Estimating Task Execution Time in EHRs Using the Keystroke

advertisement

Estimating Task Execution Time in EHRs Using the Keystroke-Level Model

Louis Lee, Yuanyuan Li, MD,MS, Krisanne Graves, MSN, RN, Amy Franklin PhD, Muhammad F. Walji, PhD, Jiajie Zhang, PhD

The University of Texas Health Science Center at Houston – School of Biomedical Informatics

Introduction

• A major barrier to EHR adoption is that clinicians find these systems inefficient and difficult to use[1]

• There is a lack of objective and reliable measures of user performance for common clinical tasks in EHRs

• Keystroke Level Models (KLM) can be used to analyze work flows and identify factors for work flow optimization, and have been demonstrated to accurately predict skilled user performance time [2, 3]

Research Goals

• Measure skilled user performance time for tasks within NIST meaningful use cases (MUCs) [4]

• Establish benchmarks for MUC time

• Compare and analyze work flows for MUCs across EHRs

Method

• CogTool [5], a KLM tool, was used to analyze 6 MUCs (e.g., ePrescribing, smoking status) across 3 EHRs

• Task execution times were predicted by CogTool, and resulting times were compared across EHRs

• Time for each physical operator (see Figure 2) was analyzed for variability

• Recommendations for improving task efficiency were generated

For additional information, please contact

SHARPC@uth.tmc.edu

Results

EHR3

EHR2

EHR1 ePrescribing

Smoking Status

Problem List

BMI

Demographics

Vital signs

0 20 40 60 80 100 120 140

Time (s)

Figure 1

. Predicted task execution time for 6

MU objectives across 3 EHRs

Type

Move Mouse

Move and Hover

Left Triple-Click

Left Double-Click

Left Click

EHR3

EHR2

EHR1

Home Keyboard

Home Mouse

0 5 10 15 20 25 30 35 40

Time (s)

Figure 2.

Time for physical operators in

Recording Problem List across 3 EHRs

Portion of work flow for recording problem list for EHR1

Patient

Dashboard: click “medical records”

“Problem List”

Tab: Click “Add

Diagnosis”

Medical Records Webpage:

Maximize the screen, click

“Problem List”

“Problem List”

Tab: Click “Add

Diagnosis”

“Patient Diagnosis” pop-out window:

Left Click entry field; Type “tia”

Left Click “search”

Fig.5. Problem List Recording Structure for EHR1

Left Click the correct diagnosis from pop-up search result box

Left Click the “Acute” status from the pull down list

Left Click “Save”

Continues

Summary of Conclusions

 Although KLM results were different across MUCs, there was surprising consistency in task time across the 3 EHR products

• Problem List had the longest execution time and greatest variability across EHRs

Next Steps

• Compare and analyze clinical task work flows in different EHRs

• Optimize the work flow according KLM to reduce task execution time

• Collecting data across larger samples will allow creation of benchmarks for NIST MUCs

References

1. Smelcer JB, Miller-Jacobs H, Kantrovich L. Usability of Electronic

Medical Records. Journal of Usability Studies.2009;4:70-84.

2. John BE, Suzuki S. Toward cognitive modeling for predicting usability.

To appear in HCI international 2009.

3. John BE. Using predictive human performance models to inspire and support UI design recommendations. Proceeding of ACM CHI’11 session on predicting & modeling human behaviors 2011.

4.

http://www.nist.gov/index.html

(accessed Oct 14th, 2011).

5.

http://cogtool.hcii.cs.cmu.edu/

(accessed Feb 15th, 2011).

Acknowledgement

This project was supported by Grant No. 10510592 for Patient-

Centered Cognitive Support under the Strategic Health IT

Advanced Research Projects (SHARP) from the Office of the

National Coordinator for Health Information Technology.

Download