CUSUM-type methods aimed at catching deviating performances

Join Professor Elena Kulinskaya for this seminar discussing her research on cumulative sum control chart -type methods aimed at catching deviating performance.

  • Thu 8 Feb 18

    13:15 - 14:45

  • Colchester Campus

    Room LTB B

  • Event speaker

    Professor Elena Kulinskaya, UEA, Norwich

  • Event type

    Lectures, talks and seminars

  • Event organiser

    Mathematical Sciences, Department of

  • Contact details

    Dr Andrew Harrison

Professor Elena Kulinskaya is a Professor in Statistics at the University of East Anglia. Her research interests include Foundations of statistics, Statistical methods for discrete and skewed data, Asymptotic methods, Applied statistics, with applications in ecology, medicine and actuarial sciences, and Meta-analysis and research synthesis.

Alexander Begun1, Elena Kulinskaya1 and Alex MacGregor2

University of East Anglia

1School of Computing Sciences and 2Norwich Medical School

Continuous monitoring of healthcare, and increasingly, social care across various providers such as consultants, units, NHS Trusts, GP surgeries, nursing homes etc. is an important task of the healthcare regulator, such as the Care Quality Commission (QCC) in the UK. Additionally, a number of professional bodies and registers take on the same function for their clinical discipline. For instance, in regards to joint replacement, surgeon- and unit-level outcomes are compiled by National Joint Registry for England and Wales (NJR).

The methods of continuous monitoring of production quality have been initially developed and employed in quality control in industry. One of the most popular methods is the cumulative sum (CUSUM) chart, a graphical method based on sequential monitoring of cumulative performance over time. This method allows timely identification of a deterioration in performance. A number of CUSUM-based systems are used in surveillance of the healthcare quality by QCC and by Dr Foster unit at Imperial College. In this talk we expand the CUSUM methodology so that it can be applied to the time-to-event data from the institutions with varying casemix. The CUSUM increments are based on the scores from a Weibull survival model with a shared frailty term accounting for the same care provider, and the control limits are found by bootstrap. The methods are applied to monitoring the performance of hip prostheses based on the NJR data.

This work was funded by the ESRC, grant number ES/L011859/1.