PhD Conference 2020

September 3-4, 2020

Online event

Rotterdam, The Netherlands



Cost-sharing and mental health care inequalities in the Netherlands - are poor young adults particularly vulnerable?

Background: Continuity of mental health care in the transition to adulthood is critical and currently a challenge for most health care systems. Treatment discontinuation in early adulthood is likely to increase the risk of mental illness relapse or recurrence, with potential long-term consequences on health, wellbeing and realization of young people’s potential. Little is known about the impact of cost sharing on the use of mental services by youth, in particular for vulnerable groups. We evaluated the impact of increasing health insurance deductibles on mental health care use by young adults and whether there are heterogeneous effects across socioeconomic groups and previous mental health treatment.

Methods: We used individual administrative records for young adults living in the Netherlands linking data on demographics, income, health expenditure and medication use, between 2009 and 2014. We identified the impact of increasing deductibles on mental health care use at age 18 using a difference-in-discontinuity approach that exploits the different deductible amounts in subsequent years and the exemption for those below the age of 18 years. Finally, we study heterogeneous treatment effects by household income, previous medication for mental disorders and previous level of mental health care expenditure.

Results: Our results show that increasing the deductible by 100 euros reduces the probability of mental health care use at 18 by 11.5% (CI 95% -14.3, - 8.6), 15.1% (CI 95% -22.04, -13.54) for females and 6.3% (CI 95% -8.79, 1.58) for males. The effect is the largest amongst low income females, with reductions of 17.5% (CI 95% -24.3, -10.7) and 19.9% (CI 95% -26.4, -13.3) for those in the first and second income quartiles. No effect is observed for the highest income group. We do also find a negative effect of increasing the deductible on mental health take up by subpopulations that are more likely to need care, as defined according to their previous consumption of psychotropic medicines and treatment intensity.

Discussion: Our findings suggest that demand-side cost sharing during critical periods of the life cycle reduces mental health care uptake disproportionally among low- and middle-income individuals. The reduction is also large among previous users of intense mental health services, suggesting that individuals are likely to forgo high-value care. These findings highlight the risk of applying cost-sharing tools regardless of individual socioeconomic and health status and should inform current debates about targeted financial health policies.

Francisca Lopes, Erasmus Medical Center

Discussant: Rachel Meacock

Does extending the freedom of choice of health provider reduce the waiting times? A methodological approach

In Spain the choice of doctor is a right guaranteed to all population by the nation-wide legislation. In specialised care, this regulation only allows patients to choose among specialist doctors of their referral hospital within their referral health areas. Nevertheless, the different Spanish regions can enact regional legislation to extend the right of choice of people. So far, most of regions have not developed regional regulation and have applied the national legislation. An exception is the Community of Madrid. In November 2009, a regional law transformed the healthcare organisation system in this region into a Single Health Area. Since then patients in Madrid can choose freely among any of the specialists of any hospital or specialities centre of the region. This reform extended the right of freedom of choice for patients and at the same time it significantly reduced the administrative barriers to exercise it.

This study is aimed at measuring the impact of this reform on the waiting times for specialised care. Previous studies using synthetic control methods suggest that this policy had a positive effect on the perceived quality of specialised services. In order to assess the influence of the method used for the empirical analysis on the results obtained, we reconsider the analysis of the impact of the reform in the Community of Madrid by using three alternative methods of policy evaluation: Difference-in-difference (DD) method; Synthetic control (SC) method; and the Manski and Pepper (MP) estimator. Whereas the DD method selects a control group based on the scientific evidence, the SC method employs data-driven procedures to calculate the most suitable comparison group for the unit treated. The MP estimator relaxes the invariance assumptions of the other methods and sets a bounded impact by using a range of admissible effects of the treatment. We use a sample of cross-sectional microdata from the Spanish Healthcare Barometer for the period 2004-2016.

Preliminary results suggest that, regardless of the method applied, the freedom of choice policy implemented in the Community of Madrid achieved to reduce the waiting times to be seen by the specialist doctor. For instance, the DD and SC methods indicate that the waiting times were reduced by around one month in 2011, whereas the MP estimator suggests a drop between 27 and 29 days. The results of this study can provide health policy-makers with more evidence about the impact of freedom of choice policies, as well as evidence about the influence on the results of applying several impact evaluation methods.

Ángel Fernández-Pérez, University of Granada

Discussant: Nicolas Sirven

Well-being right before and after a nursing home admission

Background: Although nursing home admissions are a major life event and about half of the population will be admitted at some point during their life, insight into the impact of nursing home admission on the well-being of this group is scarce and inconclusive.

Methods: We compare the well-being of nursing home entrants in the months before and after admission to gain better insight into this topic. We make use of the Dutch Health Monitor surveys of 2012 and 2016, which contains various self-reported measures on health, disabilities, loneliness and anxiety. We combine these measures with administrative data to control for any pre-existing differences in health and socio-economic status through multivariate regression analyses.

Results: A nursing home entry is not associated with large differences in well-being. Nursing home residents report slightly worse scores for general health, loneliness, and the risk of anxiety and depression just after admission compared to individuals observed 6-12 months before admission. Yet individuals who are already institutionalized 12-18 months report similar well-being scores compared to those interviewed in the months before an admission, except for functional limitations.

Conclusion: Well-being issues are equally prevalent before and after nursing home admission indicating that, for those elderly who choose to move to a nursing home, large health shocks are neither a common trigger for a nursing home admission nor a common consequence of them.

Judith Bom, Erasmus University Rotterdam

Discussant: Leonie Sundmacher

Does informal care delay nursing home entry? Evidence from Dutch linked survey and administrative data

Nursing homes are costly. To contain long-term care public spending, governments encourage aging in place and the provision of care by relatives, thought as a way to delay nursing home (NH) entry. The extent to which informal care is effective at postponing NH entry remains empirically uncertain. We assess the effect of informal care receipt on the probability of transitioning to a NH within a two-year period using Dutch linked survey and administrative data. We use children's characteristics to retrieve exogenous variation in informal care receipt. Our results indicate that, on average, receiving informal care does not causally affect nursing home entry within two years among the 65+. However, for individuals reporting having a bad health or severe functional limitations, the probability of NH admission is increased. These findings contrast with previous evidence. Therefore, policy makers - from the Netherlands and other institutionally and culturally similar countries - should not merely expect that promoting informal care will result in a lower rate of NH admission. This could even hasten the NH admission of the elderly with a bad health condition.

Julien Bergeot, Universite de Cergy Pontoise

Discussant: Alain Paraponaris

Short- and Medium-Term Effects of Early Child Care on Children’s Non-cognitive Development: Evidence from Germany

In this paper we explore the early child care (ECC) reform in Germany that substantially expanded supply of child care places to children under the age of three. We investigate the effect of early child care expansion on children’s non-cognitive development in the short and medium run using difference-in-difference approach. We use data from the German Survey of Youth and Adolescents (KiGGS) and find evidence of some negative short-term effect for the cohort of children with increased universal early child care access, especially pronounced in boys and children from higher educated families. However, when these kids reach primary-school ages we find no evidence of effect, thus indicating that the initial negative effect is not persistent in the medium term.

Maryna Ivets, CINCH

Discussant: Fabrizio Mazzona

Development and Harmonization of a Generic Resource Use Measurement Instrument for the Measurement of Broader Resource Use: PECUNIA Project

Objectives. Measuring true resource-use quantities is important for generating valid cost estimates in economic evaluations. Due to the absence of acknowledged guidelines, a measurement method is often chosen based on practicality rather than methodological evidence. Furthermore, few instruments focus on the measurement of broader resource-use in sectors outside health care and their development process is rarely described in the existing literature. This study aims to describe the development process of a generic multi-sectoral multi-national self-reported resource-use measurement (RUM) instrument developed within the European PECUNIA project, the PECUNIA RUM.

Methods. For the development and harmonization of the PECUNIA RUM, the methodological approach was based on the best practices. The process was structured into five steps starting from the definition of the instrument attributes. Methodological literature regarding RUM was reviewed to develop a harmonized approach. The main cost driving elements in each sector were identified and matched with questions based on the existing instruments where possible. Questionnaire modules with questions in each sector were combined and harmonized concerning format and wording.

Results. The PECUNIA RUM instrument comprises eight modules: residence, health and social care, medication, unpaid help, education, employment, (criminal) justice, and personal expenses. The modular structure allows for selecting relevant modules and adapting the instrument to specific settings. The setup of each module and the routing are intended to reduce respondent fatigue by allowing the respondents to skip irrelevant modules.

Discussion. The PECUNIA RUM instrument can be used with a compatible valuation tool for producing comparable cost data in economic evaluations across different settings. This is the first study that transparently describes the development process of a generic multi-sectoral RUM instrument and it can provide guidance for researchers who undertake RUM instrument development.

Irina Pokhilenko, Maastricht University

Discussant: Sandy Tubeuf

Does long-term care provision reduce hospital bed-blocking? Evidence from a policy reform in Portugal

Hospital bed-blocking occurs when patients are clinically fit to be discharged but require some support outside the hospital which is not readily available, resulting in longer lengths of stay. I study whether long-term care (LTC) provision reduces hospital bed-blocking. Using individual data on emergency inpatient admissions at Portuguese hospitals during 2000-2015, I implement a triple-differences design. This design exploits variation in the timing of entry of LTC providers across regions originating from the
staggered introduction of the public LTC Network. It also exploits variation in lengths of stay between regular patients and patients exhibiting social factors that put them at risk of bed-blocking, such as living alone, having no family to care, and having inadequate housing. I find that the entry of the first home-care team in a region reduces the length of stay of individuals living alone and those with inadequate housing by 4 days relative to regular patients. These length of stay reductions do not affect the treatment received while at the hospital. Reductions in length of stay upon the entry of the first nursing home occur only for patients with high care needs. The beds freed up by bed-blockers are used to admit more elective patients.

Ana Moura, Tilburg University

Discussant: Marianne Tenand

Choice of Complementary Health Insurance Coverage and Healthcare Consumption

Objectives: In France, national health insurance (NHI) offers universal coverage but its scope is restricted. Voluntary complementary health insurance (CHI) is required to cover the shortfall. This paper explores the relationship between the levels of CHI coverage and healthcare consumption with an explicit treatment of the self-selection issue.

Methods: Cross-sectional data from a sample of 50,163 individuals insured in 2018 by one of the main French CHI providers were used. Data entail socio-demographic characteristics (e.g., gender, age, marital status, beneficiaries, …), level of coverage and reimbursement claims. The issues of endogeneity and potential non-linearity of the relationship between the level of coverage and healthcare consumption are addressed with the help of endogenous switching regression models, estimated with full information maximum likelihood (FIML). The level of coverage (low (LC), medium (MC) or high (HC)) is the switching variable expected to significantly distort the amount of expenditures reimbursed by CHI, making it possible: 1) to identify the sole contribution (through the estimation of the average treatment effect (ATE)) of the level of coverage to the amount of healthcare consumption, a part of which is eligible to CHI reimbursement, and 2) to assess the respective contribution of each control variable.

Results: The probability of choosing MC is the highest (P_mc=0.53, P_lc=0.21 and P_hc=0.26 ). Being female, older and employed, living as a couple and needing specialised care are found to be positively and significantly linked to choosing HC (p < 0.001). There exists a significant and negative self-selection effect: unobserved variables that positively explain the level of coverage may also lead to lower healthcare consumption. As expected, individuals with HC are more likely to consume healthcare, compared to individuals with MC (ATE_hc/mc=67.84€  ), but surprisingly HC and MC insured  spend less than LC individuals (ATE_hc/lc=-262.58€ and ATE_mc/lc=-200,62€).

Discussion: Studies on the choice of CHI levelcoverage level are not numerous. Here, the CHI choice is certainly distorted by the NHI reimbursements that cover the main part of expenses, thus leading more deprived individuals to prefer LC. Yet, our results may be due to intertwined effects that deserve to be disentangled. The negative self-selection effect may be explained by advantageous selection on unobservables, and fueled by the unobserved individual risk aversion, income or education level. In addition, there may exist some kind of inertia in coverage choices.

Anne-Kim Ristori, MGEN Foundation for Public Health

Discussant: Matt Sutton

Effect of tuberculosis on labor market participation: recent evidence from South Africa

We study the effect of tuberculosis (TB) on labor market, income and household consumption in South Africa during the period of 2008 to 2014. TB is an infectious disease in its active state and carries a high mortality rate. However, some disease carriers may be active in the labor market, if they carry the non-infectious latent disease, or have initiated treatment. We evaluate whether infection with TB has an impact on the labor market. To study this, we construct a two-period individual-level data-set with pre-balanced covariates through coarsened exact matching (CEM), supplemented with a propensity score (PS) in stratification. We evaluate models of tuberculosis on changes in employment, household income and household consumption after matching. We also compare the results with a standard regression adjustment model, and propensity score based regression adjustment. We find that TB leads to a 3 to 4% higher probability of losing jobs, and 6 to 8% lower probability of finding jobs. The effect is more pronounced in older workers (> 35 years), males, urban workers, those with less education, less skills and have bank accounts. The effect of the disease through the employment channel is significant on individual and household income. However, household consumption is not significantly affected.

Helena Ting, Graduate Institute Geneva

Discussant: Carlos Riumallo Herl

Preferences for distributing health care: Balancing efficiency and equity principles

Publicly funded health care sectors around the world face the same problem of how to set priorities in allocation of scare health care resources. For welfare states, a central concern is how to define an optimal allocation of health care resources balancing efficiency and equity considerations. These objects are seldom aligned for heterogeneous patients groups and trade-offs are inevitable (Olsen 1997, Ahlert and Schwettmann, 2017). 

The present study sets out to elicit the general public’s preferences for conflicting priority principles allocating health care resources to heterogeneous patient groups. More specifically, we test whether the general public are willing to prioritize health care to specific patient groups in order to either ensure maximization of total health gain or ensure equity objectives compromising maximization of total health gain. We apply a stated preference approach with illustrative binary choices clarifying the principles’ distribution of health care and resulting health gain. Importantly, the patient groups bearing the opportunity costs can be identified.

A web-based survey was conducted in 2018/2019 using a random sample of 1.791 members of the general public in Denmark. Overall, we find that respondents are willing to prioritize resources for specific patient groups. However, we find that preferences are heterogeneous and to some extent driven by socio-economic differences as minorities. Despite the heterogeneity, most support is found for the principle prioritizing equal treatment to patient groups in poorest health, but the support depends on the total health gain.

Lise Desireé Hansen, VIVE (The Danish Center for social science research) and DACHE (The Danish Centre for Health Economics)

Discussant: Owen O'Donnell

Diffusion of a New Drug among Ambulatory Physicians in Germany - The Influence of Patients’ Pathways

When new prescribing drugs enter the market, physicians’ prescribing behavior plays a crucial role in the diffusion process. Even though regulations to foster economically efficient prescribing exist, physicians have some degrees of freedom in their choice of medication and are, at the same time, subject to various influencing factors. The aim of the present analysis is to investigate how the interaction among patients and physicians affects the diffusion of new drugs. We look at three different ways of how patients might induce prescriptions and examine these effects for the diffusion of Sacubitril/Valsartan (SV). Based on administrative data from Germany, we identify physicians who prescribe SV and the month of their first prescription since market admission within two years (2016-2017). In a survival model, we estimate the impact of the patient-physician interaction on the physicians’ adoption time, controlling for various influencing factors. For each physician we therefore determine whether he treated patients with premedication and how many physicians already prescribing S/V are being connected through patients’ pathways in patient-sharing networks. Our main findings are that patients induce adoption by demanding follow-up prescriptions at new physicians, and that patients establish connections among physicians, which ultimately lead to new prescriptions and the diffusion of the drug. Our results suggest that the diffusion of a new drug is significantly influenced by patients’ pathways.

Ronja Flemming, LMU Munich for the WirtMed consortium

Discussant: Amani Thomas Moril

Health misperception and healthcare utilisation among older Europeans

Health perception biases can have serious consequences on health. Despite their relevance, the role of such biases in determining healthcare utilisation is severely underexplored. Here we study the relationship between health misperception, doctor visits, and concomitant out-of-pocket expenditures for the population 50+ in Europe. We conceptualise health misperception as arising from either overconfidence or underconfidence, where overconfidence is measured as overestimation of health and underconfidence is measured as underestimation of health. Comparing objective performance measures and their self-reported equivalents from the Survey of Health, Ageing and Retirement in Europe, we find that individuals who overestimate their health visit the doctor 14% less often than individuals who correctly assess their health, which is crucial for preventive care such as screenings. Lower healthcare utilisation is accompanied by lower out-of-pocket spending (38% less). In contrast, individuals who underestimate their health visit the doctor more often (28% more) and have higher out-of-pocket spending (17% more). We project that underestimating health of the population 50+ will cost the average European country Intl$ 71 million in 2020 and Intl$ 81 million by 2060. The results are robust to several sensitivity tests and, more important, to various conceptualisations of the misperception measure.

Sonja Spitzer, Vienna Institute of Demography

Discussant: Dorte Gyrd-Hansen

Discrepancy between cost and payment of an innovative medical procedure in DRG-based payment system: the role of patients and hospitals’ characteristics

OBJECTIVE: This study aims to evaluate the difference between the production cost and the reimbursement tariff of mechanical thrombectomy (MT) procedures with Stent Retriever (SR) or Contact Aspiration (CA) and to identify its determinants.

METHODS: Cost analysis was based on a multicenter, prospective randomized controlled clinical trial, ASTER1 (Contact Aspiration vs Stent Retriever for Successful Revascularization). In order to estimate the production cost of MT procedure, individual data on medical resources used during the MT procedure were collected based on a micro-costing approach and valued by their unit costs. The time horizon was the duration of the MT procedure. Hospital perspective was adopted. Multiple linear and logistic regressions were performed in order to identify respectively the determinants of the procedure cost and of the difference between procedure cost and reimbursement tariffs. Costs were estimated in 2016 euros. Analyses were performed with the SAS software, version 9.4.

RESULTS: A total of 381 patients from 8 high volumes hospitals were included in the ASTER1 study: 192 in the SR group and 189 in the CA group. Mean procedure cost on the overall sample was €5,855 (SD=€1,706) in the SR group and €5,461 (SD= €2,136) in the CA group. Device prices were the main component of MT procedures cost. SR procedure was significantly more expensive than CA procedure by €394 [€22.9 - €787.6]. The mean difference between the procedure cost and its tariff was €1,568 and €1,515 respectively in the SR and CA group. Mean procedure cost per hospital was superior to its reimbursement tariff for almost all hospitals. Specifically, the procedure cost is more likely to be greater than the reimbursement tariff when patients were women, had a history of cholesterol, had a successful revascularization, or had experienced a failure of the first technique. Regarding hospitals’ characteristics, those, which had more neurosurgeons and beds in neurology ward or had more neurosurgeons and beds in neurology ward, were less likely to have a procedure cost greater than the reimbursement tariff.

CONCLUSIONS: The comparison between the estimated procedure costs of MT and its reimbursement tariffs showed that hospitals with a neuroradiology platform would potentially experience a financial loss when producing these procedures. This result highlights a potential risk of hospitals, being less proned to offer mechanical thrombectomy due to funding issues.

KOSSI Dédé Sika, Paris Descartes University-Interdisciplinary Laboratory for Applied Research in Health Economics

Discussant: William Hollingworth

Do mergers affect hospital outputs and outcomes? Evidence from the English secondary care sector

There is a substantial amount of literature assessing that hospital mergers will bring improvements in terms of services, performance and quality. We examine whether such improvements hold when we consider different combinations of services and different measures of performance. We examine the impact of mergers on a large set of outputs including inpatient admissions, elective admissions, emergency admissions, outpatients, day cases and various combinations of the above and two different measures of performance. We find some evidence that merging activity positively affects hospital outputs and improves performance. Given that mergers reduce the scope for competition between hospitals the findings suggest that further merger activity may be the appropriate way of dealing with poorly performing hospitals.

Vanessa Cirulli, Sapienza University of Rome

Discussant: Marco Varkevisser

Cost-of-illness progression before and after MS diagnosis: a nationwide register-based study in Sweden of people newly diagnosed with MS and a population-based matched reference group

Objectives: Multiple sclerosis (MS) commonly onsets when one is of working-age and previous research has identified high levels of sickness absence and disability pension among people with MS even prior to MS diagnosis. We aimed to explore cost of illness progression before and after MS diagnosis in comparison to the healthcare costs and productivity losses of a population-based matched reference group.

Methods: A nationwide Swedish register-based cohort study of working-aged people first diagnosed with MS in 2010-2012 (n=2024) and references without MS (n=8095) randomly selected with stratified matching (sex, age, type of living area, and country of birth) was conducted. Observation was for up to 9 years in 2006-2016, from 4 years prior to 4 years after the year of MS diagnosis. Annual healthcare costs (inpatient and specialised outpatient care, patient co-payments, and community dispensed prescribed drug costs) and productivity losses (sickness absence and disability pension net days) calculated using human capital approach were estimated. All-cause costs per person were estimated. Mean annual cost estimates were calculated with 95% confidence intervals (CI), and Student t-tests were used to measure the mean difference for the respective costs between the MS cohort and matched references for each study year. Generalised estimating equation models will be constructed to further investigate the cost of illness development between the MS cohort and reference group. Costs are presented in Euros (EUR), inflated to 2019 values.

Results: The mean age at MS diagnosis was 40 years. Among the MS cohort, n=1988 (98.2%) had complete follow-up, the corresponding number among the reference group was n=7981 (98.6%).
The mean difference in annual productivity losses per person 4 years before MS diagnosis was 1639 EUR (95% CI: 949-2329) more among the MS cohort than the reference group. In the MS diagnosis year, the mean difference in productivity losses was 12,157 EUR (95% CI: 11,400- 12,911), and four years after it was 11,821 EUR (95% CI: 10,963-12,678).

Discussion: Preliminary results indicate that there is an excess cost due to MS already prior to MS diagnosis. The results suggest a cost burden already early in the MS disease course. Differences between the MS cohort and matched references were observed regarding productivity losses, however it is likely differences also exist regarding healthcare costs. Adjusted analyses are required to further understand the cost development.

Chantelle Murley, Division of Insurance Medicine, Department of Clinical Neuroscience, Karolinska Institutet

Discussant: Annika Herr

Errors in financial expectations and unhealthy behaviours - Evidence from Russia

According to the World Health Organization (2014, 2016), unhealthy behaviours (especially smoking and drinking) are the leading causes of the mortality crisis in Russia. This paper investigates the unknown drivers of unhealthy behaviours by examining the relationship between errors in financial expectations and unhealthy behaviours. Using the Russia Longitudinal Monitoring Survey (RLMS) with over 120,000 observations spanning 16 years, I employ a wide range of estimation methods including fixed effects LPM, conditional fixed effects Logit, fixed effects Quantile regressions, fixed effects Negative Binomial regressions and the generalized method of moments (GMM). I find that extreme optimism (overestimating their future household finances) is associated with higher probabilities of smoking and drinking, and higher level of cigarettes and alcohol consumption. Consistent with the tension reduction theory, I find that psychological ill-being is the channel through which extreme optimism is associated with unhealthy behaviours. Finally, I examine the differential effects of gender on the association between extreme optimism and smoking and drinking. I find that the associations are stronger and more profound in the males for drinking and females for smoking, which indicates Russian males and females have different copying strategies in response of stress. The results are robust using both the static and dynamic models.

Wei Jin, University of Birmingham

Discussant: Stephanie von Hinke

Responses of family doctors to shorter waiting times for specialist diagnosis of cancer

Waiting times are a common feature of publicly funded health systems due to capacity constraints. Reductions in waiting times will produce a demand response and, depending on the elasticity of demand, this may offset the increase in capacity. Much of the previous literature looks at the elasticity of demand to waiting times for elective surgery. This paper explores the demand response to a change in hospital waiting times for cancer diagnosis. In England in 2017, five hospitals became pilot locations for a new cancer waiting time standard, aiming to ensure patients will receive a definitive diagnosis within 28 days from the point of urgent GP referral for suspected cancer. We exploit this pilot to examine the GP demand response to an exogeneous shift in hospital waiting times policy. Shorter waiting times could lead GPs to alter their threshold for referral. Referral rates may increase because of the guarantee of obtaining a diagnosis quickly, with GPs eager to rule out cancer. Alternatively, referral rates may fall if GPs can now delay referring patients that they do not suspect have cancer in the knowledge that they can enter the cancer diagnostic pathway more quickly at a later date if other avenues of investigation do not result in an alternative diagnosis. This paper addresses two questions: did the pilot affect waiting times? And did this have an effect on GP referral behaviour? We use data on 148 acute hospital trusts and 6,904 GP practices in England between the years 2012/13 to 2018/19 to answer these questions. We use difference-in-differences methods with care provider (either hospital trust or GP practice) and year fixed effects, additionally controlling for time-varying care provider variables. The results of this study are suggestive of an increase in GP demand in response to shorter waiting times. For all cancers, there is weak evidence of improvements to waiting times, and weak evidence of a GP response. For LGI specific analysis, there is stronger evidence of improved waiting times, and a significant increase in referrals.

Helen Hayes, University of Manchester

Discussant: Mikael Svensson

Development of a brief, generic, modular resource-use measure: cognitive interviews with patients

Self-report resource-use measures (RUMs) are often used to collect healthcare use data from participants in healthcare studies. However, RUMs are typically adapted from existing measures on a study-by-study basis, resulting in a lack of standardisation which limits comparability across studies. The validity and reliability of existing RUMs are often uncertain as psychometric testing is rarely conducted. This paper reports on cognitive interviews with patients to test the content validity and acceptability of a new RUM (ModRUM). ModRUM is a brief, generic RUM with a core module (8 questions) on healthcare use and modules to increase depth and breadth.

A purposeful sampling strategy with maximum variation was used to recruit patients from primary care practices to participate in “think-aloud” interviews with retrospective probing. Participants were asked to verbalise their thought processes as they completed ModRUM, which allowed errors (issues with completion) to be identified. After completion, the interviewer asked follow-up and probing questions to investigate errors, clarity and acceptability.

Interviews were audio-recorded and transcribed verbatim. Three members of the research team independently scored transcripts to identify errors in comprehension, recall, judgement and response. Members met to agree on final scores. Interview transcripts were analysed qualitatively using techniques of constant comparison, to identify common themes and ideas for improvement. Data collection and analysis were performed concurrently and in rounds, which allowed ModRUM items to be revised and tested in further interviews.

Twenty participants were interviewed between December 2019 and March 2020. Interviews were conducted in three rounds, with revisions to ModRUM made iteratively and in response to interview findings. Seven participants completed the core module and 13 completed the core module plus depth questions. Of 71 issues, 28 were in comprehension, 14 in retrieval, 10 in judgement, 18 in response and 1 uncategorised. Most issues (21 issues by 2 participants) were due to inclusion of the participant’s family’s healthcare use. Other issues included using an incorrect recall period (5 issues by 2 participants) and overlooking questions leading to missing responses (9 issues by 5 participants). Common participant suggestions included highlighting important details (e.g. recall period) and providing additional definition or examples for some terms (e.g. outpatient). In probing, most participants reported that the length, content and layout of ModRUM were acceptable.

A brief, generic RUM is needed to increase comparability of health economic research. RUM development requires thorough qualitative and quantitative testing to demonstrate and enhance validity and reliability. Using cognitive interviewing we have demonstrated the acceptability and content validity of ModRUM.

Kirsty Garfield, University of Bristol

Discussant: Marjon van der Pol

Reductions in out-of-pocket costs and moral hazard delays in health care

Understanding how price affects individual behavior and health care demand remains highly relevant for researchers and policy makers. Little is known about how patients dynamically respond to a planned reduction in out-of-pocket prices and whether forward-looking individuals induce a moral hazard delay in the demand for health care. We study a Swedish cost-sharing policy, where small to moderate copayments are eliminated for patients at age 85, creating a sharp price decrease for primary care at one’s 85th birthday. We develop a “Kronut” Regression Discontinuity (RD) design, which combines an RD analysis with a Donut Regression and a Regression Kink design, and we estimate the policy effect using the full population of about 40,000 individuals in the age range 81 to 87 from two Swedish regions where prior copayments were moderate (Stockholm) and small (Västra Götaland). We find that the elderly in the Stockholm region reduced their primary care visits in the months approaching the copayment elimination. At the age threshold of 85, we find a sharp increase in primary care visits, but little impact on long-term primary care use. The fact that the elderly delayed primary care visits until after they were free-of-charge is evidence of forward-looking behavior with regard to health care prices. However, the effect is only observed in the short-term, which implies limited evidence of a pure price effect, only a moral hazard delay. We further find that in the Västra Götaland region the elimination of smaller copayments did not affect primary care use before, at or after the threshold. Sharp elimination of moderate or high copayments may be suboptimal as they induce health care delays which may impact on health care outcomes, though we find little evidence of an impact on ambulatory care or mortality in the current study.

Naimi Johansson, University of Gothenburg

Discussant: Pieter Bakx

The interplay between maternal smoking and genes in offspring birth weight

It is well-established that both the child’s genetic endowments as well as maternal smoking during pregnancy impact offspring birth weight. In this paper we move beyond the nature versus nurture debate by investigating the interaction between genetic endowments and this critical prenatal environmental exposure – maternal smoking – in determining birth weight. We draw on longitudinal data from the Avon Longitudinal Study of Parents and Children (ALSPAC) study and replicate our results using the UK Biobank. Genetic endowments of the children are proxied with a polygenic score which is constructed based on the results of a recent genome-wide association of birth weight. We instrument the maternal decision to smoke during pregnancy with a genetic variant (rs1051730) located in the nicotine receptor gene CHRNA3. This genetic variant is associated with the number of cigarettes consumed daily, and we present evidence that this is plausibly the only channel through which the maternal genetic variant affects the child’s birth weight. Additionally, we deal with the under-reporting of maternal smoking by using measures of cotinine, a biomarker of nicotine, collected from the mothers during their pregnancy. We confirm earlier findings that genetic endowments as well as maternal smoking during pregnancy significantly affects the child’s birth weight. However, we do not find evidence of meaningful interactions between genetic endowments and an adverse fetal environment, suggesting that one’s genetic predisposition cannot cushion the damaging effects of maternal smoking.

Rita Pereira, Erasmus University Rotterdam

Discussant: Stefanie Schurer

The hidden cost of Bananas: Pesticides effects on newborns' birthweight

We study the effects of air fumigation in banana plantations on newborns birth weight, during the period 2015-2017 in Ecuador. We exploit the geographic proximity to banana plantations, and the presence of plantations surface, to construct an individual measure of the exposure to pesticides. We propose three alternative identification strategies to address endogeneity of pesticides exposure. First, we propose a difference in differences strategy that exploits seasonal variations in the use of pesticides across provinces. Second, a difference in differences model which considers geographical variations in the exposure to pesticides in banana plantations relative to the exposure to other crops. Finally, we use a mother fixed effects model that accounts for those mothers who gave birth to at least two children during the period examined, and that had different exposure levels due to the mother’s change of residential address. Our results are robust to these specifications and show that birth weight is reduced by around 20 to 80 grams when newborns are exposed to air fumigation. The effect is more relevant in females, and slightly higher for non-educated mothers. Results from these heterogeneous effects analysis imply a birth weight reduction by around 70 to 200 grams.

Bernard Moscoso, University of Barcelona

Discussant: Martina Bozzola

A Novel Application of an EQ-5D-5L Valuation Study in Examining Preference Subgroup Analysis using a Nonparametric Bayesian Method: Comparing Migrants to Natives

A multitude of policy-related research questions emerges from national EQ-5D-5L valuation studies that pertain to health state utility valuations. More specifically, these policy-related research questions are often interested in examining subgroups and preference heterogeneity that exists within the outcomes of health state utility valuation data. The present study uses a novel smaller design EQ-5D-5L valuation study to examine preference heterogeneity using composite time trade-off (cTTO) data from a group of Polish migrants residing fulltime in Ireland and a group of native Irish.

We used an orthogonal design of 25 health states plus five mild states to compare the health preferences of a group of Polish migrants residing fulltime in Ireland and a group of native Irish coupled with a nonparametric Bayesian method using cTTO data.

A total of 243 interviews were completed (119 Polish and 123 Irish). This resulted in a total of 2655 observations being used in the analysis. The Polish migrants’ predicted mean utility valuations are lower for mild states and higher for intermediate and severe states compared to the native Irish on average.

Preference heterogeneity is evident between Polish migrants and native Irish, which is seen through the differences in their respective health state utility valuations. The smaller design valuation study and methodology used in this study provides an efficient approach to examine preference heterogeneity and the health preferences of a nation to determine whether the national value set requires updating to ensure more accurate and fair healthcare resource allocation decisions.

Dan Kelleher, National University of Ireland, Galway

Discussant: Daphne Voormolen

Investigating the effect of student psychological fragilities on control beliefs

Perceived control has been shown to play a major role in human capital investment decisions. A growing number of evidence show that these control beliefs are impacted by individual's health. Many voices are rising to report increasing mental health problems among university students all around the world. This paper studies how control beliefs on future success are impacted by psychological fragilities. To estimate the impact of psychological fragilities on control beliefs, a cross-section survey on French higher education students is used. Control beliefs are distinguished from the weights on success of internal factors (on which individual as control on) and of external factors (beyond individual's control). The relative difference between these weights are especially of interest in the light of an effort decision. Using sunshine as an instrumental variable, the strategy implemented uncover a positive effect of psychological fragilities on internal factors weight and a negative effect on external factors weight. Considering the relative difference, it appears that the positive effect on internal factors weight is greater than the negative one on external factors weight. This result is in line with psychological literature results that depressed individuals blame themselves more, as if they had control over everything.

Doriane Mignon, LEDa-LEGOS, Université Paris-Dauphine, PSL

Discussant: Apostolos Davillas

The Impact of the Crisis-Induced Reduction in Air Pollution on Infant Mortality in India: A Policy Perspective

Credible estimates of the health effects associated with changes in air pollution exposure are of considerable importance for research and policy agenda, especially for the developing countries. The paper estimates the causal impact of the sharp reduction in particulate air pollution driven by the Global Financial Crisis 2008 on the district-level infant mortality in India. Utilizing plausibly exogenous geographic variation in the crisis-induced changes in air quality and novel data from household surveys and satellite-based sources, we find that infant mortality rate fell by 24% more in the affected districts, implying 1338 fewer infants deaths than would have occurred in the absence of the crisis. Analysis of the mechanism indicates that the effect is strongest in the postneonatal period and specific for respiratory diseases. Our calculations suggest that the estimated decline in infant mortality translates into a three-year after crisis total of 312.5 million U.S. dollars. Resulting health benefits can be used as a benchmark for assessing the efficiency of the policies designed to improve air quality in India.

Olexiy Kyrychenko, CERGE-EI: Center for Economic Research and Graduate Education - Economics Institute

Discussant: Joan Calzada

Household Perception Factors That Influence the Decision to Enroll in Health Insurance Scheme in Tanzania - a Cross-Sectional Study

Background: Several countries including Tanzania, have established voluntary non-profit insurance schemes, commonly known as community-based health insurance schemes (CBHIs), that typically target rural populations and the informal sector. This paper considers the importance of household perceptions towards a CBHIs in Tanzania and their role in explaining the enrolment decision of households.

Methods: This is a cross-sectional household survey that involves 722 households located in Bahi and Chamwino districts in Dodoma region. A three-stage sampling procedure was used, and the data were analyzed using both factor analysis (FA) and principal component analysis (PCA). Statistical tests such as Bartlett’s test of sphericity, Kaiser-Meyer-Olkin (KMO) for sampling adequacy, and the Cronbach’s alpha to measure internal consistency and scale reliability were performed to examine the suitability of the data for PCA and FA. Finally, multivariate logistic regressions were run to determine the associations between the identified factors and the insurance enrolment status. 

Results: PCA identified 7 perception factors while FA identified 4 factors. The quality of healthcare services, preferences (social beliefs), and accessibility to insurance scheme administration (convenience) were the common most important factors identified by the two methods. Multivariate logistic regressions showed that the factors identified from the two methods differed somewhat in importance when considered as independent predictors of the enrollment status. The most important perception factors in terms of strength of association (odds ratio) and statistical significance were accessibility to insurance scheme administration (convenience), preferences (beliefs), and the quality of health care services. Age and income were the only demographic characteristics that were significant.

Conclusion: Household perceptions influence households’ decisions to enroll in CBHIs. Policymakers should recognize and consider the role of perceptions when designing policies that aim to increase the enrolment into CBHIs.

Alphoncina Kagaigai, University of Oslo

Discussant: Igna Bonfrer

The relationship between adolescent non-cognitive skills, adult allostatic load and premature mortality: a mediation analysis

There is a growing consensus that non-cognitive skills are an important predictor of health and mortality. However, much of the current evidence focuses on non-cognitive skills in adulthood, and there is little understanding of the mediating factors in these relationships. In this study, we decompose the total association between adolescent non-cognitive skills and allostatic load at ages 44-45 into direct associations and indirect associations operating through employment status, education, smoking behaviour, alcohol consumption and relationship status in mid-adulthood. We then replicate this analysis for mortality before age 58 using allostatic load as a further mediator.  We use data on 4,306 individuals from the National Child Development Study, a birth cohort study following individuals born in a single week in 1958. We use inverse probability weights to correct for non-random attrition. We draw on three dimensions of the ‘Big Five’ definition of non-cognitive skills measured at age 16: conscientiousness, agreeableness and neuroticism. We find some evidence of an indirect relationship between our adolescent non-cognitive skills and both allostatic load and mortality. We find that adolescent conscientiousness and agreeableness have a negative indirect relationship with allostatic load, and adolescent neuroticism has a positive indirect relationship. Our results show that these are driven largely by smoking behaviour at age 33, and also in part by education level. We also find that adolescent agreeableness has a negative indirect relationship with mortality, and adolescent neuroticism has a positive indirect relationship. These are largely driven by smoking behaviour at age 33. Our results suggest there may be a need for policy that targets the improvement of conscientiousness and agreeableness and reduces neuroticism, alongside policies to reduce smoking and improve educational attainment, to improve adult health and longevity.

Rose Atkins, Health Organisation, Policy and Economics (HOPE), University of Manchester

Discussant: Murat A Mercan

Do soda taxes affect the consumption and health of school-aged children? Evidence from France and Hungary

This paper examines the effect of soda taxes on the consumption behaviour and health of school-aged children in Europe: Hungary imposed a "Public Health Product Tax" on several unhealthy products, including sodas, in 2011. France introduced solely a tax on sodas, containing sugar or artificial sweeteners, in 2012. In order to exploit spatial variation, I use a semi-parametric difference-in-differences (DID) approach. Since the policies differ in Hungary and France, I analyse the effects separately by using a neighbouring country without a soda tax as control group. The results show a counter-intuitive positive effect of the tax on the soda consumption in Hungary. Reasons for this finding might be a substitution of unhealthy products as well as the decreased amount of sugar in sodas. The effect of the soda tax in France is as expected negative, but insignificant which might be caused by a low tax rate. The body mass index (BMI) is not affected by the tax in any country. Consequently, policy makers should think carefully about the design, aim, tax rate, and the possible reaction of manufacturers before implementing a soda tax.

Selina Gangl, Health Organisation, Université de Fribourg

Discussant: Edel Doherty

An RCT of a Time Preference Intervention to Increase Physical Activity

Background: Decisions about engaging in physical activity often involve trade-offs between current costs and future benefits. Economic theory suggests that these decisions may be influenced by an individual’s time preference rate (how heavily they discount the value of future events) and how present biased they are (how much they overweight the value of an immediate outcome). Psychology points to the discount rate being related to how much an individual considers the future   consequences of their behaviour and how connected they are to their future self while present bias might involve the conflict between temptation and self-control. We test a novel intervention designed to increase physical activity by improving an individual’s consideration of future consequences and their connection to their future-self and to counteract temptation by using commitment. 

Design: A 1:1:1 randomised controlled trial, run on-line and randomised via Qualtrics with no blinding possible.

Participants: Healthy adults (n= 129) aged between 30 and 65, achieving no more than 90 minutes of moderate activity per week, with access to internet and email.
Interventions:  (a) 2Cs – Consideration of long-term consequences of being inactive plus Connection to future-self via visualisation techniques plus walking plan; (b) 3Cs – as 2CS plus offer of a Commitment contract: (c) Control – walking plan alone.

Objective: To investigate the effectiveness of 2Cs and 3Cs versus Control over a four week intervention period plus a further four week follow up period.
Outcome: Pedometer-based average daily step count for each week / Self-reported activity minutes for each week.

Results: 129 participants were recruited via flyers between February 2019 and January 2020. Retention at week eight was 71%. Results show increased activity at week eight of 2100 steps per day/275 minutes per week (p < .01) across all arms relative to baseline. Neither the 2Cs nor the 3Cs had significant differences at any week compared to control.

Conclusions: All groups were successful in increasing their physical activity over baseline, however neither intervention showed a significant difference relative to the control.

Uma Thomas, Health Organisation, University of Aberdeen

Discussant: Hans van Kippersluis