Inclusion
The data from this study were part of a larger project in three Dutch and three Belgian burn centres that focused on the social impact of burns. Previous work focused on PTSD symptoms and quality of life in burn survivors themselves [36]. Patients and their partners were recruited between October 2013 and October 2015 and followed for 18 months. Inclusion criteria for patients were: hospital stay of >24 h following the burn event, age of 18 years or older and proficiency in Dutch. The latter two criteria also applied to partners. Exclusion criteria were: psychiatric problems that interfere with the comprehension of questionnaires (e.g., psychosis, cognitive problems), and inhalation injury without external burns. The first criterion also applied to partners.
Sample and missing data
A total of 120 partners enrolled in the study. Of these partners, 111 (92.5%) completed the predictor measures and PTSD symptom measures in the hospital (denoted as T1) and were included in the final analyses. The excluded partners (n = 9) comprised relatively more men (n = 5, 62.5%) compared to the included partners (n = 22, 19.8%), ꭓ2(1) = 7.75, p = .02, but did not differ significantly (p > .05) from included partners with respect to age, number of surgeries, or any of the five PTSD measurements.
Over time, missing data on some of the PTSD measurements was present. The number of partners that completed (at least 19 of the 22 items of the) PTSD measures at 3, 6, 12 and 18 months was 94 (84.7%), 90 (81.1%), 76 (68.5%) and 79 (71.2%) respectively. Sixty-nine (62.2%) partners completed all five measurements. Respondents lost to follow-up did not differ from those participating in the study at 18 months in terms of gender, ꭓ2(1) = 0.03, p = 1.00, number of surgeries, t(109) = -0.15, p = .88, total body surface area affected, t(118) = -0.23, p = .82, and acute PTSD symptoms at T1, t(109) = -0.03, p = .98, but drop-outs were significantly younger, M = 39.0 versus M = 45.8, t(109) = -2.36, p = .02. Little’s Missing Completely At Random (MCAR) test in the final sample showed that missing data were random, ꭓ2(72) = 81.63, p = .21.
Procedure
Patients and their partners were invited to participate in the study by a local researcher while patients stayed in the burn center. After they received oral and written information about the study, they provided written informed consent. Patients and partners completed T1 during hospitalisation of the patient and the follow-ups at 3, 6, 12 and 18 months postburn (T2 to T5) by postal mail. The study was approved by ethics boards in the Netherlands and Belgium (NL44682.094.13 and B670201420373).
Measures
Post-Traumatic Stress Disorder Symptoms. The Impact of Event Scale-Revised (IES-R) [37] was used to asses partners’ PTSD symptoms. The IES-R is a self-report questionnaire with 22 items that measure symptoms in the past week. Answers were given on a 5-point Likert scale and summed to obtain a total score ranging from 0-88, with scores of 33 and higher indicating a possible diagnosis of PTSD [38]. If at least 19 of the 22 items were completed, sum scores were calculated based on the mean of the completed items. The IES-R has been validated in Dutch trauma populations and showed good psychometric properties [39]. Reliability of the IES-R in the current study was excellent, with Cronbach’s alpha between .94 and .96, at the five measurements. Partners’ PTSD symptoms were assessed during hospitalization of the patient, and at 3, 6, 12 and 18 months postburn.
Subjective appraisal of life threat and emotions. Partners reported their appraisal of the life-threatening nature of the injury through a single question (yes/no): “At any time, did you think your partner would not survive the burn event?” Psychometric properties of this item were not assessed, but previous studies have supported the validity of the measure [e.g. 40]. Emotions directly related to the burn event were assessed with the following question: “To what extent do the following emotions apply when you think about the accident that caused the burn?”. From the assessed emotions, guilt and anger were evaluated in the present study. Answers were rated on a 5-point Likert scale ranging from 0 (not at all) to 4 (a lot). This measure was previously used in burn studies [19, 21] and was assessed during hospitalisation of the burn survivor.
Rumination. The rumination scale of the Cognitive Emotion Regulation Questionnaire (CERQ) [41] was used to assess to which extent partners use this cognitive coping strategy in response to their partner’s burn event. The rumination scale comprises four items, e.g. “I am preoccupied with what I think and feel about what I experienced”. Answers were rated on a 5-point Likert scale ranging from 1 ‘(hardly) ever’ to 5 ‘(hardly) always’. In the current study, mean scores on the rumination scale ranged from 1 to 4.75. The Dutch version of the CERQ demonstrated good factorial validity and reliability (cronbach’s α = .83) in the general population [41]. Cronbach’s alpha in this study was .87. Partners’ rumination was assessed during hospitalisation of the burn survivor.
Burn characteristics. Number of surgeries, total body surface area (TBSA) burned, length of stay in the hospital, and whether the patient received mechanical ventilation (yes/no, duration) were recorded from the patient’s medical file. Number of surgeries indicates the number of skin graft procedures that was required to cover the wounds and is considered an indicator of burn severity. TBSA is the estimated percentage of the body covered with partial and full thickness burns.
Statistical analysis
Descriptive analyses were conducted in IBM SPSS v24. The potential predictors were correlated with PTSD symptom scores at each time point. Longitudinal trajectories of PTSD symptoms among partners of burn survivors were estimated using linear growth curve modeling (LGM) in Mplus 8.3 [42]. Full information maximum likelihood (FIML) was used to handle missing data in the main analyses, because Little’s MCAR test showed that data were missing completely at random. To account for the non-normality of some of the variables, Maximum Likelihood (ML) estimation with bootstrapped confidence intervals was used with 10,000 draws.
To find a model that best described the data, the adequacy and model fit of different growth models was evaluated. The slope growth factors represented the timing of the measurements since the burn event. Because a standard growth model with a single intercept plus linear slope did not fit the data well, and, to our knowledge, this was the first study to apply such a model to data on partners, we decided to explore a series of models to identify the best fitting curve. A complete overview of the consecutively evaluated models can be found in an additional file [see Additional file 1], including model fit, warnings and conclusions for each model. The results section provides a summary of this process and a choice for a final model. Note that because of the explorative nature of our approach, in the discussion section we highlight the need for replication of our model.
The predictors included in the final model were gender, number of surgeries, perceived life threat, anger, guilt and rumination. Age was not included in the final model, because addition of this variable resulted in an inadequate model with a bad fit (see Additional file 1, models 8-12) – again, replication in future research is needed. Anger, guilt and rumination were grand mean centered to aid interpretation of the intercept and slope estimates. Model fit was evaluated with the comparative fit index (CFI), the Tucker-Lewis Index (TLI), and the root mean square error of approximation (RMSEA). Models with a TLI and CFI >0.90 and RMSEA <0.08 indicate an acceptable fit, and models with TLI and CFI values >0.95 and RMSEA values <0.05 indicate good fit to the data [43, 44].
To evaluate the association of the predictors with PTSD symptoms 18 months postburn, a multiple regression analysis was performed in Mplus in two steps. In the first step, all predictors were added to the regression. In the second step, to evaluate prediction of the baseline-adjusted change of PTSD symptoms, PTSD symptoms at T1 was included in the model to correct for initial levels of PTSD symptoms.