Building deep learning systems does not only depend on quantity of data used but type and quality. The challenge is that time- years are needed to test and fine tune tools with thousands of cases for them to continue learning. The process is unfortunately not automatic, requiring human intervention and oversight [4]. Similar online tools, expert systems were used in the 80s in aviation, finance and medicine including MYCIN, a tool that advised which antibiotic to use for which infection [5,6]. Another glaring challenge is tool accuracy, which depends on foundations of the clinical engine, a challenge that expert systems also confronted [6]. Organizations providing online tools to patients should pay attention to the tools` clinical engine foundations. Untested and unvalidated clinical engines tend to give erroneous information and recommendations [1]. How many online forward triage tools (OFTTs) in use have been peer reviewed for accuracy is not clear? Engines that are vetted medically, should be able to meet all individual consumer experiences, incorporating both self-diagnosis and triaging patients and directing them to the appropriate level of care. One study of online tools revealed deficits in both diagnosis and triage, with triage advice being risk averse- recommending medical care to patients who could have done with self-care [7].

The following are suggested as questions that need to be answered before providing online services [1]:

Another challenge of OFTTs is that the pandemic linchpin lies in population-wide testing, a service telemedicine cannot offer currently [2]. The following additional challenges were revealed in a symptom checker systematic review:

Consumers seem to want the best possible medical and health information and seem to value the same highly tested information tools made available to care providers, but presented in simpler terms [1]. One poll revealed that 86% of respondent said the more information including symptoms, they supplied about their condition, the more appropriate the treatment they will get from their health care provider [1]. What is however not clear, is the type of patients polled. Were these patients with chronic conditions like chronic pain, or were these patients who had an acute condition eg, shortness of breath and needed to know what to do? In case of the latter scenario, we are of the mind that those with an acute condition would most probably be in a physically and cognitively compromised state, to self-assess using an OFTT.

.  Photo: Purchased by the authors from canstockphoto.

Should COVID-19 OFTTs, an acute condition, a pandemic situation be different from a general OFTT such as one for chronic pain? We would like to suggest-yes. How much time, on average should patients spend filling these in? This again, we would like to say depends on the condition; acute vs chronic. What is the optimal length of an OFTT questionnaire, how many questions should be asked, average, minimum, maximum? What is the optimal time a patient should spend filling in responses? What is the maximum acceptable time, a patient should spend filling OFTT responses, before being told to call 911? The purpose of this paper is to stimulate a discussion and hopefully generate answers to the above questions. One paper proposes questions to ask when setting up an OFTT [4], but does not give the average number of questions, average time nor do they differentiate between acute and chronic conditions.

We suggest a differentiation between OFTTs for chronic vs those for acute conditions. A patient with chronic pain might have become accustomed to the pain, and might find that giving all the particular differences of pain intensity and duration and triggers etc, will make the doctor better understand their pain. A patient, who on the other hand is in an acute state, has severe symptoms, is distressed and wants to know what to do next, might be discouraged or be in physical danger if asked to respond to 50 questions, until he gets a recommendation to call 911. In acute conditions time is critical in several ways;

We are therefore of the opinion that there should be cases, in which patients should be advised to discontinue the OFTT and get immediate emergency rescue services, lest they lose precious time. For some acute conditions like myocardial infarction, we are of the mind that such OFTTs, in fact be discouraged.

Do OFTTS have room in all fields of Medicine? The value of OFTTs in chronic disease, screening, diagnosis and follow-up has been demonstrated [4]. The role of OFTTs in paediatrics, neurology and acute settings are still contested, as states often change suddenly. More studies are needed in this regard. As OFTT use increases, we propose the following research questions: How many questions are optimal or how much time is optimal, for patients, to spend filling an OFTT in acute and chronic cases? Our hypothesis is that the longer the time spent on filling in OFTT responses and the higher the number of questions, the lower the questionnaire completion rates, the lower the proportion of recommendations followed and consequently poor outcomes. We also propose the assessment of COVID-19 OFTTs in Switzerland, comparing the number of questions, the average length of time spent filling in OFTT responses and outcomes like questionnaire completion rates and proportion of recommendations followed.

As OFTT use increases, their optimal lengths have not been explicitly defined. We therefore propose that frameworks for OFTT evaluation be updated to include among others, dropout rates, optimal number of questions and time spent filling in OFTT responses, validation status of clinical engines along with accuracy of diagnosis and triage.