Context The association of an adult tele-intensive care unit (ICU) intervention with hospital mortality, length of stay, best practice adherence, and preventable complications for an academic medical center has not been reported.
Objective To quantify the association of a tele-ICU intervention with hospital mortality, length of stay, and complications that are preventable by adherence to best practices.
Design, Setting, and Patients Prospective stepped-wedge clinical practice study of 6290 adults admitted to any of 7 ICUs (3 medical, 3 surgical, and 1 mixed cardiovascular) on 2 campuses of an 834-bed academic medical center that was performed from April 26, 2005, through September 30, 2007. Electronically supported and monitored processes for best practice adherence, care plan creation, and clinician response times to alarms were evaluated.
Main Outcome Measures Case-mix and severity-adjusted hospital mortality. Other outcomes included hospital and ICU length of stay, best practice adherence, and complication rates.
Results The hospital mortality rate was 13.6% (95% confidence interval [CI], 11.9%-15.4%) during the preintervention period compared with 11.8% (95% CI, 10.9%-12.8%) during the tele-ICU intervention period (adjusted odds ratio [OR], 0.40 [95% CI, 0.31-0.52]). The tele-ICU intervention period compared with the preintervention period was associated with higher rates of best clinical practice adherence for the prevention of deep vein thrombosis (99% vs 85%, respectively; OR, 15.4 [95% CI, 11.3-21.1]) and prevention of stress ulcers (96% vs 83%, respectively; OR, 4.57 [95% CI, 3.91-5.77], best practice adherence for cardiovascular protection (99% vs 80%, respectively; OR, 30.7 [95% CI, 19.3-49.2]), prevention of ventilator-associated pneumonia (52% vs 33%, respectively; OR, 2.20 [95% CI, 1.79-2.70]), lower rates of preventable complications (1.6% vs 13%, respectively, for ventilator-associated pneumonia [OR, 0.15; 95% CI, 0.09-0.23] and 0.6% vs 1.0%, respectively, for catheter-related bloodstream infection [OR, 0.50; 95% CI, 0.27-0.93]), and shorter hospital length of stay (9.8 vs 13.3 days, respectively; hazard ratio for discharge, 1.44 [95% CI, 1.33-1.56]). The results for medical, surgical, and cardiovascular ICUs were similar.
Conclusion In a single academic medical center study, implementation of a tele-ICU intervention was associated with reduced adjusted odds of mortality and reduced hospital length of stay, as well as with changes in best practice adherence and lower rates of preventable complications.
Patient needs and societal costs of adult critical care have increased as predicted from population-based models,1-3 and more efficient methods of delivery of care are needed. A tele-intensive care unit (ICU) is a promising technological approach designed to systematically alter processes of care that affect outcomes. Tele-ICU can be defined as the provision of care to critically ill patients by health care professionals located remotely. Tele-ICU clinicians use audio, video, and electronic links to assist bedside caregivers in monitoring patients, to oversee best practice adherence, and to help create and execute care plans. Tele-ICU programs have the potential to target processes that are associated with better outcomes, including shorter response times to alarms4 and abnormal laboratory values,5 more rapid initiation of life-saving therapies,6,7 and higher rates of adherence to critical care best practices.8
Studies of the effects of tele-ICU programs to date have focused primarily on community and rural hospitals and have yielded both positive and negative results. Several studies have highlighted low levels of acceptance of the tele-ICU intervention but few studies have identified the care processes or ICU structural elements that were part of the intervention; these issues have limited the ability to compare studies and identify processes associated with improved outcomes.9 To provide insight into which tele-ICU–related process changes are associated with better outcomes, we examined the association of a tele-ICU intervention with the risk of dying in the hospital and length of stay, and the contributions of best practice adherence and preventable complications to these associations. Our study focused on changes in the processes of care rather than ICU structure because the critical care team structure and governance had been previously well characterized10 and were not changed during the intervention.
We performed a prospective, unblinded, stepped-wedge study of a tele-ICU intervention at an academic medical center. To focus on processes of care rather than on structural elements, critical care governance, team structure (including an intensivist-led closed model), call schedules, interdisciplinary rounds, and staffing models were established before study enrollment. We standardized medical center best practices for the prevention of venous thrombosis,11 cardiovascular complications,12,13 ventilator-associated pneumonia,14,15 and stress ulcers16 through a task force–driven consensus-building process and codified them in our medical center policies.10 Before the start of the study, we introduced a form for ICU daily goals17 as part of the medical record and completed a supporting educational program on March 5, 2005, across 7 adult ICUs (3 medical, 3 surgical, and 1 mixed cardiovascular) on 2 campuses of an 834-bed academic medical center.
For the primary analysis, a representative sample of preintervention cases was obtained by identifying consecutive hospital discharge cases from an administrative database for cases managed in each of the 7 ICUs. All adults admitted to an ICU bed were included in the tele-ICU group. The tele-ICU intervention implementation dates for the ICUs were staggered and occurred on June 30, 2006, July 5, 2006, July 17, 2006, November 2, 2006, November 7, 2006, November 8, 2006, and April 16, 2007. The first preintervention patient was admitted to the medical center on April 26, 2005, and the last patient was discharged on February 7, 2007. The first tele-ICU intervention case was admitted on August 6, 2006, and the last case was discharged on September 30, 2007. Patients younger than 18 years and those admitted to a preintervention and tele-ICU intervention unit during the same hospital stay were excluded (Figure). The study was powered to have an 80% probability for detecting a 3.5% improvement in mortality using a 1:3 allocation of patients in the preintervention group and in the tele-ICU group at a significance level of .05. The University of Massachusetts Medical School human subjects committee approved the study and determined the study to be exempt from the informed consent requirement.
Admission, discharge, and laboratory information was abstracted electronically. Clinical information that was not in electronic form was collected by trained abstractors using the procedures detailed in the eSupplement. Randomly selected subsets of individual records were abstracted electronically or manually in duplicate to measure concordance among raters and to compare electronic with hard-copy abstraction. Adherence to prespecified and well-established best practices and rates of complications were measured (described in the eSupplement). The Acute Physiology and Chronic Health Evaluation (APACHE) III system was used to measure acuity; the period from the date of admission of the first case in each group was calculated as a time-trend factor to adjust for secular trends. Tele-ICU intervention episode type,18 severity, and determination of bedside vs off-site clinician initiation were electronically recorded. Validation procedures and data completeness are presented in the eSupplement. The guideline for managing intravenous catheters was adapted from Pronovost et al.19 To determine whether the associations of the tele-ICU intervention on length of stay could be attributed to higher rates of transfer of patients into chronic care facilities, discharge location was measured.
The tele-ICU team participated in key critical care delivery processes throughout the entire day (24 hours every day). The off-site team included an intensivist and used tele-ICU workstations. The tele-ICU team serially reviewed the care of individual patients, performed real-time audits of best practice adherence, performed workstation-assisted care plan reviews for patients admitted at night, monitored system-generated electronic alerts, audited bedside clinician responses to in-room alarms, and intervened18 when the responses of bedside clinicians were delayed and patients were deemed physiologically unstable. The care process elements during the preintervention and those added by the tele-ICU reengineering process are presented in Table 1. The off-site team had the ability to communicate with bedside clinicians or directly manage patients by recording clinician orders for tests, treatments, consultations, and management of life-support devices.20
Cases in the preintervention group who were admitted during nighttime hours were reviewed by telephone by the attending physician at the discretion of the bedside ICU house staff or affiliate practitioner after they had evaluated the case. The process was different for tele-ICU cases. The off-site intensivist reviewed the electronically available information including laboratory values and radiographic images, supported preadmission management, and assigned the case to an appropriate ICU team. The intensivist assessed the patient using real-time video, communicated with the ICU nurse and patient using audio links, responded to alerts and alarms, reviewed the response to the initial plan of care in real time, and shared responsibility for altering the care plan when the patient's condition failed to respond. All off-site clinical personnel also had rotations at other times in the medical center adult ICUs. The tele-ICU tools used to facilitate and track this intervention were licensed from Visicu Inc (Baltimore, Maryland), Cerner Healthcare Solutions (Kansas City, Missouri), or developed by our team at the University of Massachusetts (Criticalware).
Case-mix and severity-adjusted hospital mortality was prespecified as the main study outcome. Other outcomes included ICU mortality, hospital and ICU length of stay, rates of adherence to best practice, complication rates, and whether responses to alerts were initiated by bedside or by off-site team members. We performed post hoc analyses of cases matched on acuity score, ICU, season of the year, and operative status to determine whether cases of the same type and acuity had different outcomes when managed in the same ICU. Mediation analyses for hospital and ICU mortality were performed for prespecified best practice adherence and complication metrics.
The proportion of the effects mediated by these factors was estimated from differences in model coefficients using the probability density function method.21 Standardized coefficients were used for dichotomous outcomes.22 To determine if the associations were different between medical and surgical patients, we performed secondary analyses by ICU type. To investigate whether tele-ICU intensivist care plan review for patients admitted during nighttime hours contribute to the associations, we measured associations of the intervention with hospital and ICU mortality, length of stay, and duration of mechanical ventilation for patients admitted during the daytime compared with the nighttime.
The nearly 1-year period between implementation in the first and last surgical ICU allowed comparison of concurrent cases managed with and without tele-ICU support for this subset. Descriptive statistics were calculated for continuous variables and univariate comparisons between groups for continuous outcomes were made using the t test. Comparisons between groups on categorical variables were made using the Fisher exact test or the χ2 test or by logistic regression.
Dichotomous outcomes including hospital and ICU mortality were modeled using multivariable logistic regression. Continuous outcomes were modeled using general linear mixed models23,24 with the tele-ICU intervention as a fixed effect, an ICU identifier variable as a random effect, and using restricted estimation by maximum likelihood and by Cox proportionate hazards modeling. For continuous outcomes, type 3 F tests of effects were used to evaluate the significance of the contribution of predictors to each model and the minimum deviance (−2 × the model log likelihood) was used to select the best fitting models.
Likelihood ratio tests for nested models25 were used to select the appropriate predictors for each logistic model. Adherence and complication factors that were significantly affected by the intervention and the mortality effects were included with appropriate interaction terms into the model of tele-ICU effects on hospital mortality for mediation analyses. Analyses that included other appropriate interaction terms in the models were performed for effects by ICU type and admission time of day. After these models were obtained, each was refitted to data sets augmented by multiple imputations.26,27
Intraclass correlation coefficients were calculated to assess concordance among raters and for hard-copy and electronic medical record review.28 All hypothesis tests were 2-sided, were not adjusted for multiple comparisons, and the statistical significance was set at a P value level of .05. Statistical analyses were performed using Stata version 10.0 (StataCorp, College Station, Texas) and SAS version 9.2 (SAS Institute Inc, Cary, North Carolina) and used the LOGISTIC, PHREG, GLM, MIXED, and GLIMMIX procedures.
A total of 6290 qualifying cases were identified from 6465 electronic ICU admission registrations during the preintervention period and during the tele-ICU intervention period based on their dates of service. Patients who were not admitted to an adult ICU or were not adults were excluded (Figure). The tele-ICU intervention group had a larger percentage of patients with a primary admission diagnosis classified as medical rather than surgical, slightly more abnormal laboratory and physiological values than cases in the preintervention group, and accordingly higher acuity scores (Table 2) (eTable 1). These differences in APACHE III acuity score did not appear to be related to any single component or to the method of abstraction. The intraclass correlation coefficient for adherence measures was 0.83 for hard-copy abstraction and 0.86 for hard-copy vs electronic abstraction (intraclass correlation coefficients for acuity scores were larger and appear in the eSupplement).
Unadjusted ICU mortality was significantly lower in the tele-ICU group (P = .01) than in the preintervention group; hospital mortality showed a nonstatistically significant reduction (P = .07; Table 3). The hospital mortality rate was 13.6% (95% confidence interval [CI], 11.9%-15.4%) during the preintervention period compared with 11.8% (95% CI, 10.9%-12.8%) during the tele-ICU intervention and yielded an odds ratio (OR) of 0.40 (95% CI, 0.31-0.52; P = .005) after adjustment for acuity, locus of care, physiological parameters, laboratory values, and time trend. The ICU mortality rate was 10.7% for the preintervention group and 8.6% for the tele-ICU group and yielded an OR of 0.37 (95% CI, 0.28-0.49; P = .003) after adjustment for acuity, locus of care, physiological parameters, laboratory values, and time trend.
Mortality for concurrent surgical ICU cases in ICUs that had the intervention and in ICUs that had not yet implemented the intervention yielded effect estimates equivalent to those observed in the before and after comparison (absolute hospital mortality: 13.3% for the preintervention group and 7.3% for the tele-ICU group; OR, 0.21 [95% CI, 0.05-0.76]; P = .02) (absolute ICU mortality: 9.3% and 3.7%, respectively; OR, 0.17 [95% CI, 0.04-0.73]; P = .02).
Unadjusted analyses revealed that both hospital and ICU length of stay were significantly lower in the tele-ICU group than in the preintervention group (Table 3). The mean length of hospital stay was 13.3 days in the preintervention group and 9.8 days in the tele-ICU group and after adjustment for acuity, time trends, physiological parameters, laboratory values, and locus of care; hospital length of stay was significantly shorter in the tele-ICU group (hazard ratio [HR] for discharge, 1.44 [95% CI, 1.33-1.56]; P <.001). The mean length of ICU stay was 6.4 days in the preintervention group and 4.5 days in the tele-ICU group and after adjustment for all of the previously listed variables, the HR was 1.26 (95% CI, 1.17-1.36; P <.001).
Comparison of length of stay for concurrent surgical ICU cases managed in units that had the intervention with units that had not yet implemented the intervention yielded likelihood estimates of discharge that were similar to those observed in the before and after comparison (hospital length of stay: 12.6 days in the preintervention group and 12.4 days in the tele-ICU group; HR, 1.8 [95% CI, 1.1-2.9]; P = .01) (ICU length of stay: 6.9 days and 3.6 days, respectively; HR, 1.50 [95% CI, 0.95-2.40]; P = .08). For the tele-ICU group compared with the preintervention group, the fraction of patients requiring mechanical ventilation was significantly lower (Table 2) and the duration of mechanical ventilation was significantly shorter (5.70 days [95% CI, 5.27-6.13 days] and 8.50 days [95% CI, 7.47-9.53 days], respectively; P <.001). Compared with the preintervention group, the use of noninvasive ventilation was higher for the tele-ICU group (eTables 2-4).
A total of 1350 cases in the preintervention group were matched by ICU, season of the year, operative status, and APACHE III acuity score on a ratio of 1:1 to cases in the tele-ICU group (details are provided in the eSupplement). The hospital mortality was 8.5% for the tele-ICU group compared with 14.4% for the preintervention group (χ2 test P <.001) and ICU mortality was 4.8% vs 11.1%, respectively (P <.001). The mean (SD) length of hospital stay was 10.9 (13.4) days for the tele-ICU group compared with 14.2 (18.3) days for the preintervention group (paired t test P <.001) and the mean (SD) length of ICU stay was 4.2 (7.6) days and 6.9 (11.0) days, respectively (P <.001). Tele-ICU group cases of similar type and acuity had better outcomes than their preintervention group counterparts.
To understand how tele-ICU team activities affected care processes and to evaluate the degree to which the association of the intervention with changes in mortality could be attributed to these changes in process, we performed analyses of adherence to best practices, incidence of common ICU complications, intensivist involvement for cases admitted during nighttime hours, responses to alerts, and by ICU type. The associations were not attributable to any single ICU or type of ICU (data are presented in eTables 5-8).
The tele-ICU intervention was associated with significantly higher adherence to deep vein thrombosis prevention best practice and cardiovascular protection best practice, and lower rates of catheter-related bloodstream infection and ventilator-associated pneumonia (Table 4 and Table 5). We found that these factors also were associated with significantly lower ICU and hospital mortality. The magnitude of these changes in mortality was the size expected based on consensus reports of best practices.11-13 We then estimated the contribution of adherence to best practices and lower rates of complications to the lower mortality rates observed for the tele-ICU group by comparing standardized coefficients from our model estimates of tele-ICU associations for ICU and hospital mortality with and without inclusion of these factors. Analyses adjusted for these factors still demonstrated significant associations for hospital mortality (adjusted OR, 0.54 [95% CI, 0.39-0.73]; P <.001) and ICU mortality (adjusted OR, 0.58 [95% CI, 0.41-0.83]; P = .002), but the magnitude was smaller without adjustment for adherence and complications (OR, 0.40 [95% CI, 0.31-0.52] for hospital mortality and OR, 0.37 [95% CI, 0.28-0.49] for ICU mortality). The proportion of the tele-ICU association with lower mortality that could be attributed to adherence to these best practices and complication measures was estimated to be 25% for hospital mortality and 30% for ICU mortality.
We performed analyses by time of admission because worse outcomes for ICU patients admitted after daytime hours have been reported.29 We compared outcomes for 705 of 1529 cases in the preintervention group (46%) and 2287 of 4761 cases in the tele-ICU group (48%) that were admitted between 8 PM and 8 AM compared with those patients who were admitted during the daytime hours. In the preintervention group, the unadjusted hospital mortality rate for patients admitted after 8 PM was 16.1% compared with 11.5% for those admitted after 8 AM (P = .01). For the tele-ICU group, the unadjusted hospital mortality rate for patients admitted after 8 PM was 12.7% compared with 11.1% for those admitted after 8 AM (P = .11). There were significant interactions of the intervention with nighttime admission for hospital mortality (P = .04) and length of stay (P = .03).
Adjusted analyses demonstrated larger and significant associations of the intervention for those admitted after 8 PM (OR, 0.33 [95% CI, 0.18-0.59]; P <.001) compared with those admitted after 8 AM (OR, 0.79; 95% CI, 0.39-1.58). Cases admitted after 8 PM had longer hospital length of stay (14.3 days [95% CI, 12.99-15.57 days] in the preintervention group vs 12.4 days [95% CI, 11.22-13.58 days] in the tele-ICU group; P = .04) and ICU length of stay (7.7 days [95% CI, 6.77-8.63 days] vs 5.5 days [95% CI, 4.86-6.08 days], respectively; P <.001) than patients admitted after 8 AM. The diurnal differences were not significant in hospital length of stay (9.98 days [95% CI, 9.18-10.08 days] in the preintervention group vs 9.63 days [95% CI, 9.57-10.34 days] in the tele-ICU group; P = .27) and in ICU length of stay (4.39 days [95% CI, 4.15-4.63 days] vs 4.60 days [95% CI, 4.30-4.91 days], respectively; P = .24).
The magnitude of the association of the tele-ICU intervention with shorter hospital length of stay was greater for cases admitted after 8 PM (HR, 1.61 [95% CI, 1.35-1.92]; P <.001) than for patients admitted during daytime hours (HR, 1.26 [95% CI, 1.05-1.51]; P = .004). Analyses of ICU length of stay yielded similar results. The duration of mechanical ventilation was longer for patients in the preintervention group admitted after 8 PM (10.2 days; 95% CI, 9.16-11.24 days) than those admitted after 8 AM (6.9 days [95% CI, 6.34-7.46 days]; P <.01). Patients in the tele-ICU group admitted after 8 PM had a duration of mechanical ventilation of 5.8 days (95% CI, 5.24-6.36 days) compared with 5.5 days (95% CI, 4.99-6.01 days); P = .61) for those admitted after 8 AM.
Responses to alerts and alarms can be initiated by bedside personnel or by off-site personnel. During the preintervention period, there was no off-site patient monitoring and no off-site–initiated interventions. During the tele-ICU period, there were 6.80 (95% CI, 6.50-7.10) alerts for physiological instability per patient per day. Among these alerts for physiological instability per patient per day, 5.05 (95% CI, 4.77-5.33) alerts were managed by bedside clinicians without tele-ICU intervention and 1.75 (95% CI, 1.69-1.81) alerts were managed with tele-ICU intervention. Most interventions were initiated by the tele-ICU team. Among 24 426 interventions that affected the diagnostic or therapeutic plan, 23 943 were initiated by off-site clinicians and 483 interventions were initiated by bedside clinicians (ratio of 50:1). Among these interventions, 1633 were documented with progress notes that included a rating of the severity of the physiological disturbance; 76% of these were classified as major (eg, requiring initiation of a vasoactive medication), 17% intermediate (eg, evaluation of an arrhythmia), and 7% as minor (eg, electrolyte correction).
Intervention-associated differences in length of stay and mortality rates could not be attributed to higher rates of transfer of cases to institutions providing care to those recovering from acute illness. Patients in the tele-ICU group were 8% more likely to go home, 6% less likely to go to a rehabilitation or long-term care facility, and 2% more likely to go to a skilled nursing facility than patients in the preintervention group (Table 6).
The main findings of this study are that a tele-ICU intervention was associated with lower hospital and ICU mortality and shorter hospital and ICU lengths of stay. The tele-ICU intervention also was associated with significantly higher rates of adherence to critical care best practices and lower rates of complications. This study also identified more rapid response to alerts for physiological instability, and off-hours, off-site intensivist care plan review as critical care process elements that may have contributed to the lower mortality and shorter lengths of stay associated with the tele-ICU intervention. These new care processes (Table 1) entail collaboration among bedside clinicians and off-site support team personnel. The association of lower mortality and shorter length of stay at an academic medical center implies that there are benefits of tele-ICU care beyond providing bedside intensivist expertise to ICUs. It is unique because it attributes the association of the intervention with lower mortality to best practice adherence and decreased complication rates targeted by the intervention.
The tele-ICU intervention changed the way our institution promoted adherence to critical care best practices that are widely accepted as improving outcomes.8 For example, tele-ICU tools were developed to provide real-time auditing and reconciliation. The tele-ICU intervention was incremental and applied after checklist-based and traditional education-based programs had been used that targeted these same practices. The new care processes that did not primarily involve these best practices, including timely and consistent responses to alarms4 and intensivist case involvement,30,31 also were part of the intervention. The procedures used before and after the intervention are detailed in Table 1. We found that implementation of these new processes, made possible by the tele-ICU intervention, was associated with significantly higher rates of best practice adherence and lower rates of preventable complications than the educational outreach and checklist-based reminders that we had used before the tele-ICU intervention and were reported by others.17 We used mediation analysis to estimate the portion of the association of the intervention with lower mortality that could be attributed to best practice and complication measures. The contributions of best practice and complication measures were significant in these analyses and explained between 25% and 30% of the association for the intervention. Because significant associations remained after adjustment for best practice and complication measures, we performed additional exploratory analyses.
Studies reporting higher mortality for ICU patients admitted at night29 suggested the hypothesis that part of the lower mortality and shorter length of stay may be attributable to having a rested on-duty intensivist to assist with the creation of care plans and to monitor the initial progress of patients admitted at night using tele-ICU workstations and tools. We found larger associations of the tele-ICU intervention for patients admitted during the nighttime than those admitted during the daytime, suggesting that intensivist care plan review of off-hours cases was an important contributor to the association of the intervention with lower mortality and shorter lengths of stay observed for this study. These findings are consistent with studies that associate better outcomes with higher levels of intensivist involvement30 and intensivist-directed care.31
Involvement of the tele-ICU intensivist also was associated with more cases being managed with noninvasive ventilation and lower use of conventional mechanical ventilation. We speculate that the use of noninvasive ventilation increased when emergency department physicians perceived that patients responding favorably to noninvasive ventilation could continue to be managed safely with this approach due to the additional surveillance by the tele-ICU team.
We also explored how the tele-ICU intervention functioned with regard to the responses to alerts and alarms for episodes of physiological instability. Although we found that bedside clinicians responded to most episodes of physiological instability without prompting from the off-site team an average of 1.7 times per patient per day, the tele-ICU team prompted the bedside team when the timeliness of response was deemed to be unsafe. Taken together with studies associating better outcomes for critically ill adults receiving earlier interventions like antimicrobials32 and resuscitation for sepsis,7 these findings suggest that earlier responses are an additional mechanism by which tele-ICU programs can be associated with better outcomes.
One explanation for why some studies have failed to detect significant associations for their tele-ICU interventions is that most reported low rates of collaboration among tele-ICU and bedside physicians (34%-36%),33,34 and likely did not involve the tele-ICU team in case sign out, collaborative review of bedside practices, or initial responses to alarms. We speculate that bedside and off-site team collaboration is an important determinate of favorable associations of a tele-ICU intervention with outcomes. This hypothesis is supported by the observation that studies that failed to detect significant favorable tele-ICU associations with outcomes included a pathway for bedside physicians to decline participation of the off-site team in patient care, while our workflow processes and other studies reporting significant favorable associations with outcomes for a tele-ICU intervention20,35-37 did not provide an opt-out pathway.
Lower lengths of stay for tele-ICU group cases was not related to shifting patients to the chronic care system; rather, patients in the tele-ICU group had lower rates of complications, recovered more quickly, and were significantly more likely to be discharged to their homes than patients in the preintervention group. These findings suggest that critical care programs that implement processes that increase adherence to best practice, lower rates of complications, shorten response times to alerts and alarms, and support early intensivist case involvement will provide better care at a lower cost. Our study suggests that the introduction of a tele-ICU program that collaborates with and supports bedside clinicians is one way to accomplish these aims.
This study has important limitations and reports associations rather than cause and effect relationships. The fact that this is a single academic medical center study should be taken into account when considering its generalizability. Our study design was not a prospective, randomized, and blinded trial with its inherent protections against heterogeneity among the groups; some differences were detected among the study participants. Nevertheless, because this approach does not exclude heterogeneous patients, it has the potential to be more reflective of outcomes achieved in actual practice than those observed in many randomized controlled trials. With respect to differences in acuity between the 2 groups, we believe that they were due in part to the staggered implementation of the tele-ICU intervention and to higher rates of transfer of higher acuity medical cases from outside hospitals for the tele-ICU group; however, these differences could also be due to other unidentified factors. We also were aware that the effects of our educational outreach over time could have been mistaken for interventional effects. However, when we investigated this possibility, we learned that our findings persisted and were robust with regard to confounding by secular trends (details are provided in the eSupplement).
We detected significant tele-ICU intervention effects both in time factor–adjusted analyses and in interrupted time-sequence analyses. In addition, comparison of concurrent surgical ICU cases revealed better outcomes for those managed using tele-ICU resources than for cases managed without them. The reduction in unadjusted hospital mortality did not reach statistical significance but generated a statistically significant and substantial adjusted OR. The major driver of the difference after adjustment was the higher acuity scores that we observed for patients in the tele-ICU group. We investigated the possibility that the association was only due to differences in acuity score or case mix by performing an analysis of cases matched on these attributes and found significant differences associated with the tele-ICU intervention that could not be attributed to acuity or case mix.
Implementation of a tele-ICU program has challenges that are distinct from those of instituting a clinical practice guideline or participation in a clinical trial. The changes in ICU process that were associated with lower hospital mortality and shorter lengths of stay were linked to a high rate of program acceptance by our ICU community. Our approach to implementation included features that are widely accepted for quality improvement, including being focused on patient-centered outcomes, having strong executive support, and targeting changes that were known to improve outcomes. It also included aspects that are relevant to interdisciplinary care, including robust participation of all critical care disciplines in the planning and implementation phases and a process10 that empowered local leaders to define critical care best practices.
In conclusion, an adult tele-ICU intervention at an academic medical center that had been previously well staffed with a dedicated intensivist model and had robust best practice programs in place before the intervention was associated with lower mortality and shorter lengths of stay. Only part of these associations could be attributed to following best practice guidelines and lower rates of preventable complications. This suggests that there are benefits of a tele-ICU intervention beyond what is provided by daytime bedside intensivist staffing and traditional approaches to quality improvement such as the process changes presented in Table 1.
Corresponding Author: Craig M. Lilly, MD, University of Massachusetts Memorial Medical Center, 281 Lincoln St, Worcester, MA 01605 (craig.lilly@umassmed.edu).
Published Online: May 16, 2011 (doi:10.1001/jama.2011.697).
Author Contributions: Dr Lilly had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lilly, Zhao, Baker, Irwin.
Acquisition of data: Lilly, Cody, Zhao, Landry, McIlwaine, Chandler.
Analysis and interpretation of data: Lilly, Zhao, Landry, Baker, Irwin.
Drafting of the manuscript: Lilly, Zhao.
Critical revision of the manuscript for important intellectual content: Lilly, Cody, Zhao, Landry, Baker, McIlwaine, Chandler, Irwin.
Statistical analysis: Lilly, Zhao, Baker.
Administrative, technical, or material support: Lilly, Cody, Landry.
Study supervision: Lilly, Chandler.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This project was funded in its entirety by the University of Massachusetts.
Role of the Sponsors: No funding organization or sponsor was involved in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
University of Massachusetts Memorial Critical Care Operations Group: Nicholas A. Smyrnios, MD, Stephen O. Heard, MD, Nicholas Hemeon, Timothy A. Emhoff, MD, Peter H. Bagley, MD, Sara E. Cody, Michael J. Davis, Cheryl Lopriore, Greg Wongkam, Diane Henry, J. Matthias Walz, MD, Margaret Naughton, BSN, RN, Michelle M. Fernald, MS, RN, Debra Lynn Svec, RN, Karen Ostiguy, MSN, Nam Heui Kim, MD, Cheryl H. Dunnington, MS, RN, Nancy Simon, MS, RN, M. Elizabeth Colo, MS, RN, Bruce J. Simon, MD, Karen Shea, MS, RN, Wiley R. Hall, MD, Robert Spicer, RN, Lynn Harrison, MD, Naomi F. Botkin, MD, Craig Smith, MD, Gail Frigoletto, BSN, RN, Melinda Darrigo, MS, NP, Cathy Pianka, MS, RN, Linda Josephson, MS, RN, Khaldoun Faris, MD, Scott E. Kopec, MD, Scott Leonard, MBA, RRT, Cynthia T. French, MS, ANP-BC, Helen M. Flaherty, MS, RN, Sara Fine, and Walter H. Ettinger Jr.
2.Milbrandt EB, Kersten A, Rahim MT,
et al. Growth of intensive care unit resource use and its estimated cost in Medicare.
Crit Care Med. 2008;36(9):2504-251018679127
PubMedGoogle ScholarCrossref 3.Angus DC, Shorr AF, White A, Dremsizov TT, Schmitz RJ, Kelley MA.Committee on Manpower for Pulmonary and Critical Care Societies (COMPACCS). Critical care delivery in the United States: distribution of services and compliance with Leapfrog recommendations.
Crit Care Med. 2006;34(4):1016-102416505703
PubMedGoogle ScholarCrossref 5.Kuperman GJ, Teich JM, Tanasijevic MJ,
et al. Improving response to critical laboratory results with automation: results of a randomized controlled trial.
J Am Med Inform Assoc. 1999;6(6):512-52210579608
PubMedGoogle ScholarCrossref 6.Kumar A, Roberts D, Wood KE,
et al. Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock.
Crit Care Med. 2006;34(6):1589-159616625125
PubMedGoogle ScholarCrossref 7.Rivers E, Nguyen B, Havstad S,
et al; Early Goal-Directed Therapy Collaborative Group. Early goal-directed therapy in the treatment of severe sepsis and septic shock.
N Engl J Med. 2001;345(19):1368-137711794169
PubMedGoogle ScholarCrossref 8.Institute of Medicine; Committee on the Quality of Health Care. Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: National Academy Press; 2001
10.McCauley K, Irwin RS. Changing the work environment in ICUs to achieve patient-focused care: the time has come.
Chest. 2006;130(5):1571-157817099040
PubMedGoogle ScholarCrossref 11.Hirsh J, Dalen J, Guyatt G.American College of Chest Physicians. The sixth (2000) ACCP guidelines for antithrombotic therapy for prevention and treatment of thrombosis.
Chest. 2001;119(1):(Suppl)
1S-2S11157638
PubMedGoogle ScholarCrossref 12.Cannon C. Improving acute coronary syndrome care: the ACC/AHA guidelines and critical pathways.
J Invasive Cardiol. 2003;15:(Suppl B)
22B-27B, discussion 27B-29B12724583
PubMedGoogle Scholar 13.Moser DK, Kimble LP, Alberts MJ,
et al; American Heart Association Council on Cardiovascular Nursing and Stroke Council. Reducing delay in seeking treatment by patients with acute coronary syndrome and stroke: a scientific statement from the American Heart Association Council on Cardiovascular Nursing and Stroke Council.
J Cardiovasc Nurs. 2007;22(4):326-34317589286
PubMedGoogle ScholarCrossref 14.Collard HR, Saint S, Matthay MA. Prevention of ventilator-associated pneumonia: an evidence-based systematic review.
Ann Intern Med. 2003;138(6):494-50112639084
PubMedGoogle ScholarCrossref 15.Isakow W, Kollef MH. Preventing ventilator-associated pneumonia: an evidence-based approach of modifiable risk factors.
Semin Respir Crit Care Med. 2006;27(1):5-1716508876
PubMedGoogle ScholarCrossref 17.Pronovost P, Berenholtz S, Dorman T, Lipsett PA, Simmonds T, Haraden C. Improving communication in the ICU using daily goals.
J Crit Care. 2003;18(2):71-7512800116
PubMedGoogle ScholarCrossref 19.Pronovost P, Needham D, Berenholtz S,
et al. An intervention to decrease catheter-related bloodstream infections in the ICU.
N Engl J Med. 2006;355(26):2725-273217192537
PubMedGoogle ScholarCrossref 20.Rosenfeld BA, Dorman T, Breslow MJ,
et al. Intensive care unit telemedicine: alternate paradigm for providing continuous intensivist care.
Crit Care Med. 2000;28(12):3925-393111153637
PubMedGoogle ScholarCrossref 21.Freedman LS, Graubard BI, Schatzkin A. Statistical validation of intermediate endpoints for chronic diseases.
Stat Med. 1992;11(2):167-1781579756
PubMedGoogle ScholarCrossref 22.Mackinnon DP, Dwyer JH. Estimating mediated effects in prevention studies.
Eval Rev. 1993;17:144-158
Google ScholarCrossref 23.McLean RA, Stroup WW. A unified approach to mixed linear models.
Am Stat. 1991;45(1):54-64
Google Scholar 24.Robinson GK. Thast BLUP is a good thing: the estimation of random effects.
Stat Sci. 1991;45(1):15-51
Google ScholarCrossref 25.Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: Wiley; 1989
26.Harel O, Zhou XH. Multiple imputation: review of theory, implementation and software.
Stat Med. 2007;26(16):3057-307717256804
PubMedGoogle ScholarCrossref 27.Schafer JL, Olsen MK. Multiple imputation for multivariate missing data problems: a data analyst's perspective.
Multivariate Behav Res. 1998;33(4):545-571
Google ScholarCrossref 28.Fleiss JL. Statistical Methods for Rates and Proportions. New York, NY: Wiley; 1981
29.Bhonagiri D, Pilcher DV, Bailey MJ. Increased mortality associated with after-hours and weekend admission to the intensive care unit: a retrospective analysis.
Med J Aust. 2011;194(6):287-29221426282
PubMedGoogle Scholar 30.Pronovost PJ, Angus DC, Dorman T, Robinson KA, Dremsizov TT, Young TL. Physician staffing patterns and clinical outcomes in critically ill patients: a systematic review.
JAMA. 2002;288(17):2151-216212413375
PubMedGoogle ScholarCrossref 31.Hanson CW III, Deutschman CS, Anderson HL III,
et al. Effects of an organized critical care service on outcomes and resource utilization: a cohort study.
Crit Care Med. 1999;27(2):270-27410075049
PubMedGoogle ScholarCrossref 32.Kumar A, Ellis P, Arabi Y,
et al; Cooperative Antimicrobial Therapy of Septic Shock Database Research Group. Initiation of inappropriate antimicrobial therapy results in a fivefold reduction of survival in human septic shock.
Chest. 2009;136(5):1237-124819696123
PubMedGoogle ScholarCrossref 33.Thomas EJ, Lucke JF, Wueste L, Weavind L, Patel B. Association of telemedicine for remote monitoring of intensive care patients with mortality, complications, and length of stay.
JAMA. 2009;302(24):2671-267820040555
PubMedGoogle ScholarCrossref 34.Morrison JL, Cai Q, Davis N,
et al. Clinical and economic outcomes of the electronic intensive care unit: results from two community hospitals.
Crit Care Med. 2010;38(1):2-819730249
PubMedGoogle ScholarCrossref 35.Breslow MJ, Rosenfeld BA, Doerfler M,
et al. Effect of a multiple-site intensive care unit telemedicine program on clinical and economic outcomes: an alternative paradigm for intensivist staffing.
Crit Care Med. 2004;32(1):31-3814707557
PubMedGoogle ScholarCrossref 36.McCambridge M, Jones K, Paxton H, Baker K, Sussman EJ, Etchason J. Association of health information technology and teleintensivist coverage with decreased mortality and ventilator use in critically ill patients.
Arch Intern Med. 2010;170(7):648-65320386011
PubMedGoogle ScholarCrossref 37.Kohl B, Gutsche J, Kim P, Sites F, Ochroch E. Effect of telemedicine on mortality and length of stay in a university ICU.
Crit Care Med. 2007;35(12 suppl):A22
Google Scholar