Bulletin January 2018 (Vol. 19 No. 2)

Big data has also enabled him to study antibiotic prescriptions in the hospital system, using a dataset of about 36 million prescriptions over 15 years. This showed that prescriptions were rising for the more potent, last-resort antibiotics. “With a dataset like this, while we can’t evaluate whether the antibiotic use was appropriate in individual cases, we are able to flag a concern,” he said. “More antibiotic use means that resistance has a higher chance to emerge, so is there a better way to manage the use of antibiotics in hospitals? Although already the hospitals are quite careful.” The limitations of big data is why Professor Cowling also works with smaller cohorts to get detailed answers from individuals. An example of this is his ongoing research into the effectiveness of flu vaccines. Caution needed Hospital records ostensibly show how many people are admitted with flu, who was vaccinated, and the laboratory results of swabs Professor Ben Cowling of the School of Public Health is an expert in the spread and prevention of infectious diseases. To understand the trends, he draws in part on electronic hospital records from the Hospital Authority which hold millions of pieces of data. But, as he has found, there are limitations to what even a rich trove of data can tell you. taken from each patient. However, this information may not always be accurate. In patients who arrive more than four or five days after the onset of symptoms – perhaps because their condition has worsened into pneumonia or a bacterial infection – the flu virus will likely not be present in their swab because the body will have fought it off. Moreover, the history of vaccinations may not be complete, particularly if the patient is admitted with a serious condition and the doctor is pre-occupied with more urgent and On the one hand, he has been able to reveal the impact of influenza on the hospital system. His studies of hospital records have shown that when flu activity increases, so do admissions of pneumonia, lower respiratory chronic diseases such as asthma, and heart attacks and heart disease. “A viral infection has a lot of knock-on effects on people’s health,” he said. complicated issues. In such circumstances, it will not be possible to accurately gauge the effectiveness of the flu vaccine. To overcome these problems, Professor Cowling and his team interview each person in their study cohorts, which can mean up to 1,000 people, to ensure there are no false negatives of flu infection or incorrect information. Typically, they find the vaccine to be about 50 per cent effective, although this changes each year with each new vaccine strain or outbreak and tends to be lower after repeated vaccinations. There is a misunderstanding about big data among some scientists who think it is this wonderful resource and we no longer have to collect our own information. That’s not true. It’s advantageous to have both perspectives. Professor Ben Cowling “There is a misunderstanding about big data among some scientists who think it is this wonderful resource and we no longer have to collect our own information. That’s not true. It’s advantageous to have both perspectives on things like influenza, so we can see what happens at the population and the individual level. But in general we need to be quite cautious about what we can and cannot learn from big data,” he said. █ In academia, as the generation and analysis of big data have become the norm, there has also been a demand to demonstrate inputs to research are acquired in ethical ways, and the results are reproducible. HKU has taken a number of steps to stay ahead of the curve and ensure the data produced by its scholars is properly acquired, stored and accessible. The key depository is the HKU Libraries, which has set up a website with information about data management, forms and informational videos. The University has also organised talks for different groups on campus and workshops for staff and students with overseas experts. As of September, 2017, all MPhil and PhD students who enrol at HKU must produce data management plans before their degrees will be confirmed. Recipients of General Research Fund grants also have to produce such plans before the University will release their funds. All data collected must include consent forms, which could make using certain kinds of big data a challenge for researchers. But there are still many options for using big data, as exemplified by the work of Professor Cowling and other examples described on these pages. Professor Danny Chan, S Y and H Y Cheng Professor in Stem Cell Biology and Regenerative Medicine, of the School of Biomedical Sciences, who is part of the HKU team promoting responsible conduct of research, says data management and research integrity are intertwined. “In many cases around the world, where there are allegations about responsible conduct and research, it often comes down to aspects of improper data management,” he said. Big data can suggest the bigger picture of public health, but it still cannot match the rich information that comes directly from each patient. A HEALTHY BALANCE Through the ‘Integrated Programme for Influenza Prevention among Community Elderly’, the team aims to share their experience and achievements with the elderly in Hong Kong, enhance their ability to prevent influenza and identify the most suitable vaccination strategy for them. Responsible Conduct of Research is promoted throughout HKU’s academic community. (Left) President Peter Mathieson speaking at a seminar on Responsible Conduct of Research. (Right) Various initiatives have been introduced to enhance awareness and enable the sharing of values and best practices, including a research ethics course for research students, and Responsible Conduct of Research seminars for staff. Integrity and data management go hand in hand at HKU 09 | 10 The University of Hong Kong Bulletin | January 2018 Cover Story