Noninvasive Ventilation in Acute Respiratory Failure
Noninvasive Ventilation in Acute Respiratory Failure
We conducted a population-based retrospective cohort study among consecutively admitted adult (≥ 18 years) patients with ARF at the Mayo Clinic medical and surgical ICUs in Rochester, MN, from January 1st 2006 to December 31st 2006. Olmsted county residents were identified based on the ZIP codes of their primary residence and verified with the REP database. The REP database is a medical record-linkage system, which links together the medical records of almost complete Olmsted County population, irrespective of any demographic or regional characteristics. If a patient had multiple hospital admissions, only the first ARF episode was considered for analysis. The study protocol was approved by the Mayo Clinic Institutional Review Board. All eligible individuals who gave research authorization to review their medical records for research were included. Patients, who declined the use of their medical records for research, required invasive mechanical ventilation for less than 12 hours after surgical procedure and those who used CPAP treatment for sleep apnea were excluded.
Trained critical care clinical and research fellows abstracted the data from the electronic medical records (EMR) using a standardized protocol. The causes of NIV use were identified according to the standard definitions. The data on demographics, code status preferences, underlying severity of pulmonary and nonpulmonary organ dysfunctions, ventilation type and interface (noninvasive, endotracheal tube or tracheostomy) were extracted from the EMR. The Multidisciplinary Epidemiology and Translational Research in Intensive Care ICU datamart is an integrative database to extract ICU data from the hospital EMR. Patient's baseline characteristics, comorbidities and severity of illness (Acute Physiology and Chronic Health Evaluation [APACHE] III) scores, were collected from the EMR using the ICU datamart. Hemodynamic variables, fluid, drug infusion, laboratory parameters and ventilator settings were extracted from the ICU electronic database. All the relevant data from the patients' medical records and bedside flow charts were reviewed from ICU admission to ICU discharge. The hospital mortality at discharge and hospital length of stay (LOS) was collected from the electronic database by manual chart review.
The use of NIV was defined as the acute need of positive pressure ventilatory support through a tightly fitted facial or nasal mask for more than one hour. Acute respiratory failure was defined as the acute need of IMV support for more than twelve hours or NIV including CPAP for more than one hour. Palliative NIV was defined as patients who themselves signed the "Do Not Intubate" form and those in whole the health care stuff considered NIV as the ceiling therapy to alleviate the symptoms of respiratory failure and offer a possible chance of survival. The success of NIV was defined as the recovery from acute respiratory without endotracheal intubation among patients with full code (without limitation of therapy).
The primary outcome was to identify the incidence of use of NIV in patients with ARF and to identify the factors associated with the failure of the NIV among the same. The secondary outcome was to identify the long term survival among the patients with ARF who were on palliative NIV. The survivors after the hospital discharge were followed till the end of 2010. The death date was identified from the EMR or death registration record of Minnesota, in case of out-of-hospital deaths.
All the continuous data was summarized as median (interquartile range [IQR]). Categorical data was summarized as counts and percentages. Age and gender-specific incidence rate (95% confidence interval [CI]) for NIV use in the ICU was calculated assuming that the entire population of Olmsted County (≥18 years) was at risk. The incidence rate was adjusted to the projected 2006 United States population (utilizing the data from the 2000 U.S. population census and calculating an expected 1.9% population growth per year). Kaplan-Meier survival analysis with log-rank test was used to assess the long term survival among the ARF patients who were initiated on palliative NIV. A univariate and multivariate logistic regression analysis was used to identify the factors associated with failure of NIV in full code patients. Significant variables on the univariate analysis (p < 0.2) were included in the multivariate logistic regression model. Stepwise forward and backward procedure was used to select variables included in the final analysis. Non-significant factors (p > 0.05) were eliminated (one at a time) until all remaining factors had a significant association with NIV failure. JMP statistical software (version 8.0, SAS, Cary, NC) was used for all the data analyses. The level of significance for all statistical tests was 2-sided, with P <0.05.
Methods
We conducted a population-based retrospective cohort study among consecutively admitted adult (≥ 18 years) patients with ARF at the Mayo Clinic medical and surgical ICUs in Rochester, MN, from January 1st 2006 to December 31st 2006. Olmsted county residents were identified based on the ZIP codes of their primary residence and verified with the REP database. The REP database is a medical record-linkage system, which links together the medical records of almost complete Olmsted County population, irrespective of any demographic or regional characteristics. If a patient had multiple hospital admissions, only the first ARF episode was considered for analysis. The study protocol was approved by the Mayo Clinic Institutional Review Board. All eligible individuals who gave research authorization to review their medical records for research were included. Patients, who declined the use of their medical records for research, required invasive mechanical ventilation for less than 12 hours after surgical procedure and those who used CPAP treatment for sleep apnea were excluded.
Data Abstraction and Management
Trained critical care clinical and research fellows abstracted the data from the electronic medical records (EMR) using a standardized protocol. The causes of NIV use were identified according to the standard definitions. The data on demographics, code status preferences, underlying severity of pulmonary and nonpulmonary organ dysfunctions, ventilation type and interface (noninvasive, endotracheal tube or tracheostomy) were extracted from the EMR. The Multidisciplinary Epidemiology and Translational Research in Intensive Care ICU datamart is an integrative database to extract ICU data from the hospital EMR. Patient's baseline characteristics, comorbidities and severity of illness (Acute Physiology and Chronic Health Evaluation [APACHE] III) scores, were collected from the EMR using the ICU datamart. Hemodynamic variables, fluid, drug infusion, laboratory parameters and ventilator settings were extracted from the ICU electronic database. All the relevant data from the patients' medical records and bedside flow charts were reviewed from ICU admission to ICU discharge. The hospital mortality at discharge and hospital length of stay (LOS) was collected from the electronic database by manual chart review.
Identification of Noninvasive Mechanical Ventilation
The use of NIV was defined as the acute need of positive pressure ventilatory support through a tightly fitted facial or nasal mask for more than one hour. Acute respiratory failure was defined as the acute need of IMV support for more than twelve hours or NIV including CPAP for more than one hour. Palliative NIV was defined as patients who themselves signed the "Do Not Intubate" form and those in whole the health care stuff considered NIV as the ceiling therapy to alleviate the symptoms of respiratory failure and offer a possible chance of survival. The success of NIV was defined as the recovery from acute respiratory without endotracheal intubation among patients with full code (without limitation of therapy).
Outcomes
The primary outcome was to identify the incidence of use of NIV in patients with ARF and to identify the factors associated with the failure of the NIV among the same. The secondary outcome was to identify the long term survival among the patients with ARF who were on palliative NIV. The survivors after the hospital discharge were followed till the end of 2010. The death date was identified from the EMR or death registration record of Minnesota, in case of out-of-hospital deaths.
Statistical Analysis
All the continuous data was summarized as median (interquartile range [IQR]). Categorical data was summarized as counts and percentages. Age and gender-specific incidence rate (95% confidence interval [CI]) for NIV use in the ICU was calculated assuming that the entire population of Olmsted County (≥18 years) was at risk. The incidence rate was adjusted to the projected 2006 United States population (utilizing the data from the 2000 U.S. population census and calculating an expected 1.9% population growth per year). Kaplan-Meier survival analysis with log-rank test was used to assess the long term survival among the ARF patients who were initiated on palliative NIV. A univariate and multivariate logistic regression analysis was used to identify the factors associated with failure of NIV in full code patients. Significant variables on the univariate analysis (p < 0.2) were included in the multivariate logistic regression model. Stepwise forward and backward procedure was used to select variables included in the final analysis. Non-significant factors (p > 0.05) were eliminated (one at a time) until all remaining factors had a significant association with NIV failure. JMP statistical software (version 8.0, SAS, Cary, NC) was used for all the data analyses. The level of significance for all statistical tests was 2-sided, with P <0.05.
