645 Abstracts 1 Erasmus Medical Centre, Rotterdam, The Netherlands; 2 Maastricht University, Maastricht, The Netherlands OBJECTIVES: To develop and test a modular questionnaire for the measurement and valuation of all relevant aspects of pro- ductivity costs to be applied in economic evaluations of health care programmes and occupational interventions. METHODS: Modifying existing instruments and developing new ones, we con- structed a comprehensive modular measurement instrument cov- ering absence from work, compensation mechanisms, efficiency losses and productivity costs on the level of organisations. RESULTS: PRODISQ (PRODuctivity and DISease Question- naire) has integrated, modified and further validated instruments such as the Health and Labour Questionnaire (HLQ), the QQ- instrument for efficiency loss and the compensation questions as developed by Severens. In addition new instruments were devel- oped and tested. PRODISQ has the following modules: 1) General information (demography etc); 2) Profession, working situation and income; 3) Absence from work during the last 3 months; 4) Compensation mechanisms in case of absence for paid work; 5) Productivity costs at work (efficiency loss); 6) Productivity costs at the organisational level; and 7) Administrative and manage- ment costs. This summer PRODISQ will available for free on the internet. As compared to other existing instruments, one can say that PRODISQ is a comprehensive instrument, covering all rele- vant aspects of productivity costs: Absence from work, compen- sation mechanisms, efficiency loss, the link with the type of work and other employers costs. PRODISQ has been successfully tested in several Dutch populations of patients and workers: Patients with low back pain, dyspepsia, chronic fatigue syndrome (CFS), psoriasis and rheumatoid arthritis; construction workers and workers in several industrial companies. We will present key results concerning validity and relevance of productivity costs. CONCLUSIONS: All elements of productivity costs can be mea- sured in a valid way. Many elements turn out to have a relevant impact on productivity costs for different populations of patients and/or workers. MC7 USING STATED RISK-BENEFIT TRADEOFF PREFERENCES TO IDENTIFY POTENTIAL THERAPEUTIC RISK-MANAGEMENT PROBLEMS J ohnson FR 1 , Andrews EB 2 1 Research Triangle Institute, Research Triangle Park, NC, USA; 2 Research Triangle Institute OBJECTIVES: Off-label use and adherence problems can occur when there are systematic differences between physicians’ and patients’ perceptions and regulators’ explicit or implicit judg- ments regarding relative risks and benefits. The objective of this study is to quantify subjective risk-benefit tradeoffs to inform risk-management decision making. METHODS: Stated-Prefer- ence (SP) methods, such as conjoint analysis elicit patient and physician assessments of hypothetical risk-benefit tradeoffs provide a necessary common utility metric for comparing risks to benefits. We administered a web-enabled SP survey to a con- venience sample of health-care professionals registered for a major epidemiology conference (N = 444). The instrument required subjects to complete 6 tradeoff tasks involving paired hypothetical pain medications relative to a constant, standard- treatment option. Background risks were set to approximate mean male population myocardial infarction (MI) risks. Efficacy, MI risks, and costs were varied according to a D-optimal statis- tical design. RESULTS: Part-worth utility weights were estimated using conditional logit analysis. All parameters were statistically significant at the 0.01 level or better. “Somewhat-controlled” and “well-controlled” efficacy were significantly different (p = 0.005), but there was no significant difference between “well- controlled” and “fully-controlled” efficacy levels. Maximum acceptable risk (MAR) is the increase in risk that exactly offsets the increase in utility from improved pain control. Mean MAR for an increase in efficacy from somewhat-controlled to well-con- trolled is 0.0385, or about 15 times background risk. The cor- responding mean estimate for maximum acceptable cost or willingness to pay (WTP) is $1465 per month. CONCLUSIONS: A convenience sample of health care professionals provided the- oretically consistent, well-structured risk-benefit stated prefer- ences for pain-control therapy. Both MAR and WTP estimates indicate strong preferences for pain control. However, maximum acceptable risk is much larger than typical regulatory standards. Such differences between patient and regulatory risk-benefit tradeoffs signal potential problems for adherence, off-label use, and other impediments to effective risk management. MC8 COMPARING THE INTERVAL PROPERTIES OF HEALTH-STATE VALUATIONS MEASURED USING THE VISUAL ANALOGUE SCALE AND RANK-BASED SCALING Krabbe PFM 1 , Salomon JA 2 , Murray CJL 2 1 University Medical Centre Nijmegen, Nijmegen, The Netherlands; 2 Harvard University, Boston, MA, USA OBJECTIVES: The visual analogue scale (VAS) is a relatively simple technique for eliciting health-state valuations. However, several biases may affect VAS values (e.g. context effects, end- aversion, and anchoring effects). Moreover, the supposed inter- val measurement level of VAS mean values has been questioned. An alternative methodology is introduced based on the ranking of health states combined with associated scaling methods that are used to transform aggregate ranking data to interval values. METHODS: Data were collected in a Dutch EuroQol EQ-5D valuation study, in which a representative sample (n = 212) from the Dutch population valued a set of 17 EQ-5D health states. Three computational steps were undertaken: 1) Euclidean dis- tance measures between each pair of health states were computed based on individual VAS values; 2) These measures were trans- formed to ranks; and 3) Non-metric multidimensional scaling (MDS) was applied for scaling (monotone transformations) the ranks on one dimension. RESULTS: The analysis supports un- dimensionality of the raw VAS data, with 94% of the variance explained with one dimension. Scaling of ranked distances between health states using MDS produced metric values that were nearly identical to mean VAS valuations, with the rank- based values explaining 98% of the variance in the VAS. CON- CLUSION: Ordinal ranks implied by VAS values, transformed using scaling techniques, provide strong concordance with mean VAS values. Our methodology based on simple ordinal ranking exercises may offer a useful approach for quantifying health- state valuations in a more feasible and bias-free manner. However, the selection of the set of health states may affect our ranking based methodology, and the findings must be confirmed based on directly-elicited ranking data. Therefore, further research is needed to investigate the performance of this new methodology. QUALITY OF LIFE II QL5 VALIDATION OF THE GERMAN PROSTATE SPECIFIC MODULE (PSM) Bestmann B 1 , Rohde V 2 ,Weidner W 2 , Kuechler T 1 1 Reference Center Quality of Life in Oncology, Kiel, Schleswig-Holste, Germany; 2 University of Giessen, Giessen, Hessen, Germany OBJECTIVES: Every year in Germany more than 25.000 men are diagnosed with prostate cancer. Theoretically all these