H567/03 Section B Option 3 - Drews and Doig (2014) and workplace design

The cognitive problem in intensive care

Drews and Doig investigated whether a configural vital signs display could help intensive-care nurses recognise a patient's state faster and more accurately. Nurses must integrate changing physiological information while dealing with interruptions and multiple patients. Separate numerical readings can require repeated searching, remembering earlier values and mentally combining variables. The study applies the cognitive area to an environmental problem: equipment design can change the processing demands of work.

A configural display arranges information so that visual relationships among variables help reveal the overall state. Its value comes from meaningful integration, not from adding decoration. The researchers developed their display using interviews with nurses, relevant literature and data-visualisation experts. Three experienced ICU nurses reviewed prototypes repeatedly.

The final CVS display retained numerical information and added readily visible trends and variability. Trend information covered heart rate, oxygen saturation and blood pressure measures. A current-state representation and a continuously updated rectangle representing the previous hour's high and low values helped convey state and variability. The central numerical panel was identical to the control display. The full original layout should not be reconstructed from this verbal account: the assessable design principle is integration of current numbers, change over time and variability.

An alarm based on one reading crossing a threshold may miss the importance of a pattern across readings. The study therefore tested information presentation, rather than merely making an alarm louder. No clinical diagnosis or treatment instructions are required for this psychology lesson.

Design and participants

The evaluation was a laboratory experiment in the Applied and Basic Cognition Laboratory at the University of Utah. The display format was manipulated: traditional numerical display versus CVS display. Response time and correct identification of the patient's state were dependent measures. The laboratory compared performance in simulations; it did not measure whether real patients recovered better.

There were 42 registered nurses with critical-care training and at least one year's ICU experience: 21 used each display. Display format was a between-participants factor. Each nurse encountered all four scenarios once, making scenario a within-participants factor. This is a mixed design, not a repeated-measures comparison in which every nurse used both display types.

The sample's mean age was 44.59 years, with a range of 25–64; 69% were female and 31% male. Mean ICU experience was 8.48 years. Experienced nurses are relevant users of this equipment, giving the evaluation more practical relevance than a student sample. However, 42 trained participants cannot represent every hospital, novice worker or profession.

Participants were randomly assigned to a display condition until a condition was full. Scenario orders were randomised and matched across conditions: a nurse in one group received the same order as a nurse in the other group. This reduces a systematic difference in scenario order. It does not turn display format into a within-participants factor or remove every individual difference.

Procedure and measurement

A physician and an ICU nurse who had not designed the display validated the scenarios. Three involved deterioration: early sepsis, septic shock and pulmonary embolism. The fourth represented a stable patient. Relevant information for a correct assessment was available in both display conditions, although access and integration demands differed.

All nurses received standardised 20-minute training explaining their display and the task. The session lasted about 90 minutes. Before a scenario, nurses received a paper record giving the patient's clinical context, including medical history and medication information. Vital signs appeared on a 15-inch computer screen. Nurses were asked to evaluate the physiological state verbally, interpret the information and recommend appropriate interventions as quickly and accurately as possible.

Each scenario allowed 300 seconds. Response time ran from scenario onset to the nurse's verbal assessment. When no assessment occurred within the limit, the response was coded as 300 seconds. Accuracy concerned correct identification of the patient state. Researchers also recorded whether numerical-display users opened the separate trend screen.

After the four scenarios, nurses completed the NASA-TLX workload assessment and rated scenario realism and display desirability on seven-point scales. These self-reports provide evidence about perceived demand and acceptability; they do not replace the performance measures. Institutional review-board approval was obtained. Simulated patients allowed a new interface to be studied without exposing real patients to an untested display.

Standardised training, timings and materials make the task reproducible. Capping response times at 300 seconds also creates a limitation: a nurse who would take much longer is recorded at the same maximum as someone only slightly beyond five minutes. This compresses differences among the slowest responses.

Findings and their limits

The results supported CVS for some aspects of interpretation, with important exceptions.