Risk modelling 2: Risk modelling for quality improvement in the critically ill
Background
When patients are critically ill, they are rarely able to choose where they are treated. We must therefore ensure that all hospitals deliver high quality critical care. One way to do this is through clinical audit by comparing the outcome of care between patients treated in different hospitals. It is essential that this comparison takes into account the different patients treated in different hospitals. For example, if one hospital admits many more very sick or very elderly patients, then one might expect the death rate in that hospital to be higher. To do this, we use risk prediction models. These statistical models take information about the patient known before, or soon after, the start of their illness and make a prediction of their likely outcome based on many thousands of previous similar patients.
Large amounts of information (data) are routinely collected about patients using NHS services, but often we do not make the best possible use of this data to improve patient care. Data is held by different organisations in different databases. Joining up these different databases (data linkage) can give us a more complete picture of what has happened to a patient.
Design
The Intensive Care National Audit & Research Centre (ICNARC) is an independent charity that runs national clinical audits to monitor and improve care for critically ill patients. ICNARC coordinates two national clinical audits: The Case Mix Programme, a national clinical audit of adult critical care; and the National Cardiac Arrest Audit, a national clinical audit of in-hospital cardiac arrest. This research study used data linkage to improve our risk prediction models and ensure that the audits provide accurate and up-to-date information back to the hospitals to support quality improvement.
Results
By linking data from the Case Mix Programme with national death registrations, we were able to predict how many patients die by particular points in time (30 days, 90 days and one year after their critical illness). By linking with routine hospital data, we were able to take better account of how sick patients were before they became critically ill and also look at how many days they spent in hospital in the year after their critical illness and the costs of these hospital stays. By linking with two other national clinical audits (the UK Renal Registry and the National Diabetes Audit), we were able to develop new models to predict important problems of renal failure and diabetes that patients can experience after critical care. For cardiothoracic critical care (caring largely for patients who have had heart surgery), by linking data from the Case Mix Programme with the National Adult Cardiac Surgery Audit we were able to get a more complete picture of how sick these patients were before they were admitted to critical care, helping us to improve our risk models to make fairer comparisons for these patients.
Conclusion
By linking data from the National Cardiac Arrest Audit with routine hospital data, we were able to get a better picture of how sick these patients were before their cardiac arrest, and by linking with death registrations, we were able to develop new models to predict their longer-term survival. Finally, by linking data from the National Cardiac Arrest Audit with the Case Mix Programme, we were able to understand better which patients need to spend longer in critical care after their cardiac arrest.
Who is leading the study?
Prof David Harrison, Head Statistician ICNARC
Further information
This study is funded by the National Institute for Health Research (NIHR) – Health Services and Delivery Research (HS&DR) Programme (Project: 14/19/06).
Publications
Bedford JP, Ferrando-Vivas P, Redfern O, Rajappan K, Harrison DA, Watkinson PJ, Doidge JC. New-onset atrial fibrillation in intensive care: epidemiology and outcomes. Eur Heart J Acute Cardiovasc Care 2022; 11(8):620-8. http://dx.doi.org/10.1093/ehjacc/zuac080 Bedford JP, Ferrando-Vivas P, Redfern O, Rajappan K, Harrison DA, Watkinson PJ, Doidge JC. New-onset atrial fibrillation in intensive care: epidemiology and outcomes. Eur Heart J Acute Cardiovasc Care 2022; 11(8):620-8. http://dx.doi.org/10.1093/ehjacc/zuac080 Bedford JP, Johnson A, Redfern O, Gerry S, Doidge J, Harrison D, Rajappan K, Rowan K, Young JD, Mouncey P, Watkinson PJ. Comparative effectiveness of common treatments for new-onset atrial fibrillation within the ICU: Accounting for physiological status. J Crit Care 2022; 67:149-56. http://dx.doi.org/10.1016/j.jcrc.2021.11.005 Ferrando-Vivas P, Shankar-Hari M, Thomas K, Doidge JC, Caskey FJ, Forni L, Harris S, Ostermann M, Gornik I, Holman N, Lone N, Young B, Jenkins D, Webb S, Nolan JP, Soar J, Rowan KM, Harrison DA. Improving risk prediction model quality in the critically ill: data linkage study. Health Soc Care Deliv Res 2022; 10(39). http://dx.doi.org/10.3310/eqab4594 Shankar-Hari M, Rubenfeld GD, Ferrando-Vivas P, Harrison DA, Rowan K. Development, Validation, and Clinical Utility Assessment of a Prognostic Score for 1-Year Unplanned Rehospitalization or Death of Adult Sepsis Survivors. JAMA Netw Open 2020; 3(9):e2013580. http://dx.doi.org/10.1001/jamanetworkopen.2020.13580 Shankar-Hari M, Harrison DA, Ferrando-Vivas P, Rubenfeld GD, Rowan K. Risk Factors at Index Hospitalization Associated With Longer-term Mortality in Adult Sepsis Survivors. JAMA Netw Open 2019; 2(5):e194900. http://dx.doi.org/10.1001/jamanetworkopen.2019.4900
