It is the primary aim of this study to identify symptoms and/or specific words (trigger words) indicative of spontaneous subarachnoid haemorrhage (sSAH) during emergency telephone calls to the Emergency Medical Service Copenhagen (EMS).

Further, it is the aim to determine the association between the symptoms/trigger words and sSAH, the sensitivity of the symptoms/trigger words and finally, to identify factors in the telephone visitation that may influence the level of activated prehospital response

Study design: Nested case-control study.

Data sources:

A cohort of patients with confirmed sSAH is established. It will consist of patients admitted between January 1, 2014 and December 31, 2018 registered in the Danish National Patient Register with a primary discharge diagnosis of spontaneous subarachnoid haemorrhage (International Classification of Disease code I60.0-I60.9). Using central person register-numbers, their medical records will be extracted and reviewed to validate the diagnosis. Validation is performed by two independent investigators with the assistance of a predeveloped case report form with diagnostic criteria (named "CRF Diagnosis validation"). At the time of the study design, the cohort contains data on patients from year 2014 from a previous study. For the present study data must be collected on patients from year 2015 to 2018. The Danish National Patient Register is accessed online via the Danish Health Data Board (Sundhedsdatastyrelsen).

Controls are identified in the EMS Copenhagen database of emergency telephone calls.

Audiotaped emergency medical telephone calls are kept on record at the EMS Copenhagen. Audio files of calls from patients with confirmed sSAH and selected controls are extracted.

The Central Person Register is daily updated with regards to vitality. This register can be accessed online via the Danish Health Data Board (Sundhedsdatastyrelsen).

The Danish National Patient Register contains information on diagnoses related to hospital admissions. These will be used to calculate the Charlson Morbidity Index score.

Variables to be extracted:

- The Danish National Patient Register:

For patients with a primary discharge diagnosis of sSAH, admitted between years 2015 and 2019:

• CPR-number
• Admission date and time
• Admission hospital To calculate the Charlson Morbidity Index score, a second data extraction is performed for patients with sSAH validated by medical record review. This time, data on prior diagnoses are extracted.

Existing sSAH cohort (patients admitted in year 2014):

• CPR-numbers of patients with sSAH
• Date and time of first hospital admission in relation to the sSAH.
• Activated EMS response.

EMS Copenhagen:

• CPR-numbers of matched controls.
• Activated EMS response for patients with sSAH in the years 2015-2018.

Audiotaped EMS telephone calls:

• In a pilot study a Case Report Form (named "CRF: symptoms") will be designed by replaying twenty randomly selected audiotaped emergency calls from patients with sSAH. Members of the research group will listen to the selected telephone calls and discuss which symptoms and information should be extracted. A RedCap data collection template is then coded according to the CRF. For the main study, all telephone calls from patients with sSAH and controls without sSAH are then replayed and data extracted to the RedCap data collection system using "CRF: Symptoms".

The Central Person Register:

• 30-day vital status on all patients with confirmed sSAH.

Data quality:

Since 1977 the Danish National Patient Register has registered all inpatient, outpatient and emergency department visits to Danish hospitals. It is considered a high quality database. Specifically for sSAH however, we have found a positive predictive value of only 63%. To minimize the inaccuracy of using the these unvalidated data, we have chosen to review all medical records twice to verify the sSAH diagnosis. Accuracy is therefore expected to be very high.

Audiotaping of emergency telephone calls became an automated process by January 2014. At certain time periods, not all calls were recorded due to changes in the data storing system. If patients are believed to have called the EMS in any of these periods, this will be addressed in the final manuscript.

Blinding:

For the diagnosis validation, investigators will be blinded to the others´ findings as they will enter data into individual electronic spreadsheets. In case of disagreement a third investigator will review the case.

Emergency telephone calls from cases and controls are anonymized before data collection by renaming the audiofiles according to a predeveloped key. Data entered in RedCap will be entered under the anonymized file name. Data are unblinded when statistical analyses are performed.

Statistics:

Reproducibility of Case Report Form: Several investigators will divide the task of replaying telephone calls and extracting data. To ensure a uniform data collection, reproducibility of categorizations in the "CRF: Symptoms" is tested and instructions and definitions adjusted if required.

To assess reproducibility, selected variables in the CRF is compared between two investigators. Kappa-values above 0.5 is considered acceptable. Binary variables are tested with kappa-statistics, ordered categorical variables are tested with weighted kappa statistics and numerical variables are tested using the intraclass correlation coefficient.

Descriptive statistics: Continuous data will be reported as medians with inter quartile range (IQR). Proportions of patients with each symptom/ trigger words will be reported as percentages with 95% confidence intervals (CI).

Statistical analyses: Associations between specific symptoms / trigger words and sSAH will be calculated as odds ratios. Sensitivities of identified symptoms or trigger words are calculated. P<0.05 will be considered statistically significant. To identify predictors for dispatching an ambulance or physician-staffed mobile response unit with the highest priority level, a logistic regression analysis will be performed with dispatch yes/no as the outcome and age, sex and variables obtained in the CRF as possible predictors.

Under the assumption that the symptoms of interest occur in 25% of sSAH patients and 5% of controls a statistically significant difference can be detected with a power of 80% at the 5% significance level, providing we include 195 cases and 390 controls (ratio 1:2).

Statistical analyses will be performed in collaboration with the department statistician.

Missing data:

In case of more than 10% missing data for any variable a plan for imputation will be made.

Bias:

An unknown proportion of patients with confirmed sSAH is expected not to be registered with an emergency telephone call to the EMS Copenhagen. This can be because they were not admitted via the EMS in which case they may constitute a certain group of patients with regards to their general use of the health care services.

sSAH patients without a registered emergency medical telephone call might also reflect the nature of the acute situation in which a CPR-number may not always be obtainable. This can create an inclusion bias as audiotaped telephone calls cannot be identified. These patients are potentially among the most acutely affected.

To compare how patients without a registered contact with the EMS Copenhagen differ from those with a contact, they will be compared by age, sex, Charlson Comorbidity Index scores and 30-day mortality.

Inclusion Criteria:

• Having called the emergency medical service Copenhagen between January 1, 2014 and December 31, 2018.
• Must have a Danish Central Person Register-number.
• Cases must be first-ever subarachnoid haemorrhages.
• Cases must have survived to admission.

Exclusion Criteria:

• Patients transferred to a to a hospital in the Capital Region of Denmark from abroad or from another Danish geopolitical region.
• Subarachnoid haemorrhages that have occurred during hospitalization are also excluded.

People aged 18 years or older with a Danish Central Person Register-number (CPR-number), living, working or visiting the Capital Region of Denmark between January 1, 2014 and December 31, 2018.