Costs and effectiveness of tele-ICUs in reducing morbidity and mortality in intensive care units

A substantial amount of healthcare is delivered in intensive care units (ICUs); amounting to over 13 million ICU bed-days per year1. Intensive care is a costly and growing expense in the US healthcare system, with ICUs comprising between 5–10% of hospital beds, consuming 20–34% of all acute care resources, and a total of 1% of the US gross domestic product2. These costs are increasing due to more severely ill patients and new technologies3.

ICUs are sites where extremely ill patients are cared for and where morbidity and mortality are common4. For example, mortality rates in ICUs are usually ≥10%, 10–15% of ventilated patients develop pneumonia, 15–35% of ICU patients develop stress ulcers and upper gastrointestinal bleeding, 5% develop acute renal failure and 10–30% develop thromboembolic complications. Because of these high rates of resource use, morbidity and mortality, researchers have examined ways to improve ICU quality.

A powerful influence on the quality of ICU care is the presence of critical care physicians (intensivists) in the unit. High-intensity intensivist staffing (required intensivist consultation or closed ICU where intensivists see all patients) is associated with a 29% reduction in hospital mortality and a 49% reduction in ICU mortality5. These observations have led influential organisations to promote intensivist staffing6,7. The Leapfrog Group estimated that 53,000 lives could be saved annually by intensivist staffing in all ICUs8. However, only 10–15% of US hospitals have high-intensity staffing by intensivists, primarily due to an existing and worsening shortage of intensivists9. Thus, models of care must be developed that do not rely on full-time onsite intensivist staffing.

Telemedicine, a common form of health information technology, has been used to provide remote intensivist monitoring for ICUs. Remote ICUs connected via telemedicine technology (tele-ICUs) allow intensivists to simultaneously monitor more patients than possible by standard onsite care, and to extend intensivist care to patients in ICUs where intensivists would otherwise be unavailable, such as rural and small community hospitals. Furthermore, tele-ICUs may have decision support software to help identify subtle trends (rising creatinine, falling oxygen saturation) that need to be addressed to prevent complications.

However, the costs of acquiring, installing and operating tele-ICUs are considerable. The cost of setting up a tele-ICU system is about $2–$5 million10,11 and the average cost of hardware and software per ICU bed connected to the tele-ICU system has been estimated to be $48,00010. Operational costs are also high. In one study, the 6-month operating costs were $248,000 for expenses to operate the system plus $624,000 for physician staffing12. Hospital managers have reported annual operating costs of >$1.5 million10. In addition to the costs of installing and operating the tele-ICUs, some hospitals have to shoulder the additional costs of implementing electronic records systems or of updating existing electronic records systems to be compatible with tele-ICU software.

Because of the high costs, hospital managers must carefully weigh the evidence on the benefits of tele-ICUs before buying into a system. However, the evidence on the effectiveness of the tele-ICU in delivering on its promises is scant. To date, only two studies have been published evaluating tele-ICU programmes12,13. A report by the New England Healthcare Institute (NEHI)10 also evaluated tele-ICUs using published and unpublished data by early adopter hospitals, and a recent paper by Breslow11 reviewing the current status of tele-ICUs reported some clinical outcomes based on personal communications with hospital administrators. Only Breslow et al12 and the NEHI report10 discussed the costs of tele-ICU programmes.

A comprehensive published evaluation was carried out by the founders of the VISICU company, the for-profit company who developed the eICU, a proprietary tele-ICU programme12. VISICU is one of the three tele-ICU manufacturers marketing the programme to hospitals and is the leading vendor in the US. It was recently purchased by Phillips. The findings from that study are predominately displayed on the company's website in its attempts to promote their product. The design was a before and after study conducted in two adult ICUs in a single large urban tertiary care teaching hospital. The intervention was 19 hours/day of remote ICU monitoring using VISICU's eICU. In the pre-eICU period, 1,396 patients were studied, whilst 744 were studied in the post-eICU period. ICU demographics and patient acuity (measured by APACHE III) were the same in the pre and post periods. During the post-eICU, period there was a 27.1% reduction in hospital mortality for all ICU patients, and a 26.7% reduction in ICU mortality. ICU length of stay was shorter, costs per case were lower by 25%, and hospital revenues were higher (from increased volumes of patients treated in the post-eICU period). The authors conclude that the programme was associated with improved clinical outcomes and was cost saving since higher hospital revenues generated by increased ICU cases more than covered the programme costs.

However, a more careful examination of the results indicated that not all results were favourable: overall ICU mortality and surgical ICU mortality were not statistically significantly lower and hospital length of stay was unchanged overall and in the medical ICU. Finally, all economic outcomes were reported without confidence intervals, making it impossible to evaluate the statistical significance of the pre–post differences. The study suffers from further limitations14. As the authors acknowledge, despite efforts to control for patient mix and severity of illness in the pre and post periods, “the use of historical controls always raises concerns about possible changes in patient mix”. Despite checking for similarities in ICU admission criteria and APACHE III scores in the two periods, there may still remain unaccounted differences in patient mix between the two periods. In addition, as admitted by the authors, the actual basis for the changes observed in the post period compared to the pre period is unknown. It could be due to the attention paid by the hospital to the ICU in the period surrounding the implementation of the eICU that may have changed practices in a subtle way. Finally, the generalisation of the findings from one tertiary care teaching hospital that previously has intensivists on site to community hospitals or rural hospitals without any intensivist coverage is problematic.

Results from other early adopters of tele-ICU programmes have not been able to replicate improvements in clinical and economic outcomes of the same magnitude as Breslow et al12

Decreases in mortality ranged from none (Sutter Health System) to 10–15% (Leigh Valley Health System in Pennsylvania and Cornell Medical Center in New York)10. Preliminary data at Memorial Hermann in Houston, Texas, also provided mixed results. Length of stay decreased slightly in two ICUs, increased slightly in two others and was unchanged in one10. Mortality decreased at all five eICUs, although the changes were very minor and no confidence intervals were given. Zawada et al13 reports that mortality, ICU length of stay and hospital length of stay were significantly less than predicted by APACHE III severity scores in the period after the initiation of the eICU in a rural health system, Avera McKennan in South Dakota. However, the authors did not provide baseline data so it is unknown if the clinical outcomes were also more favourable than predicted by APACHE III before the introduction of the eICU. Breslow11, in the report on >30 sites that have implemented the eICU since 2003, reports decreases in severity-adjusted mortality of 13% at Maine Medical Center and 38% at OhioHealth. However, those reports were based on personal communications and no outcomes were reported for the remaining sites.

Information on the cost effectiveness of tele-ICUs is very sparse. Only Memorial Hermann System provided preliminary data on financial effects of the eICU and the economic outcomes were mixed10. The average cost per case decreased in two ICUs, increased in two ICUs and was unchanged in one. Revenue per case increased in all but one ICU.

It may be concluded that tele-ICUs have the potential to improve clinical outcomes but to a lower extent than first predicted. The information on cost is not yet available to assess the cost effectiveness of tele-ICUs. In addition, there are several barriers to the increased implementation of tele-ICUs. These include: physician, nurse and organisational acceptance of the eICU; the lack of insurance reimbursement for tele-ICUs, necessitating the hospital to cover all the costs; and the high start-up costs and operating costs that put tele-ICUs out of reach for hospitals with fewer resources or in rural areas, where the potential benefits may be greater.

While ICU telemedicine can be viewed as a means to providing intensivist coverage when intensivist onsite care is not possible, and may partly address the shortage of intensivists for onsite care in the US, more well-conducted and peer-reviewed clinical and economic evaluations of the programme are needed before recommending tele-ICUs broadly across US hospitals. These studies should follow sound study design and collect extensive clinical and economic outcomes in a variety of hospital types and with different care models. Given the difficulty of randomly assigning a large number of ICUs to tele-ICUs, carefully conducted pre–post design studies could be used, despite their limitations. It would also be very informative to compare a tele-ICU to other proven methods of quality improvement, especially ones that are less technology intensive15. These studies should control for ICU demographics and patient acuity, be carried out in a range of settings, and document clinical and economic outcomes in order to better assess tele-ICUs.

Acknowledgements

Declaration of interest: This research was supported by AHRQ grant #1RO1HS015234-01.