Journal of Pain & Palliative Care Pharmacotherapy. 2013;27:350–358. Copyright © 2013 Informa Healthcare USA, Inc. ISSN: 1536-0288 print / 1536-0539 online DOI: 10.3109/15360288.2013.846955
ARTICLE
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
Quality of Postoperative Pain Management in American Versus European Institutions C. Richard Chapman, Duncan A. Stevens, and Arthur G. Lipman AB STRACT Management of postoperative pain remains an important clinical problem throughout the world. Using the PAINOUT acute pain registry database to examine perioperative pain management in orthopedic surgery patients, we compared patient-reported outcomes (PROs) in a pooled sample obtained from four American hospitals (N = 473) with PROs in a pooled sample of 20 European institutions (N = 8799). Most American hospitals consistently assess acute pain in surgical patients due to Joint Commission accreditation guidelines. Therefore, we hypothesized that this practice would create a climate of clinical staff sensitivity to patients’ pain and a greater readiness to intervene when pain is higher than one would find in Europe as a whole. American institutions might then provide better control of postoperative pain after orthopedic surgery than European institutions. Because of the large sample sizes, our analyses focused on effect size rather than statistical significance. Evaluation of the pain PROs revealed that European patients reported much lower Worst Pain on the first day after orthopedic surgery than American patients. The mean Worst Pain (± SD) for Europeans was 5.4 (2.5) but for Americans the mean was 7.4 (2.7), p < .0001, a large effect size. Europeans also reported significantly less emotional discomfort, less interference of pain with activity and lower Least Pain. Nonetheless, 98.3% of American patients received opioids on the ward on the first postoperative day compared to 70.2% of European patients, and 41.1% received regional analgesia on the ward while 15.9% of European patients received regional analgesia (both small effect sizes). Overall, the results are clear in demonstrating much better pain control in the ensemble of European countries as compared to the United States. KEYWORDS acute, perioperative, postoperative, pain, PAIN-OUT, patient-reported outcomes, Europe, USA
million). This translates into about one surgery each year for every 25 people.1 Unfortunately, postoperative pain remains a common problem despite notable advances in the scientific understanding of pain in recent decades. A review of data on postoperative pain on the first day following surgery in 105 German hospitals, 2 found that high pain scores can result from almost all surgical procedures, including those that appear minimally traumatic. Poorly controlled pain can contribute to postoperative complications, impede recovery, extend hospitalization, and cause needless suffering. Importantly, unresolved acute postoperative pain can progress to debilitating chronic pain. Kehlet et al. estimated that unresolved acute pain after surgery leads to chronic postoperative pain in 10% to 50% of patients after common operations.3 Orthopedic surgeries are among the major generators of intense postoperative pain. Gerbershagen et al. noted that 22 orthopedic/trauma surgeries on the
INTRODUCTION The global volume of surgical procedures is large, and perioperative pain care is important. A study of 56 World Health Organization member states estimated that annually 234.2 million major surgical operations take place worldwide (95% CI 187.2–281.2 C. Richard Chapman, PhD, is Professor of Anesthesiology, Pain Research Center; Duncan A. Stevens is in the Medical Laboratory Science graduate program; Arthur G. Lipman, PharmD, is University Professor of Pharmacotherapy and Adjunct Professor of Anesthesiology, University of Utah. Support for this work came from a grant from the Department of Anesthesiology at the University of Utah. PAIN-OUT provided the data upon which we have based this report. The authors thank Ruth Zaslansky, DSc, for guidance on the development of this manuscript and Ryan Nielsen, PharmD and Daniel Lee, PharmD, for data collection assistance. Address correspondence to: C. Richard Chapman, Pain Research Center, 615 Arapeen Drive, Suite 200, Salt Lake City, UT 84108, USA. (E-mail: [email protected]).
350
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
C. R. Chapman et al.
extremities ranked among the top 40 most painful operations in a set of 179 classifications.2 This observation is consistent with earlier findings. Kalkman et al. observed that 34% to 84% of patients undergoing a variety of orthopedic procedures in The Netherlands reported moderate to severe postoperative pain.4 Sommer et al. found that 20% to 71% of Dutch orthopedic patients reported moderate or severe pain on the first day following surgery on the extremities as did 30% to 64% of back or spine surgery patients.5 Because orthopedic operations occur frequently and are reliably painful, they appear well suited for cohort studies of postoperative pain. Postoperative pain is a widespread problem in the United States. In 2010, Americans underwent 51.4 million inpatient surgeries according to the federal Centers for Disease Control and Prevention.6 How American patients experience postoperative pain in comparison to patients in other developed nations is largely unknown, but it is now possible to address this question. PAIN-OUT is a newly developed international acute pain registry that operates under the aegis of the International Association for the Study of Pain (http://www.pain-out.eu). Like other registries, PAIN-OUT maintains a large and constantly growing database of variables that characterize postoperative pain and its management. Beginning as a 4-year project sponsored by the European Commission’s 7th Framework Program for Research and Technological Development, PA INOUT is now a free-standing entity supported by its membership.7 Health care institutions register with PAIN-OUT and contribute data obtained from surgical patients. In return, they receive feedback and benchmarking related to other participant institutions that allow them to appraise the relative quality of care they deliver. In addition, they have access to the large international postoperative pain database as well as their own accumulated data for studies that they may wish to undertake. At participating institutions, trained surveyors follow an established protocol and administer a data collection tool, i.e. the International Pain Outcomes Questionnaire (IPOQ).7 This tool exists in 16 languages and is validated in eight. Registry data fall into two broad categories: (1) demographic and clinical care data; and (2) patient-reported outcomes (PRO) data. This paper reports a retrospective cohort study based on the PAIN-OUT database. We compared pooled data on postoperative pain control obtained on the first day following orthopedic surgery in our own and three other American institutions with similar pooled data obtained from an ensemble of 20 institutions in multiple developed European nations. C
2013 Informa Healthcare USA, Inc.
351
American hospitals ubiquitously and consistently assess acute pain in surgical patients because it has been a requirement for hospital accreditation by the Joint Commission for over a decade.8 We postulated that this practice has created a climate of sensitivity to patients’ pain among the clinical staff and a greater readiness to intervene when pain is high than one might find in Europe as a whole where no consistent pain management standards exist for institutional accreditation. If this is the case, we hypothesized that American institutions should demonstrate better control of postoperative pain after orthopedic surgery than European institutions. We also evaluated three secondary predictions: (1) A larger proportion of patients receive opioids on the first day after orthopedic surgery in the United States than in Europe; (2) A greater proportion of American patients receive regional analgesia for postoperative pain management than European patients; and (3) Female patients report more pain and emotional distress than male patients with no difference between American and European samples.
METHODS Data Collection Network The network of data contributors for this study comprised 20 European hospitals in France, Germany, Italy, Spain, Sweden, Switzerland, and the United Kingdom (UK). Four American university hospitals also contributed data. At each participating site, collaborators obtained approval for collecting nonidentified patient data from their local ethics committees or Institutional Review Boards.
Subject Inclusion Criteria Patients included in this study met the following criteria: (1) Underwent orthopedic surgery; (2) were at least 18 years of age or 16 years of age in the UK; (3) had been back on the clinical hospital floor (out of the postanesthesia care unit [PACU]) for at least 6 hours at the time of interview; and (4) agreed to participate in the survey. Exclusion criteria included patients who could not communicate verbally with the surveyor or read a language in which the questionnaire was available. Surveyors did not exclude patients because of gender or advanced age.
Subjects The four participating institutions in the United States contributed data from 473 patients. The 20
352
Journal of Pain & Palliative Care Pharmacotherapy
TABLE 1.
Subset of Patient-Reported Outcome Items from the International Pain Questionnaire.
Patient-reported outcome
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
Pain Worst Pain Least Pain Percent time in severe pain Interference with movement in bed Interference with movement out of bed Percent pain relief Emotional comfort Anxiety Helplessness Satisfaction with care Satisfied with pain management Wanted more treatment Received information about treatment options
Questionnaire item
Scaling
What was the worst pain you had since your surgery? What was the least pain you had since your surgery? How often were you in severe pain since your surgery? How much, since your surgery, has pain interfered with doing activities in bed? How much, since your surgery, has pain interfered with doing activities out of bed? Since your surgery, how much pain relief have you received?
Numerical-rating scale Numerical-rating scale Percentage units Numerical-rating scale
How much, since your surgery, has pain caused you to feel anxious? How much, since your surgery, has pain caused you to feel helpless?
Numerical-rating scale
How satisfied you are with the results of your pain treatment since your surgery? Would you have liked more pain treatment than you received? Did you receive any information about your pain treatment options?
Numerical-rating scale
European sites contributed data from 8799 patients. Each participating site contributed data for at least 40 patients. The percentage of the sample that each European nation contributed is as follows: France (22%), Germany (35%), Italy (8%), Spain (11%), Sweden (7%), Switzerland (14%), and the UK (3%). The mean age for American subjects was 58.3 (±14.3) ranging from 18 to 96 years. For European subjects, the mean age was 55.9 (±17.4) ranging from 18 to 107. In the American sample, 53% of patients were female, and in the European sample 52% were female. This is consistent with our hypothesis that American institutions may have a greater focus on pain than European counterparts, data from patient charts indicated that 99.5% of American patients had their postoperative pain assessed and 75.4% of European patients had their postoperative pain assessed.
Data Collection, Transmission, and Storage Surveyors visited patients in their hospital rooms. Surveyors were nurses, medical residents, medical students, or graduates from other programs in the life sciences or social sciences who had no clinical duties on the wards where they collected data. At our institution, the surveyors were doctorally-trained clinical pharmacists. Surveyors gathered demographic and clinical data from patient charts and administered the IPOQ to patients to obtain the PRO data. The PRO items assess four categories of patient experience: (1) pain; (2) emotional comfort; (3) satisfac-
Numerical-rating scale Percentage units
Numerical-rating scale
Categorical (Yes or No) Categorical (Yes or No)
tion with care; and (4) treatment-related side effects. Some PRO items required scaled responses while others asked patients for categorical Yes or No responses. Table 1 lists the subset of IPOQ PRO items studied in this report. The surveyors at all sites underwent training and followed the PAIN-OUT Standard Operating Procedures to assure consistency across participating sites. In order to minimize interviewer bias, patients completed the IPOQ independently of the surveyor, family members, and the staff. However, if a patient requested assistance or was physically unable to write, the surveyor assisted by reading the questions aloud and writing down the answers. The surveyors or other members of the study team at each institution entered the data from the IPOQs into a web-based portal secured by a password, thus contributing nonidentified patient records into the PAIN-OUT central data repository at the Institute for Medical Informatics, Statistics and Epidemiology at the University of Leipzig, Germany.
Statistical Analysis We performed descriptive and inferential statistical operations on the data using JMP PRO V.10 (SAS Institute, Cary, NC) and Aabel V.3 (Gigawiz Software, Tulsa, OK). Because of the large sample sizes, statistical significance is of minimal importance in this study. When samples are large, it is possible to achieve statistical significance in situations where the observed Journal of Pain & Palliative Care Pharmacotherapy
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
C. R. Chapman et al.
differences are clinically meaningless. In this study, differences between American and European samples may achieve statistical significance even if the difference is so small as to be meaningless clinically. A complementary statistic, the effect size, conveys the estimated magnitude of the effect that one variable exerts upon another, and it defines the substantive, as opposed to statistical, significance of a research result.9,10 We therefore draw our inferences primarily from effect size. Numerous indices of effect size exist, and these vary with the nature of the data and the statistical test employed. Investigators often use effect size indicators to classify outcomes into small, medium, and large effects, although the criteria for such classifications are purely arbitrary. The most familiar effect size indices are Cohen’s d, the standardized mean difference between two groups; R2 , which specifies the amount of variance in one variable for which the other variable can account; and Cram´er’s V, which gauges the strength of association between two categorical values. Odds ratios (ORs) also gauge effect size for categorical variables. Cohen’s d is suitable for contrasts based on the t-test,9 and the conventional criteria for small, medium, and large effect sizes are, respectively, .2, .5, and .8. It is possible to convert most effect sizes into Cohen’s d values, and we do this here to simplify and clarify the outcomes. We report 95% confidence intervals (CIs) for the Cohen’s d values and reason that, if the criterion for a given effect size falls within the d value CI for a given measure, then the outcome for that measure merits that effect size. For example, if Cohen’s d for a given measure falls below the .5 criterion for a medium effect size, but the 95% CI extends beyond .5, then the TABLE 2.
outcome is a medium effect. This classification like all effect size classifications is purely arbitrary, and we offer it simply as a guide for interpreting the outcomes. In general, we regard outcomes with small effect sizes to be academically interesting but not clinically meaningful. There are, however, exceptions. Occasionally, a small effect size is associated with an outcome that clinicians recognize as meaningful. It is important, therefore, to interpret small effect sizes thoughtfully.
RESULTS Preoperative Pain and Medication Differences In the American sample, 80.9% of patients reported that they experienced pain before their orthopedic surgery, but in the European sample only 66.2% reported pre-existing pain. For the Americans, the mean pain intensity (±SD) on an 11-point (zero10) numerical-rating scale (NRS) for this pre-existing pain was 6.1 (±2.2). The European patients reported a mean pre-existing NRS pain intensity of 6.5 (±2.2). Of the American patients reporting pre-existing pain, 32.6% used an opioid medication before admission for surgery. In contrast, 15.7% of European patients used an opioid medication for pre-existing pain prior to orthopedic surgery.
Correlations Among the PRO Measures Table 2 displays the pairwise Pearson correlations among the scaled PRO variables. When the correlation between two variables is high, it is reasonable to assume that they are redundant measures of the same
Pairwise Correlation Coefficients for the Scaled Patient-Reported Outcomes.
Correlations Effect sizes ≥ .10 = small .30 = medium .50 = large Feel part of pain treatment Feel anxious Feel helpless Pain interferes with movement in bed Pain interferes with movement out of bed Worst pain Least pain Percent pain relief Satisfaction with pain treatment Percent of time in severe pain
C
353
Feel Pain Pain part of interferes interferes pain Feel Feel movement movement treatment anxious helpless in bed out of bed
Satisfaction Percent of with Time in pain Severe treatment Pain
Least pain
Percent Pain Relief
1.000 .520 1.000 −.255 −.375 −.287 −.317
1.000 .452
1.000
−.369
−.359
Worst pain
1.000 −.042 −.007 −.073
1.000 .660 .402
1.000 .473
1.000
.001
.471
.507
.620
1.000
−.030 −.060 .102 .277
.403 .384 −.255 −.302
.416 .361 −.265 −.304
.551 .388 −.229 −.2493
.506 .425 −.290 −.293
−.070
2013 Informa Healthcare USA, Inc.
.4358
.411
.408
.505
.555
.591
1.000
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
354
Journal of Pain & Palliative Care Pharmacotherapy
underlying construct. Correlation coefficients in Table 2 exceed r = .6 in two instances. This suggests that the two measures of psychological state, anxiety, and helplessness measure a common latent construct, and the same holds for the two measures of pain interference in and out of bed. Therefore, we combined the highly correlated variables by averaging them to form two composite variables termed Negative Affect and Pain Interference. Figures 1 and 2 illustrate the differences between American and European samples on the scaled PROs. Both present the data from left to right according to effect size. Therefore, the variables that best discriminate the two samples are those to the left sides of the figures. Figure 1 shows the means [± standard errors (SEs)] for the scaled PRO variables, revealing both the direction and magnitude of the observed differences between the American and European samples. Because two PRO measures are scaled in percentage units, we multiplied them by 10 to conform to the 11point NRS scales for the purposes of graphical display in Figure 1. Figure 2 displays the effect sizes as Cohen’s d values with 95% CIs. Table 3 provides further detail including the statistical significance of the contrasts.
FIGURE 1 Mean PRO scores for European (solid) and American (striped) samples. PROs rank from left to right according to effect size. Multiplication of PROs scaled in percentage units by 10 makes them visually comparable to the PROs based on the 11point NRS.
Evaluation of the Primary Prediction We predicted that American patients would have more favorable pain outcomes than Europeans on the scaled PRO measures. For worst pain, the results took the opposite direction of our prediction with European institutions achieving better pain control than American institutions, and the effect size was large. The same direction of outcome was evident for pain interference and least pain, but with small effect sizes as Figure 2 illustrates. On the Negative Affect composite scale, the European patients reported markedly less emotional discomfort than their American counterparts, and the effect size was medium. We observed a barely meaningful difference in satisfaction, which had a small effect size. European patients were slightly more satisfied than American patients. Percent time in severe pain and pain relief failed to discriminate the two samples. The American sample was superior to the European sample on only one-scaled PRO outcome, “Feel a Part of Pain Treatment.” We evaluated the categorical PRO variables with chi-square tests. In response to the question asking whether they had received sufficient information about their treatment, 84.2% of Americans responded affirmatively while 68.0% of European
FIGURE 2 Effect sizes with 95% CI for the scaled PRO.
Journal of Pain & Palliative Care Pharmacotherapy
C. R. Chapman et al. TABLE 3.
Statistical Significance of the Scaled Patient-Reported Outcomes.
Patient-reported outcome
Statistics
United States
Europe
Feel part of pain treatment
Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD
7.75 2.91 3.57 3.04 4.89 3.04 7.39 2.49 2.62 2.20 .66 .25 7.79 2.32 .31 .28
6.10 3.83 2.13 2.57 3.76 2.63 5.40 2.74 1.73 1.80 .68 .26 8.09 2.17 .28 .25
Negative affect Pain interference Worst pain Least pain Pain relief J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
355
Satisfaction with pain treatment Percent of time in severe pain
t
p
−11.5323
ARTICLE
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
Quality of Postoperative Pain Management in American Versus European Institutions C. Richard Chapman, Duncan A. Stevens, and Arthur G. Lipman AB STRACT Management of postoperative pain remains an important clinical problem throughout the world. Using the PAINOUT acute pain registry database to examine perioperative pain management in orthopedic surgery patients, we compared patient-reported outcomes (PROs) in a pooled sample obtained from four American hospitals (N = 473) with PROs in a pooled sample of 20 European institutions (N = 8799). Most American hospitals consistently assess acute pain in surgical patients due to Joint Commission accreditation guidelines. Therefore, we hypothesized that this practice would create a climate of clinical staff sensitivity to patients’ pain and a greater readiness to intervene when pain is higher than one would find in Europe as a whole. American institutions might then provide better control of postoperative pain after orthopedic surgery than European institutions. Because of the large sample sizes, our analyses focused on effect size rather than statistical significance. Evaluation of the pain PROs revealed that European patients reported much lower Worst Pain on the first day after orthopedic surgery than American patients. The mean Worst Pain (± SD) for Europeans was 5.4 (2.5) but for Americans the mean was 7.4 (2.7), p < .0001, a large effect size. Europeans also reported significantly less emotional discomfort, less interference of pain with activity and lower Least Pain. Nonetheless, 98.3% of American patients received opioids on the ward on the first postoperative day compared to 70.2% of European patients, and 41.1% received regional analgesia on the ward while 15.9% of European patients received regional analgesia (both small effect sizes). Overall, the results are clear in demonstrating much better pain control in the ensemble of European countries as compared to the United States. KEYWORDS acute, perioperative, postoperative, pain, PAIN-OUT, patient-reported outcomes, Europe, USA
million). This translates into about one surgery each year for every 25 people.1 Unfortunately, postoperative pain remains a common problem despite notable advances in the scientific understanding of pain in recent decades. A review of data on postoperative pain on the first day following surgery in 105 German hospitals, 2 found that high pain scores can result from almost all surgical procedures, including those that appear minimally traumatic. Poorly controlled pain can contribute to postoperative complications, impede recovery, extend hospitalization, and cause needless suffering. Importantly, unresolved acute postoperative pain can progress to debilitating chronic pain. Kehlet et al. estimated that unresolved acute pain after surgery leads to chronic postoperative pain in 10% to 50% of patients after common operations.3 Orthopedic surgeries are among the major generators of intense postoperative pain. Gerbershagen et al. noted that 22 orthopedic/trauma surgeries on the
INTRODUCTION The global volume of surgical procedures is large, and perioperative pain care is important. A study of 56 World Health Organization member states estimated that annually 234.2 million major surgical operations take place worldwide (95% CI 187.2–281.2 C. Richard Chapman, PhD, is Professor of Anesthesiology, Pain Research Center; Duncan A. Stevens is in the Medical Laboratory Science graduate program; Arthur G. Lipman, PharmD, is University Professor of Pharmacotherapy and Adjunct Professor of Anesthesiology, University of Utah. Support for this work came from a grant from the Department of Anesthesiology at the University of Utah. PAIN-OUT provided the data upon which we have based this report. The authors thank Ruth Zaslansky, DSc, for guidance on the development of this manuscript and Ryan Nielsen, PharmD and Daniel Lee, PharmD, for data collection assistance. Address correspondence to: C. Richard Chapman, Pain Research Center, 615 Arapeen Drive, Suite 200, Salt Lake City, UT 84108, USA. (E-mail: [email protected]).
350
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
C. R. Chapman et al.
extremities ranked among the top 40 most painful operations in a set of 179 classifications.2 This observation is consistent with earlier findings. Kalkman et al. observed that 34% to 84% of patients undergoing a variety of orthopedic procedures in The Netherlands reported moderate to severe postoperative pain.4 Sommer et al. found that 20% to 71% of Dutch orthopedic patients reported moderate or severe pain on the first day following surgery on the extremities as did 30% to 64% of back or spine surgery patients.5 Because orthopedic operations occur frequently and are reliably painful, they appear well suited for cohort studies of postoperative pain. Postoperative pain is a widespread problem in the United States. In 2010, Americans underwent 51.4 million inpatient surgeries according to the federal Centers for Disease Control and Prevention.6 How American patients experience postoperative pain in comparison to patients in other developed nations is largely unknown, but it is now possible to address this question. PAIN-OUT is a newly developed international acute pain registry that operates under the aegis of the International Association for the Study of Pain (http://www.pain-out.eu). Like other registries, PAIN-OUT maintains a large and constantly growing database of variables that characterize postoperative pain and its management. Beginning as a 4-year project sponsored by the European Commission’s 7th Framework Program for Research and Technological Development, PA INOUT is now a free-standing entity supported by its membership.7 Health care institutions register with PAIN-OUT and contribute data obtained from surgical patients. In return, they receive feedback and benchmarking related to other participant institutions that allow them to appraise the relative quality of care they deliver. In addition, they have access to the large international postoperative pain database as well as their own accumulated data for studies that they may wish to undertake. At participating institutions, trained surveyors follow an established protocol and administer a data collection tool, i.e. the International Pain Outcomes Questionnaire (IPOQ).7 This tool exists in 16 languages and is validated in eight. Registry data fall into two broad categories: (1) demographic and clinical care data; and (2) patient-reported outcomes (PRO) data. This paper reports a retrospective cohort study based on the PAIN-OUT database. We compared pooled data on postoperative pain control obtained on the first day following orthopedic surgery in our own and three other American institutions with similar pooled data obtained from an ensemble of 20 institutions in multiple developed European nations. C
2013 Informa Healthcare USA, Inc.
351
American hospitals ubiquitously and consistently assess acute pain in surgical patients because it has been a requirement for hospital accreditation by the Joint Commission for over a decade.8 We postulated that this practice has created a climate of sensitivity to patients’ pain among the clinical staff and a greater readiness to intervene when pain is high than one might find in Europe as a whole where no consistent pain management standards exist for institutional accreditation. If this is the case, we hypothesized that American institutions should demonstrate better control of postoperative pain after orthopedic surgery than European institutions. We also evaluated three secondary predictions: (1) A larger proportion of patients receive opioids on the first day after orthopedic surgery in the United States than in Europe; (2) A greater proportion of American patients receive regional analgesia for postoperative pain management than European patients; and (3) Female patients report more pain and emotional distress than male patients with no difference between American and European samples.
METHODS Data Collection Network The network of data contributors for this study comprised 20 European hospitals in France, Germany, Italy, Spain, Sweden, Switzerland, and the United Kingdom (UK). Four American university hospitals also contributed data. At each participating site, collaborators obtained approval for collecting nonidentified patient data from their local ethics committees or Institutional Review Boards.
Subject Inclusion Criteria Patients included in this study met the following criteria: (1) Underwent orthopedic surgery; (2) were at least 18 years of age or 16 years of age in the UK; (3) had been back on the clinical hospital floor (out of the postanesthesia care unit [PACU]) for at least 6 hours at the time of interview; and (4) agreed to participate in the survey. Exclusion criteria included patients who could not communicate verbally with the surveyor or read a language in which the questionnaire was available. Surveyors did not exclude patients because of gender or advanced age.
Subjects The four participating institutions in the United States contributed data from 473 patients. The 20
352
Journal of Pain & Palliative Care Pharmacotherapy
TABLE 1.
Subset of Patient-Reported Outcome Items from the International Pain Questionnaire.
Patient-reported outcome
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
Pain Worst Pain Least Pain Percent time in severe pain Interference with movement in bed Interference with movement out of bed Percent pain relief Emotional comfort Anxiety Helplessness Satisfaction with care Satisfied with pain management Wanted more treatment Received information about treatment options
Questionnaire item
Scaling
What was the worst pain you had since your surgery? What was the least pain you had since your surgery? How often were you in severe pain since your surgery? How much, since your surgery, has pain interfered with doing activities in bed? How much, since your surgery, has pain interfered with doing activities out of bed? Since your surgery, how much pain relief have you received?
Numerical-rating scale Numerical-rating scale Percentage units Numerical-rating scale
How much, since your surgery, has pain caused you to feel anxious? How much, since your surgery, has pain caused you to feel helpless?
Numerical-rating scale
How satisfied you are with the results of your pain treatment since your surgery? Would you have liked more pain treatment than you received? Did you receive any information about your pain treatment options?
Numerical-rating scale
European sites contributed data from 8799 patients. Each participating site contributed data for at least 40 patients. The percentage of the sample that each European nation contributed is as follows: France (22%), Germany (35%), Italy (8%), Spain (11%), Sweden (7%), Switzerland (14%), and the UK (3%). The mean age for American subjects was 58.3 (±14.3) ranging from 18 to 96 years. For European subjects, the mean age was 55.9 (±17.4) ranging from 18 to 107. In the American sample, 53% of patients were female, and in the European sample 52% were female. This is consistent with our hypothesis that American institutions may have a greater focus on pain than European counterparts, data from patient charts indicated that 99.5% of American patients had their postoperative pain assessed and 75.4% of European patients had their postoperative pain assessed.
Data Collection, Transmission, and Storage Surveyors visited patients in their hospital rooms. Surveyors were nurses, medical residents, medical students, or graduates from other programs in the life sciences or social sciences who had no clinical duties on the wards where they collected data. At our institution, the surveyors were doctorally-trained clinical pharmacists. Surveyors gathered demographic and clinical data from patient charts and administered the IPOQ to patients to obtain the PRO data. The PRO items assess four categories of patient experience: (1) pain; (2) emotional comfort; (3) satisfac-
Numerical-rating scale Percentage units
Numerical-rating scale
Categorical (Yes or No) Categorical (Yes or No)
tion with care; and (4) treatment-related side effects. Some PRO items required scaled responses while others asked patients for categorical Yes or No responses. Table 1 lists the subset of IPOQ PRO items studied in this report. The surveyors at all sites underwent training and followed the PAIN-OUT Standard Operating Procedures to assure consistency across participating sites. In order to minimize interviewer bias, patients completed the IPOQ independently of the surveyor, family members, and the staff. However, if a patient requested assistance or was physically unable to write, the surveyor assisted by reading the questions aloud and writing down the answers. The surveyors or other members of the study team at each institution entered the data from the IPOQs into a web-based portal secured by a password, thus contributing nonidentified patient records into the PAIN-OUT central data repository at the Institute for Medical Informatics, Statistics and Epidemiology at the University of Leipzig, Germany.
Statistical Analysis We performed descriptive and inferential statistical operations on the data using JMP PRO V.10 (SAS Institute, Cary, NC) and Aabel V.3 (Gigawiz Software, Tulsa, OK). Because of the large sample sizes, statistical significance is of minimal importance in this study. When samples are large, it is possible to achieve statistical significance in situations where the observed Journal of Pain & Palliative Care Pharmacotherapy
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
C. R. Chapman et al.
differences are clinically meaningless. In this study, differences between American and European samples may achieve statistical significance even if the difference is so small as to be meaningless clinically. A complementary statistic, the effect size, conveys the estimated magnitude of the effect that one variable exerts upon another, and it defines the substantive, as opposed to statistical, significance of a research result.9,10 We therefore draw our inferences primarily from effect size. Numerous indices of effect size exist, and these vary with the nature of the data and the statistical test employed. Investigators often use effect size indicators to classify outcomes into small, medium, and large effects, although the criteria for such classifications are purely arbitrary. The most familiar effect size indices are Cohen’s d, the standardized mean difference between two groups; R2 , which specifies the amount of variance in one variable for which the other variable can account; and Cram´er’s V, which gauges the strength of association between two categorical values. Odds ratios (ORs) also gauge effect size for categorical variables. Cohen’s d is suitable for contrasts based on the t-test,9 and the conventional criteria for small, medium, and large effect sizes are, respectively, .2, .5, and .8. It is possible to convert most effect sizes into Cohen’s d values, and we do this here to simplify and clarify the outcomes. We report 95% confidence intervals (CIs) for the Cohen’s d values and reason that, if the criterion for a given effect size falls within the d value CI for a given measure, then the outcome for that measure merits that effect size. For example, if Cohen’s d for a given measure falls below the .5 criterion for a medium effect size, but the 95% CI extends beyond .5, then the TABLE 2.
outcome is a medium effect. This classification like all effect size classifications is purely arbitrary, and we offer it simply as a guide for interpreting the outcomes. In general, we regard outcomes with small effect sizes to be academically interesting but not clinically meaningful. There are, however, exceptions. Occasionally, a small effect size is associated with an outcome that clinicians recognize as meaningful. It is important, therefore, to interpret small effect sizes thoughtfully.
RESULTS Preoperative Pain and Medication Differences In the American sample, 80.9% of patients reported that they experienced pain before their orthopedic surgery, but in the European sample only 66.2% reported pre-existing pain. For the Americans, the mean pain intensity (±SD) on an 11-point (zero10) numerical-rating scale (NRS) for this pre-existing pain was 6.1 (±2.2). The European patients reported a mean pre-existing NRS pain intensity of 6.5 (±2.2). Of the American patients reporting pre-existing pain, 32.6% used an opioid medication before admission for surgery. In contrast, 15.7% of European patients used an opioid medication for pre-existing pain prior to orthopedic surgery.
Correlations Among the PRO Measures Table 2 displays the pairwise Pearson correlations among the scaled PRO variables. When the correlation between two variables is high, it is reasonable to assume that they are redundant measures of the same
Pairwise Correlation Coefficients for the Scaled Patient-Reported Outcomes.
Correlations Effect sizes ≥ .10 = small .30 = medium .50 = large Feel part of pain treatment Feel anxious Feel helpless Pain interferes with movement in bed Pain interferes with movement out of bed Worst pain Least pain Percent pain relief Satisfaction with pain treatment Percent of time in severe pain
C
353
Feel Pain Pain part of interferes interferes pain Feel Feel movement movement treatment anxious helpless in bed out of bed
Satisfaction Percent of with Time in pain Severe treatment Pain
Least pain
Percent Pain Relief
1.000 .520 1.000 −.255 −.375 −.287 −.317
1.000 .452
1.000
−.369
−.359
Worst pain
1.000 −.042 −.007 −.073
1.000 .660 .402
1.000 .473
1.000
.001
.471
.507
.620
1.000
−.030 −.060 .102 .277
.403 .384 −.255 −.302
.416 .361 −.265 −.304
.551 .388 −.229 −.2493
.506 .425 −.290 −.293
−.070
2013 Informa Healthcare USA, Inc.
.4358
.411
.408
.505
.555
.591
1.000
J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
354
Journal of Pain & Palliative Care Pharmacotherapy
underlying construct. Correlation coefficients in Table 2 exceed r = .6 in two instances. This suggests that the two measures of psychological state, anxiety, and helplessness measure a common latent construct, and the same holds for the two measures of pain interference in and out of bed. Therefore, we combined the highly correlated variables by averaging them to form two composite variables termed Negative Affect and Pain Interference. Figures 1 and 2 illustrate the differences between American and European samples on the scaled PROs. Both present the data from left to right according to effect size. Therefore, the variables that best discriminate the two samples are those to the left sides of the figures. Figure 1 shows the means [± standard errors (SEs)] for the scaled PRO variables, revealing both the direction and magnitude of the observed differences between the American and European samples. Because two PRO measures are scaled in percentage units, we multiplied them by 10 to conform to the 11point NRS scales for the purposes of graphical display in Figure 1. Figure 2 displays the effect sizes as Cohen’s d values with 95% CIs. Table 3 provides further detail including the statistical significance of the contrasts.
FIGURE 1 Mean PRO scores for European (solid) and American (striped) samples. PROs rank from left to right according to effect size. Multiplication of PROs scaled in percentage units by 10 makes them visually comparable to the PROs based on the 11point NRS.
Evaluation of the Primary Prediction We predicted that American patients would have more favorable pain outcomes than Europeans on the scaled PRO measures. For worst pain, the results took the opposite direction of our prediction with European institutions achieving better pain control than American institutions, and the effect size was large. The same direction of outcome was evident for pain interference and least pain, but with small effect sizes as Figure 2 illustrates. On the Negative Affect composite scale, the European patients reported markedly less emotional discomfort than their American counterparts, and the effect size was medium. We observed a barely meaningful difference in satisfaction, which had a small effect size. European patients were slightly more satisfied than American patients. Percent time in severe pain and pain relief failed to discriminate the two samples. The American sample was superior to the European sample on only one-scaled PRO outcome, “Feel a Part of Pain Treatment.” We evaluated the categorical PRO variables with chi-square tests. In response to the question asking whether they had received sufficient information about their treatment, 84.2% of Americans responded affirmatively while 68.0% of European
FIGURE 2 Effect sizes with 95% CI for the scaled PRO.
Journal of Pain & Palliative Care Pharmacotherapy
C. R. Chapman et al. TABLE 3.
Statistical Significance of the Scaled Patient-Reported Outcomes.
Patient-reported outcome
Statistics
United States
Europe
Feel part of pain treatment
Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD
7.75 2.91 3.57 3.04 4.89 3.04 7.39 2.49 2.62 2.20 .66 .25 7.79 2.32 .31 .28
6.10 3.83 2.13 2.57 3.76 2.63 5.40 2.74 1.73 1.80 .68 .26 8.09 2.17 .28 .25
Negative affect Pain interference Worst pain Least pain Pain relief J Pain Palliat Care Pharmacother Downloaded from informahealthcare.com by McMaster University on 10/23/14 For personal use only.
355
Satisfaction with pain treatment Percent of time in severe pain
t
p
−11.5323
