Improving emergency department performance by revising the patient–physician assignment process

Abstract

Emergency departments (EDs) are continuously exploring opportunities to improve their efficiency. A new opportunity lies in revising the patient–physician assignment process by limiting the number of patients simultaneously assigned to a single physician, which is defined as the application of a case manager approach with limited caseloads. The potential of introducing a case manager approach with limited caseloads as a way to improve physician productivity, and consequently ED performance, is investigated by use of a discrete-event simulation model based on a real-life case study. In addition, as the case manager system is characterised by three parameters that can be customised and optimised (i.e. caseload limit, pre-assignment queueing discipline and internal queueing discipline), the impact of these parameters on the effectiveness to improve ED performance in terms of length-of-stay and door-to-doctor time is evaluated. To the best of our knowledge, this paper is the first to examine the potential of a case manager system with limited caseloads in a complex service system like a real-life ED, and to investigate the impact of the three system parameters on the results. The outcomes of the study show that performance can be improved significantly by introducing a case manager system, and that the system parameters have an impact on the effect size.

Appendices

Appendix 1: Electronic health record data

See Table 5, 6, 7 and 8.

Table 5 Timestamp attributes in extracted input file of EHRs (T = time)

Table 6 Numerical attributes in extracted input file of EHRs

Table 7 Categorical attributes in extracted input file of EHRs

Table 8 Free text attributes in extracted input file of EHRs

Appendix 2: Validation

See Figs. 12 and 13.

Fig. 12

Validation boxplot for the LOS of patients in the non-ambulant and ambulant zone

Fig. 13

Graphical validation of hourly number of patient arrivals

Appendix 3: Statistical analysis

This online appendix provides the results of Mauchly’s test of sphericity and the repeated-measures full factorial ANOVA. For all main effects and 2-way interactions in the ANOVA, the most appropriate F-statistic is determined by the results of Mauchly’s test of sphericity Tables 9 and 18. In case the results of Mauchly’s test provide evidence for the violation of the sphericity assumption at the 5% significance level (p value < 0.05), the G–G estimate of the F-statistic is used in the ANOVA. Otherwise, the sphericity assumed estimate of the F-statistic is used.

3.1 Scenario without multitasking effect

See Tables 9, 10, 11, 12, 13, 14, 15, 16 and 17.

Table 9 Mauchly’s test results on sphericity—no multitasking effect

Table 10 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on DTDT TC2—no multitasking effect

Table 11 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on DTDT TC3—no multitasking effect

Table 12 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on DTDT TC4—no multitasking effect

Table 13 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on DTDT TC5—no multitasking effect

Table 14 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on LOS TC2—no multitasking effect

Table 15 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on LOS TC3—no multitasking effect

Table 16 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on LOS TC4—no multitasking effect

Table 17 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on LOS TC5—no multitasking effect

3.2 Scenario with multitasking effect

See Tables 18, 19, 20, 21, 22, 23, 24, 25 and 26.

Table 18 Mauchly’s test results on sphericity—multitasking effect

Table 19 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on DTDT TC2—multitasking effect

Table 20 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on DTDT TC3—multitasking effect

Table 21 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on DTDT TC4—multitasking effect

Table 22 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on DTDT TC5—multitasking effect

Table 23 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on LOS TC2—multitasking effect

Table 24 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on LOS TC3—multitasking effect

Table 25 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on LOS TC4—multitasking effect

Table 26 $2\times 2\times 3\times 25$ full factorial repeated-measures ANOVA on LOS TC5—multitasking effect

Appendix 4: Results scenario without multitasking effect

See Figs. 14, 15, 16, 17 and Tables 27 and 28.

Fig. 14

Mean DTDT per TC as a function of caseload. Note DTDT at caseload 1 is very high because of the large amount of physician idle time, and will never be used in practice. These values are not presented in the figures for clarity purposes. The DTDT at caseload 1 equals (in minutes): TC2: 1855.96, TC3: 1867.53, TC4: 2893.89, TC5: 8738.38

Fig. 15

Mean LOS per TC as a function of caseload. Note LOS at caseload 1 is very high because of the large amount of physician idle time, and will never be used in practice. These values are not presented in the figures for clarity purposes. The LOS at caseload 1 equals (in minutes): TC2: 2138.78, TC3: 2133.13, TC4: 3021.15, TC5: 8925.27

Fig. 16

Caseload range resulting in significant DTDT improvement in comparison with no caseload limit for each priority factor combination (per triage code)—no multitasking effect

Fig. 17

Caseload range resulting in significant LOS improvement in comparison with no caseload limit for each priority factor combination (per triage code)—no multitasking effect

Table 27 Significant potential KPI improvements under the current queueing disciplines (TC–TC-Equal) with corresponding caseload limit—no Multitasking effect

Table 28 Significant KPI improvements under the current queueing disciplines (TC–TC-Equal) for caseload limit 2—no multitasking effect