Digital Decision Support in the Management of Patients With Chest Pain
- Conditions
- Chest PainChest--DiseasesAcute Coronary Syndrome
- Registration Number
- NCT05767619
- Lead Sponsor
- Vastra Gotaland Region
- Brief Summary
The goal of this observational study is to develop a decision support system in patients presenting with chest pain in the prehospital setting. The main question it aims to answer is:
• Performance of a machine learning based model for decision support of patients in contact with emergency medical services due to chest pain
Participants will be asked to:
* respond to questions asked by the clinician at the scene regarding previous known risk factors and pain characteristics
* consent to the collection of routinely available data from medical records
* consent of taking one blood sample capillary or venous (if perifer catheter is placed for standard care reasons) troponin and glucose which is measured at the scene, disposed, and the result is entered in the clinical report form.
- Detailed Description
Prehospital emergency care has undergone dramatic changes in recent decades. From the fact that the ambulance nurse has previously started care at the site of the illness, with the goal of transporting the patient to the nearest emergency room, prehospital care has become increasingly differentiated. This means that "care at the right level of care" has become a watchword and a number of different levels of care have become relevant, ranging from fast track and rapid investigation and treatment in hospital for heart attack, acute myocardial infarctions and strokes to the patient being left at the scene with advice on self-care.
In principle, there is three overall levels of care:1) Need for inpatient resources;2) Need for primary care contact or visit by mobile healthcare team within the next day and 3) Referral to home care, provide support for self-care or treatment on site.
This places competence requirements on the ambulance nurse with requirements for a prehospital assessed condition compatible with a level of care where the patient's needs can be met. This approach puts patient safety in focus in a different way than before. Because with this procedure, patients with time-sensitive conditions (such as stroke, heart attack, acute myocardial infarction and sepsis) run an increased risk of being left at the scene with advice on self-care, due to inappropriate prehospital assessment. In prehospital emergency care, the assessment of the severity of the patient's condition takes place in two stages: 1) At the emergency dispatch centre when the patient has called the national emergency number and 2) At the scene of the illness by the ambulance nurse after arrival at the patients side.
The basis for prehospital decision support is a) identification of time-sensitive conditions, i.e. conditions where the time to initiation of causal treatment can affect the prognosis, and b) identification of predictors, i.e. factors that are already prehospitally characteristic of the condition (disease or accident) itself, but also of the severity of the condition. The classic examples of time-critical conditions are manifestations of cardiovascular diseases such as heart attack ,acute myocardial infarction and stroke, but also serious infectious diseases such as sepsis and severe trauma. Predictors can be identified via measurement of vital parameters such as pulse, blood pressure and oxygen saturation, medical history (previously known diseases and current onset), current symptom picture, clinical manifestations ( pallor) examination findings (ECG) and analysis of biochemical markers by capillary blood test (glucose, troponin).
Current studies indicate that for many patients in contact with the emergency medical service due to acute chest pain, other options than the emergency department, e.g. follow-up in primary care, may be more beneficial for the patient and less resource-intensive for the ambulance and the emergency care. In these cases, a decision support system based on gender, medical history, symptoms and clinical observations including ECG and and biochemical markers, (troponin) could provide support for the ambulance nurse. Chest pain is one of the most common search causes and constitutes about 10% of assignments in the Swedish emergency medical service.
Linked to the assessment in the prehospital environment is patient safety. An inappropriate prehospital assessment can compromise patient safety and risk delaying time to treatment. Primarily, it refers to patients with time-sensitive conditions who are not transported by ambulance directly to hospital after the initial assessment. At the same time, patient safety can be compromised by transporting frail elderly people to an emergency room, where long wait times can increase the risk of complications. Transportation of low-risk patients can also increase the risk of crowding in the emergency department and also have a crowding-out effect (lack of ambulance availability in case of high priority cases). An example of displacement effects is extended response times for the ambulance in the event of sudden unexpected cardiac arrests.
Patient safety is poorly studied in the prehospital setting. Own experiences indicate that the risk of adverse events in so-called "Prio1 assignments" (highest priority) is particularly high and that in about 20% of these assignments the prehospital assessment can be questioned.
The goal of the present study is to evaluate and further develop a decision support system developed within the Emergency Medical Service in Region Halland, Sweden with collection of variables to validate the previously developed model on unseen data and to further develop a machine learning model for classification. The idea is that such decision support should provide support for the ambulance nurse in the assessment of the patient at the scene, partly to optimize the possibility of the patient quickly getting to the right level of care and partly to increase patient safety. The objective is to only collect data at the scene (blood test + questionnaire) together with routine data from medical records. Patients consenting to be part of the study will not receive any other care than standard care according to guidelines which constitutes transport to the emergency department for further examination.
Recruitment & Eligibility
- Status
- RECRUITING
- Sex
- All
- Target Recruitment
- 2000
- Patient in contact with the emergency medical service and patient main symptom is chest pain or chest discomfort
- Primary assignment (not assessed by physician in primary care, hospital)
- Assignment taking place outside participating emergency medical service organisation geographical catchment area
- Under 18 years of age
- Unwillingness to participate
- Unable to participate (language, dementia, etc.)
- Other
Study & Design
- Study Type
- OBSERVATIONAL
- Study Design
- Not specified
- Primary Outcome Measures
Name Time Method Clinical outcome prediction of the decision support model based on a composite outcome Time-sensitive condition: from EMS inclusion to hospital discharge follow-up time 1 day up to 100 days; Death: from EMS inclusion up to seven days; Adverse events: from EMS inclusion up to 72 hours. The models predictive ability for classifying low risk and high risk patients. Including features collected from the clinician at the scene, patient characteristics, previous medical history, pain characteristics, time duration of pain/discomfort, ECG interpretation, vital signs,Troponin and Glucose measurements.
A composite outcome is based on the following:
1. Time-sensitive diagnose at hospital discharge
2. Death within seven days
3. Adverse events within 72 hours
Adverse events is defined as : cardiac arrest, ventricular arrhythmias, shock, convulsions, heart failure ,hypotension, syncope and loss of consciousness and associated with a grade 4 or 5 according to the CCTAE v 6.0.
Multiple models will be developed with an ensemble of classifiers and evaluated determining the best model performance. Measures reported are discriminative performance of the model roc-AUC, calibration, sensitivity, specificity, accuracy, positive predictive value, negative predictive value.
- Secondary Outcome Measures
Name Time Method
Trial Locations
- Locations (3)
Department of Prehospital Emergency Care,Region Halland
🇸🇪Kungsbacka, Region Halland, Sweden
Department of Prehospital Emergency Care, Skaraborg
🇸🇪Lidköping, Region Vastra Gotaland, Sweden
Department of Prehospital Emergency Care, Sahlgrenska University Hospital
🇸🇪Gothenburg, Region Vastra Gotaland, Sweden