top of page

Ina Kostakis : Twitter Takeovers

Curated by Ina Kostakis Research Fellow in Healthcare Modelling and Informatics at the University of Portsmouth @ @ina_kostakis on the 23rd of Aug 2021

Every week the global OHT Twitter account is curated by a wonderful member of the OHT community. They share with us how they do what they do, what they're interested in, their top tips and general learnings. We like to turn these Tweets into blogs as there is so much goodness in them!

Go on... sign up to curate our account on a Monday know you want to!


Ina Kostakis is a Research Fellow in Healthcare Modelling and Informatics at the University of Portsmouth. Her research focusses on maximising the benefits of routinely collected clinical data for both patients and doctors through analysis, modelling and digital solutions. With a background and PhD in Marine Science, Ina now uses her analytical and data science skills to model clinical outcomes and predict risk for patient in hospital. Ina is a big supporter of early career researchers and an advocate for a healthy work/life balance (an issue in the academic world).

In her Twitter takeover, she discusses the value of routinely collected clinical data, ongoing and past research projects, free resources for using R for analysis of healthcare data, and practical advice on making research and presentations more accessible.


Good morning, everybody! I'm Ina @ina_kostakis and I'm a research fellow in the UK. I use routinely collected hospital data to understand and model patient risk. I'm also interested in new tech to improve care & patient experience #OHT21#TwitterTakeover

Today, I will share my journey of working in academic healthcare research - incl. how I got here, resources that helped me along the way and examples of how data analysis can help us understand patient risk and improve patient outcomes #OHT21

Let me know if you would like to learn anything in particular about working in academia or with clinical data. Thanks for having me!

I don't have a healthcare/tech background: for the first 6 yrs I worked in marine science. But when I moved from Glasgow to Portsmouth (more, less) 2 yrs ago, I switched to #healthcare research and now use my #DataScience & analytical skills to improve outcomes for patients

On how our analysis of clinical data can assist doctors and nurses to make better decisions on patient care. In summary, I look for patterns in data to identify relationships with outcomes, e.g. length of hospital stay, death, success of surgery, quality of life... #OHT21 1/n

Typically, patients are regularly observed by nursing staff to identify as early as possible if their condition is worsening. Monitoring of vital signs (temperature, blood pressure, pulse, oxygen saturation etc.) can help spot potential deterioration & aid timely intervention 2/n

Therefore, vital signs are checked multiple times a day. Some countries and hospitals use risk models or scores to turn individual vital sign measurements into an metric that reflects how serious a patients condition is and whether it is improving or declining. 3/n

Scores also often dictate how care should be escalated, e.g. whether monitoring frequency should be improved or whether patient need to be reviewed. In the UK, this score is called NEWS (the National Early Warning Score) which is now mandated across all trusts. 4/n

Last year, we showed that NEWS performs equally well in patients who tested positive for COVID-19 which is great because it means there was no need to develop and test a new system at a time when the NHS was already under a lot of pressure:… 5/n

Electronic systems simplify the collection & storage of vital sign data and automate the calculation of aggregate risk scores helping clinical staff to make decisions on care. It also means that we have very rich datasets of patient observations that we can use for research 6/n

Another example: ongoing project investigates whether observations prior to surgery can help surgeons decide on the most appropriate treatment/procedure - particularly important for elderly patients who might never fully recover from major surgery… 7/n

What are your experiences with developing digital or data-driven tools for patients or healthcare workers? How do you identify what works and doesn't work?

But is is also important to think about how our findings can be implemented into #practice. What tools are clinicians and patients looking for (apps, web services, ...)? What is practical? What's be best way to #communicate information? #OHT21 8/n

As already mentioned, identifying patterns in routinely collected data has multiple benefits: large datasets support robust analysis and generalisable results, new models are applicable to large number of patients without increasing demand on lab/staff/IT resources 9/nAs already mentioned, identifying patterns in routinely collected data has multiple benefits: large datasets support robust analysis and generalisable results, new models are applicable to large number of patients without increasing demand on lab/staff/IT resources 9/n

Routine blood tests are another useful data source. For example, we tried to see which commonly performed blood tests can help prioritise patients with a high risk of colorectal cancer to ensure they get access to necessary diagnostic tests as quickly as possible. 10/10

I want to highlight at this point that research is a #team sport and none of my work would have been possible without the incredible support from both academic and clinical colleagues and strong #collaborations between my #university and the local #hospital

Moving away from my research projects and onto another passion of mine: #coding I love to code because it challenges me to continuously learn new things, improve and develop my skill set.

I only started using R 2 years ago after using Matlab for a few years. It was a steep learning curve but I now use it for all my analysis and almost on a daily basis. I am super impressed by its power, versatility and its supportive user community.

It is important to recognise how much we can learn from each other by openly #sharing our code, successes and struggles. It does not only advances skill and knowledge of the healthcare tech & analytics sector but also increases #inclusivity.

Talking about #diversity and #inclusion, sometimes it might seem impossible to have any impact on DEI in your field/sector/institution. But we all can make a real difference if we question our own prejudices & assumptions and put ourselves into other people's shoes.

So in my last sub-threat for today, I want to share a few examples of how I, as an analyst and academic, can make my work, presentations and visualisations more accessible (1/n) #OHT21 Replying to @One_HealthTech

1. Openly share my code and analysis using #repositories like github or gitlab #SharingIsCaring (2/n)

2. Publish my work in open access journals #OpenScience

3. Make datasets available through data repositories (not applicable when working with sensitive patient data but true for other datasets) (4/n)

4. Use colour-blind-friendly colour schemes:… Most languages (python, R, Matlab) have libraries to implement ColorBrewer color scales and maps: (5/n)

5. Add a descriptions/alt-text to graphs and images I post on twitter. Here are a couple of great posts on the topic:…… (6/n)

7. Make visualisations more accessible:… (7/n)

That was it from me for today* I'd be delighted to connect with some of you directly and learn about your #HealthTech stories. Lastly, thanks to

@One_HealthTech for the opportunity to host this #MondayTakeover *I'll check in later to respond to any last questions or comments


bottom of page