These results show that as much as 60% of the resources spent by the SNHSCI on physical treatments for non-specific chronic NP, BP and SP between 2004 and 2007, were allocated to treatments that had been found to be either ineffective or for which there was no evidence of effectiveness. This means that, in that period, over 3 million Euros from taxpayers' contributions was handed over to private centres which had been awarded contracts by the government to apply treatments which lacked any evidence of effectiveness, or which had shown to be ineffective.
This is a very conservative estimation of the total cost, as it does not factor in data on the costs deriving from physical treatments administered in public hospitals and primary care practices run by the SNHSCI; only data from patients who were referred to private practices were included, representing 70% of patients receiving physical treatment for these conditions. In addition, it was sufficient for a therapy to be considered effective in a single CPG or SR, to be categorised as such in this study. Any form of treatment labelled as "exercise" was considered to be effective, irrespective of the type of exercise, number of sessions, patients' compliance or whether it was administered appropriately or inappropriately. Exercise was considered to be effective irrespectively of its effect size, whereas previous studies suggest that improvements in pain or disability of less than 30% of baseline value are clinically irrelevant for patients with NP and BP [50–52]. Moreover, only the provider's personnel and equipment costs directly related to physical therapies used for SP, NP and BP were taken into account, disregarding other costs (e.g., other provider's costs-facilities, financial costs, etc.- and profit), societal costs (work absenteeism associated with receiving physical therapy sessions, etc.), costs incurred by patients (transport, loss of earnings, etc.) and other costs for other institutions (insurance companies dealing with work-related accidents, private healthcare, etc.).
All the SRs and CPGs reviewed conclude that the quality of most clinical trials on physical treatments was low. The main shortcomings were related to randomization process, comparison treatments, masking procedures, losses to follow-up and missing data. This is likely to have led to an overestimation of the effect of physical treatments considered to be effective .
Inconsistencies in the prescription of different forms of treatments for patients with the same condition, may be due to the preferences of the different types of clinicians involved (primary care, rheumatology, rehabilitation, etc.), and the ready availability of these techniques. However, there are no validated criteria for selecting patients for whom a particular physical treatment would be indicated, or to decide the number of sessions they would need. In fact, there is even no evidence supporting the effectiveness of most of these forms of treatment or the cost/effectiveness of any of them, alone or in combination. These facts suggest that there may be ample room for improving the efficiency of the management of these syndromes in the SNHS. Ineffective technologies expose patients to risks, delays, expectations and costs which are unjustified . Spending public funds on such procedures is inappropriate, and even more so in a context in which health resources are limited.
A number of recommendations have been issued regarding the procedures that health authorities should follow in order to ensure that the health technologies they finance are safe, effective and efficient . Implementing these recommendations is likely to improve the efficacy of treatments and reduce wastage of healthcare resources, hence increasing the efficiency of available resources. In the context of an economic crisis, it is necessary to apply these measures before considering whether healthcare spending needs to be increased. This may require communication skills to explain it to clinicians, patients and public at large .
The search for SRs and CPGs was restricted to the 2003-2007 periods. It is possible that studies published afterwards may have changed, or will change, the evidence supporting the use of procedures included in this study. However, this study focused on quantifying the costs of procedures used in routine practice, despite the lack of supporting evidence at the time of use. In this regard, even if the evidence were to change based on future research, the use of these technologies in the 2004-2007 periods would continue to be inappropriate.
The results of this study illustrate the rationale for recommending disinvestment in specific health technologies. Disinvestment refers to the processes of reducing or discontinuing utilization of selected procedures and treatments . Proactive disinvestment requires prior analyses of inappropriate variation in clinical practice, development of valid CPGs designed to update clinicians' education and practice, adequate information for patients, in order to improve consumer behaviour, organisational support within the health services for implementing these changes, economic incentives for providers, and incentives for improving the efficiency of purchasing decisions, including the elimination of funding for technologies which are ineffective and/or unsafe. Clinicians, consumers and providers of potentially discarded technologies may feel threatened by disinvestment, while beneficiaries, such as taxpayers and patients themselves, may be less aware of the advantages of such decisions. As a result, incentives to perpetuate the status quo may override the call for change .
However, disinvestment in technologies which are not evidence-based, would make funds available for implementing technologies which are not currently available in the SNHSCI despite having proven effective and cost/effective for NP and BP within the SNHS, such as neuro-reflexotherapy [58–60], and for further expanding the use of technologies proven effective for the conditions reviewed in this study, such as exercise (Table 1, Figure 1). Funds made available through disinvestment could also be used for supporting further research on the effectiveness and cost/effectiveness of physical procedures, which is sparse and of low scientific validity (Table 2).
In fact, results from this study suggest that there is a need for high-quality studies assessing the effectiveness and cost/effectiveness of different forms of physical therapy, individually and in combination. These should be randomized controlled clinical trials, with homogenous and sufficiently large samples, in which validated instruments should be used to assess clinically meaningful variables, and in which randomization, patients' assessment and data analysis should be masked. These trials should analyze both the statistical significance and clinical relevance of results. Since data on comparative effectiveness of different procedures are difficult to interpret when neither has previously shown superiority versus "sham" or placebo, these studies should compare different forms of physical treatment to the appropriate "sham" procedures as well as to other interventions. Taking into account that some treatments formerly believed to be useful for patients with acute BP proved to actually be harmful, when appropriate, these studies should consider including groups without any intervention . The assessment of the results should also be made from the patient's point of view [54, 62, 63].
This study has some strengths. It contributes to the existing literature on non-adherence to CPGs by describing its economic consequences. Analyzed data were obtained through an information system which gathers information directly from routine clinical practice, leading to results which are directly applicable. Moreover, it analyzes data retrospectively, avoiding potential changes in referral patterns and selection of treatments which may occur when clinicians feel observed.
This study also has certain limitations. The Canary Islands' ISHC was one of the first health information systems to be implemented within the Spanish National Health Service, and it does not gather data on the criteria behind the prescription of any particular form of treatment. Moreover, it does not record additional treatments which are applied once the prescribed ones have started to be administered; therefore, patients may have actually undergone more forms of treatment than those which were identified in this study. The coding procedure is likely to be accurate, since it determines payments, but it may be influenced by the experience of the coders and their interest in administrative and/or clinical details. However, this is the only systematic and validated mechanism for post-implementation surveillance of physical therapies within the Spanish National Health Service.
This study included patients who, in addition to suffering from NP, BP or SP, also presented other concomitant musculoskeletal disorders which might have required physical treatment. In order to reduce contamination, the forms of physical treatment which were not used for NP, BP or SP were excluded. Nevertheless, it could be argued that some of the physical therapies attributed to the treatment of NP, BP or SP might have actually been used to treat concomitant conditions, leading to inappropriate assumptions. However, such co-morbidity was only present in 6% of cases. Therefore, we believe that this potential inaccuracy is unlikely to modify the results significantly.
These results were obtained from a single Spanish region, so generalisation of these findings should be discussed. The Canary Islands has a population of approximately 2 million, and a similar demographic structure, populations' age distribution, life expectancy and infant mortality to the average values of Spain . The National Health Service in all of the 17 Spanish regions offers universal coverage and free access to treatments consistently across regions and includes the same physical treatments for SP, NP and BP . The ratios of healthcare resources (professionals, infrastructure and technological equipment) in the Canary Islands are in line with the average values in the rest of Spain . Moreover, the structure of healthcare and referral process to physical therapies and rehabilitation services is fairly homogeneous across the country, with a high proportion of services provided by private outpatient clinics subcontracted by regional health authorities, and the clinical management of BP patients is consistent across regions . All this suggests that generalisability of these results to the rest of the Spanish National Health Service should not be a major concern.
No studies quantifying the amount of resources potentially wasted on non evidence-based physical therapies for SP, NP and BP in other countries have been found. However, approximately 50%, 65% and 94% of physiotherapists in the UK, US and Canada, respectively, use procedures which were classified in this study as non-evidence based (Table 2) . In addition, approximately 55% of primary care practitioners in the US, recommend those forms of treatment . In fact, only in the US more than 200 treatments are offered only for BP, most of which are not evidence-based [70, 71]. With this non-evidence based approach, the cost of health care provided to low back pain patients in the US increased by 65% (constant dollars) from 1997 to 2005, without generating any improvements in outcomes . In contrast to this, in The Netherlands, yearly costs derived from low back pain represented 1.7% of the Gross Domestic Product in 1995, and were reduced to 0.9% in 2002 and to 0.6% in 2007. This decrease was achieved without noticing deterioration in outcomes, and is attributed to the progressive implementation of an evidence-based clinical management [8, 73]. Although available data do not reveal the exact amount of potential savings which could be made in other countries, results from this study are likely to be generalisable to other industrialised countries, suggesting that disinvesting in not-evidence based physical therapies for SP, NP and BP, would reduce costs without worsening patients' evolution, thus substantially improving efficiency.
This study did not assess actual patient's outcomes, and it might be argued that even therapies classified as "ineffective" could have some effects which previous clinical trials failed to detect. However, this study did not have a control group, and it would be inappropriate to assume that potential patients' improvement in routine practice necessarily corresponds to benefits deriving from the treatment, as opposed to unspecific factors (such as natural history or placebo). In fact, this study was not a clinical trial designed to assess the effectiveness of each form of treatment. It was a study of clinical practice designed to quantify the resources allocated to physical treatments which are supported by the existing evidence, and those which are not. Therefore, it classified the evidence supporting the use of each of treatment, based on the evidence from existing systematic reviews and evidence-based clinical guidelines. To this purpose, it seems appropriate to assume that a potential effect which is small or uncommon enough to remain undetected in the available trials, is likely to be clinically irrelevant.