Participants
The questionnaire was completed by 1002 participants. Thirty-five responses were removed as they were either completed in under 330 seconds or had duplicate IP addresses. The remaining 967 valid responses were used for analysis. Table
1 details the sociodemographic characteristics of the respondents. The average age of the respondents was 44.41 (
SD = 12.14). The distribution of sex was roughly balanced with 50.8% female respondents and 49.1% male respondents. The ethnic distribution of the sample is also approximately representative of the ethnic makeup of the Singapore population in 2021 [
23], with 75.0% of the respondents being Chinese, 14.5% being Malay, and 7.4% Indian.
Table 1
Sociodemographic characteristics of participants
Sex |
Female | 491 (50.8) |
Male | 475 (49.1) |
Residency Status |
Singapore Citizen | 846 (87.5) |
Singapore Permanent Resident | 121 (12.5) |
Age |
21–29 | 130 (13.4) |
30–39 | 234 (24.2) |
40–49 | 257 (26.6) |
50–59 | 223 (23.1) |
60 and above | 123 (12.7) |
Ethnicity |
Chinese | 725 (75.0) |
Malay | 140 (14.5) |
Indian | 72 (7.4) |
Others | 30 (3.1) |
Education |
No formal schooling | 4 (0.4) |
PSLE or equivalent | 9 (0.9) |
GCE O/N Level | 155 (16.0) |
GCE A Level/ diploma | 263 (27.2) |
Degree/higher education | 536 (55.4) |
Monthly household income (In SGD) |
Below $1000 | 37 (3.8) |
1000–4999 | 257 (26.6) |
5000–9999 | 329 (34.0) |
10,000–14,999 | 179 (18.5) |
15,000–19,999 | 78 (8.1) |
20,000 and over | 61 (6.3) |
Do not wish to answer | 26 (2.7) |
Knowledge about antibiotics and AMR
Respondents correctly answered an average of 12.5 (SD = 5.01) of the 28 questions on antibiotics. On the general statements asked about antibiotics, −respondents correctly answered an average of 3.61 (SD = 2.81) of the 11 questions. On the side effects of antibiotics, respondents correctly identified an average of 2.00 (SD = 1.33) of the 7 side effects of antibiotics, with fever, bloating, and vomiting being the three least known side effects. As for AMR knowledge, respondents correctly answered an average of 3.15 (SD = 1.71) of the 7 questions.
Predicting adherence to antibiotics with PMT
Table
3 presents the descriptive statistics of the PMT constructs and their correlations. Of the five PMT constructs, perceived response cost has the lowest response rating with an average of 1.48 (
SD = .82) on a 5-point Likert scale. This means that respondents, on average, tend to perceive adherence to antibiotic as an action with low cost. Perceived susceptibility to AMR has the second-lowest response rating with an average of 3.11 (
SD = 0.77) on a 5-point Likert scale. This means that most respondents tended to be neutral about whether AMR will affect themselves or the people around them. Perceived response efficacy, on the other hand, has the highest response rating of 3.92 (
SD = 0.74), meaning that respondents tended to believe that adherence to antibiotics is effective in coping with the threat of AMR.
Table 3
Descriptive statistics and intercorrelations of PMT constructs and adherence to antibiotics
1. Adherence to antibiotics | – | | | | | | | | 4.57 (0.73) | 1–5 |
2. Susceptibility | .27** | – | | | | | | | 3.11 (0.77) | 1–5 |
3. Severity | .08* | .005 (p = .12) | – | | | | | | 3.72 (0.73) | 1–5 |
4. Self-Efficacy | .11** | −.07* (p = .03) | .49** | – | | | | | 3.82 (0.67) | 1–5 |
5. Response-Efficacy | .23** | 0.04 (p = .21) | .53** | .48** | – | | | | 3.92 (0.74) | 1–5 |
6. Response Cost | −.68** | −.28** | −.11** | −.10** | −.20** | – | | | 1.49 (0.82) | 1–5 |
7. Threat appraisal | .25** | .74** | .71** | .28** | .38** | −.28** | – | | 3.41 (0.54) | 1–5 |
6. Coping appraisal | .51** | .13** | .53** | .70** | .78** | −.65** | .44** | – | 2.08 (0.52) | 1–5 |
The results also show that adherence to antibiotics is significantly correlated with all five PMT constructs. Coping and threat appraisals are significantly correlated with antibiotics adherence (r =. 51and .25 respectively, p < .01). Among the five constructs, the perceived response cost of adherence to antibiotics has the strongest correlation with adherence to antibiotics (r = −.69, p < .001). It is also the only construct negatively correlated with adherence to antibiotics. On the other hand, the perceived severity of AMR has the weakest correlation with adherence to antibiotics (r = .08, p = .01).
Table
4 shows the results of the hierarchical regression analysis to adherence to antibiotics based on the PMT model.
Model 1 served as the baseline that informs us how participants’ demographic traits could influence their adherence to antibiotics. It accounted for only 6.3% of the variance in adherence to antibiotics,
F (3, 842) = 19.82,
p < 0.001 with
R2 = .066. For
Model 2, the inclusion of the knowledge about antibiotics and AMR improved the effect size of the model, from
R2 = .066 to
R2 = .14,
F (2, 840) = 34.62,
p < 0.001, specifically. Finally,
Model 3, with five PMT constructs added to the analysis, presented the best predictive model with the biggest
R2 change,
F (5, 835) = 126,
p < 0.001, which increased the
R2 by .37. As shown in the results in Model 3, perceived response cost emerged as the most salient predictor of adherence to antibiotics,
t = − 22.57,
p < .001. One unit increase in perceived response cost decreases the average adherence score, which ranges from 1 to 5, by .61 ((
β = .61,
p < 0.001). Perceived response efficacy of adherence to antibiotics (
t = 3.1,
p = .002;
β = .096), perceived susceptibility to AMR (
t = 3.6,
p < .001
; β = .097), education level (
t = − 3.02,
p < .001,
β = −.076), antibiotic knowledge (
t = 2.55,
p = .01;
β = 0.073), and perceived severity of AMR (
t = − 2.17,
p = .03,
β = −.069) also emerged as significant predictors of adherence to antibiotics.
Table 4
Summary of hierarchical regression analysis for variables predicting adherence to antibiotics
Age | 0.013 | 0.002 | 0.222** | 0.011 | 0.002 | 0.183** | 0.003 | 0.002 | 0.049 (p = .06) | 0.005 | 0.002 | 0.077* (p = .01) |
Sex | 0.043 | 0.049 | 0.029 (p = .38) | 0.012 | 0.048 | 0.008 (p = .80) | −0.031 | 0.036 | −0.021 (p = .40) | 0.003 | 0.043 | 0.002 (p = .94) |
Education level | −0.079 | 0.032 | −0.084 (p = .01) | − 0.09 | 0.031 | −0.095** (p = .004) | −0.072 | 0.024 | −0.076** (p = .003) | −0.073 | 0.028 | −0.078** (p = .009) |
Antibiotics knowledge | | | | 0.044 | 0.005 | 0.287** | 0.011 | 0.004 | 0.073* (p = .01) | 0.016 | 0.005 | 0.104** (p = .002) |
AMR knowledge | | | | −0.033 | 0.015 | −0.077* (p = .03) | −0.018 | 0.012 | −0.041 (p = .15) | −0.064 | 0.014 | −0.148** |
Susceptibility | | | | | | | 0.091 | 0.025 | 0.097** | | | |
Severity | | | | | | | −0.070 | 0.032 | −0.069** (p = .03) | | | |
Self-Efficacy | | | | | | | 0.041 | 0.033 | 0.036 (p = .22) | | | |
Response-Efficacy | | | | | | | 0.095 | 0.031 | 0.096** (p = .002) | | | |
Response Cost | | | | | | | −0.544 | 0.024 | −0.61** | | | |
Threat appraisal | | | | | | | | | | 0.055 | 0.023 | 0.082* (p = .02) |
Coping appraisal | | | | | | | | | | 0.212 | 0.016 | 0.455** |
R2 | 0.066 | 0.137 | 0.508 | 0.316 |
Adjusted R2 | 0.063 | 0.132 | 0.502 | 0.310 |
F for change in R2 | 19.82** | 34.62** | 126.00** | 109.22** |
In the last regression (Model 4), the five PMT dimensions were replaced with their overarching constructs – coping appraisal and threat appraisal. The three sociodemographic variables, two knowledge variables, together with the coping appraisal, and threat appraisal were entered together in the regression. The final model is also significant, F (7, 838) = 55.18, p < 0.01, However, it yields a smaller R2 change = .18 (F (2, 838) = 109.22, p < .001) as compared to that of model 3 (R2 change = .37) with the five dimensions of PMT included. Therefore, model 3 is the model with better predictive power.