We demonstrate that a SSA that includes an automated text message to a mobile phone for acute stroke patients is technically feasible, and that the system maintains its stability. Only on two consecutive days were text messages (15 = 3.4%) missed as a result of technical modifications in CIS. As these adaptions were not made known to the users beforehand, no adjustments had been carried out in advance on our filter settings. Immediately after notification of CIS changes, filter settings were modified. No text messages were missed after these modifications.
In contrast to previous investigations
[15–17], our approach was to focus on the special requirements for acute trials, which mandate start of treatment within a very limited time window: within 6 or 9 hours of symptom onset. To comply with selection criteria, patients must be screened immediately on arrival in hospital. In addition, relevant information, e.g., NIHSS, is essential for reaching a decision regarding a patient’s trial eligibility.
The SSA was implemented in the screening process of the neurological department and was designed as a technical interface to achieve rapid interaction between the physician on-call – who is responsible for routine diagnostic and treatment – and a specialized trial team physician. The SSA was based on information routinely documented in CIS and thus did not entail a higher workload for the treating physician. We think that this is of utmost relevance for the acceptance and feasibility of our screening tool in the emergency setting.
In contrast to other approaches, where selection criteria had to be defined beforehand to identify patients for specific trials
, our method aimed at a more general technique: every single patient with suspected stroke was identified by the system and a text message with the most relevant screening data was generated. Text messages were sent to the trial physician on call, who was responsible for the screening and recruitment process for a multitude of stroke trials. Based on the information provided – patient sex and age, date and time of admission, and NIHSS – the trial physician pre-selected the most likely eligible patients for a variety of stroke trials. Naturally, a detailed screening process followed: the screening information supported the trial physician in deciding which patient to see first, i.e., in which case time was most pressing.
As in previous approaches, some patients were identified by the automated screening system which had not been identified in a manual search
[16–18]. However, a further analysis showed that these additional patients were stroke mimics or were not admitted to our department, and were thus not eligible for acute stroke trials.
An advantage of our reporting tool was that absolutely all stroke patients were reported to the trial physician. Even patients who failed to meet inclusion criteria at admission were recorded and surveyed later to catch changes in medical condition, e.g., in NIHSS, and later trial eligibility.
As our alerting system did not depend on specific selection criteria, it was easily expanded for new trials or after modification of selection criteria in ongoing trials; no adaptions to the system were necessary. This was especially advantageous for repeated changes in selection criteria or for high turnover of trials.
The main focus of our approach was not only on capturing absolutely all patients but also generating the most rapid alert possible. However, the time from admission to text-messaging still exceeds 60 minutes in the majority of patients, which is rather long. In the emergency setting, documentation had to take second place to acute diagnosis and treatment. This led to delays in documentation and is in fact the major obstacle to using the automated screening tool for acute stroke intervention trials; for trials with a very limited time window, delays exceeding 60 minutes post-hospital admission can reduce trial eligibility. Nevertheless, in our case, delays were due not to technical flaws but rather to suboptimal organization. We aim to solve this problem with better training for the physicians in ED. Nonetheless, we have to take into account that delays in documentation will continue because therapy and treatment are more relevant for the patient than data documentation. In addition, future adaptions to the standardized stroke sheet will be mainly in the area of supplementary information (e.g., time of symptoms onset) for the screening process and the decision on trial eligibility.
A major benefit of the automated screening tool is its availability outside regular working hours. In our survey, 118 (23.0%) patients were admitted at weekends and a further 127 (24.8%) on working days between 6 pm and 6 am. We estimate that the majority of these 245 (47.8%) patients would not have been screened at all for trial eligibility without the automated notification tool.
Additionally, the notification tool might be of interest as an automated alerting system for patients with suspected stroke in standard care as the tool could inform all relevant services within minutes once an appropriate trigger is determined. The screening tool requires an IT-based ED file and HL7. As HL7 is highly available in most systems, the tool could be used at other hospitals as well. Costs of the notification system depend on the number of sent text messages and the specified charges.
We deliberately did not perform a comparison of total trial recruitment numbers with or without the notification tool, so no quantitative correlation for total recruiting numbers is provided.