Abstract
Background
Patients requiring prolonged invasive mechanical ventilation are prone to complications, such as infections, tracheal stenosis and death. It has been proposed that early tracheostomy could have a role in preventing these outcomes, but the proper identification of patients at risk can be difficult.
Purpose
The aim of this study was to develop a multivariate model that allows the early detection of patients that will require prolonged ventilatory support.
Patients and methods
A retrospective cohort study was undertaken in the intensive care unit of the Hospital Naval Almirante Nef, Chile, between June 2011 and June 2012. The charts of all intubated patients were reviewed in search for early predictors of prolonged intubation (> 7 days). Multivariate logistic regression analysis was used to detect statistically significant associations and to assess potential confounders.
Results
A total of 349 patients were admitted to the intensive care unit during the study period and 142 (40.7%) required invasive mechanical ventilation. Most of them were male (60.5%), with a mean age of 65.8 ± 16.7 years. Thirty-five patients (24%) required to be ventilated for 7 days or more, and 16 (46%) were tracheostomized for this reason. The regression model showed that older age ( p = 0.026), a Pa/Fi ratio of less than 200 ( p = 0.046), and the presence of chronic pulmonary disease ( p = 0.035) or hypernatremia ( p = 0.012) on intubation day were significantly associated with the requirement of prolonged intubation.
Discussion
Invasive mechanical ventilation is a common reason for admittance to the ICU. The abovementioned predictors can be of assistance when selecting patients that could benefit from early tracheostomies, and are in agreement with earlier reports. Although the model’s discriminating capacity was good, it is necessary to formally validate it before recommending its widespread use.
1
Introduction
Mechanical ventilation is often imperative in many cases of acute respiratory failure, and thus a common reason for admission to an intensive care unit (ICU). Recent data have shown that approximately one in three patients in an ICU is admitted because of the need of ventilatory support, and of these, 5 to 10% will require prolonged mechanical ventilation .
With a standardized definition lacking in the literature, researchers have used wide cutoff points when referring to the abovementioned entity, which often range from 24 hours to 21 days . In spite of this situation, these patients are consistently reported to be at an increased risk of death, cognitive decline, and functional dependence and to have an overall poorer quality of life .
In order to prevent these complications, many authors have considered tracheostomies to be a suitable option. This procedure has been shown to shorten overall hospital stays, decrease the incidence of ventilator-associated pneumonia, upper airway injuries (including vocal cord ulceration) and in-hospital mortality; although the evidence has to date yielded mixed results . These benefits are theoretically explained in clinical benefits that are to be obtained from tracheostomy, which include lower airway resistances, easier and safer tracheal suction, improved patient comfort and communication, the possibility of oral feeding and faster weaning from the ventilator . All of these benefits are likely to be improved by optimal procedure timing, with an increasing interest in the medical literature for early, rather than late, tracheostomies. A recent systematic review that included both randomized and non-randomized trials found that early tracheostomy significantly reduced duration of artificial ventilation (mean difference of − 8.5 days, 95% confidence intervals, 95% CI: − 15.3 to − 1.7 days) and length of stay in intensive care (mean difference of − 15.3 days; 95% CI − 24.6 to − 6.1 days) . Beneficial results were also suggested in another systematic review by the Cochrane Collaboration composed of four randomized trials, although the authors warned that their estimates could be biased due to a perceived high risk of systematic error amongst included studies .
Being able to predict which patient will require prolonged ventilatory assistance could prove important for clinical practice. The proper recognition of the patient at risk would permit the early establishment of potentially useful therapies, such as tracheostomies. Therefore, we aimed at developing a predictive model that would help clinicians detect patients at increased risk of prolonged mechanical ventilation.
2
Patients and methods
A retrospective cohort study was undertaken amongst patients admitted to the intensive care unit of the Hospital Naval Almirante Nef between June 2011 and June 2012. All patients older than 18 years who required mechanical ventilation and were therefore subject to an orotracheal intubation were considered for inclusion. Patients who had been intubated in other hospitals or ICUs, those that had an intubation method other than orotracheal (i.e. nasopharyngeal, larynx mask) and patients that already had a tracheostomy were excluded from this study.
The main outcome was the development of a prolonged intubation. This was defined as the requirement of an orotracheal intubation for 7 days or more . A patient was considered to have been successfully extubated when he or she is able to remain for a minimum of 48 hours free from the ventilator. Those that required to be reintubated within this time frame were considered as not extubated. The requirement of reintubation for further ventilatory assistance after 48 hours was recorded in the database, but no patient was included more than once in this study.
2.1
Data collection
The charts of all included patients were reviewed for clinical data. These included gender, age, relevant physical examination findings (systolic, diastolic and mean blood pressure, pulse and respiratory rates, axillary temperature) commorbidities (as defined by Charlson’s Commorbidity Index ), Acute Physiology And Chronic Health Evaluation II (APACHE-II) and Therapeutic Intervention Scoring System (TISS) scores, primary diagnosis, and results on routine laboratory tests (complete blood counts, arterial blood gases, serum creatinine, uremia, serum electrolytes and C-reactive protein levels). We also registered the length of ICU stay, total ventilation time and vital status at discharge for each patient.
2.2
Statistical analysis
2.2.1
Sample size
Based on previous research , a prolonged intubation event rate of 25% was expected in this cohort study, with precision limit estimates of ± 5%, and 5% two-sided significance levels. The study population was expected to be of 150 intubated patients (per year), based on information available at the study centre. Therefore, a minimum sample size of 100 patients was calculated in order to accurately estimate our event rate, but it was aimed to review the charts of at least 120 patients because of potential issues with missing data.
2.2.2
Inferential analyses
Descriptive statistics are expressed as means and standard deviations (SD) or medians and interquartile ranges (IQR) for quantitative variables, or absolute frequencies and percentages for qualitative data. Bivariate comparisons between groups were performed using Fisher’s exact test for categorical variables and Student’s t test or Mann–Whitney’s test for continuous data. Ninety-five percent confidence intervals (95% CI) were also calculated for relevant outcomes.
A multivariate logistic regression model was used to assess the association between a prolonged intubation and early clinical predictors and also to adjust for confounders. Potentially relevant variables were tested with the abovementioned bivariate strategy, and all factors whose p -value was less than 0.20 were considered for inclusion in the multivariate model. All models evaluated the existence of interactions between independent variables. The principles of regression analysis were evaluated graphically and through residual analysis. Goodness of fit was assessed with Hosmer’s and Lemeshow’s statistic and the overall diagnostic capacity of the model with a receiver operating characteristic (ROC) curve. Adjusted odds ratios (aOR) with their corresponding 95% confidence intervals and p -values are reported for the final multivariate model. A p -value of less than 5% was considered to be statistically significant.
2
Patients and methods
A retrospective cohort study was undertaken amongst patients admitted to the intensive care unit of the Hospital Naval Almirante Nef between June 2011 and June 2012. All patients older than 18 years who required mechanical ventilation and were therefore subject to an orotracheal intubation were considered for inclusion. Patients who had been intubated in other hospitals or ICUs, those that had an intubation method other than orotracheal (i.e. nasopharyngeal, larynx mask) and patients that already had a tracheostomy were excluded from this study.
The main outcome was the development of a prolonged intubation. This was defined as the requirement of an orotracheal intubation for 7 days or more . A patient was considered to have been successfully extubated when he or she is able to remain for a minimum of 48 hours free from the ventilator. Those that required to be reintubated within this time frame were considered as not extubated. The requirement of reintubation for further ventilatory assistance after 48 hours was recorded in the database, but no patient was included more than once in this study.
2.1
Data collection
The charts of all included patients were reviewed for clinical data. These included gender, age, relevant physical examination findings (systolic, diastolic and mean blood pressure, pulse and respiratory rates, axillary temperature) commorbidities (as defined by Charlson’s Commorbidity Index ), Acute Physiology And Chronic Health Evaluation II (APACHE-II) and Therapeutic Intervention Scoring System (TISS) scores, primary diagnosis, and results on routine laboratory tests (complete blood counts, arterial blood gases, serum creatinine, uremia, serum electrolytes and C-reactive protein levels). We also registered the length of ICU stay, total ventilation time and vital status at discharge for each patient.
2.2
Statistical analysis
2.2.1
Sample size
Based on previous research , a prolonged intubation event rate of 25% was expected in this cohort study, with precision limit estimates of ± 5%, and 5% two-sided significance levels. The study population was expected to be of 150 intubated patients (per year), based on information available at the study centre. Therefore, a minimum sample size of 100 patients was calculated in order to accurately estimate our event rate, but it was aimed to review the charts of at least 120 patients because of potential issues with missing data.
2.2.2
Inferential analyses
Descriptive statistics are expressed as means and standard deviations (SD) or medians and interquartile ranges (IQR) for quantitative variables, or absolute frequencies and percentages for qualitative data. Bivariate comparisons between groups were performed using Fisher’s exact test for categorical variables and Student’s t test or Mann–Whitney’s test for continuous data. Ninety-five percent confidence intervals (95% CI) were also calculated for relevant outcomes.
A multivariate logistic regression model was used to assess the association between a prolonged intubation and early clinical predictors and also to adjust for confounders. Potentially relevant variables were tested with the abovementioned bivariate strategy, and all factors whose p -value was less than 0.20 were considered for inclusion in the multivariate model. All models evaluated the existence of interactions between independent variables. The principles of regression analysis were evaluated graphically and through residual analysis. Goodness of fit was assessed with Hosmer’s and Lemeshow’s statistic and the overall diagnostic capacity of the model with a receiver operating characteristic (ROC) curve. Adjusted odds ratios (aOR) with their corresponding 95% confidence intervals and p -values are reported for the final multivariate model. A p -value of less than 5% was considered to be statistically significant.
3
Results
During the observation period, a total of 349 patients were admitted to the ICU. One hundred forty-two (40.7%, 95% CI: 35.5–46%) were intubated due to the need of invasive mechanical ventilation, and all of these patients were included in this retrospective cohort.
The study sample consisted of mostly male inpatients (60.5%) with a mean age of 65.8 ± 16.7 years. The overall burden of comorbidities was considered to be moderate, with a median Charlson Index of two points (IQR: 1–3 points). Anemia (43.7%), diabetes mellitus (28.9%) and cancer (21.8%) were the most common comorbidities. The median APACHE-II score was 19 in both groups, and no significant differences were seen amongst them ( p = 0.87). The main reason for intubation was the development of respiratory failure (60 patients, 42.3%), followed by the postoperative requirement of invasive ventilatory support (51 patients, 36%) and upper airway compromise (23 patients, 16.2%). Additional information regarding patient characteristics is shown on Table 1 .
| Characteristic | Patients intubated < 7 days ( n = 108) | Patients intubated > 7 days ( n = 34) | p value |
|---|---|---|---|
| Mean age ± SD, years | 64.6 ± 17.5 | 64.6 ± 13.5 | 0.12 a |
| Male gender, no. (%) | 68 (63) | 18 (53) | 0.32 a |
| Median APACHE-II score (IQR) | 19 (14–24) | 19 (15–25) | 0.87 b |
| Median TISS score (IQR) | 27 (20–31) | 30 (25–34) | 0.03 b |
| Comorbidities a | |||
| Charlson Comorbidity Index, median (IQR) | 2 (0–3) | 2 (1–3) | 0.46 a |
| Anemia, no. (%) | 49 (47.1) | 13 (39.4) | 0.55 c |
| Cancer, no. (%) | 21 (22.6) | 10 (31.3) | 0.35 c |
| Metastatic cancer, no. (%) | 2 (2.4) | 1 (3) | 1 c |
| Heart failure, no. (%) | 10 (10.2) | 1 (3) | 0.29 c |
| Chronic obstructive pulmonary disease, no. (%) | 23 (21.3) | 12 (35.3) | 0.11 c |
| Chronic kidney disease, no. (%) | 17 (19.5) | 2 (6.5) | 0.15 c |
| Acute myocardial infarction, no. %) | 14 (15.7) | 4 (13.3) | 1 c |
| Dementia, no. (%) | 3 (3.6) | 1 (3) | 1 c |
| Diabetes mellitus, no. (%) | 34 (35.1) | 7 (22.6) | 0.27 c |
| Peripheral vascular disease, no. (%) | 8 (9.6) | 3 (10) | 1 c |
| Mild liver disease, no. (%) | 4 (4.8) | 1 (3.3) | 1 c |
| Severe liver disease, no. (%) | 8 (9.5) | 1 (3.3) | 0.44 c |
| Mesenchymopathies, no. (%) | 1 (1.2) | 1 (3.3) | 0.46 c |
| Peptic ulcer disease, no. (%) | 5 (6) | 1 (3.3) | 1 c |
| Lymphoma, no. (%) | 1 (1.2) | 0 (0) | 1 c |
| Hemiplegia, no. (%) | 1 (1.2) | 0 (0) | 1 c |
| Acquired Immunodeficiency Syndrome, no. (%) | 1 (1.2) | 0 (0) | 1 c |
| Laboratory | |||
| Serum sodium (mEq/Lt) (SD) | 140 ± 5.5 | 143 ± 6 | < 0.01 a |
| Hypernatremia, no. (%) | 13 (12.3) | 12 (36.4) | < 0.01 c |
| Serum potassium (mEq/Lt) (SD) | 4.2 ± 0.75 | 4 ± 0.70 | 0.27 a |
| Serum creatinine (mg/dL) (SD) | 1.43 ± 1.3 | 1.3 ± 1 | 0.65 a |
| Uremia (mg/dL) (SD) | 58.8 ± 38 | 61.5 ± 28 | 0.72 a |
| C reactive protein (mg/Lt) (SD) | 89.6 ± 90.9 | 145.7 ± 118.2 | < 0.01 |
| Hematocrit (%) (SD) | 31.7 ± 6.6 | 29.5 ± 8.2 | 0.13 a |
| White cell count (cells/mm 3 ) | 11.770 ± 5540 | 11.520 ± 6150 | 0.82 a |
| Platelet count (cells/mm 3 ) | 183600 ± 73400 | 184800 ± 73000 | 0.94 a |
| Median arterial oxygen (PaO 2 ) (mmHg) (IQR) | 116.7 (80–170.6) | 84.3 (76.5–163.5) | 0.35 b |
| Median arterial carbon dioxide (PaCO 2 ) (mmHg) (IQR) | 38.4 (33–43.8) | 36.1 (32.3–41.3) | 0.52 b |
| Median Pa/Fi ratio (IQR) | 260 (187–352) | 192 (128–321) | 0.12 b |
| Median bicarbonate (HCO 3 – ) (mEq/Lt) (IQR) | 20.3 (17.4–23) | 21 (19–23.1) | 0.43 b |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
