مطالعه کنترل شده تصادفی از یک مداخله پیشگیرانه-درمانگر مبتنی بر اینترنت نشان دهنده برای اختلال حزن و اندوه طولانی مدت
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|37467||2014||12 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Behaviour Research and Therapy, Volume 61, October 2014, Pages 23–34
Abstract This trial assessed the feasibility, acceptability, tolerability, and efficacy of an Internet-based therapist-assisted cognitive-behavioral indicated prevention intervention for prolonged grief disorder (PGD) called Healthy Experiences After Loss (HEAL). Eighty-four bereaved individuals at risk for PGD were randomized to either an immediate treatment group (n = 41) or a waitlist control group (n = 43). Assessments were conducted at four time-points: prior to the wait-interval (for the waitlist group), pre-intervention, post-intervention, 6 weeks later, and 3 months later (for the immediate group only). Intent-to-treat analyses indicated that HEAL was associated with large reductions in prolonged grief (d = 1.10), depression (d = .71), anxiety (d = .51), and posttraumatic stress (d = .91). Also, significantly fewer participants in the immediate group met PGD criteria post-intervention than in the waitlist group. Pooled data from both groups also yielded significant reductions and large effect sizes in PGD symptom severity at each follow-up assessment. The intervention required minimal professional oversight and ratings of satisfaction with treatment and usability of the Internet interface were high. HEAL has the potential to be an effective, well-tolerated tool to reduce the burden of significant pre-clinical PGD. Further research is needed to refine HEAL and to assess its efficacy and mechanisms of action in a large-scale trial.
نتیجه گیری انگلیسی
Results Participant characteristics The randomized and treated sample (N = 84: nimmediate = 41, nwaitlist = 43) consisted of 27 (32%) men and 57 (68%) women. On average, participants were 55.37 years of age (SD = 10.30 years), reported that they had known the deceased for 32.31 years (SD = 15.30 years), and 78% were the diseased patient's spouse/partner. At the time of the first assessment (Time 1), average time since loss was 8.38 months (SD = 2.97 months). On baseline measures of alcohol use, scores on the AUDIT were well below the cut-off score of 8 for hazardous or harmful drinking (M = 3.01, SD = 2.56). Scores on the DAST were similarly negligible (M = .16, SD = .43), falling well below a score of 1, which would indicate a low level of drug use. Due to low base-rates and floor effects seen on the AUDIT and DAST at baseline, further analyses were not performed on these measures. Randomization check We conducted independent sample t-tests, correlations, or one-way ANOVAs (where appropriate) on Time 1 PG-13, BDI, PCL, and BAI scores, demographic variables, relationship to deceased, and time since loss. No significant differences were found between conditions on any of these variables (ps > .56) with the exception of the BAI. Scores on the BAI in the immediate group were higher (M = 35.38, SD = 11.16) than in the waitlist group (M = 34.47, SD = 7.52), although this difference was only marginally significantly different, t(81) = .19, p = .06. Despite the equivalence of the groups at randomization, there was one important difference between the groups. Unexpectedly, participants took substantially longer than 6 weeks to complete the HEAL intervention. The mean time to complete HEAL for participants in the immediate group was 24.15 weeks (SD = 13.92). Because the waitlist interval was fixed at 6 weeks, the time interval between the two conditions was not equivalent, complicating the comparison between the two groups. Consequently, prior to comparing these two groups, we took a number of steps to examine the potential impact time may have had on these comparisons. First, we examined the within-group correlations between time and outcome (as well as the plots of time versus outcome). Within-group correlations between time and outcome variables (adjusting for Time 1 scores) were all non-significant (ps > .16), and the plots did not suggest nonlinearity. To further bolster confidence that the time it took to complete HEAL was not a confounding variable, we also conducted analyses using the session-by-session PG-13 score that most closely corresponded to 6-weeks since beginning the program as the Time 2 score for the immediate group. This allowed us to compare the effects of time and condition using comparable data across the two groups. These findings are reported together with the primary analyses described below. Time since loss Because there are discordant opinions about the required amount of time since loss required to diagnose PGD and the role of time in the resolution of symptoms, we examined the potential impact of time since loss on the primary and secondary outcome variables prior to conducting the main analyses. Partial correlations between time since loss and Time 2 (post-treatment for the immediate group, post-6-week waiting period for the waitlist group) outcomes scores were all non-significant (ps > .21). We also examined the correlations between time since loss and both pre- and post-treatment scores (collapsed across conditions). Again, all correlations were non-significant (ps > .17). Given these non-significant correlations, time since loss alone cannot account for changes that may be observed between pre-treatment and post-treatment. Intent-to-treat analysis To examine whether participants in HEAL evidenced significant symptom improvement over their counterparts on a 6-week waiting list, we conducted an intention-to-treat (ITT) analysis. We used linear mixed models (LMM) with restricted maximum likelihood (REML) estimation in order to take advantage of all available data at each time point. Condition, time, and the condition by time interaction were entered in each model; education and income were included as covariates. Analyses were conducted on PG-13 scores (primary outcome) as well as BDI, PCL and BAI scores. Controlled between group effect sizes were calculated as a Cohen's d, subtracting the mean change from Time 1 to Time 2 for the waitlist group from the mean change in the immediate group, and dividing by the pooled Time 1 standard deviations ( Morris, 2008). HEAL resulted in significant reductions in PG-13 symptoms from Time 1 to Time 2 in the immediate treatment group compared to the waitlist group (Fig. 2 shows a graphic representation of the condition by time interaction for PG-13 scores). The condition by time interaction was also significant for each secondary outcome variable. That is, HEAL also resulted in significant reductions in depression, anxiety, and post-traumatic stress symptoms from Time 1 to Time 2 compared to the waitlist group. Table 2 shows the means by condition for the primary and secondary outcome variables, the F-statistic for each interaction, and the between-group effect sizes. Importantly, substituting the 6-week PG-13 session-by-session data for Time 2 scores for the immediate group yielded the same statistically significant interaction F(1, 78.15 = 15.29, p < .001) as when the post-intervention score was used as the second time point. Notably, after 6 weeks in the HEAL program, on average, PG-13 scores dropped 7.29 points, while scores for the waitlist group dropped only 2.11 points (t (78.15) = −3.91, p < .001). Change in prolonged grief symptom severity attributable to treatment. ... Fig. 2. Change in prolonged grief symptom severity attributable to treatment. PG-13 = Prolonged Grief Inventory. Figure options Table 2. Efficacy of HEAL: immediate versus waitlist comparisons. Time 1 Time 2 Time × condition d [95% CI] Waitlist Immediate Waitlist Immediate M SD M SD M SD M SD PG-13 34.99 7.46 34.39 8.11 32.84 9.11 24.70 8.33 F(1, 74.10) = 29.04** 1.10 [.63, 2.27] BDI 37.65 8.01 38.08 8.20 36.15 8.67 30.80 7.60 F(1, 72.63) = 14.19** .71 [.27, 1.15] PCL 38.33 11.28 39.73 11.99 37.31 12.74 28.11 10.06 F(1,71.87) = 27.68** .91 [.46, 1.36] BAI 31.52 7.52 35.22 11.16 30.31 6.78 29.18 9.39 F(1,73.99) = 10.68* .51 [.07, 94] Note. *p < .01. **p < .001. PG-13 = Prolonged Grief Inventory; BDI = Beck Depression Inventory; PCL = Posttraumatic Checklist; BAI = Beck Anxiety Inventory. Table options Clinical significance We used two methods to explore clinical significance. First, we conducted chi-square tests comparing the number of participants in each group who met criteria for PGD at Time 1 to those meeting criteria at Time 2. PGD caseness was defined in the following manner: (1) significant separation distress (i.e., feelings of longing, pangs of grief) experienced at least once a day over the past month, and (2) at least five additional cognitive, emotional, and behavioral symptoms (e.g., confusion about one's role in life; difficulty accepting the loss; avoidance, distrust, bitterness, or emotional numbing related to the loss) rated as “quite a bit” or “overwhelmingly” over the past month. Analyses suggested no significant differences between the immediate treatment and waitlist groups in the number of participants meeting criteria for PGD at Time 1 (χ2 = 1.19, p = .28, nimmediate = 15 (37%), nwaitlist = 11 (26%)). However, there was a significant difference between conditions at Time 2 (χ2 = 4.12, p = .04); after completing HEAL, only two participants (6%) in the immediate treatment group continued to meet criteria for PGD, compared to 10 (24%) in the waitlist condition. Second, we explored clinical significance by following the procedures outlined by Jacobson and Truax (1991). Because norms on the PG-13 for a functionally normal population are not yet available, we used the more conservative CS cut-off of two standard deviations (in the functional direction) from the pre-treatment mean of the immediate group (CS cut-off = 19.22). To calculate the reliable change index (RCI), we estimated the test-retest reliability of PG-13 using scores from the first and third sessions of HEAL (approximately 1–2 weeks apart, on average; α = .86), and used the formula presented by Jacobson and Truax (1991; pg. 14). We then used the following classification scheme to estimate clinical change. Participants were deemed: (1) recovered if they passed CS and RCI criteria; (2) improved if they passed RCI only; (3) unchanged if they did not pass CS or RCI; or (4) deteriorated if they passed RCI in negative direction. Based on this conservative scheme, in the immediate group, 8 participants were classified as recovered, 11 were classified as improved, 12 were unchanged, and one participant deteriorated. In contrast, only 1 participant was classified as recovered in the waitlist condition, with 5 improving, 34 remaining unchanged, and 2 deteriorating. Pooled study findings To examine the overall impact of HEAL over time, we pooled ITT data across conditions and used LMM with REML to compare pre-intervention scores to post-intervention, 6-week follow-up, and 3-month follow-up scores (immediate group only; Time 4 assessment for waitlist group was conducted at 6 weeks). Controlled within-group effect sizes were calculated as Cohen's ds by dividing the mean difference between pre-intervention scores and each follow-up score by the appropriate pooled standard deviation. Table 3 contains the means, standard deviations, t-tests for comparisons to pre-intervention scores, and within group effect sizes for the outcome variables for all participants prior to beginning HEAL, upon completing HEAL, 6 weeks later, and 3 months later for the immediate group. Compared to pre-intervention scores, HEAL resulted in significant reductions to PG-13, BDI, PCL, and BAI scores at each follow-up assessment, ps < .02. Table 3. Efficacy of HEAL: pooled groups over time. Pre-intervention Post-intervention 6-week Follow-up 3-month Follow-upa M SD M SD t(df) d [95% CI] M SD t(df) d [95% CI] M SD t(df) d [95% CI] PG-13 33.52 8.42 24.08 8.70 t(68.86) = −10.14 1.10 [.81,1.39] 23.35 8.09 t(70.44) = −10.85 1.23 [.92, 1.54] 23.07 8.04 t(67.86) = −11.12 1.24 [.79, 1.74] BDI 36.95 8.07 30.52 6.99 t(74.21) = −7.51 .85 [.58,1.12] 30.15 6.99 t(75.77) = −7.45 .90 [.60, 1.19] 30.42 7.72 t(71.48) = −6.40 .83 [.37, 1.27] PCL 38.27 11.97 28.31 9.58 t(74.53) = −9.02 .92 [.66, 1.17] 27.37 9.92 t(74.94) = −9.37 .99 [.71, 1.26] 27.84 9.88 t(69.19) = −8.26 .95 [.52, 1.37] BAI 32.68 9.50 28.47 8.52 t(72.61) = −4.90 .47 [.26, .67] 27.24 7.63 t(73.67) = −5.91 .63 [.39, .87] 27.85 8.81 t(67.40) = −4.55 .53 [.20,.85] a Immediate group only. T-statistic is derived from the fixed effects estimates of the LMM. PG-13 = Prolonged Grief Inventory; BDI = Beck Depression Inventory; PCL = Posttraumatic Checklist; BAI = Beck Anxiety Inventory. Table options Clinical significance To further examine the clinical significance of these findings over time, we first conducted chi-square tests comparing the number of participants who met criteria for PGD prior to starting HEAL to those meeting criteria at post-intervention, 6-week, and 3-month follow-ups (3-month data again was only available for immediate group). Upon completing HEAL, significantly fewer participants met criteria for PGD, and this reduction was maintained over time (ps < .004). Specifically, prior to receiving the intervention, 25 participants across both conditions (30.1%) met criteria for PGD. Upon completion of the intervention, only six (9.0%) continued to meet criteria for PGD (χ2 = 10.13, p < .001), and that number was further reduced to five (7.6%) by the 6-week follow-up assessment (χ2 = 11.62, p < .001). Among those assessed at 3 months (immediate group only), the number of participants meeting criteria for PGD was reduced from four at 6-week follow-up to two (6.5%) by the 3-month follow-up (χ2 = 8.14, p = .004). Then, utilizing the conservative approach to Jacobson and Truax (1991) criteria described above, we found 19 participants were classified as recovered, 18 as improved, 28 as unchanged, and 2 as deteriorated at the post-test assessment. At the six week follow-up, 18 were classified as recovered, 23 as improved, 24 as unchanged, and 1 as deteriorated. At the three month follow-up (limited to those who were in the Immediate condition), 9 participants were classified as recovered, 13 as improved and 9 as unchanged (no participants were classified as deteriorated). Therapist contact Overall, most participants had minimal contact with the study therapist and primarily used the website to guide them through the intervention. The average number of therapist-initiated phone calls was considerably less than one per week (.24; SD = .26, range = 0–1.91), roughly once every four weeks, for each participant during the intervention-phase. These calls included the initial welcome/orientation call, final feedback call, and calls made in response to user alerts or online requests for assistance. Calls typically ranged from 2 to 10 min. During the intervention, the study therapist also initiated an average of .60 emails per week/per participant (SD = .38, range = .27–3.10) and left an average of .15 voicemails per week/per participant (SD = .20, range = 0–1.27). Emails consisted primarily of minor logistical or administrative contacts, brief acknowledgments of progress on homework assignments, or prompts to logon to the website after periods of inactivity. There were no significant differences between the average number of telephone calls or emails sent from the study therapist per group (waitlist versus immediate start). We examined whether each type of therapist contact, average number of contact per week, and total number of contact were related to the outcome variables, controlling for pre-intervention scores. Although there were several significant correlations (and several approaching significance), these correlations were all positive, suggesting that increased therapist contact was associated with greater symptomatology rather than improvements in outcome (see Table 4). This is not surprising given that much of the therapist contact was initiated in response to high symptom self-report by the participant. Table 4. Correlations between therapist contact and outcome variables controlling for time 1 scores – all participants completing HEAL. Time 2 score PG-13 BDI PCL BAI Emails −.077 −.085 .004 .069 Voicemails .241* .218† .206 .315* Telephone calls .225† .231† .220† .104 Average emails per week −.017 .071 .000 −.107 Average voicemails per week .221† .159 .167 .155 Average phone calls per week .312* .266* .230† −.020 Total therapist contact .084 .075 .131 .182 Average weekly therapist contact .228† .227† .173 −.008 Note. *p < .05. †p < .08. Correlations were run separately for each outcome variable and controlled for Time 1 scores. Table options Participant feedback and experience All participants who completed HEAL (N = 67) provided feedback about their experience of the intervention and website as well as the perceived quality of the intervention content. On the Protocol Evaluation Questionnaire, the vast majority reported a “somewhat positive” (44.8%) to “extremely positive” (43.3%) reaction to the website. Only two participants endorsed “somewhat negative” (1.5%) and “extremely negative” (1.5%) reactions. Most users endorsed the idea of using a website to help with grief problems as “quite a bit appealing” (46.3%) or “extremely appealing” (32.8%), whereas a minority found it “just acceptable” (14.9%), “somewhat unappealing” (4.5%), or “extremely unappealing” (1.5%). The ease with which participants were able to move around the website was rated as “extremely easy” (58.2%), “moderately easy” (34.3%), and “moderately difficult” (7.5%). The PSSUQ subscale scores were consistent with these findings. Each PSSUQ item is rated from one to seven, the latter indicating a strongly negative reaction. The mean System Usefulness subscore for all participants was 3.02 (SD = 2.16), which indicates that users found it fairly easy and efficient to learn to navigate the website and use it to complete tasks. Similarly, the mean Information Quality subscore was 2.95 (SD = 2.06), suggesting that participants mostly found the information about how to use the website to be clear and easy to follow. There were no significant differences on the PSSUQ between the immediate treatment and waitlist groups. Participant feedback about the value of the information and activities provided in HEAL was also very positive. Based on the Protocol Evaluation Questionnaire, participants found the content of HEAL to be logical (M = 7.16 out of 9.0, SD = 1.7). Most (77.6%) participants reported that the HEAL program provided “just the right amount” of information, while 16.4% would have preferred more information, and 6.0% thought there was “somewhat too much” information. Further, 77.6% rated that the instruction level on HEAL “just right” while a smaller percentage found it “somewhat too basic” (20.9%) or “far too basic” (1.5%). Participants were largely satisfied with the content of the program, with most participants reporting that they learned a “moderate” (53.7%) or “large” amount (35.8%) from the program. Further, when asked how interesting they found the content, 43.3% responded that it was “extremely interesting,” and another 53.7% rated it as “somewhat interesting.” Overall, over 90% consistently rated the individual components of HEAL as “moderately” to “extremely” valuable. Participants also reported they would be “quite confident” in recommending HEAL to a friend experiencing similar problems (M = 7.37 out of 9.0, SD = 1.89). Completers versus non-completers We also were interested in examining factors potentially related to participants' failure to complete the intervention. On average, participants who did not complete HEAL spent 12.18 weeks (SD = 9.56) in the program and completed a mean of 4.06 sessions (SD = 3.26). There were no differences between completers and non-completers on any of the demographic variables, pre-intervention outcome measures (PG-13, BDI, PCL, BAI), or average weekly therapist contact while in the program (ps > .11), with two exceptions. Non-completers received more average weekly email (M = .96, SD = .76) from the therapist while in the program compared to completers (M = .53, SD = .18), although this difference was only marginally significant (t(12.26) = 1.97, p = .07) and not unexpected due to the fact that infrequent logins prompted therapist contact. The other difference was in employment status as reported at Time 1. A larger percentage (58.8%) of non-completers were employed full-time than completers (43.3%; χ2 = 12.22, p = .032). This suggests that competing commitments in the work place may have made it difficult for some participants to engage in the program.