Factors associated with anxiety in family caregivers of children with chronic diseases

In 2018, a cross-sectional study was conducted involving 446 family caregivers of children with chronic diseases hospitalized at the Federico Gómez Children’s Hospital of Mexico, National Institute of Health. Among the participants, 83% identified as female and 17% as male, and the average age was 32.23 years (SD = 8.65). The inclusion criteria for the present study required that the participants were over 18 years of age, were providing care for a child with a chronic disease who was hospitalized at the National Institute of Health, and read and signed an informed consent form prior to enrollment in the study. The study design was nonexperimental, cross-sectional, and ex post facto, with nonprobabilistic sampling [43].

Participants responded in writing to A Sociodemographic Variables Questionnaire (Q-SV) that was developed for research on family caregivers of children with chronic diseases [24]. This instrument is composed of 20 items that measure individual, familial, and caregiver factors, such as age, sex, and marital status. In addition, this instrument collects information on the child’s sex, age, diagnosis, and time of hospitalization, among other information. Additionally, they completed a battery of 18 self-reported instruments. Family caregiver anxiety was assessed using the Beck Anxiety Inventory (BAI). Detailed information regarding these instruments can be found in Table 1.

This study was approved by the Research, Ethics, and Biosafety Commissions at the Hospital Infantil de México Federico Gómez National Institute of Health in Mexico City. We adhered to the ethics rules and considerations for research with humans currently in place in Mexico [44] as well as those outlined by the American Psychological Association [45]. Participation was voluntary, and all participants were informed of their rights under the Declaration of Helsinki prior to enrollment [46]. All family caregivers were informed of the objectives and scope of the investigation and of their research rights. Caregivers who agreed to participate in the study signed a consent letter. Consenting caregivers were given instructions and completed the questionnaires independently at their child’s hospital, and the battery of tests was administered individually.

The research team, which was composed of psychologists and social workers employed by the hospital, visited the waiting rooms of the different hospitalization services and spoke with potential participants. Then, the researchers met with each individual and provided information about the study, informed participants of their research rights, and gave them the informed consent form. Finally, the individuals who agreed to participate were given instructions on how to respond to the battery of self-reported instruments. The researchers took precautions to ensure that the participants understood that their participation was voluntary and that their responses would be kept confidential and anonymous. Each participant completed the sociodemographic questionnaire and battery of 18 instruments during one 30- to 50-min session. The data collection stage lasted approximately three months.

The analyses for this study were conducted using SPSS (v.24, IBM Inc., USA) to describe the sample, frequencies, percentages (%), arithmetic means (M), and standard deviations (SD). Descriptive analyses (Ms, SDs, and ranges) were used to describe the distributions for each instrument included in this study. Pearson’s product-moment correlation coefficient (r) was used to investigate the associations between anxiety and the psychosocial, sociodemographic, and clinical variables (potential predictors). Point-biserial correlation was performed with dichotomous variables. The significance of the correlations was tested using Fisher’s Z transformation. A linear regression model to predict family caregiver anxiety was calculated with significant correlates using the backward-elimination method. The overall significance of the model was tested with the F-test, and the significance of each predictor was assessed using a t-test. The explanatory strength of the model was determined with the squared multiple correlation coefficient (R²). The multicollinearity of each predictor was evaluated based on the variance inflation factor (VIF). The residual analysis included the Durbin-Watson test to confirm the lack of autocorrelation between consecutive residuals (in order of sampling). Cook’s distance was calculated to identify influential values, and a histogram and Fisher’s skewness and kurtosis coefficients were used to assess whether the residuals were normally distributed. The distribution of residuals was considered normal if the histogram presented a bell-shaped curve and the symmetry and kurtosis coefficients were between − 0.5 and 0.5. The distribution was considered to be approaching normal if the histogram showed a bell-shaped curve, the coefficient of asymmetry was less than 1, and the kurtosis was less than 2 [47]. Following Cohen’s cut-off points, the effect size or strength of the association was interpreted as trivial when the standardized regression weights (β) or correlation coefficients (r) had values between − 0.099 and 0.099, low when between − 0.299 and − 0.10 or between 0.10 and 0.299, medium when between − 0.499 and − 0.30 or between 0.30 and 0.499, high when between − 0.699 and − 0.50 or between 0.50 and 0.699, very high when between − 0.899 and − 0.70 or between 0.70 and 0.899, and unitary when between − 1 and − 0.90 or between 0.90 and 1 [48].

The descriptive analysis results for the sociodemographic characteristics of the family caregivers, pediatric patients, and two of the three clinical variables are presented in Table 2. Most family caregivers were women (87%), married (65.5%), housewives (65.5%), and mothers of the patient (77.1%) in addition to having a secondary education (63%) and school-age children (59.2%).

The mean time of hospitalization was 1.51 months (95% CI: [1.40–1.62]), and the average time since diagnosis was 3.5 months (95% CI: [3.31–3.69]). Regarding the diagnoses, 330 of the 446 the children (74%) suffered cancer, 31 (7%) had a persistently patent arterial duct, 21 (4.7%) had nephrotic syndrome, 18 (4%) had chronic terminal renal insufficiency, 13 (2.9%) had asthma, 12 (2.7%) had tricuspid atresia, 9 (2%) had Down syndrome, 4 (0.9%) had tetralogy of Fallot, 3 (0.7%) had HIV/AIDS, 3 (0.7%) had a liver or kidney transplant, and 2 (0.4%) had cystic fibrosis.

The average anxiety score among the family caregivers was 14.48 (95% CI: [13.29–15.67], range 0–63, SD = 12.74), with no significant difference between men and women.

Fifteen of the 16 psychosocial variables correlated with caregiver anxiety, and the strength of these associations ranged from low (r = 0.102, p = 0.031) to high (r = 0.526, p <  0.001), with a mean (in absolute values) of 0.266 (low). The correlation between the variable self-efficacy beliefs/optimism and caregiver anxiety was not significant. The means, standard deviations, ranges, and r values are presented in Table 3.

Anxiety was independent of the four sociodemographic variables of the caregivers (sex, age, education, and family income) as well as the sex (rbp = − 0.038, 95% CI: [− 0.130, 0.055], p = 0.425) and age of the children (r = − 0.076, 95% CI: [− 0.168, 0.017], p = 0.108).

Regarding the clinical variables, the child’s hospitalization period had a significant correlation with caregiver anxiety, but its strength of association was trivial (r = 0.097, 95% CI: [0.004, 0.188], p ≤ 0.05). The correlation between the time since diagnosis and caregiver anxiety was not significant (r = 0.018, 95% CI: [− 0.075, 0.111], p = 0.705). After eliminating four diagnoses with fewer than five cases, the comparison of means in caregiver anxiety among the remaining seven diagnostic categories was not significant (F[6, 427] = 1.127, p = 0.270, η² = 0.018; assuming equality of variances by Levene’s test: F[6, 427] = 1.583, p = 0.150). The correlation was also not significant when dichotomizing the variable “diagnosis” for cancer (0) and another chronic disease (1) (rbp = 0.014, 95% CI: [− 0.079, 0.106], p = 0.776).

The 16 correlating variables (15 psychosocial variables and one clinical variable) were introduced as potential predictors of caregiver anxiety. The residual analysis of this first model did not satisfy the normality assumption. The skewness of the residuals was greater than 1, and the coefficient of kurtosis was greater than 2. Consequently, the response variable data were transformed by adding one unit and subsequently obtaining the natural logarithm (Ln [BAI total score + 1]). The analysis of variance showed the overall significance of the new regression model (F [5, 439] = 58.90, p <  0.001). This new model was composed of five psychosocial variables as significant predictors (p-values lower than 0.05 using a two-tailed t-test). Depression, caregiver burden, emotional well-being, self-esteem, and a negative coping style accounted for 39.5% of the variance in caregiver anxiety. The effect size of the predictors varied from trivial (β = 0.09) to medium (β = 0.31). There was no evidence of multicollinearity among variables, because the VIF values ranged from 1.04 to 1.42 (Table 4). Cook’s distance also did not reveal any problematic data, since the values were all below the cut-off of 1.0. When sorting the data based on their sampling sequence, the Durbin-Watson test revealed that the residuals were independent (DW = 2.08). The scatter diagram of residuals, which was plotted with the standardized predicted values on the x-axis and the standardized residuals on the y-axis, showed homogeneity in the residual variance. The histogram of the residuals had a bell-shaped curve, and the distribution was mesokurtic (K = 0.275, 95% CI: -0.178, 0.728) but was slightly skewed to the left (Sk = − 0.529, 95% CI: -0.756, − 0.302); therefore, the distribution was approaching normal.