Background
Healthcare workers (HCWs), including nurses/midwives and community health volunteers (CHVs), are prone to life challenges affecting the general population and those unique to their work [1]. In addition to uncertainties about life like individuals across other professions, HCWs may be in constant fear of acquiring infections at their workplaces [2, 3] or may experience psychological strains resulting from exposure to traumatic events [4]. In Tanzania, community health volunteers (CHVs) are an essential part of the HCWs workforce recommended by village members to be the connection between the community and the healthcare system [5]. CHVs are supervised by a focal person working at the nearest health centre and are involved in promoting the uptake of essential services such as vaccinations, antenatal services, nutritional programs, and health education initiatives in their respective villages [5, 6].
Resilience is a construct that may promote adapting and bouncing back from the difficult situations HCWs face. The construct describes one’s ability to recover and positively adapt to stressful and challenging events [7]. Resilience is thought to provide a coping mechanism against challenges and enhance a sense of acceptance of conditions and situations that may be non-recoverable, regulate emotions, and promote healthy interpersonal relationships [8, 9]. It is generally thought that lived experiences may play a role in how an individual deals with life events in addition to their innate characteristics [10]. In low and middle-income countries like Tanzania, nurses/midwives and CHVs work in highly demanding environments, often exposed to burnout and stress due to higher workloads resulting from the shortage of staff, lack of adequate equipment, uncooperating clients, poor remuneration, and lack of professional development opportunities [11‐13]. Resilience may be needed to promote adapting to these problems and challenges and ensure the delivery of quality care.
Anzeige
Previous studies have demonstrated the potential beneficial effect of resilience among HCWs across different contexts and settings. For example, high resilience during the COVID-19 pandemic was associated with lower anxiety levels among HCWs in China, the Philippines, and Spain [14‐16]. In addition, high resilience had an inverse association with PTSD, depression, and stress [16‐18]. These positive effects of resilience among HWCs necessitate the use of psychometrically sound and culturally appropriate tools while measuring resilience. The Connor-Davidson Resilience Scale is a widely used resilience measurement scale (CD-RISC) [19, 20]. Originally, the CD-RISC scale consisted of 25 items with a five-factor structure [21]. However, studies across diverse populations reported divergent findings, especially on the factorial structure, leading to the development of the ten-item version (CD-RISC 10) [22]. The CD-RISC 10 has demonstrated adequate validity and reliability properties and a largely one-factor structure in different populations and settings [23‐28].
The context and population to which CD-RISC 10 is applied may influence the scale’s structure [29]. Studies on the psychometric properties of the CD-RISC 10 scale among HCWs in African settings are lacking, and the existing ones among related populations have reported divergent findings on the factorial structure [30, 31]. Based on exploratory and confirmatory factor analysis approaches, one study found a bi-dimensional structure (“toughness” and “motivation”) of the CD-RISC 10 scale among student nurses in Southwestern Nigeria [30]. On the contrary, a one-factor model was described as the best fit for the data among family caregivers of patients with psychiatric disorders in the exploratory factor analysis in the same region of Nigeria [31]. More studies are therefore needed to ascertain the reliability and construct validity of the CD-RISC 10 scale and inform the accuracy of measuring resilience among HCWs in different settings in Africa.
Nurses/midwives and CHVs are an integral part of the healthcare system in Tanzania, where Swahili is the official and primary language. To our knowledge, studies investigating the psychometric properties and the factors associated with the Swahili version of the CD-RISC 10 scale across cadres of HCWs are lacking. This study specifically aimed to determine the internal consistency, construct validity, age, and gender measurement invariance properties of the CD-RISC 10 using the classical test theory. We also assessed the scale’s item characteristics using the item response theory and the factors associated with self-reported resilience as measured by the CD-RISC 10 scale using regression modeling among nurses/midwives and CHVs in Tanzania. Our findings may shed light on the accuracy of measuring resilience, guide the assessment of various interventions aiming to improve resilience and inform training programs among nurses/midwives and CHVs in Tanzania and other Swahili-speaking settings.
Methods
Study design and participants
The study was undertaken as part of a cross-sectional study designed to describe the prevalence and predictors of common mental health problems among HCWs in Tanzania. The study recruited facility-based HCWs (medical doctors, nurses/midwives, pharmaceutical technologists, pharmacists, laboratory technicians, clinical officers, and physiotherapists) and CHVs across the five broad regions of Dar es Salaam, Dodoma, Arusha, Mbeya, and Tabora using a multi-stage stratified sampling approach. The study team worked with the health secretaries to randomly select health facilities in each of the five regions. The proportion of health facilities was as follows: Dar es Salaam (36%), Dodoma (18%), Arusha (17%), Mbeya (15%), and Tabora (14%) according to a master list provided by the health secretaries. As such, the sample size was distributed proportionately to the number of facilities in the five regions. Next, we purposefully selected the health facilities. We considered the regional referral hospitals, district hospitals, and zonal referral hospitals as the major facilities and health centres and dispensaries as the smaller facilities. We selected more of the smaller facilities compared to the major facilities because we anticipated that we would get fewer numbers of healthcare workers in the smaller facilities compared to the major ones. In general, we used the ratio 4:6 (major facility: smaller facility) in a region. In the selected health facilities, individual facility-based healthcare workers were then randomly approached and recruited into the study. The sample size was allocated proportionately by cadre and region. A total of 1444 facility-based HCWs were recruited into the study, out of which 1062 were nurses/midwives.
Anzeige
Regarding CHVs, a sample size of 367 participants was proportionately distributed across the five regions listed above. Although CHVs largely work with the communities, their base of operation is the nearest health centre. A random sample was then selected from each region until the targeted sample size was achieved. For both the facility-based HCWs and CHVs, the inclusion criteria for the study were at least 18 years of age, relevant work experience of > 6 months at the time of data collection, and willingness to participate in the study. There were no restrictions on the level of training. The present analysis included CHVs, and nurses/midwives, given the small sample size across the other cadres of facility-based HCWs. Additionally, the present analysis included participants who chose Swahili as their preferred language for participation in the study (89% of nurses/midwives and 96% of CHVs).
Measures
Sociodemographic measures
We designed and used a sociodemographic tool to collect participants’ information on age, gender, education, marital status, religion, facility type, work experience, health insurance, monthly income, and working hours/day.
Connor-Davidson resilience scale 10 (CD-RISC 10)
We assessed resilience among nurses/midwives and CHVs using the CD-RISC 10 scale. Participants scored each of the ten items on a 5-point Likert scale with the options 0 (not true at all) to 4 (true nearly all the time) concerning their experiences over the last 30 days [22]. The sum score of the ten items ranges from 0 to 40 and indicates a participant’s resilience level, with higher scores indicating higher resilience levels. The scale has an approved Swahili version that is available for use.
Patient health questionnaire – 9 (PHQ-9)
Participants’ depressive symptoms were assessed using PHQ-9 [32]. Participants were asked to score each of the nine items on a Likert Scale ranging from 0 (never) to 3 (almost every day). The sum score for a given participant ranged from 0 to 27, with a high score indicating high depressive symptoms. The scale had acceptable internal consistency among nurses/midwives and CHVs (Cronbach’s alpha = 0.81 for each dataset) in our sample. The scale has been previously validated among nurses/midwives and CHVs in Kenya [33].
Generalised anxiety disorder – 7 (GAD-7)
We used the GAD-7 scale to screen participants for generalised anxiety symptoms [34]. Participants scored the seven items on a Likert scale ranging from 0 (not at all) to 3 (nearly every day). We then summed the scores for all seven items to obtain an anxiety score (ranging from 0 to 21) for all the participants, with higher scores indicating higher generalised anxiety symptoms. Cronbach’s alpha values were 0.84 and 0.85 for nurses/midwives and CHVs, respectively in the present study. The scale has been previously validated among nurses/midwives and CHVs in Kenya [33].
Perceived stress scale-10 (PSS-10)
The level of perceived stress among the study participants was assessed using the PSS-10 scale [35]. Each item is rated on a Likert scale with options 0 (almost never) to 4 (very often), and a sum score (range 0 to 40) is obtained by summing the respective responses for all ten items. Higher scores are an indication of high perceived stress. Some items were reverse-coded. Studies involving medical students and maternity healthcare providers in Kenya have previously used the scale [36, 37]. Cronbach’s alpha values were 0.71 and 0.68 for nurses/midwives and CHVs, respectively, in the present study.
Oldenburg burnout inventory (OLBI)
We assessed burnout symptoms among the study participants using the OLBI [38]. The tool has 16 items with a Likert scale ranging from 1 (strongly agree) to 4 (strongly disagree). Some items were reverse-coded. A sum score in the range of 16–64 denotes a participant’s burnout level, with higher scores indicating high burnout. Cronbach’s alpha values were 0.76 and 0.69 for nurses/midwives and CHVs, respectively, in our sample. The scale has been previously used in a similar setting [33].
The multidimensional scale of perceived social support (MSPSS-12)
The level of perceived social support from family, friends, and significant others was assessed using the MSPSS-12 tool [39]. The tool has 12 items on a 7-point Likert scale (1 = very strongly disagree to 7 = very strongly agree) with a sum score ranging from 12 to 84. A higher sum score indicates higher perceived social support. The scale had acceptable internal consistency among nurses/midwives and CHVs in our sample (Cronbach’s alpha = 0.86 for each dataset).
The World Health Organization – five-item well-being index (WHO-5)
Participants’ mental well-being was assessed using the WHO-5 scale [40]. The tool has five items each on a 5-point Likert scale (0 = at no time to 5 = all the time) and a sum score with the range 0–25. Higher sum scores indicate higher mental well-being. The Swahili version of the scale has been validated among adults living with HIV and epilepsy in rural coastal Kenya [41]. Cronbach’s alpha values were 0.85 and 0.84 for nurses/midwives and CHVs, respectively, in our sample.
Utrecht work engagement scale-9 (UWES-9)
UWES-9 was used to assess the level of engagement among the study participants at work. The tool has nine items with a 7-point Likert scale (0 = never to 6 = every day) and a sum score ranging from 0 to 54 [42]. Higher sum scores indicated high work engagement. The scale has demonstrated good validity and reliability across diverse cadres of HCWs in different settings [43, 44]. Cronbach’s alpha values were 0.70 and 0.68 for nurses/midwives and CHVs, respectively.
Data collection procedures
Potential participants were approached by the study team and details of the study were explained to them including the potential benefits and risks of participating in the study and their right to withdraw from the study at any time. Subsequently, informed consent was obtained from participants before participation in the study. Data was then collected through face-to-face interviews in a private and quiet place to ensure confidentiality, minimise distractions and encourage a candid engagement. Data was collected between August and November 2023 by trained research assistants using the Open Data Kit (ODK) mobile data collection application. All data collection devices were encrypted, and the data server was only accessible to the study’s data manager.
The linguistic and cultural relevance of the CD-RISC 10 scale was ascertained by an independent translator alongside the field team that consisted of local residents. The accuracy, reliability, cultural relevance, and appropriateness of all tools included in the study was ascertained through a pilot test.
Statistical analyses
We conducted all statistical analyses in R statistical software version 4.3.2 [45] in four broad steps: descriptive statistics and comparisons by cadre, classical test theory (CTT), item response theory (IRT), and linear regression. All analyses were separate for nurses/midwives and CHVs except for invariance analyses, where the two datasets were combined. Continuous and categorical variables were summarised using mean (standard deviation (SD)) and frequency (%), respectively. We compared the means of continuous variables by cadre (nurses/midwives vs. CHVs) using the student’s t-test and proportions using the chi-squared test.
For CTT, the association between the psychological and mental health measures was assessed using Pearson’s correlation. Internal consistency was assessed using Cronbach’s alpha (α) and McDonald’s omega (ω), with values > 0.70 considered acceptable [46]. We used confirmatory factor analysis (CFA) to confirm the one-factor structure of the scale among nurses/midwives and CHVs. In addition, we fitted the two-factor model previously described in the literature, with items 1, 2, 3, and 5 loading to the “motivation” factor and items 4, 6, 7, 8, 9, and 10 loading to the “toughness factor [30]. We considered the Root Mean Square Error of Approximation (RMSEA) < 0.08, the Comparative Fit Index (CFI) > 0.95, and the Tucker-Lewis Index (TLI) > 0.95 as indications of a good-fitting model [47]. We then tested the scale’s configural, metric, and scalar invariance across gender and age groups using a multigroup CFA. The configural, metric, and scalar are nested models, and a change in CFI of less than 0.01 in the comparison of the metric versus the configural invariance model and the scalar versus the metric invariance model suggests that a scale is invariant across the groups of interest [48, 49].
The Two-parameter logistics (2PL) graded response models (GRM) were fitted for IRT analyses. The 2PL-GRM models estimate the probability of a participant n endorsing a response category c, or higher in item i with an ordinal Likert scale as shown in Eq. 1 [50]. This probability is a function of the item discrimination (ai), the threshold value for a category c of item i (bic), and a participant’s ability level theta (Θn; in this case, resilience level). The model assumes that the scale has a unidimensional structure.
$$\:P\left({X}_{ni}\ge\:c\right)=\frac{\text{e}\text{x}\text{p}\left({a}_{i}\right({\theta\:}_{n}-{b}_{ic}\left)\right)}{1+\text{e}\text{x}\text{p}\left({a}_{i}\right({\theta\:}_{n}-{b}_{ic}\left)\right)}$$
Item discrimination describes how well an item distinguishes between participants with varying ability levels. A range of 0.8 to 2.5 is considered acceptable for a discrimination parameter [51]. The threshold parameter shows the ability level required for a participant to endorse a response category c or higher with a 0.5 probability [52]. An item with k response categories has k-1 threshold parameters. The 2PL-GRM model also estimates a participant’s ability score (Θn) and maps it on the same scale (z-scale with mean = 0 and standard deviation = 1) with the threshold parameters [53]. Consequently, direct comparisons of Θs and bs can be made. Item characteristic and item information curves can then be plotted to visualise the item thresholds, probability of endorsement, and regions of the ability continuum that specific items and the scale in general measure with high precision.
Multivariable linear regression models were used to assess the factors associated with self-reported resilience (sum score of the items). Sociodemographic variables and psychological measures were included as the explanatory variables in the models. We assessed the linearity and homoscedasticity of residuals assumptions using scatter plots, and the normality of residuals assumptions using Q-Q plots. We also investigated the presence of multicollinearity (high correlation between two or more explanatory variables) using the variance inflation factor (VIF). VIF values < 5 suggested the absence of multicollinearity. We set the significance level at 5%.
Anzeige
Results
Missing data
There were no missing data for the CD-RISC 10 items for the nurses/midwives or the CHVs datasets. Regarding the covariates, missing responses were < 3% of the nurses/midwives and CHVs dataset, as shown in Table 1. We therefore conducted a complete case analysis for all our analyses.
Characteristics of study participants
A total of 950 nurses/midwives and 355 CHVs were included in the analyses. The majority of the nurses/midwives (81.5%) and CHVs (72.4%) were females and worked in a public healthcare facility. Among nurses/midwives, the mean (SD) age was 39.5 (10.2) years, while CHVs averaged a mean (SD) age of 42.1 (11.6) years. Table 1 shows the characteristics of the study participants and the results of comparisons by cadre.
CHVs were on average, 2.7 years older than nurses/midwives (P < 0.001). However, nurses/midwives had on average, 5.1 more years of experience compared to CHVs (P < 0.001). There were statistically significant differences between CHVs and nurses/midwives on gender, education level, marital status, work duration, work hours per day, and having health insurance, but not on health facility management (Table 1).
Table 1
Characteristics of study participants
Characteristic | Nurse/midwives N = 950 | Community health volunteers N = 355 | P-value |
|---|---|---|---|
Age (years), Mean (SD) | 39.49 (10.24) | 42.14 (11.64) | < 0.001 |
Age categories, n (%) | 0.005 | ||
18–30 years, n (%) | 228 (24.0) | 65 (18.3) | |
31–40 years | 304 (32.0) | 95 (26.8) | |
41–50 years | 236 (24.8) | 106 (29.9) | |
>51 years | 182 (19.2) | 89 (25.1) | |
Gender, n (%) | < 0.001 | ||
Female | 774 (81.5) | 257 (72.4) | |
Male | 176 (18.5) | 98 (27.6) | |
Highest education level, n (%) | < 0.001 | ||
Primary | - | 137 (38.6) | |
Secondary | 42 (4.5) | 160 (45.1) | |
Certificate | 215 (22.9) | 35 (9.9) | |
Diploma | 532 (56.7) | 16 (4.5) | |
Bachelor or above | 149 (15.9) | 7 (2.0) | |
Missing | 12 | 0 | |
Marital status, n (%) | < 0.001 | ||
Single | 279 (29.4) | 104 (29.3) | |
Married | 604 (63.6) | 195 (54.9) | |
Divorced/widowed/widower | 67 (7.1) | 56 (15.8) | |
Monthly income (T.sh.), n (%)a | - | ||
0-700,000 | 291 (31.1) | - | |
700,001–1,300,000 | 515 (55.1) | - | |
>1,3,000,000 | 129 (13.8) | - | |
Missing | 15 | - | |
Work duration (years), Mean (SD) | 13.23 (9.86) | 8.13 (5.81) | < 0.001 |
Working hours/day, n (%) | |||
<8 | 39 (4.1) | 236 (66.5) | |
>8 | 911 (95.9) | 119 (33.5) | |
Health facility management, n (%) | 0.725 | ||
Public | 907 (95.6) | 331 (95.1) | |
Private | 42 (4.4) | 17 (4.9) | |
Missing | 1 | 7 | |
Health insurance (yes), n (%) | 842 (88.6) | 93 (26.2) | < 0.001 |
Anzeige
Item characteristics and internal consistency
Table 2 shows the mean (SD) score for each of the ten items on the CD-RISC 10 scale. The item means ranged from 2.43 to 2.81 and 2.10 to 2.74 for nurses/midwives and CHVs, respectively. Cronbach’s α and McDonald’s ω were 0.82 (95% CI: 0.81, 0.84) and 0.82 (95% CI: 0.81, 0.84) for nurses/midwives, and 0.81 (95% CI: 0.78, 0.84) and 0.81 (95% CI: 0.78, 0.84) among CHVs, respectively (Table 2). The results of correlation analyses are shown in Table 3. Additional file 1: Table S1 shows the mean (SD) of the psychological measures included in the study.
Table 2
CD-RISC-10 item characteristics
Item | Nurses/midwives | CHVs | ||
|---|---|---|---|---|
Item score Mean (SD) | Loadings† | Item score Mean (SD) | Loadings† | |
CD-RISC 1 | 2.45 (0.92) | 0.54 | 2.25 (1.02) | 0.53 |
CD-RISC 2 | 2.43 (0.88) | 0.58 | 2.35 (0.96) | 0.65 |
CD-RISC 3 | 2.46 (0.96) | 0.49 | 2.24 (1.03) | 0.52 |
CD-RISC 4 | 2.22 (1.07) | 0.54 | 2.10 (1.13) | 0.46 |
CD-RISC 5 | 2.61 (0.99) | 0.54 | 2.34 (1.05) | 0.50 |
CS-RISC 6 | 2.81 (0.92) | 0.65 | 2.69 (0.97) | 0.57 |
CD-RISC 7 | 2.47 (1.03) | 0.59 | 2.55 (1.01) | 0.61 |
CD-RISC 8 | 2.71 (1.07) | 0.54 | 2.74 (1.08) | 0.63 |
CD-RISC 9 | 2.77 (0.88) | 0.64 | 2.70 (0.98) | 0.57 |
CD-RISC 10 | 2.57 (0.96) | 0.60 | 2.49 (1.10) | 0.46 |
Mean (SD)a | 25.49 (6.08) | 24.44 (6.29) | ||
Table 3
Correlations of resilience with other measures among nurses/midwives and CHVs
1 | 2 | 3 | 4 | 5 | 6 | 7 | |
|---|---|---|---|---|---|---|---|
Nurses/midwives | |||||||
1. Resilience | 1.00 | ||||||
2. Depression | -0.15** | 1.00 | |||||
3. Anxiety | -0.16** | 0.74** | 1.00 | ||||
4. Perceived stress | -0.29** | 0.51** | 0.54** | 1.00 | |||
5. Burnout | -0.27** | 0.41** | 0.42** | 0.41** | 1.00 | ||
6. Work engagement | 0.23** | -0.17** | -0.20** | -0.22** | -0.39** | 1.00 | |
7. Mental well-being | 0.24** | -0.38** | -0.41** | -0.41** | -0.40** | 0.28** | 1.00 |
8. Perceived social support | 0.19** | -0.17** | -0.19** | -0.19** | -0.22** | 0.16** | 0.21** |
Community health volunteers | |||||||
1. Resilience | 1.00 | ||||||
2. Depression | -0.11* | 1.00 | |||||
3. Anxiety | -0.10* | 0.72** | 1.00 | ||||
4. Perceived stress | -0.18** | 0.56** | 0.49** | 1.00 | |||
5. Burnout | -0.25** | 0.40** | 0.39** | 0.31** | 1.00 | ||
6. Work engagement | 0.17** | -0.06 | -0.10 | -0.14** | -0.20** | 1.00 | |
7. Mental well-being | 0.30** | -0.39** | -0.36** | -0.37** | -0.31** | 0.12* | 1.00 |
8. Perceived social support | 0.21** | -0.27** | -0.19** | -0.22** | -0.16* | 0.15* | 0.26** |
Factorial structure
Our data was adequate for factor analysis based on Kaiser-Meyer-Olkin (KMO) values of 0.89 and 0.84 and significant P-values (P < 0.001) from Bartlett’s test of sphericity for the nurses/midwives and the CHVs datasets, respectively. The fit indices of the unidimensional CFA models were as follows: RMSEA = 0.046, CFI = 0.982, and TLI = 0.976 for nurses/midwives and RMSEA = 0.046, CFI = 0.981, and TLI = 0.975 for CHVs, all within the acceptable cutoff thresholds. In addition, the standardised factor loadings for the one-factor model for the nurses/midwives and the CHVs datasets were > 0.45 (Table 2). We found comparable fit indices for the two-factor models (RMSEA = 0.037, CFI = 0.988, and TLI = 0.985 for nurses/midwives and RMSEA = 0.037, CFI = 0.988, and TLI = 0.984 for the CHVs datasets) and a very high correlation (0.83) between the factors, suggesting that the respective two-factor models did not provide additional information relative to the one-factor models Additional files 2 and 3: Figs. S1 and S2).
Anzeige
Measurement invariance
We assessed measurement invariance across gender and age groups using a combined dataset of the nurses/midwives and CHVs. Results from the analyses showed adequate fit for the configural, metric, and scalar invariance models across gender and age and that the unidimensional scale was scalar invariant across gender (RMSEA = 0.042, TLI = 0.981, CFI = 0.981, ∆CFI < 0.001) and age (RMSEA = 0.049, TLI = 0.974, CFI = 0.972, ∆CFI = 0.006; Table 4).
Table 4
Fitness indices for the measurement invariance models
Invariance model | RMSEA | TLI | CFI | ∆CFI |
|---|---|---|---|---|
Gender | ||||
Configural | 0.048 | 0.974 | 0.980 | |
Metric | 0.044 | 0.978 | 0.981 | 0.001 |
Scalar | 0.042 | 0.981 | 0.981 | < 0.001 |
Age | ||||
Configural | 0.039 | 0.984 | 0.987 | |
Metric | 0.046 | 0.977 | 0.978 | 0.009 |
Scalar | 0.049 | 0.974 | 0.972 | 0.006 |
IRT analyses
Discrimination parameters (the slopes of the curves) ranged from 1.14 to 1.89 for nurses/midwives and 1.00 to 1.68 for CHVS, all within the desired range, suggesting that all the items distinguished participants with varying levels of resilience well (Table 5). The threshold parameters (resilience level required for a participant to endorse a given category or higher of an item with a 0.5 probability) progressed orderly from low to high values, indicating a higher probability of endorsing low categories among participants with low resilience and a high probability for higher categories among resilient participants (Table 5).
Figures 1 and 2 are item characteristic curves showing the probability of endorsing a given response category for an item with varying resilience levels. The left side of the curves represents the probability of endorsing an item category for participants with low resilience levels and the right side represents the probability for participants with high resilience levels. The probability of endorsing category 0 (not at all) was lower than the probability of endorsing category 1 (rarely true) or higher even for participants at the lowest point on the resilience continuum for half of the items (items 1, 2, 3, 9, 10) among nurses/midwives and six items (items 1, 2, 5, 7, 9, and 10) among CHVs.
Figure 3 shows the total information curves for the CD-RISC 10 scale. For both datasets, the scale measured resilience with considerably high precision across a wide range of the resilience continuum. However, the increasing standard errors for participants beyond 1.5 SDs above the sample mean suggests that the scale would measure resilience with slightly lower precision for highly resilient participants. Item information curves showed that items provided varying amounts of information, with less information being provided for highly resilient participants across all items. Items 6 and 9 and 8 and 2 provided the most information among the nurses/midwives and CHVs, respectively (Additional file 4: Fig. S3).
Table 5
Item discrimination and threshold parameters for nurses/midwives and CHVs
Item | Sample | a | b1 | b2 | b3 | b4 |
|---|---|---|---|---|---|---|
CD-RISC 1 | Nurses/midwives | 1.24 (0.09) | -4.68 (0.44) | -1.75 (0.13) | 0.10 (0.07) | 1.90 (0.14) |
CHVs | 1.19 (0.15) | -4.57 (0.69) | -1.02 (0.15) | 0.34 (0.12) | 1.99 (0.24) | |
CD-RISC 2 | Nurses/midwives | 1.43 (0.10) | -3.77 (0.29) | -1.69 (0.11) | 0.02 (0.06) | 2.02 (0.13) |
CHVs | 1.63 (0.18) | -3.11(0.34) | -1.20 (0.14) | 0.11 (0.10) | 1.86 (0.19) | |
CD-RISC 3 | Nurses/midwives | 1.14 (0.09) | -3.51 (0.27) | -1.85 (0.14) | -0.09 (0.07) | 2.04 (0.15) |
CHVs | 1.16 (0.14) | -3.12 (0.38) | -1.24 (0.17) | 0.24 (0.12) | 2.31(0.28) | |
CD-RISC 4 | Nurses/midwives | 1.24 (0.09) | -2.51 (0.17) | -1.18 (0.10) | 0.26 (0.07) | 2.17 (0.15) |
CHVs | 1.00 (0.13) | -2.82 (0.37) | -0.91 (0.16) | 0.35 (0.13) | 2.57 (0.33) | |
CD-RISC 5 | Nurses/midwives | 1.36 (0.09) | -3.12 (0.22) | -1.86 (0.12) | -0.38 (0.07) | 1.45 (0.11) |
CHVs | 1.12 (0.14) | -3.36 (0.42) | -1.38 (0.19) | 0.12 (0.12) | 2.00 (0.25) | |
CD-RISC 6 | Nurses/midwives | 1.89 (0.12) | -3.04 (0.20) | -1.86 (0.11) | -0.62 (0.06) | 0.99 (0.07) |
CHVs | 1.46 (0.17) | -3.03 (0.35) | -1.90 (0.20) | -0.52 (0.11) | 1.31(0.16) | |
CD-RISC 7 | Nurses/midwives | 1.51 (0.10) | -2.68 (0.17) | -1.42 (0.10) | -0.17 (0.06) | 1.55 (0.10) |
CHVs | 1.55 (0.18) | -3.19 (0.36) | -1.46 (0.16) | -0.25 (0.10) | 1.41(0.16) | |
CD-RISC 8 | Nurses/midwives | 1.45 (0.10) | -2.87 (0.19) | -1.67 (0.11) | -0.67 (0.07) | 1.07 (0.09) |
CHVs | 1.68 (0.19) | -2.74 (0.29) | -1.54 (0.15) | -0.51(0.10) | 0.84 (0.12) | |
CD-RISC 9 | Nurses/midwives | 1.83 (0.12) | -3.44 (0.25) | -1.95 (0.11) | -0.55 (0.06) | 1.13 (0.08) |
CHVs | 1.43 (0.17) | -3.59 (0.45) | -1.75 (0.19) | -0.53 (0.11) | 1.24 (0.15) | |
CD-RISC 10 | Nurses/midwives | 1.55 (0.10) | -3.35 (0.24) | -1.57 (0.10) | -0.29 (0.06) | 1.48 (0.10) |
CHVs | 1.00 (0.14) | -3.73 (0.51) | -1.61(0.22) | -0.25 (0.13) | 1.72 (0.24) |
Fig. 1
(A). Item characteristic curves for the ten items of the CD-RISC 10 scale among nurses/midwives. Response options are 0 (not true at all; black), 1 (rarely true; red), 2 (sometimes true; green), 3 (often true; blue), and 4 (true nearly all the time; turquoise). The x-axis represents the resilience level (Θ), and the y-axis represents the probability of endorsing a response category
×
![]()
Fig. 2
(B). Item characteristic curves for the ten items of the CD-RISC 10 scale among community health volunteers. Response options are 0 (not true at all; black), 1 (rarely true; red), 2 (sometimes true; green), 3 (often true; blue), and 4 (true nearly all the time; turquoise). The x-axis represents the resilience level (Θ), and the y-axis represents the probability of endorsing a response category
×
![]()
Fig. 3
Total test information curves for CD-RISC 10 for (A). nurses/midwives and (B). community health volunteers. The x-axis represents resilience level (Θ), and the primary y-axis (solid line) represents measurement information. The secondary y-axis (dotted line) represents the standard error of measurement
×
![]()
Multivariable linear regression analyses
Among nurses/midwives, males had higher resilience levels on average than female participants (β = 1.69; 95% CI: 0.68 to 2.70, P = 0.001). A unit increase in perceived stress (β=-0.25; 95% CI: -0.34 to -0.16, P < 0.001) and burnout (β=-0.18; 95% CI: -0.27 to -0.10, P < 0.001) levels was associated with decreasing resilience. Higher work engagement (β = 0.11; 95% CI: 0.05 to 0.17, P < 0.001), mental well-being (β = 0.10; 95% CI: 0.03 to 0.18, P = 0.009), and perceived social support (β = 0.07; 95% CI: 0.02 to 0.11, P = 0.004) scores were associated with increasing resilience levels. We found no significant association between resilience with age, education level, marital status, income, facility management, health insurance, working hours/day, depression, and anxiety in the multivariable model for nurses/midwives (Table 6).
Regarding CHVs, a year increase in work experience was associated with a mean increase in resilience of 0.27 (95% CI: 0.12 to 0.41, P < 0.001). Higher burnout was associated with lower resilience levels (β=-0.26; 95% CI: -0.44 to -0.08, P = 0.004), while higher mental well-being was associated with higher resilience levels (β = 0.32; 95% CI: 0.17 to 0.46, P < 0.001). Age, gender, level of education, marital status, facility management, health insurance, working hours/day, depression, anxiety, perceived stress, work engagement, and perceived social support were not significantly associated with resilience (Table 6). Additional file 5: Table S2 presents results from univariable regression models.
Table 6
Results from multivariable linear regression model among healthcare workers in Tanzania
Characteristic | Nurses/midwives | Community health volunteers | ||
|---|---|---|---|---|
β (95% CI)1 | P-value | β (95% CI)1 | P-value | |
Age (years) | ||||
18–30 | Ref. | Ref. | ||
31–40 | 0.32 (-0.93 to 1.56) | 0.618 | 0.54 (-1.79 to 2.86) | 0.651 |
41–50 | 0.71 (-0.97 to 2.39) | 0.408 | -0.55 (-3.25 to 2.15) | 0.690 |
>51 | 0.42 (-1.78 to 2.62) | 0.708 | 0.14 (-2.82 to 3.11) | 0.923 |
Gender | ||||
Female | Ref. | Ref. | ||
Male | 1.69 (0.68 to 2.70) | 0.001 | -0.05 (-1.71 to 1.61) | 0.952 |
Highest level of education | ||||
Primary | Ref. | |||
Secondary | Ref. | 0.06 (-1.63 to 1.74) | 0.947 | |
Certificate | -1.31 (-3.17 to 0.56) | 0.170 | 0.23 (-2.82 to 3.29) | 0.881 |
Diploma | -0.63 (-2.42 to 1.16) | 0.490 | 0.86 (-2.93 to 4.65) | 0.657 |
Bachelor or above | 1.85 (-0.16 to 3.86) | 0.072 | 3.83 (-1.22 to 8.88) | 0.136 |
Marital status | ||||
Single | Ref. | Ref. | ||
Married | -0.61 (-1.57 to 0.36) | 0.218 | -1.18 (-3.08 to 0.72) | 0.223 |
Divorced/widowed/widower | -0.22 (-2.00 to 1.57) | 0.812 | -1.74 (-4.28 to 0.80) | 0.179 |
Monthly income (T.Sh.) a | ||||
0-700,000 | Ref. | — | ||
700,001–1,300,000 | -0.45 (-1.40 to 0.51) | 0.356 | — | |
>1,3,000,000 | -0.40 (-1.83 to 1.03) | 0.583 | — | |
Work experience (years) | -0.02 (-0.10 to 0.05) | 0.570 | 0.27 (0.12 to 0.41) | < 0.001 |
Health facility management | ||||
Public | Ref. | Ref. | ||
Private | -1.10 (-2.88 to 0.68) | 0.226 | 0.33 (-2.58 to 3.23) | 0.825 |
Health Insurance | ||||
No | Ref. | Ref. | ||
Yes | 1.24 (-0.20 to 2.68) | 0.090 | -1.40 (-3.06 to 0.26) | 0.098 |
Working hours/day | ||||
<8 | Ref. | Ref. | ||
>8 | 0.16 (-1.70 to 2.01) | 0.869 | 1.27 (-0.27 to 2.80) | 0.105 |
Depression | 0.11 (-0.04 to 0.26) | 0.158 | 0.06 (-0.21 to 0.33) | 0.656 |
Anxiety | 0.08 (-0.10 to 0.27) | 0.383 | 0.13 (-0.15 to 0.42) | 0.349 |
Perceived stress | -0.25 (-0.34 to -0.16) | < 0.001 | -0.07 (-0.25 to 0.11) | 0.439 |
Burnout | -0.18 (-0.27 to -0.10) | < 0.001 | -0.26 (-0.44 to -0.08) | 0.004 |
Work engagement | 0.11 (0.05 to 0.17) | < 0.001 | 0.11 (0.00 to 0.22) | 0.058 |
Mental well-being | 0.10 (0.03 to 0.18) | 0.009 | 0.32 (0.17 to 0.46) | < 0.001 |
Perceived social support | 0.07 (0.02 to 0.11) | 0.004 | 0.06 (-0.02 to 0.14) | 0.147 |
Discussion
Summary of findings
This study evaluated the reliability, construct validity, factor structure, gender and age-invariance, item characteristics, and the factors associated with the Swahili CD-RISC 10 scale among nurses/midwives and CHVs in Tanzania. The findings support that the Swahili CD-RISC 10 is a valid and reliable measure of resilience, demonstrated by acceptable internal consistency, construct validity, and measurement invariance by age and gender. In addition, the CD-RISC 10 items distinguished participants with varying levels of resilience well and measured resilience with precision across a wide range on the resilience continuum, evidenced by the acceptable item discrimination and threshold parameters.
In addition, among nurses/midwives, females had lower mean resilience levels than males. Higher perceived stress and burnout levels were associated with lower resilience levels, while higher work engagement, higher perceived social support, and stronger psychological well-being were associated with higher resilience levels. Among CHVs, higher work experience and stronger psychological well-being were associated with higher resilience levels, while higher burnout had an inverse association with resilience levels.
Internal consistency
Findings from the present study suggest that the CD-RISC 10 scale had adequate internal consistency, with Cronbach’s α and McDonald’s ω > 0.80 for both nurses/midwives and CHVs. Our findings were comparable with those reported on the original CD-RISC 10 scale (Cronbach’s α = 0.85) [22], and with those reported across diverse study populations, where Cronbach’s α or McDonald’s ω values ranged between 0.81 and 0.93 [28, 54‐56].
Factorial structure and measurement invariance
In agreement with the original study [22] and a recent study among nurses in Greece [24], results from confirmatory factor analysis indicated that the CD-RISC 10 measured a single construct among nurses/midwives and CHVs in our sample. In addition, acceptable factor loadings (> 0.45) revealed a strong association between the items and the construct.
Results from the multi-group CFA suggested that the CD-RISC 10 scale was gender and age-invariant among adults. Consequently, the results provide evidence that in our population, a comparison of resilience by gender and age is not biased and reflects true differences in the construct. The gender invariance property of the scale has been demonstrated in studies among nurses [24], university students [26, 54], and military personnel [28]. We provide additional evidence that among adults, CD-RISC 10 is age-invariant.
IRT analyses
We found acceptable discrimination values of the CD RISC 10 items and a high measurement precision of the scale over a wide range of the resilience continuum. Our findings on the factorial structure and acceptable item characteristics across the resilience continuum concur with those reported among Vietnamese students using a 2PL-GRM approach [57]. However, the scale may have slightly low precision in measuring resilience among subjects with considerably high resilience levels.
Additionally, we found a lower probability of endorsing category 0 (not true at all) than the probability of endorsing category 1(rarely true) or higher for half of the items, even for participants at the lower side of the resilience continuum. Among the Vietnamese students where a similar modeling approach was used, the probability of selecting category 0 was consistently lower than category 1 for the least resilient participants across all the ten items [57]. These results may imply that there may be a need to re-evaluate the scale’s response categories.
Factors associated with resilience levels among nurses/midwives and CHVs
Sociodemographic factors
We found evidence of lower resilience levels among females compared to male nurses/midwives but not among CHVs. Our findings are consistent with what has been reported in the literature on statistically significantly lower resilience levels among females relative to male hospital nurses in China and Iran [58, 59]. Differences in resilience levels by gender among nurses/midwives may potentially point to cultural dynamics in gender roles. In Tanzania and many African communities, women are largely responsible for caring for children and the elderly and attending to household and familial duties [60]. Consequently, female nurses may be exposed to the double burden of household and professional duties, potentially elevating their burnout and stress levels, which may, in turn, lower their resilience levels.
Gender differences in resilience among nurses/midwives and not CHVs may reflect differences in roles and social support dynamics. CHVs are recommended by village members and primarily work within community settings [5] and may therefore enjoy more social support from community members. The higher social support may buffer against gender disparities in resilience. Further studies, potentially mixed methods, where the gendered aspects of resilience are studied in greater detail, may provide greater clarity on the source of these differences. Nonetheless, it is worth noting that there were statistically significant differences in sociodemographic factors between nurses/midwives and CHVs while comparing these two populations.
In agreement with our findings among CHVs, previous studies have also reported a positive association between years of experience and resilience [58, 61, 62]. A possible explanation would be that as work experience increases, individuals learn various ways of adapting and dealing with challenges. Nonetheless, similar to other studies [63, 64], we did not find a statistically significant association between work experience and resilience among nurses/midwives, perhaps suggesting differences in resilience dynamics across different cadres of HCWs and work settings. For instance, across Tanzania, CHVs work on a volunteership or semi-volunteership basis, primarily driven by the urge to serve their communities [11]. Over time, CHVs may become satisfied and more fulfilled with their impact on communities, potentially gaining coping mechanisms to stresses and challenges. Although nurses/midwives may also be fulfilled by their impact on individuals, their motivation and job satisfaction may also be dependent on job factors such as promotion, compensation, and opportunity for further training and education [65]. In working environments where these job factors are not greatly emphasised, work experience may not necessarily influence resilience levels.
Psychological factors
Perceived stress and burnout were associated with lower resilience levels, which is consistent with findings from previous studies among diverse populations of HCWs [20, 58, 59, 66]. We hypothesise that stress and burnout resulting from prolonged heavy workloads, inadequate staffing, and lack of adequate equipment in Tanzania, especially among nurses/midwives [13] maybe associated with decreased resilience levels by straining interpersonal relationships, impairing interactions with colleagues, and social support-seeking patterns and behaviours. Intervention should focus on stress management programs such as relaxation exercises, providing long-lasting solutions to systemic issues of staffing and workload, and resilience-building programs such as team-building activities and mentorship.
Our findings confirm those of previous studies on the positive association between social support, work engagement [67‐69], and mental and psychological well-being [70]. Social support from supervisors, colleagues, and family may promote resilience by encouraging the sharing of experiences and coping strategies [71]. Work engagement, which describes an individual’s involvement in their work, may promote a sense of purpose and belonging and encourage a positive attitude while facing challenges and adversities at work [72]. Indeed, engaged workers tend to mobilise and utilise personal and external job resources, such as seeking help from colleagues to achieve their work objectives [73]. In addition, resilience and the willingness to face and deal with difficult life situations and challenges may be elevated by some aspects of mental and psychological well-being, such as optimism and self-esteem [74]. It is worth highlighting that the association between psychological measures included in the present study may be bi-directional [75, 76]. Therefore, interventional programs may focus on specific or collective psychological aspects to improve resilience among HCWs.
Limitations
Despite being considerably large and the first study to investigate the psychometric properties and factors associated with the CD-RISC 10 resilience scale among HCWs in Tanzania, findings from the present study should be interpreted with caution. The cross-sectional study design limited causal inference, suggesting that the findings of the study should only be interpreted as associations. Additionally, our findings should be interpreted with the one-time-point limitation in mind. As such, our findings may not reflect a stable state of association.
Resilience and the other psychological measures used in the study were self-reported and were therefore prone to social-desirability bias, with the potential to distort the reported associations between variables. Additionally, although the present study recruited a generalisable sample across Tanzania, our findings may not necessarily be generalised to other Swahili-speaking countries, as the construct validity of the scale may depend on culture and setting.
Despite these limitations, the study was a first step in understanding the validity, factorial structure of the CD-RISC 10 scale, and the factors associated with self-reported resilience in the absence of any previous studies in Tanzania. Future studies should incorporate both qualitative and longitudinal quantitative approaches to better understand the dynamics of resilience among HCWs. A qualitative approach may enhance understanding of the resilience process and the mechanism of association with predictors of interest and the triangulation of results from self-reported measures while a longitudinal design may shed light on the impact of the predictors over time. Additionally, more studies may be needed to assess whether some psychological measures may have mediating roles in the association between other psychological measures and resilience.
Conclusion
Findings from the present study suggest that preliminarily, the Swahili version of the CD-RISC 10 scale is a valid and reliable measure of resilience among nurses/midwives and community health volunteers in Tanzania. As such, CD-RISC 10 may be used to accurately measure resilience and in the assessment of the effectiveness of interventions designed to improve resilience among HCWs in Tanzania. However, consideration should be made to re-evaluate the scale’s response categories.
Different socio-demographic and psychological measures may have different effects among different cadres of HCWs. As such, supervisors of HCWs should take note of subgroups likely to be less resilient and develop appropriate training modules. Additionally, training programs should incorporate different strategies to target different psychological predictors of resilience.
Acknowledgements
We would like to thank all participants for taking part in the study, as well as the data collection team and research assistants.
Declarations
Ethical approval and consent to participate
Ethics approval for the study was granted by the Ethics Review Committee of The Aga Khan University, Dar es Salaam, Tanzania, and the National Institute for Medical Research, Tanzania. Permission to conduct the study was also obtained from the President’s Office, Regional Administration, and Local Government, Tanzania. Written informed consent was obtained from all participants before they participated in the study, in accordance with the declaration of Helsinki.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
