Factors influencing changes in employment among women with newly diagnosed breast cancer

Abstract

BACKGROUND:

Although studies have demonstrated that women are less likely to work after they are diagnosed with breast cancer, the influence of cancer treatments on employment is less clear. The authors of this report assessed whether chemotherapy or radiation therapy was associated with a disruption in employment during the year after a breast cancer diagnosis.

METHODS:

Using a database of health insurance claims that covered 5.6 million US residents, 3233 women aged ≤63 years were identified who were working full time or part time when they were diagnosed with breast cancer between 1998 and 2002. All changes in employment during the year after a breast cancer diagnosis were identified. Using a Cox proportional hazards model that incorporated time-varying treatment variables, the authors evaluated the impact of chemotherapy and radiation therapy on the likelihood of experiencing an employment disruption.

RESULTS:

Although most women (93%) continued to work, chemotherapy recipients were more likely than nonrecipients to go on long-term disability, stop working, or retire (hazards ratio, 1.8; P < .01). Women aged ≥54 years were more likely to experience a change in employment than women aged ≤44 years (P < .01). Radiation therapy did not influence employment (P = .22).

CONCLUSIONS:

In this population of employed, insured women, chemotherapy had a negative impact on employment. This finding may aid treatment decision making and could foster the development of interventions that support a patient's ability to continue working after treatment. It also reinforces the need to assess the impact of treatments, especially new treatments, on patient-centered outcomes such as employment. Cancer 2009. © 2009 American Cancer Society.

A substantial proportion of adult cancer survivors—40% to 54%—reduce their work hours or stop working altogether after their cancer diagnosis.1-3 Working women with breast cancer are no exception.4-7 Older women; African-American women; and those who have physically demanding jobs, less accommodating employers, more advanced cancer, and more comorbidities are especially likely to experience a disruption of employment.5, 6, 8-12 Although temporary changes in employment may be needed to complete therapy and could be welcomed by patients, permanent changes could lead to the loss of income, work-related benefits, social connections, and satisfaction and may precipitate anxiety or depression. The resulting financial strain and psychologic distress could have a substantial detrimental impact on quality of life.2, 13-16

Many of the ≥200,000 women who are diagnosed with breast cancer each year in the US are employed.4, 6, 17 Unfortunately, relatively little is known regarding the influence of chemotherapy and radiation therapy on a woman's desire and ability to work. Clinical trials usually do not report the effects of treatment on employment. Although 2 retrospective surveys of employed women with newly diagnosed breast cancer indicated that chemotherapy did not result in a reduction in employment,8, 18 1 prospective cohort study indicated that receipt of chemotherapy was associated with more lost wages and longer absence from work.19

Whether or not chemotherapy and radiation therapy lead to permanent changes in employment status remains unclear. These treatments are being given increasingly frequently for ever smaller absolute benefits, and newer more intensive treatment regimens could be more likely to affect employment than traditional regimens. Therefore, understanding the impact these treatments have on employment could help patients make informed treatment decisions. In the current study, our objective was to determine whether chemotherapy or radiation therapy was associated with a major disruption in employment during the year after a breast cancer diagnosis.

MATERIALS AND METHODS

The Medstat MarketScan Commercial Claims and Encounters Research Database served as the data source for this analysis. Medstat is a medical information company that compiles data from health plans that provide insurance to large companies, state and local governments, and public organizations in the US. The MarketScan database includes claims and enrollment records for >5.6 million individuals receiving employer-sponsored health insurance. Similar to a previous analysis,20 we used data from January 1, 1998 through December 31, 2002 to identify women ages 18 years to 63 years with at least 2 breast cancer diagnosis codes (174.x from the International Classification of Diseases, ninth edition [ICD-9]). The codes had to be dated at least 30 days apart, at least 1 code had to be from a face-to-face encounter with a healthcare provider, and women had to be enrolled continuously for at least 3 months before through 12 months after their first breast cancer diagnosis. Patients with other cancer diagnoses were excluded. On the basis of a previously developed algorithm for identifying incident breast cancer using administrative data,21 women had to have at least 1 code for a breast cancer biopsy/surgery.

Employment status was recorded on a monthly basis for up to 12 months after the first breast cancer diagnosis. We categorized employment status using definitions created by Medstat, which characterized employment status monthly based on data provided by the employer-sponsored health plans. The 8 Medstat categories were full time, part time, early retiree, retiree, long-term disability, Comprehensive Omnibus Reconciliation Act (COBRA) insurance, or unknown. Women whose employment data were unavailable and those who were not working full time or part time at the time of their first diagnosis of breast cancer were excluded from the analysis. For women who were working full time, we recorded a change in employment for those whose work status changed to part time, early retiree, retiree, long-term disability, COBRA, or unknown. For women who were working part time, we recorded a change in employment for those whose work status changed to early retiree, retiree, long-term disability, COBRA, or unknown. We assumed that each of these employment changes reflected a decrease in hours worked or a discontinuation of employment. This included women whose employment status changed to COBRA or unknown, because we knew that these women were no longer working in their old jobs and we believed they were unlikely to start new jobs while receiving cancer therapy. Patients were censored after their first employment status change. We did not assess return to work, because the dataset did not capture employment at other employers and did not allow the linkage of records between employers. We focused our analysis only on the year after the breast cancer diagnosis, because the majority of treatments occur or start within this time frame.

We extracted data on age (categorized into quintiles defined as ages ≤44 years, 45-49 years, 50-53 years, 54-57 years, or 58-63 years), health plan type (basic comprehensive, health maintenance organization, point of service, preferred provider organization, or point of service with capitation), and region of residence (Northeast, Midwest, South, West, or unknown) for each woman at her first diagnosis of breast cancer. Noncancer comorbidity was categorized using the Klabunde modification of the Charlson score.22-24 A comorbid condition was considered present if 2 claims for that condition were made at least 30 days apart from 3 months before to 12 months after the first breast cancer.20 Metastatic status was identified using ICD-9 codes for secondary malignant neoplasms (codes 197-199); because this analysis was based on insurance claims, more detailed stage data were not available. Variables were categorized as listed in Table 1. Medical claims during the 12 months after a breast cancer diagnosis were used to identify chemotherapy, radiation therapy, and hospitalization for serious chemotherapy-related adverse events; the specific claims that were used included Current Procedural Terminology codes, ICD-9 codes, Diagnosis Related Group codes, and J&Q codes, as described previously.20 Chemotherapy medications included all major cytotoxic agents that were used to treat breast cancer. Because anthracyclines have been associated with additional adverse effects,25 the receipt of an anthracycline was coded separately.

Table 1. Characteristics of Breast Cancer Patients*

Characteristics	Percentage of Patients
ER indicates emergency room. * Based on 3233 women who were employed either full time or part time at the time of their first breast cancer diagnosis using data from the MarketScan Commercial Claims and Encounters Research Database. † Age groups represent approximate quintiles of the entire cohort. ‡ Metastatic status was identified using International Classification of Diseases, ninth edition codes for secondary malignant neoplasms (codes 197-199). § Comorbid diagnoses were considered present if 2 claims were made at least 30 days apart during the 3 months before and the 12 months after the first breast cancer diagnosis (other cancer diagnoses were excluded). ‖ Includes only 3229 patients; data regarding type of health plan were absent for 4 records.
Age, y†
≤44	18.5
45-49	21.1
50-53	23.6
54-57	19.8
58-63	17
Metastatic status‡
Nonmetastatic	88.7
Metastatic	11.3
Comorbidity score§
0	92.6
≥1	7.4
Health plan type‖
Basic/comprehensive	34.8
Health maintenance organization	7.6
Point of service	13.9
Preferred provider organization	31
Point of service with capitation	12.8
Region of residence in US
Northeast	14
Midwest	20.3
South	59.6
West	2.4
Unknown	3.8
Cancer-directed therapies
No chemotherapy	46.2
Chemotherapy	53.8
No radiation therapy	42.1
Radiation therapy	57.9
Hospitalizations/ER visits for chemotherapy-related adverse events
0	89.1
≥1	10.9

We calculated the incident rate of change in employment, expressed as the number of employment changes per 1000 patient-years, and used Student t tests and analyses of variance to compare this rate by patient characteristics and receipt of treatments. Claims for chemotherapy and radiation therapy were used to derive on-treatment intervals for these therapies. The 30-day period after a treatment was considered “time on treatment,” and all other periods were considered “time off treatment.” When a patient received a second treatment within the 30-day window, which typically is the case (for example, most chemotherapy regimens consist of 4 to 8 treatments given every 2 to 3 weeks), the on-treatment period was considered continuous until 30 days after the last treatment. An employment change was attributed to a therapy only if it occurred during an on-treatment period for that therapy.

A Cox proportional hazards model was used to assess the effect of chemotherapy and radiation therapy on the likelihood of experiencing a change in employment while on treatment. Time-varying treatment variables for chemotherapy and radiation therapy were included in the model, so women could contribute information to the treatment groups when they were receiving therapy and to the control group when they were not receiving therapy. The model also included the covariates described above. To explore whether other factors might help explain the association between chemotherapy and change in employment, we modified the base model in several ways. First, we added a time-varying covariate for women who experienced hospitalizations for serious chemotherapy-related adverse events to test whether these events explained any part of the association between chemotherapy and change in employment. Second, we added a variable reflecting the receipt of anthracycline chemotherapy to assess whether these agents were responsible for any part of the association between chemotherapy and employment change. Third, to test whether proportional hazards was a reasonable assumption, we added time-by-predictor interaction variables (eg, time*age, time*chemotherapy, time* radiation, time*metastatic status). The interaction terms tested whether the effects of the predictors changed after a certain time had elapsed; such effects are referred to commonly as change points. Because no significant interaction effects were identified, we concluded that the proportional hazards assumption was reasonable.

In sensitivity analyses, we restricted the cohort to the 70% of employment changes to early retiree, retiree, and long-term disability status (ie, excluding changes to COBRA and unknown); the results were similar and are not presented. We also repeated analyses restricting the cohort to women who were working full time at diagnosis (99% of women); again, the results were similar and are not presented. Finally, to explore whether the relation between chemotherapy and employment varied for younger women versus older women, we repeated analyses after stratifying at the median age (51 years). Model results are presented as hazards ratios (HRs) with 95% confidence intervals (95% CIs). Statistical analyses were performed using SAS software (version 9.2; SAS Institute, Inc, Cary, NC). All statistical tests were 2-sided, and P values <.05 were considered significant.

RESULTS

Of the 5.6 million insured lives in the MedStat dataset from January 1998 to December 2002, there were 4068 women aged <64 years who had newly diagnosed breast cancer and at least 1 year of follow-up data. Seventy-nine percent of these women (3233) were working full time or part time when they first were diagnosed with breast cancer; these women comprised the main cohort that we analyzed in the current study. The majority of the 835 patients who were not included in the final cohort were early retirees at the time of their breast cancer diagnosis. They tended to be older (mean age, 58.3 years vs 50.6 years; difference, 7.7 years [95% CI, 7.2-8.2 years]) and were less likely to receive chemotherapy (40.5% vs 53.8%; P < .001).

The baseline characteristics of the 3233 women who were working full time or part time are listed in Table 1. Most women had nonmetastatic disease and no comorbidities (Table 1). Approximately 54% received chemotherapy, 58% received radiation therapy, and 6.6% experienced a change in employment. The most common employment change was from full time to early retiree (67%), followed by full time to unknown (12%), full time to COBRA (9%), full time to retiree (6%), and full time to long-term disability (5%). Only 2% of all changes in employment were from part time to another status. Not controlling for other predictors, women who were older and those who had metastatic cancer or a comorbidity score ≥1 demonstrated a higher rate of employment change than others (Table 2). Chemotherapy recipients also demonstrated a higher rate of employment change than nonrecipients, but this unadjusted difference was not statistically significant.

Table 2. Rate of Changes in Employment (Unadjusted)*

Characteristic	Changes in Employment per 1000 Patient-Years	P†
* Based on 3233 women employed either full or part time at the time of their first breast cancer diagnosis, using data from the MarketScan Commercial Claims and Encounters Research Database. † P values were obtained using Student t tests for binary variables and analyses of variance-based F tests for categoric variables. ‡ Age groups represent quintiles of the entire cohort. § Metastatic status was identified using International Classification of Diseases, ninth edition codes for secondary malignant neoplasms (codes 197-199). ‖ Comorbid diagnoses were considered present if 2 claims were made at least 30 days apart during the 3 months before and the 12 months after the first breast cancer diagnosis (other cancer diagnoses were excluded). ¶ Includes only 3229 patients; data regarding type of health plan were absent for 4 records.
Age, y‡
≤44	43	<.01
45-49	48
50-53	58
54-57	73
58-63	174
Metastatic status§
Nonmetastatic	75	.33
Metastatic	94
Comorbidity score‖
0	73	.06
≥1	113
Health plan type¶
Basic/comprehensive	60	.24
Health maintenance organization	90
Point of service	71
Preferred provider organization	87
Point of service with capitation	87
Region of residence in US
Northeast	75	.36
Midwest	88
South	75
West	33
Unknown	117
Cancer-directed therapies
Off chemotherapy	75	.67
Receiving chemotherapy	80
Off radiation therapy	77	.47
Receiving radiation therapy	68

Controlling for other observed patient characteristics and using time-varying treatment variables to examine the effects of treatments on employment, women who were receiving chemotherapy had 1.8-fold greater risk of experiencing a change in employment versus women who were not receiving chemotherapy, and older women were more likely to experience a change in employment than younger women (Table 3). Radiation therapy, having more comorbid conditions, and having metastatic cancer were not found to be associated with employment change. In additional models, receiving anthracycline chemotherapy and being hospitalized for a chemotherapy-related adverse effect also were not associated with employment change. None of the interaction terms that were added to the base model (including chemotherapy*age, radiation therapy*age, chemotherapy* metastatic status, and age*metastatic status) were statistically significant. When we repeated the model stratified by age, a significant association between chemotherapy use and employment change was observed among women aged >51 years (HR, 1.9; 95%CI, 1.2-3), but not among women aged ≤51 years (HR, 1.5; 95%CI, 0.5-3.9). There was less power to detect an association in the younger women, because the relative difference between the event rates in the chemotherapy and nonchemotherapy subgroups was smaller.

Table 3. Hazards Ratios for Change in Employment (Adjusted)*

Characteristic	HR for Change in Employment	95% CI	P
HR indicates hazards ratio; 95% CI, 95% confidence interval; Ref, referent group. * Based on 3233 women who were employed either full time or part time at the time of their first breast cancer diagnosis using a Cox proportional hazards model controlling for age, metastatic status, comorbidity, health plan type, region of residence, chemotherapy, and radiation therapy. † Age groups represent quintiles of the entire cohort. ‡ Metastatic status was identified using International Classification of Diseases, ninth edition codes for secondary malignant neoplasms (codes 197-199). § Comorbid diagnoses were considered present if 2 claims were made at least 30 days apart during the 3 months before and the 12 months after the first breast cancer diagnosis (other cancer diagnoses were excluded). ‖ Includes only 3229 patients; data regarding type of health plan were absent for 4 records. This variable may serve as a proxy for type of employer, because different employers adopt different health plans and different benefit packages.
Age, y†
≤44	Ref	Ref	Ref
45-49	1.3	0.8-2.2	.35
50-53	1.6	0.9-2.6	.08
54-57	2.0	1.2-3.3	<.01
58-63	4.8	3.1-7.6	<.01
Metastatic status‡
Metastatic (vs nonmetastatic)	1.0	0.7-1.6	.88
Comorbidity score§
≥1 (Vs none)	1.2	0.8-1.8	.50
Health plan type‖
Basic/comprehensive	Ref	Ref	Ref
Health maintenance  organization	1.6	0.9-2.8	.08
Point of service	1.2	0.8-1.9	.36
Preferred provider   organization	1.4	1.0-2.0	.03
Point of service with   capitation	1.9	1.2-2.9	<.01
Region of residence in US
Northeast	Ref	Ref	Ref
Midwest	1.1	0.7-1.8	.58
South	1.0	0.7-1.5	.93
West	1.7	0.8-3.6	.15
Unknown	0.3	0.1-0.9	.04
Cancer-directed therapies
Chemotherapy (vs no  chemotherapy)	1.8	1.2-2.5	<.01
Radiation therapy (vs no  radiation therapy)	1.3	0.9-1.9	.22

DISCUSSION

We observed that chemotherapy recipients were more likely to go on long-term disability, stop working, or retire compared with women who were not receiving chemotherapy even after controlling for other patient characteristics and treatment variables. Although our analysis could not characterize the exact mechanism by which chemotherapy affects employment, we suspect that chemotherapy-related adverse events decrease patients' desire or ability to work or both. It is noteworthy that hospitalizations for acute chemotherapy-related adverse events did not appear to mediate this correlation. This finding suggests that chronic chemotherapy-related adverse events are more likely to influence employment than short-term chemotherapy-related side effects, even when those short-term side effects are serious enough to cause hospitalization. We also observed that the association between chemotherapy and employment was greater for older women than for younger women, although the relation for younger women was estimated imprecisely because of the small number of employment changes in that group. No other patient characteristics were associated with a greater risk of experiencing a change in employment, nor was receipt of radiation therapy.

Quality-of-life studies identify chemotherapy as the breast cancer treatment that causes the most long-term physical and emotional morbidity.26-30 Many women with breast cancer who stop working within 6 months of their diagnoses do so because of treatment-related symptoms.6 However, 2 previous studies (a survey of 145 breast cancer patients who were diagnosed between 1986 and 1991 and a study of 416 breast cancer patients who were diagnosed during 2001 and 2002) reported no association between the receipt of chemotherapy and employment.8, 18 Several factors could explain why we observed an association when these other studies did not. First, our analysis may have had greater power to detect a relation, because the sample size was larger. Moreover, the multivariate model accounted for predictor variables that could have masked the effect of chemotherapy and incorporated time-varying covariates that may have helped isolate the component of variation attributable to chemotherapy. Second, we obtained information on cancer treatments and employment directly from patients' insurance companies rather than patient surveys, which may be subject to response and recall bias. Finally, patients with more advanced cancer and those who have more difficulty with chemotherapy may have been less likely to complete surveys but more likely to experience employment changes. Our results build on and reinforce the findings of a recently reported, contemporaneous study indicating that chemotherapy recipients had, on average, 19 additional weeks of absence from work and 8% less compensation during their absence.19

In our large, population-based cohort, 93% of women who were working when they were diagnosed with breast cancer still were working 12 months later. This is higher than several previous estimates of employment among women with newly diagnosed breast cancer. One study compared 646 breast cancer survivors with 890 controls and observed that 79% of the cancer survivors and 85% of the controls were working 3 years later.4 A second study, a longitudinal analysis of working women with newly diagnosed breast cancer identified from the Metropolitan Detroit Cancer Surveillance System, observed that 82% were employed 12 months after their diagnosis.6, 8 Our cohort may have been less likely to experience an employment change than the general population of working women with breast cancer for several reasons. First, we required continuous enrollment in a health plan for 12 months. Women who disenrolled before 12 months may have experienced an employment change but were not included in our analysis. However, women who left the workforce but maintained their health insurance benefits through COBRA were included in the analysis. Second, our analysis only included women who worked for employers who were sponsoring the health plans. Women who were classified as spouses or dependants may have been more likely to stop working but were excluded from the analysis, because the Medstat file did not contain their employment information. Because all of the women in our study risked losing their benefits if they stopped working, they had an incentive to keep working. Finally, everyone in our cohort worked for a large employer. Compared with small employers or women who are self-employed, large employers may be more able to accommodate employees and, in some cases, are required to do so by law.8, 31, 32

Our analysis has several limitations. First, although employment status was defined in a nontraditional way, the categories we used (full time, part time, early retiree, retiree, COBRA, and long-term disability) came directly from employer-sponsored health plans as coded by Medstat and reflected major changes in employment. Moreover, our findings were robust to different definitions of employment change. In other words, restricting the analysis to women who were employed full time when they were diagnosed with breast cancer or to women whose employment status changed only to retiree, early retiree, or long-term disability (ie, excluding COBRA and unknown) yielded similar results. We were not able to control for return to work at other employers or to assess employment changes that occurred more than 1 year after a breast cancer diagnosis. However, at least 1 study suggests that few employment changes occur more than 1 year after a breast cancer diagnosis.7

Second, information concerning several potentially important covariates was not available from the Medstat file. Some of these variables, such as tumor grade and receptor status, are unlikely to have a direct impact on the relation between chemotherapy use and employment change. They may indirectly affect employment through their influence on treatment decisions. However, treatments were included in the model, and therefore we believe that excluding grade and receptor status did not limit the validity of our findings. Other variables, such as socioeconomic status, race/ethnicity, occupational characteristics, marital status, spousal employment, other family members' insurance coverage, second jobs, or return to work, cannot be excluded as potential confounding or mediating variables. Future analyses that include these variables would help to assess their impact on employment change.

Third, we could control for stage only to a limited degree, because detailed stage data were not available. Any impact that stage could have on employment probably would be mediated either through treatment-related adverse effects caused by more aggressive treatment or cancer-related symptoms. Our current analysis controlled for treatments and metastatic status (patients with metastatic disease have the greatest chance of experiencing cancer-related symptoms). Although we did not observe an association between advanced stage (ie, metastatic disease) and change in employment, previous studies have reported such an association.2, 3, 8 Our study may have lacked sufficient power to identify an association, because only a small fraction of the patients had metastatic disease. Although it is possible that unmeasured variations in stage could explain the relation that we observed between chemotherapy and employment status, this seems unlikely, because changes in employment status were attributed to chemotherapy only if they occurred within 30 days of receiving chemotherapy. Because disease severity could shape the decisions made by patients and providers, its influence on employment should be studied further.

For many patients with breast cancer, the trend is to offer more aggressive therapies that lead to more frequent and more serious consequences (eg, taxanes and trastuzumab for the adjuvant treatment of women with lymph node-positive and HER-2–positive breast cancer, respectively). Clinical trials provide few details regarding patients' ability to continue working—a practical consequence of therapy that may be of particular interest to patients. Future clinical trials should assess the impact of treatment on patient-centered outcomes such as employment. In addition, research studies should strive to gain a better understanding of the reasons for the association between chemotherapy and employment (ie, do changes in employment result directly from toxicities, or do they reflect a priori choices made by patients receiving chemotherapy), which chemotherapy-related toxicities have the greatest impact on employment, and which factors prevent women who want to return to work from doing so.

Although knowing that chemotherapy might affect employment is unlikely to have a substantial impact on the decisions made by many women with breast cancer, this information could prove particularly valuable for women in whom chemotherapy offers relatively modest benefits and employment change confers significant detrimental consequences. For example, although it is doubtful that an association between chemotherapy and employment would have an impact on the decision making of a woman aged 46 years with stage III, HER-2–positive breast cancer who works part time and is enrolled in a health plan through her husband's employer, this information could influence the decision making of a woman aged 62 years with stage II, estrogen receptor-positive breast cancer who is employed full time and could lose her health insurance if she stops working. Regardless, knowledge of a link between the receipt of chemotherapy and experiencing a change in employment could lead to stronger efforts, by employers and others, to support women while they are being treated for cancer and to provide rehabilitative services that help women return to work if that is what they desire.

Conflict of Interest Disclosures

Supported by Grants PO1-HS10803 from the Agency for Healthcare Research and Quality and R01 CA104118 from the National Cancer Institute.

Dr. Hassett received salary support from Grant R25 CA092203 and an American Society of Clinical Oncology Career Development Award.