As healthcare continues to head towards evidence-based practice, physiotherapy remains the key sector where its application can be seen first hand. Considering this, it must be quite right to assume that physiotherapy is paving the way in evidence-based research. And yet, you can’t be further away from the truth. In reality, physiotherapy research is one of the most biased and inaccurate databases to draw results from – something around 72% of physios share concerns about.
But where did physiotherapy research lag?
There are several factors that dented its credibility.
It begins at the heart of where most physiotherapy research is conducted – clinics. Considered to be the best place to keep track of epidemiology, physios have underutilized the information they get from their patients while conducting research. The biggest flaw is the small sample size. A median sample size of 30 per study where more than 65% of studies don’t even have 50 participants means there is poor elimination of biases and overgeneralization of treatment. This reduces the accuracy of outcomes and studies often end up contradicting each other – something both dangerous and embarrassing.
If the problem of sample sizes isn’t big enough, there comes the nature of the research itself. Physiotherapy research is a time and labor-intensive process, with some cases having a monitoring period of as long as six months! From manually extracting information through video recordings and weekly/monthly reports to analyzing them, data processing and its translation into practice is frustratingly delayed. Mix this with poor documentation issues and low adherence rate, most studies either lose participants mid-way or generate unwanted biases.
While overgeneralization is a problem in itself, physiotherapy is still attempting to customize treatment, hoping they’d better cater to individual patients. But they still open the same can of worms – time-consuming, expensive and potentially even ineffective, especially for patients with unique needs. Research based on customization of treatments is lacking and mostly depends on the expertise of the physio.
But now that we’ve demoralized ourselves by discussing these problems, how do we solve them? The answer lies in a term we have come across far too often in recent times – Artificial Intelligence.
With the gigantic strides advanced Machine Learning algorithms have taken, it’s not surprising that ML models managed to find its application in physiotherapy too. Having the ability to deal with extremely large datasets, the seamless integration of AI-based algorithms into Electronic Health Records (EHR) can solve both the documentation and analysis issues. In fact, the accuracy of such models in seconds is stunning – with some models even outperforming traditional statistical ones!
Computer vision algorithms, natural-language processing and data clustering techniques can come clutch in aiding physios in their research. While its large-scale adoption will take considerable time, using existing equipment with tried-and-tested algorithms like Fitknees and Fitmust can massively add to the credibility of research! With predictions estimating AI to a $150B industry within healthcare by 2026, it’s only a matter of time before it becomes mainstream.
So if you’re a physio interested in research, you can take the first plunge and solve most conventional issues by investing in our equipment!
- O’Connor et al. (2015), Journal of Orthopaedic & Sports Physical Therapy
- Accenture (2017), “Artificial Intelligence: Healthcare’s New Nervous System”
- Mentiplay et al. (2018), Journal of Physiotherapy
- Buchbinder et al. (2002), “A randomized trial of vertebroplasty for painful osteoporotic vertebral fractures”
- Rajpurkar et al. (2018), “Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologists”
- Foster et al. (2017), “Machine learning regression analysis outperforms traditional clinical risk factors in the prediction of 5-year mortality in incident dialysis patients”