Productivity Losses Due to Migraine in Slovenia: An Analysis of Absenteeism and Presenteeism Costs Based on Administrative and Self-Reported Data

We adopt the human capital method that takes the earnings of an individual as a proxy for their productivity (17). To calculate productivity costs, we use the daily labour costs per worker separately for both genders in 2016, given that the studied administrative data on absenteeism refer to this year. In Slovenia, the average daily labour costs amounted to EUR 90.17 for women and EUR 95.81 for men in 2016 (18).

Average daily labour costs are calculated by dividing respective annual labour costs by the 252 working days in Slovenia in 2016. Annual labour costs per worker are calculated based on the labour cost structural statistics of the Statistical Office of Slovenia. They comprise wages and salaries, including social contributions payable by employees, social contributions paid by employers and other expenditures paid by the employers. In Slovenia, persons with compulsory insurance are entitled to a sickness cash benefit that for patients with migraine typically amounts to 80% of their earnings. The sickness cash benefit is paid either by the employer or the Health Insurance Institute of Slovenia, depending on the length of absence. By estimating absenteeism costs using daily labour costs per worker, we capture both the 20% share of the burden carried by patients and the 80% share of absenteeism costs incurred by either the employer or the social insurance fund. We thus calculate absenteeism costs from a social perspective regardless of who incurs them.

One important pitfall of the human capital approach is that differences in absenteeism costs between different diseases may be a consequence of wage differences rather than the actual length of absences and the levels of impairment (9). In such circumstances, estimates of indirect costs for illnesses affecting mostly manual uneducated labour or women may underestimate their burden of illness. This is relevant also for migraine since prevalence, incidence and remission rates vary by sex, socioeconomic status and other factors (19). This is why we use both average overall daily labour cost and average gender-specific daily labour costs in Slovenia. In 2016, the average daily labour cost per worker for both genders combined equalled EUR 93.21 (18). Educational-level and profession-specific labour costs are not used, as available datasets do not incorporate this information about the studied migraineurs.

Absenteeism is estimated using the national sick-leave database, coupled with data from the self-reported study that includes the Migraine Disability Assessment Test (MIDAS) and Work Productivity and Activity Impairment (WPAI) items, measuring absenteeism and presenteeism. We use the administrative database of the NIPH on sick leave due to migraine for 2016. Surveys reflect patients’ subjective perceptions about the type of their headaches, whereas in our administrative database the diagnoses are recorded by medical professionals at the time of sick leave occurrences according to the International Classification of Diseases (Version ICD-10-AM). This database includes the entire population of employed individuals and others entitled to sickness benefits that are on formal sick leave due to migraine rather than a specific sample. Part-time employees and self-employed individuals as well as employees with flexible work schedules, who take formal sick leave less often, represent a small share. The database for 2016 consists of 2,416 records of absence from work of full-time employees due to migraine. These records refer to 1,743 patients. Besides the duration of sick leave, information on gender, age group and month of the sick leave termination are also available. Absenteeism costs based on administrative data are estimated using Equation 1. (Eq. 1)Annual absenteeism cost per absentee=∑i=1N=1743(length of sick leavei× average (overall) daily labour costs) ÷N\eqalign{& Annual\,absenteeism\,cost\,per\,absentee \cr & \matrix{{}& {= \sum\limits_{i = 1}^{N = 1743} {(length\,of\,sick\,leave_i }}\cr {}& {\times \,average\,(overall)\,daily\,labour\,costs)\, \div N}}}

The length of sick leave is the yearly number of days off work for an individual patient with diagnosed migraine (code G43 of ICD-10-AM). Average overall labour costs are specified in Section 2.1. Eq. 1 is modified by replacing the average overall daily labour costs with the average gender-specific daily labour costs to determine the gender-specific absenteeism costs.

Absenteeism is also studied based on the data from a wider self-reported study (20). A web survey using a recall period of 3 months was conducted between 15 November and 12 December 2017 to analyse the prevalence of headache and migraine, their impact on the working population and the treatment processes. A total of 1,207 adult employees with headaches, in addition to 102 general practitioners and 50 neurologists, were surveyed.

There were 146 migraineurs out of 1,207 respondents. After excluding outliers (i.e., patients outside the range of three standard deviations from the mean number of days at work affected by migraine) 142 migraineurs were included in the analysis. The survey comprises 20 questions referring to the diagnosis, the current disease treatment and management, the use of headache diaries, migraine’s adverse effects and the productivity losses. One item capturing absenteeism is from the MIDAS instrument, which is one of the most widely used questionnaires to measure the impact of headaches on work functioning and is also included in the Slovenian national guidelines for treatment of migraine (21).

This item measures the number of days lost form work or school in the last 3 months. Absenteeism costs estimated using this approach are calculated using Equation 2. (Eq. 2)Absenteeism cost per absentee in a 3-month period= ∑i=1n=142(no. of days out of work in the last 3 monthsi× average (overall) daily labour costs)÷n\eqalign{& Absenteeism\,cost\,per\,absentee\,in\,a\,3 - month\,period \cr & \matrix{{}& {= \,\,\sum\limits_{i = 1}^{n = 142} {(no.\,of\,days\,out\,of\,work\,in\,the\,last\,3\,months_i } }\cr {}& {\times \,\,\,average\,(overall)\,daily\,labour\,costs) \div n}}}

Eq. 2 is also modified by replacing the average overall daily labour costs with the average gender-specific daily labour costs from Section 2.1.

An important challenge of using self-reported data with a 3-month recall is the calculation of absenteeism costs at an annual level. The national-level data on sick leave indicate that the frequency and intensity of sick leave is not constant throughout the year. To extrapolate results from the 3-month period covered by the survey to the annual level, we introduce conversion factors estimated from the administrative data on sick leave. Conversion factors are estimated by comparing average absences from work in a corresponding 3-month period to the average absences on an annual level.

Presenteeism is studied using the MIDAS and WPAI instruments, each including one presenteeism item (22).

In the WPAI presenteeism item, 0 indicates no effect on work and 10 complete prevention from working. The estimated presenteeism costs based on this WPAI item are calculated using Equation 3. (Eq. 3)Presenteeism cost per patient in a 3-month period= ∑i=1n=142(no. of work days with reduced productivity in the last 3 monthsi× impairment while working with migrainei / 10× average (overall) daily labour costs) ÷ n\eqalign {&{Presenteeism\,cost\,per\,patient\,in\,a\,3 - month\,period}\cr &{= \,\sum\limits_{i = 1}^{n = 142} {(no.\,of\,work\,days\,with\,reduced\,productivity\,in\,the\,last\,3\,months_i}}\cr & {\times \,\,\,impairment\,while\,working\,with\,migraine_i \,/\,10}\cr &{\times \,\,\,average\,(overall)\,daily\,labour\,costs)\,\, \div \,n}}

The number of working days with reduced productivity in the last 3 months is determined as the difference between the number of migraine days and the number of days of absence from work in this same period.

MIDAS identifies the number of days at work with substantially reduced productivity (i.e. more than 50%) over a 3-month period. The estimated presenteeism costs based on this MIDAS item are calculated using Equation 4. (Eq. 4)Presenteeism cost per patient in a 3-month period= ∑i=1n=142(no. of work days with at least 50% productivity lossi× 0.5 × average (overall) daily labour costs) ÷ n\eqalign{& Presenteeism\,cost\,per\,patient\,in\,a\,3 - month\,period \cr & = \,\sum\limits_{i = 1}^{n = 142} {(no.\,of\,work\,days\,with\,at\,least\,50\% \,productivity\,loss_i} \cr & \times \,0.5\, \times \,average\,(overall)\,daily\,labour\,costs)\, \div \,n}

To enable a comparison between estimated presenteeism and absenteeism costs, the former need to be extrapolated from a 3-month period to an annual level. We do this using conversion factors from Section 2.2. We thus assume that migraine intensity within the observed 3-month period compared to the whole year had a comparable effect on both absenteeism and presenteeism.

Figure 1 shows the distribution of the 1,743 migraineurs on sick leave from the national database in 2016 by age and gender (85.4% women, 14.6% men). Over 80% of migraine-related absences lasted for up to five days. The average annual length of absence from work due to migraine was 5.70 days, 5.65 for women and 5.96 for men.

Figure 1

The average annual absenteeism cost per absentee estimated using Eq. 1 amounted to EUR 531. The total annual absenteeism cost for all 1,743 migraineurs on formal sick leave in 2016 is thus estimated at EUR 925,762. Table 1 shows the results for both genders combined and stratified by gender, estimated at labour costs from Section 2.1.

We also estimate absenteeism costs using the self-reported data and Eq. 2. The self-reported average length of absence from work in a 3-month period was 4.40 days (4.18 for women and 4.89 for men).

These results are not directly comparable to the results based on the national database due to differences in the covered period. To enable a comparison, we use the administrative data to determine the average absence from work only for the period from September to November 2016. This time span is observed because the autumn period is also included in the survey and this helps us control possible seasonal triggers of migraine. In the above-mentioned 3-month period, there were 594 patients recorded in the national sick leave database with approximately the same gender proportion as in the whole year of 2016. Their 3-month average absence length was 3.77 days, 3.78 and 3.75 for women and men, respectively. By comparing the average length of absences from work in the three months in autumn to the annual average absences reported in Table 1, we calculate the 1-year-to-3-month ratios equalling 1.510 for both genders combined, 1.497 for women and 1.588 for men for 2016.

The estimated self-reported absenteeism costs calculated on the annual level amount to EUR 626 per absentee for both genders combined, thus exceeding the cost estimated using administrative data by 18% (11% for women and 30% for men).

According to Eq. 3, we need both the number of working days with reduced productivity and the efficiency while working with a migraine to estimate the presenteeism cost per absentee in a 3-month period (Figure 2). Over 50% of migraineurs stated that their level of impairment while working with a migraine was between 5 and 7 (out of 10).

Figure 2

According to the survey, most respondents, i.e. 75.4% (72.8% for women and 82.1% for men), report that the average number of migraine days per month is less than 4. One problem in the calculation of the number of working days with reduced productivity arises from the survey’s design. Specifically, the question regarding the length of absence provided different pre-determined intervals. To overcome this problem, we calculate the presenteeism costs by considering several different scenarios shown in Table 2, i.e. 1) taking the lower bounds of the intervals, 2) taking the middle points of the intervals and 3) taking the upper bounds of the intervals as the relevant value for the number of migraine days per month. The upper limit of the last interval is set at the maximum number of working days within a month.

The estimated annual presenteeism costs in three different scenarios are presented in Table 3. The survey enables estimates for the observed 3-month period. To enable comparisons of presenteeism costs with absenteeism costs, we converted the 3-month values to the annual level using the conversion factors calculated in Section 3.1.

Given that both the conversion factors and the daily labour cost per worker are for 2016, the estimated presenteeism costs in Table 3 also refer to 2016 to enable comparability with estimated absenteeism costs.

The results presented above refer to the WPAI presenteeism item. By using Eq. 4, we estimate presenteeism costs using the MIDAS presenteeism item as well (Table 4). A specific number of days with substantially reduced productivity was reported directly by the surveyed patients. Again, the 3-month values were extrapolated to the annual level by using the conversion factors from Section 3.1.