Quality Improvement and Behavior Analysis: A Dynamic Duo

Abstract

Behavior analysis as a discipline prides itself on the systematic use of consistent, measured observations to drive specific and clearly defined changes in behavior. The need for diversification of practice is becoming increasingly focused on the topic. We posit that one such area of growth for behavior analysts could be quality improvement (QI). The field of health care QI utilizes specific tools and techniques to drive advancement in the quality and delivery of health care. There are deep corollaries between the methodologies used in QI and behavior analysis. We describe these corollaries through explanation and examples across the 7 dimensions of behavior analysis. We conclude that strong similarities exist between these fields, providing potential pathways for behavior analysts to expand our scope while maintaining the field’s core values.

Behavior analysis can save the world. Behavior analysts who have been engaging in high-fidelity practice do not just believe this to be true—we know it is. There is remarkable compatibility between behavior analysis and quality improvement (QI) in health care. Behavior analysis is one of the best researched evidence-based practices and boasts a strong history of rigorously conducted and replicated studies. Owning such a thorough and methodical science could potentially be a barrier when considering the integration of other approaches into behavior-analytic practice. Despite the perhaps unlikely pairing, both fields pride themselves on many of the same core values and utilize some similar methodologies. We discuss these same core values and similar methodologies in this article. Not only what the two disciplines have in common but also how behavior analysts could enhance an incredible number of lives by becoming involved with health care QI should be considered.

QI in health care, relying heavily on W. Edwards Deming’s framework, is said to be based on three primary factors:

1. What do we want to accomplish?

2. If change is made, how do we know we improved?

3. What type of changes should we make? (Deming, 2000; Langley et al., 2009; Varkey, Reller, & Resar, 2007)

QI intends to examine and manage processes in a measurable way, to build high reliability, and to have highly safe health care. Organizational behavior management (OBM) examines the contingencies in the environment that support both appropriate and problematic behaviors to facilitate more effective work performance (Sundberg). Although arguably the fields of QI and OBM are different in practice, conceptually, they share many primary features.

The roots of QI started in factory work and manufacturing (Deming, 2000). Individuals concerned with quality, in assembly line formats, observed variations in the final product. To reduce variations, changes were made—and sometimes the variations got worse. As a result, continuous measurement and process improvement were born. QI implementers identified a problem or a goal, made sure it was objectively measurable, implemented interventions to see what happened to the continuous process, and, perhaps most importantly, determined if their change resulted in improvement. Such a model is nothing new to behavior analysts either. Baseline data, experimental control, and making data-driven changes are hallmarks of our field (Cooper, Heron, & Heward, 2007). We also recognize that there is normal variability in a process or behavior—that ups and downs do not necessarily mean we need to overhaul the factory—but stable trends and shifts may indeed indicate a need for changing the intervention.

Diversification of practice is at the forefront of many behavior-analytic initiatives (LeBlanc, Heinicke, & Baker, 2012; Normand & Kohn, 2013). We have begun to tackle building the pathway for behavior analysts in fields outside the treatment of developmental disabilities. QI is an opportunity for those interested in applied behavior analysis (ABA) and OBM to diversify our practice. By virtue of being trained as behavior analysts, we have theoretical knowledge and foundational skill sets conducive to learning about QI (e.g., understanding antecedents and consequences, obtaining baseline data, continuous measurement). It is relatively straightforward for behavior analysts to carve a path in QI, and there is so much for organizations and systems to gain in doing so. Here, we discuss the parallels between QI and the seven dimensions of behavior analysis. We also reflect on what behavior analysts can contribute to QI by examining each of the seven dimensions of ABA with comparisons and examples (Baer, Wolf, & Risley, 1968).

Applied

As behavior analysts, we first and foremost look to the applied nature of our interventions. The social significance and real-life benefits are at the heart of the change we seek to make. Similarly, QI strategies seek meaningful change in imperfect systems. Traditional research designs examine pre- and postmeasures under, often, unnaturally controlled conditions. QI methodology recognizes that change occurs within moving systems and that monitoring a dynamic system over time can demonstrate meaningful impacts of interventions (Lloyd & Provost, 2011; Provost, 2011). Statistical process control (SPC) methodology is utilized to reflect upon the significance of the change—the QI equivalent of experimental control. SPC methods unite the rigor of classical statistics with chronological analysis and display of data, which helps QI practitioners detect process changes and trends earlier (Benneyan, Lloyd, & Plsek, 2003). There is some disagreement among behavior analysts about the value of statistical analyses (e.g., Ator, 1999; Branch, 1999; Crosbie, 1999), with some behavior analysts identifying the value of statistical analyses to our field. Nonetheless, some behavior analysts may be deterred from QI because this is a notable difference—QI uses group research designs, whereas behavior analysts traditionally use single-case research designs. However, behavior analysts should not be deterred. The nature of the SPC methodologies utilized in QI is inherently tied to an applied and observable change in the process being monitored. Specifically, SPC refers to the use of statistics to determine data points that are more than three standard deviations outside of the mean or current “level” of the phase or process. These outliers provide statistical reliability that there was something unique (positive or negative) occurring during that data point, suggesting further investigation. As an example, in a QI study to improve antipsychotic medication safety, the QI team used a control chart (a Shewhart chart), an SPC tool to measure and improve compliance with safety guidelines (Figure 1). The chart allows the team to observe change over time and examine data points (i.e., Quarters 2 and 3, 203) that are outside SPC in the positive direction, capitalizing on the effective interventions in those time periods that led to statistically significant change.

Fig. 1.

Control chart, an SPC tool used to measure and improve medication safety for children and adolescents. A “p-chart” is a specific type of control chart used in this particular study

Statistics does need to be a taboo word—we suggest that behavior analysts’ concern with statistical significance arises when conclusions are drawn due to a discrete comparison and then are used to make inferences about socially meaningful change. The use of statistical significance in QI facilitates the identification of meaningful and sustained improvement.

Another example of applied QI work in a health care setting includes reductions in infections. Patients who have catheters, central lines, or other ports where infection has a high probability of entering are at constant risk. Health care providers have been challenged to establish best practices, protocols, and interventions to reliably protect patients from secondary infections associated with the use of such invasive lines. This change is socially significant and meaningful—but it is also occurring in the context of a hospital system that cannot shut down, or “stop” using current infection-control best practices to establish experimental control. This concept is similar to common ABA intervention targets, where the need for experimental control is still critical, but certain skills cannot be untaught for the purpose of demonstrating control, or certain behaviors cannot be allowed to continue due to danger for the client.

Behavioral

Baer et al. (1968) articulate that behavior is composed of physical events that can be measured and observed—that the behavior of an individual must be substantiated, rather than assumed based on a statement that behavior occurred. Behavior analysts highly value the reliability of measurement, realizing that the value in conclusions drawn about the change is irrevocably linked to the accuracy of the data. Drawing actionable information from data is also a foundational concept in QI. For example, the Institute for Healthcare Improvement (IHI) methodology for QI builds around an “aim statement” that clearly articulates a measurable, observable, specific, and time-oriented goal. QI methodology stresses the use of a data plan, which allows the team to focus on all possible elements of the data, including any exclusions, visual representation, and determination of the data source (Langley et al., 2009). A data plan, per IHI methodology, requires the delineation of clear operational definitions of the target data to be collected, the populations to be included or excluded, the time period in which a single data point will be encapsulated, the designated data collection method or source, the way in which data will be displayed, and the way in which the data will be normalized (i.e., discrete or ratio vs. continuous or rate).

Examples of observable and clear data collection within QI can be found in specific aims. A “specific aim” includes (a) what is to be increased or decreased, (b) in what group or population, (c) from what, (d) to what, (e) by when, and (f) how long it will be sustained. Examples of aims in QI, in the area of behavioral health, include reducing the percentage of patients who were readmitted to a specific psychiatric facility 30 days following their admission (and sustaining these numbers for a specific amount of time) or increasing the number of new patient intake or diagnostic appointments that are scheduled each month.

Analytic

The regular analysis of data is critical to good behavior-analytic practice. If we are going to make meaningful change in the lives of those we serve, we must constantly monitor pertinent behavior and make the changes to meet the goal. Similarly, those who engage in QI practice also recognize that a pre- and postmeasure are not sufficient to examine the pattern of behavior—two data points do not a trend make. In QI, data are tracked daily, monthly, quarterly, and annually (referenced earlier as “the time period in which a single data point will be encapsulated”)—with whatever frequency has underlying statistical reliability. Specifically, for a data point to have statistical reliability, there needs to be a large enough denominator to make that data point valid, which was previously referenced when discussing how a mean is used to determine SPC through identifying outliers using standard deviations. The use of categorical analysis is another method within QI that supports precision intervention and the meaningful interpretation of data. Categorical analysis in QI is essentially breaking down a larger data set into smaller parts to better understand the variables that may be impacting the overall total. Specific tools such as Pareto charts and process maps (flowcharts of processes) allow the QI team to examine different variables impacting a process, similar to a component analysis (Bucchianico, 2016). A Pareto chart is a tool that is used to break down the impacted variables into more manageable subcomponents. This can be better explained by the “80:20” principle. This concept is related to choosing the best variable to target to have the largest impact. For example, if 80% of delayed client intake is attributable to a single funding source, and the remaining 20% is minutely dispersed across the remaining funding sources, it would be considerably more efficient to target the first intervention with the 80%, getting a “bigger bang for your buck.” A process map is a tool to plan for an efficient workflow or to identify inefficiencies in existing workflows. By identifying where most of the process inefficiencies are, or the area with the highest percentage of error, subsequent interventions can be targeted more effectively. Pareto analysis can be conducted using data from a variety of sources, such as the frequency of occurrences from observations or data from patient medical records, whereas process mapping can be achieved by observing a process or consulting an individual who is embedded in a process that needs to be mapped.

To provide an example, a clinical organization may be struggling with getting meetings scheduled in schools to provide follow-up consultation. The inability to close the loop is leading to poor patient and student outcomes. By using Pareto analysis (Bucchianico, 2016), the organization would be able to save a great deal of time, and implement a more focused intervention, by learning that 50% of their delayed meetings are attributed to School A and another 25% of delayed meetings are attributed to School B. By focusing interventions on these two schools (out of the 15 schools with whom they consult), the organization can intervene on 75% of their total delayed meetings. To further explain the application of Pareto analysis, a second example is provided in Figure 2.

Fig. 2.

A Pareto chart showing the frequency of seclusions in a pediatric inpatient psychiatry unit with categories of age on the x-axis. The primary y-axis has the number (or frequency) of seclusions, and the secondary y-axis has the percentage of seclusions. From this chart, the QI team deduced that 8- and 10-year-old children should be the focus of any interventions, which can result in a 53% improvement in reducing the number of seclusions

Technological

In the world of behavior analysis, there is a strong overlap between high-quality clinical work and high-quality research. Precise language, description, and replicability are key to a solid procedure. A comprehensive protocol should allow for a novel user to effectively implement the desired intervention, even in the case of a variety of unexpected circumstances. In health care QI, this is equally important. Health care systems are incredibly vast and interact with multiple other complex and layered systems. In the effort to reach high-reliability health care, an entire system must be able to implement procedures in the face of other process change, with minimal variability no matter the change of staff, shifts, leadership, or roles. To accomplish such a feat, highly detailed, defined procedures are critical, as are the appropriate interventions to teach the desired behaviors to the necessary individuals in a way that is sustainable and generalizable.

One example of how QI in health care is necessarily technical is reflected in the use of process maps. Using a process map for a new intervention, such as making caring-contact phone calls after psychiatric admissions, allows each clinician to provide a reliable “product” and to complete all necessary steps. It also makes it much easier to later expand the process to other units and areas, with the understanding that the quality of the intervention will be the same and the process will be easily replicable. In the instance of caring-contact (Luxton, June, & Comtois, 2013) follow-up responses, the failure of a novel staff member to complete one step in the process may result in danger to the patient. If a staff member were to fail to execute all steps with a family, following a nonresponse from a caring-contact message, the patient may be acutely suicidal without the family being aware. The absence of a critical step in a given process, or lack of procedural fidelity, could lead to tragic, unanticipated consequences (Figure 3). Process-mapping techniques can be applied to a wide range of health care operations, from handwashing to complex billing operations, to help identify gaps and precise solutions to improve health systems.

Fig. 3.

Example of a QI process map in health care that includes the process and steps for making caring contact

Conceptually Systematic

The mechanisms and principles by which change is achieved in behavior analysis are the foundational pillars of our field. By using both technological descriptions and systematic language, Baer et al. (1968) state that we are best able to represent our interventions. Although the tenet of behavior-analytic change may be theoretically different from that of health care quality, in that terms such as reinforcement and punishment are not integral to the process of improvement, the need for conceptually systematic language remains a constant. Behavior analysis seeks to execute change based on specific mechanisms of behavior, but the effective implementation of these protocols cannot be assumed. Without standardized language and precise definitions of procedures, the execution of a critical intervention may be mismanaged. Similarly, QI is underwritten with the tenet that not all change is improvement, and individuals should carefully control processes to effect change. Both fields use terminology and definitions and describe concepts in a systematic fashion to communicate desired behaviors of staff, minimizing confusion. Within the IHI model for change, one major concept is a plan-do-study-act (PDSA) cycle (Cleghorn & Headrick, 1996). The PDSA approach stresses small, systematic tests of change before large-scale application, to devise the most effective system—and to reduce the possibility of undesired outcomes. This is similar to ongoing phase changes in a single-subject experimental design, as well as the need for program protocols for basic to increasingly complex procedures. Additionally, the concept of high-reliability systems and reaching Six Sigma—a level of consistency in processes and systems equivalent to that of nuclear power plants or airline companies—is reflected throughout the QI literature. The way data are graphically represented, how goals are set, and the approach to selecting the next intervention all relate back to the conceptual foundations of the field, and like behavior analysis, without that anchoring language, the ability to describe the project is undermined.

An example of how QI can be conceptually systematic in health care industries is illustrated by how staff who are members of a project team communicate about the process being targeted. First, conceptually systematic language is used to describe the team. For example, there is the project leader who is responsible for coordination and controlling pace and direction. There is also a project sponsor, whose role is to promote the project at an executive-leadership level to remove barriers and advocate for larger changes or allowances that the project leader may not be able to accomplish. Within the QI field, there is also the concept of a burning platform—similar to an “elevator speech” to communicate the project’s importance. The team all must understand and connect to the burning platform in order to maintain momentum throughout the project. An example of a burning platform can be found in a project promoting parent training during an acute hospitalization:

The first 30 days following an inpatient hospitalization are the most critical. Research tells us that over 30% of suicides that occur, in the year following an acute hospitalization, happen within the first 30 days following discharge. In pediatric populations, family members are the critical link to promote safety and supervision and notice warning signs. By providing real-time education during a psychiatric hospitalization, we are able to equip parents with some necessary tools before discharging to the less controlled home environment.

Throughout the project, the team uses the conceptually systematic language to describe team member roles and the project goals and to conduct the integral processes throughout the project. Terms such as PDSA, control charts, aim statements, and key drivers are all language that forms the foundation of the discipline.

Effective

Change is only meaningful if the interventions produce an effect large enough to have social significance to the individual, group, or system being impacted. Behavior analysis remains an incredibly strong evidenced-based practice because it is not only precise but also effective. Effectiveness is defined not only by the behavior analyst or QI project team but also by those who are impacted by the change that is being made. In QI, the individuals immediately implementing the process, or the patients whose outcomes are directly impacted, are essential to the project team. QI teams generally start their improvement work with writing specific aim statements and defining measures that will indicate if the changes are effective and result in actual improvement. For example, a team working to “increase combined appointments to save families time” defines their measure as “the percentage of patients who see multiple clinics who were able to have two or more clinic appointments in a single day.” These QI measures are collected over time and displayed on time-series charts using SPC methods (Provost, 2011). Good QI work relies on a socially valid measure of what a “good outcome” or “effective process” should be, and change that does not meet the needs of the system or individual is quickly discarded.

How change is initiated is critical within QI and behavior analysis. In many systems, change is initiated through organizational leadership identifying a problem, developing a potential solution, and then implementing that solution, often on a large scale. Unfortunately, this approach does not test for the effectiveness of the proposed solution and often does not take into account the expertise and perspective of the individuals on the front lines who have the most experience with the problem. This can lead not only to inefficiency in impacting the problem but also to possible resistance to change and poor follow-through on the part of staff in future interventions. A project illustrating this concept is the development of parent support groups on behavioral health topics. It is critical that a parent be a part of the implementation team to design the most effective and valid service to facilitate attendance for a new clinical offering. Without the presence of a consumer, elements impacting attendance and engagement could be missed. Families may suggest things such as the groups including dinner to save time, help identify topics that have little interest to the intended consumer population, or make suggestions about the appropriate duration to enhance engagement. Without this input, the interventions can be less effective or ineffective.

Generality

There are several challenges inherent in programming for generalization within the field of behavior analysis. Within QI, the equivalent concept is sustainability. As discussed earlier, small tests in a short duration allow us to observe the impact on the behavior or process, without overly investing resources and time. When making a project meet the requirements for the dimension of generality, or being sustainable, the interventions are integrated into larger and larger areas of the system until there is little variation in the process and it becomes self-sustaining (the new behavioral norm) without the constant input of the QI team. An example of this in the health care field may be handwashing—once the behavior has been taught, it must be sustained under a variety of conditions. This is not an area in which slow drift over time is acceptable. Therefore, the behavior needs to be monitored for sustainability and must continue to generalize across settings (operating room, outpatient clinics) and stimuli (antiseptic rinse, soap, gel, foam).

Discussion

Behavior analysis and health care QI are similar in many ways. We suggest that the nature of our similarities results in ways in which one field can enhance the other. One example is in the scale of the studies. When looking historically at the dissemination of our science, we see that behavior analysis tends to focus on case studies and very specific and procedural applications of small-scale interventions (e.g., Valentino, LeBlanc, & Raetz, 2018). We recognize this observation is not universal, as translational and broad-concept research is becoming more prevalent (Geiger, Carr, & LeBlanc, 2010; LeBlanc, Nosik, & Petursdottir, 2018; Mace & Critchfield, 2010)—though by far the exception and not the norm. Although the dimensions of our field lend themselves to these microtests of change, we postulate this is not exclusively necessary. Health care QI can use behavior-analytic practices as described previously while impacting large groups by changing the behavior of those providing the care. The “case” or individual can be a system, a unit, a business, or an organization—and although the application of the interventions will be quite specific to that entity (much like a child learning a discrete skill), the total number of individuals influenced has the potential to be exponentially larger. Behavior analysts have a right to be proud of the impacts of our interventions and arguably a responsibility to utilize our science on a large scale. Some current research in behavior analysis illustrating the importance of targeting behavior change on a large scale is that on environmental sustainability (Chance & Heward, 2010; Heward, 2013). Taking notes from the QI field may offer a way to expand upon our significant OBM heritage and provide opportunities for behavior analysts. Table 1 offers a list of resources that behavior analysts can access to learn more about QI.

Table 1.

Resources for behavior analysts wishing to learn more about QI

Resource	Type	Access
The Improvement Guide, 2nd Edition	Book	For purchase
Institute for Healthcare Improvement	Website/organization	www.ihi.org
IHI Open School	Self-paced online course	http://app.ihi.org/lms/mycatalogs.aspx
American Society of Quality (ASQ)	Self-paced online courses	http://asq.org/healthweb/
American Society of Quality (ASQ)	Online or classroom-based Lean and Six Sigma certification	https://asq.org/training/lean-six-sigma-black-belt-lssbbd01ms
Agency for Healthcare Research and Quality	Quality tool kits and education	https://www.ahrq.gov/professionals/quality-patient-safety/index.html

IHI and Lean Six Sigma (Sehwail & DeYong, 2003; Series & Kilo, 1998) employ diverse individuals who work in the field. Engineers, physicians, statisticians—all have considerable expertise and skill. Behavior analysts, particularly those in the clinical field, have an enormous skill set to support health care QI. Not only is there familiarity and comfort in the parallel values and concepts, but behavior-analytic science is that of changing behavior. Our tools are frequently referenced in QI; for example, behavioral economics has a strong role in system change (Mehrotra, Sorbero, & Damberg, 2010). At a time in our field when our training routes new professionals most often to the treatment of developmental disabilities, considering the expansion and diversification of our practice seems essential (Normand & Kohn, 2013). We are uniquely positioned to contribute to the field of QI. Embedding behavior analysis within the health care field and facilitating the improvement of measurable behaviors, resulting in enhanced outcomes and safer practices, align with our core identity as a discipline. We encourage behavior analysts to be open, to consider the great things our science can do, and to seek training and opportunities in similar and parallel areas where our skills and abilities can be valued and assist in, of course, saving the world.

Compliance with Ethical Standards

Conflict of Interest

The authors declare that they have no conflict of interest.

Ethical Approval

This study did not involve any human participants.