Absenteeism and indirect costs during the year following the diagnosis of an operable breast cancer

Introduction The consequences of disease on work for individual patients as well as the consequences of absenteeism from work are subjects of interest for decision-makers. Methods We analyzed duration of absenteeism and related indirect costs for patients with a paid job during the year following diagnosis of early-stage breast cancer in the prospective OPTISOINS01 cohort. A human capital and friction costs approaches were considered for evaluation of lost working days. For this analysis, the friction period was estimated from recent French data. Statistical analysis included simple and multiple linear regression to identify determinants of absenteeism and indirect costs. Results 93% of patients had at least one period of sick leave, with an average of 2 periods of sick leave and a mean total duration of 186 days. 24% of patients returned to work part-time after an average sick leave of 114 days (i.e. 41 LWD). Estimated indirect costs were €22,722.00 and €7,724.00 per patient, for the human capital and friction cost approaches, respectively. In the multiple linear regression model, factors associated with absenteeism were: invasive tumor (p=.043), mastectomy (p=.038), redo surgery (p=.002), chemotherapy (p=.027), being a manager (p=.025) or a craftsman (p=.005). Conclusion Breast cancer is associated with long periods of absenteeism during the year following diagnosis, but almost all patients were able to return to work. Major differences in the results were observed between the friction cost and human capital approaches, highlighting the importance of considering both approaches in such studies.


Introduction
Breast cancer (BC) is the most common cancer in women in France with an estimated incidence of 54,062 new cases in 2015 (standardized incidence rate of 97.4 cases per 100,000 women per year). Costs associated with health care interventions are usually divided between direct costs, referring to resource utilization, and indirect costs, representing loss of productivity or the time cost due to illness. Various Health Technology Assessment (HTA) agencies have developed guidelines identifying the cost components to be included in economic evaluation, and generally recommending exclusion of indirect costs for various reasons. Alternatively, the French Health Authority (HAS) has proposed considering the production costs of health care interventions, including the standard direct costs and the time cost of delivery of care to patients, but excluding the time cost due to the underlying condition [1].
BC generates major direct costs for health care systems. A study conducted in France on the national health insurance database concluded that BC was the most expensive type of cancer, representing 18.5% of all cancer expenditure [2]. In the US, the estimated lifetime treatment costs ranged from $20,000 to $100,000 [3]. Moreover, since BC is mostly prevalent in women of working age, it induces considerable indirect costs, which should be taken into account to ensure a more comprehensive approach to this disease from a societal perspective.
The methodology of indirect cost measurement raises a number of issues, especially estimation of lost productivity related to the time lost due to illness. The human capital (HC) approach, defined by Rice, evaluates the entire time lost, regardless of the duration of time off work [4]. In contrast, the friction costs (FC) approach assumes that indirect costs only occur during a limited period before the employer company adjusts to the worker's absence, defined as a friction period [5]. Although both of these approaches can provide useful information, the two approaches provide estimates that can differ dramatically, especially when considering a lifetime horizon [6][7][8]. 4 Data concerning absenteeism and indirect costs are useful for decision-makers when considering cost-of-illness studies or model-based economic evaluations on innovative treatments or screening strategies. Few published studies have focused on the indirect costs of BC and absenteeism, with estimates ranging from $8,068 [9] to $21,086 per patient [10], highlighting the importance of the methodology used and the country in which the study is conducted. In France, only one study has reported the indirect costs associated with adjuvant chemotherapy in BC [11]. Only limited data are also available concerning the individual characteristics driving the level of indirect costs. It may be justified to maintain employment or promote early return to work in order to decrease the economic burden of the disease from a societal perspective, but also to limit the impact of the disease on the patient's personal life (social environment, long-term career goals, etc.). In some countries, such as France, return to work is considered to be a public health priority [12-13] and could be promoted by means of multidisciplinary interventions and less toxic innovative treatments [14].
The objective of this study was to describe the indirect costs of absenteeism and their determinants in the first year following the diagnosis of early BC in a French population-based prospective cohort study, using both the HC and FC approaches.

Population
We analyzed indirect costs in early BC during the year following diagnosis. This study was part of a global research on BC pathways and burden of disease of the French multicenter OPTISOINS01 study. The design of this prospective trial has been previously described [15].
Female patients with histologically confirmed, previously untreated and primarily operable BC (exclusion of metastatic, locally advanced or inflammatory BC as defined by the AJCC) were 5 included in this cohort by 8 centers (three University hospitals, four local hospitals, one comprehensive cancer center) in three departments of the Ile-de-France region between 2014 and 2016. Patients were prospectively followed for one year with data collection concerning three work packages: 1/resource utilization and costs of pathways, 2/patient satisfaction and work reintegration, 3/quality, coordination and access to innovation. Individual social and economic characteristics were also recorded at the beginning of the study. Informed consent was obtained from all individual participants included in the study.
All patients of the initial OPTISOINS01 cohort reporting a paid job at the time of the diagnosis were included in the present study. Patients with missing data on wage or absenteeism were excluded.
All components of absenteeism were included as part of indirect costs: days of sick leave, part-time return to work, early retirement, mortality. Indirect costs related to presenteeism and unpaid work were excluded from the scope of this study. A specific questionnaire in the second work package was used to assess time lost due to the disease during the year of the survey: dates of work and absence from work during treatment, work arrangements, on-shift status (e.g., recognition of disability at work, applications for disability allowance, retirement, and layoff). Indirect costs for relatives were not taken into account.

Methods used to calculate indirect costs
Periods of absenteeism were considered by both the HC and FC approaches, applied separately. In the HC approach, we assumed that indirect costs were generated during the entire period of absenteeism [4]. According to the FC approach [6], after a friction period, the level of productivity was restored due to replacement of the sick worker, incurring no indirect costs.
However, medium-term global macroeconomic consequences may arise in an international 6 competitive labor market with an impact on macroeconomic indicators. During the friction period and as a result of various parameters (diminishing returns to labor, internal labor reserve within firms, and delaying work after the period of absence), production losses are lower than estimates based on the HC approach. With the FC approach, indirect costs were restricted to the friction period and a friction coefficient was used to reduce the value of lost production.
Medium-term macroeconomic consequences were assumed to be insignificant for this cohort.
According to Koopmanschap's method, job vacancy duration estimates (i.e. length of recruitment processes) increased by a 30-day time lag (for the decision to recruit and the time between recruitment and the first working day) were used as a proxy for the duration of the friction period [6]. As job vacancy duration depends on job categories, job vacancy duration estimates were stratified for this factor (managers vs. other job categories). Data concerning job (using the estimated elasticity for annual labor time versus labor productivity), which has never been subsequently updated, was used for the friction coefficient [6].
Periods of absenteeism, including days off (weekends), were recorded in the patient's questionnaire and converted into Lost Working Days (LWD) for the analysis: five LWD for 7 sick days, 2.5 LWD for 7 days of part-time return to work, 251 LWD for death or early retirement (considering 5 working days per week during a year). LWD, representing lost productivity, were estimated in euros based on individual daily wages plus employers' and employees' social welfare contributions (computed by an online tool [18]). Wages, expressed 7 as monthly net income, were self-reported by the patients. When multiple periods of absenteeism, interspersed by periods of work, were recorded in the FC approach, an entire new friction period was assumed for each period of absenteeism.

Statistical analysis
Various simple and multiple linear regression models were performed using Stata\IC 14.0 software (StataCorp, College Station, Texas). In the first model, the independent variable was the number of LWD estimated by the HC approach. In the second model, the independent variable was the amount of indirect costs. As HC and FC approaches result in different values, two distinct statistical analyses were performed. The following dependent variables were integrated in a simple linear regression: marital status, job category, cancer histology, surgical treatment, adjuvant therapy for BC. Factors correlated with the independent variable (with a p-value<.1) were integrated in the multiple linear regression. A p-value < .05 was considered significant.
A univariate sensitivity analysis was conducted on the parameters estimated by the FC approach (duration of the friction period, friction coefficient) by applying ±20% variation and a Tornado diagram was constructed to illustrate the effect on indirect costs. This analysis was also conducted for subgroups: managers and other job categories.
To facilitate the use of our data in other health economics analyses, average indirect costs according to subgroups defined by care pathway and job category were also estimated.

Results
Six hundred four of the 617 screened patients were included in the OPTISOINS01 cohort. Some patients were excluded from the study either due to the absence of a paid job during the study period (n=307) or due to missing data (wages for 42 patients and length of absenteeism for 87 patients). A total of 168 patients were included in the analysis (Figure 1).
Patient characteristics are presented in Table 1.  According to both approaches, in the simple linear regression model, individual indirect costs were correlated with invasive cancer, treatment characteristics (mastectomy, axillary lymph node dissection, chemotherapy) and socioeconomic characteristics (highly educated vs. poorly educated, being manager vs. being a salaried employee) (see Table 2. For details and differences between HC and FC approach). According to the HC approach, the following factors were significantly associated with indirect costs in the multiple linear regression model: mastectomy (p=.046), redo surgery (p=.003) and being a craftsman (p=.023). Indirect costs in the model were increased by being a manager, having an invasive tumor or having received chemotherapy, but with a lower degree of statistical significance. According to the FC approach, being a 12 manager was the only significant factor associated with indirect costs in the multiple linear regression model (p<.001) ( Table 2). [95%CI] p-value Coef. [95%CI] p-value Coef. [95%CI] p-value Coef. [95%CI] p-value showed that BC is often associated with periods of sick leave and part-time return to work during the year following diagnosis, resulting in considerable indirect costs even in the absence of mortality. However, the use of two different approaches resulted in discordant estimated indirect costs per patient: €22,722.00 for the HC approach vs. €7,724.00 for the FC approach.
The following factors were correlated with the number of LWD and the sum of indirect costs 15 in the HC approach: invasive tumor, mastectomy, redo surgery, chemotherapy, being a manager or a craftsman. Indirect costs estimated by the FC approach were mostly driven by job category, as managers had higher wages and longer friction periods.
The mean age at the diagnosis of BC corresponds to the mean age of retirement in industrialized countries (62 years in France) [20]. Moreover, retirement age is currently increasing in many countries, as a result of increasing life expectancy [21]. Absenteeism, indirect costs and, more generally, the work consequences associated with BC constitute major concerns for decision-makers. In the initial OPTISOINS01 cohort, one-half of patients had a paid job during the study period. These patients spent an average of almost six months on sick leave during the year after diagnosis and 24% of patients returned to work part-time for a mean period of 114 days. A mean of 125 LWD per patient was recorded and 26 patients were absent from work for the entire year of analysis, highlighting the impact of BC on paid employment.
Only a few studies have previously reported these data. Two US retrospective cohort studies found that working patients with BC had significantly longer periods of absenteeism compared to "control" patients (without BC) [9,22]. However, they observed a shorter period of absenteeism in the year following diagnosis when using an HC approach: about 10 days of sick leave and between 25 and 45 days of short-term disability (according to the severity of the disease). In Sweden, patients reported on average 271 working hours lost during the 3 months preceding the interview during the first year after a diagnosis of primary BC [10]. In line with our results, mortality and early retirement were rare during the first year following diagnosis.
The present study is the first prospective cohort study focusing on the cost of BC-related absenteeism. The large initial sample size of OPTISOINS01 allowed us to exclude working patients with missing data, which could have affected the quality of the analysis. Due to the geographical design of the study, most patients in this cohort lived near Paris and may have presented different socio-economic characteristics to those of the national population. For example, the proportion of managers in this sample was larger than in the French working population (41.6% vs. 17.1%), and the average wage was also higher (€2,214 vs. €1,926) [23].
The timeframe was limited to one year after diagnosis, but most indirect costs related to BC occur during the first year, except when the disease progresses to metastatic stage or recurrence [10,24].
This study only considered indirect costs due to absenteeism, but there is a growing interest in presenteeism, i.e. workers who have returned to work, but with reduced capacity and productivity due to illness, and unpaid work, which generate indirect costs [25]. with no justification of their sources [30]. However, Koopmanschap's estimation is based on outdated data (1988)(1989)(1990) restricted to the Dutch labor market. The time to fill job vacancies (i.e. the length of recruitment processes) increased by a time lag of four weeks (decision to recruit and time between recruitment and the first working day) was used to estimate the friction period. The duration of the friction period is therefore strongly correlated with the unemployment rate and the economic environment and varies over time and across countries.
The duration of the friction period must also be stratified according to job category, as the recruitment process is longer for managers than for employees [31]. authors have also discussed the risk of double counting in cost-effectiveness evaluation when productivity losses are included. Do individuals take the potential impact on their income and career into account when assessing the impact of their disease on quality of life, included in the denominator of the incremental cost-effectiveness ratio (ICER) [39,40]. Separate analysis of direct and indirect costs can help to guide decision-makers in relation to these scientific and equity issues in healthcare resource utilization.
The relationship between disease and work also needs to be studied in more detail in order to limit the consequences of illness on work. The low rate of early retirement and the absence of dismissal in our cohort should reassure BC patients concerning their future capacity to work and should encourage them to return to work part-time when necessary. Another study showed good readjustment to the workplace for patients with BC [41]. In agreement with many authors, we therefore believe it is crucial to report absenteeism, productivity losses and indirect costs. Moreover, accurate data on the type and duration of absenteeism are essential, particularly for economic studies based on simulated models and using a societal perspective.