Regional ICU Audit Coordinator, The Anaesthetics Unit, St Bartholomew's and the Royal London School of Medicine and Dentistry, The Royal London Hospital, Alexandra Wing, 4th Floor, Whitechapel, London E1 1BB, UK
Goldhill, Senior Lecturer and Consultant Anaesthetist
A ‘patient-at-risk team’, established to allow the early identification of seriously ill patients on hospital wards, made 69 assessments on 63 patients over 6 months. Predefined physiological criteria were not able to reliably predict which patients would be admitted to the intensive care unit. The incidence of cardiopulmonary resuscitation before intensive care admission was 3.6% for patients seen by the team and 30.4% for those not seen (p < 0.005). Of admissions seen by the team, 25% died on the intensive care unit compared with 45% of those not seen (not significant, p = 0.07). Among those not seen by the team, mortality was 40% for those who did not require resuscitation and 57% for those who did (not significant). Many critically ill ward patients had abnormal physiological values before intensive care unit admission. Identification of critically ill patients on the ward and early advice and active management are likely to prevent the need for cardiopulmonary resuscitation and to improve outcome.
Patients admitted from hospital wards to an intensive care unit (ICU) have a higher overall percentage mortality than patients admitted from other areas of the hospital . Despite being on a hospital ward, often for several days, a high percentage of these admissions receive cardiopulmonary resuscitation (CPR) before their ICU admission . Before cardiac or respiratory arrest, ward patients often have severe physiological abnormalities which are known to the medical and nursing staff [2–4]. Similarly, many patients admitted from the wards to ICU have abnormal physiological values in the hours preceding ICU admission . Because of this we thought that criteria based on physiological values may be useful in identifying critically ill ward patients at risk of suffering a cardiac or respiratory arrest, or requiring imminent ICU admission. We hypothesised that, if we could identify these patients early enough, action could be taken to prevent further physiological deterioration and improve outcome.
A ‘patient-at-risk team’ (PART), similar to the medical emergency team described by Lee et al. , was established in order to respond to these patients. The PART assessed patients who fulfilled certain physiological criteria, as well as other patients causing concern to medical and nursing staff. The PART aimed to improve care for these patients by providing advice and support to those responsible for them on the wards, by facilitating early ICU admission when appropriate, and by preventing unnecessary ICU admissions, thereby releasing valuable beds for use by patients in greater need.
The aims of this study were to see whether the physiological criteria used to call the PART were appropriate and useful in determining the necessity for admission to ICU, and whether early review or intervention improved patient outcome.
Ethics committee approval was obtained for this prospective study which took place between 1 June 1997 and 30 November 1997 at the Royal London Hospital. An earlier audit of patients admitted to ICU from the wards  described physiological abnormalities associated with patients requiring ICU admission. Based on these findings a committee with surgical, medical, intensive care and nursing input agreed criteria for identifying high-risk patients on the wards. The PART consisted of the ICU consultant or deputy, a senior ICU nurse and the duty medical or surgical registrar as appropriate. The criteria for alerting the PART are given in Table 1. With the support of the nursing directorate and the hospital's consultant staff this protocol was introduced onto the wards. Every ward was visited, laminated copies of Table 1 were placed on the ward notice boards and information about the PART was circulated to nurses and doctors within the hospital.
Table 1. The PART protocol
The intention of the PART protocol was to alert the doctor directly responsible for a seriously ill patient. A combination of physiological abnormalities was used to prompt the ward nurses to call the doctor. If the doctor required support in caring for the patient or, in exceptional circumstances, the nurses were unable to obtain a suitable response from the doctor, the PART was contacted. Doctors were also advised that they could call the PART for any seriously ill patient causing acute concern.
The early identification of a patient at risk permitted discussion on suitability for resuscitation and ICU admission, allowed for earlier intervention to prevent physiological deterioration and aided planning of ICU admissions and resources. After assessment some patients were transferred directly to ICU. If the patient remained on the ward, the PART would advise on management and decide whether regular review was necessary. If no intervention was required this was also noted.
Information on each patient seen by the PART was recorded. This included details of when the patient was seen, their age, sex and hospital admission date, and the reason for the request. A record was made of previous surgery or ICU admission, and defined interventions and therapy. If available, a record was made of the patient's most recently recorded temperature, heart rate, arterial systolic and diastolic pressures, respiratory rate, Glasgow Coma Score (GCS) and oxygen saturation. If results were available from the previous 24 h, the most recent values of serum sodium, potassium, haemoglobin, white cell count and creatinine were recorded. Values of pH, and Pao2 and Paco2 were recorded if an arterial blood sample had been taken within 6 h of the assessment. A record was made of the findings on initial assessment, of the advice given and the action taken. All patients seen by the PART were followed up to record subsequent ICU admissions, the date of hospital discharge and outcome.
The data were examined to see whether the predefined combination of physiological values for calling the PART were of any value. The physiological values recorded by the PART were used to assign Acute Physiology and Chronic Health Evaluation II (APACHE II) points . An APACHE II score  was calculated by summating the points for the physiological values and adding points for age and chronic health problems. We compared the physiological values, APACHE II scores, procedures and interventions for the PART assessments admitted to ICU within 48 h with those who remained on the ward.
Data were also collected prospectively on all patients admitted to ICU from the wards during the study period. This included personal details of each patient, the date of ICU and hospital admission and discharge. Each patient's notes and the pathology computer were examined for information on physiological values, procedures and interventions for the 24 h preceding ICU admission. Information on the ICU stay and patient outcome was obtained from the intensive care database. Further details recorded included: diagnosis, history of chronic health problems, previous surgery, ICU admission or CPR during the present admission. An APACHE II score  was calculated from the most extreme physiological values in the 24 h before and after ICU admission. In patients who received CPR before ICU admission, APACHE II scores were calculated using preresuscitation physiological values.
For all patients admitted from the wards to ICU, we compared those patients seen by the PART within 48 h of admission with those who were not seen. Comparisons included physiological values and the incidence of CPR in the 24 h before ICU admission, the APACHE II scores before and after ICU admission, and ICU outcome. For comparison of the physiological values we used the highest values of temperature, heart rate and respiratory rate and the lowest values of mean arterial pressure (MAP). Similar comparisons were made in patients admitted to ICU who were not seen by the PART in the 48 h before admission between those patients who did and those who did not receive CPR.
Categorical data were analysed using the Chi-squared test, with Yates' correction where applicable, and continuous data were evaluated using a Mann–Whitney or t-test. Standard deviation (SD) or range are given where appropriate.
Ward assessments by the PART
During the 6-month study period, the PART was called 69 times to see 63 patients. Follow-up visits are not included in this number. Six patients were each assessed twice with a range of 7–74 days between assessments.
The disposition and outcome of patients assessed by the PART are given in Fig. 1. Forty-four per cent of those assessed were admitted to ICU within 48 h and they had a 32% mortality. Of those not admitted to ICU within 48 h of their assessment, 29% were admitted later during their hospital stay and 26% died in hospital.
Following assessment and appropriate discussion, ‘do not resuscitate’ orders were written for two patients. One was 76 years of age and had bowel cancer with liver metastases. She had been in hospital for 11 days at the time of the assessment and had undergone surgery 9 days previously followed by an overnight stay in ICU. She was discharged alive from hospital 13 days after being seen. The other patient was aged 49 years and had suffered an intracranial haemorrhage. He had been in hospital for 21 days at the time of the assessment, had not undergone surgery or been previously admitted to ICU. He died on the day he was seen.
Seven patients were admitted to ICU more than 48 h after assessment, two after 3 days, three after 5 days, one after 7 days and one after 33 days.
Table 2 contains details of the PART assessments grouped by whether they were followed by admission to ICU within 48 h. No statistically significant differences were found. Before assessment, five patients had undergone intracranial surgery, three spinal surgery, 11 intestinal surgery, one orthopaedic surgery, two vascular surgery and six patients had undergone other operations including: surgical placement of central vascular access, drainage of an abscess and wound debridement.
Table 2. Details of PART assessments ICU, admissions to ICU within 48 h of assessment; no ICU, not admitted to ICU within 48 h of assessment; hospital stay before assessment, median (interquartile range) days in hospital before assessment.
Serious chronic health problems were defined according to the APACHE II scoring system . They were common, occurring in 22 (35%) of the 63 patients. The criteria for chronic health problems are specific and describe a severe restriction on activity or risk to life.
The percentage of admissions where a physiological value was recorded at the time of assessment is given in Table 3.
Table 3. Percentage of assessments with a physiological value recorded ICU, admissions to ICU within 48 h of assessment; no ICU, not admitted to ICU within 48 h of assessment; Temp., temperature; HR, heart rate; BP, arterial pressure (systolic and diastolic); Resp., respiratory rate; GCS, Glasgow Coma Score; Spo2, oxygen saturation; ABG, arterial blood gas analysed within 6 h of assessment; Na, sodium; K, potassium; Hb, haemoglobin; WBC, white cell count; Creat., creatinine.
At least one predefined criterion for calling the PART was fulfilled at assessment by 97% of patients admitted to ICU within 48 h and by 82% of patients who remained on the wards (Table 4). Of those admitted to ICU within 48 h of assessment, two or more of the criteria were fulfilled by 80% and three or more by 27%. The equivalent figures for those who stayed on the wards were 59% and 33%. There was no significant difference in any of the individual criteria between those patients admitted to ICU within 48 h and those not. The commonest abnormality was tachypnoea, followed by a depressed level of consciousness and tachycardia. The median (interquartile range) of the APACHE II scores was 14 (11–19.5) for those admitted to ICU within 48 h of assessment and 12 (7.25–18) for those not.
Table 4. Percentage of patients fulfilling criteria for calling the PART The criteria are those described in Table 1. There were six main physiological criteria as well as secondary criteria. ICU, admissions to ICU within 48 h of assessment; no ICU, not admitted to ICU within 48 h of assessment; Resp., respiratory rate (breaths.min−1); HR, heart rate (beats.min−1); SBP, arterial systolic pressure (mmHg); GCS, Glasgow Coma Score; Spo2, % oxygen saturation; Urine, % with urine output <100 ml over 4 h before assessment.
On assessment, a decision was made to admit the patient immediately to ICU, to review at least 4-hourly, to review at least daily but less frequently than every 4 h, or not to keep under regular review (Table 5). The results suggest that admission to ICU and frequent review was more likely in patients with a high respiratory rate (criterion 1) and a depressed level of consciousness (criterion 4). Many of the patients, whether admitted immediately to ICU, reviewed later or not reviewed, had other physiological abnormalities. Advice and practical help was given to the ward team responsible for the patient, primarily in the management of respiratory problems and hypovolaemia. Three patients were seen with acute surgical problems. Two were resuscitated and admitted to ICU after surgery. The third was resuscitated in ICU before the operation.
Table 5. Decision made at assessment by physiological criteria and APACHE II score ICU, % admitted to ICU within 48 h of assessment; main criteria (1,2,3,4,5,6), % fulfilling main criteria for calling team, see Table 4; APACHE II, median (interquartile range) APACHE II score.
At assessment, many of the patients were already being monitored and treated more intensively than normal for the wards. Of those admitted to ICU within 48 h of assessment over 90% were being monitored with a pulse oximeter and 87% were being given oxygen. Oximetry was used in 71% of those not admitted to ICU within 48 h with 79% receiving oxygen. Over 70% of patients were having noninvasive arterial pressure measured regularly, nearly 40% were attached to an electrogardiogram (ECG) monitor and over 30% had an urinary catheter in place.
ICU admissions from the wards
Over the 6-month study period 422 patients were admitted to ICU, 119 from the accident and emergency department, 160 from the operating theatres and recovery, 99 from the hospital wards and 44 as transfers from other hospitals or critical care facilities. Data on two patients admitted from the wards, one within 48 h of a PART assessment, were not recorded at ICU admission and their notes were subsequently unobtainable. They are not included in the analysis. Following PART assessments, 31 patients were admitted to ICU within 48 h. Two of these patients had surgery after the assessment and before ICU admission. The following analysis therefore concerns 28 patients admitted directly from the wards to ICU having been assessed by the PART within 48 h of admission, and 69 patients who were not assessed by the PART before admission (Table 6).
Table 6. Details of patients admitted from the wards to the ICU Seen by PART, seen by the PART within 48 h of ICU admission; CPR, cardiopulmonary resuscitation; age, mean age in years (SD); days in hospital, median (interquartile range) of days in hospital before ICU admission; ICU stay, median (interquartile range) days in ICU; ICU stay survivors, median (interquartile range) days for those who survived to leave ICU; pre-APACHE, median APACHE II score (interquartile range) from values from the 24 h before ICU admission; ICU APACHE, median APACHE II score (interquartile range) from values from the 24 h after ICU admission. NS, not significant.
One patient had a cardiorespiratory arrest on the ward 30 min after the PART arrived. At this time the patient was in the care of the ICU team, had already been intubated and was receiving intensive resuscitation. We have not included this patient in those who arrested before intensive care began. There was a significant difference in the number of patients who received CPR before ICU admission (one patient reviewed by the PART versus 21 not reviewed). Although there was a difference in percentage ICU mortality, this did not reach statistical significance. The data demonstrate that the majority of patients were in hospital for some days before their ICU admission. Many had previously undergone surgery or a previous ICU admission.
A minority of patients were seen by the PART before ICU admission and they may have presented with different problems from the other admissions from the wards. The differences between those who were or were not seen by the PART may reflect the treatment available in ICU, such as advanced respiratory care, as well as the practice of individual medical and surgical firms (Table 7). In this way the PART may have been called more frequently to patients with severe respiratory or airway problems, and critically ill neurosurgical and other patients may be routinely looked after on the wards and only come to the attention of the ICU when admission is imminent.
Table 7. Reasons for ICU admission Seen by PART, seen by the PART within 48 h of ICU admission; COAD, chronic obstructive airways disease; CCF, congestive cardiac failure; IHD, ischaemic heart disease; GI, gastrointestinal.
The percentage of patients for whom a physiological value was recorded during the 24 h before ICU varied widely (Table 8). For many of the physiological parameters the average worst values were well outside the normal range. The pH, Pao2 and Paco2 required analysis of an arterial blood sample, although a value was available for over 60% of admissions. The GCS was most likely to be recorded in patients with a depressed level of consciousness and a value for the 24 h before admission could only be found for 19%. It is therefore likely that the results overestimate the percentage of admissions with a GCS < 15 in the 24 h before ICU admission. Several patients not seen by the PART had very high levels of Pao2, which probably reflected the inspired oxygen concentration and is unlikely to be clinically relevant.
Table 8. Physiological values recorded in the 24 h before ICU admission Seen by PART, admitted to ICU within 48 h of the PART assessment; % available, % of ICU admissions for whom a value for the physiological variable was found in the notes or on the pathology computer; temp., temperature (°C); MAP, mean arterial pressure (mmHg), HR, heart rate (beat.min−1), resp., respiratory rate (breaths.min−1); Na, sodium (mmol.l−1); K, potassium (mmol.l−1); Hb, haemoglobin (g.dl−1); WBC, white cell count (×109.l−1), Pao2 and Paco2 in kPa; GCS, Glasgow Coma Score, median and interquartile range.
Based on physiological values recorded in the 24 h before ICU admission, at least one predefined criterion for calling the PART was fulfilled by 89% of patients admitted to ICU who had been assessed by the PART within the previous 48 h and by 81% of patients who had not been assessed. Of those admitted to ICU within 48 h of assessment, two or more of the criteria were fulfilled by 85% and three or more by 60%. The equivalent figures for those who were not seen by the PART before ICU admission were 54 and 35%.
In Table 9, ICU admissions not seen by the PART have been subdivided by whether they received CPR in the 24 h before ICU.
Table 9. Details of patients not seen by the PART and admitted from the ward to the ICU CPR, cardiopulmonary resuscitation; age, mean age in years (SD); days in hospital, median (interquartile range) of days in hospital before ICU admission; ICU stay, median (interquartile range) days in ICU; ICU stay survivors, median (interquartile range) days for those who survived to leave ICU; pre-APACHE, median APACHE II score (interquartile range) from values from the 24 h before ICU admission; ICU APACHE, median APACHE II score (interquartile range) from values from the 24 h after ICU admission.
The introduction of the PART was prompted by the high percentage mortality in patients admitted to the ICU from ward areas . As these patients are already within the hospital, and therefore accessible, it is possible to introduce changes in management to provide early intervention with the aim of reducing mortality. The PART protocol was a simple way of trying to identify critically ill patients on the wards. Similar teams have already been established elsewhere . The physiological criteria, based on ICU admissions from the wards , were similar to the physiological abnormalities found to precede cardiac arrest in other studies [2–4]. These included an abnormal respiratory rate, abnormal pulse rate, hypotension and an altered level of consciousness.
The PART was well publicised and considerable effort went into informing and educating the ward and ICU nursing and medical staff about its existence and purpose. Despite the availability of the PART, the majority of patients were not assessed before admission to ICU and a high percentage of these patients received CPR before admission. It is possible that some doctors and nurses were unaware of the existence of the PART. Alternatively, as there was no compulsion to call the PART, the doctor caring for the patient may have decided not to call the team. The data do not suggest that there was a marked difference in physiological values between patients who were seen by the PART before ICU admission and those who were not, although there may have been some differences in the reasons for ICU admission. Many of the patients not seen by the PART had physiological abnormalities and were being actively monitored and treated in the 24 h before ICU admission.
The criteria for calling the PART identified seriously ill patients, many of whom had multiple physiological abnormalities (Table 4) and were already being intensively monitored at the time of assessment. The use of monitoring and the high incidence of interventions indicate that many of these patients were recognised by the ward staff as being critically ill before the PART were called. After a patient had been seen by the PART they were either admitted directly to the ICU or management advice was given with or without a plan for review by the PART at a later date. The ICU was often full and some patients the PART would like to have admitted were managed on the ward because of a lack of intensive care beds. The mortality of patients not admitted to ICU within 48 h of assessment was high and many were admitted to ICU later.
The commonest physiological abnormalities in the patients admitted to ICU were tachypnoea and an altered level of consciousness. It is possible that we chose to admit these patients because they would benefit from the therapy available on ICU, such as ventilation. Interestingly, oxygen saturation was not very useful at determining ICU admission. An oxygen saturation of < 90% was more common in patients who were not admitted to ICU than those who were. A possible explanation for this is that single system respiratory problems are often managed on the ward with physiotherapy, antibiotics, oxygen and in some cases CPAP.
APACHE II scores were calculated for each patient from the physiological values obtained on the wards. Scores are related to the degree of physiological abnormality, selected severe chronic health disorders and the age of the patient . The APACHE scoring system, as originally described, uses information obtained during a patient's first 24 h on ICU, not prior to admission. It can be used to predict outcome in a cohort of ICU patients but should not be used to predict individual outcome [8, 9]. Although there is a superficial relationship between the APACHE II score and the decisions made by the PART (Table 5) the score is unlikely to be useful when making decisions with respect to individual patients.
Of the 97 patients admitted to ICU directly from the ward, 28 were seen by the PART within the preceding 48 h. The incidence of CPR among this group was strikingly and significantly lower than among those patients not assessed by the PART. Our data and other studies in patients who suffer an in-hospital cardiac arrest demonstrate a very high mortality rate , and there is likely to be considerable benefit from preventing the need for CPR .
The physiological criteria used to notify the PART are likely, by themselves, to be unsatisfactory for identifying critically ill patients on the wards. Some patients admitted to the ICU fulfilled none, and many fulfilled fewer than three, of the six main criteria. Many of these patients must have been recognised as being seriously ill with a high percentage being monitored and receiving oxygen on the ward before ICU admission.
The Royal London Hospital and its ICU are not typical and have a higher than usual number of emergency, trauma and seriously ill patients. There are no high-dependency beds within the hospital, which may restrict access to critical care facilities. The workload and impact of the PART may therefore not be directly applicable to some other hospitals. The study demonstrates that seriously ill patients, requiring intensive treatment, are to be found on the hospital wards. Many of the patients were already known by the ward staff to be seriously ill. The PART probably only saw a proportion of patients who could potentially benefit from improved care. Useful decisions were made, including issuing ‘do not resuscitate’ orders, and knowing about critically ill patients on the wards undoubtedly helped in planning and organising ICU admissions. Admission to a high-dependency or ICU area would have been desirable for many of the patients assessed and options were limited by the lack of facilities. The PART appeared to be successful in preventing the need for CPR and may help decrease the mortality of critically ill patients on the wards. There were 101 cardiac arrest calls to the ward during the study period and 133 in the preceding 6 months. There is therefore considerable potential to further improve the outcome of critically ill patients on the wards.
A recent study of pre-ICU care showed that management of critically ill patients was often suboptimal . Much more could be done to prevent and treat physiological abnormalities in these critically ill patients on the wards and to prevent the need for cardiopulmonary resuscitation. Early identification of these patients is essential, combined with timely, appropriate treatment on the wards or in high-dependency or intensive care facilities.
We are grateful for the advice and cooperation of our colleagues who helped draw up the PART protocol. The intensive care nurses and ward nurse at the Royal London Hospital enthusiastically supported the team and helped us identify and follow-up the patients we assessed. We should also like to thank the intensive care residents who carried much of the burden of completing the data collection forms at the initial assessment of the ward patients.