Abstract
Monitoring medical processes gained importance and researchers attempted to reduce death rates by quick detection mortality rate of surgical outcomes in recent years. The patient time until death (survival time) depends on risk factor of each patient, which reflects the patients' health condition prior to surgery. Ignoring differences in risk factors among specific patients, risk adjusted control charts could be considered as a corrective tool to minimize false alarms related to inhomogeneity in patients' health condition. A number of risk adjusted charting procedures have been developed on both phase I & II monitoring of aforementioned outcomes. This chapter will review both models and focus on phase-I risk-adjustment models in medical setting with a particular emphasis on monitoring for surgical context and describe each method's unique properties.
TopBackground: Risk-Adjusted Charts
Unlike monitoring of processes in the manufacturing industry, the monitoring of surgical performances is different and presented a unique problem because the patients are not homogeneous and hence the necessity of risk-adjustment. The Parsonnet scoring system (Parsonnet, Dean, & Bernstein, 1989) is widely used for estimating the risk of death of a patient who undergoes a cardiac operation. A competing scoring system was developed by Roques et al. (1999). Based on a patient’s gender, age, morbid obesity, blood pressure, etc. integer scores are given and the total score is called the Parsonnet score. A Parsonnet score is an integer that ranges from 0 to 100; hence it is a discrete random variable. A small Parsonnet score represents a small risk and a high score represents a high risk.
Key Terms in this Chapter
Dummy Variable: One that takes the value “0” or “1” to indicate the absence or presence of some categorical effect that may be expected to shift the outcome.
Risk-Adjusted Control Chart: A monitoring chart that used to level the playing field regarding the reporting of patient outcomes by adjusting for the differences in risk among specific patients.
Parsonnet Score: A competing integer score based on a patient’s gender, age, morbid obesity, blood pressure, etc. that estimates his/her risk of death.
Phase-I Monitoring: Monitoring historical data to remove out-of-control points from it and produce ‘cleaned’ data-set.
Logistic Regression Model: A regression model where the dependent variable is categorical and can take only two values, “0” and “1” (the case of a binary dependent variable).
Phase-II Monitoring: Monitoring: Monitoring future data based on control limits which derived from the data of in control process (cleaned data-set).
Accelerated Failure Time (AFT) Model: A parametric model that assumes the effect of a covariate is to accelerate or decelerate the life course of a disease by some constant.