Introduction
Health services research (HSR), health economics and financing, and research of quality and equality in healthcare require comparable data on service provision (Husereau et al., Reference Husereau, Drummond, Petrou, Carswell, Moher, Greenberg, Augustovski, Briggs, Mauskopf and Loder2013; Raine et al., Reference Raine, Fitzpatrick, Barratt, Bevan, Black, Boaden, Bower, Campbell, Denis, Devers, Dixon-Woods, Fallowfield, Forder, Foy, Freemantle, Fulop, Gibbons, Gillies, Goulding, Grieve, Grimshaw, Howarth, Lilford, McDonald, Moore, Moore, Newhouse, O'Cathain, Or, Papoutsi, Prady, Rycroft-Malone, Sekhon, Turner, Watson and Zwarenstein2016). However, the reporting methods can differ substantially, and HSR faces significant problems regarding the terminology of services (Salvador-Carulla et al., Reference Salvador-Carulla, Alvarez-Galvez, Romero, Gutiérrez-Colosía, Weber, McDaid, Dimitrov, Sprah, Kalseth, Tibaldi, Salinas-Perez, Lagares-Franco, Romá-Ferri and Johnson2013), its implication for the measurement of resource utilisation (Thorn et al., Reference Thorn, Coast, Cohen, Hollingworth, Knapp, Noble, Ridyard, Wordsworth and Hughes2013) and its monetary valuation (Moreno et al., Reference Moreno, Sanchez and Salvador-Carulla2008; Barnett, Reference Barnett2009). Terminology is defined as ‘a set of designations belonging to one special language’(Roche, Reference Roche2012), and its main purpose is to eliminate ambiguity from technical languages by means of standardisation. The main terminological problems in scientific research are unclarity due to ambiguity or vagueness of scientific terms. Terminological ambiguity exists when a term (the dyad of a name and its definition) can reasonably be interpreted in more than one way (e.g. two different codes of a reference classification system can be assigned to the same entity), and vagueness occurs when a word or phrase is underspecified and therefore admits borderline cases or relative interpretation (e.g. typically more than three codes can be assigned to the defined entity) (Castelpietra et al., Reference Castelpietra, Simon, Gutierrez-Colosía, Rosenberg and Salvador-Carulla2021). Disambiguation is the act of making something clear and this takes place in science by using taxonomies, for example, an ontology-based classification coding system and its related glossary of terms or dictionary. Ontoterminology is the discipline that studies disambiguation of technical and scientific terms using classifications, glossaries of terms and the related standard instruments (Castelpietra et al., Reference Castelpietra, Simon, Gutierrez-Colosía, Rosenberg and Salvador-Carulla2021).
The two major terminological problems in HSR are the non-commensurability bias and terminological unclarity. Non-commensurability is due to research involving different units of analysis that are not comparable like-with-like. For example, it occurs when the costs of an outpatient psychotherapy unit (i.e. a ‘service’) are compared to the costs of psychotherapy as an ‘intervention’ in another setting. The problem of terminological unclarity is also widespread. For example, the term ‘service’ can refer to a range of elements such as the provider, the facility, an organisational unit within the facility or a combination of functions, programmes and resources provided in this clinical unit (Salvador-Carulla et al., Reference Salvador-Carulla, Alvarez-Galvez, Romero, Gutiérrez-Colosía, Weber, McDaid, Dimitrov, Sprah, Kalseth, Tibaldi, Salinas-Perez, Lagares-Franco, Romá-Ferri and Johnson2013). Another problem refers to the lack of a formal definition of the target population the service has been designed for (i.e. diagnosis group and age group) (Salinas-Pérez et al., Reference Salinas-Pérez, Gutiérrez-Colosía, Romero López-Alberca, Poole, Rodero-Cosano, García-Alonso and Salvador-Carulla2020), and the variability in the typology of services depending on location (different areas, regions, countries, etc.) and time of evaluation (Salvador-Carulla et al., Reference Salvador-Carulla, Amaddeo, Gutiérrez-Colosía, Salazzari, Gonzalez-Caballero, Montagni, Tedeschi, Cetrano, Chevreul, Kalseth, Hagmair, Straßmayr, Park, Sfetcu, Wahlbeck and Garcia-Alonso2015).
Twenty years ago, Maciejewski et al. identified significant terminology problems in the methods for HSR. To overcome these problems, these authors produced a list of terms commonly used in HSR methods following a scoping review of the literature and internal and external expert consultations (Maciejewski et al., Reference Maciejewski, Diehr, Smith and Hebert2002). These terminology problems have been also described in mental health service evaluation (Salvador-Carulla et al., Reference Salvador-Carulla, Atienza, Romero, Guimón and Sartorius1999; Salvador-Carulla and Hernández-Peña, Reference Salvador-Carulla and Hernández-Peña2011) and are critical in the standardisation of international resource use measurement (RUM) instruments (Thorn et al., Reference Thorn, Coast, Cohen, Hollingworth, Knapp, Noble, Ridyard, Wordsworth and Hughes2013; Noben et al., Reference Noben, De Rijk, Nijhuis, Kottner and Evers2016).
Despite previous efforts, the terminology problems in HSR remain largely unnoticed and unaccounted for. For instance, in the USA, the Institute of Medicine (IoM) prioritised different areas of comparative effectiveness research but did not mention this source of systematic bias (Iglehart, Reference Iglehart2009). Likewise, there is a substantial degree of variation in the applied valuation methods in health economic studies and guidelines in Europe, but the terminological variability is rarely being mentioned (van Lier et al., Reference van Lier, Bosmans, van Hout, Mokkink, van den Hout, de Wit, Dirksen, Nies, Hertogh and van der Roest2018; Mayer et al., Reference Mayer, Fischer, Zechmeister-Koss, Ostermann and Simon2020). The World Health Organization's Family of International Classification (WHO-FIC) has recently incorporated the International Classification of Health Interventions (ICHI) (Fortune et al., Reference Fortune, Madden and Almborg2018) to its list. However, the classification of the services where these interventions occur is still missing, and alternative solutions such as the Service Availability and Readiness Assessment (SARA) (O'Neill et al., Reference O'Neill, Takane, Sheffel, Abou-Zahr and Boerma2013) are too vague and broad to be used in comparative effectiveness or for disambiguation. The System of Health Accounts (SHA 2.0) (OECD Eurostat WHO, 2017) includes separate components of Health Providers and Health Functions but it lacks a reference taxonomy and a standard glossary and shows major consistency problems (Salvador-Carulla and Hernández-Peña, Reference Salvador-Carulla and Hernández-Peña2011).
The lack of specific studies on terminological in HSR is surprising. It could be attributed to the complexity of the analysis required to measure the extent of this problem (Maciejewski et al., Reference Maciejewski, Diehr, Smith and Hebert2002); and the absence of a proper framework of reference until very recently. In the last decade, ontoterminology has been proposed in information technologies (Roche, Reference Roche2012) and adapted to disambiguation in HSR (Castelpietra et al., Reference Castelpietra, Simon, Gutierrez-Colosía, Rosenberg and Salvador-Carulla2021). Apart from providing an adequate framework for the analysis of terms in a given field, a classification using a hierarchical taxonomy with a coding system provides a reference framework to code definitions as acceptable, or as ambiguous, vague or confusing (i.e. wrong or mistaken) in a reproducible way (Castelpietra et al., Reference Castelpietra, Simon, Gutierrez-Colosía, Rosenberg and Salvador-Carulla2021).
The aim of the study was to identify the magnitude of the bias of non-commensurability and terminological unclarity in health and social services research by applying an international classification system for coding human services and adapting it as needed to the newly emerging requirements for health economics research from a societal perspective. A complementary objective of this study was to demonstrate the usability of the ontoterminology approach to disambiguation in complex topics in healthcare research.
Methods
This study was part of the PECUNIA project (ProgrammE in Costing, resource use measurement and outcome valuation for Use in multi-sectoral National and International health economic evaluAtions) conducted from 2018 to 2021. PECUNIA was aimed at developing standardised multi-sectoral, multi-national and multi-person RUM instruments, unit cost valuation templates, reference unit costs and outcome assessment tools to improve the methodology and comparability of economic evaluations in the European Union with a special focus on mental health (Mayer et al., Reference Mayer, Berger, Konnopka, Brodzsky, Evers, Hakkaart-van Roijen, Gutiérrez-Colosía, Salvador-Carulla, Park, Hollingworth, García-Pérez and Simon2022). The PECUNIA consortium coordinated by the Medical University of Vienna consisted of ten partners for health economics and health systems research, located in six European countries (Austria, Germany, Hungary, Spain, The Netherlands and UK) (The PECUNIA group, 2018). Due to their high disease burden and economic relevance, three mental disorders (depression, schizophrenia and posttraumatic stress disorder) were chosen as reference disorders to analyse the applicability of the newly developed methods and tools. This study concentrated on the disambiguation of services in the health and social care cluster relevant for mental health and was carried out in parallel to other activities of the project.
Procedure
This ontoterminology study was performed using the Standards for Quality Improvement Reporting Excellence (SQUIRE 2.0) (Ogrinc et al., Reference Ogrinc, Davies, Goodman, Batalden, Davidoff and Stevens2016). It used a mixed-methods approach and followed a multistep process to assess the clarity of terms on health and social services. This process included three steps: (I) a systematic review to produce a preliminary listing of service terms based on scientific and grey literature; (II) an expert survey and a consecutive revision for the production of the final listing; (III) disambiguation of the terms included in the final listing adapting the previously tested method of Maciejewski et al. (Reference Maciejewski, Diehr, Smith and Hebert2002). Steps I and II are fully described elsewhere (Fischer et al., Reference Fischer, Mayer, Perić and Simon2022; Mayer et al., Reference Mayer, Berger, Konnopka, Brodzsky, Evers, Hakkaart-van Roijen, Gutiérrez-Colosía, Salvador-Carulla, Park, Hollingworth, García-Pérez and Simon2022).They involved three working groups and two expert panels. Working group A included three experts in health economics from the University of Hamburg (PH, AK, CD) that led step I and participated in step II. Working group B comprised three members from the Medical University of Vienna (JS, CF, SM) who led step II and participated in step I. Internal expert panel included the PECUNIA country leads and participated in steps I and II. Finally, the external expert panel was composed of 29 health and social service researchers, health economists, and planners from public agencies and other stakeholders in every participating country that provided an external validation of terms identified previously in step I.
Working group C consisted of two experts in health system terminology and coding from Psicost in Spain and Australia (MGC and LSC) who carried out the disambiguation analysis (step III). The whole process and the activities performed by the working groups and the expert panels are shown in Fig. 1.
Multistep process for the ontoterminology study.
Ontoterminology tools
We used an updated glossary of terms based on the Psicost and REFINEMENT glossaries (Montagni et al., Reference Montagni, Salvador-Carulla, Mcdaid, Straßmayr, Endel, Näätänen, Kalseth, Kalseth, Matosevic, Donisi, Chevreul, Prigent, Sfectu, Pauna, Gutiérrez-Colosia, Amaddeo and Katschnig2018) and an international classification of human services, the Description and Evaluation of Services and Directories (DESDE system) (Salvador-Carulla et al., Reference Salvador-Carulla, Alvarez-Galvez, Romero, Gutiérrez-Colosía, Weber, McDaid, Dimitrov, Sprah, Kalseth, Tibaldi, Salinas-Perez, Lagares-Franco, Romá-Ferri and Johnson2013) for the disambiguation of terms. The REFINEMENT glossary provides consensus-based operational definitions of the basic terms relevant for the disambiguation process in health services. DESDE has been used for the comparison of mental health service typologies across countries (Alonso-Solís et al., Reference Alonso-Solís, Ochoa, Grasa, Rubinstein, Caspi, Farkas, Unoka, Usall, Huerta-Ramos, Isohanni, Seppälä, Reixach, Berdún, Corripio, Alcalde, D'amico, Almazán, Bitter, Baccinelli, Bonizzi, Bulgheroni, Mendivelso, Coenen, Cohen, Constant, Escobar, Fazekas, Feldman, Gimenez, van der Graaf, Herman, Hospedales, Jääskeläinen, Jewel, Juola, Jämsä, Kaye, Kokkinakis, Koponen, Marcó, Mentzas, Miettunen, Moilanen, Papas, Paraskevopoulos, Roldán, Rubio-Abadal, Sebú, Seppälä, Simonetti, Stevens, Tauro, Triantafillou, Unoka, Vella, Vermeir and de Vita2020), analysis of disambiguation of complex terms in health care, such as psychotherapy, (Castelpietra et al., Reference Castelpietra, Simon, Gutierrez-Colosía, Rosenberg and Salvador-Carulla2021) and for the content analysis of the national classifications system compared to an international standard (Rosen et al., Reference Rosen, Rock and Salvador-Carulla2020). Previous research has shown that the DESDE instrument scores high in feasibility, consistency, inter-rater reliability as well as face, content and construct validity (Salvador-Carulla et al., Reference Salvador-Carulla, Alvarez-Galvez, Romero, Gutiérrez-Colosía, Weber, McDaid, Dimitrov, Sprah, Kalseth, Tibaldi, Salinas-Perez, Lagares-Franco, Romá-Ferri and Johnson2013), as well as its applicability in health economics studies (Romero-Lopez-Alberca et al., Reference Romero-Lopez-Alberca, Gutierrez-Colosia, Salinas-Pérez, Almeda, Furst, Johnson and Salvador-Carulla2019).
These two related tools facilitate different types of disambiguation. On the one hand, the glossary of terms provides an operational definition of services for the identification of these units of analysis (commensurability) (Montagni et al., Reference Montagni, Salvador-Carulla, Mcdaid, Straßmayr, Endel, Näätänen, Kalseth, Kalseth, Matosevic, Donisi, Chevreul, Prigent, Sfectu, Pauna, Gutiérrez-Colosia, Amaddeo and Katschnig2018).
On the other hand, the DESDE uses its multiaxial coding system for clarity in the definition of terms. Two DESDE axes were used in the disambiguation process. The ‘target’ axis includes a code sub-thread to define the specific target population for whom the service is intended (age, gender, ICD coding, functioning and severity). The ‘service’ axis includes the specific code of the service type and its qualifiers. This code is based on the principal function of the service described as ‘Main Types of Care’ (MTCs). There are six main branches that describe the type of care: Residential care, Day care, Outpatient care, Self-help support, Information and Assessment, and Accessibility of care (Romero-Lopez-Alberca et al., Reference Romero-Lopez-Alberca, Gutierrez-Colosia, Salinas-Pérez, Almeda, Furst, Johnson and Salvador-Carulla2019). The DESDE hierarchical taxonomy includes 106 codes in five levels of granularity (main branch of care, acute/non-acute care, mobile/non-mobile care, physician/non-physician cover, intensity of care) labelled with an alphanumeric code. In this study, we requested for disambiguation only the two first digits of the label (letter + one number, e.g. O5 in Fig. 1), instead of using the full five levels of granularity of the MTC taxonomy.
In the example presented in Fig. 2, the code thread refers to the BSIC type called ‘Assertive Community Treatment team’ in the Basque Country (Spain) (García-Alonso et al., Reference García-Alonso, Almeda, Salinas-Pérez, Gutiérrez-Colosía, Uriarte-Uriarte and Salvador-Carulla2019).
Basic DESDE structure.
Disambiguation
Steps I and II yielded a list of 56 key services relevant to mental health based on extensive literature review and selected by international expert panels. Terms and definitions were classified into the following categories: accurate (the term could be classified using one code); ambiguous (the term was labelled with more than one – typically two – code), vague (the term could be coded with a series of codes), confusing (the term was wrong or incomplete according to the reference classification system as it required additional significant interpretation from the experts). Definitions were analysed at different levels. In level 1, the two raters confirmed that the definitions corresponded to services and not to another unit of analysis in HSR such as procedures, interventions or professionals, to ensure the commensurability of the terms included in the listing (Salvador-Carulla et al., Reference Salvador-Carulla, Amaddeo, Gutiérrez-Colosía, Salazzari, Gonzalez-Caballero, Montagni, Tedeschi, Cetrano, Chevreul, Kalseth, Hagmair, Straßmayr, Park, Sfetcu, Wahlbeck and Garcia-Alonso2015). In level 2, the two raters analysed the information on the target population for which the service was intended. This included age, diagnosis group, functioning or other characteristics influencing health status and contact with health services (e.g. homelessness, domestic violence). Finally, level 3 of disambiguation included the definition of the service typology using DESDE taxonomy based on MTC codes. For further clarification, a full DESDE code was provided for every service. When problems in the definition of the service did not allow the assignment of a code, a prototype code was generated based on the interpretation made by the two evaluators. For example, in the list provided, ‘outpatient healthcare service’ is defined as a contact with the provider. This was categorised as confusing because ‘contact’ refers to an activity conducted by a professional. Codification was based on the most exemplary instance of that type of service according to expert analysis and interpretation. It has been underlined and written in italics (online Supplementary Table 1). Interobserver agreement was analysed using the standard interrater reliability analysis of Cohen's κ for categorical variables; results were interpreted following Landis and Koch's criteria (Reference Landis and Koch1977). Data were analysed by using SPSS Statistics for Windows.
Ethics
The listing of terms and definitions used in this study did not require ethical approval or consents in the participating countries as data were obtained from a review of the scientific and grey literature and did not include information on individual patients.
Results
Disambiguation of the final service listing
The basic listing included 35 terms corresponding to generic services for any health condition relevant to persons experiencing a mental disorder, and 21 that were specific services for mental health. This list was the basis for the master service list used for all next steps in the development of the PECUNIA costing tools. Initially only 13 terms (32%) had accurate definitions (not confusing, ambiguous or vague) according to the classification of reference. The disambiguation was analysed at three different levels:
Level 1: commensurability (the unit of analysis actually refers to services and not to other entities such as interventions or medical products)
Fifty terms (89%) were considered accurate at level 1 (online Supplementary Table 1). Six terms (11%) were considered confusing regarding the unit of analysis included in the definition. As an example, ‘emergency ambulance ride: a special vehicle used to take sick or injured people to a hospital or other health care facility in case of emergency’, defined a device or medical good (a medicalised vehicle) but not the actual care provided. If this definition refers to the use of the vehicle, then it defines an intervention and not the ambulance service. The comparison between services targeting generic health (ICD-10) and mental health (F0-F9) showed no remarkable differences at level 1.
Level 2: target population clarity
The definition of the target population was considered accurate for 29 terms (52%). Three terms (6%) were judged ambiguous as they referred to two not-linked population groups (defined by age and/or diagnosis) at the same time, without explaining the service specificities for each of these groups. For example, ‘nursing home: an inpatient care facility that offers care for elderly OR disabled persons’. Twenty-one terms (37%) were classified as vague as the definition of a population target was missing or it admitted too many possibilities (e.g. ‘rehabilitation facility: a center or clinic where people recovering from illness, injury or addiction are treated’). Finally, three terms (5%) were considered confusing because they did not allow a clear classification of the target population, ‘Sheltered housing for mentally ill persons: A sanctuary for temporary housing, set up to provide for the needs of homeless people/women with mental disorders, often including shelter, food, sanitation and other forms of support’. The meaning of ‘homeless people/women with a mental disorder’ is not clear and it did not match the name of the service (online Supplementary Table 1).
Level 3: service type clarity
The type of care provided by the service was judged accurate for the 23 terms (41%) that could be translated into a single MTC code. Fifteen terms (27%) were rated as ambiguous because they needed two codes or admitted several code ranges. For example, ‘polyclinic: a clinic that provides both general and specialist examinations and treatments’ was coded as outpatient non-acute health-related care: O8.1-O10 (this range is used to address different intensities of frequency of care). However, it could also be coded as ‘outpatient acute health-related care for a fixed number of hours’: O4.1. Eleven terms (20%) were considered vague because a series of codes in different main care branches were necessary for classifying the term. Looking at the term ‘rehabilitation facility: a center or clinic where people recovering from illness, injury or addiction are treated’, the definition was so wide that the classification of the term required several codes from residential, day and outpatient care branches. Finally, six (11%) definitions were judged as confusing. For example, the two terms ‘Outpatient healthcare at workplace: e.g. company physician, company nurse’ and ‘outpatient healthcare service at school: e.g. school physician, school nurse’, included examples of professionals delivering the care but not an actual definition of the type of care provided. Additionally, two different names: ‘vocational training’ and ‘individual vocational qualification’ included the same definition ‘Individual qualification training for a specific type of job’ and therefore presented a problem of synonymity (online Supplementary Table 1). The six terms classified as confusing in level 1 (different unit of analysis) required expert interpretation in level 3 (coding service type), this was expressed in italics. The comparison between service targeting generic health (ICD-10) and mental health (F0-F9) showed remarkable differences regarding accuracy (22% v. 42%).
In total, 43 terms of the basic listing (77%) presented some kind of terminological inaccuracy.
The interrater reliability analysis showed statistically significant agreements for level 1 (k: 0.642, p < 0.001) and level 3 (k: 0.778, p < 0.001), while for level 2 (k: 0.875, p < 0.001) agreement was almost perfect. Agreement on the prototype DESDE codes was also high (k: 0.746, p < 0.001) (see online Supplementary material).
Discussion
This study aimed to identify the magnitude of the bias of non-commensurability and terminological unclarity bias in HSR and health economics by applying an international classification system to a set of services used by persons experiencing a mental health condition. The results are meant to be used for further processing of service terms for the development of the multi-national, multi-sectoral costing tools in the PECUNIA project. The approach was not comparing variation country by country but identifying an international basic listing of services relevant for mental health care. Despite an extensive process of revision prior to disambiguation, only 13 terms (23%) of the 56 were judged accurate. Eleven per cent of the terms in the final listing were not services according to the definition provided by the DESDE system and the related glossary of terms. In addition, 43 terms were unclear, and could not be used for international comparability. Nearly half of the terms lacked a clear definition of the target population and around 60% had problems in the definition of service types that impeded matching them to an MTC code even though we opted for broad categories within the MTC taxonomy to facilitate matching.
Our findings indicate that the terminology problem in HSR is extensive. Surprisingly enough, health economic guidelines provide detailed information on the study designs, methods of analysis and interpretation of results but they do not mention this fundamental problem for regional or international comparability and for aggregation purposes (Simon, Reference Simon, Geddes, Andreasen and Goodwin2020). Similarly, the problem of service terminology is not even mentioned in international strategies that necessarily require comparison of service delivery such as the WHO Mental Health Gap Action Programme (mhGAP) (World Health Organization, 2008). A gap analysis cannot be conducted without a standardised description of local mental health services to allow aggregate comparisons of care systems across regions and countries. The approach presented in this paper contributes to minimise differentiation between services and to support financial incentives for effective health promotion and disease prevention. Health economic studies on services and their utilisation are key for RUM and cost calculation for efficiency (cost-effectiveness), equality (access and utilisation) and quality research.
Our experience replicated some of the findings described by Maciejewski et al. (Reference Maciejewski, Diehr, Smith and Hebert2002) A detailed analysis of services in terms of cost measurement for economic evaluations reveals the necessity and usefulness of defining services using a coding system and taxonomical criteria rather than by ‘text-based descriptions’.
Limitations
Firstly, the analysis of the terminology bias in healthcare is extremely challenging and may have problems with corroboration, even when we adapted a previously tested method (Maciejewski et al., Reference Maciejewski, Diehr, Smith and Hebert2002) and used a standardised procedure. Secondly, the findings cannot be fully generalised to all areas of healthcare. We selected mental health care as case study due to its highest complexity of care provision stretching across numerous sectors (Salvador-Carulla et al., Reference Salvador-Carulla, Haro and Ayuso-Mateos2006) including a mixture of health and social care services, the high variation and diversity in service provision (Johnson and Salvador-Carulla, Reference Johnson and Salvador-Carulla1998), and its high ambiguity in key aspects such as diagnosis (Keil et al., Reference Keil, Keuck and Hauswald2016) and treatment interventions (Castelpietraa et al., Reference Castelpietraa, Salvador-Carulla, Almborg, Fernandez and Madden2017; Castelpietra et al., Reference Castelpietra, Simon, Gutierrez-Colosía, Rosenberg and Salvador-Carulla2021). Third, we opted for a broad approach to disambiguation selecting the lower level of granularity in the MTC taxonomy and avoiding a detailed definition of the different subtypes of ambiguity and vagueness (Castelpietra et al., Reference Castelpietra, Simon, Gutierrez-Colosía, Rosenberg and Salvador-Carulla2021). The disambiguation data are related to one frame of reference (DESDE system) and cannot be generalised to other frames (e.g. Systems of Health Accounts 2.0, or SNOMED). However, the validity and the formal ontology conditions of the classification of services within these other frames have not been tested. Finally, we limited our analysis to English and did not account for the variation of terminology across other languages and contexts. In any case the reference tools ESMS and DESDE have been translated into Finnish, French, German, Italian, Polish, Portuguese, Norwegian and Spanish; and the reference coding system has been used across a wide variety of contexts in over 34 countries (Romero-Rodriguez-Alberca et al., Reference Romero-Lopez-Alberca, Gutierrez-Colosia, Salinas-Pérez, Almeda, Furst, Johnson and Salvador-Carulla2019).
Research and policy implications
Currently, the majority of comparative healthcare studies rely on official service names, without taking into account other key features of every service. Service health research, health economics, care gap analysis, quality and equality research should address terminological variability as a main source of systematic bias, particularly, but not only, in bottom-up international comparative studies. For example, cost-effectiveness and comparative effectiveness research should compare the same units of analysis of service provision, and use a common vocabulary, which is feasible with a coding system such as the one provided by DESDE. This bias is also relevant in equity studies as equal access is a critical component of equity (Raine et al., Reference Raine, Fitzpatrick, Barratt, Bevan, Black, Boaden, Bower, Campbell, Denis, Devers, Dixon-Woods, Fallowfield, Forder, Foy, Freemantle, Fulop, Gibbons, Gillies, Goulding, Grieve, Grimshaw, Howarth, Lilford, McDonald, Moore, Moore, Newhouse, O'Cathain, Or, Papoutsi, Prady, Rycroft-Malone, Sekhon, Turner, Watson and Zwarenstein2016).
Finally, an international glossary of service terms and a classification of service should be incorporated into the WHO International Family of Classifications as related classifications. Likewise, national classifications of services should provide an analysis of their semantic interoperability with international standards.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2045796022000403
Data
Supplementary materials regarding HA1 of the PECUNIA project are available in the project website https://www.pecunia-project.eu/, data regarding raw lists of terms and process of disambiguation including κ analysis are available upon request.
Acknowledgement
This manuscript was written on behalf of the PECUNIA Group. Members of the PECUNIA Group are: Medical University of Vienna: PI: Judit Simon; team members: Michael Berger, Claudia Fischer, Agata Łaszewska, Susanne Mayer, Nataša Perić; University Medical Center Hamburg-Eppendorf: PI: Hans-Helmut König; team members: Christian Brettschneider, Marie Christine Duval, Paul Hinck, Johanna Katharina Hohls, Alexander Konnopka, Louisa-Kristin Muntendorf; Corvinus University of Budapest: PI: Valentin Brodszky; team members: László Gulácsi; Maastricht University: PI: Silvia M.A.A. Evers; team members: Ruben M.W.A. Drost, Luca M.M. Janssen, Aggie T.G. Paulus, Irina Pokhilenko; Erasmus University Rotterdam: PI: Leona Hakkaart-van Roijen; team members: Kimberley Hubens, Ayesha Sajjad; Servicio de Evaluación del Canario de la Salud: PI: Pedro Serrano-Aguilar; team members: Lidia García-Pérez, Renata Linertová, Lilisbeth Perestelo-Pérez, Cristina Valcárcel-Nazco; PSICOST Scientific Association: PI: Luis Salvador-Carulla; team members: Nerea Almeda, Pilar Campoy-Muñoz, Carlos R. García-Alonso, Mencía R. Gutiérrez-Colosía, Cristina Romero-López-Alberca; Giulio Castelpietra, Jose Alberto Salinas-Pérez, London School of Economics and Political Science: PI: A-La Park; University of Bristol: PI: William Hollingworth; team members: Sian Noble, Joanna Thorn. Authors want to acknowledge the collaboration of the following public health officers in Spain: Federico Alonso (Department of Sociall Affairs, Andalusia), Álvaro Iruin, Carlos Pereira and Juan Jose Uriarte (Department of health, Basque Country); and Bibiana Prat (Department of Health, Catalonia)
Author contribution
All authors are members of the H2020 European Project ‘PECUNIA’, and contributed to the paper through face to face meeting, teleconferences and emails exchange. MGC, LSC and JS made substantial contributions to the conception and interpretation of data for the work as well as in drafting the manuscript; AK, SM, CF, PH, SE, LHvR, AP, WH, VB, HHK revised it critically for important intellectual content and approved the final version of the manuscript. JAS contributed with statistical analysis.
Financial support
The PECUNIA project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 779292.
Conflict of interest
None.
Disclaimer
The content is solely the responsibility of the authors and does not necessarily represent the official views of PECUNIA.
Open access
