Results
Study population– The study included 46 physicians: 21 (AS, n=11; GP, n=10) attending the AIT-WS in Rome and 25 (AS, n=7; GP, n=18) in Pordenone. All participants were present throughout the entire AIT-WS, completed the full set of surveys, and provided informed written consent. For demographic characteristics of the respondents, please seeTable 1. No relevant differences were detected in terms of gender and age. Physicians were asked about the duration of their previous work experience with allergic patients and no relevant differences were found among the groups (mean±SD, years; AS, 15.4±8.9 and 19.0±6.1; GP, 22.6±11.9 and 25.4±9.5, respectively in Rome and Pordenone). Most of the AS (100% and 86%) and a minority of the GP (30% and 39%) were familiar with the concept of CRD; and similarly, 82-86% of the AS and 10%-17% of the GP, have used CRD as a diagnostic tool in pollinosis, respectively in Rome and Pordenone. Furthermore, the results show that 64% (n=7) and 71% (n=5) of AS, as well as 60% (n=6) and 50% (n=9) out of the GP had previous knowledge about electronic clinical diaries. However, only part of them have already used an eDiary in their own clinical practice: most of the physicians with previous experience being AS (46%, n=5, and 43%, n=3) and only a few GP (20%, n=2, and 22%, n=4, respectively in Rome and Pordenone). Some respondents declared previous knowledge of CDSS (AS 18% and 43%; GP 10% and 17%) but none indicated any previous experience with them in the management of pollinosis patients [Table 1] .
Spectrum of clinically relevant pollen(s) and hypothetical AIT prescription results – The pollen(s) identified as clinically relevant according to guidelines2 and considering the information sequentially added through the potential CDSS tools are shown in Table E2 together with the most frequent allergen source(s) selected by allergy specialists on the basis of primary data[Table E1] for hypothetical AIT prescription (if any). In patients with only one relevant allergen source, the AIT agent most frequently prescribed by AS coincided with the pollen identified as clinically relevant according to guidelines taking into account the information obtained through all potential CDSS tools (i.e. history+SPT+CRD+eDiary) (n=8) [Table E2] . When this procedure (i.e. after considering the full set of information given: history+SPT+CRD+eDiary) led to the identification of two allergens (n=6), one or both of them have been also prescribed by the allergy specialists. In the case of no (n=2) or ≥ 4 (n=4) clinically relevant allergens, no AIT has been prescribed, with one exception (case 4, Rome)[Table E2] . Eighteen patients (90%) underwent nasal provocation testing with one or more pollen(s) among the clinically relevant ones. The NAPT results were all positive, confirming the final decision based on the full set of information considered for a CDSS.
Trend and concordance between AS and GP in hypothetical AIT prescription – For each step of the algorithm and each medical category (i.e. AS and GP), the hypothetical AIT prescription was compared per individual case to the most frequent AIT prescription decided by AS at the final step (i.e. history+SPT+CRD+eDiary), as “gold standard”. In both groups (AS and GP), the hypothetical prescription of AIT changed significantly through the three diagnostic steps proposed in our ”pyramid” model (p<.01), as shown in Figure 2 . Through this evolution, the AIT decisions harmonized within the AS groups and GP groups (p<.01) [Figure 2] . In particular, taking into account the total amount of available choices (n=110 and n=70 for Rome and Pordenone respectively), 54% (Rome) and 59% (Pordenone) of AITs prescribed by AS after the first step of CDSS corresponded to the gold standard choice of AIT. These percentages increased to 66% (Rome) and 83% (Pordenone) when AS expressed their AIT hypothetical prescriptions considering also CRD data, and furtherly to 86% (Rome) and 87% (Pordenone) including also eDiary results. An analogous trend was observed among GP. Considering the total amount of available choices (n=100 and n=180 for Rome and Pordenone, respectively), the percentages of correct prescriptions of AIT increased from 37% and 39% to 57% and 63%, for Rome and Pordenone, respectively, after considering CRD data, in addition to anamnesis and SPT results. When also eDiary data were evaluated, these percentages among GP increased to 79% (Rome) and 83% (Pordenone). Furthermore, the hypothetical AIT prescriptions of GP became consistently closer to those of the AS: finally, no statistically significant differences could be observed between both groups [Figure E2] .
CRD and eDiary impact on AIT prescription by participating doctors – Overall, the number of hypothetical AIT prescriptions increased when, in addition to anamnesis and SPT, physicians considered also the CRD, and finally also eDiary results [Figure 3 and Table E3 ] . This general trend was observed in both GP (hypothetical AIT prescription: Rome, 25%56%63%; Pordenone, 29%59%72%) and AS (Rome, 49%71%65%; Pordenone, 53%77%87%) [Figure 3 and Table E3] . At individual level it was possible to count only a few cases with an inverse trend: 2/10 cases in Rome (for both medical categories) and 2/10 cases in Pordenone (only for GP) [Table E3] .
Furthermore, we evaluated if any change occurred in the number of potential AIT prescription(s) and in its composition (if applicable) on the basis of anamnesis and SPT (only), by adding data referred to CRD and eDiary [Table E4] . Specifically, for each clinical case we considered the most frequent decision taken by each medical category (i.e. AS and GP) at each step of the @IT.2020 algorithm.
In particular, when evaluated only on the basis of anamnesis and SPT, 40% (4/10) and 0% (0/10) of clinical cases presented in Rome would have received a potential AIT prescription according to AS and GP, respectively. Among patients considered eligible for AIT prescription (n=4) by AS, 25% (1/4) would have received a hypothetical AIT prescription with a different composition after considering CRD results and 50% (2/4) when eDiary was considered in addition to the previous information in comparison to the most prevalent composition at the first step (i.e. anamnesis and SPT). In Pordenone, 5 (50%) and 2 (20%) of patients would have received a hypothetical AIT prescription according to AS and GP, respectively. Only one (20%) would have received a hypothetical AIT prescription with a different composition after CRD data and two (40%) after considering CRD and eDiary results[Table E4] .
Feedback survey on doctors’ perception regarding diagnostic tools – Doctors filled a questionnaire on the role of each diagnostic tool or step in their own decisions expressed in the AIT-prescription survey. Additionally, their opinion on the algorithm proposed for our innovative CDSS has been assessed. All physicians considered the application of a CDSS useful and recognized its potential in ameliorating the traditional diagnostic procedures [Figure 4, Table E5] . There was agreement also concerning the role of molecular diagnostics in improving the accuracy of AIT prescription (100%). The reliability of the retrospective assessment of clinical histories was assessed lacking (70-100%) and optimizable by an electronic clinical diary (82-100%). In addition, all respondents judged the latter as easier to be filled by patients and to be interpreted by physicians in comparison to a paper diary. Furthermore, the majority of doctors agreed on a potential role of an electronic diary in the diagnostics of other allergic diseases (e.g. asthma and food allergy). Significant discrepancies were registered between the two medical categories (AS vs GP) in one center (Pordenone) concerning the physicians´ opinion on eligibility to AIT of patients sensitized to more than four aeroallergens (p = .03). Participants were overall satisfied by the workshop (tutorial, clinical cases and feedback survey) in terms of content and general organization [Figure 4] .