2.6 to 0.1) 0.061 −1.3 (−2.6 to 0.0) 0.058  Alcohol use 1.1 (−3.0 to 5.2) 0.602 1.0 (−3.1 to 5.0) 0.631 3.4 (0.1–6.8) 0.046 3.3 (−0.1 to 6.6) 0.054  Occupation   Sales and Services, Elem. Occ. −0.2 (−2.3 to 2.0) 0.891 −0.2 (−2.4 to 2.0) 0.884 0.8 (−0.7 to 2.2) 0.289 0.8 (−0.7 to 2.2) 0.285   Not working −1.9 (−4.0 to 0.2) 0.079 −2.1 (−4.4 to 0.1) 0.060 −0.8 (−1.9 to 0.2) 0.122 −0.8 (−2.0 to 0.3) 0.150  Household owns   Automobile −2.7 (−4.8 to −0.7) 0.010 −2.8 (−4.8 to −0.8) 0.006 1.5 (−2.4 to 5.3) 0.451 1.4 (−2.4 to 5.3) 0.468   Motorcycle 1.7 (−2.7 to 6.1) 0.453 1.8 (−2.7 to 6.2) 0.431 1.0 (−0.8 to 2.8) 0.278 1.1 (−0.8 to 2.9) 0.257   Bicycle −0.9 (−2.5 to 0.8) 0.303 −1.0 (−2.6 to 0.7) 0.239 0.6 (−0.6 to 1.8) 0.352 0.5 (−0.7 to 1.7) 0.402 Direct association of high wealth with risk factors  Alcohol use 5.6 (2.0–9.2) 0.002 5.6 (2.0–9.1) 0.002  Occupation   Sales and Services Elem. Occ. −2.0 (−6.9 to 2.8) 0.409 −2.2 (−7.0 to 2.6) 0.369   Not working −5.4 (−10.5 to −0.2) 0.043 −4.7 (−9.7 to 0.3) 0.068  Household Owns:   Automobile 23.9 (18.4–29.5) <0.001 23.8 (18.2–29.4) <0.001   Motorcycle −0.3 (−2.5 to 1.8) 0.765 −0.4 (−2.6 to 1.7) 0.714   Bicycle 8.0 (3.4–12.6) 0.001 7.7 (3.2–12.2) 0.001 Direct association of low wealth with risk factors  Alcohol use 1.3 (−1.8 to 4.3) 0.405 1.3 (−1.7 to 4.3) 0.400 6.1 (3.2–9.1) <0.001 6.2 (3.3–9.2) <0.001  Occupation   Sales and Services, Elem. Occ. −4.4 (−10.0 to 1.1) 0.118 −4.8 (−10.1 to 0.5) 0.073 −1.3 (−4.7 to 2.0) 0.438 −1.0 (−4.4 to 2.3) 0.544   Not working −1.8 (−6.9 to 3.3) 0.479 −0.6 (−5.5 to 4.2) 0.799 35.7 (30.7–40.8) <0.001 34.5 (29.7–39.3) <0.001  Household owns   Automobile −1.2 (−1.6 to −0.7) <0.001 −1.4 (−2.0 to −0.9) <0.001 −3.4 (−4.0 to −2.7) <0.001 −3.3 (−4.0 to −2.7) <0.001   Motorcycle −3.8 (−5.6 to −2.0) <0.001 −3.8 (−5.6 to −2.0) <0.001 −8.2 (−10.1 to −6.3) <0.001 −8.2 (−10.1 to −6.3) <0.001   Bicycle −6.4 (−9.6 to −3.2) <0.001 −7.1 (−10.4 to −3.9) <0.001 −26.2 (−28.3 to −24.0) <0.001 −26.2 (−28.4 to −24.0) <0.001 Total indirect association between wealth and crashes  High −0.6 (−1.2 to 0.0) 0.072 −0.6 (−1.2 to 0.0) 0.053  Low 0.1 (−0.2 to 0.3) 0.544 0.1 (−0.2 to 0.3) 0.543 −0.4 (−1.0 to 0.2) 0.217 −0.4 (−1.0 to 0.3) 0.256 Total difference in crashes by wealth  High −1.8 (−3.8 to 0.2) 0.073 −1.9 (−3.9 to 0.2) 0.075  Low −2.6 (−4.4 to −0.8) 0.005 −2.6 (−4.4 to −0.8) 0.005 −1.6 (−2.9 to −0.4) 0.012 −1.7 (−3.0 to −0.4) 0.011 Urban females Rural females Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Direct association with road crash  Wealth   High −1.3 (−3.7 to 1.1) 0.297 −1.3 (−3.7 to 1.1) 0.301   Low −2.7 (−4.5 to −0.8) 0.009 −2.7 (−4.5 to −0.9) 0.004 −1.3 (−2.6 to 0.1) 0.061 −1.3 (−2.6 to 0.0) 0.058  Alcohol use 1.1 (−3.0 to 5.2) 0.602 1.0 (−3.1 to 5.0) 0.631 3.4 (0.1–6.8) 0.046 3.3 (−0.1 to 6.6) 0.054  Occupation   Sales and Services, Elem. Occ. −0.2 (−2.3 to 2.0) 0.891 −0.2 (−2.4 to 2.0) 0.884 0.8 (−0.7 to 2.2) 0.289 0.8 (−0.7 to 2.2) 0.285   Not working −1.9 (−4.0 to 0.2) 0.079 −2.1 (−4.4 to 0.1) 0.060 −0.8 (−1.9 to 0.2) 0.122 −0.8 (−2.0 to 0.3) 0.150  Household owns   Automobile −2.7 (−4.8 to −0.7) 0.010 −2.8 (−4.8 to −0.8) 0.006 1.5 (−2.4 to 5.3) 0.451 1.4 (−2.4 to 5.3) 0.468   Motorcycle 1.7 (−2.7 to 6.1) 0.453 1.8 (−2.7 to 6.2) 0.431 1.0 (−0.8 to 2.8) 0.278 1.1 (−0.8 to 2.9) 0.257   Bicycle −0.9 (−2.5 to 0.8) 0.303 −1.0 (−2.6 to 0.7) 0.239 0.6 (−0.6 to 1.8) 0.352 0.5 (−0.7 to 1.7) 0.402 Direct association of high wealth with risk factors  Alcohol use 5.6 (2.0–9.2) 0.002 5.6 (2.0–9.1) 0.002  Occupation   Sales and Services Elem. Occ. −2.0 (−6.9 to 2.8) 0.409 −2.2 (−7.0 to 2.6) 0.369   Not working −5.4 (−10.5 to −0.2) 0.043 −4.7 (−9.7 to 0.3) 0.068  Household Owns:   Automobile 23.9 (18.4–29.5) <0.001 23.8 (18.2–29.4) <0.001   Motorcycle −0.3 (−2.5 to 1.8) 0.765 −0.4 (−2.6 to 1.7) 0.714   Bicycle 8.0 (3.4–12.6) 0.001 7.7 (3.2–12.2) 0.001 Direct association of low wealth with risk factors  Alcohol use 1.3 (−1.8 to 4.3) 0.405 1.3 (−1.7 to 4.3) 0.400 6.1 (3.2–9.1) <0.001 6.2 (3.3–9.2) <0.001  Occupation   Sales and Services, Elem. Occ. −4.4 (−10.0 to 1.1) 0.118 −4.8 (−10.1 to 0.5) 0.073 −1.3 (−4.7 to 2.0) 0.438 −1.0 (−4.4 to 2.3) 0.544   Not working −1.8 (−6.9 to 3.3) 0.479 −0.6 (−5.5 to 4.2) 0.799 35.7 (30.7–40.8) <0.001 34.5 (29.7–39.3) <0.001  Household owns   Automobile −1.2 (−1.6 to −0.7) <0.001 −1.4 (−2.0 to −0.9) <0.001 −3.4 (−4.0 to −2.7) <0.001 −3.3 (−4.0 to −2.7) <0.001   Motorcycle −3.8 (−5.6 to −2.0) <0.001 −3.8 (−5.6 to −2.0) <0.001 −8.2 (−10.1 to −6.3) <0.001 −8.2 (−10.1 to −6.3) <0.001   Bicycle −6.4 (−9.6 to −3.2) <0.001 −7.1 (−10.4 to −3.9) <0.001 −26.2 (−28.3 to −24.0) <0.001 −26.2 (−28.4 to −24.0) <0.001 Total indirect association between wealth and crashes  High −0.6 (−1.2 to 0.0) 0.072 −0.6 (−1.2 to 0.0) 0.053  Low 0.1 (−0.2 to 0.3) 0.544 0.1 (−0.2 to 0.3) 0.543 −0.4 (−1.0 to 0.2) 0.217 −0.4 (−1.0 to 0.3) 0.256 Total difference in crashes by wealth  High −1.8 (−3.8 to 0.2) 0.073 −1.9 (−3.9 to 0.2) 0.075  Low −2.6 (−4.4 to −0.8) 0.005 −2.6 (−4.4 to −0.8) 0.005 −1.6 (−2.9 to −0.4) 0.012 −1.7 (−3.0 to −0.4) 0.011 N.B. Direct association with road crash represents the change in road crash likelihood, controlling for the other factors. Direct association of wealth with risk factors represents the change in each risk factor associated with the difference in wealth. The total indirect association is the total difference in road crash likelihood comparing wealth levels that operates through the change in risk factors. The total difference in crashes by wealth is the overall difference in crash likelihood comparing wealth levels. Adjusted models also control for age. The difference in crash risk by wealth category varies inconsequentially from the values in the text because of listwise deletion of observations with missing data in the SEMs. In age-adjusted models, urban men in the wealthiest three ventiles were 2.9 percentage points (95% CI: −5.0 to −0.7) less likely to report a road crash in the last year than other urban men. Wealthier men were more likely not to be working (4.4 percentage points, 95% CI: 1.0–7.7), which was associated with 4.0 percentage point (95% CI: −7.4 to −0.7) lower likelihood of crash. They were also a marginally significant 3.2 percentage points (95% CI: −7.0 to 0.6) less likely to drive for a living, which was associated with 9.9 percentage points (95% CI: 4.8–15.0) higher crashes. However, these were counterbalanced by marginally significantly more bicycle ownership (4.0 percentage points, 95% CI: −0.2 to 8.2), which was associated with 9.4% more crashes (95% CI: 3.5–15.3), leaving the net reduction in urban male crashes unexplained. The poorest three ventiles of rural women in age-adjusted models were 1.7 (95% CI: −3.0 to −0.4) percentage points less likely to report a road crash. They reported being 6.2 (95% CI: 3.3–9.2) percentage points more likely to consume alcohol, which was associated with a marginally significant 3.3 (95% CI: −0.1 to 6.6) percentage point increase in crashes. However, vehicle ownership and occupation were not associated with road crashes for rural women. The poorest 20% of urban women were 2.6 (95% CI: −4.4 to −0.8) percentage points less likely to report a road crash than those in the middle three quintiles. They were 1.4 (95% CI: −2.0 to −0.9) percentage points less likely to own an automobile, which was associated with a 2.8 (95% CI: −4.8 to −0.8) percentage point lower likelihood of crash; this, however, does not meaningfully explain the difference in reported crashes. The wealthiest 20% of urban women were a marginally significant 1.9 (95% CI: −3.9 to 0.2) percentage points less likely to report a crash. This operated partially through greater automobile ownership (23.8 percentage points, 95% CI: 18.2–29.4). The relationship between hypothesized direct risk factors and road crash varied by context (Tables 2 and 3). For men, motorcycle ownership and occupational driving were the strongest risk factors and consistently associated with greater crash risk, and not working was associated with lower risk. Alcohol consumption was associated with greater risk in rural but not urban areas for both men and women; the association was only marginally significant for women in the fully adjusted model. Automobile ownership was associated with lower risk only for urban women, and bicycle ownership was associated with greater risk only for urban men. Results were robust to modeling assumptions. While the polychoric PCA-based wealth index explained ~5-fold more variance than the Filmer–Pritchett approach and 25% of households were categorized in different wealth quintiles, the overall relationship between relative wealth and road crash differed little between the approaches (Supplement 1). Differences between linearized and bootstrapped standard errors were negligible (Supplement 1). Discussion Main findings of this study This paper provides evidence that non-fatal road crashes were not commonest among adults in the least wealthy Kenyan households. Rather, crashes were most common in the middle of the national household income distribution for both men and women. However, this inverted-U-shaped pattern resulted from markedly different relationships between relative household wealth and road crashes between rural and urban areas. In rural areas, crashes were least likely in the poorest households—among both men and women. In urban areas, male crashes were lowest among the wealthiest, while they peaked in the middle of the female urban wealth distribution. These patterns are partially explained by differences in occupation and vehicle ownership, especially for male respondents. What is already known on this topic Prior studies have found differing relationships between wealth and road crash. Several studies from SSA find that road crash is more common among those who are relatively poorer, which is consistent with findings for urban men,6,7 while others did not.8,9 Studies from the continent have directly measured economic position relatively rarely, and studies using proxies (such as education and location of residence) have also produced mixed findings.10,12,13 Studies assessing road crash risk as lower-income countries develop have generally found increases in road crash.14–17,51 Additionally, identified mediators of crash are broadly consistent with prior literature, including occupational risk to drivers due to increased transport exposure and financial incentives to operate substandard vehicles, speed and overload vehicles.36–38 Similarly, motorcycle use both tends to be more dangerous than other modes of transport and enables people to travel farther and faster than they otherwise could.34 Limitations This analysis several limitations. First, inferring causality from cross-sectional data is tenuous. There may be instances where road crashes earlier in the recall period caused undetectable changes to wealth and other predictors. Because asset indices produce a long-term measure of economic position, this is a lesser concern than for alternative measures—such as current income—but reverse causation may persist for relatively serious injuries.22,52 Second, fatal crashes could not be captured. More crashes may be fatal among rural and poorer subpopulations, with greater barriers to treatment or a greater likelihood of involvement as vulnerable road users. If so, this paper may overestimate the observed lower crash risk among the rural poor (though an unexpectedly high fraction of crashes would have to be fatal to eliminate observed differences). Conversely, these data may underestimate urban differences. Third, two studies from SSA have found that 12-month recall periods are suboptimal because minor, though not severe, injuries go under-reported with longer recall periods. One study found no association between recall error and any respondent characteristic,53 but the other found that recall errors were more likely for rural respondents.54 This suggests that the road crash levels observed in this study may be underestimates—particularly in rural areas. However, it also suggests that associations between crashes and wealth may not be biased by recall once stratified by location of residence. Fourth, many relevant variables were not available from DHS surveys, such as average daily distances traveled, principal modes of transport and some risk behaviors. In rural areas, it is likely that the poorest travel the least due to transportation costs. In urban areas, the relatively wealthy are least likely to use more dangerous modes of transportation such as matatus (public minibuses).7,55 Similarly, the survey did not distinguish crashes as pedestrians or vehicle occupants. Fifth, women’s risk differences were particularly poorly explained, partially because occupational categories allowed less granular identification of women’s jobs. Street vending—previously been found to be a significant risk factor42,43,45—was not distinguishable from other low-skill sales and services occupations in the survey. Relatedly, vehicle ownership imperfectly represents use, particularly for women, who may be less likely to have control over household vehicles in some settings. Finally, the survey does not ask how many crashes a person experienced over the past 12 months, nor does it ask about the existence or the severity of injuries. As a result, a cumulative incidence of crashes over the past year can be calculated, but the incidence rate of crashes or crash-related injuries cannot be. Similarly, crash risk per kilometer traveled cannot be calculated without additional road-use exposure data. What this study adds This study provides evidence of a complex and nuanced relationship between relative household wealth and non-fatal road crash risk in Kenya. It extends the findings of the Kenyan government and because the sample is larger than previous studies in similar setting, it is able to identify and partially explain differing associations with crash risk in rural and urban areas and between men and women. It suggests that the relationship between wealth and road crash in SSA should be considered differently for various subpopulations and in rural and urban areas. Road crash prevention should remain a priority in developing countries. In particular, as wealth acquisition is associated with greater rural crash risk, it should be a focus of rural development efforts. Supplementary data Supplementary data are available at the Journal of Public Health online. References 1 Institute for Health Metrics and Evaluation . Road Injuries: Both Sexes, All Ages. Vol. 2017; 2017 . 2 Adeloye D , Thompson JY , Akanbi MA et al. . The burden of road traffic crashes, injuries and deaths in Africa: a systematic review and meta-analysis . Bull World Health Organ 2016 ; 94 ( 7 ): 510 – 521A . doi:10.2471/BLT.15.163121 . Google Scholar CrossRef Search ADS PubMed 3 Haagsma JA , Graetz N , Bolliger I et al. . The global burden of injury: incidence, mortality, disability-adjusted life years and time trends from the Global Burden of Disease study 2013 . Inj Prev J Int Soc Child Adolesc Inj Prev 2016 ; 22 ( 1 ): 3 – 18 . doi:10.1136/injuryprev-2015-041616 . Google Scholar CrossRef Search ADS 4 Gosselin RA , Spiegel DA , Coughlin R et al. . Injuries: the neglected burden in developing countries . Bull World Health Organ 2009 ; 87 ( 4 ): 246 – 246a . doi:S0042-96862009000400002 . Google Scholar CrossRef Search ADS PubMed 5 Lagarde E . Road traffic injury is an escalating burden in Africa and deserves proportionate research efforts . PLoS Med 2007 ; 4 ( 6 ): e170 . doi:06-PLME-ES-0915 . Google Scholar CrossRef Search ADS PubMed 6 Azetsop J . Social justice approach to road safety in Kenya: addressing the uneven distribution of road traffic injuries and deaths across population groups . Public Health Ethics 2010 ; 3 ( 2 ): 115 – 127 . doi:10.1093/phe/phq013 . Google Scholar CrossRef Search ADS PubMed 7 Nantulya VM , Muli-Musiime F . Uncovering the social determinants of road traffic accidents in Kenya. In: Evans T , Whitehead M , Diderichsen F , Bhuiya A , Wirth M (eds) . Challenging Inequities: From Ethics to Action . Oxford : Oxford University Press , 2001 , 210 – 25 . Google Scholar CrossRef Search ADS 8 Moshiro C , Heuch I , Astrom AN et al. . Injury morbidity in an urban and a rural area in Tanzania: an epidemiological survey . BMC Public Health 2005 ; 5 : 11 . doi:1471-2458-5-11 . Google Scholar CrossRef Search ADS PubMed 9 Grimm M , Treibich C . Socio-economic determinants of road traffic accident fatalities in low and middle income countries . ISS Work Pap SeriesGeneral Ser 2010 ; 504 : 1 – 44 . 10 Labinjo M , Juillard C , Kobusingye OC et al. . The burden of road traffic injuries in Nigeria: results of a population-based survey . Inj Prev J Int Soc Child Adolesc Inj Prev 2009 ; 15 ( 3 ): 157 – 162 . doi:10.1136/ip.2008.020255 . Google Scholar CrossRef Search ADS 11 Kraemer JD . Urbanization and unintentional injury in low- and middle-income countries. In: Ahn R , Burke TF , McGahan AM (eds) . Innovating for Healthy Urbanization . Springer US : Boston, MA , 2015 : 209 – 244 . doi:10.1007/978-1-4899-7597-3_10 . Google Scholar CrossRef Search ADS 12 Naci H , Chisholm D , Baker TD . Distribution of road traffic deaths by road user group: a global comparison . Inj Prev J Int Soc Child Adolesc Inj Prev 2009 ; 15 ( 1 ): 55 – 59 . doi:10.1136/ip.2008.018721 . Google Scholar CrossRef Search ADS 13 Mock CN , Tiska M , Adu-Ampofo M et al. . Improvements in prehospital trauma care in an African country with no formal emergency medical services . J Trauma 2002 ; 53 ( 1 ): 90 – 97 . Google Scholar CrossRef Search ADS PubMed 14 Wang S-Y , Chi G-B , Jing C-X et al. . Trends in road traffic crashes and associated injury and fatality in the People’s Republic of China, 1951–1999 . Inj Control Saf Promot 2003 ; 10 ( 1–2 ): 83 – 87 . doi:10.1076/icsp.10.1.83.14105 . Google Scholar CrossRef Search ADS PubMed 15 Al-Reesi H , Ganguly SS , Al-Adawi S et al. . Economic growth, motorization, and road traffic injuries in the Sultanate of Oman, 1985–2009 . Traffic Inj Prev 2013 ; 14 ( 3 ): 322 – 328 . doi:10.1080/15389588.2012.694088 . Google Scholar CrossRef Search ADS PubMed 16 Bishai D , Quresh A , James P et al. . National road casualties and economic development . Health Econ 2006 ; 15 ( 1 ): 65 – 81 . doi:10.1002/hec.1020 . Google Scholar CrossRef Search ADS PubMed 17 Paulozzi LJ , Ryan GW , Espitia-Hardeman VE et al. . Economic development’s effect on road transport-related mortality among different types of road users: a cross-sectional international study . Accid Anal Prev 2007 ; 39 ( 3 ): 606 – 617 . doi:S0001-4575(06)00181-3 . Google Scholar CrossRef Search ADS PubMed 18 Kopits E , Cropper M . Traffic fatalities and economic growth . Accid Anal Prev 2005 ; 37 ( 1 ): 169 – 178 . doi:S0001-4575(04)00068-5 . Google Scholar CrossRef Search ADS PubMed 19 Hyder AA , Peden M . Inequality and road-traffic injuries: call for action . Lancet Lond Engl 2003 ; 362 ( 9401 ): 2034 – 2035 . doi:S0140-6736(03)15145-8 . Google Scholar CrossRef Search ADS 20 Kenya National Bureau of Statistics , Ministry of Health, National AIDS Control Council, Kenya Medical Research Institute, National Council for Population and Development, I. C. F. International. Kenya Demographic and Health Survey 2014. 2015 . http://dhsprogram.com/pubs/pdf/FR308/FR308.pdf. 21 I. C. F. International . Sampling and Household Listing Manual: Demographic and Health Surveys Methodology. 2012 . 22 Filmer D , Pritchett LH . Estimating wealth effects without expenditure data–or tears: an application to educational enrollments in states of India . Demography 2001 ; 38 ( 1 ): 115 – 132 . Google Scholar PubMed 23 Rutstein SO. Steps to constructing the new DHS wealth index. Rockv MD ICF Int. 2015 . 24 Rutstein SO . The DHS Wealth Index: Approaches for Rural and Urban Areas. 2008 . http://www.dhsprogram.com/pubs/pdf/WP60/WP60.pdf. 25 Kolenikov S , Angeles G . Socioeconomic status measurement with discrete proxy variables: is principal component analysis a reliable answer? Rev Income Wealth 2009 ; 55 ( 1 ): 128 – 165 . doi:10.1111/j.1475-4991.2008.00309.x . Google Scholar CrossRef Search ADS 26 Rutstein SO , Rojas G Guide to DHS statistics: demographic and health surveys methodology. Meas DHSICF Int. 2006 . 27 Measure DHS/ICF International . Standard Recode Manual for DHS 6. 2013 . http://www.dhsprogram.com/pubs/pdf/DHSG4/Recode6_DHS_22March2013_DHSG4.pdf. 28 International Labour Organization . Inventory of Current Definitions of Rural/Urban. Vol. 2017 . 29 Kenya Ministry of Land , Housing, and Urban Development, Kenya National Bureau of Statistics. 2012/2013 Kenya National Housing Survey Basic Report. 2014 . http://statistics.knbs.or.ke/nada/index.php/catalog/83/download/520. 30 Odero W . Alcohol-related road traffic injuries in Eldoret, Kenya . East Afr Med J 1998 ; 75 ( 12 ): 708 – 711 . Google Scholar PubMed 31 Chalya PL , Mabula JB , Dass RM et al. . Injury characteristics and outcome of road traffic crash victims at Bugando Medical Centre in Northwestern Tanzania . J Trauma Manag Outcomes 2012 ; 6 ( 1 ): 1 . doi:10.1186/1752-2897-6-1 . Google Scholar CrossRef Search ADS PubMed 32 Mogaka EO , Ng’ang’a Z , Oundo J et al. . Factors associated with severity of road traffic injuries, Thika, Kenya . Pan Afr Med J 2011 ; 8 : 20 . Google Scholar PubMed 33 Forjuoh SN , Zwi AB , Mock CN . Injury control in Africa: getting governments to do more . Trop Med Int Health 1998 ; 3 ( 5 ): 349 – 356 . doi:10.1046/j.1365-3156.1998.00240.x . Google Scholar CrossRef Search ADS PubMed 34 World Health Organization . World Report on Road Traffic Injury Prevention. 2004 . http://apps.who.int/iris/bitstream/10665/42871/1/9241562609.pdf. 35 International Labour Organization . ISCO-88. Vol. 2017. 2004 . 36 Habyarimana J , Jack W . Results of a large-scale randomized behavior change intervention on road safety in Kenya . Proc Natl Acad Sci U S A 2015 ; 112 ( 34 ): E4661 – 70 . doi:10.1073/pnas.1422009112 . Google Scholar CrossRef Search ADS PubMed 37 Munala G , Maina K . Fatigue and the road users: travellers’ account! Int J Inj Contr Saf Promot 2010 ; 17 ( 4 ): 257 – 265 . doi:10.1080/17457300.2010.498641 . Google Scholar CrossRef Search ADS PubMed 38 Raynor NJ , Mirzoev T . Understanding road safety in Kenya: views of matatu drivers . Int Health 2014 ; 6 ( 3 ): 242 – 248 . doi:10.1093/inthealth/ihu034 . Google Scholar CrossRef Search ADS PubMed 39 Odero W , Khayesi M , Heda PM . Road traffic injuries in Kenya: magnitude, causes and status of intervention . Inj Control Saf Promot 2003 ; 10 ( 1–2 ): 53 – 61 . doi:10.1076/icsp.10.1.53.14103 . Google Scholar CrossRef Search ADS PubMed 40 Mock CN , Forjuoh SN , Rivara FP . Epidemiology of transport-related injuries in Ghana . Accid Anal Prev 1999 ; 31 ( 4 ): 359 – 370 . doi:S0001-4575(98)00064-5 . Google Scholar CrossRef Search ADS PubMed 41 Grieco M , Turner J , Kwakye EA . Informal public transport and the woman trader in Accra, Ghana. Seventh World Conference on Transport Research. 1995 . https://pdfs.semanticscholar.org/4504/d0e88ec5167922027632907ed89d7fc3ddc3.pdf. 42 Ogunrinola IO , Fadayomi TO , Amoo EO et al. . Occupational Health and Safety for Informal Sector Workers: The Case of Street Traders in Nigeria. Vol. 13; 2012:173–183. http://eprints.covenantuniversity.edu.ng/4110/1/2012-%20IAABD%20Conference%20Proceedings%20(Page%20173-183).pdf#page=186. 43 Ayikai OM . Self-reported Health Status Among Street Vendors in Accra. 2013 . http://ugspace.ug.edu.gh:8080/bitstream/123456789/5867/1/Owusu%20Monica%20Ayikai_Self-Reported%20Health%20Status%20Among%20Street%20Vendors%20in%20Accra_2013.pdf. 44 Kenya National Bureau of Statistics . Kenya—Integrated Household Budget Survey 2005 —KNOCS Codes. 45 Ghaffar A , Hyder AA , Masud TI . The burden of road traffic injuries in developing countries: the 1st national injury survey of Pakistan . Public Health 2004 ; 118 ( 3 ): 211 – 217 . doi:10.1016/j.puhe.2003.05.003 . Google Scholar CrossRef Search ADS PubMed 46 Saidi HS. Risk factor profile and the cost of traffic injury in a tertiary hospital in Kenya. Traffic Safety on Three Continents. Statens väg-och transportforskningsinstitut; 2001 . http://vti.diva-portal.org/smash/get/diva2:673699/FULLTEXT06.pdf. 47 Royston P , Ambler G , Sauerbrei W . The use of fractional polynomials to model continuous risk variables in epidemiology . Int J Epidemiol 1999 ; 28 ( 5 ): 964 – 974 . Google Scholar CrossRef Search ADS PubMed 48 Royston P , Altman DG . Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling . J R Stat Soc C Appl Stat 1994 ; 43 ( 3 ): 429 – 467 . doi:10.2307/2986270 . 49 Xu C , Chen J , Mantell H . Pseudo-likelihood-based Bayesian information criterion for variable selection in survey data . Surv Methodol 2013 ; 39 : 303 – 322 . 50 Wen X , Kleinman K , Gillman MW et al. . Childhood body mass index trajectories: modeling, characterizing, pairwise correlations and socio-demographic predictors of trajectory characteristics . BMC Med Res Methodol 2012 ; 12 : 38 – 2288-12-38 . doi:10.1186/1471-2288-12-38 . Google Scholar CrossRef Search ADS PubMed 51 He H , Paichadze N , Hyder AA et al. . Economic development and road traffic fatalities in Russia: analysis of federal regions 2004-2011 . Inj Epidemiol 2015 ; 2 ( 1 ): 19 . doi:10.1186/s40621-015-0051-6 . Google Scholar CrossRef Search ADS PubMed 52 Rutstein SO , Johnson K . The DHS wealth index. DHS comparative reports no. 6. Calverton ORC Macro. 2004 . 53 Mock C , Acheampong F , Adjei S et al. . The effect of recall on estimation of incidence rates for injury in Ghana . Int J Epidemiol 1999 ; 28 ( 4 ): 750 – 755 . Google Scholar CrossRef Search ADS PubMed 54 Moshiro C , Heuch I , Astrom AN et al. . Effect of recall on estimation of non-fatal injury rates: a community based study in Tanzania . Inj Prev J Int Soc Child Adolesc Inj Prev 2005 ; 11 ( 1 ): 48 – 52 . doi:11/1/48 . Google Scholar CrossRef Search ADS 55 Mutongi Kenda . Matatu: A History of Popular Transportation in Nairobi. 2017 . © The Author(s) 2018. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Public Health Oxford University Press

Relative household wealth and non-fatal road crashes: analysis of population-representative data of Kenyan adults

Journal of Public Health , Volume Advance Article – May 18, 2018

Loading next page...
 
/lp/ou_press/relative-household-wealth-and-non-fatal-road-crashes-analysis-of-NFUlAD0tJ5
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1741-3842
eISSN
1741-3850
D.O.I.
10.1093/pubmed/fdy089
Publisher site
See Article on Publisher Site

Abstract

Abstract Background This study aims to examine potential road crash disparities across relative wealth and location of residence in Kenya by analyzing population-representative Demographic and Health Survey data. Methods Relative wealth was measured by household assets, converted into an index by polychoric principal components analysis. Location and sex-stratified associations between wealth quantiles and crashes were flexibly estimated using fractional polynomial models. Structural equation models were fit to examine whether observed differences may operate through previously identified determinants. Results In rural areas, crashes were least common for both the poorest men (−5.2 percentage points, 95% CI: −7.3 to −3.2) and women (−1.6 percentage points, 95% CI: −2.9 to −0.4). In urban areas, male crashes were lowest (−3.0 percentage points, 95% CI: −5.2 to −0.8) among the wealthiest, while they peaked in the middle of the female wealth distribution (2.0 percentage points, 95% CI: 0.3–3.8). Male differences operate partially though occupational driving and vehicle ownership. Urban female differences operate partially through household vehicle ownership, but differences for rural women were not explained by modeled determinants. Conclusions Relative wealth and road crash have opposite associations in rural and urban areas. Especially in rural areas, it is important to mitigate potential unintended effects of economic development. accidents, public health, transport Introduction Road crashes are estimated to kill ~150 000 people in sub-Saharan Africa (SSA) each year.1 Unlike most other regions, SSA has not experienced marked reductions in road crash injuries over the past 25 years, and the region’s rapid economic growth and urbanization will likely drive increased road crash risk in the coming years.2,3 Nonetheless, road cash in SSA remains understudied.4,5 One particular area with relatively little research is how wealth and poverty affect with road injury risks in SSA. There is some evidence that risk tends to be greatest among those with economic disadvantages, because they are more likely to use less safe transport options and to more time vulnerably exposed to vehicular traffic as pedestrians;6,7 however, this has not been a consistent finding.8,9 One Nigerian study found that higher education—a correlate of wealth—was associated with higher road crash risk.10 Studies of rural–urban disparities—often a partial proxy for wealth—have found differing associations, suggesting context-specific relationships.11–13 Globally, studies often find that increasingly motorized transport results in increased road crashes as developing countries become wealthier.14–16 Simultaneously, cross-country analyses suggest that road crash fatalities increase as countries reach middle-income status, likely because of greater access to motorcycles and the greater risk that motorized transport will strike vulnerable road users (pedestrians, cyclists and motorcyclists), though risk may remain concentrated in populations’ lower income segments.17–19 The enactment of Kenya’s 2010 constitution, which devolved authority to 47 counties, created an opportunity to examine factors associated with road crash in more detail. The 2014 Kenyan Demographic and Health Survey, which includes a module on injury, was increased greatly in size to be able to produce county-specific indicators for key development priorities. This resulted in the largest-ever African population-based survey with road crash information. The Kenyan government’s bivariate results were tantalizing, finding a general trend toward greater road crash frequency among wealthier than the poorest women, but the opposite trend among men. It also found differences by rural and urban residence, with urban women but rural men having higher risk.20 Because location of residence tends to be highly correlated with wealth, these findings may be entangled. This paper aims to extend existing analyses of the 2014 Kenyan DHS survey. First, it seeks to understand the relationship between relative wealth and road crashes with more granularity and stratified by sex and residence. Second, to the extent associations with relative wealth are identified, it seeks to understand the extent to which they operate through well-established causes of road crash risk, such as vehicle ownership and certain occupations. Methods Participants, sampling and survey administration This study makes secondary use of cross-sectional data collected through Kenya’s 2014 Demographic and Health Survey (DHS), which has been fully described previously.20 Its essential features are briefly recapped here. The survey employed a stratified, two-stage cluster sample design. For each of Kenya’s 47 counties, rural and urban strata were developed (except for the entirely urban Nairobi and Mombasa counties). Within each stratum, clusters were selected probability proportionate to size and then 25 households were randomly selected per cluster.20,21 Half the selected households were asked a longer questionnaire that included an unintentional injury module. Enumerators invited all women aged 15–49 years and men aged 15–54 years in selected households to participate. Among households selected for the injury module, 98.8% participated and, within them, 96.2% of eligible women and 90.2% of eligible men participated.20 Because risk factors for youth are likely different, respondents under age 18 were excluded from this analysis. Data were collected via interviewer-administered surveys between May and October 2014. Interviewers received approximately three weeks of classroom and field training. Surveys were written in English and translated to 16 additional languages. Accuracy of translation was assessed during pretesting and altered if necessary. All data were double-entered for quality assurance.20 Data used for this article were downloaded from a public repository maintained by ICF International at http://www.dhsprogram.com/Data/. Measures The main outcome variable was self-reported involvement ‘in a road traffic accident as a driver, passenger, pedestrian or cyclist’ within 12 months preceding the survey. The main predictor variable was relative household wealth, as measured through an asset index. DHS datasets include a standard wealth index score derived via a principal component analysis (PCA) technique developed by Filmer and Pritchett.22–24 However, Kolenikov and Angeles recently identified potential bias from this approach’s handling of discrete variables (such as water source) as a series of mutually exclusive indicator variables.25 This article, therefore, uses two modifications they recommend to the standard DHS approach. First, some asset index items, such as water source and toilet type, are recategorized as ordinal variables. Second, polychoric correlations, which correctly handle dichotomous and ordinal variables, were used for the PCA. The wealth index score was calculated as the score corresponding to the first principal component. This score was converted to three separate ventile (1/20th quantile) rankings: one for the overall Kenyan population and separate rankings for rural and urban areas. The residence-specific rankings were used to maximize the ability to distinguish between ranks in each area. See the Methods appendix provided in Supplement 1 for more detail on the wealth index construction and quality assurance. Additional variables were included in standard DHS form as described previously.26,27 Respondents’ location of residence was considered a potential moderator of associations between wealth and road crash risk and dichotomized as urban or rural, with urban defined by the Kenyan government as settlements with at least 2000 residents.28,29 Variables through which associations between wealth and crash risk were expected to operate included self-reported alcohol use (dichotomized as reporting any alcohol use in the last 14 days or not);30–33 household bicycle, motorbike, automobile ownership (dichotomous);32,34 and occupations hypothesized to alter exposure to road crash risk. Occupations were classified according to the International Standard Classification of Occupations,35 and dichotomous variables were included for drivers7,36–40 and sales and services elementary occupations, for which a large fraction are street and informal vendors frequently exposed to crash risks.41–45 Occupational drivers were almost exclusively men and therefore not included in female analyses. Unemployment, which was hypothesized to reduce transport for work, was also included.46 Age, which may correlate with wealth and with a variety of risk-taking behaviors,8,20 was included as a categorical control variable in 5-year increments. Statistical methods Descriptive analyses were conducted to characterize the relationship between relative wealth and reported road crashes, stratified by gender and rural/urban location. With no strong a priori expectation that associations would take a particular form, the flexible fractional polynomial approach recommended by Royston and colleagues was followed.47,48 Logistic regression models with all possible first and second-degree fractional polynomials with power transformations from −2 to 3 were fit and the best fitting, parsimonious model was chosen using an adaptation of Bayesian Information Criterion for survey data.49,50 Subsequently, modeled probabilities of road crash and confidence intervals were calculated and plotted. The Methods appendix in Supplement 1 has more details. The resulting plots were then examined, and portions of the wealth distribution were identified that had markedly different crash rates than the rest. For males, these were the lowest four rural ventiles and highest three urban ventiles. For females, it was the lowest three rural ventiles and the highest and lowest four urban ventiles. Indicator variables were constructed for these portions of the wealth distribution and significant difference was determined using a sampling design-corrected test for difference in proportions. If found to be significant, structural equation models were fit to determine whether associations between road crash and wealth were attributable to differences in factors expected a priori to increase crash risk. In basic models, the wealth category was modeled to predict road crash risk as well as vehicle ownership, occupation and alcohol consumption, each of which was also modeled to predict road crash risk. In adjusted models, age was also included as a control variable. Relationships between variables were modeled with the assumption they are additive on a probability scale. All analyses incorporated sampling weights and used Taylor linearization to adjust standard errors for clustering. Analyses used Stata 15.0. Statistical code to fully replicate the paper is provided as Supplement 2. Sensitivity analyses Two analyses were conducted to check robustness to modeling assumptions. In the first, the analysis is replicated using the ventiles of the original DHS wealth index. In the second, the structural equation models’ variance estimation was based on 10 000 bootstrap replications. Both are described in more detail in Supplement 1. Because this study made secondary use of anonymized data, Georgetown University’s institutional review board did not require review. Results A total of 24 011 respondents met inclusion criteria and completed the survey version that included injury items. Of these, 12 940 were female and 14 602 lived in rural areas. Participant characteristics are provided in Table 1. Table 1 Participant characteristics Urban (n = 9409) Rural (n = 16 402) Total (n = 24 011) n Weighted % (95% CI) n Weighted % (95% CI) n Weighted % (95% CI) Sex  Male 4429 48.6 (47.1–50.1) 6642 45.0 (44.2–45.8) 11 071 46.6 (45.8–47.4)  Female 4980 51.4 (49.9–52.9) 7960 55.0 (54.2–55.8) 12 940 53.4 (52.6–54.2) Wealth index score, mean (CI) 9409 −0.2 (−0.3 to −0.1) 16 402 1.7 (1.5–1.8) 24 011 0.6 (0.5–0.7) Vehicle ownership  Automobile 708 7.6 (6.1–9.3) 421 3.4 (2.8–4.0) 1129 5.2 (4.5–6.0)  Motorcycle 762 6.3 (5.4–7.2) 1468 10.3 (9.3–11.4) 2230 8.5 (7.9–9.3)  Bicycle 1785 17.9 (16.1–19.9) 3824 28.5 (26.8–30.3) 5609 23.8 (22.6–25.2) Drinks alcohol 1732 22.0 (20.7–23.4) 2261 16.0 (15.2–16.9) 3993 18.7 (17.9–19.4) Relevant occupations  Driver 390 4.4 (3.9–5.1) 334 2.5 (2.2–2.9) 724 3.3 (3.0–3.7)  Sales and Services Elem. Occ. 2127 25.8 (23.9–27.8) 2106 15.0 (13.9–16.2) 4233 19.7 (18.7–20.8)  Not working 1935 17.6 (16.4–19.0) 3222 19.4 (18.4–20.4) 5157 18.6 (17.8–19.5) Age  25 and under 3091 33.8 (32.0–35.5) 4567 32.0 (30.9–33.0) 7658 32.8 (31.8–33.7)  26–30 2129 23.6 (22.2–25.1) 2716 18.2 (17.4–19.0) 4845 20.6 (19.8–21.4)  31–35 1473 16.5 (15.3–17.8) 2083 14.3 (13.5–15.1) 3556 15.3 (14.6–16.0)  36–40 1174 11.8 (10.8–12.9) 2094 13.7 (13.1–14.4) 3268 12.9 (12.3–13.5)  40–45 801 8.0 (7.2–8.8) 1585 11.1 (10.5–11.8) 2386 9.8 (9.3–10.2)  46 and over 741 6.3 (5.6–7.2) 1557 10.6 (10.0–11.4) 2298 8.8 (8.2–9.3) Urban (n = 9409) Rural (n = 16 402) Total (n = 24 011) n Weighted % (95% CI) n Weighted % (95% CI) n Weighted % (95% CI) Sex  Male 4429 48.6 (47.1–50.1) 6642 45.0 (44.2–45.8) 11 071 46.6 (45.8–47.4)  Female 4980 51.4 (49.9–52.9) 7960 55.0 (54.2–55.8) 12 940 53.4 (52.6–54.2) Wealth index score, mean (CI) 9409 −0.2 (−0.3 to −0.1) 16 402 1.7 (1.5–1.8) 24 011 0.6 (0.5–0.7) Vehicle ownership  Automobile 708 7.6 (6.1–9.3) 421 3.4 (2.8–4.0) 1129 5.2 (4.5–6.0)  Motorcycle 762 6.3 (5.4–7.2) 1468 10.3 (9.3–11.4) 2230 8.5 (7.9–9.3)  Bicycle 1785 17.9 (16.1–19.9) 3824 28.5 (26.8–30.3) 5609 23.8 (22.6–25.2) Drinks alcohol 1732 22.0 (20.7–23.4) 2261 16.0 (15.2–16.9) 3993 18.7 (17.9–19.4) Relevant occupations  Driver 390 4.4 (3.9–5.1) 334 2.5 (2.2–2.9) 724 3.3 (3.0–3.7)  Sales and Services Elem. Occ. 2127 25.8 (23.9–27.8) 2106 15.0 (13.9–16.2) 4233 19.7 (18.7–20.8)  Not working 1935 17.6 (16.4–19.0) 3222 19.4 (18.4–20.4) 5157 18.6 (17.8–19.5) Age  25 and under 3091 33.8 (32.0–35.5) 4567 32.0 (30.9–33.0) 7658 32.8 (31.8–33.7)  26–30 2129 23.6 (22.2–25.1) 2716 18.2 (17.4–19.0) 4845 20.6 (19.8–21.4)  31–35 1473 16.5 (15.3–17.8) 2083 14.3 (13.5–15.1) 3556 15.3 (14.6–16.0)  36–40 1174 11.8 (10.8–12.9) 2094 13.7 (13.1–14.4) 3268 12.9 (12.3–13.5)  40–45 801 8.0 (7.2–8.8) 1585 11.1 (10.5–11.8) 2386 9.8 (9.3–10.2)  46 and over 741 6.3 (5.6–7.2) 1557 10.6 (10.0–11.4) 2298 8.8 (8.2–9.3) N.B. Sample sizes (n) are unweighted observations, while proportions are weighted. Wealth index score is the component score for the first principal component calculated via polychoric PCA. Table 1 Participant characteristics Urban (n = 9409) Rural (n = 16 402) Total (n = 24 011) n Weighted % (95% CI) n Weighted % (95% CI) n Weighted % (95% CI) Sex  Male 4429 48.6 (47.1–50.1) 6642 45.0 (44.2–45.8) 11 071 46.6 (45.8–47.4)  Female 4980 51.4 (49.9–52.9) 7960 55.0 (54.2–55.8) 12 940 53.4 (52.6–54.2) Wealth index score, mean (CI) 9409 −0.2 (−0.3 to −0.1) 16 402 1.7 (1.5–1.8) 24 011 0.6 (0.5–0.7) Vehicle ownership  Automobile 708 7.6 (6.1–9.3) 421 3.4 (2.8–4.0) 1129 5.2 (4.5–6.0)  Motorcycle 762 6.3 (5.4–7.2) 1468 10.3 (9.3–11.4) 2230 8.5 (7.9–9.3)  Bicycle 1785 17.9 (16.1–19.9) 3824 28.5 (26.8–30.3) 5609 23.8 (22.6–25.2) Drinks alcohol 1732 22.0 (20.7–23.4) 2261 16.0 (15.2–16.9) 3993 18.7 (17.9–19.4) Relevant occupations  Driver 390 4.4 (3.9–5.1) 334 2.5 (2.2–2.9) 724 3.3 (3.0–3.7)  Sales and Services Elem. Occ. 2127 25.8 (23.9–27.8) 2106 15.0 (13.9–16.2) 4233 19.7 (18.7–20.8)  Not working 1935 17.6 (16.4–19.0) 3222 19.4 (18.4–20.4) 5157 18.6 (17.8–19.5) Age  25 and under 3091 33.8 (32.0–35.5) 4567 32.0 (30.9–33.0) 7658 32.8 (31.8–33.7)  26–30 2129 23.6 (22.2–25.1) 2716 18.2 (17.4–19.0) 4845 20.6 (19.8–21.4)  31–35 1473 16.5 (15.3–17.8) 2083 14.3 (13.5–15.1) 3556 15.3 (14.6–16.0)  36–40 1174 11.8 (10.8–12.9) 2094 13.7 (13.1–14.4) 3268 12.9 (12.3–13.5)  40–45 801 8.0 (7.2–8.8) 1585 11.1 (10.5–11.8) 2386 9.8 (9.3–10.2)  46 and over 741 6.3 (5.6–7.2) 1557 10.6 (10.0–11.4) 2298 8.8 (8.2–9.3) Urban (n = 9409) Rural (n = 16 402) Total (n = 24 011) n Weighted % (95% CI) n Weighted % (95% CI) n Weighted % (95% CI) Sex  Male 4429 48.6 (47.1–50.1) 6642 45.0 (44.2–45.8) 11 071 46.6 (45.8–47.4)  Female 4980 51.4 (49.9–52.9) 7960 55.0 (54.2–55.8) 12 940 53.4 (52.6–54.2) Wealth index score, mean (CI) 9409 −0.2 (−0.3 to −0.1) 16 402 1.7 (1.5–1.8) 24 011 0.6 (0.5–0.7) Vehicle ownership  Automobile 708 7.6 (6.1–9.3) 421 3.4 (2.8–4.0) 1129 5.2 (4.5–6.0)  Motorcycle 762 6.3 (5.4–7.2) 1468 10.3 (9.3–11.4) 2230 8.5 (7.9–9.3)  Bicycle 1785 17.9 (16.1–19.9) 3824 28.5 (26.8–30.3) 5609 23.8 (22.6–25.2) Drinks alcohol 1732 22.0 (20.7–23.4) 2261 16.0 (15.2–16.9) 3993 18.7 (17.9–19.4) Relevant occupations  Driver 390 4.4 (3.9–5.1) 334 2.5 (2.2–2.9) 724 3.3 (3.0–3.7)  Sales and Services Elem. Occ. 2127 25.8 (23.9–27.8) 2106 15.0 (13.9–16.2) 4233 19.7 (18.7–20.8)  Not working 1935 17.6 (16.4–19.0) 3222 19.4 (18.4–20.4) 5157 18.6 (17.8–19.5) Age  25 and under 3091 33.8 (32.0–35.5) 4567 32.0 (30.9–33.0) 7658 32.8 (31.8–33.7)  26–30 2129 23.6 (22.2–25.1) 2716 18.2 (17.4–19.0) 4845 20.6 (19.8–21.4)  31–35 1473 16.5 (15.3–17.8) 2083 14.3 (13.5–15.1) 3556 15.3 (14.6–16.0)  36–40 1174 11.8 (10.8–12.9) 2094 13.7 (13.1–14.4) 3268 12.9 (12.3–13.5)  40–45 801 8.0 (7.2–8.8) 1585 11.1 (10.5–11.8) 2386 9.8 (9.3–10.2)  46 and over 741 6.3 (5.6–7.2) 1557 10.6 (10.0–11.4) 2298 8.8 (8.2–9.3) N.B. Sample sizes (n) are unweighted observations, while proportions are weighted. Wealth index score is the component score for the first principal component calculated via polychoric PCA. The association between relative household wealth and road crash incidence is shown in Fig. 1. When rural and urban areas are combined, associations had an inverted-U shape for both men and women. However, when disaggregated, rural and urban areas had markedly different associations. In rural areas, crashes were least common among the poorest respondents for both men (5.2 percentage points lower, 95% CI: −7.3 to −3.2) and women (1.6 percentage points lower, 95% CI: −2.9 to −0.4). In urban areas, risk was lowest among the wealthiest men (3.0 percentage points lower, 95% CI: −5.2 to −0.8), whereas it peaked for women in the middle of the wealth distribution (2.0 percentage points, 95% CI: 0.3–3.8). Fig. 1 View largeDownload slide One-year cumulative road crash incidence, Kenyan adults. Fig. 1 View largeDownload slide One-year cumulative road crash incidence, Kenyan adults. Across all populations, results were similar whether adjusted for age or not (Tables 2 and 3). In age-adjusted models, the poorest rural men were 5.4 percentage points (95% CI: −7.5 to −3.3) less likely to report a crash in the last year. Of this, 1.0 percentage point (95% CI: −1.6 to −0.3) of the reduction was explainable through other variables. Poorer men were less likely to be occupational drivers (−3.5 percentage points, 95% CI: −5.2 to −1.8), and drivers were 7.0 (95% CI: 1.9–12.2) percentage points more likely to be in a crash. They were also 10.9 (95% CI: −12.5 to −9.2) percentage points less likely to own a motorcycle, and motorcycle owners were 8.2 (95% CI: 4.7–11.6) percentage points more likely to be in a crash. Table 2 Structural equation model estimates for men Urban males Rural males Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Direct association with road crash  Wealth   High −3.3 (−6.0 to −0.6) 0.017 −3.2 (−5.8 to −0.6) 0.017   Low −4.3 (−6.5 to −2.2) <0.001 −4.5 (−6.6 to −2.3) <0.001  Alcohol use 1.7 (−0.6 to 4.0) 0.138 1.6 (−0.7 to 3.9) 0.161 2.6 (0.7–4.6) 0.009 3.5 (1.4–5.5) 0.001  Occupation   Driver 9.8 (4.7–14.9) <0.001 9.9 (4.8–15.0) <0.001 7.8 (2.5–13.0) 0.004 7.0 (1.9–12.2) 0.008   Sales and Services Elem. Occ. −2.3 (−4.9 to 0.3) 0.088 −2.2 (−4.8 to 0.4) 0.094 0.5 (−1.9 to 3.0) 0.659 0.2 (−2.3 to 2.6) 0.903   Not working −3.5 (−7.0 to −0.1) 0.017 −4.0 (−7.4 to −0.7) 0.018 −1.5 (−4.1 to 1.1) 0.250 −3.5 (−6.5 to −0.6) 0.019  Household owns   Automobile −0.2 (−4.8 to 4.5) 0.938 0.2 (−4.5 to 4.9) 0.929 −3.2 (−8.0 to 1.5) 0.182 −2.8 (−7.5 to 1.9) 0.244   Motorcycle 8.3 (3.4–13.3) 0.001 8.3 (3.3–13.2) 0.001 8.3 (4.8–11.8) <0.001 8.2 (4.7–11.6) <0.001   Bicycle 3.8 (−0.5 to 8.1) 0.086 4.0 (−0.2 to 8.2) 0.059 −0.8 (−2.7 to 1.2) 0.451 −0.6 (−2.5 to 1.4) 0.567 Direct association of wealth with risk factors  Alcohol use 11.2 (3.6–18.8) 0.004 10.3 (2.6–17.9) 0.009 0.4 (−4.2 to 5.1) 0.851 −0.4 (−5.0 to 4.2) 0.863  Occupation   Driver −2.8 (−6.5 to 0.8) 0.129 −3.2 (−7.0 to 0.6) 0.099 −3.2 (−4.8 to −1.6) <0.001 −3.5 (−5.2 to −1.8) <0.001   Sales and Services Elem. Occ. −6.6 (−11.7 to −1.6) 0.010 −7.1 (−12.1 to −2.1) 0.006 3.2 (−0.7 to 7.0) 0.111 2.8 (−1.0 to 6.7) 0.149   Not working 2.2 (−1.7 to 6.0) 0.276 4.4 (1.0–7.7) 0.011 0.1 (−2.7 to 3.0) 0.920 1.6 (−1.0 to 4.3) 0.225  Household owns   Automobile 28.1 (22.2–34.0) <0.001 28.0 (22.1–33.9) <0.001 −4.0 (−4.8 to −3.2) <0.001 −4.0 (−4.8 to −3.2) <0.001   Motorcycle 0.4 (−2.4 to 3.1) 0.786 0.4 (−2.4 to 3.1) 0.799 −10.6 (−12.3 to −8.9) <0.001 −10.9 (−12.5 to −9.2) <0.001   Bicycle 10.1 (4.1–16.0) 0.001 9.4 (3.5–15.3) 0.002 −24.2 (−27.3 to −21.1) <0.001 −23.8 (−27.0 to −20.7) <0.001 Total indirect association between wealth and crashes 0.4 (−1.0 to 1.7) 0.615 0.3 (−1.1 to 1.7) 0.677 −0.8 (−1.4 to −0.1) 0.017 −1.0 (−1.6 to −0.3) 0.005 Total difference in crashes by wealth −2.9 (−5.1 to −0.7) 0.010 −2.9 (−5.0 to −0.7) 0.009 −5.1 (−7.2 to −3.0) <0.001 −5.4 (−7.5 to −3.3) <0.001 Urban males Rural males Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Direct association with road crash  Wealth   High −3.3 (−6.0 to −0.6) 0.017 −3.2 (−5.8 to −0.6) 0.017   Low −4.3 (−6.5 to −2.2) <0.001 −4.5 (−6.6 to −2.3) <0.001  Alcohol use 1.7 (−0.6 to 4.0) 0.138 1.6 (−0.7 to 3.9) 0.161 2.6 (0.7–4.6) 0.009 3.5 (1.4–5.5) 0.001  Occupation   Driver 9.8 (4.7–14.9) <0.001 9.9 (4.8–15.0) <0.001 7.8 (2.5–13.0) 0.004 7.0 (1.9–12.2) 0.008   Sales and Services Elem. Occ. −2.3 (−4.9 to 0.3) 0.088 −2.2 (−4.8 to 0.4) 0.094 0.5 (−1.9 to 3.0) 0.659 0.2 (−2.3 to 2.6) 0.903   Not working −3.5 (−7.0 to −0.1) 0.017 −4.0 (−7.4 to −0.7) 0.018 −1.5 (−4.1 to 1.1) 0.250 −3.5 (−6.5 to −0.6) 0.019  Household owns   Automobile −0.2 (−4.8 to 4.5) 0.938 0.2 (−4.5 to 4.9) 0.929 −3.2 (−8.0 to 1.5) 0.182 −2.8 (−7.5 to 1.9) 0.244   Motorcycle 8.3 (3.4–13.3) 0.001 8.3 (3.3–13.2) 0.001 8.3 (4.8–11.8) <0.001 8.2 (4.7–11.6) <0.001   Bicycle 3.8 (−0.5 to 8.1) 0.086 4.0 (−0.2 to 8.2) 0.059 −0.8 (−2.7 to 1.2) 0.451 −0.6 (−2.5 to 1.4) 0.567 Direct association of wealth with risk factors  Alcohol use 11.2 (3.6–18.8) 0.004 10.3 (2.6–17.9) 0.009 0.4 (−4.2 to 5.1) 0.851 −0.4 (−5.0 to 4.2) 0.863  Occupation   Driver −2.8 (−6.5 to 0.8) 0.129 −3.2 (−7.0 to 0.6) 0.099 −3.2 (−4.8 to −1.6) <0.001 −3.5 (−5.2 to −1.8) <0.001   Sales and Services Elem. Occ. −6.6 (−11.7 to −1.6) 0.010 −7.1 (−12.1 to −2.1) 0.006 3.2 (−0.7 to 7.0) 0.111 2.8 (−1.0 to 6.7) 0.149   Not working 2.2 (−1.7 to 6.0) 0.276 4.4 (1.0–7.7) 0.011 0.1 (−2.7 to 3.0) 0.920 1.6 (−1.0 to 4.3) 0.225  Household owns   Automobile 28.1 (22.2–34.0) <0.001 28.0 (22.1–33.9) <0.001 −4.0 (−4.8 to −3.2) <0.001 −4.0 (−4.8 to −3.2) <0.001   Motorcycle 0.4 (−2.4 to 3.1) 0.786 0.4 (−2.4 to 3.1) 0.799 −10.6 (−12.3 to −8.9) <0.001 −10.9 (−12.5 to −9.2) <0.001   Bicycle 10.1 (4.1–16.0) 0.001 9.4 (3.5–15.3) 0.002 −24.2 (−27.3 to −21.1) <0.001 −23.8 (−27.0 to −20.7) <0.001 Total indirect association between wealth and crashes 0.4 (−1.0 to 1.7) 0.615 0.3 (−1.1 to 1.7) 0.677 −0.8 (−1.4 to −0.1) 0.017 −1.0 (−1.6 to −0.3) 0.005 Total difference in crashes by wealth −2.9 (−5.1 to −0.7) 0.010 −2.9 (−5.0 to −0.7) 0.009 −5.1 (−7.2 to −3.0) <0.001 −5.4 (−7.5 to −3.3) <0.001 N.B. Direct association with road crash represents the change in road crash likelihood, controlling for the other factors. Direct association of wealth with risk factors represents the change in each risk factor associated with the difference in wealth. The total indirect association is the total difference in road crash likelihood comparing wealth levels that operates through the change in risk factors. The total difference in crashes by wealth is the overall difference in crash likelihood comparing wealth levels. Adjusted models also control for age. The difference in crash risk by wealth category varies inconsequentially from the values in the text because of listwise deletion of observations with missing data in the SEMs. Table 2 Structural equation model estimates for men Urban males Rural males Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Direct association with road crash  Wealth   High −3.3 (−6.0 to −0.6) 0.017 −3.2 (−5.8 to −0.6) 0.017   Low −4.3 (−6.5 to −2.2) <0.001 −4.5 (−6.6 to −2.3) <0.001  Alcohol use 1.7 (−0.6 to 4.0) 0.138 1.6 (−0.7 to 3.9) 0.161 2.6 (0.7–4.6) 0.009 3.5 (1.4–5.5) 0.001  Occupation   Driver 9.8 (4.7–14.9) <0.001 9.9 (4.8–15.0) <0.001 7.8 (2.5–13.0) 0.004 7.0 (1.9–12.2) 0.008   Sales and Services Elem. Occ. −2.3 (−4.9 to 0.3) 0.088 −2.2 (−4.8 to 0.4) 0.094 0.5 (−1.9 to 3.0) 0.659 0.2 (−2.3 to 2.6) 0.903   Not working −3.5 (−7.0 to −0.1) 0.017 −4.0 (−7.4 to −0.7) 0.018 −1.5 (−4.1 to 1.1) 0.250 −3.5 (−6.5 to −0.6) 0.019  Household owns   Automobile −0.2 (−4.8 to 4.5) 0.938 0.2 (−4.5 to 4.9) 0.929 −3.2 (−8.0 to 1.5) 0.182 −2.8 (−7.5 to 1.9) 0.244   Motorcycle 8.3 (3.4–13.3) 0.001 8.3 (3.3–13.2) 0.001 8.3 (4.8–11.8) <0.001 8.2 (4.7–11.6) <0.001   Bicycle 3.8 (−0.5 to 8.1) 0.086 4.0 (−0.2 to 8.2) 0.059 −0.8 (−2.7 to 1.2) 0.451 −0.6 (−2.5 to 1.4) 0.567 Direct association of wealth with risk factors  Alcohol use 11.2 (3.6–18.8) 0.004 10.3 (2.6–17.9) 0.009 0.4 (−4.2 to 5.1) 0.851 −0.4 (−5.0 to 4.2) 0.863  Occupation   Driver −2.8 (−6.5 to 0.8) 0.129 −3.2 (−7.0 to 0.6) 0.099 −3.2 (−4.8 to −1.6) <0.001 −3.5 (−5.2 to −1.8) <0.001   Sales and Services Elem. Occ. −6.6 (−11.7 to −1.6) 0.010 −7.1 (−12.1 to −2.1) 0.006 3.2 (−0.7 to 7.0) 0.111 2.8 (−1.0 to 6.7) 0.149   Not working 2.2 (−1.7 to 6.0) 0.276 4.4 (1.0–7.7) 0.011 0.1 (−2.7 to 3.0) 0.920 1.6 (−1.0 to 4.3) 0.225  Household owns   Automobile 28.1 (22.2–34.0) <0.001 28.0 (22.1–33.9) <0.001 −4.0 (−4.8 to −3.2) <0.001 −4.0 (−4.8 to −3.2) <0.001   Motorcycle 0.4 (−2.4 to 3.1) 0.786 0.4 (−2.4 to 3.1) 0.799 −10.6 (−12.3 to −8.9) <0.001 −10.9 (−12.5 to −9.2) <0.001   Bicycle 10.1 (4.1–16.0) 0.001 9.4 (3.5–15.3) 0.002 −24.2 (−27.3 to −21.1) <0.001 −23.8 (−27.0 to −20.7) <0.001 Total indirect association between wealth and crashes 0.4 (−1.0 to 1.7) 0.615 0.3 (−1.1 to 1.7) 0.677 −0.8 (−1.4 to −0.1) 0.017 −1.0 (−1.6 to −0.3) 0.005 Total difference in crashes by wealth −2.9 (−5.1 to −0.7) 0.010 −2.9 (−5.0 to −0.7) 0.009 −5.1 (−7.2 to −3.0) <0.001 −5.4 (−7.5 to −3.3) <0.001 Urban males Rural males Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Unadjusted Δ% (95% CI) P Adjusted Δ% (95% CI) P Direct association with road crash  Wealth   High −3.3 (−6.0 to −0.6) 0.017 −3.2 (−5.8 to −0.6) 0.017   Low −4.3 (−6.5 to −2.2) <0.001 −4.5 (−6.6 to −2.3) <0.001  Alcohol use 1.7 (−0.6 to 4.0) 0.138 1.6 (−0.7 to 3.9) 0.161 2.6 (0.7–4.6) 0.009 3.5 (1.4–5.5) 0.001  Occupation   Driver 9.8 (4.7–14.9) <0.001 9.9 (4.8–15.0) <0.001 7.8 (2.5–13.0) 0.004 7.0 (1.9–12.2) 0.008   Sales and Services Elem. Occ. −2.3 (−4.9 to 0.3) 0.088 −2.2 (−4.8 to 0.4) 0.094 0.5 (−1.9 to 3.0) 0.659 0.2 (−2.3 to 2.6) 0.903   Not working −3.5 (−7.0 to −0.1) 0.017 −4.0 (−7.4 to −0.7) 0.018 −1.5 (−4.1 to 1.1) 0.250 −3.5 (−6.5 to −0.6) 0.019  Household owns   Automobile −0.2 (−4.8 to 4.5) 0.938 0.2 (−4.5 to 4.9) 0.929 −3.2 (−8.0 to 1.5) 0.182 −2.8 (−7.5 to 1.9) 0.244   Motorcycle 8.3 (3.4–13.3) 0.001 8.3 (3.3–13.2) 0.001 8.3 (4.8–11.8) <0.001 8.2 (4.7–11.6) <0.001   Bicycle 3.8 (−0.5 to 8.1) 0.086 4.0 (−0.2 to 8.2) 0.059 −0.8 (−2.7 to 1.2) 0.451 −0.6 (−2.5 to 1.4) 0.567 Direct association of wealth with risk factors  Alcohol use 11.2 (3.6–18.8) 0.004 10.3 (2.6–17.9) 0.009 0.4 (−4.2 to 5.1) 0.851 −0.4 (−5.0 to 4.2) 0.863  Occupation   Driver −2.8 (−6.5 to 0.8) 0.129 −3.2 (−7.0 to 0.6) 0.099 −3.2 (−4.8 to −1.6) <0.001 −3.5 (−5.2 to −1.8) <0.001   Sales and Services Elem. Occ. −6.6 (−11.7 to −1.6) 0.010 −7.1 (−12.1 to −2.1) 0.006 3.2 (−0.7 to 7.0) 0.111 2.8 (−1.0 to 6.7) 0.149   Not working 2.2 (−1.7 to 6.0) 0.276 4.4 (1.0–7.7) 0.011 0.1 (−2.7 to 3.0) 0.920 1.6 (−1.0 to 4.3) 0.225  Household owns   Automobile 28.1 (22.2–34.0) <0.001 28.0 (22.1–33.9) <0.001 −4.0 (−4.8 to −3.2) <0.001 −4.0 (−4.8 to −3.2) <0.001   Motorcycle 0.4 (−2.4 to 3.1) 0.786 0.4 (−2.4 to 3.1) 0.799 −10.6 (−12.3 to −8.9) <0.001 −10.9 (−12.5 to −9.2) <0.001   Bicycle 10.1 (4.1–16.0) 0.001 9.4 (3.5–15.3) 0.002 −24.2 (−27.3 to −21.1) <0.001 −23.8 (−27.0 to −20.7) <0.001 Total indirect association between wealth and crashes 0.4 (−1.0 to 1.7) 0.615 0.3 (−1.1 to 1.7) 0.677 −0.8 (−1.4 to −0.1) 0.017 −1.0 (−1.6 to −0.3) 0.005 Total difference in crashes by wealth −2.9 (−5.1 to −0.7) 0.010 −2.9 (−5.0 to −0.7) 0.009 −5.1 (−7.2 to −3.0) <0.001 −5.4 (−7.5 to −3.3) <0.001 N.B. Direct association with road crash represents the change in road crash likelihood, controlling for the other factors. Direct association of wealth with risk factors represents the change in each risk factor associated with the difference in wealth. The total indirect association is the total difference in road crash likelihood comparing wealth levels that operates through the change in risk factors. The total difference in crashes by wealth is the overall difference in crash likelihood comparing wealth levels. Adjusted models also control for age. The difference in crash risk by wealth category varies inconsequentially from the values in the text because of listwise deletion of observations with missing data in the SEMs. Table 3 Structural equation model estimates for women Urban females Rural females Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Direct association with road crash  Wealth   High −1.3 (−3.7 to 1.1) 0.297 −1.3 (−3.7 to 1.1) 0.301   Low −2.7 (−4.5 to −0.8) 0.009 −2.7 (−4.5 to −0.9) 0.004 −1.3 (−2.6 to 0.1) 0.061 −1.3 (−2.6 to 0.0) 0.058  Alcohol use 1.1 (−3.0 to 5.2) 0.602 1.0 (−3.1 to 5.0) 0.631 3.4 (0.1–6.8) 0.046 3.3 (−0.1 to 6.6) 0.054  Occupation   Sales and Services, Elem. Occ. −0.2 (−2.3 to 2.0) 0.891 −0.2 (−2.4 to 2.0) 0.884 0.8 (−0.7 to 2.2) 0.289 0.8 (−0.7 to 2.2) 0.285   Not working −1.9 (−4.0 to 0.2) 0.079 −2.1 (−4.4 to 0.1) 0.060 −0.8 (−1.9 to 0.2) 0.122 −0.8 (−2.0 to 0.3) 0.150  Household owns   Automobile −2.7 (−4.8 to −0.7) 0.010 −2.8 (−4.8 to −0.8) 0.006 1.5 (−2.4 to 5.3) 0.451 1.4 (−2.4 to 5.3) 0.468   Motorcycle 1.7 (−2.7 to 6.1) 0.453 1.8 (−2.7 to 6.2) 0.431 1.0 (−0.8 to 2.8) 0.278 1.1 (−0.8 to 2.9) 0.257   Bicycle −0.9 (−2.5 to 0.8) 0.303 −1.0 (−2.6 to 0.7) 0.239 0.6 (−0.6 to 1.8) 0.352 0.5 (−0.7 to 1.7) 0.402 Direct association of high wealth with risk factors  Alcohol use 5.6 (2.0–9.2) 0.002 5.6 (2.0–9.1) 0.002  Occupation   Sales and Services Elem. Occ. −2.0 (−6.9 to 2.8) 0.409 −2.2 (−7.0 to 2.6) 0.369   Not working −5.4 (−10.5 to −0.2) 0.043 −4.7 (−9.7 to 0.3) 0.068  Household Owns:   Automobile 23.9 (18.4–29.5) <0.001 23.8 (18.2–29.4) <0.001   Motorcycle −0.3 (−2.5 to 1.8) 0.765 −0.4 (−2.6 to 1.7) 0.714   Bicycle 8.0 (3.4–12.6) 0.001 7.7 (3.2–12.2) 0.001 Direct association of low wealth with risk factors  Alcohol use 1.3 (−1.8 to 4.3) 0.405 1.3 (−1.7 to 4.3) 0.400 6.1 (3.2–9.1) <0.001 6.2 (3.3–9.2) <0.001  Occupation   Sales and Services, Elem. Occ. −4.4 (−10.0 to 1.1) 0.118 −4.8 (−10.1 to 0.5) 0.073 −1.3 (−4.7 to 2.0) 0.438 −1.0 (−4.4 to 2.3) 0.544   Not working −1.8 (−6.9 to 3.3) 0.479 −0.6 (−5.5 to 4.2) 0.799 35.7 (30.7–40.8) <0.001 34.5 (29.7–39.3) <0.001  Household owns   Automobile −1.2 (−1.6 to −0.7) <0.001 −1.4 (−2.0 to −0.9) <0.001 −3.4 (−4.0 to −2.7) <0.001 −3.3 (−4.0 to −2.7) <0.001   Motorcycle −3.8 (−5.6 to −2.0) <0.001 −3.8 (−5.6 to −2.0) <0.001 −8.2 (−10.1 to −6.3) <0.001 −8.2 (−10.1 to −6.3) <0.001   Bicycle −6.4 (−9.6 to −3.2) <0.001 −7.1 (−10.4 to −3.9) <0.001 −26.2 (−28.3 to −24.0) <0.001 −26.2 (−28.4 to −24.0) <0.001 Total indirect association between wealth and crashes  High −0.6 (−1.2 to 0.0) 0.072 −0.6 (−1.2 to 0.0) 0.053  Low 0.1 (−0.2 to 0.3) 0.544 0.1 (−0.2 to 0.3) 0.543 −0.4 (−1.0 to 0.2) 0.217 −0.4 (−1.0 to 0.3) 0.256 Total difference in crashes by wealth  High −1.8 (−3.8 to 0.2) 0.073 −1.9 (−3.9 to 0.2) 0.075  Low −2.6 (−4.4 to −0.8) 0.005 −2.6 (−4.4 to −0.8) 0.005 −1.6 (−2.9 to −0.4) 0.012 −1.7 (−3.0 to −0.4) 0.011 Urban females Rural females Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Direct association with road crash  Wealth   High −1.3 (−3.7 to 1.1) 0.297 −1.3 (−3.7 to 1.1) 0.301   Low −2.7 (−4.5 to −0.8) 0.009 −2.7 (−4.5 to −0.9) 0.004 −1.3 (−2.6 to 0.1) 0.061 −1.3 (−2.6 to 0.0) 0.058  Alcohol use 1.1 (−3.0 to 5.2) 0.602 1.0 (−3.1 to 5.0) 0.631 3.4 (0.1–6.8) 0.046 3.3 (−0.1 to 6.6) 0.054  Occupation   Sales and Services, Elem. Occ. −0.2 (−2.3 to 2.0) 0.891 −0.2 (−2.4 to 2.0) 0.884 0.8 (−0.7 to 2.2) 0.289 0.8 (−0.7 to 2.2) 0.285   Not working −1.9 (−4.0 to 0.2) 0.079 −2.1 (−4.4 to 0.1) 0.060 −0.8 (−1.9 to 0.2) 0.122 −0.8 (−2.0 to 0.3) 0.150  Household owns   Automobile −2.7 (−4.8 to −0.7) 0.010 −2.8 (−4.8 to −0.8) 0.006 1.5 (−2.4 to 5.3) 0.451 1.4 (−2.4 to 5.3) 0.468   Motorcycle 1.7 (−2.7 to 6.1) 0.453 1.8 (−2.7 to 6.2) 0.431 1.0 (−0.8 to 2.8) 0.278 1.1 (−0.8 to 2.9) 0.257   Bicycle −0.9 (−2.5 to 0.8) 0.303 −1.0 (−2.6 to 0.7) 0.239 0.6 (−0.6 to 1.8) 0.352 0.5 (−0.7 to 1.7) 0.402 Direct association of high wealth with risk factors  Alcohol use 5.6 (2.0–9.2) 0.002 5.6 (2.0–9.1) 0.002  Occupation   Sales and Services Elem. Occ. −2.0 (−6.9 to 2.8) 0.409 −2.2 (−7.0 to 2.6) 0.369   Not working −5.4 (−10.5 to −0.2) 0.043 −4.7 (−9.7 to 0.3) 0.068  Household Owns:   Automobile 23.9 (18.4–29.5) <0.001 23.8 (18.2–29.4) <0.001   Motorcycle −0.3 (−2.5 to 1.8) 0.765 −0.4 (−2.6 to 1.7) 0.714   Bicycle 8.0 (3.4–12.6) 0.001 7.7 (3.2–12.2) 0.001 Direct association of low wealth with risk factors  Alcohol use 1.3 (−1.8 to 4.3) 0.405 1.3 (−1.7 to 4.3) 0.400 6.1 (3.2–9.1) <0.001 6.2 (3.3–9.2) <0.001  Occupation   Sales and Services, Elem. Occ. −4.4 (−10.0 to 1.1) 0.118 −4.8 (−10.1 to 0.5) 0.073 −1.3 (−4.7 to 2.0) 0.438 −1.0 (−4.4 to 2.3) 0.544   Not working −1.8 (−6.9 to 3.3) 0.479 −0.6 (−5.5 to 4.2) 0.799 35.7 (30.7–40.8) <0.001 34.5 (29.7–39.3) <0.001  Household owns   Automobile −1.2 (−1.6 to −0.7) <0.001 −1.4 (−2.0 to −0.9) <0.001 −3.4 (−4.0 to −2.7) <0.001 −3.3 (−4.0 to −2.7) <0.001   Motorcycle −3.8 (−5.6 to −2.0) <0.001 −3.8 (−5.6 to −2.0) <0.001 −8.2 (−10.1 to −6.3) <0.001 −8.2 (−10.1 to −6.3) <0.001   Bicycle −6.4 (−9.6 to −3.2) <0.001 −7.1 (−10.4 to −3.9) <0.001 −26.2 (−28.3 to −24.0) <0.001 −26.2 (−28.4 to −24.0) <0.001 Total indirect association between wealth and crashes  High −0.6 (−1.2 to 0.0) 0.072 −0.6 (−1.2 to 0.0) 0.053  Low 0.1 (−0.2 to 0.3) 0.544 0.1 (−0.2 to 0.3) 0.543 −0.4 (−1.0 to 0.2) 0.217 −0.4 (−1.0 to 0.3) 0.256 Total difference in crashes by wealth  High −1.8 (−3.8 to 0.2) 0.073 −1.9 (−3.9 to 0.2) 0.075  Low −2.6 (−4.4 to −0.8) 0.005 −2.6 (−4.4 to −0.8) 0.005 −1.6 (−2.9 to −0.4) 0.012 −1.7 (−3.0 to −0.4) 0.011 N.B. Direct association with road crash represents the change in road crash likelihood, controlling for the other factors. Direct association of wealth with risk factors represents the change in each risk factor associated with the difference in wealth. The total indirect association is the total difference in road crash likelihood comparing wealth levels that operates through the change in risk factors. The total difference in crashes by wealth is the overall difference in crash likelihood comparing wealth levels. Adjusted models also control for age. The difference in crash risk by wealth category varies inconsequentially from the values in the text because of listwise deletion of observations with missing data in the SEMs. Table 3 Structural equation model estimates for women Urban females Rural females Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Direct association with road crash  Wealth   High −1.3 (−3.7 to 1.1) 0.297 −1.3 (−3.7 to 1.1) 0.301   Low −2.7 (−4.5 to −0.8) 0.009 −2.7 (−4.5 to −0.9) 0.004 −1.3 (−2.6 to 0.1) 0.061 −1.3 (−2.6 to 0.0) 0.058  Alcohol use 1.1 (−3.0 to 5.2) 0.602 1.0 (−3.1 to 5.0) 0.631 3.4 (0.1–6.8) 0.046 3.3 (−0.1 to 6.6) 0.054  Occupation   Sales and Services, Elem. Occ. −0.2 (−2.3 to 2.0) 0.891 −0.2 (−2.4 to 2.0) 0.884 0.8 (−0.7 to 2.2) 0.289 0.8 (−0.7 to 2.2) 0.285   Not working −1.9 (−4.0 to 0.2) 0.079 −2.1 (−4.4 to 0.1) 0.060 −0.8 (−1.9 to 0.2) 0.122 −0.8 (−2.0 to 0.3) 0.150  Household owns   Automobile −2.7 (−4.8 to −0.7) 0.010 −2.8 (−4.8 to −0.8) 0.006 1.5 (−2.4 to 5.3) 0.451 1.4 (−2.4 to 5.3) 0.468   Motorcycle 1.7 (−2.7 to 6.1) 0.453 1.8 (−2.7 to 6.2) 0.431 1.0 (−0.8 to 2.8) 0.278 1.1 (−0.8 to 2.9) 0.257   Bicycle −0.9 (−2.5 to 0.8) 0.303 −1.0 (−2.6 to 0.7) 0.239 0.6 (−0.6 to 1.8) 0.352 0.5 (−0.7 to 1.7) 0.402 Direct association of high wealth with risk factors  Alcohol use 5.6 (2.0–9.2) 0.002 5.6 (2.0–9.1) 0.002  Occupation   Sales and Services Elem. Occ. −2.0 (−6.9 to 2.8) 0.409 −2.2 (−7.0 to 2.6) 0.369   Not working −5.4 (−10.5 to −0.2) 0.043 −4.7 (−9.7 to 0.3) 0.068  Household Owns:   Automobile 23.9 (18.4–29.5) <0.001 23.8 (18.2–29.4) <0.001   Motorcycle −0.3 (−2.5 to 1.8) 0.765 −0.4 (−2.6 to 1.7) 0.714   Bicycle 8.0 (3.4–12.6) 0.001 7.7 (3.2–12.2) 0.001 Direct association of low wealth with risk factors  Alcohol use 1.3 (−1.8 to 4.3) 0.405 1.3 (−1.7 to 4.3) 0.400 6.1 (3.2–9.1) <0.001 6.2 (3.3–9.2) <0.001  Occupation   Sales and Services, Elem. Occ. −4.4 (−10.0 to 1.1) 0.118 −4.8 (−10.1 to 0.5) 0.073 −1.3 (−4.7 to 2.0) 0.438 −1.0 (−4.4 to 2.3) 0.544   Not working −1.8 (−6.9 to 3.3) 0.479 −0.6 (−5.5 to 4.2) 0.799 35.7 (30.7–40.8) <0.001 34.5 (29.7–39.3) <0.001  Household owns   Automobile −1.2 (−1.6 to −0.7) <0.001 −1.4 (−2.0 to −0.9) <0.001 −3.4 (−4.0 to −2.7) <0.001 −3.3 (−4.0 to −2.7) <0.001   Motorcycle −3.8 (−5.6 to −2.0) <0.001 −3.8 (−5.6 to −2.0) <0.001 −8.2 (−10.1 to −6.3) <0.001 −8.2 (−10.1 to −6.3) <0.001   Bicycle −6.4 (−9.6 to −3.2) <0.001 −7.1 (−10.4 to −3.9) <0.001 −26.2 (−28.3 to −24.0) <0.001 −26.2 (−28.4 to −24.0) <0.001 Total indirect association between wealth and crashes  High −0.6 (−1.2 to 0.0) 0.072 −0.6 (−1.2 to 0.0) 0.053  Low 0.1 (−0.2 to 0.3) 0.544 0.1 (−0.2 to 0.3) 0.543 −0.4 (−1.0 to 0.2) 0.217 −0.4 (−1.0 to 0.3) 0.256 Total difference in crashes by wealth  High −1.8 (−3.8 to 0.2) 0.073 −1.9 (−3.9 to 0.2) 0.075  Low −2.6 (−4.4 to −0.8) 0.005 −2.6 (−4.4 to −0.8) 0.005 −1.6 (−2.9 to −0.4) 0.012 −1.7 (−3.0 to −0.4) 0.011 Urban females Rural females Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Unadj. Δ% (95% CI) P Adj. Δ% (95% CI) P Direct association with road crash  Wealth   High −1.3 (−3.7 to 1.1) 0.297 −1.3 (−3.7 to 1.1) 0.301   Low −2.7 (−4.5 to −0.8) 0.009 −2.7 (−4.5 to −0.9) 0.004 −1.3 (−2.6 to 0.1) 0.061 −1.3 (−2.6 to 0.0) 0.058  Alcohol use 1.1 (−3.0 to 5.2) 0.602 1.0 (−3.1 to 5.0) 0.631 3.4 (0.1–6.8) 0.046 3.3 (−0.1 to 6.6) 0.054  Occupation   Sales and Services, Elem. Occ. −0.2 (−2.3 to 2.0) 0.891 −0.2 (−2.4 to 2.0) 0.884 0.8 (−0.7 to 2.2) 0.289 0.8 (−0.7 to 2.2) 0.285   Not working −1.9 (−4.0 to 0.2) 0.079 −2.1 (−4.4 to 0.1) 0.060 −0.8 (−1.9 to 0.2) 0.122 −0.8 (−2.0 to 0.3) 0.150  Household owns   Automobile −2.7 (−4.8 to −0.7) 0.010 −2.8 (−4.8 to −0.8) 0.006 1.5 (−2.4 to 5.3) 0.451 1.4 (−2.4 to 5.3) 0.468   Motorcycle 1.7 (−2.7 to 6.1) 0.453 1.8 (−2.7 to 6.2) 0.431 1.0 (−0.8 to 2.8) 0.278 1.1 (−0.8 to 2.9) 0.257   Bicycle −0.9 (−2.5 to 0.8) 0.303 −1.0 (−2.6 to 0.7) 0.239 0.6 (−0.6 to 1.8) 0.352 0.5 (−0.7 to 1.7) 0.402 Direct association of high wealth with risk factors  Alcohol use 5.6 (2.0–9.2) 0.002 5.6 (2.0–9.1) 0.002  Occupation   Sales and Services Elem. Occ. −2.0 (−6.9 to 2.8) 0.409 −2.2 (−7.0 to 2.6) 0.369   Not working −5.4 (−10.5 to −0.2) 0.043 −4.7 (−9.7 to 0.3) 0.068  Household Owns:   Automobile 23.9 (18.4–29.5) <0.001 23.8 (18.2–29.4) <0.001   Motorcycle −0.3 (−2.5 to 1.8) 0.765 −0.4 (−2.6 to 1.7) 0.714   Bicycle 8.0 (3.4–12.6) 0.001 7.7 (3.2–12.2) 0.001 Direct association of low wealth with risk factors  Alcohol use 1.3 (−1.8 to 4.3) 0.405 1.3 (−1.7 to 4.3) 0.400 6.1 (3.2–9.1) <0.001 6.2 (3.3–9.2) <0.001  Occupation   Sales and Services, Elem. Occ. −4.4 (−10.0 to 1.1) 0.118 −4.8 (−10.1 to 0.5) 0.073 −1.3 (−4.7 to 2.0) 0.438 −1.0 (−4.4 to 2.3) 0.544   Not working −1.8 (−6.9 to 3.3) 0.479 −0.6 (−5.5 to 4.2) 0.799 35.7 (30.7–40.8) <0.001 34.5 (29.7–39.3) <0.001  Household owns   Automobile −1.2 (−1.6 to −0.7) <0.001 −1.4 (−2.0 to −0.9) <0.001 −3.4 (−4.0 to −2.7) <0.001 −3.3 (−4.0 to −2.7) <0.001   Motorcycle −3.8 (−5.6 to −2.0) <0.001 −3.8 (−5.6 to −2.0) <0.001 −8.2 (−10.1 to −6.3) <0.001 −8.2 (−10.1 to −6.3) <0.001   Bicycle −6.4 (−9.6 to −3.2) <0.001 −7.1 (−10.4 to −3.9) <0.001 −26.2 (−28.3 to −24.0) <0.001 −26.2 (−28.4 to −24.0) <0.001 Total indirect association between wealth and crashes  High −0.6 (−1.2 to 0.0) 0.072 −0.6 (−1.2 to 0.0) 0.053  Low 0.1 (−0.2 to 0.3) 0.544 0.1 (−0.2 to 0.3) 0.543 −0.4 (−1.0 to 0.2) 0.217 −0.4 (−1.0 to 0.3) 0.256 Total difference in crashes by wealth  High −1.8 (−3.8 to 0.2) 0.073 −1.9 (−3.9 to 0.2) 0.075  Low −2.6 (−4.4 to −0.8) 0.005 −2.6 (−4.4 to −0.8) 0.005 −1.6 (−2.9 to −0.4) 0.012 −1.7 (−3.0 to −0.4) 0.011 N.B. Direct association with road crash represents the change in road crash likelihood, controlling for the other factors. Direct association of wealth with risk factors represents the change in each risk factor associated with the difference in wealth. The total indirect association is the total difference in road crash likelihood comparing wealth levels that operates through the change in risk factors. The total difference in crashes by wealth is the overall difference in crash likelihood comparing wealth levels. Adjusted models also control for age. The difference in crash risk by wealth category varies inconsequentially from the values in the text because of listwise deletion of observations with missing data in the SEMs. In age-adjusted models, urban men in the wealthiest three ventiles were 2.9 percentage points (95% CI: −5.0 to −0.7) less likely to report a road crash in the last year than other urban men. Wealthier men were more likely not to be working (4.4 percentage points, 95% CI: 1.0–7.7), which was associated with 4.0 percentage point (95% CI: −7.4 to −0.7) lower likelihood of crash. They were also a marginally significant 3.2 percentage points (95% CI: −7.0 to 0.6) less likely to drive for a living, which was associated with 9.9 percentage points (95% CI: 4.8–15.0) higher crashes. However, these were counterbalanced by marginally significantly more bicycle ownership (4.0 percentage points, 95% CI: −0.2 to 8.2), which was associated with 9.4% more crashes (95% CI: 3.5–15.3), leaving the net reduction in urban male crashes unexplained. The poorest three ventiles of rural women in age-adjusted models were 1.7 (95% CI: −3.0 to −0.4) percentage points less likely to report a road crash. They reported being 6.2 (95% CI: 3.3–9.2) percentage points more likely to consume alcohol, which was associated with a marginally significant 3.3 (95% CI: −0.1 to 6.6) percentage point increase in crashes. However, vehicle ownership and occupation were not associated with road crashes for rural women. The poorest 20% of urban women were 2.6 (95% CI: −4.4 to −0.8) percentage points less likely to report a road crash than those in the middle three quintiles. They were 1.4 (95% CI: −2.0 to −0.9) percentage points less likely to own an automobile, which was associated with a 2.8 (95% CI: −4.8 to −0.8) percentage point lower likelihood of crash; this, however, does not meaningfully explain the difference in reported crashes. The wealthiest 20% of urban women were a marginally significant 1.9 (95% CI: −3.9 to 0.2) percentage points less likely to report a crash. This operated partially through greater automobile ownership (23.8 percentage points, 95% CI: 18.2–29.4). The relationship between hypothesized direct risk factors and road crash varied by context (Tables 2 and 3). For men, motorcycle ownership and occupational driving were the strongest risk factors and consistently associated with greater crash risk, and not working was associated with lower risk. Alcohol consumption was associated with greater risk in rural but not urban areas for both men and women; the association was only marginally significant for women in the fully adjusted model. Automobile ownership was associated with lower risk only for urban women, and bicycle ownership was associated with greater risk only for urban men. Results were robust to modeling assumptions. While the polychoric PCA-based wealth index explained ~5-fold more variance than the Filmer–Pritchett approach and 25% of households were categorized in different wealth quintiles, the overall relationship between relative wealth and road crash differed little between the approaches (Supplement 1). Differences between linearized and bootstrapped standard errors were negligible (Supplement 1). Discussion Main findings of this study This paper provides evidence that non-fatal road crashes were not commonest among adults in the least wealthy Kenyan households. Rather, crashes were most common in the middle of the national household income distribution for both men and women. However, this inverted-U-shaped pattern resulted from markedly different relationships between relative household wealth and road crashes between rural and urban areas. In rural areas, crashes were least likely in the poorest households—among both men and women. In urban areas, male crashes were lowest among the wealthiest, while they peaked in the middle of the female urban wealth distribution. These patterns are partially explained by differences in occupation and vehicle ownership, especially for male respondents. What is already known on this topic Prior studies have found differing relationships between wealth and road crash. Several studies from SSA find that road crash is more common among those who are relatively poorer, which is consistent with findings for urban men,6,7 while others did not.8,9 Studies from the continent have directly measured economic position relatively rarely, and studies using proxies (such as education and location of residence) have also produced mixed findings.10,12,13 Studies assessing road crash risk as lower-income countries develop have generally found increases in road crash.14–17,51 Additionally, identified mediators of crash are broadly consistent with prior literature, including occupational risk to drivers due to increased transport exposure and financial incentives to operate substandard vehicles, speed and overload vehicles.36–38 Similarly, motorcycle use both tends to be more dangerous than other modes of transport and enables people to travel farther and faster than they otherwise could.34 Limitations This analysis several limitations. First, inferring causality from cross-sectional data is tenuous. There may be instances where road crashes earlier in the recall period caused undetectable changes to wealth and other predictors. Because asset indices produce a long-term measure of economic position, this is a lesser concern than for alternative measures—such as current income—but reverse causation may persist for relatively serious injuries.22,52 Second, fatal crashes could not be captured. More crashes may be fatal among rural and poorer subpopulations, with greater barriers to treatment or a greater likelihood of involvement as vulnerable road users. If so, this paper may overestimate the observed lower crash risk among the rural poor (though an unexpectedly high fraction of crashes would have to be fatal to eliminate observed differences). Conversely, these data may underestimate urban differences. Third, two studies from SSA have found that 12-month recall periods are suboptimal because minor, though not severe, injuries go under-reported with longer recall periods. One study found no association between recall error and any respondent characteristic,53 but the other found that recall errors were more likely for rural respondents.54 This suggests that the road crash levels observed in this study may be underestimates—particularly in rural areas. However, it also suggests that associations between crashes and wealth may not be biased by recall once stratified by location of residence. Fourth, many relevant variables were not available from DHS surveys, such as average daily distances traveled, principal modes of transport and some risk behaviors. In rural areas, it is likely that the poorest travel the least due to transportation costs. In urban areas, the relatively wealthy are least likely to use more dangerous modes of transportation such as matatus (public minibuses).7,55 Similarly, the survey did not distinguish crashes as pedestrians or vehicle occupants. Fifth, women’s risk differences were particularly poorly explained, partially because occupational categories allowed less granular identification of women’s jobs. Street vending—previously been found to be a significant risk factor42,43,45—was not distinguishable from other low-skill sales and services occupations in the survey. Relatedly, vehicle ownership imperfectly represents use, particularly for women, who may be less likely to have control over household vehicles in some settings. Finally, the survey does not ask how many crashes a person experienced over the past 12 months, nor does it ask about the existence or the severity of injuries. As a result, a cumulative incidence of crashes over the past year can be calculated, but the incidence rate of crashes or crash-related injuries cannot be. Similarly, crash risk per kilometer traveled cannot be calculated without additional road-use exposure data. What this study adds This study provides evidence of a complex and nuanced relationship between relative household wealth and non-fatal road crash risk in Kenya. It extends the findings of the Kenyan government and because the sample is larger than previous studies in similar setting, it is able to identify and partially explain differing associations with crash risk in rural and urban areas and between men and women. It suggests that the relationship between wealth and road crash in SSA should be considered differently for various subpopulations and in rural and urban areas. Road crash prevention should remain a priority in developing countries. In particular, as wealth acquisition is associated with greater rural crash risk, it should be a focus of rural development efforts. Supplementary data Supplementary data are available at the Journal of Public Health online. References 1 Institute for Health Metrics and Evaluation . Road Injuries: Both Sexes, All Ages. Vol. 2017; 2017 . 2 Adeloye D , Thompson JY , Akanbi MA et al. . The burden of road traffic crashes, injuries and deaths in Africa: a systematic review and meta-analysis . Bull World Health Organ 2016 ; 94 ( 7 ): 510 – 521A . doi:10.2471/BLT.15.163121 . Google Scholar CrossRef Search ADS PubMed 3 Haagsma JA , Graetz N , Bolliger I et al. . The global burden of injury: incidence, mortality, disability-adjusted life years and time trends from the Global Burden of Disease study 2013 . Inj Prev J Int Soc Child Adolesc Inj Prev 2016 ; 22 ( 1 ): 3 – 18 . doi:10.1136/injuryprev-2015-041616 . Google Scholar CrossRef Search ADS 4 Gosselin RA , Spiegel DA , Coughlin R et al. . Injuries: the neglected burden in developing countries . Bull World Health Organ 2009 ; 87 ( 4 ): 246 – 246a . doi:S0042-96862009000400002 . Google Scholar CrossRef Search ADS PubMed 5 Lagarde E . Road traffic injury is an escalating burden in Africa and deserves proportionate research efforts . PLoS Med 2007 ; 4 ( 6 ): e170 . doi:06-PLME-ES-0915 . Google Scholar CrossRef Search ADS PubMed 6 Azetsop J . Social justice approach to road safety in Kenya: addressing the uneven distribution of road traffic injuries and deaths across population groups . Public Health Ethics 2010 ; 3 ( 2 ): 115 – 127 . doi:10.1093/phe/phq013 . Google Scholar CrossRef Search ADS PubMed 7 Nantulya VM , Muli-Musiime F . Uncovering the social determinants of road traffic accidents in Kenya. In: Evans T , Whitehead M , Diderichsen F , Bhuiya A , Wirth M (eds) . Challenging Inequities: From Ethics to Action . Oxford : Oxford University Press , 2001 , 210 – 25 . Google Scholar CrossRef Search ADS 8 Moshiro C , Heuch I , Astrom AN et al. . Injury morbidity in an urban and a rural area in Tanzania: an epidemiological survey . BMC Public Health 2005 ; 5 : 11 . doi:1471-2458-5-11 . Google Scholar CrossRef Search ADS PubMed 9 Grimm M , Treibich C . Socio-economic determinants of road traffic accident fatalities in low and middle income countries . ISS Work Pap SeriesGeneral Ser 2010 ; 504 : 1 – 44 . 10 Labinjo M , Juillard C , Kobusingye OC et al. . The burden of road traffic injuries in Nigeria: results of a population-based survey . Inj Prev J Int Soc Child Adolesc Inj Prev 2009 ; 15 ( 3 ): 157 – 162 . doi:10.1136/ip.2008.020255 . Google Scholar CrossRef Search ADS 11 Kraemer JD . Urbanization and unintentional injury in low- and middle-income countries. In: Ahn R , Burke TF , McGahan AM (eds) . Innovating for Healthy Urbanization . Springer US : Boston, MA , 2015 : 209 – 244 . doi:10.1007/978-1-4899-7597-3_10 . Google Scholar CrossRef Search ADS 12 Naci H , Chisholm D , Baker TD . Distribution of road traffic deaths by road user group: a global comparison . Inj Prev J Int Soc Child Adolesc Inj Prev 2009 ; 15 ( 1 ): 55 – 59 . doi:10.1136/ip.2008.018721 . Google Scholar CrossRef Search ADS 13 Mock CN , Tiska M , Adu-Ampofo M et al. . Improvements in prehospital trauma care in an African country with no formal emergency medical services . J Trauma 2002 ; 53 ( 1 ): 90 – 97 . Google Scholar CrossRef Search ADS PubMed 14 Wang S-Y , Chi G-B , Jing C-X et al. . Trends in road traffic crashes and associated injury and fatality in the People’s Republic of China, 1951–1999 . Inj Control Saf Promot 2003 ; 10 ( 1–2 ): 83 – 87 . doi:10.1076/icsp.10.1.83.14105 . Google Scholar CrossRef Search ADS PubMed 15 Al-Reesi H , Ganguly SS , Al-Adawi S et al. . Economic growth, motorization, and road traffic injuries in the Sultanate of Oman, 1985–2009 . Traffic Inj Prev 2013 ; 14 ( 3 ): 322 – 328 . doi:10.1080/15389588.2012.694088 . Google Scholar CrossRef Search ADS PubMed 16 Bishai D , Quresh A , James P et al. . National road casualties and economic development . Health Econ 2006 ; 15 ( 1 ): 65 – 81 . doi:10.1002/hec.1020 . Google Scholar CrossRef Search ADS PubMed 17 Paulozzi LJ , Ryan GW , Espitia-Hardeman VE et al. . Economic development’s effect on road transport-related mortality among different types of road users: a cross-sectional international study . Accid Anal Prev 2007 ; 39 ( 3 ): 606 – 617 . doi:S0001-4575(06)00181-3 . Google Scholar CrossRef Search ADS PubMed 18 Kopits E , Cropper M . Traffic fatalities and economic growth . Accid Anal Prev 2005 ; 37 ( 1 ): 169 – 178 . doi:S0001-4575(04)00068-5 . Google Scholar CrossRef Search ADS PubMed 19 Hyder AA , Peden M . Inequality and road-traffic injuries: call for action . Lancet Lond Engl 2003 ; 362 ( 9401 ): 2034 – 2035 . doi:S0140-6736(03)15145-8 . Google Scholar CrossRef Search ADS 20 Kenya National Bureau of Statistics , Ministry of Health, National AIDS Control Council, Kenya Medical Research Institute, National Council for Population and Development, I. C. F. International. Kenya Demographic and Health Survey 2014. 2015 . http://dhsprogram.com/pubs/pdf/FR308/FR308.pdf. 21 I. C. F. International . Sampling and Household Listing Manual: Demographic and Health Surveys Methodology. 2012 . 22 Filmer D , Pritchett LH . Estimating wealth effects without expenditure data–or tears: an application to educational enrollments in states of India . Demography 2001 ; 38 ( 1 ): 115 – 132 . Google Scholar PubMed 23 Rutstein SO. Steps to constructing the new DHS wealth index. Rockv MD ICF Int. 2015 . 24 Rutstein SO . The DHS Wealth Index: Approaches for Rural and Urban Areas. 2008 . http://www.dhsprogram.com/pubs/pdf/WP60/WP60.pdf. 25 Kolenikov S , Angeles G . Socioeconomic status measurement with discrete proxy variables: is principal component analysis a reliable answer? Rev Income Wealth 2009 ; 55 ( 1 ): 128 – 165 . doi:10.1111/j.1475-4991.2008.00309.x . Google Scholar CrossRef Search ADS 26 Rutstein SO , Rojas G Guide to DHS statistics: demographic and health surveys methodology. Meas DHSICF Int. 2006 . 27 Measure DHS/ICF International . Standard Recode Manual for DHS 6. 2013 . http://www.dhsprogram.com/pubs/pdf/DHSG4/Recode6_DHS_22March2013_DHSG4.pdf. 28 International Labour Organization . Inventory of Current Definitions of Rural/Urban. Vol. 2017 . 29 Kenya Ministry of Land , Housing, and Urban Development, Kenya National Bureau of Statistics. 2012/2013 Kenya National Housing Survey Basic Report. 2014 . http://statistics.knbs.or.ke/nada/index.php/catalog/83/download/520. 30 Odero W . Alcohol-related road traffic injuries in Eldoret, Kenya . East Afr Med J 1998 ; 75 ( 12 ): 708 – 711 . Google Scholar PubMed 31 Chalya PL , Mabula JB , Dass RM et al. . Injury characteristics and outcome of road traffic crash victims at Bugando Medical Centre in Northwestern Tanzania . J Trauma Manag Outcomes 2012 ; 6 ( 1 ): 1 . doi:10.1186/1752-2897-6-1 . Google Scholar CrossRef Search ADS PubMed 32 Mogaka EO , Ng’ang’a Z , Oundo J et al. . Factors associated with severity of road traffic injuries, Thika, Kenya . Pan Afr Med J 2011 ; 8 : 20 . Google Scholar PubMed 33 Forjuoh SN , Zwi AB , Mock CN . Injury control in Africa: getting governments to do more . Trop Med Int Health 1998 ; 3 ( 5 ): 349 – 356 . doi:10.1046/j.1365-3156.1998.00240.x . Google Scholar CrossRef Search ADS PubMed 34 World Health Organization . World Report on Road Traffic Injury Prevention. 2004 . http://apps.who.int/iris/bitstream/10665/42871/1/9241562609.pdf. 35 International Labour Organization . ISCO-88. Vol. 2017. 2004 . 36 Habyarimana J , Jack W . Results of a large-scale randomized behavior change intervention on road safety in Kenya . Proc Natl Acad Sci U S A 2015 ; 112 ( 34 ): E4661 – 70 . doi:10.1073/pnas.1422009112 . Google Scholar CrossRef Search ADS PubMed 37 Munala G , Maina K . Fatigue and the road users: travellers’ account! Int J Inj Contr Saf Promot 2010 ; 17 ( 4 ): 257 – 265 . doi:10.1080/17457300.2010.498641 . Google Scholar CrossRef Search ADS PubMed 38 Raynor NJ , Mirzoev T . Understanding road safety in Kenya: views of matatu drivers . Int Health 2014 ; 6 ( 3 ): 242 – 248 . doi:10.1093/inthealth/ihu034 . Google Scholar CrossRef Search ADS PubMed 39 Odero W , Khayesi M , Heda PM . Road traffic injuries in Kenya: magnitude, causes and status of intervention . Inj Control Saf Promot 2003 ; 10 ( 1–2 ): 53 – 61 . doi:10.1076/icsp.10.1.53.14103 . Google Scholar CrossRef Search ADS PubMed 40 Mock CN , Forjuoh SN , Rivara FP . Epidemiology of transport-related injuries in Ghana . Accid Anal Prev 1999 ; 31 ( 4 ): 359 – 370 . doi:S0001-4575(98)00064-5 . Google Scholar CrossRef Search ADS PubMed 41 Grieco M , Turner J , Kwakye EA . Informal public transport and the woman trader in Accra, Ghana. Seventh World Conference on Transport Research. 1995 . https://pdfs.semanticscholar.org/4504/d0e88ec5167922027632907ed89d7fc3ddc3.pdf. 42 Ogunrinola IO , Fadayomi TO , Amoo EO et al. . Occupational Health and Safety for Informal Sector Workers: The Case of Street Traders in Nigeria. Vol. 13; 2012:173–183. http://eprints.covenantuniversity.edu.ng/4110/1/2012-%20IAABD%20Conference%20Proceedings%20(Page%20173-183).pdf#page=186. 43 Ayikai OM . Self-reported Health Status Among Street Vendors in Accra. 2013 . http://ugspace.ug.edu.gh:8080/bitstream/123456789/5867/1/Owusu%20Monica%20Ayikai_Self-Reported%20Health%20Status%20Among%20Street%20Vendors%20in%20Accra_2013.pdf. 44 Kenya National Bureau of Statistics . Kenya—Integrated Household Budget Survey 2005 —KNOCS Codes. 45 Ghaffar A , Hyder AA , Masud TI . The burden of road traffic injuries in developing countries: the 1st national injury survey of Pakistan . Public Health 2004 ; 118 ( 3 ): 211 – 217 . doi:10.1016/j.puhe.2003.05.003 . Google Scholar CrossRef Search ADS PubMed 46 Saidi HS. Risk factor profile and the cost of traffic injury in a tertiary hospital in Kenya. Traffic Safety on Three Continents. Statens väg-och transportforskningsinstitut; 2001 . http://vti.diva-portal.org/smash/get/diva2:673699/FULLTEXT06.pdf. 47 Royston P , Ambler G , Sauerbrei W . The use of fractional polynomials to model continuous risk variables in epidemiology . Int J Epidemiol 1999 ; 28 ( 5 ): 964 – 974 . Google Scholar CrossRef Search ADS PubMed 48 Royston P , Altman DG . Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling . J R Stat Soc C Appl Stat 1994 ; 43 ( 3 ): 429 – 467 . doi:10.2307/2986270 . 49 Xu C , Chen J , Mantell H . Pseudo-likelihood-based Bayesian information criterion for variable selection in survey data . Surv Methodol 2013 ; 39 : 303 – 322 . 50 Wen X , Kleinman K , Gillman MW et al. . Childhood body mass index trajectories: modeling, characterizing, pairwise correlations and socio-demographic predictors of trajectory characteristics . BMC Med Res Methodol 2012 ; 12 : 38 – 2288-12-38 . doi:10.1186/1471-2288-12-38 . Google Scholar CrossRef Search ADS PubMed 51 He H , Paichadze N , Hyder AA et al. . Economic development and road traffic fatalities in Russia: analysis of federal regions 2004-2011 . Inj Epidemiol 2015 ; 2 ( 1 ): 19 . doi:10.1186/s40621-015-0051-6 . Google Scholar CrossRef Search ADS PubMed 52 Rutstein SO , Johnson K . The DHS wealth index. DHS comparative reports no. 6. Calverton ORC Macro. 2004 . 53 Mock C , Acheampong F , Adjei S et al. . The effect of recall on estimation of incidence rates for injury in Ghana . Int J Epidemiol 1999 ; 28 ( 4 ): 750 – 755 . Google Scholar CrossRef Search ADS PubMed 54 Moshiro C , Heuch I , Astrom AN et al. . Effect of recall on estimation of non-fatal injury rates: a community based study in Tanzania . Inj Prev J Int Soc Child Adolesc Inj Prev 2005 ; 11 ( 1 ): 48 – 52 . doi:11/1/48 . Google Scholar CrossRef Search ADS 55 Mutongi Kenda . Matatu: A History of Popular Transportation in Nairobi. 2017 . © The Author(s) 2018. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Journal of Public HealthOxford University Press

Published: May 18, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off