Expert Review of Cardiovascular Therapy Downloaded from informahealthcare.com by Nyu Medical Center on 07/05/15 For personal use only.
Original Research
Incident depression increases medical utilization in Medicaid patients with hypertension Expert Rev. Cardiovasc. Ther. 13(1), 111–118 (2015)
Ian Michael Breunig1, Fadia T Shaya*1, Justin Tevie1 and David Roffman2 1 Pharmaceutical Health Services Research, University of Maryland School of Pharmacy, 220 Arch St, 12th Floor, Baltimore, MD 21201, USA 2 Department of Pharmacy Practice and Science, University of Maryland School of Pharmacy, 20 N. Pine Street, Room N423, Baltimore, MD 21201, USA *Author for correspondence: Tel.: +1 410 706 5392 Fax: +1 410 706 5394 [email protected]
Hypertension is an important risk factor for cardiovascular disease and occurs disproportionately among patients with depression. Few studies have rigorously examined outcomes specifically among hypertensive patients with newly diagnosed comorbid depression. Aim: We hypothesized that incident depression would exacerbate hypertensive disease and that this would be evident through greater utilization of medical services than would otherwise occur in the absence of depression. Methods: Claims data for hypertensive patients enrolled in Maryland Medicaid (2005–2010) were used to estimate the change in annualized utilization following incident depression, compared to a matched cohort of hypertensive patients never diagnosed with depression. Multivariate regression was used to adjust for changes in antihypertensive medications, adherence and comorbidity that followed depression onset. Results: While medical utilization increased after incident depression, additional encounters tended to be for nonacute medical care and there was no significant increase in encounters specifically for cardiovascular or hypertension-related conditions. Discussion: The results contribute to the discussion on the relationship between depression and cardiovascular disease and will inform future studies that aim to look at longer term outcomes in patients with hypertension. KEYWORDS: depression • hypertension • Medicaid • medical utilization
Comorbid depression is common in cardiovascular disease (CVD) [1]. Research to date has demonstrated that depression is a risk factor for coronary heart disease, stroke, peripheral artery disease and heart failure, and that it is associated with greater morbidity, mortality and utilization of health care resources when concurrent with these conditions [2–9]. A growing body of literature has put forth plausible biological mechanisms linking depression to these conditions (e.g., reduced heart rate variability, endothelial dysfunction, inflammatory processes, platelet function, hyperactivity of hypothalamic–pituitary–adrenal axis, increased catecholamine levels) but evidence in support of one or more mechanisms is largely inconclusive [3–7,10]. It is more certain that the most critical mediators of the relationship are modifiable behavioral factors, such as smoking, physical inactivity, obesity and medication nonadherence [3–6,8,9]. informahealthcare.com
10.1586/14779072.2014.969712
Hypertension is one important risk factor for more severe manifestations of CVD and it occurs disproportionately among patients with depression [6,8,11]. Yet, it remains unclear whether hypertension is simply coprevalent with depression and other modifiable risk factors for CVD, or whether depression may potentiate hypertension [5,6,12]. The evidence is mixed regarding a physiological pathway linking depression with blood pressure, but it is theorized that hyperactivity of the autonomic nervous system in patients with depression has a pressor effect on the cardiovascular system [5,11]. Inflammation may be a common etiological link between hypertension and depression. An increase in inflammatory markers has been shown to be positively associated with both essential hypertension [13] and the severity of depressive symptoms [14–16]. Much of the evidence linking depression with complications of hypertension suggests that
2015 Informa UK Ltd
ISSN 1477-9072
111
Expert Review of Cardiovascular Therapy Downloaded from informahealthcare.com by Nyu Medical Center on 07/05/15 For personal use only.
Original Research
Breunig, Roffman, Tevie & Shaya
depression impairs the management of hypertension [11]. Depression has been shown to more than double the risk of nonadherence to medication and medical advice [17–20]. Some studies have clearly shown nonadherence to antihypertensive medication to be associated with depression [17,20–24] but other studies fail to find this association [25–27]. Antidepressant medications, particularly selective serotonin reuptake inhibitors, have been shown to be safe and effective in treating depression in patients with CVD, but some antidepressant medications may have adverse side effects that increase systolic and diastolic blood pressure or the risk for heart disease [3,28,29]. Studies have also shown that patients with psychiatric disorders may experience greater barriers to quality care although sometimes reflective of patient characteristics and other times reflective of the health care system [9,30–37]. Inconsistent study designs and heterogeneous study populations across studies make it difficult to discern any causal link [20]. This study aims to contribute to the understanding of the added burden of depression on hypertensive CVD. We hypothesized that comorbid depression would exacerbate hypertensive CVD and that this would be evident through greater utilization of medical services than would otherwise occur in the absence of depression. One study conducted a cross-sectional analysis of over 1.2 million Medicare beneficiaries in 1999 and found that hypertensive patients with depression were 2.8-times more likely to visit the emergency department and 3.5-times more likely to be hospitalized [38]. A cross-sectional analyses, however, could not distinguish whether greater utilization by depressed patients was directly attributed to depression or whether depression was perhaps more likely to manifest itself as more severe hypertensive disease. Few studies have rigorously examined outcomes among hypertensive patients with newly diagnosed comorbid depression. We tested our hypothesis by using claims data from a U.S. state Medicaid population to determine whether changes in the annual rates of medical utilization by patients diagnosed with hypertension tend to occur after incident depression. We further assessed whether changes observed after incident depression were different from changes in annual rates of medical utilization observed among a randomly matched sample of patients diagnosed with hypertension alone. Because depression might also influence medical utilization rates in hypertensive patients by aggravating coprevalent conditions, we examined the annual frequency of all medical encounters as well as encounters specifically related to CVD and, specifically, to hypertension. Methods Study population
A retrospective cohort analysis was conducted within a Medicaid population diagnosed with essential hypertension between 1, July 2005 and 30 June 2010. The Maryland Medicaid database retains all claims for medical inpatient, outpatient and pharmacy services submitted to the state for beneficiaries enrolled in fee-for-service and managed care plans. Claims for medical services identified up to three International 112
Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes with first and last dates of service as well as information on age, gender, race/ethnicity and county of residence. Pharmacy claims were used to identify drug class defined by the American Hospital Formulary Service (AHFS) codes, dates dispensed and days supplied. We identified a cohort of patients, aged 18–64 years, first diagnosed with depression (ICD-9: 296.2, 296.3, 298.0, 309.1, 311) or anxiety associated with depression (ICD-9: 300.4) at least 1 year after a hypertension diagnosis (ICD-9: 401) was first observed, and with no record of antidepressants (AHFS: 28.16.04) dispensed before that time. Depression was defined as a primary diagnosis on an inpatient or outpatient claim, a secondary or tertiary diagnosis on an inpatient claim or at least two diagnoses within a 12-month period on any claims. Patients were excluded if there was any gap in Medicaid enrollment during the year before (‘Year 1’) or after (‘Year 2’) the first diagnosis of depression (‘index date’), or if they spent more than 30 cumulative days in long-term care within the 2-year period. The remaining patients were retained if they had at least one pharmacy claim for an antihypertensive medication in Years 1 and 2. Antihypertensive medications include those of the following drug classes: a-blockers, b-blockers, calcium channel blockers, central a-agonists, direct vasodilators, diuretics and renin-angiotensin-aldosterone system (RAAS) inhibitors. We also retained patients who were observed for at least 2 years of continuous Medicaid enrollment after a hypertension diagnosis (ICD-9: 401) was first observed, and who had no record of depression or antidepressant use for any time observed in the Maryland Medicaid database. These cases were eligible for a hypertensive control cohort, matched to the depression cohort using the process described below, and used to control for secular trends in medical utilization. Index dates were randomly generated for the potential control cohort. Patients with at least one pharmacy claim for an antihypertensive medication in Years 1 and 2 and fewer than 31 days of cumulative long-term care in the 2-year period were retained. Study design
We estimated the effect of incident depression on annual rates of medical utilization among patients diagnosed with hypertension using a semiparametric difference-in-difference analysis [39]; a quasi-experimental approach to estimating average effects from observational studies since it mitigates biases in ‘treatment-control’ group comparisons arising from permanent differences between the groups, or biases in pre–post comparisons resulting from secular trends unrelated to the ‘treatment.’ Each patient’s utilization rate after onset of depression is contrasted with his or her own utilization rate before onset of depression to mitigate confounding due to heterogeneity in the underlying severity of hypertension as well as other constant unobservable risk factors. Changes in utilization rates after incident depression are compared to changes observed in a cohort of patients diagnosed with hypertension but never depression. Expert Rev. Cardiovasc. Ther. 13(1), (2015)
Expert Review of Cardiovascular Therapy Downloaded from informahealthcare.com by Nyu Medical Center on 07/05/15 For personal use only.
Incident depression increases medical utilization in Medicaid patients with hypertension
Propensity score matching was used to match incident depression cases to hypertension control cases according to observed baseline (Year 1) characteristics potentially associated with hypertension control and severity, and thus medical utilization. These include age (18–24, 25–35, 35–44, 45–54, 55–64), gender, race/ethnicity (white, black, Hispanic, other), indicators for use of each hypertensive medication in the baseline year as well as adherence to those medications and comorbidity status. Many patients filled multiple antihypertensive medications concurrently and an overlapping fraction of the sample had one or more periods without any medication. We measured adherence to medication as the number of days in the year where the patient possessed and presumably took at least one antihypertensive medication (‘number of days treated’). Thus, dispensed dates and days supplied were used to compute the total duration of the sequence of all prescriptions filled in the year [20]. Days on which patients were admitted to inpatient care were included in the days supplied with medication since we assumed that appropriate therapy was being provided by the hospital during a stay. The Charlson Comorbidity Index (CCI) was used as a generic marker for comorbidity [40]. Quan et al.’s [41] enhanced algorithm was used to identify the 17 disease categories included in the CCI using ICD-9 diagnosis codes on medical claims in Year 1. Depression and control cases were also matched on indicators for the presence of selected comorbidities that are not components of the CCI but are either cardiovascular-related, mental-health related or are conditions for which antihypertensive medications are sometimes prescribed (cardiac dysrhythmia, hypertensive heart disease, dyslipidemia, other heart diseases, edema, benign prostatic hyperplasia [42], schizophrenia, episodic mood disorder, delusional disorder, other psychoses and alcohol- and drug-induced mental illness). A multivariate logistic propensity score model was used to estimate the probability of incident depression, based on the factors described above. To account for possible contextual factors for onset of depression or medical utilization, the propensity score model also included indicators for patients’ county of residence and their index month and year. Incident depression cases were matched 1-to-1 with control cases using nearest neighbor matching, without replacement and imposing a common support by dropping depression cases whose propensity score was higher (lower) than the maximum (minimum) score of the control cohort. After matching, all observed patient characteristics were balanced between the two cohorts, that is, differences in means were statistically insignificant (p < 0.05). Multivariate regression models were used to estimate the more direct impact of depression on the change in annual medical utilization rates. Regression covariates included indicators for the changes in antihypertensive medication across years, changes in adherence to hypertension medication, changes in CCI and indicators for new comorbidities present after the onset of depression. informahealthcare.com
Original Research
Outcome measures & statistical analysis
The main outcome variable, the annual change in medical utilization, was defined as the change in the annual sum of inpatient and outpatient (hospital setting, emergency room setting, nonhospital setting with nonmental health care providers, nonhospital setting with mental health care providers) encounters for health care services in Years 1 and 2. If more than one encounter was recorded in a day, they were recorded as one event and the type of encounter was categorized in a hierarchical fashion in the order presented above. We also examined the impact of depression onset on CVD-specific utilization and hypertension-specific utilization, defined as medical encounters (as defined above, and not including the routine dispensing of CVD/hypertension-related medication) with a primary diagnosis of any CVD (ICD-9: 390–459) and essential hypertension (ICD-9: 401), respectively. We compared the distributions of comorbidity- and medication-related variables between the two cohorts using the interquartile range for CCI and number of days treated, and frequencies and percentages for the dichotomous variables. Changes in these distributions from Year 1 to Year 2 were compared using t-tests when normally distributed and the Wilcoxon–Mann– Whitney rank sum test otherwise. All measured changes in the annual frequencies of medical encounters were normally distributed. Thus, multivariate ordinary least squares regressions with robust variances were conducted to estimate the impact of incident depression on annualized medical utilization adjusted for newly diagnosed comorbidity in Year 2, and concurrent changes in medication regimen or medication adherence. Results
The total study sample meeting the inclusion criteria included 2517 hypertensive patients with incident depression and 14,495 hypertensive patients free of recorded depression. After matching, the depression and control cohorts each consisted of 2511 patients. The total matched sample was 28% male, 63% African American, 32% white, 1% Hispanic and 4% other race/ethnicity and 2% were aged 18–24 years, 8% aged 25–34 years, 26% aged 35–44 years, 42% aged 45–54 years and 22% aged 55–64 years. TABLE 1 shows the baseline levels of medication use and comorbidity status as well as changes after the index date. If left unaccounted for, some of the changes in medical utilization associated with depression might have be attributed to worse underlying health or changes in therapy use that were associated with the onset of depression. The annualized number of days treated increased for the control group and after incident depression (mean change = +10 days vs +15 days; p = 0.050). Antihypertensive therapy was more likely to be changed by adding or switching to central a-agonists (p = 0.016) and/or direct vasodilators (p = 0.002) after the onset of depression. It was interesting to note that a-agonists which could make depression worse were used more frequently after the diagnosis of depression. The health status of incident depression cases seemed to worsen to a greater extent than control cases. While 113
Original Research
Breunig, Roffman, Tevie & Shaya
Table 1. Comparison of therapy use and comorbidity between hypertensive patients with depression onset and propensity-matched control patients. Year 1
Change from year 1 to year 2
No depression
Depression
274 (150, 342)
273 (150, 342)
No depression
Depression
p-value
10 (92.5)
15 (93.6)
0.050
Therapy use Number of days treated
Expert Review of Cardiovascular Therapy Downloaded from informahealthcare.com by Nyu Medical Center on 07/05/15 For personal use only.
Median (IQR) Mean (SD) Therapy, n (%) a-blockers
39 (1.6)
37 (1.5)
1 (0.04)
8 (0.3)
0.33
b-blockers
990 (39.4)
1035 (41.2)
28 (1.1)
58 (2.3)
0.20
Calcium channel blockers
933 (37.2)
919 (36.6)
52 (2.1)
77 (3.1)
0.30
Central a-agonists
323 (12.9)
336 (13.4)
–21 (–0.8)
21 (0.8)
0.016
Direct vasodilators
268 (10.7)
247 (9.8)
–13 (–0.51)
45 (1.8)
0.002
Diuretics
1311 (52.2)
1318 (52.5)
3 (0.1)
36 (1.4)
0.23
RAAS inhibitors
1627 (64.8)
1612 (64.2)
63 (2.5)
73 (2.9)
0.69
1 (0, 3)
1 (0, 3)
0 (0, 1)
0 (0, 1)
Original Research
Incident depression increases medical utilization in Medicaid patients with hypertension Expert Rev. Cardiovasc. Ther. 13(1), 111–118 (2015)
Ian Michael Breunig1, Fadia T Shaya*1, Justin Tevie1 and David Roffman2 1 Pharmaceutical Health Services Research, University of Maryland School of Pharmacy, 220 Arch St, 12th Floor, Baltimore, MD 21201, USA 2 Department of Pharmacy Practice and Science, University of Maryland School of Pharmacy, 20 N. Pine Street, Room N423, Baltimore, MD 21201, USA *Author for correspondence: Tel.: +1 410 706 5392 Fax: +1 410 706 5394 [email protected]
Hypertension is an important risk factor for cardiovascular disease and occurs disproportionately among patients with depression. Few studies have rigorously examined outcomes specifically among hypertensive patients with newly diagnosed comorbid depression. Aim: We hypothesized that incident depression would exacerbate hypertensive disease and that this would be evident through greater utilization of medical services than would otherwise occur in the absence of depression. Methods: Claims data for hypertensive patients enrolled in Maryland Medicaid (2005–2010) were used to estimate the change in annualized utilization following incident depression, compared to a matched cohort of hypertensive patients never diagnosed with depression. Multivariate regression was used to adjust for changes in antihypertensive medications, adherence and comorbidity that followed depression onset. Results: While medical utilization increased after incident depression, additional encounters tended to be for nonacute medical care and there was no significant increase in encounters specifically for cardiovascular or hypertension-related conditions. Discussion: The results contribute to the discussion on the relationship between depression and cardiovascular disease and will inform future studies that aim to look at longer term outcomes in patients with hypertension. KEYWORDS: depression • hypertension • Medicaid • medical utilization
Comorbid depression is common in cardiovascular disease (CVD) [1]. Research to date has demonstrated that depression is a risk factor for coronary heart disease, stroke, peripheral artery disease and heart failure, and that it is associated with greater morbidity, mortality and utilization of health care resources when concurrent with these conditions [2–9]. A growing body of literature has put forth plausible biological mechanisms linking depression to these conditions (e.g., reduced heart rate variability, endothelial dysfunction, inflammatory processes, platelet function, hyperactivity of hypothalamic–pituitary–adrenal axis, increased catecholamine levels) but evidence in support of one or more mechanisms is largely inconclusive [3–7,10]. It is more certain that the most critical mediators of the relationship are modifiable behavioral factors, such as smoking, physical inactivity, obesity and medication nonadherence [3–6,8,9]. informahealthcare.com
10.1586/14779072.2014.969712
Hypertension is one important risk factor for more severe manifestations of CVD and it occurs disproportionately among patients with depression [6,8,11]. Yet, it remains unclear whether hypertension is simply coprevalent with depression and other modifiable risk factors for CVD, or whether depression may potentiate hypertension [5,6,12]. The evidence is mixed regarding a physiological pathway linking depression with blood pressure, but it is theorized that hyperactivity of the autonomic nervous system in patients with depression has a pressor effect on the cardiovascular system [5,11]. Inflammation may be a common etiological link between hypertension and depression. An increase in inflammatory markers has been shown to be positively associated with both essential hypertension [13] and the severity of depressive symptoms [14–16]. Much of the evidence linking depression with complications of hypertension suggests that
2015 Informa UK Ltd
ISSN 1477-9072
111
Expert Review of Cardiovascular Therapy Downloaded from informahealthcare.com by Nyu Medical Center on 07/05/15 For personal use only.
Original Research
Breunig, Roffman, Tevie & Shaya
depression impairs the management of hypertension [11]. Depression has been shown to more than double the risk of nonadherence to medication and medical advice [17–20]. Some studies have clearly shown nonadherence to antihypertensive medication to be associated with depression [17,20–24] but other studies fail to find this association [25–27]. Antidepressant medications, particularly selective serotonin reuptake inhibitors, have been shown to be safe and effective in treating depression in patients with CVD, but some antidepressant medications may have adverse side effects that increase systolic and diastolic blood pressure or the risk for heart disease [3,28,29]. Studies have also shown that patients with psychiatric disorders may experience greater barriers to quality care although sometimes reflective of patient characteristics and other times reflective of the health care system [9,30–37]. Inconsistent study designs and heterogeneous study populations across studies make it difficult to discern any causal link [20]. This study aims to contribute to the understanding of the added burden of depression on hypertensive CVD. We hypothesized that comorbid depression would exacerbate hypertensive CVD and that this would be evident through greater utilization of medical services than would otherwise occur in the absence of depression. One study conducted a cross-sectional analysis of over 1.2 million Medicare beneficiaries in 1999 and found that hypertensive patients with depression were 2.8-times more likely to visit the emergency department and 3.5-times more likely to be hospitalized [38]. A cross-sectional analyses, however, could not distinguish whether greater utilization by depressed patients was directly attributed to depression or whether depression was perhaps more likely to manifest itself as more severe hypertensive disease. Few studies have rigorously examined outcomes among hypertensive patients with newly diagnosed comorbid depression. We tested our hypothesis by using claims data from a U.S. state Medicaid population to determine whether changes in the annual rates of medical utilization by patients diagnosed with hypertension tend to occur after incident depression. We further assessed whether changes observed after incident depression were different from changes in annual rates of medical utilization observed among a randomly matched sample of patients diagnosed with hypertension alone. Because depression might also influence medical utilization rates in hypertensive patients by aggravating coprevalent conditions, we examined the annual frequency of all medical encounters as well as encounters specifically related to CVD and, specifically, to hypertension. Methods Study population
A retrospective cohort analysis was conducted within a Medicaid population diagnosed with essential hypertension between 1, July 2005 and 30 June 2010. The Maryland Medicaid database retains all claims for medical inpatient, outpatient and pharmacy services submitted to the state for beneficiaries enrolled in fee-for-service and managed care plans. Claims for medical services identified up to three International 112
Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes with first and last dates of service as well as information on age, gender, race/ethnicity and county of residence. Pharmacy claims were used to identify drug class defined by the American Hospital Formulary Service (AHFS) codes, dates dispensed and days supplied. We identified a cohort of patients, aged 18–64 years, first diagnosed with depression (ICD-9: 296.2, 296.3, 298.0, 309.1, 311) or anxiety associated with depression (ICD-9: 300.4) at least 1 year after a hypertension diagnosis (ICD-9: 401) was first observed, and with no record of antidepressants (AHFS: 28.16.04) dispensed before that time. Depression was defined as a primary diagnosis on an inpatient or outpatient claim, a secondary or tertiary diagnosis on an inpatient claim or at least two diagnoses within a 12-month period on any claims. Patients were excluded if there was any gap in Medicaid enrollment during the year before (‘Year 1’) or after (‘Year 2’) the first diagnosis of depression (‘index date’), or if they spent more than 30 cumulative days in long-term care within the 2-year period. The remaining patients were retained if they had at least one pharmacy claim for an antihypertensive medication in Years 1 and 2. Antihypertensive medications include those of the following drug classes: a-blockers, b-blockers, calcium channel blockers, central a-agonists, direct vasodilators, diuretics and renin-angiotensin-aldosterone system (RAAS) inhibitors. We also retained patients who were observed for at least 2 years of continuous Medicaid enrollment after a hypertension diagnosis (ICD-9: 401) was first observed, and who had no record of depression or antidepressant use for any time observed in the Maryland Medicaid database. These cases were eligible for a hypertensive control cohort, matched to the depression cohort using the process described below, and used to control for secular trends in medical utilization. Index dates were randomly generated for the potential control cohort. Patients with at least one pharmacy claim for an antihypertensive medication in Years 1 and 2 and fewer than 31 days of cumulative long-term care in the 2-year period were retained. Study design
We estimated the effect of incident depression on annual rates of medical utilization among patients diagnosed with hypertension using a semiparametric difference-in-difference analysis [39]; a quasi-experimental approach to estimating average effects from observational studies since it mitigates biases in ‘treatment-control’ group comparisons arising from permanent differences between the groups, or biases in pre–post comparisons resulting from secular trends unrelated to the ‘treatment.’ Each patient’s utilization rate after onset of depression is contrasted with his or her own utilization rate before onset of depression to mitigate confounding due to heterogeneity in the underlying severity of hypertension as well as other constant unobservable risk factors. Changes in utilization rates after incident depression are compared to changes observed in a cohort of patients diagnosed with hypertension but never depression. Expert Rev. Cardiovasc. Ther. 13(1), (2015)
Expert Review of Cardiovascular Therapy Downloaded from informahealthcare.com by Nyu Medical Center on 07/05/15 For personal use only.
Incident depression increases medical utilization in Medicaid patients with hypertension
Propensity score matching was used to match incident depression cases to hypertension control cases according to observed baseline (Year 1) characteristics potentially associated with hypertension control and severity, and thus medical utilization. These include age (18–24, 25–35, 35–44, 45–54, 55–64), gender, race/ethnicity (white, black, Hispanic, other), indicators for use of each hypertensive medication in the baseline year as well as adherence to those medications and comorbidity status. Many patients filled multiple antihypertensive medications concurrently and an overlapping fraction of the sample had one or more periods without any medication. We measured adherence to medication as the number of days in the year where the patient possessed and presumably took at least one antihypertensive medication (‘number of days treated’). Thus, dispensed dates and days supplied were used to compute the total duration of the sequence of all prescriptions filled in the year [20]. Days on which patients were admitted to inpatient care were included in the days supplied with medication since we assumed that appropriate therapy was being provided by the hospital during a stay. The Charlson Comorbidity Index (CCI) was used as a generic marker for comorbidity [40]. Quan et al.’s [41] enhanced algorithm was used to identify the 17 disease categories included in the CCI using ICD-9 diagnosis codes on medical claims in Year 1. Depression and control cases were also matched on indicators for the presence of selected comorbidities that are not components of the CCI but are either cardiovascular-related, mental-health related or are conditions for which antihypertensive medications are sometimes prescribed (cardiac dysrhythmia, hypertensive heart disease, dyslipidemia, other heart diseases, edema, benign prostatic hyperplasia [42], schizophrenia, episodic mood disorder, delusional disorder, other psychoses and alcohol- and drug-induced mental illness). A multivariate logistic propensity score model was used to estimate the probability of incident depression, based on the factors described above. To account for possible contextual factors for onset of depression or medical utilization, the propensity score model also included indicators for patients’ county of residence and their index month and year. Incident depression cases were matched 1-to-1 with control cases using nearest neighbor matching, without replacement and imposing a common support by dropping depression cases whose propensity score was higher (lower) than the maximum (minimum) score of the control cohort. After matching, all observed patient characteristics were balanced between the two cohorts, that is, differences in means were statistically insignificant (p < 0.05). Multivariate regression models were used to estimate the more direct impact of depression on the change in annual medical utilization rates. Regression covariates included indicators for the changes in antihypertensive medication across years, changes in adherence to hypertension medication, changes in CCI and indicators for new comorbidities present after the onset of depression. informahealthcare.com
Original Research
Outcome measures & statistical analysis
The main outcome variable, the annual change in medical utilization, was defined as the change in the annual sum of inpatient and outpatient (hospital setting, emergency room setting, nonhospital setting with nonmental health care providers, nonhospital setting with mental health care providers) encounters for health care services in Years 1 and 2. If more than one encounter was recorded in a day, they were recorded as one event and the type of encounter was categorized in a hierarchical fashion in the order presented above. We also examined the impact of depression onset on CVD-specific utilization and hypertension-specific utilization, defined as medical encounters (as defined above, and not including the routine dispensing of CVD/hypertension-related medication) with a primary diagnosis of any CVD (ICD-9: 390–459) and essential hypertension (ICD-9: 401), respectively. We compared the distributions of comorbidity- and medication-related variables between the two cohorts using the interquartile range for CCI and number of days treated, and frequencies and percentages for the dichotomous variables. Changes in these distributions from Year 1 to Year 2 were compared using t-tests when normally distributed and the Wilcoxon–Mann– Whitney rank sum test otherwise. All measured changes in the annual frequencies of medical encounters were normally distributed. Thus, multivariate ordinary least squares regressions with robust variances were conducted to estimate the impact of incident depression on annualized medical utilization adjusted for newly diagnosed comorbidity in Year 2, and concurrent changes in medication regimen or medication adherence. Results
The total study sample meeting the inclusion criteria included 2517 hypertensive patients with incident depression and 14,495 hypertensive patients free of recorded depression. After matching, the depression and control cohorts each consisted of 2511 patients. The total matched sample was 28% male, 63% African American, 32% white, 1% Hispanic and 4% other race/ethnicity and 2% were aged 18–24 years, 8% aged 25–34 years, 26% aged 35–44 years, 42% aged 45–54 years and 22% aged 55–64 years. TABLE 1 shows the baseline levels of medication use and comorbidity status as well as changes after the index date. If left unaccounted for, some of the changes in medical utilization associated with depression might have be attributed to worse underlying health or changes in therapy use that were associated with the onset of depression. The annualized number of days treated increased for the control group and after incident depression (mean change = +10 days vs +15 days; p = 0.050). Antihypertensive therapy was more likely to be changed by adding or switching to central a-agonists (p = 0.016) and/or direct vasodilators (p = 0.002) after the onset of depression. It was interesting to note that a-agonists which could make depression worse were used more frequently after the diagnosis of depression. The health status of incident depression cases seemed to worsen to a greater extent than control cases. While 113
Original Research
Breunig, Roffman, Tevie & Shaya
Table 1. Comparison of therapy use and comorbidity between hypertensive patients with depression onset and propensity-matched control patients. Year 1
Change from year 1 to year 2
No depression
Depression
274 (150, 342)
273 (150, 342)
No depression
Depression
p-value
10 (92.5)
15 (93.6)
0.050
Therapy use Number of days treated
Expert Review of Cardiovascular Therapy Downloaded from informahealthcare.com by Nyu Medical Center on 07/05/15 For personal use only.
Median (IQR) Mean (SD) Therapy, n (%) a-blockers
39 (1.6)
37 (1.5)
1 (0.04)
8 (0.3)
0.33
b-blockers
990 (39.4)
1035 (41.2)
28 (1.1)
58 (2.3)
0.20
Calcium channel blockers
933 (37.2)
919 (36.6)
52 (2.1)
77 (3.1)
0.30
Central a-agonists
323 (12.9)
336 (13.4)
–21 (–0.8)
21 (0.8)
0.016
Direct vasodilators
268 (10.7)
247 (9.8)
–13 (–0.51)
45 (1.8)
0.002
Diuretics
1311 (52.2)
1318 (52.5)
3 (0.1)
36 (1.4)
0.23
RAAS inhibitors
1627 (64.8)
1612 (64.2)
63 (2.5)
73 (2.9)
0.69
1 (0, 3)
1 (0, 3)
0 (0, 1)
0 (0, 1)
