Factors associated with return-to-work outcomes in inpatient rehabilitation – a systematic scoping review

ABSTRACT

Inpatient rehabilitation is common in Germany to improve return-to-work outcomes. The objective of this systematic scoping review was to identify factors associated with return-to-work outcomes in musculoskeletal, psychological and oncological health conditions to improve tailoring of rehabilitation therapies. A search was completed in Embase, Medline, PsycInfo and AMED until May 2023 for articles investigating inpatient rehabilitation including working-age patients with oncological, musculoskeletal, or psychological diseases using a quantitative design and reporting factors associated with return-to-work outcomes. Screening of all titles and abstracts was completed by one reviewer, full texts were read by two reviewers. Quality appraisal and data extraction was completed by two reviewers. Data was analysed using a narrative synthesis. Eighteen studies of moderate quality were included. The review identified a wide range of return-to-work parameters including employment status, work ability, sickness absence, retirement status and duration of employment since rehabilitation. In addition, 48 psychological, health- and work-related factors associated with return-to-work parameters were identified. Only one RCT investigated the relationship between a depression prevention intervention and a return-to-work outcome (work ability), which showed a significant effect. In addition to the depression prevention intervention, only the factor ‘health literacy’ could be considered modifiable and be addressed as part of an inpatient rehabilitation programme. Furthermore, gradual work reintegration programs and/or workplace interventions in addition to inpatient rehabilitation should be further explored to improve return-to-work outcomes.

KEYWORDS:

• Inpatient rehabilitation
• return to work
• cancer rehabilitation
• musculoskeletal rehabilitation
• mental health rehabilitation

Introduction

The World Health Organization (WHO) estimates that currently, 2.4 billion people are living globally with a health condition that would benefit from rehabilitation (WHO, 2021). The WHO defines rehabilitation as ‘a set of interventions designed to optimize functioning and reduce disability in individuals with health conditions in interaction with their environment’ (WHO, 2021). Rehabilitation is considered a multidimensional process mainly influenced by personal characteristics and work-related variables (Gragnano et al., 2018). Depending on health condition and health service, rehabilitation therapies can be delivered in the community, in out- or inpatient clinics though there is limited evidence regarding the superiority of effectiveness of one setting over the other (Mittag & Welti, 2017). In Germany, rehabilitation therapy in inpatient clinics is offered for a wide range of health conditions to prevent early retirement, with musculoskeletal and oncological diseases as well as mental health problems being the most common reasons for referral (Deutsche Rentenversicherung, 2019). These health conditions comprise about 73% of all referrals (Deutsche Rentenversicherung, 2019).

A review of reviews identified four common facilitators (job control, work ability, perceived good health and high socioeconomic status) and six barriers (job strain, anxiety, depression, comorbidity, older age and low education) for returning to work, independent of disorder and setting (Gragnano et al., 2018). These factors could be used to support tailoring of the intervention to individual patient needs and understanding who might require additional support (Farin et al., 2006).

Given the frequency of musculoskeletal, psychological and oncological health issues qualifying for inpatient rehabilitation and the similarity of associated barriers and facilitators, our review question was: Which factors are associated with return-to-work outcomes of patients with musculoskeletal, psychological or oncologic diseases after inpatient rehabilitation programmes?

Methods

A scoping review using a narrative synthesis (Popay et al., 2006) was completed. The final systematic search was conducted on 21 May 2023. The review was not registered.

Systematic search

Databases (searched and combined in Ovid)

• Embase (1974 to 25 May 2023)

• Medline (1946 to 25 May 2023)

• PsycInfo (1906 to May week 2 2023)

• AMED (1985 to May 2023)

Search terms [abstract, keywords, MeSH term, subject heading, title]

Based on pilot searches identifying key words in the relevant literature, we determined the following search terms:

• (rehabilitation or return to work or return-to-work or back-to-work or back to work).ab,ti.

AND

• (cancer or newplasm* or carcino* or onco* or tumor or chronic pain or musculoskeletal or psycho* or mental health).ab,ti.

AND

• (predict* or factor* or determinant* or contributer*).af.

AND

• (inpatient*).af.

Where possible, the search was limited to humans. Duplicates were removed.

*ab = abstract, ti = title, af = all fields

Eligibility criteria

For eligibility criteria, we used the CoCoPop structure (Condition, Context, Population) (Munn et al., 2020). Exclusion criteria were only stated if different to the reversal of inclusion criteria.

Condition

Inclusion: studies including patients with oncological, musculoskeletal or psychological diseases.

Exclusion: any other disease group, although overlaps (e.g. lung cancer) would qualify the study for inclusion.

Context

Inclusion: inpatient rehabilitation delivered in rehabilitation centres; for studies that consider both inpatients and outpatients, the proportion of inpatients was required to be above 50% or to include a separate analysis.

Population

Inclusion: participants of working age (18–65 years). More precisely, the reported mean age of the study participants, as well as the standard deviation, should be between 18 and 65 at baseline.

Types of studies

Inclusion: quantitative (pre- and post-study designs, quasi-experimental designs or randomized control trials) peer-reviewed primary research studies in English, German, Italian, Spanish, French or Dutch.

Exclusion: all other study designs (non-pre- post designs), reviews, abstracts, protocols, editorials, discussion papers, comments (unless relating to one of the included studies) conference papers (if they consist of less than 300 words), editorials, book chapters and similar publications. Mixed-method studies were included if the quantitative and qualitative parts are analyzed separately.

Outcome

Inclusion: studies investigating factors associated with return-to-work outcomes (number of sick days, ability to work (e.g. Work Ability Index (El Fassi et al., 2013)), pension applications, number of days until return-to-work (RTW) occurred). Due to the lack of conceptual consistency of measuring RTW (Wasiak et al., 2007), we employed an inclusive approach to RTW outcomes.

Data selection

Selection of studies

All identified records were imported into a literature management program (Zotero) and screened for eligibility using the in- and exclusion criteria by one researcher (KF or VvdW). Duplicates were deleted and numbers recorded. If eligible or if the eligibility was uncertain, the article was retained. Next, full-text articles for all candidate eligible studies were retrieved and assessed by two reviewers (KF and VvdW). Each full-text was examined using a screening checklist based on in- and exclusion criteria. Reasons for non-eligibility were recorded.

Data extraction

Data were extracted using a custom-designed form. The data included author, year, country, study design, sample, intervention/comparator (if applicable), outcomes (RTW and associated factors included in the analysis) as well as results.

Data analysis

We completed a descriptive analysis of the included papers and a quality analysis of the studies based on the Critical Appraisal Skills Programme (CASP) checklist for cohort studies (Brice, n.d..). The appraisal tool was used for all studies to assess quality independently by two authors (KF and VvdW).

Due to the diversity of outcomes, we could not analyse the findings for the different illness groups (musculoskeletal, oncological, mental health). A narrative synthesis (Popay et al., 2006) was completed to investigate factors associated with RTW outcomes.

Results

The search identified in total 4301 articles in the databases including 78 duplicates (see Figure 1). After screening titles and abstracts based on the eligibility criteria, 4138 records were excluded. One article could not be obtained; therefore, 84 full texts were examined, resulting in 18 studies being included in the review.

Figure 1. Flow-chart for the systematic review process.

^†multiple exclusion reasons may have occurred.

The articles were published between 1990 and 2023. They included two RCTs and 16 cohort designs (analysis of insurance data records (n = 3), prospective studies (n = 8), a retrospective cohort study (n = 2) and longitudinal studies (n = 2)). Most studies were conducted in Germany (n = 13), followed by Norway (n = 3), Finland (n = 1) and the U.S.A. (n = 1). Studies examining rehabilitation in musculoskeletal disease analysed the data of 3,986 participants, studies investigating people with mental health issues included data of 593 participants and in oncological diseases the data of 2,440 participants was analysed. In addition, four studies included participants with multiple (oncological, musculoskeletal, mental health) diseases. These analysed the data of 253,323 participants (including one retrospective database study with 252,591 participants).

Quality appraisal

Overall, the quality of the studies was moderate (see Table 1). None of the studies explained considerations regarding all potentially confounding factors, seven studies did not present all outcome variables reported to be in the analysis (Bernardy & Sandweg, 2003; Hampel & Neumann, 2022; Holstiege et al., 2017; Meixner et al., 2022; Meng et al., 2021; Simmel et al., 2023; Tan et al., 1997). In seven articles relevant statistics such as non-significant variables and/or confidence intervals were omitted (Bernardy & Sandweg, 2003; Gissendanner et al., 2019; Hampel et al., 2009; Holstiege et al., 2017; Meixner et al., 2022; Meng et al., 2021; Tan et al., 1997).

Table 1. Assessment of quality according to the CASP criteria.

Author, year	Clear focus	Acceptable recruitment	Exposure measured to minimize bias	Outcome accurately measured to minimize bias	Identifica-tion of all confounding factors	Confoun-ding factors are taken into account in design/analysis	Follow-up complete enough^1	Follow-up long enough^2	Results precise; results believable	Applicable to local population	Results fit with other evidence
Bernardy and Sandweg (2003)	Yes	Yes	Yes	Yes	No	Yes	No (65%)	Yes	No^3,4; Yes	Yes	Yes
Ebinger et al. (2002)	Yes	n/a	No	Yes	No	Yes	n/a	Yes	Yes; Yes	Yes	Yes
Gissendanner et al. (2019)	Yes	Yes	Yes	Yes	No	No	No (65%)	yes	No^4; No	Yes	Yes
Hampel et al. (2009)	Yes	Yes	No	Yes	No	Yes	No (64%)	Yes	No^4; Yes	Yes	Yes
Hampel and Neumann (2022)	Yes	Yes	Yes	Yes	No	Yes	No (64%)	Yes	Yes^3; Yes	Yes	Yes
Holstiege et al. (2017)	Yes	n/a	No	No	No	Yes	n/a	Yes	No^3,4; Yes	Yes	Yes
Jacobsen et al. (2020)	Yes	Yes	Yes	Yes	No	No	No (65%)	Yes	Yes; Yes	Yes	Yes
Kus et al. (2022)	Yes	Yes	Yes	Yes	No	Yes	Yes (81%)	Yes	Yes; Yes	Yes	Yes
Meng et al. (2021)	Yes	Yes	No	Yes	No	Yes	Yes (98%)	Yes	No^3,4; Yes	Yes	Yes
Mehnert et al. (2017)	Yes	Yes	Yes	Yes	No	Yes	No (65%)	Yes	Yes; Yes	Yes	Yes
Meixner et al. (2022)	Yes	Yes	Yes	Yes	No	No	Yes (82%)	Yes	No^3; No^4	Yes	Yes
Neuner et al. (2013)	Yes	n/a	No	Yes	No	Yes	Yes (83%)	Yes	Yes; Yes	Yes	Yes
Oyeflaten et al. (2008)	Yes	Yes	Yes	Yes	No	Yes	Yes (83%)	Yes	Yes; Yes	Yes	Yes
Rick (2021)	Yes	Yes	Yes	Yes	No	Yes	Yes (82%)	Yes	Yes; Yes	Yes	Yes
Saltychev et al. (2014)	Yes	Cannot tell	No	Yes	No	Yes	No (57%)	Yes	Yes; Yes	No	Yes
Simmel et al. (2023)	Yes	Yes	Yes	Yes	No	No	Yes (82%)	Yes	No^3 Yes	Yes	Yes
Skagseth et al. (2021)	Yes	Yes	No	Yes	No	Yes	Yes (85%)	Yes	Yes; Yes	No	Yes
Tan et al. (1997)	Yes	Cannot tell	Yes	Yes	No	Yes	Yes (81%)	Yes	No^3,4; Yes	Cannot tell	Yes

^aConsidered enough if follow-up assessments were completed by more than 80% of the participants; ²Considered enough if follow-up interval longer than 3 months, ³ Not all variables reported; ⁴95% CI not given.

Return-to-work outcomes

Return-to-work (RTW) related outcomes included employment status (Bernardy & Sandweg, 2003; Ebinger et al., 2002; Jacobsen et al., 2020; Kus et al., 2022; Mehnert et al., 2017; Meixner et al., 2022; Oyeflaten et al., 2008; Rick, 2021; Tan et al., 1997), work ability (Gissendanner et al., 2019; Hampel & Neumann, 2022; Meng et al., 2021; Saltychev et al., 2014), duration of sickness absence (Hampel et al., 2009; Simmel et al., 2023; Skagseth et al., 2021), duration of employment since rehabilitation (Holstiege et al., 2017), and time until early retirement (Neuner et al., 2013). Meta-analyses were not possible due to varying outcome measures (different categories of employment status or scales) and different follow-up intervals (between 6 months and 5 years).f

Factors associated with RTW

Factors reported to be associated with RTW are shown in Table 2. In total, 48 factors were shown to be significantly associated with RTW outcomes. Factors were summarised into demographic, physiological health-related, psychological/work-related dimensions.

Table 2. Included articles.

Author, year, country	Study design, follow-up duration and sample characteristics (clinical group, sample size, age, gender)	Intervention (IG), comparison (CG; if applicable)	Return-to-work outcome	Predictors included in analysis	Results†
Bernardy and Sandweg (2003)	Retrospective database analysis with 5 yrs f-up; MSK; n = 226; mean age 47.1 yrs; 77.5% women.	IG: 7-wk interdisciplinary inpatient rehabilitation incl. psychotherapy and diagnosis-guided PT. No CG.	Employment status (working, early retirement).	Self-ratings (duration of sick leave before treatment, work satisfaction, work stress, pain intensity, mental health, wish to retire); Physician assessment (motivation to receive therapy, outcome prognosis, personality and relationship characteristics, mental health status);Orthopedist assessment (imaging results, laboratory parameters, MSK diagnosis)	Results of a logistic regression analysis indicated that clear pathological imaging results (OR 3.81^1), duration of sick leave (OR 1.12^1) and wish to retire (OR 0.28^1) at baseline sig. predicted employment status. The other predictors did not contribute significantly to the model.
Ebinger et al. (2002)	Retrospective database analysis with 1–2 yrs f-up; (mainly) MSK; n = 841; mean age 48.3 yrs; 25.4% women.	IG: inpatient rehabilitation. No CG.	Employment status (working, work disability/early retirement)	Routinely recorded data incl. demographic characteristics, health data, health service use, referrals, health-related well-being.	A logistic regression showed that employment status was sig. predicted by higher age (OR 1.11 (95%CI 1.07–1.14)); sick leave in the last 12 mths compared to less than 3 mths: (3–6 mths: OR 1.13 (95% CI 0.6–2.1), over 6 mths: OR 3.15 (95% CI 1.89–5.24), not working: OR 4.25 (95% CI 1.19–15.17)); increase in physicians’ estimate at start of treatment of risk for inability to work compared to baseline (OR 2.96 (95% CI 1.74–5.03); and increase in degree of difference between physicians estimate at start of treatment and end of treatment regarding risk for inability to work (OR2.69 (95%CI 1.4–5.16). The other predictors did not contribute significantly to the model.
Gissendanner et al. (2019), Germany	Prospective cohort study with 13 mths f-up; MH; n = 401; mean age 49.0 yrs; 71.6% women.	IG: 5 wks of multidisciplinary inpatient rehabilitation incl. psychotherapy and work-related motivation support. No CG.	Work ability: binary composite variable including duration of sick leave, reduction in earning capacity and work disability.	Demographic characteristics; health status; resilience; symptom burden; self-regulation; intrinsic work motivation; subjective prognosis of work ability.	Work ability was significantly associated with self-reported expectation of employment status at the start and end of rehabilitation (both OR 1.6), symptom burden (OR 2.6) and intrinsic motivation at the start of rehabilitation (OR 0.7) and self-regulation (OR 1.1) at the end of rehabilitation Other predictors did not sig. contribute to the model.
Hampel et al. (2009), Germany	Prospective non-randomised controlled study with 12 mths f-up; MSK; n = 199; mean age 49.0 yrs; 41.2% women.	IG: standard multimodal and multidisciplinary treatment (3–4 wks) plus cognitive-behavioural pain management and cognitive-behavioural management of depressive symptoms; the IG included only people with moderate and severe depressive symptoms. CG: standard multimodal and multidisciplinary treatment (3–4 wks) and cognitive-behavioural pain management. 2 CGs: 1) patients with no and mild depression symptoms, and 2) patients with moderate and severe depressive symptoms.	Duration of sick leave in days over prior 3 mths at 6 and 12 mths.	Demographic characteristics; stage of chronicity of MSK pain; global improvement; pain intensity; depressive symptoms; health-related quality of life; somatization.	In the CG group including people with no or mild depressive symptoms, men but not women showed sig. less days of sick leave at both follow-up assessments compared to baseline; in the CG group including people with moderate or severe depressive symptoms, no sig. effects were shown at 6 mths, at 12 mths women but not men had sig. less days of sick leave compared to baseline. In IG, both genders had sig. improved in days of sick leave at 6-month follow-up assessment but not 12 mths. The other predictors did not contribute significantly to the model.
Hampel and Neumann (2022), Germany	RCT with 24mths f-up; MSK, MH; n = 422; mean age 53.8 yrs; 81.5% women.	IG: standard pain competence training (4 training units and 3 educative presentations) plus depression prevention training (4 training units and 8 small group workshops). CG: standard pain competence training (4 training units and 3 educative presentations). No CG.	Work ability index at baseline, 6, 12 and 24 mths.	Demographic characteristics; depression; anxiety, pain-related self-efficacy; self-reported expectation of future employment status; pain intensity.	Work ability sig. improved in the IG at all time points but did not improve in the CG but there was no sig. difference in work ability change over time between groups. Work ability was sig. higher in those with lower depression (Cohen’s d 0.69).
Holstiege et al. (2017), Germany	Retrospective pension database study with 4 yrs f-up; mixed (MH, MSK and CV); n = 252,591; median age 48 yrs; 47.2% women.	IG: standard inpatient rehabilitation treatment (15 to 42 days). No CG.	Number of months of employment in the 4 years after completing the rehabilitation therapy.	Age at end of rehabilitation therapy; gender; citizenship (German, non-German); request by health insurance to enter rehabilitation therapy (yes/no), mths of sick leave within 12 mths before start of rehabilitation; benefits paid per day in 12 mths before start of rehabilitation.	Sig. predictors for less mths of employment within 4 yrs after rehabilitation were being a woman, age (<30 yrs or 50–61 yrs), being classified as ‘not German’, request by pension insurance to enter rehabilitation, sickness absence for 6 mths or more, not receiving benefits per day or receiving less benefits, and being unemployed at mth 24 after rehabilitation.
Jacobsen et al. (2020), Norway	Nested cohort study within an RCT with 56 wks f-up; MH; n = 137; mean age 43.0 yrs; 81% women.	IG: 3.5 wks of inpatient occupational rehabilitation programme including mental and physical training as well as work-related problem solving. No CG.	Re-entry into workforce.	Metacognitive Beliefs (MCQ-30) with 5 subscales (cognitive confidence, positive believes, cognitive consciousness, uncontrollability, need to control thoughts)	Those with higher MCQ-30 scores had sig. higher odds to re-enter the workforce (OR 1.05 (95% CI 1.00–1.10)). When analysed at sub-scale level only the dimension ‘need to control thoughts’ remained sig. (OR 1.20 (95% CI 0.03–1.39)).
Kus et al. (2022), Germany^2	Prospective cohort study with 78 wks f-up; MSK; n = 761; mean age 47.5 yrs; 26.3% women.	IG: MSK trauma inpatient rehabilitation. No CG.	Re-entry into workforce.	Age, gender, education, vocational training, social status, cultural background, employment status, type of employment, type of injury, health status, physical and emotional functioning, pain intensity, addictive behaviour, environmental factors, subjective prognosis of RTW, demand for pension claim, having legal disputes, having financial concerns, inability to work before accident, personality traits, self-efficacy, attitude to life, appraisal of consequences of the accident, life satisfaction pre-accident, resilience, coping with the injury, disease gain, quality of life.	Higher rates of RTW were sig. predicted by higher scores of current state of health (OR 1.84 (95% CI 1.10–3.07)) and higher scores on the personality traits neuroticism (OR 1.36 (95% CI 1.08–1.73)). Lower rates of RTW were sig. associated with higher pain intensity (OR 0.45 (95% CI 0.27–0.73)), working in a blue collar profession (OR 0.47 (95% CI 0.28–0.81)), having ongoing legal disputes (OR 0.43 (95% CI 0.27–0.69)), having financial concerns (OR 0.46 (95% CI 0.26–0.81)), higher scores on the personality trait conscientiousness (OR 0.69 (95% CI 0.49–0.96)), having higher life satisfaction before the accident (OR 0.33 (95% CI 0.12–0.78)), and the demand for a pension claim (OR 0.09 (95% CI 0.03–0.30)).
Mehnert et al. (2017), Germany	Prospective cohort study with 12mths f-up; OD; n = 750; mean age 48.7 yrs; 85.7% women.	IG: 3–4 weeks cancer rehabilitation (incl. patient education, PT, relaxation training, psychosocial counselling, individual psychotherapy). No CG.	Early retirement status.	Age, gender, marital status, registered partnership, socio-economic status, time since diagnosis, tumor diagnosis, remission, activities of daily living, number of therapies, status of remission, number of physical impairments, number of comorbidities, pain during last week, status sick leave, number of days absent from work, type of cancer rehabilitation, desire to retire early, work satisfaction, perceived work productivity, depressive symptoms, physical and mental quality of life.	Based on a logistic regression, early retirement was sig. predicted by older age (OR 1.09 for each yr (95% CI 1.02–1.16)), higher levels of pain during last week (OR 1.19 (95% CI 1.02–1.38)), being on sick leave (OR 6.50 (95% CI 1.97–21.47)), desire to retire early (OR 5.61 (95% CI 2.73–11.52)) and being absent from work for 90 days or more (OR 3.15 (95% CI 1.05–9.49)). Perceived work productivity (OR 0.38 (95% CI 0.18–0.83)), work satisfaction (OR 0.36 (95% CI 0.17–0.77)) and cancer remission (OR 0.23 (95% CI 0.10–0.53)) were protective factors. Lower mental quality of life (OR 0.94 (95% CI 0.91–0.98)) was also associated with early retirement (protective factor). The other predictors did not contribute significantly to the model.
Meixner et al. (2022), Germany	Retrospective cohort study with 3, 12 and 18 mths f-up; OD; n = 424; median age 61 (28–95) yrs; 100% women.	IG: Inpatient cancer rehabilitation supported by social service consultations to provide knowledge and interdisciplinary assistance for physical, psychological, administrative, social and financial burdens, to simplify access to rehabilitation services, equipment and legal claims; post-treatment rehabilitation was performed by physicians and physiotherapists. No CG.	Return to work status.	Demographic characteristics, comorbidities, lifestyle factors, smoking status, current working status, practiced profession, toxicity (Common Terminology Criteria for Adverse Events), oncological outcomes (review of clinical data, imaging results, referrals), depression, fatigue, cancer stage according to the International Federation of Gynecology and Obstetrics classification (FIGO), body mass index (BMI).	RTW rates were sig. lower in participants with higher FIGO stage, definitive treatment concepts, those who experienced fatigue during and following radiotherapy, those who had acute pain during radiotherapy and those with depression. RTW was sig. higher in those who were self-employed and those within normal BMI range at the end of radiotherapy. Odds ratios and 95%CI were not reported.
Meng et al. (2021), Germany	Prospective cohort study with 3 and 9 mths f-up; OD; n = 449 with n = 216 in work; mean age 58.8 yrs; 63% women.	IG: oncological inpatient rehabilitation (no further details provided). No CG.	Self-reported work ability (Likert scale item).	Health literacy, age, gender, education, marital status, diagnosis, health status and type of rehabilitation.	Based on a linear regression analysis, higher health literacy was sig. related to higher self-reported work ability at 9 mths (ß = 0.21; p < 0.01); also, improvements in health literacy was sig. related to higher self-reported work ability at 9 mths (ß = 0.23; p < 0.01) when controlling for age, gender, education, marital status, diagnosis, health status and type of rehabilitation.
Neuner et al. (2013), Germany	Prospective cohort study with 5 yrs f-up; MSK; n = 1,268; mean age = 50.9 yrs; 31.9% women.	IG: 3-week multidisciplinary inpatient rehabilitation programme: No CG.	Time until early retirement.	Self-rated at baseline: somatic status, functional status, mental health, pain status, therapy goals.	Based on a Cox proportional hazard regression, not achieving the following therapy goals predicted early retirement: goal to reduce pain (HR 1.93 (95% CI 1.31–2.85)); goal to improve spine motility (HR 1.59 (95% CI 0.93–2.72)); goal to improve muscle strengths (HR 2.32 (95% CI 1.14–4.70)).
Oyeflaten et al. (2008), Norway	Prospective cohort study with 1 yr f-up; mixed (MSK, MH and unspecific health complaints); n = 135; mean age 45.0 yrs; 64% women.	IG: 4-week inpatient multidisciplinary vocational rehabilitation incl. a combination of individual and group-based interventions with physical activity, education, and cognitive behavioural modification. No CG.	RTW status (RTW defined as return to work-related activity).	Gender, age, education, subjective health complaints (MSK, pseudoneurology, gastrointestinal, allergy, influenza), illness perceptions questionnaire (identity, timeline acute/chronic, timeline cyclical, consequences, personal control, cure control, illness coherence, emotional representations and cause), physical fitness, physical activity, fear-avoidance beliefs (for physical activity and work), coping (perceived self-efficacy scale, instrumental mastery-oriented coping, emotion focused coping, coping expectancy and self-efficacy, hopelessness, quality of life).	Based on a logistic regression, not returning to work at 12 mths was predicted by education (≤12 yrs; OR 0.2 (95% CI 0.09–0.64)), higher scores in fear avoidance beliefs for work (OR 6.9 (95% CI 2.30–20.91)) and higher instrumental mastery-oriented coping scores (OR 3.1 (95% CI 1.20–8.24). The other predictors did not contribute significantly to the model.
Rick (2021), Germany	Retrospective analysis of cohort data with variable f-up intervals (5.7–8.5 mths); OD; n = 817; mean age 58 yrs; 74% women.	IG: standard oncological inpatient rehabilitation with or without a gradual reintegration to work program. No CG.	RTW at follow up.	gradual reintegration program, self-rated work ability, work stress, gender, age, education, health events and status, numbers of chemotherapy cycles, social-economic status.	A logistic regression analysis showed that RTW was predicted by participation in the gradual reintegration program (OR 20.12, 95% CI 8.60–47.06), less work stress (OR 0.22, 95% CI 0.1–0.48) and higher numbers of cycles of chemotherapy (OR 1.17, 95% CI 1.06–1.29). The other predictors did not contribute significantly to the model.
Saltychev et al. (2014), Finland	Prospective cohort study with variable f-up intervals (up to 9 yrs); MSK; n = 854; mean age 51.0 yrs; 87% women.	IG: in total, 15–33 days of rehabilitation in 2–4 inpatient stays over a period of 1–2 years. The rehabilitation programme consisted of physical training and psychological education. Between the inpatient periods, participants were encouraged to exercise at home. Rehabilitation teams consisted of physicians, physiotherapists, psychologists, and vocational rehabilitation specialists. No CG.	Self-reported work ability at mean intervals of 1.5 (short term) and 7.13 (long-term) yrs.	Gender, age, type of rehabilitation, occupation grade, marital status, education, job contract, length of job contract, shift work, intention to leave work, job security, job control, obesity, physical inactivity, smoking, alcohol consumption, sickness absence, chronic severe condition, use of antidepressants or painkillers, anxiety, suboptimal perceived health, psychological distress, diagnosis of fibromyalgia.	Changes (group x year) of self-reported work ability were shown using a repeated measures-regression analysis for occupation grade, education, job contract (permanent/fixed term), sickness absence, use of painkillers, anxiety, suboptimal perceived health and psychological distress. Parameters for regression analysis results and other predictor outcomes not reported.
Simmel et al. (2023), Germany^2	Prospective cohort study with 78 wks f-up; MSK; n = 612; mean age 48.5 yrs; 7.5% women.	IG: MSK trauma inpatient rehabilitation. No CG.	Duration of work incapacity in days.	Age, gender, education, vocational training, social status, cultural background, employment status, type of employment, type of injury, health status, physical and emotional functioning, pain intensity, addictive behaviour, environmental factors, subjective prognosis of RTW, demand for pension claim, having legal disputes, having financial concerns, inability to work before accident, personality traits, self-efficacy, attitude to life, appraisal of consequences of the accident, life satisfaction pre-accident, resilience, coping with the injury, disease gain, quality of life.	The results showed a sig. negative correlation between quality of life duration of work incapacity (r = 0.27; p < .0001).
Skagseth et al. (2021), Norway	RCT with 12 mths f-up; MSK and MH; n = 175; mean age 45.5 yrs; 78.9% women.	IG: 4-week (2 weeks at rehabilitation centre, one week at home and one week at the centre) multimodal inpatient rehabilitation (IMORE) including acceptance and commitment therapy, PT, work-related problem solving and creating an RTW plan and the workplace intervention (WI); CG: workplace intervention (WI): workplace meeting, summary meetings and writing of a report after the workplace meeting.	Sickness absence (number of days during 12 months f-up).	Return to work self-efficacy (RTWSE-19; primary outcome for RCT) incl. 3 subscales (meeting job demands, modifying tasks, communication needs).	A linear regression analysis adjusted for sex, age, education and intervention/control group included the entire sample (IG and CG) and showed a sig. association between number of sick days and total RTWSE score. Compared with participants with low RTWSE scores at the end of the rehabilitations, those with medium RTWSE scores had 42 fewer absence days (95% CI −76 to −8), and participants with high RTWSE scores had 85 fewer absence days (95% CI −33 to −37). The effect of the intervention on sickness absence was not reported. There was not sig. difference between IG and CG over time in RTWSE scores (0.04; 95% CI −0.69 to 0.78).
Tan et al. (1997), USA	Prospective cohort study with 6 mths f-up; MSK; n = 59 mean age 39.6 yrs; 62.7% women.	IG: inpatient pain management programme in university setting.	Return to work (any type of gainful employment) within the 6 mths f-up.	Age (<45), sex, ethnicity, education (>12^th grade), marital status (single), length of unemployment (≤16mths), time at job, number of premorbid jobs, duration of pain (10–30 mths), number of surgeries, RTW goal, compensation status, circumstances of onset (further details regarding outcomes and use as binary variables not reported).	Sig. relationships were shown between RTW and younger age (phi^2 .38), relationship status ‘single’ (phi^2 .30), education (phi^2 .30), initial RTW goal (phi^2 .53), shorter duration unemployment (phi^2 .37), and shorter duration of pain (phi^2 .35). The other predictors did not contribute significantly to the model.

†reported as significant if p < 0.05; ¹ confidence interval not reported;² analyses in Kus et al. (2022) and Simmel et al. (2023) are based on the same dataset; CG: control group; CV: cardiovascular disease f-up: follow-up; MH: mental health issues; MSK: musculoskeletal disease; mth: month; OD: oncological disease; PT: physiotherapy; RCT: randomised controlled trial; RTW: return to work; wk: week; yr: year.

Demographic factors

Demographic factors included gender, age, marital status, ethnicity, education and health literacy. Being a woman (Hampel et al., 2009; Holstiege et al., 2017), older age (Ebinger et al., 2002; Mehnert et al., 2017), being classified as ‘not German’ (Holstiege et al., 2017), being classified as blue collar worker (Kus et al., 2022), having ongoing legal disputes (Kus et al., 2022), having financial concerns (Kus et al., 2022), demanding pension claims (Kus et al., 2022), shorter education (Oyeflaten et al., 2008; Tan et al., 1997) and a lower level of health literacy (people’s competency to understand and apply health information) (Meng et al., 2021) decreased the likelihood of returning to work. The relationship status being classified as ‘single’ (Tan et al., 1997) and being a self-employed (Meixner et al., 2022) significantly increased the likelihood of returning to work.

Different outcomes were used to assess work absence at baseline or during follow-up periods including status and duration of sick leave (Bernardy & Sandweg, 2003; Ebinger et al., 2002; Mehnert et al., 2017; Skagseth et al., 2021), status and duration of not working/unemployment (Ebinger et al., 2002; Holstiege et al., 2017; Mehnert et al., 2017). These studies indicated that less work absence at baseline or during the follow-up period was significantly related to a higher likelihood of returning to work at the final follow-up assessment. Other work-related factors negatively associated with returning to work were ‘request by the pension insurance to enter rehabilitation (Holstiege et al., 2017) and ‘not receiving/receiving less sickness benefit payments per day’ (Holstiege et al., 2017). These results emerged in studies examining musculoskeletal as well as oncological diseases but were not investigated in the studies including patients with mental health problems.

Physiological health-related factors

While parameters indicating the severity of the illness were included in 12 out of 18 studies (Bernardy & Sandweg, 2003; Ebinger et al., 2002; Gissendanner et al., 2019; Hampel et al., 2009; Mehnert et al., 2017; Meixner et al., 2022; Meng et al., 2021; Neuner et al., 2013; Oyeflaten et al., 2008; Rick, 2021; Saltychev et al., 2014; Tan et al., 1997), only eight studies found illness severity parameters associated to return to work or sickness absence outcomes.

For studies including patients with musculoskeletal diseases, physical health-related factors predicting returning to work included clear pathological imaging results showing a positive association with early retirement (Bernardy & Sandweg, 2003); the physicians’ estimate of work inability risk at the start of the treatment and the difference of this estimate at start and end of treatment was positively associated with the risk of early retirement or reduced working capacity (Ebinger et al., 2002); shorter duration of pain was positively associated with earlier return to any type of gainful employment (Tan et al., 1997) and not achieving the therapy goals to reduce pain as well as not achieving the goal to improve spine mobility and muscle strengths predicted early retirement (Neuner et al., 2013).

For studies including patients with mental health problems, a lower ‘symptom burden’ was associated with a higher likelihood to return to work (Gissendanner et al., 2019). In patients with oncological diseases, lower pain (Mehnert et al., 2017; Meixner et al., 2022), remission (Mehnert et al., 2017), participation in a gradual reintegration programme (Rick, 2021), a normal BMI at the end of radiotherapy (Meixner et al., 2022) and higher numbers of cycles of chemotherapy (Rick, 2021) were related to a higher likelihood of returning to work. The likelihood was lower in cancer patients with a worse classification for cancer stage (Meixner et al., 2022), with a definitive treatment concept after all other choices had been considered (Meixner et al., 2022), if they experienced fatigue during radiotherapy and at follow-up (Meixner et al., 2022) and if they had pain during radiotherapy (Meixner et al., 2022).

Psychological and work-related factors

A wide range of psychological predictors was included in the analyses across all three illness groups: this included the wish to retire (Bernardy & Sandweg, 2003; Mehnert et al., 2017) and depression (Hampel & Neumann, 2022; Meixner et al., 2022), which were related to a lower likelihood of returning to work; self-reported expectation of being employed assessed at the start and end of the rehabilitation therapy showed a positive relationship with returning to work at follow-up (Gissendanner et al., 2019). Higher levels of neuroticism (Kus et al., 2022), intrinsic motivation (Gissendanner et al., 2019), self-regulation (Gissendanner et al., 2019), perceived work productivity (Mehnert et al., 2017), work satisfaction (Mehnert et al., 2017), generic quality of life (Simmel et al., 2023) and mental quality of life (Mehnert et al., 2017), and higher scores in instrumental mastery-oriented coping (Oyeflaten et al., 2008) were significantly related to a higher likelihood of returning to work. Higher levels of conscientiousness (Kus et al., 2022), better ratings of life satisfaction pre-accident (Kus et al., 2022), fear avoidance beliefs for work (Oyeflaten et al., 2008) and work stress (Rick, 2021) as well as the meta-cognitive belief ‘need to control thoughts’ (Jacobsen et al., 2020) were related to a lower likelihood to return to work. A randomised controlled trial showed that depression prevention training added to standard rehabilitation improved work ability over a period of 24 months (Hampel & Neumann, 2022).

Discussion

Return-to-work and work absence were assessed with a wide range of outcome parameters including employment status, work ability, sickness absence, retirement status and duration of employment since rehabilitation. In addition, the studies included in this review investigated a large variety of demographic, work-related, health-related and psychological factors associated with RTW. Due to the range of different RTW outcomes and follow-up intervals, we could not compute any meta-analyses to indicate the strength of these associations.

The findings of the review of reviews investigating RTW outcomes across settings reflects the range of potential factors: Gragnano et al. included 14 reviews and 32 additional cohort studies investigating 58 associated factors (Gragnano et al., 2018). Comparing these results to our findings, only older age, shorter education and depression were common factors associated with RTW outcomes with those factors being barriers for returning to work. However, the range of factors also shows the complexity of issues supporting people to return to work and the need for individualised and tailored support.

Only very few of the associated factors can be considered modifiable: for cancer patients, health literacy (people’s competency to understand and apply health information) and for people with chronic back pain and depressive symptoms, a depression prevention training could be added to inpatient rehabilitation programmes. Perceived work productivity, work satisfaction and work stress could be targets for workplace related interventions, which have shown to be effective for musculoskeletal and mental health disorders (Hoosain et al., 2019; Joyce et al., 2016). Independent of disease groups, the preference to retire was shown to be associated with RTW outcomes (Bernardy & Sandweg, 2003; Mehnert et al., 2017). This finding is reflected in the general population: a population-based study in Sweden demonstrated that the preference to retire early is related to the actual time of retirement (Örestig et al., 2013). However, it is unclear if this preference could be changed as part of an inpatient rehabilitation programme. While health has been found to be a significant contributing factor to early retirement in men, this was not the case in women (von Bonsdorff et al., 2010). In addition, further work-related factors such as higher levels of physically demanding work, lower job satisfaction, less control over their work (i.e. decision authority) and less net income significantly contribute to a stronger preference to retire early (Carr et al., 2016; Pilipiec et al., 2022; von Bonsdorff et al., 2010). Although in most studies gender is significantly associated with the wish to retire early, the relationship seems unclear. Depending on the study, the results showed men or women preferring to retire early (Pilipiec et al., 2022). An intervention to change the preference to retire early should therefore exceed inpatient rehabilitation and include workplace related support. While there is not sufficient evidence of workplace interventions regarding early retirement (Cloostermans et al., 2015), a systematic review meta-analysis indicated a small positive effect on work ability (Oakman et al., 2018).

Strengths and limitations

This systematic scoping review had several strengths and limitations. The review identified a wide range of RTW related outcomes and variables investigating mediating factors demonstrating the need for a consensus-derived set of outcomes (core outcome set) for inpatient rehabilitation studies to facilitate comparisons between studies. In addition, the review revealed a lack of randomised controlled trials evaluating the effectiveness of inpatient rehabilitation. However, the scoping review also used a narrow search for ‘inpatient’, other search terms such as ‘clinic’ or ‘hospital’ could have been added to the terms to complete a full systematic review. Due to time constraints, this was not possible but ‘inpatients’ is the correct MeSH descriptor for this patient group.

Conclusion

Only two modifiable factors associated with RTW that could be addressed in inpatient rehabilitation programmes were identified, health literacy and depression prevention training for those with musculoskeletal problems and depressive symptoms. In addition, gradual reintegration programmes and/or workplace interventions should be further explored as additions to inpatient rehabilitation. Future research should include RCTs based on Medical Research Council guidelines for developing and evaluating complex interventions (Skivington et al., 2021), including process evaluations (Moore et al., 2015) and adhering to reporting guidelines. Given the wide range of RTW outcomes, a core outcome set should be developed to enable comparison and meta-analysis of findings across studies (Kirkham et al., 2017).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by the German Federal Ministry of Labour and Social Affairs [661R0053K1].