Comparing what the clinician draws on a digital pain map to that of persons who have greater trochanteric pain syndrome

Introduction

In clinical practice of musculoskeletal pain, clinicians typically record information about the size, shape and location of pain areas described by their patients on a body chart [1], [2], [3], [4]. However, there is limited empirical knowledge on how this information is used to guide clinical decision making and patient management.

Pain is an unpleasant sensory and emotional experience associated with, or resembling that association with, actual or potential tissue damage [5]. As pain is influenced by biological, psychological and social factors, there is potential for diversity in how the patient reports the pain experience [6]. Diversity in pain drawings can also be caused by the assessors’ perceptions and ability to draw verbal and physical cues on a body chart.

Although digital pain drawings are a commonly used tool, to the authors’ knowledge, no scientific evidence exists on how pain drawings are used in clinical practice. Notably, from the authors’ clinical experiences, not the patient, but the clinician is often the one to complete the pain drawing in clinical practice. This has implications for diagnosis and treatment. Specifically, differences in perceived pain distributions between clinician and patient may cause misinterpretation and miscommunication at the diagnostic and treatment stage [7].

Digital pain drawings have been shown to enable patients and clinicians to analytically understand the location, area and distribution of pain [8]. Digital pain mapping is rapidly evolving and demonstrates good usability, reliability, and repeatability in acute pain [9, 10] and persistent pain conditions of the knee, neck and back [11], [12], [13].

To be able to inform and strengthen clinical practice through the use of (digital) pain drawings by clinicians and/or patients, we need to know the similarities and differences in pain drawings between clinicians and patients. A better understanding of the similarities and differences may improve communication, benefit management and improve patient – clinician interaction/experience satisfaction with treatment [9, 14].

Greater trochanteric pain syndrome is one of the most commonly treated tendinopathies in middle-aged, overweight females [15]. It is classically characterised as pain on the lateral side of the hip over the greater trochanter [16]. Some patients also experience pain in the buttocks, lower back and/or distally down the leg [17]. The diverse distribution of pain may reflect different anatomical sources of that pain, that are identified on imaging of greater trochanteric pain syndrome. Alternatively, the diversity in pain distributions may stem from varied pathologies associated with greater trochanteric pain syndrome, such as hip osteoarthritis and referred spinal pain [18], [19], [20], central nervous system changes [21, 22], personal characteristics [23, 24], or a combination of these factors. Regardless of the mechanisms underpinning the diversity of clinical presentation, it is important that patients and clinicians can evaluate the distribution of pain in a consistent manner.

To compare patient and clinician representations of pain experienced by individuals who have persistent pain due to greater trochanteric pain syndrome, we explored the agreement between full-body digital pain drawings completed by the patients and a clinician. The specific aims of this study were to (1) assess the agreement between the drawn pain areas, and (2) assess any differences in shape and location of pain drawings. We also collated all pain maps to provide an overall representation of this cohort’s pain distribution.

Methods

Results

Similarities in area between pain drawings using Bland–Altman plots

The Bland–Altman plots comparing patient and clinician drawings showed a mean difference between −0.45 and −0.12% of the total pixel counts for all charts. Mean differences (LoA) showed systematic, negative fixed biases for the back (−0.45%; LoA −2.36%, 1.56%), front (−0.32%; LoA −1.08%, 0.44%), left (−0.28%; LoA −2.11%, 1.56%), and right (−0.12%; LoA −1.49%, 1.24%) full body drawings, indicating that clinician pain areas were marginally smaller compared to patients’ (Figure 1) [30]. A high proportional bias was found for the front charts (PCC −0.88, p<0.001). This indicates that differences between patient–clinician pain drawings for the front charts (e.g., fixed negative biases) increases with larger areas of pain. Low and negligible proportional biases were observed for the left (PCC −0.37), back (PCC −0.11), and right (PCC 0.06) charts (all p>0.09). But not for back, left and right charts.

Figure 1:

Bland–Altman plots showing the level of agreement in pain area between patient and clinician pain drawings for full-body views of the (A) back, (B) front, (C) left side, (D) right side. The limits of agreement are marked with dashed lines.

Differences between pain drawings in shape (bounding box) and location (Jaccard index)

Overlays illustrating the pain distribution for the female patients (n=21) and the clinician drawings are shown in Figure 2. Visual inspection of the location and overall pain area suggest minor differences in location and shape of patient and clinician drawings.

Figure 2:

Visual comparison of patient (left column) and clinician (right column) drawings for female full-body charts of (A) the back, (B) front, (C) left and (D) right views (n=21). Men overlay images are presented separately to the females, because the male and female avatars are unique and could thus not be overlayed (Figure S3 for male overlay images).

No differences in the bounding box(x, y) between participant and clinician drawings were found for any of the chart views (all p>0.17, Table 2). The mean Jaccard indexes were 0.09 (95% CI 0.04, 0.16) for the back view, 0.11 (95% CI −0.01, 0.23) for the front, 0.15 (95% CI 0.01, 0.29) for the left and 0.18 (95% CI 0.00, 0.35) for the right view (all p<0.05), indicating the overlap of a patient and clinician drawings were very small compared to the sum total of the areas drawn.

Table 2:

Mean differences (95% confidence intervals), p-value (2-sided paired t-test) in bounding box (BBx, y) values between patient and clinician drawings (mean [standard deviation]).

Bounding box	Patient	Clinician	Difference
Back view
BBx	144.91 (100.75)	116.48 (96.83)	−28.44 (−70.21, 13.24), 0.17
BBy	236.70 (226.20)	222.04 (228.44)	−14.65 (−133.15, 103.84), 0.80
Front view
BBx	110.74 (111.04)	76.48 (97.76)	−24.26 (−85.57, 17.05), 0.18
BBy	184.83 (175.81)	123.04 (167.39)	−61.78 (−153.10, 29.53), 0.18
Left view
BBx	68.61 (56.58)	63.65 (46.03)	−4.96 (−24.33, 14.42), 0.60
BBy	188.00 (210.77)	137.91 (168.45)	−50.09 (−147.99, 47.82), 0.30
Right view
BBx	81.74 (58.92)	72.35 (49.43)	−9.39 (−37.49, 18.70), 0.50
BBy	170.04 (168.39)	152.17 (192.14)	−17.87 (−118.24, 82.50), 0.72

Discussion

Our findings suggest differences in location of pain drawings reflected by a minimal overlap in patient and clinician drawings based on the Jaccard index, but do not suggest differences in the area and shape of the pain drawings reflected by the Bland–Altman plots and bounding box respectively.

Three other studies have reported similarities in patient and clinician pain drawings; two with the Jaccard index [9, 31], and one by categorizing drawn pain areas [32] (see Table S1 for a Table of comparison). The Jaccard index measures the overlap between two drawings (Figure S1), and our study found a mean overlap in drawn areas ranging from 9–18%. To our knowledge, no clinical important difference for the Jaccard index exists, although an overlap between of less than 20% seems very minimal. Two studies that have used the Jaccard index to compare overlap between patient and clinician drawings, finding Jaccard indexes ranging from 19–22% (Table S1) [31]. In our study, lateral images had Jaccard index values of 15–18%, which seem in line with the above studies (Table S1). A minimal overlap could be explained by the clinician being more focussed on the treatment area (in our case the lateral hip), placing the drawing in a slightly different location, and by potentially dismissing other areas of pain that the patient focuses on, which is supported by the individual drawings shown in Figure S2. Jaccard indexes of the back and front full-body charts in our study were much lower (9 and 11%, respectively), which may be explained by the fact that different pain conditions were mapped, that the clinical examination was focused on the assessment of greater trochanteric pain syndrome, and by the timing of the drawings. Another study reported accuracy in pain drawings of 49% between full-body paper drawings of 36 chronic pain patients and their doctor, analysed as drawings being either “same” or “different” [32]. Differences in findings may have been influenced by many factors, including the type of patients (acute pain [9] vs. non-acute/persistent pain [31, 32]), electronic vs. paper drawings, experience in completing pain drawings, the ability to perceive and draw pain, the type of charts and the instructions given to patients and clinicians (Table S1). These factors highlight the importance for clinicians and patients to communicate differences and similarities in their pain drawings to ensure information is not lost or changed.

In our study, clinician drawings were, on average, smaller compared to patient drawings. We could speculate that this negative fixed bias could result from patients drawing their current areas of pain, whereas the clinician projected a potential diagnosis. In this study, the bias may have been exaggerated because the clinician made drawings during the clinical examination and the patient did theirs after the clinical examination. The patients plausibly had been cued into their areas of pain after the physical examination as the clinical examination was focussed on stressing (provoking pain) the associated structures involved in greater trochanteric pain syndrome as well as those likely to refer to the hip region. The differences in pain drawings may reflect mis-communication between the patient and clinician, although differences like negative fixed biases (e.g., smaller clinician drawings compared to patient drawings) are to be expected when people perform tasks on unfamiliar devices [33]. In this study, all patients and the clinician were instructed on how to complete the drawing but were unfamiliar with the device and/or application. One study revealed that fixed biases could be overcome by repeated use [33], and therefore future research should incorporate attempts to familiarise patients and clinicians with the device and technology before further investigating these biases.

The use of clinician and patient drawings may be a beneficial tool in clinical practice. Shaballout et al. [9] studied pain drawings in an acute pain setting. They had patients and clinicians independently draw the pain on charts and then discussed both drawings during the consultation. This study showed that doctors felt they had improved understanding of the patient’s pain and to a lesser extent this impacted on their clinical decisions, even though drawings showed fair to good similarity [9]. An earlier study by Cummings et al. [32] showed that patient drawings alone are insufficient in clinical decision making and that a shared-decision approach to communication between patient and clinician over the pain distributions is desirable [32]. Further, in clinical practice it is often the clinician – not the patient – that completes the pain drawings. Our findings of differences between patient and clinician pain drawings reinforces the notion that at the least, the patient should be making the drawing, and perhaps both the patient and clinician should be completing pain drawings as suggested by Shallabout et al. [7] and Cummings et al. [31]. This may facilitate conversation between clinician and patient and subsequently improve diagnosis and treatment. Further research should examine if allowing the patient and clinician to discuss their pain maps does indeed lead to better clinical decisions and improved care. This approach is supported in other aspects of health care that show detailed communication between patient and clinician about the clinical examination and associated tests (e.g., images, blood tests) will improve quality of care [34].

Conclusions

Our findings suggest that clinicians and patients will draw somewhat different pain distributions, with differences in location but not in the area or shape of the digital pain drawings. Perceived patient–clinician inconsistencies in the location of pain may influence decision-making and subsequent management, and therefore should be acknowledged and addressed when using (digital) pain drawings in clinical practice. The impact on the decision-making and management of patients remains to be determined.