Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jan 29.
Published in final edited form as: Ann Epidemiol. 2015 Aug 4;25(11):832–838. doi: 10.1016/j.annepidem.2015.07.011

Working with Cancer: Health and Employment among Cancer Survivors

Tainya C Clarke 1, Sharon L Christ 2, Hosanna Soler-Vila 1,3, David J Lee 1, Kristopher L Arheart 1, Guillermo Prado 1, Lora E Fleming 1,4
PMCID: PMC9884513  NIHMSID: NIHMS713099  PMID: 26320705

Abstract

Background

Cancer affects a growing proportion of US workers. Factors contributing to whether they continue or return to work after cancer diagnosis include: age, physical and mental health, health insurance, education, and cancer site. The purpose of this study was to assess the complex relationships between health indicators and employment status for adult cancer survivors.

Methods

We analyzed pooled data from the 1997–2012 US National Health Interview Survey (NHIS). Our sample included adults with a self-reported physician diagnosis of cancer (n=24,810)) and adults with no cancer history (n=382,837). Using structural equation modeling (SEM), we evaluated the relationship between socio-demographic factors, cancer site, and physical and mental health indicators on the overall health and employment status among adults with a cancer history.

Results

The overall model for cancer survivors fit the data well (χ2 (374)=3654.7, p <0.001; CFI = 0.98; RMSEA=0.04). Although black cancer survivors were less likely to report good-to-excellent health, along with Hispanic survivors they were more more likely to continue to work after diagnosis compared with their white counterparts. Health insurance status and educational level were strongly and positively associated with health status and current employment. Age and time since diagnosis were not significantly assocated with health status or employment, but there were significant differences by cancer site.

Conclusion

A proportion of cancer survivors may continue to work because of employment-based health insurance, despite reporting poor health and significant physical and mental health limitations. Acute and long term health and social support are essential for the continued productive employment and quality of life of all cancer survivors.

INTRODUCTION

With major improvements in cancer detection and treatment, increasing numbers of cancer survivors return or continue to work post cancer diagnosis (1,2). Recent literature reflects this trend among survivors, many of whom continue to work even during active cancer treatment (3). However, cancer survivors’ ability to work strongly depends on their overall health status which, in turn, is influenced by age, cancer, stage at diagnosis, sequelae, treatment, access to healthcare, and health insurance (4–7).

Previous research has reported that most working cancer survivors in the U.S. are highly educated, middle-aged individuals reporting two or more functional limitations (8). Since a large proportion of cancer survivors are of working age, understanding the impact of the physical and mental health status of this population on their employment is important for employers, healthcare providers, public health planners, and cancer patient advocates, as well as the survivors themselves and their families. Most studies of cancer survivorship lack in-depth examinations of the relationships between employment and reported health status, activities of daily living (ADLs), and physical and psychological limitations. Although these factors have been studied independently (3,9–11), their joint impact on health and employment status, remain unknown. Thus, our main aim was to fill this gap by exploring the relationships between selected health indicators and health and employment status among adult cancer survivors.

METHODS

Participants

We analyzed pooled data from the 1997–2012 U.S. National Health Interview Survey (NHIS), merging data from the sample adult files, person files, and the cancer control modules (12). The NHIS collects demographic and health information from a representative sample of noninstitutionalized U.S. civilian population on an annual basis. Information is collected by household; one adult per family is randomly selected and administered questions related to health, including questions about cancer history. Annual adult response rates to the NHIS data used in this study averaged 71.3% (range: 69%–80%) (13). All data were self-reported and participants with missing information were excluded.

Variables

Participants were categorized as individuals with a cancer history (or cancer survivors) if they responded “yes” to the question: “Have you ever been told by a doctor or other health professional that you had cancer or a malignancy of any kind?” For this subsample, further questions permitted stratification by cancer site. For the purposes of this study, we looked at five of the most frequently diagnosed cancers within the U.S. (breast, prostate, lung, bladder, and colorectal cancers) (14). The rest of reported cancer sites were grouped under “other” (referent). Persons with non-melanoma skin cancer were not included in the study population.

The outcome variables of interest were employment and health status. Employment was dichotomously coded as currently “employed” or “not employed,” determined by participants’ response as to whether or not they were working (paid and unpaid) during the week prior to their NHIS interview. Health status was defined by self-reported health status measured on a Likert scale. To facilitate comparison with previous research, scale values were combined to create a dichotomous variable where health status was classified as “poor-to-fair” and “good-to-excellent”.

We included health-related predictors referring to both physical and psychosocial health. A functional limitation results from a substantial impairment in an individual's ability to complete a range of tasks or major life activities for daily functioning, whether simple or complex; as such, functional limitations form the link between impairment and disability (15). Respondents were asked about the level of difficulty experienced when performing a named task by themselves without using any special equipment. The response categories (ranging from 1–5) were “not at all difficult,” “only a little difficult,” “somewhat difficult,” “very difficult,” and “can’t do at all.” Responses to these nine self-reported items (i.e. walking, standing, stooping, carrying, grasping, climbing, sitting, pushing and reaching) were combined into a single “latent variable” (described below) measuring functional limitations. Three additional questions with the same response categories were captured by a second latent variable to assess performing instrumental activities of daily living (IADLs); these are complex skills needed for successful independent living, i.e. social functions, relaxing, and shopping (16).

Limitations in performing basic activities and the effects of chronic conditions also encompass psychosocial factors, such as psychological distress, known to afflict cancer survivors. Based on the Kessler 6 (K-6) scale of psychological distress (17), participant responses to six questions “How often did you feel …. (nervous, sad, restless, hopeless, worthless, or ‘that everything was an effort’)?” were measured by a single latent variable, psychological distress. The response options were: “none of the time,” “a little of the time,” “some of the time,” “most of the time,” and “all of the time being,” yielding a score between 1 and 5.

Socio-demographic variables included gender, age, Hispanic origin and race, education, and health insurance status. Age was measured as a continuous variable in years. Education was treated as 3-level categorical variable: less than high school (referent), high school diploma or equivalent, and some college or higher education. Hispanic origin and race were divided into four mutually exclusive categories: non-Hispanic white (white) (referent), non-Hispanic black (black), Hispanic, and non-Hispanic other/ multi-race (other race). Based on their health insurance, participants were classified as privately insured (referent), publicly insured, and uninsured, using the definitions and classification commonly used by the National Center for Health Statistics (NCHS) (18).

Data Analytic Plan

For the analyses, we used structural equation modeling (SEM) because of the complexity of the pathways and variables (including latent variables) that were being examined. The use of SEM allowed the simultaneous evaluation of all the relationships of the variables, as opposed to the individual effects of each predictor in separate multivariable logistic regression models. Also in SEM, latent variables may be used to measure multiple item scales independent of random measurement error, resulting in improved measurement reliability. Data management and descriptive analyses were conducted using SAS statistical software package version 9.2(19), we used Mplus version 5.2 (20) to test study hypotheses within a SEM framework. Due to the complex sample survey design, we performed all analyses adjusting for sample weights and design effects (12, 21). Records from each survey year were weighted according to person-level weights provided in annual NHIS data files. Weights were adjusted according to the number of representative years used in the analyses (12). The distribution of the variables was examined through frequency tabulations to look for outliers and examine the data errors and variables were tested for collinearity.

Since the focus of this study was on cancer survivors, the SEM analysis included only those individuals, 18 years and older, who both participated in the 1997–2012 NHIS surveys and had been diagnosed with cancer (except non-melanoma skin cancer) (n=24,810). Figure 1 displays the measurement model examined in this study. First, using confirmatory factor analyses (CFA) we estimated individual latent variable models. Second, we measured the fit of all three latent variables in a three-factor model to validate the latent variable measures using model fit indices. Third, we estimated the hypothesized structural equation model.

Figure 1.

Figure 1

Open in a new tab

Measurement model for structural equation modeling (SEM) for adult cancer survivors: National Health Interview Survey 1997–2012.

Preliminary analyses were conducted to determine whether there were sufficient associations (i.e. covariation) among the variables to permit model estimation, or if the correlations of these variables were high. Next, the distribution of antecedent variables (i.e., demographic and cancer-related characteristics), functional limitations, IADLs, and psychological distress was provided for participants with and without a cancer history by employment status was examined (Table 1). CFA were then fitted for each latent variable separately and together in a three-factor model. Finally, we estimated the hypothesized structural equation model (Figure 1).

Table 1.

A comparison of Employment and Socio-demographic characteristics of persons in the U.S. with and without a previous cancer diagnosis: National Health Interview Survey, 1997–2012

Persons with cancer diagnosis n=24,810 Persons without cancer n=382,837
Not employeda (n=11,037) Employeda (n=13,773) Employeda (n=283,7) Not employeda (n=99,070)
Age
  18–39 3,086 22.4% (20.90–23.90) 947 8.6% (7.11–10.09) 147,984 52.2% (50.18–54.22) 35,499 35.8% (33.79–37.81)
  40–64 6,293 45.7% (44.98–46.42) 2,721 24.7% (22.31–27.09) 110,865 39.0% (37.02–40.98) 28,237 28.5% (26.57–30.43)
  ≥ 65 4,394 31.9% (31.27–32.53) 7,369 66.7%(64.15–69.25) 24,918 8.8% (6.93–10.67) 35,334 35.7% (33.48–37.92)
Gender
  Male 5,115 37.2% (34.41–39.99) 3,801 34.4% (32.53–36.27) 136,429 48.1% (45.64–50.56) 40,595 41.0% (38.63–43.37)
  Female 8,658 62.8% (60.24–65.36) 7,236 65.6% (63.58–67.62) 147,338 51.9% (48.71–55.09) 58,475 59.0% (56.56–61.44)
Race/ethnicity
  Hispanic 1,139 8.3% (7.36–9.24) 821 7.4% (6.12–8.68) 52,248 18.4% (15.98–20.82) 18,853 19.0% (17.07–20.93)
  Non-Hispanic White 11,046 80.1% (76.98–83.22) 8,780 79.6% (74.84–84.36) 177,879 62.7% (59.47–65.93) 57,908 58.5% (56.36–60.64)
  Non-Hispanic Black 1,272 9.4 % (9.24–9.56) 1,156 10.5% (9.48–11.52) 41,127 14.5% (11.27–17.73) 17,005 17.1% (15.23–18.97)
  Non-Hispanic Otherb 316 2.2% (2.15–2.25) 280 2.5% (2.37–2.63) 12,513 4.4% (3.41–5.39) 5,304 5.4% (4.69–6.11)
Education
  <High school 3,680 29.2% (24.90–33.50) 3,075 29.4% (26.17–32.63) 77,367 34.0% (32.01–35.99) 30,814 37.4% (35.59–39.21)
  High school/GED 3,437 27.3% (24.88–29.72) 3,344 31.9% (29.87–33.93) 51,151 22.5% (20.42–24.58) 23,351 28.4% (26.53–30.27)
  Some college/higher 5,476 43.5% (40.26–46.74) 4,056 38.7% (36.67–40.73) 99,024 43.5% (41.08–45.92) 28,109 34.2%(32.24–36.16)
Health insurance status
  Private 9,323 70.0% (65.99–74.01) 1,754 16.2% (14.23–18.17) 208,918 75.9% (72.65–79.15) 32,774 34.0% (32.25–35.75)
  Public 2,490 18.7% (18.02–19.38) 8,488 78.4% (74.33–82.47) 23,947 8.7% (6.82–10.58) 49,354 51.2% (49.16–53.24)
  Uninsured 1,492 11.2% (8.41–13.99) 584 5.4% (4.08–6.72) 42,114 15.3% (13.24–17.36) 14,266 14.8% (14.32–15.28)
Difficulty with instrumental activities of daily living
  Yes 399 2.9% (2.17–3.63) 671 6.1% (5.20–7.00) 2,783 1.0% (0.96–1.04) 3,599 3.6% (3.58–3.62)
  No 13,374 97.1% (93.36–00.84) 10,366 93.9% (90.00–97.80) 280,984 99.0% (95.76–102.2) 95,471 96.4% (93.92–98.88)
≥ 2 Functional limitations
  Yes 4,959 36.0% (32.78–39.22) 6,799 61.6% (58.39–64.81) 11,635 4.1% (3.58–4.62) 26,055 26.3% (25.01–27.59)
  No 8,815 64.0% (60.63–67.37) 4238 38.4% (35.90–40.90) 272,132 95.9 % (92.06–99.74) 73,015 73.7% (71.16–76.24)
Psychological distress
  Yes 1,939 14.1% (12.30–15.90) 1,931 17.5%(16.23–18.77) 24,809 8.7% (7.75–9.65) 14,207 14.3% (12.4416.16)
  No 11,874 85.9% (82.74–89.06) 9,106 82.5% (78.69–86.31) 258,958 91.3% (87.85–94.75) 84,863 85.7% (83.22–88.18)
Health Status
  Good-excellent 11,362 82.5% (79.62–85.38) 7,285 66.1% (63.59–68.61) 265,173 93.5% (90.20–96.80) 78, 635 79.3% (77.28–81.32)
  Poor-fair 2,401 17.5% (15.58–19.42) 3,734 33.9% (31.87–35.93) 18,479 6.5% (6.37–6.63) 20,366 20.7%(19.63–21.77)
Time since diagnosis of cancer
  0–5 4,759 42.4% (40.30–44.50) 7,627 39.0% (35.75–42.25) -- --
  6–10 2,362 21.2% (19.10–23.30) 3,803 20.4% (18.00–22.80) -- --
  >10 4,076 36.4% (33.52–39.28) 7,976 40.6% (38.00–43.20) -- --
Cancer site
  Bladder 227 2.3% (1.74–2.86) 546 2.9% (2.85–2.95) -- --
  Breast 2,189 18.3% (15.31–21.29) 4,386 20.9% (18.42–23.38) -- --
  Colorectal 605 5.4% (3.47–7.33) 1,885 9.4% (8.23–10.57) -- --
  Lung 241 2.1% (0.78–3.42) 881 4.5% (4.04–4.96) -- --
  Prostate 980 9.7% (7.79–11.61) 2,904 16.2% (14.35–18.05) -- --
  Other cancers c 6,955 62.2% (57.54–66.86) 8,804 46.1% (43.56–48.64) -- --
Open in a new tab
a

Unweighted frequencies, column percentage (95% confidence interval)

b

Non-Hispanic other is a very broad and varied category of persons from a variety of races and ethnicities, including persons of multiple races. This group may be more diverse than the other race-ethnicity groups.

c

All other cancer sites not listed, excluding non-melanoma skin cancer.

Model fit for latent variables and the full SEM was evaluated in terms of: a) the comparative fit index (CFI), which compares the hypothesized model to a null model with no paths or latent variables; and b) the root mean square error of approximation (RMSEA), which estimates the extent to which the covariance matrix specified in the model deviates from the covariance matrix observed in the data. CFI values of 0.95 or greater and RMSEA values of 0.05 or less were used as indicators of good model fit (22, 23). The chi-square statistic, while reported, was not used in model fit interpretation due to its sensitivity to testing the null hypothesis of perfect fit when the sample size is large (24, 25). Modification indices (which guide minor modifications to model specifications) were used to improve model fit in the measurement model by correlating error variances where appropriate. Unless otherwise stated, estimates were compared using two-sided t-tests at the 0.05 level, assuming independence.

Confirmatory Factor Analysis (CFA)

Instrumental activities of daily living: CFA identified three indicators that loaded significantly onto a single latent construct. The loadings (regression weights) were 0.80 for relaxing, 0.94 for shopping, and 0.98 for socializing. This model was saturated (i.e., the number of observed variables were equal to the number of parameters being estimated), therefore model fit statistics are not available.

Both the functional limitations and psychological distress latent varibles had good model fit. Functional limitations: χ2 (22) = 1116.56, p <0.0001; CFI = 0.99; RMSEA 0.05. All nine indicators had statistically significant factor loadings with the latent construct; the loadings ranged from 0.76 (grasp) to 1.01 (stand). Psychological distress:: χ2 (9) = 2304.05, p <0.0001; CFI = 0.98; RMSEA < 0.0001

All six indicators had statistically significant factor loadings with the latent construct; the loadings ranged from 0.96 (restless) to 1.11 (hopeless).

The three latent variables (functional limitations, IADLs, and psychological distress) were combined into one model and evaluated. The model provided a good fit to the data: χ2 (125) = 3670.46, p <0.0001; CFI = 0.99; RMSEA = 0.03, indicating that all three factors (including ADL) were good representations of the data.

Measurement model

Overall, the measurement model depicted in Figure 1 provided an adequate fit to the data: χ2 (374)=3654.73, p<0.0001; CFI=0.98; RMSEA=0.04. We tested our full model among all cancer survivors only, with reported health status and employment status as the outcomes (Table 2). These analyses of anonymized data were approved by the University of Miami School of Medicine Human Subjects Committee.

Table 2.

Estimates of the effects of demographic factors, and functioning and limitation variables on Health status and Employment among U.S. adult cancer survivors: National Health Interview Survey, 1997–2012 (n=24,810).

Instrumental activities of daily
living		 Functional limitation Psychological distress Health status Employed status
β (95% C.I.) β (95% C.I.) β (95% C.I.) Odds ratios Odds ratios
Demographics
Age <0.01(0.002–0.002) *** 0.02 (0.020–0.020) *** −0. 01(−0.010–−0.010) *** 1.0 (0.91–1.13) <1.0 (0.94–0.96) ***
Male (referent) 1.00 1.00 1.00
Female 0.03 (0.027–0.029) *** 0.25 (0.240–0.260) *** 0.11 (0.104–0.116) *** 1.3 (1.22–1.28) *** <1.0 (0.88–1.02)
Race-Ethnicity
Non-Hispanic white (referent) 1.00 1.00 1.00
Non-Hispanic black 0.08 (0.079–0.081) *** 0.37 (0.360–0.380) *** 0.07 (0.057–0.077) *** 0.6 (0.62–0.66) *** 1.1 (1.01–1.11) **
Hispanic 0.05 (0.035–0.055) *** 0.18 (0.166–0.194) *** 0.12 (0.112–0.128) *** 0.9 (0.88–1.02) 1.2 (1.11–1.25) *
Non-Hispanic other/multi-race 0.04 (0.032–0.044) * 0.12 (0.106–0.134) * 0.09 (0.077–0.097) * 0.8 (0.71–0.80) * 0.9 (0.83–0.93) *
Education
Less than high school (referent) 1.00 1.00 1.00
High school/ GED −0.10 (−0.102–−0.102) *** −0.45 (−0.464–−0.436) *** −0.22 (−0.226–−0.214) *** 1.6 (1.49–1.75) *** 2.0 (1.88–2.04) ***
Some college/College degree −0.16 (−0.157–−0.155) *** 0.72 (−0.739–−0.701) *** −0.34 (−0.340–−0.340) *** 2.2 (2.07–2.35) *** 2.6 (2.56–2.70) ***
Health Insurance
Private insurance (referent) 1.00 1.00 1.00
Public insurance 0.05 (0.042–0.058) *** 0.24 (0.222–0.258) *** 0.19 (0.187–0.191) *** 0.5 (0.49–0.57) *** 0.3 (0.28–0.32) ***
Uninsured 0.003 (0.003–0.003) *** 0.002 0.001–0.003 *** 0.002 (0.002–0.002) *** 0.9 (0.81–0.91) ** 1.0 (0.99–1.03)
Cancer History
Time since diagnosis <0.01 (0.000–0.002) *** 0.01 (0.003–0.011) *** <0.01(−0.004–0.008) *** 1.0 (0.91–1.09) 1.0 (0.98–1.02)
Other cancers (referent) 1.00 1.00 1.00
Bladder cancer 0.02 (0.024–0.024) 0.16 (0.143–0.173) * <0.01 (0.001–0.008) 1.0 (0.87–1.03) 1.0 (0.85–1.15)
Breast cancer −0.01 (−0.009–−0.010) 0.02 (0.009–0.037) −0.07 (−0.082–−0.058) *** 1.0 (0.89–1.11) 1.0 (0.82–1.24)
Colorectal cancer 0.05 (0.042–0.058) *** 0.21 (0.196–0.232) *** −0.01 (−0.014–−0.006) 1.0 (0.89–1.09) 0.8 (0.70–0.82) **
Lung cancer 0.17 (0.154–0.186) *** 0.08 (0.077–0.081) *** 0.16 (0.159–0.161) *** 0.8 (0.72–0.84) ** 0.7 (0.69–0.79) **
Prostate cancer −0.02 (−0.042–0.002) ** −0.12 (−0.128–−0.112) *** −0.08 (−0.122–−0.118) ** 1.1 (1.06–1.18) * 0.8 (0.69–0.81) ***
Difficulty with functioning and limitations
Instrumental activities of daily living 0.9 (0.79–0.91) *** 0.8 (0.75–0.83) ***
Functional limitation 0.6 (0.55–0.61) *** 0.8 (0.74–0.82) ***
Psychological distress 0.5 (0.50–0.58) *** 0.9 (0.93–0.99) ***
Health status 2.2 (2.21–2.25) ***
Open in a new tab
*

p< 0.05;

**

p < 0.01,

***

p < 0.001

| 95% confidence intervals presented for beta values and odds ratios.

RESULTS

Cancer survivors versus individuals without a cancer history

The overall study sample included all persons 18 years and older who participated in the 1997–2012 NHIS surveys and had been diagnosed with cancer (n=24,810) as well as those without a previous cancer diagnosis (n=382,837). Among those currently working, 13,773 were cancer survivors and 283,767 were adults without a cancer history (Table 1).

The largest group of working cancer survivors (45.7%) was between 40–64 years of age versus 39.7% of working individuals with no cancer diagnosis in the same age range. White survivors (80.1%) were more likely to work than those without cancer (62.7%); Hispanic (8.3%) and black (9.4%) cancer survivors were less likely to work than those without a cancer history (18.4% and 14.5%, respectively). There was a comparatively higher percentage of survivors with some college or higher education among working cancer survivors compared to the not employed survivors (43.5% vs. 38.7%); this pattern, was similar among individuals without a cancer history. Cancer survivors, both employed (11.2%) and not employed (5.4%) were less likely to be uninsured compared to those without cancer whether employed (15.3%) or not employed (14.8%).

Eighty-three percent of working survivors reported good-to-excellent health; this was approximately 15% higher than the not employed survivors, but 10% lower than working persons with no previous cancer diagnosis. A significantly larger proportion of cancer survivors reported having two or more functional limitations compared to those without cancer; the percentage of functional limitations reported among not employed survivors (61.6%) was almost twice that of employed survivors (36.0%), and more than twice that of not employed persons without a cancer history (26.3%). The level of reported psychological distress (14%) was similar among employed cancer survivors and not employed persons without cancer, although less than that of not employed cancer survivors (17%).

Cancer survivors only

Among employed cancer survivors, 42% were diagnosed 5 years or less prior to the NHIS interview, 21.2% were diagnosed 6–10 years prior to their interview, and 36.4% had a cancer diagnosis over 10 years before the interview. The distribution of time since diagnosis was similar among not employed cancer survivors. However, whereas less than 40% of employed cancer survivors had one of the five most frequently diagnosed cancers, 60% of their not employed peers had been diagnosed with one of these cancers.

Health status

As Table 2 shows, female cancer survivors were significantly more likely to report good-to-excellent health status compared with male survivors (Odds Ratio (OR)=1.3; 95% confidence interval 1.22–1.28). There were notable differences between Hispanic origin and race groups with regards to good-to-excellent health status among cancer survivors: compared with whites, blacks (OR=0.6; 95% C.I. 0.62–0.66) and persons of other races (OR=0.8; 95% C.I. 0.71–0.80) were significantly less likely to report good-to-excellent health. Education was positively associated with health status: survivors with a high school diploma or General Education Development certificate (GED) (OR=1.6; 95% C.I. 1.49–1.75) and those with at least some college education (OR=2.2; 95% C.I. 2.07–2.35) were significantly more likely to report good-to-excellent health compared to survivors with less than a 12th grade level education. Survivors with public insurance (OR=0.5; 95% C.I. 0.49–0.57) and those who were uninsured (OR=0.9; 95% C.I. 0.81–0.91) were significantly less likely to report good-to-excellent health compared to those with private insurance. Having a functional limitation (OR=0.6; 95% C.I. 0.55–0.61), trouble performing IADLs (OR=0.9; 95% C.I. 0.79–0.91), and psychological distress (OR=0.5; 95% C.I. 0.50–0.58) were all significantly inversely related to health status.

Employment Status

Black (OR=1.1; 95% C.I. 1.01–1.11) and Hispanic cancer survivors (OR=1.2; 95% C.I. 1.11–1.25) were somewhat more likely to be employed compared with white cancer survivors. Education was positively associated with current employment: survivors with a high school diploma or GED (OR=2.0; 95% C.I. 1.88–2.04) and those with at least some college education (OR=2.63; p<0.001) were more likely to be employed compared to survivors with less than a 12th grade level education. Survivors with public health insurance were less likely to be employed compared to those with private health insurance (OR=0.3; 95% C.I. 0.28–0.32). Persons with functional limitations (OR=0.8; 95% C.I. 0.74–0.82), IADLs (OR=0.8; 95% C.I. 0.75–0.83), and psychological distress (OR=0.9; 95% C.I. 0.93–0.99) were significantly less likely to be employed. Adults who reported good-to-excellent health had an increased likelihood of current employment (OR=2.2; 95% C.I. 2.21–2.25) compared with those who reported poor to fair health.

Cancer History

For every increase in year since diagnosis, the reported impacts on a cancer survivors’ IADLs, functional limitations psychological distress worsened significantly in terms of statistical significance, but the changes in these health indicators were not substantial (β= 0.001, p <0.001; β= −0.01, p <0.001; and β= −0.002, p <0.01 respectively). There was no observable difference in time since diagnosis and a survivors’ reported health status (OR=1.0; 95% C.I. 0.91–1.09) or employment status (OR=1.0; 95% C.I. 0.98–1.02) (Table 2).

Compared to survivors of “other” cancers, colorectal and lung cancer survivors were significantly more likely to have difficulty performing IADLs (β=0.05, p<0.001; and β=0.17, p<0.001; respectively) and to have functional limitations (β=0.21, p<0.001; and β=0.08, p<0.001; respectively), but lung cancer survivors did report more psychological distress (β=0.16, p<0.001). Prostate cancer survivors were significantly less likely to have difficulty performing IADLs, functional limitations, and psychological distress (β= −0.02, p<0.01; β= −0.12, p<0.001; and β= −0.08, p<0.01; respectively). Bladder cancer survivors were more likely to have functional limitations (β=0.16, p <0.05) and breast cancer survivors were significantly less likely to report psychological distress (β= −0.07, p<0.001).

Compared with survivors of other cancers, survivors with a history of lung cancer were significantly less likely to have good-to-excellent health (OR=0.8; 95% C.I. 0.72–0.84), while prostate cancer survivors were more likely to report good-excellent health good-to-excellent health (OR=1.1; 95% C.I. 1.06–1.18). Survivors of lung (OR=0.7; 95% C.I. 0.69–0.79), colorectal (OR=0.8; 95% C.I. 0.70–0.82) and prostate cancer (OR=0.8; 95% C.I. 0.69–0.81) were significantly less likely to be employed compared with survivors of other cancers (Table 2).

DISCUSSION

Although factors such as age, cancer site, time since diagnosis, and gender may have an independent impact on an individual cancer survivor’s potential to continue to work during or post cancer diagnosis and treatment, our research suggests that educational level and private health insurance coverage may have a greater influence on employment decisions. During the study period, the mixed health insurance system in the U.S. was predominantly employment-based. Our model examining the effects of insurance on cancer survivors indicated that publicly insured survivors were significantly less likely to report good-to-excellent health or being employed compared to the privately insured. These findings support previous studies reporting that a majority of survivors elect to return, and/or continue to, work in order to have access to affordable healthcare and thus afford cancer treatment (26, 27). A greater disparity was undoubtedly diluted by access to public insurance, such as Medicaid among the older and qualified disabled cancer survivors (28).

Cancer survivors with at least some college education or a college degree were less likely to report functional limitations, trouble with performing IADLs, and psychological distress compared to those with lower education level. These findings are similar to past research reporting a positive link between higher education and lower levels of psychological distress (29) or functional limitations (30). When considering all other variables in the model, higher educational level had a significant and positive association with both good-to-excellent health status and with employment among cancer survivors. Education may influence psychosocial and behavioral mechanisms throughout life, including the prevention of physical and mental functional disorders; in fact, higher education has been associated with engaging in preventive health measures and healthy behaviors (31). Although the model did not investigate type of job, it is very likely that higher educated survivors have access to occupations with lower physical demands and greater likelihood of accommodation in the workplace (3, 32,33), thus facilitating a higher rate of return to worker compared with survivors in less accommodating blue collar jobs.

Independently, functional limitations and psychological distress were significant predictors of health status among cancer survivors, and were strong predictors of employment in this and other research (8, 34). While simultaneously considering the other sociodemographic factors examined, time since diagnosis did not have the expected inverse relationship associated with difficulty performing IADLs, having functional limitations, or experiencing psychological distress; this may be due to the fact that there was non participation in the NHIS by those hospitalized, or very sick and who had recently died as a result of the disease.

This suggests that other sociodemographic factors may play a greater role in a survivor’s limitations, overall health status, and the ability to continue working after cancer diagnosis. It is possible that many survivors return to work out of necessity, and encounter factors which make their work environment less conducive to work and recovery. Although many survivors who return to work may have a reduced workload, the stress of trying to prove they are able to perform their assigned tasks while recuperating from the disease or even undergoing treatment may be daunting (9, 35). The physical and psychological impacts on health and employment vary by cancer site. Furthermore, a combination of sociodemographic factors (e.g. education) may increase or decrease the negative associations between type of cancer (e.g. lung cancer) and reported health status or current employment for different individuals.

Our results indicate that when all other factors were considered, the associations between education, or race, and health and employment status differed between persons with and without a cancer history. It is possible that the ethnic disparities in reported health status and current employment among survivors may reflect the unequal distribution of employment benefits (such as early retirement packages and disability benefits) usually associated with certain occupations, higher education and better insurance plans. These employment benefits have been shown to be better among whites compared to other race and ethnic groups (36, 37). Despite a greater proportion of diagnoses at a later stage of cancer (38), black and Hispanic cancer survivors are as likely to remain in the workforce as their counterparts with no cancer diagnosis despite their reports of poorer health compared to whites.

Our findings provide an example of the underestimated inequalities present in the U.S. (39, 40), and highlight the need to simultaneously investigate factors that may mediate the relationships between sociodemographic factors, poor health, and employment among cancer survivors. In addition, it will be important to monitor the implementation of the Affordable Care Act as to whether the ability to retain or gain health coverage despite having cancer as a pre-existing conditions is associated with an increase in the number of lower income Americans who can afford insurance coverage and thus return to work for reasons other than an insurance based access to care.

Additionally, this study illustrates the need for workplace interventions to facilitate cancer survivors staying in or returning to work after a cancer diagnosis and treatment. Effective interventions would reduce work days lost, increase productivity, and reduce psychosocial stresses associated with unemployment and treatment costs among cancer survivors.

Limitations

These findings should be interpreted in the context of the study’s limitations. The NHIS data are cross-sectional, which precludes the differentiation between correlative associations and causal relationships when examining persons with a history of cancer. Although the NHIS survey asks persons to report a physician diagnosis of cancer, the data were self-reported and, thus, contingent on participant recall and willingness to report this diagnosis; thus, current employment and cancer diagnosis were not validated. Important clinical variables (such as stage at diagnosis or recurrences of cancer) were also not available; and type of treatment received was made available only in the 2010 NHIS cancer control module.

Nevertheless, the NHIS is representative of the entire U.S. civilian population; as such there are significant strengths to the conclusions derived from these analyses. Our analyses span 14 years of data, thus providing a large sample size and allowing for the estimation of outcomes among a variety of population subgroups. Furthermore, our findings are generalizable to the U.S. population. While we did not model all possible associations between health indicators and employment, the use of SEM allowed the simultaneous evaluation of all the relationships of the variables completely and simultaneously.

CONCLUSIONS

In summary, overall the majority of survivors return to or continue to work post cancer diagnosis. As cancer control and treatment continue to improve and more people work past the traditional retirement age, the number of cancer survivors in our rapidly aging workforce will continue to increase. Health policy analysts, oncologists, epidemiologists, employers, the U.S. Department of Labor, and insurance companies will have to work collectively on various aspects of improving access to and extent of health insurance coverage for persons diagnosed with cancer across all occupational sectors. Our findings clearly point to the need for further research on access to healthcare and disability benefits available to cancer survivors within the workplace. If working cancer survivors experience better health, improved benefits, and worksite-accommodations among those jobs/occupations that do not usually offer these services, the burden on the individual and the healthcare system would be reduced as more survivors would have the ability to return to work or to continue to work throughout their treatment.

Acknowledgments

Funding: This work was supported in part by a National Cancer Institute (NCI) fellowship at the National Institutes of Health (1F31CA153937); the National Institute of Occupational Safety and Health (NIOSH) R01 OH03915); and the European Regional Development Fund Programme and European Social Fund Convergence Programme for Cornwall and the Isles of Scilly (the University of Exeter Medical School).

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

REFERENCES

  • 1.Hoffman B. Cancer survivors at work: a generation of progress. CA Cancer J Clin. 2005;55:271–280. doi: 10.3322/canjclin.55.5.271. [DOI] [PubMed] [Google Scholar]
  • 2.Horner MJ, Ries LAG, Krapcho M, Neyman N, Aminou R, Howlander N, et al. National Cancer Institute. SEER Cancer Statistics Review, 1975–2006 [Google Scholar]
  • 3.Amir Z, Brocky J. Cancer survivorship and employment: epidemiology. Occup Med (Lond) 2009;59(6):373–377. doi: 10.1093/occmed/kqp086. [DOI] [PubMed] [Google Scholar]
  • 4.Spelten ER, Sprangers MAG, Verbeek JHAM. Factors reported to influence the return to work of cancer survivors: a literature review. Psychooncology. 2002;11:124–131. doi: 10.1002/pon.585. [DOI] [PubMed] [Google Scholar]
  • 5.Short P, Vasey JJ, Tuncelli K. Employment pathways in a large cohort of adult cancer survivors. Cancer. 2005;103:1292–1301. doi: 10.1002/cncr.20912. [DOI] [PubMed] [Google Scholar]
  • 6.Hewitt M, Rowland JH, Yancik R. Cancer survivors in the United States: age, health, and disability. J Gerontol A Biol Sci Med Sci. 2003;58:82–91. doi: 10.1093/gerona/58.1.m82. [DOI] [PubMed] [Google Scholar]
  • 7.Bouknight RR, Bradley CJ, Luo Z. Correlates of Return to Work for Breast Cancer Survivors. Journal of Clinical Oncology. 2006;24(3):345–353. doi: 10.1200/JCO.2004.00.4929. [DOI] [PubMed] [Google Scholar]
  • 8.Rowland JH, Yancik R. Cancer survivorship: the interface of aging, comorbidity, and quality care. J Natl Cancer Inst. 2006;98:504–505. doi: 10.1093/jnci/djj154. [DOI] [PubMed] [Google Scholar]
  • 9.Dellapasqua S, Colleoni M, Goldhirsch A. Focusing on physical function limitations in elderly women surviving cancer: any opportunity for improvement? Nature Clinical Practice Oncology. 2006;3:598–599. doi: 10.1038/ncponc0633. [DOI] [PubMed] [Google Scholar]
  • 10.Sweeney C, Schmitz KH, Lazovich D, Virnig BA, Wallace RB, Folsom AR. Functional Limitations in Elderly Female Cancer Survivors. J Natl Cancer Inst. 2006;98(8):521–529. doi: 10.1093/jnci/djj130. [DOI] [PubMed] [Google Scholar]
  • 11.Sehl ME, Satariano WA, Ragland DA, Reuben DB, Naeim A. Attribution of functional limitation to cancer decreases in the year following breast cancer diagnosis in older patients. Crit Rev Oncol Hematol. 2009;71(1):62–69. doi: 10.1016/j.critrevonc.2008.10.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Botman SL, Moore TF, Moriarity CL, Parsons VL. Design and estimation for the National Health Interview Survey, 1995–2004. National Center for Health Statistics. Vital Health Stat. 2000;2(130) [Google Scholar]
  • 13.National Health Interview Survey (NHIS) Division of Health Interview Statistics, National Center for Health Statistics. Hyattsville, Maryland: Centers for Disease Control and Prevention. U.S. Department of Health and Human Services; [Accessed July 19, 2012]. Survey Description documents. http://www.cdc.gov/nchs/nhis/quest_data_related_1997_forward.htm . [Google Scholar]
  • 14.American Cancer Society. Atlanta: American Cancer Society; 2012. Cancer Facts & Figures 2012. [Google Scholar]
  • 15.Verbrugge LM, Jette AM. The disablement process. Social Science and Medicine. 1994;38:16–14. doi: 10.1016/0277-9536(94)90294-1. [DOI] [PubMed] [Google Scholar]
  • 16.National Cancer Institute. Dictionary of Cancer terms. [Accessed June 6, 2011]; http://www.cancer.gov/dictionary?cdrid=430402 .
  • 17.Kessler RC, Andrews G, Mroczek D, Ustun TB, Wittchen HU. The World Health Organization Composite International Diagnostic Interview Short-Form (CIDI-SF) Int J Methods Psychiatr Res. 1998;7:171–185. [Google Scholar]
  • 18.Cohen RA. Using the National Health Interview Survey to monitor health insurance and access to care. Proceedings from the Annual meeting of the American Statistical Association: Section on Health Policy statistics. JSM. 2009:108–118. 2009. [Google Scholar]
  • 19.SAS Institute. Survey Data Analysis. [Accessed November 10, 2014]; Available at: http://support.sas.com/documentation/cdl/en/statug/66859/HTML/default/viewer.htm#statug_introsamp_sect019.htm. [Google Scholar]
  • 20.Muthén B, Muthén L. Mplus User's Guide. Sixth Edition. Los Angeles, CA: Muthén & Muthén; 1998–2010. [Google Scholar]
  • 21.Parsons VL, Moriarity CL, Jonas K, Moore TF, Davids KE, Tompkins L. Design and estimation for the National Health Interview Survey, 2006–2015. National Center for Health Statistics. Vital Health Stat. 2014;2(165) [PubMed] [Google Scholar]
  • 22.Kline RB. Principles and Practice of Structural Equation Modeling. 3rd ed. New York: Guilford Press; 2010. [Google Scholar]
  • 23.Kenny DA. Assessing model fit. [Accessed September 23, 2010]; 01/22/2010. http://davidakenny.net/cm/fit.html. [Google Scholar]
  • 24.Bentler PM, Bonnet DC. Significance Tests and Goodness of Fit in the Analysis of Covariance Structures. Psychological Bulletin. 1980;88(3):588–606. [Google Scholar]
  • 25.Jöreskog K, Sörbom D. LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language. Chicago, IL: Scientific Software International Inc.; 1993. [Google Scholar]
  • 26.Bradley CJ, Neumark D, Luo Z, Bednarek HL. Employment-contingent health insurance, illness, and labor supply of women: evidence from married women with breast cancer. Health Econ. 2007;16:19–37. doi: 10.1002/hec.1191. [DOI] [PubMed] [Google Scholar]
  • 27.Finkelstein EA, Tangka FK, Trogdon JG, Sabatino SA, Richardson LC. The personal financial burden of cancer for the working age population. Am J Manag Care. 2011;15:801–806. [PubMed] [Google Scholar]
  • 28.Nekhlyudov L, Madden J, Johnson Graves A, Zhang F, Soumerai SB, Ross-Degnan D. Cost-related medication nonadherence and cost-saving strategies used by elderly Medicare cancer survivors. J Cancer Surviv. 2011;5(4):395–404. doi: 10.1007/s11764-011-0188-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Brännlund A, Hammarström A. Higher education and psychological distress: a 27-year prospective cohort study in Sweden. Scand J Public Health. 2014 Mar;42(2):155–162. doi: 10.1177/1403494813511559. [DOI] [PubMed] [Google Scholar]
  • 30.Zimmer Z, House JS. Education, income, and functional limitation transitions among American adults: contrasting onset and progression. Int J Epidemiol. 2003;32:1089–1097. doi: 10.1093/ije/dyg254. [DOI] [PubMed] [Google Scholar]
  • 31.Fletcher JM, Frisvold DE. Higher Education and Health Investments: Does More Schooling Affect Preventive Health Care Use? J. Hum. Cap. 2009 Summer;3(2):144–176. doi: 10.1086/645090. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Center for economic and policy research. [Accessed October 26, 2014];Hard work? Patterns in Physically demanding labor among older workers. http://www.cepr.net/documents/publications/older-workers-2010-08.pdf . [Google Scholar]
  • 33.Bureau of Labor Statistics. U.S. Department of Labor, The Economics Daily, White-collar workers more likely to receive disability benefits in small establishments. [Accessed October 26, 2014]; http://www.bls.gov/opub/ted/1999/May/wk3/art04.htm .
  • 34.Clarke TC, Soler-Vila H, Lee DJ, Arheart KL, Ocasio MA, LeBlanc WG, Fleming LE. Working with cancer: Health and disability disparities among employed cancer survivors in the U.S. Prev Med. 2011;53(4–5):331–334. doi: 10.1016/j.ypmed.2011.07.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Pransky G. Work and Cancer Survivors. J Occup Environ Med. 2010;52(1):111. [Google Scholar]
  • 36.U. S. Bureau of Labor Statistics. Labor force characteristics by race and ethnicity, 2008. Washington, DC: U.S. Department of Labor; 2009. [Accesseed September 23, 2010]. http://www.bls.gov/cps/cpsrace2009.pdf . [Google Scholar]
  • 37.Toldson IA, Snitman A. Education Parity and Economic Disparities: Correcting education attainment discrepancies among Black people in the United States. [Accessed September 10, 2011];Journal of Negro Education. 2010 Available at http://www.highbeam.com. [Google Scholar]
  • 38.Soler-Vila H, Kasl SV, Jones BA. Prognostic significance of psychosocial factors in African-American and white breast cancer patients: a population-based study. Cancer. 2003;98(6):1299–1308. doi: 10.1002/cncr.11670. [DOI] [PubMed] [Google Scholar]
  • 39.Wilson WJ. The Truly Disadvantaged: The inner city, the underclass, and public policy. Chicago: University of Chicago Press; 1987. [Google Scholar]
  • 40.Wilson WJ. Social Theory and the Concept 'Underclass'. In: Grusky DB, Kanbur R, editors. Poverty and Inequality. Stanford: Stanford University Press; 2006. [Google Scholar]

RESOURCES