- Email: [email protected]
Systematic Review Toolbox
A398
VA L U E I N H E A LT H 1 9 ( 2 0 1 6 ) A 3 4 7 – A 7 6 6
allocation. Analysis: Firstly, several attributes proposed for inclusion within MCDA can be disputed as spurious. For example, it is questionable whether innovation, disease rarity and budget impact should be accepted as valid attributes. Secondly, the inclusion of attributes in a weighted sum MCDA approach requires quantification of the attributes’ level. How such attributes should be quantified is unclear in many cases. Thirdly, there are good reasons to suggest that cost-effectiveness should be represented within MCDA using a net benefit measure, rather than cost-effectiveness ratios. Net benefit is an absolute metric that varies with the size of the patient population. Consequently it is unclear how an absolute measure should be weighed against more abstract concepts such as disease severity or unmet need. Finally, to achieve a coherent resource allocation framework it is essential that any decision rule employing MCDA account for the opportunity cost of health effects and other benefits foregone in the interventions not funded. Accounting for a broader scope of benefits in the opportunity cost requires a reduction in the cost-effectiveness threshold. Conclusions: There are strong arguments in favour of expanding the scope of benefits included in CEA beyond health effects. Nevertheless, this work shows that there are a number of profound questions for the application of MCDA in healthcare resource allocations that must be resolved before it can be considered a suitable replacement for current CEA methods. PRM223 Systematic Review Toolbox Marshall C1, Sutton A2 Health Economics Consortium Ltd, York, UK, 2University of Sheffield, Sheffield, UK
1York
Background: Systematic reviews facilitate the rigorous identification and synthesis of evidence in healthcare. However, they can be time consuming, logistically challenging and labour intensive to undertake. Such challenges have led to the development of various software tools to support the systematic review process. It has remained difficult, however, for researchers to easily discover what tools are currently available to support their reviews. In response, a free, online resource of tools to support systematic reviews has been developed. Concept: Systematic Review (SR) Toolbox [1] is a community-driven, searchable, web-based catalogue of tools to support the systematic review process. The resource aims to help reviewers find appropriate tools based on the stage(s) of the systematic review process. Users can perform a simple keyword search (i.e. Quick Search) to locate tools, or a more detailed search (i.e. Advanced Search) allowing users to select various criteria to find specific types of tools. Users are also able to submit new tools they have found, used or developed, to the SR Toolbox. The resource was developed using PHP, SQL and JavaScript and uses Twitter [2] to manage its community. Results: The SR Toolbox stores information on over 100 software tools (June 2016) to support the systematic review process. These include freely-available tools based on data visualisation, text mining and machine learning approaches. The SR toolbox also catalogues a number of more substantial commercial and not-for-profit software packages that manage substantial portions of the systematic review process. Paper-based tools, such as guidelines, checklists and reporting standards, are also included. Recommendations: The SR Toolbox is a useful way to keep up to date with new software tools which have potential to make the SR process more efficient. We recommend researchers consider using the SR Toolbox as a resource for finding tools, and encourage software developers to catalogue any new tools. [1] http:// systematicreviewtools.com/ [2] http://twitter.com/SRToolbox PRM224 A Patient Level Simulation Model Gharibnaseri Z1, Davari M1, Cheraghali A2, Eshghi P3, Ravanbod R4, Spendar R1 University of Medical Sciences, Tehran, Iran (Islamic Republic of), 2Baghyatollah Medical Sciences University, Tehran, Iran (Islamic Republic of), 3Shahid Beheshti University of Medical Sciences, Tehran, Iran (Islamic Republic of), 4Tarbiat Modares University, Tehran, Iran (Islamic Republic of) 1Tehran
Understanding how treatment outcomes differ among individuals with heterogeneous characteristics is critical in creating simulation models that are closer to reality. Although Markov structures can be designed in such a way to reflect these differences, the unavoidable increase in model complexity can be seen as a disadvantage to using such models. Another limitation of Markov chains is the fact that these models are developed by defining discrete states of health supported by literature on patients’ transitions from one state to another. However, in real practice, patients’ health states usually represent a continuous spectrum rather than discrete health states. In this paper, we present a patient level model which has overcame these shortcomings using patient centered outcomes. This model was used for performing a cost-effectiveness analysis on alternative treatment protocols in haemophilia patients. Cross sectional data on patient’s characteristics, health status and quality of life was gathered and transformed to longitudinal data using the simulation model. Accumulated costs and health outcomes during patients’ adulthood was calculated and compared in both treatments. The presented model can be adjusted to other settings and chronic diseases. PRM225 Interactive Health Economic Models Configured With Google Analytics. Data And Metrics Applicable For A Healthcare Research O1, Siabro V2, Volovyk A2
Topachevskyi 1Digital Health Outcomes, Brussels, Belgium, 2Digital Health Outcomes, Kiev, Ukraine
Budget impact health economics models are often transformed into Web/iPad applications in order to improve model transparency, end user experience and communication of modeled outcomes. Google Analytics (GA) campaigns configured for health economics models provide metrics and insights that are useful for a healthcare research. GA allows setting up of custom campaigns dedicated to track specific events, usage metrics and detailed profiles of target audiences (payers, healthcare professionals) of economic models. Other important and useful information relates to specific user behavior patters within the model, user acquisition channels, geography, time spend on particular pages and other metrics. This information allows
to better understand payers perceptions and general interest to the particular parts of the economic value story. When GA is paired with embedded questioners or interface elements designed to record reaction, the service may provide data and insights required to continuously refine and upgrade an economic model to reflect real world payers perceptions. GA events is a powerful tracking tool also allowing to understand what data was changed and what were the input values entered by decision makers in a particular setting. GA also provide means to collect direct medical cost data during payer facing. Client-server software architecture of interactive economic models enable integration of different analytics web services. The majority of modern analytics services provide API for integration of necessary functionality in web based or standalone interactive health economics models. PRM226 A Conceptual Approach For Including Productivity Losses In Health Economic Evaluations From A Societal Viewpoint Müller M, Hofmann S WifOR GmbH, Darmstadt, Germany
Background: When adopting a societal perspective in health economic evaluations, productivity losses contribute in a significant way to the total costs. Nevertheless, they are rarely considered. If they are considered, they mostly include paid work only, although unpaid work (e.g. housework, informal care or voluntary work) accounts for a significant proportion of people’s lifetime. Furthermore, most severe conditions onset in the age of retirement, which means that a growing proportion of patients is not active in the labor market. Practical Implications: The restriction to paid work neglects the contribution of patients who are not active in the labor market, because they do not contribute to a nation’s added value in the classical notion in terms of paid work. However, if innovative therapies restore the ability to perform both paid and unpaid work, a consideration of productivity gains only with respect to paid work, underestimates the welfare maximizing contribution of unpaid work within a society. Taking a societal perspective, it is not appropriate to exclude unpaid work from evaluations, especially when a welfare maximizing decision is aspired. Recommendation: To value unpaid work, first, it is important to determine the amount of time people spend for unpaid work. We propose a practical approach using time surveys to identify this amount of time. Based on this, we suggest to apply a replacement cost approach and value these specific activities with sector specific wages in monetary terms, respectively equivalent market services. Findings of clinical trials or estimations by experts constitute an adequate basis to estimate the “reconstructed” time within a patient group. Health economic evaluations taking a societal perspective should consider all relevant costs. For this reason, both paid and particularly unpaid work losses should be included. We therefore recommend to value unpaid work by using national time use surveys and sector specific wages. PRM227 Guidance On Selecting Appropriate Methods When Considering Adjusting Overall Survival For Treatment Switch In Oncology Studies Watkins CL1, Latimer N2, Wang J3, Wright EJ4 Consulting Limited, Alderley Edge, UK, 2Sheffield University, Sheffield, UK, 3Celgene International, Boudry, Switzerland, 4Roche, Basel, Switzerland 1Clarostat
Several methods are available for adjusting overall survival for treatment switch. Those most commonly used in health technology assessment are described in NICE Decision Support Unit Technical Support Document 16 (Latimer and Abrams 2014). No method is universally suited to all diseases, treatments and studies, and so a situation specific assessment is required. Hence there is a need for practical guidance on the selection and implementation of the methods. Building on previous guidance, we, a PSI (Statisticians in the Pharmaceutical Industry) working group on treatment switch, have suggested a process to guide the method selection and implementation. We describe this process, including a set of useful summaries, plots and analyses that can help to determine which methods are (1) feasible based on the available data, and (2) appropriate based on their underlying assumptions. Those assumptions include no unmeasured confounders (e.g. for inverse probability of censoring weighting [IPCW] models) or common treatment effect (e.g. for rank preserving structural failure time [RPSFTM] models). Suggestions for methods to explore the viability of those assumptions are provided. Even when methods are both feasible and appropriate, model-fitting issues may occur, and we highlight some potential problems and the situations where they may arise, for awareness. The guidance provided can aid analysts and their project teams to select which methods, if any, are appropriate when considering adjusting overall survival for treatment switch in oncology studies. PRM228 Using Evidence Reviews To Develop Focused Systematic Review Questions Upton CM, Chadda S, Chapman R, Curry A SIRIUS Market Access, London, UK
Systematic reviews (SRs) have a number of important functions, but are essential for supporting Health Technology Assessments. SR questions are based on PICO elements which specify types of participants (P), interventions (I) comparators (C), and outcomes (O) of interest. Evidence reviews (ERs) can act as stand-alone documents and allow the researcher to acquire a detailed understanding of a disease area and literature base, in order to develop robust PICO elements. These PICO elements are then used to develop focused SR questions. ERs are also useful for determining if the proposed SR question has been previously answered within the literature. ERs typically take between 6 and 12 weeks to complete and, if being used to develop PICO, initially involve understanding the end-purpose of the SR (e.g. to support a network meta-analysis, or economic model). Once this is understood, researchers can agree contents of the ER. For example if the ER will be used to develop PICO for a safety and efficacy SR, it would need to include sections outlining current treatments, clinical
VA L U E I N H E A LT H 1 9 ( 2 0 1 6 ) A 3 4 7 – A 7 6 6
allocation. Analysis: Firstly, several attributes proposed for inclusion within MCDA can be disputed as spurious. For example, it is questionable whether innovation, disease rarity and budget impact should be accepted as valid attributes. Secondly, the inclusion of attributes in a weighted sum MCDA approach requires quantification of the attributes’ level. How such attributes should be quantified is unclear in many cases. Thirdly, there are good reasons to suggest that cost-effectiveness should be represented within MCDA using a net benefit measure, rather than cost-effectiveness ratios. Net benefit is an absolute metric that varies with the size of the patient population. Consequently it is unclear how an absolute measure should be weighed against more abstract concepts such as disease severity or unmet need. Finally, to achieve a coherent resource allocation framework it is essential that any decision rule employing MCDA account for the opportunity cost of health effects and other benefits foregone in the interventions not funded. Accounting for a broader scope of benefits in the opportunity cost requires a reduction in the cost-effectiveness threshold. Conclusions: There are strong arguments in favour of expanding the scope of benefits included in CEA beyond health effects. Nevertheless, this work shows that there are a number of profound questions for the application of MCDA in healthcare resource allocations that must be resolved before it can be considered a suitable replacement for current CEA methods. PRM223 Systematic Review Toolbox Marshall C1, Sutton A2 Health Economics Consortium Ltd, York, UK, 2University of Sheffield, Sheffield, UK
1York
Background: Systematic reviews facilitate the rigorous identification and synthesis of evidence in healthcare. However, they can be time consuming, logistically challenging and labour intensive to undertake. Such challenges have led to the development of various software tools to support the systematic review process. It has remained difficult, however, for researchers to easily discover what tools are currently available to support their reviews. In response, a free, online resource of tools to support systematic reviews has been developed. Concept: Systematic Review (SR) Toolbox [1] is a community-driven, searchable, web-based catalogue of tools to support the systematic review process. The resource aims to help reviewers find appropriate tools based on the stage(s) of the systematic review process. Users can perform a simple keyword search (i.e. Quick Search) to locate tools, or a more detailed search (i.e. Advanced Search) allowing users to select various criteria to find specific types of tools. Users are also able to submit new tools they have found, used or developed, to the SR Toolbox. The resource was developed using PHP, SQL and JavaScript and uses Twitter [2] to manage its community. Results: The SR Toolbox stores information on over 100 software tools (June 2016) to support the systematic review process. These include freely-available tools based on data visualisation, text mining and machine learning approaches. The SR toolbox also catalogues a number of more substantial commercial and not-for-profit software packages that manage substantial portions of the systematic review process. Paper-based tools, such as guidelines, checklists and reporting standards, are also included. Recommendations: The SR Toolbox is a useful way to keep up to date with new software tools which have potential to make the SR process more efficient. We recommend researchers consider using the SR Toolbox as a resource for finding tools, and encourage software developers to catalogue any new tools. [1] http:// systematicreviewtools.com/ [2] http://twitter.com/SRToolbox PRM224 A Patient Level Simulation Model Gharibnaseri Z1, Davari M1, Cheraghali A2, Eshghi P3, Ravanbod R4, Spendar R1 University of Medical Sciences, Tehran, Iran (Islamic Republic of), 2Baghyatollah Medical Sciences University, Tehran, Iran (Islamic Republic of), 3Shahid Beheshti University of Medical Sciences, Tehran, Iran (Islamic Republic of), 4Tarbiat Modares University, Tehran, Iran (Islamic Republic of) 1Tehran
Understanding how treatment outcomes differ among individuals with heterogeneous characteristics is critical in creating simulation models that are closer to reality. Although Markov structures can be designed in such a way to reflect these differences, the unavoidable increase in model complexity can be seen as a disadvantage to using such models. Another limitation of Markov chains is the fact that these models are developed by defining discrete states of health supported by literature on patients’ transitions from one state to another. However, in real practice, patients’ health states usually represent a continuous spectrum rather than discrete health states. In this paper, we present a patient level model which has overcame these shortcomings using patient centered outcomes. This model was used for performing a cost-effectiveness analysis on alternative treatment protocols in haemophilia patients. Cross sectional data on patient’s characteristics, health status and quality of life was gathered and transformed to longitudinal data using the simulation model. Accumulated costs and health outcomes during patients’ adulthood was calculated and compared in both treatments. The presented model can be adjusted to other settings and chronic diseases. PRM225 Interactive Health Economic Models Configured With Google Analytics. Data And Metrics Applicable For A Healthcare Research O1, Siabro V2, Volovyk A2
Topachevskyi 1Digital Health Outcomes, Brussels, Belgium, 2Digital Health Outcomes, Kiev, Ukraine
Budget impact health economics models are often transformed into Web/iPad applications in order to improve model transparency, end user experience and communication of modeled outcomes. Google Analytics (GA) campaigns configured for health economics models provide metrics and insights that are useful for a healthcare research. GA allows setting up of custom campaigns dedicated to track specific events, usage metrics and detailed profiles of target audiences (payers, healthcare professionals) of economic models. Other important and useful information relates to specific user behavior patters within the model, user acquisition channels, geography, time spend on particular pages and other metrics. This information allows
to better understand payers perceptions and general interest to the particular parts of the economic value story. When GA is paired with embedded questioners or interface elements designed to record reaction, the service may provide data and insights required to continuously refine and upgrade an economic model to reflect real world payers perceptions. GA events is a powerful tracking tool also allowing to understand what data was changed and what were the input values entered by decision makers in a particular setting. GA also provide means to collect direct medical cost data during payer facing. Client-server software architecture of interactive economic models enable integration of different analytics web services. The majority of modern analytics services provide API for integration of necessary functionality in web based or standalone interactive health economics models. PRM226 A Conceptual Approach For Including Productivity Losses In Health Economic Evaluations From A Societal Viewpoint Müller M, Hofmann S WifOR GmbH, Darmstadt, Germany
Background: When adopting a societal perspective in health economic evaluations, productivity losses contribute in a significant way to the total costs. Nevertheless, they are rarely considered. If they are considered, they mostly include paid work only, although unpaid work (e.g. housework, informal care or voluntary work) accounts for a significant proportion of people’s lifetime. Furthermore, most severe conditions onset in the age of retirement, which means that a growing proportion of patients is not active in the labor market. Practical Implications: The restriction to paid work neglects the contribution of patients who are not active in the labor market, because they do not contribute to a nation’s added value in the classical notion in terms of paid work. However, if innovative therapies restore the ability to perform both paid and unpaid work, a consideration of productivity gains only with respect to paid work, underestimates the welfare maximizing contribution of unpaid work within a society. Taking a societal perspective, it is not appropriate to exclude unpaid work from evaluations, especially when a welfare maximizing decision is aspired. Recommendation: To value unpaid work, first, it is important to determine the amount of time people spend for unpaid work. We propose a practical approach using time surveys to identify this amount of time. Based on this, we suggest to apply a replacement cost approach and value these specific activities with sector specific wages in monetary terms, respectively equivalent market services. Findings of clinical trials or estimations by experts constitute an adequate basis to estimate the “reconstructed” time within a patient group. Health economic evaluations taking a societal perspective should consider all relevant costs. For this reason, both paid and particularly unpaid work losses should be included. We therefore recommend to value unpaid work by using national time use surveys and sector specific wages. PRM227 Guidance On Selecting Appropriate Methods When Considering Adjusting Overall Survival For Treatment Switch In Oncology Studies Watkins CL1, Latimer N2, Wang J3, Wright EJ4 Consulting Limited, Alderley Edge, UK, 2Sheffield University, Sheffield, UK, 3Celgene International, Boudry, Switzerland, 4Roche, Basel, Switzerland 1Clarostat
Several methods are available for adjusting overall survival for treatment switch. Those most commonly used in health technology assessment are described in NICE Decision Support Unit Technical Support Document 16 (Latimer and Abrams 2014). No method is universally suited to all diseases, treatments and studies, and so a situation specific assessment is required. Hence there is a need for practical guidance on the selection and implementation of the methods. Building on previous guidance, we, a PSI (Statisticians in the Pharmaceutical Industry) working group on treatment switch, have suggested a process to guide the method selection and implementation. We describe this process, including a set of useful summaries, plots and analyses that can help to determine which methods are (1) feasible based on the available data, and (2) appropriate based on their underlying assumptions. Those assumptions include no unmeasured confounders (e.g. for inverse probability of censoring weighting [IPCW] models) or common treatment effect (e.g. for rank preserving structural failure time [RPSFTM] models). Suggestions for methods to explore the viability of those assumptions are provided. Even when methods are both feasible and appropriate, model-fitting issues may occur, and we highlight some potential problems and the situations where they may arise, for awareness. The guidance provided can aid analysts and their project teams to select which methods, if any, are appropriate when considering adjusting overall survival for treatment switch in oncology studies. PRM228 Using Evidence Reviews To Develop Focused Systematic Review Questions Upton CM, Chadda S, Chapman R, Curry A SIRIUS Market Access, London, UK
Systematic reviews (SRs) have a number of important functions, but are essential for supporting Health Technology Assessments. SR questions are based on PICO elements which specify types of participants (P), interventions (I) comparators (C), and outcomes (O) of interest. Evidence reviews (ERs) can act as stand-alone documents and allow the researcher to acquire a detailed understanding of a disease area and literature base, in order to develop robust PICO elements. These PICO elements are then used to develop focused SR questions. ERs are also useful for determining if the proposed SR question has been previously answered within the literature. ERs typically take between 6 and 12 weeks to complete and, if being used to develop PICO, initially involve understanding the end-purpose of the SR (e.g. to support a network meta-analysis, or economic model). Once this is understood, researchers can agree contents of the ER. For example if the ER will be used to develop PICO for a safety and efficacy SR, it would need to include sections outlining current treatments, clinical