Ai Systematic Review Automation

Systematic Review Foundations

Doctoral work examines structured methods for locating and combining research evidence. Systematic reviews underpin guidelines and public policy decisions.

Evidence Synthesis Research

Research examines bringing findings from many studies into one coherent account. Synthesis converts scattered literature into usable knowledge.

Review Typology Research

Doctoral study examines the many differing forms evidence reviews can take. Type choice determines what methods and effort a review requires.

Rapid Review Research

Research examines abbreviated methods delivering evidence within short timeframes. Speed is achieved by accepting deliberate methodological shortcuts.

Scoping Review Research

Doctoral work examines mapping the extent and nature of available literature. Scoping suits broad questions where synthesis would be premature.

Umbrella Review Research

Research examines reviews that synthesise other systematic reviews together. Umbrella methods handle fields where many reviews already exist.

Realist Review Research

Doctoral study examines synthesis explaining how and why interventions work. Realist methods suit complex interventions in varying contexts.

Mixed Method Review

Research examines combining quantitative and qualitative evidence within one review. Combination answers questions neither approach could address alone.

Qualitative Synthesis Research

Doctoral work examines combining findings from qualitative research studies. Qualitative synthesis resists the automation that numbers permit.

Network Synthesis Research

Research examines synthesis comparing many interventions simultaneously together. Network methods rank options that were never directly compared.

Review Protocol Research

Doctoral study examines documented plans specifying review methods beforehand. Protocols constrain the flexibility that produces biased conclusions.

Protocol Registration Research

Research examines publicly recording review plans before the work commences. Registration permits detecting departures from the original stated plan.

Question Formulation Research

Doctoral work examines framing questions that a review can actually answer. Question structure determines every subsequent methodological choice.

Eligibility Criteria Research

Research examines rules determining which studies a review will include. Criteria wording determines what evidence the review ultimately contains.

Scope Definition Research

Doctoral study examines setting boundaries around what a review will cover. Scope choices trade comprehensiveness against feasible workload.

Review Planning Research

Research examines organising resources and timelines for a review project. Planning quality determines whether reviews are actually completed.

Review Automation Foundations

Doctoral work examines computational support for evidence synthesis tasks. Automation addresses workload that is growing faster than capacity.

Task Decomposition Research

Research examines breaking review work into separately automatable components. Decomposition reveals which steps are genuinely ready for automation.

Human Machine Division

Doctoral study examines allocating review tasks between people and software. Allocation determines both efficiency gained and errors introduced.

Automation Readiness Research

Research examines which review tasks current technology can reliably perform. Readiness assessment prevents deploying tools that are not ready.

Search Strategy Research

Doctoral work examines designing searches that locate all relevant studies. Search design determines the ceiling on review comprehensiveness.

Query Formulation Research

Research examines constructing database queries expressing a review question. Query construction is skilled work requiring specialist expertise.

Search Term Generation

Doctoral study examines identifying words and phrases describing a topic. Missing terminology causes relevant studies to be entirely overlooked.

Controlled Vocabulary Research

Research examines standardised indexing terms applied within literature databases. Vocabulary use improves retrieval where indexing is consistent.

Search Translation Research

Doctoral work examines converting a search between differing database systems. Translation errors are common and reduce retrieval substantially.

Database Selection Research

Research examines choosing which literature sources a review should search. Source choice strongly determines what evidence is eventually found.

Grey Literature Search

Doctoral study examines locating research outside conventional published journals. Unpublished work frequently reports less favourable findings.

Preprint Search Research

Research examines including manuscripts shared before any formal peer review. Preprints provide currency and carry genuinely uncertain reliability.

Trial Registry Search

Doctoral work examines searching records of studies that were formally registered. Registries reveal studies that were conducted and never published.

Regulatory Document Search

Research examines evidence held within regulatory submission documents. Regulatory files contain data that published papers frequently omit.

Citation Chasing Research

Doctoral study examines locating studies through their citation relationships. Citation methods find work that keyword searching entirely misses.

Backward Citation Research

Research examines examining reference lists of already identified studies. Reference checking is simple and reliably finds additional studies.

Forward Citation Research

Doctoral work examines finding later work citing an already identified study. Forward searching locates the most recent relevant literature.

Similarity Based Retrieval

Research examines finding studies resembling already known relevant examples. Similarity methods require no explicit query to be constructed at all.

Semantic Search Research

Doctoral study examines retrieval based upon meaning rather than exact words. Semantic methods find studies using entirely differing terminology.

Embedding Retrieval Research

Research examines numerical representations supporting meaning based retrieval. Representation quality determines what similarity search can find.

Search Recall Research

Doctoral work examines what proportion of relevant studies a search retrieves. Recall is the property that matters most within systematic review.

Search Precision Research

Research examines what proportion of retrieved records are actually relevant. Low precision creates the screening burden that reviews then face.

Search Validation Research

Doctoral study examines testing whether a search finds known relevant studies. Validation catches strategies that would miss important evidence.

Search Peer Review

Research examines specialists checking search strategies before they are run. Peer checking catches errors that would otherwise go unnoticed.

Reproducible Search Research

Doctoral work examines recording searches so others can repeat them exactly. Incomplete reporting makes most published searches unrepeatable.

Deduplication Research

Research examines removing repeated records retrieved from several databases. Duplication is substantial and manual removal consumes real time.

Record Matching Research

Doctoral study examines recognising when two records describe one study. Matching is complicated by inconsistent and incomplete bibliographic data.

Identifier Resolution Research

Research examines using persistent identifiers to link bibliographic records. Identifiers resolve matching problems that text comparison cannot.

Reference Management Research

Doctoral work examines software organising the records a review must handle. Reference tools shape how whole review teams actually work together.

Search Reporting Research

Research examines describing searches within the published review report itself. Reporting standards exist and are followed rather inconsistently.

Search Efficiency Research

Doctoral study examines balancing retrieval completeness against screening workload. Efficiency matters because reviewer time is severely limited.

Search Automation Evaluation

Research examines assessing whether automated searching performs adequately. Evaluation must measure recall rather than mere convenience alone.

Adaptive Search Research

Doctoral work examines searches refined as relevant studies are discovered. Adaptive approaches use findings to improve subsequent retrieval.

Search Stopping Research

Research examines deciding when searching has become sufficiently complete. Stopping decisions currently rest upon convention rather than evidence.

Screening Research

Doctoral study examines assessing retrieved records against eligibility criteria. Screening is the most time consuming stage of most reviews.

Title Abstract Screening

Research examines the first screening stage using only brief record summaries. This stage removes the great majority of all retrieved records.

Full Text Screening Research

Doctoral work examines assessing complete articles against eligibility criteria. Full text assessment requires obtaining articles that may be restricted.

Screening Prioritisation Research

Research examines ordering records so relevant ones are assessed earliest. Prioritisation lets reviewers stop before examining every record.

Active Learning Research

Doctoral study examines models improving as reviewers label further records. Active approaches reduce screening effort really quite substantially.

Relevance Ranking Research

Research examines ordering records by their estimated likelihood of relevance. Ranking quality determines how much screening effort is saved.

Classifier Training Research

Doctoral work examines building models that distinguish relevant from irrelevant records. Training data quantity and quality determine model performance.

Labelled Data Research

Research examines the human judgements used to train screening models. Label quality bounds what any trained model could eventually achieve.

Class Imbalance Research

Doctoral study examines learning where relevant records are extremely rare. Severe imbalance defeats methods designed for balanced problems.

Screening Stopping Criteria

Research examines deciding when enough records have actually been assessed. Stopping too early risks missing studies that would change conclusions.

Recall Estimation Research

Doctoral work examines estimating how many relevant studies remain unfound. Estimation supports stopping decisions that would otherwise be guesswork.

Missed Study Research

Research examines relevant studies that reviews failed to identify at all. Missed studies can measurably change the conclusion a review reaches.

Screening Error Research

Doctoral study examines mistakes reviewers make when assessing the records. Human screening error rates are higher than is commonly assumed.

Single Reviewer Screening

Research examines one person assessing records rather than two independently. Single screening halves effort and measurably increases error.

Dual Reviewer Screening

Doctoral work examines two reviewers independently assessing the same records. Independent duplication is the established standard for screening.

Disagreement Resolution Research

Research examines settling differences between independently screening reviewers. Resolution approach affects which studies eventually enter a review.

Agreement Measurement Research

Doctoral study examines quantifying consistency between screening reviewers. Measured agreement is frequently lower than review teams expect.

Screening Burden Research

Research examines the human effort that screening actually demands from teams. Burden is the principal reason that reviews take so very long.

Screening Interface Research

Doctoral work examines software through which reviewers assess the records. Interface design measurably affects both working speed and error rates.

Reviewer Fatigue Research

Research examines performance declining across long screening sessions. Fatigue effects are documented and rarely accounted for in practice.

Language Restriction Research

Doctoral study examines reviews limiting inclusion to particular languages. Restriction excludes evidence and introduces clear geographic bias.

Multilingual Screening Research

Research examines assessing records written in many differing languages. Machine translation now makes multilingual review far more feasible.

Study Design Classification

Doctoral work examines automatically identifying what design a study used. Design determines eligibility and shapes all subsequent appraisal.

Population Identification Research

Research examines recognising who was studied within a given research report. Population matching is central to determining study eligibility.

Intervention Identification Research

Doctoral study examines recognising what was done to study participants. Intervention descriptions are frequently vague and inconsistently worded.

Outcome Identification Research

Research examines recognising what was measured within a research study. Outcome recognition is essential for both eligibility and synthesis.

Data Extraction Research

Doctoral work examines recording information from included studies systematically. Extraction is laborious and a frequent source of review error.

Automated Extraction Research

Research examines software pulling information from published research articles. Automated extraction remains less reliable than automated screening.

Table Extraction Research

Doctoral study examines recovering structured data from tables within articles. Most numerical results appear in tables rather than in text.

Figure Extraction Research

Research examines recovering numerical values presented only within graphs. Values existing only in figures are otherwise entirely inaccessible.

Numerical Extraction Research

Doctoral work examines recovering quantities needed for statistical synthesis. Numerical errors propagate directly into the eventual pooled result.

Statistical Result Extraction

Research examines recovering effect estimates and their measures of precision. Reported statistics are frequently incomplete or inconsistently presented.

Entity Recognition Research

Doctoral study examines automatically identifying named concepts within text. Entity recognition underpins nearly all extraction automation.

Relation Extraction Research

Research examines identifying relationships between concepts within articles. Relations capture meaning that isolated entities cannot convey.

Document Structure Research

Doctoral work examines recognising sections and organisation within articles. Structure knowledge tells extraction tools where they should look.

Full Text Access Research

Research examines obtaining the complete articles that any review requires. Access restrictions obstruct reviews outside wealthy institutions.

Format Conversion Research

Doctoral study examines converting articles into machine processable formats. Conversion quality determines what extraction tools can actually read.

Extraction Verification Research

Research examines checking extracted information against the source article. Verification catches errors that would otherwise reach synthesis.

Extraction Error Research

Doctoral work examines mistakes made when recording study information. Extraction errors are common and frequently affect published results.

Extraction Form Design

Research examines structuring the templates reviewers use for recording data. Form design strongly influences the completeness of extraction.

Data Standardisation Research

Doctoral study examines converting extracted information into consistent formats. Standardisation is prerequisite for combining studies quantitatively.

Outcome Harmonisation Research

Research examines reconciling outcomes measured in differing ways across studies. Harmonisation determines which studies can actually be combined.

Unit Conversion Research

Doctoral work examines converting measurements onto a common reporting scale. Conversion errors are a documented source of synthesis mistakes.

Missing Data Research

Research examines handling information that included studies never reported. Missing information is pervasive within published research articles.

Author Contact Research

Doctoral study examines requesting missing information from study authors. Response rates are low and requests are frequently ignored entirely.

Supplementary Data Research

Research examines information published alongside but outside the main article. Supplementary files hold data that reviews frequently overlook.

Individual Participant Data

Doctoral work examines synthesis using original participant level records. Participant data permits analysis that summary figures cannot support.

Data Sharing Research

Research examines availability of study data supporting evidence synthesis. Sharing is promised very frequently and delivered rather rarely.

Extraction Reporting Research

Doctoral study examines documenting what was extracted and from exactly where. Reporting permits others to verify the synthesis that followed.

Provenance Research

Research examines tracing every synthesised figure back to its source. Provenance tracking is essential once automation enters the workflow.

Large Language Model Research

Doctoral work examines general language models applied to review tasks. These models handle many tasks and produce confident incorrect output.

Prompt Design Research

Research examines how instructions to language models affect their output. Instruction wording changes results more than practitioners expect.

Model Reliability Research

Doctoral study examines consistency of automated output across repeated runs. Inconsistent output undermines the reproducibility reviews require.

Fabricated Output Research

Research examines models generating plausible information that is simply false. Fabricated references and figures are a documented serious risk.

Verification Workflow Research

Doctoral work examines checking automated output before it enters a review. Verification effort can exceed the effort automation actually saved.

Risk Of Bias Research

Research examines assessing whether study findings may be systematically distorted. Bias assessment determines how much weight evidence deserves.

Bias Tool Research

Doctoral study examines structured instruments for appraising study limitations. Tool choice measurably changes the assessments reviewers reach.

Automated Bias Assessment

Research examines software judging study limitations without human appraisal. Automation here is harder than screening and considerably less mature.

Domain Level Assessment

Doctoral work examines appraising limitations within separate methodological domains. Domain assessment is more informative than any overall score.

Randomisation Assessment

Research examines judging whether allocation to groups was genuinely random. Reported randomisation frequently lacks sufficient supporting detail.

Blinding Assessment Research

Doctoral study examines judging whether participants and assessors knew allocation. Knowledge of allocation distorts subjectively measured outcomes.

Attrition Assessment Research

Research examines judging the effect of participants leaving a study early. Differential loss between groups can entirely reverse conclusions.

Selective Reporting Assessment

Doctoral work examines detecting outcomes measured but never actually reported. Selective reporting biases the literature toward favourable results.

Interest Disclosure Assessment

Research examines declared interests within studies entering a review. Funding source relates measurably to reported findings in some fields.

Study Quality Research

Doctoral study examines overall methodological standard of included research. Quality and bias risk are related but genuinely distinct concepts.

Reporting Quality Research

Research examines completeness of description within published study reports. Poor reporting prevents appraisal of otherwise sound research.

Trustworthiness Research

Doctoral work examines whether included studies are genuine and reliable. Reviews increasingly must consider whether studies actually happened.

Fabricated Study Detection

Research examines identifying invented studies within the published literature. Fabricated studies have contaminated several published evidence syntheses.

Retraction Checking Research

Doctoral study examines detecting withdrawn articles among the included studies. Withdrawn work continues to be cited within published reviews.

Integrity Screening Research

Research examines systematic checks for problems within candidate studies. Integrity screening is rapidly becoming a standard stage within reviews.

Meta Analysis Research

Doctoral work examines statistically combining results from separate studies. Combination increases precision and can conceal real differences.

Effect Size Research

Research examines quantities expressing how large an observed effect was. Effect measure choice affects both interpretation and pooled results.

Synthesis Model Choice

Doctoral study examines statistical models used when combining study results. Model choice reflects assumptions about how studies actually differ.

Heterogeneity Research

Research examines variation between the results of the combined studies. Substantial variation questions whether pooling is appropriate at all.

Subgroup Analysis Research

Doctoral work examines whether effects differ between defined participant groups. Subgroup claims are frequently unreliable and widely overinterpreted.

Meta Regression Research

Research examines relating study characteristics to their observed effects. Regression across studies requires far more studies than usually available.

Network Meta Analysis

Doctoral study examines comparing many treatments through connected evidence. Network methods rest upon assumptions that are difficult to verify.

Indirect Comparison Research

Research examines comparing treatments never studied directly against each other. Indirect estimates depend heavily upon the connecting evidence.

Dose Response Synthesis

Doctoral work examines combining evidence across differing exposure levels. Dose response synthesis extracts more information than pooling alone.

Diagnostic Accuracy Synthesis

Research examines combining studies of how well tests identify conditions. Accuracy synthesis requires methods distinct from treatment synthesis.

Prognostic Synthesis Research

Doctoral study examines combining evidence about predicting future outcomes. Prognostic reviews face particularly severe reporting problems.

Prevalence Synthesis Research

Research examines combining estimates of how common a condition actually is. Prevalence pooling is complicated by genuinely differing populations.

Qualitative Evidence Synthesis

Doctoral work examines methods combining findings from qualitative studies. Interpretive synthesis resists the automation numbers readily permit.

Narrative Synthesis Research

Research examines structured description where statistical pooling is inappropriate. Narrative methods risk vagueness without a disciplined structure.

Publication Bias Research

Doctoral study examines unpublished studies distorting the available evidence. Missing studies systematically favour positive reported findings.

Small Study Effect Research

Research examines smaller studies reporting systematically larger observed effects. This pattern has several possible and competing explanations.

Sensitivity Analysis Research

Doctoral work examines whether conclusions survive differing analytical choices. Sensitivity testing reveals how fragile a conclusion actually is.

Certainty Assessment Research

Research examines judging confidence in a body of combined evidence. Certainty judgements determine how strongly recommendations can be made.

Evidence Grading Research

Doctoral study examines structured frameworks rating evidence strength. Grading frameworks make judgement explicit rather than merely implicit.

Summary Table Generation

Research examines producing concise tables presenting the key review findings. Summary tables are what most readers of reviews actually read.

Plain Language Summary

Doctoral work examines communicating review findings to nonspecialist audiences. Accessible summaries determine whether evidence reaches the public.

Synthesis Visualisation Research

Research examines graphical presentation of combined evidence and uncertainty. Visual choices strongly influence what readers eventually conclude.

Forest Plot Research

Doctoral study examines the standard display showing individual and pooled results. Plot design affects how readers weigh individual studies.

Evidence Map Research

Research examines visual overviews showing where evidence exists and where not. Maps guide research funding toward genuinely unanswered questions.

Evidence Gap Research

Doctoral work examines identifying questions the literature has never addressed. Gap identification is a principal output of scoping methods.

Living Review Research

Research examines reviews maintained continuously as new evidence appears. Living methods keep syntheses current within fast moving research fields.

Continuous Surveillance Research

Doctoral study examines ongoing monitoring for newly published relevant studies. Surveillance is what makes continuously maintained reviews possible.

Review Refresh Research

Research examines revising an existing review to incorporate newer evidence. Refresh effort is substantial and frequently never actually undertaken.

Trigger Criteria Research

Doctoral work examines deciding when a review genuinely requires revision. Triggers avoid both premature and indefinitely postponed revision.

Review Maintenance Research

Research examines sustaining reviews across long periods and changing teams. Maintenance depends upon funding that is rarely made available.

Review Obsolescence Research

Doctoral study examines how quickly published reviews become out of date. A substantial share are already outdated within a rather short period.

Review Overlap Research

Research examines reviews covering substantially the same body of studies. Overlap wastes effort and confuses those seeking any clear guidance.

Duplicate Review Research

Doctoral work examines redundant reviews addressing already answered questions. Redundancy is widespread and represents substantial wasted effort.

Review Comparison Research

Research examines methods for comparing reviews addressing similar questions. Comparison reveals which methodological differences actually mattered.

Discordant Review Research

Doctoral study examines reviews reaching differing conclusions on one question. Discordance usually arises from differing methods rather than evidence.

Review Software Research

Research examines applications supporting the conduct of evidence reviews. Software choice shapes the workflow that a review team can follow.

Workflow Platform Research

Doctoral work examines integrated systems supporting whole review projects. Platforms reduce the manual transfer between otherwise separate tools.

Tool Interoperability Research

Research examines separate review tools exchanging data with one another. Poor interoperability forces manual work between the automated steps.

Data Standard Research

Doctoral study examines agreed formats for exchanging review information. Standards determine whether separate tools can genuinely work together.

Machine Readable Review

Research examines publishing reviews in formats software can process directly. Machine readable outputs permit reuse that documents prevent.

Structured Reporting Research

Doctoral work examines organising review reports into identifiable components. Structure supports both automation and eventual reader comprehension.

Reporting Guideline Research

Research examines standards specifying what review reports must contain. Guidelines improve completeness only where journals genuinely enforce them.

Transparency Research

Doctoral study examines making review decisions and data openly available. Transparency permits others to check what a review has actually done.

Reproducibility Research

Research examines whether reviews can be repeated from their own reports alone. Most published reviews cannot currently be reproduced by anybody.

Review Replication Research

Doctoral work examines independently repeating a review to check its findings. Replication is rare and consistently reveals substantial disagreement.

Method Validation Research

Research examines demonstrating automated review methods perform acceptably. Validation must measure what any given method would actually miss.

Benchmark Dataset Research

Doctoral study examines shared collections used to compare automation methods. Benchmark quality determines whether comparisons mean anything.

Evaluation Metric Research

Research examines measures used to judge automated review tool performance. Metric choice determines which tools appear to perform the best.

Comparative Evaluation Research

Doctoral work examines head to head assessment of competing review tools. Independent comparison is scarce within this rapidly growing area.

Error Analysis Research

Research examines characterising the mistakes automated review tools make. Error patterns matter considerably more than aggregate performance figures.

Cost And Time Research

Doctoral study examines resources that conducting evidence reviews actually consumes. Reviews commonly take far longer than teams originally planned.

Efficiency Gain Research

Research examines effort actually saved by adopting automated review tools. Claimed savings frequently exceed those measured in real projects.

Quality Impact Research

Doctoral work examines whether automation changes the quality of reviews produced. Faster reviews are worthless if their conclusions become unreliable.

Trust In Automation

Research examines whether reviewers rely appropriately upon automated assistance. Both excessive and insufficient reliance produce poor outcomes.

Human Oversight Research

Doctoral study examines what supervision automated review steps genuinely require. Oversight design determines whether automation remains safe.

Accountability Research

Research examines responsibility for conclusions that automated tools shaped. Responsibility must remain clear however much work is automated.

Reviewer Skill Research

Doctoral work examines expertise required to conduct evidence reviews well. Automation shifts rather than removes the expertise that is required.

Review Training Research

Research examines preparing people to conduct systematic reviews competently. Most reviews are conducted by people with very limited training.

Information Specialist Research

Doctoral study examines librarians and searchers supporting review teams. Specialist involvement measurably improves search quality achieved.

Team Composition Research

Research examines the mix of expertise a review team actually requires. Composition determines which methodological errors are caught early.

Stakeholder Involvement Research

Doctoral work examines involving those affected in shaping review questions. Involvement makes reviews address questions people actually have.

Patient Involvement Research

Research examines patients contributing to health evidence review projects. Patient priorities frequently differ from those researchers assume.

Review Commissioning Research

Doctoral study examines how reviews are requested, funded and then specified. Commissioning arrangements shape which questions actually get answered.

Guideline Development Research

Research examines reviews feeding into formal recommendation development. Guidelines are the principal route by which reviews affect practice.

Policy Application Research

Doctoral work examines evidence synthesis informing government policy decisions. Policy questions rarely match the questions researchers studied.

Clinical Application Research

Research examines reviews supporting decisions about patient treatment choices. Clinical synthesis is where these methods originally developed.

Public Health Application

Doctoral study examines synthesis supporting population level health decisions. Population interventions resist the designs clinical synthesis expects.

Environmental Application Research

Research examines evidence synthesis within environmental science and management. Environmental evidence is heterogeneous and frequently observational.

Social Science Application

Doctoral work examines synthesis methods within social research fields. Social evidence resists pooling that clinical methods take for granted.

Education Application Research

Research examines evidence synthesis informing teaching and education policy. Educational contexts vary enormously and complicate any generalisation.

Software Engineering Application

Doctoral study examines synthesis methods within computing research fields. This field adopted systematic review methods comparatively recently.

Regulatory Application Research

Research examines synthesis supporting regulatory approval and safety decisions. Regulatory synthesis faces particularly demanding evidential standards.

Equity Consideration Research

Doctoral work examines whether reviews address effects across differing populations. Reviews frequently ignore who benefits and who does not.

Global Coverage Research

Research examines geographic representation within synthesised research evidence. Evidence originates overwhelmingly from a few wealthy countries.

Low Resource Setting Research

Doctoral study examines conducting reviews where resources are severely limited. Access barriers and tool costs exclude many potential reviewers.

Open Infrastructure Research

Research examines freely available systems supporting evidence synthesis work. Open infrastructure reduces dependence upon commercial platforms.

Sustainability Research

Doctoral work examines sustaining review infrastructure beyond initial funding. Much critical infrastructure depends upon precarious funding arrangements.

Governance Research

Research examines oversight of automated tools within evidence synthesis. Governance determines what automation is permitted and exactly where.

Economic Evaluation Research

Doctoral study examines value delivered by investment in evidence synthesis. Reviews are expensive and their downstream benefit is very diffuse.

Implementation And Adoption

Research examines why automation reaches review practice or fails to do so. Adoption depends on trust and workflow fit as much as performance.