Adherence of People with Chronic Disease Instrument

The Theory of Adherence of People with Chronic Disease Instrument is based on a theoretical model of chronically ill patients developed and tested by Kyngäs (1999) among diabetic adolescents. The theoretical model and ACDI have been developed further and used as a theoretical framework in studies related to adherence among young and adult patients with chronic diseases (Kyngäs et al. 2000; Lunnela et al. 2011; Kääriäinen et al.

2013). ACDI includes 37 items (Table 6) measuring adherence to treatment. According to the original Theory of Adherence of People with Chronic Disease (Kyngäs 1999), adherence to treatment consisted of two mean sum variables: adherence to medication (2 items) and a healthy lifestyle (4 items). These two variables were explained with nine mean sum variables: responsibility (2 items), motivation (2 items), cooperation (2 items), results of care (2 items), fear of complications (2 items), sense of normality (7 items), support from next of kin (5 items), support from nurses (4 items) and support from physicians (4 items).

The construct validity of the ACDI was verified with an exploratory factor analysis (EFA) using Principal Axis Factoring and Promax rotation, which produced a factor solution with satisfactory statistical values (Table 6). Missing values were replaced with each item’s mean value. Eleven factors explained 65 % of the total variance, communalities varied between 0.20 – 0.80, and the factor loadinds were between 0.30 – 0.90. One original item related to responsibility was removed, because it did not load to any factor. Internal consistency of the mean sum variables was evaluated by Cronbach’s alpha values, which varied between 0.40 – 0.90. The alpha of the whole instrument was 0.84, which represent acceptable value. (Burns & Grove 2009).

Based on the EFA results, and according to the original Theory of Adherence of People with Chronic Diseace (Kyngäs 1999), which was verified to be suitable for assessing adherence to treatment among patients with CHD after PCI (Kähkönen et al. 2015), 11 mean sum variables were formatted: adherence to treatment (adherence to medication and adherence to a healthy lifestyle) which was explained with nine mean sum variables (responsibility, motivation, cooperation, results of care, fear of complications, sense of normality, support from next of kin, support from nurses and support from physicians).

These mean sum variables were rated on a 5-point Likert scale ranging from ‘definitely disagree’ (1), ‘disagree’ (2), ‘uncertain’ (3), ‘agree’ (4) and ‘definitely agree’ (5). As in previous studies (Kyngäs & Rissanen 2001; Kääriäinen et al. 2013) that used the same instrument, the mean sum variables were categorised into two classes: good adherence to treatment and reduced adherence to treatment. Same categorisation was used so that the research results would be comparable with previous results on adherence to treatment.

Those with values of 1 - 3.5 were combined and assigned a value of 1, indicating a low level of adherence to treatment. Values ranging from 3.51 to 5.0 were combined and recoded with a value of 2, representing a high level of adherence to treatment.

22

Table 6. Factors, factor loadings and Chronbach’s alphas related to mean sum variables of adherence.

MEAN SUM VARIABLES AND FACTOR

Item 1: Related to patient’s adherence to medication instructions

Item 2: Related to patient’s medication changes

Factor 2: Adherence to healthy lifestyle 0.37 – 0.52 1.3 0.53 Item 3: Related to patient’s smoking habits

Item 4: Related to patient’s alcohol consumption Item 5: Related to patient’s physical activity Item 6: Related to patient’s diet

MEAN SUM VARIABLES RELATED TO ADHERENCE TO TREATMENT

Factor 3: Cooperation 0.37 – 0.87 1.6 0.71

Item 7: Related to patient’s secondary prevention follow-up treatment

Item 8: Related to patient’s possibility to discussion with physician

Item 9: Related to patient’s possibility to discussion with nurse

Factor 4: Responsibility 0.40 1.2 0.41

Item 10: Related to patient’s own responsibility Item 11: Related to patient’s willigness to good self-care

Factor 5: Support from next of kin 0.30 – 0.86 1.9 0.60

Item16: Related to support from next of kin to patient's self-care

Item 25: Related to acceptance and support from next of kin

Item 26: Related how next of kin are interested of patient's life

Item 27: Related how the next of kin remind patient of treatment

Item 28: Related to how next of kin motivate patient to self-care

Factor 6: Sense of normality 0.26 – 0.66 7.2 0.88

Item 14: Related to patien’st refusal of treatment regimens

Item 18: Related to patient’s inability to live normal life

Item 19: Related to patient’s willingness to stay at home because of illness

Item 20: Related how patient’s experience self-care as a part of their life

Item 21: Related how self-care limits patient's independence

Item 22: Related how self-care limits patient's the daily routine

Item 23: Related how self-care causes dependence of next of kin

23

Data analysis

In the first phase of the study, the suitability of the Theory of Adherence of People with Chronic Disease for assessing adherence to treatment and related factors among patients with CHD after PCI was tested using structural equation modelling (SEM; Article I). SEM was a reasonable method for testing the theoretical depiction of relationships among the concepts (Cheung & Rensvold 2002). SEM software programmes produce various statistics pertaining to the fit of the model, of which all are unnecessary to report. However, there is no consensus on which indices are the best to evaluate the goodness of fit and the correspondence between the theoretical model and the observed correlation matrix. Chi-square tests and their derivatives (χ² test; Kline 1998; Schumacker & Lomax 2004) are one of the most frequently recommended and generally used modification indices of the goodness of fit. The purpose is to test the theoretical model produced by the covariance matrix and the observed covariance matrix compatibility. The null hypothesis under the chi-square test complies with the χ² distribution. Chi-square tests assess the value of the test by determining its degrees of freedom, and the result should be less than 3 (Kline 2005.) However, it rejects null hypothesis too easily with large sample size (> 200), in which case Hoelter's ‘critical N’ test is recommended (Hoelter 1983). In addition, it is recommended to repot one baseline comparison index and the root mean square error of approximation

Item17: Related to the maintenance of health status

Item 24: Related to wellbeing

Factor 9: Support from nurses 0.62 – 0.90 3.5 0.60

Item 33: Related to nurse’s ability to make complete plan for the patient's care

Item 34: Related to nurse’s complete interest of patient

Item 35: Related to nurse’s ability to motivate patient

Item 36: Related to nurse’s interaction skills

Factor 10: Support from physicians 0.61 – 0.87 2.5 0.88

Item 29: Related to physician’s ability to make complete plan for the patient's care

Item 30: Related to physician’s complete interest of patient

Item 31: Related to physician’s ability to motivate patient

Item 32: Related to physician’s interaction skills

Factor 11: Fear of complications 0.88 – 0.89 1.4 0.88

Item 37: Related to patient's fear of cardiac events Item 38: Related to patient's fear of comorbidities

Note: Modified Adherence of Chronic Disease Instrumen has been described in accordance with copyright agreement.

24

(RMSEA; Steiger 1989). The baseline comparison index that is generally used is the comparative fit index, which was also used is this study (CFI; Bentler 1990). These indices might be less influenced by large sample sizes, and are proposed as alternatives to the chi-square test (Cheung & Rensvold 2002).

A sufficiently good model should have a CFI of 0.90 and RMSEA should be 0.06 – 0.07.

(Cheung & Rensvold 2002; Kankkunen et al. 2005; Shreiber et al. 2006; Metsämuuronen 2011; Kellar & Kelvin 2012.) The numerical estimates of the parameters can be represented by standardised or non-standardised values. Standardisation is generally given as an option, because then estimates can be interpreted in the same scale. Standardised estimates include correlations (standardised covariance) and path coefficients. The effect of standardised estimates is interpreted as weak if their values are < 0.10. Estimates with a medium effect have values ~0.30, and estimates with a value > 0.50 are interpreted as having a major effect. (Kline 1998; Schumacker & Lomax 2004).

The second objective in the first phase of the study was to identify the predictive factors (sociodemographic, health behavioural, and disease-specific) of adherence to treatment in patients with CHD after after undergoing a PCI (Article II) using the Adherence of People with Chronic Disease Instrument (ACDI). Descriptive statistics (frequencies, percentages, means, standard deviations [SD], ranges and medians) were used to describe the level of adherence to treatment (Polit & Beck 2012). Cross-tabulations with chi-square tests were used to identify statistically significant and independent sociodemographic, health behavioural and disease-specific predictors of dependent mean sum variables explaining adherence to treatment in the univariate model. In cases in which a chi-square test was not appropriate (no more than 20% of the cells should have < 5), Fisher’s exact test was used Statistically significant independent predictors in the univariate model were recoded as dummy variables (0, 1) and entered into a multivariate logistic regression model (backward stepwise selection) to confirm standardised sociodemographic, health behavioural and disease-specific factors predicting adherence to treatment. P-values < 0.05 were considered statistically significant. The goodness of fit was evaluated using chi-squared distribution and Nagelkerke R-square values (Burns & Grove 2009; Kellar & Kelvin 2012; Polit & Beck 2012).

5.2.2 EuroQoL five-dimensional scale and EuroQoL visual analogue scale instruments