This was published in the Journal of General Internal Medicine, Volume 11, March 1996, pp. 182-184. An interesting P.S. in an e-mail sent me by Fred: P.S. People around here are congratulatory to us about getting an article in JGIM. It has a reputation for being tough to publish in. Yea. Say hi to all. --------------------------------------------------------------------- Improving Outpatient Clinic Staffing and Scheduling with Computer Simulation Fred Hashimoto, MD, Stoughton Bell, PhD ABSTRACT Patient flow in an appointment-based, outpatient internal medicine clinic involving multiple, sequential providers-- registrar, nurse, physician, and discharger--was studied using computer simulation. Provider task time distributions were obtained through a time-motion study and then input into the computer program, which simulated the clinic situation well. Time interval and sensitivity analyses yielded insights into staffing levels, appointment times and clinic dynamics. A bottleneck provider was shown, and patient time in clinic was related to time of appointment and slowed by having too many doctors in the clinic. Subsequent operational changes significantly decreased average, observed patient total time in clinic from 75.4 (s.d. 34.2) minutes to 57.1 (s.d. 30.2) minutes (p < 0.001, t-test). key words: outpatient, computer, simulation J GEN INTERN MED 1996:11:182-184 INTRODUCTION Coordinating patients and service in an outpatient general medical clinic is an old and continuing problem. The primary issue has been patient waiting in the clinic versus doctor and staff idle time (1-7). As health care has become a seriously competitive business, patient convenience including minimized waiting time has risen as a priority. Simultaneously, clinic facilities, which have grown and become more complex, need to be efficient to keep costs down. Fiscal survival requires balancing these factors well. Modelling a system and using a computer to simulate its operation has been used to study medical situations (2-12). With simulation, problems are identified, interventions tried, and limitations discovered in a controlled, non-invasive way. Models of patient flow have been designed for the simple clinic system involving only patients and physicians (2-7). This study applied computer simulation techniques to a contemporary, multi-provider, appointment-based clinic and resulted in improved patient flow through the facility. METHODS The Clinic The Internal Medicine University Faculty Clinic (UFC) at the University of New Mexico Hospitals is an appointment-based clinic where faculty general internists see about 13,000 private patients per year. After arriving in the clinic, a patient first talks with a registrar, then is triaged and directed into an examination room by a nurse, is examined by the physician and then visits with a discharger. The UFC is usually staffed with two registrars, four nurses, four physicians and two dischargers. During a half-day session, each physician uses two exam rooms and sees eight patients who are scheduled at twenty-minute intervals. Although some providers are cross-trained for several jobs, during any given session they perform only one. The patient no-show rate averages ten percent. Time-Motion Study Data concerning provider times, idle times, patient waiting times and total time spent in the clinic were collected over two separate one-week periods. Providers recorded start and finish times of their service on data sheets carried by each patient. Computer Model and Simulation The computer program was written in Turbo Pascal 6.0 (Borland International, Scotts Valley, CA) and run on personal computers. Provider task times distributions were input in histogram format into the program. For the examples in this paper, the sample size was two-hundred and fifty sessions, which represents clinic activity in one year. Studies Using the Simulation Program Numbers of specific provider-patient encounters were collected during each hour to gain a sense of how busy providers were at various times during the session. Also for each hour, the average patient total time in clinic was computed for patients having appointment times during that hour. Sensitivity analyses were used to learn how clinic staffing, patient load, appointment interval, no-show rate and provider speed and variability influences total patient time in the clinic, session length and provider idle times. RESULTS The first time-motion study obtained data on 452 patient visits. How patients spent time in the UFC is shown in FIGURE 1. Figure 1. The model and simulation program fitted clinic observations and worked well. Using the model, the total patient time in clinic is 78.9 minutes compared with the measured 75.4 minutes (s.d. 34.2) (or 78.4 minutes when the component means were summed). Predicted and measured wait times also correlated closely. The computer program showed that the number of patient- provider encounters is high during hour one, peaks during hour two and thereafter rapidly declines. The inefficiency in closing the clinic for lunch and restarting it again for the afternoon is evident. Patients arriving in the clinic during the first, second and third hour take an average of 69.7, 93.1 and 104.1 minutes for their total clinic time. Increasing the number of persons in any provider group increases the total idle time of that group and variably affects idle times of other providers, the patient total time in clinic and session length (TABLE 1). When the number of physicians working during a session increased with or without an increase in the total number of scheduled patients, individual patients, on the average, spent more time in the clinic. If six physicians are working instead of four, eighteen patients are scheduled per hour instead of twelve. Since the number of other servers is not proportionately increased, patient wait times to see those servers will increase resulting in an increased patient total time in clinic from 78.9 minutes to 87.5 minutes. Changing the number of dischargers showed a critical threshhold point. With two dischargers working, the patient total time in clinic is 78.9 minutes; with only one, the patient total time rises to 157.2 minutes; with three, the patient total time is 69.3 minutes. Using knowledge gained from the time-motion study and simulation program, clinic managers made some operational changes: having two dischargers present at all times; re-training providers to do their job more efficiently; and limiting to four (or occasionally five) the number of physicians present in the clinic at any one time. Six months after the clinic changes were implemented, another time-motion study on 562 patient visits was performed. The mean patient total time in clinic was 57.1 (s.d. 30.2) min which was significantly decreased (p < 0.001, t-test) from before. The following mean values were obtained: registrar task time 3.3 min, triager time 5.8 min, physician time 22.7 min and discharger time 7.0 min. Given the new provider task time distributions, the computer simulation program projected the patient total time in clinic to be 53.0 min. DISCUSSION Simulating patient flow with a computer was helpful in studying our clinic system and operating it better. An understanding was gained of system sensitivity to staffing levels, appointment intervals, provider task times and no-show rates so that intelligent changes could be implemented that improved patient flow. Although our Medicine Clinic staffed by resident physicians has the same provider structure, the physician and discharger task times will have larger means and variances. Resident physicians are less experienced than the faculty and generally order more tests. The data can be input in histogram format into the program in order to study dynamics of patient flow within the clinic. Changes can be introduced to optimize patient flow relative to patient wait and staffing levels. Previous simulation studies (2-7) didn't incorporate sequential providers in an appointment-based medicine clinic. Sophisticated, management science analyses (4-7) focused on a one provider system and used no measured data. Classic analytical distributions could not approximate well our observed provider task time distributions. Although such analyses address the important issues, which are included in our study, the investigators admit that analytical queuing theory is either intractable or inadequate or both (4,6). This study's results do not give the final solution. Most clinics--including our's--are changing over time with respect to clientele, management priorities, providers and their job responsibilities. Experimenting with operational changes in the clinic can be disruptive, time-consuming, costly and difficult to monitor. However, they can be examined using simulation programs to see "What if?". Such computer applications can help health care facilities now to function better and in the future to adapt to changes. REFERENCES 1. Welch JD, Bailey NTJ. Appointment systems in hospital outpatient departments. Lancet (May 31) 1952; 1105. 2. Goitein M. Waiting patiently. N Engl J Med 1990; 323: 604. 3. Fetter RB, Thompson JD. Patients' waiting time and doctors' idle time in the outpatient setting. Health Services Res 1966: 1: 60. 4. Ho C, Lau H. Minimizing total cost in scheduling outpatient appointments. Management Sci 1992; 38: 1750. 5. Fries B, Marathe V. Minimizing total cost in scheduling outpatient appointments. Oper Res 1981; 29: 324. 6. Katz J. Simulation of outpatient systems. Commun ACM 1969; 12: 215. 7. O'Keefe R. Investigating outpatient departments: implementable policies and quantitative approaches. J Oper Res Soc 1985; 36: 705. 8. Levy JL, Watford BA, Owen VT. Simulation analysis of an outpatient services facility. J Soc Health Syst 1989; 1: 35. 9. Saunders CE, Makens PK, Leblanc LJ. Modeling emergency department operations using advanced computer simulation systems. Ann Emerg Med 1989; 18: 134. 10. Vemuri S. Simulated analysis of patient waiting time in an outpatient pharmacy. Am J Hosp Pharm 1984; 41: 1127. 11. Rising EJ, Baron R, Averill B. A systems analysis of a university-health-service outpatient clinic. Oper Res 1972; 21: 1030. 12. Reid RA. Simulation and evaluation of an experimental rural medical care delivery system. Socio-Econ Plan Sci 1975; 9: 111. Legends for Figures in UFC Simulation Paper Figure 1 Provider and wait times in the clinic. The arrow indicates entry into the clinic; direction is clockwise around the pie. TABLE 1 ----------------------------------------------------------------------------------------- IF there is an THEN increase in: the idle time of the following: patient session total time length --------------------------------------------------- in clinic registrar nurse doctor discharger ----------------------------------------------------------------------------------------- #registrars inc dec dec dec dec dec #nurses 0 0 0 0 0 0 #doctors dec dec dec dec inc dec #dischargers 0 0 0 inc dec dec #pts seen/doc inc inc inc inc inc inc appointment inc inc inc inc dec inc interval no-show rate inc inc inc inc dec dec doc task time 0 0 dec inc inc inc doc task time 0 0 0 inc inc inc variance ----------------------------------------------------------------------------------------- Table footnote: See text for explanation. inc = increase; 0 = no change; dec = decrease