Pain Assessment–Can it be Done with a Computerised System? A Systematic Review and Meta-Analysis

Background: Mobile and web technologies are becoming increasingly used to support the treatment of chronic pain conditions. However, the subjectivity of pain perception makes its management and evaluation very difficult. Pain treatment requires a multi-dimensional approach (e.g., sensory, affective, cognitive) whence the evidence of technology effects across dimensions is lacking. This study aims to describe computerised monitoring systems and to suggest a methodology, based on statistical analysis, to evaluate their effects on pain assessment. Methods: We conducted a review of the English-language literature about computerised systems related to chronic pain complaints that included data collected via mobile devices or Internet, published since 2000 in three relevant bibliographical databases such as BioMed Central, PubMed Central and ScienceDirect. The extracted data include: objective and duration of the study, age and condition of the participants, and type of collected information (e.g., questionnaires, scales). Results: Sixty-two studies were included, encompassing 13,338 participants. A total of 50 (81%) studies related to mobile systems, and 12 (19%) related to web-based systems. Technology and pen-and-paper approaches presented equivalent outcomes related with pain intensity. Conclusions: The adoption of technology was revealed as accurate and feasible as pen-and-paper methods. The proposed assessment model based on data fusion combined with a qualitative assessment method was revealed to be suitable. Data integration raises several concerns and challenges to the design, development and application of monitoring systems applied to pain.


Introduction
Chronic pain accounts for billions of dollars in annual medical expenditures [1]; in addition to that, the resulting decreased workers' productivity contributes to indirect costs [2][3][4], and the loss of quality of life has to be mentioned as a critical related effect.As chronic pain persists over a long period of time [5], its management results expensive due to the need of long-term rehabilitation and multi-disciplinary treatments [6].In fact, the chronic condition of pain is determined by an arbitrary interval which may vary between twelve weeks, and six months [5].However, it's hard to come up with an immediate and precise assessment that leads to the right treatments, avoiding inadequately assessed and undertreated cases [7,8].Firstly, pain is a highly subjective experience for each individual [9].Secondly, due to its duration, the assessment is often accomplished at the patient's home, and this represents a challenge for the accuracy of the treatment and the cost-effectiveness of the monitoring.Therefore, as self-report is considered the most accurate pain assessment method [10,11], patients should be asked to periodically rate their pain severity and related symptoms.Unsurprisingly, in the last few years, handheld devices and Internet-delivery treatment (IdT) methods were increasingly used to enable chronic pain monitoring.These systems were used for many different purposes [12], including education, reminders, feedback (in both directions between healthcare professionals and patients), and disease control.
The ubiquity of mobile devices and the Internet raised the paradigm of the new care model based increasingly more on contacts rather than visits [13].In fact, the ability to interact with the system anywhere and anytime thoroughly changes the coordinates of time and place, and offers invaluable opportunities for new approached to healthcare delivery.Moreover, mobile devices have shown significant advances in storage capacity, battery efficiency, portability [14] and the ability to access internet-based resources [15], therefore increasing their suitability for use in healthcare systems.The adoption of technology has allowed the development of electronic pain diaries (ED) as a computerised version of paper pain diaries (PD).These systems enable patients either to report complaints close in time to the event causingpain, called ecological momentary assessment (EMA), or to address retrospective pain, that consists in pain recall over some period of time.Rather than an isolated value, pain results from multiple aspects [16][17][18][19][20], such as sensory (e.g., location, intensity), affective (e.g., depression, anxiety) and cognitive (e.g., quality of life) ones.For this reason, chronic pain patients are invited to answer many questionnaires and scores (e.g., McGill Pain Questionnaire, Visual Analogue Scale), and/or to adopt specific behaviours as a way to treat their pain in all its dimensions.For example, the monitoring program may include self-monitoring of pain, adherence to prescribed medications, regular exercise, and weight control.In summary then, the monitoring of chronic pain patients leads to many challenges across a range of topics such as technology (e.g., to collect and send data), clinical settings (e.g., duration of treatment, momentary pain or recall pain), and multi-dimensional pain assessment (e.g., questionnaires, scales).
The aims of this study are to describe ED implemented through mobile and web-based systems applied to chronic pain monitoring, and to determine the benefits obtained from adopting these technologies, in comparison to traditional pen-and-paper methods.This is carried out by means of an extensive review of the English-language literature about computerised systems related to chronic pain complaints.

Research Questions
The primary question in this review was (RQ1) Can ED replace PD for patients' monitoring?The secondary questions were (RQ2) which ubiquitous systems have been used in the monitoring of chronic pain patients?and (RQ3) which data (e.g., questionnaires and scales) are collected?

Inclusion and Exclusion Criteria
Studies were included in this review when they met the following criteria: (1) they dealt with computerised systems related to chronic pain complaints, (2) they included data about pain assessment, namely pain intensity; and (3) they were conducted using electronic means that included mobile devices (e.g., smartphone, Personnal Digital Assistant (PDA), tablet Personnal Computer) or web-based forms; (4) preliminary or definitive results were presented; and (5) they were written in English.These criteria were also applied to studies obtained from reference tracking.Reviews, study protocols, and researches where data acquisition relied exclusively on e-mails or chats were excluded.There were no age or disease restrictions: participants could be either adults or children, they might comprise chronic pain patients or healthy individuals with pain complaints.

Search Strategy
The team conducted a systematic search over the following electronic databases: BioMed Central, Pubmed Central, and ScienceDirect.Only the studies published from 2000 up until 30 June 2012 meeting the inclusion criteria were included.Every study was independently evaluated by two reviewers (Nuno Pombo and Nuno Garcia) and its suitability determined with the agreement of both parties.A third reviewer was considered to adjudicate on differences of opinion, but it was not required because a consensus was reached.The studies were also examined to identify and isolate clusters reporting the same data, so as to avoid the risk of bias [21].

Extraction of Study Characteristics
The data extracted from the studies, were tabulated (see Appendix A1) and grouped into mobile and web-based systems.For each study, details about year of publication, age of studied population (median and standard deviation (SD)), and number of participants were reported.The data managed (collected and/or complementary) by the system were grouped into three categories: pre-treatment (data obtained during the recruitment of participants were excluded), treatment and post-treatment (also includes follow up).However, data related to intervention quality and satisfaction assessment were omitted from this review.Finally, the meta-analysis included studies comprising randomised controlled trials (RCTs) that evaluated the usage of ED or IdTs and presented pre-and post-treatment comparisons.A mathematical model was used (see Section 2.7.1) to determine the effect of technology in the monitoring of pain.Firstly, the pain outcomes obtained in the RCTs' groups (intervention and control) were converted to a 0-100 scale.Secondly, a qualitative assessment (see Section 2.7.2) was performed to build an oriented analysis on pain intensity.

Quality Assessment
The methodological quality of all the studies was independently assessed by the two reviewers using a list of 10 criteria, which was formulated for the purpose of this study (see Appendix A2).Each criterion was rated as either poor/absent (=0), reasonable (=1), or good (=2).Items scores were summed to obtain a total study quality score (range 0-20).As shown in Table A1, the quality sum scores were used to classify studies into two groups, above or below an average quality threshold.

Risk of Bias Assessment
Two reviewers (Nuno Pombo and Kouamana Bousson) independently assessed the risk of bias of each RCT included in the meta-analysis (see Table A2), using the Cochrane Collaboration's risk of bias tool [22].Distinct domains were evaluated, such as: the method used to generate and to conceal the allocation sequence, the blinding of participants, personnel and outcome assessors, incomplete outcome data, selective outcome reporting and other sources of bias.

Statistical Data Fusion
The mathematical model is based on the data fusion methods described in [23][24][25] and summarized below.
Let us consider n sets of data samples, each of which has a Gaussian distribution N ( _ x i , σ i ), where _ x i and σ i are respectively the mean (or mathematical expectation) and the standard deviation of samples in the set i.Then, the probability distribution of the aggregated set is again Gaussian with a mean _ x i and a standard deviation σ computed as: where a i is defined by: and α " and σ 2 "

Qualitative Analysis
The mean and the standard deviation, computed as described in the last section, are used for the qualitative analysis method, that we proposed below, which aims to produce a more accurate outcome.Let us consider: x P `σP q for instance as shown in Figure 1 where _ x T " 3, _ x P " 2, σ T " 1.
where ai is defined by: and and

Qualitative Analysis
The mean and the standard deviation, computed as described in the last section, are used for the qualitative analysis method, that we proposed below, which aims to produce a more accurate outcome.Let us consider: σT : standard deviation of technology outcome; σP : standard deviation of pen-and-paper outcome; : mathematical expectation of technology outcome; : mathematical expectation of pen-and-paper outcome; Consider furthermore the following conditions: for instance as shown in Figure 1 where The opposite condition is presented in Figure 2 with 1, 0.9, 0.8 . The rationale of condition (P) is that since the standard deviation σ is the average magnitude of the sample dispersion with respect to its mean value (mathematical expectation), any value x that is located Example of a distribution curve when technology and pen-and-paper are qualitatively equivalent.
The opposite condition is presented in Figure 2 with _ x T " 3, _ x P " 1, σ T " 0.9, σ P " 0.8.The rationale of condition (P) is that since the standard deviation σ is the average magnitude of the sample dispersion with respect to its mean value _ x (mathematical expectation), any value x that is located at a distance from _ x less than the standard deviation (that is, |x ´_ x| < σ) may be considered as qualitatively equal to _ x.Using condition (P) described above, a qualitative analysis is performed to clarify which one among technology and pen-and-paper approach provides the best way to get fair results in pain monitoring.at a distance from less than the standard deviation (that is, |x − | < σ) may be considered as qualitatively equal to .Using condition (P) described above, a qualitative analysis is performed to clarify which one among technology and pen-and-paper approach provides the best way to get fair results in pain monitoring.Case 1: when the lower mean value (mathematical expectation) implies better results: If condition (P) is verified, then using technology or pen-and-paper gives rise to the same conclusion, even though the mean values may be different; else if ( < ) then technology provides better results than pen-and-paper; else pen-and-paper provides better results than technology.
Case 2: when the higher mean value (mathematical expectation) implies better results: If condition (P) is verified, then using technology or pen-and-paper gives rise to the same conclusion, even though the mean values may be different; else if ( > ) then technology provides better results than pen-and-paper; else pen-and-paper provides better results than technology.

Considerations for the Analysis
Several studies were excluded from this analysis due to the absence of comparisons between pre-treatment and post-treatment outcomes [26][27][28][29][30][31][32][33], or the absence of technology validation purposes [34].The remaining sixteen unique studies were assessed in terms of risk of bias (see Appendix A3).Three studies [35][36][37] were appraised to be at lowest risk of bias, as they met every criterion except the blinding of participants, personnel and outcome assessors.In fact, none of the included RCTs met this criterion.The lack of information and explanation for attrition and missing data was observed, whereas all studies clearly reported the different outcomes.These outcomes were used to implement statistical analysis across the included RCTs.During the analysis, one study was excluded due to the inexistence of SD in the reported data [38].In addition, several studies were partially excluded due to high SD in some outcomes (a.k.a.outliers) [39,40], or due to unfeasible conversion from t-scores to a continuous scale [35].Instead of a single analysis of the studies, the pre-and post-treatment data obtained from intervention groups and control groups across the different RCTs were combined using data fusion methods [23][24][25], and compared to produce a more accurate conclusion.Thus, as shown in Table 1, the adoption of technology related with pain intensity was assessed.
Example of a distribution curve when technology and pen-and-paper are qualitatively different.
Case 1: when the lower mean value (mathematical expectation) implies better results: If condition (P) is verified, then using technology or pen-and-paper gives rise to the same conclusion, even though the mean values may be different; else if ( _ x T < _ x P ) then technology provides better results than pen-and-paper; else pen-and-paper provides better results than technology.
Case 2: when the higher mean value (mathematical expectation) implies better results: If condition (P) is verified, then using technology or pen-and-paper gives rise to the same conclusion, even though the mean values may be different; else if ( _ x T > _ x P ) then technology provides better results than pen-and-paper; else pen-and-paper provides better results than technology.

Considerations for the Analysis
Several studies were excluded from this analysis due to the absence of comparisons between pre-treatment and post-treatment outcomes [26][27][28][29][30][31][32][33], or the absence of technology validation purposes [34].The remaining sixteen unique studies were assessed in terms of risk of bias (see Appendix A3).Three studies [35][36][37] were appraised to be at lowest risk of bias, as they met every criterion except the blinding of participants, personnel and outcome assessors.In fact, none of the included RCTs met this criterion.The lack of information and explanation for attrition and missing data was observed, whereas all studies clearly reported the different outcomes.These outcomes were used to implement statistical analysis across the included RCTs.During the analysis, one study was excluded due to the inexistence of SD in the reported data [38].In addition, several studies were partially excluded due to high SD in some outcomes (a.k.a.outliers) [39,40], or due to unfeasible conversion from t-scores to a continuous scale [35].Instead of a single analysis of the studies, the pre-and post-treatment data obtained from intervention groups and control groups across the different RCTs were combined using data fusion methods [23][24][25], and compared to produce a more accurate conclusion.Thus, as shown in Table 1, the adoption of technology related with pain intensity was assessed.

Results
As illustrated in Figure 3, 490 unique citations were identified, of which 378 were excluded as a result of screening, in terms of title, abstract, and keywords.The remaining 112 papers were full text evaluated, which resulted in the exclusion of 63 papers that did not match the defined criteria.Furthermore, the reference tracking allowed for the inclusion of 13 additional papers.In summary then, our review examined 62 papers, representing 55 unique studies, due to the fact that studies reporting the same data were clustered to avoid risk of bias.

Results
As illustrated in Figure 3, 490 unique citations were identified, of which 378 were excluded as a result of screening, in terms of title, abstract, and keywords.The remaining 112 papers were full text evaluated, which resulted in the exclusion of 63 papers that did not match the defined criteria.Furthermore, the reference tracking allowed for the inclusion of 13 additional papers.In summary then, our review examined 62 papers, representing 55 unique studies, due to the fact that studies reporting the same data were clustered to avoid risk of bias.The included studies encompass a total of 13,338 participants distributed by 43 studies (78%) related to mobile systems, and 12 (22%) studies dealing with web-based systems.Eighty-one percent of mobile systems (35 studies) were designed to enable usage in patients' home whereas the remaining eight studies limited their use within hospital facilities.The data were collected at intervals or during the clinical visit or at the end of the study, and transmitted to the system database by different channels, such as: Internet, SMS, or cable.Web-based systems were reported in 12 studies varying between online questionnaires and cognitive-behavioural therapy (CBT).Moreover, 10 studies (83%) used phone calls, SMS or emails as a complement of the IdT.This methodology aims to remind patients to collect data, support system handling, and to establish contact between healthcare professionals and patients.
Moreover, 16 out of the 55 studies (29%) included in this review were published before or during 2006, and among the remaining 39 studies, 27 studies were published between the beginning of 2008 and the end of 2010.Thirty-two studies (58%) included complementary data, obtained outside the system in at least one of the following phases: pre-treatment (28 studies), treatment (8 studies) or post-treatment (16 studies).
The most representative objective was validating the IdT (12 studies, 22%), the assessment of ED (12 studies), the comparison between ED and PD (nine studies), the comparison between recalled pain and EMA (six studies), and the evaluation of medication in treatment of patients suffering from pain (three studies).Eight studies reported the correlation of several pain conditions, namely: physical activity, relationship, emotional distress, fear, and sleep.
The CBT was presented in 19 studies, among which seven were related to mobile systems.The remaining 12 studies presented CBT as a support of IdT, and included tailored exercises according to participants' symptoms, multimedia content, information and lessons about physical, cognitive, behavioural and motivational topics.The main principles of CBT for chronic pain management are based on helping the patient understand how much pain is experienced, coping-skills training, and cognitive restructuring affected by cognition and behaviour [49].

Mobile Systems
Forty-three studies were related to mobile systems, out of which 35 (81%) were designed to allow their usage in patients homes during at least one phase of the intervention (pre/post-treatment, treatment).The remaining eight studies were limited to the use of the proposed system in hospital facilities during patients' visits and thereby only comparisons among sporadic records collected during the treatment period were provided.Meanwhile, 19 studies presented transmission of data to a remote server immediately after its edition.Three studies did not report this process, whereas 21 studies reported elapsed time between the editing and the subsequent delivery.Thus, data were collected at intervals, or during the clinic visit, or at the end of the study.Internet was the preferred channel for sending data (14 studies), followed by uploading through personal computer (nine studies), and SMS (three studies).Data transmission after its edition may allow real-time access to physicians, and therefore, clinical decisions supported by updated information on the patient's conditions.Moreover, undelayed data transmission may provide the enforcement of triggering messages and alerts according to the reported pain values.This method was highlighted by four studies and comprised a clinical session report generation, SMS alerts according to answers and warning messages about the activity patterns, displayed in PDA.Data storage in a Personal Health Record (PHR), wrist actigraphy in sleep assessment, and activity monitoring supported by a Body Area Network (BAN) were proposed in a single study respectively.Interactive voice recorder (IVR) was referred in two studies [44,50].Time of intervention ranged from one clinical session to 52 weeks (one year).

Web-Based Systems
Web-based systems were reported in 12 studies, out of which 11 consisted in RCTs, comprising two groups of participants called: intervention group (IG) and control group (CG).The difference between them is that a web site was used to deliver the treatment to IG participants.At the end of intervention, participants of both groups were assessed and the IdT effects were determined.The IdT consisted of online questionnaires and/or CBT.All the articles reported positive effects and improvement in health status.With the exception of [37,39], all web-based systems used emails or phone calls jointly with Internet (83%).Six studies adopted e-mails [41,42,45,46,48,51] and three of them also performed phone calls [45,46,48], to remind patients to use and/or interact with the system.In addition, emails were applied to obtain data [40,45,46,51], to support the system handling [36,41], and, together with phone calls, to establish a contact between healthcare professionals and patients [36,43].One study [40], allowed phone calls to support the system handling.Finally, [52] used SMS to remind patients to collect data.In the same study, mobile phones with Internet access were used to present a web site whereupon treatment was provided, and therefore, it has been classified as a web-based system.Time of intervention ranged from 3 to 52 weeks (one year).It should be noted that remote data transmission is not required in these systems, while it occurs in mobile monitoring applications.

Meta-Analysis
The qualitative and quantitative analysis (see Section 2.7) revealed that the benefits of technology and pen-and-paper are equivalent on pain intensity (48.67 P (50.98 ˘3.35) and 50.98 P (48.67 ˘3.49)), as presented in Table 1.Firstly, the proposed statistical data fusion model processes each study as a different sensor and computes the individual mean and SD, related with the processed variable (determined by the collected data from questionnaires and/or scores) in both arms of the study (pen-and-paper group and computerised system group).Secondly, the combination of these values resulted in an aggregate value.Thirdly, the fusion model computes all aggregate values and presents a final decision according to the rules defined in Section 2.7.

Discussion
Some potentials and risks related to mobile and web-based systems were evidenced from the full text evaluation of included studies.Firstly, the usage of ED produces more reliable data compared to PD.Secondly, ED and IdT result in real-time analyses and subsequent agile treatment adjustments.Thirdly, ED and IdT provide time-saving and enable cost-effective medical practices.Nevertheless, training for clinical staff is critical [53], and strongly recommended to promote standardised procedures and adherence [54].In addition, device failures considered in system design [55], should be addressed to avoid missing values and/or prolonged data editing.It should be noticed that due to the frequent loss of mobile devices, their use to store health records implies the risk of losing data and personal information.These topics, along with the inefficient use of collected data to improve treatment effectiveness, emerged as critical limitations.This review included 19 studies related to CBT, in which the following outcomes were observed: the effectiveness for decreasing chronic pain, in line with [49,56,57], the reduction of pain related behaviours as suggested by [58,59], and a facilitated return to work, as presented by [60,61].In spite of their absence in these studies, innovative CBT, such as: serious games [62,63] and augmented reality [64,65], seem to be promising.Serious games are the application of motivational aspects of gaming to encourage positive health behaviours [66], whereas augmented reality provides virtual environments combined with touch sensations resulting from interacting with real objects [67].Further work is needed to understand how these technologies can aid the transformation of CBT delivery models.
The use of SMS [68] to collect data, as proposed by [69][70][71][72], and to deliver CBT, as suggested by [52], may improve treatment outcomes, due to the fact that tailoring messages to individuals may lead to effective health behaviour changes [73][74][75].Only one study [76], mentioned data integration with other systems such as PHR, which suggests limitations on accessing the collected data.In addition, some mobile-based systems were designed to interact directly with patients without the presence of a healthcare professional [77,78] and/or without evidence of reliability and accuracy.However, as pain is a multifaceted experience, its therapy tends to involve many healthcare professionals and different expertises whereby the data integration may result in the reduction of not regulated self-diagnosis [79].Therefore, it is desirable that patient information may be obtained and delivered both easily and safely (e.g., avoidance of medical examination redundancy, faster patient profile acquisition, and permanent storage of clinical records) which raises some concerns and challenges related to security aspects such as privacy and confidentiality [80], and reliable communication methods between healthcare professionals and patients.
In line with this, cloud computing as an emerging technology that provides elastic infrastructure, and efficient resource utilization [81], appears to be a promising solution for design, development and integration of systems.This technology may enable scalable, portable, and interoperable mobile and web-based systems as to deliver clinical solutions to the patients, anytime and anywhere [82].In addition, social media websites have been useful in the last few years to improve networking and communication [83] (e.g., Facebook, Twitter), and represent a new source of information and knowledge.Therefore, it is expected that clinical systems will advance to interact with patients via social media, so as to provide CBT, serious games, self-help, symptoms information and multimedia content.Thus, new studies should be addressed to determine the real benefits and disadvantages of treatments delivery using social media.Furthermore, complementary studies should be carefully addressed to analyse both data and patient privacy.
Finally, our meta-analysis demonstrated that the effects of technology and pen-and-paper should be obtained not only based on the comparison of the standard deviations together with the values of the mathematical expectations, but also considering the condition (P) described in Section 2.7.In addition, the outcome of our meta-analysis is highly accurate as evidenced by the lower SD of the obtained aggregated values (resulting from the computation of the SD of all the included samples) compared with the individual SD presented in each study.In fact, we found that technology and pen-and-paper present equivalent outcomes suggesting not only that technological systems are feasible, but also there is room for improvement to produce significant effects in patients' conditions and welfare.Moreover, further studies should be promoted to determine not only the effect of technology on different dimensions of pain (e.g., anxiety, depression, catastrophizing, disability and interference) but also the side effects of the application of technology in economic, medical, educational, and social domains.

Conclusions
This review distinguished mobile and web-based systems related to chronic pain complaints.Sixty-two studies were examined and the main findings are summarised as follows: (RQ1) The qualitative analysis model, stemming from the data fusion method, combined with a quantitative model, based on the comparison of the standard deviations together with the values of the mathematical expectations, revealed that technology is equivalent to pen-and-paper in terms of effect on pain intensity monitoring; (RQ2) Sixty-two studies were included encompassing 13,338 participants.A total of 50 (81%) studies related to mobile systems, and 12 (19%) related to web-based systems; (RQ3) The data extracted from the included studies, revealed the use of almost ninety different scales and questionnaires at pre/post/during treatment.The data collected comprised, among others: location, duration, and intensity of pain, consequences as the impact on quality of life, emotional and aversive aspects.This highlights the multi-dimensional nature of pain.
Despite these findings, effects of technology on practitioners and patients outcomes remain understudied, and their promise to increase self-care and accurate monitoring remains mostly untested.In addition, data integration raises several concerns and challenges to the design, development and application of monitoring systems applied to pain.

Limitations
Some limitations of this review should be mentioned.First, only English-language publications were included.Second, the lack of technical explanations related to data acquisition, transmission and storage, restricted both analysis and extraction.Third, the null hypothesis was considered, that means, all sample data are assumed to be sufficient.
σ T : standard deviation of technology outcome; σ P : standard deviation of pen-and-paper outcome; _ x T : mathematical expectation of technology outcome; _ x P : mathematical expectation of pen-and-paper outcome; Consider furthermore the following conditions: Condition (

Figure 1 .
Figure 1.Example of a distribution curve when technology and pen-and-paper are qualitatively equivalent.

Figure 2 .
Figure 2. Example of a distribution curve when technology and pen-and-paper are qualitatively different.

Table 1 .
Comparison between pen-and-paper, and mobile and web technology using pre and post treatment results by study and overall.

Table A1 .
Cont.GSRS-IBS, IBS-QoL, ASI, GAD-Q and CPSQ, conducted at pre-and post-treatment and 3-month follow-up CBT: gastrointestinal symptoms and stress and on relaxation training, stress management, catastrophic thinking, exposure therapy and the social consequences of IBS CPAQ, PCS Pain intensity, interference of pain, planned and achieved activities, feelings, pain-related fear, avoidance, catastrophizing and acceptance, 3 times per day (morning, evening and a time randomly chosen between 11:30 a.m. and 2 p.m.) CBT: feedback SMS with praise, encouragement messages, two separate websites, one for child access and one for parent access.The child access comprised eight treatment modules (education about chronic pain, recognizing stress and negative emotions, relaxation, distraction, cognitive skills, sleep hygiene and lifestyle, staying active, relapse prevention).

Table A1 .
Cont.Pain intensity (VAS), duration, BDI, HDI, MLPC.Treatment: Number of times and the total time used for training relaxation.MINI, PD-IIP SF-36, BPI, MFI, MOS-SS, CES-D, STPI and PGIC at pre and post-treatment CBT: multimedia content following topics: educational lectures, symptom management and adaptive life style