Improving patient flow at an outpatient clinic: study of sources of variability and improvement factors

Abstract

This paper focuses on analyzing and improving patient flow at an outpatient clinic of the Indiana University Medical Group. A structured process analysis and improvement approach was used to identify sources of variability and improvement factors. A process map, that matched the flow process at the clinic, was developed and validated. Key sources of variability that had potential to contribute to congestion in flow were identified. Data on task times were collected by observing the process with stopwatch or from historical records. A simulation model corresponding to the process map was developed, and the output was validated. Several ideas to modify clinic operations were tested on the validated simulation model. The overall result was an improvement in both the mean and the standard deviation of patient wait time, as well as higher utilization of physicians’ time. The clinic has implemented several of our recommendations and experienced significant improvements.

Appendices

Appendix 1: Input probabilities and distributions used in simulation

Table 4  

Table 5  

Table 6  

Table 7  

Table 8  

Appendix 2: Explanation how clinic can process increased load with small effect on PWT

The process considered is a two-stage process (registration and see the doctor).

Let us focus on the following numbers for old patients in Table 2:

Table 9  

The patient wait time (or PWT) includes the wait time in the queue for registration, the registration time, and the wait time in doctor’s queue. Our examination of simulation output for No Improvements showed that the wait time in doctor’s queue is about twice the wait time in the registration queue. For All Improvements, most of the wait time is in doctor’s queue. The average registration time is 3 min. The table below gives the breakdown of wait times:

Table 10  

A useful tool to analyze the waiting time in queue is the “waiting time equation” for single-server system with random arrivals and random service times shown below.

An approximation for T _q (the average waiting time in queue) when the system is in steady state is given by:

$$T_q = p \times \left( {\frac{u}{{1 - u}}} \right) \times \left( {\frac{{C_a^2 + C_p^2 }}{2}} \right)$$

where p = average service time; c _a  = coefficient of variation of inter-arrival times; c _p  = coefficient of variation of service times; and u = utilization of server. We will call the term $\left( {\frac{u}{{1 - u}}} \right)$ the utilization factor and the term $\left( {\frac{{C_a^2 + C_p^2 }}{2}} \right)$ the variability factor. The utilization factors (utilization values) for different combinations are:

Table 11  

Comparing No Improvements with All Improvements for Traditional, the waiting time in registration queue has gone down from 2.4 to 0.17. There are three factors that explain this saving:

• 1. Improvement in the utilization factor: It has gone down form 0.449 to 0.365. Recall that this improvement is because PSAs do not have to handle phone calls.

• 2. Improvement in the variability factor: There is no change in $C_a^2$ because the improvement factors do not affect the arrival process for patients. There will be some saving in $C_p^2$ because PSAs do not have to handle phone calls that were source of variability in process times.

• 3. Effect of Improvement Factors: These factors mitigate the effect of variability as discussed in the paper.

Note that while the expression for T _q is helpful in explaining the direction of saving, it will not be accurate to use it to calculate the amount of change. One of the difficulties with this expression is that it assumes that the system has reached a steady state, an assumption that does not hold for our situation. Also, the expression will need to be modified to account for the effect of improvement factors.

Let us now consider the waiting time in doctors’ queues; it has gone down from 4.8 to 0.93. One factor that contributes to saving in waiting time in doctor’s queue is the saving in $C_a^2$ for the patients joining doctor’s queue. Patients departing from registration queue join doctors’ queues. Improvement in the registration process not only reduces the waiting time in the registration queue, it also streamlines the flow of patients to doctors’ queues.

The other numbers in the waiting time table can be explained similarly. For example, the waiting time in doctor’s queue has gone down from 4.8 to 2.3 even though the utilization factor has gone up from 0.585 to 0.833. As discussed above, one factor that explains this saving is the streamlined flows to doctors’ queues.