The visual time analog scale: simple, novel measurement approach to assess pain in patients with chronic pain

Abstract

Aim:

We conducted a preliminary evaluation of a newly developed, time-based visual time analog (VITA) scale for measuring pain in chronic pain patients.

Materials & methods:

40 patients with chronic back pain rated their pain over four visits using numerical (pain) rating scale (NRS) and VITA assessing pain intensity by distributing the amount of time spent on ‘not aware of pain’ (blue), ‘aware of nothing but pain’ (red) and time in between (yellow).

Results:

The NRS scores were correlated with the VITA Red but not with VITA Blue. The psychometric analyses revealed that VITA achieved greater reliability and sensitivity than did NRS.

Conclusion:

The results provide preliminary support for VITA scale for assessing pain intensity in patients with chronic pain.

Accurate and reliable pain measurement is a cornerstone of effective pain medicine practice. The numerical rating scale (NRS) and visual analog scale (VAS) are two of the most commonly used unidimensional pain rating scales [1,2]. There are a number of other psychometrically validated self-report inventories to assess multiple domains of pain experience; however, when it comes to assessing pain severity in a busy clinic setting, quick and easy use of NRS and VAS is preferred [3]. With NRS or VAS, patients typically rate their current pain or delineate a summary score of their pain severity over some period (e.g., 24 h, 1 week) on a rating scale. NRS and VAS have both been deemed reliable and valid on the basis of high correlations with one another [4] or other pain instruments [5]. Yet, the use of a unidimensional instrument to measure a multifactorial phenomenon like pain is a challenging task [6].

In the case of chronic pain, patients experience a range of dysfunction and distress associated with their pain. The task of assigning a single number to describe pain can be an onerous task as they have to weigh different factors and come up with a single number to describe a complex experience [7,8]. In other words, the patient-specific meaning of the pain experience likely plays an important role in determining a single number for pain intensity, yet the number is unable to describe what the nature of the idiosyncratic meaning is [9,10].

Another complication with these unidimensional scales is anchoring inconsistency. Pain is typically anchored at zero at the low end, as no pain, whereas the highest rating (typically 10 or 100) should represent some notion of most severe pain that patients have experienced or can imagine. Clearly, the lowest point of zero pain is consistent across individuals. However, the upper end would vary as a function of their medical and life history. This makes it difficult to equivalate one person’s pain score to another person’s score, or even scores of a same person at different times. For example, a pain rating of 4 for one patient could be interpreted as a level of 2 or 6 by another patient. Similarly, a pain level of 4/10 today may be rated at 6/10 tomorrow by the same person. Idiosyncrasy in anchoring leads to varying attribution of symptom severity to each number [11,12].

When pain scores are aggregated, as in a clinical trial, measurement errors are somewhat eased, whereas this is not the case with the individual reporting of pain scores. Such pain rating scales have yielded typical errors of measurement that are too large to feel confident that scores of these scales accurately describe patient’s pain. We [6] have estimated the standard error of measurement (SEM) of pain scales across various laboratory and clinical studies to be about 13%. With this SEM, the 95% CI would extend about 26% in both directions, covering 52% of the total scale range. This will place a true pain score for the rating of 4 in the 0–10 pain scale to fall to somewhere between 1.4 and 6.6 at the 95% CI.

Can a scale be developed using standardized anchoring that still captures the idiosyncratic, clinically relevant meaning of pain for each patient? This paper presents results from the first proof of concept study to test an intuitive new interface, one that lets patients portray the recent course of pain according to the amount of time spent in various pain states. The scale represents a range that is constant across people: time. The categorical content measurements were developed based on meaningful differences in the clinical state that are easy for patients to distinguish, while the scale format and anchors are readily customized to the needs of the practice or trial. The main goal of the study was to establish the reliability and validity of this new scale, the visual time analog (VITA) scale, particularly with respect to measurement of change, as this is of paramount clinical relevance.

Materials & methods

The research protocol was approved by the Institute Review Board of the University of Utah, UT, USA. All participants provided written consent prior to entering the study.

Participants

We recruited 40 patients with a primary complaint of chronic back pain who were regularly followed by the University of Utah Pain Management Center (UUPMC). Patients with concurrent peripheral neuropathic pain, cancer-related pain and primary complaints in regions other than back were excluded.

There were 17 males and 23 females, with a mean age of 52 (standard deviation = 16, range 22 to 86). 38 subjects were white and the majority (97%) had at least high school education. All had a primary complaint of back pain. About half (n = 19) had no other pain complaint. The rest had pain complaints in a range of areas, including hip, leg, neck, abdomen and knees. Of 40 participants, 24 had four assessments. Four subjects had only one assessment, eight had two assessments and four had three assessments. Most patients had their physician visits at one to 3 months intervals. A few long-established patients were on maintenance therapy and returned to the clinic every 6 months to see their physicians.

Materials

A VITA scale was made of a hard plastic case (31 cm × 10 cm), containing a window in the middle with a pair of movable sliders dividing the window region into three areas of colors (blue, yellow, red: see Figure 1). The research team designed the scale, and a local company (Creative Awards, UT, USA) developed a prototype.

Figure 1. . Visual time analog scale prototype.

The total length of the three-color windows in the VITA scale represents a specified timeframe. In this study, we used one week as a frame, although it can be specified at any given time length, such as 24 h or 1 month. Each color of the VITA scale represents a specific pain-related state. In this study,

• Blue: unaware of pain;

• Red: aware of nothing but pain;

• Yellow: between the blue and red states.

The areas (lengths) of the VITA scale represent the percentage of total time spent in each of the three states, totaling to 100% of the timeframe. The states need not be contiguous. Each percentage is an estimated probability that a randomly selected waking moment throughout the week would fall into the respective category. The three percentages, therefore, comprise the estimated multinomial probability for one person over the past week.

Procedures

Each patient was approached by a research nurse (RM) or study coordinator (YK) while waiting in the exam room for the scheduled physician visit at the UUPMC. The study was described, and if interested, the patient’s eligibility was confirmed and consent obtained. Each patient was assessed on up to four consecutive physician visits at the UUPMC.

The procedure for each visit was identical. Every participant at the physician visit was asked to rate his or her pain in the last week (same timeframe as the VITA scale) according to the conventional NRS from 0 (no pain) to 10 (worst pain possible). We opted to use the NRS because it was a part of the normal UUPMC protocol and the patients were familiar with the scale and several studies [2,13] reported that the NRS was a preferred scale over others.

VITA measures

Following the NRS, the project staff handed the VITA scale to the patient, explaining that their task was to make the areas of color adequately represent the amount of time they spent in each of the color states in the past week. Each of the Blue, Red and Yellow states was described.

“Looking back at the past week, what percent of your time you have been unaware of pain? Please use the blue slider to indicate that amount of time. Looking back again at the past week, what percent of your time you have been aware of nothing but pain. Please use the red slider to indicate the amount of pain. The yellow is all the rest of the time”.

The patient was encouraged to adjust the sliders interactively until the colors ‘looked like’ the percentages of time spent in the corresponding pain states. By design, the red and blue areas reflect explicit time judgments, while the intermediate yellow category is implicit. Logically, the yellow area represents all times with partial or transitory awareness of pain. Once the patient feels confident that the VITA color distribution reflected their experience in the past week, he or she returned the VITA scale to the project staff who recorded the three percentages (the lengths of the three color divisions) using the graduated scale provided on the reverse side (not shown to patients).

Scoring

Categorical scores

In this study, 100% of the scale’s length corresponded to a period of 1 week. The percentage (length) of each color category provides ‘sufficient statistics’ with the clinically descriptive information. For example, a patient may have 60% red (‘aware of nothing but pain’), 5% blue (‘unaware of pain’) and 35% yellow (somewhat aware of pain: between red and blue), indicating the time distribution of the three pain states in the past week for this patient. A unique feature of the VITA scale is that the visual snapshot of the device is fully equivalent to the multinominal measurement, succinctly portraying all measured information about pain awareness over the time period.

Summary score

Probability of painful salience (PPS): although having three color scores can be informative, it is often convenient to obtain a single numeric score that summarizes the three color-percentage scores. Summary scoring also makes it easier to compare the VITA measurement performance with conventional pain measurement scores, such as NRS scores. For the VITA scale, the summary score, PPS, corresponds to the central, expected or typical value of the distribution. For purposes of comparison, we scaled the PPS scores so that they range from 0 to 10. Although scaled similarly to the NRS scores for convenience, there are notable differences between the two scale scores. Unlike NRS scores (which are discrete and do not define a true ratio scale), the PPS scores are continuous and represent an underlying probability of experiencing a salient painful moment during the predefined time period.

Scoring: VITA partitions the moments of the day, according to salience. Moments are characterized by full awareness of pain, no awareness of pain and all the rest of the time. Considering salience as a binary variable, the probability that a patient would experience full awareness of pain in a randomly selected moment of the week would be P(Salience) = (0*B + 1*R)/10. The yellow ‘rest of the time’ category also provides valuable information, however, since that category contains those moments that cannot be classified as fully aware or completely unaware. The yellow portion is, by implication, moments of partial salience: moments for which the respondent was aware of pain, but pain did not entirely dominate his or her awareness. As an approximation, these partially aware moments can be classified as though they consisted of more fine-grained moments that are fully aware and fully unaware in equal duration (for example, 1 min of partial pain awareness can be thought of as 30 s of full awareness and 30 s of no awareness). Under this assumption, the PPS is P(Salience) = (0*B + 0.5*Y + 1*R)/10 (with B, Y and R scored as percent). This scoring is completely equivalent to computing the mean of linearly scaled VITA categories, which for familiarity with NRS can be coded as 0, 5 and 10, respectively. This generates a continuous 0–10 summary pain score with the assumption of equal intervals explicit. Other scoring methods, based only on ordinal or stochastic assumptions, are also possible, but in this paper we discuss only the PPS for simplicity.

Figure 2 describes four examples of the VITA scale color distributions and corresponding PPS scores. Equal amounts of red and blue generate PPS = 5. In the first example bar, the three amounts are roughly equal. The second and third bars kept the yellow area constant (30%) but switched the blue-red distribution of 10:60% versus 60:10%. The bar on the far right has 98% red with 1% each for blue and yellow. In all cases, the PPS occurs at the halfway point of the yellow area, scored in the ‘painful’ direction of PPS = (R + 0.5Y)/10. In addition to the scoring based upon R, B and Y areas, the PPS can be immediately read off the VITA by locating the midpoint of the Y area, which defines the score against the graduated scale on the reverse of the device, much as one would read off the distance in cm of a VAS.

Figure 2. . Visual time analog scale color distribution examples.

PPS: Probability of painful salience.

Patient evaluations

At the end of their final assessment session, patients were asked to evaluate the VITA scale method according to its perceived difficulty and accuracy on the 6 point scale (1 = very easy/accurate ∼6: extremely difficult/inaccurate). In addition, patients were asked directly which of the two approaches they preferred on the 4 point scale 1: slider (VITA) much preferred, 2: slider (VITA) somewhat preferred, 3: 0–10 scale (NRS) somewhat preferred, 4: 0–10 scale (NRS) much preferred. Likert-style categories were used for each of the evaluations, which were treated as simple descriptive summaries.

Psychometric analyses are based on a formal statistical model, discussed comprehensively in Moinpour, Donaldson, Davis et al. [14], that implements classical reliability within unified statistical estimation of trajectory change and level parameters. The intercept (at time zero, the first observation), linear (initial rate-of-change) and quadratic (curvature) coefficients characterize each individual’s trajectory, allowing precisions and reliabilities to be calculated for each coefficient. The three estimated coefficients and their precisions, definitively measure pain change and level for each patient. In addition to the trajectory coefficients, the square root of the error variance from the same unified statistical model defines the SEM at single time points.

Sensitivity to change

A key aspect of measurement is its sensitivity to change. Although frequently viewed as an aspect of reliability, sensitivity to change in clinical populations under treatment is also a form of validity. Chronic pain patients will seldom change as rapidly as acute pain patients, but it is just as critical to detect and monitor the changes that do arise with chronic pain. A measurement that detects this change more precisely and sensitively is both more reliable and more valid. Treatment can range from active intervention and medication to watchful waiting.

Under conditions of change, reliability and precision cannot be determined by examining correlations or computing standard coefficients such as test-retest or Coefficient alpha. These standard approaches are valid only under an assumption of an unchanging true status. To measure systematic change, it is necessary to assume a statistical model for what is changing, and how. In the simplest case of linear change, the rate-of-change of an individual’s pain is estimated by the slope coefficient in the regression of pain on time for that individual. The regression slope coefficient is a measurement of change and like any measured value includes a degree of measurement error. Classical psychometric theory defines the reliability of any measure as the ratio Var(True)/Var(Total) = Var(True)/[Var(True) + Var(error)] = 1 - Var(error)/Var(Total).

This classical definition of reliability applies whether the measurement is a scale score at a single time or a regression slope computed over several times. The required quantities in the definitional equation can all be estimated using mixed effects linear models. With the four assessment points of this study, it is possible statistically to model a curvilinear trajectory for each patient over time, with a starting point (intercept), linear and quadratic components of systematic change, and error variation about the systematic trajectory:

$$\begin{array}{l}{Y}_{it}=b_{0i}+b_{1i}t+b_{2i}{t}^2+{\epsilon }_{it}\\ b_{0i},b_{1i},b_{2i}\sim MVN([{\beta }_0,{\beta }_1,{\beta }_2]\prime ,\Psi ),{\epsilon }_{it}\sim NID(0,{\sigma }_{\epsilon }^2)\end{array}$$

where t is elapsed time (coded here as sequential study assessment points 0, 1, 2, and 3) and the b_i are individual coefficients normally distributed about respective β population means with unrestricted covariance matrix Ψ. Random error is assumed normally and independently distributed with variance equal to the squared SEM.

In line with the generic theoretical definition of reliability as the ratio of true to observed variance, reliabilities of the measured trajectory components are collectively defined as: $R_{b{b}^{\prime }}=\Psi {(\Psi +{\sigma }_{\epsilon }^2{({\Lambda }^{\prime }\Lambda )}^{-1})}^{-1}$ [15], where Λ is the matrix of polynomial trend coefficients, yielding R_bb′, which contains the separate reliabilities of the intercept, linear and quadratic measurements in the diagonal. Although the mixed effects model is a simultaneous analysis of the entire sample, the reliabilities estimated are for independently measured individuals (e.g., the linear reliability is for the ordinary linear trend fit separately and independently for each individual over the four time points of the study [15,16]). As standardized indices that may be directly compared between NRS and VITA, these reliability statistics define the principal criteria for measurement quality. The emphasis is on measurement of change and hence on the variance components of change (linear and quadratic). The variance component and reliability for the intercept or initial status, reflect systematic differences across people that do not change over the course of the study and which are therefore distinct from treatment-related change. For each linear or quadratic component of change, there is a separate estimated population reliability. In addition, we report a summary ratio of systematic change variance to total change variance defined as $R_{\Sigma }=1-\left|{\hat{\sigma }}_{\epsilon }^2{({\Lambda }^{\prime }\Lambda )}^{-1}\right|/\left|\hat{\Psi }+{\hat{\sigma }}_{\epsilon }^2{({\Lambda }^{\prime }\Lambda )}^{-1}\right|$ , (one minus the ratio of the determinant of the estimated sampling covariance matrix of the linear and quadratic components to the determinant of the estimated population covariance matrix of the components). This is the multivariate generalization of the univariate reliability ratio of true variance to total variance.

For purposes of estimation, we expressed the above equations as a unified mixed effects model using SAS Proc Mixed under maximum likelihood. Addition of higher-order polynomial variances and covariances was evaluated by the likelihood ratio test of the difference in nested models, with terms added only if p < 0.05.

Results

Descriptive & correlational values

Table 1 shows mean NRS and VITA PPS scores. The two scales showed opposite trends. Over time, NRS scores increased linearly (p = 0.030; note: mixed effects restricted maximum likelihood with Kenward–Roger degrees of freedom and unrestricted covariance matrix) whereas VITA PPS scores showed a linear decline (p = 0.003; note: mixed effects restricted maximum likelihood with Kenward–Roger degrees of freedom and unrestricted covariance matrix) in pain. Figure 3 shows the changes in the color distributions of the VITA categories (stacked bars) and NRS and PPS scores (lines). The figure seems to present an interesting trend. As noted also in Table 1, the NRS scores show increasing pain, while the PPS shows a slight improvement. The VITA distributions provide additional detail. VITA-blue, the time spent with no awareness of pain, shows steady increase (p = 0.002; note: mixed effects restricted maximum likelihood with Kenward–Roger degrees of freedom and unrestricted covariance matrix). VITA-red, the time spent aware of nothing but pain, declines, though not significantly (p = 0.247; Note: mixed effects restricted maximum likelihood with Kenward–Roger degrees of freedom and unrestricted covariance matrix). The compensating changes in times spent aware and unaware of pain drive down the VITA PPS summary score, while the NRS ratings drift higher.

Table 1. . Mean numerical rating scale and visual time analog summary and scale scores.

VITA scales	Visit 1 (n = 40)	Visit 2 (n = 36)	Visit 3 (n = 29)	Visit 4 (n = 26)
NRS	5.24 (1.80)	5.29 (1.91)	5.56 (1.76)	6.12 (1.42)
VITA PPS	7.08 (1.83)	6.45 (2.24)	6.37 (1.87)	5.96 (1.76)

NRS: Numerical rating scale; PPS: Probability of painful salience; VITA: Visual time analog.

Figure 3. . Mean visual time analog scale color distributions, numerical rating scale and probability of painful salience scores across visits.

NRS: Numerical rating scale; PPS: Probability of painful salience; VITA: Visual time analog.

Figure 4 shows how NRS and mean VITA color percentages roughly corresponded. Over the four visits, we have a total of 104 pairs of the assessment data. The visual inspection suggests that the changes in the RED area appear to consistently correspond to the changes in the NRS ratings at every 2 points. On the other hand, the changes in the blue area was not well reflected in the NRS. The correlational analyses yielded a significant association between the NRS and VITA PPS r(102) = 0.40; p < 0.001; note: mixed effects restricted maximum likelihood with Kenward–Roger degrees of freedom and unrestricted covariance matrix). The VITA Red percentage was positively related to the NRS (r[102] = 0.43; p < 0.001 whereas the VITA yellow percentage was negatively associated with the NRS (r[102] = -0.39; p < 0.001. The correlation between the NRS and the VITA blue percentage was not significant r[102] = -0.11 (only two of the three VITA percentages are independent). This correlational pattern suggests that VITA blue may be contributing information beyond that incorporated in the usual NRS assessment.

Figure 4. . Relationship between numerical rating scale and visual time analog probability of painful salience scores.

NRS: Numerical rating scale.

At face value, visual inspection of Figure 4 reveals that levels of NRS scores vary consistently with the summary VITA PPS scores, particularly near the pain extremes (note the large changes between yellow and red between NRS = 1 and NRS = 2 and between NRS = 7 and NRS = 8). Between the extremes, the VITA PPS produce distributions of summary scores that are continuous and relatively platykurtic compared with NRS, which generates discrete scores in a more leptokurtic distribution. Distributions of good measurements tend to be relatively platykurtic (e.g., [17,18]), as sensitivity to the full range of possible pain distinctions increases the variance of the measurement distribution.

Patient evaluations

A vast majority (99%) of patients found the VITA scale easy to use (54% very easy, 28% quite easy and 17% rather easy). Accuracy was similarly evaluated favorably by 93% of the patients (22% very accurate, 23% quite accurate and 48% rather accurate). In direct comparisons, 69% preferred the VITA scale to the NRS (31% finding VITA much better and 38% finding the VITA scale a little better).

Psychometric results

Table 2 presents the criteria for comparing the psychometric performance of the VITA and the NRS. To facilitate these comparisons, we have standardized both the NRS and the VITA. This linear transformation has no effect on the statistical properties of the scores, but allows easier comparison of variance components that would otherwise be in different metrics.

Table 2. . Psychometric analyses of numerical rating scale and visual time analog scales.

	NRS	VITA
${\hat{\psi }}_{intercept}(\text{SE})$ Estimated person variance of pain at initial visit	0.211 (0.123)	0.451 (0.134)
${\hat{\psi }}_{linear}(\text{SE})$ Estimated person variance of linear pain rate-of-change in population under treatment	0.209 (0.137)	0.943 (0.500)
${\hat{\psi }}_{quadratic}(\text{SE})$ Estimated person variance of pain trajectory curvature in population under treatment	0	0.154 (0.074)
$\hat{\theta }(\text{SE})$ Estimated variance of measurement error in a single assessment	0.718 (0.124)	0.386 (0.083)
$\sqrt{\hat{\theta }}$ Estimated standard error of measurement in a single assessment	0.847 +/- 1.5 (0–10 metric) (95% CI: +/− 3.0)	0.621 +/- 1.2 (0–10 metric) (95% CI: +/- 2.4)
ρxx′ (once) Reliability of measured pain at single occasion	0.382	0.614
ρxx′ of Intercept (Reliability of measured starting point using data from all four occasions)	0.485	0.552
ρxx′ (linear) Reliability of measured linear change in pain with treatment	0.165	0.500
ρxx′ (quadratic) Reliability of measured quadratic change in pain with treatment	0	0.615
Reliability of change (combined linear and quadratic)	0.165	0.938

NRS: Numerical rating scale; VITA: Visual time analog.

The first three rows of the table quantify the true pain variability of the chronic pain population under study. Populations differ in their intrinsic variability. The variances estimate true person-to-person variability in patients, removing the effects of measurement error. These true score variances are those that would obtain theoretically if each person provided mean scores over a large number of independent replicated measurements. Lower variabilities indicate patients are more alike, higher variabilities that they are more different. Values of 0 indicate that there was no detectable variance on that component in the population.

Both NRS and VITA scales detected systematic individual differences at the first visit (Intercept), but VITA revealed variation that was roughly twice as great. VITA was much more sensitive than NRS in detecting individual change in the population. NRS detected a small degree of variability in linear trend, but was insensitive to any degree of trajectory curvature. By contrast, VITA revealed pronounced individual differences in linear trend as well as significant variation in curvature. The low variances for the NRS change components imply that the true change pattern for any randomly selected patient from this chronic pain population would be nearly identical to that for any other patient (although NRS does reveal some differences in initial pain level). Under VITA measurement, the true change patterns in pain under treatment differ notably from patient to patient and from the overall average.

The next two rows report the measurement error variability, which is the central quantity for evaluating the precision and sensitivity of the assessment scale itself. Error variability limits the sensitivity of the scale to detect any variation that may characterize a population. The square root of the error variance, the SEM, is easier to interpret. The SEM represents the typical deviation of a measured score from a patient's true score. The SEM is the expected magnitude of error in a single measurement of a person. This typical error was 36% greater for NRS than for VITA (recalling that the scales have been standardized to facilitate this comparison). In a psychometric or clinical evaluation context, this difference is quite large, as confident assessment for individuals typically requires ± twice the standard error. Below the standard errors, which are reported in consistent standardized metric, are the corresponding rough uncertainties in the native 0–10 scaling of the measures. The first uncertainty is for the standard error itself and the second for approximate 95% confidence limits of the measurement. At a single time point, without the benefit of longitudinal modeling, an NRS measurement is expected to be off by 1.5 NRS points in either direction and 95% confidence requires an interval of 3.0 NRS points in each direction.

The last rows combine the information above to yield familiar reliability indices for pain as measured by the different scales. In each case, the reliability statistics are the ratios of estimated true variability to measured variability (see [14] for comprehensive details describing reliability calculation in longitudinal change models). At a single assessment, the reliability of NRS was inadequate for all but gross distinctions among individuals. The reliability of the single assessment measurement with VITA was much higher, though still permitting only marginally confident inference about patient pain level at a single occasion. The intercept reliabilities use data from all four occasions, based on the model, to estimate the initial status reliability.

The final rows of the table report reliabilities of the change components, the most critical criteria for evaluating how well the scales measure change. The NRS failed to measure any aspect of change in pain reliably. Reliability of change assessment was quite poor for NRS. Over four assessments, the measured linear trend achieved a reliability of only 0.165, and zero reliability for the quadratic trend. Thus the total reliability for measured change was only 0.165 for NRS. In contrast, VITA achieved much higher reliabilities for both linear (reliability = 0.500) and quadratic (reliability = 0.615) change. The combined reliability ($R_{\Sigma }$, which accounts for overlap in correlated estimation of the components) was quite reliable (0.938). VITA effectively and precisely captured individual differences in how members of the population were changing, even when trends were not linear. Measuring nonlinearity is a critical complement to linear change for chronic pain patients, since ups and downs can occur even when the overall trend is positive or negative.

Discussion

The results of the present study provide preliminary support for the time-based VAS for assessing pain. The VITA scale was well accepted and liked by chronic pain patients. Even though the concept of the VITA scale was rather new to the patients, the majority of our patients found it easy to use. They also felt that the VITA scale represented their pain more accurately. Over two-thirds of the patients preferred the use of the VITA scale to the conventional NRS scale with which they were more familiar.

The psychometric analyses demonstrated that both NRS and VITA have a sizable SEM although the VITA scale performed somewhat better. If a patient’s pain score is reported as 5, the 95% CI would place the ‘true’ pain score to fall in a range of 2–8 for the NRS scale and 2.6–7.4 for the VITA scale. The VITA scale yielded a much higher reliability value for assessing change over time than did the NRS scale. We were surprised at the small degree of systematic individual variation in how patients change their NRS pain reporting. Indeed, using a formal longitudinal model, the VITA scale (PPS scores) met the benchmark reliability of 0.90 for the change scores, whereas the NRS showed poor reliability value.

The VITA scale, strictly speaking, measures time, not intensity. The severity of the pain condition is, thus, indirectly assessed by the amount of time a person is preoccupied with pain. Because the VITA scale scores are based upon time, the scale allows us to rely on the interval and ratio properties of time. If a person spends 50% of the time in the red before treatment and 25% in the red after the treatment, for example, we can ascertain that time spent being aware of nothing but pain has been reduced by half after the treatment. On the other hand, the NRS score intervals are not likely consistent or linear [19]. This makes it difficult to interpret what clinically meaningful reduction of pain, for example at 30% reduction, actually means across individuals. Furthermore, the NRS has a disadvantage of the ‘ceiling’ issue; a patient may have defined her maximum pain at a certain intensity as 10/10 but if her pain next day exceeds that level, she has no way of rating the pain level at that time. She may have to express her pain as 12 out of 10 – an impossibility in the NRS scale. On the other hand, the VITA has a maximum set point equally for all people, which is defined by the time frame of interest (1 week in our study).

The mean trajectories over visits for the NRS and VITA scale scores showed a striking difference. Overall, the pain scores got worse with time when the pain was measured with the NRS scale, whereas the patients showed a slight improvement with the VITA scale. With no other collateral data, we cannot come to a decisive conclusion about which scale had a better representation of the true progression. However, the changes in the color distribution may offer some insights. The red area of the scale, the ‘aware of nothing but pain’ domain, changed less significantly than the blue area of the scale; ‘not aware of pain’ increased over time.

The results suggest that the VITA scale may be tapping into uniquely integrated aspects of the pain experience. Although the amount of time spent on being aware versus not aware of pain should both be reflected by the overall NRS pain score, only the red area was associated with the NRS scores. The results can lead us to speculate that the NRS pain scores may be mostly driven by the awareness of pain, but not by ‘nonawareness of pain’ as the blue area represents. This may suggest that when a patient attempts to sum up their pain experience, the memory of how bad the pain was may outweigh the memory of when the pain was not bothersome. Selective bias to rely on the severe pain memory may skew their NRS response. Chronic pain patients are known to exhibit memory bias toward noxious sensory experience [20] and such bias can impact how people summarize their pain experience in the past week into a single number by focusing more on the duration when they had severe pain.

The implication of the blue area may be clinically interesting. If treatment allows patients to have more time that they are not aware of pain, how would that impact their function, mood and quality of life in general? It is possible that their life is unchanged as long as the Red area stays the same. However, it is also possible that the VITA blue score represents an untapped domain of pain measurement that uniquely provides clinically meaningful information. As patients have more time not feeling pain, they may be able to do more life activities, or organize their day in a way that would help them become more productive. Further research is needed to investigate how the increase in the VITA blue score may impact the function and well-being of the individuals.

The NRS has become such a ubiquitous feature of clinics, doctor’s offices, and emergency rooms that it is difficult to separate production of 0–10 numbers from cultural, informal and memetic contexts. The numerical response may be used to communicate urgency or a desire for specific treatment or medication, as well as pure intensity magnitude. Patients may differ considerably, among each other and over time, in how these motivations affect production of numbers. These multifocal differences reduce the consistent systematic variability of the NRS report in ways that probably reduce reliability. VITA, though not immune from these influences, provides a fresh way of reporting, based on perceived time durations, that are relatively free of these overlays.

It is also important to remind ourselves that pain is a multifactorial phenomenon. Assessment of how bad pain is for a patient is clearly a central feature of evaluating patients with chronic pain. However, treatment planning requires a comprehensive approach to understand co-morbidities and functional aspects to understand a complete clinical picture. We believe that an investigation of how VITA scores relate to other areas of chronic pain presents an important next step.

Another issue of significance is how applicable VITA pain assessment is for a wider range of patients. Since VITA is flexible in terms of defining the time frame, it can certainly be used in an acute pain setting, such as emergency departments and post-operative setting, by shortening the time frame to 1 h or 1 day. Similarly, we do not see specific contraindication for using VITA for other pain conditions such as cancer pain and episodic pain syndromes (e.g., headaches).

There are potential situations where VITA may not be appropriate. VITA is designed to assess clinical pain, not pain response to experimentally induced pain. Thus, VITA may not work well in the experimental setting. Laboratory induced noxious stimuli are short in duration and designed to have no significant discomfort after the experimental procedure is terminated.

VITA requires a time concept; those with poor time concepts (very young children, adults with neurocognitive problems) may not be able to grasp the essence of this measurement. Pain assessment of these populations is notoriously difficult. Typically, pain is assessed by behavioral observation or pictorial representation of discomfort [21,22]. For these challenging populations, pain assessment may also require us to gather information on overall health and relative degree of behaviors indicative of discomfort in order to capture the status of a particular patient.

There are some limitations of the study. As we collected the data from the patients coming to the UUPMC to see the pain medicine physicians, we depended on their visit schedules, rather than specified time intervals defined by the study. Thus, the time intervals between the visits varied across patients. Given the sample size, we were unable to analyze whether the different time intervals influenced how they rated their pain, though alternative analysis in terms of elapsed continuous time produced similar results. The number of visits, a total of four, may be too small to show important changes from the treatments. High reliability of change implies VITA is capturing how patients change differently. Some of these individual changes could be large, but probably most would be modest. And of course there is no proof that the changes happen because of the treatment. Furthermore, this study did not measure other functional and quality of life (QOL) factors relevant to chronic pain. A greater number of follow-up visits with more systematic intervals are needed for future studies, and how pain measured by the NRS and VITA scale scores interact with other areas of chronic pain domains needs to be investigated.

In conclusion, this proof of concept study provides preliminary support for the time-based, color-coded pain scale, VITA, with good acceptability and promising psychometric properties when used to assess pain in chronic pain patients. The ultimate goal of pain management is to restore the health and well-being of the patients. If patients extend the amount of time when they are not aware of the pain, that should provide an additional sign of clinically meaningful progress to what the conventional pain scale, such as NRS, can measure with the intensity of pain.

Summary points.

• Accurate and reliable pain scales are essential for pain medicine practice.

• Numerical rating scale (NRS) and visual analog scale are commonly used in clinical settings.

• Memory bias, anchoring and standard error of measurements are serious concerns associated with unidimensional pain scales.

• The visual time analog (VITA) scale aims to assess pain severity by measuring the amount of time spent in various pain states (blue: ‘unaware of pain’, red: ‘aware of nothing but pain’, yellow: between blue and red).

• Pain scores over four visits to the pain clinic showed slight worsening of pain with NRS but slight improvement with VITA.

• VITA red and NRS show significant association but VITA blue was not correlated with NRS.

• Psychometric analyses showed better reliability indices with VITA compared with NRS.

• VITA had smaller standard error of measurement than did NRS.

• VITA, particularly VITA blue, may supplement conventional NRS by adding an uniquely integrated aspect of pain experience.