Portal de Periódicos da CAPES

Risk Factors of Significant Pain Syndrome 90 Days After Minor Thoracic Injury: Trajectory Analysis

2013; Wiley; Volume: 20; Issue: 11 Linguagem: Inglês

10.1111/acem.12248

1553-2712

Raoul Daoust, Marcel Émond, Éric Bergeron, Natalie Le Sage, Stéphanie Camden, Chantal Guímont, Laurent Vanier, Jean‐Marc Chauny,

Shoulder Injury and Treatment

The objective was to identify the risk factors of clinically significant pain at 90 days in patients with minor thoracic injury (MTI) discharged from the emergency department (ED). A prospective, multicenter, cohort study was conducted in four Canadian EDs from November 2006 to November 2010. All consecutive patients aged 16 years or older with MTI were eligible at discharge from EDs. They underwent standardized clinical and radiologic evaluations at 1 and 2 weeks, followed by standardized telephone interviews at 30 and 90 days. A pain trajectory model characterized groups of patients with different pain evolutions and ascertained specific risk factors in each group through multivariate analysis. In this cohort of 1,132 patients, 734 were eligible for study inclusion. The authors identified a pain trajectory that characterized 18.2% of the study population experiencing clinically significant pain (>3 of 10) at 90 days after a MTI. Multivariate modeling found two or more rib fractures, smoking, and initial oxygen saturation below 95% to be predictors of this group of patients. To the authors' knowledge, this is the first prospective study of trajectory modeling to detect risk factors associated with significant pain at 90 days after MTI. These factors may help in planning specific treatment strategies and should be validated in another prospective cohort. Identificar los factores de riesgo de dolor clínicamente significativo a los 90 días en los pacientes con traumatismo torácico menor dados de alta desde el servicio de urgencias (SU). Se llevó a cabo un estudio de cohorte, prospectivo y multicéntrico, en cuatro SU canadienses de noviembre de 2006 hasta noviembre de 2010. Se seleccionó de forma consecutiva todo paciente de 16 años o más con traumatismo torácico menor dado de alta desde el SU. Se llevaron a cabo evaluaciones clínicas y radiológicas en la primera y segunda semanas, seguidas de entrevistas telefónicas estandarizadas a los 30 y 90 días. Un modelo de la trayectoria del dolor caracterizó a los grupos de pacientes con las diferentes evoluciones del dolor y determinó los factores de riesgo específicos en cada grupo a través de un análisis multivariado. En esta cohorte de 1.132 pacientes, se eligieron 734 para la inclusión en el estudio. Se identificó una trayectoria dolorosa en un 18,2% de la población del estudio que experimentó un dolor clínicamente significativo (3 sobre 10) a los 90 días tras un traumatismo torácico menor. El modelo multivariable encontró que la fractura de 2 o más costillas, el fumar y la saturación de oxígeno inicial por debajo de un 95% son factores predictivos de este grupo de pacientes. Según nuestro conocimiento, este es el primer estudio prospectivo de dibuja la trayectoria para detectar factores de riesgo asociados con un dolor significativo a los 90 días tras un traumatismo torácico menor. Estos factores pueden ayudar a planificar estrategias terapéuticas específicas, y deberían validarse en otra cohorte prospectiva. Minor thoracic injuries (MTI) represent up to 42% of emergency department (ED) visits for blunt chest trauma, and they are mainly caused by falls of less than 2 meters and motor vehicle accidents.1, 2 Up to 90% of MTI patients will seek medical attention within 72 hours of trauma.3 MTI can produce significant pain that can last for several months.4 This thoracic pain, often severe at the time of the trauma,5 can significantly limit the ability to work and to perform daily activities. It has been suggested that MTI-related pain also hinders the ability to cough and breathe deeply, which can lead to atelectasis, accumulation of secretions, hypoxia, and infection.6, 7 Although a low incidence ( 3 were considered clinically significant.18-23 All data were entered in a centralized database by research assistants. First, we identified different pain intensity trajectories during our study period and classified patients in these discrete trajectories according to LCGM (Proc Traj, v9.2, SAS Institute Inc., Cary, NC). This technique accommodates missing data under the assumption that they are completely missing at random; to minimize the effect of the imputation we limited data analysis to patients with at least three pain intensity measures (including at least one at 30 or 90 days). To achieve this goal, models with two to five groups and different trajectories were examined. The model was optimized by Bayesian information criteria (BIC; which quantify how the model fits the data), average posterior probabilities (probability that a subject with a profile of change will be in a modeled trajectory), estimated number of patients in each group, and the presence of clinically significant pain intensity (>3/10). A more complex model could be retained if the BIC confirmed a better fit to the data, confidence intervals (CIs) of group trajectories demonstrated no overlap, the mean posterior probabilities in each group were more than 70%, and the estimated number of patients in each group was at least 5% of the entire population.24, 25 Final trajectory modeling was used to identify risk factors that could differentiate patients with particular trajectories.25 Depending on the number of identified trajectories, multinomial logistic or log-binomial regression was performed. Log-binomial regression was preferred to binary logistic regression because this approach estimates relative risk directly instead of odds ratios.26 The reference category was the trajectory with the lowest risk of clinically significant pain. If more than one trajectory had no clinically significant pain at 90 days, these trajectories were combined and served as reference. The following variables were examined: secondary diagnosis, comorbidities, pulmonary symptoms, trauma mechanism, vital signs, smoking status, alcohol intake, medication(s), number and location of rib fracture(s), syncope, age, and sex. All variables with p-values less than 0.15 in the bivariate association with membership to a trajectory group with significant pain at 90 days were then tested in a multivariate model. Since no risk factors of significant pain 90 days after MTI have been previously identified, we used the stepwise regression selection method as suggested by Hosmer and Lemeshow.27 Calibration of the model was evaluated by the calibration slope; a slope close to 1 indicates that the model is well calibrated.28 Type and use of analgesia were not analyzed because of their collinearity with our main outcome and frequent systematic drug prescription but not necessarily with patient compliance. The original cohort size of 1,250 was based on 11.9% prevalence (in our pilot study) of delayed complications in MTI patients after discharge from EDs and 90% sensitivity and 70% specificity of a future derived clinical decision rule. In this cohort, pain intensity was initially considered as a potential predictor of complications. With LCGM, we needed a minimum of 100 cases and preferably 300 to 500 cases.17 We recruited 1,132 patients from November 2006 to November 2010. In this cohort, 734 patients were included in our study (Figure 1). The basic population characteristics of eligible and excluded patients were clinically similar, except for a lower percentage of patients with initial oxygen saturation of <95% and history of smoking in the excluded group (Table 1). All study patients received prescriptions for one or more analgesics at ED discharge: 75.8% opioids, 54.9% nonsteroidal anti-inflammatory drugs, and 46.1% acetaminophen. Local application of ice or heat was recommended in 29.7% of cases. We found that 145 patients (20.3%) sought medical care after the second week of outpatient evaluation. Among those, 33.1% were for pain related to MTI, 40.0% were for other complications or administrative problems related to MTI, and 26.9% were not related to MTI. Based on BIC, models with three or four groups were a better fit to the data, but only the model with three cubic trajectories met all predetermined criteria (Table 2). In this model, we identified a group composed of 46.6% of our cohort who reported low pain (3 of 10 or lower) at 2 weeks after enrollment and no clinically significant pain at 30 and 90 days (trajectory 1). Another 35.2% of our cohort had moderate pain (higher than 3 of 10) at 2 weeks, but no significant pain at 90 days (trajectory 2). Finally, we identified a group with significant pain (higher than 3 of 10) at 90 days (trajectory 3), representing 18.2% of our study population (Table 3 and Figure 2). Bivariate analysis showed that the following factors greatly increased the risk of belonging to the trajectory with significant pain at 90 days: smoking, alcohol consumption, dyspnea, rib fractures detected on chest or rib radiographs, asthma, and initial oxygen saturation < 95% (Table 4). Pain intensity at the initial visit was not associated with pain at 90 days. Multivariate log-binomial regression analysis revealed that three factors were linked with significant pain at 90 days: at least two rib fractures, smoking, and initial oxygen saturation <95% (Table 4). This model had an excellent calibration (calibration slope = 1.00, 95% CI = 0.69 to 1.31). To the best of our knowledge, this is the largest cohort study of pain after ED medical assessments of MTI and the first to adopt trajectory modeling. We noted a significant trajectory group of patients with clinically significant pain at 90 days after MTI. We also established the presence of two or more rib fractures, smoking, and initial oxygen saturation of <95% as risk factors associated with this group. No prospective MTI study has analyzed pain as a main outcome. One small retrospective investigation of rib fracture patients in a trauma registry (102 patients) found no specific risk factors of chronic pain beyond 12 months. However, the differences from our results could be explained by study population dissimilarities (the majority of patients in the trauma registry study had other concomitant injuries; those patients were excluded from our cohort to isolate the effect of MTI), by its small sample size, and by its retrospective study design (no systematic evaluation or follow-up).4 Another small study (80 patients) focused on pain as a secondary outcome (the primary outcome was to compare the analgesic effect of tramadol to nonsteroidal anti-inflammatory drugs).5 These authors found that all patients had mild pain (≤3 of 10) at 1 week and reported no complications. However, they also observed that 30% of patients returned to the hospital with pain, and they excluded patients with characteristics similar to our risk factors of pain at 90 days (smokers, pulmonary disease).5 Few studies have examined pain evolution after discharge from EDs. Specialists in pain management have recognized that measuring mean pain intensity has its limits and that the rate of pain resolution could be more informative. Consequently, Chapman et al.29 attempted to chart the slope of a linear pain trajectory in patients discharged from EDs. They found distinct groups of patients with no slope (pain stayed the same), negative slope (pain got better), or positive slope (pain got worse), who were not identified by pain means (mean pain intensity went down from almost 7 of 10 to less than 4 of 10 during a 6-day study period). They concluded that pain trajectories provide more information and increase precision in comparison to conventional pain measurement. Trajectory modeling, such as LCGM, has mostly been used in epidemiology, psychology, and social medicine, but not in longitudinal pain intensity data analysis. Some authors argue that trajectory modeling is unreliable because it is based on too many assumptions that are rarely met: within-class normality, properly specified mean and covariance structure; exogenous predictors, individual trajectory parameters that are linear, randomly missed data, and equal probabilities of selection of sampled individuals.30 However, such modeling offers many advantages over measuring means and slopes of pain evolution after discharge from EDs. Pain can get worse in the first days after trauma before it gets progressively better. In other cases, pain might suddenly get better, then stabilize for a period of time and stay the same or get better (our study). Therefore, we believe that trajectory modeling offers a better fit for this kind of pain evolution. Pain trajectories could also be more informative in evaluating the effect of new medications on pain management after discharge from EDs. Pain after initial ED visit will generally improve without treatment,29 so that mean pain intensity will generally decrease. However, new medications could provide pain relief to certain patients identified only by a trajectory approach. A significant limitation of our study is the low percentage of eligible patients available for final analysis (64.8% of the original cohort). However, this percentage is similar to those in studies on chronic pain prevalence.16 Our cohort was created to estimate the complication rate of MTI and establish a clinical rule to identify patients who need a specific treatment plan. Evaluation of pain intensity during the study period was originally considered as a predictor of complications. Patients were mostly excluded from our study for having fewer than three pain intensity measurements (generally missing at the end of the follow-up period), and only 75 patients (6.6% of the original cohort) were lost at follow-up. This was done to minimize the effect of the imputation process from the SAS program; it could produce trajectories that seem to better fit the study population and yield potentially less reliable results. The probability of missing pain intensity measures is similar across the three trajectory groups (data not shown). Thus, the missing pain values do not seem closely related to the group trajectory and we assumed that the mechanism of missing data had no significant effect on our model. We analyzed patients with no missing pain intensity measures during our follow-up period and found similar groups but with larger CIs. Furthermore, 96% of these patients remained in the same trajectories as in the original analysis (data not shown). We could also hypothesize that pain during follow-up, if moderate or severe, would be measured more often or that patients would frequently seek medical care. Therefore, it is less probable that excluded patients would be part of the group with high risk of significant pain at 90 days. The fact that eligible and excluded patients have similar characteristics is also reassuring (Table 1). Only initial oxygen saturation of <95% and history of smoking were less prevalent in excluded patients, which could reduce the probabilities of them having significant pain at 3 months. To further assess the effect of our missing data, they were first reanalyzed assuming that all 398 excluded patients for missing data were in the group with significant pain at 90 days (trajectory 3) and then in the group with no clinically significant pain at 90 days (trajectories 1 and 2). With these imputations, we found the presence of the same risk factors associated with our original study cohort. Only initial oxygen saturation of <95% was not significant with the assumption that all 398 patients had significant pain at 90 days; as discussed above this scenario is unlikely (data not shown). Furthermore, other aspects of pain (catastrophizing, mood, sleep, quality of life, etc.), that could be risk factors of significant pain at 3 months were not assessed. As two of our four study sites were trauma centers, more severe MTI patients could have been selected, which could lessen the external validity of our study. Some MTI patients could have been admitted to centers less familiar with MTI. For example, performance of the primary care emergency systems for treating simple thoracic injuries was evaluated in 488 patients transferred to a trauma center. None of these patients were admitted after initial evaluation, and only 15 required minor treatment (for orthopedic injuries).2 Finally, the risk factors of moderate pain at 90 days should be validated in a prospective cohort. To the best of our knowledge, this is the first prospective study of trajectory modeling to identify potential risk factors of pain at 90 days after minor thoracic injury. We found that two or more rib fractures, smoking, and initial oxygen saturation <95% were associated with pain at 90 days. These factors should be validated in a prospective cohort so clinicians can rapidly identify these patients, devise specific treatment strategies, and follow up after discharge from EDs, possibly preventing chronic pain progression.