This page gives a summary of what I have learnt in a few online courses that I have done.
Brighton and Sussex Medical School Online Work Experience
This virtual work experience is designed to provide an insight into some of the points that you would expect to learn in real-world work experience about being a doctor in the NHS, or supplement real work experience with points that may not have been covered. I am doing this as my work experience that had been scheduled for the summer got cancelled, and, although I will hopefully be able to reschedule it to sometime soon, at the moment I hope that this is course offers a valuable experience. The course promises opportunities to reflect on our experience, so I hope that, through learning this now, when I go to real work experience, I can do this more easily.
Module One – The NHS and General Practice
INtroduction to the NHS
The NHS is in the spotlight at the moment as the most appreciated institution in the country. We first looked at this institution, and how it works.
The NHS was launched in 1948 by health minister Aneurin Bevan, based on three core principles:
- Meet the health needs of everyone – enabled everyone to get help with any health problem(s) that they had. People with different problems could access healthcare through the same system. At the time, this meant physical health problems, and it wasn’t until 1959 when the Mental Health Act came in which meant that people with mental health problems could access their services.
- Free at the point of delivery – it was free for people to enter and to access healthcare. This meant that you could see a healthcare professional for free to be directed to other services if needed. In 1952, prescription charges of 1 shilling were introduced (except for children under 16, those in receipt of National assistance, a war disability pension or with a venereal disease).
- Based on clinical need, not the ability to pay – everyone could access all the care that they needed, meaning that the amount spent on one person would differ to the amount spent on another person; this doesn’t affect the care that they receive. Care is based on need.
Over the years, the NHS and scientific community have made many groundbreaking discoveries:
- Establishing the link between smoking and lung cancer.
- Performing the first organ transplants.
- Establishing an organ donation register.
- Rolling out a national vaccination programme.
- Here are further discoveries that have shaped today’s NHS.
There are 7 underlying principles that form part of the NHS constitution, along with 6 key values. Information can be found here, and on my NHS values post.
The NHS is a government-funded, public health service. Contrast this with private healthcare, which usually requires health insurance or direct full payment. Everyone who pays taxes contributes to the running of the NHS. Here is the NHS funding flow, which shows more detail into the spending of the NHS, up-to-date as of April 2020.
As you can see from the flow above, CCG act as a medium for determining spending on a local scale. Broadly speaking, they buy services from 3 areas of care: primary care (community-based care which is one’s first approach to health services, including GP surgeries, drop-in clinics, dentists and opticians – 90% of all patient meetings, but only 7% of budget), secondary care (the provision of higher-level care in a centre with multiple specialist staff and resources, such as a hospital. They can’t usually be accessed without a referral or via emergency administration – 70% of budget), and tertiary care (very high-level care usually focusing on one discipline such as a hospice or neurorehabilitation centre – only accessed through referral from primary or secondary care).
Here are some facts and figures for the NHS. It serves approximately 500 million patients each year and has 112,031 doctors and 311,380 nurses. Regarding what people think of the NHS: 84% say their experience of general practice is ‘(very) good’, 96% would recommend their hospital/ward for healthcare, and 96% would recommend their community health services. About 77% of people believe that the NHS is crucial to British society, and we must do everything we can to maintain it. I would like to see how this has changed in light of the COVID-19 pandemic.
What does a gp do?
GPs constitute the largest branch of British medicine, and the NHS aims to train half of its doctors to be GPs. Many consider GPs to be the backbone of British Medicine, to the extent that a recent BMJ headline said: ‘if general practice fails, the whole NHS fails’.
A GP must have the ability to understand the multiple conditions that patients have, as well as their wider mental health and well-being. This will allow the NHS to manage the challenges of the changing population. Their knowledge will mean that healthcare can truly be designated around the needs of the patient; this is why NHS leaders are recognising the importance of general practice and committing greater investment in it. As time progresses, a career as a GP will become more and more intellectually and medically challenging, diverse, and fulfilling. GPs will have portfolio careers heading multidisciplinary teams, leading work in areas from geriatrics to neurology, running ‘in-reach’ to hospitals and ‘outreach’ to patients’ homes. And GPs will be closer, and more important, to their patients than ever before.
GPs deal with a whole variety of issues, from psychiatric issues to newborns, seeing 40 to 50 patients per day. They are also often involved in medical training, CCGs, research, home visits, health promotion and many more. Some work for football teams, and it is a very flexible job, offering lots of different opportunities and different roles to take on.
When a GP sees a patient, they ask them why they came in, and take a patient history, which involves asking a series of questions, and listening to obtain information about a patient to try to work out what is wrong. It takes years to master this skill, and it requires excellent communication skills, as well as medical knowledge, to know what to ask, and how to ask it. Building a rapport with patients is key to make them feel comfortable disclosing potentially sensitive information. A rapport can be built in a variety of ways: showing empathy, being non-judgmental, listening carefully, using open questions, maintaining open body language etc.
80% of diagnosis can be made on the patient history alone. The specific questions asked in this period are dependent on the presentation, but the general structure is:
- Introduction and consent – introduce yourself and ask for consent to ask questions/take notes.
- Patient Demographic – confirm key details about the patient, e.g. name and DOB.
- Presenting Complaint – discover information about why they are here today. Give patients the ‘golden minute’, where they speak uninterrupted for one minute and this is where most information comes to the surface.
- History of Presenting Complaint – Delve into further details of the complaint. To remember the key information to gather, SOCRATES mneumonic is often used:
- Site (where is the pain).
- Onset (when did it start).
- Characteristic (type of pain).
- Radiation (has the pain moved since it started).
- Associated symptoms (are there any other symptoms).
- Timing (is there any time in the day that makes it better or worse)
- Exacerbating or relieving factors (does anything make the pain better or worse).
- Severity (pain scale 1-10).
- Past Medical History – questions on previous ailments/conditions/medical procedures etc.
- Drug history and allergies – To get an idea of prescribed, or over the counter medication that the patients have been taking. The presenting complaints could be a drug allergy. Noting all known allergies is important in prescribing medication, but some chronic conditions have certain allergies associated with them, e.g. pollen, dust and mites are key triggers of asthma.
- Family History – Many risk factors and illnesses are genetic. This history can be indicative of predispositions, and also rationalise any concerns that the patient may have.
- Social History – Builds a picture of the patient’s day-to-day life, e.g. living environment, occupation, diet, smoking, alcohol, recreational drugs – being non-judgmental.
- ICE – addressing concerns of the patient, and managing expectations of treatment/management options. Then summarise information before explaining the next steps.
- What might the problem be?
- Is there anything that worries you?
- What are you hoping we can achieve?
Sometimes a GP may need to do a referral or test if they need more information to assess and manage a patient. Tests include a blood test or an ECG, and can often be done in the GP surgery. Some tests can’t be done in a GP surgery, such as an X-ray, so the GP needs to refer the patient to the hospital. If the GP thinks that a patient can only be managed with specialist input, they can refer them to secondary care services, which are normally hospital-based. The GP must therefore liaise with the hospital and exchange patient information. Hospitals often contact GPs to gather important clinical information that patients can’t always provide, or forget, such as a list of current medication. If a patient is discharged, the hospital staff will send a discharge letter to the GP summarising why they were admitted and what happened in hospital.
A GP also has to promote health to their patients. Through:
- Consultations with an individual patient.
- Engaging patients in national screening programmes that look for serious health problems that can silently develop over time, such as a mammogram (breast tissue scan).
- Participating in public events such as conferences to spread ideas on good health promotion and disease protection.
- Signposting to patients the services that focus on promoting certain types of health such as sexual health or drug and alcohol services.
- Providing written information such as leaflets, which patients can take away with them.
Some GPs are involved in the management of their practice, and are called ‘GP partners’. They are involved in organising staff and funding, ordering equipment, and complying with different rules and regulations that all GP practices in the UK should follow.
Some are involved in teaching: supervising medical students during placements, lecturing at medical schools, teaching doctors in after medical school in GP training. But perhaps the biggest role that GPs have in teaching is in educating patients. This is a huge part of the job and can be anything from advising on the risks of certain behaviours (e.g. smoking) to teaching a patient how to use an inhaler.
Sometimes GPs have to visit a patient in their home because they are too unwell to come to the GP surgery, or may have a long-term health condition that means that they can’t leave their house. It also allows the GP to assess the environment that they live in and see how this impacts their health.
A GP also has the opportunity to contribute to medical research and academia. General practice provides great opportunities to do research. They can lead all aspects of their research and present their findings at scientific gatherings, such as conferences.
GPs are increasingly being encouraged to take on additional responsibilities, such as becoming a ‘GP with Special Interests’ (GPwSI). A GPwSI has taken further training to become competent at managing more complex problems that they are interested in. For example dermatology, diabetes, geriatrics, sexual health problems, paediatrics, managing symptoms in patients with chronic pain.
These are only some of the things that a GP does!
Here is some more information into the work of a GP.
GPs can make a massive difference to people’s lives. For example, teenagers with spots (acne) can be very self-conscious, and acne can have a massive impact on their lives. By providing mild acne creams (or antibiotics) to solve this, you can build up people’s confidence and make a massive improvement on their lives. Sometimes, a GP can have to refer acne patients to specialist dermatology services, where other treatments can be prescribed.
But, when treating any disease, it is important for the doctor to take a holistic approach, listening to the impact of the disease on the patient, and what part of the patient it affects. The GP then needs to think about medications and lifestyle changes which the patient could make, or whether the patient needs to see another member of the team. Different diseases affect different people in different ways, and it can be difficult for a GP to cover everything in the 10-minute appointment.
Other common problems in medical practice (by speciality):
- Dermatology (1/3 of people have a skin problem worth treating).
- Musculoskeletal Medicine for treating muscles bones and joints (1 in 5 consultations are for musculoskeletal problems).
- Back pain
- Sprains and strains
- Injuries and rehabilitation (e.g. broken bones)
- Ear, nose and throat (ENT) – very treatable (GP expected to deal with lots of these themselves) or self-limiting, but can have a big impact on health.
- Ear infections such as otitis media (a common ear infection in children that usually clears by itself).
- Antibiotics are good for treating bacterial infections but not other types of infection. Giving patients them inappropriately is expensive and can lead to future resistance to treatments.
- Managing chronic diseases is one of the biggest roles, and challenges, of being a GP.
- 70% of the NHS budget is spent on managing chronic diseases.
- Rather than curing chronic diseases, doctors aim to manage these diseases, meaning that they aim to stop them from getting worse, and reduce their impact on patient’s life and health
- Some are curable, such as alcoholic liver failure can be cured by a liver transplant, but this doesn’t always happen (e.g. limited donated organs etc.).
- Examples include:
- High blood pressure (hypertension)
- Rheumatoid arthritis; pain and stiffness in many joints of the body.
- Chronic diseases are becoming more and more common and can be difficult to manage.
- Older patients are particularly susceptible to suffering from two chronic diseases at the same time. This is called comorbidity, and if someone is suffering from more than two, this is called multimorbidity.
Pathway to becoming a gp
Next, we looked at the pathway to becoming a GP, which involves going to medical school, foundation training, and then about 3 years of GP training, half of which is in hospital, and half of which is in a GP surgery. You then become a GP and may want to specialise (GPwSI).
who else works in a gp surgery?
Other people that work in a GP surgery include:
- Practice nurses are important for clinical activities that aren’t complicated such as taking blood or delivering vaccinations.
- Specialist nurses run clinics for patients with chronic conditions such as asthma and diabetes or provide smoking cessation or family planning services.
- District nurses work in the community more than the GP surgery. They visit homes and provide care to patients who are often less mobile and have difficulty leaving their house to access treatments, e.g. looking after patients with diabetes or providing good wound care.
- Physician Associate
- Supervised by GPs and trained to undertake clinical tasks, such as taking medical histories, performing basic physical investigations, analysing test results, assisting with management pans and providing disease prevention advice.
- They look after a woman and their baby during pregnancy in the community, with the help of doctors when required.
- Patients usually see the same midwife throughout their pregnancy and in any subsequent pregnancies.
- Healthcare Assistants
- Provide support for nurses in taking blood samples (phlebotomy), blood pressure and weight measurements for those attending various clinics.
- Dispense medication to patients safely.
- The pharmacist checks to make sure that the prescription is safe and appropriate and then dispenses the required amount of medication. They may need to give advice about taking the medication as well.
- Administration staff
- The day-to-day running of the surgery,
- Jobs include: booking appointments, answering patient queries and organising patient notes.
- Practice managers
- IT technicians
- Social workers
- Medical/healthcare students
- And many more.
Challenges of general practice
Currently, general practice is facing some of its biggest challenges in history. Record numbers of patients are stretching general practice to breaking point meaning patients are finding it harder to get appointments, even for routine problems. Some of those that are struggling are turning up to A&E, so minor problems that could be dealt with by a GP are being treated by a service for managing emergencies, putting pressure on A&E departments that are already struggling to cope with high numbers of patients, and wasting valuable, but limited, resources.
The number of elderly patients being treated by the NHS is higher than ever and is continuing to rise. Elderly patients are generally more likely to have health problems and to be on medications for these problems, and with an ageing population, the number of elderly patients is rising, with diseases that are more difficult to treat, along with many diseases occurring together, making management complex and costly. These patients will be on lots of different medications, which are expensive, difficult to balance and have lots of side effects.
Training, recruiting and retaining new GPs are now becoming more difficult than ever. Training programmes for GPs are continuously undersubscribed, meaning there are not enough GPs to fill all the jobs that are available, and 50% of GPs are predicted to retire before retirement age, meaning the GP workforce will be drastically cut in the near future. This (pg 4, 5 and 6) details what is responsible for the increasing pressure on GPs and the multidisciplinary team. Furthermore, lots of GPs are having to close. This is putting mounting pressure on individual GPs, which can be very stressful. Thisfinal publication by the National Primary Care Network (NPCN) talks about some of the different pressures of being a GP and proposes some discussion on how we approach them in modern-day general practice.
more information on being a gp
The Royal College of General Practice (RCGP) is the UK professional body for family doctors, promoting excellence in primary healthcare. They have numerous resources on all aspects of being a GP including training, job profiles, interviews with current GPs, and further career options within general practice. Visit their website for details.
Additionally, a useful publication by the RCGP called ‘Think GP‘ talks about the benefits of working as a GP in the NHS.
The GP National Recruitment Office is responsible for coordinating the process of recruiting doctors for training to become GPs. Their website has a wealth of information and particularly good case studies on the good and challenging aspects of being a GP.
Interesting questions to consider:
- If you were given £1,000,000,000 to improve one aspect of general practice in any way, how would you invest the money and why?
- How do you think the development of new technologies (e.g. internet, apps) has impacted on the way general practice can be delivered in the UK? Do you think that telephone / live video consultations between doctors and patients are a good idea?
- Some areas of the UK are struggling to attract GPs to work there. Recent plans to tackle this include paying GPs a financial incentive to move and work to places with the most desperate need for GPs. Read this article from BBC to find out more about the plan being proposed for use in Wales, UK. Do you think that this is a good idea for attracting GPs to train and work in rural areas? Can you think of any other ways of doing this?
- It has been difficult to promote general practice as a career choice for medical students, and students often prefer to train to become hospital doctors. Why do you think this is? What would you do to increase recruitment into general practice?
Module Two – Elderly Medicine
Since the elderly get diseases more easily than the younger generations, looking after the elderly is a core part of being a doctor. The complex and growing needs of older patients mean that managing their medical problems involves a coordinated management plan which is delivered by a multidisciplinary team. In addition, the population of elderly patients in the UK is rising, meaning more patients with increasingly complex medical needs demand higher levels of care. This is stretching an already pressurised workforce with limited resources, both in the community and in hospitals.
Introduction to elderly medicine
Ageing is a normal process by which the structure and function of our bodies alters with time, usually accompanied by negative impacts on our health. We don’t know what causes it but it could be genetic and environmental factors. Women tend to live 2.8 years longer than men on average.
Elderly people are at greater risk of developing disease, including chronic disease. Chronic disease puts people at risk of other complications, meaning the sick get sicker. With many diseases , there may be polypharmacy where a patient is on drugs that interact with each other in a negative way. Older people with many diseases are harder to treat successfully, which makes elderly patients among the most difficult to treat. Diseases that they may get include:
- Nervous System
- Parkinson’s Disease
- Cardiovascular System
- High Cholesterol
- Angina/Heart Attacks
- Heart Failure
- Abnormal Heart Rhythms
- Gastrointestinal System
- Reduced vision
- Hearing impairment
- Respiratory System
- Musculoskeletal System
- Brittle Bones
- Weakness and fragility
- Reduced Mobility
Ageing comes with many losses:
- Wealth (retired – no working salary)
- Mental Health (loneliness and above reasons and other age-specific conditions, e.g. dementia)
- People (lose friends, partners etc.)
- Roles in society (less able to contribute to society)
- Discrimination (unfair perceptions of old people and their increased demand on public services)
Elderly patients are looked after in hospitals or other secondary care centres, residential care homes, or homes of family and friends, and a whole variety of people look after them:
- Friends and family
- Occupational Therapists
- Speech and Language Therapists
Respect and dignity are vital values when treating the elderly as they are often at the stage of their life where they are embarrassed and feel out of control. Therefore, we need to give them time to talk and express concerns, educate people on the importance of treating the elderly with respect, give the elderly all the time they need to complete tasks, provide positive reassurance.
For nurses there are ‘6Cs’ which ensure patients are looked after with care and compassion, by competent workers who communicate well, dare to make changes that improve care and can commit to delivering this all day, every day.:
Interestingly, one challenge of working with elderly patients is that they have a changed attitude to health. For example, they don’t think smoking is too bad because the damage has already been done to them. Furthermore, there are many ethical ramifications of working with the elderly, and there are regularly moral dilemmas.
‘Medicine is far more about managing expectations than actually curing disease.’
The elderly ward round
In a ward round, a team of healthcare professionals reviews each patient that they are responsible for and makes a plan for each patient. This occurs on a daily basis. The role of the ward round is to establish:
- Why the patient is still in hospital.
- What tests that patient needs, or has had and discuss the results.
- The medications that the patient should or shouldn’t be taking.
- Whether the patient is managing with normal tasks (e.g. sleeping, eating & drinking, going to the toilet)
- A plan for the patient’s admissions, including when they can expect to be discharged.
We then looked at examples of a few patients, looking at their history, making a diagnosis and a management plan.
Other health problems in elderly medicine
Parkinson’s Disease is something that I have done a blog post on. It is a common, yet incredibly debilitating neurological condition, which the elderly are particularly vulnerable to. It is caused by a decline in the Substantia Nigra, which results in low levels of dopamine.
There is no definitive test for the disease; doctors have to rely on the clues that they pick up in the clinical history and the physical examination. Scanning the head, for example, using a CT or MRI scanner, can help with finding other causes for the symptoms, but cannot help with determining if Parkinson’s disease is happening. Though this could be about to change, with a retired nurse who claims, with successful evidence that she can smell Parkinson’s.
Arthritis is another chronic disease that the elderly often suffer with. More than just ‘stiff joints’, arthritis can have a crippling effect on all aspects of an elderly person’s life. In fact, about 10 million people in the UK suffer with some form of arthritis, which is expected to increase by 50% by 2030. There are two main types:
- Osteoarthritis – ‘wear-and-tear’ arthritis; the result of normal lifelong use of the joints (and the focus of this part of the module).
- Rheumatoid arthritis – the body’s immune system attacks the joints leading to inflammation and damage.
Arthritis means ‘joint inflammation’, joints become stiff, painful, red and swollen, and any movements about the joint become gradually more difficult. A holistic, multidisciplinary approach is important in the management of all types of arthritis.
Another problem is loneliness. Again, I have looked at this in a blog post before. Three-quarters of elderly people are lonely, and over half of them don’t speak to anyone about it. 2 million people aged over 75 years live alone in the UK, and more than a million of these often go over a month without speaking to someone close to them. Loneliness has a variety of causes and is very dangerous, and can have many ramifications, leading to further and worsening health problems, creating a downward spiral in health and well-being. Furthermore, the stigma surrounding loneliness means older people don’t reach out for help, even when they desperately need it. To tackle loneliness, advice to elder peoples should be to:
- Fill the diary.
- Get involved in the community.
- Make use of technology.
- Pick up a new skill.
- Attend events organised by charities (e.g. Age UK).
Module 3 – Mental Health
Mental health is a branch of medicine that presents its own unique challenges. In psychiatry, doctors encounter some of the most unwell patients, and the nature of mental illness means it’s incredibly difficult to spot from the outside. In fact, mental health problems in patients are encountered by doctors in all fields of medicine; not just psychiatry. The complexity of the brain means that we only have a limited understanding of the scientific basis of mental health problems, and though there are effective treatments for many patients, it can be very hard to find the absolute cure per se. Additionally, the wider stigma surrounding mental illness means sufferers are at high risk of being judged or discriminated against, making their condition even worse.
Introduction to Mental Health
Some stats to begin:
- 1/4 of people experience a mental health problem in their lives.
- The UK has among the highest rates of self-harm in the world, at 400 per 100,000 people per year.
- Men are 3/4 times more likely to commit suicide than women.
- Mental health problems are responsible for 28% of the total disease burden in the UK, more than any other health problem.
- Mental health issues are responsible for over 21% of all years lived with disability per year worldwide.
Mental health problems can be broadly grouped into:
- Mood (e.g. depression, bipolar affective disorder).
- Anxiety (e.g. generalised anxiety disorder, phobias).
- Schizotypal (e.g. schizophrenia, delusional disorders).
- Substance misuse (e.g. alcohol, illicit substances).
- Personality (e.g. emotionally unstable, borderline).
- Childhood/adolescence (e.g. autism, ADHD).
- Old Age (e.g. dementia).
- Organic (mental health problems with a clear physical cause, e.g. hormone imbalances).
All ages can be affected by these problems. They are determined by a mixture of genetic and environmental factors, known as ‘non-modifiable’ and ‘modifiable’ factors, respectively. Due to the variety of contributors, they are called multifactorial.
It is usually easy to tell if someone has a physical health problem because you can see it. However, there are few external signs that can label someone as having a mental health problem, and people with these issues may try to hide it. This is largely caused by the stigma (shame) around behaving in a certain way with a mental health problem. In fact, 9 in 10 people feel stigma affects them in a negative way. 35-50% of people with severe mental health problems in developed countries receive no treatment.
Mental health problems cost the UK about £70-100 billion, bot only due to the costs of therapies, medication and hospital admissions but, mental health problems also prevent around 200,000 people from working.
Doctors are at an increased risk of mental health problems due to:
- Poor stress management
- Overwhelming workload
- Being emotionally overwhelmed
- Ongoing legal/GMC investigations
- Exam Stress
- And all other reasons that might lead to anyone developing mental health problems, such as relationship or financial difficulties.
Therefore, doctors need to:
- Train themselves to ‘switch off’ after finishing their shift.
- Have relaxing activities and hobbies to do in their spare time.
- Speak to other doctors or HCPs to find out how to manage stress.
- Make sure to know where to get help from.
- Prepare for exams well in advance of taking them.
- If starting to struggle, identify this early.
You never know what anyone is going through, so treating everyone with kindness and respect Is vital.
Depression and anxiety
Depression is a huge problem in the UK and worldwide:
- It is the 2nd leading cause of years lived with disability around the world
- The prevalence of depression is currently 3.3% in the UK but this is rising.
- Nearly 20% of adults aged 16 years plus in the UK show signs of depression.
According to the ICD-10 (a manual used by UK doctors for classifying mental health disorders), you can diagnose depression as ‘mild’, ‘moderate’ or ‘severe’, based on the symptoms a patient experiences:
- Mild – at least 2 core symptoms + at least 2 associated symptoms
- Moderate – at least 2 core symptoms + at least 4 associated symptoms
- Severe – all 3 core symptoms + at least 5 associated symptoms
The three core symptoms:
- Low mood
- Low energy (anergia)
- Lack of enjoyment of activities you normally find enjoyable (anhedonia)
- Sleep disturbance (too much or too little sleep)
- Reduced appetite
- Trouble with concentration and attention
- Poor memory
- Sexual dysfunction
- Feelings of guilt, hopelessness or worthlessness
- Thoughts of self-harm or suicide
The things that affect their mental health can be:
- Education (e.g. performance and bullying)
- Sex and Relationships (including previous partners)
- Existing health problems
- Alcohol use
- Drug use
- Existing or previous psychiatric problems
- Financial difficulties
- Family history
A doctor may do blood tests in addition to a history in order to find any other anomalies that can affect the mood. But no scans or other tests are usually needed.
The thing about depression is that, due to the large number of variables at play, a doctor may need a completely different approach to treating one case of depression compared to another. They can:
- Offer reassurance
- Medications – there a range of different tablets that can be prescribed by GPs for people who are depressed. They are thought to be generally effective in most people, but a tablet alone won’t completely get rid of depression.
- Technology – they can direct them towards some websites and apps that can help people. Moodgym.com is a popular app that allows patients to self-provide a form of Cognitive Behavioural Therapy (CBT), to help them overcome negative thoughts that might be driving their depression. ‘Emoodji’ is an app designed specifically for university students to help with mood issues related to the particular pressures high-level education can bring.
- CBT – as talked about briefly above, CBT seeks to identify some of the thoughts and processes that drive someone’s depression. Formal CBT involves talking through these issues with another human being.
- Referral – sometimes, a GP might want to get extra help with managing a depressed patient and organise an appointment for them with a mental health expert, such as a specialist nurse or a psychiatrist.
Anxiety is far more than just worrying all the time, here are some posters by Healthy Place which aim to describe what it is like:
In it’s worst form, anxiety can manifest as ‘panic attacks’ – periods of intense worrying with mental and physical features, including:
- Dizziness & fainting
- Increased or reduced muscle tension
- Feelings of impending doom
- Low mood
- Poor sleep
- Poor concentration & memory
- Guilt & worthlessness
To help patients with anxiety there are:
- Support groups
- CBT can help people work through some of the thoughts and problems causing their anxiety, and help with prevention and/or coping mechanisms for panic attacks.
- Some medications can help with controlling anxiety, but only when other things are also being done; these medications are not for curing.
Depression and anxiety actually share lots of features, and unfortunately, they often both occur together.
Other psychiatric problems
Firstly, eating disorders, which come in two main types:
- Anorexia nervosa
- Bulimia nervosa
- Low food intake
- Smoking, to suppress appetite
- Excessive exercise
- Distorted body image
- Using weight loss medication
- Symptoms of depression
- Cachectic appearance
- Heart, arrhythmias and low blood pressure
- Digestive, stomach ulcers and constipation
- Metabolic, vitamin/mineral imbalances
- Hormonal, thyroid or sex hormone changes
- Renal, such as kidney stones and organ failure
They can be very difficult to treat as they often don’t feel that they have any problems, treatments include:
- Nutritional support
Schizophrenia is characterised by:
- Hallucinations – perceptions without an external stimulus.
- Delusions – false beliefs about things.
- Thought interference – feeling like your thoughts are being controlled.
Some of the symptoms include reduced speech, withdrawal from social situations and having low motivation to do things. 2/3 of people with schizophrenia have significant symptoms and often need to be hospitalized. And people who have had an episode of symptoms have an 80% chance of getting more symptoms in the following 5 years. The life expectancy of someone with schizophrenia is 10 years less than the general population. What’s more, of every ten people that develop the disease, 1 will successfully commit suicide.
It is diagnosed based on the patient’s history and ruling out other somethings that could be causing symptoms. Unfortunately, schizophrenia can’t be cured, and treatment is focused on stopping symptoms and helping patients manage the problem. Antipsychotic medications are hugely important in controlling symptoms but can cause other problems. CBT is also useful in helping people cope with their hallucinations and delusions. Finally, social support is vital in helping sufferers lead a normal life.
Substance misuse is the excessive use of a harmful drug. Usually, three terms are used to describe someone’s relationship with a substance:
- Tolerance – a person has a diminished intended response to the drug; they have to take more of the drug to feel that same desired effect.
- Dependence – a person has physically adapted to a continuous presence of the drug and taking away the drug results in withdrawal symptoms.
- Addiction – a persona has uncontrollable and overwhelming urges to seek and use the drug.
Bipolar Affective Disorder (BPAD) is characterised by abnormally ‘high’ and ‘low’ mood. These are called ‘manic’ and ‘depressive’ episodes, respectively. It has a lifetime risk of 0.5-1% chance of development, making it as common as other mental health problems like schizophrenia.
Here is more mental health information.
Module 4 – Surgery and Inpatient Medicine
introduction to surgery
30% of all admissions to hospital are surgery related. There are many different types of surgery speciality for different types of the body, but each surgeon needs to know 6 basic things:
- Surgical technique
- Surgical complications
- Limits of their competence
MDT (multidisciplinary team meetings discuss difficult cases with the whole surgical department so that the wealth of knowledge is brought together to find a solution. All aspects are discussed, including history and examination, results and investigations, scans, biopsies, and blood tests. Other HCPs are present too, including clinical specialist nurses, radiologists and histopathologists.
Skills of a surgeon:
- Ability to work under pressure
- Thirst for knowledge and lifelong learning
- Respect for patients and dignity
- Being hands-on and practical
- Willing to teach
- A member of a team
- Compassion and empathy
- A holistic approach to care.
Difficult scenarios in surgery
The SPIKES method is used to deliver bad news:
- Eye contact, sit down.
- Introduce yourself, establish a relationship.
- Manage time constraints.
- Gauge patients understanding of the situation
- Tailor the news to their situation and identify any misconceptions.
- Some want to hear everything, some don’t.
- Clarify what the patient wants to hear, and respect those wishes.
- Give patients facts, introduced with a warning such as: ‘I’m afraid I have some and news for you…’
- Use language that they understand.
- Speak slow and clear
- Give information in small segments
- Check patients understanding and be sensitive to their emotional state.
- Understand and empathise with their emotions
- Encourages them to share feelings and develop the doctor-patient relationship.
- Strategy and summary
- Make a plan of treatment
- Now that we know what is going on, we can start targeted treatment to the problem
- Answer any questions
- Make a plan of treatment
Module 5 – Emergency Medicine
An introduction to emergency medicine
When you get to A&E, you have to:
- Assessment (check severity)
The ABCDE approach is used in emergency medicine:
- An incision in the throat
- Head-tilt, chin-lift)
- Cardiopulmonary resuscitation)
- IV fluids
- Automated External Defibrillator
- Donated Blood
- Make sure emergency service is on their way
- Reassess patient.
They are ranked in order of which will kill a patient most quickly. In each assessment, you need to: look, listen, feel.
ABO group – based on antigen A and antigen B (type A, B, AB and O). Type A has anti-B in plasma, type B has anti-A in the plasma, type AB has no antibodies, and O has anti-A and anti-B in the plasma.
So, in blood transfusions, blood types must not come into contact with antibodies against his or her own antigens. Rh antigens are present in Rh+ persons and absent in Rh- persons. If an Rh- person receives an Rh+ transfusion, they will produce antibodies against the Rh antigen, and agglutination will occur. Rh- mothers with Rh+ babies may come into contact with their baby’s Rh+ blood and produce antibodies that would attack any subsequent foetus. RhoGAM is an anti-Rh can be given to the Rh- mother. It binds to the Rh+ foetal cells and shields them from the mother’s cells so that the mother doesn’t produce anti-Rh cells.
Palliative Medicine and Communication Skill
Palliative medicine is an approach that focuses on managing symptoms, progression, and complications of a terminal disease, rather than curing it.
introduction to palliative medicine & cancer
One of the biggest causes of death in the UK is cancer. Cancer is ‘uncontrolled cell growth’. Normally, there is a balance between our cells dividing to form new cells, and old cells dying. The rates of these are controlled by our genes. There are genes that increase or decrease the rate of cell division, and the rate of cell death.
Mutations are sudden unplanned changes to genes, which can change the function of a gene. Mutations that regulate cell division and death are dangerous, as they can cause overactive cell division and stop cells from dying. If enough mutations occur, cell division may become out of control, and cancer develops. Cancer occurs as mutations are accumulated over time, which is why older people are more likely to develop cancer than younger people. Uncontrolled cell division leads to the formation of tumours, which can press on other parts of the body and cause symptoms. If left untreated, the tumour can ‘break off’ and travel to other parts of the body in the blood, causing problems there. This is called ‘metastasis’. There are many reasons why it is so difficult to treat:
- Cancer cells look like normal cells, meaning our body often leaves them alone, rather than destroying them.
- Because they look normal, treatments often also attack our own cells.
- Sometimes it is hard to detect, as it presents with very vague or no symptoms. Often it is far too progressed to do anything about it.
- Cancer can spread easily to other parts of the body (metastasis). These can be further spread meaning new cancer forms faster than it can be destroyed.
- There’s no telling whether cancer will completely go away with treatment. Sometimes, a patient’s life may be extended by treatment, but the side effects of treatment can be difficult to cope with.
There are three treatments available:
- Removing localised cancers that haven’t spread.
- Targeting highly ionizing radiation at cancers to destroy them.
- Using medications injected into the blood to attack and destroy cancer.
Many cancers can be cured and most go into remission meaning that the cancer is undetectable in the body.
There are five priorities for a dying patient:
- Patient’s needs and wishes considered and regularly amended. Expectation and purpose of care switches from improving patient’s disease to ‘palliative’
- Sensitive communication allowing them to ask questions.
- Make sure the patient and everyone the patient wants to be included has their chance to express their wishes.
- Make them aware of support services that they can use, and listen to any cultural needs that they have.
- Plan and do
- A personalised plan is drawn up, which is regularly revisited, and make sure that all aspects of the plan are completely satisfactory.
Situations in which a doctor need to have good communication skills for include:
- Explaining problems that are normally very complex in a way that most patients could understand, such as the mechanism of a disease, or the way a new treatment will work.
- Educating patients effectively on good and bad health behaviours, such as the benefits of quitting smoking or having a well-balanced diet.
- Talking to different members of the healthcare team professionally and providing all the information they need for them to do their job in managing a patient.
- Adapting their communications style to match the needs of the patient, for example when talking to a child, or with a patient that does not speak the native language.
- Being able to optimise consultations where communication with the patient may be difficult, such as when patients are angry or withdrawn.
- Appropriately ‘breaking bad news’ to patients and being honest about treatments and expectations.
AI in Healthcare – University of Manchester
Here is a PowerPoint that I made summarising the points below.
This week we looked at:
- Introduction to the different terms and definitions under the topic of AI including sub-types such as Machine Learning
- Overview of current case studies on applications of AI
- An overview of the benefits and challenges of implementing it in healthcare
- The recommendations from the Topol Review on supporting digital transformation including changes in the relationship of the citizen to the NHS and development of digital skills for the workforce.
We began with a discussion about what AI is and concluded that it is a computer-based system carrying out a task that would usually require human intelligence. This can be applied to numerous aspects of medicine, including screening, diagnosis, population medicine, patient support and patient monitoring and tracking. In many circumstances, AI can, in fact, be much more successful than humans. In radiography, for example, a study compared pneumonia diagnosis of 112,000 X-ray images comparing to the success rate of an AI algorithm to that of 4 radiologists and found AI outperformed them.
We then looked at some examples of the transformative power of AI, not as a system of diagnosis, but rather as a system of treatment, and in this case for Parkinson’s disease. Firstly, we looked at wearable technology can be used to counteract the tremors of Parkinson’s. Haiyan Zhang designed a wearable watch which counteracted graphic designer, Emma Lowton’s tremors giving her the ability to write and draw. The second example was Rory Cellan-Jones’ app design which measured hand gestures of patients to test the severity of the disease meaning fewer hospital visits.
But questions were asked about the scalability of these developments, and how AI can be validated and trusted on a larger scale, after all, this is where it is needed.
We looked at how AI is needed to reduce the strain on healthcare professionals. In 2018, ‘Health Profile for England 2019’ outlined 9 key challenges showing the changing landscape of 21st-century healthcare. They show demographic changes, such as increased life expectancy and the associated conditions with an ageing population such as dementia – by 2021, dementia and Alzheimer’s are likely to be the most common cause of death for men and women. They also showed changing social, physical and behavioural factors and their impact including conditions such as diabetes and obesity. Furthermore ongoing political, social and economic changes which have an impact on population health now and in the foreseeable future. So AI can speed up automate tasks which are repetitive and mundane freeing up time for clinicians. It can also improve the accuracy of diagnosis. Furthermore, it can empower patients to monitor their own condition through AI and wearable technologies, and finally predict the outcomes and hence allocate resources more efficiently. So, AI could make healthcare more efficient, effective and sustainable, but how would it need to be implemented?
The Topol Review, a 102-page document from the NHS preparing the healthcare workforce to deliver a digital workforce which is 80% about managing change and 20% about the actual technology, outlined three guiding principles to embed technology throughout the NHS:
- Patients need to be central to the program and must be informed and involved with the transformation.
- Education and training are needed by healthcare workers to equip them with the necessary expertise.
- The program will free up time to enable staff to focus on patient care.
Some areas to address however are digitising and integrating patient records to support consistent data capture and bridge silos. Supporting the workforce to learn to critically appraise any data-driven technologies, and understanding ethics and governance controls. The area of ethics and regulation must be addressed to create a code of conduct to ensure ethical principles are central to any AI tools and technologies. Finally, there should be guidance on evidence-based effectiveness to ensure all stakeholders can evaluate and regulate AI-driven technology.
My view on all of this is that the human and ethical factors need to be prioritised and are currently undervalued. Many disjointed digital ventures have come about through the focus being on technology first rather than people first, and rather they should focus on what the end-user wants, rather than the users being made to fit the technology.
This graph shows the top ten technological advances impacting healthcare, found in the Topol review. Focusing on numbers 5, 6, 8, and 9 (relating to AI), we see that the impact of AI on the workforce will not start to take effect until 2025 – and for only 20% of staff. But, this doesn’t mean that it doesn’t need to be planned for now.
So, what should we be doing now? To name a couple of things we could be doing; we should get boards and primary care networks to understand what digital transformation means, and encourage boards to be willing to drive that change, and provide them with the ability to do so. We should also bring the people that are working in digital transformation together, along with clinicians to think through how they may work differently encouraging cross-team working. In addition, we should ensure that the 1.4 million people working in the NHS are up to speed with what is happening and understand how to manage data, have some basic knowledge of cybersecurity, and how to use and interpret digital information. We also need to prepare the population for this change; similar to what they have already had with shopping and banks. Next, far more partnerships are going to have to be made outside of the NHS, as industry, academia, and clinical service will all come together; this is a symbiotic relationship, as no one can do it alone. Overall I think that we need to de-myth digitisation as often people appear scared at the prospect of AI. We need to begin to talk about it with simple language and prepare people to embrace the change.
The Topol Review page 9 says that ‘within 20 years, 90% of all jobs in the NHS will require some element of digital skills.’ Obviously, not everyone will need to work closely with AI tools and techniques but it would be very handy to be conversant in how it works.
We then looked at a glossary of terms, and a timeline of AI, the former showing how many things there are to learn about the topic and the latter showing how much AI has developed since its creation in the 1950s after the Dartmouth College summer AI conference where the term was coined, and especially in the last 10-15 years.
Machine learning is a ‘bottom-up’ approach to AI (machines learning by stimulating brain cells or neural networks – multiple layers between the input and output layers allowing a machine to recognise and learn patterns, a form of deep learning), as opposed to a ‘top-down’ approach where machines are pre-programmed with rules to enable it to reason as effectively as humans. Machine Learning is an application of AI that uses algorithms to perform tasks without explicit instruction. In machine learning, computer systems learn the rules they need to operate from data that is presented to them. The applications of this is already being researched in multiple countries, mostly regarding cancer treatment but many other diseases as well. This is being done through neural networks (as above) and random forest – a classification algorithm that uses lots of decision trees to distil data. The decision trees are made up of nodes; the first being the ‘root’ node, decision points are ‘decision’ nodes and ‘terminal’ or ‘leaf’ nodes are at the end. The algorithm works through the decision nodes and partitions the data based on the features looking for the best split. The predictive power of the trees is improved by aggregating results from a number of trees to get the best accuracy. This only works on tabular data, however. For a nice comparison between neural networks and random forests, please visit the following website: https://www.kdnuggets.com/2019/06/random-forest-vs-neural-network.html.
At this stage, if you would like some further reading regarding AI in healthcare, please read the following: https://bmjleader.bmj.com/content/2/2/59.
We then looked at an introduction into the case studies that we would look more deeply into in the coming weeks. They were largely based around prediction such as the MOBILISE Project by David Jenkins aiming to recognise a person at risk from bipolar before they ever have a manic episode through the use of statistical algorithms on real-world health data. Another is the STOpFrac Project by Paul Bromiley, which focusses on osteoporosis, a disease that makes a patient’s bones weaker and causes them to suffer from fractures. In clinical practice, this disease is underdiagnosed. This project is using AI systems to help radiologists identify and diagnose these fractures on existing patient images to make the process of identifying and reporting osteoporosis more efficient so that patients with the disease can be diagnosed earlier to reduce the chances they will go on to suffer serious compromising fractures in the future. And finally studies into Paediatric Brain Tumours which have a relatively small clinical patient cohort and so lack large amounts of data. Yet, they are still working on improving diagnosis and care on an individual basis. We will look at these in more detail later on in the course.
We then had the opportunity to give our views on the pros and cons of AI in medicine. Mine are, pros:
- The ability to interpret large amounts of data quickly and easily and interpreting it into meaningful information.
- The ability to predict health patterns and where certain diseases may arise.
- The ability to offload mundane tasks and save time for healthcare professionals, and supporting the workload of healthcare professionals.
- It allows for a better allocation of resources by predicting the outcomes of patients.
- Storing large amounts of data can pose a large security risk.
- Expensive and issues surrounding how it can be scaled up to the population size.
- Ethical problems such as who to blame when AI makes a mistake.
This week we concentrated on the mechanics of AI – how it works, the workflows involved, how it could support healthcare professionals as well as identifying potential risks to implementation. This week especially focuses on machine learning, the machine learning workflow, and investigate the data in the machine learning workflow.
There are two main forms of AI; broad and narrow AI. At the moment, we only have access to narrow AI in the vast majority, if not all cases. Narrow AI is AI designed for a specific purpose, that is very good at doing that one thing. For example ‘Siri’ is able to answer questions we may have by retrieving answers from the web. Broad AI refers to a robot that would almost be indistinguishable from a human – a fully autonomous robot. We are at the point now where a lot of AI and machine learning is very good at specific tasks, but we have fewer examples of complete artificial intelligence, and strictly speaking, until the issues surrounding consciousness in AI are resolved, we won’t have examples of true broad AI. Within these categories, there are many more parts of AI including machine learning, robotics, haptics (technology that stimulates the senses of touch and motion), understanding text (what written text actually means), working with speech (both understanding humans, and speaking itself), vision (looking and classifying pictures).
As we looked briefly at last week, AI comes with many drawbacks. One which I saw, and hadn’t considered previously, in the House of Lords Artificial Intelligence Committee ‘AI in the UK: ready, willing and able?’ was that concerning prejudice, and how the data sets that we use are poorly representative of the wider population, and so the AI that learns from this may make unfair decisions. This shows AI’s dependency on the data given to it and the impact that this can have.
Another issue is more of a personal one, the extent to which we can trust AI to make decisions. We looked at a scenario where you had just had surgery and had to spend some time in the ICU. The hospital has a deep learning AI machine that predicts the risk of developing cardiac arrhythmias, which has been shown to have better performance at this task than the clinical team. A few weeks later you receive an automatically generated letter saying that the AI system classified you as having a life-threatening long QT syndrome. Personally, I feel comfortable about the way that AI is being used in patient care, as if it is more successful than the clinical team, I would rather know earlier rather than later about a life-threatening disease, to increase my chances of survival.
Another scenario is that you have been having many fainting fits and have visited your local doctor several times. At the next appointment, the doctor’s AI system has flagged an alert saying that given the number of fainting episodes on your record as well as your history of heart palpitations, there is a risk that you might have long QT syndrome. On looking through the records, the doctor feels that the risk is real. This is explained to you and an appointment is made to see a cardiology consultant. Again, if the doctor wouldn’t have noticed this alone, I feel comfortable about the way in which AI is being used to make decisions about patient care as it identifies something otherwise missed by medical staff. Many more people agreed with the way that AI is used in patient care in this scenario – perhaps because the local doctor and consultant are involved in the decision-making process.
I then saw an interesting view on the role of AI systems in informing decisions on patient diagnosis and treatment. Hugo Ford said that he saw three scenarios:
- AI is used to ‘safety net’ clinical decisions – here I see no issues.
- AI is used to identify a risk of a condition being present triggering further investigation – again I am happy with this.
- AI is used to automate an entire clinical pathway – here I would be concerned about how people are informed about the decisions the AI is making and how they can discuss the pros and cons for them of a recommended course of action.
Another issue was that surrounding the accessibility of information that people are given. Lots of people are given information which they simply don’t understand, and it subsequently impacts their quality of care. Some interesting examples and implications of this (what they call ‘health literacy’) can be found at the following website: https://www.england.nhs.uk/blog/jonathan-berry/.
A lovely summary of what machine learning is can be seen when compared to a human. As a human, we learn from experience, trying things based on the data we have, to see if they work and modifying them so that they work better next time. We assume that computers have a specific program which constrains its behaviour. Machine learning is a computer system that learns the rules needed to operate from data that is presented to them. So, the computer learns rules from the data rather than being given them explicitly. This approach is becoming more useful in a world increasingly full of data (the ‘big data’ revolution).
So how does machine learning work? There are essentially three steps:
- Create the dataset – used to train the machine (a vital step as any time bad data is used, you will end up with a bad model).
- Generate the rules – used for the machine learning to take place.
- Evaluate and refine the model – to see how well it is working.
Through three videos I saw how these three steps work on a basic level. Here is a summary: when creating the set, you create a training set to develop a classifier that the system can learn from, and then test this classifier in a test set to see whether it may be useful in a real-world setting. To generate the rule, you give the machine multiple examples from the training set, and with these examples, it comes out with a rule of some form, depending on the data that you put in (it will be a rule about one of the variables, for example, all circles have a star, all squares don’t have a star – here the variable is the star). You can then look at the rest of the set and see if this rule stands. If you don’t give it enough examples, then the rule may not work as it won’t be representative of the set, or perhaps there is just one outlier which contradicts the rule. To evaluate the data, you take the test data and make a prediction for each of the pieces of data using the rule from the training set. You then compare it to the actual and calculate a percentage of accuracy.
However, the example which was used to demonstrate these three stages was then applied to a real-world scenario, and we were shown how historical bias can be passed through a model. Imposing our own biases on machine learning models can pose potential risks when used for decision-making purposes. One example in a healthcare setting would be if a training set was not representative of a population, for example, it didn’t include members of a particular ethnicity who had a characteristic (e.g. a gene) which made them behave differently. An AI process would prescribe the wrong treatment for them. This is a serious issue that the AI4People initiative is worth looking at. https://www.eismd.eu/ai4people/. Furthermore, the correct diagnosis may not always be the best fitting model, and, where more rare diseases occur which fit other models, AI will find it difficult to spot this. There may also be an issue if you apply standard criteria to older people, such as gasometric parameters.
Having seen the importance of the data used in the machine learning process, the question was raised as to whether routinely collected health data sets are good enough to be safely used in machine learning. Data cleansing and data preparation processes are vital to ensure good quality data and to remove bias. Also, data quality checks should be implemented at the source along with staff training for inputting data into systems.
In summary, AI and machine learning will play an important role in the future of healthcare, but we need to be aware of their abilities and limitations. By their very nature, AI learns from examples we present them with. If there are errors and biases in the data we use to train and evaluate the systems, those biases and errors will be reflected back in the systems we develop. The wealth of data we have around health inequalities is clear evidence of how factors such as income can impact on health. Health data is generated, in most cases, by humans. As such it will reflect biases inherent in the people capturing the data. Understanding these biases and how they will impact on AI systems is an area of very active research. We have to be very careful that in the move to machine learning we don’t fundamentally fix existing biases and inequalities into these new systems.
So, healthcare professionals have a responsibility to understand the data on which the systems are built and that they are not predicting based on bias. People developing these systems have a responsibility to develop models which are ‘explainable’. A healthcare professional, using an AI system to assist in patient care, must be able to receive an explanation for why a decision has been made.
Data may be considered one of the most valuable resources of the modern-day. This week, we explored the data which healthcare is using, and the challenges of having untidy data that isn’t machine-readable, and the effects of this on the machine learning workflow. It then looks at how data is currently being used and how we might achieve a better quality of data in the future.
We began to learn about the complexity of the NHS – a complicated, intertwined ever-changing ecosystem. It includes multiple organisations including Public Health England, the government, health watch, NHS England, Clinical Commissioning Groups (CCGs) and many more. Each of these organisations has a different role in the health of our country, and each day they create enormous amounts of new data, which is diverse and distinct. This data comes from a variety of clinical activities; screening, diagnosis, treatment assignment and so on, or from services that support clinical activities like administration, financing, and clinical trials.
Clinical data often exists in the form of demographics, medical notes, electronic recordings from medical devices, physical examinations and clinical laboratory and images. The best-known form of clinical data is the electronic health record (EHR), which is the digital version of a patient’s medical information and history maintained overtime. EHR includes information such as:
- Demographic information like name, age, gender, etc.
- Vital signs like heart rate, respiratory rate, and temperature.
- Diagnostic-related information, like laboratory test results from blood tests, genetic tests, images like x-rays and magnetic resonance imaging (MRI).
- Treatment information like medication, doses etc.
Other forms of healthcare data include:
- Administrative data: non-clinical data captured for record-keeping of a service, such as a hospital discharge information, bed management, etc.
- Financial data: capturing budget information, insurance information, etc.
- Patient/disease registries: data that collects and tracks clinical information of specified patient populations.
- Health surveys: statistical data to help evaluate diseases, like common chronic illnesses.
- Clinical trial data: which is clinical information that has been gathered thanks to experiments associated with clinical research.
- Patient-generated lifestyle data: information captured by patient’s devices and social media.
There are many forms and formats of data; one way of describing data is whether it can be comprehended by a machine. Structured data is data that can be stored and viewed in a consistent, organised manner, allowing it to be easily read by a machine, analysed over time, and validated against expected values or biologically plausible thresholds. Examples of health data that would fall into this category include:
- numerical values, data which is measurable, like height, weight and blood pressure
- categorical values, data which are names, labels, types or categories, like blood type
- ordinal values, data which is placed into some kind of order or scale, like the stages of disease diagnosis.
Unstructured data lacks any form of organisation and precision, meaning that it can’t be easily analysed by a machine. This means that it usually has to be manually interpreted and analysed to transform it into a structured form. Examples include:
- clinician handwritten notes
- x-ray images (to avoid confusion – it can learn to read x-rays if given the data in a structured manner, which involves a manual process).
- scanned or faxed copies of structured data
Semi-structured data is, as you might expect, somewhere in the middle. It has some tags and labels that are searchable and retrievable but doesn’t quite have the same level of organisation and predictability as structured data. One notable example is the Picture Archiving and Communication System (PACS), a medical imaging technology system which provides storage, retrieval, management, and presentation of medical images. PACS contains information about images (structured), but the discrete files (images) are unstructured data.
Key factors of maximising AI’s potential are the following:
- Having high quality, accurate data.
- Having high quality data infrastructure.
- Having good labelling methods for both structured and semi-structured data.
- Clarity around the ownership of data.
- Data sharing for medical data and AI training.
- Education and engagement of healthcare professionals.
Consistent medical labelling methods is another issue. There are some existing medical coding classification systems, including the International Classification of Diseases (ICD – the standard system for classifying mortality and morbidity statistics, used to define diseases and allocate the right resources to the right providers) and the Current Procedural Terminology (CPT – published by the American Medical Association, with approximately 10,000 codes currently at use; provides a uniform data set that can be used to describe medical, surgical and diagnostic services rendered to patients).
The biggest current challenge with structured data is variety in the consistency of recording data; some data is rarely recorded and some is often recorded. Reasons for recording a particular sort of data can be affected by many things including:
- Time constraints and competing demands.
- Sometimes payment and billing depends on recording.
- Billing landscape (differential in cost of the same procedure for different types of patients).
Ways of inputting data by a clinician or coder include, and issues in brackets – in each of these stages, there is however the potential to make mistakes, some parts are less user friendly, and there may be elements of bias in the code.
- Typing it, having remembered it (misremembered or flexibility in recording data but also can result in duplicate entries through misspelling or alternate ways of recording ‘MI’ vs Myocardial Infarction. Can be time-consuming based on typing speed – some say scans of handwritten records may be more reliable as there is no chance of a simple typo generating a ‘syntax error’).
- Browsing a list or hierarchy (searching can be time-consuming and allows for misselection. Lack of flexibility, could constrain clinicians in their recordings).
- Searching with a keyword (Can be time-consuming trying to pin down the correct keywords, could depend on the literacy/time constraints of the clinician at that point in time).
- Use of templates (Lack of flexibility, could constrain clinicians in their recordings. Can be time-consuming in busy environments).
The LOAD (landscape, organisations, actors, data) model assists in identifying factors that can cause problems when introducing new functions (e.g. ML workflows) to a complex ecosystem. It describes the four dimensions that can affect the adoption of digital innovation, such as AI, and thus identifies not just the technical factors.
Here the data is the primary point of the investigation, with the peripheral dimensions also contributing.
The model was conceptualised as, after researching 18 case studies in the NHS, it was found that the issues in the development of new AI software, were largely due to human factors (with a few technical issues). This suggested that to adopt new technology, such as ML, investigating technical aspects are as important as the human and organisational factors. I will now look at each aspect in more detail:
The data dimension refers to the factors introduced by the data in order for the new system to run functionally. For example, a change of medium from written notes to an electronic form. This is usually done by clerical staff who may not fully understand the meaning of the data, and thus errors can easily be injected. In summary, in the type and form of the data and how it is captured.
The actors dimension refers to both the people and systems interacting together with the system to accomplish value. For example, data entered into a system by secretarial staff can contain errors if the information requires medical knowledge/vocabulary that the staff lack. We refer to this as the risk of a clash of grammars (the meaning of the data being altered by the change of context because of a cultural, experience or other types of reasons) which can result to lower data quality entered into the system. In summary: who collects and inputs the data, and which system stores the data.
The organisation dimension refers to the sharing data outside of the immediate organisational unit (which is paramount in health data given the issues surrounding confidentiality), which can result in a number of administrative costs, such as reaching and complying with data-sharing agreements, as well as complying with wider information governance requirements. Also, a risk of staff reluctance to share ownership of data may exist on both sides of the movement. In summary: who owns the data.
The landscape dimension refers to the environment that the other three dimensions are acting in, such as data captured in by a GP to use research project to find out why a patient has certain symptoms; or cancer data to be used for research purposes by another agency. In summary: the purpose of the data.
We then looked at another case study: the MOBILISE project, which aims to identify earlier signs of Bipolar Disorder (BD) – a disease characterised by high and low moods – for earlier detection. They use algorithms working on data to try to identify factors ( such as socioeconomic status, other conditions, age etc.) that may influence whether someone will contract bipolar, by looking back at the health records of people with BD and identifying frequently occurring health events prior to diagnosis. They have CPRD (Clinical Research Practise Data), electronic health records, including medication or diagnosis given or just what type of consultation it was, from approximately 700 practises across the UK. It is in the form of semi-structured data, so it needs elements of cleaning. They are attempting to use machine learning; they are using random forests, to begin with, and then perhaps recurrent neural networks, to predict whether a patient will get bipolar. They will then get a percentage chance and will make a cut-off, of perhaps 50%, above which a GP would intervene. They stress the importance of clean data for ease and speed, but have also seen many mis-codes or simply code missing. The best way to combat this is imputation where the machine inputs a value based on the most likely thing that would be there. But of course, psychiatric disorders are difficult to diagnose, AI or not, for many reasons, including the fact that there is a large overlap of symptoms between each of the disorders.
We then looked at another case study, the STOpFrac project which aims to help radiologists diagnose osteoporosis, a common skeletal disorder defined by a reduction in bone mineral density, increasing the risk of fractures in the hips, wrists and vertebrae. The vertebral fractures often occur earlier in the course of the disease. Identifying them allows preventative treatments that reduce the risk of a subsequent, much more serious, hip fracture. However, only about one-third of vertebral fractures present on clinical images come to clinical attention, often because the images were acquired for other purposes. The potential utility of a computer-aided system that can identify vertebral fractures is therefore considerable. They have been using clinical images, especially CT (computer tomography) images, as the NHS publishes 5 million CT images each year, and they are relatively easy to process compared to X-rays. It is therefore unstructured data. The AI techniques that they’re using are primarily statistical shape and appearance modelling techniques annotating the outlines of the vertebrae to identify any height reduction due to compression fractures. I would ask: could there be ethical issues surrounding the use of patients CT scans, originally taken for other purposes?
One interesting idea from the Christie NHS Foundation Trust that I hadn’t considered was about how, when converting from paper to digital format; if it is historical data, everyone would know what stage the patient was at in their cancer care, but in retrospect, it isn’t so clear. By using this data, however, and using already collected data for a different purpose, you don’t put any additional strain on healthcare workers.
To unlock the full potential of AI, mechanisms for data to flow safely and securely between organisations, ensuring consent, transparency and protection of the patients’ information are maintained. The Department of Health and Social Care in the UK have given five principles to combat the emerging ethical challenges with data. A summary of each is below:
- Any use of NHS data, including operational data, not available in the public domain must have an explicit aim to improve the health, welfare and/or care of patients in the NHS, or the operation of the NHS; e.g. scientific breakthroughs such as new treatments or diagnostics.
- NHS organisations entering agreements involving data must ensure fair terms for their organisation and the NHS.
- Any arrangements agreed by NHS organisations should not undermine, inhibit or impact the ability of the NHS, at a national level, to maximise the value or use of NHS data.
- Any arrangements agreed by NHS organisations should be transparent and clearly communicated in order to support public trust and confidence in the NHS and wider government data policies.
- Any arrangements agreed by NHS organisations should fully adhere to all applicable national-level legal, regulatory, privacy and security obligations, including:
- The National Data Guardian’s Data Security Standards
- The General Data Protection Regulation (GDPR)
- The Common Law Duty of Confidentiality.
On a global level, the World Health Organisation in partnership with the International Telecommunication Union (ITU) has established a Focus Group on Artificial Intelligence for Health (FG-AI4H) aiming to establish a standardized assessment framework for the evaluation of AI-based methods for health, diagnosis, triage or treatment decisions. The Focus Group on Artificial Intelligence for Health states:
‘For each domain it [the focus group] will work for the sourcing of test data, select current gold standard test success rates (e.g. how does a professional score on this test data), set benchmark rates for AI system (to be acceptable for decision support, to be acceptable for autonomous operation), and define acceptable fail modes (e.g. alert human operator if below a given confidence threshold).’
The course then stated that NHS Digital is providing datasets open to the public to gain new insight and help to improve patient care. The course gave us the following:
- Public Health England collects a range of data, made available in different formats, for example, their fingertips tool
- The Office for National Statistics collects a range of health-related microdata at their ONS virtual microdata lab
- UCI has built an open-source training dataset for machine learning (Health Data Research UK are in the process of building further training datasets)
- Health Data Finder
- NHS Digital Data Access Request Service
This week through looking at some more case studies, we looked at the ethical issues regarding AI, and at the principles and guidance being developed to support the implementation of AI, and some of the key organisations doing so. Finally, looked at the people that will be needed to champion and develop AI in healthcare and some of the innovative and creative ways of working that will be needed to take this forward into the next decade.
We first looked at the two areas of facial technology: facial recognition (who is the person in the image?), and facial verification (is this person in the image who they claim to be?). These are now widely used and one use is to identify medical disorders. The process of this has three stages: firstly it needs to find the face, so it scans over the whole image and finds which pixels have a face in them so that it can draw around the face. There may be many faces in the image, so it then works to find the main facial features (mouth, nose and eyes) to find the orientation of the face, the technology works best if you are looking frontal. Then it extracts information from the image by drawing boxes around the main facial features, to identify the useful features. It does this by running through a convolution neural network which comes up with a set of numbers at the end of it. If you show the same person, similar numbers will arise; but with a different person, different numbers will arise. To train this technology, they use millions of people with hundreds of millions of training images and give pairs of images; if they are the same people, the technology tries to learn features which are close together, and if they are different people, it tries to learn features that are far apart.
One of the most important aspects in facial recognition technology is the training set and ensuring that it is a representative set. This means that it can’t be entirely composed of white males, as it would work poorly with black women (e.g. Joy Buolamwini – BBC website). Previously, the training set had been created at universities, so there was a clear bias towards younger people. Now we have more representative sets, though they may still have ethnic bias, or gender bias (second article by the BBC). Furthermore, research has shown that the larger the set the better.
One challenge with facial verification is that of impostors. 1 in 10,000 impostor attempts are successful. However, the technology recognises the true person in good conditions (e.g. when a person is trying to get into their phone) 90% of the time. You could make it so that this is 99% but this would make it easier for the impostor to get in.
Another challenge is that there are many factors that make it harder to accurately recognise faces in a group, e.g. to look for terrorists:
- lighting – as you have little control over the lighting,
- viewpoint – people are usually further away, so the actual size of the face in the image is smaller
- people aren’t necessarily cooperative – they might not be looking directly at the camera
- They might be trying to disguise themselves
In these scenarios, the systems make many more mistakes and get many more false positives.
Another challenge is that there are many ways to avoid detection: obscuring your face, wearing makeup, wearing dark glasses, wearing a hoodie, or even just changing the expression of your face to a grin instead of a neutral expression.
The final challenge that we looked at was pretending to be someone else, the easiest way of doing so being holding a photo of them up to the camera. Although this might fool some of the technology, others have ‘liveness detectors’ to identify movement (breathing/small eye movement) and a 3D nature to the face. If these aren’t visible, the face isn’t recognised. Some people have made a full 3D head and reconstructed the head by creating a rubber mask – this may fool most technology.
One ethical concern with this technology is the lack of privacy in tracking people in public spaces, identifying people wherever they go. For this reason, San Francisco governers banned technology. An example of where the privacy issue has been taken up was Ed Bridges, photographed whilst Christmas shopping – Ed Bridges case. Another is that of false accusation if the technology makes a mistake. Having said this, they have been getting more accurate especially with deep learning – the human experts are still better than computers. One limitation is that some people look very similar, such as twins; so it is here where other forms of biometric may help such as gait recognition – this is identifying the way that people move and walk. So the technology still has room to grow, but whether it will is another matter.
The current evolutionary areas that AI is being used in medicine are:
- Supporting diagnostics in areas such as radiology and aiding clinical decision support.
- Non-clinical applications, such as increased automation of clinical and managerial tasks.
- Preventative public health based on modelling of population data.
Given the increasing use of AI, 10 ethical, social and political challenges have been identified (the bold is what that we focused on):
- What effect will AI have on human relationships in health and care?
- How is the use, storage and sharing of medical data impacted by AI?
- What are the implications of issues around algorithmic transparency/explainability on health?
- Will these technologies help eradicate or exacerbate existing health inequalities?
- What is the difference between an algorithmic decision and a human decision?
- What do patients and members of the public want from AI and related technologies?
- How should these technologies be regulated?
- Just because these technologies could enable access to new information, should we always use it?
- What makes algorithms, and the entities that create them trustworthy?
- What are the implications of collaborations between public and private sector organisations in the development of these tools?
Population data could be used to improve patient outcomes as it would feed into AI models. It may also help to develop precision medicine (tailoring specific treatments to patients to reduce unwanted reactions and make them more effective). The ethical issues that arise are how individuals feel about sharing personal health data, and the regulation that therefore needs to be in place for a form of consent, which organisations receive the data, and should individuals receive any payment for the use of data. Finally, what if you change your mind, how can the data be removed from the AI algorithm. I would be happy to share data if it was kept confidential.
There are 4 underlying ethical principles for data-driven healthcare projects, defined by the Nuffield Council on Bioethics:
- Respect for persons
- Taking into account private and public interest, involving patients in decision making in how their data is used.
- Respect for human rights
- Respecting basic rights including the limitations on the power of states and others to interfere with the privacy of the individual in the interests of the wider patient/public benefit.
- Engaging with patients to understand their expectations of they think they data should and is being used.
- Accounting for decisions
- Having clear lines of formal accountability, including reporting what has been done and explaining any breaches or deviations from policy.
This will of course be developed further by other governing bodies, as data becomes increasingly relevant. Below are the Data Protection Act 2018’s principles (other than those highlighted in bold which are additions):
- Understand users, their needs and the context of the technology
- Define the outcome and how the technology will contribute to it
- Use data that is line with guidelines for the purpose for which it is being used
- Be fair transparent and accountable about what data is being used and by whom
- Make use of current data and interoperability standards
- Be transparent about the limitations of the data used/algorithms deployed
- Demonstrate how the algorithm is working and an ethical focus on how its performance is being validated and how it will be deployed
- Generate evidence to demonstrate efficacy and economic impact
- Make security integral to the design
- Define the commercial strategy to reduce the chance of a release of personal identifiable data.
These stress the need for public involvement which I believe is very necessary, and of course important. Since it is our data, we should have a say about how it is used and should be kept informed in how everything works, if we so wish.
We then moved onto the importance of public trust in the use of our health data, first with the example of Project explAIn, which focuses on the aspects of explicability in enabling commissioners and members of the public to understand how their data is used to develop AI. Another group focusing on informing the public on how their data might be used is Understanding Patient Data. They published an article on trading patient data.
It is important to engage with the public about topics such as sharing health data for a number of reasons. Firstly, it’s data about people, so there is a moral argument that you should ask them what kind of uses are acceptable; you can then build trust in the people and embed their values and interests, and ultimately build a better system.
There are a number of stakeholders and regulators that a developer of AI must navigate in order to make the journey through development, implementation and adoption into healthcare. Working out the remit of each organisation, the order with which they should be consulted and exactly what each organisation requires is therefore pretty complex. Interestingly no one regulator has oversight of the quality of the data used to train the algorithms, meaning that bias can creep in.
New partnerships and innovative thinking will be needed to maximise the benefit from the data that healthcare holds, via techniques such as machine learning. This will involve multidisciplinary groups of professionals forming new partnerships between the NHS, academia and industry – Team Science.
Team science has gathered some momentum in the research community in the last few years but has gained less traction when applied to the healthcare setting. Within healthcare, interdisciplinary teams of individuals that might not usually work together have come together to work on a shared clinically important problem by coming to a common language and shared vision. The team may consist of:
- Clinical Scientists
- Data Scientists
- Software Engineers (AI or Other)
- Project Managers (Agile or Other)
- Industry Partners
These teams can drive forward rapid change and deliver solutions in which each team member can maximise the outcomes by contributing unique expertise. They are working synergistically towards a common goal, as they journey together with new language forms and conceptual boundaries are crossed.
Again we looked at the MOBILISE project (on bipolar disorder), and how they have adopted a Team Science model of working, developing statistical algorithms for use of real-world primary care data to assess important factors that will influence whether a patient goes on to be diagnosed as Bipolar. We watched an interview with David Jenkins who described his interdisciplinary team, and how they all bounce off each other, and fulfil different roles. They have:
- Epidemiologist (deals with the incidence, distribution, and possible control of diseases and other health factors – patterns of disease)
- Mental Health Clinician providing an insight into the results.
- Pharmacologist helping with the coding of recodes, for example: what does this drug do, and how does it work, is it meaningful in this setting.
- Expert in AI and Machine learning providing insights and code and how to interpret models
He described how crucial the synergy of the model is, but also the challenges of being spread out across the globe and struggling to organise full team meetings. He has lots of smaller meetings to discuss the stats with the stats team for example, but when the results come in, the full team convenes.
We then looked again at STOpFrac, and at an interview with Dr Paul Bromiley on the multidisciplinary team working together to develop AI systems to help radiologists identify and diagnose early fractures in patients with Osteoporosis. He spoke about how minor the research element, getting the right algorithm, is, and how commercialising is 90% of the challenge. So, he needed a team with and academic side, clinical side and commercial side, working together and learning about each other’s part so that they had a unified team. Everyone ended up under a lot of pressure, which was challenging so they had to come together to divide the tasks up, maintaining constant communication and constant team working, not just sticking to your individual role in the team.
This points towards something raised in the Topol Review; the boundary between technical and clinical blurring and making practitioners more flexible and agile. This is on an individual level but on an organisational level, visionary leaders that understand these technologies and are supportive of innovation and new models are required.
A few key points were raised on how to maximise Team Science:
- Improve communication (with a common language) bringing together interdisciplinary teams in the same working spaces.
- Improved infrastructure for data sharing.
- Improved mechanisms for recognition.
- Time to upskill staff through training programmes.
This week we experimented with a breast cancer dataset to see how machine learning can be used to diagnose cancer. We also pulled together everything that we had learned in the course to see how it would translate into practice. Finally, we looked at ways that we can prepare to be part of a digitally ready workforce.
Random forests can be used in the process of classification of data. This means that you could have, for example, multiple pictures of moles, and the machine would be able to tell you whether it is cancerous or benign. The way that this is done is the same as methods that we have already looked at: labelled data is fed into the machine in a training set, and it learns from this (supervised machine learning). It is then tested on a test set. On one hand, if the model fits the data too well, it might not generalise well to new data. This is called overfitting. On the other hand, if it doesn’t fit well enough, it might not be a good model for the data and is called underfitting. This balancing act is one of the challenges of designing a good model. When the model is run, one can see how well it performs. Then, when happy with its performance, it can be applied to new data.
We then looked at how random forests work; the training set results in essentially multiple questions being asked with two answers. Each question is a root node, where a decision is made, and it then splits into two more nodes until a terminal/leaf node, where the answer comes out, this results in a form of an upside-down tree. It works well on tabular data as it can run through each of the questions as a node.
We then looked through the random forest for the breast cancer set, where it looked at things including worst perimeter, worst concave points and many other measurements. We saw that it was quite precise, but made a few errors which could be seen in the confusion matrix:
^1 means malignant, 0 means benign.
We also looked at a graph of the most useful factors so that less useful features could be removed, this is called dimensionality reduction.
I then had the opportunity to change some of the parameters and afterwards got asked whether AI would be a replacement or a colleague. I answered colleague but then saw an interesting argument in the discussion space: in healthcare, due to a number of factors (e.g. workload, fatigue), some practitioners may be tempted to accept the “recommendation” of the software colleague without much critical thinking. On the other hand, when the AI, which has proved so efficient with the past cases, diagnoses a situation and the human colleague has another meaning, the latter may face an uphill struggle to explain, why his or her opinion differs from that of the software “expert”.
We then looked once more at the MOBILISE project (supporting earlier detection of Bipolar Disorder), and at the practicalities of embedding a new machine learning workflow within existing health care ecosystems. We found that they aim to develop tools that work alongside doctors that would alert them if the patient is at risk of developing bipolar. They also aim to develop AI to identify further risk factors. He said that AI and GPs will always be working in synergy. Firstly because the AI learns from GPs, in the training process, but also because you can never learn clinical judgement and the experience that a clinician has is needed. Furthermore, patients tend to prefer human interaction when being treated.
We then looked at STOpFrac and the key challenge of funding in order to move research innovations into clinical practice; moving from feasibility trials to a fully functioning commercial service to support the NHS, which requires funding and given that the NHS has severe funding pressure, that’s a big step to be able to overcome.
For me, I think that the biggest issue is around system integration. We are at a point where there are lots of interesting new ideas being tested in small subsets, but the challenge really comes when you want to spread the usage to the whole population level and fit into the systems that are already in place. This involves the buy-in of patients, of clinicians who haven’t worked in AI projects and might not understand the benefits, and in the initial phases requires new tech and old tech to fit together in a way that doesn’t allow anyone or anything to drop between the systems. The regulatory environment also needs to catch up. A study like MOBILISE or STOPfrac may show that an AI application works in isolation, but it’s a long road to spreading that out to everyone.
Other challenges come around:
- Communicating clearly and accurately especially to patients so that they can comprehend what is happening to their data.
- Educating and training people to use the systems.
- Governing the whole operation with clear rules and guidelines.
- Having the time and money to implement these technologies.
- NHS systems can be notoriously difficult to use and are not always user-friendly. Creating a system which is as user-friendly as possible may be a considerable challenge and may require usability-style studies.
Watching an interview on applying AI to general practice was very interesting. It was said that the online patient record system was much more efficient and easily managed. They also talked of a device that patients could take home and, when connected to their mobile phone, it could then take an ECG reading which could be sent back to the hospital. Docobo is another device which they talked of which is able to be taken home and record vital signs such as pulse, blood pressure, oxygen saturation and many more. Finally, they spoke about the importance of consistency in documentation, which especially important when patients move across departments.
Although digital technology may be integrated into our everyday lives, this isn’t the same for everyone, and many people lack access to IT equipment and lack the most basic digital skills. This is the same in healthcare, so Health Education England (HEE) is working to improve the digital capabilities of everyone working in health and social care. The best care of all individuals is only possible if these digital capabilities are fully developed and exploited, this includes an overall positive attitude towards digital technology and innovation, needed by everyone.
Digital Literacy is defined as
“Those capabilities that fit someone for living, learning, working, participating and thriving in a digital society”.
It includes many domains:
So, as you can see HEE’s focus is not simply on the technical skill but includes a range of dimensions, mapping to knowledge, skills, attitudes and behaviours across social, cultural and ethical dimensions. There are four levels for each domain from basic to expert. Overall, the framework set up by HEE promotes a positive attitude towards change. You can read the full framework here.
Another organisation with the same intention at HEE is Building a Digital Ready Workforce (BDRW) who aim to bring people together in a culture that recognises the need to innovate and the role of digital in that innovation. The programme’s mission is to help everyone in the health and care sector in England, become comfortable enough with digital tools that they can contribute to that transformation and deliver the outcomes of their role quicker, easier, safer and at a higher level of quality. By-and-large the skills, products and services that we need already exist. This means that the BDRW programme isn’t about building a new learning solution, it’s about helping to find where people have solved part of this problem and stitch them together into a cohesive offering for our whole industry. Every single organisation in health and social care has a duty for the learning and development of its own staff and we believe that digital skill and knowledge should be a core component of this. Their program is delivered through four workstreams:
1. Leadership and culture
3. Digital academy
4. Digital Literacy
Managing COVID-19 in General Practice – St George’s University of London
This course offered a practical, concise and (where possible) evidence based approach to the management of Covid-19 in primary care. This week covered
- Background of COVID-19
- Safe identification and assessment of possible cases in general practice
- Infection prevention and control measures in primary care
Covid-19 emerged in China in December 2019 and rapidly began to spread across the globe. It primarily causes respiratory problems, including a cough and shortness of breath, but in some people, it may cause gastrointestinal symptoms. The risk and severity of the condition increases with age and in those with underlying comorbidities (additional conditions). In the largest study of approximately 75,000 cases in China, 1% were asymptomatic, 81% had mild illness, 14% had severe illness and 5% were critical. But, case fatality increases exponentially with age, ranging from <1% in <50 year-olds to 15% in >80 year-olds.
Coronaviruses are a large family of diseases which cause mild to moderate upper respiratory tract illnesses in humans. Hundreds of different coronaviruses have been identified which many circulate in animals – pigs, camels, bats and cats. Only seven are known to cause human disease (4 are mild, and three can cause very severe disease). The two previous severe infections, the severe acute respiratory syndrome which emerged in 2002 and disappeared in 2004 and the Middle East respiratory syndrome (MERS) which emerged in 2012 and continues to circulate in camels, both caused outbreaks considered major global threats, as does SARS-CoV-2 which causes the disease COVID-19.
It began when a number of patients were hospitalised with pneumonia of an unknown cause in Wuhan in December 2019, by March COVID-19 cases were reported in more than 180 countries worldwide. It isn’t possible to quote up to date figures, however, the following website offers up-to-date figures: https://coronavirus.jhu.edu/map.html.
The virus itself, SARS-CoV-2 is considered to have originated in bats and pangolins and was passed to humans in this way. It binds to angiotensin-converting enzyme 2 (ACE2) on the cell surface of human respiratory epithelia. On average, someone infected transmits the infection to 2-3 other people – this is its basic reproduction number (R0). However, there is increasing evidence that a small proportion of infected individuals may be ‘super spreaders’ and cause a disproportionately large amount of secondary infections. You may also be infectious 1-2 days before symptom development. It is transmitted through respiratory droplets, which can occur in close contact (within 1 metre) with someone that has respiratory symptoms, or exposure to respiratory droplets to the oral or nasal mucosa or the conjunctiva (part of the eye). Droplet transmission may occur through fomites (objects or materials which are likely to carry infection). There were no cases of airborne transmission in 75, 465 cases in China (viable viruses in droplet nuclei resulting from the evaporation of larger droplets or existing within dust particles. These can remain in the air for long periods of time and be transmitted over distances greater than 1 metre). Airborne transmission may, however, occur during aerosol-generating procedures such as endotracheal intubation, bronchoscopy, open suctioning, administration of nebulised treatment, manual ventilation before intubation, turning patients to the prone position, disconnecting patients from the ventilator, non-invasive positive-pressure ventilation, tracheotomy, and cardiopulmonary resuscitation. Finally, COVID-19 has been confirmed in pregnant women, but the risk of transmission to the foetus seems low or none.
That gives an overview of the disease. We then moved onto an overview of the UK’s situation. In summary, the first cases appeared in late January, but these were contained with only 16 cases on 27th February 2020. In March, both cases and deaths began to increase rapidly. One day after WHO declared the outbreak as a pandemic, on March 12th, the UK risk level was raised from moderate to high. On 23rd March, the government announced tight measures, including wide-ranging restrictions made on freedom of movement, enforceable in law through the Coronavirus Act 2020, the Health Protection Regulations 2020 and other similar statutory instruments for the home nations.
Next, we looked at the clinical course of the disease; in summary:
- The median incubation period is 5 to 6 days (range 0 to 14 days)
- The median age of confirmed cases is 59 years
- Most common symptoms in adults:
- Cough 68%
- Fever 44%
- Fatigue 38%
- Myalgia/arthralgia 15%
- Headache 14%
- >80% of patients have asymptomatic to moderate disease and recover
- ~15% may get severe disease including pneumonia
- ~5% become critically unwell with septic shock and/or multi-organ and respiratory failure
- The case fatality rate is estimated at approximately 2% overall, but ranges from 0.2% in people under 50 to 14.8% in those over 80 years, and is higher among those with chronic co-morbid conditions
- Healthcare personnel appear to be disproportionally affected
In primary care, COVID-19 should be considered if the patient has one or more of a new continuous cough and a high temperature. Otherwise, COVID-19 should be considered if the patient has:
- Influenza-like illness: fever ≥37.8°C and at least one of the following respiratory symptoms of acute onset: –
- A persistent cough (with or without sputum)
- Nasal discharge or congestion
- Shortness of breath
- Sore throat
- Acute respiratory distress syndrome (a life-threatening condition where the lungs can’t provide the body with enough oxygen.
- Either clinical or radiological evidence of pneumonia
Clinicians should also always be alert to the possibility of atypical presentations in patients who are immunocompromised (have an impaired immune system).
The following infographic summarises the key symptoms and the patient flow pathway of COVID-19:
In order to prevent the spread of the virus (especially in healthcare personnel), it is crucial to prevent suspected COVID-19 cases presenting to surgeries. This can be challenging, particularly when patients are anxious or concerned, but GP practices should put measures in place to guide and support their staff in communicating with patients. These messages must be consistent and based on current governmental and scientific advice. The key messages are as follows:
- Avoid attending in person if they or someone in their household has a cough, temperature or flu-like illness.
- Patients can be directed to self-help sites such as 111, the NHS.co.uk site or the latest government advice.
- Practices should signpost self-isolation notices online and in the surgery.
- Patient advice has been provided in 16 languages to ensure effective communication.
- Practices should ensure reception staff are kept up to date and know the resources that they can signpost as well as changes in the practice policy on prescribing and dispensing. These will help to ensure that patients access the information they need efficiently and ensure clinicians’ time is spent appropriately. An example of the preventative measures taken at the reception of a GP can be found below:
Communication challenges facing GPs:
- Isolation for doctors and staff who continue to work remotely being separated from their teams yet still attempting to communicate.
- Struggles with non-face-to-face consultations either as they aren’t confident with the technology or lack the communication skills for a non-face-to-face consultation.
- Worries about GP trainees’ education as they can’t be supervised by their GPs.
- ‘Email tsunami’ – increase in amounts of emails that GPs are receiving from patients and other health professionals.
- Issues of confidentiality when dealing with non-face-to-face patients.
How GPs have and could adapt to this change:
- Telephone triage first for all patients.
- Use of mostly remote consultations.
- One raised issue is that of misinformation, so it is important that the information passed onto patients is that coming directly from Public Health England or the NHS website, both of which are being updated regularly.
How can GPs address patient and staff anxiety?
- Active listening with empathy and compassion, acknowledging that everyone at that moment is anxious
- Offer reassurance.
- Only state certain fact, and ensure that it is shared in an understandable, clear way.
How can empathy and acknowledgement be done by remote consultation?
- The tone of voice.
- Attitude towards the patient, being compassionate and calm.
- Remember that the patient will most likely be anxious, and when they get the doctor on the phone, that time is very important to them.
There are two main types of remote consultation:
- Telephone consultation used for both triage and assessment.
- Video consultation when visual cues and a therapeutic presence is required.
There are a number of steps that the GP should then take to ensure that the conversation runs smoothly when you first begin:
- Ensure that the patient can hear (and see) them.
- Ensure the consultation is clearly documented in the patient notes.
- Ensure that the patient’s identity is confirmed and documented.
- Ask the patient to ensure that they are in a quiet environment where they won’t be disturbed or have fear of being overheard.
- Start with open questions, such as ‘How can I help you today?’
The following infographic provides an overview of things to consider during telephone and video consultation for patients with suspected COVID-19 including top tips for examining remotely:
A summary of video consultation information can be found in Oxford University’s step-by-step overview.
After reading this and watching a short interview, I think that there will be a future of video consultations as most doctors and many nurses will have tried it and realised that it isn’t too bad. This could be especially helpful for a lot of chronic disease management, counselling, speech therapy and a lot of physiotherapy.
When ending the call, the following steps should be taken:
- Ensure all of the patient’s questions are answered.
- Signpost that it is the end of the conversation.
- Summarise what was discussed.
- Provide a ‘safety-net’ and ensure that the patient understands when to seek medical help if they deteriorate.
- Virtually ‘open the door’ and see the patient out.
We then continued to look at the guidance on patient management, and how to deal with patients in various scenarios. The general idea is to avoid any contact where possible with suspected COVID-19 cases and to tell them to self-isolate. If they are in critical condition, call an ambulance and inform them that it is a COVID-19 case which needs to be dealt with by the hospital. If contact is unavoidable, PPE is necessary.
We then moved onto the key steps required in infection prevention and control for frontline NHS staff in primary care involved in receiving, assessing, and where necessary, caring for patients with possible COVID-19. The guidance is based on previous pandemics and interpandemic periods. It is correct as of April 1st 2020, but may change over time; up to date advice is given by Public Health England. A variety of measures should be taken including hand hygiene, personal protective equipment, safe isolation procedures and environmental cleaning.
Firstly, hand washing. Given that the virus is spread by respiratory droplets, spread by coughing, sneezing, or touching infected surfaces, regular hand washing by both clinicians and patients is necessary, along with avoiding touching the face. Hand hygiene’s success in eliminating the virus can be seen on the Centre for Evidence-Based Medicine’s website. Patients should be encouraged to wash their hands regularly, and clinicians must wash their hands according to the guidelines for each consultation. These include before and after touching the patient, and after touching the patient’s surroundings or following exposure to body fluids. The healthcare worker should also wash hands before wearing and after removal of personal protective equipment. Below is the correct method for applying soap and water (right) and alcohol hand rub (left).
Secondly, personal protective equipment (PPE) for staff, which helps to protect staff from contracting the virus, and therefore is central to both the wellbeing of staff and limiting the onward transmission of the virus to both other patients and colleagues. The evidence for the success of PPE is briefly considered on the Centre for Evidence-Based Medicine’s website. The appropriate form of PPE depends on the healthcare setting, but it must always be put on correctly (donning) and taken off (doffing) to minimise the risk of infection. This should be carried out as follows:
Standard PPE should be worn for face-to-face contact (within 2 meters) with the patient. This looks like:
Note that aprons and gloves are for single use only. A fluid-resistant surgical mask (FRSM) and eye protection can be for single use or used for a full session. A session ends when the healthcare worker leaves the clinical care setting. A mask should be changed if it is moistened, damaged or soiled, as masks become less effective when moistened.
Thirdly, isolation for suspected cases of COVID-19 to avoid further transmission. All healthcare settings that encounter suspected cases must set up appropriate facilities to isolate these patients. If a suspected case presents to the surgery, isolation procedures should be taken. The infographic below shows the ideal isolation room:
Fourthly, environmental cleaning including medical equipment is required once the patient has left the isolation room. The infection prevention and control guidance should be followed, the key steps of which can be seen on the following infographic:
This week we looked at:
- Management of patients on home isolation including safety netting and red flags advice
- Social distancing and shielding of the vulnerable groups
- Ethical and medico-legal issues around COVID-19
- Managing patients without COVID-19 during the pandemic
- Staff wellbeing
Recently it has been interesting to see the varied response to COVID-19 internationally. Whilst some countries have taken very stringent measures, such as China, Italy and France, others, such as Sweden, have taken a more relaxed approach. This has been echoed in general practice: the use of remote triaging and consultation isn’t ubiquitous, and the management of patients and general advice has been varied. The general feeling now, however, is turning more towards lockdowns, and remote consultation. One place that I read about was using a ‘hot and cold clinic’ idea: people that need to be seen for routine things dressing, bloods go the cold clinic and of course, people with symptoms go to the hot clinic to try and take the overcrowding at hospital A & E.
Here is the advice given by the government on self-isolation:
- Anyone with a new continuous cough, fever or flu-like illness must stay indoors for seven days from the date of onset of symptoms.
- Asymptomatic household contacts should self-isolate for 14 days.
These people are advised to:
- Wash hands with soap and water for 20 seconds regularly or use a hand sanitiser.
- Stay at least two metres away from vulnerable individuals, including anyone over 70, pregnant women, and anyone under 70 with underlying medical conditions that would qualify them for annual flu vaccination on health grounds.
- Have food, medication, and supplies delivered to their home.
- Cover coughs and sneezes with tissues and put them in a bin.
- Avoid going out (except if advised to seek medical care) and not use public transport or taxis. Own vehicle may be used.
- Not have visitors at home.
- Double bag and seal all waste. Keep this aside for at least 72 hours before putting it into the usual external household waste bin.
- Choose a well-ventilated room, and keep away from other household members not displaying symptoms.
I think that those who are understandably worried about this, as it could have an impact on mental health, should be reassured that this is only temporary isolation – and ensures the safety of the population.
COVID-19 can cause a rapid deterioration in respiratory function and an escalation in symptoms, from quite mild, to severe. The red flag symptoms are below (these show that urgent assessment is necessary), as well as safety netting advice (for those living alone, the elderly and individuals with health conditions with the virus):
Social distancing means both reducing social interaction and keeping space (2m) between yourself and other people outside of your house to reduce the transmission of the virus. Public Health England has created a social distancing summary (correct as of 9th April 2020):
Shielding is a measure to protect people at very high risk of severe illness by minimising all interaction. It includes the following people:
- Solid organ transplant recipients
- Those having treatments for some cancers
- People with severe respiratory conditions such as cystic fibrosis and severe asthma/COPD
- Those with rare diseases that increase the risk of infections such as SCID and homozygous sickle cell
- People on immunosuppression therapies
- Pregnant women with significant heart disease
These groups received a formal letter from the government which advises them to do the following:
- Do not leave the house for 12 weeks from when you get an NHS letter.
- Groceries/medication should be dropped off on the doorstep by family, friends or delivery drivers.
- If your carers have symptoms of coronavirus, they must stay away.
- Use the phone or internet to keep in touch with other people including your doctor and carer.
- If you have symptoms of coronavirus (new continuous cough and/or high temperature/flu-like symptoms) contact NHS 111 online or by phone.
- Look after your wellbeing while you are at home for 12 weeks by doing some of the following as highlighted in the image below:
Regarding the medico-legal aspects of this pandemic, it is important to note that the law doesn’t change for the pandemic, but due to the pressure, many people are beginning to cut legal corners. For example, the importance of confidentiality, autonomy and consent remains true during the pandemic. But, there are also certain aspects of the law which are more pertinent during a pandemic or are very specific to a pandemic. This is a summary of the medico-legal status currently.
Some doctors are working outside of their usual field of practice. An issue which arises from a combination of this, and a large amount of pressure, means that when they are asked to do things beyond their competence, and with no option for supervision, they attempt to do things that they aren’t necessarily able to do. This is an ethical and legal issue.
Another legal issue is that arising form online consultations. Without clinical examinations, there is an increased risk of a missed diagnosis, and the doctor can then be held legally accountable. Therefore, they should err on the side of caution.
The key ethical issue in this pandemic is that around resource allocation, an issue for a variety of healthcare workers at the moment. The following principles should be taken into consideration:
Another ethical issue is the amount of risk that healthcare workers are required to take to fulfil their duties to patients. It has caused many healthcare workers to become unwell, and sadly, some have died.
To solve ethical issues – healthcare workers should seek guidance from organisations, such as the British Medical Association.
Next, we moved onto care for patients without COVID-19. Non-COVID-19 patients should still gain their essential routine care needs, but due to a diversion of resources, there may be deaths due to non-COVID causes. Some appointments will have to stop, however. The RCPG has made a traffic light approach, which gives three categories based on what treatment should continue. Green means it should continue no matter what the scale of the outbreak. Amber encourages workload to continue as capacity allows. All red tasks will be postponed. A summary is here:
Despite the ‘lockdown’ which we are currently in, physical activity is still extremely important for maintaining good wellbeing, thus we are advised to maintain the World Health Organization’s (WHO) recommendations of ‘150 minutes of moderate-intensity or 75 minutes of vigorous-intensity physical activity per week’. The UK Chief Medical Officer also says that we should do strengthening exercises at least 2 days per week, reduce our sedentary time, and do some form of balancing exercise – this is guidance for health adults (there are different guidelines for young people, disabled adults and pregnant women). The health benefits are below:
WHO recommends the following:
- Take short active breaks during the day – short exercises, domestic chores, or playing with children.
- Follow an online exercise class – can be free and readily available online.
- Walk – on the spot, in a small space, or outside maintaining at least a 2-metre distance.
- Stand up – especially if your work involved sitting for prolonged periods.
- Relax – mindfulness and relaxation techniques to maintain emotional health.
- Keep hydrated and have a balanced, healthy and nutritional diet.
It is unsure if asthma is a risk factor for COVID-19, but we know viruses are a common cause of asthma exacerbations. Therefore patients with severe asthma (asthma requiring high-dose inhaled corticosteroids and a second controller) fall into the category of extremely vulnerable and therefore should take measures to ensure they are shielded. The advice is similar to chronic obstructive pulmonary disease (COPD). For remote respiratory medication reviews, the following approach should be taken; the 4C-ABLE approach:
- Confirm diagnosis and stage disease
- Current treatment (pharmacological and non-pharmacological)
- Control – assess the level
- Compliance – assess the level
- Agree aims
- Barriers to success
- Learning and self-efficacy
- Emend and agree management
Keeping healthcare workers safe is vital. We need to ensure that those healthcare workers that are pregnant or immunosuppressed are moved to non-patient-facing roles, and seek advice from the occupational health service if available. Furthermore, staff working in reception and communal areas who are not involved in direct patient care should maintain social distancing of two meters. If this is not practical, use of a fluid-resistant surgical mask is recommended. If a member of staff develops a fever of >37.8 degrees Celsius or respiratory symptoms, or if they live in the same household as someone with symptoms, they should follow the local policy for self-isolation.
Furthermore, individual wellbeing of NHS staff is vital, so they should take measures including reducing caffeine, alcohol, screen time; staying hydrated, taking regular breaks; speaking to friends family and colleagues; avoiding the news; mindfulness; yoga; and trying a free wellbeing app for NHS staff among other things.
We then saw an interview with Dr Clare Gerada, former chair of the Royal College of
General Practitioners, where I found out about the Practitioner Health Programme – a confidential service for doctors and dentists with mental illness and addiction. It had been seeing around 11,000 patients, but, with COVID-19, this is rapidly increasing. They are seeing doctors with ‘moral injury’, presenting with fear, anxiety, sense of guilt and shame – it isn’t a mental illness, but emotion. These are normal in abnormal situations when, for example, GPs have to decide whether someone should go to hospital or stay at home. There are however doctors who have developed a mental illness due to the crisis, suffering from insomnia; tipping them into hypermania, an acute psychotic state. One issue is that doctors are, as they have been in the past, reluctant to seek help.