This field research in outpatient service operations examines original quantitative data on appointments

Transcription

1 Lean Service Operations: Reflections and New Directions for Capacity Expansion in Outpatient Clinics Linda R. LaGanga, Ph.D. Director of Quality Systems & Operational Excellence Mental Health Center of Denver 4141 East Dickenson Place Denver, CO Phone: Published in: Journal of Operations Management, Volume 29, Issue 5, July 2011, Pages Special Issue on Field Research in Operations and Supply Chain Management Available online 10 December 2010 at Abstract This field research in outpatient service operations examines original quantitative data on appointments and analyzes a lean process improvement project that was conducted to increase capacity to admit new patients into a healthcare service operation system. Analysis of 1,726 intake appointments for the year preceding and the full year following the lean project showed a 27% increase in service capacity to intake new patients and a 12% reduction in the no-show rate as a result of the transformation of service processes achieved by the lean project. This study s action research methodology leverages the researcher s involvement in redesigning a service system that greatly improved performance and led to reflection on traditional operations management (OM) approaches to appointment scheduling. The study generates insights about effective alignment of resources, develops new strategies for service operations to respond to no-shows, reveals time-related variables that have been overlooked in appointment scheduling research, and challenges traditional OM scheduling performance measures. We provide recommendations for effective and appropriate use of overbooking and identify avenues for future research to continuously improve and increase the capacity of service operations. Keywords: Lean Process Improvement, Service Operations, Appointment Scheduling, Health Care Policy, No-shows, Overbooking. 1

2 1. Introduction Service operations face the continuing challenge of matching consumer demand with provider supply, and healthcare services in particular have been the focus of much concern and attention. Timely access, responsiveness to patient needs, and availability are high priorities among healthcare system improvements that were called for by the Institute of Medicine (2001). Delays in obtaining healthcare appointments lead to patient dissatisfaction, higher costs, and possible adverse clinical consequences (Green, Savin, and Murray, 2007). Longitudinal studies confirm that delayed access to care is associated with worse outcomes (Fontanarosa, Rennie, and DeAngelis, 2007). Also contributing to higher costs are patients who have appointments but fail to show up for them. One clinic documented a total of 14,000 missed appointments in a year, causing estimated losses over $1 million (Kim and Giachetti, 2006). No-shows are a realistic consideration in clinic operations (Kaandorp and Koole, 2007), where they interrupt the flow of patient care and reduce clinic productivity. Another study estimated that resulting revenue shortfalls could constitute 3%-14% of total clinic income (Moore, Wilson-Witherspoon, and Probst, 2001). These two factors emphasize the importance of appropriate scheduling and resource allocation to meet healthcare demand (Brewer, 2008; Dolan, 2008, Green and Savin, 2008). This study analyzes a real healthcare organization s service operations, reflects on new scheduling approaches that emerged from the organization s first lean process improvement project, and develops new directions to expand service capacity. The study uses action research methodology in which the researcher, who analyzed appointments at six clinic sites within the mental health center organization, was directly involved in the action that realigned system resources, changed operations, and developed new approaches to increase appointment 2

3 scheduling effectiveness beyond those used in prior research. The lean process improvement project, Rapid Improvement Capacity Expansion (RICE), achieved dramatic results in expanding access to services. Analysis of 1,726 intake appointments for the year preceding and the full year following the lean project showed a 27% increase in service capacity to intake new patients and a 12% reduction in the no-show rate as a result of the transformation of service processes achieved by the lean project. From here on, we refer to the recipients of the mental health center s services as consumers rather than patients to conform to recovery-oriented language used in outpatient community mental health care. This paper is organized as follows. The next section presents the background of the initial service capacity problem, which motivates the use of leaner practices. A summary of prior lean practices in healthcare provides context for migrating lean approaches from inpatient to outpatient healthcare settings. The third section discusses our research methodology, provides a description of the service operations where it occurred, analyzes appointment data, and diagrams the flows of the scheduling and appointment processes. The fourth section reports the results of the lean implementation, measured in the year following the startup of the reconfigured system, and describes feedback from participants and consumers. The fifth section reflects on insights developed in the lean project to realign resources and expand capacity, reveals variables that have been overlooked in prior appointment scheduling research, and considers the need to examine and revise the traditional emphasis of appointment scheduling research and its performance metrics. We conclude with a summary and identify new directions to continuously improve and increase the capacity of service operations. 2. Service capacity problems, solutions, and reactions 3

4 The problem of balancing patient wait time and provider idle time has received considerable research attention for decades (Bailey, 1952; Cayirli and Veral, 2003; Gupta and Denton, 2008; Ho and Lau, 1992; Klassen and Rohleder, 1996). Such studies have focused on managing the variability in appointment service times. However, many providers have told us that they had significant control over the duration and variance of service times, but had less control over patients attendance. In healthcare, no-show rates vary widely from as little as 3% to as much as 80% (Rust, Gallups, Clark, and Jones, 1995). A 50% no-show rate was reported in a hospital outpatient clinic (Kim and Giachetti, 2006). During our outpatient tracking activities, we found approximately 30% of scheduled patients did not show up for their appointments. In outpatient clinics, matching doctor capacity to uncertain daily demand has been studied through alternative appointment scheduling techniques. These include allocating capacity for open access through short-term and same-day scheduling (Kopach, DeLaurentis, Lawley, Muthuraman, Ozsen, Rardin, Wan, Intrevado, Qu, and Willis, 2007; Liu, Ziya, and Kulkarni, 2010; Qu, Rardin, Williams, and Willis, 2007; Robinson and Chen, 2010), handling walk-ins (LaGanga and Lawrence, 2008b), and the investigation of overbooking to mitigate the effects of patient no-shows (LaGanga and Lawrence, 2007b). Such appointment scheduling has emphasized the allocation of existing capacity. However, for the most part, these models have not fully addressed the pressing need to expand capacity to increase patient access to services. In the next sections we consider approaches that target capacity expansion more directly and are further developed through this study Overbooking When there is a significant chance of patient no-shows, overbooking can be an important strategy for improving patient access and stabilizing revenues (Muthuraman and Lawley, 2008). 4

5 Overbooking models have been studied extensively in airline and other transportation applications in the context of revenue management (Barnhart, Belobaba, and Odoni, 2003; Lieberman, 2004, 2005; Toh and Raven, 2003; Van Ryzin and Talluri, 2003; Weatherford and Bodily, 1992). More recently, LaGanga and Lawrence (2007b) developed an overbooking utility model to capture the trade-off between the expected benefits of serving additional patients and the costs of patient waiting times and provider overtime. Controversy about overbooking medical appointments is related to concerns with achieving both customer service and managerial efficiency (Kimes, 1994) and to customers perceptions of fairness (Wangenheim and Bayón, 2007). When customers perceptions of service fall short of their expectations, this gap leads to a negative perception of service quality (Zeithaml, Parasuraman, and Berry, 1990). Suggestions for making the practice more acceptable to customers include educating customers (Kimes and Noone, 2002) and compensating patients for excessive waiting time (Dolan, 2008). The theme of fairness was repeated often among 32 on-line responses from patients and doctors (WSJ.com Forums, 2008) to a Wall Street Journal article (Brewer, 2008) that critiques the practice of overbooking. Next we consider how lean process improvement can help resolve this controversy and improve healthcare services Lean process improvement in healthcare Quality management practices such as continuous quality improvement, total quality management, six sigma, and the Malcolm Baldrige National Quality Award have been adapted from manufacturing and implemented successfully in healthcare (Gowen, McFadden, Hoobler, and Tallon, 2006). Many of these traditional quality improvement practices focus on improving processes, reducing variability, and identifying root causes of problems. Lean practices focus on patient flow, value-stream mapping, and kaizen events (Bernstein, 2008). The focus on the 5

6 reduction or elimination of waste (unproductive effort that does not create value for the end customer) led to the term lean to describe systems that consumed fewer resources and delivered superior results (Womack, Jones, and Roos, 1991). Although lean principles were developed and applied initially in manufacturing, they are being used successfully in healthcare. Lean approaches are included in the quality improvement recommendations for organization-wide system transformation in public health (Riley, Moran, Corso, Beitsch, Bialek, and Cofsky, 2010). There are many documented cases of successful lean projects conducted in hospitals since the 1990s to improve patient care by reducing errors, waiting times, and costs while improving interdepartmental interaction and employee satisfaction (Graban, 2008). However, only a few prior studies have been found that refer to lean principles in improving outpatient service operations. The migration of lean principles from manufacturing to healthcare settings was demonstrated by hospitals such as Denver Health Medical Center (Nuzum, McCarthy, Gauthier, and Beck, 2007). In 2006, the hospital saved about $2.8 million without reducing staffing or patient care (Shanley, 2007). ThedaCare, a hospital system in Wisconsin, also initiated its lean improvement by studying manufacturing systems and saved $10 million in 2005 due to its lean programs (Matzek, 2006). Lean approaches have produced other measurable improvements such as shortened time to admit patients at Prairie Lakes Healthcare (Homolo and Fuller, 2008). Three successful lean applications at the University of Pittsburgh Medical Center improved patient flow in an emergency intake process, an outpatient foot-and-ankle clinic, and the discharge of cancer patients from inpatient to ambulatory services (Martin, Greenhouse, Kowinsky, McElbeny, Petras, and Sharbaugh, 2009). Lean approaches at St. Luke s Hospital in Houston, 6

7 Texas, reduced variability, standardized healthcare, and improved quality by meeting national standards of care 100% of the time (Cook, 2008). In this study, lean techniques are applied to improve the scheduling and delivery of outpatient services in a mental health center. Through January, 2010, the mental health center conducted nine lean rapid improvement events. The resulting lean improvement projects include improvement goals focused both on direct clinical services and on indirect processes that support service delivery. For example, one of the projects increased access to services by streamlining administrative elements of the intake process and significantly reducing clinical services and documentation that did not provide value for consumers or payers (LaGanga and Lawrence, 2009). The center s first lean project, Rapid Improvement Capacity Expansion (RICE), is described in the sections below. 3. Methodology This field research uses original data from the mental health center s service operations to study the impact of the organization s first lean process improvement project. The methodology is referred to as action research because the researcher was involved in selecting and facilitating the RICE lean project. As described by Coughlan and Coghlan (2002), the researcher was not only observing action, but was making action happen and reflecting on it to develop knowledge. The study was facilitated by a robust electronic health record (EHR) system that was implemented throughout the organization s multiple clinics several years earlier. The EHR is a valuable asset to measure the performance of the service operations, test assumptions, and analyze the impacts of changes. It captures and stores data on every scheduled and actual service event. Each event record includes the date, time, appointment disposition (kept or no-show), service activity, and identification of the consumer and provider. One year of initial 7

8 retrospective data was obtained from the EHR about scheduled appointments and no-show rates for intakes and evaluations. This data set was used to guide the selection and focus of the first lean process improvement project. The researcher used the data collaboratively with the RICE team to explore alternatives and take action, illustrating the nature of action research as an emergent process that unfolds through events that can be planned but cannot be designed in detail in advance (Coughlan and Coghlan, 2002). The goal of the lean project was to improve the consumer access process, which operates as an overall system composed of multiple clinical teams that operate as interdependent components. Thus, this action research methodology was used to study the single organizational case of the mental health center and its operations, which are described in the next section. 3.1 The organization and its operations The subject organization is a large, non-profit, outpatient community mental health center with 18 outpatient clinics located throughout Denver to serve adults, children, and families. The center is not a hospital and does not provide inpatient services directly. It does coordinate and collaborate with the hospitals that serve the mental health center s consumers when inpatient services are needed. The 18 outpatient clinical teams and programs provide therapy, psychiatry, and medication services. Each clinic represents a team serving a specific population based on consumers age groups, service type and intensity, and need for assistance navigating community resources. No-show rates vary due to such differing characteristics among consumer populations (Bean and Talaga, 1995; Rust et al., 1995), differences among individual clinicians, and varying locations of clinical sites. Capacity constraints posed a significant problem, a common situation in such service systems. Nearly 80% of American children and more than 67% of American adults who need 8

9 mental health and addictions treatment services do not receive them (The National Council for Community Behavioral Healthcare, 2010). Although about 7 new consumers were admitted each day, approximately 17 others were turned away due to lack of capacity caused by funding limitations. Appointments and their attendance status (kept or no-show) were tracked throughout the entire mental health center to identify and attempt to reduce high no-show rates. The initial focus of this study was on the capacity and process to access services, which includes two specific types of appointments: intake assessments and psychiatric evaluations. The intake assessments are conducted by clinicians with master s-level or higher degrees in clinical social work, counseling, or psychology. Many are licensed or have additional certification in clinical specialties such as addictions counseling. During the intake assessment the clinician assesses the consumer s current presenting problem, documents treatment history, and formulates an initial diagnostic impression. The clinician also assesses the consumer s available resources and goals for treatment to develop an initial treatment plan collaboratively with the consumer. Before a clinical intake assessment, intake clinicians also meet with consumers for a less clinical orientation to complete paperwork and explain procedures. Most of the orientation and intake assessments are conducted at six of the mental health center s 18 outpatient clinics. After the intake assessment, consumers see a psychiatrist in a later appointment. The psychiatrist evaluates behavioral and physical symptoms, assesses the consumer s level of functioning, confirms the diagnosis, and prescribes and provides education on medications. Psychiatrists are doctors, MDs or DOs with certification in psychiatry. The psychiatric evaluations are conducted at all of the 18 outpatient clinics. Waste could occur if consumers did not complete the process and therefore did not benefit from treatment, even though they might have spent some time and consumed clinical resources in earlier 9

10 appointments. Some clinics were operating with improved configurations that eliminated some waste, but no system-wide improvement project had been implemented. Analysis in preparation for the lean improvement project, described below, confirmed and quantified the extent of the waste caused by no-shows. In the autumn of 2007, the mental health center s manager of access services participated in a Denver Health lean event that demonstrated the potential for improvement through immediate changes in key service processes. The experience motivated him to streamline the process for scheduling initial intakes and subsequent psychiatric evaluations at the 18 outpatient clinics. His subsequent analysis of 1,769 appointments revealed that 314 intake and evaluation slots (17.75%) had been wasted due to no-shows. Table 1 shows the appointments and their results (shows and no-shows) on average for each of the 18 clinics. Insert Table 1 about here The manager consulted with the researcher, who worked with him to launch and facilitate the center s first lean improvement project to reduce no-shows and increase intake service capacity. In January, 2008, the manager and researcher convened a Rapid Improvement Capacity Expansion (RICE) cross-functional team, modeled after Denver Health s Rapid Improvement Events (Nuzum et al., 2007). The RICE team included as stakeholders several intake clinicians, clinicians at the access center, program managers and directors, and consumer representatives. Psychiatric evaluation appointments were included in the early analysis of the no-show problem; however, the project and subsequent analysis were realigned to focus on the earlier points of 10

11 contact in the admission and intake process. The team members prepared to participate in the RICE project by reviewing the center s existing scheduling data along with prior studies and papers on no-shows and overbooking, provided by the researcher. 3.2 Initial data analysis The data in Table 1, sorted in decreasing order by number of appointments for each clinic, show in the seventh column, Cumulative Appointments %, that 80% of intake and evaluation appointments occurred at four clinics. Clinics C and D account for over half of all the intake and evaluation appointments. The high variability in appointment volume shown in Figure 1 is attributable to population diversity, contracts, and funding sources. These factors impose constraints on the number of consumers served at each clinic. Figure 1 also shows the high variability in no-show rates. This reflects not only consumer diversity regarding mobility, transportation issues, and initial commitment to engaging in services, but also the varied operational practices at each clinic. Insert Figure 1 about here There is no apparent relationship between number of appointments and no-show rate. Regression analysis with scheduled appointments as the independent variable and no-show rate as the dependent variable for the 18 clinics produced a coefficient of determination (R 2 ) of only.06 and the regression coefficient of no-show rate was not significant (P =.349). Although clinic D has one of the two largest appointment volumes and one of the two lowest no-show rates, clinic K has the third smallest appointment volume and the highest no-show rate. Although all clinics had 11

12 implemented the same electronic medical record system several years earlier, there were some inconsistencies between clinics with regard to front desk recording of appointment disposition and clinician coding of activities that determine whether appointments are included in the intake and evaluation data set. The mental health center is addressing these issues through the RICE project and subsequent lean improvement projects. Analysis of no-show data shows differences between clinics, days of week, and times of day. No-show rates varied from 0 to 55% among all 45 combinations of hour and day for 9 scheduling hours and 5 days of the week. Chi-squared analysis of the 45 combinations revealed that the proportions of no-shows differed significantly among hour and day combinations (p <.001). Flexibility to vary schedules by hour of day combined with day of week could increase the yield of patients served. For example, Tuesday and Thursday appointments scheduled at 10:00 a.m. have low no-show rates (8.11% and 9.09%, respectively, compared to the average 13.98% for all days). However, consumers show up for their appointments at 9:00 a.m. on Mondays and Thursdays at higher rates than on Tuesdays. Shifting additional staffing to these better-yielding day-time slots might increase patient access even without overbooking. Further investigation is needed to determine the reasons for the variation in show rates. Hourly analysis of the data in Figure 2 shows how no-show rates vary with time of day. The data suggest that scheduling intakes at 9 a.m. and 4 p.m. would yield the highest number of kept appointments. Scheduling more appointments during the 4 p.m. hour in the clinic session could be an opportunity to increase service yield, as only 1.4% of the total appointments currently occupy that time slot. This practice would, however, require providers to work later and increase the average amount of overtime incurred. 12

13 Insert Figure 2 about here Analysis of appointment and no-show data by day of week, as shown in Figure 3, revealed that the highest number of appointments was scheduled on Tuesdays and the lowest noshow rate occurred on Thursdays. Mondays had the highest no-show rates. Insert Figure 3 about here The data were valuable in determining how to improve the scheduling process to serve more consumers. Differences between clinics, days of the week, and hours of the day revealed important opportunities to realign scheduling processes, shorten lead times, and leverage effective scheduling practices for increased service access throughout the clinics. In the sections that follow, these concepts are explored through work flow analysis and descriptions of the intake process changes, which include coordinating and shifting activities to more favorable days and times. 3.3 Work flow analysis The initial intake of a consumer occurs at one of the clinical intake teams. Each of these teams allocates appointments every week for these intakes and notifies the central access center of appointment availability. A consumer seeking access calls the access center, where access clinicians screen callers, check for appointment availability and attempt to schedule the consumer with an appropriate clinical intake team. Figure 4 shows the work flow of the process of allocating appointments for initial intakes. 13

14 Insert Figure 4 about here Matching appointment supply with service demand is based on the needs and characteristics of each caller and the availability of appointments with intake clinicians on teams with appropriate training and expertise. Service intensity variation and the need for community-based supportive services are also considered. If an appropriate appointment is available when a consumer calls, the access center schedules it, removes it from the inventory of available appointments, and provides appointment information to the consumer. Otherwise, prospective consumers are informed that they can try again at a later time. A wait list is not maintained because a high proportion of callers are homeless or in unstable living situations, making it difficult to reach them later by phone. Prior to the lean improvement project, available appointments were released weekly from each clinical intake team to the access center s central inventory of available-to-schedule appointments. Once the weekly inventory of appropriate appointments was depleted, consumers would need to wait until the next week for a chance of gaining admission to the system. Appointments were not booked any later than two weeks from the time of the call because it was believed that people would be less likely to show up for appointments booked that far in advance. Empirical studies have confirmed that the longer the delay between patient request and date of appointment, the greater the chance that the patient will cancel or not show up (Gallucci, Swartz, and Hackerman, 2005; Green and Savin, 2008; Liu et al., 2010). Some intake clinics held paperwork/procedure orientations on one day and then the clinical intake appointments on a 14

15 later day. Some consumers showed up for the orientation, but did not show up for the subsequent intake. In an attempt to compensate for no-shows, clinical intake teams intuitively allocated extra intake appointment slots without specified clinicians. The assumption was that at the actual time of the appointment, a clinician who had been assigned to an appointment but was idle because of a no-show would be available to cover the appointment. Intake appointments were scheduled simultaneously with different clinicians, each of whom was scheduled for only one intake appointment in a morning. There was no overbooking of specified clinicians. The number of no-shows often exceeded the extra allocated slots, so clinicians frequently experienced idle blocks of time, which they often used to catch up on paperwork. 3.4 Intake process changes The RICE team applied lean techniques and followed a structured analysis to determine how to improve the process. The sequence of activities consisted of these steps: analyze the current process (as it existed prior to the RICE project), describe the desired target state, identify the gaps between the current process and the target state, and develop solutions to narrow or eliminate the gaps. Throughout these steps, insights were generated and recorded about the process and how it could be improved. The team referred to the list of insights and used it in developing solutions. The changes in patient flow are demonstrated below by comparing the process before and after the RICE project. In analyzing the prior process, the team identified several sources of wasted clinical capacity that occurred when intake clinicians and psychiatrists were idle during scheduled appointments because consumers failed to show up, often after multi-day delays between different types of appointments. This flow of consumers for the prior process is illustrated in Figure 5. 15

16 Insert Figure 5 about here The RICE team chose an alternative strategy that combined no-show reduction and a nontraditional form of overbooking, based on batching consumers into group orientations and overbooking extra consumers into these groups. Individual intake appointments immediately follow. The team adopted this configuration after learning about it during their analysis of the prior process, when a RICE team member explained that it had already been developed and implemented at clinic D, which he managed. This flow of consumers for the improved process, shown in Figure 6, boosts the intake service yield by eliminating the delay caused by scheduling intakes on a different day after the orientation. This delay had been largely responsible for the clinic s approximately 20% no-show rate. Insert Figure 6 about here Assigning a single clinician to conduct group orientations allows the remaining clinicians to serve other consumers. Pre-doctoral psychologist interns are deployed at little cost to handle intakes if overbooked clients showed up. With this configuration, clinic D had increased its total intakes per year by 28% without increasing staffing or other expenses. Processes that had not been handled consistently across intake teams were standardized by adopting Clinic D s successful approaches. Although some of the other intake teams did not have pre-doctoral interns, they could still create flexible capacity by using qualified staff 16

17 members who were not initially scheduled to perform intakes but could be deployed if needed. These available-on-demand providers were typically clinical program managers who have the training and credentials (e.g., clinical social worker, counselor, psychologist, psychiatric nurse) to perform intake assessments but spend most of their time in managerial and administrative activities. Another change was to move initial orientation group schedules from once a week, mostly on Mondays, to twice a week on Tuesdays and Thursdays. This schedule allows for notification to the consumer the day before the appointment. It also increases the availability of same-week access because a consumer calling during the first half of the week can be scheduled for Thursday if appointments are available. The overbooking that was done to compensate for no-shows was changed from an intuitive guess to a systematic determination of the number of appointments to schedule. For A = appointments scheduled, N = target number of kept appointments, and S = show rate, the number of allocated appointments was set to A = N/S to yield the target number of kept appointments on average (LaGanga and Lawrence, 2007b). The availability of accurate no-show data through the EHR helps clinics monitor and adjust their overbooking levels. In discussing no-shows during the gap analysis, team members shifted their analysis from asking why consumers did not show up to the more positive focus on why people kept appointments. The question was posed to one of the consumer representatives, who had admitted that he did not show up for his first several appointments. He explained that he finally attended his appointment after his clinician conveyed a genuine sense of caring for his wellbeing. The clinician s efforts to uncover the consumer s concerns and provide reassurance about 17

18 getting to an unfamiliar place influenced the consumer s decision to show up. The immediate response by clinicians on the team was to develop a welcome letter and to call consumers the day before appointments to welcome them, provide transportation information and directions, tell them what to bring (e.g., insurance cards, prescription information), answer questions, and emphasize the importance of keeping the appointment. The RICE project changes began in March, The team conducted progress reviews two weeks before the center-wide cutover and one month after startup. The process started up smoothly and clinicians were enthusiastic and supportive. Intake clinicians reported that preappointment phone calls were successful in encouraging consumer attendance at appointments. In instances where the consumer cancelled during the phone call, the appointment slot was immediately made available and rescheduled with another caller. 4. Results 4.1. Initial results Following its start-up month, initial analysis showed encouraging results with a 10.6% increase in appointments that were kept by consumers for initial intake assessments and subsequent psychiatric evaluations (LaGanga and Lawrence, 2008a). The initial analysis was adjusted later (LaGanga and Lawrence, 2009) to concentrate specifically on the initial intake appointments, which represent the admission process and were the focus of the RICE project. Results of the refocused analysis, which compared the first month of operation to the same month in the previous year, are summarized in Table 2. The number of scheduled intake appointments increased from 91 in March, 2007 to 111 in March, 2008, a 22% increase. When combined with a 5.67% decrease in no-show rate the result is a 30.26% increase in the number of 18

19 intake appointments kept (from 76 to 99). This suggests a substantial and immediate increase in service capacity. The improvements reflect systemic changes, including an increase in the number of appointments scheduled and a decrease in no-show rates, along with a redesign of the process of scheduling and conducting intakes. Additional lean analysis of estimated reduction in non-value-creating wait times is shown in the Appendix. The next section describes feedback during and after a full year of operation with the improvements. The analysis of capacity increase in that section provides more conclusive results because it compares data from a full year of operation after the improvements to the data from the previous year. Insert Table 2 about here 4.2. Ongoing results Verbal feedback about the RICE implementation was very positive from the time the team began work on the rapid improvement event and throughout the year that followed. All participants responded positively to surveys about the changes they were implementing in their work processes. Feedback about timing and scheduling was used to modify the start and end times of future rapid improvement events to make them more manageable for the participants. At a follow-up meeting held after one month of operation, clinicians said they were providing better customer service and were more responsive to consumer demand for services because more intake appointment slots were available. They enjoyed creating a welcoming experience for new consumers and saw the results when fewer no-shows occurred. Educating 19

20 new consumers on the value of the intake appointment and asking them to call ahead of time if they needed to cancel appeared to be effective. Fewer appointment slots were wasted because cancellations were communicated to access clinicians, who refilled them. After a year of operation, feedback from the project team confirmed that the changes had been sustained and the new process was operating as planned. Feedback also was obtained from consumers. The consumer participants in the project represented the mental health center s Consumer/Staff Partnership Council, which involves consumers in identifying and participating in actions that improve service delivery. The director of quality interviewed 15 consumers in October, 2009, to discuss the project and their reactions. Consumers who were admitted prior to the process change indicated that their initial admissions had been challenging and scary. Consumers admitted after the change reported a positive experience with orientation followed immediately by the intake assessment. When specifically asked about the practice of overbooking later individual appointments, consumers said that even if it caused them to wait longer in the waiting room, they could accept that inconvenience if they received quality time and attention from their doctor or clinician when the appointment began. They also found it helpful when someone at the clinic communicated with them to let them know how long an expected delay would be. Table 3 compares 816 scheduled appointments for the year March, 2007, through February, 2008, to the 910 appointments that were scheduled the following year, March, 2008, through February, The impact of the lean project is shown as the calculated differences before and after implementation in the numbers of scheduled appointments, kept appointments, no-shows, and show rates. The total number of appointments scheduled in the year after the lean improvement is 11.52% more than the year prior, with a 26.6% increase in the number of kept 20

21 appointments compared to the year before. This represents a valuable capacity increase of 187 additional people who were able to access needed services, without increasing staff or other expenses for these services. A t-test on the increase in kept appointments confirmed that the capacity increase was statistically significant (p <.05). Insert Table 3 about here Insert Table 4 about here Table 4 highlights some interesting system dynamics about the impact of the lean improvement. Overall capacity improvement resulted not only from scheduling more appointments but also from increasing the probability that scheduled patients actually showed up for their appointments, which is reflected in the reduction of no-shows by 93 after implementation. Clinics C and H actually scheduled fewer appointments but increased the number of people served by decreasing the number of no-shows. Figure 7 shows how the system operation was transformed by changing the distribution of appointments among days of the week. As indicated in early data analysis, show rates differ by day of the week. Shifting initial appointments from Monday to Tuesday allowed clinicians to create a more effective and welcoming experience by conducting reminder calls to consumers one day prior to their appointments. Adding intake appointments on Thursdays provided more options and opportunities for new callers to obtain an appointment. Such matching of consumer 21

22 demand to patient supply has been used in other areas of service operations such as call center management (Gans, Koole, and Mandelbaum, 2003). Insert Figure 7 about here Figure 7 shows that the system performs better in scheduling intakes and reducing noshows after implementation. Cross-tabulation analysis of data by attendance status confirmed the statistical significance of the overall decrease in no-show rates (p <.01). Analysis of variance tested the effects of the RICE implementation on appointment attendance for each day of the week, and revealed that the difference in no-show rates before and after implementation was significant for every day of the work week except Fridays, which had the same number of scheduled appointments as Monday (p <.01 for Tuesdays, Wednesdays, and Thursdays, and p <.05 for Mondays.) This suggests an opportunity to increase capacity by scheduling more Friday appointments, finding out what causes no-shows on Fridays, or overbooking on that day. 5. Discussion This study used action research to examine the operations of a single case study of a multi-clinic outpatient healthcare organization. Such single cases can provide value in developing understanding and explanations for some of the findings (Flynn, Sakakibara, Schroeder, Bates, and Flynn, 1990). The study of the RICE project demonstrates lean process improvement s impact in improving operations in a real outpatient healthcare organization. The process of the RICE team and researcher working collaboratively generated important insights 22

23 and novel approaches. Here we discuss how these insights contribute to academic understanding and influence research in service operations. The first insight is that no-show rates can vary within clinics by time-related factors such as time of day and day of week, which impact scheduling operations and performance. Much of the change implemented in the RICE project was based on leveraging knowledge about timerelated variations in no-show rates to realign the scheduling of services to more advantageous times. These time-related factors and their impacts have not been included in prior appointment scheduling research. The increasing use of EHR in outpatient healthcare provides opportunities for other researchers to analyze actual appointment data, test and replicate time-related effects on no-shows, and incorporate these factors into new scheduling approaches and models. Even without having actual appointment data, simulation researchers can utilize this study s findings of significant differences in show rates by day and time to conduct sensitivity analysis and analyze the robustness of new scheduling models. Such future research might also assess the suitability of proposed scheduling rules to realistic clinic operations. The second insight is about scheduling and overbooking. The researcher s initial education of team members showed that they held the popular belief that overbooking is synonymous with double-booking two (or more) patients into the same individual appointment. This practice can achieve positive net utility in serving patients, but it can be problematic especially when service time is long or highly variable (LaGanga and Lawrence, 2007b), which are characteristics of the center s intake appointments. Although double-booking may be easy to implement in manual or electronic scheduling systems, simulation studies have demonstrated that this approach favors providers but accumulates high levels of wait times for patients (LaGanga & Lawrence, 2007a). To more fairly distribute the consequences of overbooking 23

24 between patient and provider, some prior research (LaGanga and Lawrence, 2007b) uniformly compresses the time between individual appointments (the average service time) by a factor equal to the show rate. However, the RICE team s novel approach, of batching consumers when feasible to receive non-individualized services simultaneously, has not been considered in the scheduling literature. Although the operational advantages are obvious with regard to efficient use of provider resources, the full effects remain to be tested. Future research could identify clinically-appropriate opportunities to implement batch service delivery and test the effects on service quality, consumer perceptions, and healthcare outcomes. The third insight is about performance evaluation in scheduling research, which typically compares scheduling rules based on a weighted cost metric composed mostly of patient wait time and provider idle time as established by Bailey (1952) and other researchers (Ho and Lau, 1992; Klassen and Rohleder, 1996). In addition, more recent research includes provider overtime as another cost (Kaandorp and Koole, 2007; Robinson and Chen, 2010). Our study suggests that these components might not accurately capture the true costs and benefits valued by consumers, providers, and other stakeholders. For instance, the consumers in our study suggested in a focus group that they were more concerned with the quality of their time with their doctors than with the amount of time they waited at the clinic for the appointments to begin. From the provider perspective, some idle time is viewed as beneficial to accomplish other tasks such as documenting their clinical service delivery or perhaps conducting other consumer services, such as coordination and communication with other providers. Such activities are not reflected in the appointment data but are important in overall clinical service delivery. Moreover, the RICE team s adoption of flexible capacity to handle overbooked consumers is based on utilizing idle time from an auxiliary pool of providers, which avoids incurring overtime for the primary 24

25 provider. A problem with traditional counting of idle time as a cost is that providers salaries are based on their daily presence at the clinic, not on the actual utilization of their time providing direct services. Furthermore, in healthcare, reimbursement to providers by insurance and other payers is based on the number of services delivered or number of patients served rather than on the actual time spent with patients. Instead of using a traditional weighted-cost performance measure, the success of the RICE project was evaluated by considering the change in total number of consumers served to demonstrate the increase in overall capacity. In this case, consumer access to timely services was increased without increasing staffing or other costs. Thus, we recommend that outpatient healthcare scheduling research examine the performance measures and short-interval time units (hours or minutes) that have been its traditional focus and consider broader, daily or longer-term performance measures that may better reflect true capacity expansion. Consumer preferences should also be studied to model relative perceived costs and benefits as trade-offs between elapsed time from request to appointment (in days) and the wait time in the clinic (in minutes or hours). The current emphasis on healthcare reform and policy analysis creates urgency to develop and evaluate solutions that accurately capture the complexity of high-quality, cost-effective healthcare service delivery. 6. Conclusions and future directions By incorporating actual operations data to develop and test new service processes, the RICE lean process improvement project rapidly expanded service capacity. It demonstrated that no-shows can be managed effectively through consumer engagement to improve attendance, along with creative use of overbooking to serve consumers in orientation groups followed by use of flexible capacity. Scheduling orientation and intake assessment appointments on the same day and realigning appointments to better-attended days also proved effective. 25

26 The service operations system was quickly transformed to achieve measurable performance improvement that motivated further organizational commitment to conducting lean process improvement projects. Encouraged by early positive results of the RICE project, the organization updated its strategic plan with an objective to conduct one lean improvement project every month. Projects were selected for their potential to deliver significant operational and financial improvements, but this created an unexpected lean paradox. Lean techniques help project team members quickly reach agreement on the target state and required actions for implementation, but completing the implementation was lengthy and complex for the troubling and persistent problems selected for improvement. These problems tend to share a common theme of lack of systematic processes and inadequate technological support for their operations. Thus, the streamlining of workflow to eliminate front-end delays introduced bottlenecks in the back-end of the operations where implementation required the skills and efforts of a limited number of programmers. The mental health center recognized the need to adapt its approach, modified its goals to sequentially complete lean projects, and carefully prioritized its scheduling and selection of candidate projects. In future research, prioritization guidelines could be developed by studying the factors that predict the overall and enduring value of projects. The lessons learned in this action research approach could be applied to other organizations. The outpatient organization in this study was inspired by an inpatient hospital to adopt lean practices after a successful lean event that provided important benefits to their shared customers. Reducing costs is crucial to expanding access to necessary services. The project did not eliminate no-shows entirely, so there remains a need for research to study and continuously improve appointment scheduling practices. The study develops avenues for future research by revealing time-related variables associated with no-show rates as well as capacity-oriented 26

27 performance measures targeted toward serving more consumers. It also suggests other consumer-oriented variables related to consumer preferences, clinical outcomes, and perception of service quality. Recently, the organization began to track the time of each appointment request and is analyzing the elapsed time to the actual appointment, along with patient characteristics and recovery outcomes, in data mining projects to predict no-shows and optimize scheduling. This deepens understanding of no-shows to further improve consumer access and, ultimately, clinical outcomes. Improved outpatient access can reduce the need for more expensive inpatient hospital services. This example can encourage similar shared learning, continuous improvement, and policy revision in other settings. Acknowledgements I appreciate the constructive comments and suggestions of the referees and editors. I gratefully acknowledge the special issue editors for ensuring clarity and drawing out the strengths in the study s contributions to the literature. Special acknowledgement is due to Craig Iverson and the RICE team for successful launch of lean process improvement at MHCD; thanks to Dawn Hinsvark, Cecilia Richey and Antonio Olmos for help with data and statistical analysis; and thanks to Ric Durity, Lenore Ralston, and Eric Brody for editing assistance. References Bailey, N.T., A study of queues and appointment systems in hospital out-patient departments, with special reference to waiting-times. Journal of the Royal Statistical Society, Series B 14 (2), Barnhart, C., Belobaba, P., Odoni, A.R., Applications of operations research in the air transport industry. Transportation Science 37 (4), Bean, A.B., Talaga, J., Predicting appointment breaking. Journal of Health Care Marketing 15 (1),