Abstract

Objective

Hypertension increases the prevalence of depression to a certain extent and identification and diagnosis of depression frequently pose challenges for clinicians. The study aimed to construct and validate a scoring model predicting the prevalence of depression with hypertension.

Methods

6124 individuals with hypertension were utilized from the 2007 to 2020 National Health and Nutrition Examination Survey database (NHANES), including 645 subjects that were assessed to have depressive symptoms, 390 in the development group and 255 in the validation group. Univariable and multivariable analyses were applied to analyze the impact of each parameter on depression with hypertension, resulting in establishment of a predictive model. Finally, the discriminability, calibration ability, and clinical efficacy of the model were verified for both the derivation set and validation set.

Results

Ten variables comprised this model: age, gender, race, poverty to income ratio (PIR), smoke, sleep hours, exercise, diabetes, congestive heart failure, stroke. The area under the receiver operating characteristic curve for the derivation and validating sets was 0.790 and 0.723, respectively, which showed excellent discriminability. The model also fitted well with the actual prevalence of depression with hypertension in calibration and decision curve analysis (DCA) demonstrated that the depression model was practically useful.

Conclusion

This scoring model may provide an additional perspective for evaluating the underlying risk factors of depression for hypertensive individuals.

Introduction

Hypertension has emerged as a significant global public health concern, exhibiting a prevalence among adults in the United States that fluctuates between approximately 44% and 49% [1]. It continues to represent a substantial contributor to cardiovascular morbidity and mortality, underscoring the imperative of blood pressure regulation [2]. Concurrently, depression, a pervasive global mental health disorder, is recognized as a primary risk factor for mental health-related disability on a global scale, accompanied by an elevated susceptibility to suicide [3]. Notwithstanding the availability of multiple depression-screening tools, the identification and diagnosis of depression frequently pose challenges for clinicians due to the condition’s diverse presentations, concealed symptoms, and the somewhat subjective nature of assessment [4]. According to pertinent studies, a positive correlation exists between hypertension and depression [5,6,7]. Evidence indicates that the prevalence of depression among people with hypertension ranges from 26 to 42%, surpassing rates observed in those without hypertension [8, 9]. Furthermore, depression, as a psychological comorbidity, is associated with adverse cardiovascular outcomes and contributes to an increased likelihood of developing hypertension to a certain extent [10, 11].

Despite the availability of medical interventions for hypertension accompanied by depression, the failure to timely diagnose depression in hypertensive individuals carries significant harm. Depressive symptoms are presumed to impede medication adherence, thereby hindering the control of high blood pressure and exacerbating hypertension [12]. Consequently, it is imperative to proactively identify hypertensive individuals at risk of depression and implement preventive measures.

To date, no studies have precisely predicted depression in individuals with hypertension. Accurate prediction of depression incidence in hypertensive individuals enables the timely administration of medical or psychological interventions, thereby enhancing patients’ willingness to undergo high blood pressure treatment and improving overall quality of life and life expectancy. Therefore, the construction of a nomogram to predict the risk of depression in hypertensive individuals is deemed necessary. This study aimed to construct and validate a scoring model predicting the prevalence of depression with hypertension.

Patients and methods

Study design and participants

The National Health and Nutrition Examination Survey (NHANES) is a series of national surveys to evaluate the health status of US citizens with a complex, stratified, multistage, and probability sampling method. Details about the surveys index are available at wwwn.cdc.gov/nchs/nhanes. We abstracted the data from the 2007 to 2020 NHANES surveys into this study. The studies involving human participants were reviewed and approved by the National Center for Health Statistics Research Ethics Review Board (protocol number: 2021-05). The participants provided written informed consent to participate in this study.

Hypertension was determined by any of the following conditions: (1) those with diagnosed hypertension history. (2) those with non-same-day randomized records of 3 systolic blood pressure levels ≥ 130Ìýmm Hg or diastolic blood pressure ≥ 80Ìýmm Hg. (3) those with taking antihypertensive drugs. Blood pressure determinations were taken in the mobile examination center (MEC) and during home examinations using a mercury sphygmomanometer. Blood pressure measurements were taken by one of the MEC examiners. A total of 66,148 participants were selected from the 2007–2020 NHANES, in which 17,433 adults with a diagnosis of hypertension were retained. 11,309 subjects who lacked information on any variables of confounding factors were excluded, and we ultimately enrolled 6124 participants for final analysis. A total of 3530 subjects from the 2007–2014 survey comprised the development group, and 2594 subjects from the 2015–2020 survey comprised the validation group.

Measurements

The Patient Health Questionnaire-9 (PHQ-9) scale, a nine-item depression-screening tool, rates each item from 0 (not at all) to 3 (nearly every day). The total score, based on responses about the frequency of depressive symptoms in the past 2 weeks, ranges from 0 to 27. We defined a PHQ-9 score of 10 or higher as indicating depression symptoms at the time of the survey [13].

We included as covariates confounding factors that may be involved in the occurrence of depression in people with hypertension. Among these, demographic characteristics were obtained via a self-reported questionnaire, including age, gender, race, education, and marriage status. Besides, information on clinical characteristics was collected, including diabetes, stroke, congestive heart failure, coronary heart disease as well as cancer. The height and weight were self-reported and BMI was calculated as weight in kilograms divided by height in meters squared (kg/m²). Lifestyles were also included. Drinking status was based on self-report in the past 12 months, which was divided into never, 1 to 4 times a month, 5 to 8 times a month, 9 to 12 times a month, 13 to 16 times a month and more than 16 times a month. Smoking status was self-reported, classifying participants as non-smokers, ex-smokers, or current smokers. Exercise conditions were categorized as no exercise, 10Ìýmin of moderate-intensity exercise for ≤ 3 days per week, and 10Ìýmin of moderate-intensity exercise for > 3 days per week. Sleep hours on the night was categorized into less than 6Ìýh, 6–8Ìýh and more than 8Ìýh. The exercise conditions and sleep hours reflect the usual situation of participants. The poverty to income ratio (PIR) was calculated by dividing family income by the poverty guidelines specific to the survey year and state [14].

Statistical analysis

Statistical analysis was performed by Stata 15.0 software. Data are expressed as the mean ± SD and medians (interquartile range or n [%]). All continuous variables were tested for the normal distribution of data with the Shapiro–Wilk test. Group differences of continuous variables were conducted with the independent samples t-test or the Mann-Whitney U test. The chi-square test was performed to detect group differences of categorial parameters. We used univariate and multivariate logistic regression to estimate the risk of depression in population with hypertension. Multivariable analysis was performed by backward LR regression analysis. A model including age, gender, race, PIR, smoke, sleep hours, exercise, diabetes, congestive heart failure, stroke was fitted based on the results of multivariate logistic regression. The receiver operator characteristic (ROC) curve, Hosmer-Lemeshow test results, Decision curve analysis (DCA) and nomogram were obtained using Stata 15.0 software. A model will be considered to have good discriminability if the area under the ROC curve (AUC) is higher than 0.7. The higher the AUC, the better the discriminating ability of the model. All statistical analyses were two-tailed, and P < 0.05 was considered statistically significant [15].

Results

Participants characteristics

In this study, 6124 eligible subjects were split into a development group (n = 3530) and a validation group (n = 2594). Among the 6124 subjects, 645 were identified as having depressive symptoms, with 390 in the development group and 255 in the validation group. The general characteristics of the development group and validation group are shown in TablesÌý1 and 2, and the comparisons of the general characteristics between development group and validation group are shown in Table S1.

Table 1 Comparison of general characteristics in the development group

Table 2 Comparison of general characteristics in the validation group

Association between candidate predictive variables with depression

The univariable logistic regression analysis were used in the univariable analysis. The results summarized in TableÌý3 show that age, gender, marriage, education, PIR, BMI, drinking, smoking, sleep hours, exercise, diabetes, congestive heart failure and stroke were associated with depressive symptoms (P < 0.05). Further multivariate logistic regression analysis revealed that diabetes, congestive heart disease and current smoking increased the risk of depression. However, age was inversely associated with depressive symptoms. In addition, 6 to 8Ìýh of sleep and > 3 days of moderate-intensity exercise per week reduced the risk of depression.

Table 3 Univariate and multivariate logistic regression analysis of predictors for depression

Establishment of the predictive model

Based on the results of the multivariate regression analysis, we built a full model with ten independent predictive variables and nomogram was developed and presented (Fig.Ìý1). Each variable was assigned a specific score on a rating scale. The scores of each variable were summed and a vertical line was drawn downward at the location of the total score to obtain the predicted probability of depression. Higher total scores indicated a higher probability of depression. By plotting the True Positive Rate against the False Positive Rate at different cut-off points of predicted probabilities, the cut-off value that corresponds to the point maximizing the Youden’s index is 27 points.

Fig. 1

Nomogram predicting the probability of depression in the development group. The cut-off value is 27 points

Accuracy of the predictive model

The ROC analysis of the model was shown in Fig.Ìý2. In the development group, the AUC was 0.790 (95% CI: 0.766–0.814) (Fig.Ìý2A) while in the validation group, the AUC was 0.723 (95% CI: 0.693–0.753), indicating that our model has good discriminability.

Fig. 2

Receiver operating characteristic curve analyses. (A) development group, (B) validation group. AUC, area under the curve, was 0.7897 for the development group, and it was 0.7227 for the validation group

Evaluation of the calibration ability of the predictive model

To evaluate the calibration ability of the predictive model, we used the Hosmer-Lemeshow test to calculate the χ2 for the development group and the validation group. The results showed that the χ2 of the development group was 9.82, and that of the validation group was 16.33; their P values were 0.457 and 0.091, respectively. The P values of both groups were higher than 0.05, indicating that the model accurately predicted depression risk in both groups. The calibration scatter plots are shown in Fig.Ìý3. The x-axis represents the predicted depression risk, and the y-axis represents the actual diagnosed depression. According to the plots, all scattered points fluctuated around the reference line without significant deviation.

Fig. 3

Calibration curve for the nomogram. (A) development group (B) validation group. The scattered points fluctuated around the reference line without significant deviations in the development and validation groups

Evaluation of the clinical efficacy of the predictive model

In this study, the assessment of the predictive model’s clinical efficacy was conducted using Decision Curve Analysis (DCA). The DCA curves representing the two cohorts are depicted in Fig.Ìý4. Within the illustration, the gray line signifies instances where the model predicted, for extreme cases, the absence of depression in individuals with hypertension, resulting in a clinical net benefit of 0. Conversely, the black line, exhibiting a negative slope representing clinical net benefit, indicates that in extreme scenarios, the model predicted the presence of depression in all individuals with hypertension. The dotted line represents the DCA curve specific to the predictive model. As shown in the figure, if the threshold probability of an individual is < 26% in the validation group (and < 60% in development group), screening strategies based on our nomogram resulted in superior net benefit than screen-none or screen-all strategies.

Fig. 4

Decision curve analysis for the nomogram. (A) development group (B) validation group. The gray line indicates that for extreme cases, the model predicted that none of individuals with hypertension had depression and the clinical net benefit was 0. The black line indicates that for extreme cases, the model predicted that all individuals with hypertension had intermediate or high probability, the clinical net benefit is the negative slope. The black dotted line indicates that the model has a clinical net benefit. The dotted line is higher than the gray and black lines, indicating that individuals can benefit from the model

Discussion

In our study, we used data from 6124 individuals with hypertension to develop and validate a nomogram to predict the risk of depression with hypertension based on ten variables determined by a regression model, including age, PIR, gender, race, diabetes, smoke, congestive heart failure, stroke, exercise, and sleep hours. The nomogram-based risk score serves as a supplement to existing depression screening scales to assist in diagnosis of depression.

Strengths and limitations

An inherent strength of the model lies in its exclusive reliance on objective factors, such as demographic characteristics, lifestyle habits, and physical health, while excluding subjective mental and emotional variables. This characteristic enhances the utility of the model in identifying high-risk patients who have yet to manifest depressive symptoms. This study not only revealed chronic underlying medical conditions, such as diabetes and congestive heart failure are related to depression, but also identified the association between lifestyles and risk of depression.

However, several limitations warrant consideration in the interpretation of our findings. First, data accuracy and objectivity may be compromised, for a large proportion of the object factors were self-reported in relevant questionnaires from NHANES. This may cause bias in the results. Second, the depressive symptom was determined using only a purely subjective assessment of mental and emotional experiences, the PHQ-9, and no stratification based on depression severity was undertaken. Third, the model’s establishment in the U.S. population raises concerns regarding its direct applicability to diverse international populations. Validation across a broad sample from different countries is imperative to assess the model’s performance in varied cultural contexts. Fourth, due to the large sample size, significant differences in multiple characteristics can be observed between the development group and the validation group. Furthermore, the cross-sectional design of our study necessitates additional prospective investigations to delineate causal relationships.

Model interpretation

Evidently, this study elucidated that age is negatively correlated with the incidence of depressive mood in hypertensive individuals. Psychologically, some elderly have resilient personalities and better emotion regulation skills. Socially, they may have strong family support and active social circles. In terms of lifestyle, regular routines and simple life concepts contribute [16]. As a result, these factors contribute to depression onset being inversely associated with age.

Women are more than twice as likely as men to develop depression. This was supported by results of other related studies [17], which also agrees with another depression risk prediction model for type 2 diabetes mellitus patients, claiming male comparatively lower susceptibility to depression compared to female counterparts [18]. On the one hand, the reduced female adaptation to psychological problems due to hormonal effects could be the possible justification [19]. Mechanisms underlying sex difference in stress exposure and the development of anxiety also resulted in women’s vulnerability to depression [20]. Notably, reproductive transition phases, a period of vulnerability for female mood disorders, can not be ignored [21]. On the other hand, there is a large proportion of men who suffer from depression remain undiagnosed. It has been reported that men die by suicide 3 to 4 times as women frequently, in spite of less male diagnosis of depression. Gendered processes of socialization make expression of depression vary by gender, resulting in inaccurate ratio of males to females affected depression [22].

Depression is less likely to develop in blacks than in whites, according to predictive models. This result can be supported by other studies [23]. Nonetheless, these results are inconsistent with theories of social stress, which identified that individuals with disadvantaged social statuses are exposed to more stressors, leading to wore mental health than those of advantaged social statuses [24]. There was a view that highlighted selective migration [25]. In this case, healthier immigrants are more likely to migrate from countries with a majority Black population. This accounts for the lower risk of depression among Blacks. However, the specific mechanism of racial differences needs to be further studied.

PIR is a measure of household income after adjusting for inflation and household size. Our study delineated a tight connection between PIR and depression, suggesting that lower levels of household income are associated with a higher likelihood of depression, which has been recognized in other studies [26]. Low income might be indicative of low level of happiness and life satisfaction to a certain extent, leading to mood disorder and high risk of depression [27]. On the other hand, mental health care poses a unique challenge to low-income groups, with chronic stressors linked to socioeconomic challenges. Limited access to providers, poverty, exposure to crime, and poor housing conditions are common upstream factors that impact the risk for anxiety and depression [28].

This study also associated smoking with the prevalence of depression in hypertensive individuals. Smoking leads to an elevated risk of depression in our analysis, which is consistent with other related studies confirming smoking as a risk factor for depression [29]. This is supported by evidence that, once the cessation of smoking was finished, there is evident reduction of depression and anxiety [30]. Further more, not only in participants with hypertension, the effect of smoking on increased risk for developing depression was also verified in diabetic people [18]. Hence quitting smoking reduced the incidence of depression and improved positive mood and quality of life compared to continued smoking.

Impressively, hypertensive individuals with 6–8Ìýh of sleep have the lowest risk of depression whereas those who sleep less than 6Ìýh are more likely to suffer depression. Shorter sleep duration correlated significantly with a higher risk of depression, which has been recognized by other related studies [31]. Indeed, insomnia is a well-known and significant symptom of depression. Meanwhile, association between the sleep duration and risk of depression was supported by a predictive model for depression risk in the U.S. adult population [32].

According to multivariable logistic regression analysis, more than 3 times a week of moderate-intensity exercise decreased the prevalence of depression greatly, aligning with previous related findings [33]. This link could perhaps root in brain structure and functional changes in patients with depression. Studies have revealed that moderate-intensity exercise may affect depressive mood by rebuilding brain structure [34]. It is worth noting that besides the indirect mechanisms, exercise also improves sleep quality in patients with depression, thus alleviating depression [35]. Therefore, exercise is perceived as a precautionary measure and an alternative treatment for depression [33]. With regard to the adults in the U.S., including people with and without hypertension, it is also confirmed that people with a moderate and high risk of depression tend to be more sedentary than those with a low risk and higher levels of exercise indicated a lower probability of depression [32].

This study linked comorbid systems such as diabetes to the incidence of depression, suggesting that diabetes may be a strong predictor of depression, that is, hypertensive individuals with diabetes are more likely to have comorbid depression. One of the potential key mechanisms is microvascular dysfunction caused by diabetes mellitus [36]. Blood vessels in the frontal lobe and subcortical brain regions may regulate emotions in the elderly, and their damage may cause depression [37]. In addition to the effect of blood glucose level on mood, diabetes and hypertension have the characteristics of long course and poor prognosis, which seriously affect the quality of life of patients, increase their economic and psychological burden, and indirectly accelerate the occurrence of depression.

In addition, we found that hypertensive individuals with congestive heart failure were more likely to suffer from depression. This result can be explained from the perspective of the common pathological basis of congestive heart failure and depression, that is, the high reactivity of the sympathoadrenal axis will produce excessive inflammatory mediators and pro-inflammatory factors, affecting the transmission of neurotransmitters represented by 5-hydroxytryptamine, and 5-hydroxytryptamine plays a major role in emotion regulation [38]. In addition, the results were also influenced by the lack of social or family support in some patients with congestive heart failure, differences in sedentary behavior, and the type of medication [39, 40].

The prediction results showed that hypertension combined with stroke greatly increased the risk of depression, suggesting a poor outcome. This result is associated with severe and long-term physical and mental disability in stroke patients, with poor functional outcomes affecting their ability to perform daily activities and quality of life [41,42,43]. However, the economic burden and related psychological factors caused by long-term treatment may cause patients to have anxiety and stigma [44], which makes individuals with hypertension and stroke more likely to suffer from depression.

Use of the model in a clinical context

In clinical practice, the model can be used for screening and early detection of depression in hypertensive individuals. By inputting data such as age, gender, and lifestyle factors, clinicians can obtain a predicted probability of depression. This allows for the identification of high-risk individuals who may not have yet been diagnosed with depression, enabling early intervention. For example, during routine hypertension check-ups, if the model predicts a relatively high probability of depression, further in-depth assessments can be promptly conducted using more detailed psychiatric evaluation tools, rather than waiting for the patient to present with overt depressive symptoms. It also helps in tailoring treatment plans. If a individual is identified as being at high risk of depression, healthcare providers can consider integrated treatment approaches that address both the physical and mental health aspects. The model can also assist in monitoring the effectiveness of treatment. By regularly reassessing the individual and inputting updated information into the model, clinicians can determine if the interventions are having an impact on the patient’s depression risk. If the predicted risk does not decrease as expected, adjustments to the treatment plan can be made in a timely manner. Additionally, the model can assist in resource allocation, ensuring that patients with a higher risk of depression receive appropriate care and follow-up. Overall, it serves as a valuable tool to enhance the management of hypertensive patients and improve their overall health outcomes. We encourage a multidisciplinary approach involving primary care physicians, cardiologists, and mental health professionals. Primary care physicians can use the predictive model and screening tools to identify patients at risk. Cardiologists can then collaborate with mental health professionals for further evaluation and management of patients with co-occurring hypertension and depression. This approach ensures comprehensive care and addresses both the physical and mental health aspects of the patient.

Conclusion

We created a predictive model using self-reported issues, basic demographics, lifestyle choices, physical ailments, and other objective traits that predict the risk of depression in individuals with hypertension. As a supplement to existing methods of monitoring depressive disorder, it helps to screen individuals who are vulnerable to depression and can be effective in alerting physician to the occurrence of depression.