Problems With Measurement of the Minimum Clinically Significant Difference in Acute Pain in Elders
Reprints will not be available.
Presented at the Society for Academic Emergency Medicine annual meeting, Phoenix, AZ, June 2010.
Disclosures: Dr. Bijur was partially supported by The National Institute of Nursing Research 1R21NR010929.
Supervising Editor: James Miner, MD
Abstract
ACADEMIC EMERGENCY MEDICINE 2011; 18:135–139 © 2011 by the Society for Academic Emergency Medicine
Objectives: A standard value for the minimum clinically significant difference (MCSD) in pain in nonelderly ED patients has been identified and has facilitated research in this age group. It is not clear that this value is similar in older patients. The standard method for calculating the MCSD in pain is to average scores on a numerical rating scale (NRS) over contiguous time periods. This method is based on the assumption that the MCSD remains constant over time. In an earlier hypothesis-generating study of elderly patients, this assumption was not met for the arithmetic MCSD, making it difficult to identify a single benchmark for measuring efficacy in analgesia trials in elders. The proportional MCSD was more stable, suggesting that it might constitute a better measure of analgesic efficacy in elderly patients. The objective of the study was to test the hypotheses that: 1) the arithmetic MCSD in adults 65 years and older declines over time and that 2) the proportional MCSD remains constant.
Methods: This was an observational, prospective, cohort study of emergency department (ED) patients ≥65 years with acute pain. Pain intensity was rated on a standard 11-point NRS upon study entry and every 30 minutes for 2 hours. The arithmetic MCSD was defined as the mean change in pain between contiguous 30-minute intervals when change in pain was described as “a little less” or “a little more.” The proportional MCSD was calculated as the arithmetic MCSD divided by pain intensity at the beginning of the interval. We used generalized estimating equations (GEEs) to test trend over time.
Results: A total of 214 patients were enrolled: mean (± standard deviation [SD]) age was 74 (±7.5) years, 66% were female, 63% were Hispanic, and 23% were African American. The median initial NRS was 8. The MCSD decreased 2.1 NRS units (95% confidence interval [CI] = 1.7 to 2.4) between 0 and 30 minutes, 1.4 units (95% CI = 1.0 to 1.7) between 30 and 60 minutes, 1.3 units (95% CI = 1.0 to 1.5) between 60 and 90 minutes, and 0.8 units (95% CI = 0.6 to 1.0) between 90 and 120 minutes (p < 0.001 for trend). The proportional MCSD also varied from 27% (95% CI = 23% to 32%) between 0 and 30 minutes, 19% (95% CI = 13% to 24%) between 30 and 60 minutes, 22% (95% CI = 18% to 27%) between 60 and 90 minutes, and 13% (95% CI = 9% to 18%) between 90 and 120 minutes (p < 0.001 for trend).
Conclusions: Both the arithmetic and the proportional MCSD in elderly patients in acute pain declined over time. Because both measures were numerically unstable, there does not appear to be a single value for the MCSD that can be used to identify the MCSD in pain for use in analgesic efficacy trials in elderly patients. A different metric may be needed to study pain and assess comparative analgesic efficacy in elderly patients.
There is substantial evidence that older adults perceive and process painful stimuli differently than younger adults.1 Furthermore, there are many cognitive, social, and psychological age-related differences that may affect the report of pain.2,3 Because of these differences it cannot be assumed that measures of pain that are clinically significant are the same in older and younger patients.
The identification of a minimum clinically significant difference (MCSD) in pain in younger patients has facilitated research on the efficacy of analgesia in this age group. A recent study sought to identify the analogous value for older patients.4 Results from that study indicated that the MCSD declined over a 2-hour period. Averaging the arithmetic MCSD over contiguous time periods, as is traditionally done,5–7 rests on the assumption that the MCSD remains relatively constant over time. However, if the arithmetic MCSD varies over time in the elderly, the magnitude, meaning, and clinical utility of the MCSD become unclear. In contrast to the arithmetic MCSD, the proportional MCSD in the prior study of elderly patients did not vary over time.4 Instead, the proportional decline was constant at about 25% over each of the four 30-minute intervals. We reasoned that if upon replication, we found similar results, the proportional MCSD might be used in place of the arithmetic MCSD as a benchmark for assessing the efficacy of different analgesic regimens in elderly patients.
The present study replicates the design of this earlier study4 in a similar elderly population. Based on the findings of the prior hypothesis-generating study, we wish to test the following hypotheses: 1) that the arithmetic MCSD in elderly patients in acute pain declines over time and 2) that the proportional MCSD is numerically stable over time.
Methods
Study Design
This was a prospective observational cohort study. Elders presenting to the emergency department (ED) with acute pain were asked to report pain intensity every 30 minutes for 2 hours. They were also asked to compare current pain with pain in the preceding 30-minute period. The data were collected from December 2008 to June 2009. The study was approved by the institutional review board of Albert Einstein Hospital. All participants signed informed consent in either English or Spanish.
Study Setting and Population
The study was performed in the ED of Albert Einstein Hospital, the academic medical center for the Albert Einstein College of Medicine in New York City, serving an urban population that is predominantly Hispanic and African American. The ED is staffed around the clock by salaried, bilingual, technician-level research associates whose sole task is to gather data for clinical studies under the supervision of the principal investigators.
Inclusion criteria were age 65 years and older, pain score of 1 or more at triage, English- or Spanish-speaking, not cognitively impaired as indicated on a six-item validated screening test of cognitive impairment8 or by inability to give informed consent due to lack of comprehension, expected to remain in ED for 2 hours, not intoxicated as judged by a nurse or physician, and able to give informed consent. Patients were excluded if they had pain for longer than 7 days or had used opioids in the seven days before the ED visit.
Study Protocol
After informed consent was obtained, patients were asked to report their pain intensity on an 11-point numerical rating scale (NRS). Every 30 minutes after baseline (entry into the study), patients were asked to rate the intensity of their pain on the NRS. At each time point after baseline they were also asked to contrast current pain intensity with pain at the time of the previous measurement, using one of the following five categorical descriptors: “much less pain,”“a little less pain,”“about the same pain,”“a little more pain,” or “much more pain.” The arithmetic MCSD was defined in standard fashion5–7 as the mean arithmetic change in NRS associated with a description of the change in pain as either “a little less pain” or “a little more pain” over a 30-minute period. The proportional MCSD was defined as the mean arithmetic MCSD divided by the NRS score at the beginning of each 30-minute interval. Patients received analgesics at the discretion of their physicians.
Data Analysis
Characteristics of the sample were summarized by proportions, medians, interquartile ranges, means, and standard deviations (SDs) as appropriate. We used the Kolmogorov-Smirnoff test of normality of the distribution of change in pain for ratings of “a little less” or “a little more” pain. The change in pain intensity was calculated by subtracting the NRS score at the end of each 30-minute interval from the NRS score reported at the beginning of that interval. In order to calculate the MCSD, the sign of the NRS change score was reversed for ratings of “a little more pain.” Proportional change was calculated by dividing the change of NRS in a 30-minute period by the NRS at the beginning of the period. Means and 95% confidence intervals (CIs) are presented. Generalized estimating equations (GEEs) were used to test trends over time with change in NRS treated as a continuous variable. We also used GEE to estimate the relationship between time and NRS, categorized as follows: change in NRS ≤0 = 0, 1 = 1, 2 = 2, and ≥3 = 3 and proportional NRS divided into quartiles. We used SPSS version 17 (SPSS, Inc., Chicago, IL) to conduct all data analyses.
The sample size was based on calculating a 95% CI of ±0.5 NRS units around the estimated mean MCSD of 1.3 and a SD of ±3.0. A total of 140 descriptors of change in pain as “a little less pain” or “a little more pain” are needed to attain this level of precision. Based on past experience, we estimated that 200 patients, each providing five ratings of pain intensity, would be needed to provide at least 140 ratings of “a little less pain” or a “little more pain.” We used nQuery Advisor version 6.0 (Los Angeles, CA) for sample size calculation.
Results
A total of 290 patients were approached for inclusion in the study. The most frequent reason for exclusion was cognitive impairment. A total of 214 patients were enrolled; 147 of them rated their pain as “a little less pain” or “a little more pain” at one or more 30-minute time intervals. These patients provided 250 ratings of “a little less pain” and 32 ratings of “a little more pain” over the 2 hours of the study. Table 1 shows that the sample was predominantly female and Hispanic.
| Variables | Number | % |
|---|---|---|
| Sex | ||
| Female | 142 | 66.4 |
| Male | 72 | 33.6 |
| Race/ethnicity | ||
| Hispanic | 134 | 62.6 |
| African American | 49 | 22.9 |
| White | 26 | 12.1 |
| Other | 5 | 2.3 |
| Location of pain | ||
| Head/neck | 33 | 15.4 |
| Chest | 19 | 8.9 |
| Abdomen/pelvis | 75 | 35.0 |
| Extremities | 24 | 11.2 |
| Back | 24 | 11.2 |
| Other | 39 | 18.2 |
| Pain intensity at baseline NRS | ||
| 2 | 4 | 1.9 |
| 3 | 12 | 5.6 |
| 4 | 15 | 7.0 |
| 5 | 21 | 9.8 |
| 6 | 14 | 6.5 |
| 7 | 15 | 7.0 |
| 8 | 43 | 20.1 |
| 9 | 22 | 10.3 |
| 10 | 68 | 31.8 |
| Median (IQR) | 8 | 6–10 |
| Age, yr (mean, ±SD) | 74 | ±7.5 |
- IQR = interquartile range; NRS = numerical rating scale.
Table 2 shows that both the arithmetic and the proportional MCSD declined over time (arithmetic b = −0.394, p = 0.001; proportional b = −0.037, p = 0.001). The change in pain scores was not normal (Kolmogorov-Smirnov Z = 3.2, p < 0.001); thus an ordinal model was also used to test for trend. This analysis resulted in the same inference that the MCSD declines over time (arithmetic b = −0.694, p < 0.001; proportional b = −0.535, p < 0.001).
| n | Arithmetic MCSD Within Interval | Proportional MCSD Within Interval | |||
|---|---|---|---|---|---|
| Mean | (95% CI) | Mean, % | (95% CI) | ||
| Baseline–30 min | 67 | 2.1 | (1.7–2.4) | 27 | (23–32) |
| 30–60 min | 68 | 1.4 | (1.0–1.7) | 19 | (13–24) |
| 60–90 min | 75 | 1.3 | (1.0–1.5) | 22 | (18–27) |
| 90–120 min | 67 | 0.8 | (0.6–1.0) | 13 | (9–18) |
- Arithmetic MCSD = NRSt– NRSt+30 associated with descriptors of “a little less pain” or “a little more pain” in each time period.
- Proportional MCSD = NRSt– NRSt+30 associated with descriptors of “a little less pain” or “a little more pain” in each time period divided by NRS at beginning of each time period.
- MCSD = minimum clinically significant difference.
For patients with abdominal pain, the arithmetic MCSD decreased over the four periods, but the proportional MCSD did not (arithmetic b = −0.329, p = 0.008; proportional b = −0.023, p = 0.213). The ordinal models of trend for patients with abdominal pain showed a negative trend over time for both arithmetic and proportional MCSD (arithmetic b = −0.674, p < 0.001; proportional b = −0.547, p < 0.001). The magnitude of coefficients for all patients and patients with abdominal pain were similar.
Fifty-four patients did not receive any analgesics during the 2-hour period. Eight (15%) reported “a little less,” and two (4%) reported “a little more pain” in the first 30-minute period. The mean change in pain of these 10 patients was 2.1 (95% CI = 0.6 to 3.6).
Discussion
The standard approach to calculating the MCSD has been to ask patients to report their pain intensity on an NRS or visual analog scale in several sequential time periods and then average the arithmetic change in pain associated with “a little less pain” or “a little more pain” over all time periods.5–7 This assumes that the MCSD remains constant within all time periods so that the average change is consistently representative of the smallest amount of pain relief noted by the patient (i.e., the arithmetic MCSD) within any given 30-minute time interval. This assumption was unexpectedly not met in our earlier study,4 as the MCSD declined over the four 30-minute time periods. This finding was confirmed in the present study as hypothesized. Because of this apparent variation, the average value of the MCSD over time may not be clinically meaningful in assessing comparative analgesic efficacy in elderly patients. The proportional MCSD also declined over time from a high of 25% in the first time period to 13% at the last.
To assess whether the type of pain affected the findings, we examined the MCSD in the patient subgroup with abdominal pain. The decline in MCSD was smaller for patients with abdominal pain than for all patients, although the patterns were similar. As abdominal pain is more likely to fluctuate than pain in other locations, the MCSD for abdominal pain may be less stable over time than the estimate of the MCSD from patients with other types of pain.
The MCSD for 10 patients who did not receive analgesics was similar to the MCSD of all patients. This suggests that regardless of the reason for change in pain (medication or natural variation) a similar magnitude of change on the NRS is associated with the description of change in pain as “a little more” or “a little less.”
A descriptive analysis of the NRS scores revealed discrepancies between descriptors of change in pain and the numerical change in NRS scores. Twenty-four percent of descriptors of “a little less pain” or “a little more pain” in the current study were associated with no change in NRS scores, while 14% of them equaled or exceeded three NRS units. This marked difference in change in NRS associated with the same descriptors, combined with the numerical and proportional instability of the NRS in elders, may reflect the difficulty older patients have in quantifying their experience of pain. This discrepancy raises a question about the validity of the measure and calls into question the utility of any numerical scale for measuring pain intensity in older patients.
A more straightforward means of comparing analgesic efficacy may be to simply ask patients if they want more analgesia or if their pain is adequately controlled. Asking if patients want more medication is appealing because it is easily understood and patient-centered. An additional advantage of using the simple, dichotomous question: “Do you want more pain medication?” is that it allows patients to take into account a wide variety of considerations relevant to pain management that go beyond measurement of pain intensity alone, e.g., the many side effects of analgesics that are, in the judgment of that particular patient, less desirable than further reduction of pain intensity. In the end, an omnibus assessment of the overall efficacy of pain management needs to take into account the multidimensional experience of pain and its treatment that go well beyond a numerical assessment of its intensity.
Limitations
All elderly patients with pain were eligible for the study regardless of the location or etiology of pain. Some types of pain are more likely to fluctuate over time than others, such as pain from renal or biliary colic. The report of change in pain may reflect fluctuations in discomfort experienced by the patient at various points within any given interval, rather than the difference in pain intensity from the beginning to the end of the period. This would contribute to error, but does not seem likely to produce a systematic bias in either direction.
Patients in this prospective cohort study may differ from patients enrolled in clinical trials. The intensity of pain needed to be eligible for clinical trials, particularly trials of opioids, would be higher than in this observational study that included any patient with a complaint of pain at triage. Further, patients willing to be enrolled in a clinical trial may differ from patients in observational studies in terms of attitudes toward pain management, notions of risks of addiction, expectation of side effects, and a general distrust of clinical research.
To compare our results with that of other studies of the MCSD, we have included change in NRS scores for ratings of both “a little less pain” and “a little more pain.” This strategy does not directly measure the relevant clinical phenomenon of decrease in pain. Rather, it measures the smallest change in pain discernible to the patient, regardless of the direction of change.
The NRS measures pain intensity on an 11-point scale. While this reflects an underlying continuous dimension of pain, it is not a continuous variable, and as such the mean and standard error should not formally be used as measures of central tendency and dispersion. Further, it is not normally distributed. We chose to employ this scale to test a hypothesis from the earlier study that used the NRS. It is also easier for older patients to use than the visual analog scale (VAS) as they may be supine, may not have brought reading glasses, and may have difficulty writing. We do not think this use distorted the findings or affected inference. In an earlier study that compared the NRS and VAS, we found almost identical mean values and 95% CI of the MCSD: 1.3 (95% CI = 1.0 to 1.5) on the NRS and 1.4 (95% CI = 1.1 to 1.7) on the VAS.9 Also, in the current study we found a highly significant decrease in NRS when we categorized the change in pain scores and performed a nonparametric test.
This hypothesis-testing study was conducted in a sample drawn from the same patient population as the original hypothesis-generating study. Both samples came from a population that is mainly urban, underserved, and predominantly Hispanic and African American. We excluded patients with chronic pain conditions and with cognitive impairment. Thus the findings may not be generalizable to other populations.
Conclusions
The numeric rating scale measurements of both the arithmetic and the proportional minimal clinically significant difference in pain for elderly patients appear to be numerically unstable. Because both measures varied substantially over consecutive time intervals, there does not appear to be a single value for the MCSD that can be used to identify the minimum clinically significant difference in pain for use in analgesic efficacy trials in elderly patients. A different metric may be needed to assess the efficacy of treatment for pain in older patients.
