Date of Award

January 2017

Document Type

Open Access Thesis

Degree Name

Medical Doctor (MD)



First Advisor

David M. Greer


Subarachnoid hemorrhage (SAH) is a neurological emergency associated with high morbidity and mortality. It has a pre-hospital mortality rate of up to 50%, and can cause severe disability or death in 40-60% of patients. A prompt diagnosis is crucial for timely work-up and intervention. It is still frequently missed, especially in alert patients who present with an acute headache as the only chief complaint. Acute headache is the 5th leading reason for Emergency Department (ED) visits. It accounts for about 3% of all ED visits in the US. Of these, ~2% will be secondary to aneurysmal SAH.

Some guidelines have emerged in recent years to help distinguish headache due to SAH from more benign causes. However, their generalizability is limited due to inclusion and exclusion criteria of individual studies, as well as unclear definitions of terminology used in criteria. This study aims to address these shortcomings by creating a generalizable clinical decision tool using a broader patient population, and providing clear definitions in a standardized questionnaire.

In this prospective observational study, 158 patients were interviewed using a standardized 15-item questionnaire. Patients eligible were alert, able to communicate and answer the questionnaires in English, and had a headache presentation unrelated to trauma. Data was used to identify differential features of headache secondary to SAH, and these features were used to create a 5-item clinical decision tool.

A total of 583 patients were eligible. Of those, 158 (27%) were enrolled, provided consent, and completed our questionnaire. Of these, 20 had SAH. After adjusting for confounders, patients with SAH were more likely to be ≥ 50 years old, experienced the “worst headache of [their] life,” had headache onset during exertion, had peak intensity instantaneously (<1 second), or had associated neck stiffness/pain. We proposed two clinical decision tools using these 5 features, which had a sensitivity of 100%, negative predictive value of 100%, and negative likelihood ratio of 0, comparable to existing rules.

Thus, our decision rules agree with existing ones and perform similarly, but they have clearer definition for terminology used and can be more generalizable after external validation.


This is an Open Access Thesis.