Quantifying the impact of chronic conditions on a diagnosis of major depressive disorder in adults: a cohort study using linked electronic medical records

Major depressive disorder (MDD) is highly prevalent, both in the US and worldwide. Among US adults, the estimated 12-month and lifetime prevalence rates are 8.3 and 19.2 %, respectively, and MDD is twice as common in women as in men [1, 2]. The World Health Organization (WHO) considers MDD to be the third-highest cause of disease burden world-wide, and the highest cause of disease burden in the developed world [3]. MDD is associated with declines in overall health that are equivalent to cardiovascular disease, diabetes, and arthritis, and the presence of chronic disease with MDD contributes to greater disease burden and cost to society than that of each illness alone [4].

Comorbidity refers to the presence of at least one extra chronic disease along with a chronic disease of interest, and is known to be common at all ages, but especially among the elderly. Several studies have reported high rates of MDD co-occurrence with many chronic somatic conditions including cardiovascular diseases, cancer, and diabetes [5‐8]. MDD patients with comorbid chronic somatic conditions tend to have worse depression outcomes than patients without such comorbidities, including more persistent depressive symptoms, higher relapse rates, longer time to recovery [9‐11], and greater utilization of medical resources [12]. These observations may be due to negative reciprocal interactions between MDD, depressive symptoms, and chronic somatic illness. Pain and functional impairment associated with chronic conditions may increase the risk of MDD, but MDD may also exacerbate the pain and distress associated with somatic illnesses (NICE guidelines [CG91] Published date: October 2009), suggesting a bi-directional relationship between MDD and chronic somatic illness.

The majority of studies examining the relationship between MDD and chronic conditions have focused on assessing MDD prevalence for a given somatic condition, which does not permit the examination of specific chronic conditions as risk factors for MDD in the context of other co-occurring physical illnesses. Consideration of the comparative contributions of specific chronic somatic conditions to MDD risk in the context of all co-occurring conditions is important, given that patients with chronic illnesses tend to have more than one such condition, particularly if they are depressed [8].

The age of MDD onset is highly variable. Community surveys have suggested that incidence peaks occur during ones’ 20’s and 50’s [13], although incident MDD can first present at virtually any point across lifespan [14]. More recent data suggest that a substantial number of cases present very early in life, i.e., during childhood or adolescence [15]. In addition to MDD onset, age is also strongly associated with the types of commonly occurring chronic conditions [16]. On that basis, it is possible that patterns of co-occurring conditions will vary in MDD patients depending on the age of MDD onset.

In this study, we aimed to investigate the association of selected commonly-occurring chronic conditions on the risk of MDD onset depending on age, and to quantify the relative importance of each chronic condition on MDD risk in the presence of all remaining physical conditions, rather than examining each condition separately. In doing so, we utilized electronic health record (EHR) data for ascertaining MDD disease status and co-occurring chronic conditions, offering the advantage of studying a large population of over 11,000 MDD cases and an equal number of patients without MDD. We hypothesized that the somatic conditions with the greatest relative importance for MDD differ according to age.

Epidemiological studies have documented strong and significant associations between prevalent MDD and several somatic health conditions [20, 26‐30]. Most of this literature has consisted of cross-sectional studies, or of longitudinal studies focused primarily on single chronic condition pairings with MDD. The results of cross-sectional studies are limited in that documented associations between MDD and chronic conditions could reflect any of a number of combinations including (1) the incidence of chronic conditions in patients with MDD, (2) the incidence of MDD in people with specific health conditions, (3) the effect of MDD and specific conditions on the prognosis and mortality associated with one another [26]. Studies focused on MDD associations with single chronic health conditions are also limited in that the relative influence of the conditions of interest on the risk of MDD are often not considered in the context of other comorbid physical health conditions. Consistent with the current literature, we observed that almost all chronic conditions considered showed some level of association with MDD risk when considered individually, even after accounting for educational attainment. When considering all chronic conditions simultaneously, we found that the greatest relative influence for MDD risk was observed for diabetes mellitus among subjects aged 60 years or younger and RA/OA for subjects over 60 years. We also observed that the total disease burden (measured as the number of chronic conditions) was the greatest contributor for MDD risk, followed by socioeconomic status (using education attainment as a surrogate). However, this study focused on identifying individual chronic conditions with the greatest influence on MDD risk, as it is difficult to implement clinical recommendations based on general disease burden without also targeting co-occurring chronic somatic conditions individually.

Although no single chronic condition had a very large effect on the risk of MDD diagnosis, our findings support existing recommendations for MDD screening when adequate resources for follow-up diagnosis and treatment are available, and highlight the importance of collaborative depression care, especially for patients who are diagnosed with the physical health conditions that may have greater relative importance in terms of MDD risk. This is particularly important in light of evidence that most patients with chronic medical conditions do not receive proper diagnosis and treatment of psychiatric comorbidities [30].

In our study, diabetes had the greatest relative contribution to depression in non-elderly adults even after accounting for obesity. Epidemiological studies have long documented an association between depression and higher rates of type 1 and type 2 diabetes mellitus in adults. In an influential meta-analysis of 20 controlled epidemiological studies, the risk of depression was twice as high among persons with prevalent diabetes relative to nondiabetic controls [31]. The association between depression and diabetes may be particularly strong for persons with type 2 diabetes mellitus, although several prior studies failed to control for several variables associated with increased risk of clinically significant depression [32]. Numerous models have been proposed in an attempt to provide mechanistic explanations for the high rates of depression-diabetes co-occurrence. These include negative lifestyle factors and higher rates of overweight and obesity among depressed individuals that increase the risk of new-onset type 2 diabetes, poorer adherence to diabetes treatment among depressed relative to non-depressed persons, and orexigenic effects of some antidepressant medications [33‐36]. Additionally, a number of shared biological mechanisms have been hypothesized, including chronic hypothalamic-pituitary adrenal (HPA) axis dysfunction leading to abnormal glucocorticoid signaling, chronic inflammation, disruption of normal circadian rhythms, and abnormalities in neurotrophin and monamine signaling [37, 38]. The identification of targets for primary prevention of both type 2 diabetes and depressive disorders based on the ascertainment of shared etiological mechanisms is a current research priority [39].

Among subjects aged over 60 years, a significant contribution to the risk of depression was made by RA/OA. As is the case with diabetes, epidemiological studies have documented high rates of depressive disorder comorbidity among patients with RA [40], even after controlling for the overlap in symptomatology between the two conditions [41]. Pain and disability from RA has been proposed as one clinical factor that may increase the risk of comorbid depression, which may in turn worsen clinical outcome related to RA [42]. More recent studies have focused on shared biological factors between depression and RA—specifically, the systemic inflammation that is characteristic of RA and other inflammatory disorders may also have a role in the development of depression, even among persons without RA [43].

Our study has several limitations. First, case ascertainment for MDD and the chronic conditions under consideration was conducted solely based on information from EHRs. Although our validated EHR-based algorithm used multiple data sources (billing codes, clinical notes, and medication use) to identify MDD cases, neither the MDD cases nor the chronic conditions in this study were subject to confirmatory medical record review. We also acknowledge that the information recorded in the medical records may not be accurate or complete enough for a confirmatory diagnosis of MDD. Second, medical records prior to 1994 were not utilized in this study, as they were not available through our EHR system. Therefore, it is possible that a proportion of depression episodes occurred prior to 1994 would have been missed, which potentially impacts our study findings. Third, high rates of missed depression diagnoses in clinical practice have been well supported in literature. Thus, a proportion of controls may have undiagnosed depression, which may have reduced statistical power to distinguish comorbidity patterns between cases and controls. Fourth, diagnostic practice may have changed during the study period (between 2000 and 2013). This could potentially have affected the rates of diagnosis of MDD and co-occurring somatic conditions. Fifth, the severity of MDD was not incorporated in the current study, an important consideration given evidence that MDD severity may correlate with the severity of specific somatic conditions or the number of co-occurring somatic conditions [44‐48]. Sixth, the majority of our study population was Caucasian; thus, our results should be replicated in more diverse populations. As a related point, the mean age of our cohort is younger than some other studies of multimorbidity in primary care populations [49]. For most patients, MDD has a relatively early age of onset, typically less than 40 years of age [13, 15]. Therefore, persons in this age range seeking evaluation and treatment for MDD may be expected to present to their primary care physicians. This seems particularly likely if they have existing chronic medical conditions that are already being managed by their primary care physicians—a condition that would seem to apply to our study population. And finally, although education level has been used in prior research as a proxy for socioeconomic status [50], we were unable to adjust our results for more direct measures of socioeconomic status.

 Our results suggest that specific chronic conditions such as diabetes mellitus and rheumatoid arthritis/osteoarthritis may have greater influence than others on the risk of MDD. Ethics approval and consent to participate.

This study utilizing secondary EHR data was reviewed and approved by the Mayo Clinic Institutional Review Board (IRB).