A national survey of physical activity after spinal cord injury

Physical activity is a powerful modifiable risk factor for disease and mortality. Physical activity levels in people with spinal cord injury (SCI) have not been quantified relative to uninjured individuals in a large population-based sample. We aimed to quantify and compare physical activity in people with and without SCI, and to examine the associations between physical activity, lifestyle, and socioeconomic factors. The 2010 Canadian Community Health Survey (n > 57,000) was used, which includes three measures that assess physical activity levels (i.e., leisure time activity frequency, leisure time activity intensity, and transportation time activity intensity). Bivariable and multivariable logistic regressions were performed and odds ratios (ORs) with corresponding 95% confidence intervals (CIs) were estimated. The odds of physical activity in people with SCI were 0.43 (95% CI 0.3–0.61), 0.53 (95% CI 0.36–0.75), and 0.42 (95% CI 0.28–0.61), across the three measures of physical activity, respectively. These differences persisted after adjustment for lifestyle, comorbidities, and socioeconomic factors. Physical activity is reduced in the SCI population compared with the general population. This knowledge is important to direct future research and guide the allocation of health care resources.

Individuals with SCI are at greater risk of physical inactivity compared to able-bodied individuals. In individuals with SCI relative to able-bodied individuals, the unadjusted OR for leisure time activity frequency was 0.35 (95% CI 0.27-0.44) ( Table 3). After adjusting for age, sex, and BMI the OR slightly increased to 0.39 (95% CI 0.30-0.50) (Table 3). Additionally, even after inclusion of all these variables, the results remained significant (adjusted OR 0.43; 95% CI 0.30-0.61; ROC 0.69). In individuals with SCI relative to able-bodied individuals, the unadjusted OR for leisure time activity intensity was 0.49 (95% CI 0.38-0.63) ( Table 3). After adjusting for age, sex, and BMI the OR slightly increased to 0.55 (95% CI 0.43-0.71) (Table 3). Additionally, even after inclusion of all these variables, the results remained significant (adjusted OR 0.53; 95% CI 0.36-0.75; ROC 0.64). In individuals with SCI relative to able-bodied individuals, the unadjusted OR for transportation time activity intensity was 0.38 (95% CI 0.29-0.50) ( Table 3). After adjusting for age, sex, and BMI the OR slightly increased to 0.43 (95% CI 0.33-0.57). Additionally, even after inclusion of all these variables, the results remained significant (adjusted OR 0.42; 95% CI 0.28-0.61; ROC 0.65) ( Table 3). In other words, all three elements demonstrated converging results that those with SCI participate in less physical activity compared to people without SCI. These results remained significant after adjusting for age, sex, and BMI, as well as including all potential confounders (Tables 4 and 5).
The interaction between physical activity with lifestyle and socioeconomic factors in people with and without SCI. In people with SCI, there was no association between physical activity level and sex, body mass index, migraines, or smoking. This is in contrast to people without SCI where male sex, lower body mass index, not smoking, and absence of migraines were associated with increased physical activity levels ( Fig. 1). In both those with and without SCI, there was an association between increased physical activity levels and not being diagnosed with hypertension, eating more fruits and vegetables, greater household income, greater education levels, better mental health, lower stress levels, reduced likelihood of mood disorders, and reduced anxiety (Fig. 1). For details on lifestyle and socioeconomic factors, see Tables 4 and 5.

Discussion
On a population-scale, physical activity levels are reduced in people with SCI compared to uninjured individuals. This result persisted after adjustment for age, sex, BMI, as well as numerous lifestyle and socio-economic factors. Lower-income, education, and fruit and vegetable consumption and increased alcohol use were associated www.nature.com/scientificreports/ with reduced physical activity in the SCI population. As such, specific socio-economic groups within the SCI population may benefit most from physical activity promotion activities.
Physical activity levels are reduced in those with SCI. Reduced physical activity in people with SCI is likely the result of numerous physical factors, including loss of skeletal muscle control, reduced skeletal muscle mass, reduced cardiovascular reserve, as well as environmental and social barriers 23,24 . Moreover, other psychological and behavioural barriers to engagement in physical activity include a lack of knowledge, community support, beliefs in ability, coping, conflicting goals for rehabilitation and limited access to disability-related experts and accessible rehabilitation infrastructures [25][26][27] . Reduced physical activity plausibly contributes to widespread cardiometabolic disorders after SCI, increased cardiovascular disease and diabetes risk, and a range of physical, emotional, and mental health issues that affect people with SCI 1,3,28-30 . As such, it is essential that targeted physical activity interventions for individuals with SCI integrate physical, psychological and behavioural based approaches to support effective uptake of interventional strategies and ensure the mitigation of these risk factors. Furthermore, strategies to increase physical activity may be more successful if they are integrated into adapted activities and/or those that directly align with an individual's goals (i.e., gardening, walking a dog, cycling, resistance training, yoga etc.) 31 . Previous research has shown that approximately half of people with SCI report no leisure-time physical activity. However, it is not clear if this proportion differs from non-SCI populations, who also report very high rates of physical inactivity 8,9 . Another study compared physical activity levels in 40 people with SCI to age-matched uninjured controls, showing that individuals with SCI have reduced durations of dynamic activity one year after discharge from rehabilitation 32 . Our data provide additional support for these findings on a population-scale with control individuals.
There is a unique profile of lifestyle factors associated with physical activity levels within the SCI population. Within the SCI population there is no association between male sex, reduced body mass index, less likelihood of migraines, and not smoking with physical activity levels. This may be due to a variety of factors that are outside the scope of this study to quantitatively evaluate. Some of these may include the interaction between gender and self-efficacy before and after SCI [33][34][35] , as well as the interaction between body mass index and severity of disability after SCI 36 .
In the present data, greater physical activity was associated with an improved lifestyle. Greater physical activity levels in people with SCI were also associated with a reduced likelihood of being diagnosed with hypertension, anxiety and mood disorders, better mental health, and lower self-reported stress (Fig. 1). Greater household income and education, as well as lower alcohol consumption and eating more fruits and vegetables, were also factors associated with increased physical activity levels. Lower education and household income in the general population have also been associated with low physical activity levels 37 . It is reasonable to expect that these factors, in combination with the consequences of SCI, may be further exaggerated in the SCI population. These specific demographic groups within the SCI population (those with lower education and household incomes), Table 2. Characteristics of the population-based survey by physical activity status following adjustments to confounders. Data are sample sizes (percentages). Percentages are probability-weighted. SCI spinal cord injury, BMI body mass index. *The total sample size of the participants who responded to all the confounding variables.  www.nature.com/scientificreports/ should therefore be precision-targeted to understand exercise barriers, potential education programs needed and ultimately establish effective interventional strategies that promote physical activity. For example, previous public health interventions have promoted active transportation involving human energy to mobilize and travel. These programs have demonstrated success as they have been established by creating safe environments (e.g., improving community landscape, adding more sidewalks, longer pedestrian crossing times on signal lights) for the general population including pedestrians, bikers, as well as wheelchair users 38,39 . Interventional programs such as these provide equal opportunities for individuals with varying socioeconomic backgrounds to access transportation options and provide an example of an effective strategy that integrated the physical, behavioural, and social needs of multiple demographics to encourage uptake of a desired outcome for a broad population.

Limitations.
A primary strength of this study was the use of the CCHS databases, as the sample selected is designed to be representative of the Canadian population (~ 34 million adults), and therefore the data is considered highly generalizable 40 . Furthermore, the ~ 330 respondents with SCI represent ~ 0.4% of the population of Canadians living with SCI 41 . It is also unlikely that our results suffered from response bias as is expected in a single topic survey when respondents would potentially aim to answer questions in the style that the interviewers prefer. Although self-reported physical activity levels in people with SCI are subjective, they do relate to objective physical activity measures 42 . However, the specific questions asked in the CCHS have not been assessed in comparison to objective measures. The CCHS data are derived from a cross-sectional study design and it is therefore not possible to determine the causality between variables. It is possible that misclassification occurred in terms of level and severity of injury. This would be most likely for individuals with less severe SCI, who would be expected to participate in physical activity more frequently than those with higher more complete SCI 43 . Therefore, including individuals with lower severity scores may result in an underestimation of the reported effect size.
Frequency of all leisure time physical activity lasting over 15 min. To capture "leisure time activity frequency", we used the PACDFR variable from the CCHS. This variable classifies respondents as having "regular practice of leisure time activities", "occasional practice of leisure time activities" and "infrequent practice of leisure time activities" lasting over 15 min based on the monthly frequency of physical activity reported for a three-month period. The questions are described in detail here 22,45 . Responses for PACDFR were binarized (Table 6).
Leisure time physical activity index. To capture "leisure time activity intensity", we used the PACDPAI variable from the CCHS. This variable categorizes respondents as being "active", "moderately active", or "inactive" in their leisure time based on the reported total daily Energy Expenditure values (kcal/kg/day) during the past three months. The questions are described in detail here 22,45 . Responses for PACDPAI were binarized (Table 6).
Transportation and leisure time physical activity index. To capture "transportation time activity intensity", we used the PACDLTI variable from the CCHS. This variable categorizes respondents as being "active", "moderately active", or "inactive" in their transportation and leisure time based on the average daily energy expended (kcal/ kg/day) during transportation and leisure-time physical activities by the respondent in the past three months. The questions are described in detail here 22,45 . Responses for PACDLTI were binarized (Table 6). For details on physical activity level variables, see Table 6.
Comorbidities. Previous diagnosis of hypertension (CCC_072) was obtained with the following question: "Have you ever been diagnosed with high blood pressure?" Migraine status (CCC_081) was obtained with the following question: "Do you have migraine headaches?" Previous diagnosis of mood disorders (CCC_280) was obtained with the following question: "Do you have a mood disorder such as depression, bipolar disorder, mania or dysthymia?" Previous diagnosis of anxiety disorders (CCC_290) was obtained with the following question: "Do you have an anxiety disorder such as phobia, obsessive-compulsive disorder or a panic disorder?" An individual could provide a "Yes" or "No" answer to the aforementioned questions. The questions are described in detail here 22,45 . Lifestyle and socioeconomic factors. Smoking status (SMKDSTY) indicates the type of smoker the respondent is. The questions are described in detail here 22,45 . The possible answers for SMKDSTY are "Daily smoker", "Occasional smoker (former daily smoker)", "Occasional smoker (never a daily smoker or has smoked less than 100 cigarettes lifetime)", "Former daily smoker (non-smoker now)", "Former occasional smoker (at  www.nature.com/scientificreports/ least 1 whole cigarette, non-smoker now)", "Never smoked (a whole cigarette)", or "At least one required question was not answered (don't know, refusal, not stated)". Responses for SMKDSTY were binarized (Table 5). Alcohol consumption status (ALCDTTM) indicates the type of drinker the respondent is for the past 12 months. The questions are described in detail here 22,45 . The possible answers for ALCDTTM are "Regular drinker", "Occasional drinker", "Did not drink in the last 12 months", or "At least one required question was not answered (don't know, refusal, not stated)". Responses for ALCDTTM were binarized (Table 5).
Fruit and vegetable consumption (FVCGTOT) was obtained based on the derived variable FVCDTOT (indicates the total number of times per day the respondent consumes fruits and vegetables [i.e., fruit juice, fruits, green salad, potatoes, and carrots]). The questions are described in detail here 22,45 . The possible answers for FVCGTOT are "Eats fruits and vegetables less than 5 times per day", "Eats fruits and vegetables between 5 and 10 times per day", Eats fruits and vegetables more than 10 times per day", or "At least one required question was not answered (don't know, refusal, not stated)". Responses for FVCGTOT were binarized (Table 5).
Self-perceived mental health (GEN_02B) was obtained with the question: "In general, would you say your mental health is: …excellent?, …very good?, …good?, …fair?, …poor? ", or "At least one required question was not answered (don't know, refusal, not stated)". Self-perceived life stress (GEN_07) was obtained with the question: "Thinking about the amount of stress in your life, would you say that most days are: …not at all stressful?, …not very stressful?, …a bit stressful?, …quite a bit stressful?, or …extremely stressful?", or "At least one required question was not answered (don't know, refusal, not stated)". Responses for GEN_02B and GEN_07 were binarized (Table 5).
The highest level of education attained within the household (EDUDH04) is based on the highest level of education for each member of the household (EDUDR04). The questions are described in detail here 22,45 . The possible answers are "Less than secondary school graduation", "Secondary school graduation, no post-secondary education", "Some post-secondary education", "Post-secondary degree/diploma", or "At least one required question was not answered (don't know, refusal, not stated)". Responses for EDUDH04 were binarized (Table 5).
Statistical analysis. Logistic regression models were obtained separately for the binary outcome physical activity levels with SCI as the main explanatory variable, and with lifestyle and socio-economic factors as the main explanatory variable. Models were probability weighted to account for the clustering and stratification sampling design used by the CCHS (as previously reported) 2,46 . Separate logistic regression models were generated for the physical activity outcomes using the 'glm' (generalized linear model) function with the family argument set to 'binomial()' from the R Statistical Software package 'stats' . R (R Core Team, 2017) was used for all statistical analyses. Using the logistic models, unadjusted and adjusted odds ratios (ORs) with 95% confidence intervals are presented. Goodness of fit for the full model was assessed using a receiver-operating curve (ROC). The ORs were then adjusted for potential confounders using multivariable logistic regression. In the multivariable model, age, sex, and body mass index were input as additional explanatory variables to calculate the adjusted model (AOR). The sensitivity analysis included the lifestyle and socioeconomic variables described above. A fully adjusted model (AOR2) including all these potential explanatory variables is then presented. Statistical significance was defined as a p-value ≤ 0.05. Data are presented in accordance with the STROBE guidelines of reporting 47 .