This is a quantitative non-experimental, cross-sectional research with a correlational scope and a survey-type data collection model (Hernández et al., 2014).

Non-random sample of 325 older adults with a medical diagnosis of a chronic disease of at least 1 year with a mean age and standard deviation of 68.39 and 4.15 years (CV = 6.07%), respectively. Residents of the departments of Cesar and Atlántico, participating voluntarily and provided informed consent. Subjects had no cognitive impairment (Mini-mental State Examination [MMSE] scores of >23) and had reading and writing ability so that they could understand the tests.

The conditions of the population of interest made it impossible to obtain a sampling frame, forcing the researchers to select the participants through a non-probabilistic convenience sampling. Given this situation, the sample size was determined based on the maximum number of participants that could be selected within the limitations of the research.

Therefore, the study included 390 participants, of whom 65 presented cognitive impairment in the application of the Mini-Mental State Examination (MMSE), this being an exclusion criterion; The final response rate was 83.33%.

The mini-mental scale is a brief psychometric scale to evaluate the cognitive state of people and the Colombian version of this scale has been introduced by Rosselli et al. (2000). This scale was introduced into clinical practice as a practical method to assess the cognitive status of their patients. This instrument is composed of 30 dichotomous items that evaluate six cognitive processes: temporal orientation, spatial orientation, fixation memory, recall memory, attention and calculation, and language; the internal consistency is adequate (Cronbach’s alpha: 0.807).

The Fatigue Intensity scale (FSS) consists of nine items with a Likert-type response of seven possibilities of increasing intensity and scores between 1 and 7. The total is the sum of all the items. The Colombian version of Riveros Munévar et al. (2017) was used for this study.

The Colombian version of the Difficulties in Emotion Regulation Scale (DERS) was prepared by Muñoz-Martínez et al. (2006). This measures emotional dysregulation and consists of 36 items grouped into six factors as follows: (a) non-acceptance of emotional responses (non-acceptance), (b) difficulties in goal-directed behaviors when upset (goals), (c) difficulties in controlling impulsive behaviors when upset (impulsivity), (d) limited access to emotional regulation strategies that are perceived as effective (strategies), (e) lack of emotional awareness (awareness), and (f) lack of emotional clarity (clarity). The analysis of the DERS applied in a sample of Colombians revealed an internal consistency α = 0.90.

The Stanford Patient Education Research Center’s Chronic Disease Self-Efficacy (SEMCD-S) was developed by Ritter and Lorig (2014). This consists of four items on a scale of 1–10, with 1 being “Very Insecure” and 10 being “Very Confident.” This is an instrument for measuring the self-efficacy related to chronic disease management. Research reports a reliability of 0.95 (22) and for this research a Cronbach’s alpha of α = 0.791 was obtained.

MOS SF-36 Health Questionnaire was validated in Colombia by Lugo et al. (2006). This questionnaire consists of 36 items that explore eight dimensions of health status corresponding to physical function, social function, role limitations: physical problems, emotional problems, mental health, vitality, pain, and perception of general health. The reliability of the MOSSF-36 Health Questionnaire is 0.7; the physical role, physical function, and emotional role measurements obtained reliability scores equal to 0.90; in terms of content validity they had high correlations with the physical component (r ≥ 0.74).

Brief Illness Perception Questionnaire (Brief Illness Perception Questionnaire [IPQ-B]). In the Spanish version of Pacheco-Huergo et al. (2012), this questionnaire consists of nine items with Likert-type responses ranging from 0 to 10 points. It comprises eight dimensions and an ordinal open-ended question with three possible answers. These dimensions are as follows: consequences, duration, personal management, treatment management, identity, worry, emotional response and understanding of the disease. Item number 9 is an open-ended question in reference to the factors that caused his or her disease according to the person and organized accordingly of their importance. The questionnaire has adequate internal consistency indices ranging from α = 0.67–0.89.

Before starting the field phase of the research, which included the instrument application, 16 adults aged >60 years were selected and they underwent a pilot test to identify the words or questions that were difficult to understand to identify the respective comprehension in the response system and to establish the time for each of the instruments. At the end of this process, there was clarity and understanding on the part of the participants of the pilot test.

A total of 325 older adults were selected from the departments of Cesar and Atlántico based on the fact that they were older than 60 years, met the characteristics of having a diagnosis of a chronic disease, and voluntarily decided to participate.

At the outset, a brief explanation of the research was performed, explaining the objectives and procedures to be performed. After this, informed consents were signed, so that there was a written agreement with the participants to authorize their voluntary participation; This was followed by the scaling of sociodemographic data in order to find out the characteristics of the population and with these data, the MMSE was applied, which helped to rule out the subjects who presented cognitive impairment, ruling out cognitive impairment in the participants, this was followed by applying the instruments in the following order (FSS, DERS, SEMCD-S, MOS SF-36 and IPQ-B), the whole process lasted approximately 2 h per participant. The data obtained were incorporated into a database and analyzed using the SPSS-27 statistical package.

Firstly, an exploratory data analysis (EDA) was conducted in order to evaluate compliance with the assumptions needed to perform parametric tests, as well as to assess whether the use of multiple linear regression (MLR) was an appropriate option. In this regard, the normality of the variables was tested by means of the Shapiro–Wilk (SW) test, also opting for the construction of Q-Q plots as a complementary tool. Equality of variances was examined using Levene’s test, considering each of the sociodemographic characteristics as grouping factors. Independence of observations was inspected at statistical level using the median-based Wald-Wolfowitz cluster test, while the MLR conditions were validated by means of a thorough analysis of the residuals of each model.

In this regard, SW contrast was used again to check whether the residuals conformed to a normal distribution, while homoscedasticity was inspected graphically and analytically through typed scatter plots and by means of the Breusch–Pagan (BP) and White procedures. The residual autocorrelation assumption was evaluated with the Durbin-Watson (DW) statistic, while the presence of multicollinearity was checked using the variance inflation factors (VIF) and the condition index (CI). After this preliminary phase, it was concluded that nonparametric tests and regression options involving less demanding statistical assumptions should be used due to the violation of the aforementioned assumptions.

Once these conjectures were discarded, the research variables were described using basic summary measures. The raw or transformed scores were characterized through the minimum, maximum, range, interquartile range, mean, median, mode, variance, standard deviation, coefficient of variation, lower quartile, and upper quartile.

Furthermore, the sociodemographic characteristics compiled in the study, most of which were attribute variables, were presented in the form of absolute and relative frequencies. To compare the effect of these factors on the variables of interest, the Mann–Whitney U test (MWU) was selected when these characteristics consisted of two groups, choosing the Kruskal–Wallis H test (KWH) when, on the contrary, the number of levels was greater than or equal to three. On this occasion, the post hoc contrasts were conducted with the MWU method; however, the outcomes were corrected for significance via Bonferroni correction.

The dependent variables of the binary logistic regression (BLR) models were set from each of the eight dimensions of quality of life and used as predictors of the scores of the global constructs of self-efficacy, perception of disease, emotional regulation, and fatigue severity. The study was advanced in this way and each dimension was not incorporated as a regressor owing to the multicollinearity found between the factors of each scale, which had a negative impact on the VIF and the CI, increasing their values considerably.

In addition to the abovementioned statistical argument, there are numerous empirical and theoretical evidences to justify the use of global constructs of these instruments for description, classification, and prediction purposes. However, the categorization of the DVs was obtained by defining those scores below the lower quartile and above the upper quartile as low and high levels, which enabled the dichotomous characteristics necessary for the use of the BLR to be obtained. In addition, age, sex, time of disease diagnosis and the patient’s circumstances with respect to whether he/she lives alone or accompanied were introduced to the regression models to obtain adjusted estimates and odds ratios that truly reflected the magnitude of the association between the variables. Linearity between the predictors of the BLR and the logarithm of the odds ratio was checked through the Box-Tidwell procedure without finding relevant inconsistencies, while the existence of outliers and influence points was inspected by means of residuals, Cook’s and Mahalanobis distances. The presence of certain anomalies or influence data was moderate and did not impact the estimates.

Finally, data processing and analysis were performed with the IBM SPSS statistical package (version 27 for Windows 64-bit). Results whose significance was less than 0.05, 0.01 or 0.001 were considered significant.

The sample consisted of 325 people, of which 46.46% (n = 151) were female and 53.54% (n = 174) were male. Their age ranged from 60 to 78 years, with a mean of 68.39 and standard deviation of 4.15 (CV = 6.07%). In terms of time since disease diagnosis, values ranged from 5 to 60 years, with a mean of 15.79 and variability of 6.96 (CV = 44.08%). Other participant characteristics, such as the municipality of residence, marital status, whether living alone or accompanied, education, socioeconomic status, health system, psychiatric medication intake and treatment for their chronic condition are shown in Table 1.

Quality of life, self-efficacy, illness perception, emotional regulation, and fatigue.

The physical function of quality of life will be used as an example to develop this section. As can be seen in Table 2, it was found that the scores in this dimension ranged from 20 to 80 points, involving a range of 60 points. The lower and upper quartiles were 30 and 70, respectively, thus obtaining an interquartile range of 40 points. Regarding the centralization measurements, the scores averaged 49.91 points, with a median of 50 and a mode of 30. Variability was observed around the mean of 21.42 points, thus reflecting a relatively high coefficient of variation of 42.92%. For practicality, the characterization of the rest of the variables is omitted and the reader is referred to the detailed information shown in Table 2.

Tables 3–5 show the results of assessing the effect of the sociodemographic characteristics on the variables of interest for the study. It is emphasized at this point what was clarified in the statistical analysis section: nonparametric tests were used as a consequence of the non-observance of the postulates under which these tests operate. Thus, the data contrasted in these tables respond to the average assigned ranges, this is a typical transformation applied in these tests and is used to carry out the procedure. Therefore, special care must be taken when interpreting these values since they do not correspond to the raw scores.

Having said this and taking as a reference whether the patient takes psychiatric medication to cope with his or her condition, it was observed that this factor had an effect on the scores of the “personal control” dimension of the disease perception scale. It is noted that those not taking these medications reported higher assigned ranges than those who do (165.04 vs. 109.83, p < 0.05), which could be interpreted as meaning that the perception of disease is higher among those subjects who do not have the benefit of this palliative. When analyzing the health system and its influence, it can be seen that the participants belonging to the contributory system exhibit lower ratings in the “consequences” aspect of this questionnaire than those in the subsidized regime (109.57 vs. 165.59, p < 0.05), implying that the repercussions of the disease are conceived to be less serious among these individuals. However, it is important to emphasize that previous results should be interpreted cautiously because the sample size in one group is small and very different from that in the other group. As such, the validity of these findings may be questionable.

Regarding the severity of fatigue, it is noted that those using psychiatric medication reached significantly lower figures than those not doing so, this was observed in the dimensions of motivational affectation (101.50 vs. 165.36, p < 0.01), physical (92.08 vs. 174.69, p < 0.001), social (95.47 vs. 174.13, p < 0.001) and in the questionnaire total (90.47 vs. 174.96, p < 0.001).

As regards quality of life, it was found that marital status had a significant effect on the patients’ assessment of this construct. Specifically, widowers reported much lower ratings in the aspect of this variable defined as “physical function” than those who were single, separated or divorced (60.83 vs. 210.16, p < 0.001), but also than those who were married or cohabiting (60.83 vs. 156.53, p < 0.001). In turn, participants cohabiting with their partner reflected lower assigned ranks than those who were separated or single (156.53 vs. 210.16, p < 0.001). A similar situation was observed in vitality and mental health, both subscales of the MOS SF-36 instrument that measures health-related quality of life.

In an effort to avoid an excessive description of the results of this phase, the reader is encouraged to continue with this analysis by carefully reviewing the findings shown in Tables 3–5. In this regard, it should be noted that the APA methodology has been used for identifying whether two or more assigned ranges differ statistically, a procedure that involves assigning different letters as subscripts (a, b, c, etc.,). Furthermore, we have chosen to facilitate the visualization of the table by shading in gray those comparisons that were found to be significant, increasing the intensity of the shading as this property increases. Specifically: if highlighted in light gray, the value of p obtained is less than 0.05; if in medium gray, the significance observed is less than 0.01; and if in dark gray, the difference is statistically significant for values less than 0.001.

As noted in the statistical analysis section, when establishing the association between the aspects of quality of life (DV: dependent variables) and self-efficacy, disease perception, emotional regulation, and fatigue severity (IV: independent variables), BLR models were used. DVs were categorized from the lower and upper quartiles, while IVs were entered unmodified, in their raw or direct form. To obtain adjusted odds ratios (OR), covariates such as sex, age, disease duration and the patient’s living situation were incorporated into each model. The fit was assessed with Cox-Snell and Nagelkerke coefficients (pseudo R2), while the precision of the estimates was identified by displaying the bilateral confidence interval of the ORs together with their level of significance. All BLR assumptions were tested without any major irregularities or deviations being discovered.

Thus, the results indicate that self-efficacy, disease perception, emotional regulation, and severity of fatigue are variables that could impact physical function and quality of life. Note that each of these regressors reflected statistically significant odds ratios. Specifically, the perception of the disease and the severity of fatigue were identified as factors that probably negatively influence quality of life. It is noted that, if all other factors are held constant, for each point that self-efficacy increases, the probability of this happening is reduced by approximately 12% (OR = 0.88, p = 0.002, 0.81–0.95), a fact also observed under similar conditions, although less marked, in emotional regulation. In this case, if the other variables are not modified, a unit increase in this construct implies a 3% decrease (OR = 0.97, p = 0.027, 0.95–0.99) in the possibility of obtaining low levels of physical function. Conversely, both disease perception and fatigue severity were identified as risk variables. Regarding the former, one-point increases would lead to a 22% increase (OR = 1.22, p < 0.001, 1.11–1.34) in the probability of achieving low ratings in physical function; while, regarding the latter, unit increases in the fatigue severity scale would generate an increase in this risk of 13% (OR = 1.13, p < 0.001, 1.08–1.18). Judging from the pseudo R2s, about 57 to 77% of the variation in physical function would be explained by the model. To understand all the findings of this stage of the research, let us perform a similar description as in the previous paragraph with the rest of the models shown in Table 6.