# Common Pitfalls in Planning a Literature Review and How to Avoid Them Most literature reviews run into trouble before a single database is opened. The single most common cause is a research question that sounds clear in the author's head but has never been written down in a form that fixes the population, the concept of interest, and the context. Everything downstream, from the search strategy to the screening decisions to the final synthesis, inherits whatever ambiguity lives in that unwritten question. The first thing to check is simple. Try to state your review question in one sentence that a stranger could act on without asking you a follow-up question. If you cannot, or if the sentence contains vague terms such as "recent", "relevant", or "the literature on X", the review has a planning problem, not a writing problem. Published protocols show what a workable question looks like. A scoping review of nineteenth-century legal textbooks states its question as "What books were employed to teach law in universities in Anglophonic and Hispanophone countries in the nineteenth century?" [1]. A Campbell review asks whether life skills education and psychosocial interventions reduce symptoms of depression and anxiety among forcibly displaced persons in low- and middle-income countries, compared with inactive, usual care, or active control conditions [2]. Both questions name who is being studied, what is being measured, and against what comparison. This article walks through the failures that occur most often when people plan a literature review, explains the mechanism behind each one, and gives concrete fixes. It ends with a diagnostic sequence, a troubleshooting table, and prevention habits you can adopt for your next project. ## The Most Common Cause: A Question That Exists Only in Your Head A review question that lives only in the author's head produces a characteristic cascade of failures. Because the question is vague, the search terms feel arbitrary. Because the search terms are arbitrary, the author cannot explain why certain papers were included and others were not. Because inclusion decisions cannot be explained, the write-up stalls, and the author compensates by reading more and more papers in the hope that a structure will emerge on its own. This is the single most frequent pattern behind common mistakes in literature review planning. The mechanism is worth understanding. A literature review is a research project in miniature. Errors introduced at the planning stage propagate into every later stage, which is the same principle that statisticians emphasize for experimental research, where mistakes made during study design negatively affect all subsequent stages of the work [3]. In a review, the planning stage includes the question, the eligibility criteria, the search strategy, and the plan for screening and extraction. If any of these is improvised, the review loses the property that separates a review from an essay, namely reproducibility. A reader should be able to repeat your systematic literature search and land on roughly the same set of papers. Worked examples make the standard concrete. One scoping review protocol used the PCC mnemonic, which stands for population, concept, and context, to formulate its research question about point-of-care ultrasound for measuring gastric residual volume, then specified six bibliographic databases plus gray literature sources in advance [4]. A pediatric systematic review on intrapleural fibrinolytic therapy defined "early" as within 24 hours of confirmed empyema and "later" as after 24 hours, so the central comparison was fixed before searching began [5]. That is the level of precision you are aiming for. ## A Five-Step Diagnostic Sequence ![A short yes or no path that checks whether a review question names its population, concept and context before any database search begins.](/article-images/illustrations/common-mistakes-in-literature-review-planning-question-triage.webp) Run this sequence before writing anything else. It takes under an hour and will tell you which pitfall you are facing. 1. Write the question in one sentence. Include the population, the exposure or intervention, the comparison if one exists, and the outcome or concept of interest. 2. Test the question against a framing tool. For intervention questions, use PICO, which stands for population, intervention, comparison, outcome. For scoping questions about how a concept appears in the literature, use PCC, population, concept, context, as used in a published ultrasound protocol [4]. 3. Match the question to a review type. Ask whether you need an exhaustive synthesis of controlled studies, a map of what exists, or an argument built from selected sources. Section 5 covers this choice. 4. Check for a written protocol. If you have no document specifying databases, search strings, eligibility criteria, screening plan, and extraction fields, that absence is your next task, not your search. 5. Audit any search you have already run. List the databases, the date the search was run, the exact strings, and any language, date, or publication-type filters. If this list does not exist, your search cannot be reported or reproduced. If step 1 fails, fix the question first. Nothing else in the plan can be stable until the question is stable. ## Troubleshooting Table: Symptoms, Causes, and Fixes | Symptom | Likely cause | How to confirm | Fix | |--|--|--|--| | The search returns thousands of loosely related papers | The question is too broad, with undefined population or outcome | Check whether every hit shares the same population and outcome of interest | Restrict the question using PICO or PCC elements and define each element explicitly | | Relevant papers keep surfacing after screening has started | Eligibility criteria were never fixed in advance | Look for papers you debated including or excluding without a written rule | Write the criteria, apply them to every record from the start, and document any later change with a reason and date | | A supervisor or reviewer asks why an obvious database was missing | The systematic literature search was assembled ad hoc | Compare your database list against the major databases for your field | Plan a multi-database search plus gray literature before searching, as published protocols do [4] [5] | | Key papers are missing and colleagues keep finding them | The search used only one vocabulary or one database | Test whether the missing papers contain synonyms absent from your strings | Build strings from synonyms and controlled vocabulary, and search multiple databases plus trial registries where relevant [6] | | Two screeners keep disagreeing about the same records | Subjective judgment applied by a single uncalibrated rater | Have two people screen a sample independently and compare | Screen in duplicate, and hold group discussions to resolve and correct classification mistakes [7] [8] | | You cannot describe your methods in the write-up | Steps were taken without documentation | Try to reconstruct the number of records identified, screened, and included | Restart documentation from the current point, record every decision, and follow a reporting checklist | | The draft reads like a list of summaries with no argument | No organizing framework was chosen for synthesis | Check whether each paragraph has an analytical job beyond summarizing | Select a framework or conceptual structure before extraction, as protocol authors do [9] | | The scope keeps growing with each new paper | No defined boundaries on population, time, or context | Check whether recent additions would have failed your original question | Return to the written question and protocol, and either enforce the boundary or revise the protocol formally | ## Pitfall 1: An Unfocused Research Question This is the most common failure and the root of most others. The mechanism is that an unfocused question gives you no criterion for stopping. Every search feels incomplete because "complete" was never defined. Authors in this situation often respond by widening the search, which multiplies the screening burden without improving the answer. The fix is to force each element of the question to be explicit. Guidance from trial planning makes the same demand of clinical research questions, which should precisely delineate the target population, clearly define the intervention and comparator, and pre-specify a primary endpoint that addresses the knowledge gap [10]. A review question carries the same obligations. Compare "What is known about discharge planning for injured children?" with the version used in a real protocol, which asks about the perspectives of caregivers and service providers on the hospital discharge process for injured children under 18, including challenges, improvement opportunities, and inequities, framed by a specific access-to-healthcare model [9]. The second version tells you which databases to search, which study designs to accept, and which extraction fields you will need. A practical test is to hand your written question to a colleague and ask what they would search. If their search terms differ meaningfully from yours, the question is still ambiguous. ## Pitfall 2: Searching Before You Have a Written Protocol A protocol is a document, written before the review begins, that specifies the question, eligibility criteria, search strategy, screening process, data extraction plan, and method of synthesis. Skipping it is the second most common mistake, and it is the mechanism behind most irreproducible reviews. Once searching has begun, every decision about what counts as evidence is made under pressure, and pressure pushes decisions toward whatever paper is in front of you. Published protocols show the standard. A scoping review of legal pedagogy literature published its a priori protocol detailing the planned methodology before conducting the review, following Joanna Briggs Institute guidance [1]. A Campbell systematic review protocol stated its primary and secondary research questions, its population of refugees, asylum seekers, and internally displaced persons aged 13 and older, and its restriction to controlled studies, all before searching [2]. An umbrella review on climate change and mental health registered its protocol in PROSPERO, the international prospective register of systematic reviews, before data collection [11]. Registration matters for a reason beyond tidiness. It creates a timestamped record of what you planned to do, which protects you from the accusation that inclusion criteria were shaped by the results. For reviews intended for publication, registration in PROSPERO or, for scoping reviews, platforms such as the Open Science Framework is now expected. The team updating the scoping review reporting guideline registered its own protocol with the Open Science Framework and in a peer-reviewed journal before beginning [8]. If you are writing a thesis or course assignment rather than a publishable review, the same logic applies in a lighter form. A two-page protocol memo, shared with your advisor before searching, delivers most of the benefit. The deeper principle, documented across medical research, is that involving methodological expertise early in design prevents mistakes that later stages cannot undo [3]. ## Pitfall 3: A Shallow Systematic Literature Search The third pitfall is a search that relies on one database, one vocabulary, or one search string. Its mechanism is that databases index different journals and use different controlled vocabularies, so any single source is a biased sample of the evidence. A review built on one database systematically misses whatever that database does not index. Published reviews set the benchmark for a systematic literature search. A scoping review of point-of-care ultrasound planned searches in MEDLINE, CINAHL, Embase, Scopus, the Cochrane Library, and LILACS, plus gray literature sources, which are materials produced outside commercial journal publishing such as reports, theses, and preprints [4]. A pediatric meta-analysis searched five bibliographic databases, two gray literature databases, and four clinical trial registries, with no filters or restrictions applied [5]. A systematic review of delirium prevention searched MEDLINE, Embase, the Cochrane Central Register, Web of Science, ClinicalTrials.gov, and the WHO International Clinical Trials Registry Platform from inception to the planned search date, without language or date restrictions [6]. Even a narrowly scoped review, examining decentralization in single-party states, ran a structured PRISMA search across PubMed, Embase, and Web of Science, examined 859 articles, and included 18 studies [12]. Two practical habits produce most of the gain. First, write search strings as synonym sets joined with OR, for example ("hospital discharge" OR "care transition" OR "discharge planning"), and combine concept blocks with AND. Second, document the exact string and the date for every database, because reporting guidelines require it and because you will need the numbers for the screening flow diagram. Search breadth has a real cost, and the numbers show it. A scoping review of platform trials screened 32,062 records to find nine eligible platform trials [13]. A scoping review conducted for the reporting guideline update screened 8,265 records to identify 43 usable sources [8]. Plan your screening capacity against realistic yields rather than assuming the search will return a manageable handful. ## Pitfall 4: Eligibility Criteria Decided While Screening Eligibility criteria are the written rules that determine which records enter the review. The classic mistake is to begin screening with only a rough sense of what qualifies and to decide borderline cases on the fly. The mechanism is subtle. When you have read a paper and found it interesting, the natural pull is to include it, and criteria adjusted after reading stop being criteria. They become retrospective justifications. The fix is to write inclusion and exclusion criteria as part of the protocol, in the same explicit terms as the question. A discharge-process scoping review specified in advance that eligible studies would report caregiver or service provider perspectives, would investigate discharge improvement interventions, and would be published in English between 2000 and 2025 [9]. A diagnostic meta-analysis pre-specified that included studies must assess deep learning or hand-crafted radiomics models, use MRI or ultrasound, and confirm axillary lymph node status against histopathology as the reference standard [14]. A pediatric therapy review implemented a three-step screening and selection process that was planned before the search ran [5]. Two details deserve care. Decide in advance who resolves disagreements between screeners and how the resolution is recorded. And if you must change a criterion mid-review, which does happen, record the change, the reason, and the date, then report it. Silent criterion changes are the version of this mistake that reviewers and examiners punish hardest. ## Pitfall 5: Choosing the Wrong Type of Review ![Three review types compared by the question they answer, their search approach, whether a protocol is needed and an example from the article.](/article-images/illustrations/common-mistakes-in-literature-review-planning-review-type-choice.webp) Reviews come in types, and mismatching type to question wastes months. A narrative review summarizes and interprets selected literature to build an argument. A systematic review identifies all studies answering a defined question through a reproducible search. A scoping review maps the extent, range, and nature of literature on a topic. An umbrella review synthesizes existing systematic reviews. The mechanism of failure here is aspiration mismatch. Authors want the authority of a systematic review without committing to its method, or they need a mapped overview but write an essay. A narrative review of neuroimaging and artificial intelligence in neurodegenerative disease is a legitimate example of the genre. Its sources were selected from peer-reviewed journals focused on clinical applications and technical advances, without a claim to exhaustive or reproducible coverage [15]. That is acceptable for a narrative review and unacceptable if the paper claims to be systematic. The failure mode is not the method, it is the label. Choose by question shape. If the question asks whether an intervention works compared with something else, a systematic review with meta-analysis is the appropriate type. If the question asks what literature exists and how it is distributed, as in a review mapping which books taught law across two linguistic traditions in the nineteenth century, a scoping review with a published protocol is the right form [1]. If you are summarizing a fast-moving technical field to set context, a narrative review is defensible, provided you describe your selection approach honestly [15]. ## Pitfall 6: One Person Judging Every Paper Screening and data extraction involve judgment, and unexamined judgment varies between people. The classic setup for a solo author is to screen every record alone, which feels efficient and quietly introduces error. The mechanism was quantified in a study of classification reliability in ecology reviews. Five independent raters rated 23 variables across 21 peer-reviewed publications on conservation management plans, and mistakes, including overlooking information in the text, were the most common source of disagreement between them. Group discussion about the coded categories improved both reliability and accuracy, and unlike agreement statistics, discussion allowed mistakes to be identified and corrected [7]. The fix has two parts. For reviews intended for publication, screen at least a sample of records in duplicate, with two people screening independently, which is what the reporting guideline update team did for its 8,265 screened records [8]. For solo projects such as theses, calibrate yourself instead. Screen a sample of records twice, separated by at least a week, and compare your own decisions. Where your two passes disagree, write down why. Then hold structured discussions with peers or your advisor on the difficult categories, applying the finding that discussion corrects the mistakes that single-pass rating conceals [7]. Data extraction carries the same risk as screening. Build an extraction form before reading any included paper, with fields defined by your question and, where relevant, by a guiding framework such as the access-to-healthcare model used in the discharge review [9]. ## Pitfall 7: No Plan for Critical Appraisal Critical appraisal means judging how much a study's methods support its conclusions. Reviews that skip it treat all included studies as equally trustworthy, which flattens the synthesis and lets weak studies carry the same weight as strong ones. The mechanism is that appraisal feels like an optional add-on, when it is actually part of the protocol that determines what your conclusions can claim. The fix is to choose an appraisal tool when you write the protocol, matched to your study types. The decentralization review assessed quality with the MMAT 2018, a mixed-methods appraisal tool, and the SANRA scale for narrative reviews [12]. The climate and mental health umbrella review applied the Joanna Briggs Institute critical appraisal checklist and set an a priori threshold, including only reviews scoring at least 5 out of 11 [11]. A therapy review planned to rate the certainty of its evidence using the GRADE approach [5]. None of these tools was improvised during analysis. Each was named in the plan. A related trap is appraisal used selectively. Applying a quality tool only to studies you dislike, or using quality scores to discard inconvenient findings, converts appraisal into advocacy. Apply the tool to every included study by the same rule stated in the protocol. ## Pitfall 8: Treating Reporting Guidelines as an Afterthought The rarest but most expensive pitfall is discovering reporting requirements at the end. Reporting guidelines such as PRISMA, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses, define what a complete review must document, and extensions exist for specific types. The PRISMA-ScR extension covers scoping reviews, and it is under active revision, with a published scoping review identifying candidate new items related to automation and data extraction to incorporate into the update [8]. Protocol authors build compliance in from the start. A scoping review of ultrasound assessment reported that it would follow PRISMA-ScR and JBI methodology, which shaped its planning rather than merely describing it afterward [4]. A platform trials review did the same, stating its JBI methodology and PRISMA-ScR compliance in the protocol [13]. The mechanism of failure is sequencing. If you learn the reporting checklist only during writing, you will find gaps in your documented search dates, screening counts, and exclusion reasons that can no longer be filled honestly. Reading the relevant checklist before you search, and using it as your documentation template, removes that risk at zero cost. ## Common Mistakes and Limitations Several residual mistakes deserve direct mention because they survive even careful planning. First, planning documents are not self-executing. A protocol that nobody consults during screening provides no protection. Schedule a midpoint audit against your own protocol. Second, thoroughness has diminishing returns. Searching six databases plus registries and gray literature is the published standard for many questions [4] [6] [5], but the yield drops steeply after the first few sources. If your question is narrowly scoped, a justified, documented search of the three or four core databases for your field may be sufficient, and over-searching can consume the time you need for synthesis. Third, date and language limits are tradeoffs, not sins. A delirium prevention review deliberately searched from inception without language or date restrictions to maximize completeness [6]. The decentralization review restricted to English and to 2013 through 2023, a scope decision made for feasibility [12]. Either choice is defensible when it is declared in the protocol and its bias direction is acknowledged. The mistake is making the restriction silently, during screening, when a promising paper turns out to be in another language. Fourth, a single review cannot answer every question a reader might bring to the topic. Scope boundaries that exclude relevant adjacent questions are a feature of a feasible review, not a flaw, provided they are stated. Readers bringing individual cases or clinical decisions to a review should consult appropriate professional guidance for their situation, because a literature review synthesizes research and does not replace expert judgment on a specific case. Finally, expect disagreement even under good conditions. Trained experts rating the same evidence reach different judgments at meaningful rates, and the response is structured calibration and discussion, not the assumption that any one pass is definitive [7]. ## Prevention Practices The habits below prevent the pitfalls in this article and compound across projects. 1. Write the question in one sentence using PICO or PCC before any searching [4]. 2. Draft a two-page protocol covering question, eligibility, databases, search strings, screening plan, extraction fields, and appraisal tool. Register it if the review is intended for publication [11]. 3. Plan the systematic literature search as synonym sets, run it in at least three core databases for your field plus one gray literature source, and record the string and date for each [4] [5]. 4. Screen in duplicate where possible, or self-calibrate on a repeat sample, and resolve disagreements through discussion [7] [8]. 5. Log every decision with a date and a reason, and treat any mid-review criterion change as a documented protocol amendment. 6. Download the relevant reporting checklist, such as PRISMA or PRISMA-ScR, before searching, and use it as your documentation spine [8]. A review planned this way is slower in week one and faster in every week after, because the write-up becomes a matter of reporting decisions already made rather than reconstructing them. ## Frequently Asked Questions ### What is the most common mistake in literature review planning? The most common mistake is starting to search before the review question has been written down with its population, concept, and context defined. Every later step inherits the ambiguity, so fix the question first and only then build the search. ### How do I choose between a narrative, scoping, and systematic review? Match the review type to the question shape. Questions about whether an intervention works call for a systematic review. Questions about what literature exists call for a scoping review. Context-setting overviews of a technical field can be narrative reviews, as long as the selection approach is described honestly. ### How many databases should a systematic literature search cover? Published protocols commonly search four to six bibliographic databases plus gray literature sources, and trial registries for intervention questions. At minimum, cover the core databases for your field, document the exact strings and dates, and justify any database you leave out. ### Do I need to register a protocol for a thesis or course assignment? Formal registration in PROSPERO or a similar registry is expected for publishable reviews. For a thesis or assignment, a written protocol memo shared with your advisor before searching delivers most of the same protection against improvised decisions. ### What is gray literature and why does it matter? Gray literature is material published outside commercial journals, including reports, theses, preprints, and registry entries. Including it reduces the risk that your review reflects only the published and typically positive results that journals select for. ### How should I handle language and date restrictions? Decide them in the protocol, not during screening. Searching from inception without language or date restrictions maximizes completeness, while narrower limits improve feasibility. Either choice is defensible if declared in advance and acknowledged as a limitation. ### What should I do when two screeners disagree about a paper? Do not average the judgments. Discuss the record together, consult the written eligibility criteria, and record the resolution and its reason. Group discussion has been shown to improve both reliability and accuracy of classifications and to catch outright mistakes such as overlooked text. ### When should I stop adding papers and start writing? Stop when your documented search and screening are complete under your protocol criteria, not when you feel informed enough. If new relevant papers keep appearing after screening closes, run a formal update search, report it, and revise the criteria only as a documented amendment. ## Sources 1. [Mapping the literature employed to teach law in universities in Anglophonic and Hispanophone countries in the nineteenth century to explore the circulation of legal ideas: Protocol for a scoping review.](https://pubmed.ncbi.nlm.nih.gov/42945330/) 2. [PROTOCOL: Life Skills Education and Psychosocial Interventions for Anxiety and Depression in Forcibly Displaced Persons in LMICs: A Systematic Review.](https://pubmed.ncbi.nlm.nih.gov/42282092/) 3. [Statistical errors in medical research - a review of common pitfalls.](https://pubmed.ncbi.nlm.nih.gov/17299669/) 4. [Point-of-Care Ultrasound in the Assessment of Gastric Residual Volume: Protocol for a Scoping Review.](https://pubmed.ncbi.nlm.nih.gov/42046538/) 5. [Early Versus Late(r) Administration of Intrapleural Fibrinolytic Therapy in Children with Empyema: A Systematic Review and Meta-Analysis.](https://pubmed.ncbi.nlm.nih.gov/42430023/) 6. [Dual orexin receptor antagonists for preventing delirium in hospitalized older adults: protocol for a randomized-trial systematic review and meta-analysis.](https://pubmed.ncbi.nlm.nih.gov/42437946/) 7. [Group discussions improve reliability and validity of rated categories based on qualitative data from systematic review.](https://pubmed.ncbi.nlm.nih.gov/40531894/) 8. [Updating the PRISMA reporting guideline for scoping reviews: a scoping review.](https://pubmed.ncbi.nlm.nih.gov/42107664/) 9. [Perspectives on the injured child's hospital discharge process and opportunities for improvement: a scoping review protocol.](https://pubmed.ncbi.nlm.nih.gov/42303396/) 10. [Planning and feasibility of phase III surgical trials in gynecologic oncology.](https://pubmed.ncbi.nlm.nih.gov/42710321/) 11. [Impacts of climate change on mental health and its underlying mechanisms: an umberella review.](https://pubmed.ncbi.nlm.nih.gov/42121234/) 12. [Decentralisation of the Healthcare Systems and the Functioning of Public Health Programmes in Single-Party States: A Systematic Review.](https://pubmed.ncbi.nlm.nih.gov/42160024/) 13. [Platform trials in anaesthesia and perioperative medicine: a scoping review.](https://pubmed.ncbi.nlm.nih.gov/42282875/) 14. [Assessing the performance of deep learning and hand-crafted radiomics models using MRI and ultrasound in predicting axillary lymph node status in breast cancer: a systematic review and meta-analysis.](https://pubmed.ncbi.nlm.nih.gov/42319409/) 15. [Advancements in Imaging Technologies and AI Integration for Neurodegenerative Disease Management: A Narrative Review.](https://pubmed.ncbi.nlm.nih.gov/41777595/)