Whom should we rely on when assessing symptoms of critically ill patients?
Whom should we rely on when assessing symptoms of critically ill patients?
复制标题
评估危重病人的症状时应该依靠谁？
DOI：
10.1097/ccm.0b013e31825f7bc5
复制
发表时间：
2012
影响因子：
8.8
通讯作者：
Cooke,ColinR
中科院分区：
文献类型：
--
作者：
Kross,ErinK;Curtis,JRandall;Cooke,ColinR
在重症监护室（ICU）评估患者的症状可能具有挑战性，特别是当患者病情严重且无法沟通时。使用代理症状评估作为ICU中患者报告症状的替代的想法很有吸引力，但之前的工作有限，主要是检查临床医生对疼痛的评估（1）。Puntillo博士和他的同事（2）报告了一项研究，以检查ICU患者对其症状的评估与其家人，护士和医生的评估之间的一致性。作者检查了患者和代理症状评分之间的一致性，包括疼痛、疲劳、呼吸急促、不安、焦虑、悲伤、饥饿、恐惧、口渴和困惑等10种不同症状的强度和痛苦程度。他们招募了245名重症患者，其中89人能够在研究的第一天报告症状。他们收集了患者沿着的症状评分，以及患者家属、护士和医生在患者评分后2-8小时内的评分，确定同一潜在现象的两种测量方法之间一致性的最佳方法在医学文献中已经争论了几十年（3，4）。在评估一致性时，两个度量可以具有相对一致性：即，一个度量与另一个度量相关联，第二，它们可以具有绝对一致性，即，绝对接近。几乎所有的一致性评估都有局限性，科学家必须平衡一致性评估的局限性与研究目标。Puntillo和他的同事选择关注组内相关系数（ICC）。该方法通过整合有关测量之间的平均差异的信息，改进了标准皮尔逊相关系数（仅评估相对一致性测量）。解释ICC很简单，但理解它为什么有高或低的值就比较困难了。例如，患者-家庭对疼痛强度评级的ICC为0.43。我们可以将这一结果解释为，疼痛评分的总变异性中有43%是由于研究中患者之间的疼痛差异（患者间差异），疼痛评分的57%（100%-43%）是由于同一患者的评分者对疼痛的评估差异（患者内差异）。尽管研究中所有指标的症状评分的大部分变化都存在于评分者的水平上，但作者提供了几个参考文献，表明在这种情况下ICC> 37%表明具有极好的一致性（5-7）。虽然平均而言这可能是真的，但ICC的大小高度依赖于研究样本（8，9）。由于ICC计算为患者间方差除以总方差的比值，因此当评分者之间存在极好的一致性（患者内差异较小）或患者间差异较大时，ICC可能会增加。换句话说，当应用于非常异质的患者人群时，即使评级者之间的一致性较差，也可以观察到较高的ICC。由于ICC依赖于样本中的结果分布，临床医生和研究人员应谨慎地将这些结果推广到症状评分差异性不同的其他人群。
Assessing patient’s symptoms in the intensive care unit (ICU) can be challenging, particularly when patients are severely ill and unable to communicate. The idea of using proxy symptom assessments as a surrogate for patient-reported symptoms in the ICU is appealing, but previous work has been limited and has primarily examined clinician assessments of pain (1). In this issue of Critical Care Medicine, Dr. Puntillo and colleagues (2) report a study to examine the agreement between ICU patients’ assessment of their symptoms and assessments by their family members, nurses, and physicians. The authors examined the agreement between patient and proxy symptom scores for both the intensity of and the distress caused by ten different symptoms, including being in pain, tired, short of breath, restless, anxious, sad, hungry, scared, thirsty, and confused. They enrolled 245 critically ill patients, of whom 89 were able to report symptoms on the first study day. They collected symptom ratings from patients along with ratings from family members, nurses, and physicians in a range of 2–8 hrs from the patient ratings.The optimal method to determine agreement between two measures of the same underlying phenomenon has been debated in the medical literature for decades (3, 4). When assessing agreement, two measures can have relative agreement: that is, one measure is associated with the other and, second, they can have absolute agreement, that is, be close in absolute proximity. Virtually all assessments of agreement have limitations, and scientists must balance the limitations of an agreement assessment with the goals of the study. Puntillo and colleagues chose to focus on the intraclass correlation coefficient (ICC). This method improves upon the standard Pearson correlation coefficient (which assesses only relative measure of agreement), by integrating information about mean differences between measures. Interpreting the ICC is straightforward, but understanding why it has a high or low value is more difficult. For example, the ICC for the patient-family pair rating pain intensity was 0.43. One can interpret this result as showing that 43% of the total variability in pain ratings was due to differences in pain across patients in the study (between-patient differences) and 57% of variability (100%− 43%) in pain ratings was due to differences in the assessment of pain by the raters for the same patient (within-patient differences). Even though the majority of variation in symptom ratings for all measures in the study lies at the level of the rater, the authors provide several references, suggesting that an ICC> 37% in this setting suggests excellent agreement (5–7). Although this may be true on average, the magnitude of the ICC is highly dependent on the study sample (8, 9). Because the ICC is calculated as the ratio of between-patient variance divided by total variance, it can increase when either there is excellent agreement between raters (small within-patient differences) or when there are large between-patient differences. In other words, higher ICCs may be observed when applied to a population of patients who are very heterogeneous, even when agreement between raters is poor. Because the ICC depends on the outcome distribution in the sample, clinicians and researchers should be cautious generalizing these results to other populations in which the variability in symptom ratings differs.