Clinical Interpretation of Self-Reported Pain Scores in Children with Acute Pain

Abstract

Objectives:

To identify self-reported pain scores that best represent categories of no pain, mild, moderate, and severe pain in children, and a pain score that accurately represents a child’s perceived need for medication (PNM), i.e. a minimum pain score at which a child would want an analgesic.

Study design:

Prospective cross-sectional cohort study of children aged 6-17 years presenting to a pediatric emergency department with painful and non-painful conditions. Pain was measured using the Verbal Numerical Rating Scale. Receiver operating characteristic-based methodology was used to determine pain scores that best differentiated no pain from mild, mild from moderate, and moderate from severe pain. Descriptive statistics were used to determine the PNM.

Results:

We analyzed data from 548 children (51.3% female, 61.9% painful conditions). The scores that best represent categories of pain intensity are: 0-1 (no pain), 2-5 (mild), 6-7 (moderate), and 8-10 (severe) out of 10. The area under the curve for the cut points differentiating each category ranged from 0.76 to 0.88. The median pain score representing PNM was 6 (IQR 4, 7; range 0-10) out of 10.

Conclusions:

We identified population-level self-reported pain scores in children associated with categories of pain intensity that differ from those conventionally used. Implementing our findings may provide a more accurate representation of the clinical meaning of pain scores and reduce selection bias in research. Our findings do not support the use of pain scores in isolation for clinical decision making or use of a pain score threshold to represent a child’s PNM.

The clinical interpretation of self-reported pain scores is an important consideration for the treatment and management of acute pain in children. Decisions are often made based on whether the clinician interprets an endorsed pain score to represent mild, moderate or severe pain, or whether a pain score represents a desire or need for an analgesic.[1–8] Some guidelines recommend acetaminophen or ibuprofen for mild pain, and oral or parenteral opioids for moderate or severe pain.[1,2,6,7] Others recommend pharmacologic analgesics in certain contexts only if a patient has a pain score of greater than a predetermined threshold (e.g. pain score greater than 3/10).[3] However, the practice of using pain scores to guide analgesic decisions has become controversial and, in some cases, discouraged.[9–11]

There are consequential gaps in the current understanding of the clinical meaning of pain scores self-reported by children with acute pain. First, it is unclear if pain scores currently considered to represent categories of pain intensity (ie, mild, moderate, and severe) are accurate. For example, when using a pain scale scored from 0 to 10, pain scores of 0, 1-3, 4-6, and 7-10 are commonly associated with the categories of no pain, mild, moderate, and severe pain, respectively.[2,12–17] These ranges of pain scores were not systematically derived, but rather arbitrarily assigned and subsequently reinforced and perpetuated by convention. Second, it is unclear whether there is a consistent and generalizable pain score threshold representing a child’s desire for an analgesic or self-perceived need for medication to relieve their pain. The existence of such a threshold is questionable given the variability in perception of pain between individuals.[18–23]

The consequences of misinterpreting the clinical meaning of pain scales in children include the under-treatment of pain, misinformed analgesic selection, and introducing selection bias in the study of treating acute pain in children. Therefore, the primary aim of our study was to identify the self-reported pain scores that best represent children’s perceptions of no pain, mild, moderate and severe pain. Our secondary aim was to identify a pain score threshold that accurately represents a perceived need for medication (PNM) in this population. Our exploratory aim was to identify these scores within subgroups based on age and sex.

METHODS

We conducted an observational cross-sectional study in a pediatric emergency department (ED) with an annual census of approximately 55,000 visits. This current study was a planned secondary analysis of a parent study evaluating the validity and reliability of the Verbal Numerical Rating Scale (VNRS) in children with acute pain, and similar in methodology to our prior work evaluating similar outcomes for the Faces Pain Scale – Revised and the Color Analog Scale.[21,24] Our institutional review board approved the study, and verbal informed consent was obtained.

We enrolled a convenience sample of children aged 4 to 17 years from April 2014 to March 2016. These children had either a painful or non-painful condition as identified by the triage nurse and confirmed by the study team by asking children themselves whether they had “any pain” or “any hurt”. Exclusion criteria included developmental delay or neurologic impairment; intoxication; altered mental status; a pre-existing medical condition necessitating multiple painful procedures (e.g., malignancy) or chronic disease associated with pain (e.g., sickle cell disease); or if they did not speak English or Spanish. For this current study, we analyzed only children aged 6 to 17 years because we demonstrated strong validity and reliability of the VNRS in this age group in the parent study.[24] We did not analyze patients who gave a non-integer response or were identified as not understanding the VNRS (i.e., if they did not respond, responded with a nonnumeric response, or responded with a number outside the 0 to 10 range, when asked twice).

Patients were enrolled according to study team availability, which was primarily from 9 am to midnight on weekdays. To prevent bias related to enrolling only children who did (or did not) appear to have a strong understanding of the VNRS, we completed enrollment for every patient who was approached, identified as being eligible, and consented to participate in the study.

Study Protocol

In accordance with our prior work, we employed similar methodology to achieve the aims of this current study.[21,24] Trained study team members prospectively collected and recorded all data in a standardized fashion, using the same procedures and data collection forms for all patients.

The assessment of pain was performed in the child’s primary language. Children were first asked to choose a qualitative descriptor of their pain: “No pain”, “a little bit of pain”, “a lot of pain”, or “somewhere in between”. We used these phrases as proxies for the categories of no pain, mild, severe, and moderate pain, respectively.[21,25,26] Children were then asked to indicate their level of pain using the VNRS, which was administered by asking, “On a scale from zero to 10, where zero means no pain and 10 means the most or worst pain, how much pain do you have right now?” The interaction was verbal, using no materials or equipment. The word “hurt” was used interchangeably with “pain” during all assessments, depending on what seemed most understandable for each child.

To identify a pain score representing a child’s perceived need for medication (PNM) (i.e., the minimum score at which they would want an analgesic), children were asked a series of questions. First, they were asked, “If you had 10/10 pain, would you want medicine to make your pain less?” If they answered yes, the question would be repeated, but using the next pain score in descending order (e.g., 10/10 would be followed by 9/10) until the child answered no three times. The last pain score to which they responded yes would be considered the pain score representing the child’s PNM. Because this pain score best represents the minimum pain score at which a child would want an analgesic, scores above a child’s PNM also represent pain scores at which a child would want an analgesic. For example, if a pain score of 6 out of 10 best represents a child’s PNM, scores above 6 out of 10 would also represent pain scores at which a child would want an analgesic. For all interactions, the word “smaller” was used interchangeably with “less”, depending on what seemed most understandable for each child. A child’s response to the PNM assessment was considered to be non-logical if they did not respond when questioned; if they answered with anything other than a positive or negative response (e.g., “yes” or “no”); or if they gave a positive response after having already answered with at least one negative response.

To identify these scores in subgroups based on age and sex, we performed these analyses within subgroups of younger (6-7 years old) and older (8-17 years old), and subgroups of female and male. The age categories used were chosen to stratify younger children into a subgroup that has demonstrated more variability in their accuracy and responses when describing pain.[27,28]

Statistical Analyses

Frequencies and proportions were used to describe patient demographics and reported pain scores. We employed receiver operating characteristic (ROC)-based methodology to identify cut points that best defined the range of pain scores associated with each category of pain intensity.[21] We compared each pair of adjacent categories of pain intensity (e.g., no pain and mild pain; mild and moderate pain; moderate and severe pain) to identify a cut point that yielded the optimal balance (i.e. smallest absolute difference) between sensitivity and specificity. In order to evaluate the overall performance of each cut point, we reported its sensitivity, specificity, area under the curve (AUC), and associated 95% confidence intervals. Because cut points determined in this way suffer from random variability, we used bootstrapping to quantify the variability of these cut points and took this into account when comparing groups. The cut points identified for each pair of adjacent categories were used to define the range of pain scores associated with each category of pain intensity. To identify the PNM, we calculated the median pain score and associated interquartile range across all children to represent the PNM. Patients who gave a non-logical response during the PNM assessment were not included in the analysis.

The sample size was based on the parent study for which this study was a planned secondary analysis. Statistical analyses were conducted using SPSS (version 26, IBM Corp.), R (version 4.0.3, R Foundation for Statistical Computing) and the R cutpoint package.[29]

RESULTS

Patient Characteristics

We enrolled 760 children for the parent study and data from 548 were analyzed for the current study; 200 children were removed because they were 4 or 5 years of age; 6 were removed because they gave non-integer responses when their pain was assessed using the VNRS; and 6 were removed because they did not understand the VNRS. Table I shows the characteristics of the patients analyzed.

Table 1:

Patient characteristics

Characteristic	Total Sample No. (%) n = 548	Painful Conditions No. (%) n = 339	Nonpainful conditions No. (%) n = 209
Sex
Female	281 (51.3)	171 (50.4)	110 (52.6)
Male	267 (48.7)	168 (49.6)	99 (47.4)
Age, years
6	95 (17.3)	46 (13.6)	49 (23.4)
7	99 (18.1)	49 (14.5)	50 (23.9)
8	30 (5.5)	17 (5)	13 (6.2)
9	42 (7.7)	27 (8)	15 (7.2)
10	32 (5.8)	17 (5)	15 (7.2)
11	41 (7.5)	29 (8.5)	12 (5.7)
12	40 (7.3)	29 (8.5)	11 (5.3)
13	35 (6.4)	24 (7.1)	11 (5.3)
14	40 (7.3)	32 (9.4)	8 (3.8)
15	38 (6.9)	28 (8.3)	10 (4.8)
16	33 (6)	24 (7.1)	9 (4.3)
17	23 (4.2)	17 (5)	6 (2.9)
Race/ethnicity
Hispanic	438 (79.9)	268 (79.1)	170 (81.3)
Black	70 (12.8)	44 (13)	26 (12.4)
White	26 (4.7)	19 (5.6)	7 (3.3)
Other1	14 (2.6)	8 (2.4)	6 (3)
Primary language
English	494 (90.1)	309 (91.2)	185 (88.5)
Spanish	54 (9.9)	30 (8.8)	24 (11.5)
Painful condition
Soft tissue injury		90 (26.5)
Abdominal pain		80 (23.6)
Headache		42 (12.4)
Ear/throat pain		41 (12.1)
Fracture		20 (6)
Back pain		14 (4.1)
Chest pain		12 (3.5)
Abscess		10 (2.9)
Laceration		1 (0.3)
Other		29 (8.6)
Nonpainful condition
Rash			26 (12.4)
Cough			25 (12)
Fever			23 (11)
Vomiting and/or diarrhea			20 (9.6)
Shortness of breath, wheezing			20 (9.6)
Soft tissue injury, nonpainful			15 (7.1)
Dizziness and/or syncope			6 (2.9)
Other			74 (35.4)

¹American Indian/Alaska Native; Asian; Native Hawaiian/other Pacific Islander; more than one race.

Main Results

The distribution of pain scores reported by children associated with each category of pain intensity is shown in Figure 1. The median pain scores (and interquartile ranges) associated with no pain, mild, moderate, and severe pain were 0 (0, 1), 4 (2,5), 6 (5, 7) and 9 (8, 10), respectively (Figure 2; available at www.jpeds.com). Sixty (28.7%) of the 209 children who endorsed no pain reported a non-zero pain score, with 19 (9.1%) children reporting a pain score of 1 out of 10 and 17 (8.1%) reporting a pain score of 2 out of 10. Table 2 presents the optimal cut points identified for each pair of adjacent categories of pain intensity for the VNRS, and the sensitivity, specificity, and AUC associated with each cut point. Based on these cut points, the ranges of pain scores that best represent each category of pain intensity were 0-1 for no pain; 2-5 for mild; 6-7 for moderate; and 8-10 for severe pain.

Figure 1.

Pain scores associated with categories of pain intensity. The categories of pain severity represented in each graph (from top to bottom) are no pain, mild, moderate, and severe pain.

Figure 2.

Median pain scores associated with categories of pain intensity. The horizontal line within each box represents the median pain score. The outer margins of the box represent the interquartile range. The whiskers represent the range of pain scores.

Table 2:

Optimal cut points for differentiating categories of pain intensity in children

Categories of pain intensity being differentiated	n	Cut point1 (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)	AUC (95% CI)
No pain, mild	209, 92	2 (2, 2)	0.80 (0.72, 0.88)	0.80 (0.75, 0.86)	0.88 (0.83, 0.91)
Mild, moderate	92, 115	6 (5, 6)	0.60 (0.55, 0.83)	0.82 (0.52, 0.86)	0.76 (0.69, 0.82)
Moderate, severe	115, 132	8 (8, 8)	0.80 (0.73, 0.87)	0.84 (0.78, 0.91)	0.87 (0.82, 0.91)

¹The cut point identified is the pain score that best differentiates (i.e. has the highest AUC) between adjacent categories of pain intensity (i.e. no pain and mild pain, mild and moderate pain, moderate and severe pain.

Figure 3 shows the distribution of scores representing PNM endorsed by children with acute pain. There were 67 (12.2%) patients who gave a non-logical response during the PNM assessment. Of these patients, 38 (56.7%) were 6 years old, 25 (37.3%) were 7 years old, 2 (3%) were 10 years old and 2 (3%) were 12 years old. Based on the remaining 481 (87.8%) children, the median pain score (and interquartile range) associated with a child’s PNM was 6 (4, 7) out of 10.

Figure 3.

Self-reported pain scores associated with perceived need for medication.

The optimal cut points differentiating each category of pain intensity in subgroups of younger and older children resulted in pain score ranges of 0-1, 2-4, 5-6, and 7-10, and 0-1, 2-5, 6-7, and 8-10, respectively (Table 3; available at www.jpeds.com). Similarly, pain score ranges in subgroups of female and male children were 0-1, 2-5, 6-7, and 8-10, and 0-1, 2-4, 5-6, and 7-10, respectively (Table 4; available at www.jpeds.com). The PNM in younger and older children was 5 (IQR 3, 7) and 6 (IQR 4, 7), respectively. The PNM was 6 (IQR 4, 7) in both female and male children.

Table 3:

Optimal cut points for differentiating categories of pain intensity in children based on age.

	Categories of pain intensity being differentiated	n	Cut point1 (95% CI)	Sensitivity	Specificity	AUC
6 and 7 year olds (n=194)	No pain, mild	99, 40	2 (1, 2)	0.78 (0.71, 0.98)	0.86 (0.78, 0.91)	0.88 (0.82, 0.94)
	Mild, moderate	40, 22	5 (4, 5)	0.59 (0.40, 0.78)	0.50 (0.34, 0.68)	0.59 (0.45, 0.73)
	Moderate, severe	22, 33	7 (6, 7)	0.73 (0.63, 0.91)	0.82 (0.64, 0.92)	0.85 (0.74, 0.95)
8 to 17 year olds (n=354)	No pain, mild	110, 52	2 (2, 3)	0.83 (0.69, 0.90)	0.75 (0.68, 0.88)	0.87 (0.81, 0.92)
	Mild, moderate	52, 93	6 (5, 6)	0.68 (0.60, 0.87)	0.83 (0.61, 0.89)	0.81 (0.73, 0.88)
	Moderate, severe	93, 99	8 (8, 8)	0.87 (0.80, 0.93)	0.82 (0.74, 0.89)	0.90 (0.85, 0.93)

¹The cut point identified is the pain score that best differentiates (i.e. has the highest AUC) between adjacent categories of pain intensity (i.e. no pain and mild pain, mild and moderate pain, moderate and severe pain.

Table 4:

Optimal cut points for differentiating categories of pain intensity in children based on sex.

	Categories of pain intensity being differentiated	n	Cut point1 (95% CI)	Sensitivity	Specificity	AUC
Female (n=281)	No pain, mild	110, 46	2 (2, 3)	0.80 (0.68, 0.89)	0.79 (0.72, 0.88)	0.87 (0.82, 0.92)
	Mild, moderate	46, 62	6 (5, 6)	0.65 (0.54, 0.82)	0.76 (0.57, 0.85)	0.75 (0.65, 0.84)
	Moderate, severe	62, 63	8 (8, 8)	0.87 (0.80, 0.93)	0.82 (0.74, 0.89)	0.90 (0.85, 0.93)
Male (n=267)	No pain, mild	99, 46	2 (1, 3)	0.80 (0.69, 0.92)	0.82 (0.74, 0.89)	0.88 (0.81, 0.93)
	Mild, moderate	46, 53	5 (5, 6)	0.75 (0.57, 0.84)	0.59 (0.46, 0.88)	0.76 (0.67, 0.85)
	Moderate, severe	53, 69	7 (6, 7)	0.73 (0.63, 0.91)	0.82 (0.64, 0.92)	0.85 (0.74, 0.95)

¹The cut point identified is the pain score that best differentiates (i.e. has the highest AUC) between adjacent categories of pain intensity (i.e. no pain and mild pain, mild and moderate pain, moderate and severe pain.

DISCUSSION

We have systematically identified pain scores self-reported by children with acute pain representing child-perceived categories of no pain, mild, moderate, and severe pain intensity that differ from conventionally-accepted ranges.[2,12–17] We identified a wider range of pain scores representing no and mild pain (i.e. 0-5 out of 10) and higher pain scores representing severe pain (i.e. 8-10 out of 10). These findings have implications for both the clinical management and research of acute pain in children.

Our findings support adopting the pain scores associated with each child-perceived category of pain intensity determined in our study as a more accurate representation of the clinical meaning of pain scores self-reported by children. However, our findings do not support the use of pain scores in isolation for the management of acute pain in children at the level of the individual. A child’s self-reported pain intensity is a personal and individualized expression that is shaped by a child’s own perceptions, disposition, and prior experiences. As demonstrated in our study, a single pain score can mean completely different things to different patients. For example, we observed that pain scores of 3, 5, 7, 9 and 10 out of 10 have all been reported by different children to represent no pain, mild, moderate, or severe pain on at least one occasion. Our findings demonstrate a large degree of variability in how pain is perceived and reported by each individual child with a wide range of scores representing each category of pain intensity observed. Similarly, our findings corroborate previous observations in both adult and pediatric populations that a single pain score does not accurately represent a patient’s desire for an analgesic, and that there is a large variability in pain scores associated with when a patient requests or declines an analgesic.[18–20,30,31] In this current study, every pain score from 0 to 10 was endorsed by a child to represent a PNM, with pain scores of 4, 5, 6, 7, and 8 out of 10 each representing a PNM for >10% of children evaluated (Figure 3). Therefore, our findings do not support the use of pain scores in isolation for clinical decision making or the use of a single pain score to represent a child’s perceived need for medication at the level of the individual. A more accurate and patient-centered approach than using pain scores in isolation or a pain score threshold may be a multifaceted assessment that also includes asking a child if they desire an analgesic to treat their pain, taking into consideration pain quality, a child’s comfort-function goals, and incorporating any related concerns or priorities.[9,10,18,22,32–34]

The research implications of our findings include using the identified thresholds to minimize mis-categorizing study participants based on an endorsed pain score. Studies evaluating analgesics will often use a minimum pain score as an inclusion criterion in order to help identify a desired study population. For example, a minimum pain score of 4 out of 10 may be selected if a study aims to enroll participants with moderate to severe pain.[35–39] However, using a minimum pain score based on conventional ranges of pain scores may result in the enrollment of participants who are not experiencing the expected degree of pain intensity. Using the pain thresholds that we identified as being associated with categories of pain intensity would address this type of selection bias and make research results more generalizable to a desired population.

Although we have identified limitations of pain scores as a unidimensional measure when used in isolation, pain scores can still maintain a meaningful role at the population-level in clinical practice and research. First, pain scores can serve as a familiar metric that can be incorporated as part of a multidimensional assessment of a child’s pain, as they are readily understood and communicated between clinicians when providing care. Second, they can serve as a screening tool to provide an initial impression of pain intensity being experienced by a child that precedes a more thorough and multifaceted assessment of the child’s pain.[40] Third, pain scores can be used to trend changes in pain intensity and serve as one component of evaluating the response to analgesics for an individual patient. Finally, pain scores can continue to serve as a measure of pain intensity that can be useful in research, particularly when clinically meaningful changes in pain have been determined for a specific pain scale in a defined population.[41,42]

We found differences associated with age in the study. First, the optimal cut points differentiating mild from moderate pain, and moderate from severe pain, were lower in younger children. This finding is different from our prior study of the Faces Pain Scale – Revised in which there was no difference in cut points differentiating categories of pain intensity between subgroups based on age.[21] This may be related to development in cognitive abilities such as seriation and quantity estimation, which are not necessary to use the Faces Pain Scale - Revised.[27,43–45] Therefore, it may be prudent to use the cut points determined for the overall population in younger children as well, as mis-categorizing pain intensity using the lower cut points could be of greater consequence. Second, almost all of the 67 patients who gave non-logical responses during the PNM assessment were in the younger age group (6-7 years), and only 6% were older than 7 years. This may also be related to the aforementioned issues with seriation and quantity estimation, as well as the development of hypothetical reasoning required to navigate the PNM assessment.[46]

There are limitations to our study. First, we evaluated only children with acute pain. Our findings may not be generalizable to children with chronic pain, who are known to perceive pain differently.[47–49] Second, the use of the qualitative phrases, “a little bit”, “somewhere in between”, and “a lot of pain” to represent mild, moderate, and severe categories of pain intensity may have introduced imprecision into our results. However, these are the same phrases used in prior studies that aimed to identify pain scores that best represent categories of pain intensity.[21,25,26] In addition, these descriptive phrases were probably more understandable to younger children than the words “mild”, “moderate”, and “severe”. Third, a large proportion of our study population self-identified as Hispanic. Prior studies describe differences in pain perception based on ethnicity, although our prior work evaluating the clinical interpretation and validity and reliability of self-reported pain scales in children did not demonstrate any significant differences between subgroups based on ethnicity.[24,42,50–53] Finally, our study was conducted at a single center, which may limit the generalizability of our findings. However, our study did include a comprehensive spectrum of pediatric ages, a similar proportion of both male and female children, and a varied and representative cohort of both painful and non-painful conditions in otherwise healthy children.

We have systematically identified pain scores associated with categories of pain intensity that differ from those that are conventionally used. The potential benefits of implementing our findings at the population-level include the use of a more accurate range of pain scores representing each category of pain intensity that can be incorporated as part of a multifaceted assessment, and reducing selection bias resulting from mis-categorizing study participants based on their endorsed pain score. Our findings do not support the use of pain scores in isolation for clinical decision making or use of a pain score threshold as an accurate method to determine an individual child’s perceived need for medication.