Group-based trajectory analysis of longitudinal employment patterns and predictors for adults with visual impairments

1Introduction

The World Health Organization estimates that 1.2 billion people globally have an uncorrectable visual impairment; in other words, they are blind or experience low vision that cannot be corrected with surgery or glasses (World Health Organization, 2022). Visual impairment results in an economic burden that has received considerable attention and has been estimated for the United States (Frick et al., 2007; Rein et al., 2006; Wittenborn et al., 2013), many other countries (Chakravarthy et al., 2017; Köberlein et al., 2013), and globally (Marques et al., 2021). A substantial component of this burden is lost productivity, as many people with visual impairments do not work. Although visual impairment is more common in older adults, people of all ages experience it. The prevalence of visual impairment in the United States among working-age adults is estimated to be 2.1% (U.S. Census Bureau, 2022a).

A substantial gap in the employment rates between people with visual impairments and people without visual impairments exists around the world. It is perhaps most well-documented in the United States, where this gap has been tracked across multiple national surveys since 1976 (Kirchner & Peterson, 1979; McDonnall & Sui, 2019). In 2021, the employment rate for people with visual impairments in the United States was 48.2% compared to 76.5% for people without disabilities (U.S. Census Bureau, 2022b). An employment gap between people with and without visual impairments has also been documented in many other countries and regions of the world (Brunes & Heir, 2022; Chai et al., 2023; Marques et al., 2021). Two studies focused on the relationship between visual impairment and being out of the labor force (not working or seeking employment) and unemployment (not working but seeking work; McDonnall & McKnight, 2021; Sherrod et al., 2014). Both studies found that people with visual impairments were significantly less likely to be in the labor force than people without visual impairments, but they were not more or less likely to be unemployed.

Most research related to employment among people with visual impairments has focused on employment status at one time point. Little information is available about employment status over time for this population; only a few studies have addressed continuity of employment or utilized longitudinal employment data. Two such studies found that people with visual impairments were less likely than people without visual impairments to hold permanent jobs (work continuously over a year; Kirchner & Peterson, 1980; McDonnall et al., 2022a), and another study documented that 28% of older adults (ages 55–69 years) and more than 10% of adults (ages 18–54 years) with visual impairments who were employed at the first interview were not employed 1 year later (Kirchner et al., 1999). In addition to assessing continuity of employment over a 1-year period, McDonnall and colleagues (2022a) evaluated longitudinal data and determined that only 17.3% of people with visual impairments worked continuously over a 4-year period. They reported descriptive information about longitudinal work patterns but did not compare results for the general population or investigate predictors of the work patterns.

Many other studies have investigated predictors of employment for people with visual impairments. A few variables have consistently been associated with employment, with education level having the most consistent, and often largest, association (Brunes & Heir, 2022; Clements et al., 2011; Goertz et al., 2010, 2017; Lund & Cmar, 2019a, 2019b, 2020; Marques et al., 2019). As would be expected, higher levels of education are associated with higher rates of employment (McDonnall & Tatch, 2021). The presence of additional disabilities or comorbidities is another variable frequently associated with lower levels of employment for people with visual impairments (Brunes & Heir, 2022; Clements et al., 2011; Goertz et al., 2010, 2017; Lund & Cmar, 2019b; McKnight et al., 2021). Although not investigated as often, receipt of government disability benefits has also been consistently associated with employment, with people who received these benefits much less likely to be employed than those who did not receive benefits (Crudden & McKnight, 2022; Lund & Cmar, 2019b; McDonnall et al., 2022b; McKnight et al., 2021).

Health is another variable that has been investigated less frequently but has often been associated with employment when considered. People with visual impairments who experience poorer health have typically had lower levels of employment than those in good or better health (Cimarolli & Wang, 2006; Crudden et al., 2023; Kirchner et al., 1999; McDonnall, 2010), although one study did not find health to be a predictor of employment for people with visual impairment based on measured visual acuity, while it was for people with presenting (correctable) visual impairment due to uncorrected refractive errors (McDonnall & McKnight, 2021). Another study determined that health was not a statistically significant predictor of employment for people who worked prior to experiencing disability, although it approached significance (p < .10) (McKnight et al., 2021). When evaluating the causes of being out of the labor force, a majority of people with visual impairments identified poor health as their reason (Crudden & McKnight, 2022). Greater health-related quality of life has also been associated with employment among people with visual impairments (Marques et al., 2019).

Other demographic variables have been evaluated regarding their relationship with employment for people with visual impairments, but results have been much less consistent. Although female gender has often been associated with lower rates of employment, there was no relationship between gender and employment in several studies (Goertz et al., 2010; Lund & Cmar, 2019a, 2019b, 2020). Findings regarding relationships between employment and (a) race, (b) ethnicity, and (c) age have been even less consistent (Goertz et al., 2010; Lund & Cmar, 2019a, 2019b, 2020), suggesting that there is no association between these variables and the likelihood of employment for people with visual impairments.

Only two studies were identified that utilized longitudinal analysis methods to study predictors of employment for people with visual impairments, and both focused on transition-age youth (Connors et al., 2014; McDonnall, 2010). Both studies also employed a common approach to longitudinal analyses: estimating one average trajectory for the entire sample and using random effects to capture individual variability. A potential problem with this type of analysis is that it may mask patterns of change for subgroups (Andruff et al., 2009; Nagin, 2005). An alternative approach that resolves this issue is trajectory modeling, which includes growth mixture modeling and group-based trajectory modeling (GBTM). This approach allows researchers to identify distinct groups of people who follow similar patterns of change for a specified outcome over time (Nagin, 2005; Nagin & Odgers, 2010). GBTM has been used to investigate longitudinal employment trajectories of several populations (Christiansen & Moan, 2022; Helgesson et al., 2018; Hou et al., 2012; Kuitto et al., 2019; Leinonen et al., 2019; Sun & Chen, 2017), but it has not been used to examine longitudinal employment trajectories of people with visual impairments.

The purpose of this study was to investigate longitudinal employment trajectories among adults with visual impairments in the United States and compare their employment patterns to those of people without visual impairments. Further, we sought to identify variables that are associated with the employment trajectories of people with visual impairments. We utilized two research questions to guide our study:

2Materials and methods

3Results

When building the basic model for the full sample, each increase in the number of trajectory groups from one to five yielded an improvement in model fit, but the five-group model included one group that comprised less than 5% of the sample. Therefore, we rejected the five-group model in favor of the more parsimonious four-group model. The final model had four quadratic trajectories that reflect distinct patterns of employment: (a) Low, (b) Rising, (c) Declining, and (d) High. The four employment trajectories are displayed in Fig. 1, with solid lines representing the observed trajectories and dotted lines representing the predicted trajectories.

Fig. 1

Observed and predicted employment trajectories for people with and without visual impairments (N = 1,632).

The Low trajectory group had very low employment rates across all four waves; most people in this group did not work at all during the study time frame. The Rising group had a steep upward employment trajectory, characterized by a steady increase in employment rates from moderately low in Wave 1 to high in Wave 4. The Declining group had a steep downward employment trajectory, with high employment rates in Wave 1 that decreased sharply and steadily before leveling out near zero in Wave 4. The High group had a relatively stable employment trajectory, characterized by high employment rates across all four waves.

The final, basic models for the Visual Impairment and Comparison groups yielded four trajectories with similar employment patterns to those identified for the full sample. The Visual Impairment group model had three quadratic trajectories and one flat trajectory (the linear and quadratic effects were not significant for the High trajectory), and the Comparison group model had four quadratic trajectories. The predicted trajectories from both models are depicted in Fig. 2, with dashed lines representing the trajectories for people with visual impairments and dotted lines representing the trajectories for people without visual impairments. Compared to people without visual impairments, people with visual impairments were more likely to be assigned to the Low trajectory group and less likely to be assigned to the Rising, Declining, and High groups (see Table 2). The diagnostic statistics reported in Table 2 support the adequacy of all three basic models and indicate high accuracy of the group assignments according to Nagin’s criteria (Nagin, 2005).

Fig. 2

Predicted employment trajectories for people with visual impairments (n = 816) and a matched comparison group (n = 816).

Table 3 provides estimates from the four-group multinomial logit model with visual impairment as a predictor of trajectory group membership. In this model, the Low group is the reference group. All three coefficient estimates for visual impairment were negative and statistically significant, indicating that visual impairment decreased the probability of membership in the Rising, Declining, and High groups relative to the Low group. The Wald test of the equality of the visual impairment coefficients for the Rising, Declining, and High trajectory groups was not significant (χ² (2) = 0.03, p = .98), which implies that visual impairment did not differentiate between membership in those three groups.

Table 4 presents the baseline characteristics of people with visual impairments by trajectory group assignment. Individuals assigned to the Low group were the oldest, had the highest rates of SSDI and SSI receipt, and were the most likely to have two or more additional disabilities and report fair or poor health. Individuals assigned to the Rising group were the youngest and the most likely to be Hispanic. Individuals in the Declining group were the most likely to be female and have an associate degree. Individuals in the High group were the most likely to be male, White, and not Hispanic; have a bachelor’s degree or higher; and report good or better health. The High group also had the lowest rates of SSDI and SSI receipt.

Table 5 provides model estimates from the analysis of factors associated with employment trajectories for people with visual impairments, with the Low group serving as the reference group. Controlling for the other predictors, SSDI receipt, SSI receipt, and number of non-visual disabilities significantly decreased the probability of membership in the Rising, Declining, and High groups relative to the Low group. The Wald test of the equality of the non-visual disabilities coefficients for the Rising, Declining, and High trajectory groups was significant (χ² (2) = 7.12, p = .03), which indicates that number of non-visual disabilities differentially predicted membership in those three groups. However, disability benefit receipt did not differentiate between membership in the Rising, Declining, and High groups according to the non-significant Wald tests for SSDI receipt (χ² (2) = 2.98, p = .23) and SSI receipt (χ² (2) = 3.46, p = .18). Relative to the Low group: older age decreased the probability of membership in the Rising group, female gender decreased the probability of membership in the High group, and White race and having a bachelor’s degree or higher increased the probability of membership in the High group.

Fair or poor health was associated with a significant decrease in all four employment trajectories over time (Table 5), but it had the most substantial impact on the Rising and Declining trajectories. To illustrate the impact of health on the Rising trajectory, Fig. 3 depicts three sub-trajectories: (a) the group average based on observed values for all members of that group, (b) the predicted trajectory (based on model estimates) for people who reported fair or poor health in Waves 1–4, and (c) the predicted trajectory (based on model estimates) for people who reported fair or poor health in Wave 1 and good or better health in Waves 2–4. For a change from fair or poor health at Month 11 to good or better health at Month 12, the expected employment rate increased by about 23 percentage points and continued an upward trajectory over subsequent months, remaining higher than the predicted trajectory for people with ongoing fair or poor health.

Fig. 3

Estimated impact of health on employment for people with visual impairments in the rising trajectory group.

Figure 4 depicts three sub-trajectories for the Declining group: (a) the group average based on observed values for all members of that group, (b) the predicted trajectory for people who reported good or better health in Waves 1–4, and (c) the predicted trajectory for people who reported good or better health in Wave 1 and fair or poor health in Waves 2–4. For a change from good or better health at Month 11 to fair or poor health at Month 12, the expected employment rate decreased by about 40 percentage points and continued to decline until reaching zero at the end of Wave 3.

Fig. 4

Estimated impact of health on employment for people with visual impairments in the declining trajectory group.

4Discussion

The purpose of this study was to investigate longitudinal employment trajectories of people with visual impairments. We utilized 4 years of monthly employment data to model longitudinal employment trajectories, compare employment trajectories of people with and without visual impairments, and identify factors associated with employment trajectories for people with visual impairments. We found that people with and without visual impairments had similar employment patterns, consisting of Low, Rising, Declining, and High trajectories. Most people with visual impairments were in the Low employment trajectory group, meaning that they did not work at all during the 4-year period of the study. In addition, people with visual impairments were more likely to be in the Low employment trajectory group than people without visual impairments. Factors that differentiated trajectory group membership for people with visual impairments were federal disability benefit receipt, number of non-visual disabilities, age, gender, race, and having a bachelor’s degree or higher. Furthermore, self-reported health was associated with changes in all four employment trajectories for people with visual impairments.

Our study is unique for several reasons. First, we compared longitudinal employment trajectories between people with visual impairments and a matched sample of people without visual impairments. Matching participants on gender, age, and number of non-visual disabilities was important to create a sample that was comparable on certain characteristics that would not be similar if the entire general population was utilized as a comparison group, as typically done in previous research. Second, this study is the first investigation of predictors of longitudinal employment trajectories for adults with visual impairments, which adds to the literature by focusing on factors associated with employment over time rather than at a single time point. Third, we utilized trajectory modeling to analyze the data, an approach that has not been used in other studies focusing on this population. A key advantage of the trajectory modeling approach was its ability to capture multiple distinct employment patterns.

Despite matching on three relevant traits, our comparison sample differed from our visual impairment sample on several characteristics that are important to employment. People with visual impairments were more likely to be in poorer health, members of a minority group, less educated, and receiving disability benefits. These differences may explain the fact that visual impairment was a risk factor for not working for multiple consecutive years. Research has documented that older adults with visual impairments are significantly more likely to experience chronic conditions, and that older adults with a visual impairment and a chronic condition are more likely to experience fair or poor health than older adults without visual impairments who have the same chronic condition, even when controlling for confounding factors (Crews et al., 2017). Visual impairment is also a comorbidity of conditions that typically result in poorer health for older adults (e.g., diabetes, multiple sclerosis, stroke; Court et al., 2014; Crews et al., 2017). Those previous studies focused on adults aged 65 years or older, but the same may be true for working-age adults, particularly those in the upper age range of our sample.

Our findings raise the question of whether lower employment rates are related to visual impairment itself or to the chronic conditions and poor health more commonly experienced by people with visual impairments. For people for whom visual impairment is a comorbidity of another serious condition, their primary challenge to work may not be their visual impairment. Recent research investigating reasons for being out of the labor force for working-age adults with visual impairments suggests that poor health and chronic conditions may be primary factors (Crudden et al., 2023; Crudden & McKnight, 2022). These studies found that the majority of adults with visual impairments who were out of the labor force either reported being unable to work (Crudden et al., 2023) or reported that they were not working due to health reasons (Crudden & McKnight, 2022). The current study supports these and other previous findings (Cimarolli & Wang, 2006; Kirchner et al., 1999; McDonnall, 2010) regarding the importance of good health to employment for people with visual impairments.

Although fair or poor health does not preclude people from working, it was strongly associated with employment trajectories, and 74% of people who reported fair or poor health at Wave 1 consistently did not work during the 4-year period. People in the Rising trajectory group whose health improved to good or better at Wave 2 were projected to join or rejoin the workforce sooner than those who remained in fair or poor health. Conversely, people in the Declining trajectory group who experienced fair or poor health at Wave 2 were projected to leave the workforce sooner than those who remained in good or better health. Still, our findings confirm that long-term employment is possible even for people who perceive themselves to be in fair or poor health.

Another finding from this study that supports previous research is the association between other disabilities and employment for people with visual impairments (Brunes & Heir, 2022; Clements et al., 2011; Goertz et al., 2010, 2017; Lund & Cmar, 2019b; McKnight et al., 2021). Our study documents that the number of non-visual functional disabilities has an increasing impact on the odds of being in the Low employment trajectory group. The impact of the number of non-visual disabilities on the odds of being in the Low trajectory group differed for the High employment trajectory compared to the Rising and Declining trajectories. For all three groups compared to the Low employment trajectory group, the effect of having one non-visual disability was small. The odds of being in the Low trajectory group continued to increase as the number of non-visual disabilities increased, but this effect was the strongest for comparisons between the Low and High trajectory groups. This finding indicates that people with visual impairments and multiple non-visual disabilities were much more likely to be in the Low trajectory group than the other three groups, particularly the High group.

Receipt of disability benefits had the strongest relationship with employment trajectories. Individuals who received disability benefits had tremendously higher odds of being in the Low employment trajectory group than in any of the other trajectory groups, but, not surprisingly, the odds were the largest compared to the High trajectory group. This finding is not unexpected given that SSDI receipt is contingent upon limited work earnings and SSI receipt is contingent upon having limited income and assets. Once people start receiving these disability benefits, very few terminate them due to work (Levere et al., 2018). A few other studies have documented similarly large odds ratios associated with lack of employment and disability benefit receipt among people with visual impairments (Crudden & McKnight, 2022; McKnight et al., 2021). Smaller effects of disability benefit receipt have been observed in studies of state-federal vocational rehabilitation consumers (McDonnall, 2016; Steinman et al., 2013), which may reflect differences in characteristics between people who do versus do not seek these services and receive the benefits of these services (Giesen & Hierholzer, 2016; O’Neill et al.,2015).

Although our study documents that disability benefit receipt, number of non-visual disabilities, and health are all independently associated with employment trajectory, these variables are also closely related to each other. Many people with visual impairments who receive disability benefits may have additional conditions which negatively impact their health. A study of SSDI and SSI recipients who had recently been employed found that the unpredictability of their health was a substantial barrier to future work; health concerns and concerns about health insurance discouraged them from pursuing full-time work that would cease benefits (O’Day et al., 2016). Another study documented the various and complex factors that impact SSDI beneficiaries’ decisions about returning to work (Taylor & Blackburn, 2020). Because many beneficiaries have a limited understanding of work incentives and options for retaining health insurance after benefit cessation (O’Day et al., 2016), they need accurate information and support to make informed decisions about work. Programs that offer financial and benefits counseling may encourage recipients to consider pursuing employment (Taylor & Blackburn, 2020). Benefits counseling may be especially beneficial for people with visual impairments who have poor health, chronic conditions, or both.

Finally, our results also support previous research documenting the importance of education for employment of people with visual impairments (Brunes & Heir, 2022; Clements et al., 2011; Goertz et al., 2010, 2017; Lund & Cmar, 2019a, 2019b, 2020; Marques et al., 2019). Having a bachelor’s degree or higher level of education was associated with membership in the High trajectory group. In other words, when controlling for other factors, having a college degree predicted consistent employment across the 4-year time period of the present study. This finding emphasizes the importance of earning at least a bachelor’s degree to obtaining and retaining employment for people with visual impairments, regardless of health status. Professionals should encourage and support people with visual impairments in getting a college degree, as the association between higher education and employment for this population is clear and consistent.

5Conclusion

Further investigation of employment trajectories of people with visual impairments using other data sources is needed to confirm and extend our findings. It is important to continue to evaluate employment longitudinally, as the employment rate observed during one month of this study overestimates employment over time for this population. It would be valuable to account for level of visual impairment and other vision-related information in future longitudinal employment studies. It would also be valuable to evaluate earnings over time for employed people with visual impairments. Our findings related to the impact of health and non-visual functional disabilities suggest the need to investigate employment separately for younger people and older people within the working-age range, as more chronic conditions and health problems typically occur as people age. Finally, some people in fair or poor health are employed, and they work consistently over time. Future research could investigate this group to determine what motivates and enables these individuals to continue to work despite healthproblems.