Mechanical ventilation is a common intensive care unit (ICU) modality used to treat respiratory failure from a variety of causes. Each year in the United States, more than 1 million persons admitted to ICUs receive mechanical ventilation, usually for less than 48 hours (Cox, Carson, Govert, Chelluri, & Sanders, 2007). However, approximately 34% of these patients require prolonged ventilatory support, and the rate of prolonged ventilatory support is increasing (Cox et al., 2007).
Although intubation and mechanical ventilation are necessary to support respiratory function and life, these technologies create many distressful physiological and psychological experiences for patients (Li & Puntillo, 2006; Rotondi et al., 2002). To be mechanically ventilated is to be grossly uncomfortable at best (McCartney & Boland, 1994). Patients have referred to mechanical ventilation as the most inhumane treatment ever experienced (Gries & Fernsler, 1988) and admit to being miserable most of the time while intubated (Logan & Jenny, 1997). Patients who were mechanically ventilated for more than 48 hours recall the endotracheal tube itself, being unable to talk, being thirsty, feeling tense, not being in control, difficulty swallowing (Rotondi et al., 2002), and moderately intense anxiety (Chlan, 2004; Li & Puntillo, 2006) as being most distressing. Anxiety is a state marked by apprehension, agitation, increased motor tension or activity, autonomic arousal, and fearful withdrawal (McCartney & Boland, 1994). It is one of the most common symptoms reported by patients receiving mechanical ventilatory support (Li & Puntillo, 2006; Rotondi, et al., 2002). Anxiety develops in response to these distressful experiences associated with mechanical ventilation (McCartney & Boland, 1994). Thus, anxiety compounded by fear causes increased sympathetic nervous system stimulation, increased work of breathing, increased oxygen demand, and myocardial stimulation (Johnston & Sexton, 1990).
Nurses caring for critically ill patients believe that treating anxiety is important. The most frequently used therapy is the administration of antianxiety sedative medications (Frazier et al., 2003). Critically ill patients receive a wide variety of intravenous sedative medications from disparate drug classes over the course of ventilatory support that can influence anxiety ratings (Weinert & Calvin, 2007). These medications are administered to patients to promote breathing synchrony with the mechanical ventilator, to reduce anxiety, and to promote comfort. Adjunctive, nonpharmacologic interventions used to treat anxiety include empathic touch, control of environmental stressors, providing choices with respect to care to enhance the patient's sense of control, music, and relaxation techniques (Frazier et al., 2003). Although symptom management for ventilated ICU patients can be a great challenge, it is imperative that nurses implement evidence-based strategies. However, for ICU nurses to effectively manage patient anxiety, an awareness of the dynamic nature of this distressful symptom over the course of ventilatory support must first be known to intervene appropriately.
Information about the course of anxiety across the duration of mechanical ventilation in the ICU is limited. Previous investigations of interventions to reduce anxiety in response to mechanical ventilatory support have used preintervention-postintervention measurement of anxiety, (Chlan, 1998; Wong, Lopez-Nahas, & Molassiotis, 2001) or have reported cross-sectional snapshots of anxiety ratings at one point during ventilatory support (Chlan, 2003). Our impression from practice is that ICU clinicians believe that anxiety may decrease over the course of ventilatory support as the patient adjusts to this invasive treatment modality. However, little data about the reported experience or intensity of anxiety over the course of mechanical ventilatory support in the ICU are available.
The purposes of this study were to describe anxiety ratings over the duration of study enrollment in a sample of critically ill patients receiving mechanical ventilatory support, to identify any pattern of change in anxiety ratings, to determine if anxiety decreases over the course of ventilatory support, and to explore the influence of sedative exposure on anxiety ratings.
Design, Setting, and Sample
Persons included in this sample are a subgroup of participants enrolled in a multisite, ICU-based randomized trial testing music interventions for anxiety self-management in patients receiving mechanical ventilatory support. Participants for the multisite trial were recruited from five medical centers (12 separate ICUs) located throughout the Minneapolis-St. Paul urban area. Patients receiving mechanical ventilatory support for a primary pulmonary problem (eg, pneumonia, respiratory distress, respiratory failure), making their own daily care decisions, and who were alert and interacting appropriately with nursing staff at time of enrollment were invited to participate in the study. Participants remained enrolled in the study until extubation or up to 30 days, until they choose to withdraw, or until they died. The multisite clinical trial was approved by the University of Minnesota's Institutional Review Board and by the institutional review board of the participating sites.
A descriptive design was used for this article. Participants included in this secondary analysis are those randomized to the usual care control condition. Usual care consists of the standing medical orders and standardized nursing care protocols for each respective ICU whereby registered nurses provide care in a 1:1 or 1:2 nurse-to-patient ratio.
Each participating ICU was screened daily by a research nurse for potential study participants. Once a new participant was enrolled, the study protocol commenced, and data collection began. Study entry data collection consisted of severity of illness determination, length of ICU stay (days), length of ventilatory support (days), ventilator settings, and all medications abstracted from the medical record.
Anxiety was measured with the Visual Analog Scale-Anxiety (VAS-A) at study entry and then once daily as close to the same time each day as possible over the duration of study participation for all enrolled participants. The research nurse supported participants in completion of the VAS-A as needed by assisting participants with marking their current level of anxiety on the VAS-A. Not all participants provided anxiety ratings each day enrolled in the study, including day of enrollment, due to being unable to complete the VAS-A because of fatigue, need to leave the ICU for a diagnostic procedure, altered mental status or level of alertness, or refusal to complete the assessment due to other nonspecified reasons.
Given the influence of sedative and analgesic medications on anxiety ratings, all medications were abstracted from the medical record. For this study, dosing and frequency of dose administration over each 24-hour period were obtained for midazolam, lorazepam, fentanyl, morphine, dexmedetomidine, hydromorphone, propofol, and haloperidol.
All participants were visited each day by a research nurse who conducted the anxiety assessment via the VAS-A, reviewed the medical record for additional study data on ventilator settings, and recorded all medications. Participants remained on protocol as long as they were receiving mechanical ventilatory support up to 30 days. Participants contained in this sample were extubated at different time points, which marked study completion. Thus, a varying number of anxiety assessments were included for each study participant.
Variables and Measurement
Anxiety is a state marked by apprehension, agitation, increased motor tension or activity, autonomic arousal, and fearful withdrawal (McCartney & Boland, 1994). Participants rated their current level of anxiety on the VAS-A on a 100-mm vertical line that was anchored on each end by statements not anxious at all to the most anxious I have ever been. The VAS-A had a vertical orientation, thought to be more sensitive and easier for participants to use, particularly for those with a narrowed visual field or when under stress (Cline, Herman, Shaw, & Morton, 1992; Gift, 1989). A visual analog scale with a vertical orientation, analogous to a thermometer, to perform repeated measurement of anxiety in mechanically ventilated patients was reported to be less burdensome for participants to complete than other instruments with a Likert-based response format (Chlan, 2003; Chlan, Savik, & Weinert, 2003). Scores were derived by measuring the distance in millimeters from the bottom edge of the line anchor to the mark placed by the participant (Knebel, Janson-Bjerklie, Malley, Wilson, & Marini, 1994; Lee & Kieckhefer, 1989).
Visual analog scales are appropriate for tracking a participant's clinical course because they are easily administered and easy for participants to see. Few words are used, which minimizes the possibility of different interpretations (Gift, 1989; Wewers & Loew, 1990). The VAS-A has been used by investigators to measure anxiety in patients receiving mechanical ventilation (Cline et al., 1992) and to measure changes in anxiety in ventilated patients undergoing weaning trials (Knebel et al., 1994). The VAS-A and the Spielberger State Anxiety Inventory were moderately correlated in ventilated ICU patients (r =.49; Chlan, 2004) and in patients undergoing ambulatory surgical procedures (r =.82; Vogelsang, 1988). These results show that concurrent validity of the VAS-A. Stability (ie, test-retest reliability) is not relevant due to the expected dynamic nature of state anxiety (Lush, Janson-Bjerklie, Carrieri, & Lovejoy, 1988; Wewers & Loew, 1990). Of crucial importance is the reproducibility of ratings obtained from these scales (Wewers & Loew, 1990). The VAS-A is an accurate and sensitive measure of state anxiety, capable of reproducing reliable measures of anxiety in ventilated patients undergoing weaning trials (Knebel et al., 1994) and ambulatory surgical procedures (Vogelsang 1988).
Critically ill patients receive a wide variety of intravenous sedative and analgesic medications from disparate drug classes over the course of ventilatory support that can influence anxiety ratings, referred to as sedative exposure (Weinert & Calvin, 2007). These medications are administered to patients to promote breathing synchrony with the mechanical ventilator, to reduce anxiety, and to promote comfort. To summarize the medications that mechanically ventilated patients may receive from disparate drug classes, which are not amenable to dose-equivalent calculations, we used the following approach: A dose frequency count of all sedative and analgesic medications documented in the medical record each day of study participation was used to calculate an aggregate dose, which yielded a sedation intensity score (SIS; Weinert & Calvin, 2007).
Data were abstracted from the medical record on all sedative and analgesic medications received during a 24-hour period. Dose frequency was determined by dividing the calendar day into six 4-hour time blocks (00:00, 04:00, 08:00, 12:00, 16:00, and 20:00), and for each medication (midazolam, lorazepam, fentanyl, morphine, dexmedetomidine, hydromorphone, propofol, and haloperidol), the occurrence(s) in which a drug was administered at least once during that interval was summed. Frequency of medication doses was then summed for each participant over each of the six 4-hour time blocks daily to yield a dose frequency count.
Likewise, the SIS was based on aggregate doses of medication(s) received over the same 24-hour period as described previously. The SIS is a validated measure that addresses the problem of aggregating sedative exposure across disparate drug classes (Weinert & Calvin, 2007). The weight-adjusted dose was first calculated based on an individual participant's kilogram of body weight for each medication administered during a 4-hour time block. The dose was then categorized as 1-4 based on the quartile within the distribution of that drug for one time block. For instance, if 0.1 mg/kg of lorazepam and 0.2 mg/kg of morphine were given during a 4-hour interval and 0.1 mg/kg fell into the second quartile of the distribution of all 4-hour lorazepam doses in the entire group and 0.2 mg/kg of morphine was in the third quartile, then the SIS for that time block was 2 (second quartile) + 3 (third quartile) = 5. A participant's mean SIS (quotient of sum of participant's SIS values and number of 4-hour intervals on mechanical ventilation) represents the average sedative exposure per hour relative to all other participants.
Severity of Illness
The severity of illness of each participant was ascertained at study entry via the Acute Physiology, Age, and Chronic Health Evaluation III to establish comparability of illness severity among groups at baseline; data for Acute Physiology, Age, and Chronic Health Evaluation III scores were abstracted from the medical record from the first day of ICU admission. Scoring details are described elsewhere (Knaus, et al., 1991).
Time was measured in days. Day 0 was set as the day of study enrollment. However, the date of initiation of mechanical ventilation was variable for each participant. In many cases, the initiation of mechanical ventilatory support occurred a number of days prior to study enrollment. Thus, the number of study days on which anxiety measurements were obtained varied for each participant. Length of time on protocol and number of daily measurements were limited to 30 days as 30 days was the established limit for study protocol participation.
Descriptive statistics for interval and ordinal data were presented as medians with ranges, given the skewed distributions of the data. Categorical data were presented as frequencies. In an initial analysis, anxiety trajectories were graphed for each participant to discern pattern of change.
Mixed effects models were used for analysis as they accommodate correlated and nonhomogeneous residuals, which would be expected in repeated measures. Mixed models are an ideal analysis for dealing with disparate assessment time points, missing data points, or both from participants being unable or unwilling to complete daily anxiety assessments due to medical status, mental status, or level of fatigue. A series of models were estimated to determine the best model of change for the VAS-A in this study. Model parameters are defined in Table 1, and the estimated models are listed in Table 2. In each model, Yij is the VAS-A score for person i on Day j.
The unconditional means model was estimated to determine if further modeling was appropriate. Each outcome Yij is a linear combination of the grand mean (γ00) plus the individual deviations from the grand mean (ζ0i) and a random error term (ϵij). The unconditional means model assesses two null hypotheses: (a) no change across occasions and (b) no variation between participants. Rejecting these null hypotheses warrants performing further analysis.
An unconditional growth model with DAY added as a predictor incorporated estimation of change coefficients. Models with several within-person error covariance structures that were compatible with the correlation pattern between VAS-A scores at different time points were explored. Correlations seemed to decrease as the lag time increased, which is indicative of an autoregressive (AR) structure. Three covariance structures were considered. The unstructured covariance model presupposes heterogeneous variance in VAS-A scores over time and, thus, no pattern in the covariance structure. This model is useful as it is usually the model that fits the data best and can serve as a baseline for evaluating other structures. The downside of this model is that it requires the estimation of the most parameters and thus reduces power. The AR structure appeared to best fit the correlation pattern but does imply that correlation between measures on the same individual ultimately approach zero. The AR + random effects (RE) model specifies that covariance between observations comes from two sources, the AR structure and the fact that the measures come from the same subject. The AR + RE structure does not assume that the correlations will approach zero.
An unconditional growth model with a quadratic term was also explored to assess if there were discernable nonlinear changes in VAS-A scores over time. The (AR + RE) error covariance structure was used.
Two conditional models were estimated to explore the effect of sedation frequency and sedation intensity. Sedation frequency scores and SIS were incorporated as time-varying covariates in linear growth models.
Analysis was performed using SPSS Version 17 and Proc Mixed in SAS Version 9.2 (Singer, 1998). Final parameter estimates were considered significant at p < .05. Aikake's information criterion (AIC) and the Bayesian information criterion (BIC) were used to select the best model for this sample.
Description of Study Sample
In this sample (n = 57), participants had been in the ICU for a median of 8 days (range = 1-29 days) and had been receiving mechanical ventilatory support for a median of 6 days (range = 1-27 days) prior to enrollment. Participants randomized to the usual care group remained enrolled in the study for a median of 4.1 days (range = 1-30 days). Table 3 summarizes the demographic characteristics of the participants and other variables.
Description of Anxiety Ratings and Frequency of Missing Anxiety Data
Participants reported moderate anxiety at study entry (median VAS-A = 57.5) with a wide range in anxiety from 0 (not anxious at all) to 96 (near the maximum score of 100; Table 3). Participants reported varying levels of anxiety over the course of study enrollment as illustrated in Figure 1. There is no discernable single pattern to the anxiety ratings for those participants who provided at least three anxiety ratings. For some participants, the pattern is highly variable, with increases and decreases in anxiety ratings over the study enrollment period, whereas other participants' anxiety ratings decrease, increase, or remain essentially at the same level over time.
Not all enrolled participants were able to provide anxiety ratings each study day, even at study entry. The two reasons most often cited were that participants were too tired to complete the paper-and-pencil instrument or were sedated on subsequent study days. There was no relationship between first VAS-A score and the number of days receiving ventilatory support prior to study enrollment (ρ = −.04, p =.79). Overall, the mean number of missing daily VAS-A scores was 3.6 (SD = 4.9), with a mode of one missing assessment. The numbers of participants able to provide anxiety ratings each study day are presented in Figure 2. All available anxiety scores (VAS-A) were used in the growth modeling analyses.
Modeling results are presented in Table 4. The first model explored for the analysis was the unconditional means model. This resulted in estimates of variance for both the average VAS-A score between participants and variance in the average within person over time mean. The intercept for this model was 49.5 (p < .001). Both variance parameter estimates were significant (p < .001), indicating that further modeling would be appropriate. The intraclass correlation coefficient indicated that 33% of the total variance resulted from differences between participants and 67% was due to the variance over time within participant.
Unconditional linear growth models were estimated next. Among these, AIC selected the model with an unstructured within-person error variance-covariance matrix, whereas BIC identified the model with the AR + RE structure as best. In the unconditional growth model, with AR + RE within-person error covariance matrix, the estimated average starting VAS-A value was 57 (SE = 5.5). Change was estimated at −0.50 (SE =.38) points per day; this was not statistically significant (p < .18). A model with a DAY2 term was also generated. The coefficient for the quadratic DAY2 term was not significant, and both AIC and BIC increased, so the quadratic change model received no further consideration.
A conditional linear growth model was then generated to account for the influence of sedative and analgesic medications on anxiety (sedative dose frequency and SIS). The daily sedation frequency count score and SIS were entered into separate unconditional growth models with DAY predicting VAS-A. The time-varying effect of sedative exposure did not improve relative model fits and was not significant for either the dose frequency or the sedation intensity. Sedative exposure (dose frequency and sedation intensity) left the estimates of the intercept and slope virtually unchanged. These models all left a significant amount of unexplained variance both within person over time and between participants. Figures 3 and 4 depict median anxiety ratings with sedation dose frequency and sedation intensity over total days on study protocol.
The purposes of this study were to describe anxiety ratings of critically ill patients receiving mechanical ventilatory support, to discern any pattern of change in daily anxiety ratings, to determine if anxiety decreases over the course of ventilatory support, and to explore the influence of sedative exposure on anxiety ratings. Participants in this sample were receiving prolonged periods of ventilatory support prior to study enrollment and reported moderate levels of anxiety when first measured at study entry, despite receiving sedative and analgesic medications known to influence anxiety.
The anxiety data reported in this study reflect those participants randomized to the usual care condition only and do not consider covariates such as illness severity, length of ventilatory support, or length of ICU stay. The individual anxiety ratings reported by participants indicated patterns of highly individual and variable anxiety. Reported anxiety ratings decreased for some participants over time; others reported anxiety ratings that fluctuated or increased.
The overall pattern of anxiety ratings for this group of participants over the duration of study enrollment suggested a possible slight decline over time with a highly variable pattern of this symptom experience. Participants demonstrated a general pattern of moderate anxiety over the course of study enrollment. These data are similar to findings from previous descriptive, cross-sectional work that showed that patients receiving mechanical ventilatory support for 22 or more days had the highest anxiety ratings, followed by those in the 6-21 day group (Chlan, 2003).
Results of the mixed-models analysis were that VAS-A ratings slowly decreased over time. However, there was not a statistically significant decrease in these anxiety ratings over study enrollment in this group of participants.
Sedative exposure did not significantly influence the participants' daily anxiety ratings. Neither dose frequency nor sedation intensity explained a statistically significant amount of variance within person or over time on anxiety ratings.
Study limitations include the number of missing data points on the VAS-A when participants were too fatigued to complete the assessments, were sedated, or were too ill to provide daily anxiety assessments. Measurement of subjective symptoms remains a challenge in nonverbal, critically ill patients with profound physiological and psychological limitations. Anxiety arises from numerous physiological and psychological factors in ventilated patients. This study did not attempt to discern the sources of anxiety; anxiety ratings reported here provided only one assessment time point per day.
Another limitation is the various entries into study time points. Participants were enrolled at various times during their ICU stay and course of mechanical ventilatory support. Thus, it is not known how anxious a participant might have been on the first day of receiving mechanical ventilation in comparison with their first study enrolled day. However, there was no relationship between first anxiety rating obtained and days receiving ventilatory support prior to study enrollment.
Lastly, an influence of sedative and analgesic medications not seen in the analysis was possibly due to the relatively uniform pattern of dose frequency but a varying pattern of sedation intensity. Further consideration of these issues is warranted.
Summary and Conclusions
Although findings from this study do suggest that anxiety does decrease over time for some patients receiving mechanical ventilatory support, other patients do not readily adjust to the ventilator, the ICU environment, or both and do not experience lessening anxiety over the course of treatment. Critical care clinicians should not expect that anxiety decreases over time for all ventilated patients. Ongoing nursing assessment and appropriate, individualized interventions with patients receiving mechanical ventilatory support are needed to appropriately address anxiety symptom management.
Chlan, L. (1998). Effectiveness of a music therapy intervention on relaxation and anxiety for patients receiving ventilatory assistance. Heart & Lung: The Journal of Acute and Critical Care
Chlan, L. (2003). Description of anxiety levels by individual differences and clinical factors in patients receiving mechanical ventilatory support. Heart & Lung: The Journal of Acute and Critical Care
Chlan, L. (2004). Relationship between two anxiety instruments in patients receiving mechanical ventilatory support. Journal of Advanced Nursing
Chlan, L., Savik, K., & Weinert, C. (2003). Development of a shortened state anxiety scale from the Spielberger State-Trait Anxiety Inventory (STAI) for patients receiving mechanical ventilatory support. Journal of Nursing Measurement
Cline, M. E., Herman, J., Shaw, E. R., & Morton, R. D. (1992). Standardization of the visual analogue scale. Nursing Research
Cox, C., Carson, S., Govert, J., Chelluri, L., & Sanders, G. (2007). An economic evaluation of prolonged mechanical ventilation. Critical Care Medicine
Frazier, S., Moser, D., Daley, L., McKinley, S., Riegel, B., & Garvin, B. (2003). Critical care nurses' beliefs about and reported management of anxiety. American Journal of Critical Care
Gift, A. G. (1989). Visual analogue scales: Measurement of subjective phenomena. Nursing Research
Gries, M. L., & Fernsler, J. (1988). Patient perceptions of the mechanical ventilation experience. Focus on Critical Care
Johnston, M., & Sexton, D. (1990). Distress during mechanical ventilation: Patients' perceptions. Critical Care Nurse
Knaus, W. A., Wagner, D. P., Draper, E. A., Zimmerman, J. E., Bergner, M., Bastos, P. G., et al (1991). The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults. Chest
Knebel, A. R., Janson-Bjerklie, S. L., Malley, J. D., Wilson, A. G., & Marini, J. J. (1994). Comparison of breathing comfort during weaning with two ventilatory modes. American Journal of Respiratory and Critical Care Medicine
Lee, K. A., & Kieckhefer, G. M. (1989). Measuring human responses using visual analogue scales. Western Journal of Nursing Research
Li, D., & Puntillo, K. (2006). A pilot study on coexisting symptoms in intensive care patients. Applied Nursing Research
Logan, J., & Jenny, J. (1997). Qualitative analysis of patients' work during mechanical ventilation and weaning. Heart and Lung
Lush, M. T., Janson-Bjerklie, S., Carrieri, V. K., & Lovejoy, N. (1988). Dyspnea in the ventilator-assisted patient. Heart and Lung
McCartney, J. R., & Boland, R. J. (1994). Anxiety and delirium in the intensive care unit. Critical Care Clinics
Rotondi, A., Chelluri, L., Sirio, C. A., Mendelsohn, A., Schulz, R., Belle, S. et al (2002). Patients' recollections of stressful experiences while receiving prolonged mechanical ventilation in an intensive care unit. Critical Care Medicine
Singer, J. D. (1998). Using SAS Proc Mixed to fit multilevel models, hierarchical models, and individual growth models. Journal of Educational and Behavioral Statistics
Vogelsang, J. (1988). The visual analog scale: An accurate and sensitive method for self-reporting preoperative anxiety. Journal of Post-Anesthesia Nursing
Weinert, C., & Calvin, A. (2007). Epidemiology of sedation for mechanically ventilated patients. Critical Care Medicine
Wewers, M., & Loew, N. (1990). A critical review of visual analog scales in the measurement of clinical phenomena. Research in Nursing & Health
Wong, H., Lopez-Nahas, V., Molassiotis, A. (2001). Effects of music therapy on anxiety in ventilator-dependent patients. Heart and Lung
© 2011 Lippincott Williams & Wilkins, Inc.