Why are literature reviews important in healthcare

When it comes to decision-making in healthcare, the implications are huge, with most affecting a significant number of patients. It’s important for all decisions to be informed and backed up with verified research and data.

With so much research already carried out within healthcare, a systematic literature review (SLR) is a useful way to consolidate this work.

By identifying, evaluating and summarising the findings from many different existing research studies on a specific topic we can answer questions about a healthcare intervention or diagnostic technique, identify information to be incorporated in economic models, or review validated methods to apply to current projects. Combining the results of several studies gives a more reliable and accurate estimate of effectiveness than an individual study. In a nutshell, it’s about informing evidence-based healthcare.

What is a systematic literature review (SLR)?

An SLR is a review of a clearly formulated question that uses systematic and explicit (valid and replicable) methods to identify, select and critically appraise relevant research and analyse data from the studies included in it.

It is a rigorous process which involves at least two researchers. Depending on the number of papers to be reviewed, an SLR can be a lengthy process. It’s a thorough deep-dive into data already available on the topic being investigated. As well as setting out what we already know about a particular area, an SLR can also highlight where knowledge is lacking, which can be used to guide further research. Statistical methods such as meta-analysis may be used to analyse and summarise the results of the included studies.

Stages of a systematic literature review

The very first step in an SLR is defining the question to be considered. This is done using the PICOS model, for which the concepts are as follows:

• Population – these are the people with the condition of interest
• Intervention – the new drug, medical procedure or health policy under scrutiny
• Comparators – current standard treatment (this may not be applicable)
• Outcomes – effectiveness, adverse effects, cost-effectiveness
• Study design– randomised control trial CT or cohort study etc

Once the question has been clearly defined, a protocol – a summary of all the steps to be undertaken in the research – is written. This is a ‘live’ document and can be updated as necessary.

Before searching for literature evidence it’s important to come up with a search strategy which aligns with the PICOS model. The databases to be searched can be pre-selected depending on the research question. The most commonly used databases for healthcare are Medline, Psych INFO and Embase. Other databases may be added as required but it’s best to limit the databases to between four and five for practical reasons. A strategy which aligns to the relevant PICOS concepts and involves developing keyword searches/synonyms and using Boolean operators in combination will be applied to all the different databases (slight differences may be observed using Boolean operators and applying limitations).

Typically a minimum of two reviewers select relevant studies (based on predefined eligibility criteria – agreed inclusion/exclusion criteria) from the databases and if there is any disagreement between them a third reviewer can act as arbitrator. This approach ensures the quality of the work.

An initial first screening of abstracts/titles is carried out followed by a full screening of articles to select the ones to be included for further review.

Once relevant studies have been selected for further review, a quality assessment of the studies must be undertaken. Again, this must be done by two reviewers for quality control purposes. A range of quality assessment tools may be used depending on what kind of SLR is being undertaken.

Structuring results

Once the data has been analysed, the results need to be reported before a conclusion is drawn and recommendations are made. The results section of the SLR report will usually include information about:

• Study selection
• Study characteristics
• Risk of bias within and across studies
• Evidence available for each review outcome
• Results – for each outcome and associated risk of bias/additional analysis
• Themes across outcomes

There are three tables that must always be included in any SLR. These are:

1. Study characteristics
2. Risk of bias
3. Result tables

In some cases, it may be appropriate to carry out a meta-analysis. This process statistically combines the results from two or more separate studies in order to increase the power and precision of the estimate effects and derive conclusions about the body of research.

Meta analyses are used regularly when compiling evidence from studies which contribute towards health technology appraisals (HTAs), for example to provide more precise estimates of outcomes including estimates of treatment effects or risk factors. They can also be applied outside of HTAs, for example combining the results from studies comparing the relative risk of an intervention such as medicines review against standard practice.

How does this support decision-making?

Currently we use SLR or targeted reviews to inform decisions on:

• Methods for analytic and evaluation approaches
• Including parameters and deriving estimates for use in economic models
• Identifying best practice e.g. for new models of care
• Identifying similar work done across the sector and wider
• Scoping innovation such as new digital health products and application
• Informing novel economic approaches such as Social–Technical Allocation of Resources (STAR) analysis, which looks at how an intervention will translate to ongoing positive change in the real world.

The written report produced as part of an SLR will provide the best evidence available on the subject, which can assist in informing clinical judgement and ultimately recommendations made as a result will help shape services to ensure the best patient outcomes.

Literature reviews are a way of identifying what is already known about a research area and what the gaps are. To do a literature review, you will need to identify relevant literature, often through searching academic databases, and then review existing literature. Most often, you will do the literature review at the beginning of your research project, but it is iterative, so you may choose to change the literature review as you move through your project.

Searching the literature

The University of Melbourne Library has some resources about searching the literature. Leonie spoke about how she met with a librarian about searching the literature. You may also want to meet face-to-face with a librarian or attend a class at the library to learn more about literature searching. When you search the literature, you may find journal articles, reports, books and other materials.

Filing, categorising and managing literature

In order to manage the literature you have identified through searches, you may choose to use a reference manager. The University of Melbourne has access to RefWorks and Endnote. Further information about accessing this software is available through the University of Melbourne Library.

Writing a literature review

The purpose of the literature review is to identify what is already known about a particular research area and critically analyse prior studies. It will also help you to identify any gaps in the research and situate your research in what is already known about a particular topic.

Resources

• Aveyard, H. (2010). Doing a literature review in health and social care: A practical guide. London, UK: McGraw-Hill Education. Retrieved from Proquest https://ebookcentral.proquest.com/lib/unimelb/detail.action?docID=771406
• Reeves, S., Koppel, I., Barr, H., Freeth, D., Hammick, M. (2002). Twelve tips for undertaking a systematic review. Medical Teacher. 24(4), 358-363.
• Grant, M.J. and Booth, A. (2009). A typology of reviews: an analysis of 14 review types and associated methodologies. Health Information & Libraries Journal.
• Jesson, J., & Lacey, F. (2006). How to do (or not to do) a critical literature review. Pharmacy Education, 6(2), 139-148.

Begin your qualitative journey

• MDHS Facebook
• MDHS Twitter

This search identified 4019 articles, of which 241 underwent full-text screening; 73 articles met the inclusion criteria for the larger review. Five systematic literature reviews were excluded as was one article which presented duplicate results; this left a total of 67 articles eligible for review. See Fig. 1 for the PRISMA diagram describing study screening.

PRIMSA flow diagram (attached)

This systematic literature review was originally developed to identify attitudes towards secondary use and sharing of health administrative and clinical trial data in breast cancer. However, as there was a paucity of material identified specifically related to this group, we present the multidisciplinary results of this search, and where possible highlight results specific to breast cancer, and cancer more generally. We believe that the material identified in this search is relevant and reflective of the wider attitudes towards data sharing within the scientific and medical communities and can be used to inform data sharing strategies in breast cancer.

Eighteen [18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35] of the 67 articles addressed the perspectives of clinical and scientific researchers, data custodians, and ethics committees and were analysed for this paper (Table 2). The majority (n = 16) of articles focused on the views of researchers and health professionals, [18,19,20,21,22, 24,25,26, 28,29,30,31,32,33,34,35], only one article focused on data custodians [27] and ethics committees [23] respectively. Four articles [18, 19, 21, 35] included a discussion on the attitudes of both researchers and healthcare professionals and patients; only results relating to researchers/clinicians are included in this analysis (Fig. 1).

Several study methodologies were used, including surveys (n = 11) [24,25,26,27, 29,30,31,32,33,34,35], interviews and focus groups (n = 6) [18,19,20,21,22,23], and mixed methods (n = 1) [28]. Studies were conducted in a several countries and regions; a breakdown by country and study is available in Table 3.

In addition to papers focusing on general health and sciences [18, 21, 22, 24,25,26, 29,30,31,32,33,34], two articles included views from both science and non-science disciplines [27, 28]. Multiple sclerosis (MS) [19], mental health [35], and human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS)/tuberculosis (TB) [20] were each the subject of one article.

Study quality

Results of the quality assessment are provided in Table 2. QualSyst [15] scores ranged from 0.7 to 1.0 (possible range 0.0 to 1.0). While none were blinded studies, most provided clear information on respondent selection, data analysis methods, and justifiable study design and methodology.

Themes

Four key themes, barriers, facilitators, access, and ownership were identified; 14 subthemes were identified. A graphical representation of article themes is presented in Fig. 2. Two articles reflect the perspective of research ethics committees [23] and data custodians [27]; concerns noted by these groups are similar to those highlighted by researchers and healthcare professionals.

Graphic representation of key themes and subthemes identified (attached)

Eleven articles identified barriers to data sharing [20, 22, 24, 25, 27, 29,30,31,32,33,34]. Concerns cited by respondents included other researchers taking their results [24, 25], having data misinterpreted or misattributed [24, 27, 31, 32], loss of opportunities to maximise intellectual property [24, 25, 27], and loss of publication opportunities [24, 25] or funding [25]. Results of a qualitative study showed respondents emphasised the competitive value of research data and its capacity to advance an individual’s career [20] and the potential for competitive disadvantage with data sharing [22]. Systematic issues related to increased data sharing were noted in several articles where it was suggested the barriers are ‘deeply rooted in the practices and culture of the research process as well as the researchers themselves’ [33] (p. 1), and that scientific competition and a lack of incentive in academia to share data remain barriers to increased sharing [30].

Insufficient time, lack of funding, limited storage infrastructure, and lack of procedural standards were also noted as barriers [33]. Quantitative results demonstrated that the researchers did not have the right to make the data public or that there was no requirement to share by the study sponsor [33]. Maintaining the balance between investigator and funder interests and the protection of research subjects [31] were also cited as barriers. Concerns about privacy were noted in four articles [25, 27, 29, 30]; one study indicated that clinical researchers were significantly more concerned with issues of privacy compared to scientific researchers [25]. The results of one qualitative study indicated that clinicians were more cautious than patients regarding the inclusion of personal information in a disease specific registry; the authors suggest this may be a result of potential for legal challenges in the setting of a lack of explicit consent and consistent guidelines [19]. Researchers, particularly clinical staff, indicated that they did not see sharing data in a repository as relevant to their work [29]

Trust was also identified as a barrier to greater data sharing [32]. Rathi et al. identified that researchers were likely to withhold data if they mistrusted the intent of the researcher requesting the information [32]. Ethical, moral, and legal issues were other potential barriers cited [19, 22]. In one quantitative study, 74% of respondents (N = 317) indicated that ensuring appropriate data use was a concern; other concerns included data not being appropriate for the requested purpose [32]. Concerns about data quality were also cited as a barrier to data reuse; some respondents suggested that there was a perceived negative association of data reuse among health scientists [30].

Reasons for sharing

Eleven articles [19,20,21,22, 24, 25, 29,30,31,32,33] discussed the reasons identified by researchers and healthcare professionals for sharing health data; broadly the principle of data sharing was seen as a desirable norm [25, 31]. Cited benefits included improvements to the delivery of care, communication and receipt of information, impacts on care and quality of life [19], contributing to the advancement of science [20, 24, 29], validating scientific outputs, reducing duplication of scientific effort and minimising research costs [20], and promoting open science [31, 32]. Professional reasons for sharing data included academic benefit and recognition, networking and collaborative opportunities [20, 24, 29, 31], and contributing to the visibility of their research [24]. Several articles noted the potential of shared data for enabling faster access to a wider pool of patients [21] for research, improved access to population data for longitudinal studies [22], and increased responsiveness to public health needs [20]. In one study, a small percentage of respondents indicated that there were no benefits from sharing their data [24].

Analysis of quantitative survey data indicated that the perceived usefulness of data was most strongly associated with reuse intention [30]. The lack of access to data generated by other researchers or institutions was seen as a major impediment to progress in science [33]. In a second study, quantitative data showed no significant differences in reasons for sharing by clinical trialists’ academic productivity, geographic location, trial funding source or size, or the journal in which the results were published [32]. Attitudes towards sharing in order to receive academic benefits or recognition differed significantly based on the respondent’s geographic location; those from Western Europe were more willing to share compared to respondents in the USA or Canada, and the rest of the world [32].

Views on sharing

Seven articles [19,20,21, 29, 31, 33, 34] discussed researchers’ and healthcare professionals’ views relating to sharing data, with a broad range of views noted. Two articles, both qualitative, discussed the role of national registries [21], and data repositories [31]. Generally, there was clear support for national research registers and an acceptance for their rationale [21], and some respondents believed that sharing de-identified data through data repositories should be required and that when requested, investigators should share data [31]. Sharing de-identified data for reasons beyond academic and public health benefit were cited as a concern [20]. Two quantitative studies noted a proportion of researchers who believed that data should not be made available [33, 34]. Researchers also expressed differences in how shared data should be managed; the requirement for data to be ‘gate-kept’ was preferred by some, while others were happy to relinquish control of their data once curated or on release [20]. Quantitative results indicated that scientists were significantly more likely to rank data reuse as highly relevant to their work than clinicians [29], but not all scientists shared data equally or had the same views about data sharing or reuse [33]. Some respondents argued that not all data were equal and therefore should only be shared in certain circumstances. This was in direct contrast to other respondents who suggested that all data should be shared, all of the time [20].

Differences by age, background, discipline, professional focus, and world region

Differences in attitudes towards shared data were noted by age, professional focus, and world region [25, 27, 33, 34]. Younger researchers, aged between 20–39 and 40–49 years, were less likely to share their data with others (39% and 38% respectively) compared to other age groups; respondents aged over 50 years of age were more willing (46%) to share [33]. Interestingly, while less willing to share, younger researchers also believed that the lack of access to data was a major impediment to science and their research [33]. Where younger researchers were able to place conditions on access to their data, rates of willingness to share were increased [33].

Respondents from the disciplines of education, medicine/health science, and psychology were more inclined than others to agree that their data should not be available for others to use in the first place [34]. However, results from one study indicated that researchers from the medical field and social sciences were less likely to share compared to other disciplines [33]. For example, results of a quantitative study showed that compared to biologists, who reported sharing 85% of their data, medical and social sciences reported sharing their data 65% and 58% percent of the time, respectively [33].

One of the primary reasons for controlling access to data, identified in a study of data custodians, was due to a desire to avoid data misuse; this was cited as a factor for all surveyed data repositories except those of an interdisciplinary nature [27]. Limiting access to certain types of research and ensuring attribution were not listed as a concern for sociology, humanities or interdisciplinary data collections [27]. Issues pertaining to privacy and sensitive data were only cited as concerns for data collections related to humanities, social sciences, and biology, ecology, and chemistry; concerns regarding intellectual property were also noted [27]. The disciplines of biology, ecology, and chemistry and social sciences had the most policy restrictions on the use of data held in their repositories [27].

Differences in data sharing practices were also noted by world region. Respondents not from North American and European countries were more willing to place their data on a central repository; however, they were also more likely to place conditions on the reuse of their data [33, 34].

Experience of data sharing

The experience of data sharing among researchers was discussed in nine articles [20, 24,25,26, 28,29,30,31,32,33]. Data sharing arrangements were highly individual and ranged from ad hoc and informal processes to formal procedures enforced by institutional policies in the form of contractual agreements, with respondents indicating data sharing behaviour ranging from sharing no data to sharing all data [20, 26, 31]. Quantitative data from one study showed that researchers were more inclined to share data prior to publication with people that they knew compared to those they did not; post publication, these figures were similar between groups [24]. While many researchers were prepared to share data, results of a survey identified a preference of researchers to collect data themselves, followed by their team, or by close colleagues [26].

Differences in the stated rate of data sharing compared to the actual rate of sharing [25] were noted. In a large quantitative study (N = 1329), nearly one third of respondents chose not to answer whether they make their data available to others; of those who responded to the question, 46% reported they do not make their data electronically available to others [33]. By discipline, differences in the rate of refusal to share were higher in chemistry compared to non-science disciplines such as sociology [25]. Respondents who were more academically productive (> 25 articles over the past 3 years) reported that they have or would withhold data to protect research subjects less frequently than those who were less academically productive or received industry funding [32].

Attitudes to sharing de-identified data via data repositories was discussed in two articles [29, 31]. A majority of respondents in one study indicated that de-identified data should be shared via a repository and that it should be shared when requested. A lack of experience in uploading data to repositories was noted as a barrier [29]. When data was shared, most researchers included additional materials to support their data including materials such as metadata or a protocol description [29].

Two articles [28, 30] focused on processes and variables associated with sharing. Factors such as norms, data infrastructure/organisational support, and research communities were identified as important factors in a researcher’s attitude towards data sharing [28, 30]. A moderate correlation between data reuse and data sharing suggest that these two variables are not linked. Furthermore, sharing data compared to self-reported data reuse were also only moderately associated (Pearson’s correlation of 0.25 (p ≤ 0.001)) [26].

Predictors of data sharing and norms

Two articles [26, 30] discussed the role of social norms and an individual’s willingness to share health data. Perceived efficacy and efficiency of data reuse were strong predictors of data sharing [26] and the development of a ‘positive social norm towards data sharing support(s)[ed] researcher data reuse intention’ [30] (p. 400).

Policy framework

The establishment of clear policies and procedures to support data sharing was highlighted in two articles [22, 28]. The presence of ambiguous data sharing policies was noted as a major limitation, particularly in primary care and the increased adoption of health informatics systems [22]. Policies that support an efficient exchange system allowing for the maximum amount of data sharing are preferred and may include incentives such as formal recognition and financial reimbursement; a framework for this is proposed in Fecher et al. [28].

Research funding

The requirement to share data funded by public monies was discussed in one article [25]. Some cases were reported of researchers refusing to share data funded by tax-payer funds; reasons for refusal included a potential reduction in future funding or publishing opportunities [25].

Access and ownership

Articles relating to access and ownership were grouped together and seven subthemes were identified.

Access, information systems, and metadata

Ten articles [19,20,21,22, 26, 27, 29, 33,34,35] discussed the themes of access, information systems, and the use of metadata. Ensuring privacy protections in a prospective manner was seen as important for data held in registries [19]. In the setting of mental health, researchers indicated that patients should have more choices for controlling access to shared registry data [35]. The use of guardianship committees [19] or gate-keepers [20] was seen as important in ensuring the security and access to data held in registries by some respondents; however, many suggested that a researcher should relinquish control of the data collection once curated or released, unless embargoed [20]. Reasons for maintaining control over registry data included ensuring attribution, restricting commercial research, protecting sensitive (non-personal) information, and limiting certain types of research [27]. Concerns about security and confidentiality were noted as important and assurances about these needed to be provided; accountability and transparency mechanisms also need to be included [21]. Many respondents believed that access to the registry data by pharmaceutical companies and marketing agencies was not considered appropriate [19].

Respondents to a survey from medicine and social sciences were less likely to agree to have all data included on a central repository with no restrictions [33]; notably, this was also reflected in the results of qualitative research which indicated that health professionals were more cautious than patients about the inclusion of personal data within a disease specific register [19].

While many researchers stated that they commonly shared data directly with other researchers, most did not have experience with uploading data to repositories [29]. Results from a survey indicated that younger respondents have more data access restrictions and thought that their data is easier to access significantly more than older respondents [34]. In the primary care setting, concerns were noted about the potential for practitioners to block patient involvement in a registry by refusing access to a patient’s personal data or by not giving permission for the data to be extracted from their clinical system [21]. There was also resistance in primary care towards health data amalgamation undertaken for an unspecified purpose [22]; respondents were not in favour of systems which included unwanted functionality (do not want/need), inadequate attributes (capability and receptivity) of the practice, or undesirable impact on the role of the general practitioner (autonomy, status, control, and workflow) [22].

Access to ‘comprehensive metadata (is needed) to support the correct interpretation of the data’ [26] (p. 4) at a later stage. When additional materials were shared, most researchers shared contextualising information or a description of the experimental protocol [29]. The use of metadata standards was not universal with some respondents using their own [33].

Curation

Several articles highlighted the impact of data curation on researchers’ time [20,21,22, 29, 33] or finances [24, 28, 29, 33, 34]; these were seen as potential barriers to increased registry adoption [21]. Tasks required for curation included preparing data for dissemination in a usable format and uploading data to repositories. The importance of ensuring that the data is accurately preserved for future reuse was highlighted; it must be presented in a retriable and auditable manner [20]. The amount of time required to curate data ranged from ‘no additional time’ to ‘greater than ten hours’ [29]. In one study, no clinical respondent had their data in a sharable format [29]. In the primary care setting, health information systems which promote sharing were not seen as being beneficial if they required standardisation of processes and/or sharing of clinical notes [22]. Further, spending time on non-medical issues in a time poor environment [22] was identified as a barrier. Six articles described the provision of funding or technical support to ensure data storage, maintenance, and the ability to provide access to data when requested. All noted a lack of funding and time as a barrier to increased sharing data [20, 24, 28, 29, 33, 34].

Consent

Results of qualitative research indicated a range of views regarding consent mechanisms for future data use [18,19,20, 23, 35]. Consenting for future research can be complex given that the exact nature of the study will be unknown, and therefore some respondents suggested that a broad statement on future data uses be included [19, 20] during the consent process. In contrast, other participants indicated that the current consent processes were too broad and do not reflect patient preferences sufficiently [35]. The importance of respecting the original consent in all future research was noted [20]. It was suggested that seeking additional consent for future data use may discourage participation in the original study [20]. Differences in views regarding the provision of detailed information about sharing individual level data was noted suggesting that the researchers wanted to exert some control over data they had collected [20]. An opt-out consent process was considered appropriate in some situations [18] but not all; some respondents suggested that consent to use a patient’s medical records was not required [18]. There was support by some researchers to provide patients with the option to ‘opt-in’ to different levels of involvement in a registry setting [19]. Providing patients more granular choices when controlling access to their medical data [35] was seen as important.

The attitudes of ethics and review boards (N = 30) towards the use of medical records for research was discussed in one article [23]. While 38% indicated that no further consent would be required, 47% required participant consent, and 10% said that the requirement for consent would depend on how the potentially identifying variables would be managed [23]. External researcher access to medical record data was associated with a requirement for consent [23].

Acknowledgement

The importance of establishing mechanisms which acknowledge the use of shared data were discussed in four articles [27, 29, 33, 34]. A significant proportion of respondents to a survey believed it was fair to use other researchers’ data if they acknowledged the originator and the funding body in all disseminated work or as a formal citation in published works [33]. Other mechanisms for acknowledging the data originator included opportunities to collaborate on the project, reciprocal data sharing agreements, allowing the originator to review or comment on results, but not approve derivative works, or the provision of a list of products making use of the data and co-authorship [33, 34]. In the setting of controlled data collections, survey results indicated that ensuring attribution was a motivator for controlled access [27]. Over half of respondents in one survey believed it was fair to disseminate results based either in whole or part without the data provider’s approval [33]. No significant differences in mechanisms for acknowledgement were noted between clinical and scientific participants; mechanisms included co-authorship, recognition in the acknowledgement section of publications, and citation in the bibliography [29]. No consentient method for acknowledging shared data reuse was identified [29].

Ownership

Data ownership was identified as a potential barrier to increased data sharing in academic research [28]. In the setting of control of data collections, survey respondents indicated that they wanted to maintain some control over the dataset, which is suggestive of researchers having a perceived ownership of their research data [28]. Examples of researchers extending ownership over their data include the right to publish first and the control of access to datasets [28]. Fecher et al. noted that the idea of data ownership by the researcher is not a position always supported legally; ‘the ownership and rights of use, privacy, contractual consent and copyright’ are subsumed [28] (p. 15). Rather data sharing is restricted by privacy law, which is applied to datasets containing data from individuals. The legal uncertainty about data ownership and the complexity of law can deter data sharing [28].

Promotion/professional criteria

The role of data sharing and its relation to promotion and professional criteria were discussed in two articles [24, 28]. The requirement to share data is rarely a promotion or professional criterion, rather the systems are based on grants and publication history [24, 28]. One study noted that while the traditional link between publication history and promotion remains, it is ‘likely that funders will continue to get sub-optimal returns on their investments, and that data will continue to be inefficiently utilised and disseminated’ [24] (p. 49).

Page 2

1. *Search includes ‘wildcards’ or truncation