There’s no point buying something if you’re not going to use it; clothes, kitchen appliances and food are purchases that we often don’t make good use of
Similarly, the healthcare industry has a plethora of data at its fingertips, but what is the point in collecting this data if it is not put to good use? Here, Paul Ricci, one of Kolabtree’s registered freelance scientists, explains how the healthcare industry can be improved using data analytics.
Healthcare services around the world are facing increasing pressures to be more efficient and improve clinical outcomes. Data analytics can be used to inform better decision making on a clinical and operational level and help the industry to meet these demands.
A recent study found that the median cost of pivotal clinical trials that lead to drug approval is $19 million. The industry must find ways to increase the efficiency of clinical trials to reduce this cost. There are several ways that data analytics can be used to increase clinical trial efficiency.
Thanks to recent advancements in data analytics capabilities, clinical trials can now have much larger sample sizes. It is also easier to identify meaningful patterns in data that may otherwise be missed. As a result of these developments, clinical trial data can be more thorough, accurate and reliable, which is important when applying for MHRA or FDA approval.
Data analytics can also support better decision making in clinical trials. We can look at recent trends and predicted outcomes to make better decisions that increase trial efficiency, reduce costs and ensure greater patient safety.
In addition, we can use data analytics to make the most of every data set. During old clinical trials, data was not analysed as thoroughly as it would be now. Retrospective studies are commonly conducted to reanalyse this data using advanced data analytics techniques, which can uncover patterns that were not originally identified. Retrospective studies may also be conducted to test a secondary hypothesis — an affordable way to obtain more information about a drug without collecting more data.
As the scope of what we can do with data increases, real-time patient monitoring becomes more feasible.
Wearable devices could monitor patient parameters, such as blood pressure and heart rate, and transmit information to healthcare professionals across the cloud.
This could reduce, or even eliminate, the need for regular patient visits and tests, resulting in considerable cost savings and increased clinical trial efficiency.
Selection bias in a clinical trial can invalidate the results, so make sure your sample of patients fairly represents the population you are interested in. You can avoid selection bias by comparing the demographics of your sample with census data for the population of interest and ensuring there are no discrepancies. If your sample is biased, it may be possible to correct it by giving under-represented samples more weight than the over-represented samples.
Natural language processing technology automates the analysis of millions of medical data sets, which makes it easier to predict and prevent disease. For example, information from pharmacies and general practitioners about prescriptions sold and diagnoses made can be used to detect a disease outbreak and act quickly to prevent it spreading further.
In the future, electronic health records (EHRs) may be fully digital and connected across the cloud, so that anyone with authorisation can access them. Patients could receive alerts when an appointment is due, or test results are available, and healthcare professionals could monitor the health of their patients remotely. However, for this vision to become a reality, there are data security and confidentiality issues to address.
After your data has been analysed, you must look for the statistically significant result. At this stage, there are two common errors people make. A type one error is when you find a pattern in the data that does not exist. A type two error is when you don’t find a pattern that does exist.
Imagine you are in a court of law, where your null hypothesis is that the accused is innocent until proven guilty. A type one error would be if you find the person guilty when they are innocent. A type two error would be finding them innocent when they are guilty.
To ensure you don’t make any errors, don’t jump to conclusions — be thorough in your analysis to make sure your conclusion is accurate.