Visensia for early detection of deteriorating vital signs in adults in hospital

Study component

Description

Objectives/hypotheses

To assess whether using an integrated monitoring system (Visensia) that continuously amalgamates 4 single monitoring parameters input (including HR, RR, BP, and SpO2) into an instability index value (INDEX), correlates with a single‑parameter cardio‑respiratory instability concern criterion, and whether nurse response to INDEX alert for patient attention is associated with instability reduction.

Study design

Prospective, single‑centre, before‑and‑after evaluation in sequential 8‑, 16‑, and 8‑week phases (phase 1, phase 2, and phase 3 respectively), with a total project duration of 32 weeks (November 2006 to August 2007).

Phase 1: patients received continuous single‑channel monitoring (heart rate, respiratory rate, blood pressure, and peripheral oxygen saturation) and standard care; INDEX background was recorded but not displayed.

Phase 2: INDEX was background‑recorded; members of staff were educated on use.

Phase 3: staff used a clinical response algorithm for INDEX alerts.

All phase 1 and 3 data were scored blindly off‑line and independent of phase.

The best balance between cardio‑respiratory instability concern criteria and the INDEX threshold was at a sensitivity of 70.5% and specificity of 71% when the INDEX alert threshold was set at ≥3.2. This SDU‑specific INDEX alert threshold was sensitive and specific to detect cardio‑respiratory instability as defined by single‑parameter vital sign criteria (i.e. an HR of <40 or >140 beats/min, an RR of <8 or >36 breaths/min, systolic BP of <80 or >200 mm Hg, diastolic BP of >110 mm Hg, and SpO₂ of <85%) for instability concern, which also served as medical emergency treatment activation triggers. The ≥3.2 INDEX value was used for staff alert to an instability state in phase 3.

Setting

A 24‑bed trauma step‑down unit in a single urban tertiary care centre in the USA.

Inclusion/exclusion criteria

Not reported. Subjects were all monitored patients.

Outcomes

The correlation between the monitoring system INDEX and instability states according to the University of Pittsburgh Medical Centre cardio‑respiratory instability concern criterion (INSTABILITY);

Comparison of phase 1 and 3:

the mean numbers of times admissions in each phase experienced an episode of instability and mean durations of times admissions in each phase were in unstable states

the number of admissions experiencing at least one episode of an instability state

the cumulative number of occurrences in which patients were above the thresholds of instability concern criteria for each phase

the cumulative duration of time in which patients were above the thresholds

instability for patients who experienced it in phase 1 and phase 3.

The categorisation of instability was performed independently by 2 critical care medicine physician investigators experienced in cardio‑respiratory instability with the data blinded to data phase and using only the parsimonious definitions.

Statistical methods

Analyses were performed by using Student's t tests, chi‑square or Fisher exact comparisons, and Spearman's rho correlation. Significance corresponded to p<0.05.

Patients included

All potentially unstable patients who were transferred out of intensive care units ICUs to step‑down units for further care, including 326 patients in phase 1 and 306 patients in phase 3 during these intervals.

Results ^a

Compared with phase 1, phase 3 had statistically significant reduction in average duration of any episode of instability (INSTABILITY_hit) per admission, average duration of physiologically plausible instability (INSTABILITY_min) per admission, and average number of full instability (INSTABILITY_full) per admission. There was no statistically significant difference between the phases in the other reported INSTABILITY categories.

Authors' conclusions

The integrated monitoring system INDEX correlated significantly with cardio‑respiratory instability concern criteria, usually occurred before overt instability, and was associated with decreased cardio‑respiratory instability concern criteria when coupled with a nursing alert in step‑down unit patients.

Abbreviations: BP, blood pressure; CI, confidence interval; ICUs, Intensive care units; HR, heart rate; n, number of patients; OR, odds ratio; RR, respiratory rate; SpO₂, peripheral oxygen saturation.

^a INSTABILITY, vital sign monitoring parameters were beyond instability‑concern thresholds; _hit, vital‑sign‑monitoring parameters across instability concern thresholds for any cause, including artifact; _min, subset of _hit for instability that was physiologically real (i.e. nonartifactual) even if transient; _full, subset of min for instability judged as serious and persistent and in need of intervention.

Phase 3 (Visensia)

Phase 1

(usual care)

Analysis

Number of patients monitored

n=306

n=326

Efficacy ^a

n=303

n=319

Selected outcomes

Average number of INSTABILITY_hit per admission (mean±SD) ^b

2.5±5

4.0±9

p=0.068

Average duration of INSTABILITY_hit per admission (minutes, mean±SD) ^b

25±57

57±129

p=0.007

Average number of INSTABILITY_min per admission (mean±SD) ^b

1.5±4

2.2±6

p=0.898

Average duration of INSTABILITY_min per admission (minutes, mean±SD) ^b

13±41

28±76

p=0.018

Average number of INSTABILITY_full per admission (mean±SD) ^b

0.4±1.5

0.9±3

p=0.033

Average duration of INSTABILITY_full per admission (minutes, mean±SD) ^b

7±27

16±53

p=0.050

Number of admissions experiencing at least one INSTABILITY_min episode (n, %) ^b

158 (51%)

173 (52%)

p=0.57

Number of admissions experiencing at least one INSTABILITY_hit episode (n, %) ^b

102 (33%)

114 (35%)

p=0.565

Number of admissions experiencing at least one INSTABILITY_full episode (n, %) ^b

48 (15%)

68 (20%)

p=0.09

Death (n, %)

3 (1%)

7 (2%)

Not reported

Unexpected death

0

6

Not reported

Abbreviations: CCU, critical care unit; CI, confidence interval; ICU, intensive care unit; ITT, intention to treat; NR, not reported; n, number of patients or events; RR, relative risk.

^a Total number of admissions alive at discharge.

^b INSTABILITY, vital sign monitoring parameters were beyond instability‑concern thresholds; _hit, vital‑sign‑monitoring parameters across instability concern thresholds for any cause, including artifact; _min, subset of hit for instability that was physiologically real (i.e., nonartifactual) even if transient; _full, subset of min for instability judged as serious and persistent and in need of intervention.

Study component

Description

Objectives/hypotheses

To assess whether mandated electronic vital signs monitoring reduced the frequency of adverse events in high risk medical and surgical patients outside of critical care areas, compared with that in a control group receiving usual ward care.

Study design

Randomised controlled trial. Ward and study staff were not blinded to the intervention.

For the group with mandated electronic vital signs monitoring, BioSign was only used to record vital signs data which was evaluated retrospectively. True episodes of severe physiological abnormalities were determined by 2 senior clinicians.

Setting

Medical or surgical acute ward in John Radcliffe Hospital, Oxford, UK; between September 2003 and September 2005. Follow‑up: from recruitment for 72 hours or until the patient or caring nurse requested removal.

Inclusion/exclusion criteria

Adult 'high risk' patients admitted as medical or surgical emergencies or undergoing major elective surgery at the study hospital between September 2003 and September 2005 were recruited. 'High risk' patients were those where the expected rate of complications (including death) from the primary illness or procedure exceeded 5% in published case series, the control groups of trials, or local audit data. Patients scheduled to receive their initial postoperative care on an intensive care unit were included but were not assigned until they were discharged to the acute ward. Patients were not included if they were expected to be sufficiently mobile to leave the bed space unaided within 72 hours of operation or recruitment.

Primary outcomes

Proportion of patients experiencing a major event (urgent staff calls, a change to a higher care level, cardiac arrest, or death) in 96 hours following randomisation.

Statistical methods

The sample size was estimated to be 405 to detect a clinical significant difference in event rates of 15%. Chi‑squared tests were used to determine the statistical significance of any difference in proportions of patients with major adverse events.

Patients included

Patients admitted as medical or surgical emergencies or undergoing major elective surgery (n=405).

Results

Patients with a major event by 96 hours: 113 (56%) monitored patients compared with 116 (58%) control patients; OR 0.94 (95% CI 0.63–1.40), p=0.76. Out of 690 transitions from normal to abnormal physiological activity, 652 were considered true episodes. There were technical problems with the BioSign device that prevented recording for the whole monitoring period in 33 (17%) patients. The patients with severe abnormalities that would have caused an alert by the BioSign device were more likely to have a major event (sensitivity 63%, specificity 52%).

Authors' conclusions

Mandated electronic vital signs monitoring in high risk medical and surgical patients had no effect on adverse events or mortality.

Abbreviations: CI, confidence interval; OR, odds ratio; n, number of patients.

ID

Design/method

Findings

Sen et al. 2009; Sen et al. 2010a ^a

Retrospective analysis of a trauma registry data. N=117 patients admitted to a level 1 trauma centre over a 6‑month period. Vital signs were obtained from the pre‑hospital run‑sheets and upon arrival to the emergency department. An initial pre‑hospital VSI and an emergency department VSI were calculated. Pre‑hospital life‑saving interventions (LSIs) and those carried out within 6 hours of arrival to the trauma centre (fluid bolus, CPR drugs, intubation, transfusion etc.) were considered outcome variables.

Univariate analysis: pre‑hospital VSI >3 was predictive of trauma patients who needed LSI (OR1.8, 95% CI 1.1–4.2).

Multivariate analysis: VSI had significant likelihood ratios for life‑saving interventions including endotracheal intubation, blood transfusion, CPR and use of resuscitation drugs (p<0.001). The model had an area under ROC of 0.76 in discriminant analysis. VSI outperformed other independent variables like heart rate, blood pressure, injury severity score and base deficit.

Sen et al. 2010b ^a

Retrospective analysis of a trauma registry data. N=297 patients admitted to a level 1 trauma centre over a 6‑month period. Vital signs were obtained from the pre‑hospital run‑sheets. Pre‑hospital VSI was calculated based on the vital signs in a blinded manner. Pre‑hospital life‑saving interventions and those carried out within 6 hours of arrival to the trauma centre (CPR, resuscitative drugs, intubation, blood transfusion, chest tubes, emergency laparotomy etc) were considered outcome variables.

Multivariate analysis: pre‑hospital VSI >3 was predictive of trauma patients who needed LSI after arrival to a trauma centre in the (OR1.8, 95% CI1.3–3.4; p<0.05). The model had an area under ROC of 0.79 in discriminant analysis. VSI outperformed other independent variables like heart rate, systolic blood pressure, oxygen saturation and Glasgow Coma Scale.

Choukalas et al. 2011

A cohort study of the 20 most recent consecutive patients suffering cardiac arrest requiring ACLS level care, in a mixed medical‑surgical‑cardiac ICU in a large, urban, tertiary‑care, academic teaching hospital. Data were collected sufficient to calculate a VSI at 5‑minute intervals for the 20 hours prior to cardiac arrest. The primary outcome measure was the lead‑time between the first episode of instability (defined as a VSI≥3.2) and cardiac arrest. Patient records were also reviewed to identify the first point of nursing documentation of patient instability within the 20 hours prior to arrest.

Of the 20 most recent cardiac arrests, 6 were excluded because they did not require ACLS care. Of the remaining 14 arrests, 9 were attributed to cardiac causes and the remainder respiratory. Of the 14 arrests 8 were fatal. The mean lead‑time of the VSI alert prior to arrest was 15.1 (SD=6.6) hours. Nurses documented instability an average of 9.3 (SD=7.1) hours prior to arrest.

Choukalas et al. 2015

A controlled cohort study in a large, urban, academic teaching hospital with 18 ICU beds. Data were extracted to calculate a VSI at one‑minute intervals for the 24 hours prior to arrest for patients undergoing cardiac arrest in the ICU between 2005 and 2011 as identified by a hospital quality improvement database. Control patients were all patients in the ICU during the 24 hour periods which defined cases. Hourly average VSI were calculated and compared between cases and controls using mixed effects linear models with a random effect for each subject.

Sixty one cases and 729 controls were included in the analysis. Cases were more ill than controls in that a greater proportion of them had co‑morbidities such as congestive heart failure and recent myocardial infarction. VSI showed no difference between the 2 groups at the beginning of the observation period, but became significantly higher for cases starting at 10 hours prior to arrest (p<0.05). VSI for cases continued to rise and separate from that of controls consistently leading up to the end of the observation period, when cases experienced cardiac arrest. The averaged vital signs did not show trends or changes in the hours leading up to the eventual cardiac arrest.

Abbreviations: ACLS, advanced cardiac life support; CI, confidence interval; CPR, cardiopulmonary resuscitation; ICU, intensive care unit; LSI, life‑saving intervention; n, number of patients; ROC, receiver‑operating characteristic; RR, relative risk; SD, standard deviation; VSI, Visensia Safety Index.

^a Both retrospective studies were conducted by the same authors analysing registry data on Visensia index for the prediction of life‑saving interventions in pre‑hospital trauma patients. It was unclear whether the Sen et al. 2010b study combines data from the Sen et al. 2009 and Sen et al. 2010a studies.