Tracheostomy Practices in Adult Intensive Care Units: A Multicenter Point-Prevalence Study in Izmir (IZ-TRACH)

Design and populations

This was a multicenter point-prevalence survey on 8 September 2025 with a subsequent assessment of recorded in-hospital mortality on 8 October 2025. The census included 282 adults present in eleven participating ICUs across six hospitals. Forty-six had an established tracheostomy and contribute to the point-prevalence numerator. The main descriptive sample comprises 42 patients whose tracheostomy was established during the current admission. Four with pre-admission tracheostomies remain in the census and prevalence counts but are excluded from the main patient-level summaries. Patients who died or left the ICU before the index day and patients who had not undergone tracheostomy by that date are absent from this sample.

Data collection and compilation

The initial compilation contained 30 tracheostomy patients from seven ICUs. Records for another 16 patients from four additional ICUs within one participating hospital were subsequently incorporated, yielding 46 tracheostomy patients across eleven ICUs. Their postponement reports were obtained on the same index day, 8 September 2025; the records were merged later. Of these 16 patients, three had pre-admission tracheostomies. The remaining 13 contributed to the main analysis, and four had a family-related reason.

Six hospital contact physicians submitted the online forms using information from eleven primary treating physicians, one for each ICU. At the hospital comprising six ICUs, the contact physician collated information from all six primary physicians, including the contact physician's own ICU. Each of the other five hospitals contributed one ICU. Participation in the original tracheostomy decision was not recorded, and reports may have concerned decisions made by colleagues.

Physician-reported reasons and intervals

The index-day form asked: "Please specify any factors that caused tracheostomy to be performed late." Examples included high-dose inotropic support and lack of consent from relatives. This wording assumed lateness and may have influenced responses. Fourteen patients had an explicit "No documented delay" response; none of the 42 responses was blank. These patients are described as having no reported postponement reason, which does not independently establish absence of postponement. Other recorded categories were a family-related reason, hemodynamic instability, awaiting surgical eligibility and bleeding diathesis. Multiple reasons and an "Other" free-text option were allowed; neither was used. Investigators did not select a predominant reason from multiple responses.

A family-related reason represents the physician's account of deliberation or agreement involving relatives or a surrogate decision-maker, not a demonstrated avoidable delay. The measured interval is total time from intubation to tracheostomy. Intended procedure dates, recommendation dates, discussion/agreement dates and clinical readiness dates were unavailable, so time attributable to an individual reason cannot be calculated.

Recorded complications and mortality

Complications were identified retrospectively from records through the index day. The form presented separate early- and late-complication items with prelisted event types. No seven-day boundary was communicated to respondents. These questionnaire categories were retained in the analysis; individual event dates were unavailable, so classification as occurring within or after seven days cannot be verified. Recorded early-item events were hemorrhage, subcutaneous emphysema and hypoxia/desaturation; late-item events were granuloma, stoma infection and tracheoesophageal fistula. Observation periods were unequal and unmeasured; these are recorded category counts rather than standardized procedure-anchored incidence estimates.

Recorded in-hospital deaths between 8 September and 8 October 2025 were ascertained from local records and the hospital information system. This is 30-day post-index in-hospital mortality, not mortality within 30 days of tracheostomy or ICU admission. A zero mortality code denotes no recorded in-hospital death; it does not establish survival or continued hospitalization at follow-up. Individual calendar death dates and discharge/transfer histories were unavailable in the retained analysis data. Tracheostomy-to-death intervals are summarized only among patients with a recorded death; comparable procedure-to-censoring intervals are unavailable for the others.

Descriptive analyses

The analyses report counts and percentages with explicit denominators and medians with interquartile ranges and ranges. Patient-level summaries use the 42 current-admission patients. Reason-group outcome counts and admission APACHE II summaries by recorded in-hospital death status are exploratory and were not prospectively specified. Technique is described as a procedural characteristic; outcomes are not compared by technique. No hypothesis tests, adjusted comparisons, causal effects, equivalence claims or survival analyses are presented. Earlier ROC/cutoff and early-versus-late timing analyses are not retained.

Registration chronology

Data collection was completed on 8 October 2025. The initial registry submission was on 5 February 2026 and first posting on 11 February 2026. That entry described 30 tracheostomy patients in seven ICUs and mortality within 30 days after tracheostomy as its only structured outcome. An amendment submitted on 16 June 2026 and posted on 22 June 2026 changed enrollment to the 282-patient census, described eleven ICUs and the 46-patient tracheostomy subgroup, changed the mortality anchor to the index date, and added structured outcomes. The change from 30 to 282 also changed the population represented by the enrollment field, from a tracheostomy subgroup to the full ICU census.

The explanation released for PRS review on 19 October 2026 that the discrepancy was solely a registry-entry error was incorrect. The compilation history is clarified above: 16 records collected on the index day were incorporated after the initial 30-record compilation. This amendment also corrects the unsupported seven-day complication definition and distinguishes form submitters from clinical information sources. These retrospective entries and amendments do not establish prospective specification of the analyses. Historical versions and their dates remain part of the record.