Statistical Process Control Charts to Improve Health Care: Hypertension Case

Established national standards for patient care safety and quality help hold health care systems accountable and ensure the delivery of consistent, impactful medical care across diverse clinical settings from acute care to ambulatory care, skilled nursing facilities, and home care.1–3 These standards encourage the implementation of quality improvement practices that continuously monitor patient outcomes to illuminate inefficiencies and areas for targeted intervention. The development of effective methods for data management and interpretation remains crucial to capturing relevant, insightful process-related information.4–6 Furthermore, proper data analysis informs the iterative evaluation of quality and safety measures. But interpreting health data (such as from electronic medical records, health devices, and community-based organization reporting) can be challenging due to its sheer volume, complexity, and inconsistent standards. With the diversity of data present in health care systems, determining appropriate systems for monitoring and tracking care processes for quality improvement can prove challenging.7 Graphical tools, like charts, help visualize trends, identify gaps in quality measures, and track progress toward goals. Appropriate visual representations can be critical tools in improving the care delivery process, distinguishing random fluctuations from process-driven changes and deficiencies. However, understanding how to create such visual representations can be difficult.This article aims to demonstrate how statistical process control (SPC) charts—by taking into consideration time, improvement cycle intervals, process measures, and control limits—can provide accurate monitoring of care processes to effectively improve patient outcomes. Using blood pressure as a case study, we highlight the challenges associated with interpreting trends in continuous outcome measures, such as blood pressure, that may be affected by a host of processes measured in various ways—dichotomously (eg, yes, no), ranked (eg, 1-5), categorically (eg, completely, partially, none), or even narratively.In quality improvement work, run charts and SPC charts are commonly employed visual tools.7 They allow teams to monitor performance measures over time, assess how interventions impact the measures, and provide a visual rendering of audit results as feedback for the team and key stakeholders in the organization. Plotting data over time provides a dynamic representation of care delivery processes, which is a key role that run and SPC charts can play.8Run charts are a graphical display of data plotted in some type of order, typically with time intervals on the x-axis and a variable of interest on the y-axis. A run chart could track average patient wait times in a clinic over a month, plotting daily wait times on a line graph to reveal trends, patterns, and variations in the process. Run charts are used at the beginning of a project to establish an initial baseline or identify the current state of a process, particularly when fewer than 10 data points are available. As such, they are often used as preparatory work in the subsequent development of SPC charts.Statistical process control charts, in contrast, provide the same type of data as a run chart, but in addition, they include control limits that help detect variation in the data.7 The control limits are horizontal lines that represent 3 standard deviations above and below the average of the data points. In practice, run charts with 10 to 12 data points present a useful snapshot of a current state or baseline function of a quality or safety process. Once more data are collected (> 15 data points), the SPC chart can be created and control limits added to the run chart trend to provide a more granular evaluation of the stability of the data trend. Combined, run and SPC charts can serve as versatile tools to help health care institutions gauge productivity, implement interventions, and improve the processes associated with many health outcomes.Improving blood pressure management is a national and global priority across health care system settings, as persistent high blood pressure increases the global burden of stroke, heart disease, and kidney failure.9,10 Successful implementation of local quality improvement approaches fosters the development of more meaningful care monitoring processes for continuous patient outcomes such as blood pressure. The following case uses a common example of a quality improvement initiative in health care systems: interventions to improve patient blood pressure. The same approach could be applied for other continuous outcome measures in critical care, such as continuous blood glucose monitoring, oxygenation saturation, or sleep quality indicators.To begin, every quality improvement initiative requires a baseline assessment of the current state. Run charts are often used for this purpose, to depict a simple visual representation of a continuous outcome measure like blood pressure. In Figure 1A, the run chart displays the average systolic blood pressure of a clinic system’s patient population aggregated across multiple clinical sites or units over the course of 12 months, which is one way a health care system could evaluate blood pressure outcomes. In situations with 10 data points or fewer, run charts provide a reliable baseline indication of the current functioning of a quality or safety process.However, a few key concerns arise with run charts, particularly when data are aggregated across multiple units. To provide accurate insight into the acceptability of specific process measures as they are reported and tracked over time, a more contextualized depiction of the data is required to show the stability of the system in which the process is occurring. For example, what is the desired target for the measure, and how far from or close to that target do the current data points fall? In contrast to the simple trend depicted by a run chart, an SPC chart (Figure 1B) incorporates statistical control limits around the mean and is therefore more effective for identifying meaningful variation in the observed trend, addressing many of the limitations encountered when using the run chart alone.For example, most health care systems track and report blood pressure control as a clinical quality measure across settings over time. The aggregation of data across various clinic sites within a health care system interferes with the interpretability of the run chart’s observed trends. In most scenarios, individual clinics operate uniquely with varying populations, staff, workflows, and procedures. Consequently, comparing outcomes on the macro level of an entire system of clinics fails to adequately illuminate areas of insufficiency or excellence. If the system experiences rising average systolic blood pressures, the shift in outcomes is not indicative of parallel changes within each clinic. As such, aggregated data do not offer insight that can be functionally converted into actionable, efficient plans for process improvement. To better understand the specific units or processes driving trends in systems-level data, data visualization methods could parse out singular units or clinics to evaluate processes at a more granular level. Should a particular clinic be experiencing concerning fluctuations in outcomes such as average systolic blood pressure, further investigation could take place to evaluate that individual clinic’s hypertension care–related processes in depth.Additionally, specific to the analysis of continuous variables, the selected time intervals may be inappropriate to sufficiently assess process effects. Wider time intervals (eg, quarterly or monthly) hinder the visibility and thus the understanding of potentially more important shifts that occur within shorter time intervals. Comparatively, an evaluation of outcomes and processes on a weekly time interval would provide more data points to elucidate the fluctuation observed in shorter time intervals. This approach also provides a larger set of data points that can more accurately gauge stability of the processes within the larger clinic or health system and provide a better picture of progress toward a target or goal.Charts plotted on longer time intervals may also omit valuable data points that are ultimately helpful in discerning whether a process should be modified. A month-to-month observation compared to a week-to-week alternative would be less sensitive to variations in the outcome of interest. In addition, although a run chart offers a snapshot of baseline trends, SPC charts incorporate statistical control limits (3 standard deviations above and below the mean) and thus are more effective at identifying meaningful variations. By focusing on blood pressure data collected at shorter intervals (eg, weekly) and applying standard quality definitions (eg, Healthy People 2030 guidelines11 for hypertension), SPC charts depicting upper and lower control limits can be used to visualize actual variation in reported blood pressure. If certain weeks show significant deviation beyond these limits, an outlier or special-cause variation can be suspected as the cause, rather than a system-wide, process-related common-cause fluctuation that could occur to anyone in the same system. Thus, using an SPC chart to depict a process helps clarify when targeted interventions or adjustments may be necessary.While Figure 1B is effective in highlighting variation in blood pressure, it fails to elucidate the care processes that are in place to achieve the desired outcome, which is essential for targeting interventions accurately. Process measures are a way to evaluate how systems and care processes within systems work (or do not work) to achieve a desired outcome. For example, if the outcome of interest is blood pressure, process measures associated with the outcome may include measurable aspects of the blood pressure reporting workflow (Figure 2).Intervening events, be they planned or unplanned, should be tracked and indicated on the run or SPC chart. Examples include the introduction of new educational programs or changes in clinical activities associated with the process measures, such as pharmaceutical interventions.Stratification of the data based on the location of where the blood pressure was obtained (eg, acute care unit, ambulatory clinic, or home measurement), how it was obtained (eg, manually, automatic cuff, or patient self-report), and who reported it (clinician, patient, or other) may also be collected to ensure consistent and accurate comparison of the data. Differentiating outcomes based on the level of process measurement is important to understanding each outcome. Different processes can widely influence the interpretation of the outcome and can be either continuous or attribute.Although both run and SPC charts are useful for process evaluation, selecting the appropriate chart type within a given context is key to accurately interpreting trends in the data. The appropriateness of a chart is dependent on the process of interest and the data being represented. Accurate representations of process outcomes and evolution can be particularly difficult when working with continuous data. Continuous variables are calculated on scales that allow for an infinite number of values, while discrete variables only allow whole numbers, so the values are finite.12 Because of this wide range of values, continuous data, like blood pressure measurements, have a variety of methods for determining averages, categorical assessments, and other statistical analyses. The variability of continuous measures such as blood pressure can complicate the procedure for determining appropriate graphical representations and limits of normal fluctuation. However, establishing consistent methods for data collection and interpretation can provide a meaningful way of working with continuous variables and implementing effective quality improvement processes.Run charts and SPC charts are ultimately concerned with the evaluation of processes. Therefore, when using these tools for improving care delivery, it is critical that graphical representations inform viewers about the nature of the exhibited changes in data and any key events or environmental changes that potentially intervened in the observed outcomes or trends. To accurately determine process stability, evaluators must successfully identify data points that are extraneous and cannot be accounted for by existing processes. In order for health systems to make impactful process changes, assessments must provide a means of recognizing fluctuations that are driven by implemented structures. Disaggregating data and focusing on specific process measures can help identify where interventions are needed. For continuous outcomes like blood pressure, this approach will guide more targeted improvement efforts.To illustrate how process measures can be integrated into SPC chart construction, consider a quality improvement intervention focused on improving no-show rates at follow-up appointments for patients with elevated blood pressure at a single clinic (Figure 3). The steps for beginning a clinical inquiry project for this question were reviewed, and include: After these steps, the SPC chart reveals whether changes in weekly follow-up no-show rates for patients with high blood pressure are due to random special-causes events, such as a weather-related storm that kept patients away from clinic that week, or a common-cause variation such as the system-wide implementation of an intervention to automate reminders for blood pressure follow-up appointments. For instance, if weekly follow-up attendance consistently falls inside the control limits established on the SPC chart, it may indicate that the intervention has introduced a significant and measurable impact, either positive or negative. This allows clinic administrators to evaluate whether specific changes in their processes—such as sending automated SMS reminders, offering transportation assistance, or adjusting appointment schedules—have effectively addressed attendance barriers.Focusing on follow-up show rates at the clinic level offers valuable opportunities for quality improvement. An SPC chart can reveal whether weekly attendance patterns fall below the lower control limit, indicating the need for targeted interventions. These might include revising reminder message content, addressing logistical barriers such as clinic hours, or implementing personalized outreach through patient navigation services. By identifying trends that stabilize within the adjusted control limits based on shifting means following an intervention, SPC charts also provide evidence of successful changes in care processes.By consistently tracking follow-up show rates using SPC charts and applying interventions as needed, the clinic can ensure continuous improvement in patient engagement and ultimately improve health outcomes for patients with elevated blood pressure. The integration of consistent denominators, such as the total number of scheduled follow-up visits and the number of available providers, as well as the use of precise data collection intervals, allows for reliable and interpretable data visualization that can drive targeted actions.In sum, SPC charts can serve as an effective and standardized tool to capture relevant, insightful process-related data that impact continuous clinical outcome measures in health care systems. This column describes the use of stratified or disaggregated SPC charts as more informative alternatives to simple run charts for the evaluation of processes related to continuous variables like blood pressure and for the measurement, monitoring, and evaluation necessary to improve the quality of care.