Acute care hospitals are highly complex environments in which many factors contribute to the achievement of quality patient outcomes. Care delivered within such environments is an interdependent and interactive process that is influenced by nurses, physicians, support staff and the infrastructure and technology designed to support patient care. Providing clinically effective care has been defined as doing the right thing in the right way for the right patient at the right time.1 The ability to achieve this level of patient care requires advanced technical and professional expertise as well as scientific knowledge.

Organisations such as nuclear power plants and aviation are considered highly reliable. In part, they gained this designation due to a continuous review process that attempts to identify the root cause of adverse or near-miss events and implement changes to avoid these events. Although the majority of patient care is delivered with a successful outcome, as many as 3.7% of hospitalised patients experience an adverse event.2

Silber and colleagues introduced failure-to-rescue as an outcome measurement defined as a death that occurs after a patient develops a complication in the hospital that was not present on admission.3^–7 They found failure-to-rescue to be influenced by characteristics other than patient acuity, such as care provided by a board-certified anaesthesiologist and registered nurse staffing. The assumption was that development of a complication is most strongly influenced by patient characteristics, but once a complication develops, provider and organisational characteristics are more strongly linked to patient survival. Using the same surgical patient population as Silber and colleagues,67 Aiken and colleagues8 reported that, after adjusting for patient and hospital characteristics, an assignment of one additional patient per nurse was associated with a 7% increase in the likelihood of failure-to-rescue.

The presence of unrecognised deterioration before cardiac arrest has been well reported in the literature, with as many as 66–70% of patients who experience a cardiac arrest showing evidence of deterioration 6–8 h preceding the event.9 Medical emergency teams (METs) were developed in response to evidence that such delays might result in a preventable death. Subsequent reports of a reduction in cardiac arrest and unexpected intensive-care unit (ICU) admissions led many hospitals to introduce a MET system.10^–12 In their 100 000 Lives Campaign, the Institute of Healthcare Improvement (IHI) specifically cited such METs, referred to as rapid response teams, as an intervention with the potential to reduce preventable hospital deaths.

While effective, the MET system does not address the events which precede the need to call for a response. We were unable to identify any studies that attempted to identify unique reoccurring characteristics in medical and surgical patients who preceded the need to initiate a MET response. Such information could provide direction for interventions designed to avert the need to initiate a MET response or identify the need to do so more quickly. Accordingly, we analysed patient, nurse, and organisational characteristics associated with a MET response involving patients admitted to a general medical or surgical ward in a tertiary care institution affiliated with an academic medical centre. The specific aims were: (1) to compare differences in nurse, patient and organisational characteristics in medical and surgical patients requiring a MET activation; and (2) to identify nurse, patient, and organisational variables that predict a >30 min delay in activating the MET system once the patient meets criteria for such a response.

METHODS

Site and sample

The setting for this study was the University of Pittsburgh Medical Center (UPMC) Health System, Presbyterian campus. UPMC Presbyterian, a tertiary care facility, admits approximately 32 000 patients per year. This study examined a convenience sample of 108 MET activations that occurred between January 2004 and January 2006 involving 51 medical and 57 surgical patients. Approval to conduct this research was obtained from the institution’s Total Quality and Patient Safety Committee.

The MET system was introduced as a component of care at this institution in 1989. Responders to a MET activation include an Intensive Care Physician, two Intensive Care Nurses and a Respiratory Therapist. The team can be summoned by anyone in the hospital (registered nurse, nursing assistant, transport aide, information desk clerk) at any time they have concern for patient care by calling a designated number. All clinical units have access to a list of specific criteria for which calling the team is considered appropriate. The MET activation creates an electronic page and overhead speaker announcement placed by the hospital operator. The operator records the location which is logged in a database using a pre-established protocol.

The criteria for selecting the MET activation required that it occur on one of five medical or five surgical units within the hospital. Cardiology, step-down and intensive care unit events were excluded, as were events which did not occur on an inpatient unit. To collect study data, the log of events was reviewed, and events meeting the selection criteria were forwarded to an independent nurse reviewer. All MET activations on the day prior to each scheduled work day were analysed. There was no pattern to scheduling of work days for the nurse reviewer.

Study variables

The study examined selected nurse, patient and organisational characteristics associated with a MET call. Nurse characteristics were defined as: (1) education (BSN, non-BSN); (2) experience (total years in nursing, total years on unit); and (3) whether the nurse cared for the patient day before. Nurses with preparation at the master’s level were excluded due to low numbers, as were licensed practical nurses. Patient characteristics were defined as: (1) age (years), (2) gender, (3) days in hospital prior to the event, (4) status change for MET activation (new onset dyspnoea, chest pain, acute mental status change, unilateral weakness, seizure, change in colour of extremity), (5) pulse oximetry monitoring, (6) vital signs (RR>36 or <8; systolic blood pressure <80 mm Hg; HR>160 bpm or 140 bpm with symptoms or <40 bpm), (7) cardiac monitoring, (8) pain control (patient controlled analgesia), (9) procedure (within 24 h), (10) respiratory status (oxygen, respiratory treatments), and (11) physical restraint (within 24 h). Organisational characteristics were defined as: (1) admission site (Emergency Department, other); (2) shift when the rescue event occurred (07:00 to 19:00 (day) or 19:00 to 07:00 (night)); (3) day of the week (weekday/weekend); (4) time of event related to intensive care transfer; (5) number of patients nurse was assigned; and (6) type of unit (medical, surgical). In addition, events were characterised as delayed or not delayed. A delayed event was defined as an event for which there was documented evidence in the medical record that pre-established clinical criteria for calling the MET were present for >30 min prior to the call.

DATA MANAGEMENT AND ANALYSIS

SPSS (version 12, SPSS, Chicago) was used to perform the data analysis. Group comparisons between events of medical and surgical patients were performed using Pearson’s chi-square tests of independence or t tests as appropriate. The dichotomous response variable delayed MET activation (yes or no) allowed for logistic regression to be used to evaluate the predictive ability of shift and number of patients assigned. Pearson’s chi-square statistic was examined to determine significance. Testing was two-sided with level of statistical significance set at 0.05. Complete data were not available for all variables, as indicated in the tables.

RESULTS

Patient characteristics

There were 108 MET activation events recorded for patients included in the study. Compared with medical patients, surgical patients had more days in the hospital before the event occurred (p = 0.005), were more likely to be monitored using pulse oximetry (p = 0.01) and were more likely to receive patient-controlled analgesia (p = 0.045) (table 1). No significant between-group differences were found for the remaining variables.

Nurse characteristics

There was a significant difference in the mean number of patients assigned when comparisons were made between medical and surgical patients (p = 0.044); however, the actual difference was small (5.6 (SD 1.02) vs 5.2 (1.06), respectively) (table 2). In concert with previous research examining staffing ratios, patient assignments were categorised into ⩽5 patients and >5 patients. No significant differences were found (p = 0.927); however the analysis for an assignment of ⩽5 patients was underpowered. There were no significant differences related to education, experience or continuity of care.

There was little variation in nurse/patient ratio across the units. The modal patient assignment was six patients, with the majority (86%) of nurses assigned to four to six patients (fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Patient/nurse ratio during the shift when medical emergency team activations occurred.

Organisational characteristics

Significant between-group differences were associated with the patient type and unit match (table 3). Twelve of the 51 (24%) events for medical patients occurred while they were being cared for on a non-medical unit, while only two of the 57 (3.5%) occurred off a surgical unit (p = 0.005). There were no between-group differences in regard to recent (⩽24 h) ICU transfer (p = 0.655). As anticipated, a higher percentage of medical patients were admitted through the emergency department than surgical patients (p = 0.018).

Reason for event and outcome

The reason for activating the MET system was categorised based upon the established clinical criteria used in the institution. The highest frequency for the reason for activating the MET system for both groups was pulmonary (shortness of breath or desaturation), accounting for 36% and 43% of the events for medical and surgical patients, respectively. The next most frequent reasons were change in heart rate, blood pressure or chest pain (29% for medicine patients and 35% for surgery patients). Changes in mental status that were not present on admission were found to be low (18% for medicine and 13% for surgery). There were no significant between-group differences in the reason for the event.

The outcome for the MET event was most frequently associated with a transfer to the ICU (68%), followed by transfer to a higher level of care (16%) or no change in location (16%). Only one patient did not survive the rescue event. The majority of medical (82%) and surgical (93%) patients were discharged alive (table 4). There were no significant between-group differences for outcome.

Frequency and pattern of event occurrence

Using the binomial test and assuming an equal distribution of event occurrence, the frequency and pattern of event occurrence were examined for a 24 h interval. The frequency of events was greater during daylight hours (07:00 to 18:59) compared with night hours (p⩽0.009). Sixty-three per cent of the events occurred during day shift, compared with 37% during the night shift. This pattern was true for medical patients (p = 0.049) but not the surgical patients (p = 0.111). Of the 68 events occurring during the daylight shift, 51% occurred during four particular hours (08:00, 13:00, 16:00 and 18:00), as shown in fig 2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Number of responses to medical emergency team activations on the five medical and five surgical units by time of day.

Of the 108 events, 47 (44%) were delayed events. Delayed events were examined in regard to the shift when they occurred and the number of patients assigned to the nurse when the event occurred (table 5). Delayed events were found to have a significant relationship with shift, with more delays occurring on the night shift (p = 0.020). Delayed events were not significantly related to the number of patients assigned. However, there was a trend for more delays when more patients were assigned. Nurses assigned four patients accounted for 21% of the delays, compared with 36% with a five-patient assignment, and 43% with a six-patient assignment (data not shown).

Delayed events were also examined in relationship to outcome. Of the 47 patients who experienced a delayed event, 68% were transferred to the ICU, 13% were transferred to a higher level of care, 17% had no change in the care delivery site, and one patient died. Of those with no delay, 66% were transferred to the ICU, 19% were transferred to a higher level of care, and 15% had no change in care delivery site. No significant differences were seen in the need for change in the location of care (p = 0.591); however, the only death occurred among patients experiencing a delay.

Relationship and predictors of delays

To determine if any of the variables examined could predict likelihood of a delayed event, seven predictor variables were examined: shift (07:00 to 19:00, 19:00 to 07:00); length of stay; case type (medical, surgical); cardiac monitoring; pulse oximetry; number of patients assigned (⩽5 patients, >5 patients) and care provided on designated unit (yes, no). When analysed using a forward stepwise method, shift and care provided on the designated unit were significant. No other variables added to the prediction of a delayed event. Shift was associated with a significance level of 0.009 and a 3.25 greater likelihood (95% CI 1.34 to 7.9) of a delay occurring on the 19:00 shift. The patient and unit type match was associated with a significance level of 0.014 (OR 0.07; 95% CI 0.009 to 0.579). The model correctly predicted delayed event accurately 68% of the time.

DISCUSSION

The aim in conducting this study was to identify differences in patient, nurse and organisational characteristics between medical and surgical patients who required a MET intervention. Major findings of the study were: (1) MET activations occurred significantly more often on the day shift than the night shift, for medical, but not surgical, patients; (2) delayed events occurred more often during the night shift; (3) there was a 3.2 greater likelihood of the MET activation being delayed on the night shift; and (4) two variables were found to predict greater odds of a delayed event: shift and unit placement match. In addition, significant differences between rescue events of medical and surgical patients were found for days in hospital prior to event, frequency of pulse oximetry monitoring and the use of patient-controlled analgesia. Also, there was a small but statistically significant difference between the groups in the number of patients assigned.

The pattern of deterioration that precedes a MET activation is not well understood.13^–15 Consistent with our findings, several studies have reported a pattern characterised by more MET activations during the day shift, compared with the night shift, as well as clustering of MET activations at times associated with routine observations and nursing handover.141617 Non clinical or support staff availability is typically greater during the day shift, which would increase the number of people supporting the registered nurse and visits to patient rooms. Accordingly, it is possible that we are “seeing” the patient with more vigilance during the day. There also appeared to be a pattern of frequency, with 51% of the day events occurring during 4 h of the shift (08:00, 13:00, 16:00, 18:00). This suggests that MET activations were also prompted by visits to patient rooms as part of routine care visits. These findings support the need to improve methods to identify patients who are in crisis and better respond to provide more clinically effective care.

As might be expected, delayed events, defined as an event for which documented clinical criteria for MET activation was present for >30 min prior to the call, were more common during the night shift. Like the pattern for total events, 45% of the delayed events are clustered around certain times. Of interest, the times associated with a greater number of delayed events (01:00; 06:00, 10:00, 16:00–19:00) correspond with times for routine nursing activities such as medication administration, charting and between-shift communication. Potentially, nursing personnel were occupied in the completion of routine activities for a group of patients and reduced the time for intensity of surveillance and response.

Consistent with studies reporting involving patients who experience a cardiac arrest, our findings suggest that patients have identifiable signs of deterioration for a considerable period of time prior to MET activation. METs were designed to improve patient safety by addressing “failure to rescue” issues17 but are dependent on mobilisation. At the time of the MET event, there was a significant difference between the medical and surgical patients for use of pulse oximetry, with surgical patients more likely to receive this monitoring. This finding raises questions regarding the sense of security that might be provided by this technology which, in turn, might result in less urgency in summoning help. Although patients who experienced a delayed event did not differ in the need for a higher level of care, the only death occurred among these individuals. New bedside monitoring systems can obtain a continuous feed from the bedside monitor, compare alterations with a patient normality model and sound an alarm that alerts staff to untoward changes. Such systems may provide earlier detection of adverse changes and therefore reduce the number of MET activations.18

Our study is the first to report a significant difference in MET activation associated with service/unit mismatch, that is, whether care was provided on a designated unit. More than 20% of the events for medical patients occurred while they were not being cared for on a medical unit. Conversely, less than 5% of events for surgical patients occurred when they were not on a surgical unit. Specialisation of nursing care provides patterns of knowing and recognition of clinically pertinent findings for particular patient types. The lack of familiarity of patient problems and anticipated response could influence the nurses’ recognition or realisation of the need for immediate intervention. The accuracy of patient assessment is likely more difficult and time-consuming, and it may be more difficult to identify all relevant details.1920 This finding highlights the potential vulnerability of patients when cared for by nurses who are less familiar with providing care to that population and suggests the need for further study.

There were no differences in education level or years of experience in the nurses caring for medical and surgical patients experiencing a MET event. In essence, medical and surgical patients in this study received care by a group of nurses with equivalent training and experience. The small percentage of nurses (<20%) who cared for the patient the day before the event, though not significant, is of interest in terms of the potential beneficial influence of continuity in recognising and responding to patient needs. This finding requires further exploration.

Our findings indicated a statistically significant difference in the mean number of patients assigned between the medical and surgical groups (5.6 (1.02) vs 5.2 (1.06) patients, respectively). However, the difference was small and judged not to be clinically significant. We also examined differences when events were categorised into those that occurred when the nurse was assigned ⩽5 or >5 patients and found no significant relationship with delays. However, there was a trend for more delays with larger assignments—eg, nurses assigned four patients had 21% of the delays, compared with 36% with a five patients and 43% with a six patients. Further, 75% of the events occurred when the patient:nurse ratio was 5:1 or 6:1. These findings suggest the need to further examine the relationship between the number of assigned patients and delayed events.

Our findings differ from those of Aiken and colleagues8 and Needleman and colleagues,21 who found significant differences in failure-to-rescue rates related to nurse education and the number of patients assigned. Both studies examined outcomes in large retrospective sample of patients with specific diagnoses, whereas our study compared differences in medical and surgical patients in a single institution with an established MET system. Therefore, one possible reason for this difference relates to differences in the design of the study.

We did not examine the influence of support from the multidisciplinary team or non-clinical or support staff availability, which could influence the time the registered nurse spends on direct patient observation. Also, there were likely shift-related differences in the availability of physician consultation and the availability of respiratory care personnel, a factor that could be critical given that rescue events were most commonly called for a cardiopulmonary reason. These considerations suggest that determining the appropriate number of patients assigned to a nurse may require the evaluation of multiple factors in addition to the nurse–patient ratio.

LIMITATIONS

There were several limitations of this study. First, the sample was moderate in size and was a convenient sample determined by the availability of the independent reviewer. It is possible that findings might differ if a larger sample of consecutive patients were examined. Second, the study examined outcomes on general medical and surgical units. Findings might differ if specialty units, such as cardiology, were included in the study. Third, the setting was a single university-affiliated hospital with well-defined criteria for activating the MET system. Findings might differ if events were examined in institutions with a teaching and non-teaching or community setting. Expanding this study to a multi-hospital study would provide stronger evidence of the associations found. The lack of variability in the data regarding number of patients prevents conclusions from being drawn regarding the influence on nurse staffing on events or delayed events.

CONCLUSION

Study findings suggest that patient, nurse and organisational characteristics can influence the timely rescue of hospitalised medical–surgical patients. Our findings add to the body of evidence that suggests that characteristics other than nurse–patient staffing ratios influence patient outcomes. At a time when states are developing policy on nurse staffing ratios, studies such as this demonstrate the need to examine more critically patient, nurse and organisational characteristics such as shift, unit, and time-related factors, that have the potential to increase the risk of a rescue event and, in particular, delayed events. This body of research has important implications for clinical practice and administrative decision-making regarding the organisation and delivery of medical–surgical patient care.