Prediction and monitoring of in-hospital cardiac arrest

Abstract

Background: In-hospital cardiac arrest (IHCA) is a global health concern of major importance, associated with a poor prognosis. IHCA is frequently heralded by a deterioration of vital signs, and many cases are considered preventable. Hence, prevention has become a key strategy. The overall aim of this thesis was to study the prevention of IHCA, by means of prediction and monitoring, with a view to improve patient safety.

Methods: Study I and III are observational cohort studies, based on the Swedish Registry of Cardiopulmonary Resuscitation (SRCR). In study III, we also collected additional data from medical records in a small, hypothesis-generating group of patients.

Study II and IV are prospective, observational cohort studies based on patients reviewed by Rapid Response Teams (RRTs) in 26 and 24 Swedish hospitals, respectively. In study IV, additional data on long-term survival was obtained from either medical records or the personal information directory, containing population registration data.

Specific study aims and results: In study I, we investigated how 30-day survival after IHCA was influenced by ECG monitoring at the time of collapse, as well as clinical factors that determined whether patients were ECG monitored adjacent to cardiac arrest (CA). In all, 24,790 patients were enrolled in the SRCR between 2008 and 2017. After applying the exclusion criteria, 19,225 patients remained, of which 52% were monitored at the time of collapse. In all, 30-day survival was 30%. ECG monitoring at the time of CA was associated with a Hazard Ratio of 0.62 (95% Confidence Interval 0.60-0.64) for 30-day mortality. The strongest predictor of ECG monitoring adjacent to IHCA was location in hospital. There were tangible variations in the frequency of patients who were ECG monitored at the time of collapse between Swedish regions and across hospitals.

In study II, we investigated the predictive power of NEWS 2, as compared to NEWS, in identifying patients at risk of Serious Adverse Events (SAEs) within 24 hours of an RRT-review. In all, 1,065 patients, reviewed by RRTs in general wards during the study period between October 2019 and January 2020, were included. After applying the exclusion criteria, 898 patients were eligible for complete case analyses. In all, 37% of the patients were admitted to the Intensive care unit (ICU) within 24 hours of RRT-review. In-hospital mortality and IHCA were uncommon (6% and 1% respectively). The Area Under the Receiver Operating Characteristic (AUROC) for both NEWS and NEWS 2 was 0.62 for the composite outcome, and 0.69/0.67 for mortality. Regarding the outcome unanticipated ICU admission, the AUROC was 0.59 and 0.60, respectively, while the AUROC for IHCA was 0.51 (NEWS) and 0.47 (NEWS 2), respectively.

In study III, we investigated 30-day survival and ROSC in patients suffering from IHCA, who were reviewed by an RRT within 24 hours prior to the CA, as compared to those without such review. Furthermore, we studied patient centred factors prior to RRT activation, the timeliness of the RRT-review as well as the reason for the RRT-review. We also investigated the association between RRT interventions and outcome.

During the study period between 2014 and 2021, 19,973 patents were enrolled in the SRCR. After applying the exclusion criteria, 12,915 patients remained. Among these IHCA patients, there was an RRT/ICU contact within 24 hours prior to the CA in 2,058 cases (19%).

The adjusted 30-day survival was lower among patients reviewed by an RRT prior to IHCA (25% vs. 33%, p <0.001). Regarding ROSC, we did not observe any difference between the groups. The propensity score based Odds Ratio for 30- day survival was 0.92 for patients who were reviewed by an RRT (95% CI 0.90 to 0.94, p <0.001), as compared to those who were not RRT- reviewed within 24 hours prior to IHCA. A respiratory cause of CA was more common among IHCA patients who were reviewed by an RRT. In the small, explorative subgroup (n=82), 24% of the RRT activations were delayed, and respiratory distress was the most common RRT trigger. We observed a significantly lower 30-day survival among patients triaged to remain at ward compared to those triaged to a higher level of care (2% vs. 20%, p 0.016).

In study IV, we explored the impact of age on the ability of NEWS 2 to predict IHCA, unanticipated ICU-admission, or death, and the composite of these three SAEs, within 24 hours of review by an RRT. Furthermore, we aimed to investigate 30-, 90- and 180-day mortality, and the discriminative ability of NEWS 2 in the prediction of long-term mortality among RRT-reviewed patients. In this multi-centre study based on data prospectively collected by RRTs, the NEWS 2 scores of all patients were retrospectively, digitally calculated by the study team. Age was analysed as a continuous variable, in a spline regression model, and categorized into five different models, subsequently explored as additive variables to NEWS 2. The discriminative ability of NEWS 2 in predicting 30-day mortality improved by adding age as a covariate (from AUROC 0.66, 0.62-0.70 to 0.70, 0.65-0.73, p=0.01). There were differences across age groups, with the best predictive performance identified among patients aged 45-54 years. The 30-, 90-and 180-day mortality was 31%, 33%, and 36%, respectively.

Conclusion: ECG monitoring at the time of IHCA was associated with a 38% reduction of adjusted mortality. Despite this finding, only one in two IHCA patients were ECG monitored. The most important factor influencing ECG monitoring was which type of hospital ward the patient was admitted to. The tangible variations in the frequency of ECG monitoring adjacent to IHCA observed between Swedish regions and across hospitals need to be investigated in future studies. Guidelines for the monitoring of patients at risk of CA could contribute to an improved outcome.

The prognostic accuracy of NEWS 2 in predicting mortality within 24 hours of an RRT-review was acceptable, whereas the discriminative ability in prediction of unanticipated ICU-admission and the composite outcome was rather weak. Regarding the prediction of IHCA, NEWS 2 performed poorly. There was no difference in the prognostic accuracy between NEWS and NEWS 2; however, the discriminative ability was not considered sufficient to serve as a triage tool in RRT-reviewed patients.

In-hospital cardiac arrest among patients who were reviewed by an RRT prior to CA was associated with a poorer prognosis, and a more frequent respiratory aetiology of the CA. In the explorative sub-group of patients, RRT activation was frequently delayed, the most common trigger for RRT-review was respiratory distress, and escalation of the level of care was associated with an improved prognosis. Early identification of patients with abnormal respiratory vital signs, followed by a timely response, may have a potential to improve the prognosis for patients reviewed by an RRT and prevent IHCA.

Adding age as a covariate improved the discriminative ability of NEWS 2 in the prediction of 30-day mortality among RRT-reviewed patients. The ability differed across age categories. Overall, the long-term prognosis of RRT-reviewed patients was poor. Our results indicate that age merits further validation as a covariate to improve the performance of NEWS 2.