Journal:Definitions, components and processes of data harmonization in healthcare: A scoping review

Full article title	Definitions, components and processes of data harmonization in healthcare: A scoping review
Journal	BMC Medical Informatics and Decision Making
Author(s)	Schmidt, Bey-Marrié; Colvin, Christopher J.; Hohlfeld, Ameer; Leon, Natalie
Author affiliation(s)	South African Medical Research Council, University of Cape Town, University of Virginia, Brown University
Primary contact	Online contact form
Year published	2020
Volume and issue	20
Page(s)	222
DOI	10.1186/s12911-020-01218-7
ISSN	1472-6947
Distribution license	Creative Commons Attribution 4.0 International
Website	https://bmcmedinformdecismak.biomedcentral.com/articles/10.1186/s12911-020-01218-7
Download	https://bmcmedinformdecismak.biomedcentral.com/track/pdf/10.1186/s12911-020-01218-7 (PDF)

Abstract

Background: Data harmonization (DH) has is increasingly being used by health managers, information technology specialists, and researchers as an important intervention for routine health information systems (RHISs). It is important to understand what DH is, how it is defined and conceptualized, and how it can lead to better health management decision-making. This scoping review identifies a range of definitions for DH, its characteristics (in terms of key components and processes), and common explanations of the relationship between DH and health management decision-making.

Methods: This scoping review identified more than 2,000 relevant studies (date filter) written in English and published in PubMed, Web of Science, and CINAHL. Two reviewers independently screened records for potential inclusion for the abstract and full-text screening stages. One reviewer did the data extraction, analysis, and synthesis, with built-in reliability checks from the rest of the team. We developed a narrative synthesis of definitions and explanations of the relationship between DH and health management decision-making.

Results: Of the 181 studies ultimately included in this scoping review, 61 included synthesis definitions and concepts of DH in detail. From these, we identified six common terms for data harmonization: "record linkage," "data linkage," "data warehousing," "data sharing," "data interoperability," and "health information exchange." We also identified nine key components or characteristics of data harmonization: it involves (a) multi-step processes; (b) integration and harmonization of different databases; (c) the use of two or more databases; (d) the use of electronic data; (e) pooling of data using unique patient identifiers; (f) different types of data; (g) data found within and across different departments and institutions at facility, district, regional, and national levels; (h) different types of technical activities; and (i) a specific scope. The relationship between DH and health management decision-making is not well-described in the literature. Several studies mentioned health providers’ concerns about data completeness, data quality, terminology, and coding of data elements as barriers to data use for clinical decision-making.

Conclusion: To our knowledge, this scoping review was the first to synthesize definitions and concepts of DH and address the causal relationship between DH and health management decision-making. Future research is required to assess the effectiveness of data harmonization on health management decision-making.

Keywords: data harmonization, health information exchange, health information system, scoping review

Background

Data harmonization (DH) in healthcare is a digital, technology-based innovation that can potentially help routine health information systems (RHISs) function at their best. It can help organize and integrate large databases containing routine health information. Designing, developing, and implementing DH interventions has the potential to strengthen aspects of the health system, by enhancing RHISs to a state of high-quality and relevant information that can support decisions, actions, and changes across all components and levels of the health system. When RHISs are functioning properly, they can help health practitioners and managers identify and close gaps in health service delivery, as well as inform their planning, implementation, and monitoring of interventions. They can also help address problems related to using different variables and indicators for collecting, analyzing, and reporting health information across healthcare administration and management programs, which is common in low-and-middle-income (LMIC) settings. Other challenges to effective RHIS functioning include the production of poor-quality data that cannot easily be exchanged, as well as programmatic fragmentation across levels of the health system, which can result in the duplication and excessive production of data.

Lack of standardized data production processes, fragmentation of databases, and errors and duplication in data production are only some of the challenges of RHISs, which may, at first glance be categorized as technical challenges. Solutions to such apparently technical challenges include introducing new data forms, setting up warning systems to detect potential errors, and developing algorithms for integrating different databases.

However, DH interventions for RHISs may not be used effectively if data production and utilization processes are viewed as merely technical. Given that RHISs are embedded in complex health systems, DH interventions to improve RHIS functions are also influenced by the broader setting, in which dynamic and complex social and technical factors interact. There is a need to consider the influence of social factors as well. These may include people’s competencies in dealing with new data production processes, institutional values about data utilization, and existing relationships between data producers and decision-makers.

There is growing recognition that the development and implementation of DH interventions occurs in multiple technical and social contexts, and that DH interventions may differ in definition, purpose, and intended outcomes. As such, various terms are used to describe interventions with similar aims and activities to data harmonization. For example, terms such as "record linkage," "data warehousing," "data sharing," and "health information exchange" are all used to describe data harmonization-type activities; it is not always clear to which extent these interventions are similar in practice, scope, and relevance. The use of multiple terms may not be a problem in itself, but a common understanding of the components and processes will bring more clarity about what actually constitutes "data harmonization," and it will make it easier to compare and appraise the relevance and usefulness of DH interventions across settings.

Although DH has the potential to enhance RHISs, it is still unclear whether or how it affects health management decision-making. In some cases, DH interventions may not directly improve management decision-making, especially when interventions are more focused on the technical aspects of data production and less on the organizational and behavioral aspects of data use for decision-making. The scope of this review is to therefore understand the different ways in which DH is defined, to identify its components and processes, and to describe whether or how DH can affect health management decision-making. Greater clarity about the range of definitions, components, and processes of DH interventions, as well as its intended outcomes, can help to better evaluate DH's relevance, usefulness, and impact.

Methods

This scoping review was conducted according to the methods outlined by Arksey and O’Malley. They recommend a process that is “not linear but, requiring researchers to engage with each stage in a reflexive way” to achieve both "in-depth and broad" results. This review followed the standard steps for systematic reviews: identifying the research question, identifying relevant studies, selecting studies for inclusion, extracting data, and synthesizing data. These are detailed in our published study protocol.

Study objectives

This scoping review appraised the definitions, components, and processes of data harmonization activities, and it provided a broad explanation of the relationship between data harmonization interventions and health management decision-making. The specific objectives are:

1. to identify and provide an overview of the key components and processes of data harmonization studies;

2. to identify and synthesize the various definitions of data harmonization in healthcare; and

3. to describe the relationship between data harmonization interventions and health management decision-making.

We took a stepped approach in addressing these objectives. All included studies were used to address the first objective. To address the second objective, we sampled studies that were using alternative terms for DH interventions and used those to identify, synthesize, and compare similarities and differences in definitions. While executing these two objectives, we identified a smaller number of studies that contributed to the third objective.

Identifying relevant studies

Eligibility criteria

Peer-reviewed studies and gray literature were considered eligible for inclusion into the scoping review if they provided a definition or description of DH, and or, a more detailed conceptual explanation (in the form of a model, framework, or process) of a DH intervention. Additionally, studies were eligible if they provided an explanation of the causal relationship between DH and health management decision-making (such as through improved quality and accessibility of harmonized information for management and/or the use of harmonized health information for management decision-making). We considered any studies concerned with different technical activities of DH (such as linking, merging, cleaning and transferring). After screening, only studies for which we could access full-text articles were eligible for inclusion in the review.

Search strategy

A systematic literature search was conducted in PubMed, CINAHL, and Web of Science for eligible studies from January 1, 2000 to September 30, 2018. We limited our search to as far back as 2000, as digital technology-based innovations began during this period (such as health information exchange) in high-income countries (predominantly in the United States of America), and researchers and health system managers in LMICs became more interested in the integration of large digital databases. We present the search strategy in the study protocol. Based on preliminary searches, we anticipated that these databases would yield the highest results. The search strategies included a combination of keywords and Medical Subject Headings (MeSH) related to data harmonization (concept A) and health information systems (concept B). There were no geographic restrictions, but for logistical reasons of time and resources, we only searched for English studies.

Selecting studies for inclusion

Screening records

The first reviewer (BS) conducted all the searches with the help of a librarian and collated the records in the EndNote reference management program, where duplicates were removed. Two reviewers, (BS) and second reviewer (AH), then independently screened the records (titles and abstracts) to assess eligibility for full-text review. BS and AH resolved conflicts that emerged at this stage by talking through the inclusion criteria and arriving at a joint decision.

The full-texts of potentially eligible studies were retrieved and assessed by the two reviewers (BS and AH). Final inclusion into the review was based on whether at a minimum the study had a definition or description of a DH intervention or referred to its relationship with health management decision-making. The first reviewer read all full-texts and the second reviewer only read a sample (roughly a third) of the full-texts to verify the first reviewer’s decision about inclusion. BS and AH disagreed on four studies, and after discussion, agreed to exclude the studies.

After finalizing screening, the two reviewers then mapped out the characteristics of included studies in an Excel spreadsheet. They recorded the name of the first author; the date; the type of study (primary, review, conceptual, commentary); the term used for the intervention they described (DH or alternative); the country in which the study took place; the level at which the intervention was implemented (frontline, management, research); and indicated whether or not there was a conceptual model, framework, diagram, or process description of DH and health management decision-making. This detailed mapping of study characteristics was useful for informing sampling options for the second and third objectives.

Sampling of studies

A scoping review aims to map the literature on a particular topic rather than to provide an exhaustive explanation of a particular phenomenon of interest. Thus, the number of included studies was expected to be high in the scoping reviews. To manage the high numbers for a scoping review such as this one (where the aim was to provide definitions and concepts), it was necessary to make use of a qualitative sampling approach. A qualitative sampling approach for this review aimed for variation and depth rather than an exhaustive sample; reviewing too large a number of studies can impair the quality of the analysis and synthesis. We used two types of purposive sampling techniques called maximum variation sampling and theoretical sampling. These techniques were used to identify both the range, variation, and similarities or differences in definitions and concepts and intervention descriptions (as per the second objective), and to provide a rich synthesis of explanations of causal relationships between DH and health management decision-making (as per the third objective). For the first objective, we did not apply a sampling strategy. Thus, we included all the studies that at a minimum provided a definition or description of a DH intervention.

Data extraction

BS extracted data for the first objective from all the included studies (n = 181). AH independently extracted data from 81 (45%) of included studies to verify data extraction done by the first reviewer. We used an MS Excel spreadsheet for data extraction, as presented in Figure 1. AH and BS extracted a few studies before clarifying the items in the spreadsheet. Once data extraction was complete, the reviewers were able to filter according to the individual items extracted in order to synthesize and compare the studies. Given the objectives of the scoping review, we did not extract any information relevant to conducting risk of bias or quality assessment. Not conducting risk of bias or quality assessment is consistent with scoping reviews of similar aims and methodological approaches.

Figure 1. Extract of the Excel data extraction form

Data synthesis: Collating, summarizing, and reporting findings

The first reviewer (BS) conducted data analysis using manual coding and the filter option in MS Excel. Another reviewer (NL) reviewed the data analysis work on an ongoing basis as an additional quality check. For the first objective, we conducted a numerical analysis to provide an overview of the characteristics of all the included studies. For the second objective, we conducted a qualitative analysis to provide a narrative synthesis of the different DH definitions and concepts, and to identify different components or activities that are considered part of the DH processes. For the third objective, we reviewed data related to intentions, suggestions, and/or explanations of how DH may lead to improved health management decision-making. We extracted and analyzed data relevant to the second and third objectives at the same time. We first created a list of all the different terms used to describe DH interventions and then compared definitions across alternative terms by looking for similarities or differences in the definitions or descriptions of DH interventions. We then coded key components, processes, and outcomes of DH interventions and the factors reported as important in the relationship between DH and health management decision-making.

The findings are structured according to three themes matching the three study objectives: an overview of the key characteristics of included studies, alternative terms and definitions of DH, and a narrative synthesis of the relationship between DH and health management decision-making.

Reflexivity

Throughout the review, the authors were aware of their own positions and reflected on how these could influence the study design, search strategy, inclusion decisions, data extraction, and analysis, as well as the synthesis and interpretation of the findings. The review authors are trained in anthropology, epidemiology, health systems, and evidence synthesis research. The first author was involved in participant observation of an innovative DH project in the Western Cape Department of Health in South Africa as part of her doctoral research, where she grappled with questions that informed the objectives of this review. Three of the authors (BS, AH, and NL) were involved in a Cochrane Collaboration systematic review on RHIS interventions when this scoping review was conceptualized, so they were familiar with some of the health information systems (HIS) literature and had some appreciation for the conceptual and methodological complexities of studying the field of health information management. This experience informed the way the first author developed the search strategy. She used an iterative approach to narrow down the search as much as possible because of her prior knowledge that it was difficult to balance sensitivity and specificity when developing a search strategy for HIS literature that is often multi-disciplinary in nature.

Results

Results of the search

Figure 2 shows a PRISMA diagram of the search results. We screened a total of 1,331 records: 1,232 titles and abstracts identified from searching three electronic databases, and 99 from screening for a Cochrane Collaboration systematic review assessing the effectiveness of RHIS interventions on health systems management, as well as gray literature. Almost a quarter (289 of 1,331) were deemed potentially eligible for full-text screening. We accessed full-texts for 275 studies, and of those, 181 were included in the scoping review for the first objective. We excluded 94 full-text articles because they did not meet the minimum criteria; that is, they did not provide a definition or description of a DH intervention or activity. We sampled 61 studies from the 181 for the second and third objectives. We arrived at 61 studies by including all reviews (systematic or literature reviews) and all studies (irrespective of the type of study) that also had a process description, conceptual framework, or a theory of a DH intervention (that is, in addition to the minimum criteria for the first objective).

Figure 2. PRISMA diagram of eligible studies

An overview of key characteristics of data harmonization studies

A total of 181 studies were included into this scoping review for the first objective (see Table 1). Given the high number of included studies, we decided to only map the following key characteristics of those studies: first author, date, type of study, intervention term (DH or alternative), country, and level of the health care system. Most included studies (126 of 181) were either primary studies assessing various aspects of developing and implementing DH interventions (quantitative studies n = 86), or patient, providers, or stakeholders’ perspectives (qualitative studies n = 34), or a combination of both (mixed methods studies n = 6).

Table 1. Characteristics of included studies (n = 181)   CDE = clinical data exchange, CIE = clinical information exchange, DH = data harmonization, DL = data linkage, DS = data sharing, DW = data warehouse, EHR = electronic health record, HDE = health data exchange, HIE = health information exchange, IE = information exchange, IO = interoperability, LMICs = low-to-middle-income countries, RL = record linkage Study name Date Type of study Intervention term Country Level of the healthcare system Commentary Burris 2017 Commentary HIE USA Frontline: hospitals Figge 2010 Commentary HIE USA Management McIlwain 2009 Commentary HIE USA Management Murphy 2010 Commentary HIE USA Management Overhage 2007 Commentary HIE USA Management Rudin 2010 Commentary HIE USA Frontline: workers Conceptual Boyd 2014 Conceptual RL Australia Research Carr 2013 Conceptual HIE USA Frontline: hospitals Cimino 2014 Conceptual HIE USA Management Deas 2012 Conceptual HIE USA Management Del Fiol 2015 Conceptual HIE USA Frontline: prisons, hospitals Dimitropoulos 2009 Conceptual HIE USA Management Downs 2010 Conceptual HIE USA Management Feldman 2017 Conceptual HIE USA Management Frisse 2010 Conceptual HIE USA Frontline: patients, workers Frisse 2008 Conceptual HIE USA Frontline: organizations Frohlich 2007 Conceptual HIE USA Management Godlove 2015 Conceptual HIE USA Fronline: patients Greene 2016 Conceptual HIE USA Management Grossman 2008 Conceptual HIE USA Management Haarbrandt 2016 Conceptual DW USA Management Hu 2007 Conceptual DS USA Management Jones 2012 Conceptual DS USA Management Kuperman 2013 Conceptual HIE USA Management Langabeer 2016 Conceptual HIE USA Management Liu 2011 Conceptual HIE China Management McDonald 2009 Conceptual HIE USA Management McMurray 2015 Conceptual HIE USA Management Miller 2014 Conceptual HIE USA Frontline: hospitals Nelson 2016 Conceptual HIE USA Frontline: prisons, hospitals Politi 2014 Conceptual HIE n/a Management Ranade-Kharkar 2014 Conceptual HIE USA Management Shapiro 2016 Conceptual HIE USA Frontline: workers, organizations Shapiro 2006 Conceptual HIE USA Management Thorn 2013 Conceptual HIE USA Frontline: healthcare workers Thorn 2014 Conceptual HIE USA Frontline: healthcare workers Vest 2010 Conceptual HIE USA Management Williams 2012 Conceptual HIE USA Management Yaraghi 2014 Conceptual HIE USA Management Zafar 2007 Conceptual HIE USA Management Zaidan 2015 Conceptual HIE Malasia Management Primary studies Abramson 2012 Primary: quantitative EHR, HIE USA Frontline: hospitals Adjerid 2011 Primary: quantitative HIE USA Management: states Adler-Milstein 2011 Primary: quantitative HIE USA Frontline: organizations Adler-Milstein 2013 Primary: quantitative HIE USA Management: organizations Adler-Milstein 2016 Primary: quantitative HIE USA Management Alexander 2015 Primary: qualitative HIE USA Frontline: healthcare workers Alexander 2016 Primary: qualitative HIE USA Frontline: healthcare workers Ancker 2012 Primary: quantitative HIE USA Frontline: consumers Bahous 2016 Primary: quantitative HIE Israel Frontline: hospitals Bailey 201 Primary: quantitative HIE USA Frontline: hospitals Ben-Assuli 2013 Primary: quantitative HIE USA Frontline: hospitals Boockvar 2017 Primary: quantitative HIE USA Frontline: hospitals Butler 2014 Primary: qualitative HIE USA Frontline: prisons, communities Campion 2012 Primary: quantitative HIE USA Frontline: healthcare workers Campion 2013 Primary: quantitative HIE USA Frontline: communities Campion 2013 Primary: quantitative HIE USA Frontline: clinics, hospitals Campion 2014 Primary: quantitative CDE USA Frontline: organizations Carr 2014 Primary: quantitative HIE USA Frontline: hospitals Carr 2016 Primary: quantitative HIE USA Frontline: hospitals Cochran 2015 Primary: qualitative HIE USA Frontline: clinics, communities Cross 2016 Primary: qualitative DM USA Management: organizations Dalan 2010 Primary: qualitative DM USA Research Dimitropoulos 2011 Primary: quantitative HIE USA Frontline: consumers Dixon 2011 Primary: quantitative HIE USA Frontline: laboratories Dixon 2013 Primary: quantitative HIE USA Frontline: hospitals Downing 2017 Primary: quantitative HIE USA Management: policies Dullabh 2013 Primary: qualitative HIE USA Management: organizations Elysee 2017 Primary: quantitative HIE, IO USA Frontline: hospitals Foldy 2007 Primary: quantitative HIE USA Management: organizations Fontaine 2010 Primary: qualitative HIE USA Frontline: primary healthcare French 2016 Primary: quantitative HIE USA Management: organizations Fricton 2008 Primary: quantitative HIE USA Frontline: patients, workers Frisse 2012 Primary: quantitative HIE USA Frontline: organizations Furukawa 2013 Primary: quantitative HIE USA Frontline: hospitals Furukawa 2014 Primary: quantitative HIE USA Frontline: healthcare workers Gadd 2011 Primary: quantitative HIE USA Frontline: healthcare workers Garg 2014 Primary: quantitative HIE USA Frontline: hospitals Gill 2001 Primary: quantitative DL South Africa Frontline: patients, disease Grinspan 2013 Primary: quantitative HIE USA Frontline: patients Grinspan 2014 Primary: quantitative HIE USA Frontline: healthcare workers Grinspan 2015 Primary: quantitative HIE USA Frontline: patients Hassol 2014 Primary: quantitative HIE USA Frontline: healthcare workers Herwehe 2012 Primary: quantitative HIE USA Frontline: healthcare workers Hincapie 2011 Primary: qualitative HIE USA Frontline: healthcare workers Holman 2008 Primary: quantitative DL USA Frontline: organizations, research Hypponen 2014 Primary: quantitative HIE Finland Frontline: healthcare workers Ji 2017 Primary: quantitative HIE Korea Frontline: hospitals Johnson 2011 Primary: mixed HIE USA Frontline: hospitals Jung 2015 Primary: quantitative HIE USA Frontline: hospitals Kaelber 2013 Primary: quantitative HIE USA Frontline: hospitals Kierkegaard 2014 Primary: qualitative HIE USA Frontline: healthcare workers Kierkegaard 2014 Primary: qualitative HIE USA Management Kim 2012 Primary: qualitative HIE Korea Management Knaup 2006 Primary: quantitative DS Germany Frontline: hospitals Kralewski 2012 Primary: qualitative CIE USA Frontline: organizations, workers Laborde 2011 Primary: quantitative HIE USA Frontline: hospitals Lee 2012 Primary: quantitative HIE South Korea Frontline: healthcare workers Li 2011 Primary: quantitative CDE Japan & China Frontline: organizations Liu 2010 Primary: qualitative DH China Management Lobach 2007 Primary: quantitative HIE USA Management Maenpaa 2012 Primary: quantitative HIE Finland Frontline: hospitals Maiorana 2012 Primary: mixed HIE USA Frontline: workers, disease Marinez 2015 Primary: quantitative HIE USA Frontline: hospitals Massoudi 2016 Primary: qualitative HIE USA Frontline: organizations Masterbroek 2016 Primary: quantitative HIE Netherlands Frontline: healthcare workers Mastebroek 2017 Primary: qualitative HIE Netherlands Frontline: patients Mastebroek 2017 Primary: qualitative HIE Netherlands Frontline: patients Matsumoto 2017 Primary: qualitative HIE USA Frontline: workers, hospitals Medford-Davis 2017 Primary: quantitative HIE USA Frontline: pateients, hospitals Mello 2018 Primary: qualitative HIE USA Management: policies Merrill 2013 Primary: quantitative HIE USA Fronline: managers Messer 2012 Primary: mixed HIE USA Frontline: clinics, organizations Miller 2012 Primary: qualitative HIE USA Frontline: consumers, organizations Miller 2017 Primary: quantitative HIE USA Frontline: disease, workers Moore 2012 Primary: quantitative HIE USA Fronline: workers, hospitals Motulsky 2018 Primary: quantitative HIE Canada Fronline: workers Myers 2012 Primary: qualitative HIE USA Frontline: disease, workers Obeidat 2014 Primary: quantitative IE Jordan Frontline: hospitals O'Donnell 2011 Primary: quantitative HIE USA Frontline: workers Onyile 2013 Primary: quantitative HIE USA Frontline: patients Opoku-Agyeman 2016 Primary: quantitative HIE USA Frontline: hospitals Overhage 2017 Primary: quantitative HIE USA Management Ozkaynak 2013 Primary: qualitative HIE USA Frontline: hospitals, workers Park 2013 Primary: quantitative HIE South Korea Frontline: clinics, hospitals Park 2015 Primary: quantitative HIE South Korea Frontline: clinics, hospitals Patel 2011 Primary: quantitative HIE USA Frontline: clinics, hospitals Politi 2015 Primary: quantitative HIE Israel Frontline: hospitals Ramos 2014 Primary: qualitative HIE USA Frontline: patients Ramos 2016 Primary: mixed HIE USA Frontline: patients Reis 2016 Primary: quantitative HDE USA Frontline: hospitals Richardson 2014 Primary: qualitative HIE USA Frontline: organizations, workers Ross 2010 Primary: qualitative HIE USA Frontline: clinics Ross 2013 Primary: quantitative HIE USA Frontline: workers, clinics, hospitals Rudin 2009 Primary: qualitative HIE USA Frontline: healthcare workers Rundall 2016 Primary: qualitative HIE USA Frontline: policy makers, leaders Saef 2014 Primary: quantitative HIE USA Frontline: hospitals Santos 2017 Primary: quantitative HIE Brazil Frontline: clinics, hospitals Shade 2012 Primary: quantitative HIE USA Frontline: clinics, hospitals Shank 2012 Primary: quantitative HIE USA Frontline: healthcare workers Shapiro 2007 Primary: quantitative HIE USA Frontline: healthcare workers Shapiro 2013 Primary: quantitative HIE USA Frontline: hospitals Sicotte 2010 Primary: qualitative HIE Canada Frontline: workers, hospitals Sprivulis 2007 Primary: quantitative HIE Australia Frontline: workers, organizations Squire 2002 Primary: mixed HIE USA Frontline: healthcare workers Sridhar 2012 Primary: quantitative HIE USA Frontline: hospitals Thornewill 2011 Primary: mixed HIE USA Frontline: consumers, organizations Unertl 2012 Primary: qualitative HIE USA Frontline: clinics, hospitals Vest 2009 Primary: quantitative HIE n/a Frontline: workers, patients Vest 2010 Primary: qualitative HIE USA Frontline: hospitals Vest 2011 Primary: quantitative HIE USA Frontline: patients, hospitals Vest 2013 Primary: quantitative HIE USA Frontline: hospitals Vest 2013 Primary: qualitative HIE USA Management: policy makers Vest 2014 Primary: quantitative HIE USA Frontline: patients, hospitals Vest 2014 Primary: quantitative HIE USA Frontline: hospitals Vest 2015 Primary: qualitative HIE USA Frontline: consumers, organizations Vest 2017 Primary: qualitative HIE USA Frontline: consumers, organizations Vest 2017 Primary: quantitative HIE USA Frontline: consumers, organizations Vreeman 2008 Primary: quantitative HIE USA Frontline: laboratory, radiology Webn 2010 Primary: quantitative HIE USA Frontline: patients Winden 2014 Primary: quantitative HIE USA Frontline: clinical care Wright 2010 Primary: quantitative HIE USA Frontline: healthcare workers Yeager 2014 Primary: qualitative HIE USA Frontline: consumers Yeaman 2015 Primary: quantitative HIE USA Frontline: hospitals Zech 2015 Primary: quantitative HIE USA Frontline: patients, organizations Zech 2016 Primary: quantitative HIE USA Frontline: patients, organizations Zhu 2010 Primary: quantitative HIE USA Research Study protocol Dixon 2013 Protocol: mixed HIE USA Frontline: organizations Reviews Akhlaq 2016 Review HIE LMICs Management: countries Dixon 2010 Review HIE USA Research Esmaeilzadeh 2016 Review HIE n/a Management: policies Esmaeilzadeh 2017 Review HIE n/a Frontline: patients Fontaine 2010 Review HIE n/a Frontline: primary healthcare Hopf 2014 Review DL n/a Frontline: healthcare workers Kash 2017 Review HIE n/a Frontline: hospitals Mastebroek 2014 Review HIE USA Frontline: disease, workers Parker 2016 Review HIE USA Research Rahurkar 2015 Review HIE n/a Frontline: hospitals Rudin 2014 Review HIE USA Frontline: clinical care Sadoughi 2018 Review HIE n/a Management Vest 2012 Review HIE n/a Management

Of the 181 included studies, nine were not country-specific (these were global reviews), 151 were from the USA, and the rest were from other countries (specifically Australia, Brazil, Canada, China, Finland, Germany, Israel, Japan, Jordan, Korea, Malaysia, Netherlands, South Africa, and South Korea). In terms of the level of the health care system, 128 studies were on a DH intervention or activity that was concerned with the frontline level (health service providers), 48 studies were concerned with health system factors or policy-related activities at the managerial level, and five studies focused on DH interventions specifically for research purposes. Most studies (92%) used the term health information exchange (HIE), while the remaining studies (8%) used a variety of terms to describe various DH interventions and activities, specifically, record linkage, data mining, data linkage, data warehousing, data sharing, and data harmonization.

Definitions, components, and processes of data harmonization

In this subsection, we first discuss the alternative terms and definitions of DH, and then we summarize key components and processes of DH using studies sampled from the 61 studies identified for the second and third objectives. Table 2 presents identifying details of the 61 studies, including the type of study design, the intervention terms, the country, the level of the health care system, and the purpose of the study. These studies were concerned with the challenges and opportunities of DH, the barriers and facilitators of DH, the various factors affecting DH (such as technical and financial factors), the outcomes of DH (such as patient safety and quality of care), and the privacy and security issues of patient information.

Table 2. Characteristics of sampled studies (n = 61)   CDE = clinical data exchange, CIE = clinical information exchange, DH = data harmonization, DL = data linkage, DS = data sharing, DW = data warehouse, EHR = electronic health record, HDE = health data exchange, HIE = health information exchange, IE = information exchange, IO = interoperability, LMICs = low-to-middle-income countries, RL = record linkage Study name Date Type of study Intervention term Country Level of the healthcare system Purpose of the study Akhlaq 2016 Review HIE LMICs Management: countries Barriers and facilitators of HIE Boyd 2014 Conceptual RL Australia Research Functions of record linkage Burris 2017 Commentary HIE USA Frontline: hospitals Benefits of HIE Campion 2012 Primary: quantitative HIE USA Frontline: healthcare workers Push and pull of HIE Cimino 2014 Conceptual HIE USA Management Debates around consumer-mediated HIE Dalan 2010 Primary: qualitative DM USA Research Possibilities for clinical data mining and research Dimitropoulos 2009 Conceptual HIE USA Management Privacy and security of interoperable HIE Dixon 2010 Review HIE USA Research Costs, effort and value of HIE Downing 2017 Primary: quantitative HIE USA Management: policies Relationship between HIE and organisational HIE policy decisions Downs 2010 Conceptual HIE USA Management Improving laboratory services through HIE Dullabh 2013 Primary: qualitative HIE USA Management: organizations Experience of HIE implementation Elysee 2017 Primary: quantitative HIE, IO USA Frontline: hospitals Relationship between HIE, interoperability and medication reconciliation Esmaeilzadeh 2016 Review HIE n/a Management: policies HIE assimilation and patterns for policy Esmaeilzadeh 2017 Review HIE n/a Frontline: patients Patients’ perceptions of HIE Fontaine 2010 Review HIE n/a Frontline: primary healthcare HIE for primary healthcare practices Fontaine 2010 Primary: qualitative HIE USA Frontline: primary healthcare Barriers and facilitators of HIE in primary care practices Frisse 2010 Conceptual HIE USA Frontline: patients, workers Impact of HIE on patient-provider relationships Gadd 2011 Primary: quantitative HIE USA Frontline: healthcare workers Users’ perspectives on the usability of a regional HIE Gill 2001 Primary: quantitative DL South Africa Frontline: patients, disease Linkage of non-communicable diseases data Greene 2016 Conceptual HIE USA Management Technical and financial aspects of HIE Grossman 2008 Conceptual HIE USA Management Barriers to stakeholder participation in HIE Haarbrandt 2016 Conceptual DW USA Management Approaches for a clinical data warehouse Herwehe 2012 Primary: quantitative HIE USA Frontline: healthcare workers Implementation of an electronic medical record and HIE Hincapie 2011 Primary: qualitative HIE USA Frontline: healthcare workers Physicians’ opinions of HIE Hopf 2014 Review DL n/a Frontline: healthcare workers Healthcare professionals’ views of linking routinely collected data Hu 2007 Conceptual DS USA Management Challenges in implementing an infectious disease information sharing and analysis system Hypponen 2014 Primary: quantitative HIE Finland Frontline: healthcare workers User experiences with different regional HIE Ji 2017 Primary: quantitative HIE Korea Frontline: hospitals Technology and policy changes for HIE Jones 2012 Conceptual DS USA Management An overview of electronic data sharing Kash 2017 Review HIE n/a Frontline: hospitals Hospital readmission reduction and the role of HIE Kierkegaard 2014 Primary: qualitative HIE USA Frontline: healthcare workers Applications of HIE information to public health practice Kierkegaard 2014 Primary: qualitative HIE USA Management Health practitioners’ needs and HIE Kuperman 2013 Conceptual HIE USA Management Potential unintended consequences of HIE Liu 2010 Primary: qualitative DH China Management Defining data elements for HIE Maiorana 2012 Primary: mixed HIE USA Frontline: workers, disease Trust, confidentiality, and acceptability of sharing HIV data for HIE Massoudi 2016 Primary: qualitative HIE USA Frontline: organizations HIE for clinical quality measures Mastebroek 2014 Review HIE USA Frontline: disease, workers HIE in general care practice for people with disabilities Mastebroek 2016 Primary: quantitative HIE Netherlands Frontline: healthcare workers Priority setting and feasibility of HIE Masterbroek 2017 Primary: qualitative HIE Netherlands Frontline: patients Experiences of people with intellectual disabilities in HIE Matsumoto 2017 Primary: qualitative HIE USA Frontline: workers, hospitals HIE in managing hospital services Parker 2016 Review HIE USA Research The use of HIE in supporting clinical research Politi 2014 Conceptual HIE n/a Management Use patterns of HIE Rahurkar 2015 Review HIE n/a Frontline: hospitals Impact of HIE on cost, use and quality of care Ramos 2016 Primary: mixed HIE USA Frontline: patients HIE consent process in an HIV clinic Ranade-Kharkar 2014 Conceptual HIE USA Management Improving data quality integrity through HIE Ross 2010 Primary: qualitative HIE USA Frontline: clinics Motivators, barriers, and potential facilitators of adoption of HIE Rudin 2014 Review HIE USA Frontline: clinical care Use and effect of HIE on clinical care Rundall 2016 Primary: qualitative HIE USA Frontline: policy makers, leaders Information-sharing needs and HIE Sadoughi 2018 Review HIE n/a Management Quality and cost-effectiveness, and the rates of HIE adoption and participation Santos 2017 Primary: quantitative HIE Brazil Frontline: clinics, hospitals HIE for continuity of maternal and neonatal care Shade 2012 Primary: quantitative HIE USA Frontline: clinics, hospitals HIE for quality and continuity of HIV care Shapiro 2016 Conceptual HIE USA Frontline: workers, organizations HIE in emergency medicine Shapiro 2006 Conceptual HIE USA Management Approaches to patient HIE and their impact on emergency medicine Vest 2010 Conceptual HIE USA Management Challenges and strategies for HIE Vest 2012 Review HIE n/a Management National and international approaches of health information exchange Vest 2015 Primary: qualitative HIE USA Frontline: consumers, organizations HIE to change cost and utilisation outcomes Williams 2012 Conceptual HIE USA Management Strategies to advance HIE Yaraghi 2014 Conceptual HIE USA Management Professional and geographical network effects on HIE growth Yeager 2014 Primary: qualitative HIE USA Frontline: consumers Factors related to HIE participation and use Zaidan 2015 Conceptual HIE Malasia Management Security framework for nationwide HIE Zhu 2010 Primary: quantitative HIE USA Research Facilitating clinical research through HIE

Alternative terms and definitions of data harmonization

For the second objective, we describe alternative terms and definitions of DH. We sampled 21 studies from the 61 studies identified for objectives two and three. The alternative terms and definitions are summarized in Table 3. During data analysis we realized that most studies (53 of 61) used the term "health information exchange," with similar definitions. We sampled 13 of the 53 studies to contribute to the composite HIE definition. These 13 studies were chosen to represent the term HIE because they were review studies, and we assumed that reviews provided synthesized definitions of interventions. Using maximum variation sampling, we included eight more studies (21 studies in total; see Table 3), because they provided a range of different terms for DH activities, besides the term HIE.

Table 3. Alternative terms and definitions of data harmonization interventions; where multiple studies used a similar definition, the review authors synthesized the data from similar definitions into the composite definition for each term, as presented in this table. Contributing studies Definition Liu 2010 Data harmonization is the process of integrating a person's life-long health data that are distributed in inhomogeneous information systems through identifying, reviewing, matching, redefining, and standardizing information. This process involves two steps. Firstly, identify whether all the information necessary for a single electronic platform is available in existing systems, where the information is, and how the information is defined and formatted. And secondly, make the heterogeneous information recorded by various systems consistent or at least comparable with one another by reviewing, matching, redefining, and standardizing each data item. Boyd 2014 Record linkage is the process of bringing together data relating to the same individual from within and between different datasets. When a unique person-based identifier exists, linkage can be achieved by simply merging datasets on the identifier. However, when a person-based identifier does not exist, then some other form of data matching or record linkage is required for integrating data. Gill 2001, Hopf 2014 Data linkage can be used to construct a register for a specific geographic area and disease (e.g., a district non-communicable disease register). Linkage of routine datasets by unique patient identifiers can provide an opportunity for identifying adverse drug reactions and tracking exposed individuals in real time. Routine data linkage can also enable the creation of exposure cohorts to monitor long-term outcomes and enable a more efficient screening for adverse drug reactions due to an ever-increasing data pool. Haarbrandt 2016 Data warehousing is the process of establishing specialized databases by integrating information systems (the authors specifically referred to hospital information systems) to facilitate secondary use of data. Clinical data warehouses are generally built on one of two predominant architectural paradigms: either, data is directly extracted, transformed, and loaded from applications systems and databases into a data mart (an integrated view over a defined subject), or it is stored in a centralized data repository from which data marts can be established. Both approaches rely on a process to extract data from sources, transform it appropriately, and load it (or copy it) to a specific database. Hu 2007, Jones 2012 Data sharing is based on the need for a more robust method for defining and sharing expected and actual patient outcomes. It must leverage existing informatics tools since a great deal of patient-specific information is already available in medical record, billing, and administrative systems. One type of data sharing system is an infectious disease informatics (IDI) system. An IDI system should encompass sophisticated algorithms for the automatic detection of emerging disease patterns and the identification of probable threats or events. It should also have advanced computational models that overlay health data for spatial–temporal analysis to support public health professionals’ analysis tasks. Elysee 2017 Data interoperability is one of two functionalities of an advanced electronic health record (EHR). The first function is health information exchange, which is the ability to electronically share patient-level information among unaffiliated providers across organizational boundaries. The second function is interoperability, which is the ability to produce standardized patient-level health information that can be integrated into unaffiliated health care providers’ EHRs. Akhlaq 2016, Dixon 2010, Esmaeilzadeh 2016, Esmaeilzadeh 2017, Fontaine 2010, Hopf 2014, Kash 2017, Mastebroek 2014, Parker 2016, Rahurkar 2015, Rudin 2014, Sadoughi 2018, Vest 2012 Health information exchange (HIE) is a type of health information technology (HIT) intervention. It involves the electronic mobilization of clinical and administrative data or information within or across data repositories or organizations in a community or region, between various systems as per recognized standards. This is to ensure that the HIE maintains the authenticity and accuracy of the information being exchanged, thereby enabling stakeholders to make informed decisions to enhance healthcare quality and delivery of patients and populations. Sharing clinical data can potentially improve patient safety, care coordination, and quality of care and efficiency; facilitate public health efforts; and reduce mortality and healthcare costs. Lastly, HIE involves multi-stakeholder organizations that oversee the business, operational, and legal issues involved in the exchange of information.

There is overlap between the terms and definitions. Definitions for data harmonization, record linkage, and data warehousing explicitly state that these interventions involve a process of having to integrate different or "homogeneous" databases or information systems. Data linkage and record linkage both focus on "linkage" as a core activity in combining different databases using a unique patient identifier. HIE is described as a key outcome of data interoperability, that is, where the focus is on technical linkage of different electronic databases. Data sharing, where the focus is on data accessibility and use, is described as a key outcome of HIE.

Based on the literature, we identified elements found in the various definitions of data harmonization. DH is considered a multi-step process with a range of activities (such as identifying, reviewing, matching, redefining, and standardizing information). Data harmonization interventions rely on interoperability between databases and systems, which means copying standardized patient-level data into a separate repository. Data linkage and record linkage are activities of a broader intervention (data harmonization), using mechanisms (such as unique patient identifiers) for integrating large datasets. Data warehousing is concerned with extracting, transforming, and loading large datasets using information technology (IT) platforms, application systems, and data displays (data marts or data dashboards). Data sharing (through the accessing and exchanging of electronic health information), can be considered an outcome of HIE interventions. The aim of these interventions is to integrate and make data accessible across different platforms (such as clinical and financial systems), and to allow for the sharing of this data across the patient care trajectory. The ultimate aim of DH, it would seem, is to improve patient outcomes, the coordination of health services, quality of care, and efficiency while facilitating public health interventions.

In reviewing the definitions, we identified nine characteristics of DH. No single study included all these characteristics, and there are no specific factors such as study design, country, or level of the health care system associated with the definitions. DH is characterized by the following characteristics:

• Any type of DH intervention or activity is a process of multiple steps involving both technical and social processes.
• The goal of a DH intervention or activity is to integrate, harmonize, and bring together different electronic databases into useable formats.
• There are at least two or more databases involved in any DH intervention or activity.
• A data harmonization intervention or activity involves electronic data (no reference is made to data found in paper-based sources).
• Data harmonization occurs when there is an increasing availability of electronic data that can be pooled together using unique patient identifiers.
• Different types of data can be linked and shared, such as individual patient clinical, pharmacy, and laboratory data; health care utilization and cost data; and personnel-related data.
• Electronic data required for DH processes can be found within and across different departments and institutions at facility, district, regional, and national levels.
• A data harmonization process consists of different types of technical activities such as identifying, reviewing, matching, defining, redefining, standardizing, merging, linking, merging, and formatting data.
• DH interventions or activities are defined according to a specific scope and purpose such as disease surveillance, monitoring of long-term outcomes, screening for adverse events, geographic area, secondary data use, and data display mechanisms (data marts or dashboards).

Components and processes of data harmonization

To synthesize key components and processes of DH interventions (second objective) we sampled five from the 61 studies identified for the second and third objectives. We selected five studies based on the conceptual descriptions and visual illustrations of their DH interventions (see Figure 3).

Figure 3. Concepts of data harmonization interventions and processes; this figure presents the different conceptual models of data harmonization, and the review authors provide a summary of how key components and processes were described by the authors of these models.

The conceptual description by Ji et al. comes closest to a comprehensive conceptual model of a DH intervention, illustrating different types of data, different levels of the health care system (e.g., clinics and hospitals), the multiple processes of exchanging data, the multiple directions in data exchange, and the key role of the unique patient identifier in enabling the DH process. In the next model, Boyd et al. and Santos et al. both lay out the technical processes involved in the linkage of different databases, but Santos et al. specifically focuses on linking data required for individual patient clinical management into a central repository. Lastly, Elysee et al. and Hu et al. describe DH interventions with different purposes, that is, medication reconciliation and disease outbreak surveillance, respectively.

These conceptual models of DH interventions and activities highlight that there are various steps involved in the integration of databases and in the transformation of data into useable formats. Integrating databases means bringing together data of the same individual from within and between different electronic databases, through various activities involving identifying, reviewing, matching, redefining and standardizing data. Once data is harmonized, it can be categorized by various criteria of interest, such as geographic area, disease, or patient population, and transformed into different formats such as graphs, tables, or dashboards to make it easier for users to access and use the information. There may be different ways that the data is harmonized; in some studies, DH was described as a linear and one-directional process, while other studies described it as an iterative and multi-directional process.

The relationship between data harmonization and health management decision-making

We sampled nine studies from the 61 studies (identified for the second and third objectives) that provide an explanation of the relationship between DH and health management decision-making. These nine studies were selected because they referred to the intended benefit, or directly referred to the relationship between DH and health management decision-making. We present extracts of explanations of the relationship in Table 4.

Table 4. The relationship between DH interventions and health management decision-making; the review authors directly quoted text from the primary studies where a description of the link between data harmonization and health management decision-making was provided (bold text emphasis ours). Contributing study Quote Cimino 2014 “Data completeness: A promise of HIEs is to use consolidated information over time and across providers to improve medical decision-making for the patient. When presenting a medical timeline for a patient, how does a provider know whether the HIE presentation of history is missing information? The consequences to patients can be devastating.” Downs 2010 “… community-based approach to establish a common pathway based on common data standards to facilitate the incorporation of interoperable, clinically useful genetic or genomic information and analytical tools into EHRs to support clinical decision-making for the clinician and consumer.” Grossman 2008 “… the exchanges going beyond core clinical data exchange activities that give physicians access to data at the point of care to offering physicians clinical decision support, reminders and other quality improvement tools aimed at individual patients.” Kuperman 2013 “Ideally, a physician would have access to complete, accurate and timely patient data to support optimal decision making. Health information exchange capabilities will reduce the extent of data fragmentation but will not eliminate it entirely.” Politi 2014 “In this scenario, an HIE system is likely to have a significant impact on clinical decision making if information is readily accessible; the need for rapid decisions might render the scrutiny of an HIE system impractical.” Vest 2010 “The anticipated benefits of more data to inform physician decision making, sparing patients of needless tests, helping organization identify inappropriately managed patients, and improving the health of the public will only be achieved by HIE that does not exclude providers in an area, limit what data elements are available, or restrict exchange to specific subpopulations.” Shapiro 2006 “The goal of a nationwide health information network would be to deliver information to individuals– consumers, patients, and professionals –when and where they need it, so they can use this information to make informed decisions about health and health care.” Vest 2015 “Improved access to more comprehensive information may support decision-making, inform providers of additional medications or allegories, and help avoid repeated or duplicate testing.” Zaiden 2015 “Combined with data mining and statistical analysis tools, these repositories of health information can greatly advance medical knowledge, healthcare quality, and good strategic management.”

According to Eylsee et al. (the study providing the most detail), there is a positive relationship between increased availability of electronic data sets and the ability of clinicians to deal with high volumes of data. This necessitates interoperability between electronic databases at different hospitals in order to improve timeliness, accuracy, and completeness of information sharing. According to Ji et al., Boyd et al., Santos et al., and Hu et al., the main benefit of DH is health management decision-making, including clinical decision-making. Across the studies, there is agreement that DH interventions make it possible for health providers to use data over time and across organizations to support clinical management decision-making. There is acknowledgement that DH interventions were sometimes unable to deal effectively with inconsistencies, incompleteness, and poor quality of data.

From the nine studies in Table 4, we identified three types of health management decision-making that DH contributed to. These are:

• clinical decision-making for individual patient clinical management or clinical support and quality improvement tools;
• pperational and strategic decision-making for health system managers and policy-makers; and
• population-level decision-making for disease surveillance and outbreak management.

The first level involves frontline clinicians being able to access their patients’ medical information, treatment data, and timelines (datasets of longitudinal, clinically relevant individual-level data) through DH interventions. In these situations, DH can make it easier for frontline clinicians to develop tools for reminding them about patients’ performance in treatment and care services, as well as help them improve the quality of health care services. At the operational and strategic decision-making level, DH interventions have the potential to support high-level health managers in decision-making involving a wide network of stakeholders (consumers, patients, and professionals). Lastly, disease surveillance and outbreak management decision-making rely on harmonized data to plan, monitor, and evaluate population-level interventions.

Discussion

Synthesis of findings

This scoping review aimed to provide an overview of the key characteristics of DH studies; identify definitions, alternative terms, components, and processes of DH interventions; and provide explanations of the relationship between DH and health management decision-making. Of the 181 studies that at a minimum provided a definition or description of a DH intervention or activity, 86 were primary quantitative studies, 151 were studies conducted in the USA, and 128 were aimed at improving frontline-level health services.

A key finding is that the term "health information exchange" or HIE was the most frequently used in the literature, especially for studies for the USA. Other terms used were "data harmonization," "record linkage," "data linkage," "data warehousing," "data sharing," and "data interoperability." Terms like "data harmonization" and "data warehousing" seem to describe a more comprehensive approach to DH interventions (involving both data production and data utilization aspects), whereas terms like "record linkage" and "data linkage" described specific activities within health information exchange. The term "data interoperability" focuses on the technical aspects that allow for different electronic databases to be linked and for data to be integrated, which then allow for synthesis and analysis of health information. Even though different studies used different terms, there was consensus that DH is a useful tool for health management decision making and can support improvements in patient and health system outcomes.

We identified nine characteristics of DH interventions and activities. Using these nine characteristics, DH can be summarized as a process that aims to integrate two or more electronic databases involving different types of data captured within and across various institutions at different health care system levels, with varying activities required to pool together data using unique patient identifiers for the purpose of providing information support for health management decision-making. The review identified three types of health management decision-making that DH contributes to: (a) clinical decision-making for individual patient management, clinical support, and quality improvement tools; (b) operational and strategic decision-making for health system managers and policy-makers; and (c) population-level decision-making for disease surveillance and outbreak management.

Drawing on the definitions and the conceptual models of DH identified in this review, we developed a concept map (see Figure 4) to explain how different aspects of DH interventions and activities work together to support health management decision-making. The concept map consists of different types of databases (1 to 5) containing different types of data such as demographic, clinical, pharmacy, laboratory, administrative, financial, and terminology data. A technical process involving different types of activities (such as matching, merging, and linking) takes place to integrate the different types of data using a unique patient identifier. The central repository, where the data is harmonized, is defined according to specific criteria such as a geographic area or disease outcomes. The data kept in the repository should be accessible to data users, who can then use this harmonized data as an information and analytic tool to support health management decision-making for clinical, operational, strategic, and or population-level decision-making.

Figure 4. A concept map of data harmonization and its relationship to health management decision-making

Study limitations

There are two main differences between the published protocol and this scoping review. We did not search the Global Health database as planned; we realized too late that none of the reviewers had permissions to access the database, and gaining access was not affordable. We did however manage to search at least three electronic databases, as is the convention in reviews. Due to the large volume of studies included for full-text screening, it was not feasible to conduct the full text screening in duplicate as planned. The first reviewer (BS) assessed all full-texts, and then the second reviewer (AH) verified the decisions of the first reviewer in a third of the included studies, which allowed for additional quality checks.

There are two main limitations of the review. Firstly, we restricted our literature search to English. We did not have the resources required for reviewing non-English studies. Most studies identified were from the USA, but it is possible that studies from other non-English speaking, high-income countries with extensive electronic health systems (such as France) may have been missed. Secondly, although sampling aimed to identify variety, comprehensiveness, and meaningfulness of the definitions and explanations, there is a possibility that due to sampling, we may have missed relevant studies for the second and third objectives.

Implications for research and practice

There is a need to understand what DH interventions and activities are comprised of in diverse settings and contexts, especially in LMICs. There were fewer studies from LMICs, which may be due to a lower prevalence of electronic health information systems in those settings. Nevertheless, DH interventions hold promise for improving the informational support in LMICs; studies in these contexts could usefully expand the evidence base.

The review highlights the importance of providing detailed descriptions of DH interventions, to allow for better comparisons and to improve the transferability of study results. Additionally, many resources are spent on the technical development of DH projects, with the implicit assumption that this will provide the informational and analytic support for health management decision-making, but this assumption is seldom tested in the research. There is a need for qualitative research on the health system factors of implementing DH and for formative work to inform design of DH interventions. Finally, primary research and evidence synthesis of the experiences of key stakeholders involved (implementers and users of harmonized data) would improve our understanding of the causal mechanisms between data harmonization and health systems strengthening.

Conclusion

This review aimed to widen our understanding about the range of definitions, components, and processes of DH interventions, and how it can contribute to health management decision-making. Most studies of DH interventions and activities were conducted in high-income settings and used the term "health information exchange." The review described the processes, technical activities, types of data, mechanisms for integrating data, and purpose of the DH interventions. DH interventions contributed to three types of health management decision-making, that is, clinical decision-making, operational and strategic decision-making, and population-level surveillance decision-making. We have provided a concept map of the components of DH and have made recommendations for future research.

Abbreviations

CDE: clinical data exchange

CIE: clinical information exchange

DH: data harmonization

DL: data linkage

DS: data sharing

DW: data warehouse

EHR: electronic health record

HDE: health data exchange

HIE: health information exchange

HIS: health information system

IE: information exchange

IO: interoperability

IT: information technology

LMICs: low-to-middle-income countries

MeSH: Medical Subject Headings

RHISs: routine health information systems

RL: record linkage

Acknowledgements

We would like to thank Ms. Gill Morgan, University of Cape Town, who assisted with developing the search strategy.

Contributions

BS was involved in all the tasks of conducting the scoping review. She drafted the manuscript with help from CC and NL. AH contributed to searching, screening, and data extraction processes. All authors reviewed and approved the final manuscript before final submission for peer review.

Funding

Time to write this paper was supported by the U.S. National Institute of Mental Health [grant number 1R01 MH106600] and the South African Medical Research Council (SAMRC). The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the U.S. National Institutes of Health or the SAMRC.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests

The authors declare that they have no competing interests.

Notes

This presentation is faithful to the original, with only a few minor changes to presentation. In some cases important information was missing from the references, and that information was added. The "Study objectives" section in the original only states two objectives, leading to confusion when encountering references to three objectives; the intended three objectives were inferred from the rest of the text and restated for this version. The original tables didn't include an abbreviation key; the meaning of the abbreviations was inferred from the rest of the text (and the cited papers) and added for this version. Additionally, the author has been contacted asking for clarification. The original Table 4 was converted into Figure 3, bumping the existing Figure 3 to Figure 4 and making the original Table 5 into Table 4.