Oral Presentation Victorian Integrated Cancer Service Conference 2015

Application of risk-adjusted sequential probability ratio test to analyse institutional positive surgical margin rates in radical prostatectomy (#50)

Varuni Kumari 1 , Arul Earnest 2 , Damien Bolton 3 , Jeremy Millar 2 4 , Sue Evans 2
  1. School of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria, Australia
  2. School of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria, Australia
  3. Department of Surgery, University of Melbourne, Austin Hospital, Melbourne, Victoria, Australia
  4. Alfred Health Radiation Oncology, Alfred Health, Melbourne, Australia, Alfred Health, Mwlbourne, Victoria, Australia

Introduction:

Sequential process control charts provide a longitudinal analysis of sequential data in order to evaluate variation in the process. The Victorian prostate cancer registry enables the use of these process control charts to monitor surgical outcome metrics. The aim of this study was to investigate the use of one of these process control charts, the risk-adjusted sequential probability ratio test (RASPRT), in monitoring the cumulative occurrence of positive surgical margins in radical prostatectomy at one Victorian hospital.

Methods:

We reviewed pathology data from the Victorian Prostate Cancer Registry prospectively captured between 2009 and 2013 from one hospital. We ran two different models: one a risk-adjusted sequential probability ratio test that was designed to detect a doubling of odds ratio of PSM with the false positive and false negative error rate set at 1%, and secondly a NCCN risk-adjusted test. When the accumulating risk-adjusted SPRT statistic crosses upper threshold levels we conclude that the odds for a PSM are twice the risk-adjusted prediction.

Results:

The median age of the 277 of patients was 60.6 years (IQR: 61, 66). The accumulating RASPRT statistic identified a general upward slope until January 2012, although this did not breach the upper threshold levels. There were 3 instances where the chart crossed the lower threshold level in the unadjusted model, and only 2 when adjusted for NCCN.

Conclusions: .

The RASPRT detects higher/lower rates of PSM by using accumulating evidence over time to allow rapid recognition of variation. The use of patient risk-adjusted models will help to tease out possible confounders and noise in the data. This in turn allows timely interventions to improve and then to continually monitor “process control”.