The accuracy of self-reported adherence to an activity advice

Abstract Objective. The aim of this cross-sectional observational study was to investigate whether patients referred for non-specific low back pain accurately report whether they follow home-based activity advice. Research suggests that inadequate adherence to home-based activity advice during an intervention period might diminish the effectiveness of an intervention. Insight into patient adherence to home-based activity advice is hindered by a lack of objective data. Most data consist of self-reporting and interviews. Methods. Participants were advised to complete a home-based activity for 1 week. We validated self-reporting of home-based activity with data obtained from an activity monitor worn by 51 patients who reported non-specific low back pain. The proportion of accurate self-reporting and kappa values were calculated to determine the level of accuracy of self-reporting. Results. An analysis of 357 (51 × 7) walking days found accurate participant reports on 233 days; only 22% of the participants reported accurately on all 7 consecutive days. Low kappa scores between the barrier list and the activity monitor show the poor reliability of participants’ self-reporting. Conclusion. Self-reporting about the performance of advised home-based activity for the management of non-specific low back pain in a physiotherapy practice setting is rather inaccurate.


Introduction
Low back pain is the most prevalent health problem seen in physiotherapy clinics in the Netherlands [1]. Current evidence supports the use of exercise-based physical therapy treatment for non-specifi c low back pain (NSLBP). This encourages patients to take an active role in their recovery [2][3][4][5][6], but there is little specifi c guidance about what works best for whom [7].
Exercise therapy includes the explanation and practice of exercise programmes in the clinical setting. The prescription of home-based exercise programmes in combination with general activity advice (AA) has become the mainstay in physical therapy treatment of NSLBP [8 -11]. Asking a patient to perform a prescribed activity outside the clinical setting and without supervision of the health-care provider implies an appeal to the patient to adhere to therapy. However, fall outside the scope of Medical Research Involving Human Subjects Act (WMO) and therefore did not require approval from an ethical committee or registration in a trial register .

Participants
We included patients who presented with NSLBP, which we defi ned as pain or discomfort localized below the costal margin and above the inferior gluteal folds, with or without leg pain according to the guideline for NSLBP [33]. Pain can be increased by changes in posture, movements or external load [2]. Patients were excluded from the study if they were unable to read, write or speak the Dutch language, or if they had a known medical history of cancer, osteoporosis, rheumatoid arthritis, tuberculosis, trauma and fractures in the lumbar spine, recent infections in the musculoskeletal system or visceral abnormalities. We excluded patients who presented with red fl ags at intake as described in the guidelines for low back pain [8,34] and those who indicated their inability to adhere to the walking advice. Included patients received a standardized intake form on which information about pain history, age, gender, education, work, aetiology, signs, symptoms and clinical status of pain and disability was recorded (see Supplementary Appendix 1 available online at http://informahealthcare.com/doi/abs/ 10.3109/ 21679169.2015.1075588).

Measurement tools
We used the visual analogue scale (VAS) for pain and the Quebec Back Pain Disability Scale (QBPDS), which have both been validated for assessing pain and disability [35][36][37][38]. We performed a physical examination to collect information about segmental hypomobility and aberrant movement of the spine [39,40]. Inspection of active fl exion and extension of the spine, assessment of possible aberrant movement with outcomes such as painful arc, deviation during movement, thigh support of the patient and manual passive segmental examination of the spine to assess normal or decreased segmental mobility are all standard procedures for examining the spine in manual therapy. Outcomes of these procedures informed the examiner about functioning of the spine [8,9,39 -42]. This examination, in combination with the use of the VAS for pain and the QBPDS questionnaire, were considered to be a suffi cient assessment of each participant ' s clinical status in the context of this study.
The results of the examination and assessments also enabled us to assess whether a participant ' s clinical status was serious. The clinical status was considered to be serious if a participant ' s VAS and The instrument of self-reporting used in this study is a barrier list. Barriers have been recognized to impede adherence with AA [24,25]. Given that a diary potentially infl uences study outcomes [18,26], a barrier list seemed to be a feasible alternative. The use of a barrier list is simpler than the use of a diary: by fi lling in (or not fi lling in) a score of 0-100 on the barrier list every day, patients record whether they followed the AA that day. Compared to a diary, a barrier list is less prone to inaccuracy due to misinterpretation, social desirability, and reliance on accurate recall of the intensity, frequency and duration of daily activities [19].
The literature suggests differences in the accuracy of self-reporting between demographic groups, such as according to gender, age, education and work [27 -32]. We were also interested in whether the degree of suffering NSLBP could infl uence the accuracy of self-reporting. We could not fi nd cut-offs for the degree of suffering NSLBP in the literature and decided to make a cut-off between serious and nonserious clinical status for the purpose of this study. We saw this as the only way to inform the reader about the infl uence of the degree of suffering NSLBP on the accuracy of self-reporting.
The primary aim of this study was to investigate the accuracy of self-reporting by patients suffering from NSLBP. The secondary aim was to investigate differences in the accuracy of self-reporting adherence with AA between demographic groups of patients. The research question was whether we could determine the accuracy of self-reported data about adherence to AA compared to data from a validated activity monitor in patients suffering from NSLBP.

Study design and sample size
Using an observational study design, we prospectively studied 51 patients who were suffering from NSLBP and had been referred for physiotherapy treatment. In addition to their usual treatment based on the guideline for NSLBP, they received the following AA: walk for a half-an-hour per day for 7 consecutive days, while keeping track of barriers and wearing the DynaPort MoveMonitor for 24 h per day during this period. We studied the data collected from these patients over 1 week. The study was performed in two centres for physical therapy in Papendrecht and Rotterdam, the Netherlands.

Ethics
The Medical-Ethical Testing Committee at Atrium-Orbis-Zuyd considered this observational study to QBPDS scores were above 60 and at least one of the following problems was found during the physical examination: segmental hypomobility or aberrant movement of the lumbar spine. However, this cut-off was only determined for the purpose of this study; it is not mentioned in the literature and its reliability and validity have not been investigated.
The barrier list, which corresponds with the setting of this study, contained barriers mentioned in recent adherence research [24,25]. Participants were asked to score a 0 on the items on the barrier list on the days that they followed the walking advice correctly (in their opinion). On the days when they did not follow the walking advice, participants were asked to give each item on the barrier list a score of between 0 and 100 points (see Supplementary Appendix 1 available online at http://informahealthcare.com/doi/ abs/ 10.3109/21679169.2015.1075588). If a patient recorded a score of 0 on a particular day, the researchers concluded that the patient was reporting having walked according to the advice that day. If one or more items were scored on a particular day, the researchers concluded that the participant was reporting not having walked according to the advice that day.
The DynaPort MoveMonitor (McRoberts BV, The Hague, the Netherlands) was used to record daily activity. This activity monitor has been validated to measure intensity and duration of daily activities such as walking, shuffl ing, sitting, standing and lying [22,23,43,44]. It also registers periods during which the monitor is not worn. Data from the activity monitor were transformed into individual activity reports ( Figure 1). The subjects were instructed to wear the activity monitor continuously during 1 week and it was emphasized that it must be worn 24 h a day. The activity monitor was attached at the level of the processus spinosus L3, using an elastic belt around the waist and under the clothes. Patients were instructed to take it off only when it was in danger of getting wet (e.g. showering, swimming), but to refasten it immediately after such an activity. Because of the simplicity of using the device, no further instruction was needed. The activity reports, processed from the activity monitor data, show daily information about periods of walking. These periods are expressed with a start and end time, the number of steps, the continuousness of locomotion and the acceleration.

Intervention
Previous research has shown that walking is an acceptable type of exercise because it does not require training or equipment and can be performed at an individual ' s own location and time, with little risk of injury [45]. The AA was for patients to walk outdoors for at least half-an-hour hour per day at a normal walking speed on 7 consecutive days. A " normal walking speed " is walking with a purpose, destination or goal to walk towards, not shuffl ing as one does when shopping, but also not speed-walking as if it were a competition. The patient could choose what time to start taking a walk. The only purpose of this half-hour walk was to follow the advice given. To prevent misinterpretation of the advice and confounding, the patient was not allowed to combine it with another task (e.g. walking the dog, shopping).

Procedures
After screening and intake, eligible patients received written and verbal information about the study. Patients who decided they wanted to be included in the study were given an informed consent form and were scheduled for a second contact appointment. This appointment was in addition to the assessment strategy recommended for NSLBP according to the Dutch guideline for low back pain [46] and was used only for this study.
During the second contact, after patients returned the signed informed consent form, they received advice about how to follow the AA. It was explained that we were interested in the movement behaviour of people suffering from NSLBP. To avoid socially desirable behaviour, we gave them no specifi c explanation about the walking registration. Each participant fi lled in all the items on the standardized intake form, assisted by the researcher. We then handed over the activity monitor and the barrier list, and gave them instructions about how to complete the barrier list daily.
During the third contact (1 week after the AA started), the patient returned the activity monitor and the barrier list. Further treatment took place according to the guideline for low back pain.

Data analysis
The amount of walking that would be measured by the activity monitor was expected to differ across participants because of variations in height, weight, age, lifestyle and health conditions. Therefore, we had to determine a cut-off value for a minimum number of steps per half-hour. Considering that slow-paced walking [with a minimum speed of 3.2 km/h (47)] is acceptable and taking the average length of a walking step [set at 80 cm (48)] into account, we created a formula [(3200 m/h/2)/0.80 m] that returned a cutoff value of 2000 steps/half-hour (66.6 steps/min).
All walking periods reported in the daily activity reports were assessed for duration of the walking period, number of steps, continuity of walking and the movement intensity of walking. Walking periods that lasted for half-an-hour or more, consisted of 2000 or more steps per half-hour and had no interruptions in the continuity or acceleration of walking were judged to adhere with the AA. We were aware of the possibility of short interruptions during a halfhour walk (e.g. traffi c, traffi c lights) but the outcomes of the activity reports are corrected for these interruptions [22]. To investigate adherence to AA, for the purpose of the study the AA had clearly to be distin- guished from normal daily activities to prevent information bias. In this study the aim of the half-hour walk was considered as a home-based exercise in the treatment of NSLBP and this was also explained to the patient. The criterion " no interruption " is not defi ned because a cut-off was chosen of 2000 steps/ half-hour instead of walking speed. This permits participants to stop for traffi c and other short-lasting obstacles during the half-hour walk. These interruptions are shown in the activity report but are not recorded as an interruption of the continuity of the walking period.
After this, the barrier lists were scored to avoid expectation bias. A patient reported accurately on a day whether no barriers were reported on a day where the patient was adherent to the AA according to the activity monitor, or barriers were reported on a day where the patient was not adherent to the AA according to the activity monitor.
In this study, we did not compare outcomes of measurement instruments to judge the accuracy of the measurement instruments, but we were interested in the accuracy of self-reported AA by comparing outcomes of the barrier list with outcomes of the movement monitor. For this reason, we defi ned dichotomous cut-off values for both measurement instruments.
We performed two analyses. In the fi rst analysis, we measured the accuracy of participants ' self-reports on all 357 measured days (51 participants ϫ 7 days) and the number of days that was accurately reported per participant. For each day separately, we calculated the percentage of accurate self-reporting and kappa (standard error) as a measure of agreement between outcomes of the activity monitor and selfreporting. In the second analysis, we assessed the infl uence of demographic determinants (gender, age (18 -39, 40 -59, Ն 60 years old), education level (low, medium, high), work (white/blue-collar work) and clinical status (serious/non-serious) on the proportion of participants who accurately self-reported on all 7 consecutive days, using Fisher ' s exact test (see Table II

Results
We introduced 58 eligible patients with NSLBP to the study. Of those, six refused to join the study, one discontinued the assessment period because of infl uenza, and 51 completed the study and handed in their activity monitors and barrier lists. The average age of the participants was 48.7 years (SD ϭ 14.3). On the days for which participants scored no barriers and walked for half-an-hour or more, they walked 79.9 steps/min (SD ϭ 21.3) on average.
In the fi rst analysis, the participants scored no barriers on 247 of the total 357 measured days, so based on their self-report they had walked according to the AA. The walking limit of 30 min (2000 steps/ half-hour) was confi rmed by the activity monitor on 131 of those 247 days (53%), and was therefore reported accurately by the participants on those days. On the other hand, on 110 days participants reported on the barrier list that they had not walked according to the walking advice, while according to the activity monitor, participants did not walk for half-an-hour or more on 102 out of these 110 days (93%). So, in total, the participants reported accurately on 233 out of 357 days (65%) ( Table I).
Using each participant ' s analysis, 11 out of 51 participants (22%) reported accurately on 7 consecutive days (last column in Table I). The percentage of accurate self-reporting varied from 53% (day 2) to 76% (day 5) and showed no clear increase or decrease over the 7 days (second-to-last row in Table I). Observed kappa values showed a poor ( k 5 0.24 on day 2) to moderate ( k 5 0.48 on day 5) agreement (last row in Table I). Again, no clear increase or decrease over time was found. A comparison of demographic variables with the outcomes of the second analysis showed no signifi cant differences between groups in the proportion of participants who accurately self-reported on all 7 consecutive days (Table II).

Discussion
The literature mentions that following AA is benefi cial to managing NSLBP. However, investigators have documented poor adherence in following daily AA [10 -13]. In adherence research outside the clinical setting, data were merely obtained from interviews and self-reports [13,16]. These subjective data may be inaccurate (e.g. recall bias) and infl uenced by social desirability [17,19 -21]. Nowadays, modern technology provides validated activity monitors small enough for patients to wear without limiting their daily activities [22,23]. These devices are therefore suitable for an adherence study outside the clinical setting and provide more robust data. The activity monitor was used as the gold standard in this study.
Although a lot of work has been done on selfreporting compared to objective measurement by movement monitors, only a few studies have been performed in physiotherapy clinics and only one in primary care [49]. To make our study relevant to physiotherapists working in primary care, the study was conducted in private physiotherapy clinics where patients suffering from NSLBP attending these clinics were recruited in order of appearance during the  The numbers before the slash represent the measured walking time (minutes) and the numbers after the slash represent the number of steps measured during this walking time.
To judge whether a participant followed the activity advice (AA) daily, participants had to walk for at least half-an-hour, with at least 2000 steps per half-hour. Therefore, the number of steps has to be divided by the walking time and multiplied by 30 to assess whether 2000 steps per half-hour were actually performed.
An asterisk in a cell indicates that barriers are scored by the participant, which implies that the participant did not execute the AA; no asterisk in a cell indicates that no barriers are scored, implying that the participant executed the AA. If a cell contains no asterisk and walking time is Ն 30 min and/or the number of steps is Ն 2000 steps/half-hour, the self-report on that day is accurate. If a cell contains an asterisk and walking time is Ͻ 30 min and the number of steps is Ͻ 2000/half-hour, the self-report on that day is also accurate. a Indicates participants with 7 days of accurate self-reporting.
investigation period. It is therefore helpful for physiotherapists working in primary care to know that their patients (suffering from NSLBP) are inaccurate in self-reporting adherence to AA.
To control for misinterpretation, social desirability and reliance on accurate recall of the advised activity, a barrier list was used instead of a diary. All barrier lists were returned and correctly fi lled out, and therefore inaccurate memory was not an item in this study. We think that completing the barrier list can be potentially infl uenced by social desirability, but does not infl uence the accuracy of self-reporting. Despite this, all analyses confi rmed the hypothesis of poor accuracy of self-reporting. The rate of poor accuracy was surprisingly high, with only 22% of the participants reporting accurately on all 7 consecutive days. The 7 day period of the AA was suffi cient to determine the accuracy of self-reporting, because the course of a participant ' s accuracy scores during the week showed no decline or progress and was therefore not infl uenced by indolence or loss of interest by the participants. A risk of response bias by indolence or loss of interest by the participants might have appeared if this measurement period had lasted longer than 1 week.
The investigated population can be generalized to a physiotherapy setting in primary care in the Netherlands for the following reasons. First, the study was performed in two physiotherapy clinics: one in a working-class neighbourhood in a big city and one in a middle-to upper-class neighbourhood in a village. Secondly, participants were included in order of appearance at the clinics.
Low kappa scores between the barrier list and the activity monitor corroborate the poor reliability of participants ' self-reporting that is mentioned in the literature [17,19--21]. Our assumption that a barrier list would promote more accurate self-reporting than a diary was therefore not confi rmed.
No barriers were scored on 47% of the reported days, meaning that participants believed that they followed the AA on most days. However, this activity was only confi rmed in 53% of these days by the activity monitor. As mentioned in the literature [18], this kind of inaccurate self-reporting may be caused by the participant overestimating his or her performance of the advised activity.
However, underestimation of performance hardly occurred in this study. Participants who scored barriers and therefore reported not following the AA were accurate 93% of the time (Table I). On 8 days (7%) barriers were reported inaccurately, six of these by a single participant. This participant followed the AA but reported barriers. Overall, this means that participants had good accuracy when scoring barriers and correctly estimating their lack of following the AA.
It is diffi cult to analyse walking advice given to a population with variations in body height and weight. The advantage of the activity report from the activity monitor is that outcomes are reported as the number of steps during a time period and not as walking speed and distance. Therefore, a minimum number of steps per half-hour was calculated based on the literature. To verify whether this cut-off value was realistic, we calculated the average number of steps per minute taken by the participants who scored no barriers and walked for half-an-hour or more. We found only a 20% difference between the cut-off value (66.6 steps/min) and the average number of steps (79.9 steps/min). Therefore, the cut-off value was not too strict for the participants and we considered it to be realistic.
This study had several limitations. One was related to the interpretation of the participants ' walking activity and how to fi ll out the barrier list. To minimize false interpretations, we deliberately asked each participant to walk for the purpose of this investigation and not to combine it with other activities, and participants were instructed fully on how to fi ll out the barrier list. We cannot exclude the possibility of misinterpretations in this study. Misinterpretation was not measured in the setting of this study, and may infl uence these study outcomes. However, when participants were asked after the instructions about performing the AA and how to fi ll out the barrier list, they were found to have understood the instructions. Furthermore, all barrier lists were returned correctly fi lled out. We were also aware of the bias that wearing an activity monitor would create with relation to following and reporting the AA. To avoid or minimize this bias, we explained to the participants that this study was about patients ' movement behaviour. However, the infl uence of wearing the device on the participants ' self-reporting behaviour was not questioned after the 7 day walking period and no data are available on this subject. We assume that self-reporting would have been more accurate if specifi c information about the walking registration had been provided, but this remains a topic for further research. The researcher ' s attitude and communication skills in the professional relations with the patient, which have been mentioned as infl uencing factors in a patient ' s adherence to therapy [50], were not included in this study.
The clinical statuses of the participants were determined for the purpose of this study and were not validated or investigated for their reliability. However, the procedures of the patient examinations have been well described and VAS scores and the QBPDS questionnaire have been externally validated, which supports the internal validation of the outcomes of the clinical status. Thereby, clinical status showed no signifi cant infl uence on the accuracy of self-reporting.
Because of the small sample size, we could not calculate the independent infl uence of all demographic variables and clinical status on the accuracy of self-reporting in one model. To achieve reliable statistical analysis, more participants need to be included in future research.
The results of the present study support the evidence that self-reporting of following AA outside the clinical setting is inaccurate. In conclusion, we recommend the use of a validated activity monitor as a measurement instrument in adherence research outside the clinical setting.

Declaration of interest:
The authors report no confl icts of interest. This paper presents independent research for which there were no external sources of support or funding.