Virtual human technology: patient demographics and healthcare training factors in pain observation and treatment recommendations

Abstract

Background

Patients’ sex, race, and age have been found to affect others’ perception of their pain. However, the influence of these characteristics on treatment recommendations from laypersons and healthcare providers is understudied.

Design

To address this issue, 75 undergraduates and 107 healthcare trainees (HTs) used a web-based delivery system to view video clips of virtual human (VH) patients presenting with different standardized levels of pain. Subjects then rated the VHs’ pain intensity and recommended the amount of medical treatment the VHs should receive.

Results

Results indicated that, compared with undergraduates, HTs perceived African Americans and older adults as having less pain but were more willing to recommend medical treatment for these patients than were undergraduate participants. HTs and undergraduates rated female, African American, older, and high-pain-expressing adults as having greater pain intensity than male, Caucasian, younger, and lower-pain-expressing adults. Moreover, they also recommended that female, older, and high-pain-expressing adults receive more medical treatment than male, younger, and lower-pain-expressing adults.

Conclusions

This study found that the characteristics of the VHs and whether the participants were undergraduates or HTs influenced the ratings of pain assessment and treatment recommendations. The findings are consistent with the previous VH literature showing that VH characteristics are important cues in the perception and treatment of pain. However, this is the first study to identify differences in pain-related decisions between individuals who are pursuing healthcare careers and those who are not. Finally, not only does this study serve as further evidence for the validity and potential of VH technology but also it confirms prior research that has shown that biases regarding patient sex, race, and age can affect pain assessment and treatment.

Keywords:

• pain
• virtual human
• race
• age
• gender
• perception

Introduction

Pain is the number one reason why individuals seek medical attention, and it is the number one cause of disability.¹ However, healthcare professionals find pain particularly challenging to assess and treat, because pain is subjective.^2,3 Thus, healthcare professionals must rely primarily on patient self-reporting and medical-related variables.^4,5 Previous research has indicated that biases regarding sex, race, and age can affect pain assessment and treatment.^6–14

Virtual human (VH) technology is a novel way of investigating differences in pain assessment. Two studies have used this technology to examine whether participants assess and would treat the pain of VHs (who differ by sex, race, age, and pain expression) differently.^6,15 The advantage of using VH technology is that the facial features and pain expressions can be standardized without the biases of interest being present in the construction of the stimuli. VH technology may also increase the likelihood that a healthcare professional will report his/her perceptions and treatment opinions with less social desirability bias, because the patient is not present.

Hirsh et al found that sex, race, age, and the expression of pain were prominent predictors of pain intensity and the recommendation for medical treatment in a sample of 75 undergraduates.⁶ When the VHs were either women or high-pain-expressing adults, they were rated by participants as having higher pain intensity. The study also found that VHs who were female, older adults, and expressing higher pain were more likely to be recommended for medical treatment.

According to Hirsh et al, nurses who assessed the VH profiles indicated that they perceived female, African American, older, and higher-pain-expressing adults to be experiencing greater pain. Moreover, the nurses were more likely to suggest opioid treatment if the VH was female, African American, older, and expressing more pain.¹⁵

Although the previous work indicated that both laypersons and healthcare trainees (HTs) are influenced by patient demographic cues when making pain-related decisions, it is not clear whether the influences of these cues are the same or different in these two groups. This question is best addressed by comparing the decision policies of laypersons and HTs in the same statistical analysis. If there is a difference between the decisions of laypersons and HTs, that difference might suggest that healthcare training or self-selection as a healthcare professional is related to different rates of cue use or potential bias in pain observation.

If patient characteristics (sex, race, or age) influence healthcare professionals in assessing or treating a patient, it could adversely affect the patient’s outcome.^10,15–18 Healthcare professionals not only treat patients but also they serve as models and educators for future generations of healthcare professionals. Thus, the current study will examine whether healthcare trainees (physical therapy, nursing, medical, and dental students) use patient race, sex, and age cues differently than do undergraduate students when making judgments about pain. The results of such analyses may ultimately lead to improved patient care and education efforts aimed at reducing biases among providers and the general public.

Methods

Participants

Participants included 75 undergraduate students and 107 HTs from the University of Florida, Florida, USA. The undergraduate population consisted of 53 women and 22 men and included 62 Caucasians and 13 African Americans. The average age of the undergraduate students was 21.01 years, with an age range of 18–28 years. The HT group consisted of 34 physical therapy students, 30 nursing students, 25 medical students, and 18 dental students. The HT population was made up of 83 women and 24 men and included 74 Caucasians, 10 Hispanics, 13 Asians, 3 African Americans, 6 “others”, and 1 participant who did not identify his/her race. The average age of the sample was 24.62 years, with an age range of 19–48 years. All participants were compensated $15 for their participation.

Procedure

This study was approved by the University of Florida Institutional Review Board. The study used a web-based delivery model. After giving consent electronically, participants completed a demographic questionnaire and then observed the VH profiles. The participants viewed each video for 20 seconds. For each VH profile, participants read a clinical vignette either about a VH with chronic lower back pain (HTs) or a VH with abdominal pain (undergraduates) and viewed the vital signs of each VH video.

The VHs were created with the People Putty software program that has been used in previous studies.^6,15 The VHs had four personal characteristics that were systematically manipulated: sex (male, female), race (Caucasian, African American), age (young adult, older adult), and pain expression (low, high). The VHs expressed pain through facial expressions that were coded based on the Facial Action Coding System (FACS).¹⁹ The FACS is based on facial muscle movements and distinguishes 44 different action units (AUs). However, an abbreviated version of FACS was used in this study. The study focused on 4 AUs that represent the core features of pain expression (brow lowering, tightening of the orbital muscles surrounding the eye, nose wrinkling/upper lip raising, and eye closure).^6,15,20,21 Figure 1 is an example of one of the VH faces that was used in this study.

Figure 1 A young, Caucasian, female virtual human showing a high pain expression.

To control for order effects, VH videos were presented randomly. The participants were required to complete one VH profile before viewing the next one. Also, they were not permitted to revisit a completed VH video. The participants read a set of instructions that provided information on how to answer the pain assessment and treatment ratings using the visual analog scale (VAS). The participants rated each VH on two VASs on a 1–100 scale (anchored at “no pain” to “most intense pain sensation possible”). The first VAS was “Please rate the pain intensity that the patient is experiencing in the video.” The VAS range was from “no pain” to “most intense pain sensation possible.” The second VAS was “Likelihood of recommending medical treatment for the patient in the video.” The VAS range was from “not at all likely” to “completely likely.”

A total of 16 unique scenarios were created to represent all possible cue combinations. The undergraduate students viewed 16 VH profiles. The HTs observed 32 patient profiles where they observed each cue combination twice. In order for the data to be comparable, only the first patient profile of the pair was used; thus, only 16 VH profiles were included in the study. The undergraduate students and the HTs viewed the same 16 VH videos and profiles. The study took approximately 1–1.5 hours for the participants to complete. The participants were then debriefed regarding the concept of the study.

Statistical analyses

All data analyses were performed using SPSS for Windows (Version 17). Descriptive statistics were conducted to summarize the demographic and background characteristics of the sample. Because the two groups differed significantly by age (3.6 years; r = 0.459, P = 0.000), correlations between age and the key dependent variables were conducted. These correlations were used to determine whether assumptions of covariance analyses were met to include age as a covariate. Analysis of variance (ANOVA), under the general linear model, was performed to examine the group differences (undergraduate students vs HTs) in the rating of two dimensions of pain (pain intensity and the recommendation for medical help) as a function of the VHs’ personal characteristics (sex, race, age, and pain expression). Where appropriate, age was used as a covariate to control for group differences in age. This study consists of a secondary analysis of the data from two previous dissertations conducted by our laboratory; however, the aims of the current study are distinct, and the questions addressed were not previously investigated in the prior work.

Results

Correlation of age and VHs’ personal characteristics

A correlation was conducted to examine the association between pain intensity rating and participant age. The correlation was modest but statistically significant (r = −0.156, P < 0.05). Therefore, participant age was used as a covariate in the ANOVA model for pain intensity rating. A correlation was also conducted to examine the association between medical recommendation rating and participant age. Because the correlation was not significant (r = 0.005, P < 0.95), age was not used as a covariate in the ANOVA model for recommending medical help.

Pain intensity ratings

VH group

The results of the ANOVA on pain intensity ratings indicated a main effect for participant group. Collapsed across the 4 VH cues (sex, race, age, and pain expression), undergraduate student participants gave significantly higher pain intensity ratings to VH patients than did HT participants (F[1180] = 4.81, P < 0.05, partial η² = 0.03).

VH sex

A main effect for VH sex also emerged, with female VHs perceived as experiencing more pain than male VHs (F[1180] = 22.35, P < 0.001, partial η² = 0.11). The interaction of patient sex and participant group was not significant for ratings of pain intensity (F = 0.004, P > 0.05, partial η² = 0.000).

VH race

A significant main effect for patient race was also found, with African American VHs rated as experiencing more pain intensity than the Caucasian VHs (F[1180] = 9.36, P < 0.01, partial η² = 0.05). The results also identified a significant race by group interaction (F[1180] = 13.11, P < 0.001, partial η² = 0.07). Specifically, compared with HTs, undergraduate participants gave higher pain intensity ratings to African American VHs than to Caucasian VHs.

VH age

Similarly, a main effect for patient age also emerged. The pain intensity of older VHs was rated significantly higher than that of younger VHs (F[1180] = 36.53, P < 0.001, partial η² = 0.17). A significant age by group interaction was also found. Compared with HTs, the undergraduate participants rated the pain of older VHs as significantly more intense than that of younger VHs (F[1180] = 21.60, P < 0.001, partial η² = 0.11).

VH pain expression

We also found a main effect for VH pain expression. As expected, VHs with a high pain expression were rated as having higher pain intensity than those with a low pain expression (F[1180] = 519.95, P < 0.001, partial η² = 0.74). The expression by group interaction was not significant (F = 0.05, P > 0.05, partial η² = 0.00).

Age covariate

As noted above, participant age was significantly, albeit modestly, correlated with pain intensity ratings. The results of the ANCOVA were essentially the same as those above and indicated that age was not a significant factor in the model. For this reason, we have not included the ANCOVA results in the manuscript and only consider the results presented in the discussion section.

Healthcare recommendations

Participant group

The results of the ANOVA on recommendations for medical help ratings indicated no main effects for participant groups. Collapsed across the 4 VH cues (sex, race, age, and pain expression), there were no group differences (F[1180] = 1.26, P < 0.05, partial η² = 0.01).

VH sex

There was a main effect of VH sex in the recommendation for medical help. Female VHs received significantly higher recommendation ratings than male VHs (F[1180] = 2.32, P > 0.05, partial η² = 0.01). The interaction of patient sex and participant group was not significant for ratings of recommending medical help (F = 2.01, P > 0.05, partial η² = 0.01).

VH race

There was no main effect of VH race in the recommendation for medical help (F[1180] = 6.277, P < 0.01, partial η² = 0.03). There was a significant interaction between VH race and participant group. A larger race effect was seen for HTs than for undergraduates (F[1180] = 6.277, P < 0.01, partial η² = 0.03], such that HTs more frequently recommended the African American VHs for more medical help than the Caucasian VHs.

VH age

Results also indicate a main effect for VH age in participants’ recommendations for medical help (F[1180] = 38.92, P < 0.001, partial η² = 0.18]; older VHs received significantly higher recommendation ratings than did younger VHs. A significant interaction between age and group was also indicated (F[1180] = 9.03, P < 0.01, partial η² = 0.05). HTs more often recommended medical treatment for the older VHs than did undergraduate participants.

VH pain expression

Finally, there was a significant main effect for the VH pain expression, such that VHs expressing a high level of pain were more often recommended for medical treatment than VHs with a low pain expression (F[1180] = 357.43, P < 0.001, partial η² = 0.67). The expression by group interaction was not significant (F = 9.03, P > 0.05, partial η² = 0.05).

Detailed results of the analyses discussed here are presented in Table 1.

Table 1 Means (M) and standard deviations (SD) of pain intensity and recommendation for medical help

	Pain intensity				Recommend medical help
	Undergraduate students		Healthcare trainees		Undergraduate students		Healthcare trainees
	N = 75		N = 107		N = 75		N = 107
	M	SD	M	SD	M	SD	M	SD
Sex
Men	38.65	14.53	33.74	15.78	44.06	21.16	48.79	25.00
Women	41.14	13.87	36.16	16.46	47.15	20.95	50.13	25.18
Race
Caucasian	40.06	13.97	32.97	15.51	45.92	21.18	48.15	25.27
African American	39.73	14.46	36.93	17.07	45.29	20.72	50.76	24.15
Age
Young	39.42	13.66	31.32	15.00	44.35	21.18	45.89	25.03
Old	40.37	14.91	38.59	17.89	46.85	20.94	53.03	24.74
Pain intensity cue
Low pain intensity	25.94	14.80	21.26	14.65	31.94	23.99	37.91	26.65
High pain intensity	53.86	16.99	48.64	20.31	59.26	20.91	61.00	24.90

Discussion

Using VH technology, this study examined the question of whether undergraduate students and HTs assess pain and suggest treatment differently. The overall results of this study demonstrate the ability of web-based VH scenarios to elicit sex, race, age, and pain expression influences on decisions about pain assessment and recommendations for treatment. In addition, the approach was sensitive to group differences between undergraduate students and HTs. Further, this study is important because it suggests the hypothesis that healthcare training, or self-selection as a healthcare professional, is related to different rates of cue use and potential bias in pain observation.

The current study found that the VHs’ characteristics and education status (undergraduate vs HT) influenced the ratings of pain intensity and recommendations for medical help. Interestingly, compared with HTs, undergraduates rated both African American and older VHs as having higher pain. However, their recommendations for treatment ratings were lower for these same VHs. Also of note is that undergraduates consistently rated the VHs’ pain higher than did the HTs. However, the HTs consistently recommended more treatment for the VHs. This suggests that even though HTs might perceive their patients as having less pain, they nevertheless are more willing to recommend medical treatment for them. This could reflect a selection bias in terms of who pursues healthcare as a profession. Undergraduates might be particularly sensitive to the pain of others because they are exposed to it less often and thus err on the side of caution. However, HTs might be more inured to the pain of others because they frequently encounter patients in pain during their professional training.²² Some research has shown that medical students do not accurately perceive what patients believe about their own health, including about their pain.²³ Even though undergraduates might be particularly responsive to pain in others, they might not feel that it is their responsibility or that they have the expertise to make recommendations about treatment. HTs, on the other hand, might feel more comfortable with giving such recommendations. Our results could also reflect HTs having more education than undergraduates on the best practices for pain management.

The validity of computer-generated pain expression was supported in this study. The participants were able to distinguish the level of pain expressed by the VH (low or high) when evaluating the VHs’ pain. If the pain level expressed by the VH videos had been too subtle, the inconsistencies in participants’ pain intensity ratings across the two levels of pain expression would have yielded nonsignificant results. However, participants in the study consistently identified the VH videos intended to express high pain as having higher pain intensity and the VH videos intended to express low pain as having lower pain intensity. The HTs in the study also consistently reported that the vignettes and the VHs reflected accurate perceptions of what they see while they are working with patients. This study has interesting implications for public health. Although our effect sizes for group differences were modest, the use of age, sex, and race cues could still have a big impact on healthcare. Healthcare professionals typically see hundreds or thousands of patients throughout their careers. If the use of these cues reflects a bias toward one demographic group or another, patient outcomes could be adversely affected. Also, healthcare professionals frequently serve as mentors to HTs and may communicate their biases to colleagues as well. Such informal ‘learning’ experiences may be particularly influential given that HTs receive limited pain treatment education.^22,23 Indeed, previous research has found that healthcare providers prescribe less pain medication to women, African Americans, and older patients. Because healthcare professionals see so many patients and share their information with those they are mentoring, potential biases in prescribing medication can affect a large number of patients.^9–11,17 These results suggest that health educators should increase their trainees’ awareness of the differences in pain reporting and perception that can affect the understanding of a patient’s pain experience.

Several study limitations should be considered. The participants were not asked an open-ended question as to what treatment they would suggest for the VHs, nor were they given the option of gathering additional information before recommending treatment. Differences could have emerged in the type of treatment recommended to VHs of different demographic characteristics. Also, the undergraduates and the HTs read slightly different vignettes. The vignettes could have influenced the participants’ pain ratings. Further, the participants who took part in the study were a relatively homogeneous population (young and educated). In addition, it is possible that participants were able to determine the intent of the study, and thus responded in a socially desirable manner. Finally, the representativeness of the VH videos and the scenarios presented has to be considered. However, in our pilot work, over 70% of participants indicated that the VH facial expressions were realistic, and 90% indicated that the clinical scenarios were reflective of real post-operative scenarios.¹⁵

Future research is warranted to examine the causal relationships between group membership (HTs vs undergraduates) and cue use. First, we know that there is an age difference between the HTs and the undergraduate participants. Preliminary evidence after examining the results of the 2 × 2 mixed ANCOVA for the pain rating pain intensity suggests that the age of the evaluator influences the perception of an individual’s pain. However, more in-depth research should be conducted to determine how, and to what extent, age affects the participants’ perception of pain. Compelling hypotheses for the group differences include self-selection into a helping profession or the direct effects of training in healthcare. It is also possible that the demographic characteristics of the assessor interact with those of the patient to influence ratings. Future studies could be designed that investigate the role of these and other potential factors that might account for the observed differences in pain assessment and treatment ratings. Such studies might include longitudinal designs following first year students to practice, or studies might include age-matched controls and healthcare providers.

In summary, this study found that both the characteristics of the VH and the type of participants influenced ratings of pain assessment and treatment recommendations. The findings are consistent with the previous VH literature showing that VH characteristics are important cues. However, this is the first study to identify differences in pain-related decisions between individuals who are pursuing healthcare careers and those who are not. Finally, not only does this study serve as further evidence for the validity and potential of VH technology but also it confirms prior research that has shown that biases regarding patient sex, race, and age can affect pain assessment and treatment.

Acknowledgments

We would like to thank the National Institute of Dental and Craniofacial Research (R01DE013208), National Institutes of Health, National Institute of Neurological Disorders and Stroke (F31 NS049675), National Institute of Child Health and Human Development, and National Center for Rehabilitation Research (T32 HD007424) for funding our research.

Disclosure

No conflicts of interest were declared in relation to this paper.