Returning to work after dysvascular lower limb amputation—A novel multivariate approach to examine relative contributions of biopsychosocial predictors

Abstract

Background:

Returning to work is a key outcome of rehabilitation and social re-integration after lower limb amputation. It is important to understand what biopsychosocial factors contribute to returning to work after dysvascular amputation.

Objective:

Examining relative contributions of functional and contextual predictors of returning to work in participants with lower limb amputation due to diabetes and other dysvascular diseases.

Study Design:

Cross-sectional.

Methods:

Return-to-work outcome, biopsychosocial characteristics including physical functioning, self-efficacy & perceived ability, and socioeconomical support data were collected from a purposive sample (n = 57) in a multi-state collaborative research network. Grouped Weighted Quantile Sum model analysis was conducted to evaluate relative contributions of biopsychosocial predictors.

Results:

Less than 30% of the participants returned to work after their amputation. Physical functioning (odds ratio = 10.19; 95% CI 2.46−72.74) was the most important predictor group. Working before amputation, prosthetic mobility, and access to rehabilitation care were also identified as key factors associated with returning to work.

Conclusions:

Fewer than 1 in 3 participants with dysvascular amputation returned to work, despite an average age of only 54 years at the time of amputation. Physical functioning was shown to be the most important predictor, while socioeconomic factors such as a lack of access to care also contribute to not returning to work after dysvascular amputation.

Introduction

More than 1 million people live with major lower limb amputation (LLA) in the United States, most of them lost their limbs due to dysvascular diseases such as diabetes, peripheral arterial insufficiency, and other circulation-related complications.^1,2 A recent study by Geiss et al reported a 50% increase in the prevalence of nontraumatic LLA between 2009 and 2015 in the United States.³ This increase was primarily driven by a pronounced surge of amputation rates in the younger age groups (younger than 65 years).³ These relatively young individuals are typically of viable working age and have potential for years of gainful employment. However, returning to work (RTW) after LLA is challenging to individuals who have undergone LLA^4-7 and particularly to those with chronic health concerns commonly associated with dysvascular diseases.

Research on social reintegration, specifically RTW after LLA, is currently limited. A systematic review by Darter et al showed that employment rates following amputation range from 48% to as high as 89% in civilian populations and much lower for those who lost their limb during military service.⁸ Generally, greater prosthetic use (>8 hours/day) has been shown to be correlated with employment.^9-11 Demographic characteristics, physical functioning, and preinjury vocational history seem to be the most important factors associated with the RTW decision and success. However, biopsychosocial outcomes (i.e., physical functioning and prosthetic device adaptation) after amputation vary widely, and these factors are likely to be interrelated with other health and personal factors to influence a patient’s RTW decision after LLA.^8,12 These relationships have not been comprehensively previously investigated.

The interrelationships between factors that can affect RTW after amputation are complex and should be comprehensively viewed because working and having the capacity to work are intertwined with many aspects of patients’ characteristics and life situations.¹² For example, experiencing financial difficulty after LLA has been shown to be a significant predictor for diminished work participation due to physical and emotional distress.¹³ However, numerous variables may interact with financial difficulty, such as family income, vocational history, and accessibility to health care and rehabilitation after LLA.^14,15 Conversely, the loss of employment as a direct result of LLA may also cause financial difficulty. The high degree of potential intercorrelations among these personal traits may render traditional statistical modeling methods (e.g., ordinary or logistic regression analysis) inappropriate due to collinearity and variance inflation problems. Therefore, alternative analysis methods, such as Grouped Weighted Quantile Sum (GWQS) regression analysis,¹⁶ may be needed for estimating the collective contributions of health and personal factors on RTW decision after LLA.

The purpose of this study was to comprehensively examine predictors of RTW after LLA due to dysvascular diseases. Using the GWQS regression analysis, we proposed to determine how self-efficacy and perceived ability, socioeconomical support, and physical functioning factors compound to influence patients’ RTW outcomes. This novel statistical approach may facilitate identification of key factors associated with effective rehabilitation and social reintegration in this growing patient population.

Methods

The study was cross-sectional in design using a convenience sample of civilian, community-dwelling participants with LLA due to dysvascular causes. Community dwelling was defined as someone who was not living in any type of dependent care facility. The participants were sampled from 4 US states (Arizona, Florida, Nevada, and Utah), with consideration of sample representation in socioeconomic backgrounds including sex and race/ethnicity. The inclusion criteria were as follows: (1) age 18–80 years; (2) having major lower limb amputation defined as losing at least an ankle joint; (3) the cause of amputation was due to diseases that affected blood flow and/or tissue healing such as peripheral vascular diseases and diabetes; (4) at least 3 months from initial prosthetic fit; (5) able to ambulate >30 ft with appropriate assistive device as needed; (6) able to perform sit-to-stand transitions independently or with supervision; and (7) able to understand English or Spanish language. Participants were excluded if they had any of the following: (1) any open wounds on either lower limb or significant pain that interfered physical activity and (2) neurological pathology interfering with balance or gait (e.g., hemiparesis, ataxia, multiple sclerosis, etc). This study was approved by the Institutional Review Board at the Florida International University. Participants were recruited through prosthetic and physical therapy clinics and amputee support groups in the data collection regions. Informed consent was obtained before collection of self-reported and physical performance data from each participant during a 90-min visit. Data collection was conducted by physical therapy and/or prosthetic researchers who were trained to administer the outcome measures (details provided further).

Procedures

The International Classification of Functioning, Disability, and Health model was used to guide our selection of measures as applicable to the population with LLA and RTW.^17-20 Under the International Classification of Functioning, Disability, and Health framework, the outcome measures used in this study were organized into 2 parts: (1) functional factors, which is composed of domains of performance-based and self-perceived physical functioning related to activities (e.g., prosthetic mobility)¹⁹ and participation; and (2) contextual factors, in which the selected domains were socioeconomical support and health and personal factors (Figure 1).

Figure 1.

Theoretical framework of investigating factors influencing returning to work decision after dysvascular lower limb amputation.

The measures outlined in Figure 1 are commonly used to examine clinical rehabilitation outcomes in individuals with LLA.^21-23 Specific to this investigation, selected Patient-Reported Outcomes Measurement Information System (PROMIS) questionnaires were included to capture Physical Function,²⁴ Self-efficacy for Managing Daily Activities,²⁵ Ability to Participate in Social roles and Activities, and Instrumental Support. Environmental factors associated with the participants' natural and built environment and accessibility within their community were measured using 4 items from the Environmental Factors Item Bank.²⁶ Additional questions including perceived physical mobility were measured by the Prosthetic Limb Users Survey of Mobility (PLUS-M™).²¹ Employment status both before and after amputation, availability/accessibility of health services, experience of financial difficulty after amputation,^13,27,28 marital status, and other health and personal information were also collected. Specific to their employment status before and after amputation, the following 2 questions were asked: “Did you work before your amputation? (yes/no),” and “Did you return to work after your amputation (including working part-time)? (yes/no).” All questionnaires were available in both English and Spanish. A researcher was present to clarify the survey questions during data collection. REDCap data collection service was used to ensure centralized data collection across all collection sites. Each participant’s responses were securely stored at the data center maintained by Florida International University’s Center for Statistical Consultation and Collaboration.

After completing the questionnaires, participants underwent 2 performance-based mobility tests previously validated for the population with LLA: Component Timed-Up-and-Go test,²⁹ and 2-minute walk test.³⁰ In both tests, participants were allowed to use any assistive device they typically use during daily activities. Investigators in all study sites provided standardized test instructions and followed the same procedures for both tests.

Data and statistical analyses

Data were examined for outliers and distributional normality. Descriptive statistics were used to characterize the study participants who returned and did not return to work after LLA during the study. Frequencies and percentages were reported for categorical variables. Independent sample t tests and χ² tests were conducted to compare the demographic and anthropometric characteristics of the 2 groups of participants.

Based on our theoretical framework (Figure 1), we assembled 3 groups of predictors based on the measured outcomes. The first group contained predictors related to self-efficacy and perceived ability, which included the total t score from the PROMIS Short-Form 8a—Ability to Participate in Social Roles and Activities, and 2 individual item scores: “I have trouble doing all of my usual work (included work at home),” and “I have trouble doing all of the work that is really important to me (included work at home).” This group also included the total t score from PROMIS Short-Form 8a Self-Efficacy for Managing Chronic Conditions and an individual item score: “I can keep doing my usual activities at work.”

The second predictor group contained variables related to socioeconomic support factors, including total t score from PROMIS Short-Form 8a Instrumental Support, Environmental Factors Item Bank score,³¹ availability of rehabilitation services and health insurance,³² experience of financial difficulty,^13,27,28 and marital status. Availability/accessibility of health services including rehabilitation were queried using the following 2 questions in the Environmental Factors Item Bank: “Rehabilitation services and therapies are available when I need them. (yes/no),” and “Adequate insurance is available to pay for healthcare when I need it. (yes/no).”

The third predictor group contained variables pertaining to physical functioning, including performance of the Component Timed-Up-and-Go Test (s), 2-minute walk test (m), daily prosthesis use time (h), total t scores from PLUS-M and PROMIS Short-Form 8b—Physical Function, and employment status before amputation. Employment status before amputation was placed in this group because it has been shown to be an indicator of a participant’s functioning level before amputation and a significant predictor of work disability and RTW in a previous study.³³ Nonmodifiable health and personal factors, such as the participant’s age during amputation, time since amputation, sex, and amputation level, were also included in the model analysis as covariates.

To examine the influences of our defined predictor groups as related to the participants’ RTW decision, the GWQS regression model was used. The GWQS model condenses the 3 groups of predictors into 3 mixture indices and simultaneously evaluates the mixture effects on participant’s response to the question: “Did you return to work after your amputation (Yes/No).”^16,34 The rationale for using the GWQS regression model is to account for the correlations between each predictor variable and the RTW decision when condensing the predictors by group. The weight of each individual variable within a predictor group on the mixture effect reflected the contribution of that variable to the RTW outcome (Equation 1).

$$\log \left(\frac{P\left(\text{RTW}=1\right)}{P\left(\text{RTW}=0\right)}\right)={\beta }_0+{\displaystyle \sum }_{j=1}^3{\beta }_j{GWQS}_j+{\displaystyle \sum }_{k=1}^5{\gamma }_k{Z}_k$$ (1)

Equation 1: GWQS Model Equation (RTW decision 1 = yes, 0 = no; ${GWQS}_j$ represented the $j$ th mixture index composed by the summation of the weight for each corresponding predictor multiplying by its quantile. $Z_k$ represented the $k$th covariate, that is, age, age at amputation, sex, and amputation level (transtibial; transfemoral).

We performed multi-index GWQS analyses to see how the 3 groups of predictors affected the RTW decision simultaneously. A total of 1000 bootstrapping steps were conducted to make the estimated coefficients of the mixture effects more robust. The average values of bootstrap samples determined the estimated coefficients of the 3 mixture indices, and their 95% confidence intervals (CIs) were determined by the 2.5th and 97.5th percentiles of the bootstrap samples. The exponential function of an estimated coefficient of an index was explained as the index's odds ratio (OR). In addition to the mixture effects of the predictor groups, we also assessed the contributions of individual variables within each index through empirically estimated weights, where a larger weight indicated a greater contribution from the variable to the predictor group’s mixture effect. A τ value, computed as the reciprocal of the number of factors in each predictor group, determined whether the weight of an individual variable exhibited a significant contribution (i.e., τ = 0.2 for self-efficacy and perceived ability and ≈0.17 for socioeconomic support and physical functioning indices). RStudio version March 1, 1056 (RStudio, PBC, Massachusetts) was used to perform the GWQS regression analysis. The significance level was set at 0.05.

Results

Fifty-seven participants completed the data collection (K2 = 22, K3 and above = 35). Seventeen of them (30%) reported returning to work after amputation. Examining only those who were working before amputation (n = 38), 55% did not return to work. Participants’ demographic and anthropometric characteristics by RTW groups are summarized in Table 1. Overall, there were no significant differences between the return and not return groups in age, age during amputation, weight, height, and sex. The purposive recruitment yielded a participant sample that consisted of 40% female and 42% persons of color.

Table 1.

Summary of participant characteristics.

Participant characteristics	Return to work after amputation		Not return to work after amputation		p
	(N = 17)		(N = 40)
	Mean	SD	Mean	SD
Age (y)	59.91	9.24	62.06	9.65	0.438
Age during amputation (y)	52.40	10.67	55.71	10.33	0.277
Weight (kg)	96.96	22.23	92.63	34.80	0.638
Height (cm)	172.99	11.25	172.04	12.14	0.783

Sex	N	%	N	%	p
Male	11	32.35	23	67.65	0.612
Female	6	26.09	17	73.91

Abbreviation: SD, standard deviation.

Accounting for covariates in the model, the physical functioning group was shown to be the most influential index to patients’ RTW with an adjusted OR of 10.19 (95% CI, 2.46–72.74; p value = 0.0058; Table 2). The mixture index of socioeconomic support became another significant predictor group with OR = 0.25 (95% CI, 0.06–0.71; p value = 0.0225; Table 2).

Table 2.

Odds ratios of mixture indices in the multi-index GWQS model analysis.

Index	OR	95% CI	p
Self-efficacy and perceived ability	1.28	0.52–3.34	0.597
Socioeconomic support	0.25	0.06–0.71	0.023
Physical functioning	10.19	2.46–72.74	0.006
Covariates
Age during amputation (y)	0.94	0.79–1.08	0.390
Time since amputation (y)	0.98	0.85–1.13	0.803
Sex	1.40	0.24–9.92	0.713
Amputation level (ref = bilateral)
Transtibial	0.49	0.05–4.32	0.516
Transfemoral	0.36	0.01–6.69	0.503

The estimated weights for individual predictor variables within the 3 predictor groups in the multi-index model are shown in Figure 2. Our results indicated that self-efficacy for doing activities at work (weight = 0.223), ability to participate in social roles (weight = 0.219), ability to do important work including work at home (weight = 0.215), and ability to do usual work including work at home (weight = 0.179) were factors in the self-efficacy and perceived ability index with the highest weights. In the socioeconomic support index, availability of rehabilitation services (weight = 0.475) and financial difficulty after amputation (weight = 0.184) were the 2 most important factors. In the physical functioning index, employment status before amputation and PLUS-M total t score were the key factors with relative weights of 0.440 and 0.330, respectively.

Figure 2.

Estimated weights for factors included in each predictor group. The red line indicates the τ value, which is 0.20 in self-efficacy and perceived ability index and 0.17 in socioeconomic support and physical functioning index groups. Note: se_work = self-efficacy for doing activities at work; T_se = PROMIS self-efficacy total t-score; T_ability = PROMIS ability to participate in social roles total t-score; ability_usual_work = ability to do usual work including work at home; ability_important_work = ability to do important work including work at home; efib_score = environment and accessibility measured using the Environmental Factors Item Bank; T_instrumental = PROMIS instrumental support total t-score; 2MWT = 2-minute walk test performance; T_PLUSM = PLUS-M total t-score; cTUG = component Timed Up and Go test performance; hours_use = daily prosthetic use time; T_physical_fn = PROMIS physical function total t-score.

Discussion

The aim of this study was to comprehensively examine variables that contribute to the prediction of RTW after LLA by using a novel GWQS regression analysis. To our knowledge, this investigation was the first to focus solely on RTW in individuals who had undergone LLA due to dysvascular diseases, which is the most common and a surging cause of LLA in the United States.^1,3 In addition, the purposive recruitment of a sex, race, and ethnically diverse sample and the availability of questionnaires in 2 languages may also improve the participant representation and generalizability of our findings when compared with previous studies. Our main finding is that indices from 2 of the predictor groups (i.e., physical functioning and socioeconomic support) were significant in predicting RTW. In patients with dysvascular amputation, physical functioning was revealed as the most important predictor index. A closer examination of the variables within this group showed that employment before amputation, availability of rehabilitation, and physical mobility measured by PLUS-M to be the most important individual factors determining RTW after dysvascular LLA.

The finding that employment status before amputation is a strong predictor for RTW aligns with results from previous research on this topic. For example, MacKenzie et al reported that having a job tenure of >1 year and higher levels of perceived job involvement were significantly predictive of RTW after major limb trauma including amputation. Such findings are logical because someone who was not employed before LLA is unlikely to seek or gain employment due to a wide range of challenges a person would have to face after LLA.⁴ In a systematic review, Darter et al summarized that job type (clerical vs. physical), employer support, and workplace accommodations were also important vocational factors for successful RTW and maintaining long-term employment.⁸ Furthermore, our observed RTW rate of 30% was lower than the reported range of 48%–89% in the literature,⁸ possibly because the previous studies were not specifically focused on individuals with dysvascular LLA. For example, among the 25 studies reviewed by Darter et al,⁸ 44% examined only individuals with traumatic amputation and 32% focused solely on combat-related amputations (i.e., military service members). Therefore, the average age of our participants during amputation was comparatively higher than that reported in previous studies.^35,36

To understand the decision-making process of RTW after LLA, it is important to recognize the complex interactions among the multitude of challenges associated with living with LLA. Previous studies have shown that the psychological stress and financial hardship are significantly associated with underemployment and unemployment after LLA.^13,37 The theoretical framework we developed in this study (Figure 1) is a useful guide for comprehending the influences of functional and contextual factors on a patient’s RTW decision. While the most direct and apparent effect of LLA is on physical functioning affecting the patient’s mobility and ability to participate in mobility-related activities, contextual factors may alleviate or acerbate the perception of disability. For example, lack of access to the needed care for recovering function and adapting to life after LLA is likely to affect the RTW decision. Indeed, our data showed that within the socioeconomic support predictor group, availability of rehabilitation and financial difficulty are 2 of the most important contributors. We believe that these contextual factors related to health care access are perhaps even more important for those with dysvascular LLA due to the ubiquitous presence of comorbidities and the complex coordination between specialized care necessary for these individuals.^13,38,39 In practice, physical functioning, as related to RTW, may be improved through vocational training, occupational rehabilitation, and other associated psychological factors.^40,41

Our findings also revealed that traditionally conceived ability measures such as age and amputation level may not be considered as the most important gauge of a patient’s RTW potential, specifically in individuals with dysvascular LLA. Performance-based metrics of 2-MWT distance and Component Timed-Up-and-Go test time were not heavily weighted predictors of RTW in our models. This may indicate that someone who has returned to work did not make that decision based simply on their ability to walk faster and further in a standardized testing environment and that other underlying factors related to physical mobility are more influential. In fact, the self-reported PLUS-M survey, which contains various mobility scenarios for an individual with limb loss to consider,^21,42 was shown to be a more comprehensive and better physical functioning measure for predicting RTW in our study cohort.

A standard analysis strategy in health-related epidemiology studies is to evaluate the contribution of each individual predictor variable while adjusting for covariates/confounders. However, many health and personal factors interactively affect patient outcomes, which is not accounted for in traditional regression models. Our use of the GWQS analysis allowed us to assess the relative contributions of personal and health outcome variables as groups, and individual predictors are useful for identifying meaningful factors within a plethora of clinical data.¹⁶ Further studies are needed to establish the sensitivity and specificity of our identified predictors in a larger sample and to determine whether the predictive quality of these variables can be generalized to patient populations with LLA with etiologies beyond dysvascular diseases.

Limitations

The main limitation of our study is that the RTW decision was modeled as a dichotomous variable (i.e., yes vs. no). More in-depth queries regarding work history, reasons to or not to RTW, and types of employment after RTW were not examined. Therefore, the conclusions drawn from this research, though appropriate for our study purpose, must be viewed as bases on which future research can be founded. Furthermore, the cross-sectional design of the study nature limited any interpretation about causation and temporal relationships. For example, while we found that greater prosthetic mobility is a significant predictor for returning to work, it is still possible that working may promote physical activity and have a beneficial effect on mobility. Future study with a longitudinal design is needed to examine RTW both as a contributor and a result of functional recovery after LLA.

Conclusions

Fewer than 1 in 3 participants with dysvascular LLA returned to work, despite an average age of only 54 years during amputation. This RTW rate is comparatively lower than other causes of LLA previously reported. Physical functioning factors, including prosthetic mobility measured by PLUS-M and work status before amputation, were shown to be the most important predictor of RTW in this population. Socioeconomic factors such as a lack of access to postamputation rehabilitation and experience of financial difficulty also contribute to not returning to work. Our findings and the theoretical framework developed for this study may be used to guide clinical decision-making.

Funding

The authors disclosed that they received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interest

The authors disclosed no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.