HCA375: Continuous Quality Monitoring & Accreditation-Adverse Event Reporting

119 5 Data Resources Fuse/Thinkstock Learning Objectives After reading this chapter, you should be able to do the following:

• Illustrate the importance of data in quality improvement.

• Compare the different types of data available for quality improvement.

• Assess internal sources of data for quality improvement activities.

• Select appropriate external sources of data for comparative or benchmarking purposes. fin81226_05_c05_119-148.indd 119 10/30/14 7:23 PM What is Special Education? 1 iStockphoto/Thinkstock Pre-Test 1. You can use the terms disability and handicap interchangeably. T/F 2. The history of special education began in Europe. T/F 3. The first American legislation that protected students with disabilities was passed in the 1950s. T/F 4. All students with disabilities should be educated in special education classrooms. T/F 5. Special education law is constantly reinterpreted. T/F Answers can be found at the end of the chapter. Introduction Introduction Healthcare data can be used to drive improvements in organizations. In fact, data plays a cen - tral role in the quality improvement process. It can help pinpoint areas where organizations need to improve and then help demonstrate the results of those quality improvement efforts.

Let’s look at an example of how data can show how organizations are performing and where improvements are needed—in this case, in the country’s community health centers. Since 2007, the U.S. Health Resources and Services Administration (HRSA) has made public updated data on the performance of the almost 1,200 community health centers across America. When the government released data for calendar year 2012, it showed that some of these centers, which serve millions of mostly poor people, fell short on key measures—with many of the centers failing to take steps such as vaccinate children and help diabetics control their blood sugar levels.

That was troubling news because more than 21 million people rely on these federally funded community health centers for their primary medical care (National Association of Commu - nity Health Centers, 2014). This role will only grow as the Affordable Care Act (ACA) has sig - nificantly expanded the health center program to meet the increased demand for healthcare that is expected as millions of the uninsured gain health coverage beginning in 2014.

Therefore, evidence of the quality of care that these centers provide is of major interest. The federal government provides subsidies to support these centers and in 2007 began tracking quality indicators for each center. For example, vaccination rates, blood pressure control, and prenatal care are all quality measures that the federal government monitors at these clinics.

With public release of the data by HRSA, the Kaiser Commission on Medicaid and the Unin - sured (2013) partnered with George Washington University to analyze health center perfor - mance and released a 14-page report. While most research shows the centers have a high per - formance when it comes to healthcare standards, this data revealed gaps, and the numbers showed wide variation in the quality of care delivered by these clinics. For instance, based on the latest data for calendar year 2012, community health centers in New Hampshire were the most likely to keep diabetics’ blood sugar under control, while Vermont’s health centers had the best child immunization rates, and Maine’s centers had the highest percent of pregnant women getting early prenatal care (Galewitz, 2012).

Community health centers in New England—where rates of insurance coverage are higher, making it more likely people will seek care when they need it—generally performed better than centers in the South and West. Mississippi health centers had some of the highest pro - portion of low birth-weight babies, which places those infants at risk for certain health prob - lems. Wyoming and Oregon had some of the lowest child immunization rates. While there were differences in regions and states, even centers in the same city often performed differ - ently (Galewitz, 2012).

One of the key lessons from the measures collected from these community health centers is that data plays a key role in quality reporting (i.e., whether institutions achieve certain qual - ity indicators) and also in the process of quality improvement. Measuring healthcare perfor - mance can result in real improvements (National Committee for Quality Assurance, 2014). fin81226_05_c05_119-148.indd 120 10/30/14 7:23 PM Section 5.1 The Role of Data in Quality Improvement The National Committee for Quality Assurance (NCQA) measures performance by health plans on a variety of factors aimed at improving care and keeping patients healthy.

It has been estimated that quality improvements by health plans in providing beta-blocker treatment, cholesterol management, blood pressure control, and blood sugar control for dia - betics have saved thousands of lives (NCQA, 2014). Improvements in these healthcare prac - tices have real benefits for individuals. In practical terms, smokers who are more consistently advised to quit are more likely to do so. Heart attack victims are likely to live longer if their blood pressure and cholesterol are controlled. Immunized children don’t miss as many school days because of illness and grow to be healthier adults. Therefore, it is essential for students and practitioners of quality improvement to understand healthcare data and how it can be used to drive change in healthcare organizations.

5.1 The Role of Data in Quality Improvement Data is an essential component of any quality improvement process. William Thomson (Lord Kel - vin), noted in the 19th century that “when you can - not measure it . . . your knowledge is of a meager and unsatisfactory kind . . .” (Thomson, 1891, p. 80). The data-driven approach is integral to the major qual - ity improvement methods used in the United States, including the Plan-Do-Study-Act (PDSA) model on which others are based. A variety of measurement approaches are also currently used, but regardless of the specific approach to quality improvement, data assumes a central role in defining the project work.

Increasingly, the performance and reimbursement of health systems in the United States is based on quality-oriented goals, whereas in the past health systems focused on providing a high volume of healthcare services (Panzer et al., 2013). In the past, healthcare providers were reimbursed based almost solely on the quantity of services they pro - vided, giving hospitals and physicians an incentive to order diagnostic tests and services for patients.

Improving quality relies increasingly on the use of data; therefore, it is important to understand the roles played by data in any health system quality improvement enterprise.

First, sufficient data must exist to identify the problem that is the focus of an improvement activity. Having identified that an opportunity for improvement exists, additional data are typically necessary to identify potential target(s) in the care process that can be improved.

Organizational leaders will look for the root cause of a problem by asking why the problem is occurring. For instance, why are infection rates on the rise or patient falls increasing? It’s Science and Society/SuperStock Lord Kelvin stressed the value of measuring data a century before Six Sigma was created. fin81226_05_c05_119-148.indd 121 10/30/14 7:23 PM Section 5.1 The Role of Data in Quality Improvement usually not a simple answer. Looking for a root cause means delving deeply into what is behind a particular problem—peeling the onion, so to speak, to determine the real cause.

Subsequently, leaders must use data about specific care processes to plan interventions that will generate measurable changes. Once these interventions are put into place, health system leaders use data to measure the impact of the intervention. Finally, data serve a key monitor - ing role after a test of change becomes incorporated into routine processes and procedures, so that improvements in an organization’s processes will be sustained.

Use Data to Identify the Problem Consider the variety of roles that data can play in each phase of one of the best-known qual - ity improvement models, Dr. William Edwards Deming’s PDSA cycle, as shown in Table 5.1.

The PDSA model is the original quality improvement model from which later models evolved.

As discussed earlier in the book, you will sometimes see this model referred to as Plan-Do- Check-Act (PDCA), which was the original language. However, as the model evolved, Deming amended his description of PDCA to emphasize the importance of not just checking, but using or studying the knowledge to better understand the process being improved.

Deming’s PDSA involves a sequence of four steps that are intended to provide a structured process, which is to be continuous in quality improvement efforts. 1. In the “Plan” phase, there is great consideration for the desired outcome(s), and fac - tors that lead to the desired outcome(s) are generally included in the planning efforts. 2. Within the “Do” phase, the plan is then implemented. It is within this step that data is gathered for the following “Study” phase. 3. The “Study” phase analyzes the data, which is then converted into information. For instance, the data can provide trending information, which provides the decision maker with the knowledge and understanding to make modifications or corrections during the last phase of the cycle, or “Act” phase. 4. The “Act” phase allows for changes to the initial plan. Questions regarding the effec - tiveness and appropriateness of the plan are factored into the equation, and thus help in the development of a modified plan if necessary. Let’s look more closely at how data play into the PDSA model. In the Plan phase, data are inte - grally involved in identifying the problem, defining the project goal, planning the interven - tion/target, and developing the measurement strategy. As a first step in this phase, a problem or area of interest must be identified.

For example, patients who undergo surgery are frequently at risk of developing blood clots in their legs following the operation, which is a condition known as deep vein thrombosis (DVT). These blood clots can travel to the lungs (a condition called pulmonary embolism, or PE), causing severe heart or breathing problems and potentially resulting in the death of the patient. For this reason, patients typically receive therapy after surgery to prevent the forma - tion of clots—for example, by taking anticoagulation or anti-clotting medication.

The clinical entity of DVT and PE is collectively known as venous thromboembolism (VTE), and is a leading problem in healthcare quality for hospitalized patients. Not only is PE the fin81226_05_c05_119-148.indd 122 10/30/14 7:23 PM Section 5.1 The Role of Data in Quality Improvement third most common cause of hospital death, but complications of VTE are the most common preventable cause of hospital death (Ozaki & Bartholomew, 2012). The Centers for Medicare & Medicaid Services (CMS) has six performance measures related to VTE that hospitals are required to report (Centers for Medicare & Medicaid Services, 2013f ).

Therefore, an important and common quality metric for patients undergoing surgery is whether or not they received appropriate measures to prevent the occurrence of VTE. These measures might include getting the patient up and walking, using compression stockings on the legs, or giving anti-clotting medication to patients at higher risk.

Adherence to this routine, particularly after surgical patients leave the hospital to continue recovery at home, appears low: a recent study noted that only 1.5% of elderly patients under - going surgery received appropriate medication to prevent VTE (Merkow et al., 2013). There - fore, in the planning phase, nurses, clinicians, and administrators need to know how many patients are taking the steps to prevent blood clots and VTE and how it compares to the desired rate of use. When VTE prevention measures are below the target benchmark, clini - cians recognize that a problem exists. In this instance, data signal whether a problem may exist with the delivery of care.

The example shown in Table 5.1 describes how the Plan-Do-Study-Act cycle would be used by a hospital that wishes to decrease the inpatient readmission rate for patients with heart failure.

Table 5.1: Data use by stage of Plan-Do-Study-Act cycle Stage Data use Example Plan • Identify/define problem • Define project objective/goal (target) • Plan intervention • Identify metrics and plan data gathering Problem: Readmission rate for patients with heart failure is 35% Goal: Decrease inpatient readmission rate to 20% Intervention: Outpatient follow-up visit ,7 days Metric: Percent of patients with outpatient follow-up visit ,7 days after discharge Do • Collect data to evaluate intervention (measure impact) Data Collection:

• Outpatient follow-up visits • Readmissions Study • Analyze data: Did the intervention work? Was the goal achieved?

(measure impact) Analysis:

• Track outpatient follow-up visits weekly • Compare readmission rate after 3 months to target Act • If goal not achieved: Gather data to plan next intervention (re-enter Plan stage) • If goal achieved: Monitor data to sustain improvement Goal not reached: Interview patients to determine why inpatient readmission occurred; plan inter - vention based on results Goal reached: Monitor inpatient readmission rate Sources: Adapted from Langley, G. J., Moen, R. D., Nolan, K. M., Nolan, T. W., Norman, T. L., & Provost, L. P. (2009). The improvement guide (2nd ed.). San Francisco: Jossey-Bass; Victorian Quality Council, Department of Human Services. (2008). A guide to using data for health care quality improvement . Melbourne, Australia. Retrieved from http://w w w.health.vic.gov.au/qualit ycouncil fin81226_05_c05_119-148.indd 123 10/30/14 7:23 PM Ave LOS StatisticsWeekly Running Ave Ave LOS -Hospital Unit 1 Unit 2 ICU Tele Unit 3 Unit 4 1 23456 7 4.7 % Accuracy of D/C Predictions Monthly Running Ave Ave Acc ofD/C Predict Unit 1 Unit 2 ICU Tele Unit 3 Unit 4 60% 70% 80% 90% 100% 10.5% Projected Occupancy Next 3 Hrs Ave Proj.Occupancy Unit 1 Unit 2 ICU Tele Unit 3 Unit 4 0% 25% 50% 75% 100% 95% ED Disposition to Bed Placement Weekly Running Total Hrs on Division 0:00 1:12 2:24 3:36 4:48 Unit 1 Unit 2ICUTeleUnit 3Unit 4 2.1 % LWBS 3% # EDBoarders 6.7 Ave Bed TurnsMonthly Running Ave Hosp Ave/Mo 3.0 4.0 5.0 6.0 7. 0 Unit 1Unit 2ICUTeleUnit 3Unit 4 4.8 Hosp Annual Ave 57.6 Ave Dead Bed TimeWeekly Running Ave Ave perMonth Unit 1 Unit 2 ICU Tele Unit 3 Unit 4 0:00 1:26 2:24 3:21 1:27 Days Section 5.1 The Role of Data in Quality Improvement Dashboards are commonly used to alert healthcare team members to the existence of a potential or actual problem, especially when multiple metrics are involved. Like the dash - board of an automobile, dashboards in a healthcare system display key performance indica - tors in a way that is useful for bringing providers’ and administrators’ attention to a problem.

For example, a car’s dashboard features temperature gauges that alert the driver if and when the engine becomes too hot. Note that in this case the dashboard only signals to the driver that something is potentially wrong with the engine’s temperature; additional investigation (i.e., data gathering) is required to understand why the engine is overheating. A healthcare dashboard for an outpatient clinic might include data on several key quality measures, such as immunization rates, foot examinations for patients with diabetes mellitus, and wait times from patient check-in until seen by a provider. Dashboards often include real- time data, but they can also include data that is collected on a more periodic basis (i.e., weekly or monthly). Figure 5.1 is an example of a dashboard that shows data collected periodically. Figure 5.1: Healthcare dashboard Example of a healthcare dashboard, which displays summaries of key performance measures, such as length of stay, in an easily interpretable fashion. Dashboards provide administrators and clinicians with “at a glance” information about health system performance. Ave LOS StatisticsWeekly Running Ave Ave LOS -Hospital Unit 1 Unit 2 ICU Tele Unit 3 Unit 4 1 23456 7 4.7 % Accuracy of D/C Predictions Monthly Running Ave Ave Acc ofD/C Predict Unit 1 Unit 2 ICU Tele Unit 3 Unit 4 60% 70% 80% 90% 100% 10.5% Projected Occupancy Next 3 Hrs Ave Proj.Occupancy Unit 1 Unit 2 ICU Tele Unit 3 Unit 4 0% 25% 50% 75% 100% 95% ED Disposition to Bed Placement Weekly Running Total Hrs on Division 0:00 1:12 2:24 3:36 4:48 Unit 1 Unit 2ICUTeleUnit 3Unit 4 2.1 % LWBS 3% # EDBoarders 6.7 Ave Bed TurnsMonthly Running Ave Hosp Ave/Mo 3.0 4.0 5.0 6.0 7. 0 Unit 1Unit 2ICUTeleUnit 3Unit 4 4.8 Hosp Annual Ave 57.6 Ave Dead Bed TimeWeekly Running Ave Ave perMonth Unit 1 Unit 2 ICU Tele Unit 3 Unit 4 0:00 1:26 2:24 3:21 1:27 Days fin81226_05_c05_119-148.indd 124 10/30/14 7:23 PM Section 5.1 The Role of Data in Quality Improvement As shown in Figure 5.1, managers can quickly see such factors as the average length of stay on each hospital unit, projected occupancy rates, and the emergency department patients placed in beds on each unit. For instance, occupancy in the intensive care unit is at 100%, which could lead managers to question whether staff levels are adequate.

Dashboard reports enable organization leaders to quickly look at their data. A dashboard should be created by choosing key indicators that the organization will follow over a year’s time. It might include the total number of medical errors, fall rates, restraint rates, and the percentage of indicator compliance. It’s best to focus on a few vital measures. Managers can see a rolling 12 months of data in areas that interest them, rather than poring through many reports. There are multiple ways that dashboards may appear. Organizations may use a Gantt chart, a type of bar chart that typically illustrates a project schedule; a Shewhart control chart that can illustrate statistical variations; a box plot that can graphically depict groups of numerical data; and box-and-whisker diagrams that show the distribution of data along a number line.

The biggest challenge for an organization is to determine a benchmark standard and then the appropriate variation that it will allow. For instance, a nursing home might look at how the number of resident falls in its facility compares with other nursing homes in the state. In other words, they benchmark their rate against others’. Then the nursing home may decide it will allow a deviation of, for instance, plus or minus 5% from the state average as an accept - able level. Perhaps the nursing home has many residents with Parkinson’s disease who are at higher risk of falls and can therefore expect its numbers to be higher than other facilities.

When it comes to benchmarking, an organization must take into account its location and the demographics of its providers and patients. A large, metropolitan hospital in northeast America may want to benchmark itself against similar organizations rather than small, rural facilities in the Midwest. Similarly, a health clinic that serves many seasonal or migrant farm workers may be expected to have higher rates of infectious diseases, respiratory conditions, dental diseases, and child health problems because of the poverty, limited access to health - care, and hazardous working conditions these patients face.

Identify the Improvement Target(s) Once healthcare leaders and staff recognize a problem exists, they must use data to better understand the problem and identify potential interventions—again, a critical part of the Plan phase of the PDSA cycle (Table 5.1). In the case of using medication to prevent VTE, mul - tiple steps must be taken for a patient to receive the appropriate therapy. Figure 5.2 shows a flow chart that tracks the process of administering VTE medication to a patient. fin81226_05_c05_119-148.indd 125 10/30/14 7:23 PM Surgeon recognizes need for VTW prophylaxis Surgeon places order with pharmacy Pharmacy order check and medication dispensation Order transmitted to pharmacy Nurse administers medication to patient Section 5.1 The Role of Data in Quality Improvement First, a surgeon or other physician, such as a hospitalist, who is overseeing care must know that the patient needs the medication post-surgery. Second, the surgeon or physician must place the order (including appropriate medication and dosing). Next, the order must be trans - mitted to the pharmacy, either electronically or in written form. The pharmacy will confirm the dose and dispense the medication, which is then administered to the patient. Simply knowing that the VTE medication rate is below target is only sufficient to signal that a prob - lem exists; further data is required to understand which step(s) in the process can be opti - mized in order to improve the target metric.

Plan Intervention(s) Suppose that data gathered about the process of care delivery for VTE medication suggest that surgeons or hospitalists inconsistently order medication to prevent clot formation in their patients. When quality improvement leaders press the issue in interviews with key stakeholders, they discover that surgeons or hospitalists at a particular hospital are unfamil - iar with new guidelines related to the use of medication to prevent VTE, which results in a number of patients not receiving guideline-adherent care.

One possible intervention in an electronic health record system with its computerized order entry system would be to provide point-of-care education or a prompt that helps surgeons or hospitalists identify patients who should receive medication based on current guidelines.

Then, surgeons could either order the appropriate medication, or note the clinical reason why a patient meets the guideline metric’s exemption criteria—as would, for example, a patient being monitored for active bleeding. It would be dangerous to give such a patient medications that prevent blood from clotting.

Figure 5.2: Flow chart A flow chart can help illustrate the number and order of steps involved in a treatment process. This example flow chart tracks the administration of VTE prophylaxis to a patient. Surgeon recognizes need for VTW prophylaxis Surgeon places order with pharmacy Pharmacy order check and medication dispensation Order transmitted to pharmacy Nurse administers medication to patient fin81226_05_c05_119-148.indd 126 10/30/14 7:23 PM Section 5.1 The Role of Data in Quality Improvement Measure the Impact of Change Once an intervention is planned (called a “test of change” in the PDSA cycle), data play a criti - cal role in determining what impact, if any, the intervention had on the metric of interest. This is the role that data play in the Study phase of the PDSA cycle (Table 5.1).

Returning to the example of medication to prevent VTE for post-surgery patients, assume that the planned intervention involved use of a computerized prompt to remind surgeons to order VTE medication on appropriate patients. Not only would it be important to monitor the overall rate of VTE medication use for the unit (i.e., what proportion of patients receive the therapy), but a more direct measure of whether providers ordered the medication could be important. Determining appropriate, responsive, and valid measures to assess impacts of change and the costs and benefits of different types of measures are discussed further in Chapter 6.

Maintain Improvement Once a change in a process of care is instituted, and data suggest that the change is effective in achieving the desired result (i.e., the use of VTE medication reaches the target threshold for the surgical unit), it becomes important to stakeholders to sustain the improvement. When a specific care process or quality metric is the subject of intense focus, such as during a quality improvement project for VTE prevention, the increased focus and attention can lead to per - formance changes that fade over time, as organizational priorities shift to other areas.

This is one form of the Hawthorne effect , which suggests that the mere act of measuring a process can result in improvement, as stakeholders in the system become aware of measure - ment and focus attention on compliance. In this case, dashboards or other forms of continued surveillance of key quality measures become important for making sure that adherence to VTE use guidelines remains at or above target levels. If data suggest that compliance has fallen off several months after a quality improvement effort, this monitoring function serves as an indicator to re-enter the quality improvement (PDSA) cycle (Table 5.1, Act phase of PDSA cycle).

It is always important for organizations to consider and address any confounding or influenc - ing variables that may cause a measure to be out of the acceptable range and then character - ize those variables into two categories: those the organization can change and those it cannot.

For example, certain times of year, such as those around the Thanksgiving and Christmas holi - days, are typically associated with higher levels of depression among patients. Consequently, patient satisfaction with health services may be influenced by that variable and dip below the tolerance range for November and December. At face value, that may appear to show a problem as the numbers on patient satisfaction surveys drop. But really the drop in scores has nothing to do with the current quality of service the organization is providing.

However, if the satisfaction rate dropped in November and December and then remained at that level in January and February, there may be a real problem that needs improvement to increase patient satisfaction to the acceptable level or goal. fin81226_05_c05_119-148.indd 127 10/30/14 7:23 PM Section 5.2 Data Collection Overview Confounding variables can also impact the outcome of a quality improvement project. For instance, a hospital wants to know how effective surgery is in treating women for inconti - nence. Variables that may impact the outcome of that surgery might be the age and activity of the women, whether they are obese, the severity and duration of their symptoms prior to treatment, or even the type of surgical procedure.

Or a researcher is studying the relation between birth order (first child, second child, etc.) and the presence of Down Syndrome. A confounding variable in such a study would be the age of the mother, since a higher maternal age is directly associated with Down Syndrome in a child. If a researcher were studying the effect of smoking tobacco on human health, confound - ing variables might be alcohol intake and diet, as both are lifestyle activities that can also have an impact on health. Therefore, it is important to take into account these kinds of factors, which can influence an organization’s data. Questions to Consider 1. In your opinion, what step(s) of the quality improvement cycle are most reliant on valid data? Why? 2. Consider a recent customer experience that you thought could have been better (for example, a long wait for your meal at a chain restaurant). Write down what data you might use to identify the improvement target, plan an intervention, measure the impact of a change, and sustain the improvement. 5.2 Data Collection Overview It is clear that quality improvement methods rely on data in all phases of the improvement cycle. In order to maximize use of these data, it is important to understand the various types that exist, and when each type may be useful during specific stages of the quality improve - ment process. In addition, leaders in healthcare should be familiar with existing data sources and understand when it may be necessary to collect new data. Finally, healthcare leaders should be knowledgeable in a variety of methods to collect new data.

Empirical Data Several types of data exist, and each type possesses advantages and disadvantages, depending on the quality improvement question a healthcare leader attempts to address. Broadly speak - ing, data may be considered quantitative or qualitative. Quantitative data are numerical—for example, a patient’s age, the value of a laboratory test for the amount of sodium in a patient’s serum, or the charges for an inpatient hospitalization. Qualitative data include descriptive data, such as patient race, ethnicity, or gender, which may be readily expressed in data stor - age systems as numerical data. Other examples of qualitative data that may be less readily expressed numerically include opinions, as well as written or verbal descriptions of experi - ences from the patient perspective. fin81226_05_c05_119-148.indd 128 10/30/14 7:23 PM Section 5.2 Data Collection Overview In order to determine appropriate analytic methods to provide insight into quality improve - ment opportunities, let’s consider some of the categories of data. Examples of continuous data include a patient’s age and the level of a toxin in a patient’s bloodstream. The key prop - erty of continuous data is that the interval between different levels is consistent. For example, a two-year age difference exists when comparing two patients aged 17 and 19 years as well as two patients aged 63 and 65 years. The difference between age levels (or other continuous variables) is meaningful on a scale. Categorical data (sometimes referred to as nomi - nal data), as the name implies, are those in which qualities of the subject being measured may be sorted into different groups or categories. These categories could include different racial groups, or patients diagnosed with colon cancer, or those who have suffered a heart attack.

When planning analyses or presentations, it is impor - tant to understand what type of data is being con - sidered, as its type determines the best way to sum - marize and present it. Conceptually, the process of summarizing data highlights the distinction between data and information. Data are essentially raw mea - sures—for example, the length of hospital stay in days for the last 100 patients discharged from the surgi - cal unit. It is only when the individual data points are summarized that useful information is produced.

Therefore, it is important to understand a variety of potential summary measures. For example, quality improvement teams may wish to summarize continu - ous data as an average. In contrast, categorical data are usually summarized by frequency or proportions (i.e., what percent of a group is male or female). Examples of continuous and categorical data collected in the process of delivering healthcare abound. Demographic data, including age, sex, and race, are routinely collected. Other clinical data, such as laboratory tests and vital sign measurements, like temperature, heart rate, and blood pressure, are routinely available in the healthcare environment. Information about use of medications is frequently available, as is financial data. Administrative data, such as that used to report claims for healthcare services rendered, are also readily available. Almost all of these data are either categorical or continuous, and therefore it is important to understand how to identify, summarize, present, and analyze them to best support quality improvement efforts. We will look more closely at sources of internal data later in this chapter.

Experiential Data In the course of quality improvement activities, data beyond simple continuous or categorical qualitative information are frequently required. These may include concepts, thoughts, ideas, or individuals’ opinions about experiences and are known as experiential data . For exam - ple, consider a quality improvement project that is designed to reduce patient wait times at an ambulatory clinic. Quantitative data, such as wait time (the time elapsed between when age fotostock/SuperStock Continuous data may include information measured over consistent intervals of time, such as age or weight. fin81226_05_c05_119-148.indd 129 10/30/14 7:23 PM Clinic Anesthesia Pre-Op Screen Scheduling Lab Testing Pre-Op Holding Operating Room Post-Anesthesia Care Unit Home Registration Inpatient Surgical Unit Section 5.2 Data Collection Overview the patient checks in and when she is seen by the healthcare provider), are the key metric for measuring and monitoring sustainable change. However, these types of quantitative data reveal little about how patients physically move through the clinical area (patient flow). The phenomenon of patient flow may best be represented with a visual technique, such as process mapping (see Figure 5.3).

Similarly, consider a brainstorming session in which nurses, physicians, administrators, and other staff members are trying to improve the timeliness of antibiotic administration for patients with sepsis, a severe, often life-threatening systemic infection. Timely antibiotic administration reduces the death rate from sepsis, and is thus an important quality marker.

The participants in the brainstorming session could identify a number of patient, staff, or pro - cess factors that contribute to the problem of delayed antibiotic administration. For example, suppose that intravenous antibiotics are not readily available in the clinical unit, but instead must be obtained from a central pharmacy. The process of transmitting and confirming the order, preparing the antibiotic, and transporting it can pose a significant time delay between when a physician orders the antibiotic and when a patient actually receives the first dose.

No efficient method exists for capturing these potential causes in a typical empirical or variable-driven approach. Variable-driven approaches are inefficient because it is difficult to capture every circumstance that could create a delay. For example, a transport system could cease functioning, inventory may be low or not immediately available, or competing demands from another ill patient could interfere with timely antibiotic delivery. Instead of a variable- driven approach, participant descriptions of the process may better capture relationships among various components of the phenomenon. A fishbone diagram (cause-and-effect diagram) can be an effective way to visually communicate the relationships between various steps in a process, and, more importantly, identify opportunities for intervening to improve an important metric, such as timeliness of antibiotic administration (see Figure 5.4).

Figure 5.3: Process map A process map can help illustrate the number and order of steps in a treatment process, which can vary per patient. Clinic Anesthesia Pre-Op Screen Scheduling Lab Testing Pre-Op Holding Operating Room Post-Anesthesia Care Unit Home Registration Inpatient Surgical Unit fin81226_05_c05_119-148.indd 130 10/30/14 7:23 PM Environment Machine Person Material Method Error Section 5.2 Data Collection Overview Questions to Consider 1. Describe a healthcare experience where “just the numbers” (e.g., empirical data) do not fully capture the opportunity for improvement. What other factors might come into play in making an improvement? 2. Think about a recent customer or patient experience where your expectations were not met. Draw a fishbone diagram that describes how people, equipment, procedures, measurements, and materials could contribute to the problem you experienced. Which factors were most critical? Fishbone diagrams visually represent factors contributing to a defect or failure. Typically, there are five categories of potential causes of a defect: people, equipment, procedures, mea - surements, and materials (Langley et al., 2009). Examples of people factors include inadequate training, poor communication, or apathy. Examples of equipment factors include breakdown of order relay systems or transport systems. An example of a procedure breakdown could be if the antibiotic was improperly mixed (if it was dissolved in water instead of saline solution).

Measurement factors could include improper dosing. Material factors could include inventory management problems or if an antibiotic spoiled because it was not properly refrigerated.

Figure 5.4: Fishbone diagram A fishbone diagram provides a systematic framework to identify potential causes of failures in care delivery.

Environment Machine Person Material Method Error fin81226_05_c05_119-148.indd 131 10/30/14 7:23 PM Section 5.3 Internal Data Sources 5.3 Internal Data Sources Quality monitoring and improvement activities are often local, and the goal is to achieve an internally identified target for a process or outcome measure. In this case, there is little need for comparison data from other healthcare entities, such as other hospitals in the surround - ing area or state. During the planning phase of any quality improvement project, leaders must determine whether to use existing internal data sources, or whether it is necessary to collect new data to successfully implement the project. This decision is closely tied to the measures selected to track change and monitor sustained quality improvement. Existing internal data resources possess many advantages over data that requires a new collection process.

First, existing internal data are already being collected and stored for other purposes within the healthcare system. Claims for reimbursement from Medicare and private insurers include numerical codes for the problems treated (diagnoses) and the procedures performed, such as a surgical intervention. These are routinely captured in Common Procedural Terminology (CPT) and International Classification of Diseases—Clinical Modification (ICD) codes. Often there are other internal data used for local purposes, which can include inventory tracking with barcodes and proprietary systems for tracking lab results, such as the outcomes of tests for infectious bacteria (e.g., blood or urine cultures to detect bacteria infecting a patient).

For example, a project to reduce the incidence of bloodstream infections, a serious and often lethal consequence of hospital care, might take advantage of internal data systems that track and report when a blood culture sample from a patient grows bacteria.

Use of existing internal data resources therefore reduces the cost of quality improvement projects, since new data collection and storage resources are typically not required, or are minimal. Existing data resources are often easy to use and analytic work can usually be per - formed quickly. The main limitation to using existing data resources is that sometimes the measures are imperfect, or no data exist that measure the outcome of interest with enough precision to support the quality improvement project. In other words, the validity of the exist - ing data may be limited.

Collecting new data may be necessary, particularly when no data exist to adequately describe the phenomenon under study. For example, it is unlikely that any existing internal data resources adequately describe how patients move through an ambulatory surgical center, from preoperative preparation through surgery, recovery, and discharge. The main limitation of collecting new data is the time and expense required to abstract/observe, record, and store new data. Expertise in database design and maintenance may also be required if the amount of new data will be substantial, or if it will be collected for an extended period of time.

Health system administrators, physicians, and other quality improvement leaders must there - fore carefully weigh the tradeoffs between using existing internal data sources and collecting new data to support a local quality improvement effort. For example, no existing system may capture the number of times each week that a nursing unit runs out of gauze to dress patient wounds. In this case, new data collection may be warranted, even if it is takes time or costs money. In contrast, consider a project screening for bloodstream infections. It would be time consuming to check every record to determine which patients truly had a bloodstream infec - tion; instead, data for insurance claims could be coded and examined to find out if a blood - stream infection occurred. While the claims data may not perfectly identify every patient, it will be much more efficient to use because it is already being collected electronically. fin81226_05_c05_119-148.indd 132 10/30/14 7:23 PM Section 5.3 Internal Data Sources Using Existing Internal Data Sources Healthcare organizations collect a lot of internal data that is invaluable and can be used in quality improvement projects. Below we examine the various types of data that organizations collect from within their own walls.

Demographic Data Delivery of patient care is a data-intensive process. Demographic data, such as a patient’s age and sex, are routinely recorded, as are data regarding where a patient lives (home address, ZIP code) and even where they work. Demographic information can be important to many continuous quality improvement projects. For example, a hospital that has a high rate of patient falls may wish to track data on the ages of those patients. Are the majority of falls occurring in elderly patients, who are at a higher risk of falls? If so, the hospital can target its improvements to that population of patients.

Socioeconomic Data Healthcare organizations can also collect socioeconomic data on patients, including factors such as race and ethnic origin, birthplace, language, income levels, and education levels.

This information can help a hospital have a clear picture of who is using its facility. Is the hos - pital serving a poorer population, where many patients were born in a foreign country? Does that create language barriers that can lead to cases where patients do not take their medica - tions because they cannot understand instructions from their physicians? What can the hos - pital do to help these patients? Can nurses better educate patients before discharge? Can the hospital specifically assign nurses to meet with patients before they go home and ask them to repeat the information and treatment plan they are to follow when they leave the hospital?

With information in its hands, a hospital can ask if it is meeting patients’ cultural, religious, and ethnic needs.

Clinical Data Healthcare providers assess patients in many ways, such as recording their height, weight, blood pressure, temperature, and heart rate. In today’s healthcare environment, laboratory test results are typically stored and reported electronically.

Healthcare organizations are usually paid for the services they provide, and will track the entity that pays for the services (i.e., a private insurance company, Medicare, etc.). It is nec - essary to document which services an individual patient received, so the healthcare system employs coding systems that standardize the reporting of diagnoses and procedures. The U.S.

healthcare system is currently moving towards adoption of an updated coding system known as ICD-10. The code set allows more than 14,400 different codes and permits the tracking of many new diagnoses. ICD-10 is set to replace the older ICD-9 although the government has changed the implementation deadline several times. It is now scheduled for the fall of 2015.

It is easy to understand how clinical data can play a role in quality improvement. For exam - ple, how many nursing home residents are diagnosed with Alzheimer’s disease? A nursing fin81226_05_c05_119-148.indd 133 10/30/14 7:23 PM Section 5.3 Internal Data Sources home might track how many of those residents have “sundowning,” a phenomenon where residents experience a state of confusion at the end of the day and into the night, which can result in behaviors such as anxiety, aggression, or wandering. What can staff do to help reduce sundowning, such as trying to occupy residents’ time with another activity or limiting back - ground noise that residents may find agitating?

Financial Data Healthcare organizations must also track costs and bill for services, so financial information about charges to patients and insurers are also frequently available electronically. All of these data sources can be brought to bear on projects to improve the safety or quality of care deliv - ered to patients.

Example of Data Use As an example, consider how a project to reduce the incidence and severity of post-biopsy infection in men who undergo a prostate biopsy (to rule out prostate cancer) might make use of several types of existing data to improve care. Infection after prostate biopsy is relatively rare (it typically occurs in less than 3% of men undergoing biopsy), but it can result in a severe, potentially life-threatening bloodstream infection. Doctors in the urology clinic have recently become interested in this problem because of national reports of increasing antibi - otic resistance in men experiencing this complication.

Given the relatively rare occurrence of this event, it would take a long time for quality improve - ment leaders to identify enough patients experiencing the complication to gain meaningful insight (what is known as the prospective approach, in which researchers follow patients forward through time and wait for an outcome to occur). Therefore, the urologists and quality improvement administrators use a retrospective approach (that is, one that looks backward in time).

In order to identify patients at risk for an infection, the quality improvement team uses exist - ing administrative data , which is collected to document the care provided to patients for insurers and government payers, to identify all patients in the past two years who underwent a prostate biopsy. In order to be paid, the clinic must document when doctors performed a biopsy on a patient, so they will have this information in their records. Severe infections following prostate biopsies typically occur within two weeks of the date of biopsy, so hospi - tal administrative data are searched to identify patients from among those who underwent a prostate biopsy to determine which patients were seen in the emergency department or admitted as an inpatient within two weeks of the date of their biopsy. This process, all per - formed using search algorithms on existing data, identifies the number of patients who had the procedure and the number who potentially experienced the complication the hospital wants to track.

However, it is important to note that not all of the patients that had a prostate biopsy and were subsequently hospitalized definitely had an infection. Existing diagnostic codes for the hospitalization could help distinguish between who may have had a post-biopsy infection and who may have been evaluated for chest pain that coincidentally occurred a week after the prostate biopsy. Thus, the hospital may want to verify that the patient had a post-biopsy infection by looking beyond those diagnostic codes at the patient’s clinical data , which is fin81226_05_c05_119-148.indd 134 10/30/14 7:23 PM Section 5.3 Internal Data Sources generated in the course of caring for the patient and may include vital signs, symptoms, and laboratory test results.

Hospital microbiology laboratories typically maintain databases of specimens (i.e., blood or urine) that grow bacteria when the sample is cultured. In this case, bacterial growth is a strong indicator that the patient had a severe infection. Therefore, the microbiology data may validate that the post-biopsy hospital admission was indeed for an infection, rather than an unrelated medical problem. Therefore, the quality improvement team can calculate the incidence of the complication by dividing the number of patients who had a positive urine or blood culture within two weeks of the biopsy by the total number of patients undergoing prostate biopsy within a certain timeframe (e.g., two years).

Once the quality improvement team has determined the incidence of the complication, they may wish to analyze alternative strategies to reduce the incidence or severity of the compli - cation. While effectiveness of the treatment or prevention strategy is an important consider - ation, the cost of the various strategies may also be an important factor. To determine which potential treatment or prevention strategy would be most cost effective, the team could use existing financial data that reflects the costs of caring for patients who experience this complication.

Having diagnosed a problem in a well-defined population using existing data, the team can then decide upon an intervention and prospectively monitor for new occurrences of the com - plication going forward; ideally the intervention will result in a measurable decrease in the frequency or severity of this biopsy complication. This example demonstrates how quality improvement efforts can rely largely on existing, internal data rather than using valuable time and resources collecting new data.

Collecting New Data Despite the ready availability of existing internal data in the healthcare environment, at times it may be necessary to collect new data. In particular, this need arises when existing data do not provide valid or reliable measurement of an important aspect of the improvement process, or when descriptive insights into a phenomenon or process are required. A number of methods exist to gather new data, including surveys, focus groups or key informant inter - views, chart review or abstraction, use of electronic health records, and direct observation, among others.

Surveys Surveys are an important tool for gathering data from target groups, such as patients, nurses, or physicians, in an efficient and relatively cost effective manner. A survey is typically a series of questions, frequently with pre-defined response choices (e.g., yes or no, rate from 1 to 5, etc.), that query some facet of an individual’s experience. They can provide quantitative or qualitative categorical data, and through the use of free response questions, also permit the collection of descriptive data. fin81226_05_c05_119-148.indd 135 10/30/14 7:23 PM Section 5.3 Internal Data Sources For example, a survey could produce quantitative data by asking about a participant’s age, height, or weight.

Qualitative data could be generated by querying a participant’s race, occupa - tion, or sex (among other characteris - tics). Qualitative data can also include descriptions of experiences, such as a response to the prompt: “Think about the last time you had to wait longer than expected to see the doctor. How did that make you feel?” Many surveys undertaken for quality improvement efforts are not interactive, and there - fore lack the potential for two-way interactions that permit exploring and clarifying individual responses.

Once the need for a survey is identified, leaders should have clearly defined objectives for the survey design.

Designing a good, reliable survey is a challenging task, and therefore it is wise to consult with a survey design expert early in the design phase. The survey design expert should be able to help write good survey questions, which are specific, focused on a single topic, and use lan - guage that is not ambiguous (for further detail, see Chapter 7). Survey design experts can also help with strategies to administer the survey and obtain valid, representative information as efficiently as possible. For example, a sampling strategy may help obtain valid information from a subset of patients, rather than surveying all patients undergoing a specific procedure.

Using or adapting existing survey questions is one strategy for overcoming design challenges inherent to surveys.

After creating the survey instrument, the quality improvement team should pilot test the instrument in a target audience. For example, if a survey was intended to understand medical student knowledge and attitudes toward quality improvement, then medical students consti - tute the target audience. It would not make sense to test the questions in other groups, such as patients or faculty, that would not be similar to medical students in perspective or experi - ence. Often the pilot group is a small subset of the larger group, such as a set of 20 randomly selected medical students in a 200-person class. Other important issues to address are clarity and focus of questions, confidentiality of responses, protecting the identity of respondents (particularly if sensitive information, such as illegal drug use, is queried), and data entry and management.

While surveys are commonly used in the healthcare field, sometimes the data collected should be used with some caution. Depending on how the surveys are constructed and how carefully the questions are written, the data can have issues in terms of its validity and reliability.

With a written survey, healthcare organizations have the option of creating an original ques - tionnaire themselves or using a product that they can purchase from an outside vendor. Some experts recommend using surveys developed by a company because the product has likely Shironosov/iStock/Thinkstock Surveys are an efficient and typically cost effective tool for gathering data that can provide quantitative or qualitative categorical data as well as descriptive data. fin81226_05_c05_119-148.indd 136 10/30/14 7:23 PM Section 5.3 Internal Data Sources been tested and validated. A healthcare organization can do it itself, but the process can be time-consuming.

There are some surveys that are widely used in healthcare and found to be valid and reliable based on what is known as Cronbach’s alpha, which determines the internal consistency or average correlation of items in a survey to gauge its reliability.

Focus Groups Focus groups are an important mechanism to provide in-depth understanding of individual experiences, and typically comprise key informants or participants in a process that is the subject of inquiry. For example, a focus group of patients can provide detailed descriptions of their experiences at an ambulatory surgical center. Focus groups also permit interactive inquiry, so that the facilitator can explore new responses that may not have been recognized as important by the quality improvement team leaders. For example, an important part of patient satisfaction with an ambulatory surgical experience might be the quality of waiting facilities and amenities for family members. Unless leaders decide to explicitly query fam - ily member satisfaction, a survey might miss this important aspect of the experience, which could more readily arise in the give-and-take environment offered by a focus group.

Limitations for focus groups include the relatively higher cost and lower efficiency as com - pared with surveys. Focus groups frequently are more costly than surveys because a physical space is necessary for the meeting, food is often provided, participants must travel to the focus group site to participate, and it may take more time to explore participant responses. In addition, successfully running a focus group requires experience and, at times, specific inter - ventions to minimize the influence of strong opinion leaders within the group. It is important to keep in mind that focus groups are good for describing the range of experience, but all opin - ions expressed may not be representative of the group as a whole. In contrast, surveys can be done by mail, electronically, or over the telephone, do not require travel, and frequently have lower per-participant costs than do focus groups.

Health Record Review Even though a significant amount of clinical data is captured in electronic form, not all of it can be readily analyzed. Much of the data in the electronic health record is structured, such as the numerical value of the concentration of glucose (sugar) in a patient’s bloodstream at a spe - cific time. However, electronic medical records are also replete with unstructured text data, which can be difficult to analyze without someone actually reading the text—particularly on the relatively small scale of clinical quality improvement projects. (Unstructured data ele - ments include physician notes, scanned documents, and other data that can contain valuable narratives about a patient’s health and the reasons why healthcare decisions were made).

For example, consider a quality improvement project with the goal of decreasing hospital readmissions within 30 days of hospital discharge. One of the key pieces of data to under - stand is why the patient returned to the hospital. Was the return potentially preventable? For example, was the patient unable to obtain pain medication prescribed at discharge and had to go back to the hospital in uncontrolled pain? Was the patient not scheduled for a follow-up visit within an appropriate amount of time? fin81226_05_c05_119-148.indd 137 10/30/14 7:23 PM Section 5.4 External Data Sources Answers to these types of questions will often be contained in the descriptive or narrative part of the medical record, but are not captured in any administrative or other clinical records.

Therefore, it is necessary for team members to conduct a record review (chart audit) , in which they review and abstract pertinent structured data from the medical record. The pro - cess of record abstraction and database creation can be quite time intensive, so it should only be undertaken when no other method exists to obtain the necessary information.

Direct Observation Sometimes quality improvement projects will require direct observation of behaviors. Con - sider that physicians, nurses, physical therapists, and other care providers frequently do not wash their hands before and after touching patients, which is one key way infections are transmitted in the hospital. If staff members are surveyed about how frequently they wash their hands before and after patient contact, however, they may overestimate their compli - ance with this important quality measure. An alternative is to directly observe them using unobtrusive monitors that watch to see if doctors (and other providers) actually wash their hands in accordance to policy.

This method of data gathering is very resource intensive, and potentially intrusive, so team leaders should carefully weigh the risks and benefits of data collection that requires direct observation. Direct observation may be useful for studying behaviors that individuals may not otherwise accurately report (intentionally or unintentionally). Indeed, studies comparing direct observation and self-report of hand washing suggest that healthcare providers do not actually wash their hands as frequently as they report (Braun, Kusek, & Larson, 2009; Haas & Larson, 2007; Jenner et al., 2006; O’Boyle, Henly, & Larson, 2001). Questions to Consider 1. Describe tradeoffs between using existing data and collecting new data to support a quality improvement effort. 2. What types of information may be considered data to support a quality improvement effort? 5.4 External Data Sources A number of external data sources can also inform the quality improvement process. Exter - nal data describe the environment beyond the healthcare facility, such as the surrounding community or peer institutions. External sources of data are particularly valuable in the plan - ning phases of quality improvement efforts, as they permit comparison of local quality met - rics with regional or national peer institutions. In addition, external data can be used for benchmarking—that is, the process of determining how a local process or outcome compares with a regional or national standard. fin81226_05_c05_119-148.indd 138 10/30/14 7:23 PM Section 5.4 External Data Sources For example, hospital readmissions occur when a discharged patient is admitted back to a hospital, typically within a 30-day window. Hypothetically, these readmissions could indicate ineffective inpatient treatment at the initial hospital stay, poorly coordinated post-discharge care, or a number of other factors. Thirty-day readmission rates are currently used by large payers, such as Medicare, to assess quality of care at hospitals and provide financial incen - tives to reduce readmissions. Nationally, the 30-day readmission rate for patients with heart failure is 24.7% (Centers for Medicare & Medicaid Services, n.d.a). Thus, a hospital can deter - mine whether its own readmission rate is similar to the national average, higher, or lower.

A number of government and not-for-profit agencies provide these types of heath data (see Table 5.2).

Table 5.2: Data resources Source Description URL National Institutes of Health (Health Services Research Infor - mation Central) Links to high quality data sources regarding population health and quality of care http://www.nlm.nih.gov /hsrinfo/datasites.html The Commonwealth Fund International comparisons of health indicators http://w w w.commonwealth fund.org Web-based Injury Statistics Query and Reporting System Centers for Disease Control and Prevention injury statistics ht tp://w w w.cdc.gov/injury /wisqars/index.html Centers for Medicare & Medicaid Services Data regarding quality of care for Medicare beneficiaries https://data.medicare.gov Hospital Compare Data regarding hospital quality for Medicare beneficiaries http://www.medicare.gov /hospitalcompare / Kaiser Family Foundation Unbiased information regarding healthcare and healthcare policy http://www.kff.org University HealthSystem Consortium Quality of care data for member academic medical centers http://w w w.uhc.edu The Leapfrog Group Safety and cost data http://w w w.leapfroggroup.org National Committee for Quality Assurance (NCQA)/Healthcare Effectiveness Data and Informa - tion Set (HEDIS) Data assists health plans, hospitals, and other healthcare organizations to measure their quality and performance http://w w w.ncqa.org American Hospital Association Hospital and health system data resources http://www.aha.org/research /rc/stat-studies/data-and -directories.shtml American College of Healthcare Executives Resources providing current information about healthcare, including databases, statistical resources, and research http://w w w.ache.org / Medical Group Management Association Benchmarking tools, practice dashboards, and data tools http://www.mgma.com fin81226_05_c05_119-148.indd 139 10/30/14 7:23 PM Section 5.4 External Data Sources National Institutes of Health The National Institutes of Health maintains the Health Services Research Information Central website ( http://www.nlm.nih.gov/hsrinfo/datasites.html ). This site provides links to a num - ber of high quality data sources with information about several facets of health and health - care, including quality of care. It includes links to individual county health rankings that describe the health of local populations ( ht tp://w w w.count yhealthrankings.org ). The Com - monwealth Fund provides international comparisons of a variety of health indicators, includ - ing mortality and healthcare spending ( ht tp://w w w.commonwealthfund.org ). Data related to injuries can be found in the Centers for Disease Control and Prevention’s (CDC) Web-based Injury Statistics Query and Reporting System ( ht tp://w w w.cdc.gov/injury/wisqars/index .html ). So, for instance, using the CDC sta - tistics, emergency department staff could find data about the number of traumatic brain injuries reported or the numbers of burn victims in their region or state. They could use color- coded fatal injury maps that show pat - terns of death rates across national, regional, and state levels, to help iden - tify populations at high risk of injury.

Medicare The Center for Medicare & Medic - aid Services (CMS) coordinates the Medicare Hospital Compare program to help patients find and understand quality metrics at hospitals that are certified to treat Medicare patients (ht tp://w w w.medicare.gov/hospital compare /). The website allows patients to search by hospital name or geographic location, and shows how individual hospitals rank compared to the nationwide average. For each of several quality metrics, hospitals are classified as being either 1) better than the U.S. national rate; 2) no different than the U.S. national rate; or 3) worse than the U.S. national rate. Each metric is adjusted for the severity of the patients treated by individual hospitals, so that hos - pitals treating relatively healthier patients are not significantly advantaged in the rankings.

For example, a hospital treating patients for heart failure may have worse outcomes if those patients are older, suffering from diabetes that can cause complications, have chronic obstruc - tive pulmonary disease (COPD), or have had a previous heart attack. Pneumonia patients with such conditions as liver disease or prior cardiovascular events are likely to fare poorer than healthier patients.

Comparison data are provided regarding timeliness and effectiveness of care, readmissions, complications and deaths, use of medical imaging, and patient satisfaction with care. In addi - tion, information about how much Medicare pays each hospital is reported. . A9999 DB CDC James Gathany/dpa/Corbis Websites linked to government and not-for-profit agencies, such as the CDC, provide high quality data, which allow hospitals to compare statistics with their own figures. fin81226_05_c05_119-148.indd 140 10/31/14 3:18 PM Section 5.4 External Data Sources For example, a daughter who lives in the Boston area is searching for a hospital to provide care for her mother who just suffered a stroke. She can search for data about hospitals in that city and compare how those facilities scored when it comes to measures for effective stroke care. Does the hospital ensure that stroke patients needing medicine to lower cholesterol are given a prescription before discharge? Do they make sure patients or their caregivers receive written educational materials about stroke care and prevention during the hospital stay? Do they evaluate stroke patients for rehabilitation services? How well a hospital ranks on these kinds of measures can help a person determine where they want to go for services.

Or a person who is about to have elective surgery has the option of having the procedure at two local hospitals. He can use the Hospital Compare program to see what other patients have to say. How did others rate the responsiveness of hospital staff or how well staff managed pain? How did patients survey rate the cleanliness of the hospital or how quiet it was? Were other patients willing to recommend the hospital?

Data regarding other services for Medicare beneficiaries is also publically available ( h t t p : // data.medicare.gov ). Hospitals, nursing homes, physicians, home health agencies, and dialysis facilities can all be compared using this central website. Data can be downloaded in a variety of formats for analysis, including Microsoft Access and Microsoft Excel. An online interface also permits exploration of the data if users do not wish to download it. The website offers tools for creating visual displays of the information, such as bar graphs or pie charts.

Kaiser Family Foundation The Kaiser Family Foundation ( ht tp://w w w.kff.org ) is an independent foundation that pro - vides independent, unbiased information regarding healthcare and healthcare policy. Special topics covered by the Kaiser Family Foundation include Medicare, Medicaid, changes in the private insurance market, healthcare in the safety net, and other key topics. Data and reports are typically provided without charge. The foundation is not affiliated with Kaiser Perman - ente or Kaiser Industries, and is supported by its own endowment, so it is free from vested interests in the healthcare market.

University HealthSystem Consortium (UHC) University HealthSystem Consortium (UHC) comprises 118 academic medical centers and nearly 300 additional affiliated hospitals and seeks to help its members achieve national leadership in quality, safety, and cost-effectiveness ( ht tp://w w w.uhc.edu ). Member hospi - tals report data on a variety of measures, including hospital admissions, discharges, trans - fers between facilities, purchasing, and spending. UHC then provides quality assessment and standardization of the data, adjusts for differences in the severity of illness of patients treated at each hospital, and then produces quality, safety, and cost metrics to permit comparisons between facilities.

A web-based interface to the UHC database gives individual hospitals the ability to generate automated reports that serve a dashboard function. UHC data are typically updated every three months, which means that this external data source is excellent for monitoring longer- term progress, particularly in comparison to peer institutions. However, the time lag inherent fin81226_05_c05_119-148.indd 141 10/30/14 7:23 PM Section 5.4 External Data Sources in the data processing results in data that is not as useful for day-to-day monitoring of quality indicators.

National Committee for Quality Assurance (NCQA)/Healthcare Effectiveness Data and Information Set (HEDIS) The National Committee for Quality Assurance (NCQA) is a not-for-profit organization, dedi - cated to improving the quality of healthcare organizations. The NCQA developed the Health - care Effectiveness Data and Information Set (HEDIS) to assist health plans, hospitals, and other healthcare organizations to measure their quality and performance.

HEDIS is a tool used by more than 90% of U.S. health plans to measure performance on mea - sures of care and service (NCQA, n.d.). HEDIS consists of 81 measures across five domains of care. The data allows employers to compare the performance of health plans and also allows the health plans to use the results themselves to see where they need to focus improvement efforts.

HEDIS measures health issues that include asthma medication use, beta-blocker treatment after a heart attack, controlling high blood pressure, providing comprehensive diabetes care, breast cancer screening, antidepressant medication management, childhood and adolescent immunizations, and childhood and adult weight and body mass index assessment.

Data Makes a Case for Energy Savings In many healthcare quality improvement projects, more than one type of data is used. For instance, a project aimed at saving energy costs involved the use of both financial and facility data. Facility managers at a healthcare system decided they needed to make energy reduction and savings a priority, which was no easy task, as the system operates in more than 15 states and includes over 70 hospitals as well as other healthcare facilities. Though it was a chal - lenge, by looking closely at its operations, the system was able to save more than $1 million in energy costs.

The health system used a third-party bill-pay system to gather data, pay utility bills, and work with the Energy Star Portfolio Manager. The Environmental Protection Agency runs the Energy Star program that helps businesses and individuals protect the environment through encouraging energy efficiency.

The first step for the healthcare system was reviewing its energy data. The utility bill-pay ser - vice gathered data on the gas, water, and electric use at the system’s hospitals each month and interfaced with Energy Star Portfolio Manager, which has the ability to gauge which buildings use energy efficiently and which are less than efficient. The healthcare system then had to verify that data. An employee engaged in the energy management effort went on-site to each building and confirmed the data about energy use. All the data being fed to the Energy Star Portfolio Manager was verified, including how many square feet in each building, as well as how many floors and beds.

The healthcare system then began identifying opportunities for energy savings. Hospital engi - neering helped in that effort. Managers looked specifically at lighting costs, where there is a fin81226_05_c05_119-148.indd 142 10/30/14 7:23 PM Section 5.4 External Data Sources quick return on investment. Each facility was assigned a project to re-lamp existing lights with more efficient fixtures. Secondly, facility managers made sure there was a good steam trap program for all buildings that used steam heat. The steam traps capture condensation and return it to boilers at a higher temperature, allowing them to heat buildings more efficiently.

The project managers had to justify the cost of the improvements. What system adminis - trators wanted to see—in addition to the outright expense of the project—was a return on investment. Will the healthcare system end up saving enough money to pay for the improve - ments it needs to make? Generally, if a project can pay for itself in two or three years, it is an attractive project that the administration is likely to fund. Project managers calculated the return on investment, calculated the present value of that investment, and proved those numbers. For instance, with the light replacement project, managers counted fixtures and obtained quotes for the cost of the work.

By comparing past and present energy costs, the managers were able to show the savings on energy bills. The initial focus was on the hospital buildings, which are the biggest consum - ers of energy in the healthcare system. However, now that the energy saving program is well established, managers are turning attention to how they can save energy at the system’s long- term care, assisted living, and residential living facilities.

Everybody’s Problem Let’s take a look at an example where a Joint Commission requirement became a challenge that several different hospital departments needed to solve. The Joint Commission expects hospitals to address its patient communication standards—a requirement that resulted in some creative solutions on the part of many emergency management (EM) directors.

The Joint Commission requires that hospitals make provisions for effective communication with patients with limited or no English proficiency—both during everyday operations and during emergencies. Effective communication with patients with limited English proficiency can be a challenge at any time, but especially during a major emergency or disaster.

So how would a hospital with a chemical contamination situation on its hands provide decon - tamination instructions to patients with limited English or other barriers to communication?

Joint Commission guidelines state that instructions by staff should not only be verbal, but may also include posters or other visual aids for patients who are deaf or speak limited English.

To meet this standard, emergency management directors needed to expand their planning to include how the hospital would communicate in multiple languages.

As with other aspects of EM planning, one of the first steps for hospitals to comply was to assess the situation. A key was to know the hospital population. What does the local census show in terms of population-level demographic data to help determine the needs of the com - munity? What does hospital data show about the race, ethnicity, and languages, as well as disabilities of patients?

Reviewing the demographics of the local population allows a hospital not only to plan for its communication needs on a daily basis, but also to anticipate the increased level of need during an emergency. Although a data review may not identify every small group in a com - munity, it can provide a good overview of the languages that staff will most likely encounter. fin81226_05_c05_119-148.indd 143 10/30/14 7:23 PM Section 5.4 External Data Sources The greater the population base, the greater the number of ethnic groups a hospital is likely to encounter and needs to plan for.

With data in their hands, hospital managers then had to look for solutions. A key for hospitals was to evaluate existing resources in the facility. Interpreters who were currently used on a daily basis to assist patients and families in various departments were likely a resource the hospital could use in an emergency. Interpretive services already in place may be used when a hospital’s incident command system is activated and requires language translation or cul - tural interpretation. In an emergency, a hospital would be able to translate written communi - cations for patients and families into multiple languages.

The hospitals also had to consider expanding existing services. What will a hospital do if nor - mal services are not available in the case of a blizzard or hurricane that shuts down roads and communication or if additional help is needed? One possible resource is the language skills of a hospital’s own staff, who may be proficient in multiple languages and may serve as translators. Hospitals can create a database of this information, with each staff member’s assigned department and work shifts. The database can be used in an emergency when the usual trained interpreters are not available.

One word of caution, however: Identifying other external interpretive services may be help - ful, but healthcare organizations need to keep in mind the important distinction between someone who is fluent in another language and a medically trained interpreter who under - stands the privacy, cultural, and medical aspects of translating medical information.

While the Joint Commission says hospitals must ensure the competency of their language interpreters and translators, it’s not clear how that would play out in an emergency. In an emergency situation such as a devastating hurricane or tornado, hospitals must do the best job that they can under sometimes trying circumstances and may not be held to the same standard required under ordinary operations.

Hospitals can also consider the use of third-party “language line” services for emergencies.

These over-the-phone translation services offer more than 1,000 different languages. Another option is the free “language banks” offered free by some Red Cross chapters. One issue to con - sider, however, is whether translators are aware of the need to protect patient privacy and are trained to understand medical terminology. International schools, colleges, or universities might also have instructors or students who can translate between patients and providers, especially if the size and scale of a disaster overwhelms other available resources.

By brainstorming with other hospital leaders, emergency managers looked for solutions and came up with plans to satisfy the Joint Commission requirement. Then during emergency management exercises, they tested whether their plans would be successful.

Using Data Across Settings More and more quality improvement projects are joint efforts that bring together a number of healthcare settings. Take Cincinnati Children’s Hospital Medical Center, which partnered with local physician practices, to launch a large-scale initiative to improve the care of children with asthma (Institute of Medicine, 2012a). The hospital worked with 38 community-based pediatric practices to improve the health of patients with asthma, which is one of the most fin81226_05_c05_119-148.indd 144 10/30/14 7:23 PM Summary & Resources Questions to Consider 1. Describe some of the key uses of external sources of data in quality improvement activi - ties. How might this data be used in a particular quality improvement project? 2. Describe some publicly available data sources that could be used for the following purposes:

a. Understanding the implications of proposed health policies. b. Comparing heart attack survival rates in Medicare beneficiaries. c. Determining if overall risk-adjusted inpatient mortality rates are lower at hospitals affiliated with an academic medical center. common chronic conditions in children. Asthma, a respiratory disease characterized by epi - sodes or attacks of impaired breathing, affects an estimated 6.8 million children in the United States, with many at risk for emergency department visits and hospitalizations (CDC, 2012a).

The project used population segmentation to target high-risk patients and help deliver the best care through components such as multidisciplinary-practice quality improvement teams, real-time patient-practice, and network-level data reporting. Pediatric physician practices are automatically alerted if any of their patients require an emergency department or urgent care visit or need to be admitted to the hospital because of their asthma.

The result has been better care for patients, with a 92% adherence to best practices for care management and with 92% of parents rating their child’s asthma as under control.

The project has also resulted in lower costs, with 92 avoided hospital admissions in one year, resulting in $322,000 in savings, and 266 avoided emergency department/urgent care visits.

Summary & Resources Chapter Summary Data are integral to the process of quality improvement. Team leaders of quality improve - ment efforts need to incorporate data into the planning, intervention, measurement, and monitoring phases of their study. Important tradeoffs occur between using existing data and collecting new data for a quality improvement project. The resources available for a specific project, availability of existing measures, and the validity and reliability of available mea - sures are key drivers in the decision to collect additional data.

A number of types of data exist, and each serves a particular function in the quality improve - ment process. Quantitative data allows team members to take an empirical approach and to perform statistical tests to assess relationships between variables. However, descriptive data relating to the patient or healthcare worker experience also play an important role in understanding how care is delivered and identifying possible areas for intervention. The range of experiences had by each individual in the health system provides important context for understanding how to achieve sustainable increases in quality and patient safety. fin81226_05_c05_119-148.indd 145 10/30/14 7:23 PM Summary & Resources Healthcare environments are saturated with data, much of which can inform quality improvement efforts. Hospitals and other healthcare entities routinely collect demographic, clinical, administrative, and financial data about individual patients that can be used to iden - tify and conduct quality improvement projects. When necessary, a number of techniques exist to gather additional data. The technique for data collection should be well matched to the project’s purpose and available resources so as to collect valid and reliable data.

When necessary, experts in a specific data gathering technique should be consulted prior to implementing the data collection. Finally, the Internet contains a vast trove of data about healthcare topics and healthcare institutions, which can be useful for providing context and comparators for local quality improvement efforts.

Mini Case Study You have recently been hired at a community health clinic where medication non-adherence is a major issue. Patients either delay taking their medication or do not take the medica - tion that their doctor prescribes. Some patients say a lack of insurance is the reason for not taking their medication. The clinic wants to increase the prescription drug adherence rate of patients to 90% or higher. The rate is currently 75%. Following Deming’s PDSA model, here are the steps the clinic will take.

Plan. The plan is to develop an application for cell phones so that patients will be reminded to take their pills at specific times. The study target for the project is 1,000 patients. The measure for the study is that the clinic decides to monitor patients’ compliance rate with the cell phone application.

Do. The clinic downloads the application onto 1,000 patients’ cell phones. For three months, staff collect information from patients’ cell phones in order to determine the average com - pliance rates.

Study . Clinic staff analyze the compliance data and compare results to the pre-study period. They must determine if there is at least 10% improvement from the previous statistics. The 10% threshold is randomly selected for the purpose of this study, but it may be adjusted for different organizations or programs. Generally, this rate is determined based on the consen - sus of individuals involved in the project.

Act. Clinic staff follow up with patients who did not follow the recommended schedule for taking their prescriptions. They identify the three most important common reasons for non- compliance and develop additional plans for future improvements, such as: • Offering cell phones with a basic paid call plan • Offering prescription drug cards for discounts • Offering vouchers and/or samples In this case, the PDSA model is applied to a simple problem in order to improve healthcare quality. Desired outcomes and factors leading to outcomes are set forth in the “Plan” phase, implemented in the “Do” phase, analyzed in the “Study” phase, and followed by changes that are made in the “Act” phase. This cycle enables the clinic to address barriers to the desired outcomes, modify the plan, and evaluate its effectiveness. fin81226_05_c05_119-148.indd 146 10/30/14 7:23 PM Summary & Resources administrative data Data collected to document for insurers and government pay - ers the care that was provided.

categorical data Data that describes qualities of an individual in discrete groups, which are frequently mutually exclusive.

clinical data Medical data generated in the course of caring for individual patients; examples include vital signs, symptoms, and laboratory test results.

continuous data Data that are on a contin - uous scale, such as blood pressure or serum glucose levels.

dashboard A visual display of data for the purpose of monitoring current conditions and detecting when processes may be going awry.

direct observation A data collection tech - nique that involves watching individuals perform a task, such as washing their hands.

experiential data Information, typically qualitative, that individuals provide to describe how they perceive or experience a phenomenon, such as a clinic visit.

external data Information that describes the environment outside of the healthcare organization.

financial data Information regarding the cost or charges for patient care. fishbone diagram (cause-and-effect diagram) A visual technique for identify - ing contributory factors to lapses in care delivery.

focus group A forum typically of a small group of individuals who provide qualita - tive information regarding experiences or perceptions.

Hawthorne effect A term describing the phenomenon that merely observing a behav - ior (such as hand washing) may improve performance if those being observed become aware of which behaviors are being monitored.

prospective An approach in which data col - lection is initiated prior to when an outcome, such as a patient safety event, occurs.

qualitative data Data describing character - istics or experiences of patients.

quantitative data Numerical data, such as weight, time elapsed, or many laboratory test values.

record review (chart audit) A data collec - tion process whereby medical records are examined in order to extract information regarding care of individual patients.

retrospective An approach in which one looks backward in time.

survey A series of queries designed to gather data, often provided in written or electronic format. Discussion Questions 1. What was the problem and how did it affect patient care? 2. What data sources were useful in identifying and defining the problem? 3. How did the use of a cell phone application facilitate or hinder the quality improve - ment process? 4. What are some barriers to success in getting patients to take their medication? 5. What stakeholders would be important in developing support for possible solutions to this problem? Key Terms fin81226_05_c05_119-148.indd 147 10/30/14 7:23 PM Summary & Resources Critical Thinking Questions 1. In your opinion, what are the most important roles for data in the healthcare quality improvement process? 2. How should healthcare providers balance the need to provide care to patients and the need to collect and manage data for quality improvement processes? 3. How might use of internal data sources impact the cost of healthcare delivery for a hospital or clinic? 4. Many external data sources provide rankings or other information that is also avail - able to the public. Describe some of the risks and benefits of providing this informa - tion to patients and families. Suggested Websites • Agency for Healthcare Research and Quality (AHRQ):

ht tp://w w w.qualit yindicators.ahrq.gov Federal agency website with several resources describing common indicators of quality and safety in healthcare delivery. • Health Services Research Information Central (HSRIC):

ht tp://w w w.nlm.nih.gov/hsrinfo / Website with links to a wide variety of data sources regarding epidemiology, public health, and care delivery. • Institute for Healthcare Improvement:

ht tp://w w w.ihi.org Not-for-profit organization that provides many useful resources regarding quality improvement techniques, including data collection and management tools. • Medicare:

ht tps://Data.Medicare.gov Portal for public access to data resources regarding quality of care for Medicare beneficiaries. • National Guideline Clearinghouse:

ht tp://w w w.guideline.gov Searchable database of clinical guidelines for a variety of health conditions. • National Quality Measures Clearinghouse:

ht tp://w w w.qualit ymeasures.ahrq.gov / Website with evidence-based quality metrics, as well as a link to dataset of measures being used by the U.S. Department of Health and Human Services for quality mea - surement and reporting. fin81226_05_c05_119-148.indd 148 10/30/14 7:23 PM