BACKGROUND

Work ability has been defined as the ability of a person, or rather a perception of their ability, to meet the demands of their job (Ilmarinen, 2009). Inherent in this definition is the notion that work ability is not only a function of one's personal capacities, including physical, mental, and social/interpersonal abilities, but also job requirements (Cadiz et al., 2019). That is why work ability varies within the same individual depending on their position or working conditions.

Extended evidence shows that work ability is a prognostic factor in absence from work (Reeuwijk et al., 2015), including long-term absence (Török et al., 2020), risk pension status (Roelen et al., 2014), early retirement intentions (Pit & Hansen, 2014), and outcomes beyond their working life, such as disability and death in later life (von Bonsdorff et al., 2011). For these reasons, the promotion of excellent work ability has been considered a key factor in prolonging a productive working life (Lindberg et al., 2006). Since the 1980s, researchers have focused on the evaluation of work ability, and its study has progressively gained prominence in the decades since, conditioned by demographic transitions, changes in work processes, and the incursion of new technologies and changes in labor relations (Ilmarinen, 2005). These changes have had a major impact on the nursing profession since nurses work in increasingly complex and challenging contexts, with a higher demand for care of people with multimorbidity and advanced age, leading to increased physical and mental demands made on these professionals (Catton, 2020). Additionally, hospital nurses have been reported to have especially high psychological and physical job demands (Jalilian et al., 2019).

To measure work ability, the Finnish Institute of Occupational Health developed the Work Ability Index (WAI) (Tuomi et al., 1988). The WAI is the most widely accepted and commonly used instrument to measure work ability and is available in 25 languages (van den Berg et al., 2009; World Health Organization, 2012). It is composed of 60 items distributed in seven dimensions that assess (a) current work ability compared with the lifetime best (1 item), (b) work ability in relation to the demands of the job (2 weighted items), (c) number of current diseases diagnosed by a physician (out of a list of 51 diseases), (d) estimated work impairment due to disease (1 item), (e) sick leave during the past 12 months (1 item), (f) own prognosis of work ability 2 years from now (1 item), and (g) mental resources (3 items) (Tuomi et al., 1997). Scores on each dimension are added together, with a range from 7 to 49, which allow workers to be classified into four categories: poor (7–27); moderate (28–36); good (37–43); and excellent work ability (44–49). These categories were derived from the 15th and 85th percentiles of the scores obtained for a population of Finish municipal employees in 1981, and the resulting cutoffs have remained unchanged since that time (Ebener & Hasselhorn, 2019). However, since then, numerous studies have dichotomized the variable by merging the categories poor and moderate into inadequate work ability (7–36) and good and excellent adequate work ability (37–49). This could be due to statistical reasons, but also because both poor and moderate categories are the ones in which the worker has already had an imbalance between individual resources and the demands of the job, and therefore interventions to restore or improve work ability are needed.

Until now, several studies around the world have used the WAI to determine work ability among nurses; however, the prevalence rates of inadequate work ability, including poor and moderate work ability, among hospital nurses vary considerably between studies. In this sense, knowing an approximation of the global prevalence of inadequate work ability among hospital nursing personnel around the world helps to estimate the magnitude of the problem and determine the need for corrective measures. Therefore, the objective of this study is to develop a systematic review and meta-analysis to estimate the worldwide pooled prevalence of inadequate work ability among hospital nursing personnel measured by means of the WAI.

METHODS

A meta-analytic study was conducted following the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 statement (Page et al., 2021). The systematic review protocol for this study was registered with the International Prospective Register of Systematic Reviews (PROSPERO) on 26 July 2021 (pending registration number).

Eligibility criteria Studies were included in accordance with the following criteria: Study design: primary observational studies with a cross-sectional or prospective research design. Studies aimed at psychometric testing of the WAI were excluded. Prevalence data: studies that have data on the prevalence of inadequate work ability or data on which prevalence could be calculated. In studies in which such data were not found, the authors were contacted and asked to provide them. Only data from those who responded within 5 days of contact were included. Instrument: studies using the complete form of the WAI to measure work ability were included, excluding those using shortened or partial versions of the WAI (individual dimension/s or items). Population: the study population was nursing personnel who worked in hospitals around the world. Nursing personnel encompassed nurses and Nursing Assistive Personnel. Nursing assistive personnel included all categories of unlicensed personnel who are accountable to and work under the direct supervision of a nurse to implement specifically delegated patient care activities (Association of Women's Health, 2009). Studies conducted on midwives were excluded. Studies with aggregate data from hospital nurses with other categories of hospital professionals (physicians) and with nursing personnel working in other clinical settings (primary care) were also excluded unless data from hospital nursing personnel could be extracted independently.

Reports published in English, Spanish, Portuguese, German, Italian, Croatian, and Greek were included. Only four studies were excluded due to publication languages, three in Chinese and one in Persian. No restrictions on the publication date were applied. Both original peer-reviewed articles and gray literature (conference proceedings, conference abstract, and dissertations) were included.

Information sources and search strategy

A systematic search was performed to identify relevant reports in the following databases: Web of Science (WOS), Medline/PubMed (WOS interface), CINAHL, Scopus, Lilacs, PsycINFO, ProQuest Nursing and Allied Health, and Scielo (WOS interface). A partial gray literature search was also performed on Google Scholar, limited to the first 50 most relevant reports retrieved. The search strategy was developed by the research team and then peer reviewed by a researcher with experience in bibliographic searches outside the project. (“work ability index”) AND (nurse*) AND (hospital*) was the general search strategy used, adapted to the syntax and subject headings of the databases. The searches were performed separately by two experienced researchers in the literature search in all databases where no limits were imposed. More detailed information on search strategies is provided in Table 1.

TABLE 1. Search strategy Database Search strategy Results Web of Science (WOS) ([“work ability index”] OR AB = [“work ability index”] OR AK = [“work ability index”]) AND ([nurs*] OR AB = [nurs*] OR AK = [nurs*]) AND ([hospital*] OR AB = [hospital*] OR AK = [hospital*]) 54 MEDLINE/Pubmed ((“work ability index” [Tittle/Abstract]) OR (“work ability index” [Title]) OR (“work ability index” [Other Term])) AND ((nurs*[Tittle/Abstract]) OR (nurs*[Other Term])) AND ((hospital*[Tittle/Abstract]) OR (hospital*[Title])) OR (hospital*[Other Term])) 49 CINAHL (“work ability index” OR AB “work ability index” OR SU “work ability index”) AND (nurs* OR AB nurs* OR SU nurs*) AND (hospital* OR AB hospital* OR SU hospital*) 22 Scopus (TITLE-ABS-KEY [“work ability index”]) AND (TITLE-ABS-KEY[nurs*]) AND (TITLE-ABS-KEY [hospital*]) 65 Lilacs (“work ability index”) AND (nurs*) AND (hospital*) 22 PsycINFO (ab[“work ability index”] OR ti[“work ability index”]) AND (ab[nurs*] OR ti[nurs*]) AND (ab[Hospital*] OR ti[Hospital*]) 13 ProQuest nursing and allied health (ab[“work ability index”] OR ti[“work ability index”]) AND (ab[nurs*] OR ti[nurs*]) AND (ab[Hospital*] OR ti[Hospital*]) 10 Scielo (“work ability index”) AND (nurs*) AND (hospital*) 14 Google scholar “work ability index” AND nurses AND hospital -

To address the saturation of the literature, reference lists of the selected literature were also manually searched to obtain additional relevant studies. Additionally, some articles that were not recovered from the database search but were previously known to the research team to be relevant to this review were included.

Selection process

To ensure the reliability of the study selection process, it was carried out by two members of the review team independently. The selection process started with screening titles and abstracts, followed by full reading for the initially selected studies. The full texts of eligible studies were then retrieved as far as possible to corroborate this. To resolve any disagreement, a third member of the review team was consulted. Discarded studies were classified according to the reason for exclusion. Finally, the same members of the research team conducted an inverse search, searching potentially eligible items in the reference lists of included studies, repeating the previous process. See the PRISMA flow diagram (Page et al., 2021), Diagram 1.

Flow diagram

Data collection process

The following data were extracted: Authors, year of publication, origin country of the sample, type of publication (journal article, conference proceedings or dissertation), study design (cross-sectional vs. longitudinal), sampling method (convenience vs. random), sample size (N), sampling fraction (%), professional category (nurses vs. nurses/Nursing Assistive Personnel), hospital unit, age in years (mean SD), proportion of women (n—%), prevalence of poor work ability (n—%), prevalence of moderate work ability (n—%), and prevalence of inadequate work ability (n—%). Data were extracted in duplicate by two researchers, and discrepancies were resolved by consensus between the research team.

The sample size and the crude prevalence rate of poor and moderate work ability among nursing personnel were collected from the results of the studies. These categories were determined by the WAI cutoff points according to its instructions (Tuomi et al., 1988), since poor work ability is represented by a WAI score in the range of 7–27 and moderate by a WAI score in the range of 28–36. Inadequate work ability is a merged category that combines the two previous categories (low and moderate) and therefore is represented by a WAI score <37. Data from studies using a different criterion for inadequate work ability than the one described were coded separately. If the crude prevalence rate was not reported directly, two investigators independently performed the appropriate calculations based on the data provided by the study results. The corresponding authors of the studies in which prevalence data were not reported were contacted to request data. In the case of longitudinal studies, baseline prevalence data were extracted for analysis.

Risk of bias in individual studies

The instrument proposed in the critical appraisal guide of observational studies in epidemiology (Ciapponi, 2010) was used to assess the methodological quality of the included studies. All were independently reviewed by two researchers to ensure the internal validity of the studies in three dimensions of the instrument, which consisted of 13 items: participants (items 2–6), definition and measurement of key variables (items 11–14), and statistical analysis and confounder variables (items 15–18). The degree of compliance statement represented by each item was first evaluated in five categories (“very good, good, fair, poor, or not informative”). The quality of the study can be considered high if most statements are answered as “very good” or “good” (Ciapponi, 2010). Therefore, for ease of reporting, a score was assigned to each individual study. For this purpose, a point was assigned for each item whose compliance with the statement it represents was rated as "very good" or "good," and no points were assigned in all other cases. These ratings were not used as a criterion for study eligibility.

Data analysis

The Freeman–Tukey double arcsine method (Miller, 1978) was used to calculate the prevalence rate with 95% confidence intervals (CI) for each individual study. Then, an inverse-variance-weighted random-effects meta-analysis was performed by conventional methods (DerSimonian & Laird, 1986). A random effects model is recommended for the meta-analysis of prevalence when heterogeneity is observed in prevalence estimates between studies, as in the case of this study, as a fixed effects model is likely to produce misleading results in the presence of significant heterogeneity (Wang & Liu, 2016). Three independent meta-analyses were performed to calculate the pooled prevalence of poor, moderate, and inadequate work ability among hospital nursing personnel. Additionally, subgroup analyses were performed defined by the mean age range (under 40 vs. over 40), professional categories (nurses vs. nurses/Nursing Assistive Personnel), and origin country of the sample (Brazil vs. the rest of the world) to control for a possible confounding factor, due to the fact that half of the studies included and, consequently, a large part of the sample (20%) came from a single country, Brazil. No other subgroup analyses could be performed due to the small number of studies in some categories (e.g., type of publication, study design, and sampling method) and the homogeneity of studies with regard to other characteristics (e.g., proportion of women, since only one study did not exceed 70% of women in the total sample).

Heterogeneity between individual studies was evaluated using Cochran Q test statistics, considering that heterogeneity was present if the p-value is less than 0.10 (10%) (Hoaglin, 2016). The degree of inconsistency between studies was measured using Higgins’ I2 test statistic in a meta-analysis, with values of 25%, 50%, and 75% considered low, moderate, and high inconsistency, respectively (Higgins et al., 2003). The consistency of the results was assessed by performing sensitivity analyses by excluding any study one by one and seeing that the results did not change substantially. Publication bias was assessed using Egger's linear regression, with a p-value less than 0.05 implicating publication bias (Egger et al., 1997). StatsDirect software was used for all statistical analyses described previously.

RESULTS Study selection

The selection process is reported in detail in the PRIMA flow diagram (Page et al., 2021) (Diagram 1). A total of 267 reports were initially found: 259 in electronic databases and 8 through other methods. After removing duplicates from the database search result, the titles and abstracts of 104 studies were selected, and 86 were considered potentially relevant studies for full-text reading. The eight additional records identified by other methods were also considered. Five articles could not be retrieved in full text despite the efforts made to retrieve them by various means. A total of 47 articles were excluded for the nominated reasons specified in Diagram 1: 44 from databases and 3 retrieved using alternative methods. Finally, 42 studies (Akodu & Ashalejo, 2019; Capelo et al., 2012; Carel et al., 2013; Das et al., 2019; Duran & Cocco, 2004; Ehegartner et al., 2020; Fischer & Martinez, 2013; Fonseca, 2012; Garosi et al., 2018; Golubic et al., 2009; Habibi et al., 2012; Hilleshein et al., 2011; Hoe et al., 2011; Izu et al., 2016; Knežević et al., 2010; Magnago et al., 2015; Maia et al., 2014; Fischer et al., 2006; Martinez et al., 2017; Martins, 2002; Melikidou & Sourtzi, 2014; Murassaki et al., 2013; Nery et al., 2013; Nowrouzi et al., 2015; Nunes et al., 2013; Oliveira, 2016; Pereira et al., 2021; Quispe Carbajal, 2021; Raffone & Hennington, 2005; Rodrigues et al., 2019; Rongen et al., 2014; Rostamabadi et al., 2017; Rotenberg et al., 2008, 2009; Rypicz et al., 2021; Silva et al., 2016, 2018, 2019; Sopajareeya, 2020; Sorić et al., 2013; Vasconcelos et al., 2011; Vilela et al., 2013) were included in the systematic review and 35 of them in the meta-analyses performed.

Study characteristics

The characteristics of the 42 studies included are reported in Table 2. These comprised 24,728 subjects with a mean age of 38.4 years and an average of 84.6% of women among the studies that reported these sociodemographic data. In all, 13 studies were carried out on nursing staff from various hospital units (31.0%), four in the ICU (9.5%), two in the emergency department (4.8%), and one in obstetric care (2.4%) with no specific origin reported in the remaining studies. The studies were developed in 14 different countries, of which 20 (47.6%) were developed in Brazil and 5 (11.9%) in Portugal. One study was multinational (Rongen et al., 2014), conducted in several European countries, and is the one with the largest sample, 9927 participants, while the study with the smallest sample size had 24 participants (Nery et al., 2013). 97.6% of the studies were cross-sectional, and 88.1% used convenience sampling. Regarding the type of publication, 32 studies corresponded to journal articles (76.2%) and the rest to gray literature, with 6 conference proceedings (14.3%) and 4 dissertations (9.5%).

TABLE 2. Characteristics of the studies Author(s), year Country of the origin of sample Type of publication Study design type Sampling method Sample size (N) Sampling fraction (%) Professional categories Hospital unit Age in years (mean ± SD) Proportion of women n (%) Poor work ability (WAI score 7–27) n (%) Moderate work ability (WAI score 28–36) n (%) Inadequate work ability (WAI score <37) n (%) Inadequate work ability (other criteria) Quality score Akodu and Ashalejo (2019) Nigeria Journal article Cross-sectional Convenience 135 94.4% Nurses NR 40.2 ± 10.5 126 (93.3%) 8 (5.9%) 31 (23.0%) 39 (28.9%) – 10/13 Capelo et al. (2012) Portugal Conference proceedings Cross-sectional Convenience 78 70.9% Nurses-NAP NR 33.1 ± 9.6 62 (79.5%) 1 (1.3%) 16 (21.0%) 17 (23.2%) – 9/13 Carel et al. (2013) Israel Journal article Cross-sectional Convenience 515 NR Nurses Multispecialty 41.1 ± 9.8 460 (89.0%) 5 (1%) 0 (0.0%) 5 (1%) – 10/13 Da Silva et al. (2016) Brazil Journal article Cross-sectional Convenience 100 85.5% Nurses-NAP Multispecialty 39.4 ± 9.5 88 (88.0%) – – – 35 (35%) 12/13 Das et al. (2019) Bangladesh Journal article Cross-sectional Random 197 NR Nurses Multispecialty 35.9 ± 8.0 187 (94.9%) NR NR 10 (7.1%) – 13/13 Duran and Cocco (2004) Brazil Journal article Cross-sectional Convenience 54 NR Nurses-NAP ED 37.3 ± NR 40 (74.1%) 0 (0.0%) 7 (13.2%) 7 (13.2%) – 10/13 Ehegartner et al. (2020) Germany Journal article Cross-sectional Random 382 NR Nurses NR 40.1 ± 12.0 309 (81.0%) 67 (17.5%) 148 (38.7%) 215 (56.2%) – 11/13 Fischer and Martinez (2013) Brazil Journal article Cross-sectional Convenience 514 83.8% Nurses-NAP Multispecialty 35.5 ± 8.1 397 (77.2%) NR NR 58 (11.3%) – 12/13 Fischer et al. (2006) Brazil Journal article Cross-sectional Convenience 696 69.9% Nurses-NAP NR 34.9 ± 10.5 611 (87.8%) NR NR 159 (22.8%) – 12/13 Fonseca (2012) Portugal Dissertation Cross-sectional Convenience 159 NR Nurses-NAP Multispecialty 36.5 ± 9.9 132 (83.0%) 9 (5.7%) 33 (20.8%) 42 (26.4%) – 12/13 Garosi et al. (2018) Iran Journal article Cross-sectional Convenience 101 NR Nurses ICU 24.5 ± 3.6 71 (70.0%) 6 (5.9%) 27 (26.7%) 33 (32.7%) – 11/13 Golubic et al. (2009) Croatia Journal article Cross-sectional Convenience 1086 78.0% Nurses Multispecialty 38.7 ± 10.3 1010 (93.0%) NR NR 380 (35.0%) – 12/13 Habibi et al. (2012) Iran Journal article Cross-sectional Random 228 NR Nurses-NAP Multispecialty 38.4 ± 8.5 0 (0.0%) NR NR 63 (27.6%) – 10/13 Hilleshein et al. (2011) Brazil Journal article Cross-sectional Convenience 93 NR Nurses NR 41.7 ± 9.0 92 (98.9%) 0 (0.0%) 14 (15.1%) 14 (15.1%) – 9/13 Hoe et al. (2011) Australia Conference proceedings Longitudinal Convenience 768 69.1% Nurses NR NR NR 9 (1.2%) 56 (7.3%) 65 (8.5%) – 12/13 Izu et al. (2016) Brazil Journal article Cross-sectional Convenience 144 87.3% Nurses-NAP NR 46.4 ± 8.5 127 (88.2%) 5 (3.5%) 30 (20.8%) 35 (24.3%) – 11/13 Knežević et al. (2010) Croatia Journal article Cross-sectional Convenience 1210 NR Nurses-NAP NR NR NR 67 (5.5%) 357 (29.5%) 424 (35.0%) – 13/13 Magnago et al. (2015)