Intensive Care Based Interventions to Reduce Family Member Stress Disorders: A Systematic Review of the Literature

We searched the electronic databases of PubMED, EM-BASE, and CINAHL to identify empirical studies comparing outcomes without and with the implementation of family support measures for patients admitted in the ICU (last search on September 8, 2020). We searched using free text and MeSH terms including: critical care, decision-making (appendix 1). We identified 3428 possible publications. These citations were managed and screened using the Abstrackr webtool [5]. After a pilot round of 200 citations to ensure uniform application of screening criteria, all citations were screened independently by two reviewers. Conflicts were resolved in group discussions. All citations deemed eligible at the abstract level were screened in full text in duplicate and independently (SR and TA), and disagreements were resolved by a third reviewer (TT).

Eligible studies enrolled adult family members of neonatal, pediatric, and adult patients hospitalized in a medical, neurological or surgical ICU for reasons other than traumatic brain injury or severe anoxic brain injury. These conditions were excluded because they reflect a minority of patients in an ICU census and are likely to have different emotional responses to the ICU process due to the traumatic or sudden nature of their admissions.

Eligible interventions targeted family members or surrogate decision makers for ICU patients and aimed to improve outcomes of knowledge, accuracy of risk perception, anxiety, stress or depression, or satisfaction with care and decision-making. We considered both formal decision aids and non-decision-aid supporting interventions which fulfilled the above criteria. We included randomized controlled trials (RCTs), and nonrandomized comparative studies with concurrent or historic controls. We excluded studies that did not report any information about the content (e.g., journal keeping, counseling) and delivery (timing, delivery format) of the intervention, or studies that did not report empirical outcomes.

Each study was extracted by an investigator, and confirmed by another, using pre-defined data extraction forms (Appendix 2). We used digitizing software to extract information from graphs (WebPlotDigitizer) [6]. We extracted information on the provenance of each paper (first author, title, journal and year of publication), its design (sample size, study type), characteristics of the patients, caregivers, the interventions, and data on outcomes, as described below. Three reviewers assessed the risk of bias of each study using the approach described in ROBINS I tool [7]. Any disagreements were resolved by discussion among all team members.

We extracted results for validated instruments that measure emotion and affect, including four instruments which measured stress, two which measured anxiety, two which measured depression, and two which measured behavioral changes in children as indicators of stress (Table 1) [8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20]. Table 1 provides brief descriptions for these outcome scales and their validation.

For all scales, we recoded results so that increasing values imply worsening outcomes, but left each outcome reported over its typical range without recoding them on a common (e.g., 0-100) range. To help contextualize the magnitude of observed differences in each scale, we report empirically derived confidence intervals. If such estimates are not readily available, we report the interquartile range of scores.

We did not perform quantitative analyses (including sensitivity analyses) because of substantial diversity in populations, interventions, and outcomes. Instead, we characterized each intervention over key dimensions which we used as thematic entities to structure our descriptions and organize our conclusions. In all analyses we divided eligible studies into pediatric or adult patient populations.

Data was first examined and separated by the nature of the Intensive Care Unit in which the studies were conducted. Studies which were conducted in a Pediatric or Neonatal ICU were separated from studies conducted in an adult patient population, due to the different relationships of caregivers and assessment tools. In total, 17 studies were included in the final analysis (Figure 1).

Fig 1

Interventions were initially described based on their content as well as the method of intervention delivery (Appendix 2). In assessing content, we categorized interventions as containing information regarding resources/procedures, containing general information pertaining to ICU care, containing information tailored to the patient, encouraging family participation in care, formal counseling, peer counseling, or use of navigator/facilitator support. The content of each intervention was briefly described for each relevant group. Intervention delivery was described in terms of personnel and timing with regards to the patient’s hospitalization. When available, the training of personnel who delivered the intervention was included, as well as their role on the ICU team, if one existed.

Further, we characterized each the intervention as “passive” or “not passive”. Passive interventions were those that focused on delivery of information, but without active engagement of the family or patient proxies in some way. Non-passive interventions allowed or encouraged engagement of the family, e.g., through participation in care, narrative writing.

Our literature search yielded 3428 publications for possible inclusion (Figure 1). Of these, 2835 were considered relevant for full review after abstract review by two of the authors (TA and SR). For any abstracts where there was uncertainty regarding inclusion, a third author (TT) reviewed the abstract and all three authors agreed on inclusion or exclusion. Abstract review identified 110 articles which were subsequently assessed in full. Of these, 17 were identified and included in the analysis. Of the 17, 8 pertained to pediatric populations (6 in the neonatal ICU and 2 in the pediatric ICU) and 9 to the adult ICU populations.

The 17 studies that were eligible for inclusion were systematically reviewed and summarized in tabular format for the study design and characteristics, intervention content, delivery and timing, and main findings. These were reviewed by three of the authors (TA, SR, and TT) for fidelity and the results are summarized in Table 2 for pediatric and Table 3 for adult populations. Of the 17 studies included, 13 of them were randomized control trials, and 4 were defined as quasi-experimental, totaling 3471 research participants. One of the studies (Curley) was published in 1988, 5 trails were published between 2004 and 2008 (Melnyk, Browne, Moreau, Kloos, and Lautrette), and the remaining trials were published after 2011 [21, 23, 24, 31, 32, 33]. Nine of the studies included were completed in the United States (Kloos, Carson, Curtis, White, Clarke, Curley, Brown, Melnyk, Amass), 5 in Europe (Weis, Garrouste-Orgeas, Franck, Moreau, Lautrette), 1 in Iran (Kadivar), 1 in Australia (Abdel-Latif), and 1 in China (Chiang)[22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37]. For one study (Amass), the relevant data was obtained from the author and not included in the original publication [33].

Eight studies evaluated participants associated with pediatric patient populations. Outcomes measured in these populations included several validated scales. The Parental Stress Score (PSS) was used in 5 of the studies, which results with scores between 18 and 90, with higher scores indicating more stress [8]. The Parenting Stress Index (PSI) was used by Browne et al and Melnyk et al with higher scores indicating more parental stress [9]. One study (Kadivar) used the Neonatal Index of Parent Satisfaction (NIPS) which results scores between 24 and 168 with higher scores indicating more parental satisfaction [10]. Two studies (Clarke-Pounder and Melnyk) used the State-Trait Anxiety Index (STAI), which is scored between 20-80, with higher scores indicating higher state or trait anxiety levels [11]. Finally, Weis et al used the Nurse Parent Support Tool (NPST) in addition to the PSS-NICU, which generates scores between 21-105, with higher scores indicating more positive experiences of support [12]. Of note, Melnyk et al had several time points of measure ranging from during the intervention to 12 months post-intervention. To allow for comparison to the other studies included, the 6-month observations were included in the results of this review.

There were 4 pediatric studies that were determined to have an intervention that was not passive. Browne found that education of mothers with infants hospitalized in the NICU regarding their infants improved overall stress scales [21]. The study authors were contacted to request original PSI data for this analysis but were unable to provide this information. Without the original data, the study was subsequently excluded from further analysis. Kadivar demonstrated that with narrative writing, parents’ satisfaction was higher by day 10 in the intervention group vs the control as measured by NIPS [22]. Curley et al showed that a model that had nurses engage family members to gain understanding of family and patient wishes reduced parental stress as measured by the PSS:PICU [23]. Melynk et al used an intervention titled “Creating Opportunities for Parent Empowerment” (COPE), which increased parental knowledge and participation in childcare and found trends towards improvement across several domains, though none were significant [24]. At 6 months they showed a reduction in parental stress in the intervention group, as well as lower anxiety.

Four pediatric studies were categorized as passive. Franck et al found no change in parental stress at one week as measured by PSS:NICU with delivery of a booklet that contained detailed information about pain and comforting along with demonstrations by a research nurse [25]. A second study (Abdel-Latif et al) included parents on rounds and found no difference in parental stress with the PSS or subscales [26]. Clarke-Pounder et al employed an intervention which gathered information from parents about care preferences and placed them in the chart for the provider [27]. This intervention demonstrated more reduction in anxiety between baseline and 2 weeks in the control group, but no change in the intervention group. Finally, Weis et al offered reflection sheets to the parents and showed no change in parental stress or satisfaction through this intervention [28].

There were 9 studies that focused on family members of adult patients. Outcomes were also evaluated with several different validated tools. Daly et al utilized the STAI, described in the pediatric results section [11]. Six of the studies (White, Garrouste-Orgeas, Lautrette, Moreau, Carson, and Amass) used the Hospital Anxiety and Depression Scale (HADS) which results scores between 0 and 21 for anxiety or depression, with higher scores indicating more symptoms consistent with either diagnosis [13]. Five studies (White, Garrouste-Orgeas, Lautrette, Carson and Amass) also used the Impact of Events Scale- revised (IES-r), a 22-item scale producing scores between 0 and 88, with higher scores being more consistent with symptoms of Post-Traumatic Stress Disorder (PTSD) [14]. In addition to the above two scales, White et al utilized the Patient Perception of Patient Centeredness (PPPC, modified for surrogates) which produces scores from 1 to 4 with lower scores indicating more patient/family centered care [15]. Carson et al and Amass et al also evaluated family satisfaction using the Family Satisfaction in the ICU-24 (FS-ICU24) questionnaire, scored between 24 and 100 with higher scores indicating more satisfaction [16]. Chiang et al utilized the Depression Anxiety Stress Scale- Chinese (C-DASS), a 21-item scale scored between 0-21 for each depression, anxiety, and stress, with higher scores indicating higher levels of each diagnosis [17]. The final paper to utilize multiple scales, Curtis et al, used the Patient Health Questionnaire (PHQ-9), the Generalized Anxiety Disorder-7 (GAD-7) survey, and the PTSD Checklist Civilian Version (PCL) to evaluate for symptoms of depression, anxiety, and PTSD, respectively [18, 19, 20]. For each metric, a higher score indicates more symptoms of the measured state. Curtis et al and White et al additionally evaluated length of stay (LOS) and cost of care.

Five of the included adult studies were categorized as instituting interventions that were not passive. White et al found that with engaging the family members with a “PARTNER” nurse, a nurse specially trained on supporting the family and meeting with them daily as well as coordinating provider team meetings, there were no changes in symptoms of anxiety or depression or PTSD compared to the control [29]. The intervention did demonstrate a significant reduction in ICU LOS and hospital LOS as compared to the control group. Using a “nurse navigator,” a nurse specially trained in mediation and communication techniques who met with the family to engage them and help express their concerns during provider meetings, Curtis et al demonstrated a significant reduction in symptoms of depression in the intervention vs the control 6 months after discharge of the patient, without a significant change in symptoms of anxiety or PTSD over the same time period [30]. The intervention did demonstrate a significant reduction in hospital LOS without a significant change in ICU LOS, and additionally demonstrated a significant reduction in both ICU and hospital costs. Lautrette et al implemented a physician training on the VALUE mnemonic (supplemental index), which trained physicians to allow family members to express themselves more during the family meetings as well as elicit and validate their feelings [31]. With this intervention, the authors reported a significant reduction in anxiety and depression and PTSD in the intervention vs control 90 days after discharge of the patient from the ICU. Kloos and Daly reported that with an intervention which instructed family members to complete a journal of both events (eg surgery timing, patient condition) and feelings, there was no change in anxiety state in the intervention as compared to the control [32]. Finally, Amass et al demonstrated that active engagement of family members in patient care resulted in a significant reduction in symptoms of PTSD in the intervention compared to the control group 90 days after patient discharge [33]. Family members who participated in the intervention seemed to have higher overall satisfaction with the care that their loved one received in the ICU than individuals who did not participate in the intervention, although this did not reach the level of statistical significance. However, the study noted no significant change in depression or anxiety scores in the intervention vs the control at 90 days post discharge of the patients.

There were 4 studies of adult populations categorized as passive. Moreau et al reported no difference in anxiety or depression scores as measured by HADS when information was delivered by a senior physician as compared to a junior physician [34]. When compared with usual care, Carson et al reported that palliative care lead meetings in the ICU demonstrated an increase in symptoms of PTSD but did not alter symptoms of depression or anxiety [35]. They additionally reported no difference in satisfaction scores using the FSICU-24. Chiang et al reported a reduction in depression symptoms and no difference in symptoms of anxiety when a nurse delivered a tablet to the family loaded with patient disease-specific information [36]. Finally, Garrouste-Orgeas et al report a reduction in anxiety and depression scores on the HADS when nurses were trained to deliver information to the families and participation with the family meetings [37]. They also report a reduction in PTSD symptoms as measured by IES-r scores.