Risk of Post-traumatic Stress Symptoms in Family Members of Intensive Care Unit Patients

Transcription

1 Risk of Post-traumatic Stress Symptoms in Family Members of Intensive Care Unit Patients Elie Azoulay, Frédéric Pochard, Nancy Kentish-Barnes, Sylvie Chevret, Jérôme Aboab, Christophe Adrie, Djilali Annane, Gérard Bleichner, Pierre Edouard Bollaert, Michael Darmon, Thomas Fassier, Richard Galliot, Maité Garrouste-Orgeas, Cyril Goulenok, Dany Goldgran-Toledano, Jan Hayon, Mercé Jourdain, Michel Kaidomar, Christian Laplace, Jérôme Larché, Jérôme Liotier, Laurent Papazian, Catherine Poisson, Jean Reignier, Fayçal Saidi, and Benoît Schlemmer Service de Réanimation Médicale, Hôpital Saint-Louis, Paris, France Rationale: Intensive care unit (ICU) admission of a relative is a stressful event that may cause symptoms of post-traumatic stress disorder (PTSD). Objectives: Factors associated with these symptoms need to be identified. Methods: For patients admitted to 21 ICUs between March and November 2003, we studied the family member with the main potential decision-making role. Measurements: Ninety days after ICU discharge or death, family members completed the Impact of Event Scale (which evaluates the severity of post-traumatic stress reactions), Hospital Anxiety and Depression Scale, and 36-item Short-Form General Health Survey during a telephone interview. Linear regression was used to identify factors associated with the risk of post-traumatic stress symptoms. Main results: Interviews were obtained for family members of 284 (62%) of the 459 eligible patients. Post-traumatic stress symptoms consistent with a moderate to major risk of PTSD were found in 94 (33.1%) family members. Higher rates were noted among family members who felt information was incomplete in the ICU (48.4%), who shared in decision making (47.8%), whose relative died in the ICU (50%), whose relative died after end-of-life decisions (60%), and who shared in endof-life decisions (81.8%). Severe post-traumatic stress reaction was associated with increased rates of anxiety and depression and decreased quality of life. Conclusion: Post-traumatic stress reaction consistent with a high risk of PTSD is common in family members of ICU patients and is the rule among those who share in end-of-life decisions. Research is needed to investigate PTSD rates and to devise preventive and early-detection strategies. Keywords: autonomy; burden; decision making; end of life; paternalism Having a family member hospitalized in the intensive care unit (ICU) causes overwhelming stress and distress (1). Over the last decade, ICU physicians and nurses have created the concept of family-centered care (2), in which close attention is given to meeting the informational and emotional needs of family members. Family-centered care has been studied in the overall popu- (Received in original form September 29, 2004; accepted in final form January 12, 2005) This article was written on behalf of the FAMIREA Study Group: the Intensive Care Unit of the Saint-Louis Teaching Hospital and University of Paris 7, Assistance Publique Hôpitaux de Paris, Paris, France. The study was done in the 21 intensive care units listed in the Appendix. Supported by a grant (AOR1004) from the Assistance Publique Hôpitaux de Paris and the Direction Régionale de la Recherche Clinique, two nonprofit organizations funded by the French government. Correspondence and requests for reprints should be addressed to Elie Azoulay, M.D., Ph.D., The French Famirea Study Group, Service de Réanimation Médicale, Hôpital Saint-Louis, 1 Avenue Claude, Vellefaux, Paris, France. elie. azoulay@sls.ap-hop-paris.fr This article has an online supplement, which is accessible from this issue s table of contents at Am J Respir Crit Care Med Vol 171. pp , 2005 Originally Published in Press as DOI: /rccm OC on January 21, 2005 Internet address: lation of ICU patient families (3 5) and in the subset of families of dying patients (6 8). One of the goals of family-centered care is to reduce the risk of stress reactions to the traumatic ICU experience. The possibility that post-traumatic stress disorder (PTSD) may develop in family members of ICU patients is of special concern. PTSD was individualized in 1980 based on investigations of symptoms reported by veterans of the Vietnam War. The Diagnostic and Statistical Manual of Mental Disorders (fourth edition) expanded the diagnostic criteria for PTSD, which is now a well recognized clinical entity (9). Many tools have been developed for identifying PTSD or symptoms of post-traumatic stress reaction associated with a high risk of PTSD (10 12). In a study using one of these tools, the Impact of Event Scale, symptoms indicating a high risk of PTSD were common among family members of patients discharged alive from the ICU (11). Knowledge of factors associated with the risk of PTSD would be useful for designing family-care strategies capable of protecting families against the development of stress-related morbidity. Studies have shown that improvements are needed regarding interactions with family members of dying ICU patients (13 15). However, stress-related symptoms in family members after ICU management of a relative have not been evaluated in a large cohort and risk factors for such symptoms have not been sought. We conducted a longitudinal study to assess possible stressrelated morbidity among family members 90 days after ICU discharge or death of a relative, with special attention to symptoms indicating a risk of PTSD. Families were recruited in 21 ICUs. During a telephone interview, each family member completed the Impact of Event Scale, which measures symptoms associated with a risk of PTSD; the 36-item Short-Form General Health Survey (SF-36); and the Hospital Anxiety and Depression scale (HADS). Risk factors for post-traumatic stress symptoms were sought. We found that clinically meaningful post-traumatic stress reaction was common and was the strongest risk factor was sharing in end-of-life decisions for the relative in the ICU. METHODS Study Design This longitudinal study was conducted in 21 medical-surgical ICUs in France between March 2003 and November The centers were selected on the basis of homogeneous practices regarding end-of-life care and family-centered care. Family members were eligible for the study if they came to the ICU to visit a patient with a stay duration greater than 48 hours. For each patient, we included the family member who was highest in the hierarchy for substitute decision making (spouse; parents/children; others). We obtained informed consent for the 90-day interview near the time of discharge or death of the patient (16). The ethics committee of the French Society for Critical Care Medicine approved the study. Each study ICU included consecutive patients and their family members over a 15-day period. To ensure optimal quality of the data, all 90-day interviews were conducted by the same person (N.K.-B.), a sociologist with extensive interviewing experience. Because

2 988 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL TABLE 1. PATIENT AND FAMILY MEMBER CHARACTERISTICS (n 284 OF EACH) All No PTSR PTSR Patient Characteristics (n 284) (n 190) (n 94) p Value Male 158 (55.6) 101 (63.9) 57 (36.1) 0.25 Age 59 ( ) 59 (45 73) 58 (43 74) 0.66 Poor chronic health status 90 (32.8) 52 (38.3) 40 (44.4) Comorbidities Chronic obstructive pulmonary disease 59 (20.8) 38 (20) 21 (22.3) 0.64 Chronic heart failure 54 (19) 44 (23.1) 10 (10.6) 0.01 Solid cancer or hematologic malignancies 40 (14) 20 (10.5) 20 (21.3) 0.01 Neurologic disorders 24 (8.4) 12 (12.3) 12 (6.7) 0.07 Psychiatric disorders 16 (5.6) 9 (4.7) 7 (7.4) 0.41 Simplified Acute Physiology Score II score at admission 40 (27 53) 38 (24 52) 41 (32 56) 0.04 End-of-life decisions 50 (17.6) 20 (10.5) 30 (31.8) Intensive care unit mortality 56 (19.7) 28 (14.7) 28 (29.7) Family member characteristics Relationship with the patient Spouses 135 (48.2) 90 (48.4) 45 (47.8) 0.93 Children 37 (13.2) 17 (9.1) 20 (21.2) Parents 67 (23.9) 45 (24.2) 22 (23.4) 0.33 Other family members 41 (14.6) 34 (18.3) 7 (7.4) 0.28 Age 51 (41 61) 51 (40 62) 52 (44 60) 0.38 Female sex 192 (67.6) 116 (61) 76 (80.8) Effectiveness of the information provided Family felt that not enough time was allowed for information 43 (15.1) 23 (12.1) 20 (21.3) 0.04 Family felt that information was not easy to understand 45 (15.8) 21 (11) 24 (25.5) Family felt that information was incomplete 95 (33.4) 49 (25.8) 46 (48.9) Involvement in everyday patient care 71 (25) 54 (28.4) 17 (18.1) 0.06 Bathing 32 (45) Feeding 71 (100) Tracheal aspiration 3 (4.2) Involvement in everyday (not end-of-life) decisions Consent to research 69 (24.3) 36 (18.9) 33 (35.1) Decision to perform tracheotomy 20 (7) 14 (7.4) 6 (6.4) 0.39 Discussion about the appropriate level of care 18 (6.3) 9 (4.7) 9 (9.6) 0.11 Discussion about the patient s preferences and values and the patient s quality of life 26 (9.1) 11 (5.8) 15 (15.9) (20.8) 35 (18.4) 24 (25.6) 0.07 Family was only informed of the end-of-life decision 28 (56) 16 (80) 12 (40) Family shared in the end-of-life decision 22 (44) 4 (20) 18 (60) Definition of abbreviation: PTSR post-traumatic stress reaction. the estimated time needed to contact family members and to conduct the interviews was 6 months, we enrolled the patients according to a rotating pattern, with four ICUs including patients each month for 5 months and one ICU the last month. Thus the interviews were done between June and November The first phone call was made 90 days after the end of the ICU stay. Family members who did not answer the first call were called again, two to three times per day. After 15 unsuccessful attempts, the family member was considered lost to follow-up. Characteristics of Patients and Family Members When discharge or death of the patient was felt to be imminent, the study investigator in each ICU asked the eligible family member (one per family) whether he or she consented to a telephone interview aimed at assessing physical, mental, and emotional well-being and the potential burden generated by the ICU experience 90 days after the end of the ICU stay. If the family member consented, the investigator recorded one or two phone numbers for contacting the family. The characteristics of patients and family members are listed in the online supplement. Participation of family members in care and decisions was described using a semiquantitative scale reported in the online supplement. Table 1, and Table E1 in the online supplement, show additional characteristics of the family members. Participation of Family Members in End-of-Life Decisions Information was provided daily to family members of all patients, regardless of their prognosis. When a move from curative to palliative care was considered, the family members were asked to attend a family- ICU staff conference aimed at clarifying treatment objectives, in accordance with current recommendations (6 8). In general, a decision to limit curative care was taken after this conference. During the conference, the family members were free to choose the level of involvement they wanted. Further family conferences were then held to determine which treatments would be limited and which palliative measures would be used; again, these decisions were taken with the level of family involvement desired by the family. All family conferences were led by the attending intensivist in charge of the patient. At the end of the first conference, this intensivist completed the above-described scale for family involvement in decisions. The procedures for family involvement in care, everyday decisions, and end-of-life decisions are standard practice in all 21 participating ICUs. The investigator, together with the patient s nurse, used a standardized form to record whether family members shared in care (bathing, feeding, and tracheal suctioning) and decisions (informed consent to research, decision to perform tracheotomy, discussions about the appropriate level of care, and discussions about the patient s preferences and values and the patient s quality of life) at any time during the patient s ICU stay. The degree of family involvement in care was described using a scale from 0 to 3, where 0 indicated no involvement; 1, a smaller role in carrying out the task than that of the ICU staff; 2, a greater role in carrying out the task than that of the ICU staff; and 3, carrying out the task with no help from the ICU staff. The degree of family involvement in everyday treatment decisions was described using a scale from 0 to 3, where 0 indicated no involvement; 1, that the family member testified to the patient s wishes, then left the decision to the physicians; 2, that the family member stated his or her opinion, then left the decision to the physicians; and 3, that the family member made the decision after listening to the physicians opinion.

3 Azoulay, Pochard, Kentish-Barnes, et al.: Family Burden after Critical Care 989 When end-of-life decisions were implemented, a specific form was used to record whether family members were involved in the end-of-life decision, and the degree of family involvement was described using the same scale as for everyday treatment decisions. This specific form was completely separately from the forms for everyday treatment decisions, even for family members of patients who died. Telephone Interviews 90 Days After Patient Discharge or Death After introducing herself, the interviewer asked questions about the effectiveness of the information provided by the ICU staff (see online supplement). The family members then completed the Impact of Event Scale, the SF-36, and the HADS, in that order. The IES has been widely used for many years and found reliable across a broad range of traumatic events (17). It can be easily completed during a telephone interview and has been validated in French (18). The IES is not a tool for diagnosing PTSD, but instead detects symptoms indicating a risk of PTSD. Given the paucity of data on stress-related symptoms in family members of ICU patients, we elected to cast a wide net in our investigations and therefore to use the less restrictive approach offered by the IES (17, 19). An additional advantage of the IES is that it provides information, not only on the presence of stress-related symptoms, but also on their severity. Each of the 15 items is scored on a 6-point scale rated from 0 to 5 (19 21), so that the total score can range from 0 to 75. Two subscores can be determined (19): the intrusion subscore, which reflects how insistently thoughts and impressions related to the event reappear, and the avoidance subscore, which assesses behaviors aimed at avoiding people, places, or activities that act as reminders of the traumatic event. Higher scores indicate more severe post-traumatic stress symptoms. We classified patients as having low or high IES scores using 30 as the cutoff, in agreement with previous reports that IES scores higher than 30 indicated post-traumatic stress reaction with a significant risk of PTSD (22 24). We used the SF-36 and HADS both for differential diagnosis, because overlap exists between symptoms of PTSD and those of generalized anxiety disorder, and as a means of checking the clinical relevance of the IES score results. The SF-36 includes eight multipleitem scales that assess physical functioning, social functioning, physical role, emotional role, mental health, pain, vitality, and general health. Scores for each scale can range from 0 (worst possible health) to 100 (best possible health) (18). The HADS consists of two subscales that evaluate symptoms of depression (seven items) and symptoms of anxiety (seven items), respectively (6). Subscale scores range from 0 (no distress) to 21 (maximum distress) (1). Detailed descriptions of the SF-36 and HADS, and the rationale for their use, are given in the online supplement. Data Validation All missing data were retrieved by the investigator in each study center (see Appendix) at the request of the first three authors. We audited the ICU-patient database by comparing study and source documents in a random sample of 10% of the patients. Statistical Analysis We summarized continuous variables with the medians and interquartile range, and we used the Wilcoxon rank-sum test for comparing patients with and without post-traumatic stress reaction. We summarized categorical variables using proportions and used the Pearson chisquare test for comparing family members with and without post-traumatic stress reaction. We used Fisher s exact test when appropriate. All statistical tests were two-tailed. We also performed univariate linear regression to identify factors associated with an increased IES score in family members 90 days after ICU discharge or death of their relative. Variables significant in the univariate analyses (p 0.2) were considered for inclusion in a multivariable linear regression model, using backward stepwise descending selection. Information on the performance of the multivariable models was assessed by the percentage of variation (r 2 ) explained by the model. SAS software (version 8, SAS Inc., Cary, NC) was used for all statistical analyses. RESULTS Characteristics of Patients and Family Members Over the study period, we screened 459 patients who spent more than 48 hours in one of the 21 study ICUs. Reasons for Figure 1. Flow chart of the patients. noninclusion are outlined in Figure 1. Consent was obtained from 432 (94.1%) family members, who provided one or more phone numbers for the 90-day interview. When called, 60 (13%) additional family members refused to participate in the study, yielding a total refusal rate of 19%. Among the 372 (81%) remaining family members, 88 (19.2%) could not be contacted despite 15 attempts. This left 284 (61.8%) family members for whom full telephone interviews were obtained. Patient characteristics, which were available for all 459 patients, showed no significant differences between families who did and did not complete the interview. Among the 21 participating ICUs, 11 (52.4%) were in university hospitals. The median visiting time per day was 3 (2 4) hours; 5 (24%) ICUs had recently extended their visiting time to 24 hours per day. The study ICUs had 14 (12 16) beds, 5 (4 6) intensivists, and 2 (2 3) patients per nurse. Of the 284 patients, 220 (77.5%) had medical conditions and 64 (22.5%) surgical conditions; 185 (65.1%) required conventional mechanical ventilation, 64 (22.5%) noninvasive mechanical ventilation, 137 (48.2%) vasopressors, and 36 (12.6) renal replacement therapy. ICU mortality was 19.7% (56 patients) and 90-day mortality was 30.9% (88 patients). Although most of the family members were satisfied with the information they received from the ICU staff, insufficient time spent receiving information, lack of clarity, and incompleteness of information were reported by 43 (15.1%), 45 (15.8%), and 95 (33.4%) family members, respectively. Family members were involved in care of the patient in 71 (25%) cases and in decisions about the patient s management in 69 (24.3%) cases. On the four-level scale (0 3) used to evaluate the degree of involvement in care and decisions, the median for care was 1 (0 1), but the median for decisions was 3 (2 3), indicating that those family members who shared in decision making were involved to a very large extent. Among the 56 patients who died in the ICU, 50 (89.3%) received end-of-life decisions; for 28 patients, the families were simply told of the decisions; and for 22 patients the families, shared in the decisions. After the ICU stay of their relative, 24 (8.4%) family members received support from a psychologist and 24 (8.4%) took physician-ordered psychotropic medications. Tables 1 and E1 report additional data on the patients and family members. Impact of Event Scale Scores and Post-traumatic Stress Reaction As indicated in Table 2, 94 (33.1%) family members had an IES score greater than 30, indicating post-traumatic stress reaction with a significant risk of PTSD. By univariate analysis, IES scores greater than 30 were found in about half the family members who perceived the information as unsatisfactory and in more

4 990 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL TABLE 2. IES SCORES AND RATE OF OCCURRENCE OF PTSR IN FAMILY MEMBERS OF ICU PATIENTS Patients Median IES Score Patients With PTSR n (%) (25 75%) n (%) All family members 284 (100) 22 (11 34) 94 (33.1) Family members involved in patient care 71 (25) 20 (10 29) 17 (24) Family members of patients discharged alive from the ICU 228 (80.3) 21 (10 32) 66 (28.9) Family felt that not enough time was allowed for information 43 (15.1) 29 (20 39) 20 (46.5) Family felt that information was not easy to understand 45 (15.8) 33 (21 39) 24 (53.3) Family felt that information was incomplete 95 (33.4) 29 (15 37) 46 (48.4) Family members involved in everyday decisions about the patient 69 (24.3) 30 (15 36) 33 (47.8) Family members of patients who died in the ICU 56 (19.7) 30.5 (18 38) 28 (50) Family members of patients who died in the ICU after end-of-life decisions 50 (17.6) 33 (22 39) 30 (60) Family members involved in end-of-life decisions 22 (7.7) 35.5 (31 39) 18 (81.8) Definition of abbreviations: ICU intensive care unit; IES Impact of Event Scale; PTSR post-traumatic stress reaction. PTSR was defined as an IES score 30, indicating a high risk of developing post-traumatic stress disorder. Family involvement in decisions about the patient included consent to research, decision to perform tracheotomy, discussions about the appropriate level of care, and discussions about the patient preferences and values and about the patient s quality of life. than 80% of those who shared in end-of-life decisions. As reported in Table 3 and Table E2, factors independently associated with an IES score greater than 30 in family members were female sex, being a child of the ICU patient, cancer in the ICU patient, more severe acute illness in the patient as assessed by the Simplified Acute Physiology Score II score, feeling that information was incomplete, sharing in decisions in the ICU, and death of the patient in the ICU. Conversely, IES scores were lower in family members of patients with heart failure and of those admitted to the ICU for postoperative care. In the model including only patients with end-of-life decisions (Table 4), IES scores were higher in family members who felt that information was incomplete and in those who shared in end-of-life decisions. IES scores were higher in ICUs with larger numbers of beds. The r 2 for the multivariable models was 0.69 and 0.71 for Table 3 and Table 4, respectively (indicating that 69% and 71% of the variation were explained by the models). Linear regression and logistic regression produced identical results. Figures 2 and 3 report the IES scores according to the perceived effectiveness of information and to the level of involvement of family members in end-of-life decisions, respectively. Construct Validation of the IES Score in Families of ICU Patients The IES score was compared with health-related quality-of-life (SF-36) and symptoms of anxiety and depression (HADS). Measurement of health-related quality-of-life disclosed a slight decrease in general health (72% [60 80%]) but a major decrease in mental health (60% [48 72%]). Measurement of the levels of anxiety and depression showed that 140 (49.3%) family members had anxiety symptoms and 50 (20.1%) depression symptoms. As shown in Figure E1, patients with IES scores greater than 30 had significant impairments in quality of life predominantly affecting the mental health area, in which all components showed low median scores (66 [50 80] for emotional role, 70 [60 90] for social functioning, 60 [45 70] for vitality, and 60 [48 72] for mental health). As shown in Figure E2, the score on the anxiety and depression subscales of the HADS differed significantly between patients with and without IES scores greater than 30. Among the 94 family members with IES scores greater than 30, 68 (72.3%) had severe anxiety (compared with 37.9% in the other family members, p ) and 38 (40.4%) had severe symptoms of depression (compared with 10% in the other family members, p ). TABLE 3. FACTORS ASSOCIATED WITH THE IES SCORE IN THE OVERALL POPULATION OF FAMILY MEMBERS Univariate Analysis Multivariate Analysis Coefficient (SD) p Value Coefficient (SD) p Value Patient characteristics Chronic heart failure 9.08 (2.05) (1.77) Cancer and hematologic malignancies 6.96 (2.2.36) (1.98) Admission to the intensive care unit for postoperative care 6.98 (1.62) (1.40) Simplified Acute Physiology Score II score at admission (per point) 0.15 (0.04) (0.037) Death in the intensive care unit 8.41 (2.03) (1.81) Family-member characteristics Children 9.29 (2.41) (2.12) Female sex 8.02 (1.71) (1.46) Felt that information was incomplete 5.27 (1.74) (1.47) Involvement of family members in everyday decisions 6.57 (1.90) (1.64) Involvement of family members in end-of-life decisions 9.54 (3.55) The Impact of Event Scale (IES) score range is 0 75, with higher scores indicating greater risk for post-traumatic stress symptoms. The coefficients are added to the mean IES score ( ). For example, when a cancer patient ( 4.49) is dying in the intensive care unit ( 6.19), his or her daughter (6.33 for children 6.29 for female sex) has an IES score increase of In this example, the model estimates the mean IES score at

5 Azoulay, Pochard, Kentish-Barnes, et al.: Family Burden after Critical Care 991 TABLE 4. FACTORS ASSOCIATED WITH A HIGHER IES SCORE IN FAMILY MEMBERS OF CRITICALLY ILL PATIENTS WHO DIED AFTER AN END-OF-LIFE DECISION Multivariable Coefficient (SD) Analysis p Value Number of beds per intensive care unit 0.99 (0.31) Involvement of family members in end-of-life decisions (3.12) Felt that information was incomplete (4.25) The Impact of Event Scale (IES) score ranges from 0 to 75, with higher scores indicating greater risk for post-traumatic stress symptoms. The coefficients are added to the mean IES score ( ). For example, when the next of kin of a patient shared the end-of-life decision with the intensive care unit team but felt the information was incomplete, the model estimates his or her IES score at ( ). DISCUSSION The present study provides data on the rate of occurrence of post-traumatic reaction in a large number of family members of unselected critically ill patients. We found post-traumatic stress reaction, defined as an IES score greater than 30, in one third of family members 90 days after ICU discharge or death of a relative. Information perceived as unsatisfactory was associated with post-traumatic stress reaction, and more than four fifths of family members who shared in end-of-life decisions exhibited post-traumatic stress reaction. The IES detects and quantitates post-traumatic stress reaction associated with a high risk of PTSD (11, 17, 19 21, 23, 24). The association between sharing in end-of-life decisions and a high risk of PTSD found in our study raises two issues: bereavement symptoms may have been misclassified as post-traumatic stress symptoms or these symptoms may have been caused by bereavement (25, 26). However, the IES has been found effective in discriminating between post-traumatic stress reaction and bereavement-related symptoms (19). Furthermore, in our study, 50% (28/56) of family members of patients who died had posttraumatic stress reaction, but so did 28.9% (66/228) of family members of surviving patients, although this last subgroup did not have the main risk factor for post-traumatic stress symptoms; namely, sharing in end-of-life decisions. Most of the previous studies of emotional issues in family members focused on families of patients who died in the ICU (13 15), which was the case for only about one fifth of our patients. Thus factors that may have contributed to post-traumatic stress reaction in our study include having witnessed a critical situation in a loved one (26) and, in the families of survivors, experiencing PTSD from the patient (9, 27). In our 90-day interviews, in addition to post-traumatic stress, we explored other dimensions related to well-being; namely, symptoms of anxiety and depression and quality of life. The results of the HADS and SF-36 demonstrated clearly that post-traumatic stress reaction was associated with an increased prevalence of symptoms of anxiety and depression and with alterations in quality of life, in keeping with earlier studies in ICU patients (12). The high rate of post-traumatic stress reaction in our study indicates a need for seeking preventive and early-detection strategies. First, a useful step would be identification of factors detectable at the time of the ICU stay and associated with increased vulnerability to post-traumatic stress reaction. Conceivably, the high rate of symptoms of anxiety and depression found among family members of ICU patients (1) may increase or be associated with an increase in the risk of post-traumatic stress reaction. Other studies have reported similar rates of anxiety and depression and have suggested interventions to support family members (11, 28, 29). Detailed studies of the nature of the anxiety, or fear response, experienced by families during the ICU stay might help to detect abnormalities associated with progression to PTSD (9). In addition, a study of trauma patients found that marked symptoms of PTSD during the surgery ward stay predicted PTSD over the next year (12). It would be helpful to determine, using simple screening tools, whether families also exhibit stress-related symptoms while their relative is in the ICU, and whether these symptoms predict subsequent PTSD (30). Second, in the present study, perceived incompleteness of the information received in the ICU was independently associated with post-traumatic stress reaction. This finding is in agreement with previous studies of the specific needs of family members of ICU patients (31). Third, we found that participation in end-of-life decisions was a major risk factor for post-traumatic stress reaction. We hypothesize that involvement in end-of-life decision making may cause symptoms of post-traumatic stress, but also that many family members may Figure 2. Post-traumatic stress reaction and perceived effectiveness of information provided in the intensive care unit.

6 992 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL Figure 3. Post-traumatic stress reaction and level of involvement of family members in end-of-life decisions. choose to be involved in decision making even if they are aware this carries a risk of greater post-traumatic stress reaction after 90 days. This highlights an area in which improvements in family care strategies are needed. In our study, involvement of families in decisions about their relatives was conducted in accordance with current international recommendations (7, 8, 32). The first step toward improvement may be to determine whether better information of families sharing in end-of-life decisions decreases their risk of post-traumatic stress reaction. Although it is possible that the symptoms found in this study represented a normal process of grief and bereavement, meeting the informational needs of family members is particularly crucial when the patient is dying (8, 33) and when the family members are acting as surrogate decision makers for the patient (34, 35). Adequate information empowers family members in their active role within the shared decision-making process (36). Further studies are needed to define the family information strategy that minimizes the risk of post-traumatic stress reaction among family members, most notably those who agree to share in decisions about their relative (37). In addition to information, other changes may be needed regarding the manner in which family conferences are conducted, the listening and communication skills of ICU staff members, the efforts exerted to make families feel welcome in the ICU, and other areas. The other independent risk factors identified in our study (Table 3 and Table E2) are not amenable to preventive strategies. The provision of support to family members who experience post-traumatic stress reaction requires attention. We found that only 25% of family members with post-traumatic stress reaction were receiving medical care. This contains an important message for general practitioners and family physicians, because they are in a unique position for providing families with routine followup and support after the ICU experience, detecting stress-related symptoms, and ensuring that appropriate evaluation and treatment are received if needed. Our study has several limitations. First, complete telephone interviews were possible in only 284 (61.8%) family members. However, patient characteristics were available for all 459 patients and did not differ significantly between patients whose families did and did not provide complete interviews. Second, this study was performed in France, where the patient-physician relationship usually allows for only limited involvement of families in decisions (38 40). Thus, in the present study, only approximately 25% of family members overall participated in decisions and care. However, family members were informed, and many of those whose relatives were dying were involved in end-of-life decisions, indicating that the role given to families is growing in France (41). Moreover, the number of centers, patients, and family members involved in the present study suggests that our results may apply to many ICUs. Also, the factors identified are intuitively valid and should serve to remind clinicians in all cultures, even those in which families are routinely involved in decision making, of the need to improve communication between caregivers and families in the ICU. Third, evidence from the literature indicates that the incidence of PTSD decreases over time after a traumatic experience (10, 42). We conducted a single evaluation 90 days after the ICU stay. On the one hand, it is known that stress symptom intensity and the incidence of true PTSD after extreme stress exposure can increase over many months after the event. On the other hand, there is good evidence that after a climax of maximal incidence/intensity of stress symptoms (probably around 6 months) after the event(s), stress symptoms decrease in most individuals and most patients eventually recover, even from very severe traumatic experiences. Therefore, the evaluation time point selected for this study might underestimate the incidence of stress symptoms during longer follow-up periods. Furthermore, our study did not address the important question of how many highly stressed individuals subsequently experience chronic debilitating PTSD. Fourth, our study did not take into account vicarious trauma, in which family members may develop post-traumatic stress reactions in response to those experienced by other family members in the waiting room. Moreover, many relatives were family members of chronically ill patients and may have been carrying a heavy caregiver burden before the ICU admission. Thus some of the stress symptoms found at the 90-day interview were perhaps present before the admission. Fifth, in family members of survivors, a possible contribution of stress related to concern generated by caring for the patient at home was not evaluated. In conclusion, post-traumatic stress reaction consistent with a high risk of PTSD is common in family members of ICU patients 3 months after the ICU experience. Family members of patients who died in the ICU are at higher risk. Perceived inadequate information by ICU physicians and involvement in end-of-life decisions are independent determinants of the risk for posttraumatic stress reaction that might deserve consideration for preventive strategies. Further studies are needed to assess the rate of occurrence of PTSD in family members of ICU patients,

7 Azoulay, Pochard, Kentish-Barnes, et al.: Family Burden after Critical Care 993 the feasibility and usefulness of early screening, perhaps conducted by the ICU staff at the end of the ICU stay, and the level and qualitative characteristics of family information that must precede an invitation to share in decision making. General practitioners should be aware that they have a key role to play in detecting post-traumatic stress reactions in families of ICU patients. Conflict of Interest Statement : E.A. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; F.P. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; N.K.-B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; S.C. in the subject of this manuscript; J.A. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; C.A. in the subject of this manuscript; D.A. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; G.B. in the subject of this manuscript; P.E.B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; M.D. in the subject of this manuscript; T.F. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; R.G. in the subject of this manuscript; M.G.-O. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; C.G. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; D.G.-T. does not have a financial manuscript; J.H. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; M.J. does not have a financial manuscript; M.K. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; C.L. does not have a financial manuscript; J.L. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; J.L. does not have a financial manuscript; L.P. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; C.P. does not have a financial manuscript; J.R. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; F.S. does not have a financial manuscript; B.S. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Acknowledgment : The authors thank A. Wolfe, M.D., for helping with this manuscript. References 1. Pochard F, Azoulay E, Chevret S, Lemaire F, Hubert P, Canoui P, Grassin M, Zittoun R, le Gall JR, Dhainaut JF, et al. Symptoms of anxiety and depression in family members of intensive care unit patients: ethical hypothesis regarding decision-making capacity. Crit Care Med 2001;29: Azoulay E, Sprung CL. Family-physician interactions in the intensive care unit: effective communication empowers family members. Crit Care Med (In press) 3. Azoulay E, Chevret S, Leleu G, Pochard F, Barboteu M, Adrie C, Canoui P, Le Gall JR, Schlemmer B. Half the families of intensive care unit patients experience inadequate communication with physicians. Crit Care Med 2000;28: Azoulay E, Pochard F, Chevret S, Lemaire F, Mokhtari M, Le Gall JR, Dhainaut JF, Schlemmer B. Meeting the needs of intensive care unit patient families: a multicenter study. Am J Respir Crit Care Med 2001; 163: Wasser T, Pasquale MA, Matchett SC, Bryan Y, Pasquale M. Establishing reliability and validity of the critical care family satisfaction survey. Crit Care Med 2001;29: Curtis JR, Patrick DL, Shannon SE, Treece PD, Engelberg RA, Rubenfeld GD. The family conference as a focus to improve communication about end-of-life care in the intensive care unit: opportunities for improvement. Crit Care Med 2001;29:N26 N Way J, Back AL, Curtis JR. Withdrawing life support and resolution of conflict with families. BMJ 2002;325: Truog RD, Cist AF, Brackett SE, Burns JP, Curley MA, Danis M, DeVita MA, Rosenbaum SH, Rothenberg DM, Sprung CL, et al. Recommendations for end-of-life care in the intensive care unit: the Ethics Committee of the Society of Critical Care Medicine. Crit Care Med 2001;29: Yehuda R. Post-traumatic stress disorder. N Engl J Med 2002;346: Bleich A, Gelkopf M, Solomon Z. Exposure to terrorism, stress-related mental health symptoms, and coping behaviors among a nationally representative sample in Israel. JAMA 2003;290: Jones C, Skirrow P, Griffiths RD, Humphris G, Ingleby S, Eddleston J, Waldmann C, Gager M. Post-traumatic stress disorder-related symptoms in relatives of patients following intensive care. Intensive Care Med 2004;30: Zatzick DF, Jurkovich GJ, Gentilello L, Wisner D, Rivara FP. Posttraumatic stress, problem drinking, and functional outcomes after injury. Arch Surg 2002;137: Abbott KH, Sago JG, Breen CM, Abernethy AP, Tulsky JA. Families looking back: one year after discussion of withdrawal or withholding of life-sustaining support. Crit Care Med 2001;29: Cuthbertson SJ, Margetts MA, Streat SJ. Bereavement follow-up after critical illness. Crit Care Med 2000;28: Malacrida R, Bettelini CM, Degrate A, Martinez M, Badia F, Piazza J, Vizzardi N, Wullschleger R, Rapin CH. Reasons for dissatisfaction: a survey of relatives of intensive care patients who died. Crit Care Med 1998;26: Robinson SM, Mackenzie-Ross S, Campbell Hewson GL, Egleston CV, Prevost AT. Psychological effect of witnessed resuscitation on bereaved relatives. Lancet 1998;352: Sundin EC, Horowitz MJ. Horowitz s Impact of Event Scale evaluation of 20 years of use. Psychosom Med 2003;65: Jordhoy MS, Fayers P, Loge JH, Ahlner-Elmqvist M, Kaasa S. Quality of life in palliative cancer care: results from a cluster randomized trial. J Clin Oncol 2001;19: Horowitz M, Wilner N, Alvarez W. Impact of Event Scale: a measure of subjective stress. Psychosom Med 1979;41: Brady K, Pearlstein T, Asnis GM, Baker D, Rothbaum B, Sikes CR, Farfel GM. Efficacy and safety of sertraline treatment of post-traumatic stress disorder: a randomized controlled trial. JAMA 2000;283: Joseph S. Psychometric evaluation of Horowitz s Impact of Event Scale: a review. J Trauma Stress 2000;13: Sundin EC, Horowitz MJ. Impact of Event Scale: psychometric properties. Br J Psychiatry 2002;180: Aardal-Eriksson E, Eriksson TE, Holm AC, Lundin T. Salivary cortisol and serum prolactin in relation to stress rating scales in a group of rescue workers. Biol Psychiatry 1999;46: Mayou RA, Ehlers A, Hobbs M. Psychological debriefing for road traffic accident victims: three-year follow-up of a randomised controlled trial. Br J Psychiatry 2000;176: Zisook S, Chentsova-Dutton Y, Shuchter SR. PTSD following bereavement. Ann Clin Psychiatry 1998;10: Galea S, Ahern J, Resnick H, Kilpatrick D, Bucuvalas M, Gold J, Vlahov D. Psychological sequelae of the September 11 terrorist attacks in New York City. NEnglJMed2002;346: Yehuda R, McFarlane AC, Shalev AY. Predicting the development of post-traumatic stress disorder from the acute response to a traumatic event. Biol Psychiatry 1998;44: Chelluri L, Im KA, Belle SH, Schulz R, Rotondi AJ, Donahoe MP, Sirio CA, Mendelsohn AB, Pinsky MR. Long-term mortality and quality of life after prolonged mechanical ventilation. Crit Care Med 2004; 32: Im K, Belle SH, Schulz R, Mendelsohn AB, Chelluri L. Prevalence and outcomes of caregiving after prolonged mechanical ventilation in the ICU. Chest 2004;125: Sondergaard HP, Ekblad S, Theorell T. Screening for post-traumatic stress disorder among refugees in Stockholm. Nord J Psychiatry 2003; 57: Heyland DK, Rocker GM, Dodek PM, Kutsogiannis DJ, Konopad E, Cook DJ, Peters S, Tranmer JE, O Callaghan CJ. Family satisfaction with care in the intensive care unit: results of a multiple center study. Crit Care Med 2002;30: Curtis JR, Engelberg RA, Wenrich MD, Nielsen EL, Shannon SE, Treece PD, Tonelli MR, Patrick DL, Robins LS, McGrath BB, et al. Studying communication about end-of-life care during the ICU family conference: development of a framework. J Crit Care 2002;17: Azoulay E, Pochard F. Communication with family members of patients dying in the intensive care unit. Curr Opin Crit Care 2003;9: Heyland DK, Cook DJ, Rocker GM, Dodek PM, Kutsogiannis DJ, Peters

8 994 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL S, Tranmer JE, O Callaghan CJ. Decision-making in the ICU: perspectives of the substitute decision-maker. Intensive Care Med 2003;29: Azoulay E, Pochard F, Chevret S, Adrie C, Bollaert PE, Brun F, Dreyfuss D, Garrouste-Orgeas M, Goldgran-Toledano D, Jourdain M, et al. Opinions about surrogate designation: a population survey in France. Crit Care Med 2003;31: Rabow MW, Hauser JM, Adams J. Supporting family caregivers at the end of life: they don t know what they don t know. JAMA 2004;291: Azoulay E, Pochard F, Chevret S, Arich C, Brivet F, Brun F, Charles PE, Desmettre T, Dubois D, Galliot R, et al. Family participation in care to the critically ill: opinions of families and staff. Intensive Care Med 2003;10: Vincent JL. European attitudes towards ethical problems in intensive care medicine: results of an ethical questionnaire. Intensive Care Med 1990;16: Vincent JL. Cultural differences in end-of-life care. Crit Care Med 2001; 29:N52 N Sprung CL, Cohen SL, Sjokvist P, Baras M, Bulow HH, Hovilehto S, Ledoux D, Lippert A, Maia P, Phelan D, et al. End-of-life practices in European intensive care units: the Ethicus Study. JAMA 2003;290: Ferrand E, Robert R, Ingrand P, Lemaire F. Withholding and withdrawal of life support in intensive-care units in France: a prospective survey. French LATAREA Group. Lancet 2001;357: DeLisi LE, Maurizio A, Yost M, Papparozzi CF, Fulchino C, Katz CL, Altesman J, Biel M, Lee J, Stevens P. A survey of New Yorkers after the Sept. 11, 2001, terrorist attacks. Am J Psychiatry 2003;160: APPENDIX The following Famirea Study Group centers participated in the study. From the Intensive Care Units of the Saint-Louis Hospital, Paris; Henri Mondor Hospital, Créteil; Delafontaine Hospital, Saint Denis; Raymond Poincaré Hospital, Garches; Victor- Dupouy Hospital, Argenteuil; Central Hospital, Nancy; la Croix Rousse Hospital, Lyon; René Dubos Hospital, Pontoise; Saint- Joseph Hospital, Paris; Cochin Hospital, Paris; Lariboisière Hospital, Paris; CHIPS, Saint-Germain en Laye; Salengro Hospital, Lille; Le Kremlin Bicêtre Hospital, Le Kremlin Bicêtre; Sainte Marguerite Hospital, Marseille; Les Oudairies Hospital, La Roche Sur Yon; G. Montpied Hospital, Clermont Ferrand; Frejus Saint Raphaël Hospital, Frejus; Brabois Hospital, Nancy; Nemours Hospital, Nemours; and Lens Hospital, Lens.