Run Charts, Control Charts, and SPC: Basic Power Tools for Quality Improvement James I. Hagadorn MD MSc Division of Neonatology, Department of Pediatrics University of Connecticut School of Medicine Connecticut Children s Medical Center Dr. Hagadorn has no financial relationships to disclose or conflicts of interest to resolve. Objectives. Describe the role of analysis in the context of the Model for Improvement Describe how Run Charts and Control Charts can be used to identify nonrandom fluctuations in a process Describe how Control Charts can be used to improve the safety of oximeter alarms and oxygen therapy in your NICU The Model for Improvement Three fundamental questions 1. What are we trying to accomplish? 2. How will we know whether a change is an improvement? 3. What changes can we make that will result in improvement? + PDSA Cycles www.health.ny.gov Applying the Model for Improvement Five critical components are needed to apply the Model for Improvement 1. An improvement project : process-focused: save time, money or improve quality of a service or system outcome-focused: improve health status, behavior, or knowledge 2. s to be tested that are grounded in science 3. A family of measures, including impact and process measures. A few of the measures should have the potential of being tracked at least monthly. 4. People who will test the interventions and collect the data 5., usually 6 to 16 months, to allow for multiple tests of interventions The NY State DPH Left Out One Thing Your team needs a nerd Your team needs a numbers person www.health.ny.gov October 2, 2015 Page 1
SPC: Statistical Process Control A method of quality control which uses statistical methods to monitor and control a process, ensuring that it operates at its full potential with a minimum of waste Can be applied to any process where the desired outcome can be measured Key tools used in SPC include control charts, a focus on continuous improvement, and the design of interventions Deming Shewhart Run Charts and Control Charts Allow you to Identify non-random changes in data collected over time Evaluate the effect of a QI intervention upon your outcome measures Communicate these results to others visually en.wikipedia.org Random and Nonrandom Variation Run = 1 or more consecutive data points on the same side of the central line Don t include data points that fall on the median Central line = Median 11 runs in these data Random Variation Affects all process outcomes Source is intrinsic to the process design Is stable Is predictable Nonrandom Variation May affect some or all process outcomes Source is external to the process design Is unstable Is unpredictable 1. A run of 6 consecutive points on the same side of the median 2. A trend of 5 consecutive points all increasing or decreasing October 2, 2015 Page 2
3. Too many or too few runs 4. An extremely high or low data point 1. A run of 6 consecutive points on the same side of the median 2. A trend of 5 consecutive points all increasing or decreasing 3. Too many or too few runs 4. An extremely high or low data point? 1 1 J FMAM J J A S OND J FMAM J J A S OND J FMAM J J A S OND J A S O N D J F M A M J J A S O N D October 2, 2015 Page 3
1 New group of hospitals begin contributing data J FMAM J J A S OND J FMAM J J A S OND J FMAM J J A S OND Control Charts Anatomy of a Control Chart AKA Shewhart Chart More powerful than Run Chart Uses central tendency Also uses dispersal of the data ( sigma limits ) More sensitive; can identify non-random variation sooner Characterize a process as stable or unstable Upper Mean Lower Rules for Identifying Nonrandom Data in Control Charts Anatomy of a Control Chart Upper Mean Lower October 2, 2015 Page 4
X Chart: Admissions/month Monitoring Vs Improving Your Process 1 New group of hospitals begin contributing data J A S O N D J F M A M J J A S O N D Monitoring Crucial process or outcome Data collection ongoing Random variation characterized Nonrandom variation identified thru data rules Expend resources to identify and correct cause of nonrandom variation Improving Identify process or outcome for improvement Start collecting baseline data Nonrandom variation introduced = your intervention Follow up data collected Types of Control Charts Depends on type of data you re measuring Continuous: length of stay, discharge weight Yes/No: any ROP Proportion: proportion of oximeter alarms that are nonactionable Count data: Nonactionable alarms per hour https://www.spcfo rexcel.com/knowle dge/control chartbasics/selectingright control chart X charts Reducing Proportion of SpO2 values >97% c charts Reducing Non Actionable Oximeter Alarms Monitoring Number of Episodes SpO2 <80% of Duration >60s October 2, 2015 Page 5
Summary Use the Model for Improvement Make sure your team has a numbers person Select outcomes to measure Follow with Run Charts or Control Charts Establish baseline Make your interventions Follow outcomes, balancing measures Repeat References Berardinelli C. A Guide to Control Charts. Available at: http://www.isixsigma.com/tools-templates/control-charts/a-guide-tocontrol-charts/ Buttrey SE. An Excel Add-In for Statistical Process Control Charts. Journal of Statistical Software Vol. 30, Issue 13, Jun 2009. Available at www.jstatsoft.org/v30/i13 Wheeler DJ. Understanding Variation: The Key to Managing Chaos. 1993 SPC Press, Knoxville TN Wheeler DJ. Understanding Statistical Process Control. 1992 SPC Press, Knoxville TN Proportion of with SpO2 >97% Thank You! Baseline Data #1 Proportion of with SpO2 >97% Proportion of with SpO2 >97% Following #1 Following #1 s recalculated #2 #1 October 2, 2015 Page 6
Proportion of with SpO2 >97% Proportion of with SpO2 >97% Following #2 #2 Following #2 s recalculated #2 Non Actionable Desaturation Alarms/Nurse Shift Non Actionable Desaturation Alarms/Nurse Shift Non Actionable Desaturation Alarms Baseline Data Non Actionable Desaturation Alarms After Nurse Shift Nurse Shift Episodes/ of SpO2 <80% x 60s Episodes/ of SpO2 <80% x 60s 2nd >97% Nonactionable 1 st >97% Alarms Baseline Act Week October 2, 2015 Page 7