Using SPC to analyze Patient recorded vital sign data
Steven Ringer, MD Ph.D., Director of Newborn Services, Brigham and Women's Hospital, Harvard Medical School, Boston MA
Kimberly R. Zimmerman, R.Ph., Independent Consultant Montgomery, AL
Warren Beatty, Ph.D., Associate Professor of Management, University of South Alabama
Steven M. Zimmerman, Ph.D., Professor of Quality and System Management, University of South Alabama
Abstract
Patient vital sign are available from:
The issue is the value of patient recorded vital sign data relative to computer data and nurse recorded data for clinical decision making. The study uses statistical process control (SPC) methods to study the behavior of data recorded by a stroke patient to see what conclusions may be drawn from that data.
Introduction
Vital sign data are available from: 1) computer recorded data; 2) nurse (professional caregiver) recorded data; and 3) patients recorded data. Currently, nurse recorded data are considered the best and most reliable. A case study is used to examine the value of patient recorded data.
Patients want to know about their own bodies and want to understand what is happening to them when they become sick. Many patient with non-medical backgrounds have no idea what the drugs they are told to take does to their bodies. Using vital sign monitors that can be obtained from many pharmacists, patients can measure their own vital signs. This paper focuses on one method of analyzing patient data using statistical process control (SPC) and the value of the data in helping caregivers work with patients.
Self Testing
Patients have the right and the ability to test themselves. Once they have their data, how should we used the data? Can caregivers rely on the data or should they ignore anything that is not taken and recorded by a professional caregiver?
Patients need help in learning how to use the monitors they purchase and then how to react to the data they collect. In particular, they must lean about variation so that they do not pick up the phone and call their physician every time their blood pressure goes up four of five points. In addition the patient must learn how to keep records, so that the results are of use to the caregiver.
Instrument Accuracy, Consistency, and Training
Accuracy is how near to the correct value an instrument outputs. Consistency is how well an instrument replicates its results. Both are important in an home measuring device. We feel that consistency is more important than accuracy, because a primary objective of clinical monitoring is to identify changes. Both accuracy and consistency are important because the home user wants to know:
Generally, patients do not have training in using biomedical measuring equipment, however, a determined patient can usually learn to use any of the home monitors available. There is a danger that home user will react to the readings they are obtaining, i.e. a placebo effect. In our experience, the home user is more concerned to getting the reading and the placebo effect is at a minimum because of lack of understanding of what the reading means and what it should be.
The Sunbeam Model 7621, Digital Blood Pressure Monitor was selected by our patient due to ease of use. Our patient then spend hours reading and trying out the instrument before beginning to record her results.
The patient used three agents: Tenoretic (50mg), Procardia XL (30mg), Calan SR (180mg), and Calan SR (240mg). The exact dates on which each agent was started was recorded.
Data Analysis
The patient performed an statistical process control (SPC) analysis for each of the vital sign measurement sets taken: pulse, systolic, and diastolic blood pressure.. Figure 1, 2, and 3 illustrates the SPC charts. Control charts may be used to identify when a change occurs or to study the effect of changes. In our analysis we are interested in studying the effect of changes in agents, therefore, control limits are recalculated each time the agent is changed. In order to obtain a good base line, the patient searched her old medical records and obtained prior to stock data on her blood pressure. The pulse data prior to her stock was not available, therefore, this chart starts later than the two blood pressure charts
Figure 1 Pulse rate
Figure 2 Systolic blood pressure
Figure 3 Diastolic blood pressure
All three figures include two process control charts. The top chart is for the individuals process control chart and the bottom chart is for variation, a moving range process control. There are four lines in the individual process control chart are:
top dashed line is the upper control limit for individuals
middle solid line is the average of the individuals
bottom dashed line is the lower control limit for individuals
solid (up and down) line connects each individual observation
The lower chart in figure 1 uses the range as its measure of variation. We usually use the standard deviation (SD), however, for subgroup sizes of two the range and standard deviation results are identical. The lines on this chart are:
top dashed line is the upper control limit for ranges
middle solid line is the average of the ranges
bottom dashed line is the lower control limit for ranges
solid (up and down) line connects each range
Effect of Agents
The charts indicate that the diastolic blood pressure had the smallest reaction to the agents, which is consistent with clinical expectations. The first medication given to the patient to control her blood pressure was Tenoretic. The patient experienced side effects to the medication and requested a change. The second agent (Procardia) caused feelings of elevated pulse rate which may be confirmed by examination of the charts. She was the most satisfied with the third agent (Calan SR 180), but felt it needed additional adjustment. The last change was to increase its dosage from 180mg to 240mg. The patient seemed to react slowly to the increase in dosage while its overall effect seemed to be to reduce the amount of variation. In industry a reduction of variation is usually considered favorable.
Technical Issue
Sampling frequency for our patient recorded study varied from once a day during the time period when the search for an agent was taking place to once a month, or whenever the patient is feeling poorly! This sampling plan is not desirable, but it is what it is. Data analysis must take into account the reality of the situation. One common biomedical data problem is autocorrelation. Autocorrelation is a measure of the relationship of serial observation. Figure 4 illustrates the autocorrelation for our data set. An analysis of the relationship of serial points yielded a correlation of 0.0451. Computer data collection may have correlation values over 0.95. The patient recorded data collection scheme eliminates the autocorrelation problem.
Figure 4 Autocorrelation Analysis
Caregivers Use
Most caregivers are not trained in data collection, recording, and analysis. Current clinical practice does not produce quality data. Current practice is to observe and record vital signs once per hour, while vital signs can change between heart beats. Many caregivers discount the value of clinical recorded data. The lack of confidence in clinical data may be expected to effect caregivers attitudes towards patient recorded data.
Conclusions
Self testing is here. Patients have the equipment available and have the option to test themselves. The pharmacist has a role in patient monitored data. Patients need the pharmacist help in understanding and using the data they have collected on themselves. Our case study, the heart patient has demonstrated that there are graphing methods that will enhance the meaning of the data collected and help patients, pharmacists, and physicians understand better the data that are collected.
The monitors purchased by the patient should be studied by the pharmacist so that they can help individuals select the best device for their situation and so that the data collected is of maximum value.
The encouragement of self testing is a judgement call based on the patient, their skills, and their clinical condition. It is important that the patient be able to learn to use the device correctly so that the data collected are of value.
The statistical process control methods selected by the patient does indicate changes in the body's performance as a function of the agent used. It is a simple technique to understand, but does require calculations and graphing which may be performed through the use a microcomputers.
The danger of patient reaction or over-reaction to the data collected is present and requires that the best available methods be used to measure, record, analyze the data, and review the results. The pharmacists role in guiding the patient is critical. Acceptance and use of patient self tested data are growing and we expect it will continue to grow.
References
