Admission Markers Identify Adult Tetanus Patients at High Risk for Mechanical Ventilation
Adult tetanus remains an emergency in which respiratory failure can develop quickly and with little warning. In this dual-center retrospective study from Southern Jiangxi, China, investigators built a bedside nomogram designed to identify which hospitalized adults with tetanus were most likely to require mechanical ventilation. The model was built from routine admission data, which gives it immediate appeal for emergency and intensive care teams working under time pressure.
The cohort included 227 adults admitted between January 2012 and December 2024, and 123 patients, or 54.2%, ultimately required invasive ventilation for respiratory failure. In a condition where airway compromise and autonomic instability can escalate fast, a tool that separates high-risk from lower-risk patients at the door has obvious practical value. The final prediction model used five independent predictors: age, chronic kidney disease, neutrophil count, D-dimer, and CK-MB.
Five Routine Variables Carried the Strongest Signal
Feature selection started with 38 candidate predictors spanning demographic data, clinical characteristics, and laboratory biomarkers. LASSO regression narrowed the field to six variables: age, chronic kidney disease, neutrophil count, D-dimer, urea, and CK-MB. Urea did not remain in the final multivariable model, leaving five variables that independently tracked with ventilation requirement.
The strongest association came from chronic kidney disease, which carried an odds ratio of 4.939 with a wide confidence interval, reflecting both a meaningful effect size and the uncertainty that comes with a relatively modest sample. Age also contributed steadily, with each increment associated with higher odds of ventilation. Neutrophil count and D-dimer both pointed toward a state of heightened systemic inflammation and coagulation activation, while CK-MB suggested concurrent myocardial injury or stress. In tetanus, where severe muscle rigidity, sympathetic overdrive, and oxygen demand can destabilize cardiopulmonary function, those markers fit a biologically plausible pattern of deterioration.
A single bedside panel may capture more than infection burden; it may capture the physiologic price of a toxin-driven crisis.
Model Performance Was Strong in Internal Validation
Discrimination reached an area under the ROC curve of 0.813, with a 95% confidence interval from 0.757 to 0.868. For a clinical prediction model drawn from admission variables alone, that level of separation between patients who did and did not need ventilation is clinically meaningful. The optimal cutoff was 0.576, which produced a sensitivity of 0.675 and a specificity of 0.837.
Calibration was also acceptable. The Hosmer-Lemeshow test yielded a chi-square value of 11.031 with a P value of 0.200, and calibration plots supported alignment between predicted and observed risk. Bootstrap validation, using 1,000 iterations, produced an AUC of 0.776 with a very narrow reported confidence interval, suggesting the model retained useful performance after resampling. Decision curve analysis showed net clinical benefit across threshold probabilities from 0.1 to 0.8, which matters because clinicians do not treat every patient the same way at every level of risk. Some require early ICU transfer, airway planning, or closer cardiopulmonary monitoring; others may not.
Why These Biomarkers Fit the Physiology of Severe Tetanus
The included variables map onto the pathobiology of severe tetanus in a way that makes clinical sense. Age often tracks with physiologic reserve and the ability to tolerate prolonged neuromuscular stress. Chronic kidney disease may mark a narrower therapeutic margin, greater metabolic instability, and higher vulnerability to complications during a prolonged ICU course. Elevated neutrophil count and D-dimer suggest a systemic inflammatory and prothrombotic state, which can accompany critical illness and may reflect the severity of toxin-mediated stress.
CK-MB deserves particular attention. In tetanus, cardiac strain is not a peripheral issue. Sympathetic surges, tachycardia, labile blood pressure, and increased oxygen demand can produce myocardial injury or reveal pre-existing cardiac fragility. A marker of myocardial stress entering the model supports the idea that ventilation risk is not only a respiratory problem; it is part of a wider failure of physiologic compensation.
Bedside Use Depends on Simplicity as Much as Accuracy
A nomogram only matters if clinicians can use it quickly. This one relies on routinely available admission parameters, avoiding the need for exotic assays or prolonged computational steps. In emergency settings, especially where tetanus remains an unpredictable disease and ICU beds are limited, the appeal lies in triage. A patient who appears stable on arrival may still warrant earlier airway planning if the combined score is high.
The study also reflects a broader challenge in tetanus care: respiratory failure is often not a late surprise but an expected hazard that emerges from the interaction of rigidity, spasms, autonomic dysfunction, and cardiopulmonary strain. Prediction models do not replace bedside judgment, but they can sharpen it. If validated externally, this nomogram could support decisions about early transfer, closer monitoring, and proactive ventilation readiness before collapse occurs.
Internal Validation Is Useful, but Generalizability Remains Unproven
The study was retrospective and dual-center, which strengthens it relative to a single-site analysis but still leaves the model within one regional health system. The cohort came from two tertiary hospitals in Southern Jiangxi, and the model has not yet been tested outside that environment. External validation will need to determine whether the same five-variable structure performs similarly in other regions, different hospital systems, and populations with different vaccination histories, case severity, or ICU resources.
The sample size was also modest for a multivariable prediction study, and the confidence interval around chronic kidney disease was wide. Missing data were limited to 3.2% and handled with random forest-based multiple imputation, which is methodologically sound, yet any retrospective model remains dependent on the quality and completeness of the source records. None of that erases the value of the tool; it simply defines the next step. The question now is whether the model can preserve its discrimination and calibration when tested prospectively in independent adult tetanus cohorts.
For now, the study provides a clinically workable framework for identifying adult tetanus patients at high risk of mechanical ventilation using five admission markers that are already available in routine care. In a disease where delays can be dangerous, the ability to quantify risk early may be as important as the score itself.