Dr Wayne Smith, Health Economic Lead, draws on his extensive experience to explain why a systematic literature review is such an important tool for healthcare decision-making.
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:
- Study characteristics
- Risk of bias
- 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.
