Patient Safety Helping Leaders Blink Correctly: Part II Understanding variation in data can help leaders make appropriate decisions. So, what are the limitations of the Editor’s Note: Part I of this two-part monthly budget numbers are positive column can be found in the May/June and another way when they are nega- static approach to understanding variation? Aggregated data presented 2010 issue of Healthcare Executive. tive, or when patient satisfaction scores increase from one month to in tabular formats or with summary statistics will never allow you to In healthcare we tend to make quick the next? The simple explanation is decisions (“blink”) by finding pat- that most healthcare professionals understand the variation in the data or to determine the impact of quality terns in data based on narrow slices are not given sufficient training in of experience (“thin slicing”), a con- statistical methods to “extract improvement efforts. Aggregated data can only lead to blinking quickly and cept Malcolm Gladwell details in his knowledge that may be locked up book Blink: The Power of Thinking inside the data,” as Don Wheeler often leads to a decision based on judgment (see Part I for the distinc- Without Thinking (Little, Brown, illustrates in his book Understanding 2005). This approach is usually prob- Variation: The Key to Managing Chaos tion between using data for improve- ment and data for judgment). lematic because we see trends where (SPC Press, 1993). They are taught no trends exist, conclude that the to apply static rather than dynamic To truly understand the variation in data have shifted when in fact they statistical approaches to understand- display nothing more than random ing variation. your data, a dynamic approach that uses statistical process control methods variation, or spend an inordinate amount of time trying to explain a A static approach to understanding to analyze variation in data over time is most appropriate. The primary sta- single high or low data point while variation is hallmarked by the follow- ignoring the rest of the data. ing activities: tistical tools for understanding varia- tion in this context are run and While Part I of this article intro- • Presenting data in tabular or control charts. This article will focus only on control charts. duced the first two skills healthcare aggregated formats and display- leaders need to make appropriate ing this data in bar or pie charts Time is always shown on the horizon- decisions (understanding the messi- ness of improving healthcare and • Using measures of central ten- tal axis of a control chart; the measure of interest is plotted on the vertical determining why they are measuring dency (the mean, median and in the first place), this article will dis- mode) and measures of disper- axis; and the center line (CL) is the mean of the data points (see chart on cuss the remaining two skills: under- sion (the range, standard devia- standing and depicting variation and tion, variance, coefficient of page 73). The control chart also has variation, etc.) to summarize the the added advantage of having esti- translating data into information. variation in the data mates of the variation in the data. As the sample control chart on page 73 Understanding and Depicting Variation • Comparing two data points to indicates, the variation is captured by determine if they are statistically the upper and lower control limits Variation exists in all that we do, so why do we react one way when the different (UCL and LCL). Statistical rules are 72 Reprinted from Healthcare Executive JULY/AUG 2010 ache.org
then applied to the data to determine time and compute the average over create information, but data are not if the variation is common cause (i.e., time and the variation between read- information in and of themselves. random) or special cause (i.e., statisti- ings. Clinicians would never use a Charles Austin provides a clear cally different). static approach to monitor an ICU description of the distinction patient, however, because it does between these two concepts in his Consider this clinical example that not provide sufficient understanding book Information Systems for Hospital demonstrates the distinction between of variation. Instead, clinicians rely Administration (Health static and dynamic approaches to on telemetry data to understand vari- Administration Press, 1983): “Data understanding variation: monitoring ation in the patient’s key physiologi- refers to the raw facts and figures a patient’s vital signs in the ICU. cal measures (e.g., heartbeat, that are collected as part of the nor- Using a static approach, we might respiration, blood pressure or oxygen) mal functioning of the hospital. obtain the ICU patient’s blood pres- over time so that appropriate real- Information, on the other hand, is sure at two points in time (at time of time interventions can be made—a defined as data that have been pro- admission and at discharge) and then dynamic approach. cessed and analyzed in a formal, compare the two readings to deter- intelligent way so that the results are directly useful to those involved in mine if they are statistically different. Translating Data Into Information Or, we might take several blood pres- All too often we confuse data and the operation and management of the hospital.” sure readings at various points in information. Data can be used to Elements of a Control Chart 50.0 An indication of a special cause UCL=44.855 45.0 (Upper Control A Limit) 40.0 Measure—Number of Complaints B 35.0 C CL=29.250 30.0 C X (Mean) 25.0 B 20.0 A 15.0 LCL=13.645 (Lower Control 10.0 Limit) 5.0 Jan01 Mar01 May01 July01 Sept01 Nov01 Jan02 Mar02 May02 July02 Sept02 Nov02 Time—Month Source: Lloyd, Robert. Quality Health Care (Jones and Bartlett Publishers Inc., 2004); p. 275. 73 Reprinted from Healthcare Executive JULY/AUG 2010 ache.org
Patient Safety Translating data into information data collection, respondent and Step 5: Interpretation of the Results . This is the step when data occurs only as a result of a deliberate data collector bias, and data collec- process that involves the following tion methods are all critical ele- begins to transform into informa- tion and a point at which it is easy steps, which are also outlined in the ments of this step. chart on this page. to blink too quickly and make a decision based on incomplete Step 4: Data Analysis and Step 1: Theoretical Concepts . All Output . Decide who has access to a information. Interpreting results seeks to answer a very simple ques- scientific inquiry begins with theoret- statistical package to tabulate and ical concepts (ideas and hypotheses) analyze data and to produce graphi- tion: why ? This is the point at which the data and the theory and making predictions. The real test cal displays of the data. Also, deter- of any theory or hypothesis lies with mine which type of statistical should be compared. the empirical evidence that can be analysis will be conducted. Will assembled to test the validity and you merely calculate the average, Do the analytic results support the proposed theories? Are the data con- reliability of the idea. minimums and maximums, and standard deviations for the data sistent with what we have seen in the past? If not, are the data correct, or is Step 2: Select and Define (static approach), or will you ana- Measures . This is a critical step. lyze the variation in the data using the theory wrong? Do the data reflect common or special causes of varia- Define a limited set of measures (usu- run or control charts (dynamic ally between five and seven) for each approach)? If you are focusing on tion? This is also the point at which data for improvement (not judg- previous research and data play key improvement project. Do not blink too quickly in this step by selecting ment or research) then use the roles. Are your results consistent with dynamic approach. what others have found? measures that are convenient or that you always have tracked. The Process of Turning Data Into Information Develop a clear operational defini- tion for each measure (e.g., what is a Step 1 medication error? or when does sur- gery start?) so that data appropri- Theoretical Concepts ately represent the concept being (ideas & hypotheses) measured. There are no universally Step 6 Step 2 correct operational definitions, so achieving consensus and consistency Information for Select & Define is most important. Decision Making Measures Theory Step 3: Data Collection . This step requires considerable planning and and Step 5 Step 3 execution. Identifying what data Prediction will be collected is determined by Interpretation Data your defined measures. But you of the Results Collection (asking why?) (plans & methods) also must consider how the data will be collected and by whom . Also, determine where the data will Data Analysis be stored (e.g., in a database, in the and Output chart or at the nursing station). Issues such as stratification, sam- Step 4 pling, the role of pilot tests, the Source: Lloyd, Robert. Quality Health Care (Jones and Bartlett Publishers Inc., 2004); p. 153. duration and frequency of 74 Reprinted from Healthcare Executive JULY/AUG 2010 ache.org
Recommend
More recommend
