http://orcid.org/0000-0002-5085-6662
Figures
Abstract
Background
Comorbid depression is common among patients with diabetes and has severe health consequences, but often remains unrecognized. Several questionnaires are used to screen for depression. A systematic review and meta-analysis regarding the diagnostic accuracy of depression questionnaires in adults with diabetes is unavailable. Our aim was to conduct a systematic review and meta-analysis to evaluate the diagnostic accuracy of depression questionnaires in adults with type 1 or type 2 diabetes.
Methods
PubMed, Embase and PsycINFO were searched from inception to 28 February 2018. Studies were included when the diagnostic accuracy of depression questionnaires was assessed in a diabetes population and the reference standard was a clinical interview. Data extraction was performed by one reviewer and checked by another. Two reviewers independently conducted the quality assessment (QUADAS-2). Diagnostic accuracy was pooled in bivariate random effects models. The main outcome was diagnostic accuracy, expressed as sensitivity and specificity, of depression questionnaires in an adult diabetes population. This study is reported according to PRISMA-DTA and is registered with PROSPERO (CRD42018092950).
Results
A total 6,097 peer-reviewed articles were screened. Twenty-one studies (N = 5,703 patients) met the inclusion criteria for the systematic review. Twelve different depression questionnaires were identified, of which the CES-D (n = 6 studies) and PHQ-9 (n = 7 studies) were the most frequently evaluated. Risk of bias was unclear for multiple domains in the majority of studies. In the meta-analyses, five (N = 1,228) studies of the CES-D (≥16), five (N = 1,642) of the PHQ-9 (≥10) and four (N = 822) of the algorithm of the PHQ-9 were included in the pooled analysis. The CES-D (≥16) had a pooled sensitivity of 85.0% (95%CI, 71.3–92.8%) and a specificity of 71.6% (95%CI, 62.5–79.2%); the PHQ-9 (≥10) had a sensitivity of 81.5% (95%CI, 57.1–93.5%) and a specificity of 79.7% (95%CI, 62.1–90.4%). The algorithm for the PHQ-9 had a sensitivity of 60.9% (95%CI, 52.3–90.8%) and a specificity of 64.0% (95%CI, 53.0–93.9%).
Conclusions
This review indicates that the CES-D had the highest sensitivity, whereas the PHQ-9 had the highest specificity, although confidence intervals were wide and overlapping. The algorithm for the PHQ-9 had the lowest sensitivity and specificity. Given the variance in results and suboptimal reporting of studies, further high quality studies are needed to confirm the diagnostic accuracy of these depression questionnaires in patients with diabetes.
Citation: de Joode JW, van Dijk SE, Walburg FS, Bosmans JE, van Marwijk HW, de Boer MR, et al. (2019) Diagnostic accuracy of depression questionnaires in adult patients with diabetes: A systematic review and meta-analysis. PLoS ONE 14(6): e0218512. https://doi.org/10.1371/journal.pone.0218512
Editor: Wisit Cheungpasitporn, University of Mississippi Medical Center, UNITED STATES
Received: March 15, 2019; Accepted: June 4, 2019; Published: June 20, 2019
Copyright: © 2019 de Joode et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Depression among patients with diabetes is common and has severe health consequences. Depression is defined as severely depressed mood that persists for at least two weeks in combination with 5 of the symptoms (i.e. loss of pleasure, changes in sleep pattern, early rising, changes in appetite with weight loss/gain, feelings of guilt/worthlessness, low energy level, difficulty concentrating, nervousness, morning sadness)[1]. Comorbid depression is present in 12% to 19% of patients with type 1 and type 2 diabetes respectively[2]. The number of people suffering from both depression and diabetes is expected to rise sharply in the next decade[3, 4]. Comorbid depression is associated with a reduction in quality of life[1, 5], poorer self-care behavior[1, 6, 7], deterioration of glycemic control[1, 7, 8], and increased expenditure on health care costs[9, 10]. Moreover, patients with both diabetes and depression have more comorbidities[1, 7, 11] and show higher mortality rates[1, 7, 12] compared to diabetes patients without depression.
Although effective treatment options for depression in patients with diabetes are available[13, 14], comorbid depression may still be a problematic issue. Depression may remain unacknowledged and undiagnosed in more than half of the cases in both specialized diabetes centers[15] and non-specialized centers[16], thereby possibly missing appropriate intervention and treatment. The main reasons that patients and health care professionals may not discuss depression as an issue include the focus on somatic symptoms and complications, undue normalization of depressive symptoms, and a lack of opportunity to discuss mental health in routine diabetes consultations[17]. Screening for depression is recommended in clinical guidelines[18–21] and various depression questionnaires are used for screening and diagnosing purposes[22–26]. These questionnaires are often based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders III or IV (DSM-III or DSM-IV).
Some symptoms of depression (e.g., change in appetite, changes in weight, loss of energy and difficulties in concentrating) are also common in diabetes. This may result in an overestimation of depressive symptoms in diabetes patients and, higher scores on depression questionnaires, resulting in a higher false positive rate. To ensure existing depression screening questionnaires can be validly used in a population of diabetes patients, many of these have undergone psychometric testing in this specific population. Recently, a systematic review focusing on measurement properties (i.e. reliability, validity and responsiveness) of these questionnaires in a diabetes population was performed and found that, based on the current knowledge, the Centre for Epidemiological Studies Depression Scale (CES-D) is the best questionnaire for monitoring depressive symptoms[27]. However, screening purposes are related to other measurement properties (i.e. sensitivity and specificity) than monitoring purposes. The screening and diagnostic quality of a tool is determined by the diagnostic accuracy of a test, which is defined as “a test’s ability to discriminate between people with the target condition and those without” compared to a reference standard[28], such as a clinical interview for depression.
Roy et al. (2012) performed a systematic review of the literature in which they identified frequently used depression questionnaires in a diabetes population, and the corresponding sensitivity and specificity of these questionnaires. However, a meta-analysis and quality assessment were not included[29]. Practical recommendations regarding the use of specific tools could therefore not be made. Furthermore, the correlation between specificity and sensitivity was not taken into account[29], as recommended by the Cochrane Collaboration[28]. The aim of this study was to conduct a systematic review and meta-analysis to evaluate the diagnostic accuracy of depression questionnaires in adults with type 1 or type 2 diabetes.
Materials and methods
Design
This study is registered with PROSPERO, number CRD42018092950[30], and is reported according to the Preferred Reporting Items for a Systematic Review and Meta-analysis of Diagnostic Test Accuracy Studies (PRISMA-DTA) (S1 and S2 Tables)[31].
Search strategy and study selection
PubMed, EMBASE and PsycINFO were searched from inception up to February 28, 2018. The search strategy consisted of terms for diabetes and depression (S3 Table). Terms about diagnostic accuracy and questionnaires were not included because clear terms for identifying diagnostic accuracy studies in databases are lacking[28, 32] and no studies should be missed. Studies were included when the diagnostic accuracy of depression questionnaires was measured in a diabetes population (i.e. at least 80% of the population had diabetes type 1 or 2) and the reference standard was a clinical interview. There were no language restrictions. Depression questionnaires are defined as questionnaires which are developed to measure depressive symptoms. Despite the fact that the World Health Organization-Five Well-Being (WHO-5) was originally developed for the assessment of subjective psychological well-being, it was included, because this questionnaire is widely used for measuring depression symptoms[33]. Duplicate records were removed according to the recommendations of Bramer et al.[34]. The titles and abstracts of peer-reviewed full articles were screened; comments, letters, editorials, book sections and theses were excluded.
Pairs of review authors independently assessed titles and abstracts to identify relevant articles. Full-texts were retrieved when both review authors agreed that studies were relevant or when consensus was not reached. Three review authors read the full-texts to judge study eligibility, independently. Disagreements were resolved by discussion, when consensus was not reached, a fourth reviewer made the final decision. Reference lists of included studies were screened for additional relevant studies by two review authors independently.
Data extraction
Using a structured data extraction form, the following characteristics and data were extracted from included studies: sample size, age, gender, diabetes type, prevalence of depression in the sample, the country and setting in which the study was performed, depression questionnaire used, language, used thresholds with corresponding diagnostic accuracy properties (i.e. sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), area under the curve (AUC)) and data to generate two-by-two-tables. Sensitivity of a questionnaire entails “the probability of a positive test given the presence of the disease”, while specificity entails “the probability of a negative test in those without the disease”[35]. Sensitivity and specificity of a questionnaire can be calculated at several thresholds. A threshold is defined as the sum score on a questionnaire that is the turning point between having a depression or not. The result of a screening questionnaire is used by clinicians to make decisions about further testing and therapy[18–20] and is used by researchers to make decisions about eligibility for participation in studies. For this reason, the depression questionnaire with the best diagnostic accuracy should be identified in particular for clinical practices and for research among patients with diabetes. The PPV is “the probability of the presence of disease in those with a positive test result” and the NPV is “the probability of absence of disease in those with a negative test result” [35]. The AUC in diagnostic accuracy studies is the area under the receiver operating characteristic (ROC) curve that reflects the inverse relationship between sensitivity and specificity at several thresholds. Data were extracted by one review author and checked by a second review author. The percentage of agreement for the data extraction was 0.94. Primary outcome of interest was diagnostic accuracy expressed as sensitivity and specificity of depression questionnaires in an adult diabetes population.
Quality assessment
The quality assessment of included studies consisted of the following four domains according to the revised version of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2): Patient Selection, Index Test, Reference Standard and Flow and Timing[36]. In this review, Index Test refers to the specific depression questionnaire evaluated. No signaling questions were added to or omitted from the QUADAS-2 format[36]. Interpretations of the signaling questions are described in S1 Text and S4 Table. All included studies were assessed for risk of bias in each domain and for applicability concerns in the first three domains. Risk of bias was judged as “low”, “high”, or “unclear”. Applicability concern is “the concern that the study does not fit in the review question” and was also judged “low”, “high” or “unclear”[36]. The quality assessment was independently performed by two review authors. The German and Spanish article were discussed with a native German and Spanish academic colleague, respectively. When consensus was not reached, a third review author decided.
Data synthesis and statistical analysis
For the pooling of extracted data about sensitivity and specificity, at least three studies for each questionnaire with a corresponding threshold were needed. A bivariate random effects model was performed to adjust for the within- and between-study variance in sensitivity and specificity[37]. The method for the meta-analysis was based on the Stata manual of the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy[38]. Sensitivity and specificity were converted to two-by-two-tables to get data of true positives (TP), false positives (FP), true negatives (TN) and false negatives (FN). Then, data of the individual studies was plotted in a forest plot and a summary receiver operating characteristic (SROC) plot to illustrate the location and scatter of the data using RevMan (version 5.1). Analyses were conducted using the metandi option in StataSE (version 14). When the correlation between sensitivity and specificity could not be estimated, the xtmelogit option was used. These analyses resulted in a summary operating point (i.e. summary estimate for sensitivity and specificity) per questionnaire with 95% confidence region and 95% prediction region[38]. The 95% prediction region “illustrates the extent of statistical heterogeneity by depicting a region within which (assuming the model is correct) we have 95% confidence that the true sensitivity and specificity of a future study should lie”[39]. We aimed to investigate the source of heterogeneity between results using meta-regression and subgroup analysis. Prior to the analyses, variables that could lead to heterogeneity were selected. These were blinding of the reference standard, distribution of diabetes type, percentage of depression cases in the sample and setting. However, due to the low number of studies in the meta-analysis, it was not possible to perform meta-regression or subgroup analysis.
Results
Study inclusion and characteristics of included studies
Fig 1 shows the study selection process in detail according to the PRISMA-DTA[31]. In the identification phase, 8,219 records were identified through database searching (S3 Table). No additional records were identified by screening of reference lists. In the screening phase, titles and abstracts of 6,097 full articles were screened. In the eligibility phase, 127 articles were selected for full-text retrieval, of which 106 were excluded. Reasons for exclusion are described in Fig 1. This resulted in the inclusion of 21 studies[40–60] for the systematic review (N = 5,703 patients). Of these, ten studies (N = 3,026 patients) were eligible for meta-analysis[43, 48, 49, 51, 53, 54, 56, 57, 59, 60] because at least three studies per threshold per questionnaire were needed.
Table 1 displays the characteristics of the included studies. Twelve different questionnaires were identified in the included studies, of which the CES-D and the Patient Health Questionnaire 9-item version (PHQ-9) were the most frequently evaluated. S5 Table presents the characteristics of the twelve questionnaires. In 19 studies consisted the study samples of patients with diabetes[40–46, 48–50, 52–60] and in two studies the diagnostic accuracy data was reported separately for patients with diabetes[47, 51]. Distribution of diabetes type differed between studies; from 100% diabetes type 1[44] to 100% diabetes type 2[40, 42, 43, 46, 48, 49, 51, 53, 55, 56, 59, 60]. Studies varied largely in sample sizes (range 65[41, 58]– 793[48]) and were conducted in different settings. Study samples differed in proportion of men (range 31.4[42]– 67.3%[48]), mean age (range 43.3[44]– 71.4[51] years) and prevalence of depression based on the clinical interview (range 3.5[44]– 43.2%[49]). S6 Table presents the extracted data regarding the diagnostic accuracy.
Quality assessment
Table 2 presents the results of the quality assessment regarding risks of bias and applicability concerns; explanations of decisions are listed in S4 Table. The risk of bias in the domain of Patient selection was low in the majority of studies[41, 42, 44, 45, 47, 48, 52, 54–56, 58–60]. The clinical interview was interpreted with knowledge of the scores on the depression questionnaire in two studies resulting in a high risk of bias in the domain of Reference Standard[51, 57]. In the majority of studies the procedure of testing patients was not clearly described resulting in an unclear risk of bias for the Index test[40, 43, 44, 46, 48, 50, 52, 55, 59, 60] and the Reference Standard[40, 42–44, 46, 48, 50, 52–54, 56, 59, 60]. In the domain Flow and Timing the risk of bias was either unclear[40, 42–44, 46–53, 59, 60] or high[41, 45, 54–58], because the procedure was not clearly described or the drop-out rates were high. Since appropriate index tests and reference standards were specified in inclusion criteria, all studies had low applicability concerns in domains Index Test and Reference Standard.
Results of meta-analysis
Only for the CES-D and the PHQ-9 there were at least three studies available for meta-analytical procedures. Data of the CES-D were pooled at a threshold of 16. For the PHQ-9 the data were pooled at a threshold of 10 and at the threshold according to the algorithm. The algorithm for the PHQ-9 is a specific threshold for identifying depression, which is defined in accordance with DSM-IV: five or more of the nine depressive symptoms criteria are present for at least more than half the days in the past two weeks and one of the symptoms is depressed mood or anhedonia[49]. The Forest plots (Fig 2A) and SROC plots (Fig 2B) contain the data that were pooled in the meta-analysis. Table 3 displays the summary operating points per questionnaire and S1 Fig displays these results visually in SROC plots. The CES-D (≥16) had a pooled sensitivity of 85.0% (95%CI, 71.3–92.8%) and a specificity of 71.6% (95%CI, 62.5–79.2%); the PHQ-9 (≥10) a sensitivity of 81.5% (95%CI, 57.1–93.5%) and a specificity of 79.7% (95%CI, 62.1–90.4%). Finally, the algorithm for the PHQ-9 had a sensitivity of 60.9% (95%CI, 52.3–90.8%) and a specificity of 64.0% (95%CI, 53.0–93.9%).
Forest plots (A) and SROC plot (B) of the CES-D (≥16), PHQ-9 (≥10) and PHQ-9 (algorithm) (A) 95%CI = 95% confidence interval; FN = false negatives; FP = false positives; TN = true negatives; TP = true positives. a two-by-two-table was obtained after correspondence with the author. (B) Each symbol represents a pair of sensitivity and specificity from a study and the size of symbols reflects the sample size of the study.
Discussion
The results of the meta-analysis indicate that the CES-D (≥16) had the highest sensitivity and the PHQ-9 (≥10) had the highest specificity, although confidence intervals were wide and overlapping. The algorithm for the PHQ-9 had the lowest sensitivity and specificity.
In 2012, Roy et al. summarized the diagnostic accuracy of depression questionnaires among patients with diabetes in a systematic review in which 23 studies were included[29]. Only 7 of these studies were included in the current review because some studies did not meet our more strict inclusion criteria; especially the criterion that the reference standard should be a clinical interview was often not met. In the review of Roy et al. the correlation between sensitivity and specificity was not taken into account and there was no information on the exact thresholds[29]. Therefore, outcomes of the mean sensitivity and specificity from the review of Roy et al.[29] cannot be compared with the pooled outcomes of the current review.
Several reviews evaluated the diagnostic accuracy of depression questionnaires in other populations. A review from 2016 which evaluated the CES-D in the general population[61] reported a higher accuracy for the CES-D at a threshold of 20[61]. Unfortunately, this threshold was not used in any of the studies in this review. Similar to the current review, a meta-analysis from 2015 in the general population concluded that the diagnostic accuracy of the PHQ-9 at a threshold of 10 was better than for the algorithm[62]. However, the pooled specificity (94%) for the algorithm[62] was much higher than in the current review (64.0%). A possible explanation is that symptoms of depression and diabetes overlap, resulting in higher false positive and lower false negative rates at a certain threshold in patients with diabetes compared to people without diabetes. Two reviews in any population found comparable results on sensitivity (77%[62] and 78%[63]) and specificity (85%[62] and 87%[63]) as the current review (sensitivity of 81.5%; specificity of 79.7%).
Strengths and limitations
To the best of our knowledge, this is the first systematic review that included a meta-analysis to evaluate the diagnostic accuracy of depression questionnaires among patients with diabetes type 1 or 2. Furthermore, a standardized tool (i.e. QUADAS-2) was used for the quality assessment and the meta-analysis was based on the Stata manual of Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy. In addition, this systematic review and meta-analysis followed the recent PRISMA-DTA guidelines for transparent reporting.
However, there are some limitations. The number of studies per questionnaire in the meta-analysis was low (maximum of 5) because the included studies in the systematic review reported diagnostic accuracy data at different thresholds. Because of the low number of studies, meta-regression and subgroup analysis with pre-specified variables (i.e. blinding of the reference standard, distribution of diabetes type, percentage of depression cases in the sample and setting) could not be performed. Comparison between diabetes type 1 and type 2 could not be made because only one study included patients with diabetes type 1. Furthermore, the effect of the quality of the studies on the results could not be estimated, since the risk of bias in many studies was unclear in multiple domains. The diagnostic accuracy data could only be pooled at the usual thresholds. Since some symptoms of depression and diabetes overlap, the expectation was that higher thresholds would result in less false positives, and thus a higher specificity. Data about the NPV and PPV are of high value in the clinical setting. However, data about the NPV and PPV was not pooled, because these values are influenced by the prevalence of depression in the study populations.
No external ‘golden standard’ exists for diagnosing depression. A recent review by Petterson et al. suggests that the golden standard for diagnosing depression is the Longitudinal, Expert, All Data (LEAD) procedure in which all available data of a patient is taken into account as basis for diagnosis (i.e. information of family members, hospital records, psychological evaluation and laboratory results)[64]. However, a clinical interview is still the standard for diagnosing in clinical practice and was, therefore, incorporated as inclusion criterion. None of the included studies used the LEAD as reference standard.
The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach could not be applied in the review. The GRADE-approach is a tool for “rating the quality of evidence and move from evidence to a recommendation”[65]. An essential component of formulating a recommendation is the patient-related outcomes of testing positive or negative on a depression questionnaire. These outcomes were not established in the included studies. A study into screening for depression in primary care found that “no trials have found better outcomes among patients who were screened than among patients who were not screened” because of low PPVs and small treatment effects[66]. It should be noticed that the prevalence of depression is higher in patients with diabetes than in the general population[2] which improves the PPV, and effective treatments are available for patients with diabetes[13, 14]. However, the number of false positives among patients with diabetes is still high.
Recent publications on diabetes and depression show the importance of subclinical depression [67] (i.e. clinically relevant depressive symptoms without fulfilling the criteria for major depressive disorder) and of diabetes-related emotional distress [68] (i.e. symptoms of depression and anxiety and disease specific related problems), as relevant constructs associated with increased depressive symptoms in people with diabetes or other comorbid chronic diseases. Depression plays an essential role in the course and prognosis of diabetes and other chronic diseases and must be recognized and treated in an early stage. Yet, we must be aware of the potential negative consequences of screening and diagnosing of patients at risk such as false positive screening results, high costs, additional burden and stigmatization.
Conclusion
This review indicates that the CES-D (≥16) has the highest sensitivity, whereas the PHQ-9 (≥10) shows the highest specificity, yet confidence intervals were wide and overlapping.
Research implications.
The results can aid future researchers to make better decisions in choosing questionnaires for the eligibility of participants in studies with patients with diabetes. The recommendation is to use the PHQ-9 (≥10) or the CES-D (≥16). The CES-D should be evaluated further, since best support was found regarding measurement properties for this questionnaire among patients with diabetes[27]. The PHQ-9 should be incorporated as well because this questionnaire yielded comparable results regarding sensitivity and specificity. Because other questionnaires (e.g. BDI, WHO-5 and HADS) are frequently used in clinical practice[1], these should be evaluated and tested more rigorously in the future. Future research could further estimate the diagnostic accuracy of depression questionnaires in the diabetes population. Focus should be on direct comparison of questionnaires to minimize the effect of bias; the use of higher thresholds to minimize the risk of overlap between symptoms of depression and diabetes; and trials to relate screening to use of screening questionnaires to patient-related outcomes in order to apply the GRADE-approach. The Standards for Reporting Diagnostic accuracy studies (STARD) guidelines help improve completeness of reporting[69].
Clinical implications.
We suggest that the PHQ-9 (≥10) and the CES-D (≥16) are the most useful questionnaires for clinicians for the screening for depression among patients with diabetes. However, ultimately it is for clinicians to make an informed decision with a patient about the use of a depression questionnaire giving the aim, setting, time available and other relevant circumstances.
Supporting information
S3 Table. Search strategy and details of the removal of non-peer reviewed articles and duplicates.
https://doi.org/10.1371/journal.pone.0218512.s003
(DOCX)
S4 Table. Answers on signaling questions of the QUADAS-2 per study.
https://doi.org/10.1371/journal.pone.0218512.s004
(DOCX)
S5 Table. Characteristics of included questionnaires.
https://doi.org/10.1371/journal.pone.0218512.s005
(DOCX)
S6 Table. Extracted data regarding diagnostic accuracy by questionnaire.
https://doi.org/10.1371/journal.pone.0218512.s006
(DOCX)
S1 Text. Risk of bias assessment—signaling questions with interpretations.
https://doi.org/10.1371/journal.pone.0218512.s007
(DOCX)
S1 Fig.
SROC plots of the (A) CES-D (≥16), (B) PHQ-9(≥10) and (C) PHQ-9 algorithm.
https://doi.org/10.1371/journal.pone.0218512.s008
(DOCX)
Acknowledgments
The authors would like to thank Dr. Hella Brandt and Dr. Gerardo Zavala Gomez for the translations of the German and Spanish articles, respectively. We thank Lennart van der Zwaan in screening the abstracts in an early stage and Dr. Caroline Terwee for her feedback in the early stage of conducting this review.
References
- 1. Hermanns N, Caputo S, Dzida G, Khunti K, Meneghini LF, Snoek F. Screening, evaluation and management of depression in people with diabetes in primary care. Prim Care Diabetes. 2013;7(1):1–10. pmid:23280258.
- 2. Roy T, Lloyd CE. Epidemiology of depression and diabetes: A systematic review. J Affect Disord. 2012;142:S8–S21. pmid:23062861.
- 3. Ogurtsova K, da Rocha Fernandes JD, Huang Y, Linnenkamp U, Guariguata L, Cho NH, et al. IDF Diabetes Atlas: Global estimates for the prevalence of diabetes for 2015 and 2040. Diabetes Res Clin Pract. 2017;128:40–50. pmid:28437734.
- 4. Collaborators GDaIIaP. Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. The Lancet. 2016;388(10053):1545–602. https://doi.org/10.1016/S0140-6736(16)31678-6.
- 5. Adriaanse MC, Drewes HW, van der Heide I, Struijs JN, Baan CA. The impact of comorbid chronic conditions on quality of life in type 2 diabetes patients. Qual Life Res. 2016;25(1):175–82. pmid:26267523; PubMed Central PMCID: PMC4706581.
- 6. Gonzalez JS, Peyrot M, McCarl LA, Collins EM, Serpa L, Mimiaga MJ, et al. Depression and Diabetes Treatment Nonadherence: A Meta-Analysis. Diabetes Care. 2008;31(12):2398–403. pmid:19033420
- 7. Pouwer F, Nefs G, Nouwen A. Adverse Effects of Depression on Glycemic Control and Health Outcomes in People with Diabetes: a Review. Endocrinol Metab Clin North Am. 2013;42(3):529–44. pmid:24011885.
- 8. Lustman PJ, Anderson RJ, Freedland KE, de Groot M, Carney RM, Clouse RE. Depression and poor glycemic control: a meta-analytic review of the literature. Diabetes Care. 2000;23(7):934–42. pmid:10895843.
- 9. Bosmans JE, Adriaanse MC. Outpatient costs in pharmaceutically treated diabetes patients with and without a diagnosis of depression in a Dutch primary care setting. BMC Health Serv Res. 2012;12:46. Artn 46 WOS:000301931100001. pmid:22361361
- 10. Molosankwe I, Patel A, Gagliardino JJ, Knapp M, McDaid D. Economic aspects of the association between diabetes and depression: A systematic review. J Affect Disord. 2012;142:S42–S55. WOS:000309577800007. pmid:23062857
- 11. Lin EHB, Rutter CM, Katon W, Heckbert SR, Ciechanowski P, Oliver MM, et al. Depression and Advanced Complications of Diabetes: A prospective cohort study. Diabetes Care. 2010;33(2):264–9. pmid:19933989; PubMed Central PMCID: PMC2809260.
- 12. van Dooren FEP, Nefs G, Schram MT, Verhey FRJ, Denollet J, Pouwer F. Depression and Risk of Mortality in People with Diabetes Mellitus: A Systematic Review and Meta-Analysis. PLoS ONE. 2013;8(3):e57058. pmid:23472075; PubMed Central PMCID: PMC3589463.
- 13. Baumeister H, Hutter N, Bengel J. Psychological and pharmacological interventions for depression in patients with diabetes mellitus and depression. The Cochrane database of systematic reviews. 2012;12:1–75. CD008381. pmid:23235661
- 14. Baumeister H, Hutter N, Bengel J. Psychological and pharmacological interventions for depression in patients with diabetes mellitus: an abridged Cochrane review. Diabetic Medicine. 2014;31(7):773–86. pmid:24673571.
- 15. Pouwer F, Beekman AT, Lubach C, Snoek FJ. Nurses' recognition and registration of depression, anxiety and diabetes-specific emotional problems in outpatients with diabetes mellitus. Patient Educ Couns. 2006;60(2):235–40. pmid:16442465.
- 16. Rubin RR, Ciechanowski P, Egede LE, Lin EHB, Lustman PJ. Recognizing and treating depression in patients with diabetes. Curr Diab Rep. 2004;4(2):119–25. pmid:15035972.
- 17. Coventry PA, Hays R, Dickens C, Bundy C, Garrett C, Cherrington A, et al. Talking about depression: a qualitative study of barriers to managing depression in people with long term conditions in primary care. BMC Fam Pract. 2011;12(1):10. pmid:21426542; PubMed Central PMCID: PMC3070666.
- 18.
Force IDFCGT. Global Guideline for Type 2 Diabetes Brussels: IDF, 2012.
- 19. Association CD. Canadian Diabetes Association 2013 Clinical Practice guidelines for the Prevention and Management of Diabetes in Canada Canadian Journal of Diabetes. 2013;37(A3-A13):S1–S212.
- 20. Association AD. Standard of Medical Care in Diabetes—2017. Diabetes Care. 2017;40:S1–S135. pmid:27979885
- 21. Corathers SD, Kichler J, Jones N-H Y, Houchen A, Jolly M, Morwessel N, et al. Improving Depression Screening for Adolescents With Type 1 Diabetes. Pediatrics. 2013;132(5):e1395–e402. pmid:24127480.
- 22. Lenore Sawyer R. The CES-D Scale: A Self-Report Depression Scale for Research in the General Population. Applied Psychological Measurement. 1977;1(3):385–401.
- 23. Zigmond AS, Snaith RP. The Hospital Anxiety and Depression Scale. Acta Psychiatr Scand. 1983;67(6):361–70. pmid:6880820.
- 24. Kroenke K, Spitzer RL. The PHQ-9: A New Depression Diagnostic and Severity Measure. Psychiatric Annals. 2002;32(9):509–15. WOS:000178070800004.
- 25. Beck AT, Beck RW. Screening Depressed Patients in Family Practice. Postgraduate Medicine. 1972;52(6):81–5. pmid:4635613.
- 26. Zung WWK. A Self-Rating Depression Scale. Arch Gen Psychiat. 1965;12(1):63–70. WOS:A1965CCC4600008.
- 27. van Dijk SEM, Adriaanse MC, van der Zwaan L, Bosmans JE, van Marwijk HWJ, van Tulder MW, et al. Measurement properties of depression questionnaires in patients with diabetes: a systematic review. Qual Life Res. 2018;27(6):1415–30. pmid:29396653; PubMed Central PMCID: PMC5951879.
- 28. Leeflang MMG, Deeks JJ, Takwoingi Y, Macaskill P. Cochrane diagnostic test accuracy reviews. Syst Rev. 2013;2:82–8. pmid:24099098; PubMed Central PMCID: PMC3851548.
- 29. Roy T, Lloyd CE, Pouwer F, Holt RI, Sartorius N. Screening tools used for measuring depression among people with Type 1 and Type 2 diabetes: a systematic review. Diabet Med. 2012;29(2):164–75. pmid:21824180.
- 30. de Joode W, Van Dijk S, Walburg F, Bosmans J, van Marwijk H, van Tulder M, et al. Diagnostic accuracy of depression questionnaires in adult patients with diabetes: a systematic review and meta-analysis. PROSPERO 2018 CRD42018092950. Available from: http://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD42018092950.
- 31. McInnes MF, Moher D, Thombs BD, McGrath TA, Bossuyt PM,. Group atP-D Preferred Reporting Items for a Systematic Review and Meta-analysis of Diagnostic Test Accuracy Studies: The PRISMA-DTA Statement. JAMA. 2018;319(4):388–96. pmid:29362800.
- 32. Beynon R, Leeflang MM, McDonald S, Eisinga A, Mitchell RL, Whiting P, et al. Search strategies to identify diagnostic accuracy studies in MEDLINE and EMBASE. The Cochrane database of systematic reviews. 2013;(9):Mr000022. Epub 2013/09/12. pmid:24022476.
- 33. Topp CW, Østergaard SD, Søndergaard S, Bech P. The WHO-5 Well-Being Index: A Systematic Review of the Literature. Psychother Psychosom. 2015;84(3):167–76. pmid:25831962.
- 34. Bramer WM, Giustini D, de Jonge GB, Holland L, Bekhuis T. De-duplication of database search results for systematic reviews in EndNote. J Med Libr Assoc. 2016;104(3):240–3. PMC4915647. pmid:27366130
- 35.
Grobbee DE, Hoes AW. Clinical Epidemiology: Principles, Methods, and Applications for Clinical Research. Second ed. Burlington: Jones & Bartlett Learning; 2015.
- 36. Whiting PF, Rutjes AWS, Westwood ME, Mallet S, Deeks JJ, Reitsma JB, et al. QUADAS-2: A Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies. Ann Intern Med. 2011;155(8):529–36. pmid:22007046.
- 37. Reitsma JB, Glas AS, Rutjes AWS, Scholten RJPM, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol. 2005;58(10):982–90. pmid:16168343.
- 38.
Meta-analysis of test accuracy studies in Stata [Internet]. Cochrane Methods: Screening and Diagnostic Tests: Version 1.1. 2016.
- 39.
Macaskill P, Gatsonis C, Deeks JJ, Harbord RM, Takwoingi Y. Chapter 10 Analysing and Presenting Results. Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy. 1.0 ed: Cochrane Collaboration; 2010.
- 40. Ali N, Jyotsna VP, Kumar N, Mani K. Prevalence of depression among type 2 diabetes compared to healthy non diabetic controls. The Journal of the Association of Physicians of India. 2013;61(9):619–21. Epub 2014/04/30. pmid:24772698.
- 41. Awata S, Bech P, Yoshida S, Hirai M, Suzuki S, Yamashita M, et al. Reliability and validity of the Japanese version of the World Health Organization-Five Well-Being Index in the context of detecting depression in diabetic patients. Psychiatry and clinical neurosciences. 2007;61(1):112–9. Epub 2007/01/24. pmid:17239048.
- 42. Diaz-Rodriguez G, Reyes-Morales H, Lopez-Caudana AE, Caraveo-Anduaga J, Atrian-Salazar ML. [Validation of a clinimetric scale for the diagnosis for depression in patients with diabetes mellitus type 2, in primary health care]. Revista de investigacion clinica; organo del Hospital de Enfermedades de la Nutricion. 2006;58(5):432–40. Epub 2007/04/06. PubMed PMID: 17408103.
- 43. Fisher L, Skaff MM, Mullan JT, Arean P, Mohr D, Masharani U, et al. Clinical Depression Versus Distress Among Patients With Type 2 Diabetes. Diabetes Care. 2007;30(3):542–8. WOS:000244941200014. pmid:17327318
- 44. Fisher L, Hessler DM, Polonsky WH, Masharani U, Peters AL, Blumer I, et al. Prevalence of depression in Type 1 diabetes and the problem of over-diagnosis. Diabet Med. 2016;33(11):1590–7. Epub 2016/10/18. pmid:26433004.
- 45. Hermanns N, Kulzer B, Krichbaum M, Kubiak T, Haak T. How to screen for depression and emotional problems in patients with diabetes: comparison of screening characteristics of depression questionnaires, measurement of diabetes-specific emotional problems and standard clinical assessment. Diabetologia. 2006;49(3):469–77. pmid:16432706.
- 46. Hsu LF, Kao CC, Wang MY, Chang CJ, Tsai PS. Psychometric testing of a Mandarin Chinese Version of the Clinically Useful Depression Outcome Scale for patients diagnosed with type 2 diabetes mellitus. International journal of nursing studies. 2014;51(12):1595–604. Epub 2014/06/22. pmid:24951085.
- 47. Hyphantis T, Kotsis K, Kroenke K, Paika V, Constantopoulos S, Drosos AA, et al. Lower PHQ-9 cutpoint accurately diagnosed depression in people with long-term conditions attending the Accident and Emergency Department. J Affect Disord. 2015;176:155–63. Epub 2015/02/28. pmid:25721612.
- 48. Janssen EP, Kohler S, Stehouwer CD, Schaper NC, Dagnelie PC, Sep SJ, et al. The Patient Health Questionnaire-9 as a Screening Tool for Depression in Individuals with Type 2 Diabetes Mellitus: The Maastricht Study. J Am Geriatr Soc. 2016;64(11):e201–e6. Epub 2016/10/27. pmid:27783384.
- 49. Khamseh ME, Baradaran HR, Javanbakht A, Mirghorbani M, Yadollahi Z, Malek M. Comparison of the CES-D and PHQ-9 depression scales in people with type 2 diabetes in Tehran, Iran. BMC Psychiatry. 2011;11:61. Epub 2011/04/19. pmid:21496289; PubMed Central PMCID: PMC3102614.
- 50. Krille S, Kulzer B, Reinecker H, Haak T, Hermanns N. Einflüsse von Psyche und Verhalten auf den Krankheitsverlauf (F54) bei Diabetes mellitus: Prävalenz und Screeningmethoden. Verhaltenstherapie & Verhaltensmedizin. 2008;29(4):323–35.
- 51. Lamers F, Jonkers CC, Bosma H, Penninx BW, Knottnerus JA, van Eijk JT. Summed score of the Patient Health Questionnaire-9 was a reliable and valid method for depression screening in chronically ill elderly patients. J Clin Epidemiol. 2008;61(7):679–87. Epub 2008/06/10. pmid:18538262.
- 52. Lustman PJ, Clouse RE, Griffith LS, Carney RM, Freedland KE. Screening for depression in diabetes using the Beck Depression Inventory. Psychosomatic medicine. 1997;59(1):24–31. Epub 1997/01/01. pmid:9021863.
- 53. McHale M, Hendrikz J, Dann F, Kenardy J. Screening for depression in patients with diabetes mellitus. Psychosomatic medicine. 2008;70(8):869–74. Epub 2008/10/10. pmid:18842744.
- 54. Stahl D, Sum CF, Lum SS, Liow PH, Chan YH, Verma S, et al. Screening for depressive symptoms: validation of the center for epidemiologic studies depression scale (CES-D) in a multiethnic group of patients with diabetes in Singapore. Diabetes Care. 2008;31(6):1118–9. Epub 2008/03/14. pmid:18337303.
- 55. Sultan S, Luminet O, Hartemann A. Cognitive and anxiety symptoms in screening for clinical depression in diabetes: a systematic examination of diagnostic performances of the HADS and BDI-SF. J Affect Disord. 2010;123:332–6. Epub 2009/10/29. pmid:19861228.
- 56. Twist K, Stahl D, Amiel SA, Thomas S, Winkley K, Ismail K. Comparison of depressive symptoms in type 2 diabetes using a two-stage survey design. Psychosomatic medicine. 2013;75(8):791–7. Epub 2013/08/08. pmid:23922402.
- 57. van Steenbergen-Weijenburg KM, de Vroege L, Ploeger RR, Brals JW, Vloedbeld MG, Veneman TF, et al. Validation of the PHQ-9 as a screening instrument for depression in diabetes patients in specialized outpatient clinics. BMC Health Serv Res. 2010;10(1):235. pmid:20704720; PubMed Central PMCID: PMC2927590.
- 58. Yoshida S, Hirai M, Suzuki S, Awata S, Oka Y. Neuropathy is associated with depression independently of health-related quality of life in Japanese patients with diabetes. Psychiatry and clinical neurosciences. 2009;63(1):65–72. Epub 2008/12/11. pmid:19067994.
- 59. Zhang Y, Ting R, Lam M, Lam J, Nan H, Yeung R, et al. Measuring depressive symptoms using the Patient Health Questionnaire-9 in Hong Kong Chinese subjects with type 2 diabetes. J Affect Disord. 2013;151(2):660–6. Epub 2013/08/14. pmid:23938133.
- 60. Zhang Y, Ting RZ, Lam MH, Lam SP, Yeung RO, Nan H, et al. Measuring depression with CES-D in Chinese patients with type 2 diabetes: the validity and its comparison to PHQ-9. BMC Psychiatry. 2015;15:198. Epub 2015/08/19. pmid:26281832; PubMed Central PMCID: PMC4538746.
- 61. Vilagut G, Forero CG, Barbaglia G, Alonso J. Screening for Depression in the General Population with the Center for Epidemiologic Studies Depression (CES-D): A Systematic Review with Meta-Analysis. PLoS One. 2016;11(5):e0155431. Epub 2016/05/18. pmid:27182821; PubMed Central PMCID: PMC4868329.
- 62. Manea L, Gilbody S, McMillan D. A diagnostic meta-analysis of the Patient Health Questionnaire-9 (PHQ-9) algorithm scoring method as a screen for depression. General hospital psychiatry. 2015;37(1):67–75. Epub 2014/12/03. pmid:25439733.
- 63. Moriarty AS, Gilbody S, McMillan D, Manea L. Screening and case finding for major depressive disorder using the Patient Health Questionnaire (PHQ-9): a meta-analysis. General hospital psychiatry. 2015;37(6):567–76. pmid:26195347.
- 64. Pettersson A, Boström KB, Gustavsson P, Ekselius L. Which instruments to support diagnosis of depression have sufficient accuracy? A systematic review. Nord J Psychiatry. 2015;69(7):497–508. pmid:25736983.
- 65. Schunemann HJ, Oxman AD, Brozek J, Glasziou P, Jaeschke R, Vist GE, et al. Grading quality of evidence and strength of recommendations for diagnostic tests and strategies. BMJ (Clinical research ed). 2008;336:1106–10. Epub 2008/05/17. pmid:18483053; PubMed Central PMCID: PMC2386626.
- 66. Thombs BD, Coyne JC, Cuijpers P, de Jonge P, Gilbody S, Ioannidis JPA, et al. Rethinking recommendations for screening for depression in primary care. CMAJ. 2012;184(4):413–8. pmid:21930744; PubMed Central PMCID: PMC3291670.
- 67. Davidson SK, Harris MG, Dowrick CF, Wachtler CA, Pirkis J, Gunn JM. Mental health interventions and future major depression among primary care patients with subthreshold depression. Journal of Affective Disorders. 2015;177:65–73. pmid:25745837
- 68. Chew BH, Vos RC, Metzendorf MI, Scholten R, Rutten G. Psychological interventions for diabetes‐related distress in adults with type 2 diabetes mellitus. Cochrane Database of Systematic Reviews. 2017;(9). PubMed PMID: CD011469
- 69. Cohen JF, Korevaar DA, Altman DG, Bruns DE, Gatsonis CA, Hooft L, et al. STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration. BMJ Open. 2016;6(11):e012799. PMC5128957; PubMed Central PMCID: PMC5128957. pmid:28137831