User:Bisast/sandbox

= Chatbots for long-term health management = The concept of chatbots began at the early 1960s. Chatbots have evolved over the decades and are used for long-term health management and offer the potential to improve users' health by providing personalised health information and support through a conversational interface. Health chatbots now assist users in managing health conditions, tracking progress, and providing reminders.

Health chatbots use natural language processing (NLP), machine learning algorithms, and artificial intelligence (AI) to analyse the user’s data and other existing data sets to meet their unique health needs and preferences. The technology has been found to help individuals take ownership of their health by providing them with relevant and actionable advice that contributes towards their long-term health goals.

Overall, chatbot AI for long-term health has the potential to revolutionise the way we manage our health and well-being by providing us with personalised and accessible support that can help us lead healthier and happier lives. However, despite their great potential and rapid development during the Covid-19 pandemic, there continues to be hesitance about some applications.

Early Developments
Alan Turing first accredited the chatbot idea after publishing the well-known article “computing machinery intelligence” in 1950. However, based on copies of Turing's correspondence, he considered machine intelligence as far back as 1941, during the second world war. The initial Turing Test was originally called The Imitation Game and was developed to ascertain whether humans could reliably distinguish between interactions between machines and humans. If unable to tell the difference, the machine would have passed the test. The test itself has become a much-debated topic.

The earliest iteration of a chatbot used for health was ELIZA in the 1960s. The computer program simulated talking to a therapist using predetermined scripts to formulate responses to users. The early chatbot was the first to deceive users into thinking they were interacting with another human, known as the “Eliza Effect”.

In 1972, the chatbot PARRY was developed to simulate the behaviour of humans with schizophrenia for teaching purposes. The creator, Colby, believed that PARRY, loosely on ELIZA’s technology, could lead to therapeutic tools capable of regularly assisting many mental health patients. A decade later, Racter was developed. Racter created structured sentences and told stories sophisticated enough to author a book called “The Policeman's Beard is Half Constructed”.

In 1988 Jabberwacky was developed and was the first AI chatbot that learned through conversation with users. By recording content from the conversations, it could use users' words for future conversations. However, Jabberwacky could not contextualise conversations which often led to unsensible conversations with users.

Personal Computer Integration
The first chatbot can verbalise its communication to a certain extent, although at a basic level, was developed by Dr Sabaitso in the early 1990s.

Coinciding with the disruptive innovation of the personal computer and the internet, ALICE (Artificial Linguistic Internet Computer Entity) became the first online chatbot in 1995. ALICE was based on pattern-matching technology without reasoning capabilities.

In 2001 Smarterchild became the first chatbot to assist people in daily computer-based activities, with the ability to retrieve data from multiple databases.

With advances in AI and natural language processing, more sophisticated chatbot systems started to appear. These were still primarily used for providing health information but could also guide users through symptom checkers or other diagnostic tools.

The development of Artificial Intelligence Markup Language (AIML) allowed ALICE to have a programme upgrade and win the Loebner prize for being the most human-like computer programme of the year.

Messaging Apps
The development of chatbots has run parallel to other complementary innovations such as smartphones, WIFI technology, social media messaging services and stronger artificial intelligence, which has opened up the accessibility of chatbots. Chatbots advanced using machine learning and AI to create more human responses. Stronger AI chatbots allow for flexible non-scripted conversation between users. Developing smartphone-based personal assistants like Siri and Google Assistant opened up new possibilities for health-related applications. These assistants could provide health advice, remind users to take medication, and even help users contact healthcare providers.

Notably, Woebot was developed in 2017 as a chatbot designed to offer cognitive behavioural therapy through a mobile phone application. Other chatbots that could offer personalised health advice based on user-inputted symptoms, such as Ada in Germany, were developed. Please see the clinical applications below

As AI applications have become more sophisticated and the prevalence of phone use increases, more people are using chatbots such as Microsoft Xiaoice, the most used chatbot integrated with Facebook Messenger and WeChat. It was designed to mimic empathy and to build deep connections in an attempt to anthropomorphise the chatbot.

COVID-19
The COVID-19 pandemic overwhelmed healthcare systems and needed other means to get the population timely and accurate health information. The World Health Organization (WHO) created an artificial intelligence chatbot using WhatsApp and Facebook Messenger to disseminate information about COVID-19 and answer questions related to the pandemic, reaching over 4 Billion people by the end of 2020. During the pandemic, the popularity of chatbots has increased, and it has been found more patients are willing to use health chatbots in the future. The virus has been found to impact mental health and other chronic conditions that would further affect long-term health management beyond the pandemic. This also made people consider the importance of their health, making long-term health management a priority during a time of isolation and uncertainty.

Chat GPT
Natural language processors such as Chat GPT become more advanced in the 2020s. The AI processes publicly available data on the internet and mimics language. Please see the industry and systems applications below.

Political & Social Developments
An unhealthy lifestyle is linked to the development of chronic diseases, which accounts for 90% of the total patient spending on healthcare in the United States. Chronic diseases are long and progress slowly, including cancer, cardiovascular, respiratory diseases and diabetes. Therefore long Term Health Management is critical to prevent and treat chronic diseases.

In 2021, the WHO released its Global Strategy on Digital Health, which aims to improve patient health outcomes using digital health technologies such as AI and machine learning. With many people already turning to digital technology to obtain health information, the use and advancement of chatbots have increased.

The above advances have converged with a philosophical change in how medicine is conducted, from a reactive to a proactive system known as P4 (Predictive, Preventive, Personalised and Participatory) Medicine. It has been suggested that digital health measures such as chatbots can help towards health promotion and disease prevention by delivering P4 medicine.

Obesity & Weight Management
Obesity and poor lifestyle are leading causes of long-term chronic diseases; preventing these will save lives and positively impact long-term health management. Physical activity, healthy diet and weight management can prevent the development of chronic conditions, and several chatbots have been developed for this purpose.

In 2019, CoachAI was developed to gather patient information such as BMI, physical activity and diet and assign the user activities and feedback to mimic a health coach. Since then, It has been found that Chatbots can improve physical activity, but further studies are required to test the efficacy of chatbots on diet and weight management.

Conversational agents have since been used to promote physical activity and good nutrition through connecting to activity trackers to personalise the chatbot responses further.

Diabetes
Diabetes has a large economic impact on healthcare, and the number of people with diabetes is predicted to increase year on year, and the use of chatbots can assist with its prevention and treatment. In 2019, a chatbot was developed to educate current diabetes patients and provide personalised pre-diabetic predictions by analysing existing health datasets.

Mental Health
Whilst chronic conditions are often defined as physical conditions, some definitions include mental health illnesses. Mental disorders account for more economic costs than chronic diseases, and the direct healthcare costs pressure the healthcare system. Using chatbots for mental health allows patients to have 24/7 access to therapy.

In 2017 Woebot, a cognitive behavioural therapy app, was developed using an instant messenger service. After two weeks, those who used the app experienced reduced depressive symptoms. Other studies have found similar results from AI mobile phone applications such as Wysa in later years.

Oncology
IBM’s Watson was first used commercially as a virtual assistant and was eventually used as a diagnostic tool for illnesses such as cancer in 2011. The machine learning system was found to be based upon insufficient data and ultimately led to poor diagnosis and its demise. Since then, chatbots have been developed to help cancer patients in various stages, from treatment, monitoring, support and general health promotion. In 2017, a chatbot, Vik, was developed to help cancer patients, providing personalised messages to the patient. It was found to support breast cancer patients and improve medication adherence.

Other Chronic Conditions
A study which created a chatbot for patients with psoriasis shows that chatbots can be adapted for any chronic condition rather than focus on special conditions as much of the existing studies do.

Industry & System Applications
Other chatbots have been used in healthcare, which can assist patients with long-term health management in a broader sense.

Disease Diagnosis
In the UK, the National Health Service used the privately owned Babylon Health chatbot to check user symptoms to assist with diagnosing conditions. Similar services have been launched in Germany with the chatbot Ada, allowing patients to care for their health. In a study comparing doctors' diagnoses of conditions to Babylon Health, Babylon Health outperformed the doctors in the study. However, the Lancet later published an article discrediting the methodology. Despite this, it is thought that future symptom checker chatbots will allow patients to take greater ownership of their health.

Medication Adherence
Poor health management and chronic diseases are a major burden on healthcare systems, and poor medication adherence contributes to increased morbidity. Chatbots such as Emma have been developed to remind patients to take their medication, check for food interactions and collect compliance data that is effective for medication management.

Electronic Health Records
With the increasing digitisation of health records and the development of the Internet of medical things, chatbots started to be integrated directly into health systems to ensure more meaningful use of the data gathered. They could now access patient records to provide personalised advice and facilitate telemedicine appointments with permission.

Using standards such as Fast Healthcare Interoperability Resources (FHIR) is necessary to be interoperable with healthcare systems and provide quality healthcare. In 2020 a chatbot for chronic diseases was developed, which incorporated standard data-sharing models with FHIR to enable interoperability with electronic health records.

Chat GPT
It has been hypothesised that Chat GPT will not replace human capability in healthcare; however, it can be used as a valuable assistance tool in health management. Such AI applications are competent at medical note-taking, providing medical knowledge such as diagnosis information and are expected to be used by more and more clinicians and patients if potential risks are overcome.

Major Implications and Impacts
The conversational platforms help patients using Artificial intelligence to respond to queries. With ever-changing technology in the world, it is a valuable addition to healthcare that provides support to medical professionals. The COVID-19 pandemic has amplified the use and growth of Chatbot technology. Chatbots have potential benefits and some negative impacts on health care. Studies have surveyed physicians and patients on using chatbots in healthcare, which have found both positive and negative impacts.

Positive Implications
Chatbots can respond to unlimited queries by users providing a constant source of interaction. It can help users with tasks such as scheduling medical appointments, locating health clinics, and providing medical information and diagnostics support. Other long-term health applications have been to promote and increase physical activity and cognitive behaviour therapy for psychiatric and somatic disorders. Please see the clinical applications above.

Healthcare Accessibility & Patient Participation
Chatbots help end users to have prompt answers and proactive behaviours like notifications and reminders. When chatbots are coupled with social media, it offers increased access to health information and encourages patient participation in preventing illness and care.

Confidentiality
Chatbots allow patients to obtain health information on personal private health issues without talking to real humans. Therefore patients feel less shame and are comfortable interacting with AI chatbots. In a study conducted in 2020, interviews conducted with participants stated chatbots provided sexual health services in a convenient, anonymous and non-judgmental way.

Personalised Health Information
Personal Health information from wearable sensors such as smartwatches and devices with facial recognition can assess real-time biomarkers. This can integrate with chatbots to provide more valuable and personalised health information.

Cost-effectiveness & Preventative Medicine
Chatbots can cost-effectively target large audiences. Chatbots can be deployed quickly to screen many patients before accessing a healthcare facility, as seen in the COVID-19 pandemic. Chatbots can screen and prioritise patients by providing appropriate education and information to clinicians. Alternative healthcare models are more cost-effective than current healthcare systems, such as telemedicine, by removing the human element.

Diagnostic Accuracy & Predictive Medicine
Despite early failings (see clinical applications above), chatbots can provide superior diagnostic accuracy based on having access to health data from large data sets from other patients. User-facing applications also talk to patients in real-time and provide advice and instructions based on probabilities. It has been hypothesised that these tools will get more accurate over time.

Monitoring Patients & Follow Up
Chatbots are an effective solution for monitoring & following up with patients. Chatbots can interact with patients and follow up to schedule future appointments in chronic conditions like cancer. This improves patients' quality of life and saves health professionals time.

Greater Efficiency
Using chatbots in healthcare with AI technology can minimise errors and increase efficiency by providing predictive analysis of real-time information that can reduce the workload for health professionals. Clinical data analysis can drive some clinical decisions to improve health outcomes.

Negative Implications
Despite the benefits of chatbots in healthcare, it has been suggested that there could be technical and ethical debt. Ethical debt occurs when unintended consequences are not considered. There have been several published examples of this. One of the most concerning was a study which demonstrated racial bias in a hospital-based algorithm which inappropriately identified unwell patients as needing less care than those of a differing racial background.

Risk of Bias
Chatbots' overt appearance and design may lead to design biases and ethical debt towards certain ethnic groups. The knowledge base used to train the chatbot algorithm and the knowledge it acquires can create bias. Systematic errors made by software programs may provide unfair outcomes, such as preferential treatment of certain population groups. Datasets with issues such as missing or misclassified data with measurement errors or small sample sizes may misrepresent subgroups leading to inappropriate recommendations. Chatbots from different countries and cultures may introduce design biases if other demographic-specific needs are not considered. To minimise bias, software programmers need to consider the population as a whole. This can be achieved through adequate research and training algorithms to minimise the risk of bias.

Risk of harm
Due to the limitation and incompleteness of healthcare chatbots, they cannot provide detailed clinical assessments of a patient's health. If chatbots do not have a complete clinical history of the patient, they are at risk of providing incorrect responses, which can lead to poor patient outcomes. Chatbots must consider medical regulations, ethical codes and the latest research data. Patients may become overly reliant on Chatbots to self-diagnose health problems, leading to ethical issues.

Privacy
Privacy is an important aspect of healthcare delivery. With the large amount of sensitive health data collected from users, chatbot administrators must be aware of local data protection laws. The technology must inform end users of the limits to privacy and how their data is stored.

Inequitable Access
Not everyone will have access to a chatbot. In less developed countries, poor technology literacy and cost will reduce the number of users. The viability of chatbot technology relies on users' access to technology and literacy.