Abstracts NHESG Tromsø 2018

Transcription

1 Abstracts NHESG Tromsø 2018 All accepted abstracts for the Nordic Health Economist Study Group 2018 are included below. Each contribution will require one discussant for the conference. All participant are expected to be prepared to act as a discussant. We practice a first come, first serve principle. To make sure you get the opportunity to discuss a preferred paper, you do the following: 1. Go to the doodle at 2. Find the abstract you would like to discuss (numbers corresponding to numbers in this document) 3. Select the abstract you would like to discuss. 4. Close your browser Contributions without discussant will be assigned a discussant from the pool of conference participants who have not volunteered for the task by July 2 nd.

2 Abstract 2 Socioeconomic Inequality in Community Based Health Insurance Premium Contribution in Rwanda Gowokani Chijere Chirwa 1 and Marc Suhrcke, Prof 2 Objectives: Community based health insurance (CBHI) appears to have mushroomed in many low and middle income countries (LMICs). This is perhaps in response to policymakers calling for lowcost pro-poor health financing, within the various health systems, as a means to achieve universal health coverage (UHC), and therefore achieve sustainable development goals (SDGs). Existing CBHI evaluations have, however, tended to ignore the distributional aspects of the household contributions made to CBHI. One country that has made substantial positive progress in CBHI implementation is Rwanda, having experimented with two types of CBHI within a decade. In this paper, the objective is to investigate the pattern of socioeconomic inequality in CBHI premium contributions (payments) in Rwanda. In addition to this we also assess gender difference in CBHI contribution (payments). Data and Methods: The analysis methods uses three econometric approaches; decomposition of the concentration index of inequality, Blinder-Oaxaca mean-based decomposition, and unconditional quantile regression decomposition.the study uses two sets of cross-sectional data for the periods 2010/11 and 2013/14. Results: Among other findings, the key takeaway message from the results is that the categorisation of CBHI premiums into different payment groups, may have led to the CBHI being financed by the richer individuals and reduces regressivity. In both the flat rate system, and the wealth based categorised system, inequality exists but it s much more pronounced in the flat rate system. By designing a new system based on wealth categorisation, inequality in CBHI payments improved to the advantage of the poor. In terms of gender differences in CBHI payments, female headed households are likely to spend less on CBHI than male headed households. The Blinder-Oaxaca decomposition analysis shows that the difference in CBHI payments is due to group differences in the distribution of individual characteristics between the female-headed households and the male headed households. Conclusion: From the main results, it means that inequality in contribution in a CBHI system based on wealth categories is better than the flat rate system, thereby indicating that CBHI is indeed pro-poor. On gender differences, the implications is that there has to be a gender approach in design of CBHI policy. The lessons for other developing countries is that it is better to design a CBHI system that make people contribute differently basing on their wealth status. Additionally, to ensure health for all, and achieve SDGs, gender issues need to be addressed if CBHI has to be pro-poor. 1 PhD Student University of York, UK: Centre for Health economics (CHE) : gcc509@york.ac.uk

3 Abstract 4 To develop a prediction model for presenteeism using the EQ5D-5L, SF6D, and ICECAP-A Cheryl Jones 1, Suzanne Verstappen 2, Brenda Gannon 3, Alex Thompson 1, Katherine Payne 1 1 Manchester Centre for Health Economics, The University of Manchester, Manchester, United Kingdom. 2 Arthritis Research UK Centre for Epidemiology, Centre for Musculoskeletal Research, The University of Manchester, United Kingdom, 3 The University of Queensland, Centre for Business and Economics of Health, Brisbane, Australia Background and Objectives: Presenteeism is defined as the reduction in performance at work due to ill-health. Workplace interventions (WPIs) are a set of programmes, activities and equipment designed to help individuals manage their condition and improve functional ability. Employers are frequent funders of WPIs but often have limited access to information on the costs and productive benefits WPIs may produce. By comparison, national decision-makers typically have access to a large body of data from economic evaluations comparing the costs and consequences of various healthcare interventions. Ideally this existing evidence could be used to inform funding decisions made by employers who are interested primarily in improving presenteeism, as well as health. However, first the link between presenteeism and health-related quality of life and capability, as traditionally measured for healthcare intervention studies, must be established. Data and Methods: Regression-based methods, including ordinary least squares (OLS), Tobit, and Censored Least Absolute Deviation (CLAD), were used to analyse data from a sample of working people with examples of musculoskeletal conditions (rheumatoid arthritis and ankylosing spondylitis) identified from an on-line panel company and a patient support group website. A bespoke survey was designed that asked respondents to record their levels of: 1) health status, using the EuroQol Five Dimensions Five Level (EQ5D-5L and the Short-Form Six Dimension (SF6D); 2) capability, using the ICEpop Capability for Adults (ICECAP-A); and 3) presenteeism measured using the Work Productivity Activity Impairment Questionnaire (WPAI). Models were specified defining the WPAI as the dependent variable and EQ5D-5L, SF6D or ICECAP-A, respectively, as the independent variable with other selected covariates. Spearman s rank correlation was used to assess the degree of conceptual overlap between the EQ5D-5L, SF6D, ICECAP-A, and the WPAI. Observed and predicted levels of presenteeism were estimated and validated using k-fold (k=10) cross-validation. Model performance was assessed using mean error, root mean squared error, and mean absolute error. Results: The dataset comprised 542 employed individuals with RA (n=497) and AS (n=85). A strong correlation was found between the EQ5D-5L and the WPAI (ρ = ) and a moderate correlation was found between the ICECAP-A and the WPAI (ρ = ). The model that included the SF6D, age and gender using an OLS regression most accurately predicted levels of presenteeism measured by the WPAI. Conclusions: This study shows there is a quantifiable link between presenteeism and health or capability. The results from this study suggest it is feasible to prospectively predict the impact WPIs have on presenteeism by using a measure of health status or capability. The SF6D was identified to most accurately predict levels of presenteeism, a finding that is unique given previous literature where only the EQ5D-3L has been used to predict levels of presenteeism.

4 Abstract 5 How far does the peer effect on smoking behavior reach? Jørgen T. Lauridsen, Department of Business and Economics, University of Southern Denmark Peer effects in forming smoking habits are well known from literature. If living in a low smoking environment, an individual will be less tempted to smoke, and reversely if living in a high smoking environment. Based on municipality level data for 98 Danish municipalities observed in three waves in 2010, 2013 and 2017, the present study investigates whether such peer effects can be found between municipalities in the form of an endogenous spatial spillover. By using a spatial quantile regression approach, it is investigated whether peer effects are present in extreme environments (i.e. environments with low or high smoking propensities) as well as environments with a median smoking propensity. A peer effect in the form of a significantly positive spatial spillover is found for median environments. For extreme environments with low or high smoking propensities, a positive, but not significant, peer effect is found. The results indicate that the peer effect predominantly exists in environments characterized by a median smoking behavior, and that it is less relevant for environments with high or low smoking propensities.

5 Abstract 6 Information shocks and provider adaptation: Evidence from interventional cardiology Daniel Avdic, Stephanie von Hinke, Bo Lagerqvist, Carol Propper and Johan Vikström, University of Duisburg-Essen Regional health care expenditures vary substantially also after adjustment for patient characteristics, prices and consumer demand for health services. However, areas with higher spending do not, in general, have better health outcomes. This suggests that health care resources can be more efficiently allocated to reduce total expenditures without worsening patient outcomes. One largely overlooked factor that may cause such inefficiencies is discretionary provider treatment decisions. We study how cardiologists in Sweden changed their preferred choice of stents when performing percutaneous coronary interventions (PCI) after it was revealed that the newly introduced drug-eluting stents (DES), which had begun to replace the older bare-metal stents (BMS), had potentially severe side-effects. We exploit the DES information shock by splitting our sampling time frame into three mutually exclusive information regimes; an introduction phase where the popularity of the DES continuously grew; a period of adverse information in which the popularity dropped sharply; and a period characterized by the introduction of national guidelines which provided clear information about the appropriate use of DES. To study cardiologist responses, we subsequently construct a period-specific measure of adaptability to information defined as the relative rate with which each cardiologist adopted (or abandoned) DES. We then apply the resulting distribution of adaptability to explore whether patient outcomes differ by the type of cardiologist they were treated by in each information regime. To estimate our econometric models we use data from the Swedish Coronary Angiography and Angioplasty Register (SCAAR). SCAAR is a Swedish national database that registers all interventional coronary procedures in Swedish hospitals since The data contain detailed information on all medical procedures each patient received and characteristics of the specific performing physician. The SCAAR data is also linked to a set of patient clinical outcomes from the Swedish inpatient registry for up to ten years after the procedure was performed. Our results show substantial variation in the rate with which cardiologists responded to information in each information phase. Specifically, the variation in response is significantly less pronounced during the first period relative to the second period. Furthermore, the introduction of national guidelines in the final phase restricted physician choice and reduced this variability substantially. Finally, our findings suggest that the speed of response is associated with a number of patient outcomes. More specifically, we find that patients treated by slow adapters have a lower risk of complications compared to patients treated by fast adapters. These effects largely disappear once the guidelines were introduced, suggesting that variation in physician discretion and patient health outcomes are intimately linked. Our findings suggest that imposing guidelines could be superior to discretionary treatments in contexts where patient heterogeneity is relatively limited. However, at the same time, it is possible that these guidelines may not have been proposed without the evidence based on the eagerness of enthusiastic adapters' treatment decisions. Indeed, without early adapters, any potential side-effects of new treatments and how to avoid such may have never been discovered, suggesting a crucial trade-off in the policy discussion on restricting medical practice.

6 Abstract 7 Individual or enterprise liability? The roles of sanctions and liability under contractible and non-contractible safety efforts. Sverre Grepperud. Department of Health Management and Health Economics, University of Oslo, PO 1089, N-0317 Oslo, Norway. Phone: sverre.grepperud@medisin.uio.no In the literature on medical malpractice and patient safety there is a debate about the relative attractiveness of individual - and enterprise (firm or corporate) liability and some critics of medical malpractice advocate shifting liability from physicians to the organizations that deliver health care. The social attractiveness of individual and enterprise liability is analyzed in a model where accident risks are influenced by decisions made by both the enterprise and the employees of the enterprise (individuals). The regulator observes a share of accidents while the safety decision of the individual can be contractible or non-contractible for the enterprise. Given contractible individual care, sanction regimes yield the first best, whereas liability regimes produce sub-optimal solutions. Given noncontractible individual care, the combined use of an individual sanction and an enterprise sanction (joint use) produces the first best, the exclusive use of an individual sanction produces the first best if the enterprise does not suffer any direct harm, while the exclusive use of an enterprise sanction does not produce the first best. If both decision-makers are solvent and have similar liability probabilities, individual - and enterprise liability do equally well under contractible individual care while individual liability does best for non-contractible individual care.

7 Abstract 8 The importance of market conditions when paying physicians a fee for service Sibilla Di Guida1, Dorte Gyrd-Hansen2, Anne Sophie Oxholm2* 1University of Southern Denmark, Department of Business and Economics 2 University of Southern Denmark, Department of Public Health, DaCHE Danish Centre for Health Economics, *Presenter Background: In many health care systems physicians are paid a fee per service they provide. This payment form is popular because it gives physicians a direct financial incentive to exert effort into health care. However, several studies find that this payment scheme may lead to supplier-induced demand (SID), defined as over-serving of patients beyond their need of care. These studies find evidence of SID in markets where there is high-competition for patients, as resource-abundant physicians have an incentive to increase profit by inducing demand. However, in many health care systems competition is low, implying that physicians are resource constrained, i.e. the delivery of services to one patient group incurs opportunity costs to other patients. Although physicians are unable to induce overall demand for care, SID may still occur, as physicians can increase profits by allocating care from low-profit patients to high-profit patients. The literature has yet to investigate the prevalence of SID in markets with resource-constrained physicians. Objective: This experiment adds to the literature by testing physicians response to fee-forservice under both market conditions (resource abundance and resource constraint), thereby verifying whether the presence of patient opportunity costs protects against SID. We also test whether physicians' altruism is a further protective factor against SID, by observing whether physicians are less prone to SID when patients suffer a health loss from being over-served. Methods and data: We use a controlled and fully-incentivised laboratory experiment with 39 medical students. The participants are asked to decide on the number of services they wish to provide to patients who are in different need of care, who either experience no-health loss or a health loss from being over-served, and whose treatment generates either a high or low profit margin. The participants face two different market conditions: 1) resource abundance, where they can fulfil a patient s need of care and 2) resource constraint, where the cost of fulfilling a patient s need of care is reduced benefits to other patients. Results: On average, physicians supply a larger number of health care services to a patient when receiving a high-fee compared to a low-fee. This result holds both when physicians are resource abundant and resource constrained. However, on average physicians do not profit maximise, signifying that altruism also plays a significant role in physicians' treatment decisions. In cases where physicians are resource abundant, we find that the increased number of services leads to an overserving of patients; particularly, when over-serving does not lead to a decline in patients health and patients are in low need of care. When resources are constrained we observe less over-serving of patients; particularly, we find a reduction in the provision of services, which are detrimental to patients health. Conclusion: The literature has shown that patients are over-served under fee-for-service payment schemes. This study shows that the extent to which over-serving takes place is highly context specific. In the presence of patient opportunity costs and/or when over-serving harms patients, we observe highly reduced prevalence of SID.

8 Abstract 9 ATTITUDES TOWARDS GENETIC TESTING AND INFORMATION: DOES PARENTHOOD SHAPE THE VIEWS? Antti Saastamoinen1,*, Mika Kortelainen1, Juho Aaltio4, Mari Auranen2,4, Emil Ylikallio2,4, Tuula Lönnqvist3, Markus Sainio4, Anu Suomalainen4,5, Henna Tyynismaa4,6, Pirjo Isohanni3,4 1) VATT Institute for Economic Research, Helsinki Finland 2) Clinical Neurosciences, Neurology, University of Helsinki and Helsinki University Hospital, Finland. 3) Department of Child Neurology, Children s Hospital, University of Helsinki and Helsinki University Hospital, Helsinki, Finland. 4) Research Program Unit, Molecular Neurology, University of Helsinki, Helsinki, Finland. 5) Neuroscience Center, University of Helsinki, Helsinki, Finland. 6) Department of Medical and Clinical Genetics, University of Helsinki, Helsinki, Finland. * Corresponding author, antti.saastamoinen@vatt.fi, Objectives: This study examines how parents of pediatric patients might differ in their views and attitudes towards genetic technology and information when compared to adult patient views. Data and methods: The study has a twofold approach. First we conduct a health economics orientated literature review in order to interpret some of the theoretical models of health decision making presented in the literature from the perspective of parental decision making. On empirical side, we cover studies which have examined for example the effects of different risk and time preferences on decision making and on the other hand how life events such as parenthood affects to the formation of such preferences. Secondly, we conduct a survey of views on genetic technology and information for two groups, namely a group of parents of pediatric patients (n=31) and a group of adult patients (n=68). All the patients were suspected to have a neurological disease of unknown etiology and were thus found eligible to undergo clinical whole-exome sequencing within this project. In addition to our own survey instrument, we conducted several other standard form surveys, such as RAND36, Beck s Depression Inventory, and Parenting Stress Index to gain insight on respondent s psychosocial factors that might affect to their attitudes and views. Results: Literature review revealed that many models and concepts of decision making offer a valuable viewpoint to understand how people handle health related information. We argue that to better understand the paradigm change that personalized genetic medicine ought to bring, we should further examine how issues such as time and risk preferences, value and accuracy of information, anticipatory emotions, information avoidance, and societal norms, affect the adoption and acceptance of these new technologies. This is especially relevant when we analyze the decisions of parents as intergenerational disease risks are the key areas of utilization for genetic technology. The empirical studies indeed often find that parenthood induces more risk aversive behavior. The analysis of our own survey results also gives support for this view. The results suggest that parents are more concerned about their children s genetic risk factors when compared to the attitudes of adult patients about their own risk. Moreover, while both groups favored obtaining extensive results, a larger share of parents chose to obtain results faster but in this case as less extensive. For both groups negative emotional state and even depression was associated with more concerns towards genetic information. For parents, this association seemed to arise from their feelings of failure or incompetence as a parent. Conclusions: Our study provides insights on how to think about parental decision making in the context of genetic medicine. The special nature of parental judgement should be acknowledged because parents decide not only for themselves, but on the behalf of their children also. From the perspective of health policy aiming to foster acceptance of genetic technology, this parental perspective should be recognized.

9 Abstract 10 Does Accreditation Affect the Job Satisfaction of General Practitioners? A Cluster Randomised Field Experiment in Denmark Line Bjørnskov Pedersen a,b, Thomas Allen c, Frans Boch Waldorff b Merethe Kirstine Kousgaard Andersen b a DaCHE Danish Centre for Health Economics, Institute of Public Health, University of Southern Denmark, J.B. Winsløwsvej 9B, 5000 Odense C, Denmark b Research Unit for General Practice, Institute of Public Health, University of Southern Denmark, J.B. Winsløwsvej 9A, 5000 Odense C, Denmark c Manchester Centre for Health Economics, University of Manchester, Oxford Road, Manchester M13 9PL, UK Background: A critical question for policy makers in health care is whether external interventions such as accreditation or pay-for-performance systems can have the unintended consequences of lowering professionals motivation and job satisfaction. This is particularly pertinent in general practice, where many countries have a shortage of general practitioners (GPs), and where many GPs suffer from burnout. It has recently been shown that monetary incentives, in the form of pay-forperformance, do not affect the job satisfaction of GPs. The question is whether this is also the case for non-monetary incentives. Objectives: We investigate whether non-monetary incentives, in the form of accreditation, affect the job satisfaction of GPs. As accreditation is perceived as a tool for external control among most Danish GPs, it is likely, that the introduction of mandatory accreditation in general practice will negatively affect their job satisfaction. Methods and data: Accreditation of general practice in Denmark was introduced as a cluster randomised stepwise implementation from Job satisfaction was measured at three time points: before the randomisation took place, one year into the accreditation process and two years into the accreditation process. We use a balanced sample of GPs who have completed all three waves of the survey (n=846). We analyse the data in four different models: 1) A random effects ordered logit model taking account of the panel structure in the data, 2) a mixed effects ordered logit model further taking account of GPs being clustered in different general practices, 3) a mixed effects ordered logit model as in 2) but also taking account of general practices being clustered in municipalities, and 4) a mixed effects ordered logit model taking account of clustering of GPs within general practices and including a priori attitudes towards accreditation, measured in different ways, as explanatory variables. Results: Our preliminary analyses show no significant effects of accreditation on job satisfaction. There are, however, significantly positive associations between job satisfaction and having positive a priori attitudes towards accreditation, or perceiving accreditation as a tool for quality improvement, while we find significantly negative associations between job satisfaction and perceiving accreditation as a tool for external control. Conclusion: We conclude that accreditation does not affect GP job satisfaction. Hence, even when non-monetary incentives are perceived as control instruments by most GPs, the intervention is unlikely to affect the job satisfaction. However, GPs perceptions of the intervention are associated with their job satisfaction. Therefore, policy makers are encouraged to carefully inform health care professionals about new interventions before implementation to diminish unnecessary negative perceptions about the intervention, e.g. due to concerns of the unknown.

10 Abstract 11 Health-related Quality of Life and Compensating Income Variation for 18 Health Conditions in Iceland Tinna Laufey Ásgeirsdóttir, University of Iceland Kristín Helga Birgisdóttir, University of Iceland Hanna Björg Henrysdóttir, University of Iceland Þórhildur Ólafsdóttir, University of Iceland Using data from an Icelandic health and well-being survey, carried out in 2007, 2009 and 2012, we estimate the monetary compensation needed to maintain the same level of well-being with and without particular health conditions. Specifically, 18 health problems are evaluated using a compensating income variation (CIV) approach. This approach employs individual well-being measures with no hypothetical situations involved, thus offering a solution to biases of frequently-used methods to value non-marketed goods. Results from our CIV analyses indicate that 1,301,758 USD are needed per year to compensate for the presence of melancholy, 159,301 USD are needed to compensate for frequent headaches, 118,465 USD are needed per year to compensate for severely low vision and for severe monthly menstrual cramps 62,419 USD per year are needed. This research adds to the literature by employing a rarely used, but conceptually promising method to a range of health conditions. Furthermore, since several different conditions were valued with the same sample and same methodology this research provides a ranking between the conditions, aiding policy makers in prioritizing scarce resources.

11 Abstract 12 Markets and rationality: Competition results in a more predictable behavior Ge Ge 2 and Geir Godager 3 Objective: The aim of this paper is to contribute with new knowledge on how competition affects individual s behavior. Data and methods: In our model of physician behavior we assume that patient health benefit as well as physician profit influence physicians choice of medical treatment. We study three different market settings, monopoly duopoly and quadropoly. When competition is present, the treatment decisions determine market shares. We assume a symmetric game of complete information, where providers simultaneously choose a treatment alternative from a set of discrete treatment alternatives. We assume a logistic quantal response equilibrium (LQRE), which can be understood as a statistical version of the Nash equilibrium. In LQRE, players choose strategy based on relative expected utility and they assume other players do the same. Under the assumption of LQRE, best response functions are probabilistic and can be expressed in the form of logit model with scaled variance. We use data from an incentivized laboratory experiment in which subjects play the role of physician and make treatment choices in monopoly, duopoly, and quadropoly. We estimate the parameters of the model by means of maximum likelihood. We also simulate gameplay in the fixed point given by the model estimates, in order to assess the properties of the estimates. Results and conclusions: Our results suggest that the rationality parameter rises as competition intensifies, implying that behavior becomes more systematic, and further, that subjects probability of choosing the best response increases. The results show that ignoring the possibility that individual s rationality can depend on context can result in the misleading conclusion that preferences are affected by the market setting. The results provide useful nuances to the recent literature on whether markets erode social responsibility. 2 Department of Health Management and Health Economics, University of Oslo, Norway. 3 Department of Health Management and Health Economics, University of Oslo, Norway, and Akershus University Hospital.

12 Abstract 13 Low risk, high reward? Empirical evidence from physiotherapy procurements with multiple winners Visa Pitkänen Social Insurance Institution of Finland Signe Jauhiainen Social Insurance Institution of Finland Ismo Linnosmaa University of Eastern Finland Objectives: We study physiotherapy firms pricing in a service where firms participate every four years in a competitive bidding organised by the Social Insurance Institution s insurance districts. We exploit the institutional feature, that during the latest three biddings, only couple of districts rejected any firm. Thus, the risk of rejection has been very low, giving firms a possibility to use their market power. We document large price increases over the last three competitive biddings and analyze how risk and competition are related to prices. We also analyze counterfactual scenarios where a capacitybased rule of acceptance is implemented. Methods: We analyze panel data on all submitted bids (N=3,740) in the 2006, 2010 and 2014 competitive biddings. We measure the risk of rejection using variation in insurance districts rejection rates and firms closeness to the rejection cut-offs in previous competitive biddings. We compute several market power measures based on the value that each bidder would bring into the districts network, using discrete choice modelling on matched patient-provider register data. Results: We find that the price bid for a 45-minute physiotherapy service increased from an average of 39 euros in 2006 to 57 euros in Our choice models show that consumers prefer large, highquality providers within close distance. We find that higher risk of rejection is related to lower prices, whereas greater market power is related to higher prices. Our simulations show that implementing a capacitybased rule would have resulted into fiscal savings of 4.8 to 9.2 million euros in the contract period. Discussion: Accepting nearly all bidders has increased the overall price level in the service, as the risk of rejection has been very low and firms have been able to use their market power. Our key policy recommendation is to apply a capacity-based rule for acceptance, which creates incentives to bid a price that reflects bidders costs of providing the service and fosters price competition.

13 Abstract 14 Cost-effectiveness of web based peer support for young adults not in employment or education Leena Forma 1, Jussi Partanen 1, Jan Klavus 1, Pekka Rissanen 1,2 1 Faculty of Social Sciences (health sciences), University of Tampere 2 National Institute for Health and Welfare Quality of life (QoL) among young adults not in employment or education has been found to be much lower than among their age peers in employment or education. A web based intervention offering peer support and aiming to improve wellbeing of young adults not in employment or education was carried out. We study whether the intervention was cost-effective compared to usual services. 147 customers of targeted youth services at the age of were recruited into the study. They were randomized into the intervention and control groups. The intervention was an anonymous online community, moderated by youth or social workers. Control group got the services as usual. Use of health, social and employment services as well as QoL were asked in supervised interviews at baseline and after six months follow-up. National unit costs are used in calculating the costs of service use. QoL was measures by the WHOQOL-BREF instrument, and is used as effectiveness measure. Incremental cost-effectiveness ratios are calculated. Now, follow up data have been collected from 20 young adults in the intervention group (IG) and 23 in the control group (CG). Both groups were heterogeneous in terms of their life situation, service use and QoL. The average age was 25 years in IG and 23 in CG. 50% were female in IG and 48% in CG. At the baseline, the psychological dimension was statistically significantly lower in IG (42) than in CG (57). Costs of services at baseline (in previous 12 months) were higher in CG than in IG, but in the follow-up (in previous 6 months) vice versa. Improve in QoL in the IG was not found. The rest of the data will be collected during the spring 2018, and the results are ready to be written into the full paper in the beginning of July. We are considering using also other effectiveness indicators, which may be more suitable for this type on intervention, like UCLA loneliness scale, which was also included in the data collection. We will discuss the challenges of economic evaluation of soft interventions compared to e.g. medical treatments.

14 Abstract 15 Incentivising day case surgery: Spillover effects on non-targeted care Philip Britteon 1 (PhD student), Søren Kristensen, Yiu-Shing Lau, Ruth McDonald, Matt Sutton 1: Division of Population Health, Health Services Research & Primary care, The University of Manchester, England Background: Tariff payment incentives have proved to be an important mechanism in encouraging hospitals to switch from conventional to day case surgery. Yet, little is known about the wider spillover effects that incentivising day cases may have on non-targeted areas of care. Day case units may prioritise incentivised procedures over others when operating at full capacity. Alternatively, incentives may have long-term positive spillovers if day case units invest in capacity over time. We exploit variation in hospital responses to the rollout of day case incentives in England to evaluate spillovers on non-targeted procedures recommended for day case surgery. Objectives: To test whether hospital responses to day case incentives are associated with the organisation and outcomes of non-targeted procedures. Data: 1,362,154 patients admitted to 131 hospitals in England for 16 non-targeted procedures between April 2007 and March Methods: Multivariable regression relating the short- and long-term changes in day case rates and waiting times of non-targeted procedures to the day case rate of incentivised procedures treated within the same hospital or by the same consultant team. Results: A one percentage point increase in the day case rate of incentivised procedures was associated with a percentage point (CI to 0.040) short-term reduction and percentage point (CI to 0.193) long-term increase in the day case rate of non-targeted procedures treated within the same hospital. The effect of the incentives on elective waiting times for non-targeted procedures was insignificant. Conclusions: Incentives to increase the proportion of procedures performed as a day case had positive long-term spillovers on the treatment of other non-targeted procedures. Spillovers were negative but insignificant in the short-term, when capacity of day case units was fixed.

15 Abstract 16 Insurance coverage and teenage suicidality Christoph Kronenberg a,b,c a CINCH; University of Duisburg Essen, Germany b Leibniz Science Campus Ruhr, Essen, Germany c RWI Leibniz-Institute for Economic Research, Essen, Germany Economists have studied suicide since Hamermesh and Soss (1974), though economists have nearly exclusively focused on suicide and suicidality of adults. Using individual level data covering a timespan of nearly thirty years from 37 US states I analyze changes in teenage suicidality. The descriptive results show that teenage suicidality decreased during the 1990s until around 2005, but is increasing again since Looking at the ratio of those attempting suicide over those considering or planning suicide (conversion ratio hereafter) provides some interesting insights. A minority of teenagers attempt suicide without considering or planning it. The decreases in suicidality starting in the 1990s appear to be driven by those least serious about attempting suicide as the conversion ratio is increasing or only slow decreasing. The recent increase in teenage suicidality is driven by a marked increase in the conversion rate, especially for boys. Over the entire period girls have a larger conversation rate than boys, which is in contrast to adults suicides. Given recent findings adult suicides in the US are driven by white middle-aged men (Case & Deaton, 2015), while for teenager suicidality minorities have a considerably higher conversion rate than white teenagers. The reductions in suicidality among teenagers coincide with changes to the US insurance system in the 1990s and early 2000s requiring mental health benefits to be on par with physical health benefits. These so called parity laws have been shown to reduce adult suicide rates (Lang, 2013). Parity laws were introduced in different states at different times so that the effect of parity laws can be estimated net of general time and state trends. Furthermore, parity laws can be split into different categories trying to explore which aspect is most relevant for a potential reduction in suicide. Preliminary results indicate that increased mental health care benefits reduce suicidality, but not suicide attempts Corresponding author: Christoph Kronenberg, CINCH, University Duisburg-Essen, Berliner Platz 6-8, Essen, Germany Tel. +49 (0) christoph.kronenberg@uni-due.de

16 Abstract 17 Impacts of Healthy Diets on Mental Health and Wellbeing: is perception a relevant mediator in this relationship? Attakrit Leckvivilize, Health Economic Research Unit (HERU), University of Aberdeen, Scotland The relationship between healthy diet and physical health is well-documented in the literature, particularly its positive effects on disease prevention, weight and physical functioning. However, there are fewer studies exploring an association between healthy diet and psychological outcomes such as wellbeing, depression and anxiety. Moreover, one s perception on her diet might affect her eating behaviour as well as psychological outcomes. For instance, if people perceive a dish as healthy, it can have positive effects on their physical and emotional health despite its lack of real nutritional value. Three waves of Understanding Society Survey, which are longitudinal data sets for a representative UK population, are analysed with the first-difference model. Our dependent variables are indicators for mental health and wellbeing, i.e. self-rated health status, parts of the 12-item short form survey (SF-12) as well as life and health satisfaction. Healthy diet is measured by portion of fruit and vegetables consumed per day, whereas one s perception on healthy diet is proxied by two questions, i.e. participants were asked to rate from a scale 0 to 10 if they could stick with a healthy diet and their views on the benefits of eating healthy in the long run. Our preliminary results indicate that consuming more fruit and vegetables has a positive impact on wellbeing even after controlling for socio-economic factors and other health behaviours such as physical activity, smoking and alcohol consumption. Nevertheless, we do not find robust impacts of an interlink between perception and actual diet on either mental health or wellbeing.

17 Abstract 18 Is there a socioeconomic gradient in travel time to health care services? 4 Rosanna Johed 5, Kjetil Telle 6 A socioeconomic gradient in travel time to the general practitioner (GP) may give advantaged groups better possibilities to maintain own health. And if the travel time to the GP is a determinant for consulting the GP, this may lead to over- and under- treatment. This again is a major risk towards individual health. It has been shown that people with longer distance to casualty clinics in Norway visit such clinics far less often. Our aim is to analyze socioeconomic gradients in travel time and consultations to the GP. We do an individual level accessibility analysis using exact geographic coordinates. For all individuals in Norway registered to a GP in 2017 we have exact geographic location of residential home and GP o ce (N= ), and the minimum travel time by car and beeline (shortest line between the two coordinates) are calculated. We use population wide individual level registry data for Norway with information on socioeconomic- and health characteristics (visits to GP, income, gender, age, education, employment, sick leave, disability bene ts etc.). Preliminary and incomplete results nd that median (mean) travel time and beeline to the GP is 5 (20) minutes and 2.8 (1.5) km in Travel time is longest for low income groups (see Figure 1) and individuals aged 20-29, and shortest for women and elderly. Especially among men, travel time decreases with level of education. The mean number of consultations to the GP decreases with travel time and beeline. This is particulary clear among the lowest income groups and older population (age group and above), see Figure 2. There is a sorting among beeline and consultations to the GP. Individuals who regularly consult the GP Figure 1: Beedistance to GP over income and gender Figure 2: a & b have shorter distance than non-consulting individuals. The beeline to the GP (due to individuals changing GP, not due to residential reloacation) decreases remarkedly for individuals who start to consult the GP (after at least two years with no consultation). Preliminary and incomplete conclusion: There are signs of a socioeconomic gradient in travel time and consultations to the GP. Low income groups have the longest travel time to their GP and longer travel time is claerly associated with fewer consultations. This may have implications for health, and is thus important for designing policies to improve public health and maintain costs of health care. 4 This research has recieved support from the Norwegian Research Council (grand no ) 5 University of Oslo, Norway & Statistics Norway 6 Statistics Norway

18 Abstract 19 Market conditions and technology adoption in primary care Tor Iversen Department of Health Management and Health Economics, University of Oslo Objectives: Medical technology contributes to improved longevity and quality of life. Studies have shown that payment system, insurance coverage, and market conditions contribute to technology adoption by providers. These studies typically consider the entire health system or only the hospital sector. Technology adoption in primary care has not been studied much. This is a critical gap in the literature, especially when primary care is expected to play a crucial role in prevention and health maintenance of patients with chronic diseases, as, for instance, Type 2 Diabetes (T2D). The present study explores factors that contribute to technology adoption by primary care physicians (PCPs). Methods: We develop a theory of technology adoption by PCPs based on benefit and cost. We derive the market equilibrium and comparative statics, and present hypotheses. The theory is tested with a unique data set on all PCPs in Norway between 2009 and In Norway, where each resident is listed with a PCP, more than 95% of PCPs are self-employed and paid by a combination of capitation fees and fee-for-service. In 2009, a systematic surveillance program for patients with T2D was introduced. While PCPs have gradually adopted the program, still only a quarter of PCPs are making use of it. Based on our theory, we assess factors that contribute to technology adoption by individual PCPs. We estimate linear fixed-effects models and hazard functions by flexible parametric survival models with cubic splines. During the observation period, an education program for PCPs was introduced in some counties. By means of difference-in-differences models, we estimate the effect of the education program on technology adoption. We also estimate to what extent technology adoption contributes to consumers choice of PCP. Results: Results show that the technology adoption among PCPs depends positively on the number of listed patients with T2D, on being a specialist in general medicine, and the number of PCPs in the community who have adopted the technology in previous periods. We find that technology adoption depends negatively on the PCP s age and access to private specialists and hospitals. Spare capacity measured by the number of PCPs in the community who accepts additional patients, has no effect. We find further that the introduced education program has a positive effect. We do not find that technology adoption has an effect on consumers choices of PCPs. Conclusion and discussion: The empirical results give support to the theory that additional education of PCPs, organization, and access to specialized care affect technology adoption in primary care. We discuss why neither an effect of competition among PCPs nor a demand effect among patients seem to exist.

19 Abstract 20 Double Jeopardy Nicolai Simonsen a, Olaf Irgens, Trine Kjær a a DaCHE, University of Southern Denmark, Department of Public Health, J.B. Winsløws Vej 9B, 5000 Odense C, Denmark It is often assumed in the literature that an individual s marginal utility of consumption doesn t vary with the health state (Feldstein, 1973; Feldman and Dowd, 1991; Mitchell et al., 1999; Engen et al., 1999; Davidoff et al., 2005; Golosov and Tsyvinski, 2006; Scholz et al., 2006; Hall and Jones, 2007). It has however, for a long time, been recognized that if the form of the utility function differs between health states, it will impact a number of economic behaviors like insurance and saving decisions (Arrow, 1974). Also, the class of literature which evaluates health states or changes in health states by using compensating income variation are sensitive to this potential change in marginal utility (Asgeirsdottir et al., 2017; Howley, 2017; Groot and van den Brink, 2007; Ferrer-i-Carbonell and van Praag, 2002). A few empirical studies have previously investigated the relationship between marginal utility of consumption and health state using different methodologies and with mixed findings (Finkelstein et al., 2013; Gyrd-Hansen, 2016; Tengstam, 2013; Viscusi and Evans, 1990). The aim of this paper is to contribute to the literature providing new and more comprehensive evidence of state dependency using a unique combination of survey and register data. The model which we utilize is a refinement of the one proposed and used by Finkelstein et al. (2013), where we allow income and health to jointly determine utility. We define positive (negative) state dependence as being present if the marginal utility of consumption increases (decreases) when health is decreased. If the utility function doesn t change when health changes, we say that the utility function exhibits state independence of health. We estimate our model by combing data on life satisfaction from the Survey of Health, Ageing and Retirement in Europe (Bo rsch-supan et al., 2013) with data from the Danish registers including information on healthcare utilisation, education, income etc. This allows us to build a comprehensive dataset, which links self reported changes in the individual s utility (subjective well being) with rich background information on health and income collected from the registers. To the best of our knowledge we are the first to utilize this unique data source. The results show strong evidence in favour of a positive state dependence. Our main estimate suggest that the marginal utility of consumption increases by 42%, when an individual turns ill. The results are robust to different levels of risk aversions, and to assumptions regarding the mapping of the latent utility into observed utility. Our results of positive state dependence of health should be of great interest for policy makers as it entails e.g. higher level of optimal saving and insurance and can affect economic evaluations of health interventions.

20 Abstract 21 Predicting diagnostic coding in hospitals: individual level effects of price incentives Anthun, Kjartan Sarheim SINTEF Technology and Society, Department of Health Research OBJECTIVE AND BACKGROUND: This paper tests if an implicit price incentive influence the diagnostic coding in hospitals. Hospitals can group uncomplicated patients into complicated groups by way of diagnostic coding to increase reimbursement. We estimate if the probability of being coded as a complicated patient was related to a specific price incentive. Age change mechanisms have been suggested earlier as an indicator of upcoding, and this paper also tests empirically if upcoding can be linked to shifts in patient composition through proxy measures such as age composition, length of stay, readmission rates, mortality- and morbidity of patients. DATA: Data about inpatient episodes in Norway in all specialized hospitals in the years were collected, N= Our main dependant variable was the complicated status of each patient (was the patient grouped in a complicated or an uncomplicated diagnostic related group (DRG)). A fixed version DRG grouper was used. Independent variables were patient background information (age, sex, county of residence), information about the hospital episode (length of stay, case-mix, emergency/elective, surgical/medical, and major disease categories), regional hospital trust dummies, time trend and finally the price incentive (measured at the DRG pair level as the price differences between the complicated and uncomplicated groups). METHODS: First, we analyse trends in the proxy measures of diagnostic upcoding: can hospital behavioural changes be seen over time with regards to age composition, readmission rates, length of stay, comorbidity and mortality? These analyses are descriptive statistics, graphical and difference in difference regressions to test if those hospital episodes explicitly exposed to a price incentive are more upcoded. Finally, we examine only those patients grouped in DRG pairs to see if variations in the price incentive are related to probability of being coded as complicated. RESULTS: 40.5% of the discharges were grouped as complicated for the period as a total. In the first years ( ) there was a marked increase in the episode coded as complicated, while the level has become more stable in the years In the trend analyses we found some indications of upcoding as some of the outcome measures saw the predicted shifts hypothesized. In the difference in difference regressions four out of five outcome measures had negative interactions indicating that upcoding was taking place. We ran logit estimations on the probability of being grouped as a complicated patient. When run on an empty model, we found a positive effect from price incentive (OR 1.39, 95CI ). However, in the larger models with controls, the effect was much smaller at 1.04 (OR, 95CI ) for the whole period and almost none at 1.00 (OR, 95CI ) for the years CONCLUSION: We found no evidence of widespread upcoding fuelled by implicit price incentive, as other issues such as patient characteristics were more important. This study adds to previous research by testing individual level predictions. The added value of such analysis is to have better case mix control. We observe smaller price effects here than in earlier aggregate analysis. However, the effect of price is still present.