Introduction
What is new?
- 1.
The developmental stage of most prediction models is preliminary.
- 2.
The study quality of the studies in which prediction models were derived showed substantial variation.
- 3.
This review provides a clear overview of the current position of prediction models and insight into the shortcomings of these models.
- 4.
The results of this study show that present prediction models are not ready to be applied in daily practice.
- 5.
Future research should focus on validating earlier derived prediction models, rather than on developing new ones.
Prognostic research provides insight into various patient characteristics and physical examination results that may predict patient outcome. Therefore, such research can help physical therapists improve their prognostic estimations, and make their clinical decisions more transparent and uniform [1]. However, because of the variability in patients and their health problems, it is difficult to adequately estimate prognosis based on studies evaluating a single predicting variable [2], [3]. In prognostic research, this raises the need to focus on multiple prognostic factors at the same time, which allows us to construct so-called prediction models.
Prediction models can be described as a combination of patient characteristics or test results that increase or decrease the probability of treatment success, or a certain prognosis, or diagnosis [4], [5], [6], [7]. The clustering or combining of predicting variables to guide clinical decisions explains the strength of prediction models because this approach better matches the actual context of health care. Because therapeutic actions often depend on prognostic expectations, reliable estimations on this matter are of importance.
In the past decades, many prediction models have been developed that are relevant for physical therapists. We found four reviews that provide an overview of prediction models for musculoskeletal complaints that are relevant for physical therapists in primary care [8], [9], [10], [11]. One review [10] focused on prediction models that are assistive during the physical therapy management of low back pain. The other three reviews [8], [9], [11] focused on prediction models that aim to assist treatment selection. In addition to treatment selection, estimation of the probability of a certain prognosis is important in the work of physical therapists. Physical therapists use this estimation to decide whether or not a patient should be treated. For example, in case of a favorable prognosis, a physical therapist may decide that a wait and see policy is indicated. However, a comprehensive summary of prediction models to guide physical therapists during their evaluation of patient prognosis for musculoskeletal complaints in general is missing. Therefore, the present review addresses this by 1) summarizing prediction models relevant for physical therapy in primary care, 2) evaluating their methodological quality, and 3) describing their stage of development.