Details of the study selection process and search results are shown in Fig 1. We identified a total of 6,408 articles based on the systematic literature search in five databases: MEDLINE (n=1,147), EMBASE (n=2955), PsycINFO (n=447), Emcare (n=1090) and CINAHL (n=769). After removal of duplicates, the title and abstracts of 3,674 articles were screened. Forty-seven studies were selected for a full-text review. Articles were excluded, generally, because study authors did not report on the association between subthreshold depression and self-care behaviours. Two studies that met the eligibility criteria were included in the review.
Both the included studies were hospital-based. A longitudinal study of 866 primary care attenders in Germany was reported by Dirmaier et al. (43) and Shin et al. reported a cross-sectional survey of 103 outpatients in the United State (44). Table 1 shows selected characteristics of the two included studies.
How was subthreshold depression determined?
The exposure, subthreshold depression, was determined using different procedures in the included studies. Dirmaier et al. classified participants as having subthreshold depression or depression if they had a Depression Screening Questionnaire (DSQ) score of between 5 and 7 or 8 or higher, respectively (43). Shin et al. used a two-step process to make a depression diagnosis (44). Initially, a screening questionnaire, the two-item version of the Patient Health Questionnaire (PHQ-2) (45), was completed and scored. Participants with a score of three or higher were assessed by a clinician using the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders- IV (DSM-IV) Axis I Disorders, non-patient edition (46). Also, participants taking any antidepressant medication (for any reason) and those who already had an MDD diagnosis were interviewed by a researcher. Participants were determined to have minor or major depression if they met DSM-IV diagnostic criteria (47).
How were self-care behaviours determined?
In the Dirmaier et al. study, six items in the questionnaire were used to determine self-care behaviours (43): A single item was used to determined medication non-adherence and five items to determine non-adherence to other health behaviours (diet (2 items), smoking, alcohol and physical activity). Shin et al. (44) used the Summary of Diabetes self-care activities (SDSCA) (48) to determine participants level of self-care behaviours during the past seven days. Self-care behaviours were assessed in five domains, which included diet, exercise, foot care, blood sugar testing, and medication taking. The overall self-care behaviour was calculated by adding up the scores of all the domains.
Participant characteristics are shown in Table 1. Included studies involved 968 adults (aged 18 years and above) participants. In both the studies, males and females were equally represented (43, 44). Subjects in the Dirmaier et al. study (43) were older than those in Shin et al. study (44). Mean diabetes duration was around ten years in both studies (43, 44).
Quality appraisal of included studies is shown in Table 2. Since none of the included studies were of a case-control study design, the JBI critical appraisal checklist tool for cross-sectional and cohort studies were used. None of the studies was excluded based on their quality appraisal.
In the study by Dirmaier et al. , three of eleven items were rated as potential sources of bias (43). Of particular concern, key variables were determined using measures that were developed specifically for the study – e.g., medication non-adherence – despite validated measures being available. The flow of participants through the study is not clear and that authors do not state strategies to address the incomplete follow-up.
Several potentially important sources of bias were identified in the Shin et al. study (44). Six items were rated low, and two unclear risk of bias. The most important source of potential bias was related to confounding, as authors adjusted for three confounders (age, sex, and diabetes duration) in their analysis. Other potential confounders that may have impacted the observed association have not been addressed. The information about the setting of the study is not sufficiently detailed. Although the authors articulated clear hypotheses, the authors did not report a sample size calculation; consequently, it cannot be determined if the sample size was appropriate to test the expressed hypotheses.
Both the studies have used self-report tools to measure self-care behaviours (43, 44). The study by Dirmaier et al. used a single item to measure medication adherence (43). Health behaviour was measured using a tool that was not validated and the items specific to diabetes self-care behaviour such as blood glucose testing and foot care was not included to measure total health behaviour (43).
Based on these assessments, the overall methodological quality of included studies was judged to be potentially biased because of a number of reasons that included limitation in the reserch design, lack of use of validated measure, inappropriate confounders adjustments and a small sample size.
Association between subthreshold depression and self-care behaviours
Dirmaier et al. reported that subthreshold depression was associated with non-adherence to health behaviour over a period of twelve months follow-up (43). The association was retained (β= 1.01; CI: 0.62-1.40; p< 0.001) after adjusting for gender, age, marital and employment status, education, physical activity, BMI, smoking, drinking, duration of diabetes and type of diabetes treatment. No significant association between subthreshold depression and medication non-adherence was observed.
Shin et al. reported that individuals with T2D and subthreshold depression scored lower in their self-care behaviour as compared to the group without depression (44). However, there was no significant association between subthreshold depression and the overall self-care behaviour or individual self-care domains (diet, exercise, blood sugar, foot care, and medication).