Explanation for the choice of comparators

During the control phase (i.e., usual care), the CDSS is deactivated. For each hospitalized patient, pharmacists review the prescription via the computerized physician order entry (CPOE) system. If a patient is considered to be at risk of AKI or hyperkalemia, the pharmacist assesses the situation and issues a PI to the medical team in charge of the patient. This intervention will probably be subject to both inter- and intrahospital variability. Data related to the hospital stay is collected and extracted from the CPOE by the CDSS’s developers and will be cross-referenced with the e-CRF so that only patients who have given their verbal, informed consent are selected.

Intervention description

The use of the CDSS is linked to each center’s clinical pharmacy activity [27,28,29]. The CDSS conducts a real-time data analysis and produces alerts for situations where there is a risk of an ADEor a proven ADE. The clinical pharmacist in charge examines the alert in three steps: (i) verification of technical validity (comparison of the rule’s appraisal elements with the patient’s record); (ii) evaluation of the alert’s pharmaceutical relevance, with the issuance of a PI if necessary; (iii) follow-up of whether the prescribing physician has accepted or not the PI. Before the start of the stepped-wedge trial, we carried out a preliminary phase of creating, integrating and testing the rules in the CDSS. Firstly, the rules were drafted in a seminatural language by an expert group, on the basis of the scientific literature and clinical expertise [19]. We have 12 rules for preventing ADEs (six for AKI and six for hyperkalemia) and 19 rules for monitoring the event (18 for AKI and one for hyperkalemia). All the centers reviewed and corrected the rules. The rules were then coded in computer language in collaboration with the industrial partners (Keenturtle and Quinten). Next, we tested the rules in real life in the five investigating centers, by using a CDSS mode in which alerts are visible only to the centers’ principal investigators. This approach enabled us to detect errors in how the rules were drafted.

Lastly, we created a set of standardized courses of action for each rule. These measures were based on a review of the literature and subsequently proofread by a committee of experts (comprising several pharmacists, an infectious disease specialist, a geriatrician, and a nephrologist) and then by all the centers for validation. The CDSS developers and the clinical pharmacists ensured that these standardized courses of action are integrated into the CDSS so that the latter are readily available to any clinical pharmacist analyzing the alerts.

The intervention is carried out in addition to the usual care. The intervention phase starts at the beginning of period 2 when the CDSS is activated (Fig. 1). During this phase, the CDSS is used to detect a risk of AKI or hyperkalemia or confirmed AKI or hyperkalemia. An alert is triggered when the patient’s data matches all the rule’s criteria. The clinical pharmacist analyzes the alert from a technical and pharmaceutical viewpoints. Depending on the standardized course of action, the pharmacist may issue a PI to the medical team in charge of the patient.

Criteria for discontinuing or modifying allocated interventions

Criteria for discontinuing or modifying allocated interventions were not possible to define, as the interventions are allocated by hospitals.

Strategies to improve adherence to interventions

To avoid the loss of information at-risk situations that have not been analyzed by other pharmacists, a pharmacist at each center checks for non-analyzed alerts on a weekly basis.

Relevant concomitant care permitted or prohibited during the trial

No types of concomitant care is prohibited during the trial.

Provisions for post-trial care

There are no provisions for post-trial care.

Outcomes

For the primary objective, the outcome is the occurrence of one or both of the following events: AKI and/or hyperkalemia. Hyperkalemia is defined as a high blood potassium value; the cut-off is set by each hospitals’ lab but is usually > 5.0 mmol/L). The definition of AKI is based on the Kidney Disease Improving Global Outcomes criteria: elevation in creatinine by a factor of 1.5 over a period of 7 days or less (AKI stage 1) [30]. The outcomes corresponding to the above-listed SOs are as follows.

SO-1: Safety of the intervention. The PI linked to the alert may lead to a change in the prescription that could induce a subsequent ADEs (e.g., hypokalemia induced by the discontinuation of potassium supplementation, or a report of a high blood pressure following the discontinuation of an anti-hypertensive treatment during an AKI). The pharmacist reports the occurrence of any ADE associated with a change in prescription.

SO-2: Indicator of the pharmaceutical relevance. When an alert is triggered, the pharmacist analyzes the patient’s record to determine whether or not a PI should be sent to the medical team. The number of alerts leading to a PI being sent on to the medical team is divided by the total number of alerts for the same rule triggered by the CDSS and analyzed by a pharmacist.

SO-3: Indicator of medical relevance. Upon receiving the PI, the medical team assesses the relevance of the intervention and decides whether or not to accept it. The number of the PIs accepted by the medical team is divided by the total number of PIs for the same rule sent by the pharmacists following a CDSS alert. Furthermore, the CLEO scale (developed by the French Society for Clinical Pharmacy (Société Française de Pharmacie Clinique) is applied to all the PIs sent to the medical team [31]. The CLEO scale provides a standardized assessment of the clinical, economic, and organizational impact of a given PI.

SO-4 Time period for prescription changes. The respective mean time intervals between the issuance of an alert by the CDSS, the analysis of the alert by the pharmacist, the transmission of the PI to the medical team, and the decision to accept (or not) the PI by the medical team.

SO-5: Health economic assessment. This evaluation is divided into three parts: (i) the intervention cost, corresponding to the mean time spent creating and integrating the rules into the CDSS; (ii) the usage cost, corresponding to the time spent analyzing alerts triggered during the study; and (iii) the intervention cost, corresponding to the cost of preventing and managing AKI or hyperkalemia during a hospital stay.

SO-6: Organizational impact of using a CDSS. In one department of each participating center, a team of human factors specialists investigates the detection and management of AKI and hyperkalemia events. The aim is to understand the role of each professional (senior physicians, junior physicians, nurses, clinical pharmacists and medical biologists), their work processes, and the technologies used in the management of medication-related iatrogenic events (hyperkalemia and AKI). This analysis is carried out before and after the activation of the CDSS in the centers, in order to identify and explain the impact of the introduction of the DETECT-IP rules on their workflows.

Participant timeline

The patient timeline follows this process:

Screening: upon arrival at the hospital for admission to a ward selected for the study, patients aged 65 and over receive a study information note

The CRA or the pharmacist receives verbal, informed consent from an eligible patient (or from his/her next of kin), and this consent is documented in the e-CRF

During the control period (i.e., when the CDSS is deactivated), all patients are included if they consent. At-risk situations are detected during the standard review of prescriptions by the pharmacists

During the intervention period (i.e., with activation of the CDSS): all patients are included if they consent. At-risk situations are detected by the triggering of an alert in the CDSS which are analyzed by the pharmacist

Sample size

In this stepped-wedge, cluster-randomized, controlled trial, the unit of randomization is the center. The number of participants depends on the correlation between the participants in the same cluster, as estimated with the intraclass correlation coefficient (ICC). No data are available for determining the ICC for the primary endpoint ADE in a population similar to that expected in our study. However, we have estimated the ICC to be 0.005 on the basis of on the ADAMS study [32], which provides ICC values for a large number of variables in primary care. The estimateADE rate is 10% in the control group (4.5% for hyperkalemia and 8.0% for AKI, with both events concomitantly in some cases) [20]. Based on these values, we estimated that 7% in the interventional group were at risk of hyperkalemia and/or AKI (70% of ADEs are considered to be preventable, and the system will correct about 40% of these). With the five-cluster, six-phase design, a total of 4920 patients will have to be recruited (164 per cluster and per phase, in a two-sided test with an alpha risk of 5% and a power of power = 80%). All cases of AKI and hyperkalemia are included as events, regardless of the causes; this has been taken into account in the sample size calculation.

Recruitment

The recruitment process is assisted by the use of two documents: (i) a detailed study information sheet given to the patient upon arrival in the ward and (ii) a plain-language study summary flyer for the patients and their families. Patients are recruited in certain wards of each hospital, in which the pharmacists and physicians have agreed to participate in the study and are covered by the CDSS.