A Nationwide Flash-Mob Study for Suspected Acute Coronary Syndrome

We would like to thank Angel Schols and colleagues for their "A Nationwide Flash-Mob Study for Suspected Acute Coronary Syndrome". As the developers of the Marburg Heart Score (MHS) we welcome studies promising to extend the evidence base for this primary care decision tool.

In their study Schols et al. investigated patients referred by their GPs for suspected acute coronary syndrome (ACS). Investigators were approached at the national conference, via personal communication, journals etc. to take part in this 'flashmob'-study. Thus a relatively short recruitment period of two weeks resulted in a sample size of n=258. They conclude that the MHS has low diagnostic accuracy in the sample investigated.

In our view, the finding of a limited accuracy in this study is not surprising. The MHS was developed to support GPs in their assessment of the likelihood for coronary heart disease (CHD) in patients with chest pain. CHD included acute unstable as well as chronic and stable forms. If above the threshold of 2 points, further diagnostic testing and/or referral are recommended. In other words, the MHS is meant to inform the decision whether to refer at all. In the sample investigated by Schols et al., participating GPs can be assumed to have used the criteria of the MHS informally, such as age/sex, known vascular disease, worsening on exercise, failure to reproduce the pain by palpation, and the assessment by the patient herself. Diagnostic features, which have been used previously, cannot be expected to discriminate in a sample consisting of predominantly positive observations. By the referral process, the power of a diagnostic test "wears off". (1) If there is a surprise, it is the result of some diagnostic discrimination in a referred population.

The findings of the study should not discourage primary care doctors to use the MHS, although we advise GPs to watch out for additional findings and to pay attention to their intuitive assessment. The authors of a recent meta-analysis evaluating the performance of clinical decision rules for chest pain in general practice concluded that the MHS "has been most extensively tested and appears to outperform clinical judgement alone". (2)

The MHS was developed for the reference condition 'any CHD'. In those with a positive result, GPs still have to decide whether they face a stable or an unstable patient. In our studies, stable patients dominated the sample so that estimates are somewhat unstable regarding ACS. The selection process underlying the study by Schols et al. resulted in a high prevalence sample (ACS 18.5%). If we apply the accuracy measures they obtained to our study of all patients with chest pain (ACS prevalence 3.4%), the post-test probability for ACS is 2%, the negative predictive value 98%. (3) Applying their results to a sample with a more realistic disease prevalence thus results in a satisfactory negative predictive value. Despite this, larger studies evaluating clinical criteria regarding ACS are desirable. They would have to recruit and evaluate all patients presenting with chest pain, not only those the GP thinks are highly likely to suffer the condition of interest.

To conclude, the MHS was derived and validated with patients different from those recruited by Schols et al. The MHS has not been recommended for the sample they investigated. We think the work by Schols et al. has its merits above all for its research design. The recruitment of patients with conditions occurring infrequently, which is the case with suspected ACS, is a stubborn problem for primary care research. The flash-mob study is an interesting design feature, which deserves further investigation.

References

1. Knottnerus JA, Leffers P. The influence of referral patterns on the characteristics of diagnostic tests. Journal of Clinical Epidemiology 1992; 45(10):1143-54.

2. Harskamp RE, Laeven SC, Himmelreich JC, Lucassen WAM, van Weert HCPM. Chest pain in general practice: A systematic review of prediction rules. BMJ Open 2019; 9(2):e027081.

3. Bosner S, Haasenritter J, Becker A, Karatolios K, Vaucher P, Gencer B et al. Ruling out coronary artery disease in primary care: development and validation of a simple prediction rule. CMAJ 2010; 182(12):1295-300.

Competing interests: The above mentioned authors have developed and validated the Marburg Heart Score