Mapping of Family Reported Outcome Measure (FROM-16) scores to EQ-5D: algorithm to calculate utility values

OBJECTIVE

Although decision scientists and health economists encourage inclusion of family member/informal carer utility in health economic evaluation, there is a lack of suitable utility measures comparable to patient utility measures such those based on the EQ-5D. This study aims to predict EQ-5D-3L utility values from Family Reported Outcome Measure (FROM-16) scores, to allow the use of FROM-16 data in health economic evaluation when EQ-5D data is not available.

METHODS

Data from 4228 family members/partners of patients recruited to an online cross-sectional study through 58 UK-based patient support groups, three research support platforms and Welsh social services departments were randomly divided five times into two groups, to derive and test a mapping model. Split-half cross-validation was employed, resulting in a total of ten multinomial logistic regression models. The Monte Carlo simulation procedure was used to generate predicted EQ-5D-3L responses, and utility scores were calculated and compared against observed values. Mean error and mean absolute error were calculated for all ten validation models. The final model algorithm was derived using the entire sample.

RESULTS

The model was highly predictive, and its repeated fitting using multinomial logistic regression demonstrated a stable model. The mean differences between predicted and observed health utility estimates ranged from 0.005 to 0.029 across the ten modelling exercises, with an average overall difference of 0.015 (a 2.2% overestimate, not of clinical importance).

CONCLUSIONS

The algorithm developed will enable researchers and decision scientists to calculate EQ-5D health utility estimates from FROM-16 scores, thus allowing the inclusion of the family impact of disease in health economic evaluation of medical interventions when EQ-5D data is not available.

Exploring the relationship between baseline health-related quality of life (AQoL-4D utility values) and mortality during long-term follow-up for people living with a psychotic disorder

Holistic summary measures of health-related quality of life (HRQoL) could provide useful information for screening and/or monitoring high-risk individuals. This study explores the association of baseline HRQoL (AQoL-4D values) with death from natural and external causes (such as suicide, overdose, or accident) in a prevalent cohort over a mean follow-up of 6.9-years. Baseline data were obtained from the 2010 National Survey of High Impact Psychosis (SHIP) and its 2012 sister survey, the North-Metro Survey of High Impact Psychosis. Follow-up data were obtained from the National Deaths Index (to January 18, 2018). Associations between baseline HRQoL (and the 12 individual items of the AQoL-4D) and deaths by cause, stratified by sex were investigated using multivariable Cox Proportional Hazards models, with age used as the time scale. Overall, 7.6 % of the cohort (137 of 1805 people meeting ICD-10 criteria for a psychotic disorder and agreeing to linkage) died, 4.2 % from natural causes (primarily heart disease and cancer) and 3.4 % from external causes. There was an association between low HRQoL (AQoL-4D ≤ 0.20) and mortality due to natural causes, primarily driven by difficulties with mobility and vision in men, and difficulty with household tasks in women. No significant associations were obtained between AQoL-4D utilities and death from external causes. Sleep problems were associated with death from external causes for men only. As people with psychotic disorders experience multiple causes of death with complex associations which are difficult to quantify, monitoring for low HRQoL using the AQoL-4D could be a useful indicator of increased mortality risk in this population. TWITTER: Monitoring for low HRQoL (AQoL-4D ≤ 0.20) could be a useful indicator of increased mortality risk from natural causes in people living with psychotic disorders, with differing but specific items of relevance to men and women.

Quantifying health-related quality of life in Malaysian type 2 diabetes: focusing on complication types and severity

PURPOSE

There is a knowledge gap of health utility values for Type 2 Diabetes Mellitus (T2DM) complications in Malaysia. This study aimed to estimate EQ-5D-5L utility values and evaluate health-related quality of life (HRQoL) for Malaysian T2DM associated with complications and clinical characteristics.

METHODS

A cross-sectional study was conducted on T2DM patients at a tertiary hospital outpatient using the Malay and English version of the EQ-5D-5L questionnaire. Health utility values were derived using the Malaysian EQ-5D-5L value set. Ordinary least squares (OLS) multivariable regression model was used to estimate the health utility decrements associated with T2DM-related complications and clinical characteristics.

RESULTS

A total of 513 T2DM patients were recruited. Overall, pain was the most affected of all five EQ-5D-5L dimensions. Patients with foot ulcer, amputation, severe heart failure and frequent hypoglycemia reported more problems collectively in all EQ-5D-5L dimensions. Older age, lower education level, longer duration of T2DM, urine protein creatine index (UPCI) > 0.02 g/mmol, and injection therapy were significantly associated with lower EQ-5D-5L utility values (p < 0.004, Bonferroni adjusted). The lowest unadjusted utility values were reported for severe heart failure 0.65 (interquartile range, IQR 0.50), frequent hypoglycemia 0.74 (0.22) and being amputated 0.78 (0.47). In the multivariable regression model after controlling for sociodemographic and clinical characteristics, the largest utility value decrement was observed for amputation (- 0.158, SE 0.087, p = 0.05), frequent hypoglycemia (- 0.101, SE 0.030, p = 0.001), myocardial infarction (-0.050, SE 0.022, p = 0.022) and obesity (-0.034, SE 0.016, p = 0.029).

CONCLUSION

Larger utility value decrements were found for severe stages of complications. These findings suggest the value of defining severity of complications in utility elicitation studies. The utility decrement quantified for different T2DM complication severity will be useful for economic evaluations within diabetic-related fields.

A survey exploring caregiver burden and health-related quality of life in hereditary transthyretin amyloidosis

BACKGROUND

Hereditary transthyretin amyloidosis (ATTRv) is an ultra-rare, life-shortening disease with a high unmet need. This study examined ATTRv caregiver health-related quality of life (HRQoL) and productivity.

METHODS

A cross-sectional online survey, including EQ-5D-3L, Hospital anxiety and depression scale (HADS), and caregiver and patient characteristics questions, was developed to assess ATTRv caregiver burden. A companion general population survey collected EQ-5D-3L, HADS and chronic health conditions data. Caregiver-control group differences in HRQoL were assessed using t-tests and chi-square tests. Ordinary Least Squares regression was used to estimate the disutility of being a caregiver compared to controls stratified by patient ambulatory status.

RESULTS

Thirty-six caregivers and matched controls completed the survey (n = 72). The disease severity of patients they cared for was varied: 33% required no assistance walking, 58% required assistance with walking and 9% required a wheelchair/were bedridden. On average, caregivers spent 6 h daily on practical care and 4 h daily on emotional support. Fifty-six percent indicated that they had changed their employment due to providing ATTRv care. Caregivers reported lower HRQoL, as indicated by lower EQ-5D 3L utility scores (M = 0.772, SD = 0.178 vs. M = 0.849, SD = 0.218) and higher HADS anxiety (9.3 vs. 6.1, p < 0.01) and depression (7.6 vs. 4.4, p < 0.01) scores, compared with matched controls. Caregivers were also more likely to report sleep problems (33% vs. 8%, p < 0.01) and stress (42% vs. 0%, p < 0.001) as chronic conditions than controls.

CONCLUSIONS

The study results indicate that caring for a person with ATTRv can have a considerable negative impact on caregivers' HRQoL and productivity. The study findings provide important information for economic evaluations of ATTRv treatments.

Assessment of health-related quality of life in Australian patients with idiopathic pulmonary fibrosis: a comparison of the EQ-5D-5L and the AQoL-8D

PURPOSE

Idiopathic pulmonary fibrosis (IPF) is a progressive and debilitating chronic lung disease with a high symptom burden, which has a substantial impact on health-related quality of life (HRQoL). Our study aimed to assess the suitability of the EuroQol five-dimension (EQ-5D-5L) and the Assessment of Quality of Life- eight-dimension (AQoL-8D) questionnaires in measuring HRQoL as health state utility values (HSUVs) in an Australian IPF cohort.

METHODS

Data for estimation of health state utility values (HSUVs) were collected from participants of the Australian IPF Registry (AIPFR) using self-administered surveys which included the EQ-5D-5L and the AQoL-8D. Data on lung function and disease specific HRQoL instruments were collected from the AIPFR. Performance of the two instruments was evaluated based on questionnaire practicality, agreement between the two instruments and test performance (internal and construct validity).

RESULTS

Overall completion rates for the EQ-5D-5L and AQoL-8D were 96% and 85%, respectively. Mean (median) HSUVs were 0.65 (0.70) and 0.69 (0.72) for the EQ-5D-5L and AQoL-8D, respectively. There was reasonable agreement between the two instruments based on the Bland-Altman plot mean difference (-0.04) and intraclass correlation coefficient (0.84), however there were some fundamental differences. A larger range of values was observed with the EQ-5D-5L (-0.57-1.00 vs 0.16-1.00). The EQ-5D-5L had a greater divergent sensitivity and efficacy in relation to assessing HSUVs between clinical groupings. The AQoL-8D ,however, had a higher sensitivity to measure psychosocial aspects of HRQoL in IPF.

CONCLUSION

The EQ-5D-5L demonstrated superior performance when compared to AQoL-8D in persons with IPF. This may be attributable to the high symptom burden which is physically debilitating to which the EQ-5D-5L may be more sensitive.

Defining Utility Values for Chorea Health States in Patients with Huntington's Disease

Introduction

Chorea is characterized by sudden, involuntary movements that interfere with quality of life (QOL). Utility values measure preferences for different health states and reflect societal perceived disease severity. To date, no studies have reported utility values specifically for Huntington’s disease (HD) chorea. We estimated impact on QOL of HD chorea severity using utility values from the general population.

Methods

Participants were enrolled using computer-assisted telephone interviews. Participants read vignettes describing four health states for varying levels of chorea severity, with the same underlying HD severity. Time trade-off (TTO) methods were used to estimate utility values, which range from −1 (worse than death) to +1 (perfect health) and represent the number of years in an imperfect health state an individual is willing to give up to live in full health. TTO utilities were augmented with visual analog scale (VAS) participant responses. The primary outcome was HD chorea utility estimated by TTO.

Results

Mean ± SD TTO-derived utility values were 0.07 ± 0.52, 0.26 ± 0.50, 0.48 ± 0.47, and 0.64 ± 0.41 for severe, moderate/severe, moderate/mild, and mild chorea severity, respectively. Differences between each health state and its adjacent less severe health state were statistically significant (all P < 0.0001). Respondents were willing to give up 3.6, 5.2, 7.4, and 9.3 years during a 10-year life span to avoid living with mild, mild/moderate, moderate/severe, and severe chorea, respectively. VAS and TTO results were consistent.

Conclusions

Significant decreases in utility values were seen as HD chorea severity increased. These data can be leveraged for cost-effectiveness modeling to better understand the value of treatments for chorea.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12325-022-02046-z.

Examining the association between menstrual symptoms and health-related quality of life among working women in Japan using the EQ-5D

Background

Menstrual symptoms have been identified as a substantial burden among women of reproductive age, affecting their health status and quality of life globally. A range of menstrual symptoms have been studied as they affect the health-related quality of life (HRQoL), showing variations across specific menstrual symptoms and study settings. A major concern is demonstrated due to menstrual symptoms in women’s professional and social life, and consequently societal and economic loss for women and the society at large. Yet evidence is scarce that estimates the index form HRQoL score related to menstrual symptoms that is needed for health economic evaluations.

Methods

This study aims to investigate the association between menstrual symptoms and the HRQoL among working women in Japan in an index form, using a self-reporting questionnaire (n = 6048). The EQ-5D-3L (EuroQoL 5-dimension 3-level) is used that is a widely used tool to measure health outcomes for health economic evaluations globally. Multivariate regression analysis is conducted to assess the association between the HRQoL score and specific nineteen physical and mental conditions related to menstruation (e.g., pain, heavy bleeding, concentration, negative affect).

Results

The index form HRQoL score for menstrual symptoms is estimated as 0.682 in the study population (where a score one suggests perfect health). The association of the HRQoL score varies substantially across the menstrual symptoms. Several of the physical conditions and disorders show a substantial negative association with the HRQoL score. Also, most of the mental and psychological issues are significantly and negatively related to the HRQoL score.

Conclusions

This study suggests that HRQoL is substantially and negatively affected by menstruation among working women in Japan. Distinct variations of negative influences across menstrual symptoms underscore the multi-dimensional nature of menstruation and consequently the need of collective interventions to address these difficulties. The evidence of HRQoL continues to be an important area for future research on women’s health and health economic evaluations to inform effective and efficient resource allocations for relevant health policies and financing strategies.

Investigating health-related quality of life in rare diseases: a case study in utility value determination for patients with CLN2 disease (neuronal ceroid lipofuscinosis type 2)

Background

Utility studies enable preference-based quantification of a disease’s impact on patients’ health-related quality of life (HRQoL). It is often difficult to obtain utility values for rare, neurodegenerative conditions due to cognitive burden of direct elicitation methods, and the limited size of patient/caregiver populations. CLN2 disease (neuronal ceroid lipofuscinosis type 2) is an ultra-rare, progressive condition, for which there are no published utility data fully capturing all disease stages. This case study demonstrates how utility values can be estimated for ultra-rare paediatric diseases by asking clinicians to complete EQ-5D-5L questionnaires based on vignettes describing the stages of CLN2 disease.

Methods

An indirect elicitation method using proxy-reporting by clinical experts was adopted. Eighteen vignettes were developed, describing nine progressive disease stages as defined by motor and language domain scores of the CLN2 Clinical Rating Scale, in individuals treated with cerliponase alfa or standard care. Eight clinical experts with experience of treating CLN2 disease with cerliponase alfa and current standard care completed the proxy version 2 EQ-5D-5L online after reading these vignettes. Resulting scores were converted to EQ-5D-5L utility values for each disease stage, using UK, German and Spanish value sets.

Results

Utility values, which are typically anchored by 0 (equivalent to death) and 1 (full health), decreased with CLN2 disease progression (results spanned the maximum range of the utility scale). Assigned utility values were consistently higher for patients receiving cerliponase alfa than standard care; differences were statistically significant for the 6 most severe disease stages (p < 0.05). Analysis of the individual dimensions of the EQ-5D-5L showed that greatest differences between patients treated with cerliponase alfa and standard care occurred in the pain dimension (differences in mean scores ranged between no difference and 1.8), with notable differences also observed in the anxiety/depression dimension (differences in mean scores ranged between 0.1 and 1.0).

Conclusions

This study demonstrates a feasible methodology for eliciting utility values in CLN2 disease, indicating HRQoL declines with disease progression. Vignettes describing patients receiving cerliponase alfa were consistently assigned higher utility values for the same disease state, suggesting this treatment improves HRQoL compared with standard care.

Trial registration NCT01907087, NCT02485899.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13023-021-01829-x.

A systematic review of health state utility values for thyroid cancer

Purpose

Health state utility values are commonly used to inform economic evaluations and determine the cost-effectiveness of an intervention. The aim of this systematic review is to summarise the utility values available to represent the health-related quality of life (HRQoL) of patients with thyroid cancer.

Methods

Eight electronic databases were searched from January 1999 to April 2019 for studies which included assessment of HRQoL for patients with thyroid cancer. Utility estimates derived from multiple sources (EuroQol questionnaire 5-dimension (EQ-5D), time trade-off [TTO] and standard gamble [SG] methods) were extracted. In addition, utility estimates were generated by mapping from SF-36 and EORTC QLQ-30 to the EQ-5D-3L UK value set using published mapping algorithms.

Results

Searches identified 33 eligible studies. Twenty-six studies reported HRQoL for patients with differentiated thyroid cancer and seven studies for patients with general thyroid cancer. We identified studies which used different methods and tools to quantify the HRQoL in patients with thyroid cancer, such as the EQ-5D-3L, SF-36, EORTC QLQ-30 and SG and TTO techniques to estimate utility values. Utility estimates range from 0.205 (patients with low-risk differentiated thyroid cancer) to utility values approximate to the average UK population (following successful thyroidectomy surgery and radioiodine treatment). Utility estimates for different health states, across thyroid cancer sub-types and interventions are presented.

Conclusion

A catalogue of utility values is provided for use when carrying out economic modelling of thyroid cancer; by including mapped values, this approach broadens the scope of health states that can be considered within cost-effectiveness modelling.

Electronic supplementary material

The online version of this article (10.1007/s11136-020-02676-2) contains supplementary material, which is available to authorised users.

Characterizing health state utilities associated with Duchenne muscular dystrophy: a systematic review

Background

Preferences for health states for Duchenne muscular dystrophy (DMD) are necessary to assess costs and benefits of novel therapies. Because DMD progression begins in childhood, the impact of DMD on health-related quality-of-life (HRQoL) affects preferences of both DMD patients and their families. The objective of this review was to synthesize published evidence for health state utility from the DMD patient and caregiver perspectives.

Methods

A systematic review was performed using MEDLINE and Embase, according to best practices. Data were extracted from studies reporting DMD patient or caregiver utilities; these included study and patient characteristics, health states considered, and utility estimates. Quality appraisal of studies was performed.

Results

From 888 abstracts, eight publications describing five studies were identified. DMD utility estimates were from preference-based measures presented stratified by ambulatory status, ventilation, and age. Patient (or patient–proxy) utility estimates ranged from 0.75 (early ambulatory DMD) to 0.05 (day-and-night ventilation). Caregiver utilities ranged from 0.87 (for caregivers of adults with DMD) to 0.71 (for caregivers of predominantly childhood patients). Both patient and caregiver utilities trended lower with higher disease severity. Variability in utilities was observed based on instrument, respondent type, and country. Utility estimates for health states within non-ambulatory DMD are under reported; nor were utilities for DMD-related health states such as scoliosis or preserved upper limb function identified.

Conclusion

Published health state utilities document the substantial HRQoL impacts of DMD, particularly with disease progression. Additional research in patient utilities for additional health states, particularly in non-ambulatory DMD patients, is warranted.

Electronic supplementary material

The online version of this article (10.1007/s11136-019-02355-x) contains supplementary material, which is available to authorized users.

Do socio-demographic characteristics and/or health status explain the magnitude of differences between patient and general public utility values? A chronic low back pain patients case study

Background

Utility values can be obtained from different respondent groups, including patients and members of the general public. Evidence suggests that patient values are typically higher than general public values. This study explores whether the magnitude of disagreement between both values can be explained by socio-demographic characteristics and/or health status.

Methods

Data of 5037 chronic low back pain patients were used. Self-reported EQ-VAS was employed as a proxy of patients’ preference for their own health state. General public values for the patients’ EQ-5D-3L health states were obtained using the Dutch VAS-based tariff. The difference between patient and general public values was assessed using a paired t-test. Subsequently, this difference was used as a dependent variable and regressed upon dummy variables of socio-demographic and health status characteristics. Coefficients represented age, gender, education level, social support, back pain intensity, leg pain intensity, functional status, comorbidities, catastrophizing, and treatment expectations.

Results

Patient values were higher than general public values (0.069; 95%CI:0.063–0.076). The magnitude of disagreement between both values was associated with age, gender, education level, social support, functional status, and comorbidities, but not with back pain intensity, leg pain intensity, catastrophizing, and treatment expectations.

Conclusions

Patients were found to value their own health status higher than members of the general public. The magnitude of disagreement between both values was found to differ by various socio-demographic and/or health status characteristics. This suggest that patient characteristics account for a relevant fraction of the identified disagreements between patient and general public values, and that mechanisms thought to be responsible for these disagreements, such as adaptation and response shift, have a differential impact across patient sub-groups.

Social preferences for health states associated with acute myeloid leukemia for patients undergoing treatment in the United Kingdom

Background

Health state (HS) utility values for patients with acute myeloid leukemia (AML), a hematological malignancy, are not available in the United Kingdom (UK). This study aims to develop clinically sound HSs for previously untreated patients with AML and to assign utility values based on preferences of the general UK population.

Methods

This study was conducted in the UK and comprised 2 stages. During the first stage, AML HSs were drafted based on evidence from a literature review of AML clinical and health-related quality-of-life studies (published January 2000–June 2016) and patient-reported outcome measures previously used in this population. A panel of UK hematologists with AML experience validated the clinical relevance and accuracy of the HSs. During the second stage, validated HSs were valued in an elicitation survey with a representative UK population sample using the time trade-off (TTO) method. Descriptive statistics and bivariate tests were obtained and performed.

Results

A total of eight HSs were developed and clinically validated, including treatment with chemotherapy, consolidation therapy, transplant, graft-vs-host disease (GvHD), remission, relapse, refractory, and functionally cured. In total, 125 adults participated (mean age, 49.6 years [range, 18–87 years], 52.8% female). Mean (95% confidence interval [CI]) TTO preference values (n = 120), ranked from lowest (worst HS) to highest (best HS) were as follows: refractory − 0.11 (− 0.21 to − 0.01), relapse 0.10 (0.00–0.20), transplant 0.28 (0.20–0.37), treatment with chemotherapy 0.36 (0.28–0.43), GvHD 0.43 (0.36–0.50), consolidation 0.46 (0.40–0.53), remission 0.62 (0.57–0.67), and functionally cured 0.76 (0.72–0.79). Mean (95% CI) visual analog scale preference values followed the same rank order, ranging from 0.15 (0.13–0.17) for refractory to 0.71 (0.68–0.73) for functionally cured.

Conclusions

To our knowledge, this is the first study to report utility values for AML from the UK societal perspective. Participants were able to distinguish differences in severity among AML HSs, and preference values were consistent with clinical perception of HS severity. HS preference values observed in this study may be useful in future evaluations of treatment benefit, including cost-effectiveness analyses and improved patient well-being.

Challenges Associated with Estimating Utility in Wet Age-Related Macular Degeneration: A Novel Regression Analysis to Capture the Bilateral Nature of the Disease

Introduction

The estimation of utility values for the economic evaluation of therapies for wet age-related macular degeneration (AMD) is a particular challenge. Previous economic models in wet AMD have been criticized for failing to capture the bilateral nature of wet AMD by modelling visual acuity (VA) and utility values associated with the better-seeing eye only.

Methods

Here we present a de novo regression analysis using generalized estimating equations (GEE) applied to a previous dataset of time trade-off (TTO)-derived utility values from a sample of the UK population that wore contact lenses to simulate visual deterioration in wet AMD. This analysis allows utility values to be estimated as a function of VA in both the better-seeing eye (BSE) and worse-seeing eye (WSE).

Results

VAs in both the BSE and WSE were found to be statistically significant (p < 0.05) when regressed separately. When included without an interaction term, only the coefficient for VA in the BSE was significant (p = 0.04), but when an interaction term between VA in the BSE and WSE was included, only the constant term (mean TTO utility value) was significant, potentially a result of the collinearity between the VA of the two eyes. The lack of both formal model fit statistics from the GEE approach and theoretical knowledge to support the superiority of one model over another make it difficult to select the best model.

Conclusion

Limitations of this analysis arise from the potential influence of collinearity between the VA of both eyes, and the use of contact lenses to reflect VA states to obtain the original dataset. Whilst further research is required to elicit more accurate utility values for wet AMD, this novel regression analysis provides a possible source of utility values to allow future economic models to capture the quality of life impact of changes in VA in both eyes.

Funding

Novartis Pharmaceuticals UK Limited.

Electronic supplementary material

The online version of this article (doi:10.1007/s12325-017-0620-x) contains supplementary material, which is available to authorized users.

Mapping of the DLQI scores to EQ-5D utility values using ordinal logistic regression

Purpose

The Dermatology Life Quality Index (DLQI) and the European Quality of Life-5 Dimension (EQ-5D) are separate measures that may be used to gather health-related quality of life (HRQoL) information from patients. The EQ-5D is a generic measure from which health utility estimates can be derived, whereas the DLQI is a specialty-specific measure to assess HRQoL. To reduce the burden of multiple measures being administered and to enable a more disease-specific calculation of health utility estimates, we explored an established mathematical technique known as ordinal logistic regression (OLR) to develop an appropriate model to map DLQI data to EQ-5D-based health utility estimates.

Methods

Retrospective data from 4010 patients were randomly divided five times into two groups for the derivation and testing of the mapping model. Split-half cross-validation was utilized resulting in a total of ten ordinal logistic regression models for each of the five EQ-5D dimensions against age, sex, and all ten items of the DLQI. Using Monte Carlo simulation, predicted health utility estimates were derived and compared against those observed. This method was repeated for both OLR and a previously tested mapping methodology based on linear regression.

Results

The model was shown to be highly predictive and its repeated fitting demonstrated a stable model using OLR as well as linear regression. The mean differences between OLR-predicted health utility estimates and observed health utility estimates ranged from 0.0024 to 0.0239 across the ten modeling exercises, with an average overall difference of 0.0120 (a 1.6% underestimate, not of clinical importance).

Conclusions

This modeling framework developed in this study will enable researchers to calculate EQ-5D health utility estimates from a specialty-specific study population, reducing patient and economic burden.

Electronic supplementary material

The online version of this article (doi:10.1007/s11136-017-1607-4) contains supplementary material, which is available to authorized users.