a b s t r a c t

Objectives: In-hospital cardiac arrests (IHCAs) are often preceded by abnormal vital signs. Preceding abnormal vital signs might lower the physiological reserve capacity and therefore decrease survival after an IHCA.

Aim: To assess the preceding National Early Warning Score and its relation to survival after an IHCA. Material and methods: All patients >=18 years suffering an IHCA at Karolinska University Hospital between 1st Jan- uary 2014 and 31st December 2015 were included. Data regarding the IHCA, patient characteristics, calculated NEWS and 30-day survival were obtained from electronic patient records. Parameters included in NEWSs were assessed up to 12 h before the IHCA. Differences in survival were assessed with adjusted logistic regression models and presented as Odds Ratios with 95% Confidence Intervals (OR, 95% CI) between patients with NEWSs of 0-4 points (low) versus those with at least 5 points (moderate) and 7 points (high). Adjustments included hospital site, sex, co-morbidities, first rhythm and location of the IHCA.

Results: In all, 358 patients suffered an IHCA, of whom 109 (30%) survived at least 30 days and 296 (83%) had suffi- cient vital sign documentation to calculate NEWS before the IHCA. The 87 patients witha medium NEWS hada four- fold chance and those 78 with a high NEWS (22%) had an almost tenfold chance of dying after the IHCA compared to those with a low NEWS (Adjusted OR 4.43, 95% CI 1.81-10.83 and OR 9.88 95% C.I. 2.77-35.26, respectively).

Conclusion: The NEWS can be a probable proxy for estimating physiological reserve capacity since high NEWS is asso- ciated to high change of death in case of an IHCA. This information can be used when discussing prognosis with pa- tients and relatives. But even more importantly, it stresses the need for better Preventive strategies in IHCAs.

Strengthens and limitations with this study:

• Strengths include the complete information about the outcome, i.e. 30-day survival or not due to Swedish Personal Identification Numbers and registries.
• Another strength is the collection of exposure data, NEWSs, was based on a strict a priori determined protocol,

including only data documented in the medical record and therefore known to the staff before the IHCA.

• Limitations include the use of only documented vital signs, i.e. those taken but not documented, as well as the fact that information on interventions performed but not documented in the medical file are missing.
• A potential limitation is the lack of a control group with equally abnormal vital signs but no IHCA; such infor-

mation would have answered the question how predictive the NEWS is for an IHCA but this is out of the scope of this study regardless.

(C) 2017

Introduction

About one third of all cardiac arrests occur in hospitals [1]. Popula- tion-based registries show that for in-hospital cardiac arrests (IHCAs) several links in the Chain of survival are already optimised, i.e. the

* Corresponding author at: Department of Medicine, Solna, Karolinska Institutet, Stockholm, Sweden.

E-mail address: [email protected] (T. Djarv).

majority of IHCAs occur under ECG-surveillance, are witnessed, receive CPR within 1 min and are defibrillated within 3 min [1,2]. Still, survival rates are relatively low and mostly stable over recent decades [3]. The first link in the chain of survival includes prevention which is stressed in the recent European Resuscitation Council (ERC) guidelines [4]. Pre- vention of IHCA includes reducing the population at risk by, for exam- ple, influencing end-of-life decisions such as DNAR/DNACPR-orders as well as identifying high risk patients since IHCAs are rarely sudden, yet instead often occur with preceding abnormal vital signs [5-9]. In

http://dx.doi.org/10.1016/j.ajem.2017.04.072

0735-6757/(C) 2017

order to identify abnormal vital signs some kind of early warning score (EWS) is recommended [4]. In total over 100 EWSs exist and in 2012 the Royal College of Physicians in London, UK, developed a national strategy for standardisation of EWSs; the National Early Warning Score [10]. The NEWS has, during its development as well as afterwards, been shown to be better than other EWSs at discriminating risk of acute ill- ness [11-13]. The NEWS aims to be prognostic by identifying patients at risk of adverse outcomes, but since physiological reserve capacity is important for survival after an IHCA, an elevated NEWSs might even be predictive, i.e. associated with response or lack of response to resus- citation treatment [14-16].Therefore, we conducted a cohort study with the aim of assessing the prevalence of the NEWS before an IHCA as well as its association with 30-day survival after an IHCA.

Method

Study design and settings

This cohort study took place at Karolinska University Hospital be- tween 1st January 2014 and 31st December 2015. Karolinska Univer- sity Hospital is one of five large hospitals in Stockholm city, home to approximately 2,000,000 people. Karolinska University Hospital has two sites, Solna and Huddinge, which are 30 km apart. The Solna site hosts about 750 beds and is a level one trauma unit, has neuro and thoracic surgery units and provides 24/7 angiography for ST-eleva- tion myocardial infarctions. The Huddinge site also hosts about 750 beds and includes a geriatric ward and relatively fewer intensive care unit (ICU) beds. Karolinska University Hospital as a whole has about 108,000 admissions yearly and 1.8 million patient visits. Karolinska University Hospital has a mature medical emergency team (MET) using single calling criteria since 2005 [17], i.e. the hospital does not use NEWS.

Ethics

This study used a database and registry; by accepting care at Karolinska University Hospital, patients are informed about on- going studies and where they can find information including with- drawal of participation. Further, all patients surviving their IHCA were asked for informed consent and agreed to participate in the registry and on-going studies based on it. The Regional Ethical Review Board in Stockholm, Sweden approved the study, Dnr 2013/ 1959-31/4.

Participants

All cases of IHCA among adults, i.e. patients aged at least 18 years, oc- curring between January 1st 2014 and December 31th 2015 were eligi- ble for inclusion in the study and identified through the hospital’s cardiac arrest report sheet. The Karolinska University Hospital partici- pates in the Swedish Cardiac Arrest Registry (SCAR) [1,18] and collect data according to Utstein and SCARs definition of IHCA was used in the current study, i.e. a hospitalised patient who is unresponsive with apnoea (or agonal, gasping respiration) where CPR and/or defibrillation have been initiated. No patients or location of the IHCA were excluded. In the case of multiple IHCAs within the same admission, only the first event was included.

National Early Warning Score

The concept of the NEWS was released by the Royal College of Phy- sicians, London, UK in 2012 in order to standardise the assessment of acute-illness severity and for surveillance through both tracking the clinical condition over time as well as triggering a timely clinical re- sponse [10]. The NEWS is a simple scoring system for seven vital signs (respiration rate, oxygen saturation, supplementary oxygen, heart

rate, systolic blood pressure, temperature and level of consciousness). The magnitude of the score (range 0-3 points) reflects how extremely the parameter varies from the norm. A score of 3 points on a parameter is also called a RED score implying an extreme variation from the norm. The score on each parameter is aggregated to a total sum ranging be- tween 0 and 20 points. A total sum of 0-4 is categorised as a low score which should trigger a response by a nurse, a sum of at least 5 or RED score is categorised as medium score and should trigger an urgent re- view by a skilled clinician and at least 7 points is categorised as a high score prompting at least an emergency assessment by a clinical team [10]. The ERC guidelines for ACLS 2015 clearly stress the need for a track and trigger system in order to prevent IHCA [4].

Data collection and categorization

Patients were identified through the hospital’s cardiac arrest report sheet, where data on the following variables were collected: sex, age (collected in years, categorised in 10-year intervals according to the age-adjusted Charlson co-morbidity index (ACCI) [19] starting at 18- 40, 41-50 and further on to N 81 years), location of IHCA (patient ward, intermediate care unit, intensive care unit (ICU), angio lab/oper- ation theatre or other area including emergency department and radiol- ogy department), first documented heart rhythm (VT/VF or PEA/ asystole). Thereafter by entering the hospital’s electronic patient record (Take Care version 14.2.9) information on co-morbidities was gathered based on ICD-10 codes available at least at admission to the hospital and assessed according to the ACCI [20,21] and categorised into “Minimal burden of age-combined co-morbidities” if the ACCI was 0-2 points, “Low burden of age-combined co-morbidities” if the ACCI was 3-5 points, “Moderate burden of age-combined co-morbidities” if the ACCI was 6-7 points or “Severe burden of age-combined co-morbidities” if receiving at least 8 points [19,22,23]. Information to calculate NEWSs was gathered from the electronic medical record up to 12 h before the IHCA, in the case of multiple sets, the one closest in time to the IHCA was used. The NEWS was calculated as an aggregated score if at least one documented parameter was RED or if at least 4 out of 7 parameters were document- ed, otherwise the NEWS was categorised as missing.

Information on the outcome, i.e. 30-day survival (yes or no) was re-

trieved through the electronic patient record which is linked to the Swedish total population registry and automatically updated within a maximum delay of three days, which enables a complete follow-up [24]. Information on the Cerebral Performance Category score at discharge from the hospital was gathered from the SCAR and categorised as good (1-2) or poor (3-5) [25].

Statistical analyses

Characteristics of patients surviving at least 30 days and those who were deceased were compared using the two-sided Chi² test and a p- value of <= 0.05 was interpreted as statistically significant. For each pa- rameter included in the NEWS, both the prevalence in percentage and the ratio of decease were presented. Missing data was kept missing,

i.e. not imputed or estimated. In order to assess a quantitative, i.e. a de- scriptive dose-response, relationship mortality ratios for single versus a combination of two, three or more parameters were counted. Logistic regression models were used to estimate the association between cate- gory of aggregated calculated NEWSs (low versus medium or high) and 30-day survival, expressed as odds ratios (ORs) with 95% confidence in- tervals (95% CIs) with adjustment a priori decided and for known con- founders: 1) sex, 2) ACCI, 3) first documented heart rhythm, 4) location of IHCA, and 5) hospital site. Further, as a sensitivity analysis OR with 95% CI was calculated with NEWS as a continuous variable. All analyses were performed with the statistical package STATA 10.2 for Windows (STATA Corp, College Station, TX).

Results

Study participants

In all, 358 patients suffered an IHCA at Karolinska UH Solna or Hud- dinge during 2014-2015 and the overall 30-day survival rate was 30%. There was no difference between sexes regarding survival (Table 1). Two thirds of survivors were aged below 70 years while two thirds of the deceased patients were aged above 71 years (p-value b 0.01). Re- garding co-morbidities, three quarters of the survivors had a minimal or low burden of ACCI while two thirds of deceased patients had a mod- erate or severe burden of ACCI prior to the IHCA (Table 1, p-value b 0.01).

In terms of the location of IHCA, the largest portion of deceased pa- tients had suffered their IHCA in a patient ward while survivors were more scattered. Regarding the first documented heart rhythm, survivors had shockable rhythm, i.e. VT/VF, equally often as they had non-shock- able rhythm, i.e. PEA/asystole, compared to the presence of a shockable rhythm in only 11% of deceased patients (P-value b 0.01). Among the 109 survivors, only 3 (4%) patients had a poor CPC score at discharge.

Prevalence of NEWSs — single parameters

Missing data per parameter was: 44% for respiration rate, 20% for ox- ygen saturation, 21% for any supplementary oxygen, 16% for heart rate, 14% for systolic blood pressure, 20% for temperature and 10% for level of consciousness (Table 2).

The prevalence of normal (0 points) or almost normal (1 point) pa- rameters was as follows: respiration rate 28%, oxygen saturation 48%, supplementary oxygen 46%, systolic blood pressure 66%, heart rate 67%, temperature 86% and level of consciousness 89% (Table 2). The prevalence of parameters with extreme variation, i.e. a RED score, was

Table 1

Characteristics of 358 patients suffering an in-hospital cardiac arrest during 2014-2015 in Karolinska University Hospital.

most commonly derived from Low Oxygen Saturation (prevalence 24%), high respiration rate (16%), low systolic blood pressure (9%) and high heart rate (6%). Mortality ratios over 85% were found for respira- tion rates >= 21, oxygen saturation <= 91% as well as for systolic blood pres- sure <= 90 (Table 2).

Prevalence of NEWS—aggregated score

Regarding the calculated NEWS before the IHCA, in total 168 (47%) had a complete calculated NEWS and 128 (36%) has at least 4 out of 7 parameters documented <= 12 h before their IHCA and therefore included in further analysis. The NEWS was low in a third of the patients, medium in a quarter and high in every fifth patient (Table 3). Survival rates were 47% among those with low NEWS, 20% among those with medium NEWS and 10% among those with high NEWS.

Subanalysis of those with a complete NEWS showed low NEWS among 69 (41%), medium in 46 (27%) and high in 53 (32%) and survival rates for each category was identical to the total group (47%, 20% and 9% respectively) (data not shown). Combinations with >= 90% mortality ratio, were found for patients with at least 2 points on two of the follow- ing parameters in any combination: high respiratory rate, oxygen satu- ration b 90%, high heart rate and low systolic blood pressure (Fig. 1). Further, a >= 2-point score on at least three of the four above mentioned parameters formed a lethal triad with 100% mortality (Fig. 1).

Association between NEWS and survival

Patients with a medium NEWS had a more than fourfold chance of dying in their IHCA compared to those with a low NEWS (crude OR 3.59, 95% CI 1.90-6.75, as well as adjusted OR 4.43, 95% CI 1.81-10.83,

Table 3). Likewise, patients with a high NEWS had tenfold chance of dying of their IHCA compared to those with a low NEWS (crude OR 7.62, 95% CI 3.40-17.11, as well as adjusted 9.88, 95% CI 2.77-35.26,

Table 3). When NEWS was kept as a continuous variable each additional point came with an increased risk of dying (crude OR 1.37, 95% CI 1.24- 1.52, adjusted 1.43, 95% CI 1.23-1.68, data not shown).

Patients surviving at least 30 days Number (%)

109 (100)

Patients deceased Number (%)

249 (100)

Sex			0.68
Male	64 (59)	152 (61)
Female	45 (41)	97 (39)
Age category			b 0.01
18-40	5 (5)	7 (3)
41-50	16 (15)	9 (4)
51-60	17 (16)	17 (7)
61-70	34 (31)	52 (21)
71-80	25 (23)	86 (34)
>= 81 Age-adjusted Charlson	12 (11)	78 (31)	b 0.01
co-morbidity index
0-2 points	22 (20)	15 (6)
3-5 points	61 (56)	77 (31)
6-7 points	15 (14)	93 (38)
>= 8 points	11 (10)	64 (26)
Place of cardiac arrest			b 0.01
Patient ward	20 (18)	102 (41)
Intermediate care unit	32 (29)	75 (30)
Intensive care unit	13 (12)	26 (10)

P-value^a

Discussion

This cohort study illuminates a high prevalence of preceding abnor- mal vital signs as well as their highly negative association to survival after IHCA. This finding is not unexpected but of utmost clinical impor- tance since it stresses the first link in the chain of survival, namely pre- vention of IHCA. Prevention might be both to resuscitate vital signs more aggressively and to take a Do Not Attempt Resuscitation (DNAR) decision.

In line with previous research [11,13,26-29] we found a quantitative,

i.e. dose-response, pattern between a higher NEWS and a poorer out- come. Further, a small number of patients identified with lethal triads with 100% mortality ratio, i.e. at least 2 points on any three of the follow- ing parameters: respiration rate, oxygen saturation, heart rate and sys- tolic blood pressure. In contrast to previous studies [28,29] our hospital has a mature MET. Reasons for the findings of such deranged vital signs in hospitals with mature MET [17] might be explained by the fact that MET is a single criteria system coming in when the patients has a RED score while NEWS with its aggregated scores comes in already from low scores and have been shown to be more sensitive [11].

Procedure room incl angio lab./Operating rooms Others incl. Emergency

department, x-ray department

24 (22) 17 (7)

20 (18) 29 (12)

Some of the deaths from IHCA might not be preventable, but one im- portant task for the clinicians facing a patient with a high NEWS is to trigger the organisation to urgent assessment of acute illness including

considering transfer to a higher dependency care area. However, it is

First documented heart rhythm b 0.01

VT/VF 43 (40) 28 (11)

PEA/asystoli 45 (41) 186 (76)

Missing 21 (19) 35 (14)

^a P-values were assessed with Chi2-test regarding differences between 30-day survivors and deceased patients.

important to bear in mind that along the path from preceding abnormal vital signs to discharge, several important Clinical decisions have been made. So, even if we cannot say for sure that correcting abnormal vital signs would either prevent or improve outcomes, the result of this study stresses the question whether more aggressive prevention

Table 2

Prevalence of documented vital signs to calculate National Early Warning Scores (NEWSs) [10] among 358 patients suffering an in-hospital cardiac arrest at Karolinska University Hospital during 2014-2015.

	3p RED	2p	1p	0p	1p	2p	3p RED	Missing
Respiration rate	<= 8		9-11	12-20		21-24	>= 25
Prevalencea (%)	0%		6%	22%		12%	16%	44%
Portion deceased (%) Oxygen saturation	- <= 91	92-93	50% 94-95	52% >= 96		86%	86%	70%
Prevalencea (%)	24%	8%	11%	37%				20%
Portion deceased (%)	89%	89%	78%	56%				59%
Supplementary oxygen Prevalencea (%)		Yes 32%		No 46%				21%
Portion deceased (%) Systolic blood pressure	<= 90	84% 91-100	101-110	63% 111-219			>= 220	61%
Prevalencea (%)	9%	10%	11%	55%			0%	14%
Portion deceased (%) Heart rate	85% <= 40	84%	73% 41-50	64% 51-90	91-110	111-130	- >= 131	66%
Prevalencea (%)	1%		2%	51%	15%	8%	6%	16%
Portion deceased (%)	50%		29%	69%	79%	83%	73%	59%
Temperature	<= 35		35.1-36	36.1-38	38.1-39	>= 39.1
Prevalencea (%)	<= 1%		8%	62%	16%	1%		20%
Portion deceased (%)	50%		79%	69%	80%	80%		62%
Level of consciousness Prevalencea (%)				Alert 89%			Verbal, pain, unresponsive 2%	10%
Portion deceased (%)				70%			66%	65%

^a Prevalence = number of IHCAs with score for parameter/total number of IHCAs during 2014-2015.

including transfer should be carried out before, rather than around the time of resuscitation. If such a transfer is not justified for good reasons, one has to remember another important task of a clinician facing a high NEWS, namely the task to take end-of-life decisions such as DNAR and to allow natural death when CPR does not seem to in- crease the overall survival.

A major backdraw of the current study is the use of documented vital signs to calculated NEWS instead of a full set of NEWS taken and assessed according to the concept of NEWS, we cannot assume that missing data is normal, so the NEWS in our data could be seen as false low.

The high amount of missing data has two main explanations; firstly nearly half of all IHCAs occurred in intermediate or intensive care units where patients are continuously monitored for some vitals but which are not necessarily documented in the electronic patient record. Sec- ondly, it probably well reflects clinical practice in a hospital using calling criteria and MET rather than a standardised minimal set-up for vital signs such as NEWSs [10].

Strengths include inclusion of two separate different hospital-sites over two years. Further, collection of exposure data, NEWSs, was based on a strict a priori determined protocol, including only data doc- umented in the medical record and therefore known to the staff before the IHCA. Another strength is the complete information about the out- come, i.e. 30-day survival or not due to Swedish Personal Identification Numbers and registries [24].

Limitations include the use of only documented vital signs, i.e. those taken but not documented, as well as the fact that information on

interventions performed but not documented in the medical file are missing. Likewise, the limit of 12 h before the IHCA might seem long in the perspective of most patients having deranged vital signs, howev- er, this was not a interventional study but a study reflecting clinical practice. A potential limitation is the lack of a control group with equally abnormal vital signs but no IHCA; such information would have an- swered the question how predictive the NEWS is for an IHCA but this is out of the scope of this study regardless. However, even if such a con- trol group would indicate that abnormal vital signs are common in hospitalised patients, patients included in this study suffered an IHCA and their high NEWS was a bad prognostic sign that regardless of whether the hospital uses EWS or MET, ideally it would have led to staff concern and possible Escalation in care. Further limitations include residual confounding such as the absence of data relating to socio-eco- nomic status.

The results are likely generalisable since physiology across the world seems similar and as part of the intern validation of our results we per- formed separate analysis for the both included hospitals with similar re- sults (data not shown).

In conclusion, the main finding of a clearly increased chance of dying in their an IHCA among patients with a medium or high NEWS implies that the NEWS can be a valuable tool and a probable proxy for estimat- ing physiological reserve capacity when discussing prognostic and pre- dictive factors in the case of an IHCA with patients and relatives. But even more importantly, it stresses the need for focus on strategies more than single parameter calling criteria’s aimed at the prevention part of chain of survival, as well as on strategies for increasing survival

Table 3

Association between National Early Warning Scores (NEWSs) up to 12 h before an in-hospital cardiac arrest and 30-day survival at Karolinska University Hospital during 2014-2015.

	All patients Number (%) 358 (100)	Patients surviving at least 30 days Number (%a) 109 (30)	Crude OR (95% CI)	Adjusted OR (95% CI)b
NEWSc
Low, 0-4p	131 (37)	61 (47)	1.00 (Reference)	1.00 (Reference)
Medium, >= 5 or one RED parameter	87 (24)	17 (20)	3.59 (1.91-6.75)	4.43 (1.82-10.83)
High, >= 7	78 (22)	8 (10)	7.63 (3.40-17.11)	9.88 (2.77-35.26)

^a Percentage = number of survivors in NEWS category/number of patients in NEWS category.

^b Adjusted for sex, ACCI, hospital site, first documented heart rhythm and location of cardiac arrest.

^c NEWS was calculated as an aggregated score if at least one RED parameter or if at least 4 out of 7 parameters were documented.

Fig. 1. Combinations of parameters in NEWSs with high mortality ratios (n% +) among 358 in-hospital cardiac arrests at Karolinska University Hospital 2014-2015.

after an IHCA. We believe that routinely and more frequently checking vital signs is one important strategy in order to prevent IHCA, which is an important task for hospitals to do. Further the study highlights the opportunity for improving policy, legislation, metrics, accountability, and systems of care to result in better survival among IHCA.

Competing interest declarations

None for any author.

Conflict of interest

None for any author.

Contributorship

All authors of this manuscript have directly participated in the plan- ning, execution, and analyses of the study. All authors have read and ap- proved the final version of the submitted manuscript. There are no directly related manuscripts or abstracts, published or unpublished, by any of the authors of this paper.

Detailed author contribution DR TD

Study concept and design X X

Acquisition of data X X

Analysis and interpretation of data X X

Drafting of the manuscript X X

Critical revision of the manuscript for important intellectual content X X Statistical analysis X X

Obtained funding X

Administrative, technical, or material support X

Study supervision X

Data sharing statement

No additional data exists that is suitable for publication since data are based on medical records of individuals.

Acknowledgement

The study was supported by the Karolinska Institutet. TD was supported by the Stockholm County Council (clinical research appointment).

References

1. Herlitz J. Swedish CardioPulmonary Register — yearly report; 2014 2015In Swedish.
2. Shao F, Li CS, Liang LR, et al. Incidence and outcome of adult in-hospital cardiac arrest in Beijing, China. Resuscitation 2016;102:51-6 (published Online First: Epub Date) 10.1016/j.resuscitation.2016.02.002.
3. Nolan JP, Hazinski MF, Aickin R, et al. Part 1: executive summary: 2015 International Consensus on Cardiopulmonary Resuscitation and Emergency Cardiovascular Care Science with Treatment Recommendations. Resuscitation 2015;95:e1-31 (pub- lished Online First: Epub Date) 10.1016/j.resuscitation.2015.07.039.
4. Soar J, Nolan JP, Bottiger BW, et al. European Resuscitation Council Guidelines for Re- suscitation 2015: Section 3. Adult advanced life support. Resuscitation 2015;95: 100-47 (published Online First: Epub Date) 10.1016/j.resuscitation.2015.07.016.
5. Bossaert LL, Perkins GD, Askitopoulou H, et al. European Resuscitation Council Guidelines for Resuscitation 2015: Section 11. The ethics of resuscitation and end- of-life decisions. Resuscitation 2015;95:302-11 (published Online First: Epub Date) 10.1016/j.resuscitation.2015.07.033.
6. Soar J, Deakin CD, Nolan JP, et al. European Resuscitation Council guidelines for re- suscitation 2005. Section 7. Cardiac arrest in special circumstances. Resuscitation 2005;67(Suppl. 1):S135-70 (published Online First: Epub Date) 10.1016/j.resuscita- tion.2005.10.004.
7. Hodgetts TJ, Kenward G, Vlackonikolis I, et al. Incidence, location and reasons for avoidable in-hospital cardiac arrest in a district general hospital. Resuscitation 2002;54(2):115-23.
8. Churpek MM, Yuen TC, Edelson DP. Risk stratification of hospitalized patients on the wards. Chest 2013;143(6):1758-65. http://dx.doi.org/10.1378/chest.12-1605 (pub- lished Online First: Epub Date).
9. Ebell MH, Jang W, Shen Y, et al. Development and validation of the Good Outcome Following Attempted Resuscitation (GO-FAR) score to predict Neurologically intact survival after in-hospital cardiopulmonary resuscitation. JAMA Intern Med 2013; 173(20):1872-8. http://dx.doi.org/10.1001/jamainternmed.2013.10037 (published Online First: Epub Date).
10. Royal College of Physicians. National Early Warning Score : standardising the assessment of acute illness severity in the NHS. Report of a working party Lon- don: RCP; 2012.
11. Jarvis S, Kovacs C, Briggs J, et al. Aggregate National Early Warning Score (NEWS) values are more important than high scores for a single vital signs parameter for dis- criminating the risk of adverse outcomes. Resuscitation 2015;87:75-80. http://dx. doi.org/10.1016/j.resuscitation.2014.11.014 (published Online First: Epub Date).
12. Abbott TE, Vaid N, Ip D, et al. A single-centre observational cohort study of admission National Early Warning Score (NEWS). Resuscitation 2015;92:89-93. http://dx.doi. org/10.1016/j.resuscitation.2015.04.020 (published Online First: Epub Date).
13. Smith GB, Prytherch DR, Meredith P, et al. The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated

    intensive care unit admission, and death. Resuscitation 2013;84(4):465-70. http:// dx.doi.org/10.1016/j.resuscitation.2012.12.016 (published Online First: Epub Date).

    Clark GM. Prognostic factors versus predictive factors: examples from a clinical trial of erlotinib. Mol Oncol 2008;1(4):406-12. http://dx.doi.org/10.1016/j.molonc.2007.

    12.001 (published Online First: Epub Date).

    Italiano A. Prognostic or predictive? It’s time to get back to definitions. J Clin Oncol Off J Am Soc Clin Oncol 2011;29(35):4718 author reply 18-9 10.1200/JCO.2011.38. 3729. (published Online First: Epub Date).

14. Wang AY, Fang CC, Chen SC, et al. Periarrest modified early warning score predicts the outcome of in-hospital cardiac arrest. J Formos Med Assoc 2016;115(2): 76-82. http://dx.doi.org/10.1016/j.jfma.2015.10.016 (published Online First: Epub Date).
15. Konrad D, Jaderling G, Bell M, et al. Reducing in-hospital cardiac arrests and hospital mortality by introducing a medical emergency team. Intensive Care Med 2010; 36(1):100-6. http://dx.doi.org/10.1007/s00134-009-1634-x (published Online First: Epub Date).
16. Stromsoe A, Svensson L, Axelsson AB, et al. Validity of reported data in the Swedish Cardiac Arrest Register in selected parts in Sweden. Resuscitation 2013;84(7): 952-6. http://dx.doi.org/10.1016/j.resuscitation.2012.12.026 (published Online First: Epub Date).
17. Piscator E, Hedberg P, Goransson K, et al. survival after in-hospital cardiac arrest is highly associated with the Age-combined Charlson Co-morbidity Index in a cohort study from a two-site Swedish University hospital. Resuscitation 2016;99:79-83. http://dx.doi.org/10.1016/j.resuscitation.2015.11.023 (published Online First: Epub Date).
18. Charlson M, Szatrowski TP, Peterson J, et al. Validation of a combined comorbidity index. J Clin Epidemiol 1994;47(11):1245-51.
19. Sundararajan V, Henderson T, Perry C, et al. New ICD-10 version of the Charlson co- morbidity index predicted in-hospital mortality. J Clin Epidemiol 2004;57(12): 1288-94. http://dx.doi.org/10.1016/j.jclinepi.2004.03.012 (published Online First: Epub Date).
20. Terman SW, Shields TA, Hume B, et al. The influence of age and chronic medical con- ditions on neurological outcomes in out of hospital cardiac arrest. Resuscitation 2015;89:169-76. http://dx.doi.org/10.1016/j.resuscitation.2015.01.006 (published Online First: Epub Date).
21. Lee CC, Tsai MS, Fang CC, et al. Effects of pre-arrest comorbidities on 90-day survival of patients resuscitated from out-of-hospital cardiac arrest. Emerg Med J 2011; 28(5):432-6. http://dx.doi.org/10.1136/emj.2009.087593 (published Online First: Epub Date).
22. Ludvigsson JF, Otterblad-Olausson P, Pettersson BU, et al. The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research. Eur J Epidemiol 2009;24(11):659-67. http://dx.doi.org/10.1007/s10654-009-9350-y (published Online First: Epub Date).
23. Herlitz. Swedish Cardio-Pulmonary Registry — yearly report; 2014 2014(In Swedish).
24. Bell MB, Konrad D, Granath F, et al. Prevalence and sensitivity of MET-criteria in a Scandinavian University Hospital. Resuscitation 2006;70(1):66-73. http://dx.doi. org/10.1016/j.resuscitation.2005.11.011 (published Online First: Epub Date).
25. Andersen LW, Kim WY, Chase M, et al. The prevalence and significance of abnormal vital signs prior to in-hospital cardiac arrest. Resuscitation 2016;98:112-7. http://dx. doi.org/10.1016/j.resuscitation.2015.08.016 (published Online First: Epub Date).
26. Skrifvars MB, Boyd J, Kuisma M. Prearrest signs of shock and respiratory insufficien- cy in out-of-hospital cardiac arrests witnessed by crew of the emergency medical service. Am J Emerg Med 2009;27(4):440-8. http://dx.doi.org/10.1016/j.ajem. 2008.03.027 (published Online First: Epub Date).
27. Skrifvars MB, Nurmi J, Ikola K, et al. Reduced survival following resuscitation in pa- tients with documented clinically abnormal observations prior to in-hospital cardiac arrest. Resuscitation 2006;70(2):215-22. http://dx.doi.org/10.1016/j.resuscitation. 2006.01.002 (published Online First: Epub Date).