Introduction
Pain is the most common chief complaint leading to emergency department (ED) visits. 1 - Reference Todd, Ducharme and Choiniere 3 Optimal pain management is lacking in most EDs, as many patients do not receive analgesiaReference Todd, Ducharme and Choiniere 3 - Reference Pines, Shofer and Isserman 12 and time to analgesic administration is often too long.Reference Todd, Ducharme and Choiniere 3 , Reference Todd, Sloan and Chen 5 - Reference Grant 8 , Reference Singer, Garra and Chohan 10 - Reference Daoust, Paquet and Lavigne 19 Most health care professionals underestimate pain intensity.Reference Rundshagen, Schnabel and Standl 20 , Reference Mantyselka, Kumpusalo and Ahonen 21 In the absence of tachycardia or hypertension, we observed that ED staff tend to discredit self-assessments of severe pain when patients do not appear to be suffering. Health care professionals have also noted variations in heart rate (HR) and blood pressure associated with changes in pain intensity, validating their notion that pain intensity levels are linked with significant fluctuations in vital signs. However, in the prehospital and ED literature, it is unclear how useful vital signs are for predicting pain intensity.Reference Moltner, Holzl and Strian 22 - Reference Tousignant-Laflamme, Rainville and Marchand 30
Pain intensity is subjective, and with conflicting results, researchers have tried to identify unbiased parameters (including vital signs) to validate pain intensity. HR (or HR variability), more than blood pressure, has been associated with pain intensity,Reference Moltner, Holzl and Strian 22 - Reference Bendall, Simpson and Middleton 24 , Reference Loggia, Juneau and Bushnell 27 , Reference Arbour and Gelinas 31 - Reference Treister, Kliger and Zuckerman 34 but mostly in males.Reference Tousignant-Laflamme and Marchand 29 , Reference Tousignant-Laflamme, Rainville and Marchand 30 Some studies do not support this association.Reference Peckerman, Saab and McCabe 35 - Reference Ledowski, Tiong and Lee 38 Systolic (SBP), diastolic (DBP), and mean arterial pressure (MAP) have been associated with pain intensity.Reference Bendall, Simpson and Middleton 24 , Reference Landrum and Watson 26 However, as was the case for HR, other investigations reported an absence of relationship.Reference Marco, Plewa and Buderer 28 , Reference Ledowski, Reimer and Chavez 33 , Reference Peckerman, Saab and McCabe 35 , Reference Lord and Woollard 39 Most studies have failed to show an association between pain and respiratory rate.Reference Moltner, Holzl and Strian 22 , Reference Bendall, Simpson and Middleton 24 , Reference Landrum and Watson 26 , Reference Marco, Plewa and Buderer 28 , Reference Arbour and Gelinas 31 , Reference Ledowski, Reimer and Chavez 33 , Reference Fowler, Slater and Garza 37 , Reference Lord and Woollard 39 Transcutaneous oxygen, skin conductance, and the analgesia/nociception index (calculated from HR variability) have been assessed in patients experiencing pain with inconsistent conclusions.Reference Hampf 23 , Reference Loggia, Juneau and Bushnell 27 , Reference Arbour and Gelinas 31 , Reference Chang, Ma and Tsay 32 , Reference Treister, Kliger and Zuckerman 34 , Reference Ledowski, Tiong and Lee 38 , Reference Boselli, Daniela-Ionescu and Begou 40 - Reference Ledowski, Stein and Albus 42
Patients with severe pain who show no hypertension or tachycardia to substantiate their pain intensity might not receive the aggressive pain treatment they warrant. Our study aimed to determine whether vital signs are associated with pain intensity levels in both the ED population and a subgroup of patients with diagnosed conditions recognized to produce significant pain. In the same populations, we also investigated whether changes in vital signs in individual patients are coupled with variations in pain intensity levels over time. We hypothesized that ED patients, especially those with recognized painful conditions, will demonstrate a weak association between vital signs and self-reported pain intensity levels.
Methods
Study design and setting
We undertook retrospective analysis of real-time, archived data from electronic medical prescription and nursing record systems in the ED of two health care institutions:, a tertiary academic urban hospital with an annual census of approximately 60,000 ED visits per year, and a secondary regional community hospital with approximately 48,000 ED visits per year. The computerized system included demographic data, final diagnosis, triage information (reason for ED visit pain assessment, vital signs), and subsequent vital signs and pain intensity evaluation during ED stay. All information was collected in real time and time-stamped. The study was approved by the ethics review boards of both institutions.
Selection of participants and study protocol
All consecutive patients over 16 years of age with pain intensity at triage of 1 or more on an 11-point verbal numerical scale (VNS) who presented to the tertiary urban hospital between March 2008 and December 2012, or to the secondary regional hospital between March 2005 to December 2012, were included in the study. To evaluate pain intensity, nurses instructed patients to rate their pain between 0, representing “no pain,” and 10, representing “the worst pain imaginable.” Vital signs were measured by automated electronic sphygmomanometer. From the database, we extracted: gender, age, Canadian Triage and Acuity Scale (CTAS) priority level (high=1, 2; low=3, 4, 5), arrival mode (ambulance or walk-in), ED visit duration, disposition after ED (admission or release) and final diagnosis charted by ED physicians. Pain intensity, HR, SBP, and DBP recorded at triage and at first evaluation during the ED stay were selected for analyses. MAP was calculated as [(2×diastolic)+systolic]/3. As our primary outcome, we studied the associations between pain intensity categorized as mild (VNS 1–3), moderate (VNS 4–6), or severe (VNS 7–10)Reference Coniam and Mendham 43 and values of HR, SBP, DBP, and MAP.
For our secondary outcomes, we selected a subgroup of patients with two assessments of pain intensity within the first 24 hours of their ED stay and ascertained whether changes in pain intensity levels in this population were linked with variations in HR, SBP, DBP, or MAP. We also studied the same associations in patients with diagnosed conditions recognized to produce significant pain: fracture, dislocation, and renal colic. These diagnoses were identified in a focus group of ED physicians from the tertiary urban hospital. The group was instructed to find ED diagnoses that physicians would unanimously agree on as having a pain intensity generated by “genuine” painful stimuli. All analyses were repeated, excluding patients with a pacemaker or taking beta-blocker medication known to possibly affect vital signs.
Statistical analyses
Patients’ characteristics from the two hospitals were compared by χ2, Mann-Whitney, and independent t-tests, depending on the scale of the variables. Because of the large sample size, Cohen’s effect sizes (ES) were reported instead of significance levels. To compare vital signs at triage for the three pain intensity categories, a one-way analysis of variance (ANOVA) was performed for each vital sign.
A one-way ANOVA was also conducted for the subgroup of patients with two pain intensity assessments, to compare differences in vital signs associated with pain intensity category differences. Pain intensity differences were divided into four categories: 1) “major drop,” i.e., patients with severe pain intensity at triage and mild pain intensity at second evaluation; 2) “minor drop,” i.e., severe to moderate or moderate to mild pain intensity; 3) “stable,” i.e., no change in pain intensity between the two evaluations; and 4) “increased,” i.e., heightened pain intensity level at the second evaluation. We did not distinguish between minor and major increases because, in the context of pain management in EDs, very few patients presented with a major increase in pain intensity level. For all one-way ANOVAs, Cohen’s (f) ES were reported. They are described as: small=0.10, moderate=0.25, and large=0.40. An ordinal logistic regression was first performed to evaluate the contribution of vital signs to pain intensity level during triage, controlling for age and gender. However, the assumption that “the relationship between each pair of outcomes was the same” was rejected in all our analyses. Multinomial logistic regressions served as an alternative, to facilitate interpretation; only odds ratios (OR) ±95% confidence intervals (CI) of mild as opposed to severe pain intensity category were reported. To facilitate clinical interpretation, dichotomous variables were created from vital signs with the threshold suggested by Bendall et alReference Selbst and Clark 4 : HR ≥100 beats/min, SBP ≥140 mm Hg, and DBP ≥90 mm Hg. All analyses were made with SPSS version 20 (IBM, Somers, NY).
Results
During our study period, 153,567 patients aged 16 years or older were triaged with a pain intensity of at least 1/10, and 13,691 patients (8.9%) had a diagnosed condition recognized to produce significant pain (Figure 1). Patient characteristics were clinically similar for the two health care institutions, except for longer median length of stay and higher percentage of arrival by ambulance at the secondary regional hospital (Table 1). The results for all outcomes were not different between hospitals, so they were pooled. Eight percent of the patients had a pacemaker or had taken beta-blocker medication prior to or during the study. Results including or excluding those patients were similar; therefore, results for the whole sample are presented.
Figure 1 Flow chart of patients’ selection.
Table 1 Comparisons of patients’ characteristics for the two hospitals.
* HSCM: Hôpital du Sacré-Cœur de Montréal;
† HRSJ: Hôpital régional de St-Jérôme;
‡ Fracture, dislocation, renal colic.
The mean age of our entire cohort was 48.4 years (SD±19.3), 55.5% were women, and 22.3% were 65 years old or older. Median pain intensity was 7/10 (IQR±3). Because our sample size was large, statistically significant differences (p<0.001) in HR, SBP, DBP, and MAP were evident between each category of pain intensity: the difference between mild and severe pain intensity categories for HR was 2.7 beats/min (95% CI: 2.4−3.0), 0.13 mm Hg (95% CI: −0.26−0.52) for SBP, 1.5 mm Hg (95% CI: 1.2−1.7) for DBP, and 1.0 mm Hg (95% CI: 0.76−1.3) for MAP. However, these differences were not clinically significant and generated very small ES (<0.10) (Figure 2). The results were similar for diagnosed conditions recognized to produce significant pain (fracture, dislocation, and renal colic): the difference was 1.3 beat/min (95% CI: 0−2.7) for HR, 3.6 mm Hg (95% CI: 1.6−5.5) for SBP, 3.5 mm Hg (95% CI: 2.2−4.8) for DBP, and 3.5 mm Hg (95% CI: 2.2−4.9) for MAP.
Figure 2 Mean (±SD) systolic, diastolic, mean arterial blood pressure and heart rate as a function of three levels of triage pain intensity in the overall sample and in patients with recognized painful conditions. Cohen’s one-way ANOVA effect sizes are also reported.
Table 2 presents the results of multinomial logistic regressions, predicting pain severity as a function of each vital sign, adjusted for age and gender for both the entire sample and in patients with conditions recognized to produce significant pain. For the entire sample, vital signs demonstrated low odds ratio (OR) (≤1.3), with age and gender producing generally higher OR than vital signs. The results were similar for patients with conditions recognized to produce significant pain;however, age and HR were no longer significant predictors.
Table 2 Odd ratios (±95% CI) of multinomial logistic regressions (mild as opposed to severe pain intensity) for the entire sample and for patients with recognized painful conditions.
* Odd ratios in bold character are significant at p<0.001.
† Fracture, dislocation, renal colic.
The median time between two consecutive pain intensity assessments (at triage and second evaluation within 24 hours) used to examine associated vital signs changes over time was 2.9 hours (IQR=5.4). Also, because our sample size was large, changes in pain intensity were significantly associated withvariations in HR, SBP, DBP, and MAP (Figure 3). Differences between the “major drop” and “increased” categories of change in pain intensity were 3.4 beats/min (95% CI: 2.4−4.4) for HR, 8.8 mm Hg (95% CI: 7.2−10.4) for SBP, 6.2 mm Hg (95% CI: 5.0−7.4) for DBP, and 7.1 mm Hg (95% CI: 5.9−8.3) for MAP. For conditions recognized to produce significant pain, changes were generally larger (except for HR) with: 2.1 beats/min (95% CI: −0.48−4.7) for HR, 17.5 mm Hg (95% CI: 13.0−21.9) for SBP, 11.2 mm Hg (95% CI: 8.1−14.3) for DBP, and 13.5 mm Hg (95% CI: 10.3−16.6) for MAP. However, they only represented small to moderate ES.
Figure 3 Mean (±SD) changes in systolic, diastolic, mean arterial blood pressure and heart rate as a function of four levels of pain intensity change in the overall sample and in patients with recognized painful conditions. Cohen’s one-way ANOVA effect sizes are also reported.
Discussion
To the best of our knowledge, this is the largest cohort study exploring the association between vital signs and acute pain intensity. We found small ES (0.01 to 0.04) for blood pressure and for pulse (0.06), confirming that vital signs should not substantiate self-reported acute pain intensity level in the ED. Our results are similar to those of other studies in ED, prehospital, or military settings.Reference Bossart, Fosnocht and Swanson 25 , Reference Marco, Plewa and Buderer 28 , Reference Bartfield, Janikas and Lee 36 , Reference Fowler, Slater and Garza 37 , Reference Lord and Woollard 39 For example, Bendall et al demonstrated, in a retrospective prehospital study, a slightly increased likelihood of severe pain if: HR is ≥100 beats/min for patients ≤64 years old, or if SBP was ≥140 mm Hg for patients ≥65 years old (18% and 14%, respectively). However, these authors concluded: “Simple correlation fails to show clinically important association between prehospital vital signs and pain severity.”Reference Bendall, Simpson and Middleton 24 Ledowski et al established that plasma epinephrine and norepinephrine levels in a postoperative population were generally not associated with pain intensity, the only exception being norepinephrine level changes for major changes in pain intensity. They concluded that “the absence of signs of sympathetic stimulation cannot be seen as a guarantee for absence of significant pain.”Reference Lord and Woollard 39
As stated in our introduction, some literature supports a relationship between vital signs and pain intensity, but most studies had a small sample or were derived from post-operative and/or laboratory settings. A small sample can bias the finding toward a positive association if, by chance, patients with similar responses to pain are included in the study. Fazalbhoy et al,Reference Fazalbhoy, Birznieks and Macefield 44 employing intramuscular infusion of hypertonic saline to evoke pain, determined that muscle sympathetic nerve activity, blood pressure, and HR were increased in some subjects, while others presented decreased variables. Furthermore, these responses were constant over time.
Many health care professionals still believe that vital signs are useful to validate pain intensity, particularly if their impression is different from the patient’s self-evaluation. Our study sheds light on a possible explanation. For the whole population, blood pressure and pulse varied slightly with changes in pain intensity over time. ES for HR, SBP, DBP, and MAP in our sample were larger for changes of pain intensity over time than for a specific level (particularly in recognized painful conditions). Consequently, physicians notice an association between changes in vital signs and variations in pain intensity in ED patients as a group. However, standard deviations are large, which makes it difficult to predict pain intensity level or variations associated with these vital signs in a single individual.
Pain encountered in EDs is generated by a wide array of pathologies, and a significant proportion of them could influence vital signs independent of the pain they cause. In our study, age and gender were more associated with pain intensity than vital signs. When health care professionals notice sudden variation in vital signs, they should re-evaluate patients’ medical conditions, including pain intensity. However, these changes in vital signs should not be attributed primarily to variations in pain intensity.
Our study has potential limitations. Data were not recorded specifically to address the study objectives. Database accuracy depends on personnel who enter the data. However, nurses are well aware of the importance of accurate and detailed charting. Moreover, triage nurses receive comprehensive training in administering the pain intensity scale, and the electronic interface prompts them to adopt specific wording as well as completion of an 11-point numerical scale to evaluate pain intensity.
The stimuli generating pain were not standardized; however, it appears that perceived pain is more closely related to vital signs than stimuli intensity.Reference Moltner, Holzl and Strian 22 , Reference Loggia, Juneau and Bushnell 27 , Reference Tousignant-Laflamme, Rainville and Marchand 30 , Reference Arbour and Gelinas 31 Furthermore, the retrospective study design could not control for vital signs prior to the pain stimuli and for other factors that could influence changes in vital signs.
On the other hand, the study has the following advantages: vital signs were measured automatically by electronic sphygmomanometer and the record review was performed by computer. Indeed, computerized medical record review has been shown to be more efficient and accurate than manual review of ED records.Reference Biese, Forbach and Medlin 45
Conclusion
ED vital signs (heart rate, systolic, diastolic, and mean arterial blood pressures) should not be used to estimate or substantiate self-reported acute pain intensity levels. The study conclusions also apply for variations in pain intensity over time for an individual patient.
Competing Interests: The statistical analysis part of the study was supported by the Emergency Department Research Fund of Hôpital du Sacré-Coeur de Montréal.
