Effects of psychosocial support interventions on survival in inpatient and outpatient healthcare settings: A meta-analysis of 106 randomized controlled trials.

BACKGROUND: Hospitals, clinics, and health organizations have provided psychosocial support interventions for medical patients to supplement curative care. Prior reviews of interventions augmenting psychosocial support in medical settings have reported mixed outcomes. This meta-analysis addresses the questions of how effective are psychosocial support interventions in improving patient survival and which potential moderating features are associated with greater effectiveness. METHODS AND
FINDINGS: We evaluated randomized controlled trials (RCTs) of psychosocial support interventions in inpatient and outpatient healthcare settings reporting survival data, including studies reporting disease-related or all-cause mortality. Literature searches included studies reported January 1980 through October 2020 accessed from Embase, Medline, Cochrane Library, CINAHL, Alt HealthWatch, PsycINFO, Social Work Abstracts, and Google Scholar databases. At least 2 reviewers screened studies, extracted data, and assessed study quality, with at least 2 independent reviewers also extracting data and assessing study quality. Odds ratio (OR) and hazard ratio (HR) data were analyzed separately using random effects weighted models. Of 42,054 studies searched, 106 RCTs including 40,280 patients met inclusion criteria. Patient average age was 57.2 years, with 52% females and 48% males; 42% had cardiovascular disease (CVD), 36% had cancer, and 22% had other conditions. Across 87 RCTs reporting data for discrete time periods, the average was OR = 1.20 (95% CI = 1.09 to 1.31, p < 0.001), indicating a 20% increased likelihood of survival among patients receiving psychosocial support compared to control groups receiving standard medical care. Among those studies, psychosocial interventions explicitly promoting health behaviors yielded improved likelihood of survival, whereas interventions without that primary focus did not. Across 22 RCTs reporting survival time, the average was HR = 1.29 (95% CI = 1.12 to 1.49, p < 0.001), indicating a 29% increased probability of survival over time among intervention recipients compared to controls. Among those studies, meta-regressions identified 3 moderating variables: control group type, patient disease severity, and risk of research bias. Studies in which control groups received health information/classes in addition to treatment as usual (TAU) averaged weaker effects than those in which control groups received only TAU. Studies with patients having relatively greater disease severity tended to yield smaller gains in survival time relative to control groups. In one of 3 analyses, studies with higher risk of research bias tended to report better outcomes. The main limitation of the data is that interventions very rarely blinded personnel and participants to study arm, such that expectations for improvement were not controlled.
CONCLUSIONS: In this meta-analysis, OR data indicated that psychosocial behavioral support interventions promoting patient motivation/coping to engage in health behaviors improved patient survival, but interventions focusing primarily on patients' social or emotional outcomes did not prolong life. HR data indicated that psychosocial interventions, predominantly focused on social or emotional outcomes, improved survival but yielded similar effects to health information/classes and were less effective among patients with apparently greater disease severity. Risk of research bias remains a plausible threat to data interpretation.

Introduction

Decades ago, researchers found that psychosocial support interventions (e.g., survivor groups and individual nurse support sessions) may improve not only patient quality of life but also patient survival [1,2]. Subsequent evidence regarding patient survival has been mixed [3].

Adequate support among medical patients has been linked to better outcomes, while those that lack adequate support systems have poorer outcomes including greater hospitalization, mortality, and medical costs—such that evaluations of supportive psychosocial interventions have been recommended in healthcare settings [4]. Substantial epidemiological evidence supports the link between psychosocial functioning and health outcomes, including meta-analyses indicating that presence or absence of social support predict all-cause mortality to an extent equivalent to other leading indicators of health (e.g., BMI and smoking cessation) [5-7]. The accumulated research evidence meets the Bradford Hill criteria, establishing low psychosocial support as a causal risk factor for premature mortality [8]. Level of psychosocial functioning has been shown to influence health risk through both emotional coping/resilience and behavioral modeling/motivation [9,10]. However, less is known concerning whether emotional and behavioral support from healthcare professionals can improve medical patients’ survival [4]. Given mounting evidence of health consequences of poor psychosocial functioning, the medical community can benefit from evaluating which psychosocial interventions most improve patient survival [11].

Over the past 4 decades, dozens of psychosocial support interventions have been evaluated for medical patients; accumulated literature on the topic is extensive but diverse. These include interventions conducted in patients’ homes, in support groups, or via telephone/online conversations. Some psychosocial interventions focus on behavior, explicitly supporting patients’ modification of health behaviors. This is based on evidence demonstrating that social support is linked to improved medical adherence [12,13], physical activity [14], sleep [15], and healthcare service utilization [16]. Other psychosocial interventions focus more specifically on emotion, explicitly supporting patients’ coping with distress. Abundant research evidence suggests that psychosocial distress co-occurs with physical disease, with bidirectional relationships that influence disease progression (e.g., appraisal and self-regulation ability) [4]. Research indicates that psychosocial functioning not only affects relevant social capital (e.g., access to health information and improved trust of healthcare) [17] but can also reduce inflammation and improve systemic circulation [18-20]. More specifically, even short-term emotional management interventions can influence inflammatory gene expression [21]. The number of psychosocial interventions with medical patients has multiplied rapidly in recent years, with interventions including multiple overlapping components (e.g., reducing distress and enhancing healthcare utilization). Before the complexity increases further, it would be useful to take stock of extant data by comparing psychosocial interventions across study, intervention, and patient characteristics.

Prior meta-analyses of psychosocial support interventions have evaluated patient survival [22-Cancer Med. 2019 ">43]; however, these were susceptible to error due to low numbers of studies included (range = 1 to 36, M = 11.2). Also, few previous meta-analyses have identified effective/ineffective intervention attributes, and most have had limited scope (e.g., breast cancer survivor groups). Although specificity in research is usually optimal, an unintended consequence has been ignoring the reality that professionals across medical specialties use similar psychosocial interventions. Thus, to evaluate differences across contexts, we have conducted what to our knowledge is the largest meta-analytic review to date, including 3 times the number of studies of any prior meta-analysis that we could locate on the topic. We sought to evaluate the overall degree to which psychosocial support interventions improve survival among patients receiving curative or rehabilitative care—and to specifically compare psychosocial interventions emphasizing behavioral support (e.g., modeling/motivation to engage in health behaviors such as physical activity) with those focused primarily on social/emotional support (e.g., emotional resilience following surgery). We also investigated outcome differences across study risk of bias and (a) study characteristics: duration of intervention, length of follow-up, type of control group, and patient psychosocial improvement; (b) intervention type: group meetings, telephone/online support, home visits, and family inclusion; and (c) patient characteristics: age, gender, disease, and mortality rate.

Methods

Search strategy

This study is reported as per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA guidelines [44]; S1 Checklist). We sought published and unpublished studies written in any language investigating the effects of psychosocial support interventions on medical patient survival. All authors participated in searching studies completed between January 1980 and October 2020, accessed using Embase, Medline, Cochrane Library, Alt HealthWatch, CINAHL, PsycINFO, Social Work Abstracts, and Google Scholar. To locate all relevant articles, we used an extensive list of search terms, manually examined the reference sections of both prior reviews and studies meeting the inclusion criteria, and contacted authors of included studies (S1 Text).

Study selection

The meta-analysis included randomized controlled trials (RCTs) reporting data of medical patients’ survival as a function of a real-time intervention providing psychological, emotional, and/or social support. We included studies of patients with a health condition likely to result in death if untreated, and who were recruited from healthcare settings (e.g., hospitals, rehabilitation clinics, or inpatient/outpatient databases). We excluded patients with solely mental health disorders (e.g., anxiety or dementia) because those conditions contribute indirectly to mortality, and we also excluded mortality resulting from accident, suicide, or violence as well as mortality data combined with morbidity/hospitalization.

As the majority of psychosocial support interventions described in the literature involve multiple components, we included interventions with mixed components (e.g., group psychotherapy, nurse visits, and telephone support) and coded for differences to compare outcomes. We excluded those providing only psychoeducation or disease management and those consisting solely of one-on-one psychotherapy, which historically has been a distinct kind of intervention deserving separate systematic review. We similarly excluded hospice or palliative care interventions which deserved separate review because of their focus on improving quality of life, not necessarily length of life, which is the observed outcome of this meta-analysis specific to curative and rehabilitative care. S1 Table provides detailed inclusion/exclusion criteria.

Data analysis

A team of 2 raters coded each article; subsequently, another team of 2 raters independently coded the same article. Teams resolved discrepancies through manuscript scrutiny until achieving consensus. Coders extracted (a) number of participants with composition by gender and average age; (b) length of intervention and follow-up; (c) type of intervention; and (d) multiple indicators of study risk of bias. Effect size data were hazard ratios (HRs) and odds ratios (ORs); when studies reported other values (e.g., regression coefficients or Cohen’s d), we transformed them to OR using multiple effect size calculators available online. Data were extracted from the longest follow-up period; when studies contained multiple effect sizes at the same time point (e.g., across subsamples), averaged values were weighted by SE. When reports explicitly tracked mortality but no participants died in either condition, we coded the effect size as OR = 1. We sought effect sizes from multivariable models but calculated OR from survival frequency counts when statistical models were unreported. Stata 16, SPSS 25, and Comprehensive Meta-Analysis 3 were used to calculate random effects weighted models in data aggregation and in subsequent subgroup analysis and meta-regressions.

Our data analysis plan (S2 Text) was to (a) report descriptive statistics of study characteristics; (b) calculate random effects weighted omnibus HR and OR values and also indicators of between-study heterogeneity (Q and I); (c) conduct subgroup analysis across intervention type (behavior focused versus social/emotional focused); (d) report meta-regressions separately for study, intervention, and patient characteristics; and (e) estimate the likelihood of publication bias. We did not prespecify which variables to include in the meta-regressions but clustered them according to study, intervention, and participant characteristics. We reported a subgroup analysis contrasting behavioral support with social/emotional support as a result of reviewer feedback, not as a prespecified analysis. We prospectively planned to evaluate the likelihood of publication bias estimates using funnel plots, the trim and fill method, and Egger and Peters regression tests. This meta-analysis is registered with Open Science Framework (3nj8u), with data available at https://osf.io/3qydb/.

Results

Description of included studies

We located 42,054 studies and screened 909 using the full text (Fig 1). Nonredundant effect sizes were extracted from 106 RCTs [1–3, 45–147] conducted in locations as follows: 50 (47%) in Europe, with 22 in Scandinavia, 11 in the United Kingdom, 6 in the Netherlands, 4 in Germany, and 7 other; 35% in North America, with 28 in the United States and 10 in Canada; 10 in Asia; 6 in Australia; and 2 in Africa. Data involved a total of 40,280 participants, whose average age was 57.2 years (SD = 9.9, range = 11 to 78), with an average of 52% females and 48% males. Across all studies, 81 (76%) involved medical outpatients, 20 (19%) recruited hospitalized inpatients, and 5 (5%) involved both. Patients had cardiovascular disease (CVD) in 44 studies (42%), cancer in 38 (36%) studies, or other conditions in 24 (22%) studies; a total of 102 (96%) reported all-cause mortality, with 2 reporting CVD mortality, 1 reporting cancer mortality, and 1 reporting HIV-related mortality.

Fig 1

PRISMA flow diagram of study selection process.

PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RCT, randomized controlled trial.

Regarding intervention focus, 34 of the 106 studies (32%) provided psychosocial behavioral support, explicitly focusing on health behaviors, and 72 (68%) focused on social/emotional support. Many studies included only 1 form of intervention: 46 (43%) in-person group meetings, 11 (10%) telephone/online sessions, 9 (8%) home visits, and 7 (7%) in-person individual sessions; the remaining 34 (32%) provided a combination of formats. Interventions were conducted by nurses or medical staff in 37 studies (35%), social workers or mental health professionals in 32 studies (30%), peers with the same medical condition in 4 studies (4%), combinations of those groups in 24 studies (23%), and family members in 1 study, with 8 studies unspecified. Across all interventions, 67 (63%) intended to foster patient social relationships with previously unknown individuals; 20 (19%) provided support only from professional staff; and although 34 (32%) invited family members, only 10 of those focused on family–patient relationships. On average, each intervention session lasted 83 minutes (SD = 46.1; excluding 1 day-long intervention [97]), with 14.7 total sessions (SD = 15.9) over 7.5 months (SD = 8.0). Researchers followed participants after the intervention for an average of 25.6 months (SD = 38.5), during which an average of 13.6% of participants died (SD = 20.3).

Of the 106 RCTs, 87 reported survival data for discrete time periods (transformed to OR), and 22 reported data in terms of survival time (HRs), with 3 studies [74,85,124] reporting both metrics. These types of studies differed in several ways (Table 1). Compared with studies reporting only OR data, studies reporting HR data tended to have an earlier date of initiation and longer total duration, with more female participants, more sessions, longer follow-up, and a correspondingly higher proportion of patient mortality by study conclusion. S2 and S3 Tables contain detailed information about individual studies by data type.

Table 1

Study characteristics by type of survival data reported.

Variable	OR		HR
	Mean	SD	Mean	SD	t	p
Participants
Number of participants	331	503	571	1,276	0.9	0.40
Participant average age	57.5	10.0	55.7	9.9	0.7	0.46
Participant % female	47.2	30.8	69.7	31.4	3.0	0.003
Participant % attrition	9.8	11.2	9.9	15.5	0.1	0.97
Participant % mortality	8.2	11.3	36.7	31.2	4.0	0.001
Interventions
Number of sessions	12.2	14.3	24.3	18.2	2.8	0.002
Minutes of each session	85.4	48.7	77.3	35.4	0.7	0.51
Months of intervention	6.9	8.3	10.0	6.5	1.6	0.12
Months of follow-up	15.6	23.3	65.6	58.1	3.9	0.001
Year initiated	2003	9.0	1996	8.1	3.3	0.002

Note: Independent samples t tests compared 22 studies reporting HR data with 84 studies reporting only OR data; 3 studies reporting both metrics are included in the HR data.

HR, hazard ratio; OR, odds ratio.

Main analyses

Across 87 observations at fixed time periods, the average was OR = 1.20 (95% CI = 1.09 to 1.31, p < 0.001), indicating a 20% increased likelihood of survival for intervention participants compared to controls. However, the observed effects differed (Q = 9.3, p = 0.002) between the 31 psychosocial behavioral support interventions having an explicit focus on improving coping/motivation to engage in health behaviors (OR = 1.35, 95% CI = 1.20 to 1.52, p < 0.001) and the 56 interventions emphasizing social/emotional support (OR = 1.01, 95% CI = 0.87 to 1.16, p = 0.94). The effect sizes for both kinds of interventions varied substantially, with broad confidence intervals (S1 and S2 Figs). However, in separate analyses specific to effect size heterogeneity, the percentage of variance explained by between-study heterogeneity was estimated to be zero for both the 31 psychosocial interventions based on behavior support (I = 0.0; Q(30) = 27.9, p = 0.57) and the 56 focused on social/emotional support (I = 0.0; Q(55) = 47.0, p = 0.77). Given the absence of between-studies heterogeneity, no further analyses were conducted with OR data.

The 22 RCTs reporting data in terms of survival time averaged HR = 1.29 (95% CI = 1.12 to 1.49, p < 0.001), indicating a 29% increased likelihood of longer survival compared to controls (S3 Fig). As only 4 of the 22 studies focused on supporting health behaviors, we did not analyze subgroup differences. Since the HR data were characterized by a moderate percentage of between-study heterogeneity (I = 54.0; Q(21) = 45.7, p = 0.001), we conducted meta-regressions to evaluate possible moderation by study, intervention, and patient attributes.

Meta-regressions of study, intervention, and patient characteristics

Due to the limited number of studies (k = 22), we evaluated study, intervention, and participant characteristics in 3 separate meta-regression models of HR data. The first model, which evaluated study characteristics (Table 2), explained 37.5% of the variance in effect sizes and reached statistical significance (p = 0.014). The model included 2 significant predictors: control group type (β = −0.42, p = 0.048) and estimated risk of research bias (β = 0.470, p = 0.018). The 8 studies in which control group members received health information/classes in addition to treatment as usual (TAU) averaged HR = 1.14 (95% CI = 0.92 to 1.40, p = 0.23), but the 14 studies with only TAU controls averaged HR = 1.38 (95% CI = 1.17 to 1.62, p < 0.001). Studies with relatively higher risk of research bias tended to report improved patient survival as a result of the intervention; given that finding, we included risk of bias in subsequent meta-regression models.

Table 2

Random effects meta-regression of HR estimates of study characteristics on patient survival.

Variable	R2	B	SE	p	β
Study characteristics	0.375			0.014
Intervention in months		−0.004	0.010	0.69	−0.083
Follow-up in months		−0.001	0.001	0.48	−0.157
Psychosocial improvement achieved1		−0.087	0.084	0.30	−0.234
Control group receiving health information2		−0.270	0.136	0.048	−0.421
Risk of bias3		−0.059	0.035	0.02	−0.470

1Statistically significant improvement on psychosocial measures at the end of the intervention compared to controls.

2Comparison of studies with control groups receiving only TAU with control groups that received TAU plus information/classes relevant to their health condition.

3Sum of indicators of risk of bias.

β, standardized beta; B, unstandardized beta; HR, hazard ratio; SE, standard error; TAU, treatment as usual. k = 21.

The second meta-regression predicted HR data based on the type of intervention (Table 3). The model explained 10.3% of the variance in effect sizes and did not reach statistical significance (p = 0.69). Different kinds of interventions tended to yield similar likelihood of patient survival.

Table 3

Random effects meta-regression of HR estimates of intervention type on patient survival.

Variable	R2	B	SE	p	β
Type of intervention	0.103			0.69
Family support1		−0.003	0.064	0.97	−0.008
Group meetings only		−0.085	0.151	0.57	−0.126
Home visit support only		−0.271	0.214	0.22	−0.270
Telephone/online support only		−0.157	0.226	0.49	−0.145
Risk of bias2		−0.004	0.037	0.91	−0.022

1Degree of inclusion of family/partner in the intervention.

2Sum of indicators of risk of bias.

β, standardized beta; Β, unstandardized beta; HR, hazard ratio; SE, standard error. k = 22.

The third meta-regression predicted HR data from patient characteristics (Table 4). The model explained 41.0% of the variance in effect sizes (p = 0.025). One variable in the model reached statistical significance: Interventions with patients having more advanced disease severity (marked by percentage of patients dying per month) tended to yield lower effect sizes (β = −0.61, p = 0.007). That is, patients with greater disease severity tended to experience reduced benefits from a psychosocial intervention compared to participants in studies with relatively lower disease severity. To put this finding into perspective, 9 studies in which ≥0.5% of patients died per month averaged HR = 1.13 (95% CI = 0.95 to 1.34, p = 0.16), but 11 studies with lower rates of patient mortality averaged HR = 1.64 (95% CI = 1.37 to 1.97, p < 0.001). Risk of bias did not reach statistical significance; we conducted collinearity diagnostics and disconfirmed multicollinearity for all 3 meta-regressions.

Table 4

Random effects meta-regression of HR estimates of patient characteristics on patient survival.

Variable	R2	B	SE	p	β
Patient characteristics	0.410			0.025
Average patient age at recruitment		−0.003	0.008	0.71	−0.087
Percentage of female patients		0.003	0.003	0.33	0.262
CVD patients		0.339	0.284	0.23	0.386
Cancer patients		−0.136	0.181	0.45	−0.204
Patient mortality % per month1		−0.272	0.100	0.007	−0.606
Risk of bias2		0.075	0.072	0.30	0.300

1Number of patient deaths divided by total number of patients divided by total study months.

2Sum of indicators of risk of bias.

β, standardized beta; Β, unstandardized beta; CVD, cardiovascular disease; HR, hazard ratio; SE, standard error. k = 19.

Evaluation of risk of bias

Fig 2 summarizes sources of potential bias across all 106 studies (individual studies reported in S4 Fig). In intervention studies of psychosocial support, both personnel and participants know the conditions of the group to which they are assigned. However, it is difficult to limit personnel and/or participant awareness about the other arm of the study in order to diminish unbalanced expectations for improvement. Such blinding of personnel or participants occurred in very few of the 106 studies evaluated (7% blinding participants, 3% blinding personnel, and 2% blinding both). Thus, the results observed in this meta-analysis do not control for plausible expectation differences between treatment and control groups.

Fig 2

Risk of bias graph of characteristics across 106 studies.

Blinding of outcome assessment was unclear in 44% of studies. Although reports of patient death are reasonably reliable, optimally, researchers would confirm patient mortality through independent records. When independent confirmation does not occur, a plausible threat to study validity is that some participants who researchers are “unable to contact” have died. The impact of missing survival data depends on whether participant attrition remains low and balanced across groups. In this meta-analysis, medical patient attrition across all studies averaged 9.9%, with an average difference of 0.6% between the intervention and control groups, so the risk of bias due to attrition was generally low.

Most studies in this meta-analysis explicitly reported the randomization strategy (64%) and allocation concealment (61%). Participants in the intervention and control groups were typically balanced across variables measured at baseline (78%). The vast majority of studies reported intent-to-treat data (85%) as well as endpoint data on all measures administered (93%).

Estimate of publication bias

We evaluated the degree to which publication bias may have impacted the overall findings. Begg test, Egger test, and Peters test did not reach statistical significance for either HR or OR data. Inspection of the funnel plots (S5 and S6 Figs) did not suggest more than a few missing studies. Trim and fill analyses [148] of the HR data indicated only 1 missing study using the L0 estimator but 4 missing studies using the R0 estimator. When 4 studies were imputed in the distribution, the results remained statistically significant (HR = 1.22, 95% CI = 1.05 to 1.41, p = 0.009). Trim and fill analysis of the OR data identified no missing studies using the R0 estimator but 8 missing studies using the L0 estimator. When 8 studies were imputed in the distribution, the results of the OR data remained statistically significant (OR = 1.15, 95% CI = 1.03 to 1.29, p = 0.015). Overall, the results of this meta-analysis did not appear to be adversely impacted by publication bias.

Discussion

Statement of principal findings

This meta-analysis, including 106 RCTs and 40,280 participants, examined the extent to which different types of psychosocial support interventions increased survival among medical patients receiving curative or rehabilitative care. Overall, the interventions increased odds of survival (OR = 1.20) and relative length of survival (HR = 1.29), with the magnitude of these data being comparable with other tertiary prevention interventions (Fig 3).

Fig 3

Comparison of odds (lnOR) and hazards (lnHR) of mortality across several tertiary prevention interventions.

Note: lnOR = natural logarithm of the OR of patient survival. lnHR = natural logarithm of the HR of patient survival. Effect size of 0 indicates no effect, and values above 1 favor the intervention group relative to the control group. Comparison effect sizes and 95% confidence intervals were reported in meta-analyses: A = McQueen et al. [149]; B = Wu et al. [150]; C = Taylor et al. [151]; D = Ma et al. [152]; E = Kritchevsky et al. [153]; F = Mons et al. [154]; G = Taylor et al. [155]; H = Calman et al. [156]; I = Hauner et al. [157]. CVD, cardiovascular disease; HR, hazard ratio; OR, odds ratio.

Across the 87 studies reporting survival data at a fixed point in time, the 31 psychosocial behavioral support interventions (e.g., motivation for treatment adherence) improved the likelihood of patient survival, but the 56 interventions emphasizing social/emotional support yielded results no better than those of control groups. It is unclear whether behaviorally focused interventions are more effective or whether these types of interventions merely involve more components (behavioral and social/emotional), thereby providing greater diversity of support. As only 4 of the studies reporting HR data were explicitly focused on promoting patient health behaviors, a similar subgroup comparison was not advisable until additional studies reporting HR data accrue in the literature. The HR data predominantly represented interventions focused on social/emotional outcomes (18 of 22 studies).

Other differences between the studies reporting OR and HR data can inform data interpretation. A primary difference involves the nature of ORs and HRs. ORs provide a snapshot at a fixed point in time, but HRs reflect changes across time. Moreover, Cox proportional hazards regression models typically include covariates, such that other variables (e.g., initial health status and socioeconomic status) are less likely to influence the reported outcomes. In terms of our data specifically, the 22 studies reporting HR data tended to have more female participants, twice as many sessions, and 5 more years of patient follow-up, with correspondingly lower rates of patient survival than the 87 studies reporting OR data. Future research is needed to confirm whether interventions with more sessions and longer follow-up yield greater benefits, as recommended in a National Academy of Science report [4].

Analyses with HR data indicated that patient disease severity (percentage of deaths per month) moderated the overall findings. Specifically, studies in which a relatively larger percentage of patients died each month tended to report fewer benefits from the psychosocial intervention in terms of patient survival compared to control conditions. Future research can investigate if the higher mortality rates are a function of more reliable outcomes when death is not uncommon in the distribution. Alternatively, psychosocial interventions might be more effective in improving survival among patients when conducted earlier in the disease trajectory, consistent with effectiveness of other medical treatment.

Meta-regression analyses with HR data indicated that effect sizes did not differ across the format of the intervention (support groups, telephone/online conversations, family involvement, or home visits). However, in one of the meta-regressions, the findings differed as a function of study risk of bias, with studies reporting more robust results also tending to have more indicators of research bias. Having disconfirmed multicollinearity, we cannot account for why that variable reached statistical significance in only one of 3 analyses, but the result provides a caution that qualifies the overall findings reported in the literature. The overall strength of evidence was mixed (Fig 2), with the primary limitation being the neglect of blinding personnel and participants to study conditions. Thus, it is difficult to distinguish between intervention effects and expectation effects when personnel and participants have knowledge of both study conditions. This concern was reinforced by the finding that 8 psychosocial support interventions reporting HR data did not show statistically significant differences from control groups receiving health information/classes.

Limitations of the study

This meta-analysis has several limitations. First, the results varied widely across individual studies (see S1–S3 Figs). The omnibus results should be interpreted using their confidence intervals. Across the 87 studies reporting OR data, the confidence intervals for individual studies were so wide that there were no nonoverlapping values (I = 0.0). The wide confidence intervals for most of these OR studies corresponded with a numerically low percentage of patients who died across studies (8.1%, see Table 1); low mortality rates yielded high standard error values. Second, variability existed in the approach and delivery of support provided in the studies. Psychosocial support was offered via peer support groups, telephone calls, one-on-one nurse sessions, etc., with our statistical contrast being the mixed interventions. Third, only 10 of the 106 RCTs included support from naturally occurring relationships in at least half of the intervention, with 6 of those focusing specifically on family/partners, yet preexisting relationships constituted the epidemiological evidence that precipitated such interventions [5]. Strengthening preexisting close relationships may produce longer-lasting effects due to the chronic and often intimate nature of such relationships [158]; nonetheless, not all patients have supportive social networks. Fourth, we did not evaluate preexisting levels of patient psychosocial support because the literature inconsistently reported such data. Patients with strong social networks tend to fare better than others on multiple clinical markers [20,159] and outcomes [159-161]. Failure to account for preexisting differences in social resources can be corrected in future research [162]. Fifth, although many of the studies reporting HR data included other variables in statistical models, such as patient age and health status, only 3 of the studies reporting OR data statistically controlled for other variables. The HR estimates were therefore more trustworthy than the OR estimates [43].

Implications for clinicians, researchers, and policy makers

Prior meta-analyses have reported mixed results [25,27,31,36], some concluding that psychosocial interventions did not improve patient survival [23,33,37]. Therefore, a major contribution of this meta-analysis was to clarify that although the vast majority of studies did not reach statistical significance (96 of 106 [91%]; see S5 and S6 Figs), psychosocial interventions overall tended to benefit survival with results comparable to rehabilitation programs (Fig 3). However, the extent of variability in results across studies suggests that care must be taken during design and implementation to maximize patient outcomes. In particular, this meta-analysis confirmed that the minority of interventions (32%) that explicitly promoted patient motivation/coping to engage in health behaviors tended to improve patient survival, with an observed effect (OR = 1.35) corresponding with a number needed to treat of 19.6. Rather than focus solely on emotional and psychological support, future psychosocial support interventions with medical patients should also address health behaviors (e.g., motivation for treatment adherence). The accumulated data now make it questionable to neglect including behavioral support when planning psychosocial interventions with medical patients receiving curative care.

Given the concerns raised in this meta-analysis about study risk of bias adversely impacting the reported results, future research should specifically address that issue. Although blinding personnel and participants to the other study arm may be challenging, this gap needs to be addressed to advance the science beyond its current state. Other scholars have recommended that future research identify patient existing psychosocial supports and needs, evaluate specific causal pathways influencing disease progression [9,10,163], focus on strengthening naturally occurring relationships [158], and refine interventions utilizing the Multiphase Optimization Strategy [164,165].

Despite the multiple qualifications and concerns raised in this meta-analysis, psychosocial support interventions improved medical patient survival to a degree comparable with other tertiary prevention methods (Fig 3), with the findings being equivalent to a meta-analysis of epidemiological data on the effects of social isolation on mortality [6]. Taken together with prior research documenting that social isolation increases healthcare costs [166] and excessive utilization [16,167], and with increasing social isolation in recent years [168], this meta-analysis urges increased methodological rigor but tentatively supports recommendations [4] to consider psychosocial interventions in promoting health behavior in a public health framework.

PRISMA 2009 checklist.

PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

(PDF)

Click here for additional data file.

Spanish translation of the abstract by Laura Melgarejo Perez and Juan Valladares.

(PDF)

Click here for additional data file.

Traditional Chinese Characters translation of the abstract by Cheng Wai Man.

(PDF)

Click here for additional data file.

Simplified Chinese Characters translation of the abstract by Li Zhen.

(PDF)

Click here for additional data file.

Tamil translation of the abstract by Babu Manuel Abel.

(PDF)

Click here for additional data file.

Italian translation of the abstract by Claudia Mencarelli and Tommaso Cardullo.

(PDF)

Click here for additional data file.

Turkish translation of the abstract by Murat Çakır.

(PDF)

Click here for additional data file.

German translation of the abstract by Samira Herber.

(PDF)

Click here for additional data file.

Indonesian translation of the abstract by Pungki Lupiyaningdyah.

(PDF)

Click here for additional data file.

Arabic translation of the abstract by Sara Abu Al-Samen.

(PDF)

Click here for additional data file.

Portuguese translation of the abstract by Solange Andrezzo and Larissa Vecchi.

(PDF)

Click here for additional data file.

Literature search strategies and selection criteria.

(PDF)

Click here for additional data file.

Data abstraction and analyses.

(PDF)

Click here for additional data file.

PICOT inclusion criteria.

(PDF)

Click here for additional data file.

Characteristics of 87 psychosocial intervention studies reporting ORs of medical patient survival.

OR, odds ratio.

(PDF)

Click here for additional data file.

Characteristics of 22 psychosocial intervention studies reporting HRs of medical patient survival.

HR, hazard ratio.

(PDF)

Click here for additional data file.

Forest plot of 56 social/emotional support focused RCTs reporting OR.

OR, odds ratio; RCT, randomized controlled trial.

(PDF)

Click here for additional data file.

Forest plot of 31 behavioral support RCTs reporting ORs.

OR, odds ratio; RCT, randomized controlled trial.

(PDF)

Click here for additional data file.

Forest plot of 22 RCTs reporting HRs.

HR, hazard ratio; RCT, randomized controlled trial.

(PDF)

Click here for additional data file.

Risk of bias summary.

(PDF)

Click here for additional data file.

Contour-enhanced funnel plot of 89 RCTs, OR data.

OR, odds ratio; RCT, randomized controlled trial.

(PDF)

Click here for additional data file.

Contour-enhanced funnel plot of 22 RCTs, HR data.

HR, hazard ratio; RCT, randomized controlled trial.

(PDF)

Click here for additional data file.

20 Jul 2020

Dear Dr Smith,

Thank you for submitting your manuscript entitled "Psychosocial Support Interventions and Medical Patient Survival: A Meta-Analysis of 140 Randomised Controlled Trials" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by .

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Caitlin Moyer, Ph.D.,

Associate Editor

PLOS Medicine

1 Oct 2020

Dear Dr. Smith,

Thank you very much for submitting your manuscript "Psychosocial Support Interventions and Medical Patient Survival: A Meta-Analysis of 140 Randomised Controlled Trials" (PMEDICINE-D-20-03297R1) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to three independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Oct 22 2020 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

1.Prospective analysis plan: Did your study have a prospective protocol or analysis plan? If so, please provide the analysis plan as a supporting information file. Please state whether analyses were pre-planned (either way) early in the Methods section.

2.Data availability statement: “All data files will be available from the BYU Scholar's Archive database at the time of publication (URL to be determined).” Thank you for agreeing to make your data available. If the data are freely or publicly available, note this and state the location of the data: within the paper, in Supporting Information files, or in a public repository (include the DOI or accession number). If the data are owned by a third party but freely available upon request, please note this and state the owner of the data set and contact information for data requests (web or email address). Note that a study author cannot be the contact person for the data.

3.Abstract: Background: The final sentence should clearly state the study question.

4.Abstract: Methods and Findings: Please briefly summarize the patient demographics of those included studies.

5.Abstract: Methods and Findings: Please provide p values for these results, also please provide more description of the results (for example, what are the OR/HR representing here: “Across 112 RCTs reporting data for discrete time periods, the average was OR = 1.15 (95% CI = 1.03 to 1.29; I2 = 23.6, I2 CI 4 to 40). Across 29 RCTs reporting survival time, the average was HR = 1.32 (95% CI = 1.15 to 1.51; I2 = 53.6, I2 CI 29 to70).”)

6.Abstract: Methods and Findings: In the last sentence of the Abstract Methods and Findings section, please describe the main limitation(s) of the study's methodology.

7.Abstract: Conclusions: Please address the study implications without overreaching what can be concluded from the data; the phrase "In this study, we observed ..." may be useful. It is not clear what is meant by “interventions minimally improved patient survival…” in the first sentence- please clarify this to either state that the interventions did improve, or did not improve survival.

Please interpret the study based on the results presented in the abstract, emphasizing what is new without overstating your conclusions. “Psychosocial support programmes can be improved by better meeting patient needs.” this sentence does not seem to be supported by the study’s data or objectives.

9.Author Summary: At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

10.References throughout: For in-text citations, please place the reference number in brackets before the punctuation, like this [1].

11.Introduction: Line 77-78: “Thus, in conducting the largest meta analytic review to date…” Please qualify this with “to the best of our knowledge” or similar.

12.Methods: Line 88: Please revise this to “This study is reported as per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline.” Please provide the completed PRISMA checklist as a supporting information file. When completing the checklist, please use section and paragraph numbers, rather than page numbers.

13.Methods Line 91-92: Please update your search to the present time.

14.Results: Lines 146-148: Please provide the p values associated with “Across 112 estimates using fixed time periods, the average was OR = 1.15 (95% CI = 1.03 to 1.29), indicating a small increased likelihood of survival for intervention participants compared to controls.

15.Results: Lines 156-157: Please provide the OR/HR data with 95% CIs and p values for the findings excluding terminally ill patients or deaths shortly after baseline, between inpatinets and outpatients, by health condition, and attrition vs. mortality, or treatment as usual with our without additional materials. Please refer to the table where these are presented.

16.Results: Lines 167-168: Please also present the HR and 95% CI breakdown for this result, that intervention length was not associated with survival time.

17.Results: LInes 171-174: Please provide the p values for the sensitivity analyses of the high quality studies.

18.Results: Liines 177-181: Please provide p values for psychosocial functioning results.

19. Discussion: Line 316: The reference to the “current pandemic” should be removed.

20.Figure 4: Please define the abbreviations for OR, HR, CVD in the figure legend.

21.Competing interests: Can you please clarify the existing competing interests of the authors? The sentence “no financial relationships with any organisations that might have an interest in the submitted work in the previous three years” is not clear.

22.Supporting information: Please provide titles and legends for each individual table and figure in the Supporting Information.

Comments from the Academic Editor:

1. The Introduction seems to conflate several different constructs (eg., social isolation, loneliness, and social support). And those are different from other constructs in the literature. eg.,

Social integration:

Seeman TE. Ann Epidemiol 1996;6:442-451.

Tsai AC, Lucas M, Sania A, Kim D, Kawachi I. Ann Intern Med. 2014 Jul 15;161(2):85-95.

Tsai AC, Lucas M, Kawachi I. JAMA Psychiatry. 2015 Oct;72(10):987-993.

Social participation:

Obembe AO, Eng JJ. Neurorehab Neural Repair 2016;30:384-392.

Kuiper JS, et al. Ageing Res Rev 2015;22:39-57.

As a result, it is unclear to me exactly what kinds of interventions are the focus of this study. The title and Methods refer to "psychosocial support" interventions, but when I look at the list of included studies, I see several CBT interventions (eg., Choi et al ref #63)-- but I would not characterize CBT as a "social support intervention". When I get to the Limitations and find out (line 280) that only 8/140 RCTs focused on naturally-occurring relationships, then I am even less certain as to what kinds of interventions were studied. ENRICHD, for example, was focused on getting study participants to strengthen their existing relationships, and is generally regarded as a prototypical "social support intervention". Does this mean that 132/140 RCTs were more of the "prescribe a friend" type? Lines 134-139 indicate that most interventions were more of the "clinician-delivered support" type.

2. The search needs to be updated. My understanding is that PLOS generally requests for evidence searches to be out of date by no more than 6 months. Should this manuscript make it through peer review, by the time it is published, more than 3 years will have elapsed since the evidence search was conducted.

3. Line 98: The authors restricted analysis to studies of "medical patients". Perhaps they could be more specific about what this term means. Are they referring to inpatients? Outpatients? Or did the included studies need to include study participants with a defined medical diagnosis and/or acute medical illness? (Line 104 ["mental illness as the sole health condition"] seems to imply that included studies were focused on some health conditions.) Any study conducted with a participant sample recruited in a health care setting? If so, does this mean they excluded group interventions occurring outside of health care settings (eg., Alcoholics Anonymous)? I see "self-help group" in the search strategy but not "AA", "Alcoholics Anonymous", or "12 step" (and those terms would normally be included in a systematic review of AA-style interventions, cf. Kelly JF, Magill M, Stout RL. Addict Res Theory 2009;3(17):236-259). They seem to have included "home visits", so I would expect that AA-style interventions were also included--but the search strategy does not seem like it was designed to capture these.

4. Lines 101-104: There are a few things that make the exclusions a bit "squishy". (a) The authors state that they are focused on "disease" but then exclude suicide, mental illness, dementia-- is the implication that mental illness is "not disease"? (b) If the authors excluded studies of mental illness as the "sole health condition", does that mean they included studies that were focused on depression among study participants with another medical condition? I see Berkman/ENRICHD in the list of included studies, so I assume this is true. Please clarify. (c) I am in agreement with R2 that the exclusion of palliative care interventions "not intended to prolong life" seems problematic and should be clarified or amended.

5. Line 111: Does this mean that none of the studies reported data as (for example) standardized effect sizes, regression coefficients, etc.? I find this very, very hard to believe. In any case, if this is true, please indicate here in the text.

6. Line 115: What does "questionable effect size data" mean? Please be more specific here in the text. Examples (eg., in the Appendix) may help.

7. Line 116 and elsewhere: The authors use the term "multivariate" but I believe they mean "multivariable". (cf. Peters TJ. Paediatr Perinat Epidemiol 2008;22(6):506).

8. Line 116: The authors included randomized controlled trials. Why would these studies need to "control for known confounds"? Are the authors referring to randomized studies in which covariate balance was not achieved through randomization? Please clarify here in the text.

9. Line 129 seems to suggest that studies were restricted to either medical inpatients or medical outpatients. (cf comment #3 above)

10. Abstract, Line 147, Line 236, Figure 4, etc.: why are the findings (pooled OR=1.15, pooled HR=1.32) described as "minimally" improving patient survival (eg., "small increased likelihood of survival", "increased odds to a small degree")? I am _not_ suggesting that the authors revise this adjective (eg., "substantially"). I am simply requesting that the authors justify their use of this adjective. I see the comparisons to exercise, cardiac rehab, etc. But I do not find those examples too compelling. Aspirin reduces the risk of CVD events by 15% in relative terms (ie, meta-analysis OR=0.85), but I do not believe that clinicians would characterize that as a minimal improvement. Perhaps in the literature on social support interventions the bar is different?

11. Line 155: All of these subgroup analyses, and the meta-regression, should be described in the Methods section.

12. Line 158 has a typo ("statistically significant differ")

13. Line 169: I see the N's for the low-quality studies. What were the N's for the high quality studies, and what were the pooled estimates?

I did not see a Risk of Bias assessment table. This would be particularly important to know which studies (OR vs HR data) were in the "high risk of bias" vs "low risk of bias" groups, because the higher quality studies did not reach statistical significance.

How was risk of bias assessed? This is only mentioned in passing on line 109. eg Did a study get 1 point each for blinding, exclusions, attrition (what threshold?), and then "low risk of bias" = 3 points?

14. Line 175 and Figures 2 & 3: what do the authors mean by "statistically significant improvements on psychosocial variables"? Presumably variables like depression symptom severity, social role functioning, etc. were included. What kinds of variables were excluded? What do the authors mean by "inconsistent/ineffective" in Figure 2? (What if a study was "consistent" but "ineffective", or "inconsistent" but "effective"?) I see that these data are partially represented graphically in Figures 2 & 3, but it would be helpful to have an Appendix table that is analogous to Table 2 showing each study line by line (stratified by "inconsistent/ineffective" vs. "effective"), the "psychosocial variable" assessed (eg., 15-item Hopkins Symptom Checklist), and the efficacy estimate and standard error on the psychosocial variable.

15. Line 184: The average age and percentage of female patients are not study-level characteristics and should not be included in the meta-regression model.

16. Line 284: "many studies did not measure preexisting levels of support"-- I did not see this anywhere in the Results, please add.

Comments from the reviewers:

Reviewer #1: I confine my remarks to statistical aspects of this paper. The general approach is fine, but I have some concerns and suggestions before I can recommend publication.

First - As I was reading along, I thought "why aren't they doing meta-regression?" Unless I missed something, this was first mentioned on p 8. It ought to be mentioned in methods on p. 4 or 5. (The authors say 'subsequent analysis' but that is very vague).

More specific comments

Lines 50 and 52: It isn't clear what these OR and HR are, exactly. It seems like it is "support of any kind" vs. "no support" but this should be made clear.

Line 101 What does the sentence starting "Differences were coded ..." mean? What differences? Coded how?

Line 149 ff I don't think statistical significance of Isquared is particularly useful. Maybe delete the p values and CIs.

Subgroup analysis - I'm unclear on how subgroup analysis was used to conduct hypothesis tests and yield p values and so on. This is what meta-regression doe. But if you do an analysis only on one subgroup, how do you do a tst of it vs. another subgroup? This doesn't make the subgroup analysis wrong, but .... how did the authors get these results?

Fig. 2 is way too small a font. Maybe spread it over several pages.

Fig 3 I like that there is relatively little text here. But it would be good to add a vertical line at OR = 1 That would make it a lot clearer that these effects are in one direction.

Peter Flom

Reviewer #2: Review PMEDICINE-D-20-03297R1

Psychosocial Support Interventions and Medical Patient Survival: A Meta-Analysis of 140 Randomised Controlled Trials

I welcome the goal to perform a comprehensive meta-analysis to establish the state of knowledge on this topic. The heterogeneity of interventions and medical conditions included however is linked to significant problems in the present work which raise the risk of misinterpretation due to oversimplification. My points include 1) oversimplified presentation of the theoretical background, 2) need for refinements in some of the central extracted data from studies and 3) need to discuss conclusions in a much more modest way.

1) In the introduction the authors aim to establish a link from patient loneliness and lack of social connections to the eventual meta-analysis of a wide range of psychosocial support interventions, including for example several-session telephone disease management programs. The authors need to better explain how and why such interventions would be able to reduce loneliness/increase social connectedness as the assumed mechanism of prolonging survival. It is hard to imagine how low-dosis interventions of this structure that have been included in the review could change an individuals' social relationships so enduringly that survival is significantly prolonged. Yet it seems reasonable to assume a different mechanism: in-person disease management or telephone support with home visits can improve health behavior (therapy adherence, smoking, diet, alcohol consumption etc.), which in turn improves survival.

2) Inclusion criteria: The authors state that interventions "not intended to prolong life" were excluded. This is a difficult statement because, at least for the majority of the included cancer studies, the study was not conducted with a primary intention of prolonging survival but to enhance quality of life. However, survival was co-assessed as one of many potential outcomes. This needs to be clearly stated. It is difficult to understand why similar studies in palliative care populations were excluded. For example, several early-palliative-care intervention trials reported survival data along a range of outcomes and they are not represented in this review.

Page 6, line 139: The average length of intervention sessions seems very long. Do the authors mean the total length of all sessions?

Abstract/Results: It is difficult to understand, what exactly the OR of 1.15 means in terms of survival. I suggest adding a sentence explaining this in simple words. The addition of separate results for hazard ratio-studies increases the confusion. Please make both numbers more accessible by an explaining sentence.

Table 1/2:

Please report descriptive data about time of survival or the time period studied for intervention and control condition for OR and HR studies.

The tables should also entail brief information about the content and type of the intervention in addition to the format.

The tables would also be easier to read if the confidence interval were included rather than or in addition to the standard error.

More information is needed on how the authors decided on whether a study was "effective" or "ineffective" regarding psychosocial outcomes. Most studies may have been ineffective with regard to some and effective with regard to other outcomes.

Figures: Does effectiveness refer to psychosocial outcomes? Please include a note in the figure caption.

What size was the association between effectiveness in terms of psychosocial improvements and risk of bias? Did the authors control for multicollinearity of the two variables entered simultaneously to the regression models?

3) A highly relevant result is that if high-quality studies were considered alone, no effect on survival was found. This is an important finding suggesting cautious interpretation.

The conclusion to better meet patients' needs does not follow from the results.

Reviewer #3: Review, Psychosocial support Interventions and medical survival

This is a very comprehensive, large, and rigorous meta analysis of psychosocial support interventions and survival for medical patients (mostly CVD and cancer). A total of over 40,000 studies were screened, and 140 studies were included in the meta-analysis. The approach adopted for study selection and coding as well as the careful analytic approach taken are key strengths of this meta analysis. The expansion to multiple diseases is novel, and the inclusion of intervention attributes is also novel and important. The conclusions are meaningful and likely to have an impact on clinical care.

I have a few comments, which are mostly minor.

1) In the first sentence, the consensus report calling for interventions for loneliness and social isolation in medical and health care settings is interesting, but the construct of social isolation and loneliness among persons dealing with chronic illnesses. While a small distinction, if the statement is true, then it might be helpful to change the reference to something more relevant.

2) Some interventions contain content that focuses on enhancing support within the person's network using CBT (overcoming barriers to support). Other interventions are delivered by aj interventionist but would contain solely education and skill practice, but like most 1:1 treatments, a bond develops. Some group treatments are not support oriented, in that the group is more of an education delivery method, but the members obtain support after group or during group exercises in the room. Solely online interventions can contain supportive elements, via email exchanges/message boards, or even videotaped content offering normalization and narratives from other patients. Disease management support is categorized as support, but these are more educational interventions, unless I am missing something. Couples treatments may or may not include support, but may reduce conflict. It would have been very helpful to define social, emotional, and psychological support more clearly and discuss why disease management support is included. Disease management is assumedly to foster adherence to medical treatment. The supplemental materials define interventions but the above approaches would not be clearly included or excluded as the definition is very general.-

3) The conclusion that the interventions that resulted in psychological improvement achieved better outcomes in terms of survival than control interventions seems a little generic. Since the interventions may have provided support as well as coping skills (at least if my understanding of the interventions meeting criteria are correct), then the conclusion might state something more along the lines of "the intervention may not have included effective components." This leads me to the broader concern that the psychotherapeutic goals of each intervention might be better characterized in this study, as efficacy may be due to the components, and that would guide this study's implications much more clearly.

4) It might have been interesting to discuss the baseline level of loneliness/low support as a moderator of effects for future studies that focus on survival. It may be that the interventions have a stronger impact among lonelier patients, and that may be why older patients fared better.

Any attachments provided with reviews can be seen via the following link:

[LINK]

17 Jan 2021

Submitted filename: Responses to Editor PLOS Medicine revisions Jan 2021.docx

Click here for additional data file.

8 Mar 2021

Dear Dr. Smith,

Thank you very much for re-submitting your manuscript "Psychosocial Support Interventions and Medical Patient Survival: A Meta-Analysis of 106 Randomised Controlled Trials" (PMEDICINE-D-20-03297R2) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by one of the original reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

Please note, when your manuscript is accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you've already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosmedicine@plos.org.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Mar 15 2021 11:59PM.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

1. Please revise the title to indicate that the meta-analysis was restricted to studies where recruitment occurred in health care settings, e.g. "Psychosocial support interventions delivered in inpatient and outpatient health care settings: a meta-analysis of 106 randomized controlled trials"

Please ensure that the Abstract similarly emphasizes this distinction.

2.Data availability statement: Thank you for making the study data available. Please provide a complete link to the dataset here (for example: https://osf.io/3qydb/)

3.Abstract: Conclusions: Line 57: The link to the dataset registration can be removed from the abstract.

4.Throughout: Please remove spaces from within brackets of in-text citations ([1,2] instead of [1, 2] for example)

5.Introduction: Line 140: Please revise to “sought to evaluate”

6.Methods: Line 152: Please include a reference to the supporting information file containing the PRISMA checklist (S1_Checklist)

Methods: Line 170/Table 1: Please include a very brief rationale for excluding palliative care interventions, in the most appropriate point in the text (perhaps drawing on the most relevant points from the updated cover letter).

7.Methods: Data analysis section: The use of the first person here is fine; however, if possible we suggest considering a reduction in the use of “We…” as this seems to start the beginning of most sentences.

8.Table 1: Please consider if this information needs to be displayed in a table format, or can be presented as subsections within the text of the Methods section.

9.Figure 1: Please provide a more descriptive title/legend for the flow diagram.

10. Figure 2: Please provide a more descriptive title/legend for the diagram illustrating risk of bias.

11. Results: Description of included studies: Please provide the numerators/denominators (or at least the number, as the total number of studies, 106, is given earlier in the paragraph) for any percentages reported where the numbers aren’t displayed in a table.

12. Results: Main analyses: Please report p<0.001 (rather than p=0.0001 at line 248, for example) unless there is a reason to report them to this number of digits.

13. Results: Line 344-345 and Line 348-349: Please provide the p value in addition to the 95% CIs for the publication bias analysis for the 4 and 8 imputed studies from the HR and OR data.

14. Page 24: Please remove the Competing Interests and Funding sections from the main text, and ensure that all information is accurately entered in the “Competing Interests” and “Financial Disclosures” section of the manuscript submission system.

15. Reference list: Please use the "Vancouver" style for reference formatting, and see our website for other reference guidelines https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references

16. S1 Appendix: It would be helpful to provide the search criteria, analysis plan, and PRISMA checklist as separate files.

17. S2 Appendix: Please provide legends in addition to titles for all figures and tables, including those in Supporting Information files. Again, it would be helpful for each table/figure to be a separate file, to make it easier to refer the reader to the correct figure within the text.

Comments from Reviewers:

Reviewer #1: The authors have addressed my concerns and I now recommend publication

Peter Flom

Any attachments provided with reviews can be seen via the following link:

[LINK]

17 Mar 2021

Submitted filename: Response to requested revisions PLOS Medicine March 2021.pdf

Click here for additional data file.

25 Mar 2021

Dear Dr Smith,

On behalf of my colleagues and the Academic Editor, Alexander C. Tsai, I am pleased to inform you that we have agreed to publish your manuscript "Effects of Psychosocial Support Interventions on Survival in Inpatient and Outpatient Health Care Settings: A Meta-Analysis of 106 Randomised Controlled Trials" (PMEDICINE-D-20-03297R3) in PLOS Medicine.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine