iScore outperforms stroke physicians in predicting patient outcomes

HONOLULU – Neurologists, internists, and emergency physicians with expertise in acute stroke care are wildly inaccurate in predicting key clinical outcomes in patients with acute ischemic stroke, a study has shown.

Indeed, in the JURASSIC (Clinician Judgment versus Risk Score to Predict Stroke Outcomes) trial, physician estimates as to whether patients would be dead or disabled at discharge were accurate in only 16.9% of cases, or one out of six times. In contrast, a validated predictive model of stroke mortality known as the iScore was on the mark 90% of the time, Dr. Gustavo Saposnik reported at the International Stroke Conference sponsored by the American Heart Association.

Dr. Saposnik and his coworkers developed the iScore because they saw the need for an objective, simple tool to stratify mortality risk in acute ischemic stroke patients. The ability to estimate prognosis with reasonable accuracy is essential to treatment decisions, discharge planning, counseling of patients and their families, and the accuracy of health care policy makers’ comparisons of outcomes between different hospitals.

For the JURASSIC trial, the investigators involved a convenience sample of 111 Ontario physicians with expertise in acute stroke care. Half were neurologists, and the rest were internists or emergency physicians. Their mean age was 40 years, and they each saw an average of 98 stroke patients annually.

Each physician was presented with case summaries for five acute ischemic stroke patients and asked to predict their likelihood of death or disability at discharge as well as 30-day mortality. The five cases were representative of the most common clinical scenarios extracted from a pool of more than 12,000 patients admitted to Ontario stroke centers, so there were no red herrings or tricks.

The 111 physicians collectively made 1,661 outcome predictions. Only 16.9% of their predictions regarding death or disability at discharge were within the 95% confidence interval for the actual observed outcomes. The physicians’ accuracy at predicting 30-day mortality was 46.9%. The iScore-based outcome estimates were within the 95% confidence interval in 90% of cases.

Only 1 of the 111 physicians was accurate in predicting clinical outcomes in four of the five cases. None got all five correct.

The iScore utilizes information easily obtainable within hours of hospital presentation without specialized lab tests or imaging. The variables incorporated into the iScore are age, sex, stroke severity and subtype, coronary artery disease, heart failure, smoking, cancer, hyperglycemia upon admission, history of atrial fibrillation, and renal disease requiring dialysis. Thus, the user-friendly iScore is suitable for use in community hospitals as well as tertiary centers, noted Dr. Saposnik of the University of Toronto.

The score enables patients to be classified into one of five categories based on the estimated risk of mortality. In a validation study (Circulation 2011;123:739-49), the 30-day mortality risk ranged from a low of 1.19% in group 1 to 41.57% in group 5.

The iScore is available as a Web-based tool.

Dr. Saposnik reported having no relevant financial conflicts.

[email protected]

HONOLULU – Neurologists, internists, and emergency physicians with expertise in acute stroke care are wildly inaccurate in predicting key clinical outcomes in patients with acute ischemic stroke, a study has shown.

Indeed, in the JURASSIC (Clinician Judgment versus Risk Score to Predict Stroke Outcomes) trial, physician estimates as to whether patients would be dead or disabled at discharge were accurate in only 16.9% of cases, or one out of six times. In contrast, a validated predictive model of stroke mortality known as the iScore was on the mark 90% of the time, Dr. Gustavo Saposnik reported at the International Stroke Conference sponsored by the American Heart Association.

Dr. Saposnik and his coworkers developed the iScore because they saw the need for an objective, simple tool to stratify mortality risk in acute ischemic stroke patients. The ability to estimate prognosis with reasonable accuracy is essential to treatment decisions, discharge planning, counseling of patients and their families, and the accuracy of health care policy makers’ comparisons of outcomes between different hospitals.

For the JURASSIC trial, the investigators involved a convenience sample of 111 Ontario physicians with expertise in acute stroke care. Half were neurologists, and the rest were internists or emergency physicians. Their mean age was 40 years, and they each saw an average of 98 stroke patients annually.

Each physician was presented with case summaries for five acute ischemic stroke patients and asked to predict their likelihood of death or disability at discharge as well as 30-day mortality. The five cases were representative of the most common clinical scenarios extracted from a pool of more than 12,000 patients admitted to Ontario stroke centers, so there were no red herrings or tricks.

The 111 physicians collectively made 1,661 outcome predictions. Only 16.9% of their predictions regarding death or disability at discharge were within the 95% confidence interval for the actual observed outcomes. The physicians’ accuracy at predicting 30-day mortality was 46.9%. The iScore-based outcome estimates were within the 95% confidence interval in 90% of cases.

Only 1 of the 111 physicians was accurate in predicting clinical outcomes in four of the five cases. None got all five correct.

The iScore utilizes information easily obtainable within hours of hospital presentation without specialized lab tests or imaging. The variables incorporated into the iScore are age, sex, stroke severity and subtype, coronary artery disease, heart failure, smoking, cancer, hyperglycemia upon admission, history of atrial fibrillation, and renal disease requiring dialysis. Thus, the user-friendly iScore is suitable for use in community hospitals as well as tertiary centers, noted Dr. Saposnik of the University of Toronto.

The score enables patients to be classified into one of five categories based on the estimated risk of mortality. In a validation study (Circulation 2011;123:739-49), the 30-day mortality risk ranged from a low of 1.19% in group 1 to 41.57% in group 5.

The iScore is available as a Web-based tool.

Dr. Saposnik reported having no relevant financial conflicts.

[email protected]

HONOLULU – Neurologists, internists, and emergency physicians with expertise in acute stroke care are wildly inaccurate in predicting key clinical outcomes in patients with acute ischemic stroke, a study has shown.

Indeed, in the JURASSIC (Clinician Judgment versus Risk Score to Predict Stroke Outcomes) trial, physician estimates as to whether patients would be dead or disabled at discharge were accurate in only 16.9% of cases, or one out of six times. In contrast, a validated predictive model of stroke mortality known as the iScore was on the mark 90% of the time, Dr. Gustavo Saposnik reported at the International Stroke Conference sponsored by the American Heart Association.

Dr. Saposnik and his coworkers developed the iScore because they saw the need for an objective, simple tool to stratify mortality risk in acute ischemic stroke patients. The ability to estimate prognosis with reasonable accuracy is essential to treatment decisions, discharge planning, counseling of patients and their families, and the accuracy of health care policy makers’ comparisons of outcomes between different hospitals.

For the JURASSIC trial, the investigators involved a convenience sample of 111 Ontario physicians with expertise in acute stroke care. Half were neurologists, and the rest were internists or emergency physicians. Their mean age was 40 years, and they each saw an average of 98 stroke patients annually.

Each physician was presented with case summaries for five acute ischemic stroke patients and asked to predict their likelihood of death or disability at discharge as well as 30-day mortality. The five cases were representative of the most common clinical scenarios extracted from a pool of more than 12,000 patients admitted to Ontario stroke centers, so there were no red herrings or tricks.

The 111 physicians collectively made 1,661 outcome predictions. Only 16.9% of their predictions regarding death or disability at discharge were within the 95% confidence interval for the actual observed outcomes. The physicians’ accuracy at predicting 30-day mortality was 46.9%. The iScore-based outcome estimates were within the 95% confidence interval in 90% of cases.

Only 1 of the 111 physicians was accurate in predicting clinical outcomes in four of the five cases. None got all five correct.

The iScore utilizes information easily obtainable within hours of hospital presentation without specialized lab tests or imaging. The variables incorporated into the iScore are age, sex, stroke severity and subtype, coronary artery disease, heart failure, smoking, cancer, hyperglycemia upon admission, history of atrial fibrillation, and renal disease requiring dialysis. Thus, the user-friendly iScore is suitable for use in community hospitals as well as tertiary centers, noted Dr. Saposnik of the University of Toronto.

The score enables patients to be classified into one of five categories based on the estimated risk of mortality. In a validation study (Circulation 2011;123:739-49), the 30-day mortality risk ranged from a low of 1.19% in group 1 to 41.57% in group 5.

The iScore is available as a Web-based tool.

Dr. Saposnik reported having no relevant financial conflicts.

[email protected]