According to the Centers for Disease Control and Prevention, an estimated 1.8 million people were diagnosed with cancer in the US in 2020 [5]. This has important implications for employers and workers alike. However, limited data on the work absence costs or lost time are available. Using a mix of claims data and proxies for cost (based on time absent and estimated average cost per day) within the most costly health and productivity conditions from 1997 to 1999, cancer of the female breast was ranked number 15, and cancer of the colon and rectum ranked number 20 [6]. A follow-up paper reported that “any cancer” was ranked in the top 10 conditions by estimated dollar impact, based on medical and productivity losses per employee [7].

A literature review revealed studies on absence and disability costs for patients with breast cancer,[8-11] multiple myeloma [12] bone metastases, [13] and non-Hodgkin lymphoma [14]. One early study using administrative claims compared multiple types of cancer with matched controls [15]. Some researchers also analyzed medication administration (oral vs injectable therapy) [12] and self- and physician-administered cancer treatment [16]. Several studies have reported patients reduce their work time due to their cancer, with many patients eventually returning to work within 5 years following diagnosis [17-18].

This present retrospective research compares all-cause payments for medical, prescription, (short- and long-term) disability, workers’ compensation, and sick leave, and explores changes from baseline for eligible employees with cancer conditions. Different types of cancer were compared to understand whether trends were similar across cancer types.

Methods

To better understand the impact of cancer conditions on an employed population and on work absenteeism, claims from the Workpartners Research Reference Database (RRDb) were analyzed from January 1,2001-December 31, 2020. The RRDb is a national database that contains de-identified employee direct medical and prescription claims on 3.6 million US employees and 1.9 million dependents from multiple insurers in the manufacturing, retail, service, medical, transportation, technology, energy, financial, and utility industries. The RRDb also has data on employee job-related information (e.g., salary, exempt-/non-exempt status, full-/parttime status, etc.), vision and dental claims, plus payments and lost time for four absence benefits: sick leave, short-term disability, long-term disability, and workers’ compensation. Employees in the RRDb with eligibility for the absence benefits are shown in Table 1.The RRDb has been used to analyze absence payments and lost productivity and trends for various conditions [19-29].

For the present study, patients were chronologically identified in the Workpartners RRDb,based on claims with ICD9/-10 codes for the cancer categories defined by the US Agency for Healthcare Research and Quality [30] (AHRQ) (Table 2). The study claims data were analyzed over the fixed 20 calendar year periods ending 2020 for annual medical and prescription costs. Infused and injected agents paid for under the medical benefit are reported with the medical costs.

Figure 4 : Percent of Employees Filing Claims by Benefit.

Annual days of leave relative to baseline by benefit for the All cancer cohort are presented in Figure 5. During the 20-year period (from 2001 through 2020) as a percent of baseline, relative mean days of sick leave were 65.2%-199.0%, short-term disability were 81.1%-116.1%, long-term disability leaves were 96.8%330.5%, and workers’ compensation leaves were 65.5%-293.1%. Sick leave days increased, with similar slopes for all individual cancer cohorts in the study. No significant annual changes were identified for the other three absence benefits.

Figure 5 : Average Days of Leaves Per Year by Benefit (as a percent of baseline).

The ranges of average absence payments as a percent of baseline average absence payments are shown in Figure 6 for each benefit. From 2001–2020 as a percent of baseline, the average payments were 91.9%-366.5% for sick leave, 96.0%-175.5% for short-term disability, 190.5%-810.3% for long-term disability, and were 60.4%-831.0% for workers’ compensation. Analysis of the average payment trend lines slopes for the individual cancers revealed that all sick leave and all short-term disability average payments except Female and GI cancers were significantly increasing annually. Additionally, long-term disability payments for Other cancers were also increasing, whereas the average costs for each of the remaining benefit-cancer combinations did not significantly increase.

Figure 6 : Average Payment by Benefit (as a percent of baseline).

At baseline, sick leave median payments were equal to salary, and for the other benefits, the median payments as a percent of salary were highest for short-term disability, followed by workers’ compensation and long-term disability. Compared with baseline, the range of relative median payments as a percent of salary is shown in Figure 7. From 2001 to 2020, the median payments were 83.9%-130.7% of baseline for short-term disability, 136.5%331.1% for long-term disability, and 75.9%-207.3% for workers’ compensation. The slopes of the median payment lines for median long-term disability and workers’ compensation annual payments were all similar to zero (no increase) overall and for the individual cancers. Within short-term disability a decrease in annual median payments was noted for the Breast and Skin cancer categories.

Figure 7 : Median Payment as a Percent of Salary by Benefit (as a percent of 2001 baseline). Because sick leave payments are equal to salary, they are omitted from this chart.

Discussion

Although many studies report using real-world data to examine the impact of cancer on work absence, disability, and cost, few studies in the literature use real-world, person-specific absence cost and lost time data from comprehensive employee benefits and payroll systems [19-29].

Few studies cover multiple benefit types. The majority of studies focus only on disability data, [8-14] whereas some also include sick leave information [10,12,14]. Research by Goyal, [8] Wan,[10] Qian,[13] and Chang [15] focused on short-term disability and excluded long-term disability, whereas Merola [12] and Yu [14] examined long term disability. None of the studies identified included workers’ compensation data. In these papers, the researchers multiplied lost time by a constant payment for absences across benefits,[12,13,15] which appears to be inaccurate, based on the present results. Other research focused on the out-ofpocket costs for cancers, but did not address absence costs [32].

A lot has changed in cancer management since Chang et al. [15] researched seven types of cancer, and while they used matched controls and report direct medical costs specific to each the cancers, similar to our research, their absence data were aggregated due to small sample sizes. Unlike the present research, Chang et al. [15] used US bureau of labor statistics (BLS) data to estimate absence costs for sick leave and short-term disability data, estimating monthly costs of $373 for sick leave ( vs. controls $101) and $698 for short-term disability (vs . controls $25).

Chang et al. [15] reported that cancer patients had 5.2 mean monthly (62.4 annual) short-term disability days and 2.8 mean sick leave absence days per month (33.3 per year), translating to a monthly mean costs of $373 for sick leave and $698 for short-term disability ($8769 and $16,409 per year inflation adjusted to 2020 dollars, respectively). Our study found annual sick leave costs averaged $2483 (range: $1176-$4688) and short-term disability costs averaging $7579 (range: $5417-$9895), with annual sick leave days averaging 9.3 (range: 5.1-15.5) and short-term disability days averaging 43.0 (range: 34.7-49.7). The differences in days and costs between studies may be in part due to differences in overall benefit administration.

The estimates by Mariotto et al. [1] were largely based on cancer survivors aged 65 years and older, and excluded subjects who did not survive. Furthermore, these estimates project the Medicare data to those with cancer treated in the initial and end of life phases that are younger than 65 years old.1 Furthermore, this estimate [1,2] only presents cancer-attributable costs for medical services and oral prescription drugs [5] and ignores productivity losses, indirect health cost, and workers’ compensation claims.

Absences from work can have significant implications on business performance. Absence costs in the literature are often estimated using a constant rather than the actual amount paid to the employee for time missed due to their absence, and studies often assume that the percent of salary received is the same, regardless of benefit [10,12,15]. These estimates often combine disability (short- and long-term) and in some cases do not differentiate between absence benefits used [12].

Reflecting the unique information contained in the database, the Workpartners RRDb has been leveraged to publish data from all four absence benefits using real-world data based on claims and payroll data [19-29] In the present study, annual cohort inclusion, prevalence, medical costs, and CCI were based on medical claims, drug costs were derived from prescription claims, with the remaining outcomes based on absence benefits.

Many of the identified studies focused on patients with cancer. The present research focused on employees with cancer, but the results have been compared with research on employees whose spouses have cancer [33]. In comparing the prevalence of these two recent cancer studies in this database, the spouses with cancer and employees with cancer were highly correlated (Pearson correlation coefficients [r] ≥ 0.9) for All cancer, Skin cancer and Other cancer. Female cancers had a small correlation (r = 0.07), and Male, Breast, and GI cancers had r of 0.67, 0.74, and 0.78, respectively.

The present study has several strengths. This study used real-world, objective data from employer disability, workers’ compensation claims and payroll systems, and was conducted in a diverse, commercial workplace–centric database, which includes patients dispersed throughout the US. The database also includes job-related information (salary, exempt status, and part-/full-time status) and self-reported racial information not contained in other databases. The employers in the database represent a wide range of industries.

This study has several limitations. These administrative claims data are derived from employees with commercial health insurance over the study period and may not be generalizable to patients who do not have employer-sponsored health insurance or who are unemployed. Although the study did examine subgroups of cancer and compared them with cancer overall, we did not examine all of the individual cancer conditions-and similar to the variance amongst the cancer categories in the study, there were likely differences within the individual cancer categories.

The study also did not assess the impact of treatment or ascertain disease control of the patients. In addition, comparisons with baseline allow for a level of control, but the study did not include specific control groups. The Workpartners RRDb provided a convenience-based sample, and the population expanded or contracted by employers joining or leaving the database. Even though the study was conducted over a 20-year period, each year/ condition was an evolving and thus different cohort with potentially evolving absence benefit plan characteristics. Cost data associated with paid leaves are generally not normally distributed and may benefit from a regression-based approach. Finally, the RRDb cannot control for employer-based interventions to decrease sick leave and disability. This study includes 20 years of observation; therefore, it is likely that at least some of the trends were impacted by such interventions.

This study did not address presenteeism, which refers to the lost productivity that occurs when employees are not fully functioning in the workplace because of an illness, injury, or other condition. Perhaps 25% of cancer survivors feel less productive at work, with 25% having difficulty with physical tasks and 14% having difficulty with mental tasks required by their job [34].

Future research should consider comparing health issues and treatments for employees versus spouses with cancer, examining specific cancer conditions, using control cohorts (employees without the stated condition), adjusting inclusion/exclusion criteria to require multiple medical or prescription claims, and using two-part regression models controlling for employee job-related information (e.g., salary, full-/part-time status, etc.) to estimate absences and costs, which might allow the impact to be projected to the US employed population.

Conclusion

In this real-world study, the overall severity of illness in the patient population with cancer conditions, as measured by the Charlson Comorbidity Index, increased during the study period. Whereas medical and drug costs for employees with cancer increased from 2001 to 2020, utilization of absence benefits and payments for those absences as a percent of salary have not increased. Additionally, the percent of employees with cancer using disability, workers’ compensation, or sick leave in a given year varied greatly. Additionally, the lengths of leave as well as payments for leave as a percent of salary over time were highly variable and associated with benefit type. Using a constant cost or salary replacement factor over time, or for all benefits, is not accurate or appropriate in health benefit absence research. Every effort should be made to use actual person-level or claim-level absence and payment data from employer disability, workers’ compensation, and payroll data systems.

Acknowledgments

The authors thank Eric M. Rosenberg from Workpartners, LLC, Pittsburgh, PA for his assistance with the development of this manuscript.