Empirical Analysis of the Effect Residents Have on Treatment Times in an Emergency

Department

David Anderson, John Silberholz, Bruce Golden, Mike Harrington, Jon

Mark HirshonPOMS Annual Meeting, Denver 2013

2

Overview

Broad HealthcareLandscape

-Health Care Reform Bill, 2010-Americans spent $2.3 trillion on health care in 2007-Hospitals are one of the least efficient sectors

University of MarylandMedical Center (UMMC)

UMMC UMMC ED

700 beds1,182 doctors742 residents

55 beds20% admission rate

46,000 patients/year

3

Residency Model

Medical School• Four years• Classes, clinical rotations

Residency• First year: Internship, general medicine• Next 3-7 years: Specialty• Designed for teaching

Attending Physician• Private practice or hospital

4

Research Question

What effects do residents have on the efficiency of the emergency department?

- Longer or shorter patient treatment times?

Residents are in the hospital to learn, but also treat patients

One conjecture is that the teaching of residents takes time away from patient care and negatively impacts efficiency

5

Inconclusive Literature• Medicare reimbursement rates consider the direct and indirect costs of

training residents (Rosko, 1996)• It has been argued that Medicare reimbursement rates overcompensate

for the costs of training residents (Anderson & Lave, 1986; Custer & Wilke, 1991; Rogowski & Newhouse, 1992; Welch, 1987).

• Presence of residents increases faculty staffing requirements - attending physicians are required to spend time supervising and instructing the residents (DeBehnke, 2001)

• Teaching and treatment can occur simultaneously, meaning that residents can help to improve throughput (Knickman et al. 1992)

6

Observational Studies

• Residents Slow Down Treament:– Harvey et al. (2008) – New Zealand resident strike– Salazar et al. (2001) – Resident strike in the US– Lammers et al. (2003) – pre-post addition of residents

• Residents Help:– Theokary et al. (2011) – Residents increase treatment quality– Blake and Carter (1996) – When residents treat patients they help– Eappen et al. (2004) – Addition of anesthesiology residents – Offner et al. (2003) – Addition of trauma residents– Dowd et al. (2005) – As residents gain experience they increase

efficiency

7

Resident Seminars

• Residents absent every Wednesday morning for a seminar

• No replacement workers hired• Wednesday mornings provide a representative

sample of all emergency department activity– Wide range of arrival rates – All types of patients and severities– Congestion levels vary as well

8

Advantages Over Previous Studies

• Short-term absence of residents– No operational changes

• No changes in staffing• Quick turnover: easy to measure impact

9

Data

• Data was provided on 7395 patients• Information on severity score, number of lab

and radiology tests needed, arrival time, treatment time, congestion of the waiting room, and whether or not the patient was admitted to the hospital was given for each patient

• Resident presence was determined based on the time the patient was first treated

10

Arrival Rates

11

Regression Analysis• Regressed treatment time on patient and treatment

characteristics: Resident absence increases treatment times by almost 8% (exp .075 = .08)

Variable Coefficient Std. Error t-value p-value(Intercept) 5.002 0.020 247.475 <.001

NoRes 0.075 0.034 2.242 0.025Line 0.010 0.002 5.455 <.001Admit 0.088 0.015 5.819 <.001NumLab 0.032 0.001 35.847 <.001Labs 0.335 0.018 18.716 <.001NumRad 0.057 0.004 13.509 <.001Rad 0.148 0.016 9.376 <.001Weekend -0.044 0.013 -3.311 <.001Sev1 -0.148 0.096 -1.544 0.123sev2 0.048 0.017 2.730 0.006sev3 0.031 0.015 2.080 0.038sev4 -0.178 0.032 -5.511 <.001sev5 -0.543 0.090 -6.001 <.001

(Adjusted R2 = .5355, N = 7935)

12

High Severity vs Low Severity

• Residents might play different roles when treating different types of patients

• Ran regressions on high and low severity patients separately

• Residents have a strong effect on lowering treatment times when treating high severity patients, but no noticeable effect when treating low severity patients

13

High Severity ResultsVariable Coefficient Std. Error t-value p-value

(Intercept)

5.027 0.020 245.581 <.001

NoRes0.073 0.034 2.138 0.033

Line0.009 0.002 4.784 <.001

Admit0.090 0.015 5.955 <.001

Numlab0.032 0.001 35.832 <.001

Labs0.316 0.018 17.242 <.001

Numrad0.056 0.004 13.331 <.001

Rad0.143 0.016 8.881 <.001

Weekend-0.055 0.014 -4.010 <.001

sev1-0.146 0.095 -1.528 0.126

sev20.049 0.017 2.828 0.005

sev30.029 0.015 1.987 0.047

(Adjusted R2 = .5133, N = 7549)

14

Low Severity ResultsVariable Coefficient Std. Error t-value p-value

(Intercept) 4.234 0.104 40.558 <.001

NoRes 0.110 0.189 0.581 0.562

Line 0.041 0.011 3.711 <.001

Admit 0.010 0.127 0.081 0.935

Numlab 0.035 0.007 4.899 <.001

Labs 0.553 0.087 6.324 <.001

Numrad 0.133 0.037 3.610 <.001

Rad 0.144 0.093 1.559 0.120

Weekend 0.135 0.062 2.183 0.030

sev4 0.281 0.099 2.834 0.005

(Adjusted R2 = .5737, N = 341)

15

Morning Patients

• One possible source of endogeneity is that residents are only absent for patients treated in the morning

• We restrict the data to only those patients who began treatment between 7 am and 1 pm (the time of the seminars on Wednesday)

• Our results still hold

16

Morning ResultsVariable Coefficient Std. Error t-value p-value

(Intercept) 4.630 0.055 84.908 <.001

NoRes 0.068 0.034 2.008 0.045

Line 0.023 0.006 3.792 <.001

Admit 0.146 0.031 4.669 <.001

Numlab 0.030 0.002 15.628 <.001

Labs 0.328 0.038 8.750 <.001

Numrad 0.054 0.009 5.901 <.001

Rad 0.188 0.033 5.763 <.001

High 0.345 0.054 6.359 <.001

(Adjusted R2 = .5712, N = 1768)

17

Survival Analysis

• Instead of measuring treatment times, we can measure discharge rate

• We compare discharge rate on Wednesday mornings to the discharge rate on other weekdays

• A higher discharge rate would imply that residents help to increase throughput

18

Survival Analysis ResultsVariable Coefficient Standard

Errorz Pr(>|z|)

Numlab 0.0037 0.0055 0.6680 0.5044Numrad -0.0358 0.0254 -1.4090 0.1587

NoRes -0.2505 0.0860 -2.9140 0.0036Sev1 0.6403 0.4540 1.4100 0.1585Sev2 -0.0447 0.1023 -0.4370 0.6622Sev3 -0.1140 0.0836 -1.3640 0.1725Sev4 -0.0320 0.1932 -0.1660 0.8685Sev5 0.6317 0.5864 1.0770 0.2814Line 0.0327 0.0104 3.1310 0.0017Labs -0.6133 0.1067 -5.7490 0.0000Rad -0.2198 0.0905 -2.4290 0.0152

19

Different Patient Populations

0 10 20 30 40 50 60 70 80 90 1000

20

40

60

80

100

120

140

160

180

200

First Bed Time

More Severe Population

Linear (More Severe Popula-tion)Less Severe Popu-lation

Linear (Less Se-vere Population)

Percent of Patients Treated by a Resident

Aver

age

Min

utes

unti

l Firs

t Bed

Residents have a bigger impact on more severe populations

20

Conclusion

• Contrary to our original intuition, we have shown that residents speed up the treatment of patients in the emergency department

• This effect is especially strong when treating high severity patients

• We recommend that when possible, residents treat high severity patients

21

Future Work

• Quantify effect of each additional healthcare worker and compare nurses and nurse practitioners to residents

• Model doctor decisions explicitly, show how they move through ED

• Identify bottlenecks in the system• Include data gathered in person to help model doctor

movement