Bedside pulse oximeters with a clinical algorithm make economic sense in the intensive care unit

OBJECTIVE: To determine the effect on arterial blood gas (ABG) and hospital resource use by introducing a strategy of using bedside oximeters with a clinical algorithm, based on the argument that bedside pulse oximeters make economic sense in the intensive care unit (ICU) if safe patient oxygenation can be ensured at a lower cost than that of existing monitoring options. DESIGN: A before and after design was used to examine the consequences of a pulse oximeter at each bedside in the ICU along with a pulse oximeter clinical algorithm (POCA) describing use for titrating oxygen therapy and for performing ABG analysis. SETTING: A 19-bed multidisciplinary ICU with a six-bed extended ICU (EICU) available to function as a ‘step-down’ facility. PATIENTS: All patients admitted to the ICU/EICU over two 12-month periods were included. RESULTS: The strategy yielded a 31% reduction in the mean number of ABGs per patient after POCA (20.0 26.1 versus 13.8 16.7, mean SD; P<0.001) as well as a potential annual cost savings of $32,831. CONCLUSIONS: Bedside oximeters within the ICU, when used with explicit guidelines, reduce ABG use and result in hospital cost savings.

P ulse oximetry has been extensively adopted in the inten- sive care unit (ICU) over the past few years in the absence of definitive studies demonstrating efficacy (1,2).This is not unanticipated given the oximeter's ease of use and noninvasiveness, plus the widespread belief that the identification of serious arterial desaturation would be inappropriately delayed by using clinical signs or intermittent arterial blood gas (ABG) analysis.Though observational data from the operating room suggest a decreased incidence and duration of arterial desaturation with fewer adverse consequences (3,4), appropriately designed studies may never be performed within the ICU to answer the question of whether patient outcomes are improved by oximeters.Given the similarities between the two environments, it seems reasonable to assume that the results are applicable to the ICU.Nevertheless, even if it has not been firmly established that oximetry affects patient outcome positively, it has been argued that it is still worthwhile to use routine oximetry in the ICU if it accomplishes the expected tasks at a lower cost (5).For the potential user, it then becomes a question of determining the most economical manner of ensuring adequate oxygenation at the bedside.
Pulse oximetry has two advantages over clinical signs and intermittent ABG analysis: first, it can act as a warning system for detecting hypoxemia, with the potential to prevent the adverse patient consequences of arterial desaturation; and second, it can be used as an end-point for the titration of in-spired oxygen (F I O 2 ) and positive end-expiratory pressure (PEEP), hence serving as a response indicator.With these advantages in mind, we introduced a strategy of using bedside oximeters with a clinical algorithm in our ICU to determine the effect on ABG and hospital resource use.

Clinical setting:
The ICU at University Hospital is a 19-bed, multidisciplinary unit with house staff available in the unit 24 h a day.A dedicated intensive care physician is available at all times and coordinates patient management with other health care professionals.One nurse per patient is the usual staffing practice and a respiratory technologist is always available for ventilator and oxygen therapy management.An extended ICU (EICU) of six beds is available to function as a 'step-down' facility.All hospital patients requiring mechanical ventilation, excluding those in the operating room, are managed in these units.ABG analysis (Pulmonary Function Laboratory), biochemistry and hematology are performed outside the ICU/EICU by technicians, and samples require transport to the laboratories by ancillary staff.Study design: Clinical activity and ABG, electrolyte and hemoglobin use were recorded over two 12-month periods.From January 1, 1991 to December 31, 1991, five standalone oximeters were shared among the 25 beds, but without a clinical algorithm for expected use.Although ABG analysis was mandatory for all patients every morning it also fre- quently followed changes in ventilator or oxygen support.From August 1, 1992 to July 31, 1993, each bedside within the ICU had its own oximeter (Hewlett Packard Component Monitoring System with M1020A oxygen saturation modules, Hewlett Packard, Massachusetts) and the five shared units were used in the six-bed EICU; however, a pulse oximeter clinical algorithm (POCA) describing oximeter use for titrating oxygen therapy and for performing ABG analysis was included (Figure 1).It was believed that ABG and not other blood sampling frequencies would be influenced by POCA.

Oximetry accuracy and clinical algorithm development:
Oximeter accuracy was assessed before the new strategy was introduced into the usual ICU care.For a one-month period, oximeter saturation (SpO 2 ) of the M1020A oxygen saturation modules at the time of ABG sampling was recorded and compared with arterial saturation (SaO 2 ), measured using a six-wavelength co-oximeter (OSM3 Hemoximeter, Radiometer, Copenhagen, Denmark).The bias (SpO 2 -SaO 2 ) ± precision (SD) of the oximeters was determined to ensure an 'acceptable' lower limit for SpO 2 at the bedside (ie, an SpO 2 where 97.5% of the SaO 2 values were 90% or greater).A clinical algorithm describing oximeter use for ABG sampling and for titrating therapy (using this 'target' SpO 2 ) was constructed with ICU staff input, then taught and implemented (6).Staff compliance with ABG sampling was evaluated six months following introduction of the guidelines.Cost analysis: A cost minimization analysis of the new strategy, compared with the existing one, was undertaken from the viewpoint of the hospital (5).The capital cost of acquiring the pulse oximeter modules, above the cost of the bedside monitoring system, was treated as a straight-line depreciation cost per year over a period of five years, which was judged to be a conservative estimate of the clinical life of the oximeters.The annual operating cost of the oximeters was determined during the 12-month study and included actual expenses to replace damaged finger probes and cables.The annual cost reduction of the strategy was calculated using the 'controllable' cost of an ABG sample to the hospital, the expected number of ICU patients per year and the actual ABG reduction rate per patient.The controllable cost of an ABG sample to the hospital was $6.55 (April 1993).The components of this cost included direct labour worked ($3.68) and not-worked hours ($0.77), employee benefits ($0.64), direct materials ($0.26), direct overhead ($0.42) and direct equipment depreciation ($0.77).Statistical methods: Unpaired t tests (two-sided) were used to compare quantitative variables of the before and after groups.A c 2 test was used for noncontinuous variables.The Bonferroni correction for multiple comparisons was applied.P<0.05 was taken as significant.

Oximeter accuracy and clinical algorithm development:
Three-hundred and forty-eight SpO 2 /SaO 2 consecutive data sets from a one-month period were assessed to define oximetry accuracy.The bias ± SD was -0.08±1.28%for a SaO 2 range of 89 to 100%.The SaO 2 95% CI for a SpO 2 reading of  93% was determined to be 90.4 to 95.4%.In the clinical algorithm construction a target SpO 2 reading 93% or greater was used to ensure 'safe' arterial oxygenation (Figure 1).

Demographic characteristics before and after POCA:
The demographic characteristics of the patients studied during the two 12-month periods are shown in Tables 1 and 2. Patient characteristics were similar except that average patient age was higher after POCA introduction (58±15 versus 60±15 years, mean ± SD, P<0.05).

Cost analysis:
The capital cost of acquiring the oximeter modules was $66,192 but was treated as a straight-line depreciation cost of $13,239 (A) per year, over a period of five years (Table 3).An annual operating cost of $3,716 (B) was required to replace damaged finger probes and cables.Given 1100 ICU patients per year and a 6.2 ABG reduction per patient following POCA, an annual cost reduction in ABG testing of $44,671 (C) was possible.An additional annual cost reduction of $5,115 (D) was possible as a consequence of less use of gloves, syringes and blood gas kits.Hence, the total cost savings to the hospital for the 12-month period after POCA implementation was $32,831 ([C+D] -[A+B]).A greater cost savings was considered possible as an audit six months after POCA implementation revealed that the guidelines were correctly followed for ABG sampling only 65% of the time.

DISCUSSION
We found a 31% reduction in ABG use in our ICU following the introduction of bedside pulse oximeters when used with explicit guidelines.At the same time, there were no apparent differences in patient outcome, including mortality and duration of ventilatory support and ICU stay.With this strategy the hospital experienced a potential annual savings of $32,831.In fact, an actual savings for the hospital occurred as 1.5 fulltime equivalent positions were eliminated within the blood gas laboratory over the past three years, in part due to the reduced ABG workload (in that the ICU provided approximately 50% of the hospital ABG workload).Whether such savings would occur in other hospitals is uncertain and would likely be influenced by the characteristics of the patient population, the monitoring protocols in place, compliance with the guidelines, and whether ABG analysis is external or internal to the ICU.
Our findings agree with those of a recent randomized trial of continuous pulse oximetry in 35 patients after elective cardiac surgery (7).In that small study the ICU staff were instructed to obtain a mandatory ABG analysis each morning but to use the pulse oximetry data in lieu of ABG analysis whenever possible.The mean number of ABG analyses per  Our results, however, can be generalized to a multidisciplinary ICU and include an assessment of the economic benefits of the strategy.In contrast, two other recent studies did not demonstrate a change in ABG use following the introduction of oximeters within the ICU (8,9).One possible explanation for this discordance is that explicit instructions for use of the oximeter and for ABG sampling were not given to the ICU staff in those studies.Clinical algorithms designed to give rational indications for ABG analysis (10,11), without compromising patient outcome, are necessary to make oximeters economically attractive within the ICU.This study is limited by the before-after design.Although the patients studied in the two time periods were similar, we cannot be certain that other unmeasured confounders were present.Bias may also result from unappreciated changes in the ICU environment and from interventions other than oximetry being responsible for the ABG reduction.The fact that electrolyte and hemoglobin sampling frequency did not significantly change supports our argument that the intervention and not other factors was responsible for the reduction in ABG analysis.
In conclusion, use of oximeters within our ICU, in combination with explicit guidelines, was associated with a significant reduction in ABG analysis and resulted in hospital savings.

TABLE 3 Financial impact of acquiring POCA
SD) was comparable with our results (23.1±8.8 versus 20.0±26.1 without the oximetry strategy and 12.4±7.5 versus 13.8±16.7 with the oximetry strategy, respectively).
ABG Arterial blood gas; POCA Pulse oximetry with clinical algorithm; *Treated as a straight-line depreciation cost over a period of five years patient (±