Development of risk-adjustment methods to evaluate systems of emergency medical care
Lead Research Organisation:
University of Sheffield
Department Name: Health and Related Research
Abstract
Emergency care systems, involving ambulance services, hospitals and general practitioners, have a crucial role to play in delivering life-saving treatments, yet these systems vary hugely across the United Kingdom and across the world. To date, there has been little research into finding out how emergency care systems perform and what makes a good system. The research that has been undertaken has focussed upon measures, such as waiting times, that do not necessarily reflect the quality of medical care provided. Comparing death rates between emergency care systems would be a good way of comparing quality of care, but could be misleading if patients use different systems in different ways or if local populations have different underlying degrees of illness.
We plan to develop a method for comparing death rates among patients treated by different emergency care systems that takes into account differences in the health of the local population and the way they use emergency services. This will ensure that any differences we identify are likely to be due to the care provided. We can then draw conclusions about what types of emergency care systems provide the best quality of care, in terms of lives saved.
We will test patient characteristics that can be recorded by the first health professional to treat them to find out whether these characteristics predict a higher chance of dying, and will use statistical techniques to select the characteristics that are the best predictors. By using the patient?s characteristics to predict their chances of dying we can then determine, by seeing whether they actually die or not, whether the emergency care system is performing better or worse than expected.
This approach will allow us to compare systems of emergency care and find out whether certain types of system perform better than others, and whether individual systems are performing to the expected standard. We can then draw conclusions regarding what makes a good emergency care system that will guide future development of emergency ambulance, hospital and general practitioner services.
We plan to develop a method for comparing death rates among patients treated by different emergency care systems that takes into account differences in the health of the local population and the way they use emergency services. This will ensure that any differences we identify are likely to be due to the care provided. We can then draw conclusions about what types of emergency care systems provide the best quality of care, in terms of lives saved.
We will test patient characteristics that can be recorded by the first health professional to treat them to find out whether these characteristics predict a higher chance of dying, and will use statistical techniques to select the characteristics that are the best predictors. By using the patient?s characteristics to predict their chances of dying we can then determine, by seeing whether they actually die or not, whether the emergency care system is performing better or worse than expected.
This approach will allow us to compare systems of emergency care and find out whether certain types of system perform better than others, and whether individual systems are performing to the expected standard. We can then draw conclusions regarding what makes a good emergency care system that will guide future development of emergency ambulance, hospital and general practitioner services.
Technical Summary
Systems of emergency care, consisting of ambulance services, emergency departments, hospitals and general practitioners, play a crucial role in delivering time-critical life-saving medical interventions, yet these systems vary substantially. Controlled trials of emergency systems are usually impractical or unethical, so the evidence base for development is very limited. National audits of emergency care have been undertaken, focussing upon process measures (such as waiting times), but have not evaluated important measures, such as mortality.
Comparison of crude mortality between systems is limited by differences in case mix and use of emergency services. Risk-adjustment methods have been developed for specific patient groups, but limiting evaluation to these groups risks development of systems that favour a minority of patients with a single, clear pathology, whereas most emergency medical cases have multiple pathologies and co-morbidities.
We plan to develop a method for evaluating emergency care systems that takes a system-wide approach, is based upon risk-adjusted mortality, and uses variables that can be consistently and reliably measured in emergency care. We will systematically identify all potentially useful variables, test the feasibility of collecting them and measure their association with outcome. We will then derive a risk-adjustment model based upon a limited number of variables that each independently predict outcome. We will then assess the performance of the model and it?s constituent variables in a variety of different settings, and seek explanations for any discrepancies between observed and expected outcome. Finally we will refine the model to ensure applicability to a wide variety of settings. Secondary analyses will test the model and it?s constituent variables in specific subgroups or patients.
The principal outputs of this project will be: 1) Identification of the key predictor variables that should be collected and used in the analysis of any non-randomised study of emergency medical care, 2) Estimates of the predictive coefficients for each variable in the overall population and in different subgroups of patient, and 3) A risk-adjustment tool that can be used in a wide range of different settings to compare mortality among patients with medical emergencies.
This project will allow comparison of systems of care in situations where randomised comparison would be impossible, impractical or unethical, ensuring that service development is rational and evidence-based. Comparison of actual to case-mix adjusted mortality within individual systems will provide valuable quality assurance.
Comparison of crude mortality between systems is limited by differences in case mix and use of emergency services. Risk-adjustment methods have been developed for specific patient groups, but limiting evaluation to these groups risks development of systems that favour a minority of patients with a single, clear pathology, whereas most emergency medical cases have multiple pathologies and co-morbidities.
We plan to develop a method for evaluating emergency care systems that takes a system-wide approach, is based upon risk-adjusted mortality, and uses variables that can be consistently and reliably measured in emergency care. We will systematically identify all potentially useful variables, test the feasibility of collecting them and measure their association with outcome. We will then derive a risk-adjustment model based upon a limited number of variables that each independently predict outcome. We will then assess the performance of the model and it?s constituent variables in a variety of different settings, and seek explanations for any discrepancies between observed and expected outcome. Finally we will refine the model to ensure applicability to a wide variety of settings. Secondary analyses will test the model and it?s constituent variables in specific subgroups or patients.
The principal outputs of this project will be: 1) Identification of the key predictor variables that should be collected and used in the analysis of any non-randomised study of emergency medical care, 2) Estimates of the predictive coefficients for each variable in the overall population and in different subgroups of patient, and 3) A risk-adjustment tool that can be used in a wide range of different settings to compare mortality among patients with medical emergencies.
This project will allow comparison of systems of care in situations where randomised comparison would be impossible, impractical or unethical, ensuring that service development is rational and evidence-based. Comparison of actual to case-mix adjusted mortality within individual systems will provide valuable quality assurance.
