Five Mistakes That Weaken a Dissertation Literature Review

A literature review chapter rarely fails because the prose is poor. It fails because of things settled before the first paragraph was written: a review type chosen by default, a search built in an afternoon, a search nobody can reconstruct, a review that duplicates one already published, and the whole thing done alone. Juarez and Honkaniemi, writing in Pedagogy in Health Promotion, describe exactly this pattern in master’s theses — time and resource constraints push students into choosing review types without clear guidelines, which compromises transparency and reproducibility, and poor-quality theses or lower grades follow. The five mistakes below are the operational form of that problem.

What the five have in common

All five are upstream errors. None of them produces a visible problem while you are writing; each produces an unanswerable question at the viva or in peer review. That asymmetry is why they are worth naming individually rather than being folded into general advice about rigour.

They also share a cause. Juarez and Honkaniemi, in Pedagogy in Health Promotion, trace the pattern in master’s theses to time and resource constraints: students choose review types without clear guidelines, which compromises transparency and reproducibility, and a lack of awareness of alternative review types and of pedagogical guidance results in poor-quality theses or lower grades through unrealistic expectations 1. They note the problem is worse where the course sets no specific learning outcomes or grading criteria for literature reviews — which is to say, where nobody has told the student what the chapter is being marked on. If your programme has not specified that, you are working without a brief, and the five decisions below are the ones to make explicitly for yourself.

The practical framing: for each mistake, there is a decision you can make in writing this week, and a record you can keep that makes the decision checkable later. That is the whole of the fix. Our page on what a literature review is for covers the argumentative side of the chapter; this one is about the machinery underneath it.

Mistake 1: letting the review type be chosen for you

The commonest version of this is not choosing at all. The chapter gets written as “the literature review”, the methods paragraph describes a search in general terms, and the word systematic appears somewhere in the text without any of the apparatus that would make it true. The result is a chapter whose conclusions nobody can calibrate, because the reader cannot tell what the search was supposed to achieve.

Review type is a set of commitments. Grant and Booth’s typology in the Health Information and Libraries Journal characterised 14 review types by what each requires on search, appraisal, synthesis and analysis 29. Sutton, Clowes, Preston and Booth went further in the same journal, mapping the information-retrieval requirements that come with each member of the review family 24. Choosing a type is therefore choosing how exhaustive your search must be, whether you appraise included studies formally, and what you may claim at the end.

Two practical points. First, a scoping review is not a systematic review with a lower bar — it answers a mapping question and is reported differently, which is why PRISMA-ScR exists as a separate extension 27,31. Choosing it deliberately is respectable; drifting into it because the search got large is not. Second, write the type and its justification into your methods before you draft, including the honest constraint. “A scoping review was chosen because the aim was to map the extent and nature of the literature rather than to estimate an effect” is a defensible sentence. “A systematic review was conducted”, in a chapter with a one-database search and no appraisal step, is not. See scoping review and systematic review for what each actually involves.

Mistake 2: a search strategy with errors in it

This is the mistake most students do not know is possible, because a flawed search still returns results. It returns the wrong ones, or too few, and nothing in the output signals that.

The scale is documented. Salvador-Olivan, Marco-Cuenca and Arquero-Aviles examined the search strategies of 137 published systematic reviews in the Journal of the Medical Library Association and found that 92.7% contained at least one error, and 78.1% contained errors that reduce recall 23. These were published reviews, which have passed peer review. The errors are the ordinary mechanical kind: Boolean operators in the wrong precedence, AND used where OR was meant, truncation applied to a term stem that captures the wrong words, controlled vocabulary used without checking the thesaurus for the actual indexing term, field tags misapplied so a term is searched in the wrong field, spelling and plural variants omitted, and filters applied that silently exclude eligible records 23.

There is a second, less obvious failure. Burns, Shapiro, Nix and Huber, also in the Journal of the Medical Library Association, compared the same MEDLINE search run across five host platforms and found materially different results. MEDLINE accessed through different interfaces is not one searchable object; syntax, field behaviour, mapping and update frequency differ 25. So the platform is part of the method, and a methods section that names the database without the interface has under-specified the search 20,25.

Catching these requires reading the strategy line by line against what each line was supposed to do, checking retrieved counts per line for implausible drops, and testing the strategy against a handful of papers you already know should be found. If a known-eligible paper is not retrieved, the strategy has a defect and the defect is findable. Search strategy peer review is the formal version of this check.

Mistake 3: a search nobody can reconstruct

A search that cannot be re-run cannot be verified, and an unverifiable search makes every claim in the chapter provisional. This is a separate failure from mistake 2: a perfectly constructed search, badly recorded, is not usable evidence.

There is a specification for what to record. PRISMA-S, published by Rethlefsen and colleagues in Systematic Reviews, sets out sixteen reporting items, each with exemplar reporting and a rationale, so that every component of the search can be checked as completely reported and therefore reproducible 20. It covers the information sources, the dates, the full strategies as run rather than paraphrased, limits and filters, and the supplementary methods — reference checking, citation searching, contacting authors, grey literature, trial registers.

The item that causes the most trouble is the full strategy. A narrative description of a search is not a search: “we searched MEDLINE and Embase using terms relating to the intervention and the population” cannot be re-run, and nobody can tell from it whether mistake 2 is present. The full line-by-line strategy for each database, as actually executed, belongs in an appendix 20,21. The practical consequence is that this has to be captured live. Reconstructing a search months later produces a plausible strategy that is not the one you ran, and the counts will not match. Our page on reporting guidelines for literature reviews covers which guideline applies to which review type, and what PRISMA is and is not covers the common misuse of the main statement.

Mistake 4: writing a review that already exists

Checking for existing reviews is a one-afternoon task that students routinely skip, on the assumption that their question is specific enough to be unclaimed. The evidence says otherwise.

An analysis of redundant systematic reviews published in Systematic Reviews examined 1,054 registered COVID-19 protocols and found that 13% closely duplicated an existing protocol 30. The same work documents how little of the field is registered at all: 32% of systematic reviews published in 2018 reported a PROSPERO registration, and only 25% of COVID-19 treatment reviews were registered. Those two facts compound. Low registration means the registry undercounts work in progress, so searching PROSPERO alone understates how crowded your question is, and your own non-registration means somebody else will fail to find you 30.

Finding a prior review is not fatal and is often useful. It gives you a tested search strategy, a defined inclusion set and a dated boundary. The legitimate responses are to update it with a declared cut-off and a stated reason, to narrow or re-angle the question, to address a methodological weakness in the existing review explicitly, or to change the review type where the prior work left a different question open. What is not defensible is producing a near-duplicate without acknowledging the predecessor, because a reader who finds it will assume you did not look 30. Register your own protocol, and see duplicate publication for the related integrity problem at the publishing end.

Mistake 5: doing all of it alone

A dissertation is an individual piece of work, which is often read as meaning every step must be performed single-handed. The steps where a second person changes the result are specific and few, and they are worth arguing for with a supervisor.

Search design is the first. Whitney and colleagues, in the Journal of the Medical Library Association, analysed 505 reviews and found librarian collaboration associated with more reproducible searches and better reported methods 4. Their time trend is striking in both directions: involvement rose from 9.0% of reviews in 2010 to 10.4% in 2015 and 20.7% in 2021, so it is becoming normal while remaining a minority practice. Read alongside the 92.7% error rate Salvador-Olivan and colleagues found, the implication is straightforward — search construction is a specialist skill, most reviews are built without that skill, and most reviews contain errors 23.

Screening is the second. Independent duplicate screening of titles, abstracts and full texts, with a stated procedure for resolving disagreement, is what makes an inclusion set something other than one person’s reading 3,21. Where a second screener is genuinely unavailable, the honest alternatives are to double-screen a documented sample to estimate your own error rate, to have a supervisor check all exclusions at full-text stage, and to state in the methods that screening was single-handed. A declared limitation is survivable; an undeclared one found by an examiner is not. How abstract screening works end to end sets out the procedure, and handling discrepancies during data collection covers what to do when two people disagree.

What a methodological reviewer checks first

Examiners and methodological peer reviewers triage. Knowing the order they work in tells you where to spend your remaining time.

They start with whether the review type claimed in the title matches the methods described, because that single comparison exposes mistakes 1 and 3 at once. Then the search: which databases, which interfaces, what dates, and whether the full strategy is available to inspect. Then the numbers — whether the counts reconcile from records identified through to studies included, and whether full-text exclusions carry reasons 21. Then screening: how many people, independently or not, and how disagreement was resolved 34. Then whether prior reviews were identified and discussed. Only after that do they read the synthesis as prose.

The implication is uncomfortable but useful: the chapter is assessed largely on its apparatus before it is assessed on its argument. A beautifully written review with an unreconstructable search gets stopped at the second checkpoint. A plainly written one with a documented search, declared limitations and reconciled numbers gets read on its merits.

Repairing a chapter you have already written

If you are reading this with a draft already in hand, the five mistakes are not equally repairable, and it is worth knowing which is which.

Mistake 3 is almost always fixable: re-run your search, record it to the PRISMA-S items this time, and report the re-run date. You lose a day and gain a reproducible method. Mistake 1 is usually fixable by re-labelling honestly — if the chapter is a narrative or scoping review, call it that, define it, and adjust the conclusions to what that type supports. Mistake 4 is fixable at any point by finding the prior reviews and positioning against them explicitly; doing this late is awkward but far better than not doing it.

Mistake 2 is fixable but costs real time, because correcting a search means re-screening what the corrected search returns. Do it anyway if a known-eligible paper was missed, because that defect will be found. Mistake 5 is the one that cannot be retrofitted: you cannot make single-handed screening independent after the fact. What you can do is double-screen a sample now to quantify the problem, have a second person re-check full-text exclusions, and state the limitation in the methods with that quantification attached. An examiner who sees a limitation measured rather than confessed reads it as competence. Thesis editing and literature review support are where we usually pick this up.

Take-home message

How to apply this to your study

Common questions

@misc{list_the_five_mistakes_that_may_happen_2026,
  author       = {{Pubrica Academy}},
  title        = {Five Mistakes That Weaken a Dissertation Literature Review},
  howpublished = {Pubrica},
  year         = {2026},
  month        = {oct},
  url          = {https://pubrica.com/academy/research/list-the-five-mistakes-that-may-happen-while-writing-an-effective-literature-review-for-biomedical-research-dissertation/},
  note         = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}

LITERATURE REVIEW: FIVE-MISTAKE AUDIT

1. REVIEW TYPE
   Type claimed in the title: ______________________
   Type the methods actually describe: ______________________
   [ ] These are the same
   [ ] The type is defined in the methods with a justification
   [ ] Every conclusion is one this review type supports
   If the choice was driven by time or resources, write that here:
   ______________________________________________________

2. SEARCH CONSTRUCTION
   Known-eligible papers used as a test set: ____ of ____ retrieved
   [ ] Boolean precedence checked with parentheses on every line
   [ ] OR used within synonym blocks, AND only between concepts
   [ ] Truncation stems checked against retrieved titles
   [ ] Every concept checked against the database thesaurus
   [ ] Field tags verified
   [ ] Per-line retrieval counts inspected for implausible values
   [ ] Spelling, plural and synonym variants included
   [ ] Every limit and filter listed WITH a justification
   Lines I could not account for: ______________________

3. SEARCH REPORTING  (the sixteen PRISMA-S items)
   [ ] Every source named WITH its interface/platform
   [ ] Coverage dates given (inception or from-date, to-date)
   [ ] Date the search was run, and re-run date if applicable
   [ ] Full line-by-line strategy for EVERY database, as executed
   [ ] Records retrieved per database reported
   [ ] Limits and filters reported
   [ ] Supplementary methods listed individually:
         [ ] backward citation checking  [ ] forward citation checking
         [ ] hand-searching (journals: ______)
         [ ] trial registers (which: ______)
         [ ] grey literature  [ ] author contact
   [ ] Sources NOT searched stated as a limitation
   Appendix number holding the strategies: ______

4. PRIOR REVIEWS
   Registries searched: ______________  Date: ______
   Existing reviews found: ______________________
   My position relative to them:
     [ ] update, with cut-off ______ and reason ______
     [ ] narrower/different question: ______
     [ ] addresses a stated weakness: ______
     [ ] different review type: ______
   [ ] My own protocol is registered.  Registry/number: ______

5. SECOND PERSON
   Search design reviewed by: ______________________
   Title/abstract screening:  ____ screeners  independent? [ ]
   Full-text screening:       ____ screeners  independent? [ ]
   Disagreement procedure: ______________________
   If single-handed:
     Sample double-screened: ____ records, agreement ______
     Full-text exclusions re-checked by: ______________
     [ ] Limitation stated in the methods with the measured agreement