Pain is a very complex and subjective phenomenon. It is regarded as a symptom of an underlying condition or as a condition of its own. For adequate medical treatment, however, it is compulsory to classify the reported pain. Hints to the correct pain diagnosis are given by a pain assessment looking at the intensity of the pain, its distribution and duration, as well as the quality of the pain. Despite many new technological advances, however, objectification of pain is still an unsolved problem [1]. Asking patients to draw their pain has been used for half a century to overcome the complexity of communicating a subjective sensation from patient to physician. This method has different names in the literature, the most common being pain drawing (PD). Different instruments have been used to obtain PDs, starting from pen-on-paper drawings [2] and recently developing toward electronic PDs collected on tablet personal computers (PCs) [3-7]. Several studies have been conducted to test the usability, reliability, and repeatability of PDs in chronic pain situations such as shoulder pain [8], knee pain [9], back pain [10], and neck pain [11] as well as in acute low back pain [12-14], whiplash disorder [15], or experimentally triggered pain [16]. Regardless of the method used, it has been proven that using PDs together with anamnesis and physical examination can aid the differential diagnosis in many pain situations [9,10].

In this study, we examined the potential role of electronic PDs collected on a tablet PC in the clinical assessment and treatment of inpatients in acute pain situations. Therefore, we evaluated if knowing the patients’ PDs improved the pain specialists’ understanding of their patients and influenced their clinical decision making. Furthermore, we sought to identify differences between PDs of patients and doctors and to find those specific characteristics of the drawings that had the largest impact on doctors’ understanding.

To this end, we collected electronic PDs from a sample of inpatients that received a consultation by a pain specialist from acute pain service (APS) because of severe pain (average score of 7.3, SD 2.0 on an 11-point numeric rating scale [NRS]). We then asked the pain specialists to draw their own impression of the patients’ pain after anamnesis and physical examination but without having seen the patients’ drawings. PDs of the patients were then revealed to the doctors, and they were asked to evaluate how much the additional information improved their understanding of the case and how much it influenced their clinical decision. Similarities and differences of patients’ and doctors’ PDs were assessed by statistical image analysis and visual inspection. Finally, exploratory analyses were conducted to identify specific factors that had the greatest impact on doctors’ understanding of the patients.

Our study population were inpatients from different departments of Hannover Medical School. All patients were in acute pain situations that required a consultation by a pain specialist, which is provided by members of the APS. APS members visit the patients and adjust their pain management individually according to the requirements of each patient’s situation. Eligible patients were adult (age ≥18 years in Germany) inpatients of Hannover Medical School with acute pain and the ability to give written informed consent. Furthermore, they had to be physically able to complete an electronic PD on a tablet PC. We recruited 69 patients (37 females) for participation in our study, and all of them prepared a PD. Due to a technical failure of the tablet PC, 1 drawing was lost. The treating pain specialist prepared a PD for 61 of the remaining patients, all of which were included in the analysis. Of these 61 patients with complete data, however, only 47 (24 females) were discussed and rated because of the absence of some of the treating pain specialists during the meeting of the APS in the afternoon (see below). Characteristics of the final study population can be found in Table 1.

The study was approved by the Ethics Committee of Hannover Medical School (#2987-2015) and was conducted in accordance with the Declaration of Helsinki. All patients gave written informed consent after information of the purpose of the study.

Data collection was divided into 3 steps: (1) the patient draws a PD shortly before the consultation, (2) the pain specialist draws a PD of the same patient directly after the consultation based on anamnesis and bodily examination results, and (3) the pain specialist rates the patient’s PD’s influence on his or her understanding of the patient as well as on clinical decision making.

Knowing patients’ PDs significantly improved the doctors’ understanding of their patients (average rating: 4.81, SD 2.60, P<.001) and to a lesser extent their clinical decision (average rating: 2.68, SD 1.18, P<.001). Results are shown in Figure 2.

Patients drew on average 1.25 (SD 0.53) pain symptoms, a number closely matched by the doctors’ who drew 1.34 (SD 0.64) symptoms. With 3.34 (SD 2.82) different pain descriptors, patients described their pain more detailed (P=.03) than the doctors, who used 2.43 (SD 1.30) descriptors. The average pain distribution drawn by patients and doctors and the frequencies of pain descriptors are shown in Figure 3 and Table 2, respectively. Visual comparison of the averaged PDs suggested a high similarity of doctors’ and patients’ drawings. Similarity analysis of the individual PDs revealed fair reproducibility for pain extent with an ICC of 0.565 (95% CI 0.459-0.655) but poor reproducibility for the number of pain clusters with an ICC of 0.368 (95% CI 0.238-0.485). The Jaccard index was 0.217 (SD 0.171). Detailed results for each body view are listed in Table 3. The poor reproducibility of the number of pain clusters also showed when we compared PD characteristics directly between the 2 groups (Table 4). Here, we found that patients drew significantly more pain clusters when comparing unthresholded (P<.001) and thresholded clusters (P=.01). Pain extent, average pain intensity, and the number of nonempty body views on the other hand showed no significant differences between patients and doctors.

We identified 5 patients who received either very low or very high ratings from the doctors, that is, patients in which seeing the PD was of very little or very high value for the doctors understanding. The group with high ratings is shown in Figure 4, the group with the low ratings in Multimedia Appendix 1. The individual clinical cases are discussed in Table 5.

Exploratory cross-correlation analysis revealed that the understanding of pain was influenced most strongly by 2 factors: the area of the pain as drawn by the patient (r=.454, P=.001) and the WPI (r=.447, P=.001) as calculated from the PD (see Figure 5). In both cases, the correlation was positive, which means that a larger pain area and higher WPI were associated with greater improvement in understanding of pain. When testing the same factors but looking at their absolute differences in doctors’ and patients’ drawings, area of pain showed the only significant correlation with understanding of pain (r=.313, P=.03), whereas WPI showed a tendency (r=.255, P=.08).

In this study, we have tested whether knowledge of the electronic PD of a patient in acute pain can improve the doctors’ understanding of the patient’s pain and potentially influence clinical decision making. We have furthermore sought to identify similarities and differences of electronic PDs made by patients and their treating pain specialists. Finally, we wanted to find specific characteristics of the drawings that have a large impact on doctors’ understanding.

Our results show that PDs significantly improved the doctors’ understanding of their patients’ pain based on their own judgment. On average, doctors rated this improvement with 4.81 out of 10 points. The impact on clinical decision was also significant but of smaller size (2.68 out of 10 points). There are 2 possible explanations for the fact that a relatively large improvement in understanding resulted in rather modest changes in clinical decision. First, in the majority of cases, additional pain clusters drawn by the patients do not lead the physicians to new diagnoses that would require additional medical investigation or intervention. Instead, most of these clusters reveal previously diagnosed chronic pain sources that are unrelated to the acute problem. To take one of our examples (patient number 2 from Figure 4 and Table 5), a patient requiring consultation by the APS for acute leg pain from peripheral arterial occlusive disease draws pain clusters in the leg region but also adds a large cluster in the arm. The latter stems from a previously diagnosed pain syndrome of the cervical spine. Although this additional knowledge gives a more complete image of the patient and, therefore, improves the doctor’s understanding, it has little impact on the clinical decision.

The similarity of PDs from patients and their treating doctors observed in our study was considerably lower compared with studies looking at test-retest reliability of PDs in chronic pain patients. We found only fair reproducibility for pain extent (ICC: 0.57) and poor reproducibility for the number of pain clusters (ICC: 0.37). In contrast, Barbero et al and Neubert et al, using PDs from different chronic pain populations, found ICCs from 0.92 to 0.97 for pain extent [3,17]. Results regarding the reproducibility of the number of pain clusters differed even more. Here, Neubert et al report an ICC of 0.70, which is almost twice as high as the value found in our study. The Jaccard index of 0.22 (as compared with 0.46-0.49 in the above-mentioned studies) indicates an average overlap of only 22% between the PD of a patient and the associated drawing of the doctor in our study. Of course, we are comparing apples with pears here, as repeated PDs by the same person will be much more similar than PDs based on information derived from verbal and nonverbal communication. A comparison like this, however, allows us to get an estimate for how much information is lost or changed in patient-doctor communication.

The above-mentioned issue raises the rather philosophical question, which of the drawings contains the correct information. On the one hand, only the patient is able to perceive the symptoms that are expressed in the PD. As several studies have shown, generating a PD is a highly reproducible and reliable process [15,16]. Thus, there are good reasons to argue that patients’ PDs constitute the ground truth in this comparison. However, other studies have based their analysis on the assumption that the doctors’ drawings contain the correct information. For example, Cummings et al compared patients’ PDs with drawings by their doctors and found that the usage of patients’ PDs to measure pain extent, number of pain clusters, and related symptoms may lead to inaccurate diagnoses [26]. It should also be noted that anamnesis and physical examination themselves can lead to more detailed PDs by the doctors as they may reveal additional symptoms omitted by the patient.

In our opinion, both patients’ and doctors’ PDs are valid drawings in their own right. When analyzing their contents, however, it must be acknowledged that they do not represent the same thing and that there may be systematic differences that will influence the results of the analysis. These differences will be discussed in the next section.

In our study, we did not find significant differences between patients’ and doctors’ PDs regarding pain extent, average pain intensity, and the number of body views used in for the PD. Together with our findings regarding the low similarity of doctors’ and patients’ drawings, this means that both groups drew about the same number of pixels with the same intensity but in different places on the body. We further found that doctors used significantly fewer pain clusters and symptom descriptors than patients when drawing their pain symptoms. There are several possible explanations for this. On the one hand, it is likely that doctors’ drawings exhibit less specificity of pain location than the patients’, simply because some level of information loss is to be expected for verbal communication of bodily symptoms. This can be seen in Multimedia Appendix 1, where clusters drawn by doctors are clearly overlapping with those of the patients but are generally larger and, thus, less specific. On the other hand, doctors may have focused more on single symptoms with larger extent. This could be explained by the principles of the APS [27]. In contrast to outpatient care settings, inpatients are well diagnosed and known to their treating medical team. Therefore, these patients are presented with a specific question to the specialists of the APS. Pre-existing and already treated pain diagnoses are not as much in the focus of interest as they would be in an outpatient setting.

Our innovative methodology of electronic PD analysis allowed us to extract a variety of information from the acquired PDs. This included characteristics such as the area of pain, average intensity, and WPI [25] as well as the number of clusters, pain sensations, and body views used in the drawing. Availability of this information enabled us to perform an exploratory analysis to identify those characteristics that significantly improved doctors’ understanding of their patients’ pain. We found that both pain area and WPI had the largest impact on doctors’ understanding. Thus, not only drawings with more pain area received higher ratings but also those where the pain was more widespread. Both effects can be observed when comparing patients with the highest ratings (Figure 4) with those with the lowest (Multimedia Appendix 1). It is evident that the latter show much smaller pain areas and less widespreadness than the former. However, pain area and WPI also showed a high level of correlation with each other, indicating similar information content.

When looking at differences between patients’ and doctors’ PDs, however, only the absolute differences in pain area improved the doctors’ understanding. Thus, an over- or underestimation of pain area by the doctor made seeing the patient’s PD valuable for the doctor. Among other things, this finding indicates that doctors do consider information from the patients’ PD as being “correct.”

Although our study was not aimed at comparing electronic PDs with their pen-on-paper counterparts, we would nevertheless like to emphasize the advantages of using the electronic version. In the last 10 years, several research groups have developed PD apps to be used on tablet computers [3-7,28]. A central advantage of the electronic drawings acquired this way is the possibility to analyze results right after completion of the drawings and without the need for time-consuming digitization. Although measuring pain area is possible for pain-on-paper drawings by using grid-based methods [13,29], such analyses usually take several minutes and require the doctor to sit down at a table with adequate lighting. Furthermore, the calculation of more complicated but relevant variables, such as average pain intensity or pain overlap as used in our paper, would take even longer to extract from pen-on-paper drawings. Finally, well-designed PD apps allow the user to zoom in and, thus, can be used by people with severe visual impairments for whom conventional drawings would be challenging or even impossible.

Although our study has reached the planned aims, there were some limitations that we could not avoid. First, our sample consisted of inpatients in acute and often severe pain. Thus, the accuracy of completing PDs may have been lower compared with, for example, chronic pain patients that have had some time to adapt to their pain. Of course, such lower accuracy will also influence all further analyses, for example, regarding similarity of patients’ and doctors’ PDs. Second, our procedure of rating the improvement in understanding and influence on clinical decision was suboptimal as only the treating doctor was allowed to give a rating and this rating was anonymous. This made it impossible to assess potential bias, for example, by certain doctors, giving only good or bad ratings. Furthermore, we did not ask for the explicit reasons why a PD was considered helpful or why its knowledge did or did not influence clinical decision. Third, our sample size of 47 patients was rather small and may have led to false-negative results in all analyses directly comparing patients with doctors. Finally, the fact that all doctors that rated the impact of the patients’ drawings on their understanding and clinical decision also prepared drawings themselves may be seen as a confound. Although we believe that the instructions for rating were unmistakably aimed at the impact of seeing the patients’ drawing (and not, eg, of comparing the patients with their own drawing), we cannot rule out the possibility that the act of drawing may have confounded the rating in some individuals.

Future studies should assess the importance of patients’ PDs independently from that of PDs made by the doctors, that is, compare doctors using the app with those not using it. Furthermore, it would be desirable to assess improved understanding of the patient by more objective means than self-report.

We have shown that in a clinical setting, electronic PD can improve doctors’ understanding of patients in acute pain situations based on their own judgment. The ability of electronic PDs to visualize differences between doctors’ and patients’ conception of pain has the potential to improve doctor-patient communication.