Intensive care: an app that saved lives

A major issue dur­ing the pan­dem­ic has been avail­abil­i­ty of crit­i­cal care beds in inten­sive care units. At vital points dur­ing the peak in num­bers of Covid-19 patients, many hos­pi­tals were under pres­sure. The ser­vices need­ed to know where to send patients in order to pre­vent over-load­ing ICUs. 

Ini­tial­ly, health­care staff were not­ing the num­ber of beds avail­able in dif­fer­ent hos­pi­tals on Excel sheets, which they were send­ing to one anoth­er via What­sApp chats. Even though the cri­sis con­di­tions required fast action, it was nei­ther the most effi­cient nor the most secure way to share data. Espe­cial­ly since we are talk­ing about infor­ma­tion that could have caused pan­ic had it been intercepted. 

I work with doc­tors and nurs­es in those ser­vices on a dai­ly basis and, once we had dis­cussed the prob­lem, I knew we could auto­mate it. As a response, our team designed an easy-to-use dig­i­tal tool to keep track of beds. At the begin­ning of the lock­down, we worked around the clock to turn out the tool in 3–4 days. The ICU Bed Avail­abil­i­ty Mon­i­tor (ICUBAM) was imme­di­ate­ly picked up by the local health author­i­ties of the Grand Est – a region of France with a pop­u­la­tion of over 5.5 mil­lion people. 

The ICUBAM sends a text mes­sage to the mobile phones of ICU work­ers twice a day, with a link to a sim­ple online form. In a mat­ter of sec­onds, the user can input the nec­es­sary infor­ma­tion includ­ing the num­ber of avail­able beds, whether they were equipped for COVID+ patients or not, and the num­ber of peo­ple discharged. 

Data was then col­lect­ed and cen­tralised, which we used to gen­er­ate mod­els and real-time analy­ses of evo­lu­tions in the pan­dem­ic. Through­out the cri­sis, we con­tin­ued to inter­act with doc­tors, pre­sent­ing them with the dash­boards of our mod­els and fig­ures. The data was there to help them make deci­sions but of course judge­ments calls remained their own responsibility. 

ICUBAM was rel­a­tive­ly easy to put in place. All we real­ly need­ed was a place to store the data – and some fund­ing to cov­er the cost of the text mes­sages. As the num­bers came in over time, we trans­formed the data into a visu­al map of the pan­dem­ic across the coun­try. From that, we were able to go fur­ther by com­bin­ing the data from dif­fer­ent hos­pi­tals with epi­demi­o­log­i­cal mod­els of the pan­dem­ic to pre­dict avail­abil­i­ty of beds in the com­ing days. Our goal was to pro­vide vital infor­ma­tion to doc­tors who were mak­ing dif­fi­cult deci­sions about where to send high-risk patients. 

After the lock­down peri­od at the begin­ning of sum­mer 2020 we stopped run­ning the ICUBAM. How­ev­er, we were lat­er con­tact­ed by many inten­sive care health work­ers with requests for us to put it back into action in prepa­ra­tion for the sec­ond wave. On our side, we would like the author­i­ties on a nation­al lev­el to take up the sys­tem; health work­ers in ICUs need it. To face the sec­ond wave, the ICUBAM has been put back into action by INRIA and four French regions. For me, this shows our capac­i­ty to rede­ploy the tool when needed.

Dig­i­tal tools like ICUBAM can pro­vide real insights into pub­lic health issues by keep­ing track of impor­tant data for doc­tors, but also so that we may analyse it to bet­ter under­stand a sit­u­a­tion for oth­er stake­hold­ers. We cre­at­ed it in a mat­ter of days, and the for­mat could eas­i­ly be adapt­ed to oth­er cri­sis sce­nar­ios. Depend­ing on the con­text, it could be used to track avail­abil­i­ty of med­ical mate­ri­als, their stor­age loca­tions and so on. In devel­op­ing coun­tries, sim­i­lar data col­lect­ing sys­tems could help sup­port and track pub­lic health crises. 

On a pos­i­tive note, post-lock­down I have also noticed more PhD appli­ca­tions in my field. It would seem that the pan­dem­ic offers an exam­ple of how sta­tis­tics can be applied to real-life sit­u­a­tions. Young sci­en­tists want to apply their knowl­edge and there are many oth­er excit­ing chal­lenges to tack­le in the domain of arti­fi­cial intel­li­gence for health.