Introduction to Control Charts By Farrokh Alemi Ph

Introduction to Control Charts By Farrokh Alemi Ph. D. Sandy Amin HEALTH INFORMATICS PROGRAM HI. GMU. EDU

PURPOSE Provide an overview of control chart applications for common healthcare data. We assume: • User has a basic understanding of process variation • User has knowledge of simple statistics (i. e. measures of central tendency). HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

WHAT IS A CONTROL CHART? A graphical display of data over time that can differentiate common cause variation from special cause variation HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

COMPONENTS OF CONTROL CHART UCL Observations LCL HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

INTERPRETATION OF CONTROL CHARTS Points between control limits are due to random chance variation HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

SUGGESTED NUMBER OF DATA POINTS • More data points means more delay • Fewer data points means less precision, wider limits • A tradeoff needs to be made between more delay and less precision HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

FREEZING & REVISING CONTROL LIMITS HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

SELECTING APPROPRIATE CHART • • Xm. R X-bar Tukey Time-in-between P-chart Risk adjusted X-bar chart HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

EXAMPLES OF MEASURES Continuous variables Length of stay Average length of stay Average of a specific patient population Process turn around time Staff salaries Severity of medication errors Individual patient’s weights, blood sugars, cholesterol levels, temperatures, or blood pressures over time Patient Satisfaction Average Scores Infectious waste poundage generated Electrical usage Wait times Accounts receivable balances Time in restraints Time before hanging up the phone SF – 36 scores HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

EXAMPLES OF MEASURES Rates and discrete events Number of employee accidents Number of patient falls Nosocomial infection rates Percent of patients in restraints Medication error rate Adverse event rate C-Section rates Number of dietary tray errors Numbers of delinquent medical records Percent of patients with insurance Percent of patients who rated the facility as excellent Telephone abandonment rates Pressure ulcer rates Employee injuries rates Percent of records that contains appropriate documentation HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

WHICH CHART IS RIGHT? If continuous variable If one data point per time period If outliers likely: Tukey chart If outliers not likely: Xm. R chart If multiple data points per time period: Xbar chart If discrete event If event is rare: Time-in-between chart If event is not rare: P-chart HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY

RISK ADJUSTMENT When case mix changes over time, use risk adjusted control charts Instead of comparing to historical patterns, new observations are compared to expectations Risk adjusted control charts are calculated by applying the formulas for control limits to the difference of observed and expected values HEALTH INFORMATICS PROGRAM GEORGE MASON UNIVERSITY