2018 Volume 41 Issue 3 Pages 409-418
Improvement in patient waiting time in dispensing pharmacies is an important element for patient and pharmacists. The One-Dose Package (ODP) of medicines was implemented in Japan to support medicine adherence among elderly patients; however, it also contributed to increase in patient waiting times. Given the projected increase in ODP patients in the near future owing to rapid population aging, development of improved strategies is a key imperative. We conducted a cross-sectional survey at a single dispensing pharmacy to clarify the impact of ODP on patient waiting time. Further, we propose an improvement strategy developed with use of a discrete event simulation (DES) model. A total of 673 patients received pharmacy services during the study period. A two-fold difference in mean waiting time was observed between ODP and non-ODP patients (22.6 and 11.2 min, respectively). The DES model was constructed with input parameters estimated from observed data. Introduction of fully automated ODP (A-ODP) system was projected to reduce the waiting time for ODP patient by 0.5 times (from 23.1 to 11.5 min). Furthermore, assuming that 40% of non-ODP patients would transfer to ODP, the waiting time was predicted to increase to 56.8 min; however, introduction of the A-ODP system decreased the waiting time to 20.4 min. Our findings indicate that ODP is one of the elements that increases the waiting time and that it might become longer in the future. Introduction of the A-ODP system may be an effective strategy to improve waiting time.
Patient waiting time at the point of health care service delivery is a key determinant of patient satisfaction, and is used as one of the indicators for evaluation of the performance of health care facilities.1–5) In Japan, segregation of medicine prescription and dispensing functions has been implemented nationwide, and the number of pharmacies and the engaged pharmacist has increased over the past 10 years. According to a report by the Ministry of Health, Labor, and Welfare of Japan, the number of pharmacies have increased from 51233 (40.1 per 100000 population) in 2005 to 58326 (45.9 per 100000 population) in 2015, which corresponds to an increase of 113.8%.6) Similarly, the number of pharmacists has increased from 116303 (91.1 per 100000 population) in 2004 to 161198 (126.8 per 100000 population) in 2014, which corresponds to an increase of 138.6%.7) Over the years, the role of dispensing pharmacy has expanded from mere dispensing of medicines to patient education and management, which has helped improve the quality of their services. Improvement in patient waiting time is an important element as it can negatively affect both patient satisfaction and efficiency of the work of pharmacists. Patients who visit dispensing pharmacies have multifaceted needs. In particular, elderly patients are typically prescribed multiple medicines owing to several comorbid conditions, which increases the risks associated with overlapping prescriptions and drug interactions. Moreover, these patients require additional support to improve their medicine adherence.8) Several studies have documented poor medicine adherence among the elderly patients. In a community-based cohort study of 1722 Japanese elderly patients, 12.6% of patient were found to be non-adherent (adherence rate ˂80%) to the prescribed treatment.9) Similarly, in a cohort study (n=18645; mean age: 60 years) by Origasa et al., 30.5% of patients exhibited poor adherence to prescribed medicines.10) Provision of One-Dose Package (ODP) has been used as a strategy to promote medication adherence. It entails provision of multiple medicines as a single package, depending on prescribed time of administration, such as after meals or between meals.11) ODP is typically dispensed by pharmacists, based on physician’s instructions or on the request of patients, using an ODP system. Broadly, there are two types of ODP systems used in Japan: the semi-manual ODP (M-ODP) system and full-automated ODP (A-ODP) system. The M-ODP system requires the pharmacist to manually input the prescribed medicines in the system, while the A-ODP system allows for automated packaging of the prescribed medicines based on a pre-configured system. The impact of the use of ODP system on patient waiting time has not been adequately studied. Due to the projected increase in the need for ODP system owing to rapid population aging, assessment of its impact is an important issue for dispensing pharmacies.
Discrete event simulation (DES) is a computer simulation modeling technique that provides an intuitive and flexible approach to study complex systems.12–14) The DES is known to be a powerful tool to support evidence-based decision-making in a risk-free environment. It has been used in studies in the field of hospital management, health resource planning, and to improve patient flow and waiting time at hospitals.15–21) However, relatively few studies have focused on pharmacy systems.22–24)
In the present study, we aimed to clarify the impact of ODP on patient waiting time by conducting a survey at a dispensing pharmacy, and propose an improvement strategy with use of DES model. Furthermore, we predicted the waiting time and suggest improvement strategies assuming an increase in number of patients who will require ODP in the future.
A 5-d survey was conducted from November 7 to 11, 2016 (weekdays) at a single dispensing pharmacy located next to circulatory disease special hospital in Osaka City. The dispensing pharmacy operates from 9:00 to 19:00 h on weekdays, and from 9:00 to 15:00 h on Saturdays. A total of 10 pharmacists and 3 medical clerks are employed at the pharmacy, of which 8 pharmacists and 2 medical clerks worked during the study period. There are 2 units of M-ODP systems available for implementation of ODP. While using the M-ODP system, the pharmacists need to set the medicines of one dose unit at manually into the system according the patients prescriptions, and after that, the medicine packaging process was done automatically.
Workflow at the Dispensing PharmacyThe workflow at the dispensing pharmacy includes the following events: (1) Patient arrival: Patients bring their prescription to pharmacy; (2) Prescription audit: Pharmacist checks the prescription and clarifies any doubts from the physician; (3) Dispensing medicines: Pharmacist dispenses the prescribed medicines and implements ODP with the M-ODP system, if applicable; (4) Inspection medicines: Dispensed medicines are independently verified against the prescription by another pharmacist; (5) Medicine counseling: Pharmacist provides necessary information to the patient such as dosage, side effects, and the recommended precautions; (6) Patient departure: Patient leaves the dispensing pharmacy after payment.
Data AcquisitionThe descriptive data for all arrival patients during the study period were collected from the questionnaire which is filled by the pharmacist. Data pertaining to the following variables were collected: age classification (<65, ≥65); number of prescribed medicines (<3 items, 3–7 items, ≥7 items); prescription period; status of ODP implementation; status of prescription audit; patient arrival time; dispensing start time; dispensing end time; dispensing inspection start time; dispensing inspection end time; medicine counseling start time; medicine counseling end time.
Waiting Time DefinitionThe patient waiting time was defined as the time from patient arrival until the end of medicine counseling. Waiting times were separately recorded for patients who received medicines with ODP (ODP-patient) and those who received medicines without ODP (Non-ODP patient).
Simulation ApproachThe conceptual DES model for dispensing pharmacy was developed using the Simul8 Professional Software package (2016, SIMUL8 Corporation, U.S.A.). The theoretic basis and explanation of the DES model can be found elsewhere.12–14) Briefly, the Simul8 enables the replicates of a discrete event or behavior in a real-life process in the computer model, and allows the performance of various experiments on the computer. The model was constructed using the components of Entity, Resource, Activity, Queue, Start point, and End point. The Entity is the item that flows through the simulation model (in this case, the patient’s prescription). The Resource is the items that are required in order for the Activity to work (pharmacist and ODP system). The Activity is the place where work takes place by specified Resource (dispensing of medicines, inspection of medicines, and medicine counseling). The Queue is the place where Entities have to wait until the appropriate Resource or Activity becomes available. The Start point is the entry points for entities (in this case, patient arrival). The End point is the exit points for entities (patient leaving the dispensing pharmacy). Patient’s prescription is modeled as individuals with specific characteristics of age, number of medicines, status of prescription audit, and ODP. These characteristics influence the dispensing process and the flow of the patient through the model.
The DES model is constructed according to the flow chart of prescription is described in Figs. 1 and 2. At the Fig. 1, in the case of implementation of ODP, the dispensing medicine time and inspecting medicine time partially overlapped in our survey data, because there were some cases of conducting the dispensing medicine and inspecting it at the same time. Therefore, we defined the model as integrate the both dispensing medicine time and inspect medicine time. In conjunction with the probability distribution of parameters, these components are able to provide a robust representation of the actual operating conditions of dispensing pharmacy as a function of time. One working day (9:00 to 19:00 h) at the dispensing pharmacy was simulated, with 200 replications per simulation strategy; incomplete work after 19:00 h was discarded per replication. The outcome measures were the mean waiting time [with 95% confidence intervals (95% CI)] for ODP and non-ODP patients, and the percentage of patients who completed the workflow within 15 min.
The key parameters inputted to the model were calibrated with observational data from the survey. Several parameters of the model are associated with probability distributions, given the variability and uncertainty associated with some parameters. The probability distributions were estimated with the help of the StatFit software (2016, Geer Mountain Software Co., U.S.A.) used in conjunction. The mean patient arrival rate per 1 h was calculated with the underlying assumption of an exponential distribution. The model was validated following the general method, to confirm the workflow, logic, input parameters, and by running the model step-by-step to confirm the logical path of a single patient and double-checking.23) Subsequently, the output of mean, maximum and minimum waiting time and the percentage of patients who completed within 15 min from the baseline model was compared against the observed data, which was the commonly method in similar research with using the DES model.22–24)
Simulation Approach with ScenariosWe adopted two simulation approaches in this study. Simulation 1: simulation for considering the strategy to improve the current waiting time; Simulation 2: simulation for predicting the future waiting time and considering the improving strategy. For Simulation 1, the following three scenarios are presented which involve different combinations of the number of ODP systems and the pharmacists:
For Simulation 2, the simulations were performed assumes the transfer of the current non-ODP patients to ODP according to the ODP transfer rates. The ODP transfer rate has changed from 0 to 40% in 10% increments; we defined these as Prop.0 (ODP transfer rate at 0%, reference), Prop.1 (10%), Prop.2 (20%), Prop.3 (30%), Prop.4 (40%), respectively. Also, the following three scenarios which envisage different number of ODP systems are presented (the number of pharmacists was fixed at 8):
The study protocol is approved by the Ethical Committee of the Faculty of Medicine, Osaka University (Approval number: 16142, date: 24 July 2016).
A total of 673 patients presented at the pharmacy during the study period and data was collected for all patients. A summary of the survey data is presented in Table 1. As shown in table, ODP patients accounted for 18% of all patients. Further, 81.8% of ODP patients were >65 years of age. The mean prescription period for ODP patients was 67.1 d and was 26% longer than that for non-ODP patients (48.4 d). For calculation of the waiting time, 44 patients were excluded as these patients went out for a long time after presenting the prescription to the pharmacist. The mean waiting time (total patients) was 13.2 min and the percentage of patients who completed the workflow within 15 min was 70.3%. The waiting time for ODP patient was 2.0 times longer than that for non-ODP patients (22.6 and 11.2 min, respectively). Also, the percentage of ODP patients who completed within 15 min was only 18.6%. Figure 3 shows the arrival characteristics of patients at different time of the day. On average, 134.4 patients arrived per day; the first peak was observed from 10:00 to 11:00 h and the second peak was observed from 14:00 to 15:00 h. Patient arrival reached a trough between 18:00 and 19:00 h. No ODP patients were encountered after 18:00 h.
| Variable | ODP patient (n=121, 18.0%) | Non-ODP patient (n=552, 82.0%) | Total (n=673, 100%) |
|---|---|---|---|
| Age | |||
| <65 years, n (%) | 22 (18.2) | 305 (55.2) | 327 (48.6) |
| ≥65 years, n (%) | 99 (81.8) | 247 (44.8) | 346 (51.4) |
| Prescription audit, n (%) | 23 (19.0) | 43 (7.8) | 57 (18.0) |
| Prescription medicine | |||
| <3 items, n (%) | 0 (0.0) | 257 (46.6) | 257 (38.2) |
| 3–7 items, n (%) | 44 (36.4) | 235 (42.6) | 279 (41.5) |
| >7 items, n (%) | 77 (65.6) | 60 (10.9) | 137 (20.4) |
| Prescription period (day), Mean (S.D.) | 67.1 (21.1) | 48.4 (33.0) | 51.8 (32.0) |
| Waiting time (min) | |||
| Mean (S.D.) | 22.6 (9.37) | 11.2 (6.93) | 13.2 (8.6) |
| Median | 20 | 9 | 11 |
| Minimum | 8 | 4 | 4 |
| Maximum | 51 | 39 | 51 |
| Patients who complete the workflow within 15 min (%) | 18.6 | 81.2 | 70.3 |
The mean (Standard deviation) of arriving patients during the operation time from 9:00 to 18:00 per 1 h were, 7.6 (2.0), 19.4 (2.4), 14.6 (6.2), 15.8 (5.8), 13.2 (3.8), 18.0 (3.2), 17.0 (5.3), 15.0 (3.8), 9.0 (3.3), 4.8 (2.0) at overall patient respectively. Similarly, 1.2 (0.4), 3.6 (1.5), 1.6 (1.0), 3.6 (1.5), 2.6 (1.4), 3.8 (1.5), 4.2 (1.2), 2.8 (1.3), 0.6 (0.5), 0.0 (0.0) at ODP patient, and 6.4 (1.9), 15.8 (1.0), 13.0 (5.6), 12.2 (4.4), 10.6 (2.9), 14.2 (2.5), 12.8 (5.0), 12.2 (3.7), 8.4 (3.2), 4.8 (2.0) at Non-ODP patient.
The input parameters for the simulation model which are calculated from the observation data are described in Table 2. The probability distribution for dispensing medicine time, inspection medicine time, and medicine counseling time were estimated to follow the lognormal distribution. Corresponding average statistics for the observational data and the simulation results of baseline model are presented in Table 3. The results of mean waiting time were comparable to those observed; the difference was less than 1 min. Similarly, there were minor differences with respect to patient arrival, minimum and maximum waiting times and the percentage of patient who completed the workflow within 15 min.
| Variable | Values and distribution (mean, S.D.) | |
|---|---|---|
| Daily time-dependent patient arrival rate | Exponential distribution each hourly interarrival rate (Figure) | |
| Number of medications per prescription | Table 1 | |
| Percentage of Patient age | Table 1 | |
| Percentage of prescription audit | Table 1 | |
| Times (min) | ||
| Dispensing medicine for Non-ODP patient | Lognormal. (2.1, 0.9); <3 items. | Lognormal. (3.4, 2.6); <3 items.a) |
| Lognormal. (3.8, 1.7); 3–7 item. | Lognormal. (6.1, 3.4); 3–7 item.a) | |
| Lognormal. (6.1, 2.1); >7 item. | Lognormal. (9.0, 3.6); >7 item.a) | |
| Dispensing medicine for ODP patient | Lognormal. (11.6, 4.1); 3–7 item. | Lognormal. (16.0, 10.3); 3–7 item.a) |
| Lognormal. (18.2, 5.5); >7 item. | Lognormal. (21.6, 11.0); >7 item.a) | |
| Inspection medicine | Lognormal. (2.8, 0.9); <3 item. | Lognormal. (5.0, 3.7); <3 item.a) |
| Lognormal. (4.3, 2.0); 3–7 item. | Lognormal. (6.3, 2.8); 3–7 item.a) | |
| Lognormal. (7.4, 2.6); >7 item. | Lognormal. (11.4, 2.6); >7 item.a) | |
| Medication counseling | Lognormal. (3.2, 1.6); <65 age. | Lognormal. (5.2, 6.4); <65 age.a) |
| Lognormal. (4.0, 2.2); ≥65 age. | Lognormal. (9.1, 9.0); ≥65 age.a) | |
| Resources | ||
| Number of Pharmacists | Maximum available=8 | |
| Number of M-ODP system (N) | Maximum available=2 | |
a) Probability distribution for prescription audit case.
| Number of mean arrival patient | Patient waiting time, min | % Patient complete within 15 min | |||
|---|---|---|---|---|---|
| Mean | Minimum | Maximum | |||
| ODP Patient | |||||
| Observationsa) | 24.2 | 22.6 | 8.0 | 51.0 | 18.6 |
| Baseline modelb) | 22.5 (21.7–23.2) | 23.1 (22.6–23.5) | 10.2 (9.9–10.4) | 48.3 (46.1–50.6) | 19.8 (18.5–21.0) |
| Difference | −7.0% | +2.2% | +27.5% | −5.3% | +6.5% |
| Non-ODP patient | |||||
| Observationsa) | 103.2 | 11.2 | 4.0 | 39.0 | 81.2 |
| Baseline modelb) | 108.1 (106.6–109.6) | 11.5 (11.5–11.6) | 4.8 (4.7–4.9) | 35.3 (33.6–36.6) | 81.6 (81.0–82.2) |
| Difference | +4.7% | +2.7% | +20.0% | −9.2% | +0.5% |
a) 5 d mean at each variables. b) 95% confidence intervals in parentheses.
The results of Simulation 1 are presented in Table 4. In scenario 1, the reduction in waiting time for both ODP and non-ODP patients was not associated with increase in the number of pharmacists. Conversely, the waiting time increased with decrease in the number of pharmacists to <8; the maximum observed increase at the level of 6 pharmacists was 1.02 times (from 23.1 to 23.6 min) for ODP patients; and 1.06 times (from 11.5 to 12.3 min) for non-ODP patients. On the other hand, in scenario 2, the waiting time for ODP patients reduced by approximately 0.94 times from the reference level, for every single increase in the number of pharmacists. The largest predicted reduction in waiting time for ODP patient was observed in scenario 3 (approximate reduction by 0.5 times from the reference). The association between the waiting time and increase in the number of pharmacists was also observed in scenario 1. The waiting time for non-ODP patients hardly changed in all scenarios. The percentage of patients who completed the work flow within 15 min in each scenario is shown in Fig. 4. The percentages in scenario 1 and scenario 2 were comparable, but were approximately 10% higher at scenario 3 (78.9, 81.3, 82.0, 82.1, 82.2% with 6, 7, 8, 9, and 10 pharmacists, respectively).
| Number of pharmacists | Scenario 1 | Scenario 2 | Scenario 3 | |||
|---|---|---|---|---|---|---|
| Waiting time, Meana) | Rate of change from reference | Waiting time, Meana) | Rate of change from reference | Waiting time, Meana) | Rate of change from reference | |
| ODP patient | ||||||
| 6 | 23.6 (23.1–24.0) | 1.02 | 22.4 (21.9–22.8) | 0.97 | 12.4 (12.1–12.7) | 0.54 |
| 7 | 23.1 (22.7–23.6) | 1.00 | 21.7 (21.3–22.1) | 0.94 | 11.8 (11.5–12.1) | 0.51 |
| 8 | 23.1 (22.6–23.5) | Reference | 21.6 (21.2–22.0) | 0.94 | 11.7 (11.4–12.0) | 0.50 |
| 9 | 23.0 (22.6–23.5) | 1.00 | 21.6 (21.2–22.0) | 0.94 | 11.6 (11.3–11.9) | 0.50 |
| 10 | 23.0 (22.6–23.5) | 1.00 | 21.6 (21.1–22.0) | 0.93 | 11.5 (11.3–11.9) | 0.50 |
| Non-ODP patient | ||||||
| 6 | 12.3 (11.9–12.6) | 1.06 | 12.4 (12.1–12.7) | 1.08 | 12.0 (11.6–12.3) | 1.04 |
| 7 | 11.7 (11.4–12.0) | 1.01 | 11.7 (11.4–12.0) | 1.02 | 11.6 (11.3–11.9) | 1.01 |
| 8 | 11.5 (11.2–11.8) | Reference | 11.5 (11.2–11.9) | 1.00 | 11.5 (11.2–11.8) | 1.00 |
| 9 | 11.5 (11.2–11.8) | 1.00 | 11.5 (11.2–11.8) | 1.00 | 11.5 (11.2–11.8) | 1.00 |
| 10 | 11.5 (11.2–11.8) | 1.00 | 11.5 (11.2–11.8) | 1.00 | 11.5 (11.2–11.8) | 1.00 |
a) Numbers in parentheses are 95% confidence interval.
The percentages at scenario 1 from 6 to 10 pharmacists were 67.2, 70.0, 70.9, 71.1, and 71.1%, respectively. Similarly, 66.9, 70.2, 71.2, 71.5, and 71.5% at scenario 2, and 78.1, 80.7, 81.5, 81.6, and 81.6% at scenario 3.
The results of Simulation 2 are presented in Table 5. The waiting time for ODP patient at scenario 4 increased 2.46 times (from 23.1 to 56.8 min) during the Prop.0 to Prop.4. A moderate increase (1.26 times) in waiting time for ODP patients was observed in scenario 5 (increased from 21.6 to 29.0 min). The waiting time for ODP patient in scenario 5 was less than the reference time of 23.1 min until the Prop.1. The results indicate that the waiting time was more or less maintained until 10% of Non-ODP patients had transferred to ODP. The waiting time for ODP patients in scenario 6 ranged from 11.7 to 20.4 min, and were never below the reference level. In all three scenarios, the waiting time for non-ODP patients showed almost no change from the reference waiting time of 11.5 min. In order to clarify these results, we confirmed the model and found that non-ODP patients in the model were preferentially processed regardless of the queue of the ODP patients, if the pharmacists were available. This preferential processing is consistent with the process observed in actual dispensing pharmacy. The changes in the percentage of patients who completed within 15 min are shown in Fig. 5. The percentages in scenario 3 were approximately 10 to 20% higher than those in scenario 1 and scenario 2 (82.0, 80.2, 77.4, 72.8 and 65.9% during Prop.0, Prop.1, Prop.2, Prop.3 and Prop.4, respectively).
| Transfer rate | Number of arrival patients, Meana) | Scenario 4 | Scenario 5 | Scenario 6 | |||
|---|---|---|---|---|---|---|---|
| Waiting time, Meana) | Rate of change from reference | Waiting time Meana) | Rate of change from current | Waiting time, Meana) | Rate of change from reference | ||
| ODP patient | |||||||
| Prop.0 | 22.5 (21.8–23.3) | 23.1 (22.6–23.5) | Reference | 21.6 (21.3–21.9) | 0.94 | 11.7 (11.5–11.9) | 0.51 |
| Prop.1 | 34.4 (33.4–35.3) | 26.2 (25.5–26.8) | 1.13 | 22.3 (22.1–22.6) | 0.97 | 12.7 (12.5–12.9) | 0.55 |
| Prop.2 | 45.0 (44.0–45.9) | 31.5 (30.1–32.8) | 1.36 | 23.4 (23.0–23.7) | 1.01 | 14.0 (13.7–14.3) | 0.61 |
| Prop.3 | 55.7 (54.7–56.8) | 42.0 (39.9–44.2) | 1.82 | 25.5 (24.9–26.0) | 1.10 | 16.5 (15.9–17.0) | 0.71 |
| Prop.4 | 66.3 (65.2–67.4) | 56.8 (54.1–59.5) | 2.46 | 29.0 (28.0–30.0) | 1.26 | 20.4 (19.5–21.4) | 0.89 |
| Non-ODP patient | |||||||
| Prop.0 | 108.1 (106.6–109.6) | 11.5 (11.5–11.6) | Reference | 11.5 (11.5–11.6) | 1.00 | 11.5 (11.4–11.6) | 1.00 |
| Prop.1 | 96.3 (94.9–97.7) | 11.6 (11.5–11.6) | 1.00 | 11.6 (11.5–11.7) | 1.01 | 11.5 (11.4–11.6) | 1.00 |
| Prop.2 | 85.7 (84.4–84.0) | 11.6 (11.5–11.7) | 1.00 | 11.6 (11.5–11.7) | 1.01 | 11.5 (11.5–11.6) | 1.00 |
| Prop.3 | 74.9 (73.7–76.1) | 11.6 (11.5–11.7) | 1.00 | 11.7 (11.6–11.5) | 1.01 | 11.5 (11.5–11.6) | 1.00 |
| Prop.4 | 64.3 (63.2–65.4) | 11.5 (11.4–11.6) | 1.00 | 11.6 (11.5–11.8) | 1.01 | 11.6 (11.5–11.7) | 1.00 |
a) Numbers in parentheses are 95% confidence interval.
The percentages at scenario 1 from Prop.0 to Prop.4 were 70.9, 64.2, 58.0, 50.8, and 44.8%, respectively. Similarly, 71.2, 65.3, 59.9, 53.9, and 47.9% at scenario 2, and 82.0, 80.2, 77.4, 72.8, and 65.9% at scenario 3
The waiting time at the health care service delivery point is one of the most important determinants of patient satisfaction, and a commonly used indicator for evaluation of the performance of the health care facilities.1–5,16,25–29) Moreover, long waiting times at dispensing pharmacies increases the workload of pharmacists and predisposes to dispensing errors or incidents.30) Expanding the role of pharmacists, from dispensing medicines, to patient education and management of patient adherence, and improving patient waiting time are important issues related to dispensing pharmacies.
In Japan, the ODP was adopted to facilitate dispensing of multi drug combinations to elderly patients with the objective to improve medication adherence.11) Although the increase in waiting time associated with ODP is not well documented, the process of ODP, which requires the use of mechanical system, is liable to increase the patient waiting time. Furthermore, considering the rapid population aging, the number of ODP patients is projected to increase in the near future. According to the Annual Report on the Aging Society of Cabinet Office, Government of Japan, about one in three people are estimated to be over 65 years of age and one in five people to be over 75 years of age in 2025.31) Therefore, this study was conducted to clarify the impact of ODP on waiting time in the current situation, and to inform future strategies. Computer simulation is an intuitive and flexible tool to replicate complex systems in a risk-free environment, and supports evidence based decision-making for policy formulation and management.32,33) Use of computer simulation in the field of health sciences offers a distinct leverage as real-world experiments are costly and expose patients and medical staff to potentially serious risks. Rapid advances in computer technology, software, and application tools have facilitated computer simulation studies in the fields such as hospital management, resource-planning, including those aimed at reduction in outpatient waiting time and optimizing of patient flow.34–38) However, relatively few studies have focused on pharmacy system. For example, Tan et al. used the DES model to evaluate the impact of automated dispensing system on patient waiting time and indicated that the introduction of the system helped reduce the waiting time.23) Reynolds et al. suggested a design for efficient hospital pharmacy for outpatients which changed the current prescription workload, staff level, and utilization of the automatic dispensing system.24) To the best of our knowledge, few studies have attempted to improve the issues related to ODP waiting time in dispensing pharmacy, and this study is probably the first to adapt the DES model for dispensing pharmacy. In the present study, the mean waiting time for all patient at the pharmacy was 13.6 min; 18% of all patients were ODP patients. The length of waiting time in this pharmacy was comparable to those reported for other dispensing pharmacies in Japan; Yoshida et al. reported a waiting time of 14.4 min,39) and Uesawa et al. reported waiting time range from 6 to 24 min in a single dispensing pharmacy.40) The proportion of ODP patients and age distribution of patients were similar to those reported by Nakai et al. from their study covering 86 pharmacies [percentage of ODP patient: 22.8% (14535 out of 63907); and 61% of ODP patient were >65 years of age)].11) A notable finding from our study is that the waiting time for ODP patients was about 2.0 times longer than that for non-ODP patients (22.5 and 13.5 min, respectively).
The DES model for dispensing pharmacy was constructed based on parameters estimated from the observed data. As results of model validation shows, the constructed model was representative of the actual situation of the dispensing pharmacy. In the simulation 1, we found that introduction of the A-ODP system was the most effective strategy for reducing the waiting time (Scenario 3) (waiting time for ODP patients reduced by approximately half). Furthermore, the predicted impact of additional pharmacists and M-ODP system on waiting time was found to be negligible. This result is probably attributable to the processing time and bottlenecks (obstacle) associated with the ODP system. The M-ODP system requires the medicines to be manually set into the ODP system, which increases the processing time. Therefore, the prescription of ODP involves waiting for processing if the all the M-ODP systems are occupied, which leads the occurrence of bottlenecks. On the other hand, the A-ODP system improved the processing time and is free from any constraining bottlenecks. Simulation 2 showed a gradual increase in waiting time with increased in the number of ODP patients. In the scenario comprising of 2 units of M-ODP system and 8 pharmacists (scenario 4), the waiting time for ODP patient increased 2.46 times (from 22.5 to 66.3 min). Although addition of the extra M-ODP system reduces the waiting time, it became longer than current waiting time, if the ODP transfer rate increased over 20% (Prop.2). On the other hand, the waiting time for ODP patients on introducing the A-ODP system (Scenario 6) did not show any increase even if the ODP transfer rate increased to 40% (Prop.4). It was noteworthy that the waiting time for non-ODP patients showed few changes, as these were preferentially processed regardless of the queue of the ODP patient, if the pharmacists were available. This preferential processing is consistent with that observed in real-world setting, which supports the reproducibility of the model.
To summarize the simulation results, introduction of the A-ODP system showed advantages in terms of reduction in waiting time in the current as well as projected future scenarios which were greater than those achieved with implementation of the M-ODP system or use of additional pharmacists in this pharmacy. As pointed out, a limitation of the M-ODP system is the limited ability to reduce waiting times owing to the need for manual processing of medicine settings in the system. In Japan, the scope of work of pharmacists has expanded into newer areas such as home medical care service (visiting pharmacist), function as a family pharmacy or health support pharmacy. In addition, population aging is a serious problem which is likely to increase ODP patients in the future. Our results emphasized the importance of understanding the current situation and analyzing the countermeasures for the future to increase the strength of pharmacy function. Also, we showed that the DES model was a powerful and helpful tool for adapting to and solving the problems of dispensing pharmacy.
This study should be viewed in the context of several limitations. First, the bias associated with a limited survey. Although we collected data from all patients, the study subjects were confined to a specific dispensing pharmacy and the study period was relatively small. Therefore, the results may not be generalizable to other dispensing pharmacies with different demographic characteristics of patients, pharmacists, and resources. Especially, the practice of direct transmission of prescriptions from hospital to pharmacy via internet or fax has increasingly been adopted in recent years; these changes are likely to affect the workflow as well as the results of simulation under these situations. Secondly, about the composition and assumption of the simulation model. Generally, it is not feasible to construct a simulation model that includes all features of the real world. The model used in this study was been kept constructively simple and included only important factors in dispensing pharmacy; some factors such as skills of pharmacist, time of machine maintenance, and inventory management were not incorporated in the model. In addition, it should be noted that the A-ODP system process time was estimated from the commercially available A-ODP system; data from actual use was not available. Although we used probability distribution to time parameters to retain the inherent random variation, the actual processing time for A-ODP might be different and the results might be over or underestimated. Further research is required to evaluate the validity of this assumption. Thirdly, economic analyses such as cost-effectiveness analysis was not included in this study. In this study, we only focused on to the patient waiting time and did not consider the cost of ODP systems or expenses for employing pharmacists. Thus, introduction of additional ODP system may possibly have unfavorable financial implications for dispensing pharmacies. Therefore, more realistic strategies may be devised to improve patient waiting time. In addition, implementing a new system is inherently costly and often requires substantial redesign of the workflow. The resultant changes in the workflow may not necessarily be the best strategy for improvement. In future work, these factors should be considered to develop more useful models for dispensing pharmacy.
In conclusion, we found that the waiting time for ODP patients was longer at the dispensing pharmacy and introduction of A-ODP system was an effective strategy for reduction of current and future waiting time. Also, we highlights that computer simulation based on discrete events may be useful for dispensing pharmacy and facilitate a better understanding of the waiting time as well as help improve strategies for decision-making within a short time scale and at minimal costs in a risk-free environment. Future study should test the present DES model on other dispensing pharmacies with different characteristics, and to produce results with better generalizability. Finally, we believe that some of the techniques and theories in this study might be applicable to other dispensing pharmacies which face similar problems.
We would like to thank the all staff members of dispensing pharmacy and all members of the Department of Mathematical Health Science, Osaka University for their in valuable support to this research. This research supported by The Japan Society for the Promotion of Science (JSPS) KAKENHI Grant Number JP16K21152.
The authors declare no conflict of interest.
