EXECUTIVE SUMMARY. Introduction. Methods

EXECUTIVE SUMMARY Introduction University of Michigan (UM) General Pediatrics offers health services to patients through nine outpatient clinics located throughout South Eastern Michigan. These clinics face scheduling problems caused by seasonal trends in demand for scheduled visits. Scheduled visits can be categorized into three types: return, urgent and well. Urgent visits are often scheduled same-day or next-day whereas well visits are often scheduled weeks or months in advance. The clinics experience large variation in the demand for urgent and well visits throughout the year. This variation in demand often results in overbooking of patient visits or underutilization of physicians. The client would like understand the trends in demand for urgent, return and well visits in order to better match the availability of appointments to demand. The University of Michigan project team 5 investigated three UM General Pediatric clinics: 1) Briarwood Center for Women, Children and Young Adults; 2) East Ann Arbor Health and Geriatric Center; and 3) West Ann Arbor Health Center. These clinics differ in size, number of physicians and number of patient visits. The team conducted an in-depth analysis of the current scheduling system and historical scheduling data to understand the variability and trends in the demand for sick, urgent and well visits. The team summarized the observed trends and provided recommended scheduling templates that reduce overbooking and maximize physician utilization. This executive summary is a brief overview of the team s findings. Methods Determining trends in demand for visits. In order to identify trends in demand for the various visit types, the team will define the following statistics to analyze: 1) the number of each category of slots scheduled each day for each physician and 2) the utilization of each physicians time for each day. Utilization for each day will be defined as the percent of a physician s total clinic time that was scheduled for patient visits. Once these statistics are gathered, the team will perform ANOVA analyses in Minitab to identify statistically significant difference in the distributions across various factors including month, day-of-the-week and time of day. Creating Seasonal Scheduling Templates. The Assistant Director of Clinical Services at Briarwood is interested changing the scheduling template throughout the year to better accommodate seasonal changes in demand. The team will use the results of the ANOVA analysis to recommend how these seasons should be defined. This includes how many seasons should be defined and when each season occurs. These seasons may not line up with the seasons of the year. Furthermore, the team will recommend session limits on well visits and number of urgent visits to allocate for each of the seasons for each clinic. Forming conclusions and recommendations. The team will use the identified trends from the analysis to make conclusions about the demand for visits overall. Furthermore, the team will apply these conclusions to develop seasonal scheduling templates. The team will report these conclusions in the final report. 1

Percentage of total time Average percent of physician time Findings and Conclusions Based on the findings, the team concluded that there were notable trends in demand by month and day-of-week. Also, from the interviews with scheduling staff, the team recognized the key problems existing within each clinic associating with scheduling system. Trends in demand for visits by month. The team calculated the distribution of total visit time for each of the clinics by the 36 months across the three years. Since the clinics are different in sizes, the team was not able to add up the total number of visits. To take out the effect of clinic size, the team decided to use percentages. Figure 1 shows the average percent of physician time for each month for all three clinics. 60% 50% 40% 30% 20% 10% 0% Jan, 09 May, 09 Sep, 09 Jan, 10 May, Sep, 10 10 Month Jan, Figure 1: Demand for visits by month May, Sep, Urgent Visits Return Visits Well Visits No Visits Scheduled From Figure 1, the team concluded that there are noticeable seasonal trend in urgent and well visits while the seasonal trend in return visits are less pronounced. In addition, the team also calculated the idle time within each of the clinics and determined that it was relatively constant throughout each month. Trends in demand for visits by day-of-week 60% 50% 40% 30% 20% 10% 0% Briarwood Urgent Visit Well Visit Return Visit The team used the percentages of total time spent on each visit type so that the trend in demand is independent of the number of FTEs. From the graph, it was concluded that there was a fairly strong trend in demand by day-of-week. For one session limit or for one FTE, the relative Figure 2: Demand for visits by day-of-week 2

demand for urgent visits is highest on Monday, then gradually decreases but slightly rises back on Friday while the relative demand for well visits moved with a completely opposite pattern. Physician staffing level and utilization The team interviewed various clinical staff from the three clinics and concluded that each clinic has specific problems that would like be addressed. West Ann Arbor Health Clinic: Clerks are having difficulty in categorizing urgent visits and thus scheduling a patient under an urgent visit when it is not necessary. Briarwood Center for Women, Children and Young Adults: Clerks are having trouble with scheduling when physicians leave for vacation or sabbatical. East Ann Arbor Health and Geriatric Center: The medical director does not know when to switch their scheduling schedules. Switching schedule is based completely on intuition. Recommendations Our recommendation is to use proposed ranges of both the urgent and well visits per FTE for each month as shown on the seasonal scheduling template below. Team 5 believes these guidelines will allow each clinic to choose the actual patient demand for the different visit types. Our recommendation also promotes the integration of physician intuition when making the scheduling templates under different circumstances such as late or early flu season. Team 5 also recommends using the fact that the capacity for urgent visits increases on Monday. As stated in the Findings and Conclusions sections, the general pediatric clinics experience an increase in demand for urgent visits on Monday. Some examples of how to increase this capacity include increasing the minimum number of urgent visits on Monday or adding additional staff to the clinics. Team 5 has left the exact way to add capacity up to the clinics, allowing them to fit their individual needs into our recommendation. Urgent Visits per FTE Well Visits per FTE Month 25th 50th 75th 25th 50th 75th January 5.2 7 8.5 4.2 5.5 6.9 February 6.6 8.7 10.7 3.7 5.1 6.6 March 5.8 7.5 9.3 4.3 5.5 6.8 April 4.6 6.2 7.8 4.7 5.9 7.1 May 5.1 6.7 8.3 4.4 5.8 7.2 June 3.6 5.2 6.8 4.7 6.2 7.6 July 2.9 4.2 5.5 5.1 6.9 8.6 August 3.1 4.3 5.4 5.9 7.8 9.7 September 3.9 5.4 6.8 6 7.4 8.8 October 5.2 6.8 8.5 5.2 6.7 8.2 November 6.3 7.9 9.5 5.4 6.8 8.3 December 5.1 6.7 8.2 3.4 5.2 6.9 3

INTRODUCTION University of Michigan (UM) General Pediatrics offers health services to patients through nine outpatient clinics located throughout South Eastern Michigan. These clinics face scheduling problems caused by seasonal trends in demand for scheduled visits. Scheduled visits can be categorized into three types: return, urgent and well. Urgent visits are often scheduled same-day or next-day whereas well visits are often scheduled weeks or months in advance. The clinics experience large variation in the demand for urgent and well visits throughout the year. This variation in demand often results in overbooking of patient visits or underutilization of physicians. The client would like understand the trends in demand for urgent, return and well visits in order to better match the availability of appointments to demand. The University of Michigan project team 5 investigated three UM General Pediatric clinics: 1) Briarwood Center for Women, Children and Young Adults; 2) East Ann Arbor Health and Geriatric Center; and 3) West Ann Arbor Health Center. These clinics differ in size, number of physicians and number of patient visits. The team conducted an in-depth analysis of the current scheduling system and historical scheduling data to understand the variability and trends in the demand for sick, urgent and well visits. The team summarized the observed trends and provided recommended scheduling templates that reduce overbooking and maximize physician utilization. This report shows the background, methodology, finds, conclusions and recommendations. BACKGROUND UM General Pediatrics provides primary care for children patients. General Pediatrics has nine clinic divisions throughout the South Eastern Michigan. Pediatricians provide care through both sick and well visits. Urgent and return visits are scheduled for patients exhibiting symptoms of illness. Well visits include regularly scheduled physicals and health maintenance exams. The team studied three clinics: the Briarwood Center for Women, Children and Young Adults; the East Ann Arbor Health and Geriatric Center; and the West Ann Arbor Health Center. The number of physicians and the volume of visits per year at these three clinics are shown on Table 1. Table 1. Number of physicians and average number of visits per year for three clinics Location Current Physician FTE Average visits per year (full time equivalent) (2008-20) Briarwood 3.25,550 East Ann Arbor 3.88 14,964 West Ann Arbor 1.37 4,786 Table 1 indicates that the number of physicians is proportional to the volume of visit. Recently, UMHS has changed their scheduling software from Enterprise-Wide System (EWS) to Cadence by EPIC. This change in scheduling software is not expected to have a great impact to the efficiency of outpatient scheduling. The client is interested in studying the summer and winter seasons. The summer season currently is defined from May 1 through September 30. The winter season currently is defined from October 1 through April 30. Clinic revenue is 4

based on a common organizational metric call Relative Value Units (RVU). RVU indicated the relative amount of revenue for various services provided. Higher complexity services correspond to higher RVU. Sick visits often have a higher RVU then well visits. Improved outpatient scheduling will help both the UM division of General Pediatrics and the individual health centers to maximize revenue while meeting patient needs. KEY ISSUES The following key issues are the driving force of this project. Clinics experience seasonal variation in demand for sick and well visits Sick patients may be turned away if physicians do not have availability in their schedules resulting in lost revenue and potentially increased Emergency Room visits Physicians may overbook their schedules leading to over-burden of staff and patient waiting Appointment slots may go unused if too many are saved for sick visits leading to decreased physician utilization and decreased revenue GOALS AND OBJECTIVES The primary goals for the project include the following: To understand the trends and factors influencing the demand for sick and well visits at 1) Briarwood Center for Women, Children and Young Adults; 2) East Ann Arbor Health and Geriatric Center; and 3) West Ann Arbor Health Center To apply that understanding to clinic scheduling strategies to improve outcomes To achieve these goals the team has identified the following objectives: Identify trends in the demand for sick and well visits by clinic size, time-of-year and dayof-the-week Design clinic scheduling templates that incorporate trends in demand Decrease the number of overbooked patients Decrease the number of patients turned away due to the lack of appointment availability Increase physician utilization Improve patient access to care PROJECT SCOPE This project considers the demand for clinic appointments at three of the nine UM General Pediatric Clinics: the Briarwood Center for Women, Children and Young Adults; the East Ann Arbor Health and Geriatric Center; and the West Ann Arbor Health Center. The following factors will be considered in the analysis of demand: appointment type, time-of-year, day-ofweek and clinic size. The appointment types investigated in this project include urgent, return 5

and well visits. The data collected comprise of existing data obtained from existing UM scheduling databases. The team made recommendations for optimizing decision making within the existing scheduling system. All clinic activity outside the General Pediatrics division of the UM Pediatric Department will not be studied. The six UM General Pediatric Clinics not listed above are not included in this project. The following factors were not considered in stratifying demand: patient symptoms/diagnosis, specific treatment to be provided during visit, and patient gender. The team did not investigate or make recommendation to improve the broader scheduling system as it exists today, namely, the scheduling software, and standards for the duration of various appointment types, and the frequency of clinical visits. The clinical processes within a clinic appointment are considered outside the scope of this project. METHODOLOGY Using the following combinations of methods, the team was able to develop recommendations to address scheduling problems in Briarwood Center for Women, Children and Young Adults, East Ann Arbor Health and Geriatric Center and West Ann Arbor Health Center. Conduct literature search. The team conducted a literature search on reports that involve patient scheduling. The team found two documents from Google Scholar (is this good?). Although both of these documents are heavily focused on the mathematically approach of solving scheduling problems, the team was able to find ideas on how to analyze datasets and how to understand patterns. Observe clinics and interview clinical staff. The team visited the Briarwood Center for Women, Children and Young Adults, East Ann Arbor Health and Geriatric Center and West Ann Arbor Health Center to develop a basic understanding of the daily operations within the clinics. The team informally interviewed four clinical staff including clerks responsible for scheduling, nurses, and physicians. The purpose of these interviews was to understand the general census of scheduling problems. Determine and acquire necessary data. With the help of the client and coordinator, the team acquired three datasets. Enterprise Worldwide System Appointment Database. The team acquired historical data from Enterprise Worldwide System. The time period time for this data is 2009 to 20. This data included the scheduling date of appointment, date and year of the appointment, appointment procedure type, patient arrival times and patient s physician. To analyze the data, the team removed unnecessary data which included the data on nurse visits, physician assistant visits, injections, lactation visits, phone consultations, and pharmacy visits. It is important to remove these data since it is not in the scope of the project. Appointment Duration. The team acquired appointment duration data from previous study conducted within the general pediatrics clinics. The data shows the default duration and the actual 6

duration for each appointment procedure type. The team used this data to determine the time that each appointment procedure type should use each visit. Physician Work Schedule. The team acquired a snapshot of the physicians work schedule for each clinic. This snapshot included the hours that each physician is supposed to be in clinic for each day of the week. This data was for one week in 2012. The team was unable to acquire historical data for the scheduling worksheets since the clinics do not collect this information. Analyze trends in the physician full-time equivalent (FTE) 2009-20. Physician FTE measures the number of physician*hours available over a given period expressed as the number of equivalent full-time employees. For example, 2 FTE on a given day in clinic means that a total of 16 physician*hours were scheduled in that clinic across all physicians for that day since a fulltime employee works 8 hr/day. Physician FTE directly affects the demand each physician experiences for clinic visits. The team studied the physician FTE over all three clinics in order to identify significant patterns or changes to FTE. The team applied this understanding of FTE to the more general analysis of demand for clinic appointments and the development of recommendations. The team requested physician work schedules from 2009-20 but was unable to obtain this information in the required timeframe. Consequently, the team designed the following formula to measure the FTE of physician s: [est. FTE ] Physician, Day = ( 24 x ([latest apt] + [duration of latest apt] [earliest apt.] ) ) / 8 Each variable above is a time expressed as a fraction of 24 hours between 0 and 1. For example, 0 and.5 corresponded to 00:00 and 12:00 respectively. This formula allowed the team to create a lower bound on physician FTE using the data contained in the provided appointment database. This estimate fails to account for unutilized time at the first start and very end of the day. Define well, return and urgent visits. All scheduled visits with a physician fall into 12 specific procedure types depending on the length of the scheduled visit, the purpose of the visit and other factors. For the project, these procedure types have been grouped into three general visit types: well visits, return visits and urgent visits. There are important differences in the process by which each of these visits types are scheduled. These visit types are defined below. Well visits are scheduled weeks or months in advance. These visits include predominantly health maintenance exams such as checkups for newborn babies and young children. To leave room in the schedule for return and urgent visits, clinic management must limit the number of well visits scheduled each day. This limit is referred to as the session limit for well visits. Return visits are scheduled within a week or two of the visit date. These visits typically are requested by patients presenting with less urgent symptoms, for instance a patient 7

with a lingering cough or muscular soreness. All miscellaneous visits to see a physician, such as follow-up visits to an urgent visit, are scheduled the same as return visits. Urgent visits are scheduled on the day of the appointment or one day before. These visits are given to patients presenting with the most urgent symptoms such as chickenpox or extreme fever. To ensure slots for urgent visits remain available, clinic management must hold a certain number of slots open on the schedule for urgent visits. On the day of clinic, all patients presenting with symptoms are sent to the triage nurse. The triage nurse alone has the ability to schedule these patients into slots saved for urgent visits Analyze data. The team combined information from the provided datasets into an analyzable form by putting the data in a database, removing unnecessary information and formatting the structure of data. The goal of this aggregate database is to relate historical demand for return, urgent and, well visits to the factors being studied. Next, the team performed statistical tests such ANOVA tests to understand the effect of these factors on demand. Lastly, the team constructed graphs to display patterns in demand for each type of visits as well as the average FTE schedule over the three year per clinic. FINDINGS AND CONCLUSIONS Based on the findings, the team concluded that there were notable trends in demand by month and day-of-week. Also, from the interviews with scheduling staff, the team recognized the key problems existing within each clinic associating with scheduling system. Trends in demand for visits by month. The team conducted several ANOVA tests. The ANOVA tested the demand for visits by 36 months over three years for each of the three clinics. It was determined that there is a significant difference between the demand for visits by month. Table 2 summarizes the results from the ANOVA tests. Table 2. ANOVA Results Test # Clinic Visit Type F Statistic P-value 1 West Ann Arbor Urgent 6.23 0.000 2 West Ann Arbor Return 3.55 0.000 3 West Ann Arbor Well 2.35 0.000 4 East Ann Arbor Urgent 9.74 0.000 5 East Ann Arbor Return 7.66 0.000 6 East Ann Arbor Well 2.42 0.000 7 Briarwood Urgent 16.22 0.000 8 Briarwood Return 8.05 0.000 9 Briarwood Well 8.18 0.000 According to Table 2, each of the P value in the ANOVA tests is equal to 0 which implies a significant difference between the months. 8

Physician FTE by Month. Since there was insufficient data on physician work schedule, the team was able to develop a formula to calculate the FTE for each physician. Using the FTE for each of the physician, the team calculated the percentile of number of demand per physician FTE for each type of visit. Table 3-5 shows a percentile table with the types of visit by physician FTE for each of the clinics. Table 3. Briarwood FTE by Month Briarwood Urgent Return Well Month 25th 50th 75th 25th 50th 75th 25th 50th 75th Jan 5 7.5 9.1 3.1 4.8 6.5 3.4 5.1 6.8 Feb 6.4 8.6 10.9 3.3 5.2 7.1 3.2 4.9 6.6 Mar 5.4 7.5 9.5 3.8 6 8.2 3.7 5.4 7 Apr 4.6 6.2 7.8 4.7 6.7 8.7 5.2 6.5 7.7 May 4.8 6.5 8.1 4.2 6.3 8.3 4.8 6.3 7.8 Jun 3 4.6 6.1 4.1 6.1 8.1 4.9 6.4 8 Jul 2.6 3.8 5.1 3.6 5.1 6.6 5.3 6.9 8.5 Aug 2.8 3.9 5 3.3 4.9 6.5 5.9 8.1 10.3 Sep 3.5 4.9 6.2 4.1 5.9 7.7 6 7.7 9.4 Oct 5 6.7 8.4 4 5.7 7.3 5.6 7.5 9.5 Nov 6.8 8.5 10.3 4.1 5.8 7.5 6 7.7 9.4 Dec 5.4 7.1 8.8 4.5 6 7.6 4.2 6.2 8.1 Table 4. East Ann Arbor FTE by Month East Ann Arbor Urgent Return Well Month 25th 50th 75th 25th 50th 75th 25th 50th 75th Jan 5.1 6.6 8 4.1 5.8 7.5 4.3 5.7 7.1 Feb 6.7 9.2 2.8 4.9 7 3 4.7 6.3 Mar 5.3 7 8.7 4.3 6.1 7.8 4.4 5.7 6.9 Apr 4.3 6.1 7.9 3.8 5.6 7.5 4.1 5.5 6.9 May 5.1 6.9 8.6 3.8 5.4 7.1 3.8 5.3 6.9 Jun 3.7 5.5 7.3 3.1 5.1 7.1 4.5 6.1 7.7 Jul 2.6 4.2 5.8 2.5 4.5 6.4 4.2 6.4 8.6 Aug 3.1 4.6 6.1 2.4 4 5.5 5.1 7.1 9.1 Sep 4.2 5.8 7.4 3 4.3 5.7 5.9 7.2 8.6 Oct 5.2 7 8.9 3.7 5.2 6.7 4.5 6 7.4 Nov 6 7.5 9.1 4.4 6.2 8 4.7 6.2 7.7 Dec 4.7 6.3 7.9 3 5.3 7.6 2.6 4.5 6.3 9

Average percent of physician time Table 5. West Ann Arbor FTE by Month West Ann Arbor Urgent Return Well Month 25th 50th 75th 25th 50th 75th 25th 50th 75th Jan 5.4 6.9 8.5 3.9 5 6 4.8 5.7 6.7 Feb 6.8 8.4 10.1 4 5 6.1 4.7 5.7 6.7 Mar 6.6 8.1 9.7 4.2 5.4 6.5 4.7 5.6 6.4 Apr 5 6.3 7.7 4.1 5.2 6.4 4.7 5.7 6.7 May 5.5 6.9 8.3 3.9 5.2 6.5 4.7 5.9 7 Jun 4 5.5 6.9 3.7 5 6.3 4.8 6 7.2 Jul 3.6 4.6 5.7 3.1 4.2 5.2 5.9 7.3 8.6 Aug 3.6 4.4 5.2 2.9 4 5 6.7 8.2 9.7 Sep 4.1 5.5 6.9 3.5 4.6 5.7 6 7.2 8.4 Oct 5.4 6.8 8.2 4.3 5.3 6.4 5.5 6.5 7.6 Nov 6 7.6 9.2 4 5.1 6.1 5.5 6.6 7.7 Dec 5.3 6.6 8 3.4 4.8 6.1 3.5 4.9 6.2 The team analyzed each of the three tables and determined that urgent visits are highest from October to March while well visits are highest from June to November. The team highlighted these months in each of the tables. Lastly, the team determined that the variation between each of the clinics is very small despite the different sizes of each of the clinics. Distribution of total visit time by month. The team calculated the distribution of total visit time for each of the clinics by the 36 months across the three years. Since the clinics are different in sizes, the team was not able to add up the total number of visits. To take out the effect of clinic size, the team decided to use percentages. Figure 1 shows the average percent of physician time for each month for all three clinics. 60% Figure 1. Percentage of Total Visit Time by Month 50% 40% 30% 20% 10% Urgent Visits Return Visits Well Visits No Visits Scheduled 0% Jan, 09 May, 09 Sep, 09 Jan, 10 May, 10 Sep, 10 Jan, May, Sep, Month 10

Average Time per day Average Time per day From Figure 1, the team concluded that there are noticeable seasonal trend in urgent and well visits while the seasonal trend in return visits are less pronounced. In addition, the team also calculated the idle time within each of the clinics and determined that it was relatively constant throughout each month. Trends in demand for visits by day-of-week The team investigated if there existed any relationship between the trends in demand and day-ofweek. The team used the pivot table and calculated the average time spent on each visit type by day-of-week. In general, the overall volume increased on Monday and decreased throughout the week for all three clinics as shown in Figure 2-4. Figure 2. Briarwood Average Time for Visits by DOW 14:00 12:00 10:00 8:00 6:00 4:00 Briarwood Urgent Visit Well Visit Return Visit 2:00 0:00 Monday Tuesday Wednesday Thursday Friday Day of Week Figure 3. East Ann Arbor Average Time for Visits by DOW 14:00 12:00 10:00 8:00 6:00 4:00 2:00 0:00 Monday Tuesday Wednesday Thursday Friday Day of Week East Ann Arbor Urgent Visit Well Visit Return Visit

Percentage of total time Average Time per day Figure 4. West Ann Arbor Average Time for Visits by DOW 7:00 6:00 5:00 4:00 3:00 2:00 1:00 0:00 Monday Tuesday Wednesday Thursday Friday Day of Week West Ann Arbor Urgent Visit Well Visit This trend makes a reasonable sense because patients who get sick over the weekend tend to visit the clinic on Monday. Because the demand is higher for urgent visits on Monday, the clinic would provide more physicians to accommodate the demand. Eventually more physicians would create more open slots for other types of visits which means that the total demand for well and return visits would also increase. However, the problem with this analysis is its dependency on the number of FTEs. From the interviews, the team found that some clinics had different FTEs on different day-of-week. For instance, in East Ann Arbor clinic, the clinic provided more physicians in Monday afternoon because more urgent visits are anticipated in Monday afternoon from historical knowledge. Accordingly, instead of looking at the average time spent on each visit type by day-of-week, the team decided to look at the trend in demand by percentages of total time spent on each visit type. With this approach in analysis, the trends in demand will be independent of the number of FTEs. As shown on the graphs, the relative demand for urgent visits still increases on Monday, decreases next two or three days depending on the clinic and slightly increases on Friday but the proportion of well visits is highest on Wednesday. All three clinics had similar trends in demand for each visit type as shown in Figure 5-7. Figure 5. Briarwood Average Time for Visits by DOW 60% 50% 40% 30% 20% 10% 0% Briarwood Urgent Visit Well Visit Return Visit 12

Percentage of visit time Percentage of visit time Figure 6. East Ann Arbor Average Time for Visits by DOW 60% 50% 40% 30% 20% 10% 0% East Ann Arbor Urgent Visit Well Visit Return Visit Monday Tuesday Wednesday Thursday Friday Figure 7. West Ann Arbor Average Time for Visits by DOW 60% 50% 40% 30% 20% 10% 0% West Ann Arbor Urgent Visit Well Visit Return Visit Monday Tuesday Wednesday Thursday Friday Based on the findings, the team drew a conclusion although the trend is not as strong as the trend by month. For one session limit or for one FTE, the relative demand for urgent visits is highest on Monday, gradually decreases as it approaches the end of the week but slightly increases on Friday whereas the relative demand for well visits moved in the completely opposite pattern Interviews with clinic staff. The team interviewed various clinical staff from the three clinics and concluded that each clinic has specific problems that would like be addressed. Table 6 summarizes the findings from the interviews from the three clinics. Table 6. Interview Results Clinics Problems West Ann Arbor Health Clinic Clerks are having difficulty how to categorize urgent visits thus scheduling a patient under a urgent visit when it is not necessary Briarwood Center for Women, Children and Clerks are having trouble with scheduling when Young Adults physicians leave for vacation or sabbatical East Ann Arbor Health and Geriatric Center The medical director does not know when to switch their scheduling schedules. The clinic depends on the lead clerk to switch scheduling schedule. Switching schedule is based completely on intuition. 13

RECOMMENDATIONS Our original recommendation was focused on providing a hard-set number for the well visit session limit and the minimum number of urgent slots per day for the winter and summer seasons. After looking at our findings and conclusions, it was decided that a hard-set number was not the best way to construct a scheduling template, as there are too many variables to take into account. Our belief is that a hard-set number would not do the best possible job of matching physician schedules to patient demand. This is the reason for presenting our recommendation formatted as the table below. Table 7. Recommended FTE Templates Urgent Visits per FTE Well Visits per FTE Month 25th 50th 75th 25th 50th 75th January 5.2 7 8.5 4.2 5.5 6.9 February 6.6 8.7 10.7 3.7 5.1 6.6 March 5.8 7.5 9.3 4.3 5.5 6.8 April 4.6 6.2 7.8 4.7 5.9 7.1 May 5.1 6.7 8.3 4.4 5.8 7.2 June 3.6 5.2 6.8 4.7 6.2 7.6 July 2.9 4.2 5.5 5.1 6.9 8.6 August 3.1 4.3 5.4 5.9 7.8 9.7 September 3.9 5.4 6.8 6 7.4 8.8 October 5.2 6.8 8.5 5.2 6.7 8.2 November 6.3 7.9 9.5 5.4 6.8 8.3 December 5.1 6.7 8.2 3.4 5.2 6.9 Our recommendation is to have 25th, 50th, and 75th percentile guidelines of both the urgent visits per FTE and well visits per FTE for each month. These guidelines will allow the scheduling template to better fit the actual patient demand for the different visit types. Our recommendation also promotes the integration of physician intuition when making the scheduling templates at the different clinics. An example of the benefits of using guidelines can be highlighted when comparing the winter seasons of 2009 and 20. In 2009, the H1N1 pandemic was taking place and because of this, the general pediatric clinics experienced an above average demand for urgent visits. Our guidelines would allow the minimum number of urgent visits to be at the 75th percentile or even greater, insuring that the increased demand for urgent visits was met with an increase in capacity. During 20, a very mild winter was experienced, resulting in a below average amount of influenza cases. The template could then be modified to put the minimum number of urgent visits closer to the 25th percentile value. This would result in a lower chance of urgent visits going unused and allow additional well visits to be performed. Another recommendation is to increase the capacity for urgent visits on Monday. As stated in the Findings and Conclusions sections, the general pediatric clinics experience an increase in 14

demand for urgent visits on Monday. Some examples of how to increase this capacity include increasing the minimum number of urgent visits on Monday or adding additional staff to the clinics. The exact way to add capacity has been left up to the clinics, allowing them to fit their individual needs into our recommendation. EXPECTED IMPACT With the proposed recommendations, the team believes that there will be a positive impact on each of the clinics. These impacts include understanding demand patterns, developing better seasonal scheduling templates, predicting seasonal shifts and constructing more efficient physician schedule. OPPORTUNITES FOR CONTINUED RESEARCH The research and recommendations developed make for a good starting point to look deeper into influence of different seasons on the demand of urgent, return, and well visits. Due to the fourmonth time frame, there are many opportunities for further research that either were not in the scope of this project or weren t feasible given the time frame. Implement real-time monitoring strategies including patients turn away In order to fully see the benefits of this project, the best measure would be looking at the number of patients turned away. Currently the pediatric clinics do not record this data. It would take at least one year to collect this data to make sure it includes the variability in demand for urgent visits, the biggest driving factor in being turned away on any given day. Once this data is collected, it can be used to compare the current state of the clinics to the proposed future state. Formulate a detailed mathematical model Formulating a mathematical model encompassing our findings was outside the scope of the project. This model formulation would require knowledge and techniques more commonly found at the graduate level and because of this think it would be a great Ph.D. thesis. The idea behind the model is that it could constantly be fed data and provide updated scheduling templates in real-time. This would allow the clinics to better meet demand in their scheduling templates while the model looks at predictive factors to determine the optimal scheduling template. Design simulation of clinic scheduling Building a ProModel simulation of the pediatrics would be a great way to both validate our model and see how well our proposed changes to the scheduling template work. The reason simulation was not performed for this project was because the client and the coordinator did not require it. Understand and improve scheduling processes within each clinic. To improve scheduling processes within the clinic, the team recommends that the clinics continuously find opportunities to improve scheduling processes. The team proposes that the clinics have workshops each year that train the clinical staff in methods of standardized work and implements flow charts and decision trees. 15

ACKNOWLEDGEMENTS IOE Team 5 would like to thank the following individuals for their continuous support on this project. Dr. Sharon Kileny Jacquelyn Lapinski Dr. Heather Burrows Dr. Sara Sandvig Michelle Barnett WORKS CITED 16

APPENDICES Appendix 1: Duration Distribution (EWS) Procedure Type Default Duration Booked Duration Frequency 15 159 NP 30 30 1337 45 73 60 7 NP-45 45 45 14 NP-HME 30 30 5 NP- NEWBORN 30 NP-URGENT 15 NP-WCE 30 RV 15 RV-30 30 RV-45 45 RV-URGENT 15 30 578 45 1 60 1 15 84 30 154 45 10 60 2 15 1 30 2482 45 177 60 8 75 1 15 33516 30 2615 45 224 60 26 15 55 30 4026 45 14 60 10 30 2 45 161 45 330 60 46 15 49 30 34812 RV-WCE 30 45 1063 60 22 75 1 WCE-RV 30 45 1 17

FTE FTE FTE Appendix 2: Staffing level Briarwood (FTE vs Month) 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00 Jan 09' May 09' Sep 09' Jan 10' May 10' Sep 10' Jan ' May ' Sep ' Month East Ann Arbor (FTE vs Month) 4.50 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00 Jan 09' May 09' Sep 09' Jan 10' May 10' Sep 10' Jan ' May ' Sep ' Month 2.00 West Ann Arbor (FTE vs Month) 1.50 1.00 0.50 0.00 Jan 09' May 09'Sep 09' Jan 10' May 10'Sep 10' Jan ' May 'Sep ' Month 18

Appendix 3: ANOVA (FTE vs. Month) One-way ANOVA: FTE versus Month(1-36) Source DF SS MS F P Month(1-36) Error 35 59.58 1.70 1.13 0.281 2277 3442.18 1.51 Total 2312 3501.77 S = 1.230 R-Sq = 1.70% R-Sq(adj) = 0.19% Individual 95% CIs For Mean Based on Pooled StDev Level N Mean StDev -----+---------+---------+---------+---- 1 2 68 2.3 1.240 64 2.429 1.371 (-------*-------) (-------*--------) 3 4 70 2.426 1.248 66 2.565 1.246 (-------*--------) (-------*--------) 5 6 60 2.366 1.261 66 2.318 1.109 (--------*-------) (-------*--------) 7 8 66 2.209 1.073 (-------*--------) 64 2.2 1.128 (-------*--------) 9 10 63 2.562 1.232 67 2.674 1.350 (-------*--------) (-------*--------) 12 58 2.652 1.197 67 2.353 1.243 (--------*--------) (-------*--------) 13 14 62 2.315 1.130 62 2.368 1.247 (--------*--------) (--------*-------) 15 16 70 2.484 1.271 66 2.346 1.054 (-------*-------) (-------*-------) 17 18 60 2.382 1.060 66 2.338 1.186 (--------*--------) (--------*-------) 19 20 64 2.281 1.175 67 2.326 1.216 (-------*--------) (-------*--------) 21 22 64 2.306 1.187 64 2.478 1.398 (--------*--------) (--------*-------) 23 24 60 2.588 1.191 (--------*--------) 63 2.236 1.090 (--------*--------) 25 26 63 2.708 1.168 61 2.550 1.416 (-------*--------) (--------*--------) 27 28 70 2.392 1.200 64 2.379 1.289 (-------*--------) (--------*--------) 29 30 63 2.495 1.264 67 2.263 1.123 (-------*--------) (--------*-------) 31 32 60 2.343 1.262 69 2.591 1.303 (--------*--------) (-------*-------) 33 34 63 2.651 1.254 63 2.631 1.266 (--------*-------) (--------*--------) 35 36 60 2.942 1.375 63 2.536 1.323 (--------*--------) (-------*--------) -----+---------+---------+---------+---- 2.10 2.45 2.80 3.15 Pooled StDev = 1.230 19

Appendix 4: ANOVA (Visit time vs. Month) One-way ANOVA: Urgent time versus Month Source Month DF SS MS F P 35 1.95994 0.05600 16.22 0.000 Error Total 755 2.60721 0.00345 790 4.56715 S = 0.05876 R-Sq = 42.91% R-Sq(adj) = 40.27% Individual 95% CIs For Mean Based on Pooled StDev Level 1 N Mean StDev ----+---------+---------+---------+----- 26 0.18229 0.04753 (--*--) 2 3 24 0.22569 0.05644 26 0.19912 0.06714 4 5 22 0.17377 0.04815 20 0.20208 0.05327 6 7 22 0.13968 0.06570 22 0.10085 0.03713 8 9 22 0.648 0.03702 21 0.13988 0.05448 10 23 0.17074 0.07070 20 0.22448 0.08136 12 13 23 0.15716 0.06265 22 0.16525 0.06927 14 15 22 0.17945 0.08571 24 0.14366 0.05258 16 17 22 0.648 0.04542 20 0.10625 0.04554 18 19 22 0.08570 0.04153 22 0.06487 0.03734 20 21 23 0.07065 0.03794 22 0.09754 0.04161 22 23 22 0.15814 0.05968 20 0.21667 0.06409 24 25 21 0.19643 0.04472 21 0.23462 0.05088 26 27 21 0.26587 0.09568 24 0.236 0.08566 28 29 22 0.19081 0.06440 21 0.19048 0.04997 30 31 23 0.12772 0.05961 20 0.13594 0.04159 32 33 23 0.504 0.03866 21 0.14087 0.04654 34 35 21 0.18948 0.06527 20 0.22135 0.06294 36 21 0.20238 0.07623 ----+---------+---------+---------+----- 0.070 0.140 0.210 0.280 Pooled StDev = 0.05876 One-way ANOVA: Well time versus Month Source DF SS MS F P Month Error 35 4.2847 0.1224 8.05 0.000 755.4792 0.0152 Total 790 15.7639 S = 0.1233 R-Sq = 27.18% R-Sq(adj) = 23.80% Individual 95% CIs For Mean Based on Pooled StDev 20

Level N Mean StDev --------+---------+---------+---------+- 1 2 26 0.1955 0.56 24 0.2179 0.43 3 4 26 0.2436 0.1207 22 0.3286 0.0831 5 6 20 0.2625 0.32 22 0.2670 0.0866 7 8 22 0.3172 0.0677 22 0.3277 0.14 9 10 21 0.36 0.0880 23 0.3107 0.1032 12 20 0.3260 0.1221 23 0.2971 0.1348 13 14 22 0.2386 0.00 22 0.2453 0.17 15 16 24 0.2873 0.1248 22 0.3201 0.0854 17 18 20 0.3302 0.0820 22 0.3182 0.0954 19 20 22 0.3532 0.1574 23 0.4230 0.1725 21 22 22 0.39 0.1615 22 0.3712 0.15 23 24 20 0.4083 0.1217 21 0.2996 0.1523 25 26 21 0.3720 0.1059 21 0.3095 0.1516 27 28 24 0.3047 0.1278 22 0.3542 0.59 29 30 21 0.3819 0.1079 23 0.4058 0.1293 31 32 20 0.4052 0.26 23 0.5036 0.1682 33 34 21 0.4435 0.1296 21 0.4960 0.1373 35 36 20 0.4583 0.1045 21 0.3591 0.1556 --------+---------+---------+---------+- 0.24 0.36 0.48 0.60 Pooled StDev = 0.1233 One-way ANOVA: Return time versus Month Source DF SS MS F P Month Error 35 1.08549 0.03101 8.18 0.000 755 2.86233 0.00379 Total 790 3.94781 S = 0.06157 R-Sq = 27.50% R-Sq(adj) = 24.13% Individual 95% CIs For Mean Based on Pooled StDev Level N Mean StDev ----+---------+---------+---------+----- 1 2 26 0.10817 0.07217 24 0.10851 0.06530 3 4 26 0.258 0.06555 22 0.12973 0.03576 5 6 20 0.771 0.04276 22 0.12074 0.04677 (-----*----) 7 8 22 0.553 0.03706 22 0.032 0.05366 9 10 21 0.13641 0.04152 23 0.13678 0.05515 (----*-----) 12 20 0.15573 0.06763 23 0.15580 0.05175 (----*-----) 13 14 22 0.14015 0.07272 22 0.16809 0.08090 (-----*----) 15 24 0.22135 0.08569 21

16 22 0.24384 0.06593 17 18 20 0.24427 0.06452 22 0.23106 0.06514 (-----*----) 19 20 22 0.16903 0.06048 23 0.15670 0.06956 21 22 22 0.17898 0.07859 22 0.14915 0.06630 23 24 20 0.719 0.05146 21 0.13938 0.04979 (----*-----) 25 26 21 0.12946 0.04161 21 0.12897 0.06501 27 28 24 0.13672 0.06335 22 0.14299 0.06503 (-----*----) 29 30 21 0.12748 0.05259 23 0.12138 0.05427 (----*-----) 31 32 20 0.10729 0.05605 (----*-----) 23 0.141 0.04357 33 34 21 0.13988 0.07385 21 0.15278 0.06541 (-----*----) 35 36 20 0.17760 0.06496 21 0.173 0.07206 (-----*----) (----*-----) ----+---------+---------+---------+----- 0.100 0.150 0.200 0.250 Pooled StDev = 0.06157 One-way ANOVA: Urgent time versus Month Source Month DF SS MS F P 35 1.80907 0.05169 9.74 0.000 Error Total 725 3.84621 0.00531 760 5.65528 S = 0.07284 R-Sq = 31.99% R-Sq(adj) = 28.71% Individual 95% CIs For Mean Based on Pooled StDev Level 1 N Mean StDev -----+---------+---------+---------+---- 21 0.25595 0.09338 2 3 20 0.31042 0.08653 22 0.29687 0.08594 4 5 22 0.24384 0.08405 20 0.28854 0.07984 6 7 22 0.21922 0.07610 22 0.17945 0.05883 8 9 21 0.15526 0.03622 21 0.22569 0.06373 10 22 0.28788 0.07384 19 0.30318 0.08490 12 13 22 0.24621 0.09050 20 0.23229 0.08224 14 15 20 0.28073 0.08771 23 0.28578 0.08426 16 17 22 0.20549 0.05903 20 0.22865 0.07583 18 19 22 0.20975 0.06440 21 0.18204 0.06194 20 21 22 0.16477 0.05079 21 0.18452 0.07285 22 23 21 0.20833 0.06862 20 0.25104 0.07465 24 25 21 0.21032 0.03353 21 0.26190 0.07044 26 27 20 0.32188 0.09156 23 0.29484 0.08288 28 29 21 0.23661 0.06216 21 0.22669 0.05609 30 31 22 0.16241 0.08331 20 0.13802 0.04300 32 23 0.15353 0.03913 22

33 21 0.18601 0.08050 34 35 21 0.23512 0.06281 20 0.27083 0.09187 36 21 0.25992 0.06842 -----+---------+---------+---------+---- 0.140 0.210 0.280 0.350 Pooled StDev = 0.07284 One-way ANOVA: Well time versus Month Source DF SS MS F P Month 35 4.6000 0.1314 7.66 0.000 Error 725 12.4342 0.0172 Total 760 17.0342 S = 0.1310 R-Sq = 27.00% R-Sq(adj) = 23.48% Individual 95% CIs For Mean Based on Pooled StDev Level 1 N Mean StDev 21 0.4345 0.22 -+---------+---------+---------+-------- 2 3 20 0.4052 0.1006 22 0.4081 0.0985 4 5 22 0.4527 0.0896 20 0.3719 0.1351 6 7 22 0.4763 0.1367 22 0.5426 0.1525 8 9 21 0.5506 0.1551 21 0.5744 0.61 10 22 0.4678 0.0906 19 0.4550 0.0895 12 13 22 0.3447 0.1603 20 0.3625 0.1010 14 15 20 0.3979 0.0819 23 0.3659 0.0661 16 17 22 0.49 0.17 20 0.4573 0.0962 18 19 22 0.4337 0.1253 21 0.5288 0.1383 20 21 22 0.5483 0.1352 21 0.4425 0.04 22 23 21 0.4990 0.19 20 0.5365 0.1355 24 25 21 0.3343 0.1244 21 0.4435 0.0892 26 27 20 0.4313 0.1364 23 0.4284 0.1064 28 29 21 0.3740 0.1085 21 0.4365 0.1327 30 31 22 0.3920 0.55 20 0.4969 0.1571 32 33 23 0.7147 0.3228 21 0.5456 0.1297 34 35 21 0.4494 0.1207 20 0.4365 0.18 36 21 0.3770 0.1455 -+---------+---------+---------+-------- 0.30 0.45 0.60 0.75 Pooled StDev = 0.1310 One-way ANOVA: Return time versus Month Source DF SS MS F P Month 35 0.31252 0.00893 2.42 0.000 Error 725 2.67801 0.00369 23

Total 760 2.99053 S = 0.06078 R-Sq = 10.45% R-Sq(adj) = 6.13% Individual 95% CIs For Mean Based on Pooled StDev Level N Mean StDev ---------+---------+---------+---------+ 1 2 21 0.17758 0.03848 20 0.17969 0.06249 (-------*------) 3 4 22 0.19886 0.05935 22 0.18229 0.06346 5 6 20 0.18594 0.06105 22 0.16430 0.07159 7 8 22 0.15104 0.05977 21 0.14038 0.05692 9 10 21 0.18204 0.05596 22 0.20313 0.06123 12 19 0.16776 0.055 22 0.15009 0.06756 (-------*-------) 13 14 20 0.16615 0.05219 20 0.17292 0.03859 15 16 23 0.19746 0.07349 22 0.19223 0.05864 17 18 20 0.18333 0.05739 22 0.19271 0.06687 19 20 21 0.16567 0.05321 22 0.15341 0.05698 21 22 21 0.16766 0.05269 21 0.19593 0.04781 (-------*------) 23 24 20 0.17656 0.05829 21 0.15079 0.04970 25 26 21 0.19147 0.06929 20 0.18906 0.06538 (-------*------) (-------*-------) 27 28 23 0.18433 0.05380 21 0.19196 0.06658 (-------*------) 29 30 21 0.19643 0.086 22 0.18229 0.06379 31 32 20 0.13490 0.05230 (-------*------) 23 0.13542 0.05984 33 34 21 0.14583 0.060 21 0.17609 0.05974 (-------*------) 35 36 20 0.20104 0.05806 21 0.21429 0.08252 ---------+---------+---------+---------+ 0.140 0.175 0.210 0.245 Pooled StDev = 0.06078 One-way ANOVA: Urgent time versus Month Source DF SS MS F P Month Error 35 0.189459 0.005413 6.23 0.000 725 0.630030 0.000869 Total 760 0.819489 S = 0.02948 R-Sq = 23.12% R-Sq(adj) = 19.41% Individual 95% CIs For Mean Based on Pooled StDev Level N Mean StDev ---+---------+---------+---------+------ 1 2 21 0.07143 0.02167 20 0.771 0.03673 3 4 22 0.08475 0.02454 22 0.07955 0.03109 5 6 20 0.08594 0.04096 22 0.06203 0.03394 7 8 22 0.054 0.02764 21 0.05506 0.02286 9 21 0.06944 0.03250 24

10 22 0.08902 0.03178 12 19 0.08882 0.02390 22 0.08191 0.03038 13 14 20 0.07031 0.02387 20 0.09792 0.04213 15 16 23 0.07020 0.03175 22 0.06392 0.02695 17 18 20 0.07708 0.02302 22 0.07197 0.03293 19 20 21 0.06002 0.01890 22 0.05682 0.02870 21 22 21 0.06151 0.02396 21 0.08135 0.02976 23 24 20 0.08750 0.02836 21 0.07242 0.02746 25 26 21 0.08780 0.02748 20 0.09844 0.03647 27 28 23 0.08967 0.02857 21 0.06895 0.03399 29 30 21 0.07688 0.02475 22 0.05966 0.02535 31 32 20 0.03437 0.03136 23 0.04846 0.02731 33 34 21 0.07093 0.02689 21 0.07540 0.03438 35 36 20 0.08750 0.02876 21 0.06548 0.02512 ---+---------+---------+---------+------ 0.030 0.060 0.090 0.120 Pooled StDev = 0.02948 One-way ANOVA: Well time versus Month Source DF SS MS F P Month Error 35 0.38602 0.003 3.55 0.000 725 2.25122 0.003 Total 760 2.63724 S = 0.05572 R-Sq = 14.64% R-Sq(adj) = 10.52% Individual 95% CIs For Mean Based on Pooled StDev Level N Mean StDev -----+---------+---------+---------+---- 1 2 21 0.13591 0.05497 20 0.08438 0.05034 3 4 22 0.13068 0.03473 22 0.14299 0.03234 5 6 20 0.354 0.059 22 0.14205 0.05845 (-----*----) 7 8 22 0.13068 0.06173 21 0.15675 0.07672 9 10 21 0.16667 0.04370 22 0.16572 0.05359 (----*-----) 12 19 0.17325 0.04609 22 0.09186 0.06088 13 14 20 0.13750 0.04751 20 0.12188 0.06278 15 16 23 0.12681 0.05754 22 0.123 0.05220 (-----*----) 17 18 20 0.875 0.06049 22 0.13920 0.05728 19 20 21 0.16766 0.07891 22 0.16856 0.06710 21 22 21 0.16171 0.04109 21 0.12798 0.03801 23 24 20 0.12187 0.05746 21 0.806 0.07721 (-----*----) 25 21 0.12599 0.03812 (----*-----) 25

26 20 0.12083 0.05305 27 28 23 0.13859 0.036 21 0.607 0.05727 (-----*----) 29 30 21 0.14087 0.03948 22 0.14583 0.05183 (----*-----) 31 32 20 0.15208 0.08363 23 0.17029 0.06548 (-----*----) 33 34 21 0.17857 0.05574 21 0.12302 0.05056 35 36 20 0.14271 0.04127 21 0.10218 0.05702 -----+---------+---------+---------+---- 0.080 0.120 0.160 0.200 Pooled StDev = 0.05572 One-way ANOVA: Return time versus Month Source DF SS MS F P Month Error 35 0.079321 0.002266 2.35 0.000 725 0.700002 0.000966 Total 760 0.779323 S = 0.03107 R-Sq = 10.18% R-Sq(adj) = 5.84% Individual 95% CIs For Mean Based on Pooled StDev Level N Mean StDev -------+---------+---------+---------+-- 1 2 21 0.06647 0.03045 20 0.04688 0.03537 3 4 22 0.06771 0.02853 22 0.06723 0.02646 5 6 20 0.06510 0.02924 22 0.06392 0.03967 7 8 22 0.05398 0.03501 21 0.04266 0.03577 9 10 21 0.06151 0.02571 22 0.05019 0.02376 12 19 0.07292 0.02821 22 0.06108 0.03967 13 14 20 0.06667 0.03352 20 0.07240 0.03934 15 16 23 0.08197 0.03082 22 0.06866 0.03933 17 18 20 0.06354 0.03150 22 0.05587 0.03645 19 20 21 0.03869 0.03075 22 0.04593 0.02441 21 22 21 0.04712 0.01793 21 0.06349 0.02232 23 24 20 0.06094 0.03350 21 0.05754 0.02627 25 26 21 0.06944 0.02601 20 0.05312 0.02925 27 28 23 0.06295 0.02971 21 0.06052 0.02921 29 30 21 0.06200 0.02804 22 0.05729 0.02742 31 32 20 0.06354 0.03428 23 0.05027 0.01795 33 34 21 0.04315 0.02265 21 0.06944 0.02802 35 36 20 0.08229 0.02905 21 0.06647 0.04990 -------+---------+---------+---------+-- 0.040 0.060 0.080 0.100 Pooled StDev = 0.03107 26