Innovate and Validate: Design-Led Simulation Optimization to Test Centralized Registration Feasibility in a Multispecialty Clinic

Abstract

Objective:

This study utilizes a design-led simulation-optimization process (DLSO) to refine a hybrid registration model for a free-standing outpatient clinic. The goal is to assess the viability of employing DLSO for innovation support and highlight key factors influencing resource requirements.

Background:

Manual registration in healthcare causes delays, impacting patient services and resource allocation. This study addresses these challenges by optimizing a hybrid centralized registration and adopting technology for efficiency.

Method:

An iterative methodology with simulation optimization was designed to test a proof of concept. Configurations of four and five registration options within a hybrid centralized system were explored under preregistration adoption rates of 30% and 50%. Three self-service kiosks served as a baseline during concept design and test fits.

Results:

Centralized registration accommodated a daily throughput of 2,000 people with a 30% baseline preregistration rate. Assessing preregistration impact on seating capacity showed significant reductions in demand and floor census. For four check-in stations, a 30%–50% preregistration increase led to a 32% seating demand reduction and a 26% decrease in maximum floor census. With five stations, a 50% preregistration reduced seating demand by 23% and maximum floor census by 20%.

Conclusion:

Innovating introduces complexity and uncertainties requiring buy-in from diverse stakeholders. DLSO experimentation proves beneficial for validating novel concepts during design.

Keywords

centralized registration outpatient center discrete event simulation optimization

The manual registration process significantly contributes to delays in patient services, requiring additional time for paperwork completion, insurance verification, and manual data processing. These inefficiencies lead to suboptimal resource allocation, including the need for extensive waiting spaces and increased human resources for administrative tasks. Addressing these challenges through the optimization of registration processes and the adoption of technology solutions aims to enhance efficiency, reduce delays, and optimize resource utilization within healthcare settings. In a conventional medical office building (MOB) design, visitor flow management is centered on the building’s entry, typically a lobby. Each floor is occupied by medical practices with dedicated exam, therapeutic, and office spaces. Ancillary spaces, such as check-in and check-out areas, patient, and visitor lounges, are duplicated for each practice, reducing space available for clinical care, which also limits interdisciplinary exchange and resource sharing among specialists.

To address this issue, we proposed a redesigned MOB concept with a centralized check-in and registration system that would free up space for patient intake, allowing for more exam rooms and clinical treatment areas. This MOB has 14 floors with approximate total of 250,000 square feet (SF). Each floor accommodates single or multiple practices, and the clinical care space is evenly distributed across all floors, as depicted in Figure 2. The SF of each clinical floor is 16,700, ensuring a standardized and consistent basis for evaluating the MOB’s scale and capacity. This redesign separates clinical spaces from academic offices, maximizing the clinical space on practice floors and fostering collaboration among practice teams. Based on projected patient volumes and assumptions about patient companions, the centralized entry was estimated to handle approximately 2,000 people per day. This required an integrated system to accommodate the diverse flow of patients and visitors merging in the MOB welcome area while enhancing the overall experience. The redesign represented a significant departure from current practices and necessitated multidisciplinary planning for people flows, registration processes, staffing, environment design, digital technology, and security. The centralized concept had not been tested in the health system, and its implementation required reimagining existing processes. Two key factors influenced the decision-making process: an anticipated rise in daily foot traffic compared to existing locations and the necessity to integrate digital and self-service options with personalized full-service choices.

To address these challenges, the project team adopted a transparent and inclusive design process, involving various stakeholders, such as business strategists, operations teams, practice administrators, clinical leaders, architects, designers, and design researchers. The inclusive process involved extensive consultations over several months, which included team meetings, process mapping, and flow analysis. Scenario modeling using design-led simulation-optimization process (DLSO) was employed to assess the feasibility of the centralized welcome area, facilitate discussions on future operations, and support decision-making. This approach of DLSO was relatively uncommon in the context of central registration, highlighting the innovative nature of the project. The scenarios focused on two key areas of interest; firstly, the focus was on aligning patient and companion flows with available seating and check-in resources. The objective was to determine whether the allocated space could adequately accommodate five in-person and three digital check-in options, along with 70 seating, without causing significant wait times more than 10 min or congestion. Secondly, the scenarios explored the impact of a digital and full-service hybrid mix on resource requirements. By evaluating the patient population and existing digital adoption rates, the objective was to determine whether the current and projected future uptake of digital check-in (at rates of 30% and 50% of patients, respectively) would effectively showcase the viability of the centralized concept. This assessment focused on aspects, such as wait times, seating availability, and avoiding bottlenecks. The scenarios considered included both a 30% and a 50% preregistration rate, each configured with four and five check-in stations and three kiosks. Establishing baseline seating and check-in desks entailed analyzing patient volume trends, appointment scheduling patterns, and the physical layout of the clinic. Figures 1 and 2 provide a visual comparison between the conventional design and our reimagined concept.

Figure 1.

Conventional medical office building concept. Images courtesy of EwingCole.

Figure 2.

Reimagined medical office building concept (right). Images courtesy of EwingCole.

Our study focuses on two key research questions (RQs):

RQ1. How can we align patient and companion in-flows with available seating and check-in resources while maintaining minimal wait times and smooth circulation?

RQ2. What is the impact of a digital and full-service hybrid mix on resource needs, including wait times, seating, and potential bottlenecks?

To answer these questions, we implemented an extensive consultative process involving various stakeholders and used DLSO. Our approach aimed to assess the feasibility of the centralized wellness welcome area concept, stimulate discussions around future operations, and support decision-making. We modeled different scenarios to analyze the effects of patient and companion in-flows, check-in processes, and digital adoption rates. Key performance measures, such as check-in wait times, total floor occupancy, seating demand during peak and off-peak hours, and check-in station staff utilization, were utilized. This study presents our design-led and discrete event simulation (DES) methodology for optimizing central registration capacity and discusses its implications for system design. It also highlights the integration of central registration during the design process and decision-making. This article concludes with the results of our experimental runs, implications for optimal system design, and considerations for centralized registration in healthcare settings.

Background

DES in Outpatient Facility Design and Capacity Planning

The utilization of DES modeling regarding real-time support for decisions during early healthcare facility planning, concept design, and test fits for registration processes is limited in the literature. However, it has been a valuable approach for comprehending the impact of alternative designs in healthcare improvement efforts (Alvarado et al., 2018; Cai & Jia, 2019; Fun et al., 2022). The advantages of DES to incorporate stochastic variables, deal with uncertainty, and capture the connections between spaces and processes from complex patient care flows are reflected in the growing adoption of DES in healthcare operational and facilities planning. DES is informing the building layout and new operational strategies and allowing teams to validate new ideas in outpatient (clinics, emergency departments, and MOBs) and inpatient (medical-surgical and intensive care units, radiology, and surgery) settings. In the outpatient setting, DES has been used in studies seeking to improve or test the feasibility of specific or combined innovations related to capacity planning (Barros et al., 2021) and patient flow and process (Easter et al., 2019; Maass et al., 2022), staffing models (Saadouli & Ltaif, 2021), appointment scheduling (Demirli et al., 2021; Kubala et al., 2021), resource assignment strategies (Aboueljinane & Frichi, 2022), and designed environment layout options (Vahdat et al., 2019; Valipoor et al., 2022), to name a few examples.

In practice, decision-makers are increasingly expecting decision support that quantifies expected performance concerning cost and patient-centeredness. A systematic literature review (2008–2018) of methodologies applied to advance evidence-based healthcare facility design noted the burgeoning use of DES, identifying 14 papers focused on DES to inform decision-making related to architectural concepts, layout modeling, and resource allocation (Halawa et al., 2020). A 2022 review concluded that the full potential of DES in outpatient capacity planning had not been realized (Al-Kaf et al., 2022). The practical utility of DES in facility and operations planning, however, continues to be demonstrated for a broad range of decision-making and performance-related purposes. Ponis et al. (2013) apply simulation modeling to determine capacity requirements and medical service planning across multiple, networked outpatient facilities (Ponis et al., 2013). Gosavi et al. (2016) apply DES to assess and compare four layout options for a primary care clinic using two primary performance metrics: staff and patient walking distances and queueing or wait times for exam rooms. Cai and Jia (2019) present multiple use cases for DES including exam room capacity under multiple growth rate scenarios and describe DES as a tool for performance-driven design. Combining design with operational concepts, Vahdatzad (2018) proposed a simulation-optimization framework to improve the timeliness of patient care by improving the departmental layout for multiple orthopedic outpatient clinics through flexible physician assignment strategies. Challenges associated with integration of DES or a hybrid of DES and other methods in facility design and operations have also been identified. The critical involvement of stakeholders, a consultative approach, versatility, and an iterative approach have also been identified as important contributors to not only application, but “uptake” of modeling efforts (Tyler et al., 2022).

DES in Centralized Registration and Scheduling Design

Methodology

A systematic framework for DLSO was established, as illustrated in Figure 3, involving the collection and analysis of relevant patient arrival data, followed by the conceptualization and development of the model, and rigorous verification and validation. The model underwent experimentation with two scenarios, varying the preregistration rate at 30% and 50% of patients, and two subscenarios reflecting changes in daily patient volume, distinguishing between a busy day and an average day. Optimization ensured station utilization equal to or greater than 85%, with a constraint on patient waiting time set to be less than 2 min. The simulation results guided design decision-making and were updated to incorporate design and operational changes.

Figure 3.

Design-led simulation optimization methodology.

A modified action research approach entailed multiple, overlapping elements: a discovery phase to understand current state practices and identify opportunities and potential challenges; data analysis and consultation to determine system inputs for the model; architectural design and test fits; future state system design mapping, defining key questions of interest, performance metrics, and experimental scenarios; and model building, validation, and experimentation all taking place in close consultation with stakeholders. Multiple process maps and models were developed in the process of validating and revalidating key patient flows and to reflect new ideas, data availability, and design changes. Figure 4 captures that iterative approach as the concepts were refined as additional information, ideas, and insights were generated by the model.

Figure 4.

Iterative design and experimentation cycle.

Key questions of interest which were to guide the modeling and validation process included the following:

What is the practical feasibility of centralized registration? Does it allow for a smooth and welcoming flow of incoming patients and companions without excessive waiting times, queues, congestion, or bottlenecks?

What would be the expected seating demand under a centralized concept? Would it require a “bus station” type environment or would there be adequate space designed to create a welcoming environment and provide optimal seating arrangements for all?

How many check-in stations (in-person and self-service) would be required to meet the demand generated by arriving patients and would this exceed space availability or contribute to space savings? A primary hypothesis was that cross-training and shared registration would lead to savings in resources.

Is the feasibility of the centralized concept on the rate of self-service digital preregistration? More specifically, was there a threshold for adoption to make the concept work? A primary hypothesis was that a higher level of digital preregistration would be required to accommodate volumes expected in the centralized model.

At the onset of the discovery process, group and one-on-one semi-structured interviews with practice managers and administrators were held to map current state check-in processes and resources, document patient flows, and understand patient needs and future state practice vision. Visits to each practice were conducted to document current state spaces. This discovery phase mapped current state registration component processes for each practice that would relocate to the new facility. The specific needs of each patient population were identified. Following the synthesis of the current states, future state process maps for the centralized registration concept were developed and refined over 2 months encompassing the journey from patient arrival through proceeding to the appointment (see Figure 5 for a simplified version of the visual process map). Operational processes and experiential aspirations were defined for each step.

Figure 5.

Simplified centralized registration visual process map. Image Courtesy of EwingCole.

Simulation Model Development

Patient flow and check-in process

In the model, patient flow begins with the arrival of the patient and any companions and ends when the scheduled time is reached or registration and check-in is completed. Patients may arrive early, on time, or late for their appointments. Upon arrival, patients who have completed a digital preregistration complete a self-service (no staff involved) check-in using a smartphone or a kiosk. For self-check-in on a smartphone, patients receive a notification or use a dedicated app to initiate the process. The app may confirm their check-in and utilize geolocation to ensure they are physically present at the clinic. If a patient’s scheduled visit time is reached within 5 min of completing the check-in, the patient proceeds to the practice floor and exits the model. If more than 5 min remain before the scheduled time, the patient finds a seat in the wellness lounge and proceeds to the clinic floor when the scheduled time was reached. Patients who are not preregistered proceed to any staffed desk for an initial greeting and check-in and receive their forms to complete in the lounge. Companions accompany patients through this process. Cross-training of check-in staff was assumed, such that any patient could be checked in by any staff member. Once the forms are completed, the patients may return them to any desk. As with preregistered patients, patients may either proceed to the clinic floor or sit in the lounge if their scheduled time has not yet been reached.

Process durations and assumptions

The durations for components of the registration process, including the initial greeting and check-in, paperwork completion, and self-service kiosk use, were formulated with input from practice administrators, registration team members, and benchmarks. No existing data were available for these durations. The processing times for each process component were determined using triangular distributions to account for variability in duration times: 1–3 min for greeting, 5–15 min for in-person paperwork completion, and 2–7 min for self-service (smartphone or kiosk) check-in. Table 1 outlines the process durations and assumptions.

Table 1.

Model Process Durations. Subject Matter Experts (Clinical and Strategy Team).

Model Process Durations	Distribution (minutes)	Source
Initial greeting (in-person)	T(1,2,3)	SME input
Paperwork completion	T(5,10,15)	SME input
Digital check-in	T(2,4.5, 7)	SME input

Forecasted future state clinical visit volumes

Deterministic annual forecasted future state clinical visit volumes for each specialty practice were provided by the health system and used to generate expected daily arrival volumes for the centralized registration and wellness space. Annual volumes were converted into weekly figures by dividing them by the number of operating weeks. Annual expected volumes were based on initial forecasts distributed across 50 weeks. It was assumed future state scheduling would not vary significantly from current state practice and retain a nonlevel loaded pattern. Likewise, a conservative (worst case) approach was adopted to set the upper limit of visit volumes, simulating a “busy day.” This involved using the highest proportion of weekly visits on any given day to convert weekly volume into a daily figure. Companions of patients were later incorporated into the model based on historical data regarding the number of companions allowed per patient. The model then allocated the essential resources to address their needs, including additional seating and security check duration. This approach aimed to optimize the utilization of resources and ensure that both patients and companions received adequate support during their visits. Companion throughput (people accompanying a patient) was determined based on practice manager recommendations input, whereby 75% of patients could be expected to be accompanied by at least one companion, 22% of patients by two companions, and 3% of patients by three companions. Patient arrival behavior (early or late), cancellation and no-show rates, and schedules were estimated based on 10 months of practice data (n = 63,302). That deidentified scheduling and encounter data included visit status (completed, no-show, or canceled), as well as timestamps for arrival time and scheduled time by practice. Python 3.4 was used to preprocess the historical data. Based on historical patterns, 12% of all scheduled appointments are either canceled or missed (no-show) by the patient, while 4% are conducted through televisits. The highest volume clinic day reached 23% of the volume of a typical 5-day week and was used to model a “busy day.” The expected daily throughput was determined to be 866 patients and 1,109 companions. The historical median number of visits per 15-min appointment slot was calculated and used to model the appointment schedule. Figure 6 shows the forecasted bimodal arrival distribution used to simulate the proportion of total patients expected to arrive during each time slot of the day.

Figure 6.

Median value of scheduled appointments per time slot of the day (based on 10 months of data).

The historical offset between patient arrival and appointment time, measured in minutes, was calculated for each visit. An empirical distribution for patient arrivals to the clinic (early, late, and on-time) was generated to represent the historical arrival pattern of patients and visitors. An analysis of the arrival patterns showed that 64.3% of patients arrive 60 min or earlier for their appointments, 3.1% arrive on time, and 32.6% arrive less than 30 min late. Arrival less than or equal to 30 min late and less than or equal to 1 hr early were used as bounding limits. Practice managers noted that visits more than 30 min late would be considered no-shows. The best-fit distribution for arrivals was estimated using SciPy library in Python and found to be log-logistic (−80.23, 86.59, and 8.9; see Figure 7). The best-fit distribution was used to generate an empirical distribution of Fisk (c = 9.90, loc = −80.23, and scale = 86.59) with 42,400 rows in governing patient arrivals to the clinic. A summary of model parameters can be seen in Table 2.

Figure 7.

Distribution of patient arrival to appointment time off-set in minutes (early or later than scheduled visit). Based on 10 months of historical data (n = 63,302). Used to generate an empirical distribution to govern patient arrivals to the clinic.

Table 2.

Model Parameters for Simulation Model Based on Subject Matter Experts Inputs, Forecasted Encounter Data, and 10 Months of Historical Data.

Arrival Modeling Parameters	Value	Source
Operating weeks per year	50	SME input
Operating days per week	5	SME input
Busiest day of the week volume (% of weekly total)	23%	Historical data
Daily patient throughput	866	Projected annual volumes
Daily companion throughput	1,109	SME input
Cancellation/no-show rate (excluded from arrivals)	18%	Historical data
Tele-visit rate (excluded from arrivals)	7%	Historical data
Early/late arrival delay (minutes)	Fisk (c = 9.90, loc = −80.23, scale = 86.59)	Historical data
Arrival offset distribution	Empirical distribution (42,400 rows)	Historical data

Note. n = 63,302.

Test fits and space planning

As processes and data analysis for modeling inputs were being developed, architectural test fits and space planning took place, together with data analysis to establish arrival patterns and volumes, SME consultation to set assumptions regarding companions and other elements not captured in data sets, and model building and validation. Architectural test fits determined that up to eight check-in desks and seating groupings totaling 120 seats could be accommodated in the new space, along with space for four or five self-service free-standing kiosks and required circulation areas. Floor plans were created in Revit and imported into FlexSim and used as a background to create an accurate, scaled 3D visualization of the designed space in FlexSim in consultation with the architectural team. The model was created in FlexSim for healthcare. Figure 8 shows a screenshot of the simulated registration space. All volumes, operational assumptions, process flows, processing times, and the distributions used to develop the model were discussed and verified by the practice managers and the ambulatory operations team. These assumptions and workflows were documented and reviewed by the client during regular meetings to discuss updates to the model. Operational managers, designers, and stakeholders observed 3D animations of the model to check that the patients and visitors could follow the proper flow of events and queuing was feasible at different stages.

Figure 8.

Partial screenshot of the simulation model.

Model validation

Verification and validation processes play a crucial role in ensuring the accuracy and reliability of simulations (Sargent & Balci, 2017). Structural walkthroughs involved step-by-step examinations of the simulation model to verify its correctness and coherence. This process ensured that the model accurately represents the underlying system and its dynamics. The validation of the model was conducted using several methods to ensure its accuracy and reliability. The 3D graphic model was used to visually display the operational behavior of the system as it moved through time. This involved visually depicting the movements of patients and staff within the lobby during simulation runs. The graphical representation of the system allowed for the evaluation and experimentation of different configurations, layouts, and scenarios in a virtual environment. Furthermore, it enhanced stakeholder engagement and enabled designers to observe the dynamics of the system while evaluating and experimenting with various configurations, layouts, and scenarios. Event validity was ensured by comparing the occurrence of events in the simulation model to the real events observed within the clinic lobby. The number of patient arrivals, wait times, and other key events simulated by the model were compared to the corresponding data obtained from actual clinic operations. This analysis aimed to confirm that the simulation model accurately replicated real-world events. Face validity was determined through consultations with domain experts familiar with the clinic’s lobby operations. These experts evaluated the model and its behavior to assess the validity of the model, the assumptions, and outputs. The accuracy of the model’s logic and the validity of its input-output relationships were reviewed and validated through providing critical insights and contributing to the model refinement. Trace validity was conducted by closely monitoring the behaviors of specific entities within the model, such as patients and staff members, to evaluate the correctness of the model’s logic and the accuracy of its outcomes. By tracking these entities’ movements and activities, their progression was compared to the expected behaviors, providing insights into the model. Extreme scenarios were employed to test the robustness and performance of the model. By using very high or low parameter values, the model’s behavior was examined under extreme conditions. This allowed for the identification of potential vulnerabilities, bottlenecks, or failures in the system. Extreme scenarios provided valuable insights into the model’s behavior and helped refine it to handle a wide range of conditions. Sampling from an empirical distribution and utilizing patient arrival data played a crucial role in validating the model through empirical validation. This validation method involved comparing the behavior of the model to real-world data acquired from the system being simulated. In the specific case of patient arrivals, the model’s assumptions and behavior underwent testing using historical data gathered from the clinic. Historical data validation was implemented, utilizing a data set specifically collected for constructing and testing the simulation model. This data set represented the empirical distribution of patient arrival times until their appointment at the clinic over a period of 10 months. Utilizing this data set, the model simulated patient arrivals by drawing random samples from the empirical distribution (Figure 9). To achieve a comprehensive understanding of the distribution, 100,000 samples were utilized. Additionally, an additional distribution was created using all 42,401 rows of data, as opposed to the initial sampling of 2,000 rows. The results of sampling from both the 2K-row and 42K-row empirical distribution were compared to the original raw Excel data. Notably, the new empirical distribution, incorporating all 42 K rows, exhibited a perfect match with the original data, aligning with expectations. Conversely, the 2,000-row distribution exhibited a rougher representation of the data.

Figure 9.

Historical data (blue) versus sample data (green). Sampling from the created empirical distribution used to model patient arrival patterns.

Scenario Design and Optimization

Experimental scenarios and performance measures

Experimental scenarios relating to digital preregistration adoption rates and full-service registration needs were developed by the team based on baseline digital registration use (30%) and a maximum future state expected adoption rate (50%) based on input from subject matter experts. The team developed performance measures with the objective of optimizing registration station utilization and staff efficiency. The aim of the experimentation was to minimize the number of registration stations, considering two constraints: maximum wait time for check-in does not exceed 2 min and check-in staff usage should not exceed 85% of a daily shift. All check-in stations were staffed from 7 a.m. to 6 p.m., with the assumption that breaks would be covered by other staff members. To determine the optimal capacity requirements under varying digital and in-person registration rates and given the service and utilization objectives, two scenarios with two subscenarios each were considered (see Table 3). To estimate the seating demand in the waiting area, each scenario was run for 5 days with 32 replications for each week. The Optimizer engine in FlexSim seeks an optimal solution by generating values for each model configuration while running it under multiple simulations simultaneously.

Table 3.

Experimental Scenarios for Optimizer Runs.

Experimental	Patients Using Digital Check-In (%)	Patients Using In-Person Check-In (%)	Maximum Seating Capacity	Number of Kiosks	Number of In-Person Check-In Stations
Scenario 1A	30	70	120	3	4
Scenario 1B	30	70	120	3	5
Scenario 2A	50	50	120	3	4
Scenario 2B	50	50	120	3	5

Wait time and resource utilization are the two key performance measures used to optimize resource allocation in a system. The wait time metric helps assess the time customers or entities spend waiting, enabling the identification of potential bottlenecks and areas for improvement. Resource utilization measures how effectively resources are used. By analyzing resource utilization, organizations can identify underutilized or overutilized resources, allowing for better allocation and optimization. Maximum wait time was considered to account for the longest duration a patient has to wait to be registered. It is calculated by tracking the start and end times of each waiting period and identifying the longest duration among them. Location and staff utilization used to measure the efficiency and effectiveness of utilizing physical locations and staff resources within the lobby. Location utilization measures how effectively physical spaces or resources are utilized. It is typically calculated by dividing the actual usage or occupancy of a location by its maximum capacity or availability. The formula for location utilization is shown in Equation 1 as follows:

Location Utilization = (\frac{Actual Usage}{Maximum Capacity}) \times 100.

For example, if a check-in stations is used for 20 hr out of a possible 40 hr in a week, the location utilization would be (20/40) × 100 = 50%.

Staff utilization measures the efficiency of utilizing available staff resources. It is usually calculated by comparing the productive time (time spent on productive tasks) to the total available time. The formula for staff utilization is shown in Equation 2 as follows:

Staff Utilization = (\frac{Productive Time}{Total Available Time}) \times 100.

Productive time can vary depending on the context, but it typically excludes breaks, nonwork-related activities, and unproductive time. Total available time is the total scheduled work time for staff members. For example, if an employee spends 6 hr on productive tasks out of an 8-hr workday, the staff utilization would be (6/8) × 100 = 75%.

Results

Scenario 1 considered the baseline mix of 30% digital and 70% in-person registration and compared four (1A) versus five (1B) check-in stations. In Scenario 1A, the average use of check-in stations ranged from 51% to 68%, with the target value set at 85%. The maximum use was 80% at 10 a.m. and 1 p.m. The maximum floor census (total people on the floor at any time) did not exceed the seating capacity of 120, reaching only a maximum of 100–111 people during peak hours. For Scenario 1B, the average use of stations ranged from 41% to 54% over the day with a maximum of 64% at 10 a.m. and 1 p.m., the peak clinic hours (Figure 10). The maximum floor census ranged from 82 to 98 people. The baseline scenario with four check-in stations violated performance metrics for wait times, with the potential for wait times exceeding 2 min 19% of the time (see Figure 11). However, the model demonstrated a maximum wait time of 5.6 min with a 0.32% probability. With five check-in stations, wait time for check-in of more than 2 min is reduced to 1% (see Figure 12). The additional check-in station resulted in a lower total floor census (less congestion) and lower wait times for check-in and registration but also resulted in lower station use.

Figure 10.

Use of check-in and registration stations for Scenario 1A (left) and Scenario 1B (right generated by the simulation model. Data are partitioned by the hour of the day. The model was constrained to keep utilization below 80%. Results are aggregated over 32 replications.

Figure 11.

Frequency of wait times for Scenario 1 (A and B).

Figure 12.

Frequency of wait times for Scenario 2 (A and B).

Scenario 2 considered the 50% digital and 50% in-person registration mix and compared four (2A) versus five (2B) check-in stations. In Scenario 2A, the average use of check-in stations ranged from 51% to 68% during peak hours with a maximum of 80% at peak hours. The census for the seating area reached a maximum of 69–82 people during peak hours. Wait time for check-in exceeded 2 min just 2% of the time and the longest wait time recorded was 2.6 min with 0.11% probability. For Scenario 2B, the average use of stations ranged from 41% to 54% with a maximum of 64%. The census for the seating area reached a maximum of 63–74 people. Wait time for check-in exceeded 2 min 0.03% of the time and the maximum wait time was 1.6 min with 0.06% probability.

All scenarios provided evidence that the seating availability capacity of 120 chairs determined by the test fits was sufficient and could be reduced to 110 seats to provide more generous seating options for guests. Likewise, because square footage to accommodate up to eight stations was available, the model demonstrated the ability to meet service quality indicators even at the current baseline digital registration rate or lower. The low use of check-in stations during the early morning and later afternoon showed opportunities for flexible future state resource allocation strategies. The primary benefit of a higher digital adoption rate would be to reduce the check-in station requirement from five to four, which could be implemented incrementally as digital adoption increased over time. As seen in Figure 13, increasing preregistration adoption to 50% from 30% could be expected to reduce seating demand in the waiting area by 32% on average (from 76 to 52 seats) and maximum floor census by 26% (111–82 persons) when four check-in stations were in use. With five check-in stations, increasing preregistration adoption from 30% to 50% could be expected to reduce seating demand in the waiting area by 23% (96–74 seats) on average and reduce maximum floor census by 20% (69–55 persons). These results, detailed in Table 4, show that as the number of registration stations increases, there is a higher demand for seats. This is because patients and companions can complete the registration process more quickly, allowing them to move to the waiting area until their appointment.

Figure 13.

Total floor census (seating demand) for each of the four scenarios tested. Solid lines show the average number of people on the floor, and the dashed line represents the maximum number of people on the floor (including patients and companions). Aggregated over 32 replications. Under the current volume, arrival patterns, service times, and companion assumptions, five check-in stations and 110–120 seats will accommodate incoming patients during peak hours under the baseline preregistration adoption of 30% (70% of arriving patients have additional paperwork to complete).

Table 4.

Results of Experimental Scenarios.

Experimental	Patients Using Digital Check-In (%)	Patients Using In-Person Check-In (%)	Number of Check-In Stations	Maximum Floor Census	Average Utilization of Check-In Stations	Seating Demand in Waiting Area
Scenario 1A	30	70	4	100–111	51%–68%	76
Scenario 1B	30	70	5	82–98	41%–54%	96
Scenario 2A	50	50	4	69–82	51%–68%	52
Scenario 2B	50	50	5	63–74	41%–54%	74

A combined strategy of increasing check-in stations to five and reaching 50% adoption rates results in the lowest overall congestion in terms of floor census and seating use. The trade-off for more check-in stations is lowered check-in staff use rates.

Discussion

The modeling provided robust evidence of the feasibility of the centralized process to accommodate the expected daily patient and companion throughput of 2,000 people per day, even if digital preregistration and preregistration rates remain at the baseline level of 30%. Assuming a higher preregistration rate is reasonable with a well-defined strategy for increased adoption. This involves educational campaigns, digital communication channels, incentives, and streamlined processes to encourage patients. The modeling scenarios also revealed that variability in patient arrival patterns is an important factor in influencing process flows during registration, whereas digital registration adoption will influence general congestion or crowding. The modeling demonstrated the benefits of cross-trained admission staff and opportunities for achieving reasonable use of staff while meeting patient experience objectives. The simulation model allowed the team to quantify average use, wait times, and congestion over small time increments of 5–15 min and compare them to expected maximum values over the same increments. The ability to understand the flow dynamics in a more granular and intuitive way to see a “picture” of patient flows throughout the day was identified as a key benefit by the team.

As a DLSO, the process engaged an interdisciplinary group of clinical, design, and operational team members to imagine, design, and test the new centralized registration concept, as well as to gain insights into and begin planning for future state operations. The team’s commitment to integrating the voice of the multidisciplinary team members into the process was reflected in the highly iterative and transparent process, which brought multiple stakeholders into the design and analysis of options. Stakeholders were actively involved in defining processes, potential scenarios, and key performance targets. The modeling was conducted during the early design process, providing real-time analytics to inform the space planning, while also contributing to consensus-building around pursuing the centralized registration model. The hybrid DES and optimization approach during early design allowed the team to explore multiple scenarios to optimize operations and capacity requirements for combined digital and in-person registration as well as seating demand of a lobby level for a multispecialty clinic while also accounting for risk tolerance. The results allowed the architectural team to proceed to greater levels of design details with confidence. The incorporation of modeling from the initial stages of the conversation provided a framework for the development of innovative ideas and the rigorous testing and validation of their feasibility.

Limitations

While this study and using DES was useful for assessing the capacity of central registration in an MOB, it is not without its limitations. One of the primary limitations of DES that might affect the results of the capacity study is that it does not consider patient behavior. While this study accounted for late or early arrivals and demand fluctuations, it did not consider patients’ behavior such as being unwilling or unable to follow the central registration process. Second, DES may not be able to account for changes in the central registration process itself. For example, if a new technology is introduced to the central registration process, DES may not be able to accurately predict how this change will affect the overall capacity. Third, the full validation of the numerical results is challenging as the results can be affected greatly by the choice of variable parameters, representative scenarios, and uncertainty drivers. For future work, a more robust distribution-based prediction approach for patients’ arrivals will be tested and the generalization ability of the process to other similar systems and settings will be considered.

The results of the study indicate that a centralized registration with cross-trained staff can be optimized for clinics with a mix of patients who have the option to check-in in person or via self-service technolog.

The collaborative nature of simulation modeling provided direct decision inputs for the design team to identify the most efficient space design and validate estimates on the seating capacity in the wellness lounge, while considering the constraints and concerns expressed by the larger project team.

Practical Implications

Leveraging simulation modeling to validate innovative operational and design concepts: For novel operational and design processes, the simulation modeling of a system and testing various scenarios enables teams to refine innovative concepts, deepen understanding of potential risks and trade-offs, and gain insights into system interactions. DLSO incorporates extensive consultation with stakeholders and subject–matter experts within departments to validate the feasibility of newly proposed future state concepts. This promotes co-design and buy-in among multidisciplinary team members, creating more viable solutions.

Optimizing resources with DLSO for enhanced performance: allowing for the right-sizing of space and resources to meet performance goals. The results indicate optimization potential for clinics with a mix of check-in options, emphasizing the feasibility of centralized registration with cross-trained staff. Iterative team discussions enhance preparedness for future state operations, while data analysis of historical data provides insights into peak load and utilization. Multidisciplinary team meetings explore operational changes required for the new facility.

Promoting low-risk and cost-effective exploration of various design and operational configurations through DLSO: DLSO can be accomplished without extensive additional costs to the project, as the modeling is fully integrated into the planning process, particularly with access to a simulation specialist. Multiple scenarios, including stretch or “wild ideas,” can be tested in a low-risk, low-cost environment. However, it does require engagement from the full range of stakeholders, including financial, facility, clinical, ancillary, data management, security, and simulation specialists, among others. Access to volumetric and arrivals data is ideal, and gaps can be supplemented through observation, time studies, and subject–matter expertise.

Preparing teams for change management early: DLSO provides a space for the exchange of ideas and functional needs and practicalities of implementing new processes among multiple stakeholder groups. Building models and scenarios is an opportunity to uncover assumptions about staffing, space, or resource availability early and prepare the teams to pilot and prepare for change long before building occupancy.

Supplemental Material

Supplemental Material, sj-pdf-1-her-10.1177_19375867241237504 - Innovate and Validate: Design-Led Simulation Optimization to Test Centralized Registration Feasibility in a Multispecialty Clinic

Supplemental Material, sj-pdf-1-her-10.1177_19375867241237504 for Innovate and Validate: Design-Led Simulation Optimization to Test Centralized Registration Feasibility in a Multispecialty Clinic by Maryam Hosseini, Alice M. Gittler, Michael Hoak, Jonathan Cogswell and Mohammad T. Khasawneh in HERD: Health Environments Research & Design Journal

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Maryam Hosseini

Alice M. Gittler

References

Aboueljinane

Frichi

(2022). A simulation optimization approach to investigate resource planning and coordination mechanisms in emergency systems. Simulation Modelling Practice and Theory, 119, 102586. https://doi.org/10.1016/j.simpat.2022.102586

Al-Kaf

Jayaraman

Demirli

Simsekler

M. C. E.

Ghalib

Quraini

Tuzcu

(2022). A critical review of implementing lean and simulation to improve resource utilization and patient experience in outpatient clinics. TQM Journal, 35(3), 734–758. https://doi.org/10.1108/TQM-11-2021-0337

Alvarado

M. M.

Cotton

T. G.

Ntaimo

Pérez

Carpentier

W. R.

(2018). Modeling and simulation of oncology clinic operations in discrete event system specification. Simulation, 94(2), 105–121.

Barros

Weber

Reveco

(2021). Demand analysis and capacity management for hospital emergencies using advanced forecasting models and stochastic simulation. Operations Research Perspectives, 8. https://doi.org/10.1016/j.orp.2021.100208

Cai

Jia

(2019). Using Discrete Event Simulation (DES) to support performance-driven healthcare design. Health Environments Research & Design Journal, 12(3), 89–106. https://doi.org/10.1177/1937586718801910

Demirli

Al Kaf

Simsekler

M. C. E.

Jayaraman

Khan

M. J.

Tuzcu

E. M.

(2021). Using lean techniques and discrete-event simulation for performance improvement in an outpatient clinic. International Journal of Lean Six Sigma, 12(6), 1260–1288. https://doi.org/10.1108/IJLSS-09-2020-0138

Easter

Houshiarian

Pati

Wiler

J. L.

(2019). Designing efficient emergency departments: Discrete event simulation of internal-waiting areas and split flow sorting. American Journal of Emergency Medicine, 37(12), 2186–2193. https://doi.org/10.1016/j.ajem.2019.03.017

Fun

W. H.

Tan

E. H.

Khalid

Sararaks

Tang

K. F.

Ab Rahim

Md Sharif

Jawahir

Sibert

R. M. Y.

Nawawi

M. K. M.

(2022). Applying discrete event simulation to reduce patient wait times and crowding: The case of a specialist outpatient clinic with dual practice system. Healthcare (Basel), 10(2), 189. https://doi.org/10.3390/healthcare10020189

Gosavi

Cudney

E. A.

Murray

S. L.

Masek

C. M.

(2016). Analysis of clinic layouts and patient-centered procedural innovations using discrete-event simulation. Engineering Management Journal, 28(3), 134–144. https://doi.org/10.1080/10429247.2016.1195664

10.

Halawa

Madathil

S. C.

Gittler

Khasawneh

M. T.

(2020). Advancing evidence-based healthcare facility design: A systematic literature review. Health Care Management Science, 23(3), 453–480. https://doi.org/10.1007/s10729-020-09506-4

11.

Kubala

Gardner

J. R.

Criddle

Ward

Richter

G. T.

(2021). Utilizing process improvement strategies to generate clinic templates and improve patient flow in pediatric otolaryngology. International Journal of Pediatric Otorhinolaryngology, 147, 110779. https://doi.org/10.1016/j.ijporl.2021.110779

12.

Maass

K. L.

Halter

Huschka

T. R.

Sir

M. Y.

Nordland

M. R.

Pasupathy

K. S.

(2022). A discrete event simulation to evaluate impact of radiology process changes on emergency department computed tomography access. Journal of Evaluation in Clinical Practice, 28(1), 120–128. https://doi.org/10.1111/jep.13606

13.

Ponis

S. T.

Delis

Gayialis

S. P.

Kasimatis

Tan

(2013). Applying Discrete Event Simulation (DES) in healthcare: The case for outpatient facility capacity planning. International Journal of Healthcare Information Systems and Informatics (IJHISI), 8(3), 58–79.

14.

Saadouli

Ltaif

(2021). Evaluating the impact of human resource management on the patient flow at an outpatient orthopedic clinic. International Journal of Healthcare Management, 14(4), 1492–1500. https://doi.org/10.1080/20479700.2020.1788335

15.

Sargent

R. G.

Balci

(2017). History of verification and validation of simulation models. In 2017 Winter Simulation Conference (WSC), pp. 292–307. IEEE. https://doi.org/10.1109/WSC.2017.8247794

16.

Tyler

J. M.

Murch

B. J.

Vasilakis

Wood

R. M.

(2022). Improving uptake of simulation in healthcare: User-driven development of an open-source tool for modelling patient flow. Journal of Simulation, 17(6) 1–18. https://doi.org/10.1080/17477778.2022.2081521

17.

Vahdat

Namin

Azghandi

Griffin

(2019). Improving patient timeliness of care through efficient outpatient clinic layout design using data-driven simulation and optimisation. Health Systems, 8(3), 162–183. https://doi.org/10.1080/20476965.2018.1561160

18.

Vahdatzad

(2018). Simulation and optimization methods applied to outpatient health center operations [Northeastern University]. https://repository.library.northeastern.edu/files/neu:m044c567w/fulltext.pdf

19.

Valipoor

Hakimjavadi

Nobles

P. M.

(2022). Toward building surge capacity: Potentially effective spatial configurations in emergency departments. Health Environments Research & Design Journal, 15(3), 42–55. https://doi.org/10.1177/19375867221096639

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB