Journal of Critical Care (2012) 27, 688–693

Adverse outcomes associated with delayed intensive careconsultation in medical and surgical inpatientsLouay Mardini MDa, Jed Lipes MD, FRCPCb, Dev Jayaraman MD, MPh, FRCPC a,b,⁎

aMontreal General Hospital, McGill University Health Centre, Montreal, Canada, H3G 1A4bSir Mortimer B. Davis, Jewish General Hospital, Montreal, Canada, H3T 1E2

C

0h

Keywords:Delay;Transfer;Mortality;Ward;Intensive care

AbstractPurpose: The impact of delay in obtaining an intensive care unit (ICU) consult from inpatient wards isunclear. The goal of this study was to examine the effect of time to ICU consult from medical andsurgical wards on mortality and length of stay (LOS).Materials and Methods: This was a retrospective study of 241 adult medical and surgical inpatientsadmitted at 2 tertiary care ICUs in Canada between 2007 and 2009. Neither institution has medicalemergency teams (METs). Patient demographics, time when the patient would have fulfilled METcalling criteria (MET time), time of ICU consult, and ICU admission were analyzed. The main outcomevariables were 30-day mortality and ICU LOS.Results: Multivariate analysis demonstrated an increase in mortality (odds ratio, 1.8; 95% confidenceinterval, 1.1-2.9; P = .01) with increased duration from MET time to ICU consult for medical patients.There was no effect of this period on ICU LOS in medical patients. In contrast, in surgical patients, theMET time to ICU consult duration was associated with an increased ICU LOS (coefficient, 2.1 fordelay; 95% confidence interval, 0.26-3.8; P = .02) but had no effect on mortality.Conclusions: Increased duration to ICU consult from MET time is associated with adverse outcomes.These adverse outcomes are different between medical and surgical patients.© 2012 Elsevier Inc. All rights reserved.

1. Introduction

Delay in appropriate delivery of medical care has beenshown to have a negative impact in a variety of medical andsurgical diseases. From the initial “golden hour” of trauma,to door-to-balloon time in acute myocardial infarction andearly goal-directed therapy in septic shock, all have more

⁎ Corresponding author. Montreal General Hospital, Montreal, Quebec,anada H3G 1A4.E-mail address: Dev.jayaraman@mcgill.ca (D. Jayaraman).

883-9441/$ – see front matter © 2012 Elsevier Inc. All rights reserved.ttp://dx.doi.org/10.1016/j.jcrc.2012.04.011

favorable outcomes if treated in a timely fashion [1-3]. Thecritically ill patient represents an even greater challengebecause a number of interventions, both medical andprocedural, involving multiple personnel all have to occurrapidly to ensure optimal care [1,4,5]. Moreover, studieshave shown that delays in patient transfers of the critically illfrom the emergency department to intensive care have beenassociated with worse outcomes, although the results havebeen mixed [6-8].

There is limited data analyzing intensive care unit (ICU)admission delays from medical or surgical inpatient wardsand mortality. Cardoso et al [9], in a single-center study from

689Adverse outcomes of delayed intensive care consultation

Brazil, have demonstrated that every hour delay in admissionto the ICU from the time of decision to admit to the ICUincreased the risk of ICU mortality by 1.5%. However, themedian delay in admission in this study was 17.8 hours.Young et al [10] demonstrated that a greater-than-4-hourdelay in admission from the wards in a community hospitalresulted in an increased mortality and morbidity.

Indirect data from the literature surrounding medicalemergency teams (METs) also suggest that earlier access tocritical care either in the ICU or via METs may improveoutcomes, although the benefits of METs are not univer-sally recognized [5,11-14]. In fact, most hospitals in theprovince of Quebec, including the 2 centers in this study,do not have METs. Within this context, we sought toaddress 2 issues in an academic tertiary care environment.Our first goal was to investigate the hypothesis that in-creased time to ICU consultation was more likely asso-ciated with increased 30-day mortality and ICU length ofstay (LOS) than the time to ICU admission after consul-tation. Our second aim was to explore if there were differ-ences between patients admitted from surgical wards vsmedical wards in terms of delays as well as outcomes.

2. Materials and methods

This study is a structured retrospective chart review at2 tertiary university centers in Montreal. Both centers have20-bed closed medical-surgical ICUs and do not haveMETs, and all consultations were performed by the on-service ICU team. House-staff or attending physicians onwards obtain an ICU consult by calling the ICU resident oncall, and the consults are seen within 30 minutes of requestand reviewed with an intensivist or critical care fellow soonafter. The medical and surgical wards in the 2 hospitals donot have bedside monitors or high-dependency units.Nursing ratios and frequency of routine vital signs moni-toring were also similar between the wards. Patients aregenerally admitted to the medical wards from the emer-gency department with a small number transferred fromother hospital units. Patients are admitted to the surgicalwards postoperatively or through the emergency depart-ment for monitoring for acute surgical issues or awaitingsurgery. A small number of patients are transferred fromother hospital units to the surgical wards.

All ICU admissions from medical and surgical wardsduring a 12-month period were screened at both centers.Because of administrative limitations—one of the hospitalswas switching to electronic charts—the time points weredifferent in the 2 hospitals (March 2008 to March 2009 athospital A and September 2007 to November 2008 athospital B). Screened patient charts were included if theywere admitted to the ICU after being admitted for morethan 48 hours on an inpatient ward. Patients were excludedif they were admitted 48 hours or less after admission from

the emergency department or 48 hours or less postopera-tively, as the time of deterioration may have preceded theward admission. Patients admitted to the ICU from thewards without having reached any MET criteria or weretransferred from another critical care area such as thecoronary care unit or postanesthesia recovery unit were alsoexcluded. The study was approved by ethics committees ofboth hospitals.

Time points were collected via a retrospective review ofmedical charts and ICU database. Medical emergency teamtime was defined as when the patient would have firsttriggered a MET response. As neither hospital had a MET,we obtained the data retrospectively via chart review andnoting the time of physiologic deterioration (defined asMET time) with a heart rate120 or more or 40 or less beatsper minute, respiratory rate 30 or more or 10 or less breathsper minute, systolic blood pressure 200 or more or 100 orless mm Hg, temperature 39.5°C or more or less than35.5°C or new altered mental status [15]. We defined 2periods: MET time to ICU consult and ICU consult to ICUadmission. The time of ICU consult and ICU admissionwas obtained from chart records. The MET time, consult,and admission were also stratified as day or night (daydefined between 8 AM and 6 PM). All chart reviews wereperformed by 2 of the authors (JL and LM). A randomsample of 10% of the charts were then independentlyexamined by 1 of the authors (DJ) to evaluate the reliabilityof our time point assessments for delay from concern toadmission time, which yielded a Lin correlation coefficientof 0.94 [16]. Modified early warning scores (MEWSs) werecalculated at MET time and at ICU consult time [17].Acute Physiology and Chronic Health Evaluation(APACHE) II scores at ICU admission were obtainedfrom our ICU database. Because delays in obtaining anICU consult or admission to the ICU may be related to thepresence of markers of poor prognosis, we also noted ifsuch features (defined as a do-not-resuscitate order, activecancer, moderate to severe dementia, advanced Parkinsondisease, nursing home residence before hospital admission,note of end-stage cardiac or lung disease in the chart) werepresent before ICU consult.

Thirty-day mortality and ICU LOS were ascertainedfrom the hospital database. Survival status of patientsdischarged from hospital before 30 days was confirmed bysubsequent presentation at outpatient clinics or readmis-sions to the hospital.

All significance testing was 2 tailed, and significancelevel was set at .05. Proportions or percentages, means andstandard deviations, or medians and interquartile range(IQR) were used where appropriate to describe clinical anddemographic characteristics. Student t test and Mann-Whitney U test were used for parametric and nonparametriccontinuous variables, respectively. Categorical variableswere compared using χ2 test or Fisher exact test whenappropriate. Correlations between nonparametric variableswere tested by calculating the Spearman correlation

690 L. Mardini et al.

coefficient. Age, sex of patient, MEWS and APACHE IIscores, presence of poor prognosis markers, MET time andICU consult time duration and ICU consult to admission timeduration, and whether the MET time or ICU consult oc-curred during day or night were all tested as predictors of30-day mortality or ICU LOS in univariate analyses. Pre-dictors with a P value of less than .3 were included in mul-tivariate models. A stepwise backward elimination processwas used, and all covariates with P N .10 were excluded fromthe final model. Goodness of fit was evaluated using theHosmer-Lemeshow goodness of fit test [18] and receiveroperating curves.

3. Results

Overall, 367 patients were screened (Fig. 1). Eighty-sixpatients did not fit the inclusion criteria: 48 patients weretransferred to the ICU within 48 hours after admission to theward and 38 patients did reach MET calling criteria beforeadmission to the ICU. Of the eligible 281 patients, 40patients (26 from medical wards and 14 from surgical wards)were excluded because of poor documentation of timeperiods precluding reliable analysis.

Two hundred forty-one patient admissions (144 medicalpatients and 97 surgical patients) were analyzed (Fig. 1). Inmedical patients, the MET time to ICU consult durationwas almost twice as long as that from ICU consult toadmission (P=.05). In addition, the MET time to consultduration was twice as long in medical patients when

Fig. 1 A total of 367 medical and surgical ICU admissions were screeinclusion criteria, and 40 had incomplete or missing data.

compared with surgical patients (Table). There were nodifferences between the 2 groups with regard to ICU con-sult to ICU admission duration. No differences were notedbetween patients from medical wards or surgical wards interms of age, sex, APACHE II scores, or MEWS at concernor consult time. There were also no statistically significantdifferences between the 2 groups in terms of the time (dayor night) of MET, consult, or admission. Only 5 patients(3 medical and 2 surgical) had do-not-resuscitate ordersbefore ICU admission.

The 30-day mortality was considerably higher for themedical patients compared with the surgical patients (42%[95% confidence interval or CI, 33%-50%] vs 28% [95%CI, 19%-37%], P = .02). This effect persisted afteradjusting for APACHE II. Intensive care unit mortalitywas also higher in medical patients, although this did notreach statistical significance (26% [95% CI, 19%-34%] vs17% [95% CI, 9.2%-24%], P=.08). Patients who weredischarged from hospital before 30 days were all confirmedto be alive at 30 days. All patients who died, with theexception of 1 surgical and 1 medical patient, had do-not-resuscitate orders written before their deaths during theircourse in the ICU or on the wards after ICU discharge.There were no differences in ICU LOS between the 2groups for ICU survivors.

3.1. Duration between time of MET and ICU consult

The age, MEWS at MET, the change in MEWS fromMET to consult time, day vs night, and sex of the patient

ned according to inclusion criteria. Eighty-six patients did not meet

ig. 2 A, Duration (divided in tertiles) between MET and ICUonsult and mortality in medical patients. Values are expressed asean mortality ± SEM. B, Duration (divided in tertiles) betweenET and ICU consult and mortality in surgical patients. Values arexpressed as mean mortality ± SEM.

Table Baseline characteristics, clinical measures, and outcomes between medical and surgical patients

n Surgical Medical P

97 144

Male (%) 63% 59% .38Age (y), mean (SD) 69.3 (16) 66.1 (16.4) .12MEWS at MET, median (IQR) 4 (3-5) 4 (2-5) .16MEWS at consult, median (IQR) 4 (3-5) 4 (3-6) .08APACHE II, mean ± SD 23.5 (7.7) 22.9 (8.8) .50Day/nightAt MET(% day) 48% 55% .30At consult (% day) 52% 54% .69At admission (% day) 42% 53% .09DurationsMET to consult (h), median (IQR) 1.6 (0.7-4.2) 3 (0.6-12.3) .05Consult to admission (h), median (IQR) 2 (1-4.2) 1.8 (1-3.3) .23MET to admission (h), median (IQR) 4.5 (2.8-11.5) 6 (2.7-17.5) .14ICU LOS (d), median (IQR) 3 (1-9) 3.7 (1.8-9.2) .19ICU mortality (%) 18.30% 26.20% .1830-day mortality (%) 27.90% 42.10% .02

691Adverse outcomes of delayed intensive care consultation

were not correlated with MET time to ICU consult timeduration in surgical or medical patients. There were nosignificant differences in the MET time to consult timebetween hospital sites in surgical or medical patients.

For medical patients, when divided into tertiles of dura-tion from MET time to ICU consult time, we found thatmortality increased significantly from 29% in the lowesttertile to 52% in the highest tertile (Fig. 2A). There were nosignificant differences in age, APACHE II scores, oradmission diagnoses between the tertiles. Multivariateanalysis yielded an odds ratio of 1.8 (95% CI, 1.1-2.9) formortality comparing the highest to the lowest tertile (P =.01). The multivariate model included admission APACHEII score, sex of the patient, and the highest and lowest METtime to consult time tertiles. The area under the curve ofthe receiver operating curve was 0.75. The multivariateanalysis was repeated with the duration from MET to ICUconsult time as a continuous variable, which confirmed anincreased mortality with each hour increase in delay fromMET time to ICU consult on a natural log scale (odds ratio,1.3; 95% CI, 1.1-1.8; P = .05). The model includedAPACHE II score, sex of the patient, and MET time toconsult time in hours. The area under the curve for thismodel was 0.74 for the receiver operating curve.

In contrast to medical patients, there was no increase inmortality with increased MET time to ICU consult time insurgical patients (Fig. 2B). Multivariate analyses modeled astertiles or as a continuous variable did not demonstrate anyrelationship between duration of time to ICU consult afterMET time and mortality in this group.

There was no significant effect of MET time to ICUconsult duration on ICU LOS of ICU survivors in medicalpatients (P = .70; Fig. 3A). In contrast, there was an increaseICU LOS for ICU survivors in the surgical patients (P = .02;Fig. 3B). As a continuous variable, this delay was the only

significant predictor on univariate analysis of ICU LOS insurgical patients (coefficient 2.1 for MET time to ICUconsult on a log scale; 95% CI, 0.26-3.8; P = .02).

FcmMe

Fig. 3 A, Duration between MET and ICU consult and ICU LOSin medical patients. Values are expressed as median LOS+1/2 IQR.B, Duration between MET and ICU consult and ICU LOS insurgical patients. Values are expressed as median LOS+1/2 IQR.

692 L. Mardini et al.

Multivariate analyses in medical or surgical patients did notalter the above findings.

3.2. Intensive care unit consult time to ICUadmission time duration

A higher MEWS at the time of ICU consult was asso-ciated with a shorter time to admission in medical patients(Spearman ρ = −0.27, P=.001) but not in surgical patients.The delay from ICU consult to admission was not associatedwith increased mortality or ICU LOS in medical or surgicalpatients in univariate or multivariate analyses.

4. Discussion

We studied the effect of increased time to ICU consultafter MET time on ICU LOS and 30-day mortality andinvestigated the differences between surgical and medicalpatients. We found that increased duration of this periodwas associated with increased 30-day mortality in medicalbut not in surgical patients. In contrast, ICU LOS wasincreased in surgical patients but not in medical patientswith increased duration in this period. Delay of longer than

6 hours from the emergency room to ICU admission hasbeen shown to be associated with an increased mortalityand ICU LOS [7]. In another study, a delay of longer than4 hours after deterioration to ICU admission from hospitalwards in a community hospital was associated with anincreased mortality [10]. However, these studies did notdifferentiate between delays in obtaining an ICU consult vsadmission to the ICU after an ICU consult. In addition, ourstudy was conducted in tertiary care university centers withthe continuous presence of house-staff and examineddifferences between medical and surgical wards.

Considering the modest sample size, it is possible thatthe reason we did not observe an effect of time to ICUconsult on mortality in the surgical patients was that theoverall mortality of patients transferred from surgical wardswas low compared with that of medical patients. Thisdifferential mortality between surgical and medical patientshas been noted before [22,23] and may reflect theunderlying fragility of patients admitted to medical floors.Similarly, it is possible that, with a larger sample size, wemay have been able to note an effect on ICU LOS withdelay to consult in the medical patient as well. Alterna-tively, because surgical patients seem less likely to die, theeffect of delay is seen with an increased ICU LOS.

In contrast to the MET time to ICU consult period, therewas no association with 30-day mortality or ICU LOS andtime from ICU consult to admission. This may imply thatthere was either a failure to recognize the deterioration and/orinadequate therapy on the wards before obtaining an ICUconsult, and subsequent to an ICU consult, as patients aregenerally comanaged by the ICU team until the ICU admis-sion, they may have been appropriately treated. Severalstudies [8,19-21] have noted the lack of recognition of thecritically ill patient on hospital wards, and in our cohort,there is an association with this delay and mortality. Cardosoet al [9] reported an increased mortality with delay to ad-mission after decision to admit. However, the median delayof 17.8 hours was considerably longer than that in our insti-tutions. In our medical patients, the time to ICU consult wasalmost twice as long as time to ICU admission, and this mayhave also contributed to the effect on mortality of the firstperiod for the medical patients.

The time to ICU consult was longer in medical patientsthan in surgical patients. This may be due to perhaps amore insidious decline in medical patients that is harder todetect compared with surgical patients. There may also becultural differences between the practice and organizationof medical and surgical teams that may have influenced theinclination to call for an ICU consult, which will need to beexplored. On the other hand, the time to ICU admissionwas the same for the 2 groups in our institutions.

This study has limitations. We studied only patientsadmitted to the ICU from medical or surgical wards and notthose who may have been sick without an ICU consult or notadmitted to the ICU. We chose this population becausewithin the limits of a retrospective chart review, this was the

693Adverse outcomes of delayed intensive care consultation

most consistent way to identify an unstable patient on theward. In addition, this was the population who are morelikely at risk for a demonstrable adverse outcome fromdelayed care. We also had to exclude 40 patients becauseof poor documentation, such that we could not ascertainperiods reliably. We could not infer, from the informationavailable in the charts, any tendency to early or late METtime to ICU consult duration in these patients. The pro-portion of patients that were excluded because of thisreason was also similar between surgical and medicalpatients. In addition, the 30-day mortality of these patientswas similar to that of that of the surgical and medicalpatients included in the study. Although almost all patientswho died had a do-not-resuscitate order before their death,it is unlikely that they were written as a direct consequenceof a delay in ICU consult or origin from surgical or medi-cal wards. Indeed, a chart review indicated that these orderswere written as a reflection of the patients' status andprognosis during or after an ICU admission. It would havealso been interesting to ascertain the appropriateness ofactions on the wards after decompensation as defined byMET criteria. Unfortunately, we could not obtain reliableinformation concerning this from the charts in ourretrospective review.

In conclusion, for patients who are sick enough to beadmitted to the ICU, we have shown an effect of longertime to ICU consult from meeting MET criteria on hardoutcomes such as ICU LOS and mortality. The outcome ofinterest may be different for surgical and medical wards,and this may be important for the metrics used to measurethe impact of more rapid consultation in these groups. Ourinstitutions did not have METs, and indeed, our resultssuggest that rapid critical care specialist assessment after anacute clinical deterioration could improve outcomes. Theability for a medical emergency/ICU consultation team tobe successful will, however, depend highly on the rapidityin which the treating ward team is able to recognize andreact to a given clinical deterioration. Possible reasons fordelayed MET activation in general have been described[24,25]. Although it is difficult to ascertain reasons fordelay in calling the ICU in a retrospective study such asours, based on recent surveys on our wards, factors such aslack of recognition of severity or fear of criticism seem tohave been important in the delay in consulting the ICUteam in our institutions. We are in the process of ad-dressing these factors through a training program on thewards. Individual institutions will need to develop tailoredapproaches that work best to facilitate rapid critical careassessment for a given center.

References

[1] Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy inthe treatment of severe sepsis and septic shock. N Engl J Med2001;345(19):1368-77.

[2] Terkelsen CJ, Sørensen JT, Maeng M, et al. System delay andmortality among patients with STEMI treated with primary percuta-neous coronary intervention. JAMA 2010;304(7):763-71.

[3] Blow O, Magliore L, Claridge JA, et al. The golden hour and thesilver day: detection and correction of occult hypoperfusion within 24hours improves outcome from major trauma. J Trauma 1999;47(5):964-9.

[4] Kumar A, Roberts D, Wood KE, et al. Duration of hypotensionbefore initiation of effective antimicrobial therapy is the criticaldeterminant of survival in human septic shock. Crit Care Med 2006;34(6):1589-96.

[5] Sebat F, Musthafa AA, Johnson D, et al. Effect of a rapid responsesystem for patients in shock on time to treatment and mortalityduring 5 years. Crit Care Med 2007;35(11):2568-75.

[6] Saukkonen KA, Varpula M, Räsänen P, et al. The effect of emergencydepartment delay on outcome in critically ill medical patients:evaluation using hospital mortality and quality of life at 6 months.J Intern Med 2006;260(6):586-91.

[7] Chalfin DB, Trzeciak S, Likourezos A, et al. Impact of delayed transferof critically ill patients from the emergency department to the intensivecare unit. Crit Care Med 2007;35(6):1477-83.

[8] Engoren M. The effect of prompt physician visits on intensive care unitmortality and cost. Crit Care Med 2005;33(4):727-32.

[9] Cardoso L, Grion C, et al. Impact of delayed admission to intensivecareunits on mortality of critically ill patients: a cohort study. Crit Care2011;15:R28.

[10] Young M, Gooder V-J, et al. Transfers to the intensive care unit.Delays are associated with increased mortality and morbidity. J GenIntern Med 2003;18:77-83.

[11] Buist MD, Moore GE, Bernard SA, et al. Effects of a medicalemergency team on reduction of incidence of and mortality fromunexpected cardiac arrests in hospital: preliminary study. BMJ2002;324(7334):387-90.

[12] Hillman K, Chen J, Cretikos M, et al. Introduction of medicalemergency team (MET) system—a cluster-randomised controlledtrial. Lancet 2005;365:2091-7.

[13] Sharek PJ, Parast LM, Leong K, et al. Effect of a rapid response teamon hospital-wide mortality and code rates outside the ICU in achildren's hospital. JAMA 2007;298(19):2267-74.

[14] Chan PS, Jain R, Nallmothu BK, et al. Rapid response teams: asystematic review and meta-analysis. Arch Intern Med 2010;170(1):18-26.

[15] Lee A, Bishop G, Hillman KM, et al. The medical emergency team.Anaesth Intensive Care 1995;23(2):183-6.

[16] Lin LIK. A concordance correlation coefficient to evaluate reproduc-ibility. Biometrics 1989;45:255-68.

[17] Subbe CP, Kruger M, Rutherford P, et al. Validation of a modifiedEarly Warning Score in medical admissions. QJM 2001;94(10):521-6.

[18] Hosmer DW, Lemeshow S. A goodness-of-fit test for the multiplelogistic regression model. Commun Stat 1980;A10:1043-69.

[19] McGloin H, Adam SK, Singer M. Unexpected deaths and referrals tointensive care of patients on general wards. Are some cases potentiallyavoidable? J R Coll Physicians Lond 1999;33:255-9.

[20] McQuillan P. Confidential inquiry into quality of care beforeadmission to intensive care. BMJ 1998;316:1853-8.

[21] Goldhill DR, et al. Physiological values and procedures in the 24 hbefore ICU admission from the ward. Anaesthesia 1999;54:529-34.

[22] de Rooij SE, Govers A, Korevaar JC, et al. Short-term and long-termmortality in very elderly patients admitted to an intensive care unit.Intensive Care Med 2006;32:1039-44.

[23] Romo H, Amaral AC, Vincent JL. Effect of patient sex on intensivecare unit survival. Arch Intern Med 2004 Jan 12;164(1):61-5.

[24] Jones D, et al. Long term effect of a medical emergency team oncardiac arrestsin a teaching hospital. Crit Care 2005;9:R808-15.

[25] Jon L, Quach, et al. Characteristics and outcomes of patients receivinga medical emergency team review for respiratory distress orhypotension. J Crit Care 2008;23:325-31.