This study was conducted at Pusan National University Hospital, Busan, Korea, a university-affiliated tertiary care hospital with a 12-bed adult medical ICU with full cardiovascular and close airway monitoring, and a separate 7-bed adult coronary care unit. The physician staff in the medical ICU was comprised of 1 full-time specialist, 1 clinical fellow in pulmonary and critical care medicine, 2 resident physicians, and 1 intern. Overnight care was provided by fellow and resident physicians, while the nurse to bed ratio was 1:3. Full-time physical and respiratory rehabilitation therapies were available for all patients. Consultation services were available for all the subspecialists in the hospital.

All data were derived retrospectively from the medical records of patients admitted to the adult (≥18 years old) medical ICU between January 1, 2005 and December 31, 2010. Patients with PMV were defined as those who had undergone mechanical ventilation for longer than 21 days¹⁴, regardless of the tracheostomy date. Excluded from the study were patients who had been diagnosed with irreversible brain damage, regardless of length of ICU admission. Two investigators (M.H.K. and K.L.) confirmed that the study objectives and procedures were honestly disclosed and both had full access to all data. The study protocol was approved by the Institutional Review Board (IRB) of the Pusan National University Hospital (E-2011027).

The following data were gathered from the medical records of each patient: age, gender, diagnosis on initial ICU admission, Acute Physiology and Chronic Health Evaluation (APACHE) II scores on day 1 of ICU admission¹⁵, length of stay in the ICU and hospital, tracheostomy rate after admission, period from admission to the day of tracheostomy, ventilator weaning rate, ICU and hospital survival, and 6-month survival after ICU admission. The Sequential Organ Failure Assessment (SOFA) score and its components were calculated to quantify the severity of illness based on the degree of organ dysfunction¹⁶, and the simplified Therapeutic Intervention Scoring System 28 (TISS) score was calculated to assess the intensity of care on days 1 and 21 of ICU admission¹⁷. The Charlson Comorbidity Index was calculated to predict the effect of comorbid disease on patient outcome, acquired from the medical records regarding patient comorbidity before the catastrophic illness requiring ICU care¹⁸. All APACHE II, SOFA, TISS scores and Charlson Comorbidity Index were calculated from the raw data collected during chart review by the authors. Survivors were defined as patients who had survived for 6 months after ICU admission. Patients discharged without documentation of a post-discharge follow-up examination were contacted by telephone to determine the length of survival. Successful ventilator weaning was defined as breathing without assistance.

Statistical analyses were performed using the SPSS version 17.0 (SPSS Inc., Chicago, IL, USA). Descriptive results were expressed as mean and standard deviation or medians with range. Student's t-test was used to compare the continuous variables and the Chi-square or Fisher's exact tests (for small expected cell sizes) were used to compare categorical characteristics. Patient survival was analyzed using the Kaplan-Meier method, and compared using the log-rank test. The factors found to be significantly associated with survival were analyzed further with a Cox proportional hazard model to adjust for the potential confounding effect of each factor. Hazard ratios with 95% confidence intervals were used to report the results. A two-tailed p<0.05 was considered to indicate a significant difference.

During the study period, a total of 2,644 patients were admitted to the medical ICU. Of those, 136 patients (5.1%) met our definition of PMV patients. The median number of days on mechanical ventilation was 45 (range, 21~231). Clinical characteristics of the patients are shown in Table 1. The most common diagnosis of initial ICU admission was respiratory failure caused by pulmonary causes (n=97) which included pneumonia (n=69), acute exacerbation of chronic lung disease (n=15), pulmonary tuberculosis (n=9), and pulmonary edema (n=4). The medical ICU and hospital mortality rates were 45.6% and 48.5%, respectively, and the 6-month cumulative mortality rate was 58.8%. There were 96 patients who underwent tracheostomy placement after admission and their mean period from admission to the day of tracheostomy was 21.3±8.4 days (median, 22 days; range, 1st day~50th day after admission). Among the 96 patients with trachesotomy after admission, 57 subjects (59.3%) had their placement date after 21 days of mechanical ventilation. Sixty-three patients (46.3%) were successfully weaned from ventilator care; however, we could not find any factors that were associated with successful weaning from ventilator (data not shown). Of the ICU survivors (n=74), 34 patients (45.9%) were transferred to other hospitals (not university hospitals), but there was no significant 6-month cumulative mortality difference compared to patients who were transferred to the general ward (Figure 1).

Aforementioned clinical factors in data collection were evaluated with regard to their influence on survival. Three factors were associated with mortality based on Kaplan-Meier survival curves and log-rank tests (p<0.05): thrombocytopenia (platelet count≤150×10⁹/L) (p<0.001), requirement for vasopressors on day 21 (p=0.007) and tracheostomy during hospitalization (p=0.025). After adjusting for age and sex, these three factors were introduced into the Cox regression hazard model, which revealed that two variables (thrombocytopenia and requirement for vasopressors on day 21) were independent factors of survival (Table 2, Figure 2). Patients (n=57) with at least one of two factors had higher APACHE scores on admission day (25.3±7.2 vs. 22.9±6.6, p=0.044), TISS scores on day 21 (31.2±4.8 vs. 27.9±3.4, p<0.001) and 6-month cumulative mortality rates (75.4% vs.46.8%, p=0.001) than patients without these two factors.