In order to statistically derive diagnosis-specific reference values for the QUALITOUCH Activity Index (AI), two large data samples were successfully merged. The sample related to the AI was significantly smaller than the clinical data set, indicating that the AI was not issued and/or completed by all patients. Since the clinical partner also treats patients who are not the target population of the AI, this seems reasonable. Merging the data occasioned the loss of only a few entries, and these were all cases where the patient could not be identified in the clinical data set. Thus, the basic data set as derived here is as complete as possible and, due to its size, is regarded as a sound basis for further analysis, with high external validity and a strong informative value with regard to quality of care.
The patients included in the basic data set showed an average age of 50.1 years, an average BMI of 24.6 and a sex ratio of approximately one male to three females. The high proportion of women is remarkable, but age and BMI are in line with published data of the general Swiss population . With respect to these overall descriptors, it can thus be assumed that the data reflect a representative population. The clearly increased age for the subsample of patients with knee osteoarthritis seems plausible given the degenerative nature of this pathology. However, additional characteristics that might have an influence on the therapeutic progress (e.g. education, occupation, comorbidities) could not be taken into account because those factors were not documented in either database.
For the statistical analysis, four different diagnoses were chosen. These cover different body regions, can be described as common fields of application in physiotherapy and require different therapeutic approaches. Therefore, it is believed that this choice can serve well as an example to demonstrate the impact of a generic PROM.
In contrast to other studies [18, 19], only one PROM was evaluated here, but it was evaluated with respect to several different diagnoses. The generic nature of the AI allowed this comparative analysis. Considering that physiotherapy is dealing with a variety of diagnoses in clinical practice, it seems reasonable and practical to use a single generic PROM instead of several diagnosis-specific ones. At the same time, this can be a limitation as analyses of diagnosis-specific aspects become more challenging, if not impossible. For this evaluation two dimensions of the AI were chosen, i.e. two questions. Both Q1 and Q4 are relevant for all patients. The maximum pain (Q1) was evaluated because it seems easier and more reliable to estimate than average pain (Q2). Discomfort during sleep (Q3) is known to be associated with pain and therefore was not used. Complaints during leisure time (Q5) and at work (Q6) were omitted in favour of focusing on household activities (Q4), which were deemed to represent some daily activity that is similarly relevant for patients of all age groups and socio-economic backgrounds.
As expected from other studies [9, 20], the AI did highlight a decline in pain and complaints after physiotherapy. For all diagnoses the AI documented a significant improvement between the first consultation and follow-up consultations. This confirms the assumption that the AI is a suitable instrument for recognising therapeutic progress and success.
The statistical procedure to derive reference corridors for expected AI progress over time was straightforward and rather simple, using a linear approximation. This reinforces the credibility and transparency of the results. The visualization as corridors allows for an easy comparison of the response of a specific patient with the statistical expectation. Hence, the corridors can be used as a monitoring tool to support both the therapist and the patient. In this way, it can be assessed whether the course of the therapy corresponds to the norm and whether it has an effect on the patient (per the different dimensions of the AI). This tool can thus quantify the effect of the therapy on the patient and complement the hands-on experience of the therapist.
Although the amount of data available was enormous, there were a few limitations in addition to those already mentioned above. In our data sample one patient can have multiple cases, and the AI questionnaire was issued for each different case, thus all were considered in the evaluation. This means that individual patients are represented several times, which could have an influence on the AI score (e.g. if chronically ill / with multiple diagnoses). Furthermore, many patients only filled in the AI at the beginning of their therapy, with a huge drop-off in the numbers of further follow-ups. To some extent this can be explained by the fact that some patients only needed a few sessions to complete their therapy and hence they stopped returning the AI at a follow-up. Others might have had poor compliance. From the data used here, it is unknown why any given patient stopped returning the AI. From a statistical point of view, while the declining number of responses from follow-up to follow-up can be explained, it does introduce uncertainties.
Besides these limitations, the established reference corridors offer a variety of opportunities related to quality of care. The use of PROMs involves the patient and contributes to considering the patient’s needs and identifying any unmet needs. Patients who are not responding well to therapy or whose success is stagnating can be identified early and options to adjust the therapy can be considered. This might also be helpful for decision-making, e.g. when weighing up conservative therapy versus surgical intervention. Using a reference corridor to compare the individual progress against a statistical expectation might help in this respect, and also in managing patient expectations. If, for example, a patient with knee joint arthrosis shows an AI score for Q1 of 70 points at the initial consultation, the reference corridor indicates that this patient is at the upper end of the statistical expectation. With this information the therapist can ensure the patient is closely monitored and if the score is reduced below 50 points at the third follow-up, the therapist is assured that such reduction represents the norm for this patient group indicating that the therapy seems to be successful whereas other patients only start a therapy with the same score. When using a PROM that covers different dimensions, as the generic AI does, a reference corridor can also be of help in prioritizing therapeutic aims and thus personalizing the intervention based on the expected outcome.
In addition to monitoring individual progress, quality control of an entire patient cohort from, for example, one physiotherapy practice is possible, and the therapeutic success of the practice can be documented and compared to the reference cohort. This enables practices to demonstrate the quality of their therapy, e.g. for health insurers , which is in line with current trends moving the health-care system towards a pay-for-performance system.
Future research should complement AI corridors for other diagnoses and provide corridors for further PROMs. Likewise, a predictive model in the form of a factor analysis could be a possibility to investigate in greater detail the predictive power of different influencing factors on such reference corridors. Table 2 already indicates several factors, such as age or BMI, that should be included in such a factor analysis. Further dimensions of the AI and medical aspects, such as comorbidity, should then also be included. Likewise, lifestyle related factors can be integrated to specify different peer groups to whom the reference corridors can be applied. The implementation of reference corridors in clinical practice and an evaluation of its impact can further contribute to discussion about evidence-based quality management in physiotherapy.