The Relationship of Chronological Age, Age at Injury, and Duration of Injury to Employment Status in Individuals with Spinal Cord Injury

Abstract

Employment status following spinal cord injury (SCI) has important implications for financial and psychosocial well-being. Several age-related variables—in particular chronological age, duration of SCI, and age at SCI onset—have been identified as being associated with employment among individuals with SCI. Cross-sectional investigations of this topic are complicated by methodological and statistical issues associated with aging and disability. The purpose of the current study was to examine the associations between three aging variables and employment status in individuals with SCI through a series of regression analyses. Six hundred twenty individuals with SCI completed a survey that included measures of demographic characteristics, pain, psychological functioning, physical functioning, fatigue, and sleep. The results indicated that chronological age and age at SCI onset were significant predictors of employment status. A significantly greater proportion of individuals aged 45–54 were employed compared to those aged 55–64 even after controlling for biopsychosocial variables. Additionally, there was a negative linear relationship between percent employed and age at SCI onset, and this relationship was not accounted for by the biopsychosocial variables. The analyses used in this study provide one method by which to disentangle the effects of different age-related variables on important SCI outcomes in cross-sectional research. Continued research in this area is needed to better understand age-related effects on employment status, which could be used to help maximize the quality of life in individuals with SCI.

Introduction

Spinal cord injury (SCI) is associated with considerable costs related to medical treatment, attendant care, and equipment. The indirect costs associated with reduced employment can also be substantial. Employment rates following SCI range from 11.5% to 74% (Lidal et al. 2007), and lost earnings can exceed $30,000 per year (DeVivo et al. 1995). The financial consequences of reduced employment following SCI may be compounded by psychosocial costs as well since gainful employment is related to greater life satisfaction and better overall health (Hess et al. 2004; Krause 1992; Krause and Anson 1997).

A number of factors have been found to be associated with employment status following SCI, including higher levels of education (Anderson et al. 2007; Castle 1994; DeVivo et al. 1987; Dijkers et al. 1995; El Ghahit and Hanson 1979; Goldberg and Freed 1982; Kewman and Forchheimer 1998; Krause and Anson 1997; McShane and Karp 1993; Noreau and Shephard 1992), less severe injury (Anderson et al. 2007; Dijkers et al. 1995; El Ghahit and Hanson 1978; Geisler et al. 1966; Krause et al. 1999; Stover and Fine 1986), and white race (Anderson et al. 2007; DeVivo and Fine 1982; Devivo et al. 1987; James et al. 1993; Krause and Anson 1997; Krause et al. 1998, 1999). Several age-related variables have also emerged as important considerations in this context. Younger chronological age and younger age at injury onset have both been found to be positively associated with employment following SCI (Athanasou et al. 1996; Castle 1994; Dijkers et al. 1995; DeVivo and Fine 1982; DeVivo et al. 1987; El Ghahit and Hanson 1978; Kemp and Vash 1971; Krause and Anson 1997; Krause et al. 1999; Noreau and Shephard 1992). There is also evidence that the longer an individual has lived with an SCI (i.e., duration of injury), the more likely that person is to be employed (Krause 1992; Krause and Anson 1996; Krause and Broderick 2005; Krause and Coker 2006; Krause et al. 1998, 1999; Meade et al. 2004; Pflaum et al. 2006).

Continued investigation of the relationships between these aging variables and employment following SCI is important for several reasons. Unlike many other neurologic disorders such as stroke, new incidents of SCI often involve young adults at a critical stage for vocation-related decisions and activities (National Spinal Cord Injury Statistical Center 2008). In addition, because individuals with SCI, relative to those without SCI, experience age-related functional declines at an increased rate (Bauman and Waters 2004; Capoor and Stein 2005), their ability to maintain gainful employment over time and as they age may be compromised. These functional declines are compounded by the fact that the longer an individual has lived with an SCI—particularly for those 20 or more years post-injury—the greater the functional decline (Capoor and Stein 2005). Moreover, the relationship between functional capacity and these aging variables may be nonlinear, as indicated by research showing that the rate of decline increases over the lifespan in individuals with SCI (Capoor and Stein 2005). Taken together, these findings suggest a complex relationship between employment and age-related variables in individuals with SCI.

Unfortunately, conclusions about the relative importance of aging variables in the context of SCI are constrained by the cross-sectional nature of much of this research and the statistical overlap among age-related variables (Krause and Adkins 2004). The three age-related variables most often considered are chronological age, age at SCI onset, and duration of injury. When assessed cross-sectionally, individuals with SCI who are older are more likely to both (1) have a longer duration of injury and (2) have been older when they were injured compared to their younger counterparts. Consequently, if chronological age emerges as a significant predictor in statistical analyses, it is not possible to draw conclusions about the unique effect of this variable unless the other two aging variables (age at SCI onset and duration of injury) are also controlled. At the same time, however, because these three aging variables are linearly related, all three cannot be included in the same analysis, which limits our ability to distill their unique influence on the phenomenon of interest in a single analysis.

In a recent manuscript (Jensen et al. 2009), we presented an analytic approach that addresses these aging issues in cross-sectional research on sleep problems among individuals with SCI. This approach (see below) consists of a series of analyses in which each of the three aging variables is systematically examined. The results from this approach as applied to sleep data indicated that chronological age, and not SCI duration or age at SCI, was the primary aging-related variable that contributed to the prediction of sleep problems (Jensen et al. 2009). These results have important clinical implications because they provide clinicians with knowledge about when to be particularly alert for increased sleep difficulties (i.e., at a certain chronological age, as opposed to a certain duration following injury or among those injured at a certain age) and also when preventative interventions for sleep problems might be most effective. Similarly, although age-related variables themselves cannot be affected by clinical interventions, knowing which variable(s) is most closely linked to return to work following SCI allows clinicians to determine when and to whom measures that support continued work might be most effectively provided.

The purpose of the current paper was to apply the analytic strategy used successfully to tease out the relative importance of chronological age, SCI duration, and age at SCI as they predict sleep difficulties to a cross-sectional investigation of the effects of age-related variables on employment status following SCI. Although employment among individuals with SCI has received extensive empirical attention to date, the current study makes an important contribution to this line of research in that the analytic approach used offers, for the first time, an evaluation of the unique contribution of each age-related variable while taking into account (controlling for) the effects of the other age-related variables.

Based on the extant literature, we hypothesized that older chronological age, older age at SCI onset, and shorter injury duration would be associated with poorer employment outcomes in individuals with SCI. We also investigated both linear and quadratic relationships among these aging variables and employment status since the magnitude of these relationships may change in an exponential (i.e., nonlinear) manner across age groups and because quadratic relationships are hypothesized for most areas of decline with age in SCI models (Capoor and Stein 2005).

Materials and Methods

Participant Recruitment Procedures

Participants were drawn from a sample of individuals who completed a survey designed to provide data that could be used to improve the measurement of pain, fatigue, and other health symptoms in persons with disabilities. The participants were recruited through the University of Washington (UW) in Seattle, Washington, the Shepherd Center at the Virginia Crawford Research Institute in Atlanta, Georgia, and by means of web and print advertisements. Eligibility requirements for the current survey study included self-report of a SCI and age of at least 18 years.

Invitation letters were mailed to 518 individuals from the UW and 1,890 individuals from the Shepherd Center. Those who did not respond within 1 month were sent a follow-up letter. Self-report surveys were mailed to the 812 individuals from these centers, as well as the 25 individuals who responded to advertisements, who expressed interest in participating. Those who had not returned the survey after 1 month were sent a reminder letter and another copy of the survey. If the survey had not been received by the study office within 2 weeks of sending the reminder letter, a telephone call was made to the participant. At that point, individuals recruited from the Shepherd Center and via advertisement were given the option to complete the survey online. This option was not available to those recruited from the University of Washington. Eight participants completed the survey online.

A total of 620 participants (Shepherd, 404; UW, 210; advertisements, 6) completed and returned the survey (74.1% response rate for individuals who were mailed surveys and 25.7% response rate for individuals who were invited to participate). Twenty-five dollars was paid to all participants except those recruited via web and print advertisements. The study procedures were approved by the University of Washington Human Subjects Review Committee.

Measures

Demographic and SCI-Related Characteristics

Participants provided information about their sex, age, race and ethnicity, education level, marital status, and employment status. Clinical information was elicited about the duration of SCI, age at SCI onset, SCI level(s), and completeness of injury.

Pain Severity

Pain severity was assessed with the bodily pain item of the eight-item version of the Short-Form Health Survey (SF-8; Ware et al. 2001). This item asks respondents “How much bodily pain have you had during the past 4 weeks?” Responses are indicated on a six-point scale with endpoints at 1 (none) and 6 (very severe). The SF-8 has demonstrated good reliability and validity as a measure (Ware et al. 2001), and the bodily pain item has been used in previous research on pain and functioning (e.g., Chmielewski et al. 2008; Turner-Bowker et al. 2003).

Psychological Functioning

Psychological functioning was assessed with the ten-item version of the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff 1977; Andreson et al. 1994). This scale asks respondents questions about their perceived mental health status over the preceding week on a scale with endpoints at 0 (Rarely or none of the time) and 3 (Most or all of the time). A total score is calculated by summing each of the items, with higher scores indicating greater depressive symptomatology. The ten-item version of the CES-D has demonstrated strong psychometric properties in previous research (Andreson et al. 1994).

Physical Functioning

Physical functioning was assessed with the Physical Component Summary (PCS) score of the eight-item version of the SF-8 (Ware et al. 2001). The SF-8 asks respondents to rate their health during the past 4 weeks across a variety of physical and emotional health domains. The PCS is calculated from the individual items according to standardized procedures and scored on a 0–100 scale, with higher scores indicating better physical functioning. The SF-8 PCS is well-correlated with the PCS of the full 36-item measure (SF-36) and has demonstrated adequate reliability and validity for use in survey research (Ware et al. 2001).

Fatigue

Fatigue was assessed with the Fatigue Severity Scale (FSS; Krupp et al. 1989). The FSS asks respondents to rate the extent to which they experienced each of nine fatigue symptoms during the past week. Responses are provided on a scale from 1 (Completely disagree) to 7 (Completely agree). The FSS total score is the mean of the nine items, with higher scores representing greater fatigue. The FSS has been widely used in research on persons with disabilities, including SCI, and has demonstrated good reliability and validity in these groups (Anton et al. 2008; Herlofson and Larsen 2002; Kleinman et al. 2000; Krupp et al. 1989; Schepers et al. 2006; Vasconcelos et al. 2006).

Sleep

Sleep quality was assessed with the nine-item Sleep Problems Index (SLP9) from the Medical Outcomes Sleep Study scale (Hays and Stewart 1992). The overall scale contains 12 items that assess respondents’ sleep quality during the past 4 weeks. Ten items are rated on a six-level frequency scale, ranging from “None of the time” to “All of the time.” Sleep latency is assessed on a five-point scale from “0–15 minutes” to “more than 60 minutes.” The remaining item asks respondents to indicate the average number of hours slept per night during the past 4 weeks. The SLP9 comprised nine items that are scored and standardized on a 0–100 scale, with higher scores indicating more sleep problems. The psychometric properties of the SLP9 have been established in large samples (Hays and Stewart 1992; Spritzer and Hays 2003) and across a wide spectrum of ages, including the elderly (Rejas et al. 2007; Vernon et al. 2008).

Analyses

First, the survey response rate and respondents’ demographic and SCI history information were examined for descriptive purposes. Next, the univariate associations between demographic and SCI-related variables (injury level, completeness, and etiology) and the primary outcome of interest (employment status; see below) were examined to determine the need to control for any potential confounding variables in the planned regression analyses. Race and education were the only demographic variables that were significantly related to employment status. A greater percentage of white participants (34%) than non-white participants (13%) were currently employed full- or part-time (p<0.001, two-sided Fisher’s exact test). In addition, a greater percentage of participants who had attended and/or graduated from college (37%) were currently employed (p<0.001, two-sided Fisher’s exact test) compared to those who had no college-level education (14%). No SCI-related variables demonstrated significant univariate associations with employment status. Thus, participant race (white, non-white) and education (college, no college) were entered in the first step of the regressions, as described below. The associations between the three aging variables were then examined with correlation analyses.

Next, three logistic regression analyses were conducted to examine the unique contributions of chronological age, duration of injury, and age at SCI onset to the prediction of employment status. Participants’ current employment status (employed full- or part-time vs. not employed) served as the criterion variable for these analyses. Consistent with previous studies (e.g., Krause et al. 1998, 1999), participants who reported that they were currently working any number of hours per week were considered to be “employed.” Chronological age, duration of injury, and age at SCI onset were the primary predictors and were entered in the regression models as continuous variables. Linear and quadratic terms were included in these models in order to examine both linear and quadratic relationships between the aging variables and employment status. These predictor variables were centered prior to inclusion in the regression models in order to account for the increased multicollinearity that is associated with testing both linear and quadratic relationships.

Should a significant quadratic term emerge from the regression analyses, we planned to then categorize the age-related variable into discrete age groups for further analysis using groups delineated in decade (years) increments: <25, 25–34, 35–44, 45–54, 55–64, 65–74, and 75+. These categories are consistent with those used in our previous paper (Jensen et al. 2009) and were adopted for the following reasons: (1) they provide sufficient detail to examine nonlinear relationships; (2) they are large enough to include adequate numbers of participants to yield reliable statistical estimates; and (3) they represent a subset of the standard cutoffs used in research on aging and disability (e.g., <25, 25–34, and 35–44 all fall into the standard “young adult” range). Follow-up frequency and chi-square analyses were then planned to further characterize the nature of any nonlinear relationships identified in the regression analyses. Follow-up regression analyses were also planned to determine the extent to which any significant aging–employment relationships are explained by biopsychosocial variables that are thought to be important considerations in this context; the specific biopsychosocial variables we planned to examine included pain severity, physical functioning, fatigue, sleep, and psychological functioning.

Regression Considerations

All three primary age variables cannot be meaningfully examined in the same regression model due to their linear associations (Krause and Adkins 2004). That is, if all three age variables were included in the same analyses, the variable entered last would necessarily contain information about two age variables. For example, in a regression model that included chronological age in the first step, followed by age at SCI in the second step, when chronological age is controlled, age at SCI would contain information about both age at injury and injury duration (Cohen 1988). The reason for this is that injury duration can be computed from chronological age and age at SCI (injury duration = chronological age − age at SCI). The statistical implication of this fact is that after controlling for chronological age, entering either age at SCI or injury duration in the second step of the regression produces the same change in the R² value.

Given this issue, a series of three regression analyses were performed in order to determine the unique variance that each variable contributes to the prediction of the criterion (Jensen et al. 2009). In these analyses, each age variable was entered either first or second (e.g., chronological age followed by injury duration, injury duration followed by onset age, and onset age followed by chronological age). The overall pattern of results from this analytic approach allows for determination of whether the three aging variables are uniquely related to employment status. For example, if injury duration is a significant and unique (i.e., when controlling for the other aging variables) predictor, then this variable will be significant when entered prior to the other aging variables and also significant (1) when entered after either of the other two variables and (2) when a variable containing duration information is entered after one of the other aging variables.

Testing for Nonlinear Relationships

The analysis plan to examine any significant nonlinear relationships that emerged in the regression analyses was similar to that for significant linear relationships. In the analytic series outlined above, the linear terms were entered prior to the quadratic terms in their respective models. For example, in the first regression, after controlling for respondent race and education in step 1, chronological age was entered in step 2, followed by injury duration in step 3. Chronological age and injury duration are linear terms and test for the linear relationships between these variables and employment status. The nonlinear relationships are tested in subsequent steps. Continuing with the previous example, the quadratic term for chronological age (chronological age squared) was entered in step 4, followed by the quadratic term for injury duration (injury duration squared) in step 5. In this way, it is possible to determine if a nonlinear relationship among the predictors and employment status is significant after controlling for the linear relationships. Follow-up analyses, as described above, were planned in the event of a significant quadratic relationship.

Summary of Regression Approach

The logistic regression approach for this study (1) controlled for relevant baseline variables, (2) determined the unique contributions of chronological age, injury duration, and age at SCI onset to the prediction of employment status, and (3) examined nonlinear relationships among these three aging variables and employment status. The first regression included respondent race and education in the first step, followed by chronological age (linear) in the second step, injury duration (linear) in the third step, chronological age (quadratic) in the fourth step, and injury duration (quadratic) in the last step. The second regression equation consisted of race and education in step 1, injury duration in steps 2 and 4, and age at SCI in steps 3 and 5. The third regression analysis included race and education in step 1, age at SCI in steps 2 and 4, and chronological age in steps 3 and 5.

Results

Response Rate and Participant Characteristics

The 837 individuals who expressed interest in participating in this research were mailed surveys or invited to participate via the web. Of these, 620 participants completed the survey (paper: n=612; web: n=8), which is 25.7% of those who were invited to participate and 74.1% of those who expressed initial interest. The average age of participants in the final sample was 44.9 years (SD=14.3; range, 18–84). The number of years since SCI onset ranged from 1 to 56 years (mean=11.6; SD=10.2), with 86% of participants reporting that their injury occurred less than 25 years ago. The average age of SCI onset was 34.0 years (SD=14.3; range, 9–79). The three aging variables were moderately to strongly related. Chronological age was positively correlated with injury duration (r=0.35, p<0.001) and age of SCI onset (r=0.74, p<0.001), whereas injury duration and age of SCI onset were negatively correlated (r=− 0.37, p<0.001).

The most frequent single cause of SCI was reported to be a motor vehicle crash (48%), followed by a fall (18%), sports injury (7%), gunshot wound (6%), or diving (4%). The percentages of participants reporting their highest level of injury were as follows (1% of participants had missing data): C1-4, 21%; C5-8, 27%; T1-5, 11%; T6-12, 33%; and L1-S4/5, 7%. Injuries were described as incomplete by 52% of participants; 41% reported a complete injury, and 8% indicated that they did not know if their injury was complete or incomplete.

Over two thirds of participants (67%) were male. The self-reported racial categories were as follows: 79% white, 11% black, 4% Hispanic, 2% Asian or Pacific Islander, 1% Native American, and 3% mixed heritage. For statistical purposes, these racial categories were collapsed into white (79% of participants) and non-white (21% of participants). Approximately 49% of participants were currently married or living with a partner in a committed relationship. In terms of education, 28% of participants reportedly had some college education, 25% were college graduates, and 14% had graduate degrees. The remainder reported education at the level of vocational/technical school (8%), high school diploma/equivalent only (19%), and less than high school diploma/equivalent (6%). For statistical purposes, participants were grouped based on whether they had attended and/or graduated from college (67% of participants) or not (33% of participants). In regards to current employment status, 30% of participants indicated that they were employed full- or part-time, and the remainder were not employed at time of survey completion. Table 1 contains descriptive information on participants across employment status.

Table 1.

Demographic and clinical characteristics across employment status

	Employed Mean (SD) or % (N)	Not employed Mean (SD) or % (N)
Demographic variables
Sex
Male	62% (114)	68% (298)
Female	38% (69)	32% (139)
Race
White	91% (167)	74% (325)
Not white	9% (16)	26% (112)
Marital status
Married/Cohabitating	46% (84)	47% (204)
Not married/Cohabitating	54% (97)	53% (232)
Education
College	84% (153)	60% (260)
No college	16% (30)	40% (177)
Age	43.25 (12.35)	45.65 (14.98)
Injury duration	12.53 (9.91)	11.22 (10.30)
Age at SCI onset	31.49 (12.95)	35.10 (14.76)
Clinical variables
Cause of injury
Gunshot wound	4% (7)	6% (28)
Fall	15% (28)	19% (82)
Motor vehicle crash	50% (91)	47% (205)
Sports injury	10% (19)	6% (25)
Diving	6% (10)	3% (15)
Other	15% (28)	19% (81)
Injury level
L1-S4	9% (16)	7% (30)
T6-T12	37% (68)	31% (136)
T1-T5	13% (23)	11% (46)
C5-C8	24% (43)	28% (123)
C1-C4	17% (31)	22% (97)
Injury completeness
Complete	41% (75)	41% (177)
Incomplete	54% (98)	51% (222)
Don’t know	6% (10)	9% (37)
Pain severity	4.81 (2.25)	5.67 (2.16)
Physical functioning	43.94 (8.71)	36.63 (9.58)
Fatigue	3.83 (1.49)	4.38 (1.61)
Sleep	34.30 (18.30)	40.63 (21.27)
Psychological functioning	6.82 (5.70)	9.79 (6.43)

Pain severity = SF-8 bodily pain item; physical functioning = SF-8 PCS; fatigue = FSS total score; sleep = SLP9 score; psychological functioning = CES-D total score

Relative Contributions of Chronological Age, Duration of Injury, and Age at SCI Onset to the Prediction of Employment Status

As described above, three logistic regression analyses were conducted with employment status (employed full- or part-time vs. not employed) as the criterion in each. Detailed results of these analyses are presented in Table 2. Results from the first regression indicated that after controlling for participant race and education, chronological age (linear) made a significant addition to the model (Δχ² =7.63, p< 0.01). Injury duration (linear) accounted for significant, unique variance in the next step (Δχ² =3.01, p=0.08). Chronological age (quadratic) was a significant contributor in the fourth step (Δχ²=11.34, p<0.01); however, injury duration (quadratic) was not significant in the final step.

Table 2.

Logistic regression analyses predicting employment status from race, education, current age, injury duration, and age at SCI onset

Step and variables	χ2 change	Exp(B)	95% C.I. for Exp(B)
Step 1: Race	57.38****	0.30****	0.170–0.526
Education		0.31****	0.196–0.475
Step 2: Age (linear)	7.63***	0.98***	0.969–0.995
Step 3: Duration (linear)	3.01*	1.02*	0.998–1.036
Step 4: Age (quadratic)	11.34***	1.00***	0.998–0.999
Step 5: Duration (quadratic)	1.88	1.00	0.997–1.001
Step 1: Race	57.38****	0.30****	0.170–0.526
Education		0.31****	0.196–0.475
Step 2: Duration (linear)	0.42	1.01	0.989–1.023
Step 3: Age at SCI (linear)	9.43***	0.98***	0.965–0.992
Step 4: Duration (quadratic)	5.10**	1.00**	0.997–1.000
Step 5: Age at SCI (quadratic)	1.34	1.00	0.999–1.000
Step 1: Race	57.38****	0.30****	0.170–0.526
Education		0.31****	0.196–0.475
Step 2: Age at SCI (linear)	9.47***	0.98***	0.967–0.993
Step 3: Age (linear)	0.63	0.99	0.974–1.011
Step 4: Age at SCI (quadratic)	1.86	1.00	0.998–1.000
Step 5: Age (quadratic)	10.43***	1.00***	0.997–0.999

^*p<0.10,

^**p<0.05,

^***p<0.01,

^****p<0.001

The second regression equation consisted of duration of injury (linear) in step 2 and age at SCI onset (linear) in step 3. The quadratic terms for duration and onset age were entered in the fourth and fifth steps, respectively. The results of this analysis indicated that injury duration (linear) was not significant in step 2. In contrast, age at SCI onset (linear) was a significant addition in step 3 (Δχ²=9.43, p<0.01), and the quadratic term for injury duration was also significant in step 4 (Δχ²=5.10, p<0.05). Onset age (quadratic) was not significant in the last step of the regression model.

In the third regression analysis, age at SCI (linear) was a significant predictor of employment status after controlling for race and education (Δχ²=9.47, p<0.01). Neither chronological age (linear) nor age at SCI (quadratic) was a significant addition to the model in the subsequent two steps. However, in the final step, chronological age (quadratic) was a significant predictor (Δχ²=10.43, p<0.01).

In order to reduce multicollinearity, the predictor variables in the regression analyses presented above were centered. Examination of variance inflation factors, an index of multicollinearity, indicated that all values were <10 (none exceeded 2.50), which is the commonly accepted standard above which harmful multicollinearity is generally indicated.

Follow-up Analyses for Significant Variables in the Regression Models

The results from the series of regression analyses above indicated that chronological age and age at SCI onset were significant predictors of employment status. Although the linear term for injury duration was significant when entered in step 3 of the first regression and the quadratic term for this variable was significant in step 4 of the second regression, these significant findings are likely due to the fact that injury duration contained information about the other significant aging variables in these steps. Specifically, when entered in step 3 of the first regression, injury duration contains information about both injury duration and age at SCI onset. Moreover, when entered in step 4 of the second regression, injury duration contains information about both injury duration and chronological age. Because injury duration was not significant when entered prior to the other predictors (in the second regression), the significant findings above can thus be interpreted as being due to injury duration’s association with the other aging variables. Next, analyses were conducted to further examine the relationships between employment status and the two aging variables that were significant predictors in the regressions.

Chronological Age and Employment Status

The quadratic relationship between chronological age and employment status was explored by defining seven age groups delineated in decade increments (<25, 25–34, 35–44, 45–54, 55–64, 65–74, and 75+). Table 3 presents the breakdown of employment status across these age groups. Results of an omnibus chi-square analysis of employment status across the seven age groups was significant (χ²= 20.02, df=6, p<0.01). Follow-up analyses indicated that a significantly greater proportion of individuals in the 25- to 34-year age group were employed than in the <25-year age group (39% vs. 20%, respectively; p<0.05, two-sided Fisher’s exact test). Additionally, a significantly greater proportion of individuals aged 45–54 were employed compared to those aged 55–64 (36% vs. 23%, respectively; p<0.05, two-sided Fisher’s exact test). Follow-up analyses were conducted to further explore the decrease in employment status from the 45–54 to 55–64 age groups since this marked the point at which the relationship between age and employment status showed a directional change (see Fig. 1).

Table 3.

Employment status across chronological age, age at SCI onset, and injury duration groups

Years	Employed % (N)	Not employed %(N)
Chronological age
Less than 25	20% (12)	80% (48)
25–34	39% (40)	61% (63)
35–44	31% (40)	69% (90)
45–54	36% (58)	64% (104)
55–64	23% (24)	77% (81)
65–74	19% (9)	81% (39)
75 and greater	0% (0)	100% (12)
Age at SCI onset
Less than 25	34% (70)	66% (135)
25–34	32% (47)	68% (98)
35–44	25% (29)	75% (87)
45–54	29% (26)	71% (64)
55–64	19% (8)	81% (34)
65–74	17% (3)	83% (15)
75 and greater	0% (0)	100% (3)
Injury duration
Less than 11	26% (95)	74% (265)
11–20	33% (42)	67% (87)
21–30	36% (33)	64% (58)
31–40	36% (12)	64% (21)
41–50	25% (1)	75% (3)
51 and greater	0% (0)	100% (2)

Percent values represent the percentage of participants within a given age group

Fig. 1.

Relationship between employment status and chronological age

Logistic regression was employed to determine whether the decrease in employment status among the 45–54 and 55–64 age groups was related to putatively important biopsychosocial variables. In this analysis, employment status continued to serve as the criterion variable. Consistent with the analytic series above, participant race and education were entered in the first step of the model. The predictor variables that were entered in the second step included pain severity (SF-8 bodily pain item), physical functioning (SF-8 PCS), fatigue (FSS), sleep (SLP9), and psychological functioning (CES-D). Descriptive data for these variables are presented in Table 4. Chronological age group (45–54 vs. 55–64) was entered in the final step, thus allowing us to examine the relationship of this variable to employment status after controlling for the biopsychosocial variables. The results of this analysis (see Table 5) indicated that the block of biopsychosocial variables was significant in step 2 (Δχ²=52.33, p<0.001). Examination of the individual coefficients indicated that physical functioning, sleep, and psychological functioning were significant unique predictors of employment status in this step of the model. Worse physical and psychological functioning and fewer sleep problems were associated with decreased odds of being employed. In the last step of the regression model, age remained a significant predictor of employment status (Δχ²=6.25, p<0.01). Thus, participants aged 45–54 were significantly more likely to be employed than those aged 55–64 even after controlling for the group of biopsychosocial variables.

Table 4.

Mean (SD) scores for the biopsychosocial variables included in the regression analyses

	Total sample	45–54 age group	55–64 age group
Pain severity	41.19 (9.23)	40.09 (9.31)	40.65 (8.93)
Physical functioning	38.79 (9.90)	37.56 (9.54)	37.54 (9.74)
Fatigue	4.22 (1.59)	4.41 (1.56)	4.32 (1.63)
Sleep	38.76 (20.62)	40.36 (21.00)	36.40 (20.14)
Psychological functioning	8.91 (6.36)	9.49 (6.62)	8.24 (6.14)

Pain severity = SF-8 bodily pain item; physical functioning = SF-8 PCS; fatigue = FSS total score; sleep = SLP9 score; psychological functioning = CES-D total score; age=45–54 and 55–64 years

Table 5.

Logistic regression analyses predicting employment status from race, education, biopsychosocial variables, age group, and age at SCI onset

Step and variables	χ2 change	Exp(B)	95% C.I. for Exp(B)
Step 1: Race	23.54****	0.21***	0.070–0.607
Education		0.33***	0.164–0.658
Step 2: Pain severity	52.33****	0.99	0.951–1.039
Physical functioning		1.10****	1.045–1.149
Fatigue		0.93	0.726–1.192
Sleep		1.04***	1.012–1.059
Psychological functioning		0.88***	0.812–0.956
Step 3: Age	6.25***	2.33**	1.182–4.580
Step 1: Race	53.71****	0.29****	0.163–0.521
Education		0.31****	0.199–0.489
Step 2: Pain severity	68.77****	0.99	0.966–1.024
Physical functioning		1.08****	1.052–1.113
Fatigue		1.05	0.889–1.227
Sleep		1.01*	0.998–1.026
Psychological functioning		0.93***	0.891–0.980
Step 3: Age at SCI onset	3.57*	0.99*	0.972–1.001

Pain severity = SF-8 bodily pain item; physical functioning = SF-8 PCS; fatigue = FSS total score; sleep = SLP9 score; psychological functioning = CES-D total score; age=45–54 and 55–64 years

^*p<0.10,

^**p<0.05,

^***p<0.01,

^****p<0.001

Age at SCI Onset and Employment Status

To further characterize the relationship between employment and age at SCI onset, we first calculated the proportion of participants in each of the seven groups (SCI occurred at age <25, 25–34, 35–44, 45–54, 55–64, 65–74, and 75+) who were currently employed. As can be seen in Table 3, there was a steady decrease in the percent employed as the age at SCI onset increased. Participants who sustained an SCI before age 25 demonstrated the highest rate of employment at 34%, whereas only 14% of those who were injured at age 65 and over were currently employed full- or part-time.

Next, a logistic regression analysis was conducted similar to the one above. Specifically, after controlling for race and education in step 1 of the model, the same biopsychosocial variables (pain severity, physical functioning, fatigue, sleep, and psychological functioning) were entered in step 2, followed by age at SCI onset (linear) in step 3. The results (see Table 5) indicated that after controlling for race and education, the biopsychosocial variables were significant predictors of employment status in step 2 (Δχ²=68.77, p<0.001). Physical and psychological functioning were the significant contributors in this step, with worse functioning associated with decreased odds of being employed. The final step consisting of age at SCI onset was a marginally significant addition to the model (Δχ²=3.57, p=0.06). Thus, the biopsychosocial variables included in this study do not explain the decreased odds of working that are associated with a later age at SCI onset.

Discussion

The aim of the current study was to examine the associations of three age-related variables to employment status following SCI. Although these relationships are more difficult to investigate in cross-sectional research due to the linear dependence among the aging variables, it is possible to determine the unique contribution of each age-related variable to employment status by conducting a series of regression analyses. The primary results of this analytic approach as applied to a large sample of individuals with SCI indicated that chronological age and age at SCI onset, but not injury duration, were significant and unique predictors of employment status. Moreover, these relationships persisted even after controlling for the influence of a number of biopsychosocial variables.

Chronological age demonstrated a nonlinear relationship to employment status. Participants who were between 25 and 34 years of age had an employment rate of 39%, which was almost twice that of those younger than age 25. Employment rates were relatively similar for participants between the ages of 25 and 54; however, there was a marked decrease in employment after age 54. In fact, less than one fourth of those aged 55–64 and less than one fifth of those aged 65–74 were employed at time of survey completion. In addition, and consistent with a priori hypotheses and previous research (Athanasou et al. 1996; Castle 1994; Dijkers et al. 1995; DeVivo and Fine 1982; DeVivo et al. 1987; El Ghahit and Hanson 1978; Kemp and Vash 1971; Krause and Anson 1997; Krause et al. 1999; Noreau and Shephard 1992), older age at the time of SCI was linearly associated with lower likelihood of being employed in a part- or full-time position. The employment rate was highest (34% employed) among those injured before turning 25 years old and declined in all subsequent age at onset cohorts. In follow-up analyses, we explored whether these differences in employment rates were due to differences in pain severity, physical functioning, fatigue, sleep, and/or psychological functioning. Although these biopsychosocial variables have been previously shown to be related to both advancing age (Aapro et al. 2002; Avidan 2002; Gagliese et al. 1999; Gibson 2003; Gill et al. 2001; Hardy and Studenski 2008; Hedden and Gabrieli 2004; Helme and Gibson 2001; Jorm 2000; Liao and Ferrell 2000; Stuck et al. 1999) and employment (Bombardier and Buchwald 1996; Crook et al. 2002; Eriksen et al. 2001; Linton and Bryngelsson 2000; Sanderson and Andrews 2002; Sivertsen et al. 2006; van Amelsvoort et al. 2002), they did not account for the age group differences in employment status observed in the current sample.

It was somewhat surprising that the biopsychosocial variables did not account for the associations between employment status and the two aging variables. Individuals with SCI tend to experience more health problems in general, and at earlier ages, than individuals who do not have an SCI, in part due to more rapid decline of joint and organ systems (Bauman and Waters 2004; Capoor and Stein 2005). Those injured later in life also report poorer health outcomes than individuals who sustained an SCI at a younger age (DeVivo et al. 1990). Because physical functioning and general health are important determinants of sustained employment (Lidal et al. 2007), one might suspect the aging–employment relationship to be fully mediated by these health-related consequences associated with increased age (chronological and injury onset). The current findings suggest, however, that other (un-assessed) factors are critical in understanding the relationships found. It is possible that the biopsychosocial variables assessed in the current study may be important, but that their contribution cannot be determined using cross-sectional data. For example, changes in these variables over time may predict changes in employment. Longitudinal data collected from at least two time points would be needed to evaluate this possibility.

Systemic factors related to access to health care and issues around subsidy may also be important. For example, there may be an accumulation of small functional changes that lead individuals who are working to premature retirement. Because of lack of health care access (e.g., loss of employer paid insurance), individuals may find that they need to retire to disability subsidy and Medicare coverage rather than reduce their hours of work (Johnson 2009).

Differences in education also did not explain the age effects observed herein. Education has often been found to be an important predictor of return to work following an SCI (Lidal et al. 2007), and this finding was replicated in the current study wherein participants with more education were employed at a higher rate than those with less education. However, because education was not strongly associated with any of the three aging variables (data not presented), it is not surprising that the differences in employment status across chronological age and age at SCI onset groups persisted even after controlling for education. As noted below, the current sample included a relatively large number of participants with advanced education; therefore, these findings should be interpreted with caution and replicated in other samples.

Injury severity (i.e., level and completeness) did not significantly differ across employment status. Previous research has been mixed on the importance of injury severity in this context. Some studies have found that individuals with injuries that were more severe, at a higher level, and/or complete were less likely to be employed, whereas others have not found such differences (see review by Anderson et al. 2007). It appears then that variables other than injury severity are more important factors in the context of employment, although for any given individual, injury severity may in fact be a highly important determinant of return to work following an SCI.

The analytic approach used in this study and our previous one (Jensen et al. 2009) addresses some of the challenges that plague cross-sectional research on aging and disability. Specifically, it allows the researcher to disentangle the effects of chronological age, duration of injury, and age at injury onset. Because of the linear relationships among these three aging variables, they cannot all be considered in a single analysis. However, we have demonstrated that the unique influences of these variables on important SCI outcomes (employment in the current study and sleep in the previous report) can be quantified through a series of regression analyses using cross-sectional data. This is an important contribution to this specific line of research, as well as the broader disability field, since longitudinal investigations require significant resources that are not always available to researchers. Nevertheless, we strongly caution against relying on this and related analytic approaches as a substitute for well-designed longitudinal investigations since such investigations make it possible to identify causal relationships among the variables of interest.

There are several limitations of the current study that should be considered. First, these data were drawn from a cross-sectional survey which, as noted above, precludes drawing causal interpretations about the observed relationships. Second, for statistical purposes, participants who were employed part- or full-time were combined into one group, and those who were not currently employed were combined into a second group. Thus, a more fine-grained analysis on employment status was not conducted, e.g., comparing individuals who were employed full-time, part-time or not at all, or examining the relationships of aging variables to number of hours worked per week. Third, some of the participants in this study fell outside of the traditional working age range of 18–64, and 15% of these participants were employed. Although it is possible that we over-sampled from elderly individuals who were employed, when the data were reanalyzed to include only those aged 18–64, the results were not appreciably different and the overall conclusions were the same. Fourth, compared to a recent study based on the SCI model system (Pflaum et al. 2006), women, non-Hispanic whites, and individuals with post-secondary education were overrepresented in the current sample. These factors may make it difficult to compare the findings from this study to similar research from the SCI Model systems and may limit the generalizability of these findings. Relatedly, although the response rate of those who were mailed study surveys was good (74%), this represents only about one quarter of the total number of individuals who were initially invited to participate. It is possible that those who volunteered to participate may differ in important ways from non-respondents; however, comparisons between these two groups were not possible given the lack of informed consent and data from non-respondents. Thus, any conclusions that are drawn from this study about the overall population of individuals with SCI should be made with caution. Finally, the survey consisted of only self-report measures, which may be subject to reporting bias. Future research could supplement such self-report data with direct observation and/or information elicited from other sources (e.g., health care providers, family).

These limitations notwithstanding, the results of the current study suggest that chronological age and age at SCI onset, but not injury duration, are particularly (and uniquely) important variables in the context of post-SCI employment. The current findings highlight the mid 50s as a time when individuals with SCI are at risk to be the most challenged with respect to maintaining employment. Given that maintaining employment is a valuable goal for the individual, the services that have been shown to assist individuals with SCI to maintain employment may be particularly important to provide when these individuals reach their early 50s.

Interestingly, neither of the age-related associations identified in this study were explained by putatively important biopsychosocial variables. Additional work is needed that continues to address employment outcomes in individuals following SCI. In particular, longitudinal designs are needed that assess intra-individual changes in employment and putatively important biopsychosocial variables over time. There should also be an emphasis on collecting data on variables that might be expected to have an impact on retention of employment, as well as variables that lend themselves to intervention efforts (individual and/or social). Longitudinal research is also needed to differentiate between age and cohort effects, which are necessarily confounded in cross-sectional data (Krause and Adkins 2004) such as that analyzed in the current study.