A group of surgeons performing an operation in a hospital

One of our key aims is to research how best to use administrative databases to inform how the NHS is performing both overall but also at the level of hospital, surgeon and in primary care. Challenges that we tackle include how to best process these complex data sets, indicator development, risk adjustment, statistical monitoring and feeding back the results to NHS stakeholders.

Projects:

Completed projects

Risk adjustment for administrative data using comorbidity measures and machine learning

Summary

To compare hospitals fairly, we need to measure and adjust for patient factors such as age and comorbidities using the most appropriate statistical models. These increasingly use administrative information, but this relies on clinical coding to record diagnosis and operation details, which is known to vary between hospitals. In this project we built statistical models for several key indicators of quality of care using English NHS administrative data, with a focus on comorbidities.

Aims

  • To derive the most suitable comorbidity index for the NHS
  • To compare logistic regression with machine learning methods for risk adjustment for mortality, readmission and unplanned reoperation.

Outputs

  • Bottle A, Sanders RD, Mozid A, Aylin Pet al., 2013, Provider profiling models for acute coronary syndrome mortality using administrative data, INTERNATIONAL JOURNAL OF CARDIOLOGY, Vol: 168, Pages: 338-343, ISSN: 0167-5273 PubMed
  • Sharabiani MTA, Aylin P, Bottle A, 2012, Systematic Review of Comorbidity Indices for Administrative Data, Medical Care, Vol: 50, Pages: 1109-1118, ISSN: 0025-7079 PubMed
  • Gaudoin R, Montana G, Jones S, Aylin P, Bottle A. Classifier calibration using splined empirical probabilities in clinical risk prediction. Health Care Manag Sci 2015;18(2):156-65. PubMed
  • Bottle A, Gaudoin R, Goudie R, Jones S, Aylin P. Can valid and practical risk-prediction or casemix adjustment models, including adjustment for comorbidity, be generated from English hospital administrative data (Hospital Episode Statistics)? A national observational study. Health Serv Delivery Res 2014;2(40). PubMed
  • Bottle A, Aylin P, Bell D. Effect of the readmission primary diagnosis and time interval in heart failure patients: analysis of English administrative data. Eur J Heart Fail 2014;16(8):846-53. PubMed
  • Bottle A, Aylin P. Comorbidity scores for administrative data benefited from adaptation to local coding and diagnostic practices. J Clin Epidemiol 2011;64(12):1426-33. PubMed
  • Bottle A, Sanders RD, Mozid A, Aylin P. Provider profiling models for acute coronary syndrome mortality using administrative data. Int J Cardiol 2013;168(1):338-43. PubMed
Volume-outcome relations for colorectal, upper GI and urological surgery

With our longstanding surgical collaborators at Imperial NHS Trust and other hospitals, we have assessed the extent to which the number of operations that a surgeon or hospital performs each year is associated with the outcomes for their patients. This work informs the growing calls for the NHS and many other healthcare systems in the developed world to centralise services.

Department of Primary Care and Public Health

Privacy notice

The Dr Foster Unit at Imperial College London uses your health information for a number of purposes. The Dr Foster Unit GDPR Privacy Notice (PDF) provides a summary of how we use your information.

by Paul Aylin and Alex Bottle

Statistical Methods for Healthcare Performance Monitoring
A book for those measuring, monitoring, comparing and improving healthcare performance