Skip to main content

Research instrument

DCS Risk Probability Calculator

Live implementation of the published Marroni 2026 multifactorial logistic model. Adjust your dive's parameters; see the predicted DCS probability and the signed contribution of every predictor. Below the calculator, each input is paired with the underlying physiology and the peer-reviewed evidence that supports or contradicts it.

Marroni A, Kot J, Pieri M, Pelliccia R, Balestra C. Identification of DCS risk factors in recreational diving: multifactorial model based on the DAN DSL Database 2024. Int Marit Health 2026; 77(1): 1–12. doi:10.5603/imh.108038. CC BY-NC-ND 4.0. 127,957 dives, 5,907 divers, AUC 0.910.

Dive parameters

Highest M-value ratio at surfacing (Bühlmann ZH-L16C). Equivalent to "Surface GF" on a Shearwater.

Predicted P(DCS)

—

odds —

Z-equation breakdown
Dataset baseline: 0.49% · Median DCS DSSG 0.866 · Median no-DCS DSSG 0.743

Research instrument — not a personal go/no-go tool

Three predictors in this model (pre-dive exercise, thermal comfort, pre-dive tiredness) are flagged by the authors as questionnaire-confounded in directions opposite to controlled-experiment physiology. The calculator reproduces the published model faithfully — including those confounded coefficients — so absolute values must not be cited as actionable risk-reduction advice. See the per-predictor mechanism notes below.

The published Z-equation

Z = −16.2976
  + 13.9063 · DSSG
  − 0.2338 · DSSG_COMPRT
  + 1.5330 · GENDER_N        (female = 1)
  − 1.2014 · FEELING_N       (tired = 1)
  + 1.0405 · THERMAL_N       (comfortable/hot = 1)
  + 1.0537 · GAS_COUNT
  + 0.7254 · EXERCISE_N      (any pre-dive exercise = 1)
  + 0.4779 · WORKLOAD_N      (any in-dive workload = 1)
  + 0.3065 · PURPOSE_N       (technical = 1)
  − 0.1591 · BMI_CLASS_N     (−3 … +4)
  − 0.0638 · DIVE_NR
  − 0.0404 · SURF_INTERVAL   (hours)

P(DCS) = 1 / (1 + e−Z)

Model fit: AUC 0.910 · Somer's D 0.8287 · KS 0.6832 (paper, Figs. 2–3). The large DSSG coefficient reflects that DSSG values are small fractions (0.3–1.4), not that DSSG is the only meaningful predictor.

DSSG: the dominant predictor

Bühlmann ZH-L16C basis

The body is modelled as 16 hypothetical tissue compartments with nitrogen half-times from 4 min to 635 min. Each compartment has a depth-dependent M-value — the maximum inert-gas tension it tolerates before bubble nucleation.

Compt #Half-timeTissue analogue
1–34–12.5 minBlood, CNS, heart
4–818–54 minMuscle, skin
9–1377–240 minFat, poorly perfused
14–16305–635 minTendon, bone

DSSG in plain English

At the moment of surfacing, divide every compartment's inert-gas tension by its sea-level M-value. The compartment with the highest such ratio is the leading compartment; its ratio is the DSSG.

DSSG < 1.0: every compartment within Bühlmann tolerance. DSSG ≥ 1.0: at least one compartment has crossed its theoretical M-value — and the empirical DCS rate at that threshold in DSL is 37.5%.

Recomputing DSSG from real dive-computer logs is non-trivial because most logs stop sampling 0.5–1 m below the actual surface (wrist position, sample rate). The authors apply a correction collapsing trailing <0.5 m samples to a single 0 m point.

DSSG → P(DCS): empirical bins + live model

Dots: univariate empirical rate per DSSG bin (radius ∝ √n). Curve: the full model evaluated across DSSG with your current other inputs. Marker: your current DSSG slider position.

Tabular form (paper Table 3)
DSSG binNo-DCSDCSP(DCS)
0.34,1730—
0.46,7160—
0.58,3260—
0.619,34250.012%
0.734,122240.095%
0.846,3502070.720%
0.916,8463473.344%
≥ 1.09184537.532%

The twelve predictors

Each card shows the published coefficient and odds ratio, the physiological mechanism, and the peer-reviewed evidence that supports or contradicts the direction observed in DSL data.

① DSSG — DAN Surface Supersaturation Gradient

OR dominantβ = +13.91 · ↑ risk

Why it matters. The decompression model literally defines when bubbles are predicted to form. DSSG is the cleanest single-number summary of "how much supersaturation did the diver carry across the surface boundary".

Mechanism. Henry's law: dissolved gas leaves solution when its partial pressure exceeds the surrounding pressure. The Bühlmann model places compartment-specific nucleation thresholds (M-values, Workman 1965; Bühlmann 1983). VGE/DCS field work confirms bubble loads and DCS rates rise together.

Evidence. Hjelte 202319 SWEN21: 154 chamber dives, 2% clinical DCS, median peak VGE grade 3 — confirms the supersaturation/bubble/DCS axis is causal. De Ridder 202320: lower GFlo (deeper first stop) loads slow tissues more — for air dives, GFlo 100 with GFhi 75–100 best matches DCIEM/USN safety record.

Practical implication. Surface GF on a Shearwater is the diver-facing analogue of DSSG. Driving Surface GF down (longer stops, lower GFhi) is the single most actionable lever on DCS probability that this model surfaces.

② Leading compartment index

OR 0.79per +1 · β = −0.234 · ↓ risk per slower compt

Why it matters. A high DSSG on a fast compartment (e.g. #3, half-time 8 min) is a different injury than the same DSSG on a slow compartment (#14, 305 min). Fast tissues include CNS and spinal cord — the targets of Type-II DCS. Slow tissues are tendon, bone, cartilage — the targets of milder Type-I "bends".

Mechanism. Fast compartments unload bubbles into venous return rapidly, overwhelming the pulmonary filter and (in PFO carriers) increasing arterialisation. Slower compartments produce smaller, more sustained loads that are clinically milder on average.

Evidence. Matches the NEDU deep-stops literature: redistributing stop time deep (slow-compartment supersaturation) increases DCS incidence relative to shallow-biased stops — supported by De Ridder 202320.

③ Female sex

OR 4.63β = +1.533 · ↑ risk · 3.29× univariate

Why it matters. The single strongest non-physiological coefficient in the model. Females had DCS at 1.25% (212/16,703) vs 0.38% (416/110,494) for males. Adjusted, female odds remain 4.63× higher.

Animal evidence. Lautridou 20173: heritable DCS susceptibility in rats. Lautridou 20204: in males, two physiological axes (lower coagulation tendency + enhanced inflammatory response) discriminate DCS-resistant animals; in females, neither does — implying the resistance mechanism itself differs between sexes.

Human observation. Irgens 201719: professional female divers show distinct injury and physiological patterns vs male peers.

Caveat. "Female" in DSL is self-reported binary; menstrual cycle, parity, hormonal contraception, HRT not captured. The OR averages over whatever mix the DSL population happens to include.

④ BMI class

OR 0.85per +1 step · β = −0.159 · raw U-shape

Why it matters. Conventional dive medicine teaches that obesity raises DCS risk because nitrogen is ~5× more soluble in fat. The raw DSL data show a U-shape: lowest empirical DCS rate in class-II obese; highest in mildly and moderately underweight divers.

Mechanism. Two competing physiologies. (a) Heavier BMI → more lipid reservoir → more nitrogen stored on slow compartments. (b) Heavier BMI may correlate with shorter, shallower recreational dives in this population. After adjustment, heavier classes had marginally lower DCS odds.

BMI bands per WHO 20002: severely underweight (BMI < 16) … morbidly obese (BMI ≥ 40). The U-shape (paper Fig. 6) is explicitly flagged for future research.

⑤ Gas count

OR 2.87per +1 mix · β = +1.054 · ↑ steeply

Why it matters. Univariate gradient is dramatic: 1 gas → 0.40% DCS; 2 → 2.81%; 3 → 8.59%; 4 → 50.0%. After adjustment, each added mix multiplies DCS odds by 2.87.

Mechanism. Gas count is a proxy, not a cause. A dive needing helium plus one or two deco mixes is, almost by definition, deeper, longer, colder, more equipment-intensive than a single-tank recreational profile. The associated hazards travel together: greater DSSG, more deco stops, higher CO₂ load29, more cold-water time, gas-switch task loading, dehydration. The 2.87× factor is "all of these things at once".

CCR caveat. On rebreathers the concept of gas count loses meaning because PpO₂ is held constant while FO₂ varies. The 1.4% CCR slice of DSL is included but its gas-count column is not directly comparable.

⑥ Dive purpose — "technical"

OR 1.36β = +0.307 · ↑ risk

Why it matters. Even after adjusting for gas count, DSSG, leading compartment, and workload, declaring the dive "technical" added another 36% to DCS odds. This is the residual effect of the technical milieu — overhead environments, task-loading, longer exposure, equipment failure modes.

Data. 1.24% DCS in 1,989 technical dives vs 0.57% in 104,929 recreational vs 0.02% in 18,648 "other" (guidance / instructional / student dives held to tightly supervised profiles).

⑦ In-dive workload

OR 1.61β = +0.478 · ↑ risk

Why it matters. Long-established association now quantified in field data: any in-dive exertion (current, photo rig drag, victim recovery, equipment failure response) bumps DCS odds 61%. Univariate: 0.67% with workload vs 0.39% without.

Mechanism. Exertion at depth raises cardiac output and tissue perfusion → faster on-gassing of working muscle → higher tissue tensions at the start of ascent. CO₂ production rises with exertion; CO₂ is itself a DCS modifier (vasodilation at depth, micronuclei activation)29. Madden 20156 linked SCUBA-induced intrapulmonary shunting to exercise — opening pulmonary "back doors" through which venous gas emboli can arterialise even without a PFO.

Caveat. Self-reported ordinal (none/light/moderate/heavy/exhausting) treated as binary above. The 1.61× OR is the average across "any workload"; moderate-to-exhausting is likely higher.

⑧ Pre-dive exercise ⚠ confounded

OR 2.06β = +0.725 · ↑ but flagged

What the model says vs the experiments. The model attaches doubled DCS risk to "any pre-dive exercise". The controlled-experiment literature on pre-conditioning says the opposite for protocolised pre-dive exercise: a single bout 2–24 h before diving reduces vascular gas emboli and decompression stress markers.

Pre-conditioning counter-evidence:

  • Blatteau 200515: single bout of aerobic exercise 2 h before a 30 msw chamber dive reduces post-deco bubbles.
  • Madden 201414: exercise before SCUBA ameliorates decompression-induced neutrophil activation.
  • Madden 201613: pre-dive exercise reduces repetitive-dive decompression stress.
  • Wilhelm 201612: exercise intensity modulates pro-angiogenic microvesicles — candidate cellular mechanism.
  • Lambrechts 202216: short mini-trampoline session reduces post-dive VGE.
  • Germonpré & Balestra 201717: systematic review of preconditioning (vibration, sauna, HBO, exercise, dietary).

Why the dataset reads backwards. The DSL questionnaire does not distinguish protocolised pre-conditioning ("20 min on the rower this morning") from confounding exertion ("carried tanks 400 m in 30 °C heat"). The negative correlation between "exercise" and "tired before dive" (r = −0.317) suggests both interpretations sit in the same data. The 2.06 OR averages a protective and a harmful population — the harmful component dominates the sample.

Do not cite this coefficient as advice

"Don't exercise before diving" is not a conclusion this paper supports, and is contradicted by six controlled studies. The right interpretation: avoid pre-dive exhaustion; structured aerobic pre-conditioning hours before a dive has mechanistic support for being protective.

⑨ Feeling tired before dive ⚠ confounded

OR 0.30β = −1.201 · ↓ but flagged

What the model says. Self-reporting "tired" or "exhausted" before diving cut DCS odds by 70% in the multivariate model. Univariate: 0.51% if rested vs 0.29% if tired.

The behavioural reading (authors' preferred). A diver who feels poorly probably modifies the dive: shallower, shorter, more buffer, earlier ascent. The "protective" effect is a behavioural surrogate — the diver, not the body, did the protecting. Lafère 201718 and Morgan 199521 showed personality and anxiety modulate dive behaviour and risk uptake.

The pre-conditioning reading. "Tired" correlates with "exercise" (r = −0.317), so some of the "tired" group are tired because they pre-conditioned. The same feature drives both this protective coefficient and the apparently-harmful exercise coefficient — they read the same population from opposite sides.

This coefficient does not say tiredness is protective. It says the population of divers who reported feeling tired had lower DCS in DSL — most likely because they then dived more conservatively. Acting on "I'll go tired, the data says I'm safer" inverts the actual mechanism.

⑩ Thermal comfort during dive ⚠ confounded

OR 2.83β = +1.041 · ↑ but flagged

What the model says vs the experiments. Counter-intuitive headline: divers reporting comfort or hot had DCS at 0.51% (615/119,604); divers reporting cold/very cold had 0.16% (11/7,001). Adjusted OR 2.83 for comfort/hot.

Controlled-experiment counter-evidence:

  • Gerth 20159: diver thermal status modulates DCS susceptibility — warm during deco improves off-gassing through perfusion.
  • Pollock 20158: "Don't dive cold when you don't have to" — warm bottom + cold deco is the worst combination.
  • Gaustad 20217: cold decompression in a rat model worsened haemodynamic function and DCS risk.
  • Leffler 200110: surface-decompression diver outcomes modulated by ambient temperature.
  • Broome 199311: climatic and environmental factors in DCS aetiology.

Why DSL reads backwards. (1) "Comfortable" dives skew toward warm-water recreational profiles which are longer/deeper than cold-water training dives. (2) Self-report is end-of-dive recall, not core or skin telemetry. The warm-bottom→cold-deco transition that matters physiologically cannot be captured by a single end-of-dive label.

The mechanism literature is unambiguous: warm-bottom-then-cold-deco is the high-risk pattern; staying warm during decompression is protective. The DSL coefficient is best read as evidence that a single-label thermal field is too coarse for risk inference, not as evidence that being warm increases DCS.

⑪ Dive number in repetitive series

OR 0.94per +1 dive · β = −0.064 · ↓ per added rep

Why it matters. Every additional dive within the rolling 48 h repetitive window cut DCS odds by ~6%. Counter-intuitive given that bubble models (VPM, RGBM) penalise repetitive series for nuclei carry-over.

Likely interpretations. (a) Selection — divers completing many rep dives are healthier and trained; DCS-prone divers self-select out after dive 1 or 2. (b) The surface-interval variable is co-modelled, so this coefficient is the residual after adjusting for time between dives. (c) RGBM-style penalties for repetition aren't visible in this mostly-recreational dataset — consistent with LANL field comparisons showing ZHL-16C and RGBM statistically indistinguishable.

⑫ Surface interval (hours)

OR 0.96per hour · β = −0.040 · ↓ per +1 h

Why it matters. Per added hour at the surface, DCS odds drop ~4%. Mirrors Haldanean theory: longer off-gassing windows reduce residual loading and bubble nuclei populations before the next exposure. Matches DAN flying-after-diving consensus (12 h single, 18 h repetitive).

Mechanism. Slow compartments (240–635 min half-times) need many hours to fully unload. An 8 h surface interval clears most of a 240-min compartment but only ~50% of a 635-min one. Each additional hour monotonically reduces residual N₂.

What this model does not capture

Twelve predictors is a lot, but several recognised DCS risk factors are absent from the DSL feature set. The authors flag this explicitly.

Patent foramen ovale (PFO)

~25–30% of adults; right-to-left shunt path for venous gas emboli to arterialise. Germonpré 202123 prospective: increased DCS risk with right-to-left shunt. Lafère 201718: PFO tied to iterative DCS history in 209 cases. Not in DSL.

Hydration status

Plasma volume and viscosity modulate inert gas transport. Gempp 200926 RCT: pre-dive hydration measurably reduced post-dive bubbles. Wekre 202222: bioimpedance + urine SG measurements in commercial saturation. Not in DSL.

Plasma lipids / surface tension

Schellart 201525: relationships between plasma lipids, surface tension and post-dive bubbles. Candidate mechanism for sex differences (lipid handling). Not in DSL.

Static metabolic bubbles & nucleation seeds

Imbert 201924: "static metabolic bubbles" pre-dive as a possible explanation for inter-individual bubbling variability. Beyond what any field questionnaire can capture without imaging.

Author-stated limitations

  • Long data-collection window means modern continuous-telemetry streams (PpO₂, inert-gas partials, temperature, gas bubbles) are unavailable for most records.
  • Self-reported categorical exposures (workload, thermal comfort, exercise, feeling) cannot be quantitatively confirmed.
  • CCR coverage is thin (1.4% of dataset). "Gas count" is not meaningful on a closed circuit where PpO₂ is held constant.
  • Missing variables: PFO, hydration, diet, plasma lipids, sleep, hormonal status, genetic predisposition — all known DCS modifiers, absent from the DSL questionnaire.
  • "Repetitive" defined as ≤ 48 h — multi-day live-aboard exposure analyses may behave differently.
  • DCS reporting bias: mild Type-I DCS may be under-reported. The 0.49% base rate is a floor estimate.

References

Numbered references match the in-text superscripts and the Marroni paper's own reference list. Click a number in the body to jump here; click here to return.

  1. Cialoni D, Pieri M, Balestra C, et al. Dive risk factors, gas bubble formation, and decompression illness in recreational SCUBA diving: analysis of DAN Europe DSL database. Front Psychol 2017; 8: 1587. doi:10.3389/fpsyg.2017.01587 · PMID 28974936
  2. World Health Organization. Obesity: preventing and managing the global epidemic. WHO Tech Rep Ser 2000; 894: 1–253. PMID 11234459
  3. Lautridou J, Buzzacott P, Belhomme M, et al. Evidence of heritable determinants of decompression sickness in rats. Med Sci Sports Exerc 2017; 49(12): 2433–2438. doi:10.1249/MSS.0000000000001385 · PMID 28731987
  4. Lautridou J, Dugrenot E, Amérand A, et al. Physiological characteristics associated with increased resistance to decompression sickness in male and female rats. J Appl Physiol 2020; 129(3): 612–625. doi:10.1152/japplphysiol.00324.2020 · PMID 32702269
  5. Madden D, Thom SR, Dujic Z. Exercise before and after SCUBA diving and the role of cellular microparticles in decompression stress. Med Hypotheses 2016; 86: 80–84. doi:10.1016/j.mehy.2015.12.006 · PMID 26804603
  6. Madden D, Ljubkovic M, Dujic Z. Intrapulmonary shunt and SCUBA diving: another risk factor? Echocardiography 2015; 32 Suppl 3: S205–S210. doi:10.1111/echo.12815 · PMID 25693625
  7. Gaustad SE, Kondratiev TV, Eftedal I, et al. Effects of cold decompression on hemodynamic function and decompression sickness risk in a dry diving rat model. Front Physiol 2021; 12: 763975. doi:10.3389/fphys.2021.763975 · PMID 34803743
  8. Pollock NW. Re: Don't dive cold when you don't have to. Diving Hyperb Med 2015; 45(3): 209. PMID 26415074
  9. Gerth WA. On diver thermal status and susceptibility to decompression sickness. Diving Hyperb Med 2015; 45(3): 208. PMID 26415073
  10. Leffler CT. Effect of ambient temperature on the risk of decompression sickness in surface decompression divers. Aviat Space Environ Med 2001; 72(5): 477–483. PMID 11346015
  11. Broome JR. Climatic and environmental factors in the aetiology of decompression sickness in divers. J R Nav Med Serv 1993; 79(2): 68–74. PMID 8263855
  12. Wilhelm EN, González-Alonso J, Parris C, et al. Exercise intensity modulates the appearance of circulating microvesicles with proangiogenic potential upon endothelial cells. Am J Physiol Heart Circ Physiol 2016; 311(5): H1297–H1310. doi:10.1152/ajpheart.00516.2016 · PMID 27638881
  13. Madden D, Barak O, Thom SR, et al. The impact of predive exercise on repetitive SCUBA diving. Clin Physiol Funct Imaging 2016; 36(3): 197–205. doi:10.1111/cpf.12213 · PMID 25371042
  14. Madden D, Thom SR, Milovanova TN, et al. Exercise before scuba diving ameliorates decompression-induced neutrophil activation. Med Sci Sports Exerc 2014; 46(10): 1928–1935. doi:10.1249/MSS.0000000000000319 · PMID 24576865
  15. Blatteau JE, Gempp E, Galland FM, et al. Aerobic exercise 2 hours before a dive to 30 msw decreases bubble formation after decompression. Aviat Space Environ Med 2005; 76(7): 666–669. PMID 16018350
  16. Lambrechts K, Germonpré P, Vandenheede J, et al. Mini trampoline, a new and promising way of SCUBA diving preconditioning to reduce vascular gas emboli? Int J Environ Res Public Health 2022; 19(9). doi:10.3390/ijerph19095410 · PMID 35564805
  17. Germonpré P, Balestra C. Preconditioning to reduce decompression stress in scuba divers. Aerosp Med Hum Perform 2017; 88(2): 114–120. doi:10.3357/AMHP.4642.2017 · PMID 28095955
  18. Lafère P, Balestra C, Caers D, et al. Patent foramen ovale (PFO), personality traits, and iterative decompression sickness. Retrospective analysis of 209 cases. Front Psychol 2017; 8: 1328. doi:10.3389/fpsyg.2017.01328 · PMID 28824507
  19. Irgens Å, Troland K, Grønning M. Female professional divers. Int Marit Health 2017; 68(1): 60–67. doi:10.5603/IMH.2017.0010 · PMID 28357838
  20. Hunt JC. Psychological aspects of scuba diving injuries: suggestions for short-term treatment from a psychodynamic perspective. J Clin Psychol Med Settings 1996; 3(3): 253–271. doi:10.1007/BF01993911 · PMID 24226762
  21. Morgan WP. Anxiety and panic in recreational scuba divers. Sports Med 1995; 20(6): 398–421. doi:10.2165/00007256-199520060-00005 · PMID 8614760
  22. Wekre SL, Landsverk HD, Lautridou J, et al. Hydration status during commercial saturation diving measured by bioimpedance and urine specific gravity. Front Physiol 2022; 13: 971757. doi:10.3389/fphys.2022.971757 · PMID 36246118
  23. Germonpré P, Lafère P, Portier W, et al. Increased risk of decompression sickness when diving with a right-to-left shunt. Front Physiol 2021; 12: 763408. doi:10.3389/fphys.2021.763408 · PMID 34777020
  24. Imbert JP, Egi SM, Germonpré P, et al. Static metabolic bubbles as precursors of vascular gas emboli during divers' decompression. Front Physiol 2019; 10: 807. doi:10.3389/fphys.2019.00807 · PMID 31354506
  25. Schellart NAM, Rozloznik M, Balestra C. Relationships between plasma lipids, proteins, surface tension and post-dive bubbles. Undersea Hyperb Med 2015; 42(2): 133–141.
  26. Gempp E, Blatteau JE, Pontier JM, et al. Preventive effect of pre-dive hydration on bubble formation in divers. Br J Sports Med 2009; 43(3): 224–228. doi:10.1136/bjsm.2007.043240 · PMID 18308884

Supplementary references (mechanism context, not in Marroni paper)

  1. Bühlmann AA. Decompression — Decompression Sickness. Berlin/New York: Springer-Verlag; 1984. Foundational documentation of the ZH-L16 algorithm.
  2. Workman RD. Calculation of decompression schedules for nitrogen-oxygen and helium-oxygen dives. USN Experimental Diving Unit Research Report 1965; 6–65.
  3. Daubresse L, Vallée N, Druelle A, et al. Effects of CO₂ on the occurrence of decompression sickness: review of the literature. Diving Hyperb Med 2024; 54(2): 110–119. doi:10.28920/dhm54.2.110-119 · PMID 38870953
  4. De Ridder S, Pattyn N, Neyt X, Germonpré P. Selecting optimal air diving gradient factors for Belgian military divers. Diving Hyperb Med 2023; 53(3): 251–258. doi:10.28920/dhm53.3.251-258 · PMID 37718300
  5. Hjelte C, Plogmark O, Silvanius M, et al. Risk assessment of SWEN21 — a suggested new dive table for the Swedish armed forces. Diving Hyperb Med 2023; 53(4): 299–305. doi:10.28920/dhm53.4.299-305 · PMID 38091588

Not medical advice. This page is a research-grade exposition of a published probabilistic model. It implements the Marroni 2026 formula verbatim but does not constitute clinical guidance, training material, or a personal go/no-go tool. Diving decisions should be made with a qualified dive medical officer and current dive-computer guidance.