Mary Joanne Verhoef

Chapter 5 102 contrast with the meta-analysis by Downar et al.21 where sensitivity of the SQ was lower than specificity (67% versus 80%, area under the curve 0.83 (95%CI 0.79-0.87)). This is similar to two earlier studies in cancer patients on the performance of the SQ in a general practice stage IV cancer population by Moroni et al.28 and in an outpatient oncology clinic by Moss et al.27 Our finding that sensitivity of the SQ was higher at the ED (89.1%) than in the studies by Moroni (70%) and Moss (75%) confirms that patients included in this study had higher a priori risk of death within one year compared to Moroni’s and Moss’ studies. In addition, it is likely that physicians working in different healthcare settings estimate the SQ differently.21, 27, 28 In screening for palliative care needs at the ED, identifying as many truly deteriorating patients as possible (hence high sensitivity), is preferred over selecting very accurately who might not benefit from palliative care (high specificity). Specificity may be low in our study, because the ED-visit can also be accompanied with temporary deterioration, in which patients’ condition can improve after interventions at the ED. These interventions may encompass improvement of pain medication, start of antibiotics and blood transfusions. A study in the ED-setting in patients with heart failure showed comparable performance of the SQ (sensitivity 79%, specificity 57%), but with a higher c-index (0.68).35 The poor discriminative ability of the SQ for death within one year indicates that the SQ should not be used as a predictor of one-year-survival; however, with a HR of 2.1 for approaching death in NS-patients (with a median survival of three months) compared to S-patients (with a median survival of nine months), the SQ can identify those who have more urgent palliative care needs. Downar et al. suggested that the SQ would perform better with addition of other indicators of palliative care needs.21, 43 We therefore added poor physical performance status (ECOG 3-4), which resulted in increased specificity (from 40% to 92%) at cost of sensitivity (from 89% to 40%). This is probably because most patients visit the ED with problems causing a decreased physical performance (which might improve after the ED-visit), regardless of their prognosis. The high positive predictive value (95.1%) demonstrates that SQ plus ECOG 3-4 is correct in 95% (PPV) in identifying patients who die within the year, with a hazard ratio for approaching death of 2.5. The c-index of SQ plus ECOG 3-4 increased to a moderate 0.65. It may feel obvious that performance status is part of the assessment associated with answering the SQ; however, both the SQ and ECOG 3-4 were independent predictors of approaching death in multivariable analysis. Since addition of ECOG 3-4 improved the SQ’s discriminative ability, an NSanswer should be followed by an assessment of the performance status to differentiate between urgency of palliative care needs. Our study shows that the SQ plus ECOG 3-4 can discern three groups of patients with different levels of urgency for initiating palliative care. Firstly, although S-patients had the

RkJQdWJsaXNoZXIy MTk4NDMw