This Delphi consensus is part of a wider study aimed at developing an eHealth tool to assess neurological impairment at the population level, in the absence of a neurologist, to be used for epidemiological research. As a first step, a systematic review mapped existing eHealth software tools assessing one or more neurological functions [25]. This Delphi exercise represents a step further of the theoretical work needed for the development of the final eHealth tool—the NeuroEpiTool.

This study aimed to reach a consensus among experts on what the screening tests/anamnestic questions are to assess neurological function to maximise the observer’s ability to rule out the largest number of neurological signs and symptoms. Following a parsimonious principle, when all items are negative, neurological impairment can be ruled out in a participant; and if positive, further tests need to be administered. The NeuroEpiTool will then undergo firstly a comparison study at the individual level against a clinical neurological examination, followed by a validation phase at the population level to compare the ability of specific combinations of signs and symptoms to predict a neurological disease.

Given the general context in which this research develops and the need to interpret the data coming from the NeuroEpiTool without the presence of a neurologist or any instrumental diagnostic procedure, we deliberately chose to approach the neurological examination by function, rather than anatomical location. A map of the neurological function was compiled and provided to the expert for guidance (see Fig. 1). The final selection of items identified needed to meet both the feasibility and acceptability criteria, i.e. they had to be feasible to perform and interpret by a trained health worker and acceptable by potential research participants.

Mapping of the neurological function covered by the selected consensus items

The map of neurological function was organised according to five core domains: cognitive status, cranial nerve function, motor strength, sensory system and coordination [28]. Each domain branches into several functions or subfunctions (e.g. orientation and attention in cognitive status, upper and lower limb motor function) which can be selectively impaired. As the map was based on function alone, some conditions or diseases may be left out of the mapping (e.g. epilepsy); however, experts were invited to provide complementary tests/questions that screen conditions not represented in the map.

Given the overarching scope of the Delphi exercise, we identified three different groups of experts: in the neurological examination, in data collection, and in participant experience. The original aim was to gather a sample of approximately 15 experts. Experts were invited through personal network, identification from the scientific literature, and snowballing. Special attention was posed in widening the geographical location and to ensure gender balance of the participants. Experts were invited to self-identify as expert in one or more roles.

Experts in the neurological examination (i.e. clinical neurologists) were asked to identify the screening tests/anamnestic questions for assessing neurological function. A total of six experts were involved, five from Italy and one from Uruguay (CC, CDL, CK, LT, MP, PR).

Experts in data collection (i.e. researchers with experience of data collection in the field) were asked to assess the feasibility of performing and interpreting each test identified by the experts in the neurological examination. A total of five were involved, two from Portugal, one from Italy, one from Colombia, and one from Uruguay (AG, CR, CK, JA, LT). Two experts in data collection were also experts in the neurological examination.

Experts in participant experience (i.e. people living with a neurological disorder) were asked to assess the acceptability of each test. A total of three were involved, two from the Netherlands and one from the UK (HH, RS, SH).

Initial online meetings with a round of introductions were conducted to explain the scope of the study and what was required by each group of experts. These session were followed by questions and answers. Shortly after, questionnaires were circulated, and panellists had 2 weeks to consider the questionnaires and return their answers, with a reminder being sent every two working days, on Tuesdays and Thursdays. The choice of this time window to return the answers was to maintain momentum and to avoid panellist withdrawal. We were able to gather a 100% response rate at the end of each round, with no withdrawals after the Delphi start. Panellists were contacted via email on blind carbon copy (BCC) throughout the rounds. A flowchart for each of the Delphi rounds is represented in Fig. 2. All data were collected and stored through the Qualtrics XM platform. Raw data for this paper will not be available for further use, but the final NeuroEpiTool will be freely available to use. A table describing the questionnaires is provided in Table 1. The entire exercise was conducted in English.

Description of the Delphi rounds and timetable

The aim of the first round was to identify the tests able to rule out the largest number of neurological signs when conducted on a participant coming from the general population and to assess the feasibility and acceptability of administering them. The questionnaire was composed of open-ended questions, and experts in neurological examination provided a title and short description of a maximum of three tests accompanied by a description of the neurological impairment ruled out, if negative. Data acquired were then collected and organised (i.e. tests were listed and described; some items were merged) to be circulated to the experts in data collection and participant experience.

Experts in data collection rated the proposed tests in terms of the feasibility of them being performed and correctly interpreted by a trained health worker on a real-life participant. They rated feasibility on a Likert-type scale from 0 to 4 (e.g. extremely easy to be performed and interpreted by any health worker/extremely difficult to be performed and interpreted by any health worker). Similarly, the same tests were presented to the experts in participant experience, who evaluated their acceptability by the potential participant, answering in a Likert-type scale from 0 to 4 (e.g. extremely comfortable/extremely uncomfortable).

All tests collected during Round 1 were mapped against the framework of the neurological function shown in Fig. 1. This was circulated to the experts in the neurological examination for them to identify and suggest additional tests to fill any gap. Additionally, experts were invited to provide comments on the neurological function mapping. In this round, the experts in neurological examination were also asked to identify a maximum of three anamnestic questions each. The aim of the questions was—as for the tests—to rule out the largest possible neurological impairment. The anamnestic questions were aimed at assessing symptoms which could not be assessed with a test (e.g. headache). Afterwards, the experts in data collection and participant experience once again rated each of the new tests or anamnestic questions on the basis of their feasibility in administering/interpreting and acceptability. The rounds were then closed and the Delphi completed. Experts were always encouraged to provide their own feedback on the Delphi exercise, the questionnaires provided, and the neurological function mapping.

The Delphi leaders (VRF, VG) collated all data and delivered them back to the experts according to their role, from round to round. To reach a final consensus on each of the tests/questions, a threshold of at least an average 2.5 rating out of 4 for both feasibility and acceptability had been established. Tests/questions whose acceptability was ranked between 2 and 2.5 out of 4 were deemed to be potentially included albeit with a special warning to the assessor (e.g. “the following test may cause discomfort to the participant, please consider their willingness to perform the test”).