Clinical outcomes after telemedicine intensive care unit implementation*

Clinical outcomes after telemedicine intensive care unit implementation* Beth Willmitch, RN, BSN; Susan Golembeski, PhD, RN, CHRC; Sandy S. Kim, MA, MEd; Loren D. Nelson, MD, FACS, FCCM; Louis Gidel, MD, PhD, FCCP Objective: To examine clinical outcomes before and after implementation of a telemedicine program in the intensive care units of a five-hospital healthcare system. Design: Observational study with the baseline period of 1 yr before the start of a telemedicine intensive care unit program implementation at each of 5 hospitals. The post periods are 1, 2, and 3 yrs after telemedicine intensive care unit program implementation at each hospital. Setting: Ten adult intensive care units (114 beds) in five community hospitals in south Florida. A telemedicine intensive care unit program with remote 24/7 intensivist and critical care nurse electronic monitoring was implemented by a phased approach between December 2005 and July 2007. Measurements and Main Results: Records from 24,656 adult intensive care unit patients were analyzed. Hospital length of stay, intensive care unit length of stay, hospital mortality, and Case Mix Index were measured. Severity of illness using All Patient Refined-

Q

uality health care in intensive care units (ICUs) is complex and requires extensive resource use. The Leapfrog group has provided guidelines for ICUs, which recommend intensivist-led care for all patients in ICUs (1). This is not easily achievable as a result of the overall shortage of intensivists (2). In addition, the small size of many hospitals precludes their ability to support a fulltime intensivist program. Tele-ICU programs are one solution because they are capable of leveraging the skills of an ex-

*See also p. 668. From eICU© LifeGuard (BW, LG) and the Center for Research & Grants (SG), Baptist Health South Florida, Miami, FL; Stat Support, Inc. (SSK), Boulder, CO; and the University of Central Florida (LDN), Orlando, FL. Work attributed to Baptist Health South Florida, Department of eICU® LifeGuard, and Center for Research & Grants. Dr. Gidel is an employee of Baptist Health in south Florida. The remaining authors have not disclosed any potential conflicts of interest. For information regarding this article, E-mail: [email protected] Copyright © 2012 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/CCM.0b013e318232d694

450

Diagnosis Related Groups scores was used as a covariate. From the baseline year to year 3 postimplementation, the severity-adjusted hospital length of stay was lowered from 11.86 days (95% confidence interval [CI] 11.55–12.21) to 10.16 days (95% CI 9.80ⴚ10.53; p < .001), severity-adjusted intensive care unit length of stay was lowered from 4.35 days (95% CI 4.22– 4.49) to 3.80 days (95% CI 3.65–3.94; p < .001), and the relative risk of hospital mortality decreased to 0.77 (95% CI 0.69 – 0.87; p < .001). Conclusions: After 3 yrs of deployment of a telemedicine intensive care unit program, this retrospective observational study of mortality and length of stay outcomes included all cases admitted to an adult intensive care unit and found statistically significant decreases in severity-adjusted hospital length of stay of 14.2%, intensive care unit length of stay of 12.6%, and relative risk of hospital mortality of 23%, respectively, in a multihospital healthcare system. (Crit Care Med 2012; 40:450 – 454) KEY WORDS: ICU outcomes; tele-ICU; telemedicine

perienced team of critical care doctors and nurses to ICUs where bedside services are not available and provide a vehicle for broadly applying evidence-based best practice protocols to improve patient safety and outcomes. The deployment of a tele-ICU program is not the same as the deployment of a single new procedure or device, but rather represents the deployment of a whole new complex culture for management of patients in the ICU. A study by Yoo et al (3) points out that the deployment of these systems involves different elements in different hospital systems and that these subtle differences may lead to significantly different outcomes. This variability in outcomes has been reflected in published reports about the effectiveness of tele-ICU systems (2, 4 –10) and may be linked both to the availability of fully electronic medical record systems and to the extent to which tele-ICU intensivists are permitted to proactively intervene in the patient’s care. Studies reporting improved outcomes have tended to allow more proactive tele-ICU participation (5–10), whereas those reporting little change in outcomes have tended to

allow less (2, 4). Among those allowing greater participation, a study by Rosenfeld et al (8) reported decreased in severity-adjusted ICU mortality by 45% and decreased hospital mortality 30%. A study by Breslow et al (7) reported hospital mortality for ICU patients was lower during the period of remote ICU care (9.4% vs. 12.9%; relative risk 0/73; 95% confidence interval [CI] 0.55⫺0.95) and ICU length of stay was shorter (3.63 days; 95% CI 3.21– 4.04 vs. 4.35 days; 95% CI 3.93– 4.78). A study by McCambridge et al (9) reported the hospital mortality after tele-ICU deployment was reduced by 29.5%. A study by Zawada (5) reported severity-adjusted length of stay (LOS) was reduced in three regional facilities and within their tertiary care facility, tele-ICU was associated with a reduction in severity-adjusted ICU mortality (odds ratio 0.35; p ⫽ .007), decreased ICU LOS (3.79 vs. 2.08 days; p ⫽ .001), and reduced hospital LOS (10.08 vs. 7.81 days; p ⫽ .001). The New England Healthcare Institute (11) reported that at the University of Massachusetts Memorial Medical Center, ICU mortality rates decreased ⬎20% even as the severity of the patients’ conCrit Care Med 2012 Vol. 40, No. 2

ditions rose significantly, and the ICU patients’ total hospital mortality rates declined 13%. At Community Hospital 1, ICU-adjusted mortality decreased 36%. ICU patients’ LOS decreased dramatically under tele-ICU at the University of Massachusetts Memorial Medical Center with an average reduction of almost 2 days or 30%. Both Community Hospitals 1 and 2 also saw a reduction in ICU LOSs. Finally, a study by Lilly et al (12) quantified the association of a tele-ICU intervention with hospital mortality, LOS, and complications that are preventable by adherence to best practices in a single academic medical center study. They reported a hospital mortality rate of 13.6% (95% CI 11.9% to 15.4%) during the preintervention period compared with 11.8% (95% CI 10.9% to 12.8%) during the tele-ICU intervention period and hospital LOSs of 9.8 days and 13.3 days, respectively. The purpose of this study was to assess retrospectively the effect of the tele-ICU program implementation on mortality and LOS associated with implementation of a 24/7 tele-ICU program in a fivehospital, community-based healthcare system involving 24,656 ICU admissions. Baseline information collected for 1 yr before implementation is compared with outcomes data collected for 3 yrs after implementation of the tele-ICU system.

METHODS Study Setting. Baptist Health South Florida is comprised of five hospitals with 1612 inpatient beds. A phased approach was used to bring all five hospitals live with the tele-ICU program from December 2005 to July 2007. Only the largest of the five hospitals has 24/7 bedside intensivist coverage with the intensivist controlling ICU allocation. All other ICUs in the health system are open units with care managed by private practice physicians. Three of the hospitals have physically designated stepdown units. The largest of the hospitals in the healthcare system has 680 beds with 32 adult ICU beds. The smallest hospital has 42 beds, eight of which are considered ICU. Two of the hospitals are in rural locations. The two largest hospitals are Magnet-designated facilities. The tele-ICU facility is located in a separate, standalone, off-site location distant from all hospital campuses. All admitting and consulting physicians in the healthcare system were asked at the time the tele-ICU program was implemented to indicate their requested level of intervention from the tele-ICU for their own patients ranging from level I (low) through level III (high). Level I is for life-threatening care such as codes. Level II includes all best practices ad-

Crit Care Med 2012 Vol. 40, No. 2

opted by the hospital. Level III is partner level with all care being open to adjustments by the tele-ICU team members. Of the 2607 members of the medical staff, 30 selected level I, 2531 selected or were assigned by default to level II, and 46 selected level III. The tele-ICU facility operates 24/7 and is staffed by one intensivist, three critical care nurses, and one unit secretary. Each workstation consists of the following: Philips VISICU eCare Manager electronic critical care system with Admission, Discharge and Transfer, laboratory (except the microbiology data) and pharmacy (beginning in year 2) electronic interfaces, Philips VISICU Smart Alerts, Philips VISICU camera system (Philips, Amsterdam, The Netherlands) in each ICU room allowing two-way voice and one-way video communication in each ICU room, mirrored real-time Philips bedside monitors and PACS (Picture Archiving and Communications System) for radiology, Sovera archival system (Sovera Health Information Management, a part of CGI of Montreal, Quebec, Canada) for patient data from all five hospitals, and Siemens health information system (Siemens, Malvern, PA). The eCare Manager has an outbound link for all notes, orders and nursing documentation from eCare Manager to the Sovera archival documentation system, but no inbound link for any documentation from either Sovera or the bedside paper chart. Handwritten bedside physician documentation is available to the tele-ICU program when faxed from the bedside. Before implementation all nursing documentation was done on paper and transitioned to electronic format in year 2. Study Design and Sample. This observational study uses a pre and post design approved by the institutional review board at Baptist Health South Florida. The baseline pre period is 1 yr before the start date of tele-ICU implementation at each of the five hospitals

within the system. The post periods are 1, 2, and 3 yrs post tele-ICU implementation start date at each hospital. A total of 24,656 adult ICU patient records were analyzed. Statistical Analyses. The methods used for statistical analysis are described in a study by Lang and Secic (13). Basic descriptive and observed outcome data are presented as the frequency, percentage, or mean ⫾ SD. Differences were assessed by chi-square test for categorical variables. Analysis of variance and analysis of covariance were used to test differences in outcomes as a result of the presence of tele-ICU program (baseline, 1, 2, and 3 yrs postimplementation). The post hoc Bonferroni procedure was applied to measure differences for baseline, 1-, 2-, and 3-yr postimplementation periods. Logistic regression analysis was used to compare the risk of severity-adjusted hospital mortality using All Patient Refined-Diagnosis Related Groups score with the presence or absence of tele-ICU program between the study periods. All analyses were performed on SPSS 13.0 (SPSS Inc., Chicago, IL). All of the analysis was performed at the same time after all the data from the baseline year and years 1 through 3 were available.

RESULTS A total of 24,656 adult patient records from ICU were analyzed in the study (Table 1). The severity of illness increased from baseline to each of the postimplementation periods and was statistically significant (p ⬍ .001). Case Mix Index means paralleled severity of illness, which increased each year compared with baseline (p ⫽ .002). Specific unadjusted mean differences from post hoc tests are indicated in Table 1. The unadjusted total mortality rate was not significantly different by intervention periods (p ⫽ .114).

Table 1. Observed clinical characteristics of baseline and post telemedicine intervention by yeara Baseline (n ⫽ 6504) Hospital 1 2241 2 704 3 1072 4 436 5 2051 Severity of illness 2.81 Case Mix Index 2.68 Total mortality 798 Hospital length 11.38 of stay Intensive care unit 4.20 length of stay

1 Yr Post (n ⫽ 6353)

2 Yrs Post (n ⫽ 6018)

3 Yrs Post (n ⫽ 5781)

34.5% 2155 33.9% 2097 34.8% 2168 10.8% 774 12.2% 679 11.3% 783 16.5% 1108 17.4% 1035 17.2% 458 6.7% 365 5.7% 305 5.1% 410 31.5% 1951 30.7% 1902 31.6% 1962 (1.036) 2.94c (0.999) 2.96c (0.998) 2.95c (2.799) 2.73 (2.820) 2.82b (2.410) 2.84b 12.3% 806 12.7% 751 12.5% 655 (15.356) 11.92 (16.516) 11.11 (8.396) 10.36b (6.345)

4.37

(6.018)

4.02

(3.908)

37.5% 13.5% 7.9% 7.1% 33.9% (1.002) (2.661) (11.3%) (12.690)

3.86b (5.316)

a Raw data, unadjusted means, and percentages; bp ⫽ ⬍ .05; cp ⬍ .001 compared with baseline. Data are summarized as means (⫾ SD) for continuous variables and no. (percentages) for categorical variables.

451

The unadjusted hospital LOS and ICU LOS both increased from baseline periods compared with 1 yr post. Two-year postimplementation means decreased from baseline (Table 1). Unadjusted hospital LOS and ICU LOS were significantly lowered in year 3 post compared with baseline (Table 1). Likewise, when hospital LOS and ICU LOS were adjusted for severity of illness, the means decreased significantly in the second and third years (Tables 2 and 3). The relationship between the covariate (severity of illness) and the dependent variable (LOS) did not differ significantly as a function of the independent variable (implementation period). This suggested that differences in LOS are not the result of the interaction of severity of illness and implementation period. Table 3 reveals the significant mean differences for severity of illness-adjusted LOS and ICU LOS between baseline and 2 and 3 yrs postintervention. There are statistically significant decreases in LOS in the second and third year postimplementation of the tele-ICU program. ICU LOS severity of illness-adjusted means also indicate a significant decrease in ICU LOS specifically comparing baseline with 2 and 3 yrs postimplementation periods (Fig. 1). There are significant decreases between 1 yr postimplementation compared with years 2 and 3 for LOS and ICU LOS. The summary of logistic regression analysis on mortality is presented in Table 4. Severity of illness was used as a covariate. Adjusted total hospital mortality was significantly lower between baseline and 2 yrs postimplementation (relative risk, .88; 95% CI 0.78 – 0.98, p ⫽ .025) and also 3 yrs postimplementation (relative risk, 0.77; 95% CI 0.69 – 0.87; p ⬍ .001) indicating a decrease in risk of hospital death. The ICU mortality could not be calculated from this data set because the hospital medical records did not consistently indicate whether a patient died in the ICU or on the floor. Additional analysis was performed to examine the effect of long LOS outliers on the LOS results. In this subgroup analysis, patients were excluded from the sample if hospital LOS was ⬎50 days or ICU LOS ⬎30 days (14). The percentage of patients excluded by truncating the sample by these long LOSs was 2.8%, 2.6%, 2.4%, and 1.7% at the baseline year and years 1, 2, and 3 post, respectively. From the baseline year to year 3 postimplementation in this subgroup analysis, the severity-adjusted hospital LOS 452

Table 2. Length of stay and intensive care unit length of stay means adjusted for severity of illness Baseline (1 yr) Variable

Mean

95% CI

1 Yr Post Mean

2 Yrs Post

95% CI

Mean

95% CI

3 Yrs Post Mean

95% CI

Length of staya 11.86 11.52 12.21 11.81 11.47 12.16 10.88 10.53 11.24 10.16 9.80 10.53 Intensive care unit 4.35 4.22 4.49 4.34 4.20 4.48 3.95 3.81 4.09 3.80 3.65 3.94 length of staya CI, confidence interval. Covariates appearing in the model are evaluated at the following values: severity of illness ⫽ 2.92.

a

Table 3. Hospital length of stay and intensive care unit length of stay mean difference adjusted for severity of illness by intervention period Intervention Period

Variable Length of stay

Intensive care unit length of stay

95% Confidence Interval

Mean Difference

Baseline 1 yr post 2 yrs post 3 yrs post 1 yr post 2 yrs post 3 yrs post Baseline 1 yr post 2 yrs post 3 yrs post 1 yr post 2 yrs post 3 yrs post

p

0.054 0.980 1.700

⫺0.432 0.487 1.202

0.539 1.473 2.198

.829 ⬍.001 ⬍.001

0.926 1.646

0.432 1.146

1.421 2.146

⬍.001 ⬍.001

0.016 0.403 0.558

⫺0.179 0.205 0.359

0.210 0.600 0.757

.874 ⬍.001 ⬍.001

0.387 0.542

0.189 0.342

0.585 0.743

⬍.001 ⬍.001

Figure 1. A, Severity of illness-adjusted intensive care unit (ICU) length of stay (LOS). B, Severity of illness-adjusted hospital LOS.

Table 4. Logistic regression model for hospital mortality

Hospital mortality

Variable

Relative Risk

Intervention period Baseline 1 yr post 2 yr post 3 yr post

Reference 0.92 0.88 0.77

95% Confidence Interval

0.82 0.78 0.69

1.03 0.98 0.87

p

.142 .025 ⬍.001

Hosmer-Lemeshow statistics for goodness of fit for model p ⬍ .001.


was lowered from 10.06 days (95% CI 9.82–10.25; p ⬍ .001) to 9.14 days (95% CI 8.94⫺9.34; p ⬍ .001), the severityadjusted ICU LOS was lowered from 3.71 days (95% CI 3.63–3.82; p ⬍ .001) to 3.41 days (95% CI 3.31–3.50; p ⬍ .001), and the relative risk of hospital mortality decreased to 0.77 (95% CI 0.68 – 0.87; p ⬍ .001).

DISCUSSION This study sought to understand the effects of implementation of a tele-ICU program on patient outcomes, including ICU LOS, hospital LOS, and hospital mortality in a five-hospital, mostly suburban, healthcare system diverse in size, demographics, and setting. The sample size of this study was ⬎24,000 ICU patients. This study extended the postimplementation analysis period to 3 yrs and showed a statistically significant improvement in both LOS and mortality. From the baseline year to year 3 postimplementation, the severity-adjusted hospital LOS was lowered from 11.86 days (95% CI 11.52– 12.21) to 10.16 days (95% CI 9.80 – 10.53), ICU LOS was lowered from 4.35 days (95% CI 4.22⫺4.49) to 3.80 days (95% CI 3.65–3.94), and the relative risk of hospital mortality decreased to 0.77 (95% CI 0.69 – 0.87). The overall conclusion of this analysis supports claim that tele-ICU programs can improve the quality of care and shorten LOSs. However, as other authors have noted, the deployment of a tele-ICU program is a complex process consisting of hundreds of discrete elements. The sum total of these elements is the creation of a new culture for management of ICU patients. Deployment of this new culture takes time to build trust and create the extended care team. The bedside nurses need time to accept the tele-ICU nurse as clinical partners and private practice physicians need to see that the presence of tele-ICU program will not alter consulting patterns but can facilitate broader application of evidence-based best practice medicine. Our tele-ICU physicians have been encouraged to be as proactive as possible. One common theme running through existing published data is that the greater the level of participation of the tele-ICU in the care of the patient, the more improved are the outcomes. This seems to be supported by the results of this study. Furthermore, the single greatest impediment to a fully proactive teleICU program is the lack of fully electronic medical records because the tele-ICU Crit Care Med 2012 Vol. 40, No. 2

team has only limited access to any bedside paper chart, which necessarily limits their ability to be proactive. In this study, postimplementation data were tracked for 3 yrs. The data indicate that the unadjusted hospital and ICU LOS increased slightly after the first year and then significantly decreased in year 3. Severity-adjusted hospital and ICU LOS significantly decreased in 2 and 3 yrs postimplementation. Severity-adjusted hospital mortality showed improvement each year over the 3 yrs postimplementation, but this improvement did not reach statistical significance until years 2 and 3. Hospital mortality can be a stronger reflection of ICU patient outcome than ICU mortality because it takes into account the status of the patient after they leave the ICU (e.g., patients transferred from the ICU with “do-not-resuscitate” orders) without excluding them from the study (14). There was an increase in the severity of illness based on the All Patient Refined-Diagnosis Related Groups score and Case Mix Index over the 3 yrs as shown in Table 1. It is not known whether this represents a secular trend in our health system or will prove to be a cyclic process. This increase in severity of illness does not account for the improvement in the calculated severity-adjusted mortality and LOS outcomes because there was improvement in the mortality and LOS outcomes even in the unadjusted raw data over the 3-yr period postimplementation. These improvements are magnified by the increase in the severity of illness over the same period of time. A limitation of the study was that Acute Physiology and Chronic Health Evaluation severity-adjusted data were not used in this study because the Acute Physiology and Chronic Health Evaluation data for the health system were not available for the baseline year. However, the Acute Physiology and Chronic Health Evaluation scores were available for years 1 through 3 after the tele-ICU program was implemented: year 1 (55.04), year 2 (56.57), and year 3 (56.50). So although All Patient Refined-Diagnosis Related Groups and Case Mix Index may not be ideal for severity of illness adjustment for ICU patients, the small secular upward trend seen in our All Patient RefinedDiagnosis Related Groups and Case Mix Index scores is matched by a similar trend in the Acute Physiology and Chronic Health Evaluation scores from years 1 through 3. This lends credence to the use of All Patient Refined-Diagnosis

Related Groups and Case Mix Index scores as surrogates for severity of illness for the ICU patients in this study. The location of patient death for smaller hospitals was unavailable. Therefore, we could not ascertain ICU mortality pre tele-ICU program implementation. The data in this study aggregate the outcomes from all five hospitals rather than reporting individual hospital results to maximize statistical significance. The subgroup analysis that was performed by excluding patients with hospital LOS ⬎50 days or ICU LOS ⬎30 days found statistically significant decreases in hospital LOS of 9%, ICU LOS of 8%, and did not alter the findings of shortened hospital and ICU LOS after the implementation of the tele-ICU program. The relative risk of hospital mortality decreased to 0.77 (95% CI 0.68 – 0.87; p ⬍ .001), unchanged from the primary analysis, which excluded no one.

CONCLUSIONS This retrospective observational study of mortality and LOS outcomes 3 yrs after deployment of a tele-ICU program included all cases admitted to an adult ICU and found statistically significant decreases in severity adjusted hospital LOS of 14.3%, ICU LOS of 12.6%, and relative risk of hospital mortality of 23%, respectively, in a multihospital healthcare system. These results show that tele-ICU programs can be used to extend and leverage the skills and resources of intensivist physicians and nurses to provide the level of ICU care recommended by the Leapfrog group for all ICU patients. Just as important, this study also shows that improvements in severity-adjusted mortality were achieved with shorter hospital and ICU LOS indicating that the overall cost of medical care is reduced.

REFERENCES 1. The Leapfrog Group for Patient Safety: Factsheet: ICU Physician Staffing. Available at: http://www.leapfroggroup.org. Accessed January 3, 2010 2. Thomas EJ, Lucke JF, Wueste L, et al: Association of telemedicine for remote monitoring of intensive care patients with mortality, complication, and length of stay. JAMA 2009; 302:2671–2678 3. Yoo EJ, Dudley RA: Evaluating telemedicine in the ICU. JAMA 2009; 302:2705–2706 4. Morrison JL, Cai Q, Davis N, et al: Clinical and economic outcomes of the electronic intensive care unit: Results from two com-

453

5.

6. 7.

8.

454

munity hospitals. Crit Care Med 2010; 38: 2– 8 Zawada ET, Herr P, Larson D, et al: Impact of an intensive care unit telemedicine program on a rural health care system. Postgrad Med 2009; 121:160 –170 Breslow MJ: Remote ICU care programs: Current status. J Crit Care 2007; 22:66 –76 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: 31–38 Rosenfeld BA, Dorman T, Breslow MJ, et al:

Intensive care unit telemedicine: Alternative paradigm for providing continuous intensivist care. Crit Care Med 2000; 28: 3925–3931 9. McCambridge M, Jones K, Paxton H, et al: Association of health information technology and teleintensivist coverage with decreased mortality and ventilator use in critically ill patients. Arch Intern Med 2010; 170:648 – 653 10. Lilly CM, Thomas EJ: Tele-ICU: Experience to date. J Intensive Care Med 2010; 25:16 –22 11. Fifer S, Evertt W, Adams M, et al: Critical Care, Critical Choices: The Case for TeleICUs in Intensive Care. Westboro, MA: New England Healthcare Institute and Massachu-

setts Technology Collaborative, December 2010 12. Lilly CM, Cody S, Zhao H, et al: Hospital mortality, length of stay, and preventable complications among critically ill patients before and after tele-ICU reengineering of critical care processes, JAMA 2011; 305:2175–2183 13. Lang TA, Secic M: How to Report Statistics in Medicine. Second Edition. Philadelphia, PA: American College of Physicians, 2006 14. Zimmerman JE, Kramer AA, McNair DS, et al: Acute Physiology And Chronic Health Evaluation (APACHE) IV: Hospital mortality assessment for today’s critically ill patients. Crit Care Med 2006; 34:1297–1310
