The Association Between Resilience and Mental Health in the Somatically Ill

Abstract

Background

Resilience refers to an individual’s positive adaptation to the experience of adversity. The maintenance of mental health is commonly considered a sign of successful coping with adverse conditions. The goal of the present meta-analysis was to investigate the association between resilience and mental health in patients with a somatic illness or health problem.

Methods

Studies were included if they reported measures of association between resilience, as assessed using a version of Wagnild and Young’s Resilience Scale, and self-reported mental health. A systematic literature search was conducted in the Medline, Web of Science, PsycInfo, PubPsych, and ProQuest databases and in the dissertation catalogue of the German National Library. In addition, a manual search was carried out. The study was registered with PROSPERO (registration number: CRD42017054822).

Results

55 studies involving a total of 15 003 patients were included in the meta-analysis. Assuming a random-effects model, the weighted mean Pearson correlation between resilience and mental health was r = 0.43 (95% confidence interval [0.39; 0.48], p<0.001). This association was robust, although the heterogeneity among individual effect sizes was substantial (I²= 89.6%). Correlations tended to be weaker in unpublished studies than in published ones.

Conclusion

Despite substantial heterogeneity across studies, the findings suggest a strong association between resilience and mental health in the somatically ill. In clinical practice, a lack of resilience as a resource for successful coping might indicate a need for psychosocial support during treatment for somatic illness.

In psychology, resilience is the term used to describe an individual’s positive adaptation in the face of adversity, i.e. one’s success in dealing healthily with significant stressors.

The definitions of psychological resilience distinguish between resilience as a personality trait and resilience as a dynamic process (4). These definitions also represent the two major lines of psychological resilience research: the approach in personality psychology on the one hand, and the approach in developmental psychology on the other hand (5). In developmental psychology, resilience is primarily studied in children and adolescents who showed positive development despite having experienced considerable hardship or trauma (6); examples for this approach include the longitudinal studies of the research groups around Emmy Werner (7) and Ann Masten (8). By contrast, the concept of resilience as a personality trait is typically encountered in the literature on resilience in adults (2). Rooted in a psychoanalytical research tradition, it originates from the concept of ego-resiliency introduced by Block and Block (9) in the 1950ies. Current research on the personality psychology approach (10, 11) uses the term “trait resilience“ (12), describing resilience as a relatively stable personality trait, in contrast to the processual approach of developmental psychology.

A number of instruments is available to measure resilience. Of these, the Resilience Scale by Wagnild and Young (13) is the most commonly used measure (2, 14); it has been translated into numerous languages. When comparing the psychometric properties of instruments designed to measure psychological resilience as a personality trait, the most convincing evidence was found for the Resilience Scale with regard to theoretical foundation, reliability and validity (15, 16).

Wagnild and Young’s Resilience Scale measures the level of resilience as a positive personality characteristic in terms of a personal resource that enhances individual adaptation (13). The items can be attributed to the two underlying factors “personal competence“ and „acceptance of self and life“ (13). The following items are attributed to the first factor:

• Self-reliance

• Independence

• Mastery

• Resourcefulness

• Perseverance.

The second factor is described by the items:

• Adaptability

• Balance

• Flexibility and balanced perspective of life (13, 17).

The Resilience Scale consists of 25 items and a 7-point Likert-type scale (values from 1–7). Today, various short versions are available, including a German version with 13 items (18) (table 1).

Table 1. German version of the Resilience Scale, shortened 13-item version (RS-13) (18).

	1 = no Totally disagree				7 = yes Totally agree
When I make plans, I follow through with them.	1	2	3	4	5	6	7
I usually manage one way or another.	1	2	3	4	5	6	7
I usually take things in stride.	1	2	3	4	5	6	7
I am friends with myself.	1	2	3	4	5	6	7
I feel that I can handle many things at a time.	1	2	3	4	5	6	7
I am determined.	1	2	3	4	5	6	7
I take things one day at a time.	1	2	3	4	5	6	7
Keeping interested in things is important to me.	1	2	3	4	5	6	7
I can usually look at a situation in a number of ways.	1	2	3	4	5	6	7
Sometimes I make myself do things whether I want to or not.	1	2	3	4	5	6	7
When I‘m in a difficult situation, I can usually find my way out of it.	1	2	3	4	5	6	7
I have enough energy to do what I have to do.	1	2	3	4	5	6	7
It’s ok if there are people who don‘t like me.	1	2	3	4	5	6	7

The score is calculated by adding up the points. For the interpretation of the level of resilience, the recommendation is as follows: 13–66 = low; 67–72 = moderate; 73–91 = high. For norms and percentile ranks refer to (18). Re-print of the German version of the Resilience Scale courtesy of Verlag Vandenhoeck und Ruprecht (18). Permission to translate and use the original RS-25 scale was granted by its authors, Wagnild and Young.

In the definitions of both concepts of resilience—as a process and as a personality trait—two aspects are key:

• A preceding experience of adversity

• Subsequent positive adaption (1).

Positive adaption is understood as the maintenance of mental health or relatively rapid recovery after temporary disturbances (19).

While resilience as a resource is discussed in the context of a variety of adverse conditions or stressful situations (2), dealing with physical illness and health problems is the focus of this paper. Just as traumatic experiences and chronic stress, illnesses can precede the development of mental disorders (20). This risk of developing mental health problems is one and a half to two times as high in individuals living with chronic physical disease compared to both healthy individuals and the general population. In addition, subclinical symptoms of mental distress are commonly observed in patients with physical illness (21).

When patients develop symptoms of mental distress, these can have an impact on the course of disease, the compliance of the affected individual, and the success of treatment (21). Thus, understanding the factors which may help patients to successfully cope with physical illness is key. Systematic reviews reported negative associations between resilience and symptoms of mental distress in patients with physical illness (22), cancer (23), or chronic disease (24). A positive association was found between resilience and quality of life.

Meta-analyses of the association between resilience and mental health as an indicator of positive adaption to stress (25– 27) reported significant positive associations. However, these data refer to a wide range of stressful situations. In view of the diversity of adversities for which resilience is discussed as a protective factor (2), it would seem reasonable to differentiate between the respective stressors.

Thus, the aim of this study is to provide a systematic overview of the association between resilience and mental health in the physically ill and integrate the evidence by means of meta-analysis.

Methods

A detailed description of the methodology is provided in the eMethods section. Research questions, inclusion criteria and methods were pre-specified in a review protocol (PROSPERO International prospective register of systematic reviews; registration number: CRD42017054822).

Studies meeting the criteria listed in Table 2 were included in the analyses. A systematic literature search was conducted in the Medline, Web of Science, PsycInfo, PubPsych, and ProQuest databases and in the dissertation catalogue of the German National Library. In addition, a manual search was undertaken. Study selection and data extraction were performed by the two authors (FF, JR); any disagreements were resolved by consensus discussion.

Table 2. Inclusion criteria.

Criterion	Description
Population	Adults who experience acute or chronic illness; or whose physical complaints (health problems) are less serious, but distressing; or who undergo surgery or other impairing treatments
Resilience	Self-reported using a version of the Resilience Scale by Wagnild and Young (13)
Mental health	Self-reported using questionnaires on depression, anxiety or distress (in the sense of negative mental health, i.e. mental distress) or using questionnaires on mental quality of life
Statistical parameters	Cross-sectional measure of association (i.e. correlation coefficient/standardized regression coefficient) or a measure which can be converted into such measure
Publication year	No restrictions
Publication language	English, German, French, Spanish, Italian

Potential risk of bias was assessed based on the reliability of the instruments used, response rates, and completeness of reporting. The Pearson product-moment correlation coefficient was chosen as effect size measure. The weighted mean effect size for the association between resilience and mental health was calculated assuming a random-effects model. The magnitude of effect sizes is determined according to the conventions by Cohen (28). Hence, correlations from 0.10 are classed as a small effect, from 0.30 as a medium effect and above 0.50 as a large effect. In addition, the impact of potential publication bias on the determined mean effect was evaluated and subgroup, meta-regression and sensitivity analyses were performed.

Results

Study selection

Initially, the literature search identified 5592 studies; of these, n = 55 studies with i = 57 reported samples fulfilled the inclusion criteria (figure). These contained data on k = 95 associations between resilience und measures of mental health.

Figure.

Flowchart of study selection

Study characteristics

An overview of the included studies is provided in eTable 1. Altogether, studies from 18 countries with 15 003 patients published between 2006 and 2018 were taken into account. In addition to 49 published studies, 6 unpublished studies were included. 51% of study participants were female; the average age was 57.3 years.

eTable 1. Descriptive characteristics of and associations between resilience and mental health in the included studies.

Study	Sample	Country	RS version	Mental health	N	r	95% CI	p
Amler et al. 2015 (e1)	Survivors of acute myeloid leukemia after allogeneic stem cell transplantation	Germany	RS-25	Anxiety Depression Emotional functioning	41	0.727	[0.539; 0.845]	0.000
Bathke 2011 (e2)	Obese patients (BMI ≥ 30 kg/m²)	Germany	RS-13	Anxiety Depression Mental QoL	164	0.219	[0.065; 0.363]	0.005
Brix et al. 2008 (e3)	Cancer patients with radiation therapy	Germany	RS-25	Mental QoL	239	0.390	[0.277; 0.493]	0.000
Chopp-Hurley et al. 2017 (e4)	Older university employees with ‧osteoarthritis	Canada	RS-25	Depression	24	0.500	[0.121; 0.752]	0.012
Cohen et al. 2014 (e5)	Former colorectal cancer patients (Diagnosed 1–5 years prior to start of study, stages II–III)	Israel	RS-25	Anxiety Depression	92	0.625	[0.482; 0.736]	0.000
Defrancesco et al. 2013 (e6)	Obese women (BMI: M = 33 kg/m²) 1–5 years after bariatric surgery	Austria	RS-25	Depression	64	0.367	[0.150; 0.551]	0.001
Erim et al. 2015 (e7)	Living kidney donor (3 months after transplantation)	Germany	RS-13	Mental QoL	41	0.380	[0.082; 0.616]	0.014
Esteve & Ramírez-Maestre 2013 (1) (e8)	Patients with inflammatory bowel ‧disease (Crohn’s disease, ulcerative colitis)	Spain	RS-25	Depression	128	0.620	[0.500; 0.716]	0.000
Esteve & Ramírez-Maestre 2013 (2) (e8)	Patients with back pain (GP treatment)	Spain	RS-25	Depression	141	0.520	[0.388; 0.631]	0.000
Esteve & Ramírez-Maestre 2013 (3) (e8)	Patients with various pain syndromes (treatment in pain clinic)	Spain	RS-25	Depression	137	0.170	[0.002; 0.328]	0.047
Fitzke 2015 (e9)	Long-term survivors of cancer (5 and 10 years, respectively, after diagnosis)	Germany	RS-11	Anxiety Depression	354	0.396	[0.304; 0.480]	0.000
Freire de Medeiros et al. 2017 (e10)	Patients on hemodialysis	Brasil	RS-25	Depression Mental QoL	202	0.511	[0.401; 0.606]	0.000
García-Maroto Fernández 2015 (e11)	Women with breast cancer	Spain	RS-25	Anxiety	202	0.030	[−0.109; 0.167]	0.672
Gotay et al. 2007 (e12)	Survivors of single/multiple cancer diagnoses	USA	RS-25	Depression	1 076	0.120	[0.059; 0.180]	0.000
Harding 2014 (e13)	Women after breast biopsy prior to receipt of results	USA	RS-14	Anxiety Depression	128	0.576	[0.447; 0.682]	0.000
Hennig 2011 (e14)	Patients with hypertension	Germany	RS-13	Mental QoL	197	0.120	[−0.013; 0.249]	0.076
Herrmann et al. 2011 (e15)	Infertile couples at encounter for ‧fertility testing	Germany	RS-25	Mental QoL	398	0.539	[0.465; 0.605]	0.000
Jaenichen et al. 2012 (e16)	Patients after severe sepsis (2–117 months)	Germany	RS-13	Distress	87	0.734	[0.619; 0.818]	0.000
Jang et al. 2017 (e17)	Burn victims after acute phase	Korea	RS-25	Anxiety Depression	138	0.519	[0.385; 0.631]	0.000
Jegan et al. 2017 (e18)	Patients with chronic low back pain	Germany	RS-11	Depression	423	0.497	[0.422; 0.566]	0.000
Kamen et al. 2017 (e19)	Homo- and bisexual women after breast cancer	USA	RS-14	Anxiety Depression	201	0.480	[0.365; 0.579]	0.000
Keil et al. 2017 (e20)	Patients with COPD	Germany	RS-13	Anxiety Depression	531	0.551	[0.489; 0.607]	0.000
King & Orel 2012 (e21)	Homosexual men with HIV/AIDS aged 45–78 years (M = 53.5)	USA	RS-14	Mental QoL	38	0.431	[0.129; 0.660]	0.006
Krause 2011 (e22)	Patients with type 2 diabetes mellitus	Germany	RS-13	Anxiety Depression Mental QoL	192	0.409	[0.281; 0.523]	0.000
Küch et al. 2016 (e23)	Patients undergoing inpatient ‧orthopedic rehabilitation	Germany	RS-13	Anxiety Depression	107	0.449	[0.283; 0.588]	0.000
Kunschitz et al. 2017 (e24)	Patients with coronary heart disease	Austria	RS-13	Anxiety Depression	166	0.281	[0.125; 0.423]	0.001
Kurz et al. 2014 (e25)	Patients with lung cancer	Germany	RS-13	Anxiety Depression	49	0.441	[0.182; 0.642]	0.001
Leontjevas et al. 2014 (e26)	Residents of a nursing home ‧rehabilitating unit	Netherlands	RS-25	Anxiety Depression	40	0.305	[−0.007; 0.563]	0.055
Li et al. 2016 (e27)	Bladder cancer patients	China	RS-14	Emotional wellbeing	365	0.402	[0.312; 0.485]	0.000
Liu et al. 2015 (e28)	Stable heart failure patients treated on outpatient basis	Taiwan/China	RS-25	Mental QoL	128	0.379	[0.220; 0.518]	0.000
Liu et al. 2017 (e29)	Patients with ovarian cancer	China	RS-14	Anxiety Depression	198	0.390	[0.265; 0.502]	0.000
Losoi et al. 2015 (e30)	Patients with mild traumatic brain ‧injury	Finland	RS-25	Depression	74	0.188	[−0.042; 0.399]	0.109
Lossnitzer et al. 2014 (e31)	Chronic heart failure patients with highly ‧distressing symptoms	Germany	RS-25	Anxiety Depression	186	0.356	[0.223; 0.475]	0.000
Mautner et al. 2013 (e32)	Pregnant women diagnosed with ‧preeclampsia	Austria	RS-13	Depression Mental QoL	67	0.502	[0.297; 0.662]	0.000
McGowan et al. 2018 (e33)	HIV-positive individuals	United Kingdom	RS-14	Anxiety Depression	195	0.413	[0.290; 0.523]	0.000
Moe et al. 2013 (e34)	Old patients (>79 years) with chronic disease receiving outpatient nursing care	Norway	RS-25	Mental QoL	120	0.388	[0.224; 0.530]	0.000
Müller et al. 2015 (e35)	Patients with chronic kidney disease before or after transplantation	Germany	RS-25	Anxiety Depression Mental QoL	252	0.391	[0.230; 0.531]	0.000
Pascoe & Edvardsson 2016 (e36)	Patients with prostate cancer	Australia	RS-14	Anxiety Depression	209	0.567	[0.468; 0.653]	0.000
Popa-Velea et al. 2017 (e37)	Cancer patients	Romania	RS-14	Distress	178	0.562	[0.452; 0.655]	0.000
Quiceno & Vinaccia 2013 (e38)	Patients with rheumatoid arthritis	Colombia	RS-25	Mental QoL	41	0.484	[0.207; 0.689]	0.001
Radke & Franke 2016 (e39)	Patients in outpatient orthopedic ‧rehabilitation	Germany	RS-13	Anxiety Depression	110	0.562	[0.418; 0.678]	0.000
Ramírez-Maestre et al. 2017 (e40)	Patients with acute back pain	Spain	RS-25	Depression	232	0.160	[0.032; 0.283]	0.015
Rebagliati et al. 2016 (e41)	Patients in orthopedic rehabilitation after hip or knee surgery	Italien	RS-10	Depression	81	0.504	[0.321; 0.651]	0.000
Robottom et al. 2012 (e42)	Patients with Parkinson’s disease	USA	RS-15	Anxiety Depression Mental QoL	83	0.376	[0.175; 0.548]	0.000
Ruiz-Párraga et al. 2015 (e43)	Patients with chronic musculoskeletal back pain	Spain	RS-18	Distress	592	0.193	[0.114; 0.270]	0.000
Runkewitz et al. 2006 (e44)	Patients in GP practice with acute or chronic disease	Germany	RS-25	Anxiety Depression	242	0.502	[0.401; 0.591]	0.000
Schumacher et al. 2014 (e45)	Patients after allogeneic stem cell transplantation	Germany	RS-25	Anxiety Depression Emotional functioning	75	0.503	[0.309; 0.656]	0.000
Suffeda et al. 2016 (e46)	Patients on the day before undergoing middle ear or pharyngeal ‧surgery	Germany	RS-13	Anxiety Depression	82	0.479	[0.291; 0.632]	0.000
Torma et al. 2013 (e47)	Older adults diagnosed with fibromyalgia	USA	RS-25	Depression	221	0.540	[0.439; 0.627]	0.000
Vinaccia & Quiceno 2011 (e48)	Hospitalized patients with COPD	Colombia	RS-25	Mental QoL	40	0.334	[0.025; 0.585]	0.035
Vinaccia & Quiceno 2011 (e49)	Patients with chronic kidney disease	Colombia	RS-25	Mental QoL	40	0.492	[0.213; 0.697]	0.001
Wallhäußer-Franke et al. 2014 (e50)	Tinnitus patients	Germany	RS-13	Anxiety Depression	4 705	0.572	[0.551; 0.591]	0.000
Wallhäußer-Franke et al. 2015 (e51)	Tinnitus patients (recent onset)	Germany	RS-13	Anxiety Depression	28	0.286	[−0.098; 0.595]	0.142
Wang et al. 2016 (e52)	Patients with malignant disease of the hematopoietic system	China	RS-14	Anxiety Depression	227	0.337	[0.216; 0.448]	0.000
Wick et al. 2015 (e53)	Patients awaiting liver or kidney transplantation	Germany	RS-13	Mental QoL	103	0.260	[0.070; 0.432]	0.008
Wolfram 2017 (e54)	Oncological patients with supplementary homeopathic treatment	Germany	RS-13	Distress Mental QoL	40	0.670	[0.454; 0.812]	0.000
Yang et al. 2016 (e55)	Patients with bladder or kidney cancer	China	RS-14	Anxiety Depression	489	0.474	[0.402; 0.540]	0.000
Mean effect size					15 003	0.432	[0.385; 0.477]	0.000

95% CI = 95% confidence interval; AIDS = acquired immunodeficiency syndrome; BMI = body mass index; COPD = chronic obstructive pulmonary disease; HIV = human immunodeficiency virus; QoL = quality of life; M = mean; N = number of patients; RS = Resilience Scale;

In 23 samples, patients with chronic illness were studied and in 22 samples patients with acute critical illness. Four samples assessed patients undergoing a medical intervention, e.g. a surgical procedure. In 5 other samples, a health problem, such as high blood pressure, was present, while 3 samples were comprised of individuals with heterogeneous health problems, e.g. elderly patients with chronic illness.

At the time of data collection in the included studies, the respective health problem was present in 46 samples, while it had occurred in the past in 10 samples (in one study the point in time was unclear).

In 24 studies, psychological resilience was measured using the original or a translated version of the Resilience Scale with 25 items. In the remaining studies, short versions with 10 to 18 items were used, particularly often the version with 13 items (table 1).

In most studies, mental health was measured using questionnaires assessing mental distress, i.e. by means of scales on anxiety (k = 30), depression (k = 40) or distress (k = 5). In the remaining cases, scales on mental quality of life (k = 17), on emotional functioning (k = 2) or on emotional wellbeing (k = 1) were used. The instruments used are listed in eTable 2 2.

eTable 2. Instruments used to measure mental health.

Instrument	Subscale	k
Hospital Anxiety and Depression Scale (HADS)	Full scale (distress)	3
	Subscale “anxiety”	13
	Subscale “depression”	18
Brief Symptom Inventory 18 (BSI-18) or Brief	Subscale “anxiety”	4
Symptom Checklist (BSCL) or Mini-Symptom Checklist (Mini-SCL)	Subscale “depression”	4
Patient Health Questionnaire (PHQ)	Depression module (PHQ-9)	8
Generalized Anxiety Disorder 7-item Scale (GAD-7)		6
State-Trait Anxiety Inventory (STAI)	Subscale „state anxiety“	1
	Subscale „trait anxiety“	1
	Full scale (anxiety)	2
Center for Epidemiological Studies Depression Scale (CES-D)		4
Beck Depression Inventory (BDI-II)		3
CONOR Mental Health Index (CONOR-MHI)		1
Rotterdam Symptom Checklist (RSC)	Subscale “psychological distress“	1
Edinburgh Postnatal Depression Scale (EPDS)		1
Geriatric Depression Scale (GDS)		2
Self-Rating Anxiety Scale (SAS)		2
State-Trait Operation Anxiety Inventory (STOA)		1
SF-36 Health Survey	Sum score “mental health”“	6
SF-12 Health Survey	Sum score “mental health”	8
SF-8 Health Survey	Sum score “mental health”	1
World Health Organization Quality of Life Assessment (WHOQOL-BREF)	Subscale “psychological health“	2
European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ-C30)	Subscale “emotional functioning“	2
Functional Assessment of Cancer Therapy scale – bladder-cancer-specific module (FACT-BL)	Subscale ”emotional wellbeing“	1

k = number of associations

Risk of bias within studies

Reliability of resilience scales was reported for 32 samples by stating Cronbach’s a; values varied between a = 0.85 and a = 0.97 with a mean of a = 0.91. The reliability of scales measuring mental health was reported for 47 associations and ranged between a = 0.60 and a = 0.97, with a mean of a = 0.85. An increased risk of potential bias through lack of measuring accuracy affecting the measured effect sizes (Cronbach’s a<0.80) was found for k = 9 associations (9%), while for k = 40 associations (42%) a low risk and for k = 46 associations (48%) an unclear risk was found (etable 3).

eTable 3. Risk of bias within studies.

Reliability of instruments	k*1	%
Acceptable reliability/low risk	40	42.1
Questionable reliability/high risk	9	9.5
Insufficient information/unclear risk	46	48.4
Incomplete reporting	n*2	%
Satisfactory reporting/low risk	46	83.6
Incomplete reporting/high risk	9	16.4
	i*3	M (SD)
Response rate	43	0.70 (0.24)

*¹ reported for associations, *² reported for studies, *³ reported for samples; for i = 14 samples no information available

M = mean; SD = standard deviation

Results of individual studies and synthesis of results

The measures of association in the individual studies varied between r = 0.03 (e11) and r = 0.73 (e16) and showed considerable heterogeneity (Q [56] = 538.97, p<0.001, I²= 89.6%). Five individual effect sizes were not significantly different from zero. All other study effects showed a significant positive association between resilience and mental health. In 11 studies a small effect, in 25 studies a medium effect, and in 21 studies a large effect was identified (28). The weighted mean correlation across all studies was r = 0.43; (95% confidence interval: [0.39; 0.48]; p<0.001) (etable 1).

Risk of bias across studies

Visual evaluation of the funnel plot revealed a deviation of the distribution of studies from the funnel shape which one would normally expect in the absence of publication bias. Assuming a random-effects model, the trim-and-fill analysis indicated that 10 studies were missing. The mean effect size adjusted for these studies is r = 0.39 (efigure). The left-sided test for asymmetry of the funnel plot using Egger’s regression test was not significant (p = 0.062). According to the classic Fail-safe N analysis, 31 159 studies with null results would be required for the mean correlation to exceed the significance level of a>0.05.

eFigure.

Funnel Plot of Fisher´s Z by standard error

Additional analyses

In addition, a search for intervening variables was conducted which could have an effect on the association between resilience and mental health and could thus help to explain the heterogeneity of study effects. A significant negative effect of sample size was identified: The larger the study sample, the weaker was the association between resilience and mental health (p = 0.045). Furthermore, a trend towards significant differences was found for publication status (p = 0.056): In unpublished studies, weaker associations were reported compared to those in published studies.

Other important intervening variables could not be identified (etable 4). The calculated mean effect was robust in sensitivity analyses to alternative approaches with regard to individual inclusion criteria and methods of analysis (etable 5).

eTable 4. Results of subgroup analyses and meta-regression.

Variable	Beta	SE	95% CI	p
Publication year	0.010	0.011	[−0.012; 0.032]	0.363
Sample size*1	−0.003	0.000	[−0.005; -0.000]	0.045
Mean age	−0.002	0.003	[−0.008; 0.005]	0.618
Proportion of female study participants	−0.061	0.133	[−0.321; 0.199]	0.644
Number of items of Resilience Scale	−0.001	0.006	[−0.012; 0.010]	0.821
Response rate	−0.086	0.146	[−0.373; 0.200]	0.554
	i	r	95% CI	p
Disease status*2				0.221
Current	46	0.41	[0.37; 0.46]
Past	10	0.51	[0.36; 0.64]
Type of illness				0.310
Health problem	5	0.30	[0.09; 0.49]
Medical intervention	4	0.51	[0.43; 0.59]
Chronic illness	23	0.44	[0.38; 0.50]
Acute critical illness	22	0.44	[0.36; 0.52]
Mixed sample	3	0.44	[0.33; 0.54]
Publication status				0.056
Published	51	0.45	[0.40; 0.49]
Unpublished	6	0.30	[0.14; 0.45]

*1 without Wallhäuser-Franke (with >4000 patients) (e50); *2 insufficient information in 1 study Beta = standardized regression coefficient; i = number of samples; SE = standard error; 95% CI = 95% confidence interval

eTable 5. Results of sensitivity analyses.

	i	r	95% CI	p	I2
Mean effect size	57	0.432	[0.385; 0.477]	<0.001	89.6%
Without unreliable instruments and insufficient reliability information	27	0.449	[0.369; 0.522]	<0.001	94.0%
Without incomplete reporting	48	0.429	[0.376; 0.479]	<0.001	91.2%
Without positive outliers*1	51	0.403	[0.357; 0.446]	<0.001	83.1%
Without negative outliers*2	51	0.466	[0.430; 0.501]	<0.001	79.1%
Measure of mental quality of life	20	0.391	[0.311; 0.466]	<0.001	79.9%
Measure of mental distress	46	0.449	[0.396; 0.499]	<0.001	90.7%
Measure of depression	40	0.471	[0.418; 0.520]	<0.001	90.0%
Measure of anxiety	28	0.429	[0.374; 0.481]	<0.001	85.4%
Hunter & Schmidt’s integration method (e65)	57	0.440	[0.139; 0.742]	<0.001	90.3%
Without prospective associations*3	56	0.433	[0.385; 0.478]	<0.001	89.8%

*¹ Amler et al. 2015 (e1); Cohen et al. 2014 (e5); Esteve & Ramírez-Maestre 2013 (1) (e8); Jaenichen et al. 2012 (e16); Keil et al. 2017 (e20); Wallhäuser-Franke et al. 2014 (e50)

*² Bathke 2011 (e2); Esteve & Ramírez-Maestre 2013 (3) (e8); García-Maroto Fernández 2015 (e11); Gotay et al. 2007 (e12); Hennig 2011 (e14); Ramírez-Maestre et al. 2017 (e40)

*³ Erim et al. 2015 (e7)

i = number of samples; I² = measure of heterogeneity; 95% CI = 95% confidence interval

Discussion

Altogether, 57 samples of patients with various physical illnesses and health problems from 55 studies were included in the meta-analysis. 95 associations between a version of the Resilience Scale (13) and a positive or negative measure of mental health were taken into account. Across the samples, a significant effect of r = 0.43 was determined, representing a moderate correlation. The higher the resilience in individuals with physical illness, the better they considered their mental health to be. This key result of our meta-analysis is consistent with previous findings on the association between higher resilience and better mental health in other contexts.

Other meta-analyses (25– 27), which did not define specific criteria for the studied sample and consequently included data from a variety of situations, found correlations in the moderate-effect range and evidence of high heterogeneity of effect sizes.

Overall, our meta-analysis comprises—despite a narrower choice of resilience scales used and stronger restrictions on the stress context in the included studies—more studies than some of the already available reviews on the association between resilience and mental health (26, 27, 29). Consequently, its representativeness for the studied area of resilience research can be regarded as good.

The Resilience Scale proved to be a reliable instrument in the included studies. The short forms appear to be especially well suited to economically measure resilience as a personality trait. With regard to measuring resilience, the issue of a potential tautology of resilience and depression is the subject of ongoing discussion. It is argued that some items of the Resilience Scale are merely positively worded depression-specific statements. If this were the case, the specific association between resilience and depression should be significantly stronger compared to the association between resilience and anxiety. However, this was not confirmed in sensitivity analyses.

In addition, there are obvious conceptual differences: The concept of resilience includes comparatively stable personality characteristics, such as self-reliance, perseverance, adaptability, balance, and flexibility. By contrast, depression as a temporary state compromises mental symptoms, such as feeling of inner void, loss of energy, self-doubt, fears, and physical symptoms. Here, the respective instruments differ significantly in the items they include.

Limitations

A key weakness of the present meta-analysis is founded in the design of the included studies, because associations identified in cross-sectional studies do not allow to draw causal inferences about the association between resilience and mental health.

Overall, very few prospective studies were available; of these, only two studies reported associations between resilience at a certain point in time and mental health at a later point in time. However, for reasons of methodological consistency, this review focused on cross-sectional associations. A further limitation stems from the substantial statistical heterogeneity of the individual study effects which makes it more difficult to generalize the results. However, all studies found positive associations; in 81% of samples, these were interpreted as moderate to strong.

Signs of publication bias represent another limiting factor. For example, unpublished studies found smaller effects compared with published studies. However, the calculated mean correlation was only slightly overestimated, because the adjusted mean effect is also significant and of medium size.

As yet, there is no common understanding of the stability or variability of resilience. While one line of research views resilience as a personality trait, another considers it to be a dynamic and changeable process and thus proposes different, outcome-based methods of operationalization and measurement (30).

Conclusion

Despite the limitations outlined above, the results of this meta-analysis indicate that in the context of physical illness or health problems a higher level of resilience is associated with better mental health.

In view of the breadth and diversity of psychological resilience research, the homogeneity of the primary studies with regard to the underlying conceptualization of resilience as a personality trait and the use of the Resilience Scale as the instrument (13) represents a strength of this review.

Given the numerous isolated findings in this field of research, a need for synthesis was identified. This study is the first to provide a statistical integration of the effects from 55 studies. With the higher validity of its findings compared to individual studies, the study makes a contribution to resilience research. Taking a meta-analytical approach, it ensures, in contrast to the existing reviews, a more objective selection and an integration of the results backed by statistical analysis. Prospective studies designed to clarify the causal nature of the association between resilience and mental health should be the focus of future research.

In view of the practice of medical psychology, this review provides starting points for a more targeted psychosocial support to help patients cope with physical illness. The appearance of symptoms of mental distress leads to increased care requirements as part of the medical management and can influence the course of illness and the successful outcome of treatment (21).

By screening patients for their levels of resilience, it would be possible to identify patients with low resilience early in clinical practice and to offer them more support in the form of external resources (5, 25).

Supplementary Material

eMethods

Methods

Inclusion criteria

Research questions, inclusion criteria and methods were pre-specified in a review protocol (PROSPERO CRD42017054822). Studies meeting the criteria listed in Table 2 were included in this review.

Literature search

To identify relevant studies, a search was performed in the electronic databases Medline, Web of Science, PsycInfo, PubPsych, ProQuest, and in the dissertation catalogue of the German National Library. The search strategies of the electronic database search are listed in the eBox.

In addition, the references of existing reviews and the included primary studies were searched manually for further relevant studies. Furthermore, to identify unpublished research, the lists of theses of the Institutes of Medical Psychology in Jena and Leipzig, where short forms of the Resilience Scale were developed, were screened (17, 18).

Study selection and data extraction

First, the title and abstract of the identified publications were assessed for eligibility, based on the pre-specified inclusion criteria. Relevant papers were obtained in full-text and then reassessed. In case of uncertainty about the eligibility of a reference, the study authors were contacted. Subsequently, characteristics of the studies, potential limitations to study quality, and the data required for calculating effect sizes were extracted. Study selection and coding was performed by the two authors (FF, JR); any disagreements were resolved in consensus discussion.

Risk of bias in individual studies

Since no clear recommendations on the assessment of potential risk of bias in cross-sectional observational studies are available (27), the MOOSE guidelines (e56), the STROBE statement (e57) and the Quality Assessment Tool for Quantitative Studies (e58) were consulted. To assess a potential risk of bias for the reported effects, the following indicators were used:

Effect size measure

By selecting the Pearson product-moment correlation as measure of effect size, it was possible in the majority of cases to extract effect sizes for the respective associations between resilience and mental health in the form of correlation coefficients directly from the eligible studies. In those cases where rank correlation coefficients (Spearman’s ? und Kendall’s t) were given, they were converted into Pearson correlations (e59, e60). Where other measures (e.g. regression and contingency coefficients) were reported, the authors were contacted and asked to provide the corresponding correlations.

Synthesis of results

To be able to integrate the individual effect sizes, first the correlations were converted using the Fisher‘s z transformation. Where multiple effect-size measures were available for one sample (e.g. several indicators of mental health, correlations provided for men/women separately), these were first aggregated to one measure per sample. In order to determine the weighted mean effect size for the association between resilience and mental health, the integration method by Hedges and Olkin (e61) was used under the assumption of a random-effects model. Finally, the pooled Fisher’s z values were converted back to Pearson correlations to improve interpretability. All effect size estimates are stated with 95% confidence intervals [CI]. The magnitude of an effect size is determined according to the conventions by Cohen (28). Hence, correlations from 0.10 are classed as a small effect, from 0.30 as a medium effect and above 0.50 as a large effect.

Risk of bias across studies

In order to estimate the impact of potential publication bias on the calculated mean effect sizes the funnel plot was visually evaluated. In addition, the asymmetry of the funnel plot was tested using Egger´s regression test (e62). A trim-and-fill analysis (e63) provided a mean effect size corrected for publication bias. Furthermore, the classic Fail-Safe N was determined (e64).

Additional analyses

In order to assess the heterogeneity of study effects and potential contributing factors, subgroup and meta-regression analyses were conducted. In addition, sensitivity analyses were performed to evaluate the robustness of the effects against a variation of various parameters with potential impact on validity (such as risk of bias, exclusion of statistical outliers) as well as the integration method (e65). All data analyses were performed using the Comprehensive Meta-Analysis Software (version 3.0; Biostat Inc.).

• Reliability of the instruments used to determine the association between resilience and mental health.

• Response rate as a measure of completeness of results data.

• Completeness of results reporting.

The Clinical Perspective.

Serious physical illness is a stressful experience many people are confronted with during their lifetime (22). Coping with such life situations requires adequate resources which are not equally available to everyone. Experiencing ongoing physical illness is frequently associated with mental distress, even to the extent of developing mental disorders, which may go along with a less favorable course of disease and a less positive rehabilitation outcome (21). Resilience is considered a key indicator of the maintenance of mental health in stressful circumstances and at times of personal crisis. Conversely, less resilient individuals are vulnerable to mental health impairment as a result of physical disease. Consequently, the physically ill with little resources in terms of resilience should receive targeted psychosocial support to minimize mental distress and the risk of depression or anxiety disorders (25). Here, screening using a shortened form of the Resilience Scale (18) is an economical way to identify such patients in clinical practice and to offer them supportive interventions early on in treatment planning (5).

Key Messages.

• Resilience is the term used to describe an individual’s positive adaptation in the face of adversity, i.e. one’s success in dealing healthily with significant stressors.

• The Resilience Scale in its long form and various short forms is used as an internationally well-established instrument to measure resilience as a personality trait.

• All studies included in the present meta-analysis demonstrated a positive association between resilience and mental health; however, considerable statistical heterogeneity of the effect sizes was found, ranging from small to large effects.

• In the context of physical illness, a higher level of resilience is associated with better mental health.

• Prospective studies on the association between resilience and mental health are required to be able to make statements on causality.

eBOX. Search strategies for the electronic database search.

• Search strategy PsycInfo: TX (“resilienc# scale“ OR Resilienzskala) NOT AG (adolescence OR childhood OR school age OR preschool age OR infancy OR neonatal)

• Search strategy Medline: (“resilience scale“ [All Fields] OR “resiliency scale“ [All Fields] OR Resilienzskala [All Fields]) OR “resilience, psychological“ [MeSH Terms] AND “adult“ [MeSH Terms]

• Search strategy PsycInfo: (“resilience scale” OR “resiliency scale“ OR Resilienzskala) NOT (AGE= [adolescence OR childhood OR school age OR preschool age OR infancy OR neonatal])

• Search strategy Web of Science: TOPIC: (“resilienc? scale“ OR Resilienzskala)

• Search strategy ProQuest Dissertations and Theses: “resilience scale“

• Search strategy catalog of the German National Library: “Resilienz “ restricted to – catalogs/collections: Dissertations and Theses