InsilicoΣ
Drug Discovery, Cheminformatics & Bioinformatics
About Our Team Publications Mobile App Contact Us Login Register

About RoB-Σ

RoB-Σ  ·  Risk of Bias Sigma - a comprehensive sum of bias assessment methods

The Σ (capital Sigma) is the mathematical symbol for summation - it represents summing up all aspects of a complex problem into one integrated result. In RoB-Σ, this reflects the tool's function: it does not implement a single instrument in isolation, but consolidates six internationally validated risk-of-bias frameworks (RoB 2, NOS, ROBINS-I, QUADAS-2, QUIPS, AMSTAR 2) into one guided platform - the sum of methodological bias assessment across all major study designs. Every domain judgment, override, and rationale is tracked in a single unified audit trail. The Σ is also the signature of the InsilicoΣ lab, identifying tools that sum up all relevant analytical dimensions rather than addressing them piecemeal.

RoB-Σ is a guided, user-friendly tool for assessing risk of bias in studies included in systematic reviews. It supports six internationally recognized assessment instruments.


Supported Instruments

Study TypeToolJudgment ScaleReference
RCT RoB 2 Low / Some concerns / High Sterne JAC et al. BMJ 2019; 366:l4898
Cohort NOS (Cohort) Low risk / Unclear risk / High risk Wells GA et al. Ottawa Hospital Research Institute
Case-control NOS (Case-Control) Low risk / Unclear risk / High risk Wells GA et al. Ottawa Hospital Research Institute
Cohort, Case-control, Non-randomized intervention ROBINS-I Low / Moderate / Serious / Critical / No information Sterne JA et al. BMJ 2016; 355:i4919
Diagnostic accuracy QUADAS-2 Low / High / Unclear Whiting PF et al. Ann Intern Med 2011; 155(8):529-536
Prognostic QUIPS Low / Moderate / High Hayden JA et al. Ann Intern Med 2013; 158(4):280-286
Systematic review AMSTAR 2 High / Moderate / Low / Critically low Shea BJ et al. BMJ 2017; 358:j4008

How It Works

  1. Create a project for your systematic review
  2. Add studies and select the appropriate assessment instrument
  3. Answer guided questions domain by domain - the algorithm computes judgments automatically
  4. Review and override judgments with justification if needed
  5. Generate figures (traffic light plots, summary bar charts) for your manuscript
  6. Export results as CSV, Excel, or PDF

Algorithm Methodology

RoB-Σ implements each instrument's published decision-tree algorithm rather than a generic counting rule. Polarity, conditional gating, and reverse-coded questions are handled per the source paper. Algorithm output is a guide - every domain and the overall judgment can be overridden by the assessor with a recorded justification.

InstrumentDomain rule sourceOverall rule
RoB 2 Sterne 2019 Box 1: per-domain decision rules with reverse-coded questions (Q1.3, D4, etc.) and routing questions (Q2.1/2.2/4.3) explicitly handled. Sterne 2019 Box 2: worst-case across D1-D5; reviewer is alerted when multiple domains carry "Some concerns".
ROBINS-I Sterne 2016 Table 5: D1 gating (Q1.1 No -> Low directly), Q1.6 post-intervention adjustment -> Serious, Q1.4 inadequate control -> Serious, etc. for D2-D7. Sterne 2016 Table 6: worst-case across the seven domains. "Critical" is never assigned automatically - reserved for the manual override per Sterne's guidance.
QUADAS-2 Whiting 2011: signalling questions uniformly Yes = favourable. Per- domain risk + separate applicability concern. Worst-case across the four domains for risk and applicability.
QUIPS Hayden 2013 does not prescribe a counting algorithm. RoB-Σ uses a transparent heuristic (all Yes -> Low; >=50% No -> High; else Moderate) that the reviewer is encouraged to override. Worst-case across the six domains.
NOS (Cohort / Case-Control) Per Wells: stars awarded per item; comparability stars must be earned sequentially (the second star cannot be granted without the first). Star totals mapped to Low/Unclear/High using AHRQ thresholds (>=7/4-6/<4). Note: the Cochrane Handbook v6.4 ch.25 cautions against dichotomising NOS totals; reviewers should use the override field where the pattern of stars matters more than the count.
AMSTAR 2 Shea 2017: per-item Yes / Partial yes / No / No meta-analysis. "Partial yes" is treated as a non-critical weakness (not a critical flaw). "No meta-analysis" responses on Q11/Q12/Q15 are excluded from the weakness count. Shea 2017 Table 2: thresholds on critical and non-critical weaknesses to assign High / Moderate / Low / Critically low confidence.

Implementation lives in rob_sigma/services/judgment_engine.py. Source rules are cited in the docstring of each domain function. When the engine is updated (current version: 2026-05-12-decision-trees), legacy assessments can be refreshed with manage.py recompute_judgments while preserving every user override.


Key References

  • RoB 2: Sterne JAC et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ 2019; 366:l4898
  • NOS: Wells GA et al. The Newcastle-Ottawa Scale for assessing the quality of nonrandomised studies in meta-analyses. Ottawa Hospital Research Institute.
  • ROBINS-I: Sterne JA et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ 2016; 355:i4919
  • QUADAS-2: Whiting PF et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med 2011; 155(8):529-536
  • QUIPS: Hayden JA et al. Assessing bias in studies of prognostic factors. Ann Intern Med 2013; 158(4):280-286
  • AMSTAR 2: Shea BJ et al. AMSTAR 2: a critical appraisal tool for systematic reviews. BMJ 2017; 358:j4008
AI Lab