Sensitivity of the Pediatric Early Warning Score to Identify Patient Deterioration

Rationale

Code blue and RRT events were included because they represent unexpected and/or critical deterioration characteristics in some hospitalized pediatric patients. In addition, assessment and communication processes are similar. At Children's, employees, patients, parents, and families may unconditionally request urgent medical assistance (RRT) for a patient who is perceived to be in distress without advance consultation with housestaff or the patient's attending physicians. We tested the sensitivity of PEWS in this population as an objective tool to predict critical deterioration and to support structured communication of patient status between team members. RRT and code blue events provide measurable outcomes that we hope to reduce with implementation of PEWS.

This was a retrospective study design of cases before implementation of PEWS. It is the only design that is not contaminated by interventions in response to PEWS. Once PEWS is implemented, it becomes unclear as to whether the event would have occurred as a result of interventions that are responses to critical PEWS.

Assignment of PEWS

PEWS are assigned in 3 domains: behavior, respiratory, and cardiovascular. Scores in each domain can range from 0 to 3 points (Fig 1). In addition, 2 points are added for nebulizations that are continuous or every 15 minutes and 2 points for persistent postoperative vomiting. The total score can range from 0 to 13.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Children's guide to PEWS.

PEWS were calculated for 24 hours before the event at intervals of at least 4 hours for each of the identified patients. This is consistent with assessment frequencies for nurses on these units. When assessment charting was more frequent than 4 hours, additional PEWS were calculated. Behavioral scoring required the actual use of the words shown in Fig 1, such as playing, sleeping, and/or irritable. During waking shifts, this would not be customary documentation. When these words were not documented, the item was scored as missing. When data were not available for a specific domain, it was marked as missing. To ensure consistency, a single “expert” performed all scoring.

Training the Expert in PEWS

A licensed registered nurse with extensive intensive care, rapid response, and code blue experience reviewed patient charts and determined PEWS. Working with a hospital advanced practice nurse, they reviewed charts, discussed and refined the scoring rules, and then independently scored patients until interrater reliability was achieved on a sample of 10 patient records. Our electronic medical record included many more descriptors of patient status than are included in the tool, which required agreement on which terms would be recommended to staff to facilitate scoring PEWS. One example is the option for patient color. PEWS has 3 choices, and the electronic medical record has 6. By comparing their results data, interpretation was refined until agreement was achieved. This agreement of terms for scoring was also used to train staff.

PEWS Action Algorithm: Critical PEWS

We agreed to use the Monaghan¹² criteria that a total score of ≥4 or a score of 3 in any of the PEWS domains reflected a critical value that required a consultative action. We refer to these criteria as critical PEWS.

Additional Data Collected

During the 24-hour pre-event period, we collected data on increased frequency of nursing assessment and the addition of equipment, including pulse oximetry and cardiac monitoring. Consultations by another nurse, physician, or respiratory therapist were also collected. The purpose was to evaluate these as indicators of staff awareness of deteriorating patient status. An exhaustive review of multiple documentation elements such as nursing flow sheets, nursing narrative notes, physician progress notes, and physician orders was used to retrieve data elements. Demographic data included age, gender, race, diagnosis, and All Patient Refined Diagnosis Related Groups (APRDRG)¹⁶ service line and whether the patient had had surgery. APRDRG is a recognized patient classification scheme available from 3M that relates patient cost to his or her diagnosis and comorbidities. APRDRG service line classification is 1 component of this method.

Statistical Analysis

Sensitivity was defined as percentage of patients who had a critical PEWS before the RRT or code event. Median and range were used to describe continuous variables such as the time from a critical PEWS, first consultation, and first monitor added, to the time of the RRT or code blue event.

For a subgroup of patients who had (1) a critical PEWS, (2) a consultation, and (3) addition of a monitor, median time from each of these to the RRT or code event were compared by using the Friedman nonparametric test with a 2-sided P < .05 required for significance. The nonparametric Wilcoxon Signed Ranks test was used for paired comparisons of these median times where a Bonferroni correction was applied requiring a 2-sided P < .015 for significance. Because of the small number of code blue events it was not possible to compare outcomes for these events to RRT events. SPSS 15.0 (SPSS Inc, Chicago, IL) was used for statistical analysis.