Could telemedicine solve the cancer backlog?

Abstract

Becky McCall reports. Although millions of patients with cancer around the world face delays in diagnosis and treatment because of the diversion of resources during the COVID-19 pandemic, there is a growing expectation that telemedicine may play a central role in easing the backlog. The use of telemedicine comprising video, telephone, and other electronic communication has substantially risen from around 10% of general medicine consultations before COVID-19 to approximately 75% during the peak of the pandemic, in the UK. The number of remote consultations has started to level out to somewhere in between, as countries tentatively tread the slow road back to normality. However, there are indications that telemedicine is here to stay. Richard Roope, senior clinical advisor at Cancer Research UK and practising general practitioner in Sussex, UK, notes, “The unforeseen consequence [of the pandemic] is that with telemedicine, the number of cases a GP [general practitioner] can manage daily has increased tremendously.” This change became possible with the surge in the application of digital technology. “We've had five years of innovation in five weeks”, he told The Lancet Digital Health. As the pandemic loomed, remote consultations were stepped up to help to overcome the difficulties of doing patient consultations in person when the public were required to stay home. People were reluctant to attend medical facilities for fear of infection, or to avoid overburdening already strained health-care services. Screening services were suspended and delays in cancer diagnoses, referrals, and treatment accumulated. In early June, Cancer Research UK released figures showing that COVID-19 has led to a backlog of 2·4 million people requiring cancer care. Likewise, a recent study by University College London, London, UK, estimates that over 6000 additional deaths could occur in England over the next year because of delayed care and patients with cancer contracting COVID-19. Similar calculations in the USA estimate an additional 34 000 deaths due to newly diagnosed cancer over the same 12 months. A study from May, 2020, by the COVIDSurg Collaborative, estimated that 28 404 603 operations would be cancelled or postponed worldwide during the peak 12 weeks of COVID-19 disruption. However, the authors note that cancer surgery would be prioritised and that most cancellations would be for benign conditions, mostly orthopaedics. But as health-care systems move forward in tackling the backlog in cancer treatment, telemedicine will be key. Martin Marshall, CBE, chair of the Royal College of General Practitioners, cautiously predicts that up to half of general practitioner consultations will continue to be remote after the pandemic has subsided, at least in the foreseeable future. Roope echoes this prediction and notes that the adoption of telemedicine practice is reinforced by recent NHS guidance and standard operating procedures for general practice covering the provision of health-care access during COVID-19 recovery and beyond. This guidance proposes total triage whereby a face-to-face appointment is only granted after a video, telephone, or electronic consultation has been done. Likewise, the USA has also seen a surge in telemedicine. AnaMaria Lopez, medical oncologist at Thomas Jefferson University, Philadelphia, PA, USA, told The Lancet Digital Health that at her cancer centre the number of telemedicine visits during May was equivalent to that anticipated for the entire year. “Many centres have moved from 40 telemedicine visits per year to 4000 in a month”, she says. Cary Presant, a cancer specialist from the City of Hope Hospital, Los Angeles, CA, USA, says that all staff were trained in telemedicine in early March as the threat of COVID-19 closed in, and it has been widely adopted since. “Telehealth serves those patients vulnerable to infection”, which is nearly everyone in Presant's cancer practice because of older age, immunodeficiency, or comorbidities. But this swift rise in adoption of telemedicine over the pandemic comes with potential disadvantages in aspects of cancer care, especially cancer diagnosis. However, artificial, or more precisely, augmented intelligence (AI), could be used to help to optimise its use. The wealth of data and understanding pertaining to risk factors for cancer makes it particularly suitable for care enhanced by AI. Surgeon Aswini Misro is founder of the start-up company, YouDiagnose, which is developing diagnosis technology powered by AI for various cancers. He points out that unusual cancers are easy to miss on first, second, and even third consultations, and this problem is only exacerbated with telemedicine. Studies suggest that cancers such as pancreas, myeloma, and lung usually need three or more consultations face-to-face before referral, he says. “As such, cancers presenting with subtle, non-specific clinical features, or atypically lacking red flags, may be less likely to be considered as cancer and referred early by GPs”, explains Misro. Clinical decision support powered by AI can work synergistically to validate and optimise human decisions and vice versa. “This can reduce the incidence of medical errors, thereby significantly improving the prospect of early cancer diagnosis”, he points out. And with respect to telemedicine, the absence of a physical examination, the patient's physical cues, and reduced doctor–patient emotional engagement, there is a pressing need to incorporate other large data into video consultations via AI, he adds. Real-world source data used by an AI algorithm are diverse and might come from the patient's own history, including previous diagnoses and investigations; genomic profile; electronic health records, including radiography; and possibly microbiome genomics. “Data might include localised information, for example, radiation levels in some geographical areas, or high ethnic minority populations with delayed presentation. This real-world evidence could enhance the consultation with information that cannot be sourced in a nine-minute meeting”, explains Misro. Hamish Fraser from Brown Center for Biomedical Informatics, Brown University, Providence, RI, USA, agrees that additional data to the video consultation should improve the diagnostic accuracy of possible cancer. However, he cautions that “currently, accessing complete data from EHR [electronic health records] records in a form usable by AI algorithms remains challenging, and requires evidence of performance in routine use. UK primary care electronic health records have an advantage in this respect over the more fragmented systems in many US health systems.” He points out that the use of symptom checkers or chatbots for the collection and analysis of data directly from patients on symptoms, family history, and risk factors is promising. But the challenge right now is in establishing which patients are best suited to which mode of consultation—telephone, video, or face-to-face—and to what extent AI can enhance this, reflects Presant. “AI can be helpful here. Using metrics of disease, patient and physician, we can determine which patients will have outcomes enhanced or worsened by telehealth.” In cancer care, AI has been used to help to understand the disease process, and in predicting how effective, but also how toxic a treatment might be, notes Presant. “We want to mitigate these risks by using AI in a predictive way”, he explains, but adds that there is a distinct need to research and refine how to optimise use of AI in medicine. Lopez concurs that big data and AI are valuable but says it is important to uncover the crucial elements of data that most improve patient outcomes. Equally, Fraser highlights that the most important challenge is the quality, completeness, and accessibility of the key data items, rather than the development of new algorithms. The importance of the quality of the data source is exemplified by concerns around the use of AI to improve the reliability of skin cancer diagnosis. Research suggests that AI might exacerbate health-care disparities for non-white patients if inclusivity of population data source is not adopted. Data mainly come from white populations and if the algorithm is based on the appearance of lesions on fair skin, it might miss those on skin of darker colour. Roope is an advocate for greater use of AI in UK health care, but chiming with other experts, he says that more evidence for clinical use is needed. Some cancers are more challenging even in a conventional consultation, says Roope, and these are more difficult still on a video or telephone consultation. “This is where AI has tremendous potential”, Roope remarks. “E-Consult [an online consultation platform] is being used by the NHS and generates questions intuitively depending on answers entered, and providing a positive predictive value for that patient's chance of having a certain cancer.” Remote consultation has value in accessing hard-to-reach patients, and “there is no doubt AI and telehealth have a role in the future, but I feel we are still a long way from having the data to carry out the activities a real live clinician can do”, cautions Marshall. He stresses that patient consultations are about the building of a relationship founded on trust. “The face-to-face interaction is at the heart of general practice and means I can persuade a patient not to seek antibiotics for a sore throat, say, or not to have a scan when it won't add value.” Resonating with the importance of the close patient–doctor interaction, Lopez, who strongly supports telemedicine, recognises the fundamental role of the doctor–patient relationship on effective patient care. “Foremost, we want to focus on the level of patient care not the technology per se. Telemedicine is a way to bring back the house-call of old. We see the patient in their own home environment, we can even talk to the family. We can be right there with the patient.” For more on the rise in telemedicine in the UK see https://www.rcgp.org.uk/about-us/news/2020/april/around-7-in-10-patients-now-receive-gp-care-remotely-in-bid-to-keep-patients-safe-during-pandemic.aspxFor more on cancelled or postponed operations worldwide during COVID-19 see https://bjssjournals.onlinelibrary.wiley.com/doi/epdf/10.1002/bjs.11746For more on the difficulties of diagnosing cancers see Br J Gen Pract 2018; 68: e63–72For more on research that suggests AI might exacerbate healthcare disparities see JAMA Dermatol 2018; 154: 1247–48 For more on the rise in telemedicine in the UK see https://www.rcgp.org.uk/about-us/news/2020/april/around-7-in-10-patients-now-receive-gp-care-remotely-in-bid-to-keep-patients-safe-during-pandemic.aspx For more on cancelled or postponed operations worldwide during COVID-19 see https://bjssjournals.onlinelibrary.wiley.com/doi/epdf/10.1002/bjs.11746 For more on the difficulties of diagnosing cancers see Br J Gen Pract 2018; 68: e63–72 For more on research that suggests AI might exacerbate healthcare disparities see JAMA Dermatol 2018; 154: 1247–48