The study was conducted at two urban academic Emergency Departments associated with Montefiore Medical Center. Emergency Medicine residents and full-time Emergency Medicine faculty are present at both sites. One of the sites is a municipal hospital Level 1 Trauma Center managing approximately 50,000 adult patients per year. The other site is the primary academic medical center, managing approximately 80,000 adult patients per year, many of them requiring tertiary care. Both departments serve an urban population that is predominantly African-American and Hispanic.
Study Protocol Data were obtained from a convenience sample of patients from September, 2003 through November, 2006. Patients were eligible if the provider suspected an isolated long-bone fracture and long-bone films were obtained. Because it was not possible to determine which patients would have positive films in advance, all patients with suspected fractures who met other inclusion criteria were initially considered eligible and enrolled. Only those with confirmed fractures were included in the final study sample. Patients were considered to have fractures if the radiology report by an attending radiologist specified an isolated long-bone fracture defined as a fracture of the humerus, radius, ulna, femur, tibia, or fibula (ICD-9 codes: 812, 813, 821, 823, and 824).
Data collection took place from 8:00 am to midnight seven days a week at one institution and 24 hours a day, seven days a week at the other. Data were collected on a standardized data collection instrument by trained fluently bilingual (Spanish and English) Research Associates. The Research Associates attended a three-day course in research ethics, and specific data collection procedures of the study. This was followed by a two-week internship during which new Research Associates were paired with more experienced Research Associates.
Research Associates approached patients whose triage note indicated an extremity injury, extremity pain, if the Research Associate observed a physical indication of extremity injury, or if a provider identified a patient who was potentially eligible for the study. Eligible patients were asked to sign informed consent. Data were collected from patients at the time of the physician's examination of the patient (baseline) and at discharge. At discharge the treating physician was asked to rate the patient's pain when he/she first examined the patient. If the physician's shift ended before the patient was discharged he/she was asked to rate the patient's initial pain before leaving the ED.
Measurements The names of all medications, dose, route, and time administered were obtained in real time from the chart and ED staff. Patients were categorized as receiving: 1) no analgesics; 2) NSAIDS or acetaminophen; or 3) opioids. If both NSAIDS or acetaminophen and opioids were administered, patients were categorized as having received opioids. Analgesics given solely for fracture reduction were not used to categorize patients. Opioid analgesics were transformed into morphine equivalents as follows: 1 mg morphine = 3 mg of oxycodone; 0.15 mg of hydromorphone, and 13 mg of codeine. No other opioid analgesics were used with the exception of fentanyl, which was only used for fracture reduction.
The primary predictor variable was self-reported race/ethnicity of the patient. This was measured by the response to the question: "Do you consider yourself to be: Hispanic/Latino, White/Caucasian, Black/African-American, multi-racial, or other?"
Covariates: Patients' pain and provider assessment of patients' pain are determinants of treatment and thus have the potential of confounding the relationship between race/ethnicity and treatment if they vary systematically by race/ethnicity.10 Pain was measured in several ways: Patients were asked to rate their pain intensity on a validated and reproducible 11-point numerical rating scale (NRS) ranging from 0 "no pain", to 10 "worst imaginable pain".18 They were also asked to describe pain categorically as: "no pain", "mild pain", "moderate pain", or "severe pain". Change in pain was measured by response to the question: "How does your pain feel now compared to when you entered the Emergency Room?" It was also measured by the difference between the patient's rating of pain at entry and pain at discharge. At time of discharge the treating physicians were asked to recall and rate the patient's initial pain using the same verbal descriptors used by the patients We reasoned that asking doctors to rate patients' pain before they made treatment decisions might bias treatment.
Other clinically plausible confounders of the relationship between pain management and race/ethnicity include: sex, age, educational level (some high school, high school or General Equivalency Diploma (GED), some college, Bachelor's degree or higher), insurance status (self-pay, health maintenance organization (HMO) or private insurance, Medicaid, other insurance), accompanied to the ED by someone who might act as a patient advocate, mechanism of injury (fall, sports, motor vehicle, other), fracture reduction in ED, and arrival by ambulance.
Data Analysis:
Patient characteristics are reported as means with standard deviations, medians with interquartile ranges (IQR) and percentages. We used chi-square tests to compare discrete patient characteristic by race/ethnicity. One-way analysis of variance was used to compare continuous characteristics and the Kruskal-Wallis test to compare medians of non-normally distributed variables.
Logistic models were used to test for the presence of an independent (adjusted) association between race/ethnicity and the treatment of pain. We used forced entry into the model of initial patient rating of pain intensity, provider rating of patient's pain at baseline, hospital, sex, age, education, insurance status, whether accompanied to the ED, mechanism of injury, fracture reduction in ED, and arrival by ambulance. Hosmer Lemeshow's test was used to assess model fit. We used Zhang and Kai's method for estimating relative risks from the odds ratios.
Sample size calculation: There is no consensus about what constitutes a clinically significant difference in treatment between racial and ethnic groups. In the absence of such a consensus we based our estimates of effect size on previous findings.11, 12 We used estimates of 75%, 50% and 60% respectively for proportion of whites, African-American, and Hispanics treated with any analgesic, a two-tailed significance level of 0.05, and a power of 80% to calculate a sample size of at least 70 patients with long-bone fractures in each of the three groups.
Two individuals independently entered data into the Statistical Package for the Social Sciences (SPSS) Data Entry(Chicago, IL). Transcription errors were reconciled by referral to the original hardcopy of the data collection instrument. We used NQuery 6.0 ( Saugus, MA) for the sample size calculation, and SPSS version 15.0 (Chicago, IL).
