ICU Disparnumerophobia and Triskaidekaphobia: The 'Irrational Care Unit'?
Pith reviewed 2026-05-25 11:21 UTC · model grok-4.3
The pith
Intensive care clinicians show a bias toward odd ventilator settings and an aversion to the number 13.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Retrospective analysis of ventilator parameters from adult and paediatric ICUs reveals that clinicians choose odd values for PEEP, respiratory rate and inspiratory pressure at rates significantly higher than chance (ORs 0.16–0.48), while respiratory rate and inspiratory pressure show statistically significant avoidance of the number 13; PEEP settings of 13 were more common than expected.
What carries the argument
Monte Carlo simulation combined with conventional odds-ratio tests to quantify departures from uniform or random number distributions in extracted ventilator settings.
If this is right
- Clinicians should monitor their own number choices when titrating therapy.
- Similar statistical searches could reveal number biases in drug dosing, alarm limits or fluid targets.
- The Monte Carlo approach provides a general tool for detecting digit preferences across clinical data sets.
Where Pith is reading between the lines
- If number biases exist in ventilation, they may also appear in non-numeric decisions such as timing of interventions.
- Outcome studies could test whether the detected patterns correlate with measurable differences in ventilator days or complications.
- Training that makes clinicians aware of the specific numbers they favour might reduce the bias without changing overall care protocols.
Load-bearing premise
The observed ventilator settings primarily reflect clinician number choice rather than being constrained by patient physiology, equipment limits, or unmeasured clinical factors.
What would settle it
A controlled experiment in which clinicians set ventilators on identical simulated patients with no physiological or equipment constraints; persistence of the odd/even and 13 patterns would support the claim, while their disappearance would falsify it.
read the original abstract
Whilst evidence-based medicine is the cornerstone of modern practice, it is likely that clinicians are influenced by cultural biases. This work set out to look for evidence of number preference in invasive mechanical ventilatory therapy as a concrete example of subconscious treatment bias. A retrospective observational intensive care electronic medical record database search and analysis was carried out in adult general, specialist neurosciences and paediatric intensive care units within a tertiary referral hospital. All admitted, invasively mechanically ventilated patients between October 2014 and August 2015 were included. Set positive end-expiratory pressure (PEEP), respiratory rate (RR) and inspiratory pressure (Pinsp) settings were extracted. Statistical analysis using conventional testing and a novel Monte Carlo method were used to look for evidence of two culturally prevalent superstitions: Odd/even preference and aversion to the number 13. Patients spent significantly longer with odd choices for PEEP ($OR=0.16$, $p<2\times10^{-16}$), RR ($OR=0.31$, $p<2\times10^{-16}$) and Pinsp (OR=0.48, $p=2.9\times10^{-7}$). An aversion to the number 13 was detected for choices of RR ($p=0.00024$) and Pinsp ($p=3.9\times10^{-5}$). However a PEEP of 13 was more prevalent than expected by chance ($p=0.00028$). These findings suggest superstitious preferences in intensive care therapy do exist and practitioners should be alert to guard against other, less obvious but perhaps more clinically significant decision-making biases. The methodology described may be useful for detecting statistically significant number preferences in other domains.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript claims to identify evidence of number-based superstitions in the choice of mechanical ventilation parameters (PEEP, respiratory rate, and inspiratory pressure) in ICU patients. Using data from a tertiary hospital's electronic records for ventilated patients from October 2014 to August 2015, it applies odds ratios and a Monte Carlo simulation to demonstrate preferences for even numbers and, in some cases, avoidance of the number 13.
Significance. Should the central result be robust to clinical covariates, the paper would provide concrete evidence that cultural biases influence clinical choices in critical care settings. This has potential implications for training and protocol design to mitigate such biases. The use of an external patient database and a simulation-derived null without fitted parameters is a methodological strength that supports the empirical nature of the finding.
major comments (2)
- [Monte Carlo procedure (abstract and methods)] Monte Carlo procedure (abstract and methods): The null distribution does not incorporate patient physiology, compliance, oxygenation targets, or equipment limits that tightly constrain PEEP/RR/Pinsp choices; this is load-bearing because the central attribution of deviations at odd integers or 13 to superstition requires that the null captures the expected distribution absent number bias.
- [Statistical analysis (abstract)] Statistical analysis (abstract): The reported ORs (e.g., OR=0.16 for PEEP odd/even) and p-values are presented without stratification or regression on clinical covariates, so it is not possible to verify whether the Monte Carlo null has accounted for the unmodeled dependencies that the skeptic identifies as producing exactly the observed non-uniform marginals.
Simulated Author's Rebuttal
We thank the referee for the constructive comments on our manuscript. We address each major comment below and outline planned revisions where appropriate.
read point-by-point responses
-
Referee: [Monte Carlo procedure (abstract and methods)] Monte Carlo procedure (abstract and methods): The null distribution does not incorporate patient physiology, compliance, oxygenation targets, or equipment limits that tightly constrain PEEP/RR/Pinsp choices; this is load-bearing because the central attribution of deviations at odd integers or 13 to superstition requires that the null captures the expected distribution absent number bias.
Authors: The Monte Carlo null samples parameter values uniformly from the empirically observed ranges in the dataset (without fitted parameters or number preferences), thereby testing for deviations attributable to numerical properties rather than clinical constraints. Clinical factors such as oxygenation targets are expected to influence the overall range but not to produce systematic odd/even imbalances or specific aversion to 13, which has no physiological basis. We agree this assumption merits explicit discussion and will revise the methods and discussion sections to clarify the null model assumptions and note the limitation that unmodeled dependencies could contribute to marginal non-uniformity. revision: partial
-
Referee: [Statistical analysis (abstract)] Statistical analysis (abstract): The reported ORs (e.g., OR=0.16 for PEEP odd/even) and p-values are presented without stratification or regression on clinical covariates, so it is not possible to verify whether the Monte Carlo null has accounted for the unmodeled dependencies that the skeptic identifies as producing exactly the observed non-uniform marginals.
Authors: The ORs and p-values are computed by comparing observed frequencies directly against the Monte Carlo null distribution of no number preference; the simulation itself is non-parametric and does not rely on covariate modeling. While primary results are unadjusted, the specific avoidance of 13 (lacking clinical rationale) provides supporting evidence for the superstition interpretation. We will add a supplementary analysis or discussion of stratification by available covariates (e.g., age, diagnosis category) in the revised manuscript to address potential confounding. revision: yes
Circularity Check
Empirical observational analysis with simulation null; no circular derivation
full rationale
The paper conducts a retrospective database search on real ICU ventilator settings (PEEP, RR, Pinsp) and applies conventional hypothesis tests plus a Monte Carlo simulation to generate a null distribution for detecting odd/even or 13 preferences. No parameters are fitted to the target result, no equations are defined in terms of the observed preferences, and no self-citations or prior author work are invoked to justify the core claim. The reported ORs and p-values are direct empirical comparisons against the external data and the simulation-derived null, rendering the derivation self-contained.
Axiom & Free-Parameter Ledger
axioms (1)
- domain assumption The Monte Carlo simulation correctly samples the null distribution of ventilator settings under random choice without bias.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.