Table 5 Key aspects of the stepped wedge design in the HEART study

Implemen-tation decisions

(5a)*

Based on the results of previous validation studies it is likely that the HEART score will be cost-effective if applied correctly (including adherence to management recommendations). The SWD has the benefit to demonstrate the value of the HEART score in real practice and problems in implementation can be observed and documented in each cluster

(5b)*

When a formal decision would be made on nation-wide implementation of the HEART score based on cost-effectiveness estimates from the HEART Impact study, the costs of disinvestment (de-implementation) have to be considered. As the intervention under investigation is the use of a clinical prediction model disinvestment costs are likely to be very small and not much larger in the SWD than they would have been in a parallel group CRT design

Equipoise

(5c)

Earlier validation studies have demonstrated the ability of the HEART score to stratify patients with chest pain according their risk of having a serious heart condition. However, it is unclear whether actively using the HEART score in practice will indeed be safe and improve health care in terms of health care resources, patient burden and costs

Participation

(5d)

The use of risk scores in chest pain patients is recommended in (Dutch) guidelines. The SWD was attractive for hospitals as each hospital would experience using the HEART score during the trial

Preparation

(5e)

Inclusion in the HEART study started almost a year later than planned as all hospitals need to start at same time. Procedures in 1 hospital were slow, which contributed to the delayed start

Informed consent

(5f)*

No informed consent from patients was sought to determine HEART score

Informed consent was asked from patients to collect additional data

Timing of consent: during the initial evaluation by the treating physicians at the emergency department

Study design

(5g)

A mix of hospitals (with respect to size, city and rural, academic and non-academic) participates in the HEART study leading to differences in population and standard of care between hospitals. The SWD allows for a within-hospital comparison reducing the impact of these differences

(5h)

A mix of hospitals participates in the HEART study leading to noteworthy variation in numbers of included patients per hospital which has not been taken into account in the sample size calculation

Blinding

(5i)

The primary outcome is major adverse cardiac events (MACE), which has some subjective elements. There will be an adjudication committee blinded for intervention period for the main endpoints

Interim analyses

(5j)

The HEART study has been classified as a low-risk trial. Therefore, no formal interim analyses are planned. A DSMB is monitoring the trial in particular to give an independent advice to participating hospitals about continuing the use of the HEART score at the end of the study only

Sample size

(5k)

Inclusion rates have been much lower than expected. The study team considered adding clusters or time points to the study, but decided not to do this because it is unclear how to accommodate such changes properly in the final statistical analysis. Furthermore, there was considerable uncertainty about the assumptions in the initial sample size calculation

Method of analysis

(5l)

A Generalized Linear Mixed model (GLMM)-analysis is planned to take into account the hierarchical nature of the data

(5m)

No interim economic evaluation has been planned. Negative results in the health economic analysis could, at least in theory, lead to de-implementation of the HEART score. As this process requires time and money, depending on the number of hospitals already switched to HEART, performing a preliminary economic evaluation as part of an interim analysis might have been worthwhile

(5n)

The (in)consistency in effect across clusters (hospitals) will be examined in a explorative way, for instance whether the effect size varies depending on type or size of hospital