Author: Olga Zaytseva (Russian Federation)
Co-authors: Vladimir Neroev, Yurii Zhulev, Vera Petrakovskaya, Anna Paleeva, Maria Trifonova, Irina Efimenko
Purpose
The purpose of the study is to analyze Patient Voice (messages from patients with wet age-related macular degeneration (wAMD) and diabetic retinal disease (DRD), or their caregivers) from open Internet resources in the Russian Federation, using artificial intelligence (AI) methods: technologies for automated analysis of unstructured natural language texts, incl. semantic technologies. The study provides valuable information, not only on patients’ characteristics and treatment patterns in real-world management of retinal diseases, but also on patients’ attitude to the disease, diagnostic and treatment-related procedures, their needs and barriers to treatment. The need for such data is explained by the fact that there are no available comprehensive reviews of country-specific patients’ ‘journeys’ assessed from patients’ perspective and analyzed by extracting insights from ‘big data’. Moreover, recent publications show that the healthcare system is ready to recognize such information as an important source of real-world data (RWD). The intensive development of the virtual environment boosted by COVID-19 pandemic also highlighted the responsibility of healthcare stakeholders to develop digital solutions for different groups of patients (apps, remote monitoring tools, web-based platforms, etc.), and it is not achievable without understanding patients’ digital experience, their ‘digital literacy’, attitudes and preferences.
Setting/Venue
Real-life anonymized patient stories from Russian-language open Internet-sources (forums, social networks) were processed with the technologies of natural language understanding (NLU). Patients and caregivers (mainly, relatives) provide direct insights on their journey in diagnostics and treatment in an open and ‘uncensored’ way while looking for a second opinion or exchanging their experience and supporting each other. They use social networks (e.g., Facebook, Instagram, and country-specific networks), disease-related forums, and Q&A portals. There were 73,098 DRD/wAMD-related posts identified including 13,138 posts by 844 DRD patients and 358 posts by 212 wAMD patients.
Methods
The key concept was ‘Patient Voice’ - a source of the patients ‘reality’ which is applied by regulatory bodies (e.g., FDA), pharmaceutical companies, patient advocacy groups, etc. It was analyzed in several steps with semantic technologies aimed at processing large volumes of unstructured information (text semantic analysis considering not keywords, but the whole structure of the natural language and its meaning). 1. Crawling: gathering unstructured data (user posts) from open Web 2.0 sources. After additional filtering based on the system of heuristics (selecting only ‘target’ discussions), plain text data were converted into XML representation. Anonymized patient stories matching the criteria of the ‘diseases of interest’ were selected. 2. Extracting information from each post by NLU algorithms based on hybrid approach to text mining, including pattern-based approach similar to expert systems and machine learning by the Semantic Hub platform. It was driven by the formal domain models called ‘ontologies’ i.e. semantic networks or directed graphs showing the domain concepts and relations between them. Output results were structured in a JSON format, then merged with intelligent algorithms and loaded into a ‘knowledge base’. 3. After unstructured ‘big text data’ were transformed into structured base, traditional biostatistics methods and visualization tools were applied.
Results
Low ‘digital’ awareness and activity of wAMD and DRD patients was detected compared with the diseases with lower prevalence (breast cancer, multiple sclerosis, etc.). The majority of content for DRD was produced by relatives (82.6% of messages), in wAMD - by patients themselves (65%). The key focus for DRD patients was diabetic microvascular manifestations (over 42,000 posts for ‘diabetic foot’ and only 681 for ‘diabetic retinopathy’). Quality of life (QoL) was shown to be significantly affected with inability to work as a major burden for 30% of wAMD patients, and diabetes-associated comorbidities as a key factor compromising QoL in 20% of DRD patients. In wAMD patients mean time-to-diagnosis after disease manifestation was 1 year, in DRD more than a half mentioned 1-2 months. Key reasons for visiting the clinics included in-depth eye exams (OCT mentioned by 27.6% of wAMD patients) and treatment (24.1%) visits. The proportion of DRD patients seeking primary advice of ophthalmologist was higher compared to wAMD (23.3% vs 8.6%). Only 33.2% of wAMD patients and 7% of DRD patients noted that they received anti-VEGFs with affordability as a key barrier mentioned. Patients demonstrated the lack of clear understanding of prognosis and effective treatment options.
Conlusions
The study revealed low activity and awareness in wAMD and DRD patients compared to other ophthalmological and systemic conditions. The awareness on other microvascular diabetic complications in DRD was obviously higher in comparison to other eye disease. This justifies a need for increasing ‘digital’ literacy and awareness not only of patients (it may be challenging due to the age of wAMD patients and their access to digital environment in general) but of their younger relatives as well. The results confirm that vision-related quality of life is compromised among the studied population that could be used for the justification of the appropriate resources allocation in negotiations with health authorities. Based on healthcare experience reported by patients with two vision-threatening retinal diseases it has been demonstrated that there is a need for improvement in at least several points in their ‘journey’ - time-to-diagnosis, treatment choice and availability of effective treatment options. There is also a need to improve patients’ awareness on effective therapeutic options, and efficacy criteria in parallel with improving the access to anti-VEGF therapy.
Financial Disclosure
No
Comments
-