Acute pelvic pain is one of the most frequent reasons for consultation in emergency gynecologic units.1,2 Severe acute pelvic pain may sometimes be related to serious life-threatening gynecologic events, such as ruptured ectopic pregnancy,3,4 massive hemoperitoneum,5 salpingitis,6 or adnexal torsion,7 and nongynecologic events such as acute appendicitis.8 Appropriate evaluation of the pain intensity may thus help management of acute pelvic pain in gynecology.
Unfortunately, there are at present no precise recommendations for measuring the intensity of acute pelvic pain in the context of gynecologic emergencies. Assessment of the spontaneous pain intensity felt by patients is usually based on evaluation by the clinician, who usually quotes the pain as absent, present, or severe according to his own subjective impression.5,9
Standardized evaluation of the intensity of acute pain is widely used in the context of general emergency cases, and guidelines propose methods to assess this pain10–13: the pain intensity may be generally evaluated by the patient herself, through one-dimensional self-report intensity scales that include verbal rating scale, numerical rating scale14 or visual analog scale (VAS).15 However, some clinicians may be reluctant to use self-report scales because they view them as essentially subjective, elusive, and complex undertakings. Another way to measure pain intensity may be third-party assessment (or clinician rating assessment) based on observation by the medical staff of pain-related behavior (movement, facial expression, posturing, etc) and/or physiologic indicators (heart rate, blood pressure, respiratory rate, etc). Measuring pain intensity with scales based on pain behavior has proven to be feasible and valid in the context of chronic pain16 and in the context of pain delivery.17 In the context of emergency, measuring acute pain intensity with the use of pain-related behavior and/or physiologic indicators had been suggested,10,18 but structured behavioral scales of this type have been only poorly evaluated.19
The aim of this prospective observational study was to compare and assess the value of five methods for measuring pelvic pain intensity in a gynecologic emergency unit: three self-report pain intensity scales: VAS, numerical rating scale, and verbal rating scale, and two structured behavioral indices, one of which was specifically developed for assessment of acute pelvic pain in gynecologic emergency.
MATERIALS AND METHODS
All consecutive female patients aged older than 13 years presenting with or without acute pelvic pain at the gynecologic emergency unit of the Poissy-St. Germain Hospital Center from December 2002 to February 2003 were enrolled in the study. Exclusion criteria were 1) acute pain from locations other than pelvic, including perineal pain (eg, bartholinitis and skenitis) and breast pathologies; 2) pregnancy of more than 20 gestational weeks; 3) patients known to have pathologies responsible for severe chronic pelvic pain (ie, severe endometriosis).
Pain intensity measurements were performed at the beginning of the consultation before any analgesic intake. Three self-report scales were successively used in the same order for all patients: a VAS,15,20,21 an 11-point numerical rating scale,21–23 and a five-level verbal rating scale.14,20 A written form was used for all patients12 (Appendix).
In addition, two indices based on behavioral items were also used to assess the intensity of the pain indirectly. The first behavioral index (BI-1) was the one previously developed for assessing pain intensity in a context of general emergency.18,19 BI-1 consists of four different items that are behavioral criteria of the manifestation of pain: vocal complaints, facial expression, movement, and posturing. Each criterion was rated by the medical staff as absent (0), mild (1), or marked (2) (Appendix). The pain intensity score was then calculated by adding all four ratings.
The second behavioral index (BI-2) was specifically developed for the present study. It was adapted from a previously developed behavior scale that was validated in a context of chronic pain.16 It is based on eight different criteria, including pain-related behaviors (some of which were similar to those previously used for BI-1), to which we added physiologic indicators of acute pain (neurovegetative signs and anxiety)24 (Appendix). The score for pain intensity measured with BI-2 was then calculated by adding all eight ratings. The different behavioral items used in the two indices were collected at the beginning of the examination by a member of the medical staff (medical student, resident in family medicine, resident in gynecology, or senior consultant).
The relevant final diagnosis was the diagnosis at the time the patient was discharged from hospital. Final diagnoses were obtained according to the results of the different examinations performed during the consultation or hospitalization (if needed), that is, clinical examination, abdominal and transvaginal ultrasonography, routine biology (if needed), and computed abdominal tomography (if needed). Some diagnoses were based on surgical evaluation (ie, laparoscopy, dilatation, and curettage or diagnostic hysteroscopy). The diagnosis of ectopic pregnancy was based either on laparoscopic evaluation or on an algorithm.25,26 Laparoscopy was performed when the ectopic pregnancy was thought to be complicated (ie, abundant hemoperitoneum with active bleeding and/or tubal rupture). The diagnosis of salpingitis was based either on the laparoscopic evaluation, if it was thought to be necessary, or according to noninvasive diagnostic models.6,27 Some other diagnoses were also based on the laparoscopic evaluation, including abundant hemoperitoneum related to a ruptured cyst, adnexal torsion, and appendicitis.
The diagnosis of idiopathic acute pelvic pain was made when examination of the women was normal and the evolution benign, with appropriate analgesia at the end of the consultation. After discharge, there was no specific follow-up, but women were encouraged to return in case of recurrence of the pain.
For each patient with pain, the presumed mechanism of the pain was classified as one of the following: peritonitis, ischemia, colic, stretch or distension, infection, or other.28,29 When several mechanisms could apply for the same patient, we selected the one presumed to give the greatest pain according to the neurophysiology. For example, in the case of salpingitis, either infection or peritonitis may provoke the pain; given that peritonitis is acknowledged to be the most painful mechanism, we chose to state peritonitis.
Life-threatening events were those diagnoses with potentially fatal evolution in the absence of appropriate therapy. They included appendicitis, acute renal colic, acute pyelonephritis, abundant hemoperitoneum (ruptured ectopic pregnancy or ruptured corpus luteum), salpingitis, and adnexal torsion.
Factor analysis was used to study the dimensional aspect of the behavioral indices. The scree plot was used to ensure that each of the two behavioral scales was unidimensional.30 A scree plot represents graphically the eigenvalues of the correlation matrix of a set of items. These eigenvalues represent the relative importance of the dimensions of these items. If a single eigenvalue seems substantially higher than the others, the set of items may be considered as unidimensional. Factor analysis loadings were then computed; the extraction method was principal component analysis with varimax rotation.31 Cronbach’s alpha coefficient was used to measure the internal consistency of each of the two behavioral scales,32 a coefficient alpha of more than 0.70 is considered to be acceptable, and more than 0.8 is considered to be good.33
The content validity of the five methods was then assessed by the use of a new factor analysis. This analysis was performed with the five scales to ensure that they measured a single underlying dimension: the pain intensity. Cronbach’s alpha coefficient was then used to ensure that the five scales had a high internal consistency, in other words that the measurement error was small.
Descriptive statistics were used to judge the intrinsic statistical quality of each of the five scales (mean, standard deviation, number of missing data items, number of levels, saturation, etc). The Shapiro-Wilk test was used to evaluate the normality of the distribution of the scales. To allow comparison between the scales, each of them was standardized on a range from 0 to 100 by the following formula: standardized value=raw score for the scale/maximal possible raw score×100.
The predictive validity of each of the methods of pain intensity measurement was assessed as follows12: for each scale, the intensity of pain according to different groups of patients known to have high or low pain intensity scores were compared. For example, we hypothesized that a patient consulting for pain would have higher pain intensity than a patient consulting for other symptoms. Another example is that a patient suffering from a disease known to be responsible for severe pain would have a higher pain intensity score than the others. The diseases considered to be responsible for severe pain were those responsible for ischemia, colic, or peritoneal irritation.28 Pain scores were thus compared according to several factors thought to affect the pain intensity. These factors included the pathology finally diagnosed, the anatomic location of the disease, the autonomic pathways involved, the mechanism of the pain,28 the prognosis of the disease (ie, life-threatening event or not), and also demographic characteristics. Comparisons were made using nonparametric tests (Kruskal-Wallis test and Wilcoxon Mann-Whitney test).
We then compared the five methods graphically by plotting the value of the different scales according to the mechanism and to the anatomic location of the disease. The five scales were also compared by calculating the effect size of different variables that were found to significantly affect any of the pain scores. Effect sizes were computed as Cohen’s d, where a positive effect size represents an increasing pain score due to the factor.34 An effect size of 0.2 is small, 0.5 medium, and above 0.8 large.35 Ninety-five percent confidence intervals (CIs) were computed. For each factor studied, comparisons between scales were made based on the confidence intervals of the mean effect size. If the mean of one scale was not included within the 95% CI of one other scale, then the two scales differed significantly at P<.05.
The sample size was justified according the following arguments: considering the instrument of interest consists of eight items, 175 subjects lead to a half-range of a 95% CI of a Cronbach alpha equal to 0.075 (for an alpha=0.7) and to a half-range of a 95% CI of principal components loading of approximately 0.1.36 This last result may also be used to be confident in the assessment of the internal consistency of the set of the five pain intensity measurement methods. A P<.05 was considered as significant. The results were analyzed using StatView 5.0 software (SAS Institute Inc., Cary, NC).
All subjects gave informed consent. Our work complied with French statutes and regulations, which authorize epidemiologic surveys without advance approval of an ethics committee. Our survey involved no intervention and is thus excluded from the French statute on biomedical research (Huriet-Serusclat law, dated December 20, 1998). We complied with all French statutes concerning data about the subjects, confidentiality, and restrictions (eg, no religious or racial data). It was submitted to and approved by the Ethics Committee of the French National College of Gynecologists and Obstetricians.
During the study period, 177 patients met the inclusion criteria and underwent pain evaluation. The main characteristics of the study population and the final diagnoses are given in Table 1. The main diagnoses were complications of intrauterine pregnancy (miscarriage, bleeding, or other) in 72 cases (40.7%), pain related to menstrual cycle or idiopathic acute pelvic pain in 29 cases (16.4%). Nine patients had ectopic pregnancy (5.1%), nine had pelvic inflammatory disease (5.1%), two had appendicitis (1.1%), and one had adnexal torsion (0.6%).
All four items of BI-1 were unidimensional according to the scree plot (data not shown). The first dimension explains 61.9% of the total variance. The loadings for each of the four items were all more than 0.7. Cronbach’s alpha coefficient was 0.80, which is “good.” The BI-1 result was thus constructed by adding the scores for all four criteria. All eight items of BI-2 were also unidimensional according to the scree plot (data not shown). The first dimension explains 49.4% of the total variance. The loadings for each of the eight items were all more than 0.5. Cronbach’s alpha coefficient was 0.84, which is “good.” The BI-2 result was thus constructed by adding the scores for all eight criteria.
Visual analog scale, numerical rating scale, verbal rating scale, BI-1, and BI-2 were highly correlated, with Pearson’s correlations ranging from 0.54 to 0.94 (all P<.001). Because a single eigenvalue seemed substantially higher the others, the five scales could be considered as a unidimensional set according to the scree plot (Fig. 1). The first dimension explains 76.9% of the total variance. The loadings for each of the five scales were 0.93, 0.91, 0.90, 0.83, and 0.81 for VAS, numerical rating scale, VRS, BI-1, and BI-2, respectively. Cronbach’s alpha coefficient was 0.89.
The intrinsic statistical properties of the five scales are given in Table 2. Missing data rates were greater for the two behavioral scales than for the self- report scales (all had more than 99% response rate). The BI-2 had the greatest missing response rate (Table 2). The number of distinct levels varied from 74 for the VAS to five for the verbal rating scale. None of the scales showed a normal distribution except the verbal rating scale (Shapiro-Wilk test for normal data less than 0.05 for VAS, numerical rating scale, BI-1, and BI-2). Compared with the self-report scales, the behavioral scales were shifted toward nonpainful scores; the standardized median scores were lower for the behavioral scales than for the self-report scales (Table 2, Fig. 2). When the behavioral scales were used, the raw scores were less frequently below the midpoint of the scale, and the percentage of patients with maximal scoring was lower. The verbal rating scale had the highest percentage of women scoring maximum, whereas BI-2 had no patient with a maximal score (Table 2).
Whatever the method of measurement used, the pain intensity scores were related to pain physiology (Fig. 2), life-threatening event, anatomic location (Fig. 2), main complaints, time of consultation, and the presence of an accompanying person (data not shown). Variations of pain intensity according to the pain physiology or the pain location showed very similar patterns whatever the method of measurement used (Fig. 2).
Whatever the method used, pelvic pain intensity did not vary according to age, occupational category, parity, or past history of severe pain (data not shown). Pain intensity did not vary according to geographic origin except with the numerical rating scale (higher score for non–European Economic Community patients P=.048).
Pain intensity was related to the existence of an ongoing pregnancy and to the autonomic pathways involved when self-report scales were used but not when the behavioral scales were used (data not shown). The behavioral scores were higher in divorced women than others, but their self-report scores were similar.
The five methods were all sensitive to the physiology of the pain, the location and the severity of the illness, and pain as a main complaint with large effect sizes (Table 3). The effect sizes of the behavioral scales tended to be larger than those of the self-report scales for pain physiology and the severity of the illness, whereas the self-report scales were greatest for pain as a main complaint. The self-report scales but not the behavioral indices were sensitive to ongoing pregnancy, whereas the behavioral scales, but not the self-report scales, were sensitive to the fact of being a divorcee (Table 3).
The receiver operating characteristic curve area for predicting life-threatening events were 0.81, 0.78, 0.76, 0.80 and 0.79 for VAS, numerical rating scale, verbal rating scale, BI-1, and BI-2 respectively (no significant difference).
The five methods proposed here have been shown to measure pelvic pain intensity in a gynecologic emergency department with good predictive ability. Indeed, all methods of measurement were highly sensitive to the physiology and anatomy of the pain and severity of the illness, whereas demographic variables had little effect on the score of any of the pain scales. The pain behavior scales were highly concordant with the self-report scales, although slight differences do exist between them. The behavioral scales underestimated the pain intensity compared with the self-report assessment. The behavioral scales were not as easy to use and had more missing values than the self-report scales.
This study compared methods of measurement of acute pain intensity in a gynecologic emergency department. The study does have important limitations, however. First, we did not assess interobserver or intraobserver variability. In the cases of the self-assessment methods, this variability is obviously limited, because a written form was used to evaluate all patients. In the case of the behavioral indices interobserver or intraobserver variability might be of some importance because these require appreciation by the physician of the different criteria used to construct the index. However, in the context of chronic pain, the use of an index similar to ours has shown an excellent interrater reliability.16
Second, we did not assess sensitivity to change. High sensitivity of the various scales to appropriate analgesic intake or treatment of the pathology would be an important marker of the efficiency of the method to measure pelvic pain intensity adequately.37 However, self-assessment methods12,20 and also behavioral methods16,19 have proven to be sensitive to the short-term changes in pain intensity associated with emergency care.10–12
The principal component analysis indicated that the five scales each provided a concordant measure of the pelvic pain intensity. Furthermore, each of the scales provided a similar and substantial ability to detect differences on external criteria known to relate with pain intensity, such as anatomic location of the disease, pain mechanism, and unfavorable evolution. Conversely, demographic variables such as age, geographic origin, family situation and parity, and occupational category had little influence on the pain scores whatever the method used. Thus, our results show unambiguously that the behavioral methods and the self-report methods have excellent concordance for assessment of pelvic pain intensity with respect to the physiologic and neuroanatomic aspects of the pain. Previous studies, either on chronic pain16 or on acute pain,17,19 did not permit any conclusive evidence to be drawn of such a correlation between pain behavior and self-report methods.
Self-report assessments are generally considered as the criterion standard for measurement of the actual pain intensity in the emergency department.10–13 However, because skepticism is a common response of the physician confronted by a patient’s pain report, attempts to objectify the pain experience are sought.12,16 The new element provided by our study is that the predictive validity of simple instruments like self-report scales is the same as more complex instruments that aim to measure the external manifestations of pain “objectively.”
Another point is that the relationships between the physiology of pelvic pain and pain intensity suggest that measuring the pain intensity, whatever the method used, might help to detect serious, life-threatening gynecologic or nongynecologic conditions in the emergency room. Indeed, the intensity of pelvic pain was found to be severe in ruptured ectopic pregnancy,4 in abundant hemoperitoneum,5 in adnexal torsion,28,38 in urinary, or in digestive tract infections that may have serious consequences.28,39 Unfortunately, none of these earlier studies used a standardized method to assess pain intensity. Although in the present study patients with severe illness (ie, with appendicitis, ruptured ectopic pregnancy, adnexal torsion, etc) were uncommon, our results strongly suggest integrating pain intensity as a “vital sign” within the gynecologic emergency department at the time of triage.12
The advantages of behavioral scales over self-report scales remain unclear as yet. Behavioral measurements were more sensitive to severe gynecologic or nongynecologic conditions, whereas self-report scales were more sensitive to some mild aspects of the painful condition, such as the existence of an ongoing pregnancy and pain as the reason for consultation. This suggests that behavioral and self-report measurements of pain do not measure exactly the same aspect of the pain experience. When looking at the way the behavioral scales have been built, it seems obvious that the behavioral indices will relate more to the somatic aspect of the pain experience. Certain diseases might indeed interfere directly with some of the criteria thought to be behavioral manifestations of the pain itself. For example, a vasovagal reflex might be related to abundant hemoperitoneum40,41; nausea and vomiting are common in cases of appendicitis or adnexal torsion.8,42
What it is important to note is that the spectrum of pain intensity discrimination is quite different between self-report and behavioral methods. Indeed, the same patients had generally lower scores with the behavioral scales than with the self-report ones, which means that the former tend to minimize the pain compared with the latter.19 This suggests that the self-report scales could be better at discriminating mild and moderate pain, whereas behavioral scales could be better at discriminating severe pain.
Ease of administration and scoring is important, because scale users are often short of time and resources. From this point view, measuring the pain intensity with the behavioral scales was found to be cumbersome, time consuming, and requiring the involvement of medical staff. This may explain the rate of missing data for the behavioral scales, which was unacceptable compared with the self-report scales. It should be noted that the rate of missing data was more marked for BI-2 than for BI-1, which obviously demonstrates that the rate of missing data is related to the number of items. From this point of view, it is reasonable to ask whether all eight items in the BI-2 are really pertinent and to question its construction. A study19 performed in a general emergency department compared self-report scales (VAS and numerical rating scale) to a behavioral scale similar to the BI-1 and drew conclusions similar to ours: The use of behavioral scales was feasible in the emergency room, but the self-report scales were easier for administration of scoring.
The present study suggests that self-report measurements do have adequate qualities to be selected for routine use in a gynecology as in other emergency rooms.10,11,18 The VAS is theoretically the most convenient method to subjectively assess pain intensity. The VAS does not force quantum changes in pain intensity as is the case with ordinal scales.43 It provides an infinite number of possible response categories, whereas the numerical rating scale provides only 11, and the verbal rating scale provides five categories. The number of categories does not necessarily improve the sensitivity of the instrument,21 but a sufficient number of responses is considered to be necessary; this number might be as high as 20 subdivisions of pain intensity.15 In this respect, the verbal rating scale, with only five descriptors, was not fine enough in our experience and led to an unacceptable percentage of patients scoring maximum compared with the other pain scores. Although VAS was more discriminating than the numerical rating scale, the latter showed sufficient precision.
In conclusion, all methods measured the pelvic pain intensity in gynecologic emergencies adequately, ie, they were sensitive to physiologic and anatomic aspects of the pain mechanism, whereas sociodemographic variables had little influence on pain scores. The measurements obtained with the pain behavior scales were concordant with those obtained with the self-report scales, but the former were not as easy to administer and score than the latter. Among the self-report scales, VAS and the numerical rating scale seem to be the most convenient for routine use, whereas the verbal rating scale has insufficient response categories. The advantages of behavioral scales over self-report scales remain unclear as yet. Self-report measurements do indeed have adequate qualities to be selected for routine use in a gynecologic emergency.
1. Kontoravdis A, Chryssikopoulos A, Hassiakos D, Liapis A, Zourlas PA. The diagnostic value of laparoscopy in 2365 patients with acute and chronic pelvic pain. Int J Gynaecol Obstet 1996;52:243–8.
2. Walker JW. Abdominal and pelvic pain. Emerg Med Clin North Am 1987;5:425–8.
3. Mol BW, Hajenius PJ, Engelsbel S, Ankum WM, van der Veen F, Hemrika DJ, et al. Can noninvasive diagnostic tools predict tubal rupture or active bleeding in patients with tubal pregnancy? Fertil Steril 1999;71:167–73.
4. Fauconnier A, Mabrouk A, Heitz D, Ville Y. Ectopic pregnancy: interest and value of clinical examination in management policy [in French]. J Gynecol Obstet Biol Reprod (Paris) 2003;32:S18–27.
5. Fauconnier A, Mabrouk A, Salomon LJ, Bernard JP, Ville Y. Ultrasound assessment of haemoperitoneum in ectopic pregnancy: derivation of a prediction model. World J Emerg Surg 2007;2:23.
6. Kahn JG, Walker CK, Washington AE, Landers DV, Sweet RL. Diagnosing pelvic inflammatory disease. A comprehensive analysis and considerations for developing a new model. JAMA 1991;226:2594–604.
7. Hibbard LT. Adnexal torsion. Am J Obstet Gynecol 1985;152:456–61.
8. Bongard F, Landers DV, Lewis F. Differential diagnosis of appendicitis and pelvic inflammatory disease. A prospective analysis. Am J Surg 1985;150:90–5.
9. Abbott J, Emmans LS, Lowenstein SR. Ectopic pregnancy: ten common pitfalls in diagnosis. Am J Emerg Med 1990;8:515–22.
10. Jacobi J, Fraser GL, Coursin DB, Riker RR, Fontaine D, Wittbrodt ET, et al. Clinical practice guidelines for the sustained use of sedatives and analgesics in the critically ill adult [published erratum appears in Crit Care Med 2002;30:726]. Crit Care Med 2002;30:119–41.
11. Ducharme J. Emergency pain management: a Canadian Association of Emergency Physicians (CAEP) consensus document. J Emerg Med 1994;12:855–66.
12. Todd KH. Pain assessment instruments for use in the emergency department. Emerg Med Clin North Am 2005;23:285–95.
13. Fosnocht DE, Swanson ER, Barton ED. Changing attitudes about pain and pain control in emergency medicine. Emerg Med Clin North Am 2005;23:297–306.
14. Keele KD. The pain chart. Lancet 1948;2:6–8.
15. Huskisson EC. Measurement of pain. Lancet 1974;2:1127–31.
16. Richards JS, Nepomuceno C, Riles M, Suer Z. Assessing pain behavior: the UAB Pain Behavior Scale. Pain 1982;14:393–8.
17. Bonnel AM, Boureau F. Labor pain assessment: validity of a behavioral index. Pain 1985;22:81–90.
18. SFUM. Le traitement médicamenteux de la douleur de l’adulte dans un Service d’Accueil et d’Urgences. IIIème conférence de consensus en médecine d’urgence, le 24 Avril 1993. Réan Urg 1993;2:321–7.
19. Blettery B, Ebrahim L, Honnart D, Aube H. Les échelles de mesure de la douleur dans un service d’accueil des urgences. Réan Urg 1996;5:691–7.
20. Berthier F, Potel G, Leconte P, Touze MD, Baron D. Comparative study of methods of measuring acute pain intensity in an ED. Am J Emerg Med 1998;16:132–6.
21. Jensen MP, Karoly P, Braver S. The measurement of clinical pain intensity: a comparison of six methods. Pain 1986;27:117–26.
22. Downie WW, Leatham PA, Rhind VM, Wright V, Branco JA, Anderson JA. Studies with pain rating scales. Ann Rheum Dis 1978;37:378–81.
23. Kremer E, Atkinson JH, Ignelzi RJ. Measurement of pain: patient preference does not confound pain measurement. Pain 1981;10:241–8.
24. Puntillo KA, Miaskowski C, Kehrle K, Stannard D, Gleeson S, Nye P. Relationship between behavioral and physiological indicators of pain, critical care patients’ self-reports of pain, and opioid administration. Crit Care Med 1997;25:1159–66.
25. Ankum WM, Van der Veen F, Hamerlynck JV, Lammes FB. Laparoscopy: a dispensable tool in the diagnosis of ectopic pregnancy? Hum Reprod 1993;8:1301–6.
26. Mol BW, van Der Veen F, Bossuyt PM. Implementation of probabilistic decision rules improves the predictive values of algorithms in the diagnostic management of ectopic pregnancy. Hum Reprod 1999;14:2855–62.
27. Soper DE. Pelvic inflammatory disease. Infect Dis Clin North Am 1994;8:821–40.
28. Abbott J. Pelvic pain: lessons from anatomy and physiology. J Emerg Med 1990;8:441–7.
29. Guerriero WF, Guerriero CP 3rd, Eward RD, Stuart JA Jr. Pelvic pain, gynecic and nongynecic: interpretation and management. South Med J 1971;64:1043–8.
30. Cattell RB. The scree test for the number of factors. Multivariate Behav Res 1966;1:245–76.
31. Falissard B. Mesurer la subjectivité en santé. Perspectives méthodologiques et statistiques. Paris (France): Masson; 2001.
32. Cronbach L. Coefficient alpha and the internal structure of tests. Psychometrika 1951;16:297–334.
33. Lord FM, Novick MR. Statistical theories of mental test scores. Reading (MA): Addison-Wesley; 1968.
34. Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed. Mahwah (NJ): Lawrence Erlbaum Associates; 1988.
35. Rosnow RL, Rosenthal R. Computing contrasts, effect sizes, and counternulls on other people’s published data: general procedures for research consumers. Psychol Methods 1996;1:331–40.
36. Feldt LS. The approximate sampling distribution of Kuder-Richardson reliability coefficient twenty. Psychometrika 1965;30:357–70.
37. Streiner DL, Norman GR. Health measurement scales: a practical guide to their development and use. 2nd ed. Oxford (UK): Oxford University Press; 1994. p. 168.
38. Mondor H. Torsion des kystes et tumeurs solides de l’ovaire. Diagnostics urgents. 9th ed. Paris (France): Masson; 1965. p. 1050–8.
39. Eskelinen M, Ikonen J, Lipponen P. Usefulness of history-taking, physical examination and diagnostic scoring in acute renal colic. Eur Urol 1998;34:467–73.
40. Hallatt JG, Steele CH, Snyder M. Ruptured corpus luteum with hemoperitoneum: a study of 173 surgical cases. Am J Obstet Gynecol 1984;149:5–8.
41. Snyder H. Lack of a tachycardic response to hypotension with ruptured ectopic pregnancy. Am J Emerg Med 1990;8:23–6.
42. Lomano JM, Trelford JD, Ullery JC. Torsion of the uterine adnexa causing an acute abdomen. Obstet Gynecol 1970;35:221–5.
© 2009 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
43. Ohnhaus EE, Adler R. Methodological problems in the measurement of pain: a comparison between the verbal rating scale and the visual analogue scale. Pain 1975;1:379–84.