Article Text
PDF
Abstract
Objective To determine if time to disposition decisions for emergency department (ED) patients can be reduced when blood tests are processed using point-of-care (POC) devices and to conduct a cost-effectiveness analysis of POC compared with laboratory testing.
Methods This randomised trial enrolled adults suspected of an acute coronary syndrome or presenting with conditions considered to only require blood tests available by POC. Participants were randomised to have blood tests processed by POC or laboratory. Outcomes measured were time to disposition decision and ED length-of-stay (LOS). The cost-effectiveness analysis calculated the total and mean costs per ED presentation, as well as total and mean benefits in time saved to disposition decision.
Results There were 410 POC participants and 401 controls. The mean times to a disposition decision for POC versus controls were 3.24 and 3.50 h respectively, a difference of 7.6% (95% CI 0.4% to 14.3%, p=0.04), and 4.32 and 4.52 h respectively for ED LOS, a difference of 4.4% (95% CI −2.7% to 11.0%, p=0.21). Improved processing time was greatest for participants enrolled by senior staff with a reduction in time to disposition decision of 19.1% (95% CI 7.3% to 29.4%, p<0.01) and ED LOS of 15.6% (95% CI 4.9% to 25.2%, p=0.01). Mean pathology costs were $12 higher in the POC group (95% CI $7 to $18) and the incremental cost-effectiveness ratio was $113 per hour saved in time to disposition decision for POC compared with standard laboratory testing.
Conclusions Small improvements in disposition decision time were achieved with POC testing for a moderate increase in cost. Greatest benefit may be achieved when POC is targeted to senior medical staff.
- cost effectiveness
- diagnosis
- emergency care systems, efficiency
Statistics from Altmetric.com
Introduction
Reducing the time that patients stay in the emergency department (ED) is a desirable goal to reduce overcrowding, improve patient flow, improve patient satisfaction and reduce morbidity and mortality.1–5 Australian EDs also must comply with the National Emergency Access Target of 4 h for completion of ED management.6 Point-of-care (POC) testing, defined as laboratory testing near a patient location with rapid availability of results, has the potential to reduce ED length-of-stay (LOS) through short turn-around times allowing clinical decisions to be made earlier.
In this study, we tested the hypothesis that short turn-around times of POC devices would translate to improved patient processing times. The primary aim of this study was to determine if the time to make an admission or discharge decision (hereafter referred to as a disposition decision) could be reduced with common blood tests being available by POC testing in the ED. Secondary aims were to investigate improvements in processing times on several patient subgroups, and to perform a cost-effectiveness analysis of POC compared with central laboratory testing from an Australian health system perspective.
Methods
Study design and setting
This study was an open, parallel arm, randomised trial conducted in the ED of a tertiary referral and level 1 trauma centre located in Sydney, New South Wales, Australia, over a 6-month period from December 2011 to May 2012. The ED has approximately 65 000 presentations a year, and pathology services are available 24 h a day. Permission for the study was granted by the South Eastern Sydney and Illawarra Area Health Service (central network) Human Research Ethics Committee and registered with the Australia and New Zealand Clinical Trials Registry (ANZCTR#12611001228976). Funding for this study was provided by a grant from the NSW Department of Health (Ministerial Taskforce on Emergency Care ‘Taking the pressure of public hospitals’ project grants 2011/12) and from the study hospital.
Selection of participants
Patients presenting to the ED were eligible if they were ≥18 years of age and fulfilled the requirements for either of the following two groups. The first group were patients suspected of having an acute coronary syndrome (ACS group). Those with acute ST-elevation myocardial infarction were excluded. The second group (general group) were patients whom the enrolling staff member thought would only need blood tests from the selection available by POC to complete assessment and management. The POC blood tests available were creatine, electrolytes, glucose, calcium, haemoglobin, Troponin-T, D-Dimer, β-HCG and INR. The POC devices used were the Radiometer ABL-800 FLEX blood-gas analyser, Radiometer AQT-90 FLEX and the Roche CoaguChek XS-PRO. Patients who presented more than once to the ED within the study period could be re-enrolled.
Participants were enrolled by nurses, nurse-practitioners and doctors. Nurses enrolled participants as it was routine practice in this ED for nurses to ‘fast-track’ blood tests for patients waiting to be seen. Study recruitment was driven by regular education at staff meetings and encouragement by research staff in the department.
In order to determine an estimate of the proportion of patients appropriate for assessment by POC and to assist sample size determination, a pilot study was first conducted over 2 months prior to the randomised trial. Clinicians completed a survey determining inclusion/exclusion criteria for each adult patient who presented to the ED.
Method of randomisation
Patient consent was waived by the ethics committee. Participants meeting the inclusion/exclusion criteria were randomly allocated by opening sequentially numbered, sealed, opaque envelopes which contained the study allocation. Randomisation was stratified according to clinical group (ACS or general). To ensure balanced numbers of participants in each arm of the study block randomisation was used, with blocks of variable size to prevent prediction of the allocation sequence in this non-blinded study. The randomisation sequence was created using a computerised random number generator.
Interventions
For participants allocated to the intervention, in the general group all blood tests were processed in the ED using POC devices. For participants in the ACS group, only troponin was processed by POC, and other blood tests if required were sent to the central laboratory for processing. This was because we considered troponin to be the critical blood test for making a disposition decision in patients with an ACS, while other tests often are requested for ‘baseline’ measurement and infrequently influence disposition decisions. Turn-around times for the POC devices ranged from 2 to 22 min. For control participants (ACS and general groups), all blood tests were sent to the central hospital pathology service for processing. Turn-around times for laboratory tests take between 30 min and 2 h.
Following this initial set of testing, any additional pathology required was performed in the central laboratory.
Outcome measures
The primary outcome was the time from ED arrival to disposition decision. This was chosen as the primary outcome (rather than ED LOS) as delays in accessing inpatient ward beds and ultimate transfer out of ED may mask a benefit in patient processing time.
The secondary outcomes were ED LOS for the whole study population, time to disposition decision and ED LOS for the following subgroups: diagnostic group (ACS or general), disposition (discharged home, admitted to the ward, admitted to the Emergency Medicine Unit which is an ED short stay ward) and seniority of enrolling staff. A cost-effectiveness analysis of POC testing compared with central laboratory testing was conducted.
Methods and measurements
The staff member enrolling a participant entered diagnostic information on a data collection form. For those in the general group this was the provisional diagnosis, while those in the ACS group were stratified to a low, intermediate or high risk category. Demographic data and times for the primary and secondary outcomes were obtained from the ED computer management system in which the times of all significant events in the patient journey are entered. The time of admission decision was defined as the time that the clinician notified the nurse in charge to book a bed following patient acceptance by an admitting team. For patients sent home, the discharge decision time was the departure ready time as entered by the clinician into the ED computer management system.
Analysis
The required sample size was determined using an estimate of the mean and SD of the disposition decision time for the study population. This was obtained from the patients in the pilot study considered appropriate for assessment by POC. The mean time to a disposition decision was 4.4 h. POC devices were expected to provide a processing benefit of approximately 40 min (15%). A benefit of this magnitude would also be clinically relevant as this would move the average time to a disposition decision below the 4 h National Emergency Access Target.6 Using a power of 80% and an α-level of 0.05, 450 participants were needed. We required this study to be powered for subgroup analysis, in particular the ACS group. The pilot study demonstrated the ACS group and the whole study group to have a similar mean and SD, and so we determined to stop the study once 450 participants had been enrolled in the ACS group.
The primary analysis was by intention-to-treat. The outcome measures of time to disposition decision and ED LOS were positively skewed; therefore, the data were first transformed to a normal distribution by taking the natural logarithm and the analysis was conducted by comparing the means of the natural logarithm of these outcomes using linear regression. The differences in time between study groups are presented as percentage reductions in the means of the logarithmically transformed data, while the average times presented are the geometric means, which are the means of the logarithmically transformed data back-transformed using the exponential. A random effect model was included to adjust for repeated presentations over the period of the study. This analysis was conducted in Stata 12 (StataCorp LP, Texas, USA).
Economic evaluation
The cost-effectiveness analysis calculated total and mean costs per ED presentation, as well as total and mean benefits in time saved to disposition decision. All pathology and radiology tests from the time of arrival to the time of disposition decision were obtained from the pathology and radiology databases, respectively. This included any central laboratory troponins for patients randomised to the POC group. Direct unit costs from the pathology service provider and hospital casemix data were obtained for each pathology and radiology diagnostic test. Indirect costs for capital equipment (ie, POC analysers) were calculated using the equivalent annual cost method.7 A weighted average clinical staff time for POC and laboratory test processing was derived from a time-in-motion study with 25 consecutive ED presentations. The differences between costs in the two groups and the 95% CIs were then calculated. Volumes of resources and costs are reported as mean values with SDs and as mean differences with 95% CIs. Discounting was not applied. The arithmetic mean of the disposition decision time (rather than the geometric mean) was used in the calculation of an incremental cost-effectiveness ratio (ICER) for POC compared with central laboratory testing as this is the standard methodology used for economic evaluations. The ICER was calculated using the following formula: (mean cost of POC − mean cost of control)/(mean effect of POC − mean effect of control). Non-parametric bootstrapping was employed for a 95% CI around the ICER. The economic analysis was conducted in Excel 2007 (Microsoft, USA).
Results
Characteristics of study subjects
From the pilot study, 44% of adult ED presentations were eligible for the study. Over the randomised study period, 20 292 adult patients presented to the ED, giving an estimated number of eligible participants of 8928. There were 881 presentations enrolled (10% of the estimated eligible population). A total of 66 enrolment forms were not returned preventing identification of the participant. Two participants were excluded as they were enrolled in both arms of the study for the same presentation. This left 811 presentations available for the intention-to-treat analysis. There were 410 presentations randomised to POC and 401 to the control arm of the study (figure 1). In all, 19 participants presented and were enrolled more than once during the study: 17 participants had two observations, one participant had three observations and one participant had five observations. The trial was balanced with respect to baseline characteristics (table 1).
Baseline characteristics by randomisation group
Participant flow chart.
Main results
For the primary outcome, POC testing reduced the time to a disposition decision from a mean of 3.50 to 3.24 h, a difference of 0.26 h or 7.6% (95% CI 0.4% to 14.3%, p=0.04), with trends toward shorter decision making times in all subgroups analysed (table 2). There was a reduction in ED LOS of 4.4%, from 4.52 to 4.32 h. This difference was not statistically significant (95% CI −2.7% to 11.0%, p=0.21). There were trends toward shorter ED LOS in all but one of the subgroups analysed (table 3). The improvement in patient processing times were greatest for those patients enrolled by senior staff (consultants and registrars), with a reduction in the time to a disposition decision of 19.1% (95% CI 7.3% to 29.4%, p<0.01) and ED LOS of 15.6% (95% CI 4.9% to 25.2%, p=0.01). Testing for interaction was performed to determine if there was evidence that the effect of the intervention on processing time depended on the seniority of the enrolling staff (test for interaction p=0.06 and p=0.21 for disposition decision time and ED LOS, respectively).
Time from arrival to disposition decision
Length-of-stay in the emergency department
Economic outcomes
The calculation of the ICER was based on the arithmetic mean time to a disposition decision (this was 3.78 h in the POC group and 3.99 h for the control group, a difference of 0.21 h (13 min) in favour of POC testing).
Resource utilisation
Table 4 shows the utilisation of healthcare resources per ED presentation according to the study group allocation. The number of pathology, radiology and cardiology tests per presentation did not significantly differ between the groups. The ED staff time for hands-on pathology processing was significantly shorter in the POC group compared with the control group (1.34 min, 95% CI 1.22 to 1.46).
Healthcare costs
Healthcare costs per ED presentation are reported in table 4. For pathology costs, there was no significant difference in the mean volume of tests; however, the mean cost per patient was $12 higher in the POC group (95% CI $7 to $18). The overall healthcare costs per ED presentation were $174±$157 in the POC group and $150±$129 in the control group, a net difference of $24 (95% CI $4 to $44) in favour of the control group.
Mean use of healthcare resources and mean total healthcare costs per presentation for time to decision according to random allocation
Cost-effectiveness
The difference in mean costs ($24 rounded) divided by the difference in mean benefits (0.21 h saved) resulted in an ICER of $113 per hour saved in time to a disposition decision for POC compared with standard laboratory testing. This means an additional $113 was spent per patient randomised to the POC group to achieve a time saving of 1 h. Sensitivity analyses demonstrate that this finding is robust with the majority of POC presentations having both higher costs and higher benefits (ie, saved time) compared with the control group (see highlighted section, figure 2).
Cost-effectiveness plane showing 1000 bootstrap replicates of incremental cost per hour saved (time to disposition decision) for point-of-care (POC) versus central laboratory testing. Overall, 70% of replicates shown in highlighted section were in the north-east quadrant of the plane, demonstrating that in the majority of cases POC had both higher costs and higher effects (ie, saved time) compared with the control group. NE=north-east quadrant where interventions are more expensive, but more effective. SE=south-east quadrant where interventions are less expensive and more effective. SW=south-west quadrant where interventions are less expensive but less effective. NW=north-west quadrant where interventions are more expensive and less effective.
Figure 3 shows a cost-effectiveness acceptability curve of POC testing at different willingness to pay levels for 1 h of time saved to disposition decision. This curve indicates the probability that an intervention is cost-effective compared with its alternative, given the data, for a range of values up to a maximum acceptable ceiling ratio. If the Australian health system were willing to pay $120 or higher in order to save 1 h of time in the ED, then our data suggest POC testing has an 80% probability of being cost-effective.
Cost-effectiveness acceptability curve for point-of-care (POC) at different willingness to pay levels. Willingness to pay for 1 h of emergency department (ED) decision time saved (AU$).
Discussion
In this randomised trial, we were able to demonstrate a small reduction in time to reach a disposition decision but no measurable improvement in ED LOS among participants randomised to POC testing. While the improvement in the primary outcome was statistically significant, we had prespecified the minimum clinically important reduction to be 15%. In the subgroup analysis, there were trends toward small improvements in processing times with the exception of participants enrolled by senior staff where the outcomes were considerably better and exceeded our minimum clinically important reduction.
Previous studies on POC devices have had conflicting results. A before/after study design8 using POC troponin testing for ACS demonstrated shorter ED LOS and time to admission decisions, a quasi-randomised trial9 demonstrate only a trend to shorter ED LOS, while two randomised trials failed to demonstrate a benefit.10 ,11 In studies of POC testing using machines that perform a variety of blood tests, two before/after studies demonstrated shorter ED LOS12 ,13 while a third did not,14 a small randomised trial found a shorter ED LOS,15 but a large randomised trial was unable to demonstrate a difference in ED LOS, hospital LOS, admission rates or mortality.16 However, if only the randomised trials are considered, these are largely consistent with our study in finding minimal impact of POC devices on patient processing time.
There are a number of reasons why only small improvements were demonstrated. Physicians seeing several patients simultaneously may have got caught up in clinical care delaying action on an available result. Participants enrolled by nurses were having tests ‘fast-tracked’ prior to being seen by a doctor, and so the benefit of POC testing may have been nullified by prolonged waiting times. It is also possible that some clinicians had a ‘lack of faith’ in the accuracy of POC results and deferred decisions until repeat laboratory tests were available. Possibly the most important factor was that a junior doctor's ability to make decisions could be influenced more by the time taken to obtain a history, examination and consultation rather than the turn-around time of a test. As the majority of patients were enrolled by junior doctors, this would have had a strong influence towards a null effect. This is supported by the subgroup analysis of processing times according to the seniority of the clinician.
Our study has assessed the utility of POC testing simply as an alternate method of processing pathology specimens. A recent randomised study of ACS patients compared standard care with an intervention of a panel of POC cardiac biomarkers combined with a structured clinical pathway, demonstrating a greater proportion of patients being discharged from ED in the intervention arm.17 It is possible that benefits of POC devices could be realised when the faster processing times are combined with structured clinical pathways.
These are important findings for departments considering the implementation of POC devices, particularly for tertiary EDs with large numbers of junior staff, and laboratory services 24 h a day. Our results would indicate that in these settings only small benefits could be expected. However, if the use was targeted to senior staff with the experience to make rapid decisions, clinically relevant benefits could potentially be achieved. It is also important to emphasise the importance of system improvements to ensure flow of patients out of the ED as any improvements in efficiency within the ED will be quickly lost, an effect echoed in our results with minimal improvement seen in ED LOS compared with a disposition decision.
Despite these small improvements in processing time, the increased cost of POC testing to the Australian healthcare system is relatively small for the benefit of an hour saved in disposition decisions. To put our price of $113 into perspective, Australian EDs are funded based on their activity with this ED allocated $505 per patient treated.18 Participants enrolled in this study had an average LOS of approximately 4.5 h. At face value, this would equate to $112 per patient per hour of their stay, which would suggest that POC testing is a cost neutral intervention if time saved in decision making translated to time saved in the ED. Based on the estimate that 44% of adult presentations were eligible for POC testing, over 1 year this would equate to 23 496 eligible presentations. To fully implement POC testing, we estimate the additional cost over 1 year to be: 23 496 presentations × $12 each for pathology costs=$281 952 at current prices.
Limitations
It would be expected that outside the context of a study the benefits of POC testing are likely to be smaller as staff tend to be more motivated and pay more attention to participants than they would in the ‘real world’. In addition, this performance bias may affect the POC arm of the study to a greater extent given this study was not blinded. It was not possible to blind staff to the study allocation because POC testing involves the ED staff in sample processing and so it was important that all normal procedures that occur with usual use of POC devices be preserved. Another potential source of bias was the loss of 7% of enrolment forms, without which the participant who was enrolled could not be identified. If this was a random event this would be unlikely to introduce bias, but if there was a systematic reason such as staff discarding the form if they received a particular allocation, this could introduce important bias. Given the similar proportion of missing forms in each arm of the study and the balance in baseline characteristics, this is likely to have been a random event.
We estimated that approximately 10% of eligible patients were enrolled into the study. This could have introduced systematic bias if, for example, patients who were enrolled were simpler cases in which POC benefits could be directly translated into disposition decision benefits, while those not enrolled were more complex with disposition decisions influenced by factors other than test processing time.
There may also be inaccuracies in the measurement of the processing times as this relied on staff entering the time on the computer management system. When staff were diverted by more urgent priorities the time recorded may have been longer.
The generalisability of results from single centre study is always a concern as the patients or conditions unique to a particular institution may reduce the relevance when extrapolated to other sites. However, the patient population targeted by this study tended to be of lower acuity, with single system problems commonly seen in all EDs, and so our results should be relevant to a broad range of ED environments.
With regard to the economic evaluation, the cost-effectiveness results may be limited in their generalisability to tertiary EDs supported by pathology services with similar costs. A 3-month follow-up study of ACS patients is currently underway to assess the longer-tem effectiveness and cost-effectiveness of POC testing.
Conclusions
While POC devices have the potential to reduce patient processing times, in the subgroup of patients suitable for POC testing we could only demonstrate small improvements in the time to make a disposition decision, and no measurable impact on ED LOS. The small increase in costs does not appear to support more widespread adoption of POC pathology testing in the ED. This might be reconsidered when a fuller range of testing is available, and costs become more equivalent to those in a central laboratory. Restricting the use of POC testing to senior clinicians with the experience and ability to make rapid decisions is a potential niche where POC testing could produce important benefits, and warrants further investigation.
References
Footnotes
-
Funding Funding for this study was provided by a grant from the NSW Department of Health (Ministerial Taskforce on Emergency Care ‘Taking the pressure of public hospitals’ project grants 2011/12) and from the St George Hospital.
-
Competing interests None.
-
Ethics approval South Eastern Sydney and Illawarra Area Health Service (central network) Human Research Ethics Committee.
-
Provenance and peer review Not commissioned; internally peer reviewed.
-
Author contribution SEA and ACFC conceived the study and designed the trial. All authors contributed to revision/refinement of the study design. ACFC, AA and RW obtained research funding. SEA, EW, AA, DH supervised the conduct of the trial and data collection. EW managed the data. PK and RM provided statistical advice on study design and analysed the data. RW provided expertise and pathology related economic data for the economic evaluation. SEA drafted the manuscript, and all authors contributed substantially to its revision. SEA takes responsibility for the paper as a whole.