Abstract & Commentary

The Impact of Establishing a Regional Weaning Unit for Patients Requiring Prolonged Mechanical Ventilation

By Richard J. Wall, MD, MPH, Pulmonary Critical Care & Sleep Disorders Medicine, Southlake Clinic, Valley Medical Center, Renton, WA, is Associate Editor for Critical Care Alert.

Dr. Wall reports no financial relationship to this field of study.

SYNOPSIS: In this modeling study of the possible effects of establishing a regional weaning facility in an area currently without such resources, whether cost savings would likely be achieved depended on numerous factors, including how much care at the weaning facility costs in relation to care in the ICU.

SOURCE: Lone NI, Walsh TS. Prolonged mechanical ventilation in critically ill patients: Epidemiology, outcomes and modelling the potential cost consequences of establishing a regional weaning unit. Crit Care 2011;15:R102.

Although most critically ill patients require only short periods of respiratory support, a minority require prolonged mechanical ventilation (PMV). When these patients linger in the intensive care unit (ICU), they create a challenge for acute care hospitals because they tie up beds and ventilators, reduce hospital throughput, and drive up health care costs. In some U.S. cities, these patients are ultimately transferred to long-term acute care (LTAC) facilities for ventilator weaning. In other cities, however, LTAC facilities are not available. In the United Kingdom (UK), LTAC facilities do not exist.

The authors of this study used a large comprehensive database from southeast Scotland to model whether a hypothetical ventilator weaning unit would be beneficial to the local community. The main objectives of this study were to: 1) establish the incidence of PMV in a large UK region; 2) examine the characteristics and outcomes of PMV patients; and 3) model the potential impact on costs and outcomes of establishing a regional ventilator weaning unit. PMV was defined as requiring mechanical ventilation (MV) ≥ 21 days. The authors performed a retrospective cohort study using a prospectively collected anonymous dataset from a community of 900,000 who are served by three adult hospitals. Each hospital has a "closed" adult general mixed medical/surgical ICU with intensivist staff. The hospitals are essentially managed as a single organization. The dataset captures every admission episode. It has been previously validated and has 94% accuracy.

The authors ran a variety of sensitivity analyses, a technique wherein they modeled various possible scenarios and inputs. For example, they ran their model using two different definitions of PMV. In the first definition, patients were mechanically ventilated for ≥ 21 consecutive days (with a required minimum of 6 hours daily). In the second definition, patients simply had to be ventilated for 21+ cumulative days during the hospitalization. When calculating incidence, the authors used different denominator definitions. The first denominator included every ICU admission irrespective of MV status. The second denominator only included ICU admissions requiring MV at some point during their stay.

The authors ran additional sensitivity analyses in which they varied aspects of the new hypothetical weaning unit. They varied the unit's admission criteria. Some units accepted patients on renal replacement therapy (RRT), others did not. Some units accepted patients within 48 hours after discontinuing pressors, others required 7 days of hemodynamic stability. This resulted in four different units, one which took patients on RRT 48 hours after weaning pressors (least stable), one that refused patients on RRT and didn't take patients until 7 days after pressors (most stable), and two units in between. They varied the cost of a weaning bed between 50% and 100% of an acute ICU bed. They estimated likely refusal rates depending on both bed availability and the admission criteria. They used these data to determine the optimal number of beds that made most financial sense for the community.

Overall, they examined 7848 admission episodes over a 5-year period. The incidence of PMV ranged between 4.4 and 6.3 per 100 ICU admissions. These patients utilized almost one-third of all ICU bed days. The most common diagnoses for all ICU patients were pneumonia, sepsis, and trauma. The diagnoses most associated with PMV were Guillain-Barré syndrome, pancreatitis, acute respiratory distress syndrome, pneumonia, and sepsis. In general, 8%-10% of ICU beds were occupied by PMV patients who could potentially be transferred. Other characteristics of PMV and non-PMV patients are listed in the table.

Table. Characteristics and Outcomes of ICU Patients Requiring and Not Requiring Prolonged Mechanical Ventilation (PMV)

PMV

Non-PMV

P value

Age, mean (years)

59.6

56.9

0.001

Female, %

42

43

0.86

APACHE II

21

18.8

< 0.001

Surgical, %

19

25

< 0.001

MV on day 1, %

92

66

< 0.001

PaO2/FiO2, median (mmHg)

139

227

< 0.001

Tracheostomy during admit, %

63

6

< 0.001

Hospital mortality, %

40

34

0.02

ICU mortality, %

26

23

0.23

ICU length of stay, median (days)

33

2

Hospital length of stay after ICU discharge, median (days)

17

7

< 0.001

In the best-case scenario, establishing a regional weaning unit resulted in annual cost savings of approximately $600,000. In the worst-case scenario, the unit costs $56,000 annually. In addition, once a weaning unit bed reached 70% of the cost of an ICU bed, it was no longer cost saving. Of note, costs in the UK are not necessarily equivalent to the U.S. health care system.

COMMENTARY

While prospective randomized trials are excellent methods for rigorously testing discrete interventions, they are not well suited to real-world scenarios that typically involve multiple, interacting, and uncertain factors. Although observational studies using administrative data can generate hypotheses, they rarely can answer questions about the "next best step." In such cases, modeling the various hypothetical situations is often a much more practical approach.

In this study, PMV patients consumed a substantial amount of a region's health care resources. To address this issue, the authors modeled various scenarios to determine whether a community should build a weaning unit. I'm sure one could find faults with this study. For example, the authors did not account for the potential impact of the unit on patient outcomes, namely duration of MV or mortality. However, I think the authors nonetheless did a nice job. They used a preexisting database to minimize selection bias, and they performed numerous sensitivity analyses to address the uncertainty we all face in our practices. As providers and administrators are increasingly forced to make large sweeping decisions while also containing costs, I suspect we will begin to see more such studies.