Overview of the study
The pilot study was approved and granted waiver of informed consent by the Committee of Human Research, University of California, San Francisco.
This is a prospective cohort study to test the diagnostic accuracy of Card 28 alone versus Card 28 plus the Cincinnati Stroke Scale.
The study will be conducted over a one year period at the County Communication Center, Santa Clara, California. The County Communication Center was established in 1948 and serves the population of Campbell, Cupertino, Los Altos, Los Altos Hills, Los Gatos, Monte Sereno, Morgan Hill and Saratoga. The County Communication Center receives about 90,000 emergency medical dispatch calls every year. Most of these calls are 911 calls transferred from Public Service Answering Points (PSAPs) after determination of the need for interrogation of medical emergency. The study site is a National Academy of Emergency Medical Dispatchers (NAEMD) accredited center of excellence since 2002 [10
The target study population consists of all 911 callers with symptoms suggestive of stroke in the participating county. Specifically, the accessible study population includes subjects within the participating county with symptoms suggestive of stroke whose 911 calls are answered and interrogated by the emergency medical dispatchers at the County Communication Center, Santa Clara.
(1) All 911 calls transferred by local Public Service Answering Points (PSAPs) to the County Communication Center of Santa Clara County where the emergency medical dispatchers complete the interrogation using Medical Priority Dispatch System (MPDS) protocols.
(2) All 911 calls received directly at the County Communication center of Santa Clara County where emergency medical dispatchers complete the interrogation using the MPDS protocols.
(3) All 911 calls received from subjects (patient) aged ≥ 18 years or second party calls (by-stander or family who are in close proximity to the patient and can administer the tool) by subjects ≥ 18 years of age.
(1) All calls that require immediate response (ECHO level determinant for life threatening conditions such as unconsciousness, breathing difficulty) and emergency medical dispatchers cannot complete Card 28
(2) Calls answered by emergency medical dispatchers who have not completed training on the use of Cincinnati Stroke Scale.
(3) Calls originating from the cities of Palo Alto, Mountain View, Sunnyvale, Santa Clara, and San Jose that are not interrogated by the County Communication Center for Santa Clara County.
Subject recruitment and enrollment
Patient eligibility will be determined by the emergency medical dispatchers at the time of the 911 call. No informed consent will be obtained due to the emergency nature of the call. The study protocol was reviewed and approved by the Institutional Review Board and was granted waiver of informed consent for the study.
Prior to initiation of the study, all emergency medical dispatchers in the County Communication Center will undergo training in the use of the Cincinnati Stroke Scale as well as in interpreting the findings relayed by the caller. During the study period, emergency medical dispatchers will continue to use the MPDS protocols for triaging a 911 call and assigning an ambulance. For calls that are suggestive of stroke, they will complete the Card 28 questions (Figure : Card 28 Protocol For Emergency Medical Dispatchers) followed by the scripted version of the Cincinnati Stroke Scale
Case entry protocol. If caller reports any of the symptoms listed in question 4 (after reporting a chief compliant suggestive of stroke), dispatch determinant of stroke will be assigned.
(Figure - Cincinnati stroke scale for emergency medical dispatchers).
Cincinnati stroke scale for emergency medical dispatchers.
Emergency medical dispatchers will determine eligibility of the call based on chief compliant of symptoms suggestive of stroke. Common expressions used by callers to report symptoms of stroke include altered mental status, "stroke", trouble walking, impaired speech, falling or dizziness, muscle weakness and/or facial numbness [3
]. If the emergency medical dispatcher determines the chief compliant to be a stroke related symptom, they will also screen for other life threatening symptoms. If associated life threatening symptom such as trouble breathing is reported, an ambulance will be dispatched immediately. These calls will be excluded since Card 28 will not be used to interrogate these 911 calls. We will also exclude calls which are placed by callers not in close proximity to the subject (i.e. third party caller like a family member calling from an office to report symptoms of their parent situated in their residence) and cannot administer the Cincinnati Stroke Scale. Since non-stroke causes of the symptoms (shown above) are more common in children and also because the ability of children to comprehend and administer the Cincinnati Stroke Scale to a subject is unknown, we will exclude subjects and callers less than 18 years of age from this study.
Once the emergency medical dispatcher determines that the call meets the inclusion criteria, they will interrogate the caller using the questions in Card 28. After completion of question 4 in the key interrogation sequence in Card 28 and recording of the initial diagnostic determinant of stroke (Figure ), the scripted Cincinnati Stroke Scale tool will appear for use by the emergency medical dispatchers, in the Pro-QA system (computerized version of MPDS protocol). An assessment will be recorded electronically by the emergency medical dispatchers after completion of Cincinnati Stroke Scale. If during the interrogation of the call, emergency medical dispatchers determine that symptoms are not suggestive of stroke, they will complete the call by using one of the other medical priority dispatch protocol cards. Cincinnati Stroke Scale will also be applied to these calls and assessments recorded at the end of interrogation.
(Figure : Schematic representation of the study protocol)
Schematic representation of the study protocol.
This information will be used to create a screening log, which will be compared to all dispatch calls with chief complaints related to stroke symptoms to ensure eligible patients were not missed by the emergency medical dispatchers.
During the study period, we will record the following prehospital variables. The source of the data variables is the computer assisted dispatch (CAD) database, which will capture all the listed variables during the 911 call. Variables will include the demographics of the subject (age, sex, location of the caller) the chief complaint of the caller; use of Card 28 based on the chief complaint of the caller; assessments after Card 28 which will be listed as dispatch determinant of stroke (CVA-28), assessment following completion of Cincinnati Stroke Scale, time from initiation of Card 28 to completion of key questions in the protocol and time from initiation to completion of Cincinnati Stroke Scale.
The primary study outcome is the hospital based diagnosis of all subjects who were interrogated with Card 28 and Cincinnati Stroke Scale by emergency medical dispatchers. The secondary study outcomes include time to complete the two stroke protocols and rate of intravenous t-PA use in this cohort.
Data quality and management
The prehospital agency and OSHPD data will be encrypted and shipped to the Database Management Unit of Academic Research Systems, a unit of the Clinical and Translational Science Institute, University of California San Francisco. The data will be converted to SAS tables and visually inspected for inconsistencies. Using a subset of the data, probabilistic linkage will be used to link the databases (Figure ) [11
]. The merged dataset will be de-identified and will be used for statistical analysis.
(Figure : Data variables used in probabilistic matching algorithm)
Data variables used in probabilistic matching algorithm.
The primary source of individual patient outcome data is the discharge abstract file of OSHPD that also contains variables like age, date of birth, sex, race/ethnicity, zip code, county of residence, hospital zip code, admission date, month and year, principal emergency department and discharge diagnosis, principal emergency department and hospital procedure codes, and discharge date. Prehospital database contains the following data variables: date of birth, gender, race, zip Code of residence, hospital ID, date of service, county of caller residence, zip code of caller residence, county in which destination hospital is located and the service date/month/year. These variables will be used to link the prehospital data with the outcome database. The details of the methodology will be published in another paper but preliminary results of the linkage algorithm showed close to 90% unique matches.
Data analysis plan
The primary analysis will compare the sensitivity and specificity of the two protocols using logistic regression and generalized estimating equations to adjust for clustering by dispatcher. Wald test will be used to test the significance of the coefficients in the regression model. The data variables (listed in procedures section) in the computer assisted dispatch database and OSHPD database will be used to link the two databases and to calculate the diagnostic accuracy of the triage protocols. The hospital discharge diagnosis will be used as the gold standard for calculation of the performance characteristics of the protocols. Use of IV t-PA in the true stroke population will be determined by the procedure code in OSHPD and will be compared between the two protocol groups.
For the primary analysis, sensitivity will be defined as the proportion of true positives transported by EMS (stroke diagnosis by both emergency medical dispatchers and hospital discharge diagnosis) to the total number of patients with a hospital based stroke diagnosis. Specificity will be defined as the proportion of true negatives (no diagnosis of stroke by emergency medical dispatchers providers and non-stroke discharge diagnosis) to the total number of patients with hospital based non stroke diagnosis given by physicians (table ).
Variables used in measurement of test characteristics of Card 28 alone and Card 28 and Cincinnati Stroke Scale
Sample size calculations
During the one year study period, we anticipate about 90,000 emergency medical dispatch calls and about 350 eligible stroke calls. Based on published data showing a sensitivity of 40% for emergency medical dispatchers using MPDS protocol, we will have 80% power to detect an absolute 15% greater positive predictive value for emergency medical dispatchers stroke recognition using Cincinnati Stroke Scale with a two-sided alpha = 0.05