Quality Assessment

Risk of bias assessment with four validated tools (RoB 2.0, ROBINS-I, NOS, QUADAS-2) and GRADE evidence certainty rating — AI-assisted, human-confirmed.

Quality Assessment

Study quality isn't a checkbox. It's the foundation of credible evidence synthesis.

Including low-quality studies without appropriate assessment undermines the entire purpose of a systematic review. Peer reviewers scrutinize this step more than almost any other. mapped integrates four validated risk of bias tools and the GRADE framework into a single, consistent workflow — with AI assistance that ensures reliability across your entire review.


Risk of Bias Assessment

mapped supports the four major risk of bias instruments, each designed for a specific study type:

RoB 2.0 — Randomized Controlled Trials

The Cochrane Risk of Bias tool 2.0 assesses five domains:

  1. Randomization process — was allocation truly random and concealed?
  2. Deviations from intended interventions — were there protocol deviations?
  3. Missing outcome data — was attrition balanced and handled appropriately?
  4. Outcome measurement — were assessors blinded?
  5. Selection of reported result — is there evidence of selective reporting?

Each domain receives a judgment: Low risk, Some concerns, or High risk. mapped's Advanced Reasoning Engine generates preliminary assessments with supporting rationale, which you review and confirm per domain.

ROBINS-I — Non-Randomized Studies of Interventions

For cohort studies, case-control studies, and other non-randomized designs:

  1. Confounding — were important confounders controlled?
  2. Selection of participants — was selection related to intervention and outcome?
  3. Classification of interventions — was intervention status well-defined?
  4. Deviations from intended interventions — similar to RoB 2.0 but for non-randomized contexts
  5. Missing data — as above
  6. Measurement of outcomes — as above
  7. Selection of reported result — as above

Newcastle-Ottawa Scale (NOS) — Observational Studies

A star-based system (maximum 9 stars) assessing:

  • Selection (4 stars) — representativeness, selection of controls, ascertainment of exposure
  • Comparability (2 stars) — controlling for confounders
  • Outcome (3 stars) — outcome assessment, follow-up duration and adequacy

QUADAS-2 — Diagnostic Accuracy Studies

For studies evaluating diagnostic tests:

  1. Patient selection — appropriate spectrum? Consecutive enrollment?
  2. Index test — was the threshold pre-specified?
  3. Reference standard — is it likely to correctly classify the condition?
  4. Flow and timing — did all patients receive the reference standard? Appropriate interval?

AI-Assisted Assessment

mapped's Advanced Reasoning Engine assists with risk of bias assessments by:

  • Reading the full text of each included study
  • Generating domain-level judgments with supporting quotes from the paper
  • Identifying potential concerns that human reviewers might overlook (e.g., selective reporting patterns)
  • Maintaining consistency across studies — applying the same standards to study #1 and study #50

Every AI-generated judgment is presented alongside the source text. You accept, modify, or override each one. The dual-reviewer protocol applies here too — two independent assessments, followed by conflict resolution.


GRADE Evidence Certainty

After assessing individual study quality, the GRADE framework rates the overall certainty of evidence for each outcome:

Downgrade Factors

FactorQuestion
Risk of biasAre the included studies at high risk of bias?
InconsistencyDo results vary substantially across studies?
IndirectnessDoes the evidence directly address your question?
ImprecisionAre confidence intervals wide? Small sample size?
Publication biasIs there evidence that studies are missing?

Upgrade Factors

FactorQuestion
Large effectIs the effect size so large that confounding alone can't explain it?
Dose-responseIs there a clear dose-response gradient?
Residual confoundingWould residual confounders reduce rather than increase the effect?

mapped walks you through each factor for each outcome, producing a final rating: High, Moderate, Low, or Very Low.


Summary of Findings Tables

mapped generates Summary of Findings (SoF) tables that combine:

  • Effect estimates from your meta-analysis
  • GRADE certainty ratings per outcome
  • Absolute risk differences (when applicable)
  • Footnotes explaining downgrade/upgrade decisions

These tables follow Cochrane formatting and are ready for manuscript insertion.


Visualizations

Risk of bias results are automatically visualized:

  • Traffic light plots — domain-level judgments per study (green / yellow / red)
  • Summary bar charts — proportion of studies at each risk level per domain
  • Risk of bias summary tables — tabular format for manuscripts

All visualizations are exportable to PNG, SVG, or directly inserted into your manuscript draft.


Why This Step Matters

Reviewers and editors increasingly reject systematic reviews with superficial quality assessments. "All studies were assessed using RoB 2.0" isn't enough — you need domain-level justifications, inter-reviewer agreement documentation, and transparent GRADE judgments. mapped provides all of this in a structured, auditable workflow.


Next step: With quality assessed, move to Meta-Analysis →