This study extends several important aspects of previous knowledge about exercise adherence to HEPs and its predictive factors. Our study did not demonstrate different rates of adherence to both HEP components, frequency per week and duration per session. However, other authors have also found variability among exercise components .
According to our expectation, this study revealed that the predictive factors of frequency adherence were different from those associated with duration adherence. The only shared predictors were self-efficacy and participation in home exercises during previous episodes of pain. The association between self-efficacy and adherence is consistent with previous research with aerobic exercise in general population  and older adults .
The importance of barriers is consistent with the results of previous research [5, 8, 12, 22]. Barriers related to fitting the HEP into a daily routine was included in multivariate model to explain frequency adherence, but was not significant for duration. If this finding is true, it suggests that barriers influence frequency adherence but not duration. It is reasonable to expect that barriers influence initiation of a HEP session (and therefore frequency adherence), but once a HEP session is initiated barriers are no longer important and duration adherence can be fulfilled. Just as barriers differ for different populations , barriers may differ for different components of HEP. For example, it is possible that physical symptoms interfere more with duration adherence that with other components of the HEP. Further research is needed to confirm these findings more fully examine differences in specific barriers to HEP components.
It is important to note that none of the patient characteristics are significant determinants of adherence in this sample. These factors have been the focus of numerous investigations of adherence in several chronic diseases. However, age, sex and education have not been definitely associated with adherence . While it is possible that this may be influenced by response bias, the similarities in these factors between respondents and non-respondents make this possibility nominal for most factors. The only factor that may be impacted by response bias is that of age. Since there was greater response from the >59 years age group, we cannot rule out age >59 years as a predictor of either component of HEP adherence.
The influence of adherence to HEP on long-term adherence has been demonstrated . However, our findings add that previous adherence differentially influences current frequency and duration adherence. Previous adherence predicted duration adherence, but did not predict frequency adherence.
Variables related to how physiotherapists interact and communicate with their patients are key determinants of home exercises adherence. In fact, one of the major barriers to adherence described in the literature is lack of information to the patient. However, lack of information alone is not enough for creating or maintaining good adherence habits. For example, the "Information, Motivation and Behavior (IMB) model" asserts that information is a prerequisite for changing behavior, but in itself is usually insufficient to achieve this change if the patient is not motivated to perform the behavior [26, 27]. In this study, it is logical to surmise that the most motivated patients are those that ask questions during clinical encounters. Our results supports the IMB model: providing patients' required information has a decisive influence on performing home exercises to recommended frequency per week (Table 3, item "Clarifying doubts and questions from patient" increases odds of adherence to frequency multivariate OR (95% CI) = 4.1 (1.4-12)).
Intervention characteristics also have an important influence on adherence. In this sense it could be hypothesised that the greater the number of exercises prescribed for the HEP, the greater the probability that subjects do not complete them. This phenomenon is similar to those reported in other recent works about adherence in HEP and in medications for chronic conditions [28, 29].
As might have been expected, there was a significant relationship between frequency and duration adherence (Table 3, item "Good adherence to frequency per week" increases odds of adherence to duration univariate OR (95% CI) = 8.0 (3.7-17) and the same for item "Good adherence to duration per session"), suggesting that when patients meet frequency recommendations it is more probable that they also regularly meet the duration recommendations.
This study has limitations. First, because variables related with physiotherapists' behaviors were measured together with adherence measures, the direction of causality cannot be determined. Thus, this is a study of association and is not a study of cause and effect. However, it should be noted that the physiotherapy adherence literature does show a causal connection of interactions between physiotherapists and patients causing an increase in adherence in home exercise programs .
Our analysis used the completed data in the respondent sample, which may bias the findings toward increased adherence. The more conservative intent to treat analysis (include non respondents as non adherent) was not chosen as too many assumptions would influence the analysis with imputed "non-adherence" for the non-respondents.
Nevertheless, the reader should realize that we cannot rule out the possibility that there was a predictive influence of age - which was different between the respondents and non respondents.
Finally, the statistical models presented herein have been utilized to test hypotheses regarding relationships between a variety of factors with either frequency or duration adherence to HEP. These models should not, at this time, be utilized in an attempt to predict odds of adherence in clinical populations based on these factors. Such predictions will first require additional data to confirm relationships and propose predication equations; and then would require verification in an independent sample.