Precision and accuracy in measuring absence from work as a basis for calculating productivity costs in The Netherlands
Introduction
When performing economic analyses alongside clinical trials different viewpoints can be used for the analyses, such as a third party payer perspective, a health care perspective, or a societal perspective. The impact of disease on the ability of a person to perform work should be part of an economic evaluation when a societal viewpoint is used for the analysis (Drummond et al., 1997). Debate is going on about how to include the consequences of being able to work due to illness in the ratio between costs and effects as a result of a cost-effectiveness analysis (Brouwer et al., 1997a, Weinstein et al., 1997, Brouwer et al., 1997b). However, it seems to be clear that in cost-effectiveness analyses which do not use utility measures for effectiveness, the consequences of not being able to work should be reflected in monetary terms in the numerator of the cost-effectiveness ratio, the so called productivity costs (Luce et al., 1996).
When analysing productivity costs, a distinction can be made between lost productivity related to absence from paid work, reduced productivity at paid work, and lost home productivity (Roijen et al., 1996). Although the latter two can be of importance, in our study we concentrate on productivity costs related to absence from paid work. When analysing productivity costs due to absence from work, days absent from work are to be measured, after which the number of days is valued (Koopmanschap and Rutten, 1996).
To determine absence from work due to illness several instruments can be used. Sick leave registers would be a reliable source of information to obtain the number of days sick leave for participants in a study. However, when analysing the productivity cost of study participants related to a specific disease who, of course, are employed in different localities or companies this approach is not practical. To overcome this problem questionnaires are used to measure absence due to illness to be able to calculate productivity costs. Such questionnaires can be applied in a more or less prospective manner, but more often data are gathered in a retrospective way. Different recall periods to measure absence from work can be found in the literature and to our information the longest period used was 12 months (Jones et al., 1995, Agius et al., 1994, Bertera, 1991). The question arises whether using such questionnaires leads to valid results because a potential for recall bias exists in every study in which historical self-reported information from respondents is used. The imperfect memory of respondents can harm the precision (difference by chance between memory and fact) and accuracy (systematic difference between memory and fact) of the sick leave data, which, theoretically, can cause recall bias. This can influence the absolute level of costs induced by absence from work. However, recall bias is only relevant when accuracy of recall regarding the measure of interest is different between the groups of patients that are distinguished in a study (Raphael, 1987).
The purpose of our study was to study precision and accuracy of a retrospective self-administered questionnaire on sick leave by comparing the self-reported absence with company-registered absence data. Different recall periods were used to analyse the relationship between memory and length of the recall period.
Section snippets
Methods
A local branch of a pharmaceutical company involved in research, marketing, and sales participated in the study. All employees, existing largely of office workers, who were working for this company for at least 1 year, were asked to fill-in a questionnaire. This questionnaire aimed to measure retrospectively the number of days absent from work due to illness concerning five different recall periods: the past 2 weeks, 4 weeks, 2 months, 6 months and 12 months. The actual date the questionnaire
Results
At the time of this project 210 people worked at the company; all were asked to participate in the study. Of these, 155 returned the questionnaire (response rate 74%). One questionnaire was incomplete, and the reported results could not be matched to the data from the company registry, resulting in a study sample of 154. It was possible to make company-registered absence due to illness available for all respondents except one. Reported sick leave for the different recall periods was limited
Discussion
For reasons of reliability, prospective registration of absence from work is to be preferred above retrospective analysis. However, this indicates a substantial workload for participants, which can be one of the causes of missing data (Goossens et al., 1998). Using only a short period of prospective registration, and multiplying the results afterwards to estimate absence from work for a longer term leads to valid results only when large numbers of respondents participate (Roijen et al., 1996).
Acknowledgements
The authors are grateful to Mrs. I. Van Camp MSc, and Mr. H. Nelis for data collection, Mr. W. Lemmens, and Mrs. L. Lemmens for data management, and Mr. Th.M. de Boo for statistical advice. This study was supported by a grant from Astra Pharmaceutica B.V., The Netherlands.
References (15)
- et al.
Questionnaire for the identification of back pain for epidemiological purposes
Occupational and Environmental Medicine
(1994) The effects of behavioral risks on absenteeism and health-care costs in the workplace
Journal of Occupational Medicine
(1991)- et al.
Productivity costs measurement through quality of life? A response to the recommendation of the Washington panel
Health Economics
(1997) - et al.
Productivity costs in cost-effectiveness analysis: numerator or denominator: a further discussion
Health Economics
(1997) - et al.
Reliability of a questionnaire on sickness absence with specific attention to absence due to back pain and respiratory complaints
Occupational and Environmental Medicine
(1996) - Centraal Bureau voor de Statistiek (CBS) 1997. Statistisch Jaarboek 1997. CBS,...
- et al.
Methods for the economic evaluation of health care programmes
(1997)
Cited by (162)
Social and economic costs of home parenteral nutrition
2020, Clinical Nutrition ESPENCitation Excerpt :This questionnaire uses three modules to measure productivity costs: lost productivity at paid work due to absenteeism, lost productivity at paid work due to presenteeism and lost productivity at unpaid work. The iPCQ uses a recall period of four weeks [28]. Considering the variability of this information in longer periods, the data obtained with this questionnaire could not be extrapolated to one year.
Association between job strain and sick leave: a systematic review and meta-analysis of prospective cohort studies
2020, Public HealthCitation Excerpt :Recorded data on sick leave are dependent on the physician's comments and what doctors define as intervals for sick leave.86 Because self-reported data are less dependent on occupational conditions and other related factors,86 some studies have shown that self-reported data have a good validity compared with recorded data.84,87 As the findings of the present study showed, in both self-reported and recorded data, there is a relationship between job strain and sick leave.
Symptom control in chronic rhinosinusitis is an independent predictor of productivity loss
2018, Annales Francaises d'Oto-Rhino-Laryngologie et de Pathologie Cervico-FacialeSymptom control in chronic rhinosinusitis is an independent predictor of productivity loss
2018, European Annals of Otorhinolaryngology, Head and Neck DiseasesCitation Excerpt :To evaluate for productivity losses, participants were asked to report how many days of work or school were missed in the last three months due to CRS, as previously described for CRS [15–17]. This method of assessing disease-specific productivity loss has been previously validated as an accurate means of determining missed days of work or school due a chronic disease [18–20]. All participants in our study were also asked to categorize their CRS symptom control as “Not at All”, “A Little”, “Somewhat”, “Very”, and “Completely”, as previously described for assessing patient-perceived sinonasal symptom control [13,21].