Risk Scores to Predict Complication Risk in Suspected ACS
Risk Scores to Predict Complication Risk in Suspected ACS
These results indicate that the GRACE and FFE scores are able to identify a very low risk of complications in patients with chest pain admitted for suspected ACS, and that only the GRACE score may be able to identify a significant number of patients at almost no risk. To the best of our knowledge, the present study is the first to compare directly the abilities of the GRACE and FFE scores and the Goldman rule to predict a low complication risk in patients admitted with chest pain.
Of the three models, the GRACE score had the best risk prediction ability (AUC 0.76) and could theoretically have been used to identify correctly almost one in five admitted patients as having no (or almost no) risk of complications. The observation that the identified patients were discharged after a median hospital stay of only 1 day and had a lower than average age further support that they were indeed at low risk. These results thus imply that the GRACE score, after validation studies using appropriate cut-offs, may be able to identify a significant number of patients with suspected ACS that are suitable for safe investigation without monitoring, perhaps even outside the hospital. Since the use of a monitoring bed is the most costly part of the management of patients with chest pain, such a practice would save substantial resources that could instead be used for true high-risk patients. The GRACE score is easily obtainable and routine ED assessment of patients with chest pain normally includes all eight parameters of the score. Further research is needed to explore the value of combining the variables of the GRACE and FFE risk scores, or creating modified models, to attain an even better risk prediction. Studies may also identify parameters with greater predictive ability than others. For instance, age alone may predict the risk of adverse events within 30 days with an AUC of 0.66 among ED patients with chest pain.
The FFE score, previously not tested in unselected patients with chest pain, predicted the risk of complications less well than the GRACE score and only 14 patients (2.5%) could theoretically have been identified as having no (or almost no) risk of complications. This is perhaps surprising since the FFE score was superior to the GRACE score in predicting low risk in patients with NSTEMI and UA in a previous study, and since almost 90% of our patients with ACS lacked ST elevation on the ECG. The poorer performance of the FFE score in the present study is probably explained by the observations (Table 2) that UA symptoms (indicating lower risk in the FFE score) and positive troponin (included in GRACE but not FFE) were both more common in patients with complications.
For the Goldman rule, the area under the ROC curve did not differ significantly from 0.5 and the lowest Goldman risk group had a complication rate of 4.1%. Thus, in contrast to previous studies, the Goldman rule did not successfully predict a very low risk of complications in our patients. One explanation for this might be that the Goldman rule is based on ECG changes and a few clinical variables, and hence may be of limited use in a population such as the one in the present study with few cases of distinct ST elevation and a low overall risk of complications. Indeed, in this study more than 80% of the patients fell into the very low or low Goldman risk groups.
Established standard risk stratification tools in the ED such as troponin T level and the ECG were clearly insufficient as predictors of a low complication risk. In patients with complications the initial troponin T level and the ECG indicated low risk (ie, were normal) in two out of five and two out of three patients, respectively. The likelihood of ACS would probably also not be an optimal risk prediction tool since 34% of the patients with complications did not get them from ACS.
In the present study only 5.7% suffered a complication and 0.9% died. Complication rates and mortality differ distinctly in previous studies of patients admitted for suspected ACS. In two studies evaluating the GRACE score, in-hospital mortality was 9.3% and 1.4%, respectively. Studies on Goldman's rule from 1996 and 2001 showed complication rates of 8.9% and 6.6%, respectively. In a Swedish study from 1994, 18% of the admitted patients had complications. These differences may be due to different definitions of complications but are probably primarily due to different patient populations and changes in therapy over the years. For instance, high-risk patients with STEMI now often bypass the ED at many hospitals on the way to the angiography suite. These differences in complication rates and mortality support the need for multicentre studies to evaluate the true benefit of implementing risk scores for ED patients with chest pain in routine care.
In the recent decade, ED crowding and a general shortage of hospital beds have triggered a development of new diagnostic strategies for patients with suspected ACS such as immediate imaging, accelerated testing for multiple biomarkers, the chest pain unit concept and evaluation of the patient in the waiting room without monitoring. In this context, the few complications and the low mortality among the patients in the present study underscores the challenge, the importance and the potential benefit of improved risk stratification in patients with chest pain with suspected ACS. Of our patients, 94% had no complication and one might argue that, in retrospect, these patients did not truly need an in-hospital bed. However, many of these patients might have avoided complications by the nursing received during the hospital admission and some were probably hospitalised for reasons in addition to a suspected ACS. Nevertheless, we believe it is likely that many of the patients with a GRACE score below 87 in the present study could have been evaluated without monitoring, perhaps even as outpatients with a rapid return visit for additional tests. Larger studies are needed to confirm this hypothesis.
This study was performed at one university hospital and the results are not necessarily generalisable to other hospitals. Also, the number of patients and complications were limited. Before clinical application, the predictive ability of the GRACE score at specific cut-off points must therefore be tested in new cohorts and at other centres.
Discussion
These results indicate that the GRACE and FFE scores are able to identify a very low risk of complications in patients with chest pain admitted for suspected ACS, and that only the GRACE score may be able to identify a significant number of patients at almost no risk. To the best of our knowledge, the present study is the first to compare directly the abilities of the GRACE and FFE scores and the Goldman rule to predict a low complication risk in patients admitted with chest pain.
Of the three models, the GRACE score had the best risk prediction ability (AUC 0.76) and could theoretically have been used to identify correctly almost one in five admitted patients as having no (or almost no) risk of complications. The observation that the identified patients were discharged after a median hospital stay of only 1 day and had a lower than average age further support that they were indeed at low risk. These results thus imply that the GRACE score, after validation studies using appropriate cut-offs, may be able to identify a significant number of patients with suspected ACS that are suitable for safe investigation without monitoring, perhaps even outside the hospital. Since the use of a monitoring bed is the most costly part of the management of patients with chest pain, such a practice would save substantial resources that could instead be used for true high-risk patients. The GRACE score is easily obtainable and routine ED assessment of patients with chest pain normally includes all eight parameters of the score. Further research is needed to explore the value of combining the variables of the GRACE and FFE risk scores, or creating modified models, to attain an even better risk prediction. Studies may also identify parameters with greater predictive ability than others. For instance, age alone may predict the risk of adverse events within 30 days with an AUC of 0.66 among ED patients with chest pain.
The FFE score, previously not tested in unselected patients with chest pain, predicted the risk of complications less well than the GRACE score and only 14 patients (2.5%) could theoretically have been identified as having no (or almost no) risk of complications. This is perhaps surprising since the FFE score was superior to the GRACE score in predicting low risk in patients with NSTEMI and UA in a previous study, and since almost 90% of our patients with ACS lacked ST elevation on the ECG. The poorer performance of the FFE score in the present study is probably explained by the observations (Table 2) that UA symptoms (indicating lower risk in the FFE score) and positive troponin (included in GRACE but not FFE) were both more common in patients with complications.
For the Goldman rule, the area under the ROC curve did not differ significantly from 0.5 and the lowest Goldman risk group had a complication rate of 4.1%. Thus, in contrast to previous studies, the Goldman rule did not successfully predict a very low risk of complications in our patients. One explanation for this might be that the Goldman rule is based on ECG changes and a few clinical variables, and hence may be of limited use in a population such as the one in the present study with few cases of distinct ST elevation and a low overall risk of complications. Indeed, in this study more than 80% of the patients fell into the very low or low Goldman risk groups.
Established standard risk stratification tools in the ED such as troponin T level and the ECG were clearly insufficient as predictors of a low complication risk. In patients with complications the initial troponin T level and the ECG indicated low risk (ie, were normal) in two out of five and two out of three patients, respectively. The likelihood of ACS would probably also not be an optimal risk prediction tool since 34% of the patients with complications did not get them from ACS.
In the present study only 5.7% suffered a complication and 0.9% died. Complication rates and mortality differ distinctly in previous studies of patients admitted for suspected ACS. In two studies evaluating the GRACE score, in-hospital mortality was 9.3% and 1.4%, respectively. Studies on Goldman's rule from 1996 and 2001 showed complication rates of 8.9% and 6.6%, respectively. In a Swedish study from 1994, 18% of the admitted patients had complications. These differences may be due to different definitions of complications but are probably primarily due to different patient populations and changes in therapy over the years. For instance, high-risk patients with STEMI now often bypass the ED at many hospitals on the way to the angiography suite. These differences in complication rates and mortality support the need for multicentre studies to evaluate the true benefit of implementing risk scores for ED patients with chest pain in routine care.
In the recent decade, ED crowding and a general shortage of hospital beds have triggered a development of new diagnostic strategies for patients with suspected ACS such as immediate imaging, accelerated testing for multiple biomarkers, the chest pain unit concept and evaluation of the patient in the waiting room without monitoring. In this context, the few complications and the low mortality among the patients in the present study underscores the challenge, the importance and the potential benefit of improved risk stratification in patients with chest pain with suspected ACS. Of our patients, 94% had no complication and one might argue that, in retrospect, these patients did not truly need an in-hospital bed. However, many of these patients might have avoided complications by the nursing received during the hospital admission and some were probably hospitalised for reasons in addition to a suspected ACS. Nevertheless, we believe it is likely that many of the patients with a GRACE score below 87 in the present study could have been evaluated without monitoring, perhaps even as outpatients with a rapid return visit for additional tests. Larger studies are needed to confirm this hypothesis.
Limitations of the Study
This study was performed at one university hospital and the results are not necessarily generalisable to other hospitals. Also, the number of patients and complications were limited. Before clinical application, the predictive ability of the GRACE score at specific cut-off points must therefore be tested in new cohorts and at other centres.
Source...