Method

A computerised search of the literature was undertaken using Medline. Articles were searched for the textwords “chest pain observation”, “chest pain evaluation”, or “chest pain assessment”. The medical subheading (MeSH term) “chest pain” was also searched in combination with MeSH terms: “emergencies”, “observation”, “myocardial ischaemia”, “unstable angina”, and “myocardial infarction (diagnosis)”. Any article that reported costs or outcomes for patients managed on a CPOU was reviewed. The bibliography of each article was searched for related citations. Articles relating to chest pain clinics^{12, 13} (rapid access outpatient cardiology services) were not included. Although similar, these services have a different source of referral and tackle a different clinical problem to the CPOU.

The question of effectiveness was addressed in two ways. Firstly studies were selected that compared outcomes of CPOU management with those of routine patient care. The quality of these studies was assessed against standard criteria covering reporting, statistical analysis, internal validity (bias and confounding), and external validity (generalisability).^{14, 15} Particular attention was directed at determining how subjects were selected and allocated to intervention (CPOU) and control (routine care) groups, how the controls were chosen, how follow up was performed for each group, the completeness of follow up, and the range of outcomes examined.

Secondly, studies were selected that made no comparison but simply described outcomes of CPOU patients. Quality was assessed by determining how subjects were selected, the nature and completeness of follow up, and the range of outcomes examined. Although descriptive studies cannot demonstrate effectiveness on their own, they may be helpful in adding to a body of knowledge that a technology can be safely applied in a variety of settings.

To address the economic question studies were selected that compared CPOU costs with those of a comparison group. Criteria relating to the reporting of economic evaluations have recently been published.¹⁶ Unfortunately most of the literature relating to CPOUs were submitted for publication before this and the reports of economic data are, by comparison, poor. Hence rigorous examination of quality is not possible. Quality assessment was therefore focused upon essential criteria for internal and external validity of the economic comparison, such as the method of allocation to intervention and control groups, the choice of controls, the range of costs included and the costing technique used. As most of the data obtained was observational and non-randomised, no attempt to perform a meta-analysis was made. Instead the various estimates of costs were examined for heterogeneity and explanations for any differences sought.

Results

All the studies found were from the United States. Six studies compared outcomes for patients admitted to a CPOU to a control group.^17–22 One of these, which compared outcomes principally in terms of patient satisfaction,²² ran alongside a randomised trial of cost effectiveness¹⁸ and will be discussed separately. The results of the five remaining studies are summarised in table 1.

Reporting was adequate for all studies. The objectives, interventions, outcomes, main findings, and patient characteristics were well described. All studies included appropriate statistical analysis except that of Kerns et al.²¹ Regarding validity, blinding of patients and carers was inevitably absent from all the studies and represents a potential source of bias. Three trials were randomised and are therefore the most likely to be valid.^17–19 The non-randomised study by Gaspoz et al made appropriate adjustment for confounding,²⁰ but only random allocation can take into account the influence of unknown confounders. No such adjustment was made by Kerns et al and taking into account the small number and lack of statistical analysis,²¹ this study can only be considered to represent pilot data. Only the study by Farkouh et al described the full details of patients excluded from the trial.¹⁷ Without such details it is impossible to determine whether the trial population is representative of all low risk patients with chest pain and we must be cautious about applying findings to other patient groups.

The AMI rate in the study groups varied from zero to 4.9% and did not differ significantly between CPOU and control groups in any study. Baseline characteristics, in terms of age, sex, type of pain, risk factors, and history of coronary artery disease, did not differ significantly between CPOU and control groups in any of the randomised trials.^17–19 In the study by Gaspoz et al the control patients were significantly more likely to be male, have recurrent or atypical pain, have abnormal ECGs, and have a history of ischaemic heart disease.²⁰ No statistical analysis of baseline characteristics was carried out by Kerns et al.²¹

Six studies were found that described outcomes for CPOU subjects without comparison with a control group,^23–25 or used a control group for cost analysis only.^26–28 Hence there were a total of 11 studies reporting follow up of a cohort of CPOU patients. These are outlined in table 2. All studies excluded subjects with an ischaemic ECG and selected those at low risk of AMI. Often this selection involved a subjective element of physician judgment. Most studies achieved high follow up rates.^{17–20, 24, 26, 28} This follow up was typically done by mail or telephone and was therefore adequate to exclude major morbidity but not “missed AMI” by the criteria outlined above.

Nine studies comparing costs of a CPOU to routine care were found,^{18–21, 26, 28–31} two of which were only in abstract form.^{30, 31} Two of the studies consisted of two separate comparisons so there were a total of 11 comparisons to review.^{19, 29} These are outlined in table 3. One other study was found that compared resource use but no costs.³² Data from this institution, including costs, has been published elsewhere.³⁰

Table 4 outlines the diagnostic tests used in each of the CPOU protocols described in the literature.

Discussion

The effectiveness of the CPOU has been investigated by five comparative studies.^17–21 Despite differences in inclusion criteria, method of allocation to treatment, and follow up there is broad similarity in outcomes. No significant difference in any objective outcome measure has been demonstrated. The main threat to the validity of this conclusion is the small number of deaths, AMI, and complications in these low risk subjects. A larger trial might be needed to detect a small difference in these outcomes, but it appears that the CPOU does not markedly affect hard outcome measures.

More subjective outcomes should be interpreted with caution in view of the inability of researchers to institute blinding. The only significant difference in outcome detected was an increase in diagnostic certainty after CPOU assessment.¹⁸ Two studies recorded a non-significant trend towards higher reattendance rates among CPOU patients (Farkouh et al¹⁷: 8.0% v 4.2% and Roberts et al¹⁸: 6.1% v 4.5%). These are measures of processes of care and their value to the patient is debatable.

A further study by Rydman et al reported improved patient satisfaction among patients referred to a CPOU when compared with routine care,²² but this may be an example of patient preference bias.³³ Being a randomised controlled trial of a new intervention, patients who have a preference for CPOU care can only obtain it by entering the trial and risking disappointment if they are randomised to the control. While those with a preference for routine care can be sure to obtain their preference by refusing consent. Unless patients have no preferences recruitment will be biased towards those who prefer CPOU care.

It should be noted that all the aforementioned studies compare CPOU patients with those admitted. A more appropriate comparison would also include patients discharged after initial emergency department assessment so as to report the proportion of AMIs discharged. Such a study, particularly if randomised, would present significant logistic and ethical problems but must be considered the only way of providing definitive proof of the relative effectiveness of the CPOU.

Descriptive and comparative studies have now reported large numbers of patients receiving CPOU assessment.^{17–21, 23–28} Follow up by telephone or mail is reasonably comprehensive and, supported by searches of death registries, is adequate to ensure that significant symptomatic pathology is not being missed. Death and complication rates do not exceed those expected for the study population and it is reasonable to conclude that the CPOU is a safe management strategy for low risk patients. Protocols consisting of continuous ST monitoring, creatine kinase MB isoenzyme measurement, and exercise stress testing seem to be the standard practice. Most result in discharge of around 80% of CPOU patients. The low discharge rates seen in the studies by Farkouh et al¹⁷ and Roberts et al¹⁸ probably occur because these protocols stipulate admission of those with inconclusive exercise testing.

It is tempting to compare the results of CPOU follow up with the previously reported rates of missed AMI.^{1, 2} Indeed, this has been done to conclude that the CPOU reduces inadvertent discharge of AMI.²⁵ This conclusion should be resisted for two important reasons. Firstly, none of the studies of CPOUs report testing for missed AMI with clinical, ECG, and enzyme assessment at 48–72 hours after discharge. Without such rigorous follow up it is impossible to tell if an equivalent number of AMIs are missed. Secondly, estimates of missed AMI rates predate many changes in emergency management of chest pain that may have improved or increased the caution with which patients with chest pain are managed. The use of historical controls is recognised to exaggerate the effects of new interventions,^{34, 35} and any conclusion of benefit based on such a comparison should be viewed with scepticism. Though it is reasonable to conclude that the CPOU offers a safe alternative to hospital admission, there is no convincing evidence of improved outcome.

Even if the CPOU is no more effective than routine care in ensuring safe discharge of patients with chest pain, surely the evidence of cost saving provides a compelling reason to introduce this form of care to the UK? Before this can be accepted the validity of cost estimates and their applicability to the UK must be reviewed.

Economic evaluations are subject to many of the same threats to validity as clinical trials.¹⁶ The value of concealed, random allocation in preventing known and unknown confounders being over-represented in one or other group is such that for clinical trials and economic evaluations it is considered to be the gold standard.¹⁴ Non-random allocation may cause bias if subjects with a different prognosis are systematically allocated to one group.^{36, 37} The CPOU has been investigated by both randomised^{18, 19} and non-randomised methods^{20, 21, 26, 28–31} and it is noticeable that cost savings are less impressive when a randomised method is used (see table 3).

Randomised trials may also be subject to bias. Subjects may be selected for inclusion in a trial on the basis that they are deemed “suitable” for the new intervention. Management decisions may be influenced by awareness that the CPOU patient is under investigation. Patients may refuse consent if they are adverse to any risk associated with discharge and express a preference for more investigation or a longer hospital stay. If only hospital costs are recorded then significant costs incurred as outpatients may be missed. All such factors will tend to exaggerate the potential for the CPOU to reduce costs and must be considered when reviewing claims of cost effectiveness.

The potential for cost saving will also depend upon the proportion of patients normally discharged directly from the emergency department.²⁵ Most estimates of cost minimisation (even from randomised controlled trials) compare CPOU patients with those admitted.^{18, 19, 21, 26, 28–31} If the presence of a CPOU leads to enrolment of patients who would normally be discharged, then this comparison will no longer be valid and cost savings reduced. It is noticeable from table 3 that the only trial with a control group that included those discharged directly from the emergency department had a relatively low cost saving per patient.²⁰

The costing of all these trials was limited. Only hospital costs were included and only two studies looked beyond inpatient costs.^{19, 20} If, by facilitating early discharge, a CPOU simply moves investigations from an inpatient to an outpatient setting, then cost savings detected by analysis of inpatient costs only will be an overestimate. The use of patient charges to estimate costs may also introduce inaccuracy. Cross subsidising may mean that charges are a poor reflection of costs.

The application of trial findings to local circumstances must be considered. Protocols for the CPOU are usually well defined and can be transferred from one location to another. However, routine practice for hospital admission may vary greatly. It is important that local practice for patients admitted with chest pain is similar to that of the control population in a trial if cost savings are to be reproduced. For example, some studies report rates of inpatient coronary catheterisation for controls of 20%–25%.^{19, 29} Such high cost comparisons are unlikely to be found in the UK. The extent to which subsequent costs should be included in the analysis is a matter of debate and depends upon the economic viewpoint. From the A&E viewpoint it may be reasonable to only consider costs incurred in detecting or ruling out acute disease. Given the present low rate of interventional cardiology in the UK, the introduction of a CPOU that increases the detection of cases of coronary artery disease may result in more cardiology referrals and therefore greater costs. Whether this is appropriate or not requires a subjective judgment.

This review takes a critical look at the arguments in favour of the CPOU. We can conclude that CPOU care is safe and that resource use is controlled and well defined. Uncertainty remains regarding whether the CPOU can improve patient outcomes and whether cost savings can be reproduced in the UK. It should be noted that much of this uncertainty relates to a lack of comparative data on present practice in the UK. It would be perverse to use this uncertainty to conclude that there is insufficient evidence to establish CPOUs in the UK.

The problem of chest pain management in the A&E department is unlikely to diminish in the future. The potential benefits of early thrombolysis mean that patients will be encouraged to attend the A&E department early if they experience acute chest pain. We need to have strategies in place to manage these patients if the ECG is non-diagnostic. Bed availability for emergency admissions is unlikely to increase to meet this demand. Meanwhile the potential for litigation if patients with AMI are discharged is likely to increase. All this suggests that we cannot afford to be complacent about our management of patients with chest pain. The evidence base for the CPOU may have its limitations but we have little evidence to support our present approach.

Conclusion

The CPOU offers a safe alternative to routine hospital admission that may be cheaper and more effective. The potential for cost saving depends upon the proportion of patients attending the A&E department who are subsequently admitted, the typical resource use of those admitted, the proportion of those admitted who would be suitable for care on a CPOU, and the ability of the A&E department to support CPOU services in an efficient manner. Further evidence is essential to determine whether this promising new approach can be applied in the UK.