Predicting health-related quality of life (EQ-5D-5L) and capability wellbeing (ICECAP-A) in the context of opiate dependence using routine clinical outcome measures: CORE-OM, LDQ and TOP

Predicting health-related quality of life (EQ-5D-5L) and capability wellbeing (ICECAP-A) in the... Background: Economic evaluation normally requires information to be collected on outcome improvement using utility values. This is often not collected during the treatment of substance use disorders making cost-effectiveness evaluations of therapy difficult. One potential solution is the use of mapping to generate utility values from clinical measures. This study develops and evaluates mapping algorithms that could be used to predict the EuroQol-5D (EQ-5D-5 L) and the ICEpop CAPability measure for Adults (ICECAP-A) from the three commonly used clinical measures; the CORE-OM, the LDQ and the TOP measures. Methods: Models were estimated using pilot trial data of heroin users in opiate substitution treatment. In the trial the EQ-5D-5 L, ICECAP-A, CORE-OM, LDQ and TOP were administered at baseline, three and twelve month time intervals. Mapping was conducted using estimation and validation datasets. The normal estimation dataset, which comprised of baseline sample data, used ordinary least squares (OLS) and tobit regression methods. Data from the baseline and three month time periods were combined to create a pooled estimation dataset. Cluster and mixed regression methods were used to map from this dataset. Predictive accuracy of the models was assessed using the root mean square error (RMSE) and the mean absolute error (MAE). Algorithms were validated using sample data from the follow-up time periods. Results: Mapping algorithms can be used to predict the ICECAP-A and the EQ-5D-5 L in the context of opiate dependence. Although both measures can be predicted, the ICECAP-A was better predicted by the clinical measures. There were no advantages of pooling the data. There were 6 chosen mapping algorithms, which had MAE scores ranging from 0.100 to 0.138 and RMSE scores ranging from 0.134 to 0.178. Conclusion: It is possible to predict the scores of the ICECAP-A and the EQ-5D-5 L with the use of mapping. In the context of opiate dependence, these algorithms provide the possibility of generating utility values from clinical measures and thus enabling economic evaluation of alternative therapy options. Trial registration: ISRCTN22608399. Date of registration: 27/04/2012. Date of first randomisation: 14/08/2012. Keywords: ICECAP, EQ-5D, Mapping, Addiction, Mental health, Preference-based measures, Condition-specific measures, Economic evaluation * Correspondence: E.Frew@bham.ac.uk Health Economics Unit, Institute of Applied Health Research, Public Health Building, University of Birmingham, B15 2TT, Birmingham, UK Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 2 of 11 Background Leeds Dependence Questionnaire (LDQ), and the Treat- In many healthcare systems around the world, resources ment Outcomes Profile (TOP). The CORE-OM and the are scarce and the demand for healthcare outweighs sup- TOP are instruments that are used to assess the treatment ply. This scarcity warrants the need for economic evalu- outcome, whilst the LDQ is used to assess the level of de- ation to aid decision makers with information about the pendence at the time of assessment [10, 11]. These mea- most efficient use of resources in order to maximise the sures, however, are unsuitable for use within health health gained for every unit of currency spent. Within economic evaluations [12]. To enable information from the UK, and in many other country jurisdictions, the these measures to be used in economic evaluations, a most common approach to economic evaluation is the process called ‘mapping’ can be applied [13]. Mapping, cost-utility analysis. The outcomes of a cost-utility ana- quantifies the relationship between different measures lysis are expressed in quality-adjusted life years (QALYs) using appropriate statistical techniques [14, 15], and al- [1]. QALYs take into account both quality and length of lows for the estimation of HRQOL and wellbeing for use life and offer a commensurate unit that allows compari- in economic evaluations using data collected from routine sons of cost-effectiveness across different disease areas clinical measures. and interventions [2]. To measure QALYs, generic This study aims to map three clinical instruments that preference-based measures of health-related quality of are often used in the routine care of addiction and opi- life (HRQOL) are required [3] to capture a broad con- ate dependence (CORE-OM, LDQ, and TOP) onto the struct of health through key dimensions that are known EQ-5D-5 L and ICECAP-A measures, generating algo- to affect quality of life [4]. Commonly, the EuroQol-5D rithms that can be used in future studies to aid reim- (EQ-5D-5 L) measure is used in economic evaluations to bursement decisions in the absence of information estimate QALYs [5]. The EQ-5D-5 L describes HRQOL related to EQ-5D-5 L and ICECAP-A. With these map- through the dimensions of mobility, self-care, usual ac- ping algorithms, data from the three clinical measures tivities, pain and discomfort, and anxiety and depression. can be translated into health and capability scores for When considering substance use disorders, the use in an economic evaluation. The mapping algorithms broader concept of wellbeing is considered more appro- were developed using data from a pilot randomised con- priate to reflect the clinical and policy objectives [6]. trol trial (RCT) that sought to explore the effectiveness Drug dependence undermines an individual’s capability of two psychosocial interventions for heroin users re- [6], and this disempowerment is largely overlooked in ceiving opiate substitution treatment (OST) in England the health economics of addiction-related interventions [10]. This is the first study to develop mapping algo- as a result of the narrow definition of HRQOL. Until re- rithms from routine clinical outcome measures used in cently, it was hard to define and quantify wellbeing for addiction therapy onto the EQ-5D-5 L and ICECAP-A. the purposes of an economic evaluation. Amartya Sen’s work on capabilities allowed for a conceptualization of Methods wellbeing through human functionings (what an individ- The study uses data collected as part of a pilot random- ual ‘does’) and capabilities (the ability of the individual ized controlled trial designed to investigate the clinical to exercise a functioning) [7]. The development of the and cost-effectiveness of two psychological interventions ICEpop CAPability measure for Adults (ICECAP-A) delivered in addition to the usual care of individuals who based on Amartya Sen’s capability approach, means that had been receiving opiate substitution treatment for wellbeing can now be measured in a way that is compat- more than one year [10]. All trial participants met the ible for use in economic evaluations [8]. The ICECAP-A ICD-10 criteria for opioid dependence and were re- measures capability wellbeing through the dimensions of cruited if they were in opiate substitution treatment with stability, enjoyment, achievement, attachment and au- methadone or buprenorphine for more than a year but tonomy. Both EQ-5D-5 L and ICECAP-A were found to still reported heroin use during the last month. The only have the appropriate construct validity within the addic- exclusion criteria were having a physical or mental tion context, but ICECAP-A appeared to be significantly health condition that prevented engagement in the psy- more responsive to changes of key clinical indicators [9]. chosocial intervention, or an imminent period of impris- In order to reduce the burden of assessment on patients onment. A number of client outcomes, including mental and to help acquire data for clinical means, particularly in health (CORE-OM), substance dependence (LDQ), phys- the context of substance use disorder, studies tend to rely ical and psychological health (TOP), health-related qual- only on clinical context-specific measures. For substance ity of life (EQ-5D-5 L), and capability wellbeing use disorders, measures that are commonly used to assess (ICECAP-A) were assessed at baseline, 3 months and the level and impact of dependence and assess the effect- 12 months post-randomisation. These outcome mea- iveness of a treatment are the Clinical Outcomes in Rou- sures are described in detail below. The trial was con- tine Evaluation - Outcome Measure (CORE-OM), the ducted by three community drug teams in England. All Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 3 of 11 trial participants provided written informed consent, and self-care, usual activities, pain and discomfort, and anx- ethical approval was obtained from the Black Country NHS iety and depression. Participants select their functioning Research Ethics Committee (reference: 12/WM/0046). level from five options ranging from 1 to 5, with 1 repre- senting that an individual has no functioning problems Outcome measures in a given dimension, whereas 5 represents severe prob- Clinical outcomes in routine evaluation - outcome measure lems with functioning [22]. An index of health-related (CORE-OM) quality of life is generated to illustrate an individual’s The CORE-OM is widely used to assess the mental overall health status using a population tariff. This study health effects of psychological interventions [16]. It com- used the English population tariff which was developed prises 34 items across four main areas of focus: subject- based on the time trade-off and discrete choice experi- ive wellbeing, symptoms, functioning and risk. A five ment methods [23]. The health index score ranges from point rating scale has been adopted, with 0 representing − 0.281 to 1, with negative values representing health not at all, whilst 4 represents all the time [17]. A mean states worse than death, 0 representing the “dead” state, item score is commonly generated to allow an under- and 1 the “full health” state. The reliability and validity standing of the level of psychological distress of an indi- of the EQ-5D-5 L for use with the population of study vidual. The CORE-OM is a widely used measure that in this investigation has already been published [9]. generates clinically meaningful information [17], and it has been regarded as acceptable, reliable, valid [16]. ICEpop CAPability measure for adults (ICECAP-A) The ICECAP-A is a measure of capability wellbeing. It fo- Leeds dependence questionnaire (LDQ) cuses on ability to function across five key dimensions of The LDQ is used to identify an individual’s level of de- wellbeing. These are stability, enjoyment, achievement, at- pendence on a variety of substances [18]. It features 10 tachment and autonomy. Participants select their capabil- items that are rated on a scale from 0 to 3; 0 represent- ity level from four options ranging from 1 to 4, with 1 ing never, whilst 3 represents nearly always. The ques- representing that an individual has limited capability in a tions are centered around substance use and frequency; given dimension, whereas 4 represents high levels of cap- asking about desires, how substances fit into daily rou- ability [8]. An index of capability wellbeing is generated tines and any future plans of taking substances [19]. The that illustrates an individual’s overall capability levels from scores of all items are aggregated to indicate the overall 0 to 1 using a UK population tariff developed based on level of dependence. This can vary between 0 and 30 the best-worst scaling method [24]. A score of 0 suggests with the cut-offs of 10 and 22 used to classify individ- that an individual has no capability, whilst 1 represents full ual’s level of dependence into low, moderate, and severe. capability [8]. The reliability and validity of the ICECAP-A Raistrick et al. [18] describe a variety of features of LDQ for use with the population of study in this investigation that suggest it may be able to complement economic has also been published [9]. evaluation. The authors conclude that LDQ is under- standable and sensitive to change in the level of depend- Estimation and validation datasets ency over time and across all substance dependencies. A common approach across mapping studies is to split the dataset into an estimation dataset, where the mapping Treatment outcomes profile (TOP) algorithm between the source outcome measures (CORE- The TOP is used to assess the change and progress in OM, LDQ and TOP) and the target measures (EQ-5D-5 L key areas of life for individuals who are being treated for and ICECAP-A) is first derived, and a validation dataset, drug or alcohol addiction [20]. It features 20 questions, where the predictive properties of the algorithm are tested reflecting the four domains of substance use- injecting [25]. Two approaches were used to determine the estima- risk behaviour, crime and health and social care func- tion and validation dataset. In the first approach, the esti- tioning [21]. For this study, the health and social care mation sample was developed from the baseline data and functioning aspect, which included psychological health, the validation sample from the 3 month follow-up data. In physical health and the overall quality of life dimensions the second approach, the data from the baseline assess- were of importance. These three dimensions are rated ment and 3 month follow up were pooled in order to cre- on a scale of 0 (which represents poor) to 20 (which rep- ate a larger estimation sample, and the 12 month follow resents good) [20]. up data were then used for validation purposes, similar to other studies in the literature [26–28]. EuroQol – 5 dimensions – 5 levels (EQ-5D-5 L) The EQ-5D-5 L is an instrument that is used to measure Statistical analysis health-related quality of life. It is a self-reported ques- Longworth and Rowen [25] highlight a variety of regres- tionnaire that covers five dimensions of health; mobility, sion methods that are often used in mapping studies. Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 4 of 11 The type of method employed depends on whether the predictive ability and not fit should be considered, and prediction goal is for the overall index score or the di- therefore an internal and external validation of the models’ mension scores of a preference-based measure. Given predictive ability was undertaken. Internal validation that mapping onto an overall index score rather than di- involved the prediction of the EQ-5D-5 L and ICECAP-A mension scores has been found to offer better predictive index scores from each model’s outputs and evaluated how ability [29], this approach was adopted for the purposes close the predicted results were to the observed ones in the of this study. estimation dataset using the root-mean-squared-error In the first mapping approach, ordinary least squares (RMSE) and the mean absolute error (MAE) [25]. For ex- (OLS) and tobit regressions were used. The OLS regres- ternal validation, models’ coefficients were applied to the sion approach is commonly used in mapping studies scores of CORE-OM, LDQ, and TOP in the validation [25] and has been regarded as robust at predicting the dataset and the results were plotted on a graph in order to mean index score of a preference-based measure [30]. see how close the predicted EQ-5D-5 L and ICECAP-A The tobit regression is a censored regression method index scores were to the actual index scores using the providing opportunity to limit predictions within the ap- RMSE and MAE. All analyses were undertaken in Stata ver- propriate range of scores for EQ-5D-5 L (− 0.281 to 1) sion 13MP. and ICECAP-A (0 to 1). Given that the OLS regression is likely to provide predictions beyond these ranges, Results these predictions were subsequently forced to the appro- Descriptive statistics priate threshold value. This is. Table 2 presents the demographic details for the 83 trial a common approach in mapping studies [31–34]. In the participants. The mean age was 37 years and the sample second mapping approach, a cluster regression and a multi- comprised mostly men (87%). The majority were of white level mixed effects regression at an individual level were ethnicity (84%) and unemployed (79%). Nearly 82% of the used to account for within-subject dependence [35, 36]. sample received some form of state benefits. At A number of potential explanatory variables were avail- baseline, mean capability (ICECAP-A) index score was able and these were explored incrementally and in line 0.662 (SD = 0.189) and mean health (EQ-5D-5 L) index with the recommended methods guidance provided by score was 0.806 (SD = 0.204). The summary statistics Longworth and Rowen [25]. These included overall scores, for the two preference-based outcome measures and dimension scores, quadratic terms for potentially nonlin- for both the estimation and validation samples across ear relationships, interaction terms, and patient character- the different follow-up periods are shown in Additional istics (i.e. age and gender). All model specifications used file 1: Table S1. to map from CORE-OM, LDQ and TOP onto the EQ-. 5D-5 L and ICECAP-A are shown in Table 1. Mapping CORE-OM onto EQ-5D-5 L and ICECAP-A To test whether the algorithms were fit for purpose, the The performance of the different models in the internal 2 2 R , the adjusted R , the Akaike information criterion (AIC) (estimation) and external validation samples is provided and the Bayesian information criterion (BIC) were assessed. in Additional file 2: Table S2 and Additional file 3: Table For the final model choice, Brazier et al. [14]argue that S3, respectively. The results showed that most models Table 1 Summary of the model specifications used when mapping from CORE-OM, LDQ, and TOP onto EQ-5D-5 L and ICECAP-A Model CORE-OM LDQ TOP 1 Mean score Aggregate score Overall quality of life score 2 Mean score; Aggregate score; Overall quality of life score; 2 2 Mean score Aggregate score Physical and Psychological health status 3 Wellbeing; Symptoms; Best model from Model 2 plus quadratic terms Functioning; Risk above plus Age and Age 4 Model 3 plus Model 3 plus Gender Model 3 plus interaction terms quadratic terms 5 Model 4 plus Best model from above interaction terms plus Age and Age 6 Best model from Model 5 plus Gender above plus Age and Age 7 Model 6 plus Gender CORE-OM Clinical Outcomes in Routine Evaluation - Outcome Measure, LDQ Leeds Dependence Questionnaire, TOP Treatment Outcomes Profile, EQ-5D-5 L EuroQol – 5 Dimensions – 5 Levels, ICECAP-A ICEpop CAPability measure for Adults Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 5 of 11 Table 2 Patient Demographic Information Table 3 Mapping Models from the CORE-OM to the EQ-5D-5 L and the ICECAP-A Number in Study 83 Age, mean (SD) 37.1 (6.40) EQ-5D-5 L ICECAP-A Men, n (%) 72 (86.80) Model OLS (3) Tobit (2) c c White, n (%) 70 (84.30) Intercept 1.048 0.999 Employed, n (%) 17 (20.50) CORE-OM score −0.296 Married, n (%) 2 (2.40) CORE-OM score 0.041 Family Accommodation, n (%) 75 (90.40) Wellbeing 0.0005 Secondary Education or less, n (%) 56 (67.50) Symptoms −0.109 State Benefit Recipients, n (%) 68 (81.90) Functioning −0.010 n number of patients,SD standard deviation Risk −0.033 AIC −56.362 −69.824 BIC −44.451 −60.296 predicted the EQ-5D-5 L and ICECAP-A index scores 2 2 Adjusted R / Pseudo R 0.355 −1.197 well in both samples and for both estimation ap- RMSE (external sample) 0.134 0.138 proaches. For the EQ-5D-5 L most models provided predictions within a 0.03 range from the observed MAE (external sample) 0.100 0.106 health index score, and therefore not different in Statistically significant at the 1% level. AIC Akaike information criterion, BIC Bayesian information criterion, MAE mean absolute error, OLS ordinary least terms of clinical importance [37]. Exceptions were squares, RMSE root mean squared error mainly from the tobit regressions. For the ICECAP-A, most models provided predictions within a 0.01 range Using CORE-OM as an example, the EQ-5D-5 L utility from the observed capability index scores. In the in- score can be calculated from the following coefficients: ternal validation sample, model specification 5, which included the CORE-OM dimension scores plus quad- EQ−5D−5L utility score ¼ 1:048 þðÞ Wellbeing ratic and interaction terms, consistently produced the ðÞ 0:0005 −ðÞ Symptoms lowest RMSE and MAE scores across the different ðÞ 0:109 −ðÞ Functioning types of regression and the largest variability around ðÞ 0:010 −ðÞ RiskðÞ 0:033 the predicted mean EQ-5D-5 L and ICECAP-A index scores. Because of this variability, Model 5 resulted in large RMSE and MAE in the external validation sam- Mapping LDQ onto EQ-5D-5 L and ICECAP-A ple. In this sample, model specification 3, which in- Detailed information about the performance of the dif- cluded the four CORE-OM dimensions as covariates, ferent models used to predict health and capability index produced consistently the lowest RMSE and MAE re- scores in the internal and external validation samples is sults for the EQ-5D-5 L at the mean value as well as provided in the Additional file 5: Table S5 and at repeated measurements (i.e. 25th, 50th, and 75th Additional file 6: Table S6. Similar to the CORE-OM percentiles), whilst model specification 2, which in- measure, the results indicated that most models pre- cluded the mean CORE-OM score and its squared dicted the EQ-5D-5 L and ICECAP-A index scores term, showed the best performance across the differ- closely in both samples and for both estimation ap- ent models and estimation approaches (Additional file proaches. The only exception was from the tobit models, 4 Table S4). In terms of health index score, the OLS which gave predictions beyond the potentially acceptable 0. model 3 had the lowest RMSE (0.134) and MAE (0.1) 03 threshold difference from the observed EQ-5D-5 L and predicted the mean EQ-5D-5 L index score with index scores. For both internal and external samples, model a < 0.007 deviation from the observed score (0.83). In specification 4, which included the total LDQ score, age, terms of the capability index score, tobit model 2 was age squared and sex, was found to offer the best predictive found to have the best predictive properties with ability, with RMSE and MAE ranging between 0.172–0.216 RMSE and MAE scores of 0.138 and 0.106 respect- and 0.122–0.146 across the different analyses for the health ively. The coefficients for each model covariate and index score and between 0.163–0.194 and 0.133–0.154 for the model’s fit for the two mapping algorithms are the capability index score. OLS model 4 when mapping shown in Table 3.Figures 1 and 2 show graphs dis- to the EQ-5D-5 L had the better predictive ability. The playing the predicted scores in comparison to the ob- model produced a MAE score of 0.128 and RMSE score served scores for the chosen algorithms, when of 0.178. OLS model 4 also had better predictive ability mapping from the CORE-OM. The ICECAP-A is bet- when mapping to the ICECAP-A. This model produced ter predicted. aMAE scoreof 0.138 andaRMSE scoreof 0.171 Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 6 of 11 Fig. 1 The observed vs predicted scores of the EQ-5D-5 L mapped from the CORE-OM based on Model 3 (Additional file 7 Table S7). The coefficients for each ICECAP-A index scores well in both samples and for both model covariate based on the external validation sam- approaches. Tobit models gave again predictions of health ple and the model’s fit are detailed in Table 4.Figures 3 index scores that were more than 0.03 points different to and 4 show the difference between the EQ-5D-5 L and the observed EQ-5D-5 L scores but performed well in the ICECAP-A predictions. terms of predicting capability scores. Model specifications The EQ-5D-5 L scores were more dispersed and 4; which included TOP dimension scores, quadratic terms spread than the ICECAP-A scores. and interaction terms and model 6, which included the covariates in model 4 with the addition of age, age and Mapping from the TOP onto EQ-5D-5 L and ICECAP-A gender; appeared to have a better performance in the The performance of the different mapping algorithms internal sample resulting in RMSE and MAE that ranged from the TOP measure onto EQ-5D-5 L and ICECAP-A between 0.155–0.199 and 0.122–0.126 respectively but measures in both internal and external validation samples again with larger variability around the predicted mean is shown in the Additional file 8: Table S8 and Additional index scores. The model specification with the best file 9 Table S9. Most models predicted the EQ-5D-5 L and external predictive ability across the different models and Fig. 2 The observed vs predicted scores of the ICECAP-A mapped from the CORE-OM based on Model 2 Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 7 of 11 Table 4 Mapping Models from the LDQ to the EQ-5D-5 L and Discussion the ICECAP-A Policy decision makers are becoming more focused on EQ-5D-5 L ICECAP-A treatment outcomes that go beyond consideration of ab- stinence alone and capture wider treatment impact upon Model OLS (4) OLS (4) patients HRQOL [38]. These measures should include Intercept 0.415 0.958 economic measures that allow consideration of value for b b LDQ score −0.014 −0.0122 money. This study developed mapping algorithms from Age 0.033 −0.004 three key clinical measures in the context of opiate de- Age −0.0005 −0.000002 pendence (CORE-OM, LDQ, and TOP) onto the EQ- Sex (if Female) −0.016 −0.052 5D-5 L and ICECAP-A, which are recommended by the National Institute for Health and Care Excellence AIC −44.943 −53.753 (NICE) in the economic evaluation of health and social BIC −32.971 −41.781 care interventions [5]. These algorithms introduce the Adjusted R 0.250 0.214 possibility of estimating HRQOL and capability well- RMSE (external sample) 0.178 0.171 being from the information contained within each clin- MAE (external sample) 0.128 0.138 ical measure and therefore the ability to make treatment a b Statistically significant at the 5% level; Statistically significant at the 1% level. recommendations based on wider quality of life and AIC Akaike information criterion, BIC Bayesian information criterion, MAE mean wellbeing outcomes. As these instruments are focused absolute error, OLS ordinary least squares, RMSE root mean squared error exclusively upon health-related quality of life, and well- for both EQ-5D-5 L and ICECAP-A was model specifica- being effects, these algorithms provide the vehicle to tar- tion 2, which included the three TOP dimensions only get resources towards treatment that will benefit (Overall quality of life score, Physical health status, and population quality of life. The ICECAP-A was better pre- Psychological health status). Overall, the OLS model 2 dicted than the EQ-5D-5 L. This suggests that when map- predicted the observed EQ-5D-5 L (0.83) and ICECAP-A ping from the clinical measures to the ICECAP-A, there (0.69) index score with a − 0.01 point difference. For EQ- will be a greater alignment between the wellbeing aspects 5D-5 L and ICECAP-A, the RMSE scores were 0.167 and than when mapping to the EQ-5D-5 L. This presents an 0.151 respectively, while the MAE scores were 0.123 for interesting development for reimbursement decisions both mapping algorithms (Additional file 10: Table S10). within the context of opiate dependence and OST. The coefficients for each model covariate based on the ex- This is the first study to generate mapping algorithms ternal validation sample and the models’ fit are detailed in for these clinical measures. This was a particularly diffi- Table 5. Figures 5 and 6 display the graphs detailing the cult hard to reach population under study with many predicted and observed scores when mapping from the participants receiving OST at a therapeutic dose for at TOP. The EQ-5D-5 L results had a greater dispersion, least 5 years, and still reporting heroin use [11]. The and were plotted further away from the fitted value line. study benefited from good completion rates across the Fig. 3 The observed vs predicted scores of the EQ-5D-5 L mapped from the LDQ based on Model 4 Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 8 of 11 Fig. 4 The observed vs predicted scores of the ICECAP-A mapped from the LDQ based on Model 4 different measures, and although the primary trial fo- size was used. To overcome the sample size issue, obser- cused on a hard to reach population undergoing opiate vations were pooled from two time-periods but could substitution treatment but still reporting heroin use, the have led to assumptions of independence being violated distribution of health and capability scores provides between the observations and lead to ungeneralizable re- some confidence that these algorithms are likely to be sults. However, appropriate techniques were applied to generalizable to other contexts involving substance use account for within-subject dependence. disorders. The algorithms developed had good predictive Although the sample size was effectively doubled and ability and the errors identified fall within an acceptable the dependence accounted for, the pooled dataset did not range in comparison to other mapping studies [14]. produce better results relative to the algorithms created Demographic information, other than age and sex, was from the normal estimation dataset. Studies, however, not used during the mapping process. This was a delib- have been conducted with smaller samples [39, 40]. The erate choice appreciating that future use of the mapping errors reported in this study are also not significantly dif- algorithms will be maximized by the algorithms only re- ferent to ones reported within other studies [14]. At first quiring demographic information on age and sex [4]. glance, the sample may not appear representative of the The study had limitations. Given that the study relies general population but was generally representative of a on data collected within a pilot trial, a modest sample UK OST population with a high number of men, un- employed and people with mainly white ethnicity. The Table 5 Mapping Models from the TOP to the EQ-5D-5 L and EQ-5D-5 L and ICECAP-A had maximum scores of 1. the ICECAP-A Initial regression analysis identified results that led to EQ-5D-5 L ICECAP-A scores greater than one. This meant that the upper bound- Model OLS (2) OLS (2) ary had to be censored to 1. The main drawback with this b b approach is that the mean, RMSE and MAE scores are Intercept 0.463 0.348 a underestimated however there was no alternative solution Overall Quality of Life −0.005 0.014 as mean values above 1 are not possible. Physical Health Status 0.010 −0.002 Although, there are no published studies that map to b a Psychological Health Status 0.024 0.015 the measures presented, there are many with other clin- AIC −52.030 − 70.213 ical measures that have mapped to the EQ-5D, which BIC −42.402 −60.586 can offer a point of comparison. This study showed that applying an alternative regression specification such as Adjusted R 0.298 0.335 the tobit regression did not improve the results, the OLS RMSE (external sample) 0.167 0.151 models were demonstrators of goodness of fit. This con- MAE (external sample) 0.123 0.123 cords with other studies [41–44]. The predicted EQ-5D- a b Statistically significant at the 5% level; Statistically significant at the 1% level. 5 L scores generated during the mapping process were AIC Akaike information criterion, BIC Bayesian information criterion, MAE mean absolute error, OLS ordinary least squares, RMSE root mean squared error across a much smaller spread than the observed EQ-5D- Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 9 of 11 Fig. 5 The observed vs predicted scores of the EQ-5D-5 L mapped from the TOP based on Model 2 5 L scores. This has also been seen to be the case in a variety of things, drug control, physical and mental health various other studies [41, 42]. [45]. Mapping from a condition-specific measure to a trad- As the results were generated from a fairly small sam- itional generic preference-based measure could miss out ple size, it would be useful to validate the algorithms these key drivers within the recovery process. It is import- using a larger sample. It would also be important to con- ant to capture these impacts on an individual beyond the duct research into how the chosen algorithms would in- bounds of health and utilize tools such as the ICECAP-A fluence QALYs and cost-effectiveness decisions in the wellbeing measure. The use of the TOP clinical measure is realm of mental health. common practice particularly within the context of UK spe- cialist drug treatment. Having these algorithms available Conclusion provides the potential to estimate incremental QALYs and The application of the ICECAP-A could have the ability to wellbeing outcomes using routinely collected data, and thus capture mental health related quality of life outside the util- provides a framework for estimating the cost-effectiveness ity framework. In relation to mental health, recovery entails of alternative therapy options. Fig. 6 The observed vs predicted scores of the ICECAP-A mapped from the TOP based on Model 2 Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 10 of 11 Additional Files Funding The study was funded under the National Institute for Health Research (NIHR) under its Research for Patient Benefit (RfPB) Programme (Grant Additional file 1: Table S1. Descriptive statistics of the generic and Reference Number PB-PG-0610-22392). condition-specific measures in the estimation and validation datasets. Sta- tistics describing the features of the EQ- 5D-5 L, ICECAP-A, TOP, LDQ and Availability of data and materials the CORE-OM for the estimation and validation datasets. (DOCX 16 kb) The datasets used and/or analysed during the current study are available Additional file 2: Table S2. Model performance of the Internal from the corresponding author on reasonable request. Validation Sample Mapping from the CORE-OM to the EQ- 5D-5 L and the ICECAP-A. Results for each model when mapping from the CORE-OM Authors’ contributions to the EQ-5D and the ICECAP-A using the internal validation sample. JP led on the analysis of the data with support from IG and EF. ED is the (DOCX 17 kb) chief investigator for the project. ED, AC, EF and NF contributed to the design of the study. All authors drafted the manuscript and read and Additional file 3: Table S3. Model performance of the External approved the final manuscript. Validation Sample Mapping from the CORE-OM to the EQ- 5D-5 L and the ICECAP-A. Results for each model when mapping from the CORE-OM Ethics approval and consent to participate to the EQ-5D and the ICECAP-A using the external validation sample. The study received ethics approval from the National Research Ethics (DOCX 17 kb) Committee: The Black Country (REC number: 12/WM/0046; Approved 08/02/ Additional file 4: Table S4. Model performance of the best fitting 2012). Written, informed consent to participate in the study was obtained models mapping from the CORE-OM to the ICECAP-A and the EQ-5D-5 L from all participants. using the external validation sample. Results for the best fitting models, models 2 and 3, when mapping from the CORE-OM to the EQ-5D and Competing interests the ICECAP-A using the external validation sample. (DOCX 14 kb) The authors declare that they have no competing interests. Additional file 5: Table S5. Model performance of the Internal Validation Sample Mapping from the LDQ to the EQ- 5D-5 L and the ICECAP-A. Results Publisher’sNote for each model when mapping from the LDQ to the EQ-5D and the Springer Nature remains neutral with regard to jurisdictional claims in ICECAP-A using the internal validation sample. (DOCX 15 kb) published maps and institutional affiliations. Additional file 6: Table S6 Model performance of the External Validation Sample Mapping from the LDQ to the EQ- 5D-5 L and the ICECAP-A. Results Author details for each model when mapping from the LDQ to the EQ-5D and the Health Economics Unit, Institute of Applied Health Research, Public Health ICECAP-A using the external validation sample. (DOCX 15 kb) Building, University of Birmingham, B15 2TT, Birmingham, UK. Research and Additional file 7: Table S7 Model performance of the best fitting Innovation Department, Birmingham & Solihull Mental Health NHS models mapping from the LDQ to the ICECAP-A and the EQ-5D-5 L using Foundation Trust, Birmingham, UK. Addictions Department, Institute of the external validation sample. Results for the best fitting models, models Psychiatry, Psychology & Neuroscience, King’s College London, London, UK. 3 and 4, when mapping from the LDQ to the EQ-5D and the ICECAP-A School of Psychology, University of Birmingham, Birmingham, UK. using the external validation sample. (DOCX 14 kb) Department of Primary Care and Population Health, University College London, London, UK. Melbourne School of Population and Global Health, Additional file 8: Table S8. Model performance of the Internal Validation University of Melbourne, Melbourne, Australia. Sample Mapping from the TOP to the EQ- 5D-5 L and the ICECAP-A. Results for each model when mapping from the TOP to the EQ-5D and the Received: 22 March 2018 Accepted: 7 May 2018 ICECAP-A using the internal validation sample. (DOCX 14 kb) Additional file 9: Table S9. Model performance of the External Validation Sample Mapping from the TOP to the EQ- 5D-5 L and the ICECAP-A. Results References for each model when mapping from the TOP to the EQ-5D and the 1. Morris S, Devlin N, Parkin D. Economic analysis in health care. Chichester: ICECAP-A using the external validation sample. (DOCX 16 kb) Wiley; 2007. Additional file 10: Table S10. Model performance of the best fitting 2. Coast J. Maximisation in extra-welfarism: a critique of the current position in models mapping from the TOP to the ICECAP-A and the EQ-5D-5 L using health economics. Soc Sci Med. 2009;69(5):786–92. the external validation sample. Results for the best fitting models, models 3. Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. 1 and 2, when mapping from the LDQ to the EQ-5D and the ICECAP-A Methods for the economic evaluation of health care programmes. Oxford: using the external validation sample. (DOCX 14 kb) Oxford university press; 2005. 4. Chuang L, Whitehead SJ. Mapping for economic evaluation. Br Med Bull. 2012;101:1-15. doi: https://doi.org/10.1093/bmb/ldr049. 5. National Institute for Health and Care Excellence (NICE). Guide to the methods of technology appraisal. London: NICE; 2008. Abbreviations 6. Orford J. Power, powerlessness and addiction. Cambridge: Cambridge University AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion; CORE- Press; 2013. OM: Clinical Outcomes in Routine Evaluation - Outcome Measure; EQ-5D- 7. Sen A. Capability and well-being. The quality of life. Oxford: Oxford 5 L: EuroQol - 5 Dimensions- 5 Levels; HRQOL: Health-Related Quality Of Life; University Press; 1993. ICECAP-A: ICEpop CAPability measure for Adults; LDQ: Leeds Dependence 8. Al-Janabi H, Flynn TN, Coast J. Development of a self-report measure of Questionnaire; MAE: Mean Absolute Error.; n: Number of patients; capability wellbeing for adults: the ICECAP-A. Qual Life Res. 2012;21(1):167–76. NICE: National Institute for Health and Care Excellence; OLS: Ordinary Least 9. Goranitis I, Coast J, Day E, Copello A, Freemantle N, Seddon J, et al. Squares; OST: Opiate Substitution Treatment; QALY: Quality-Adjusted Life Measuring health and broader well-being benefits in the context of opiate Years; RCT: Randomised Control Trial; RMSE: Root-Mean-Squared-Error; dependence: the psychometric performance of the ICECAP-A and the EQ- SD: Standard Deviation; TOP: Treatment Outcomes Profile 5D-5L. Value Health. 2016;19(6):820–8. 10. Day E, Copello A, Seddon JL, Christie M, Bamber D, Powell C, et al. Pilot study of a social network intervention for heroin users in opiate substitution treatment: Acknowledgments study protocol for a randomized controlled trial. Trials. 2013;14(1):264. We would like to thank everyone who participated in the study and to the 11. Day E, Copello A, Seddon JL, Christie M, Bamber D, Powell C, Bennett C, Akhtar community drug and alcohol teams located at Leicester (Leicestershire S, George S, Ball A, Frew E, Goranitis I, Freemantle N. A pilot feasibility Partnership NHS Trust) and Solihull (Birmingham & Solihull Mental Health randomised controlled trial of an adjunct brief social network intervention in NHS Foundation Trust). opiate substitution treatment services. BMC Psychiatry. 2018;18(1):8. Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 11 of 11 12. Baio G. Bayesian methods in health economics. Chapman and Hall/CRC 36. Khan KA, Madan J, Petrou S, Lamb SE. Mapping between the Roland Morris Press. Boca Raton, FL, US. 2012. questionnaire and generic preference-based measures. Value Health. 2014; 13. Petrou S, Rivero-Arias O, Dakin H, Longworth L, Oppe M, Froud R, et al. Preferred 17(6):686–95. reporting items for studies mapping onto preference-based outcome measures: 37. Marra CA, Woolcott JC, Kopec JA, Shojania K, Offer R, Brazier JE, et al. A the MAPS statement. Health Qual Life Outcomes. 2015b;13(1):106. comparison of generic, indirect utility measures (the HUI2, HUI3, SF-6D, and the EQ-5D) and disease-specific instruments (the RAQoL and the HAQ) in 14. Brazier JE, Yang Y, Tsuchiya A, Rowen DL. A review of studies mapping (or rheumatoid arthritis. Soc Sci Med. 2005;60(7):1571–82. cross walking) non-preference based measures of health to generic 38. Bray JW, Aden B, Eggman AA, Hellerstein L, Wittenberg E, Nosyk B, et al. preference-based measures. Eur J Health Econ. 2010;11(2):215–25. Quality of life as an outcome of opioid use disorder treatment: a systematic 15. Petrou S, Rivero-Arias O, Dakin H, Longworth L, Oppe M, Froud R, et al. The review. J Subst Abus Treat. 2017;76:88–93. MAPS reporting statement for studies mapping onto generic preference- 39. Young MK, Ng S, Mellick G, Scuffham PA. Mapping of the PDQ-39 to EQ-5D based outcome measures: explanation and elaboration. scores in patients with Parkinson’s disease. Qual Life Res. 2013;22(5):1065–72. PharmacoEconomics. 2015a;33(10):993–1011. 40. Kontodimopoulos N, Aletras VH, Paliouras D, Niakas D. Mapping the cancer- 16. Evans C, Connell J, Barkham M, Margison F, McGrath G, Mellor-Clark J, et al. specific EORTC QLQ-C30 to the preference-based EQ-5D, SF-6D, and 15D Towards a standardised brief outcome measure: psychometric properties instruments. Value Health. 2009;12(8):1151–7. and utility of the CORE-OM. Br J Psychiatry. 2002 Jan;180:51–60. 41. Cheung YB, Luo N, Ng R, Lee CF. Mapping the functional assessment of 17. Barkham M, Mellor-Clark J, Connell J, Cahill J. A CORE approach to practice- Cancer therapy-breast (FACT-B) to the 5-level EuroQoL Group’s 5-dimension based evidence: a brief history of the origins and applications of the CORE- questionnaire (EQ-5D-5L) utility index in a multi-ethnic Asian population. OM and CORE system. Couns Psychother Res. 2006;6(1):3–15. Health Qual Life Outcomes. 2014;12(1):180. 18. Raistrick D, Bradshaw J, Tober G, Weiner J, Allison J, Healey C. Development 42. Brazier J, Connell J, Papaioannou D, Mukuria C, Mulhern B, Peasgood T, et al. of the Leeds dependence questionnaire (LDQ): a questionnaire to measure A systematic review, psychometric analysis and qualitative assessment of alcohol and opiate dependence in the context of a treatment evaluation generic preference-based measures of health in mental health populations package. Addiction. 1994;89(5):563–72. and the estimation of mapping functions from widely used specific 19. Raistrick R, Bradshaw J, Tober G, Weiner J, Allison J, Healey C. Development measures. Health Technol Assess. 2014;18(34):vii,viii, xiii-xxv):1–188. of the Leeds dependence questionnaire (LDQ): a questionnaire to measure 43. Hawton A, Green C, Telford C, Wright D, Zajicek J. The use of multiple alcohol and opiate dependence in the context of a treatment evaluation sclerosis condition-specific measures to inform health policy decision- package. Addiction. 1994;89(5):563-572. making: mapping from the MSWS-12 to the EQ-5D. Mult Scler J. 2012;18(6): 20. Marsden J, Farrell M, Bradbury C, Dale-Perera A, Eastwood B, Roxburgh M, et 853–61. al. Development of the treatment outcomes profile. Addiction. 2008;103(9): 44. Kim H, Kim D, Jang EJ, Lee M, Song HJ, Park S, et al. Mapping health 1450–60. assessment questionnaire disability index (HAQ-DI) score, pain visual analog 21. National Treatment Agency for Substance Misuse. Treatment Outcomes scale (VAS), and disease activity score in 28 joints (DAS28) onto the EuroQol-5D Profile (TOP) [Internet]; 2017. Available from: http://www.nta.nhs.uk/ (EQ-5D) utility score with the KORean observational study network for arthritis healthcare-TOP.aspx. (KORONA) registry data. Rheumatol Int. 2016;36(4):505–13. 22. Herdman M, Gudex C, Lloyd A, Janssen M, Kind P, Parkin D, et al. 45. Goranitis I, Coast J, Day E, Copello A, Freemantle N, Frew E. Maximizing Development and preliminary testing of the new five-level version of EQ-5D health or sufficient capability in economic evaluation? A methodological (EQ-5D-5L). Qual Life Res. 2011;20(10):1727–36. experiment of treatment for drug addiction. Med Decis Mak. 2017;37(5): 23. Devlin N, Shah K, Feng Y, Mulhern B, van Hout B. Valuing health-related 498–511. quality of life: an EQ-5D-5L value set for England.. 2016. 24. Flynn TN, Huynh E, Peters TJ, Al-Janabi H, Clemens S, Moody A, et al. Scoring the Icecap-a capability instrument. Estimation of a UK general population tariff. Health Econ. 2015;24(3):258–69. 25. Longworth L, Rowen D. Mapping to obtain EQ-5D utility values for use in NICE health technology assessments. Value Health. 2013;16(1):202–10. 26. Crott R, Briggs A. Mapping the QLQ-C30 quality of life cancer questionnaire to EQ-5D patient preferences. Eur J Health Econ. 2010;11(4):427–34. 27. McKenzie L, Van Der Pol M. Mapping the EORTC QLQ C-30 onto the EQ-5D instrument: the potential to estimate QALYs without generic preference data. Value Health. 2009;12(1):167–71. 28. Pinedo-Villanueva RA, Turner D, Judge A, Raftery JP, Arden NK. Mapping the Oxford hip score onto the EQ-5D utility index. Qual Life Res. 2013;22(3):665–75. 29. Gray AM, Clarke PM, Wolstenholme JL, Wordsworth S. Applied methods of cost-effectiveness analysis in healthcare. Oxford: OUP; 2010. 30. Mitchell PM, Roberts TE, Barton PM, Pollard BS, Coast J. Predicting the ICECAP-O capability index from the WOMAC osteoarthritis index: is mapping onto capability from condition-specific health status questionnaires feasible? Med Decis Mak. 2013;33(4):547–57. 31. Dakin H, Gray A, Murray D. Mapping analyses to estimate EQ-5D utilities and responses based on Oxford knee score. Qual Life Res. 2013;22(3):683–94. 32. Skaltsa K, Longworth L, Ivanescu C, Phung D, Holmstrom S. Mapping the FACT-P to the preference-based EQ-5D questionnaire in metastatic castration-resistant prostate Cancer. Value Health. 2014;17(2):238–44. 33. Sullivan PW, Ghushchyan V. Mapping the EQ-5D index from the SF-12: US general population preferences in a nationally representative sample. Med Decis Mak. 2006;26(4):401–9. 34. Young TA, Mukuria C, Rowen D, Brazier JE, Longworth L. Mapping functions in health-related quality of life: mapping from two cancer-specific health-related quality-of-life instruments to EQ-5D-3L. Med Decis Mak. 2015;35(7):912–26. 35. Rivero-Arias O, Ouellet M, Gray A, Wolstenholme J, Rothwell PM, Luengo- Fernandez R. Mapping the modified Rankin scale (mRS) measurement into the generic EuroQol (EQ-5D) health outcome. Med Decis Mak. 2010;30(3):341–54. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Health and Quality of Life Outcomes Springer Journals

Predicting health-related quality of life (EQ-5D-5L) and capability wellbeing (ICECAP-A) in the context of opiate dependence using routine clinical outcome measures: CORE-OM, LDQ and TOP

Free
11 pages

Loading next page...
 
/lp/springer_journal/predicting-health-related-quality-of-life-eq-5d-5l-and-capability-aTVmCMYgiq
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s).
Subject
Medicine & Public Health; Quality of Life Research; Quality of Life Research
eISSN
1477-7525
D.O.I.
10.1186/s12955-018-0926-7
Publisher site
See Article on Publisher Site

Abstract

Background: Economic evaluation normally requires information to be collected on outcome improvement using utility values. This is often not collected during the treatment of substance use disorders making cost-effectiveness evaluations of therapy difficult. One potential solution is the use of mapping to generate utility values from clinical measures. This study develops and evaluates mapping algorithms that could be used to predict the EuroQol-5D (EQ-5D-5 L) and the ICEpop CAPability measure for Adults (ICECAP-A) from the three commonly used clinical measures; the CORE-OM, the LDQ and the TOP measures. Methods: Models were estimated using pilot trial data of heroin users in opiate substitution treatment. In the trial the EQ-5D-5 L, ICECAP-A, CORE-OM, LDQ and TOP were administered at baseline, three and twelve month time intervals. Mapping was conducted using estimation and validation datasets. The normal estimation dataset, which comprised of baseline sample data, used ordinary least squares (OLS) and tobit regression methods. Data from the baseline and three month time periods were combined to create a pooled estimation dataset. Cluster and mixed regression methods were used to map from this dataset. Predictive accuracy of the models was assessed using the root mean square error (RMSE) and the mean absolute error (MAE). Algorithms were validated using sample data from the follow-up time periods. Results: Mapping algorithms can be used to predict the ICECAP-A and the EQ-5D-5 L in the context of opiate dependence. Although both measures can be predicted, the ICECAP-A was better predicted by the clinical measures. There were no advantages of pooling the data. There were 6 chosen mapping algorithms, which had MAE scores ranging from 0.100 to 0.138 and RMSE scores ranging from 0.134 to 0.178. Conclusion: It is possible to predict the scores of the ICECAP-A and the EQ-5D-5 L with the use of mapping. In the context of opiate dependence, these algorithms provide the possibility of generating utility values from clinical measures and thus enabling economic evaluation of alternative therapy options. Trial registration: ISRCTN22608399. Date of registration: 27/04/2012. Date of first randomisation: 14/08/2012. Keywords: ICECAP, EQ-5D, Mapping, Addiction, Mental health, Preference-based measures, Condition-specific measures, Economic evaluation * Correspondence: E.Frew@bham.ac.uk Health Economics Unit, Institute of Applied Health Research, Public Health Building, University of Birmingham, B15 2TT, Birmingham, UK Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 2 of 11 Background Leeds Dependence Questionnaire (LDQ), and the Treat- In many healthcare systems around the world, resources ment Outcomes Profile (TOP). The CORE-OM and the are scarce and the demand for healthcare outweighs sup- TOP are instruments that are used to assess the treatment ply. This scarcity warrants the need for economic evalu- outcome, whilst the LDQ is used to assess the level of de- ation to aid decision makers with information about the pendence at the time of assessment [10, 11]. These mea- most efficient use of resources in order to maximise the sures, however, are unsuitable for use within health health gained for every unit of currency spent. Within economic evaluations [12]. To enable information from the UK, and in many other country jurisdictions, the these measures to be used in economic evaluations, a most common approach to economic evaluation is the process called ‘mapping’ can be applied [13]. Mapping, cost-utility analysis. The outcomes of a cost-utility ana- quantifies the relationship between different measures lysis are expressed in quality-adjusted life years (QALYs) using appropriate statistical techniques [14, 15], and al- [1]. QALYs take into account both quality and length of lows for the estimation of HRQOL and wellbeing for use life and offer a commensurate unit that allows compari- in economic evaluations using data collected from routine sons of cost-effectiveness across different disease areas clinical measures. and interventions [2]. To measure QALYs, generic This study aims to map three clinical instruments that preference-based measures of health-related quality of are often used in the routine care of addiction and opi- life (HRQOL) are required [3] to capture a broad con- ate dependence (CORE-OM, LDQ, and TOP) onto the struct of health through key dimensions that are known EQ-5D-5 L and ICECAP-A measures, generating algo- to affect quality of life [4]. Commonly, the EuroQol-5D rithms that can be used in future studies to aid reim- (EQ-5D-5 L) measure is used in economic evaluations to bursement decisions in the absence of information estimate QALYs [5]. The EQ-5D-5 L describes HRQOL related to EQ-5D-5 L and ICECAP-A. With these map- through the dimensions of mobility, self-care, usual ac- ping algorithms, data from the three clinical measures tivities, pain and discomfort, and anxiety and depression. can be translated into health and capability scores for When considering substance use disorders, the use in an economic evaluation. The mapping algorithms broader concept of wellbeing is considered more appro- were developed using data from a pilot randomised con- priate to reflect the clinical and policy objectives [6]. trol trial (RCT) that sought to explore the effectiveness Drug dependence undermines an individual’s capability of two psychosocial interventions for heroin users re- [6], and this disempowerment is largely overlooked in ceiving opiate substitution treatment (OST) in England the health economics of addiction-related interventions [10]. This is the first study to develop mapping algo- as a result of the narrow definition of HRQOL. Until re- rithms from routine clinical outcome measures used in cently, it was hard to define and quantify wellbeing for addiction therapy onto the EQ-5D-5 L and ICECAP-A. the purposes of an economic evaluation. Amartya Sen’s work on capabilities allowed for a conceptualization of Methods wellbeing through human functionings (what an individ- The study uses data collected as part of a pilot random- ual ‘does’) and capabilities (the ability of the individual ized controlled trial designed to investigate the clinical to exercise a functioning) [7]. The development of the and cost-effectiveness of two psychological interventions ICEpop CAPability measure for Adults (ICECAP-A) delivered in addition to the usual care of individuals who based on Amartya Sen’s capability approach, means that had been receiving opiate substitution treatment for wellbeing can now be measured in a way that is compat- more than one year [10]. All trial participants met the ible for use in economic evaluations [8]. The ICECAP-A ICD-10 criteria for opioid dependence and were re- measures capability wellbeing through the dimensions of cruited if they were in opiate substitution treatment with stability, enjoyment, achievement, attachment and au- methadone or buprenorphine for more than a year but tonomy. Both EQ-5D-5 L and ICECAP-A were found to still reported heroin use during the last month. The only have the appropriate construct validity within the addic- exclusion criteria were having a physical or mental tion context, but ICECAP-A appeared to be significantly health condition that prevented engagement in the psy- more responsive to changes of key clinical indicators [9]. chosocial intervention, or an imminent period of impris- In order to reduce the burden of assessment on patients onment. A number of client outcomes, including mental and to help acquire data for clinical means, particularly in health (CORE-OM), substance dependence (LDQ), phys- the context of substance use disorder, studies tend to rely ical and psychological health (TOP), health-related qual- only on clinical context-specific measures. For substance ity of life (EQ-5D-5 L), and capability wellbeing use disorders, measures that are commonly used to assess (ICECAP-A) were assessed at baseline, 3 months and the level and impact of dependence and assess the effect- 12 months post-randomisation. These outcome mea- iveness of a treatment are the Clinical Outcomes in Rou- sures are described in detail below. The trial was con- tine Evaluation - Outcome Measure (CORE-OM), the ducted by three community drug teams in England. All Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 3 of 11 trial participants provided written informed consent, and self-care, usual activities, pain and discomfort, and anx- ethical approval was obtained from the Black Country NHS iety and depression. Participants select their functioning Research Ethics Committee (reference: 12/WM/0046). level from five options ranging from 1 to 5, with 1 repre- senting that an individual has no functioning problems Outcome measures in a given dimension, whereas 5 represents severe prob- Clinical outcomes in routine evaluation - outcome measure lems with functioning [22]. An index of health-related (CORE-OM) quality of life is generated to illustrate an individual’s The CORE-OM is widely used to assess the mental overall health status using a population tariff. This study health effects of psychological interventions [16]. It com- used the English population tariff which was developed prises 34 items across four main areas of focus: subject- based on the time trade-off and discrete choice experi- ive wellbeing, symptoms, functioning and risk. A five ment methods [23]. The health index score ranges from point rating scale has been adopted, with 0 representing − 0.281 to 1, with negative values representing health not at all, whilst 4 represents all the time [17]. A mean states worse than death, 0 representing the “dead” state, item score is commonly generated to allow an under- and 1 the “full health” state. The reliability and validity standing of the level of psychological distress of an indi- of the EQ-5D-5 L for use with the population of study vidual. The CORE-OM is a widely used measure that in this investigation has already been published [9]. generates clinically meaningful information [17], and it has been regarded as acceptable, reliable, valid [16]. ICEpop CAPability measure for adults (ICECAP-A) The ICECAP-A is a measure of capability wellbeing. It fo- Leeds dependence questionnaire (LDQ) cuses on ability to function across five key dimensions of The LDQ is used to identify an individual’s level of de- wellbeing. These are stability, enjoyment, achievement, at- pendence on a variety of substances [18]. It features 10 tachment and autonomy. Participants select their capabil- items that are rated on a scale from 0 to 3; 0 represent- ity level from four options ranging from 1 to 4, with 1 ing never, whilst 3 represents nearly always. The ques- representing that an individual has limited capability in a tions are centered around substance use and frequency; given dimension, whereas 4 represents high levels of cap- asking about desires, how substances fit into daily rou- ability [8]. An index of capability wellbeing is generated tines and any future plans of taking substances [19]. The that illustrates an individual’s overall capability levels from scores of all items are aggregated to indicate the overall 0 to 1 using a UK population tariff developed based on level of dependence. This can vary between 0 and 30 the best-worst scaling method [24]. A score of 0 suggests with the cut-offs of 10 and 22 used to classify individ- that an individual has no capability, whilst 1 represents full ual’s level of dependence into low, moderate, and severe. capability [8]. The reliability and validity of the ICECAP-A Raistrick et al. [18] describe a variety of features of LDQ for use with the population of study in this investigation that suggest it may be able to complement economic has also been published [9]. evaluation. The authors conclude that LDQ is under- standable and sensitive to change in the level of depend- Estimation and validation datasets ency over time and across all substance dependencies. A common approach across mapping studies is to split the dataset into an estimation dataset, where the mapping Treatment outcomes profile (TOP) algorithm between the source outcome measures (CORE- The TOP is used to assess the change and progress in OM, LDQ and TOP) and the target measures (EQ-5D-5 L key areas of life for individuals who are being treated for and ICECAP-A) is first derived, and a validation dataset, drug or alcohol addiction [20]. It features 20 questions, where the predictive properties of the algorithm are tested reflecting the four domains of substance use- injecting [25]. Two approaches were used to determine the estima- risk behaviour, crime and health and social care func- tion and validation dataset. In the first approach, the esti- tioning [21]. For this study, the health and social care mation sample was developed from the baseline data and functioning aspect, which included psychological health, the validation sample from the 3 month follow-up data. In physical health and the overall quality of life dimensions the second approach, the data from the baseline assess- were of importance. These three dimensions are rated ment and 3 month follow up were pooled in order to cre- on a scale of 0 (which represents poor) to 20 (which rep- ate a larger estimation sample, and the 12 month follow resents good) [20]. up data were then used for validation purposes, similar to other studies in the literature [26–28]. EuroQol – 5 dimensions – 5 levels (EQ-5D-5 L) The EQ-5D-5 L is an instrument that is used to measure Statistical analysis health-related quality of life. It is a self-reported ques- Longworth and Rowen [25] highlight a variety of regres- tionnaire that covers five dimensions of health; mobility, sion methods that are often used in mapping studies. Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 4 of 11 The type of method employed depends on whether the predictive ability and not fit should be considered, and prediction goal is for the overall index score or the di- therefore an internal and external validation of the models’ mension scores of a preference-based measure. Given predictive ability was undertaken. Internal validation that mapping onto an overall index score rather than di- involved the prediction of the EQ-5D-5 L and ICECAP-A mension scores has been found to offer better predictive index scores from each model’s outputs and evaluated how ability [29], this approach was adopted for the purposes close the predicted results were to the observed ones in the of this study. estimation dataset using the root-mean-squared-error In the first mapping approach, ordinary least squares (RMSE) and the mean absolute error (MAE) [25]. For ex- (OLS) and tobit regressions were used. The OLS regres- ternal validation, models’ coefficients were applied to the sion approach is commonly used in mapping studies scores of CORE-OM, LDQ, and TOP in the validation [25] and has been regarded as robust at predicting the dataset and the results were plotted on a graph in order to mean index score of a preference-based measure [30]. see how close the predicted EQ-5D-5 L and ICECAP-A The tobit regression is a censored regression method index scores were to the actual index scores using the providing opportunity to limit predictions within the ap- RMSE and MAE. All analyses were undertaken in Stata ver- propriate range of scores for EQ-5D-5 L (− 0.281 to 1) sion 13MP. and ICECAP-A (0 to 1). Given that the OLS regression is likely to provide predictions beyond these ranges, Results these predictions were subsequently forced to the appro- Descriptive statistics priate threshold value. This is. Table 2 presents the demographic details for the 83 trial a common approach in mapping studies [31–34]. In the participants. The mean age was 37 years and the sample second mapping approach, a cluster regression and a multi- comprised mostly men (87%). The majority were of white level mixed effects regression at an individual level were ethnicity (84%) and unemployed (79%). Nearly 82% of the used to account for within-subject dependence [35, 36]. sample received some form of state benefits. At A number of potential explanatory variables were avail- baseline, mean capability (ICECAP-A) index score was able and these were explored incrementally and in line 0.662 (SD = 0.189) and mean health (EQ-5D-5 L) index with the recommended methods guidance provided by score was 0.806 (SD = 0.204). The summary statistics Longworth and Rowen [25]. These included overall scores, for the two preference-based outcome measures and dimension scores, quadratic terms for potentially nonlin- for both the estimation and validation samples across ear relationships, interaction terms, and patient character- the different follow-up periods are shown in Additional istics (i.e. age and gender). All model specifications used file 1: Table S1. to map from CORE-OM, LDQ and TOP onto the EQ-. 5D-5 L and ICECAP-A are shown in Table 1. Mapping CORE-OM onto EQ-5D-5 L and ICECAP-A To test whether the algorithms were fit for purpose, the The performance of the different models in the internal 2 2 R , the adjusted R , the Akaike information criterion (AIC) (estimation) and external validation samples is provided and the Bayesian information criterion (BIC) were assessed. in Additional file 2: Table S2 and Additional file 3: Table For the final model choice, Brazier et al. [14]argue that S3, respectively. The results showed that most models Table 1 Summary of the model specifications used when mapping from CORE-OM, LDQ, and TOP onto EQ-5D-5 L and ICECAP-A Model CORE-OM LDQ TOP 1 Mean score Aggregate score Overall quality of life score 2 Mean score; Aggregate score; Overall quality of life score; 2 2 Mean score Aggregate score Physical and Psychological health status 3 Wellbeing; Symptoms; Best model from Model 2 plus quadratic terms Functioning; Risk above plus Age and Age 4 Model 3 plus Model 3 plus Gender Model 3 plus interaction terms quadratic terms 5 Model 4 plus Best model from above interaction terms plus Age and Age 6 Best model from Model 5 plus Gender above plus Age and Age 7 Model 6 plus Gender CORE-OM Clinical Outcomes in Routine Evaluation - Outcome Measure, LDQ Leeds Dependence Questionnaire, TOP Treatment Outcomes Profile, EQ-5D-5 L EuroQol – 5 Dimensions – 5 Levels, ICECAP-A ICEpop CAPability measure for Adults Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 5 of 11 Table 2 Patient Demographic Information Table 3 Mapping Models from the CORE-OM to the EQ-5D-5 L and the ICECAP-A Number in Study 83 Age, mean (SD) 37.1 (6.40) EQ-5D-5 L ICECAP-A Men, n (%) 72 (86.80) Model OLS (3) Tobit (2) c c White, n (%) 70 (84.30) Intercept 1.048 0.999 Employed, n (%) 17 (20.50) CORE-OM score −0.296 Married, n (%) 2 (2.40) CORE-OM score 0.041 Family Accommodation, n (%) 75 (90.40) Wellbeing 0.0005 Secondary Education or less, n (%) 56 (67.50) Symptoms −0.109 State Benefit Recipients, n (%) 68 (81.90) Functioning −0.010 n number of patients,SD standard deviation Risk −0.033 AIC −56.362 −69.824 BIC −44.451 −60.296 predicted the EQ-5D-5 L and ICECAP-A index scores 2 2 Adjusted R / Pseudo R 0.355 −1.197 well in both samples and for both estimation ap- RMSE (external sample) 0.134 0.138 proaches. For the EQ-5D-5 L most models provided predictions within a 0.03 range from the observed MAE (external sample) 0.100 0.106 health index score, and therefore not different in Statistically significant at the 1% level. AIC Akaike information criterion, BIC Bayesian information criterion, MAE mean absolute error, OLS ordinary least terms of clinical importance [37]. Exceptions were squares, RMSE root mean squared error mainly from the tobit regressions. For the ICECAP-A, most models provided predictions within a 0.01 range Using CORE-OM as an example, the EQ-5D-5 L utility from the observed capability index scores. In the in- score can be calculated from the following coefficients: ternal validation sample, model specification 5, which included the CORE-OM dimension scores plus quad- EQ−5D−5L utility score ¼ 1:048 þðÞ Wellbeing ratic and interaction terms, consistently produced the ðÞ 0:0005 −ðÞ Symptoms lowest RMSE and MAE scores across the different ðÞ 0:109 −ðÞ Functioning types of regression and the largest variability around ðÞ 0:010 −ðÞ RiskðÞ 0:033 the predicted mean EQ-5D-5 L and ICECAP-A index scores. Because of this variability, Model 5 resulted in large RMSE and MAE in the external validation sam- Mapping LDQ onto EQ-5D-5 L and ICECAP-A ple. In this sample, model specification 3, which in- Detailed information about the performance of the dif- cluded the four CORE-OM dimensions as covariates, ferent models used to predict health and capability index produced consistently the lowest RMSE and MAE re- scores in the internal and external validation samples is sults for the EQ-5D-5 L at the mean value as well as provided in the Additional file 5: Table S5 and at repeated measurements (i.e. 25th, 50th, and 75th Additional file 6: Table S6. Similar to the CORE-OM percentiles), whilst model specification 2, which in- measure, the results indicated that most models pre- cluded the mean CORE-OM score and its squared dicted the EQ-5D-5 L and ICECAP-A index scores term, showed the best performance across the differ- closely in both samples and for both estimation ap- ent models and estimation approaches (Additional file proaches. The only exception was from the tobit models, 4 Table S4). In terms of health index score, the OLS which gave predictions beyond the potentially acceptable 0. model 3 had the lowest RMSE (0.134) and MAE (0.1) 03 threshold difference from the observed EQ-5D-5 L and predicted the mean EQ-5D-5 L index score with index scores. For both internal and external samples, model a < 0.007 deviation from the observed score (0.83). In specification 4, which included the total LDQ score, age, terms of the capability index score, tobit model 2 was age squared and sex, was found to offer the best predictive found to have the best predictive properties with ability, with RMSE and MAE ranging between 0.172–0.216 RMSE and MAE scores of 0.138 and 0.106 respect- and 0.122–0.146 across the different analyses for the health ively. The coefficients for each model covariate and index score and between 0.163–0.194 and 0.133–0.154 for the model’s fit for the two mapping algorithms are the capability index score. OLS model 4 when mapping shown in Table 3.Figures 1 and 2 show graphs dis- to the EQ-5D-5 L had the better predictive ability. The playing the predicted scores in comparison to the ob- model produced a MAE score of 0.128 and RMSE score served scores for the chosen algorithms, when of 0.178. OLS model 4 also had better predictive ability mapping from the CORE-OM. The ICECAP-A is bet- when mapping to the ICECAP-A. This model produced ter predicted. aMAE scoreof 0.138 andaRMSE scoreof 0.171 Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 6 of 11 Fig. 1 The observed vs predicted scores of the EQ-5D-5 L mapped from the CORE-OM based on Model 3 (Additional file 7 Table S7). The coefficients for each ICECAP-A index scores well in both samples and for both model covariate based on the external validation sam- approaches. Tobit models gave again predictions of health ple and the model’s fit are detailed in Table 4.Figures 3 index scores that were more than 0.03 points different to and 4 show the difference between the EQ-5D-5 L and the observed EQ-5D-5 L scores but performed well in the ICECAP-A predictions. terms of predicting capability scores. Model specifications The EQ-5D-5 L scores were more dispersed and 4; which included TOP dimension scores, quadratic terms spread than the ICECAP-A scores. and interaction terms and model 6, which included the covariates in model 4 with the addition of age, age and Mapping from the TOP onto EQ-5D-5 L and ICECAP-A gender; appeared to have a better performance in the The performance of the different mapping algorithms internal sample resulting in RMSE and MAE that ranged from the TOP measure onto EQ-5D-5 L and ICECAP-A between 0.155–0.199 and 0.122–0.126 respectively but measures in both internal and external validation samples again with larger variability around the predicted mean is shown in the Additional file 8: Table S8 and Additional index scores. The model specification with the best file 9 Table S9. Most models predicted the EQ-5D-5 L and external predictive ability across the different models and Fig. 2 The observed vs predicted scores of the ICECAP-A mapped from the CORE-OM based on Model 2 Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 7 of 11 Table 4 Mapping Models from the LDQ to the EQ-5D-5 L and Discussion the ICECAP-A Policy decision makers are becoming more focused on EQ-5D-5 L ICECAP-A treatment outcomes that go beyond consideration of ab- stinence alone and capture wider treatment impact upon Model OLS (4) OLS (4) patients HRQOL [38]. These measures should include Intercept 0.415 0.958 economic measures that allow consideration of value for b b LDQ score −0.014 −0.0122 money. This study developed mapping algorithms from Age 0.033 −0.004 three key clinical measures in the context of opiate de- Age −0.0005 −0.000002 pendence (CORE-OM, LDQ, and TOP) onto the EQ- Sex (if Female) −0.016 −0.052 5D-5 L and ICECAP-A, which are recommended by the National Institute for Health and Care Excellence AIC −44.943 −53.753 (NICE) in the economic evaluation of health and social BIC −32.971 −41.781 care interventions [5]. These algorithms introduce the Adjusted R 0.250 0.214 possibility of estimating HRQOL and capability well- RMSE (external sample) 0.178 0.171 being from the information contained within each clin- MAE (external sample) 0.128 0.138 ical measure and therefore the ability to make treatment a b Statistically significant at the 5% level; Statistically significant at the 1% level. recommendations based on wider quality of life and AIC Akaike information criterion, BIC Bayesian information criterion, MAE mean wellbeing outcomes. As these instruments are focused absolute error, OLS ordinary least squares, RMSE root mean squared error exclusively upon health-related quality of life, and well- for both EQ-5D-5 L and ICECAP-A was model specifica- being effects, these algorithms provide the vehicle to tar- tion 2, which included the three TOP dimensions only get resources towards treatment that will benefit (Overall quality of life score, Physical health status, and population quality of life. The ICECAP-A was better pre- Psychological health status). Overall, the OLS model 2 dicted than the EQ-5D-5 L. This suggests that when map- predicted the observed EQ-5D-5 L (0.83) and ICECAP-A ping from the clinical measures to the ICECAP-A, there (0.69) index score with a − 0.01 point difference. For EQ- will be a greater alignment between the wellbeing aspects 5D-5 L and ICECAP-A, the RMSE scores were 0.167 and than when mapping to the EQ-5D-5 L. This presents an 0.151 respectively, while the MAE scores were 0.123 for interesting development for reimbursement decisions both mapping algorithms (Additional file 10: Table S10). within the context of opiate dependence and OST. The coefficients for each model covariate based on the ex- This is the first study to generate mapping algorithms ternal validation sample and the models’ fit are detailed in for these clinical measures. This was a particularly diffi- Table 5. Figures 5 and 6 display the graphs detailing the cult hard to reach population under study with many predicted and observed scores when mapping from the participants receiving OST at a therapeutic dose for at TOP. The EQ-5D-5 L results had a greater dispersion, least 5 years, and still reporting heroin use [11]. The and were plotted further away from the fitted value line. study benefited from good completion rates across the Fig. 3 The observed vs predicted scores of the EQ-5D-5 L mapped from the LDQ based on Model 4 Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 8 of 11 Fig. 4 The observed vs predicted scores of the ICECAP-A mapped from the LDQ based on Model 4 different measures, and although the primary trial fo- size was used. To overcome the sample size issue, obser- cused on a hard to reach population undergoing opiate vations were pooled from two time-periods but could substitution treatment but still reporting heroin use, the have led to assumptions of independence being violated distribution of health and capability scores provides between the observations and lead to ungeneralizable re- some confidence that these algorithms are likely to be sults. However, appropriate techniques were applied to generalizable to other contexts involving substance use account for within-subject dependence. disorders. The algorithms developed had good predictive Although the sample size was effectively doubled and ability and the errors identified fall within an acceptable the dependence accounted for, the pooled dataset did not range in comparison to other mapping studies [14]. produce better results relative to the algorithms created Demographic information, other than age and sex, was from the normal estimation dataset. Studies, however, not used during the mapping process. This was a delib- have been conducted with smaller samples [39, 40]. The erate choice appreciating that future use of the mapping errors reported in this study are also not significantly dif- algorithms will be maximized by the algorithms only re- ferent to ones reported within other studies [14]. At first quiring demographic information on age and sex [4]. glance, the sample may not appear representative of the The study had limitations. Given that the study relies general population but was generally representative of a on data collected within a pilot trial, a modest sample UK OST population with a high number of men, un- employed and people with mainly white ethnicity. The Table 5 Mapping Models from the TOP to the EQ-5D-5 L and EQ-5D-5 L and ICECAP-A had maximum scores of 1. the ICECAP-A Initial regression analysis identified results that led to EQ-5D-5 L ICECAP-A scores greater than one. This meant that the upper bound- Model OLS (2) OLS (2) ary had to be censored to 1. The main drawback with this b b approach is that the mean, RMSE and MAE scores are Intercept 0.463 0.348 a underestimated however there was no alternative solution Overall Quality of Life −0.005 0.014 as mean values above 1 are not possible. Physical Health Status 0.010 −0.002 Although, there are no published studies that map to b a Psychological Health Status 0.024 0.015 the measures presented, there are many with other clin- AIC −52.030 − 70.213 ical measures that have mapped to the EQ-5D, which BIC −42.402 −60.586 can offer a point of comparison. This study showed that applying an alternative regression specification such as Adjusted R 0.298 0.335 the tobit regression did not improve the results, the OLS RMSE (external sample) 0.167 0.151 models were demonstrators of goodness of fit. This con- MAE (external sample) 0.123 0.123 cords with other studies [41–44]. The predicted EQ-5D- a b Statistically significant at the 5% level; Statistically significant at the 1% level. 5 L scores generated during the mapping process were AIC Akaike information criterion, BIC Bayesian information criterion, MAE mean absolute error, OLS ordinary least squares, RMSE root mean squared error across a much smaller spread than the observed EQ-5D- Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 9 of 11 Fig. 5 The observed vs predicted scores of the EQ-5D-5 L mapped from the TOP based on Model 2 5 L scores. This has also been seen to be the case in a variety of things, drug control, physical and mental health various other studies [41, 42]. [45]. Mapping from a condition-specific measure to a trad- As the results were generated from a fairly small sam- itional generic preference-based measure could miss out ple size, it would be useful to validate the algorithms these key drivers within the recovery process. It is import- using a larger sample. It would also be important to con- ant to capture these impacts on an individual beyond the duct research into how the chosen algorithms would in- bounds of health and utilize tools such as the ICECAP-A fluence QALYs and cost-effectiveness decisions in the wellbeing measure. The use of the TOP clinical measure is realm of mental health. common practice particularly within the context of UK spe- cialist drug treatment. Having these algorithms available Conclusion provides the potential to estimate incremental QALYs and The application of the ICECAP-A could have the ability to wellbeing outcomes using routinely collected data, and thus capture mental health related quality of life outside the util- provides a framework for estimating the cost-effectiveness ity framework. In relation to mental health, recovery entails of alternative therapy options. Fig. 6 The observed vs predicted scores of the ICECAP-A mapped from the TOP based on Model 2 Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 10 of 11 Additional Files Funding The study was funded under the National Institute for Health Research (NIHR) under its Research for Patient Benefit (RfPB) Programme (Grant Additional file 1: Table S1. Descriptive statistics of the generic and Reference Number PB-PG-0610-22392). condition-specific measures in the estimation and validation datasets. Sta- tistics describing the features of the EQ- 5D-5 L, ICECAP-A, TOP, LDQ and Availability of data and materials the CORE-OM for the estimation and validation datasets. (DOCX 16 kb) The datasets used and/or analysed during the current study are available Additional file 2: Table S2. Model performance of the Internal from the corresponding author on reasonable request. Validation Sample Mapping from the CORE-OM to the EQ- 5D-5 L and the ICECAP-A. Results for each model when mapping from the CORE-OM Authors’ contributions to the EQ-5D and the ICECAP-A using the internal validation sample. JP led on the analysis of the data with support from IG and EF. ED is the (DOCX 17 kb) chief investigator for the project. ED, AC, EF and NF contributed to the design of the study. All authors drafted the manuscript and read and Additional file 3: Table S3. Model performance of the External approved the final manuscript. Validation Sample Mapping from the CORE-OM to the EQ- 5D-5 L and the ICECAP-A. Results for each model when mapping from the CORE-OM Ethics approval and consent to participate to the EQ-5D and the ICECAP-A using the external validation sample. The study received ethics approval from the National Research Ethics (DOCX 17 kb) Committee: The Black Country (REC number: 12/WM/0046; Approved 08/02/ Additional file 4: Table S4. Model performance of the best fitting 2012). Written, informed consent to participate in the study was obtained models mapping from the CORE-OM to the ICECAP-A and the EQ-5D-5 L from all participants. using the external validation sample. Results for the best fitting models, models 2 and 3, when mapping from the CORE-OM to the EQ-5D and Competing interests the ICECAP-A using the external validation sample. (DOCX 14 kb) The authors declare that they have no competing interests. Additional file 5: Table S5. Model performance of the Internal Validation Sample Mapping from the LDQ to the EQ- 5D-5 L and the ICECAP-A. Results Publisher’sNote for each model when mapping from the LDQ to the EQ-5D and the Springer Nature remains neutral with regard to jurisdictional claims in ICECAP-A using the internal validation sample. (DOCX 15 kb) published maps and institutional affiliations. Additional file 6: Table S6 Model performance of the External Validation Sample Mapping from the LDQ to the EQ- 5D-5 L and the ICECAP-A. Results Author details for each model when mapping from the LDQ to the EQ-5D and the Health Economics Unit, Institute of Applied Health Research, Public Health ICECAP-A using the external validation sample. (DOCX 15 kb) Building, University of Birmingham, B15 2TT, Birmingham, UK. Research and Additional file 7: Table S7 Model performance of the best fitting Innovation Department, Birmingham & Solihull Mental Health NHS models mapping from the LDQ to the ICECAP-A and the EQ-5D-5 L using Foundation Trust, Birmingham, UK. Addictions Department, Institute of the external validation sample. Results for the best fitting models, models Psychiatry, Psychology & Neuroscience, King’s College London, London, UK. 3 and 4, when mapping from the LDQ to the EQ-5D and the ICECAP-A School of Psychology, University of Birmingham, Birmingham, UK. using the external validation sample. (DOCX 14 kb) Department of Primary Care and Population Health, University College London, London, UK. Melbourne School of Population and Global Health, Additional file 8: Table S8. Model performance of the Internal Validation University of Melbourne, Melbourne, Australia. Sample Mapping from the TOP to the EQ- 5D-5 L and the ICECAP-A. Results for each model when mapping from the TOP to the EQ-5D and the Received: 22 March 2018 Accepted: 7 May 2018 ICECAP-A using the internal validation sample. (DOCX 14 kb) Additional file 9: Table S9. Model performance of the External Validation Sample Mapping from the TOP to the EQ- 5D-5 L and the ICECAP-A. Results References for each model when mapping from the TOP to the EQ-5D and the 1. Morris S, Devlin N, Parkin D. Economic analysis in health care. Chichester: ICECAP-A using the external validation sample. (DOCX 16 kb) Wiley; 2007. Additional file 10: Table S10. Model performance of the best fitting 2. Coast J. Maximisation in extra-welfarism: a critique of the current position in models mapping from the TOP to the ICECAP-A and the EQ-5D-5 L using health economics. Soc Sci Med. 2009;69(5):786–92. the external validation sample. Results for the best fitting models, models 3. Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. 1 and 2, when mapping from the LDQ to the EQ-5D and the ICECAP-A Methods for the economic evaluation of health care programmes. Oxford: using the external validation sample. (DOCX 14 kb) Oxford university press; 2005. 4. Chuang L, Whitehead SJ. Mapping for economic evaluation. Br Med Bull. 2012;101:1-15. doi: https://doi.org/10.1093/bmb/ldr049. 5. National Institute for Health and Care Excellence (NICE). Guide to the methods of technology appraisal. London: NICE; 2008. Abbreviations 6. Orford J. Power, powerlessness and addiction. Cambridge: Cambridge University AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion; CORE- Press; 2013. OM: Clinical Outcomes in Routine Evaluation - Outcome Measure; EQ-5D- 7. Sen A. Capability and well-being. The quality of life. Oxford: Oxford 5 L: EuroQol - 5 Dimensions- 5 Levels; HRQOL: Health-Related Quality Of Life; University Press; 1993. ICECAP-A: ICEpop CAPability measure for Adults; LDQ: Leeds Dependence 8. Al-Janabi H, Flynn TN, Coast J. Development of a self-report measure of Questionnaire; MAE: Mean Absolute Error.; n: Number of patients; capability wellbeing for adults: the ICECAP-A. Qual Life Res. 2012;21(1):167–76. NICE: National Institute for Health and Care Excellence; OLS: Ordinary Least 9. Goranitis I, Coast J, Day E, Copello A, Freemantle N, Seddon J, et al. Squares; OST: Opiate Substitution Treatment; QALY: Quality-Adjusted Life Measuring health and broader well-being benefits in the context of opiate Years; RCT: Randomised Control Trial; RMSE: Root-Mean-Squared-Error; dependence: the psychometric performance of the ICECAP-A and the EQ- SD: Standard Deviation; TOP: Treatment Outcomes Profile 5D-5L. Value Health. 2016;19(6):820–8. 10. Day E, Copello A, Seddon JL, Christie M, Bamber D, Powell C, et al. Pilot study of a social network intervention for heroin users in opiate substitution treatment: Acknowledgments study protocol for a randomized controlled trial. Trials. 2013;14(1):264. We would like to thank everyone who participated in the study and to the 11. Day E, Copello A, Seddon JL, Christie M, Bamber D, Powell C, Bennett C, Akhtar community drug and alcohol teams located at Leicester (Leicestershire S, George S, Ball A, Frew E, Goranitis I, Freemantle N. A pilot feasibility Partnership NHS Trust) and Solihull (Birmingham & Solihull Mental Health randomised controlled trial of an adjunct brief social network intervention in NHS Foundation Trust). opiate substitution treatment services. BMC Psychiatry. 2018;18(1):8. Peak et al. Health and Quality of Life Outcomes (2018) 16:106 Page 11 of 11 12. Baio G. Bayesian methods in health economics. Chapman and Hall/CRC 36. Khan KA, Madan J, Petrou S, Lamb SE. Mapping between the Roland Morris Press. Boca Raton, FL, US. 2012. questionnaire and generic preference-based measures. Value Health. 2014; 13. Petrou S, Rivero-Arias O, Dakin H, Longworth L, Oppe M, Froud R, et al. Preferred 17(6):686–95. reporting items for studies mapping onto preference-based outcome measures: 37. Marra CA, Woolcott JC, Kopec JA, Shojania K, Offer R, Brazier JE, et al. A the MAPS statement. Health Qual Life Outcomes. 2015b;13(1):106. comparison of generic, indirect utility measures (the HUI2, HUI3, SF-6D, and the EQ-5D) and disease-specific instruments (the RAQoL and the HAQ) in 14. Brazier JE, Yang Y, Tsuchiya A, Rowen DL. A review of studies mapping (or rheumatoid arthritis. Soc Sci Med. 2005;60(7):1571–82. cross walking) non-preference based measures of health to generic 38. Bray JW, Aden B, Eggman AA, Hellerstein L, Wittenberg E, Nosyk B, et al. preference-based measures. Eur J Health Econ. 2010;11(2):215–25. Quality of life as an outcome of opioid use disorder treatment: a systematic 15. Petrou S, Rivero-Arias O, Dakin H, Longworth L, Oppe M, Froud R, et al. The review. J Subst Abus Treat. 2017;76:88–93. MAPS reporting statement for studies mapping onto generic preference- 39. Young MK, Ng S, Mellick G, Scuffham PA. Mapping of the PDQ-39 to EQ-5D based outcome measures: explanation and elaboration. scores in patients with Parkinson’s disease. Qual Life Res. 2013;22(5):1065–72. PharmacoEconomics. 2015a;33(10):993–1011. 40. Kontodimopoulos N, Aletras VH, Paliouras D, Niakas D. Mapping the cancer- 16. Evans C, Connell J, Barkham M, Margison F, McGrath G, Mellor-Clark J, et al. specific EORTC QLQ-C30 to the preference-based EQ-5D, SF-6D, and 15D Towards a standardised brief outcome measure: psychometric properties instruments. Value Health. 2009;12(8):1151–7. and utility of the CORE-OM. Br J Psychiatry. 2002 Jan;180:51–60. 41. Cheung YB, Luo N, Ng R, Lee CF. Mapping the functional assessment of 17. Barkham M, Mellor-Clark J, Connell J, Cahill J. A CORE approach to practice- Cancer therapy-breast (FACT-B) to the 5-level EuroQoL Group’s 5-dimension based evidence: a brief history of the origins and applications of the CORE- questionnaire (EQ-5D-5L) utility index in a multi-ethnic Asian population. OM and CORE system. Couns Psychother Res. 2006;6(1):3–15. Health Qual Life Outcomes. 2014;12(1):180. 18. Raistrick D, Bradshaw J, Tober G, Weiner J, Allison J, Healey C. Development 42. Brazier J, Connell J, Papaioannou D, Mukuria C, Mulhern B, Peasgood T, et al. of the Leeds dependence questionnaire (LDQ): a questionnaire to measure A systematic review, psychometric analysis and qualitative assessment of alcohol and opiate dependence in the context of a treatment evaluation generic preference-based measures of health in mental health populations package. Addiction. 1994;89(5):563–72. and the estimation of mapping functions from widely used specific 19. Raistrick R, Bradshaw J, Tober G, Weiner J, Allison J, Healey C. Development measures. Health Technol Assess. 2014;18(34):vii,viii, xiii-xxv):1–188. of the Leeds dependence questionnaire (LDQ): a questionnaire to measure 43. Hawton A, Green C, Telford C, Wright D, Zajicek J. The use of multiple alcohol and opiate dependence in the context of a treatment evaluation sclerosis condition-specific measures to inform health policy decision- package. Addiction. 1994;89(5):563-572. making: mapping from the MSWS-12 to the EQ-5D. Mult Scler J. 2012;18(6): 20. Marsden J, Farrell M, Bradbury C, Dale-Perera A, Eastwood B, Roxburgh M, et 853–61. al. Development of the treatment outcomes profile. Addiction. 2008;103(9): 44. Kim H, Kim D, Jang EJ, Lee M, Song HJ, Park S, et al. Mapping health 1450–60. assessment questionnaire disability index (HAQ-DI) score, pain visual analog 21. National Treatment Agency for Substance Misuse. Treatment Outcomes scale (VAS), and disease activity score in 28 joints (DAS28) onto the EuroQol-5D Profile (TOP) [Internet]; 2017. Available from: http://www.nta.nhs.uk/ (EQ-5D) utility score with the KORean observational study network for arthritis healthcare-TOP.aspx. (KORONA) registry data. Rheumatol Int. 2016;36(4):505–13. 22. Herdman M, Gudex C, Lloyd A, Janssen M, Kind P, Parkin D, et al. 45. Goranitis I, Coast J, Day E, Copello A, Freemantle N, Frew E. Maximizing Development and preliminary testing of the new five-level version of EQ-5D health or sufficient capability in economic evaluation? A methodological (EQ-5D-5L). Qual Life Res. 2011;20(10):1727–36. experiment of treatment for drug addiction. Med Decis Mak. 2017;37(5): 23. Devlin N, Shah K, Feng Y, Mulhern B, van Hout B. Valuing health-related 498–511. quality of life: an EQ-5D-5L value set for England.. 2016. 24. Flynn TN, Huynh E, Peters TJ, Al-Janabi H, Clemens S, Moody A, et al. Scoring the Icecap-a capability instrument. Estimation of a UK general population tariff. Health Econ. 2015;24(3):258–69. 25. Longworth L, Rowen D. Mapping to obtain EQ-5D utility values for use in NICE health technology assessments. Value Health. 2013;16(1):202–10. 26. Crott R, Briggs A. Mapping the QLQ-C30 quality of life cancer questionnaire to EQ-5D patient preferences. Eur J Health Econ. 2010;11(4):427–34. 27. McKenzie L, Van Der Pol M. Mapping the EORTC QLQ C-30 onto the EQ-5D instrument: the potential to estimate QALYs without generic preference data. Value Health. 2009;12(1):167–71. 28. Pinedo-Villanueva RA, Turner D, Judge A, Raftery JP, Arden NK. Mapping the Oxford hip score onto the EQ-5D utility index. Qual Life Res. 2013;22(3):665–75. 29. Gray AM, Clarke PM, Wolstenholme JL, Wordsworth S. Applied methods of cost-effectiveness analysis in healthcare. Oxford: OUP; 2010. 30. Mitchell PM, Roberts TE, Barton PM, Pollard BS, Coast J. Predicting the ICECAP-O capability index from the WOMAC osteoarthritis index: is mapping onto capability from condition-specific health status questionnaires feasible? Med Decis Mak. 2013;33(4):547–57. 31. Dakin H, Gray A, Murray D. Mapping analyses to estimate EQ-5D utilities and responses based on Oxford knee score. Qual Life Res. 2013;22(3):683–94. 32. Skaltsa K, Longworth L, Ivanescu C, Phung D, Holmstrom S. Mapping the FACT-P to the preference-based EQ-5D questionnaire in metastatic castration-resistant prostate Cancer. Value Health. 2014;17(2):238–44. 33. Sullivan PW, Ghushchyan V. Mapping the EQ-5D index from the SF-12: US general population preferences in a nationally representative sample. Med Decis Mak. 2006;26(4):401–9. 34. Young TA, Mukuria C, Rowen D, Brazier JE, Longworth L. Mapping functions in health-related quality of life: mapping from two cancer-specific health-related quality-of-life instruments to EQ-5D-3L. Med Decis Mak. 2015;35(7):912–26. 35. Rivero-Arias O, Ouellet M, Gray A, Wolstenholme J, Rothwell PM, Luengo- Fernandez R. Mapping the modified Rankin scale (mRS) measurement into the generic EuroQol (EQ-5D) health outcome. Med Decis Mak. 2010;30(3):341–54.

Journal

Health and Quality of Life OutcomesSpringer Journals

Published: May 30, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off