Can artificial intelligence increase the efficiency of referrals from primary to specialized care?

The overload of the current healthcare model makes the search for strategies to improve process efficiency essential. An artificial intelligence (AI) program based on Natural Language Processing pipelines (1-5) was used. It analyzed the referrals from primary care to Gastroenterology in the health area corresponding to our hospital in order to identify the most frequent reasons for consultation and to assign them a protocol for the performance of complementary tests before being seen for the first time in specialized care. We compared all referrals received in the first half of 2018, prior to the implementation of the AI pathway July 2018, with those received in the first half of 2019. Our aim was to evaluate the efficiency of this program in terms of discharges, need for additional tests and the number of follow-up visits required (number of follow-up visits/number of first visits in a given time period, FU/F index). In 2018, 1799 referrals were received, 1309 within our health area and 490 from outside the area. In 2019, 2261 referrals were received, 1392 from our area and 869 out-of-area. The AI pathway was applied to 31.4% of the area-referred patients. Overall, in 2019, the number of blood tests and CT scans requested at the first visit decreased (55.3 vs 61.4% and 4.4 vs 7.4% respectively, p<0.05 for both comparisons). The FU/F index in 2019 was 1.9 ± 0.04 vs 2.26 ± 0.07 in 2018 (p<0.05). When analysing patients from our health area, a higher number of discharges at the first consultation was observed during 2019 The number of requested supplementary exams among patients referred using the AI pathway was reduced compared to 2018. The FU/F index in patients referred using the AI pathway was 1.72 ± 0.08 vs 2.25 ± 0.08 in 2018 (p<0.05) and 1.93 ± 0.07 in those referred through the standard pathway in 2019 (p=0.07). Among patients referred from outside our health area, the number of endoscopies requested in 2019 was higher. The FU/F index improved in 2019 (1.95 ± 0.06 vs 2.29 ± 0.13, p<0.05). The number of patients referred using the AI pathway remains low, which could explain the lack of differences observed in the number of discharges or tests requested compared to patients referred via the standard pathway. However, the number of endoscopies and follow up visits requested for these patients did decrease.

High SARS-CoV-2 viral load is associated with a worse clinical outcome of COVID-19 disease.

COVID-19 severity and progression are determined by several host and virological factors that may influence the final outcome of SARS-CoV-2-infected patients. The objective of this work was to determine a possible association between viral load, obtained from nasopharyngeal swabs, and the severity of the infection in a cohort of 448 SARS-CoV-2-infected patients from a hospital in Madrid during the first outbreak of the pandemic in Spain. To perform this, we clinically classified patients as mild, moderate and severe COVID-19 according to a number of clinical parameters such as hospitalization requirement, need of oxygen therapy, admission to intensive care units and/or death. Also, Ct values were determined using SARS-CoV-2-specific oligonucleotides directed to ORF1ab. Here we report a statistically significant association between viral load and disease severity, a high viral load being associated with worse clinical prognosis, independently of several previously identified risk factors such as age, sex, hypertension, cardiovascular disease, diabetes, obesity and lung disease (asthma and chronic obstructive pulmonary disease). The data presented here reinforce viral load as a potential biomarker for predicting disease severity in SARS-CoV-2-infected patients. It is also an important parameter in viral evolution since it relates to the numbers and types of variant genomes present in a viral population, a potential determinant of disease progression.