TAN YING XUN, MEGAN

Background information of the organisation

Changi General Hospital is Singapore's first general hospital for the east and north-east regions. With its logo of a blue cross identical to a “+” symbol, it represents hope for people who visit the hospital, aligning with their mission: To Deliver the Best Patient Care with Passion and Empathy.

The department involved in this research study is health services research and the Integrated Sleep Service.

The health services research department researches on various aspects of healthcare services in CGH, which focuses on health economics to find cost-saving, effective and efficient alternatives that can better serve the patients’ needs. For example, our mentor working in this department has done research on various topics from the benefits of optical nasogastric tubes to the effects of using automatic gait analysis instead of manual analysis.

The Integrated Sleep Service in CGH is under the Changi Sleep & Assisted Ventilation Centre which offers comprehensive test, diagnosis and treatment services for patients with sleep disorders. This research study is focused on the sleep laboratory located at the Integrated Building, which also holds tests like the CPAP (Continuous Positive Airway Pressure) titration sleep study, MSLT (Multiple Sleep Latency Test) and MWT (Multiple Wakefulness Test), through updated sleep test devices.

Background of project

The focus of this research is on polysomnography (PSG), a clinical test used to diagnose sleep disorders. It does so by recording the patient’s brain waves, blood oxygen levels, heart rates and breathing rates, and eye and leg movements while they sleep.

In Changi General Hospital, it is conducted at the sleep laboratory where patients visit usually at night for the test to record their night time sleep patterns.

The cycle of falling asleep starts with non-rapid eye movement (NREM) sleep, At this stage, a person’s brain waves slow down. At this stage, the eyes do not move rapidly. At later stages of sleep, such as after 1-2 hours of NREM sleep, brain activity resumes and rapid eye movement (REM) sleep begins. A person typically experiences several sleep cycles at night, switching between NREM and REM sleep in around 90 minutes, but this can be disrupted by sleep disorders.

A PSG monitors a patient’s activities while asleep. The sleep technologist then interprets the activities such as by identifying the different sleep stages and cycles of the patient to make the appropriate diagnosis. However, this process is inefficient, as the sleep technologist has to manually score the data. Scoring the data of one patient for one PSG can take from 1 to 3 hours, depending on the complexity of data. Recently, an AI-based sleep stage classification system, Neurobit, has been created to speed up the process of interpreting PSG data by scoring it automatically. By helping to identify the different sleep stages of the patient’s sleep, it can greatly speed up the process of analysis. The software aims to speed up the analysis, but the sleep technologist is still required to check and properly diagnose the patient.

As such, Changi General Hospital aims to perform this research study to assess workflow improvements in terms of time and cost in diagnosing sleep disorders using the automated scoring software versus manual scoring by a sleep technologist. This is done by determining the effect of this system on various aspects contributing the workflow efficiency:

1. Cycle time to complete the manual scoring of a PSG by sleep technologist
2. Turnaround time between completion of a sleep test and corresponding PSG report being ready for review by clinician
3. Sleep technologist to patient ratio and overall workload capacity

Resources

The personnel involved in this research study include the sleep technologists in charge of scoring the PSGs, the respiratory sleep doctors, and health services researchers.

The budget for this study is approximately $10 000.

Record of Activities Done

Hypothesis

The automated scoring software is more efficient in interpreting the data than a sleep technologist

Methodology

To assess the workflow improvement, comparison of the efficiency of manual scoring and automated scoring has to be made. Therefore, we have to time the amount of time it takes for a sleep technologist to manually score a PSG (excluding disturbances such as attending to patients, picking up phone calls, etc.) as well as the time it takes to score a PSG with the aid of the automated software for the same PSG.

To maintain confidentiality of the patient as this research study does not include patient identifiers, only certain information of the patient is taken down. This includes the age, gender, race and sleep diagnosis (if applicable) of the patient. The sleep technologist’s age and number of years of experience is also recorded.

For auto-scoring by software, the following processes are timed:

1. Running-up laptop
2. Starting software
3. Transfer of data from hospital desktop to neurobit software
4. Neurobit data analysis
5. Transfer of data from neurobit software back to hospital desktop
6. Counter-check of neurobit data by sleep technologist

Assignment

This research study will last 6 months long. However, as this attachment only lasts for the first 3 weeks of the start of the study, we cannot complete the research study within this time period. Due to the time constraint, we can only complete the timing of manual scoring of PSG by sleep technologist.

Therefore, our assignment for the 3 weeks will focus on attaching to the sleep technologists and time how long they take to manually score a PSG.

We are seated near the sleep technologist as she scores the PSG, and we take note of the time where she starts, pauses, continues and stops scoring. Afterwards, the times are recorded in an excel sheet, and the total time taken for the sleep technologist to score that particular PSG is calculated.

Hence, during this project we were able to record the time taken by the sleep technologists but not the software. Hence, we were not able to compare the results and derive a conclusion. This step will be continued by the researchers at CGH.

Challenges

Some challenges we faced were that we could not precisely record the time where they pause or resume work, as we do not know the instant when they stop and start. We also had to assume that as long as they are looking at their computers, they were scoring the PSG as we had no better metric to determine whether they were working In addition, we could not record very short breaks that last for a few seconds accurately. Hence, we had to be on alert and be patient for the 1-3 hours of scoring.

3 content knowledge / skills learnt

Health economics

Health economics focuses on issues like efficiency, effectiveness, costs and behaviour in the demands of healthcare. People working in this field specialise in understanding the healthcare systems in various places to find the best methods to use in the healthcare sectors, from diagnosis to treatment. By researching on the efficiency, effectiveness and costs of different healthcare procedures, it can reduce unnecessary expenditure in hospitals or clinics, which is important in a world where resources are scarce. By finding the most cost-savings and still effective methods, more work can be done within less time and using less money, thus more patients can be cared for.

In health economics, economic evaluation is key in conveying statistics to compare the costs and other factors of different interventions. These statistics and evaluations can provide information for the healthcare providers to decide which is the best method to adopt for them, in their particular environment.

Statistical tests

There are many statistical tests used for different types of research studies to present different types of data. Some statistical tests introduce in this study include the Hypothetical Test, T-test and Chi-Square test.

• Hypothesis Test

Hypothesis testing is when an analyst tests an assumption regarding a population. The steps for this test is:

1. Determine the null hypothesis
2. Determine the alternative hypothesis
3. Choose a significance level (alpha)
4. Using the statistics, calculate the p-value
5. Draw a conclusion by rejecting or not rejecting the null hypothesis..

The p-value is the probability of finding the observed or more extreme results when the null hypothesis of the study is true. By convention, alpha (threshold p-value) is 0.05, though it can vary depending on the analyst. If p-value < 0.05, there is a significant difference and the null hypothesis is rejected. If p-value>0.05, there is no significant difference and the null hypothesis cannot be rejected.

• T-test

T-test determines if there is a significant difference between the means of two groups, which may be related in certain features. This is the main test used in this research study, which aims to find the mean of the first group-- time taken to score the PSGs manually by the sleep technologist, and the second group-- time taken to score the PSGs with the aid of Neurobit. When the means are found, the T-test will be conducted to find out if there is a significant difference in the time taken, and a conclusion on whether to adopt Neurobit in scoring PSGs in CGH can be drawn.

• Chi-Square test (χ² test)

Chi-square test measures how the expected data compares to actual observed ones, thus it is also known as a “goodness of fit” statistic. It tests how probable an observed data is due to chance.

Formula for chi-square statistic:

2 interesting aspects of your learning

Importance of technology

Through this programme, I further understood the importance of technology in healthcare, which was interesting as I saw the various ways technology weaved into different healthcare sectors.

For example, with the existence of telehealth, it allows clinicians to closely monitor patients from the convenience of their own homes. It can even educate, care for and remind patients on what to do to properly control their diseases, at the same time training individual responsibility as they learn how to take ownership of their own health. This is especially important and can greatly increase efficiency in treating diseases. Removing the need for patients to go down to hospitals/clinics to receive their respective treatments, doctors/healthcare worker can also focus on more patients, improving workflow. In the context of Singapore where more and more people are diagnosed with diabetes, it is interesting to see the different ways technologies are used to aid us in combating these chronic illnesses.

System of the hospital

It is known to all that Singapore is tackling the problem of an ageing population, which will only be increasingly apparent in our society. With this knowledge, it was interesting to find out how the hospital improves and upgrade their facilities and different areas to cater to the ageing population, as the elderly are more prone to illnesses especially related to old age.

It was even more interesting to observe how much the hospital does to help patients get back on their feet and regain confidence in themselves. For example, the hospital even had wards that were designed to mimic that of a HDB flat, so that patients who can be discharged could get used to the environment and believe that they are able to go back to their own flats and live independently.

Although this is not as related to my research, it was enriching to gain a deeper insight into how a hospital functions and how Singapore is better catering to our ageing population by investing more time and effort into bettering our healthcare system.

Takeaway

AI is playing an increasingly important role in healthcare, making healthcare procedures more efficient and convenient

e.g. AI helps to analyse diabetic patients’ wounds, speeding up the process for nurses who do it manually

Due to us living in an era of technological transition, many doctors and nurses are not skilled in using the available technology to assist their work. Hence, new doctors and nurses should also require AI training in future.

Although there is a common fear of AI and technology, especially it causing displacements of jobs, new technology is still constantly being invented to help us with our everyday work, making things much more efficient and generally accurate. In this world where technology will only keep evolving in various fields of healthcare, a balance has to be found and we have to find ways to adapt. Furthermore, as novel methods of diagnosing and treating patients are created, health economics will play a vital role in researching and determining whether these new methods are preferred to traditional, manual ways of working. By comparing using criteria like cost, efficiency, effectiveness, accuracy and more, a choice can be made by different healthcare providers under their different environment to find the optimal balance.