Benchmarking Depressive Symptoms in Cardiac Rehabilitation

Abstract

Purpose:

Depression affects cardiac health and is important to track within cardiac rehabilitation (CR). Using two depression screeners within one sample, we calculated prevalence of baseline depressive symptomology, improvements during CR, and predictors of both.

Methods:

Data were drawn from the University of Vermont Medical Center CR program prospectively collected database. 1,781 patients who attended between January 2011 and July 2019 were included. Two depression screeners (Geriatric Depression Scale-Short Form[GDS-SF] and Patient Health Questionnaire[PHQ-9]) were compared on proportion of the sample categorized with ≥ mild or moderate levels of depressive symptoms (PHQ-9≥5, ≥10; GDS-SF≥6, ≥10). Changes in depressive symptoms by screener were examined within patients who had completed ≥9 sessions of CR. Patient characteristics associated with depressive symptoms at entry, and changes in symptoms were identified.

Results:

Within those who completed ≥9 sessions of CR with exit scores on both screeners (N=1,201), entrance prevalence of ≥mild and ≥moderate depressive symptoms differed by screener (32% and 9% PHQ-9;12% and 3% GDS-SF; Both P <.001). Patients who were younger, female, with lower cardiorespiratory fitness (CRF) scores were more likely to have ≥mild depressive symptoms at entry. Most patients with ≥ mild symptoms decreased severity by ≥one category by exit (PHQ-9=73%; GDS-SF=77%). Non-surgical diagnosis and lower CRF was associated with less improvement in symptoms on the PHQ-9 (Both P <.05).

Conclusion:

Our results provide initial benchmarks of depressive symptoms in CR. They identify younger patients, women, patients with lower CRF, and those with non-surgical diagnosis as higher risk groups for having depressive symptoms or lack of improvement in symptoms.

Condensed Abstract

Establishing entry prevalence and change in depressive symptoms during cardiac rehabilitation is important. The Patient Health Questionnaire-9 identified more patients as having at least mild, and at least moderate, depressive symptoms than the Geriatric Depression Scale-Short Form. Approximately three quarters of patients improved depressive symptom severity by cardiac rehabilit exit.

Depression is more prevalent in cardiac patients than in the general population, and depressive symptom severity is implicated in the development of cardiovascular disease (CVD), increases following acute cardiac events, is associated with worse outcomes following cardiac events, and adversely impacts patient recovery.^1,2,3 Although the mechanisms remain unclear, one hypothesis is that depression interferes with patients’ abilities to comply with secondary prevention recommendations. Following an adverse cardiac event, patients with depression are less likely to cease smoking, initiate a healthy diet, exercise, adhere to medication, and attend cardiac rehabilitation (CR). ^1,3,4

Given the importance of depression in managing heart disease, cardiac patients should be assessed and receive treatment for depressive symptoms. Cardiac rehabilitation programs are particularly well-positioned to screen for depression and implement interventions as necessary. To understand and treat depression in CR, programs need to establish normative values of (i.e., benchmark) the prevalence, impact, and trajectories of depressive symptoms within their patient populations.^5,6

Unfortunately, depression screening is wildly inconsistent. Estimates of programs that screen for depression range from 29-68%, and it is unclear how programs screen and use screening data. ^7,8 These figures can be improved by adhering to the American Association of Cardiovascular and Pulmonary Rehabilitation’s (AACVPR) performance measure guidelines¹⁰.

Starting in 2014, the AACVPR began to require identification of entrance and exit depression, as well as the ratios of patients who change depression severity categories in CR for program certification. As a result, programs that assess depression have an existing framework for standardized depressive symptom benchmarking.^2,10,11

Despite this attempt at standardization, many programs continue to utilize depression screeners that are not specified in these guidelines and have yet to be validated in CR populations. One such popular screener is the Short Form of the Geriatric Depression Scale (GDS-SF).^8,12 Many programs prefer the GDS-SF for its simple format, focus on patients with age-related cognitive decline, and focus on patient quality of life over somatic symptoms.^4,8 However, the GDS-SF is often used across patients of all ages, even though it is designed for geriatric populations. Further, it may under-identify lower levels of symptoms due to the low number of meaningful categories.^8,12 This is concerning, as even subclinical depression affects mortality rates in CR patients.⁸ Given that depression identification in CR is tied to severity category, and that AACVPR guidelines assess depression through categorical benchmarks, it is vital to compare the GDS-SF categorization options against a screener that is well-validated in CR. At current, no study has compared the GDS-SF against an AACVPR endorsed screener in the same CR sample.

Our study had two aims. First, to provide AACVPR performance improvement measure benchmarks on two depression screeners for entry, exit, and change in depressive symptoms in a large CR sample (N = 1,781) across demographic characteristics. Second, to directly compare benchmarks from two GDS-SF cutoff options against those from an AACVPR endorsed screener, the PHQ-9.

METHODS

Data were obtained from a prospectively collected clinical database of patients from the University of Vermont Medical Center Cardiac Rehabilitation Program (UVMMC CR). All records from January 2011 to July 2019 were examined for potential inclusion. If a patient attended CR multiple times only the first instance was included. Analyses of demographics in completers of a clinically important dose of CR (≥9 sessions) ^13,14 against non-completers included those who completed CR with exit scores, and those who did not complete CR but had scores on both the PHQ-9 and the GDS-SF at entry (N = 1,781). Patients were included in analyses of changes in depressive symptoms if they had completed ≥9 sessions and had exit scores on both screeners (N = 1,201). Patients were excluded from these analyses for the following reasons: 499 for not having completed exit assessments, a further 61 for having attended < 9 sessions of CR, and 20 patients for not having scores on both depression screeners.

MEASURES OF DEPRESSION

The PHQ-9 and GDS-SF are self-report depression screeners. Both were administered at the same two time periods, at entrance to CR and then at exit after 4 mo, which corresponds with the general length of a prescription of CR.

The PHQ-9 aligns with DSM-IV criteria for clinical depression, is conducive to estimating symptom severity and encapsulates somatic symptoms of depression.¹⁵ The PHQ-9 has nine items yielding scores ranging from 0-27. Recommended cutoffs for the general population result in 5 categories; 0-4 indicates minimal or no depression, 5-9 mild depression, 10-14 moderate depression, 15-19 moderate-severe depression, and 20-27 severe depression.¹⁶ Recommendations by the American Heart Association (AHA), indicate CR cutoffs of ≥ 10.⁴ However, several studies including large, recent reviews recommend the lower cutoff.^17,18,19 This study utilized both general and AHA recommended cutoffs.

The GDS-SF is a 15 item, yes/no depression screener designed for adults ≥ 63 yr that includes items relating to life-satisfaction.^20,21,22 The GDS-SF stratifies depression into three categories: scores of 0-5 are normal, 6-9 suggests mild-moderate depression symptoms, and ≥ 10 indicates severe symptoms of depression.²² Cutoffs of >5 are the most utilized and are justified by discriminant validity analyses in CR, although justifications for higher cutoffs of ≥ 10 have been presented.^20,23,24,25 To assess both recommendations, a less conservative cutoff of ≥ 6, and a more conservative cutoff of ≥ 10 were utilized.

Changes in depressive symptoms were calculated using AACVPR completion performance guidelines of change in category by recommended screener cutoffs, i.e., normal, mild, severe for CR completers.⁷ Comparisons between GDS-SF and PHQ-9 identification were made between the less conservative mild depression (PHQ-9 ≥ 5, GDS-SF ≥ 6), and also the more conservative moderate depression cutoffs (PHQ-9 ≥ 10, GDS-SF ≥ 10). Improvement was the ratio of those who decreased depressive symptom severity by at least one category at exit out of those who had ≥ mild/moderate symptoms at entry. Worsening of symptoms was calculated as those who increased by at least one depressive symptom severity category over those who were categorized below the most severe depression category at entry.

BENCHMARKING VARIABLES AND DEMOGRAPHICS

Depressive symptom scores were compared across demographic and clinical variables. Older age, female, lower cardiorespiratory fitness (e.g. peak Metabolic Equivalent of Tasks [METs_peak] and peak oxygen uptake during an exercise task [VO_2peak]), higher body mass index (BMI; kg/m²), lower education, surgical intervention, and attending fewer sessions of CR have been positively associated with depression severity and likelihood of cardiac events and were thus included.^{11,16,26,27,28}

Age, sex, and diagnosis were obtained through chart review of hospital records, educational attainment was self-reported through a single-item question, current smoking was measured as ≥ 6 ppm expired carbon monoxide (CoVita Smokerlyzer). Both VO_2peak and METs_peak were assessed through exercise tolerance tests (ETT). Expired gas analysis was measured during the ETT for determination of VO_2peak in ml/kg/min. At entry and exit from CR, patients performed a symptom-limited treadmill ETT until volitional exhaustion or medical findings that warrant termination of the test. Expired gas analysis was measured throughout ETT with an Ultima CPX (Medgraphics). The highest average 30-sec value was considered VO_2peak. METspeak was defined as the workload performed on the last completed stage of ETT.

The BMI was obtained from scale and stadiometer readings and stratified into three categories, normal (18 ≤ BMI < 25 kg/m²), overweight (25 ≤ BMI <30 kg/m²), and obese (BMI ≥ 30 kg/m²). Educational attainment was split into categories of academic achievement: less than high school or GED, high school completion, some college, terminal bachelor level of education, and postgraduate education.

ANALYSIS

Demographic characteristics were examined through Chi-squared tests (χ²) and ANOVA between those who had or had not completed ≥ 9 sessions. This dichotomy was based on previous studies showing that attending 9 sessions is representative of achieving a minimally clinically important dose of CR, and that patients in the highest risk stratification completed < 9 sessions.^13,14 Prior studies of benchmarking in CR have also used a cut-off of 9 sessions.²⁹ All other analyses included those who had attended ≥ 9 sessions and had entry and exit scores on both screeners. Χ² and ANOVA compared depressive symptom prevalence at entry by screener and differences in symptom severity across characteristics. This approach was also used to test whether improvement in depressive symptoms differed by demographics. Missing data were handled through listwise deletion. Consistent with exploratory analyses, α was set to .05 for all analyses.

McNemar χ² tests were used to compare patients’ change in depressive symptoms by the PHQ-9 or GDS-SF, for the entire sample as well as within the subset of the population who were ≥ 65 yr of age for which the GDS-SF should be more sensitive. This age was chosen as it represents the Medicare coverage cutoff, a point that introduces distinct age-related variability in patient care, health outcomes, and participation.³⁰

RESULTS

Table 1 shows characteristics of the sample separated into completers of ≥ 9 sessions with exit screening, and those who completed < 9 sessions. Overall, the sample was 98% white, 1% Middle Eastern, .5% Black, .3% Asian, and .3% reported their race as ‘other’. In the whole sample, mean age was 66 yr ±11 (19-97), mean BMI was 29.8 ± 6.0 kg/m² (15.5-61.6 kg/m²) mean PHQ-9 score was 4.5 ± 4.6 (0-27), and mean GDS-SF score was 3.2 ± 3.0 (0-18).

Table 1.

Characteristics of Sample at Entrance

Demographic	Completed 9+ Sessions with Exit Scores		Completed <9 Sessions	P value	Correlation with Number of Sessions Completed
Age, <65; ≥65 yr				<.001*	.16*
	<65	527 (44)	321 (56)
	≥65	674 (56)	259 (45)
Sex				.002*	.04*
	Male	885 (74)	388 (67)
	Female	316 (26)	192 (33)
Educational attainment				.016*	.12*
	<High School/GED	22 (4)	14 (7)
	High School	153 (25)	78 (34)
	Some College	135 (22)	49 (21)
	Terminal Bachelor’s	155 (25)	43 (20)
	Postgraduate Education	147 (25)	44 (19)
PHQ-9 at entrance				<.001*	−.11*
	None-Mild	821 (68)	332 (58)
	Mild	275 (23)	143 (25)
	Moderate	74 (6)	59 (10)
	Moderate-Severe	24 (2)	25 (4)
	Severe	7 (0.6)	21 (4)
GDS-SF at entrance				<.001*	−.10*
	None	1053 (88)	453 (78)
	Mild-Moderate	107 (9)	79 (14)
	Severe	41 (3)	48 (8)
Objective smoking at entrance				.002*	−.15*
	Does Not Smoke	84 (93)	79 (81)
	Smokes	13 (7)	18 (19)
Diagnosis				0.04*	−.07*
	Nonsurgical	785 (70)	378 (75)
	Surgical	343 (30)	128 (25)
BMI at entrance				<.004*	−.07*
	Normal	230 (19)	86 (16)
	Overweight	497 (41)	193 (35)
	Obese	472 (39)	253 (47)
Continuous Variables	Score		Score	P Value
VO2peak at entrance				.437
	19.5 (6.8)		19.2 (6.3)
METspeak at entrance				.229
	5.9 (2.2)		5.7 (2.0)

^*Denotes a significant difference (P < 0.05) for χ², Pearson Correlation and ANOVA tests.

Abbreviations: BMI, Body Mass Index; GDS-SF, Geriatric Depression Scale Short-Form; METs_peak, Peak Metabolic Equivalent of Tasks Score; PHQ-9, Patient Health Questionnaire-9; VO_2peak, Peak Oxygen Uptake.

Data are presented as (N), % of Category for categorical variables; for continuous variables data are presented as mean ± SD. Pearson Correlations for categorical variables compare first categories against second categories.

Note: Educational attainment, Objective Smoking at Entrance, Diagnosis, and BMI at entrance, contain missing data and thus do not add up to total N = 1781 The patients included in this study had both GDS-SF and PHQ-9 scores and were not excluded for missing values on other variables. 964 patients had no Educational attainment data as the UVMMC CR program began collecting Educational attainment after the onset of depression screening. 147 patients had diagnosis that could not be located through chart review (out of state or network diagnosis), and 50 patients had no BMI scores due to not attending exercise tolerance tests before CR.

Those who attended < 9 sessions of CR were more likely to be younger (P < .001), female (P = .002), have lower educational attainment (P =.016) have a non-surgical diagnosis (P = .04), be obese (P = .004), and have higher levels of depressive symptoms on both the PHQ-9 (P < .001) and GDS-SF (P < .001). Distributions of VO_2peak between completers and non-completers are shown in Figure 1.

Figure 1.

Distributions of VO₂peak scores at entry between completers and non-completers.

Abbreviations: VO₂peak, peak maximal oxygen uptake.

Of all patients, 35% measured by the PHQ-9, and 15% measured by the GDS-SF were identified as having ≥ mild levels of depression. Number of sessions attended was inversely correlated with GDS-SF and PHQ-9 entrance scores (r = −.094, −.113, both P < .001), as well as change in depression on the PHQ-9 (r =−.089, P < .001).

Table 1 shows the sample included in the analyses of change in depression severity throughout CR. In the change sample mean age was 66 yr ±11 (19-97), mean BMI was 29.4 ± 5.5 kg/m² (17.1-56.6 kg/m²) mean PHQ-9 score was 3.8 ± 3.9 (0-22), and mean GDS-SF score was 2.8 ± 2.6 (0-14). The majority were ≥ 65 yr (52%), male (71%), and did not smoke at entrance (89%). Only 4% did not complete high school/GED; the rest were distributed evenly between terminal high school (28%), some college (22%), terminal bachelor’s degree (24%), and postgraduate education (23%).

DEPRESSION PREVALENCE AT ENTRY AND IMPROVEMENT WITHIN THE CHANGE ANALYSIS SAMPLE

Proportions of patients categorized with ≥ mild depressive symptoms, and ≥ moderate symptoms across variables are shown in Table 2 (PHQ-9) and Table 3 (GDS-SF). A minority of patients in the analyses were identified as having ≥ mild or moderate depressive symptoms at entry (PHQ-9: 32%, 9%; GDS-SF: 12%, 3%; mild or greater, and moderate or greater respectively).

Table 2.

Depressive Symptoms at Entrance, and Improvement in Symptoms using the Patient Health Questionnaire-9 (PHQ-9)

PHQ-9 at Entry							PHQ-9 Improvement
Category		Score	Mild or Worse	P Value (Mild or Worse)	Moderate or Worse	P Value (Moderate or Worse)	Score Improvement	Improved by at Least One Category
Overall		3.8 (3.9)	380 (32)		105 (9)		1.7 (3.5)	73
Sex				.024*		.139
	Male	3.6 (3.9)	264 (30)		71 (8)		1.7 (3.5)	75
	Female	4.3 (3.9)	116 (36)		34 (11)		1.7 (3.5)	69
Education				.900		.023*
	<High School/GED	4.5 (5.1)	7 (32)		4 (18)		1.3 (4.7)	86
	High School	4.2 (4.5)	52 (34)		18 (12)		1.8 (4.0)	71
	Some College	3.8 (3.8)	42 (31)		11 (8)		1.8 (3.3)	81
	Terminal Bachelor’s	3.5 (3.1)	47 (30)		5 (3)		1.9 (3.1)	72
	Postgraduate Education	3.3 (3.7)	42 (29)		10 (7)		1.8 (3.5)	86
Diagnosis				.400		.373
	Surgical	4.0 (4.0)	114 (33)		24 (10)		2.2 (3.7)	85*
	Nonsurgical	3.7 (3.9)	241 (31)		65 (8)		1.5 (3.4)	69
Continuous Variable		Correlation with PHQ-9 Score		Mean Difference (Mild or Worse)		Mean Difference (Moderate or Worse)	Correlation with PHQ-9 Improvement
Age		−.07*		1.3 (10.6)		1.3 (10.6)		.04
BMI		.11*		−.40* (6.1)		−.16* (5.6)		−.04
VO2peaka		−.14*		1.7* (6.2)		2.6* (6.4)		.06
METspeaka		−.17*		.60* (2.0)		.90* (2.0)		.07*

^*Denotes a significant difference (P<.05), between groups on proportion of patients with at least mild/moderate depressive symptoms at entry, or significant correlations.

^a.ANOVA Test of mean differences

Abbreviations: BMI, Body Mass Index; METs_peak, Peak Metabolic Equivalent of Tasks Score; PHQ-9, Patient Health Questionnaire-9; VO_2peak, Peak Oxygen Uptake.

Data are presented as N (% of category) for categorical variables for PHQ-9 at entry. For PHQ-9 entrance scores and improvement, data in columns 2 and 7 are presented as mean (SD), and column 8 as % of patients with mild or worse depression scores who improved by one category.

For continuous variables (Age, BMI, VO_2peak, METs_peak) data are presented as Pearson correlations and mean differences (fitness scores of those with < mild/moderate symptoms – scores of those with mild/moderate or worse).

Notes: Denominators for Mild or Worse or Moderate or Worse in columns 3 and 5 were taken from Table 1. These ratios were calculated as the number of patients who completed 9+ sessions with Mild, Moderate or Worse scores with exit depression scores on both screeners / the total number patients in that category who completed 9+ sessions with exit scores (column 2 Table 1)., e.g., 264 males with 9+ sessions and exit scores on both screeners had mild or worse depression scores / 885 males with 9+ sessions had exit scores on both screeners.

Educational attainment, Objective Smoking at Entrance, Diagnosis, and BMI at entrance, contain missing data and thus do not add up to respective PHQ-9 category N. The patients included in this table were not excluded for missing values on variables outside of PHQ-9 scores. Many patients had no Educational attainment data as the UVMMC CR program began collecting Educational attainment after the onset of depression screening. Several patients had diagnosis that could not be located through chart review (out of state or network diagnosis), and many patients had no BMI scores or were not tested for smoking status due to not attending exercise tolerance tests before CR.

Table 3.

Depressive Symptoms at Entrance, and Improvement in Symptoms using the Geriatric Depression Scale – Short Form (GDS-SF)

GDS-SF at Entry							GDS-SF Improvement
Category		Score	Mild or Worse	P Value (Mild or Worse)	Moderate or Worse	P Value (Moderate or Worse)	Score Improvement	Improved by at Least One Category
Overall		2.8 (2.6)	148 (12)		41 (3)		.9 (2.3)	77
Sex				.005*		.128
	Male	2.6 (2.5)	95 (11)		26 (3)		.9 (2.3)	80
	Female	3.2 (2.8)	53 (17)		15 (5)		.9 (2.3)	72
Education				.027*		<.001*
	<High School/GED	4.0 (3.1)	5 (23)		2 (9)		1.8 (3.2)	100
	High School	3.0 (2.5)	20 (13)		11 (7)		.9 (2.8)	95
	Some College	3.0 (2.0)	21 (16)		2 (2)		1.0 (2.1)	71
	Terminal Bachelor’s	2.4 (2.3)	10 (7)		0 (0)		1.1 (1.7)	80
	Postgraduate Education	2.3 (2.5)	12 (9)		3 (2)		1.0 (2.0)	75
Diagnosis				.640		.359
	Surgical	2.7 (2.6)	39 (12)		29 (4)		1.3 (2.3)	87
	Nonsurgical	2.7 (2.7)	97 (11)		9 (3)		.8 (2.2)	75
Continuous Variable		Correlation with GDS-SF Score	Mean Difference (Mild or Worse)		Mean Difference (Moderate or Worse)		Correlation with GDS-SF Improvement
Age		−.02	1.3 (10.6)		3.7* (11.3)		.09*
BMI		.10*	−1.0* (5.6)		−2.7* (6.0)		−.05
VO2peaka		−.21*	2.8* (6.2)		3.3* (6.4)		.05
METspeaka		−.23*	1.0* (1.94)		1.5* (2.0)		.05

^*Denotes a significant difference (P<.05), between groups on proportion of patients with at least mild/moderate depressive symptoms, or significant correlations.

^a.ANOVA Test of mean differences

Abbreviations: BMI, Body Mass Index; Geriatric Depression Scale Short-Form; METs_peak, Peak Metabolic Equivalent of Tasks Score; VO_2peak, Peak Oxygen Uptake.

Data are presented as % of Category, (N) for categorical variables for GDS-SF at entry. For GDS-SF improvement, data in columns 2 and 7 are presented as mean (SD), and column 8 as % patients with mild or worse depression scores who improved by one category.

For continuous variables (Age, BMI, VO_2peak, METs_peak) data are presented as Pearson correlations and mean differences (fitness scores of those with < mild/moderate symptoms – scores of those with mild/moderate or worse).

Notes: Denominators for Mild or Worse or Moderate or Worse in columns 3 and 5 were taken from Table 1. These ratios were calculated as the number of patients who completed 9+ sessions with Mild, Moderate or Worse scores with exit depression scores on both screeners / the total number patients in that category who completed 9+ sessions with exit scores (column 2 Table 1)., e.g., 95 males with 9+ sessions and exit scores on both screeners had mild or worse depression scores / 885 males with 9+ sessions had exit scores on both screeners.

Educational attainment, Objective Smoking at Entrance, Diagnosis, and BMI at entrance, contain missing data and thus do not add up to respective GDS-SF category N. The patients included in this table were not excluded for missing values on variables outside of GDS-SF scores. Many patients had no Educational attainment data as the UVMMC CR program began collecting Educational attainment after the onset of depression screening. Several patients had diagnosis that could not be located through chart review (out of state or network diagnosis), and many patients had no BMI scores or were not tested for smoking status due to not attending exercise tolerance tests before CR.

The PHQ-9 and GDS-SF similarly identified that females were more likely to have ≥ mild depressive symptoms, but this relationship did not hold for the ≥10 cutoff for either measure. Those with higher education were less likely to be categorized with ≥ moderate and ≥ mild depression by the GDS-SF, but only by ≥ moderate depression by the PHQ-9, (all P < .05).

Mean PHQ-9 and GDS-SF scores improved between entrance and exit screening across all demographics. Patients reduced their scores on the PHQ-9 by 1.68 on average, and by .93 points on average on the GDS-SF. Of patients with ≥ mild depressive symptoms at entry on the PHQ-9, 73% reduced their depression by at least one symptom severity category, compared to 77% by the GDS-SF. The only significant differences in improvement between categories were that surgical patients were significantly more likely to improve by a depression category than nonsurgical patients by the PHQ-9 (85% vs. 69%, P < .001).

Higher PHQ-9, but not GDS-SF scores were associated with younger age at entry (r =−.07, P < .05), although GDS-F and not PHQ-9 score improvement was associated with older age at entry (r = .09, P < .05). Higher scores on both measures were associated with greater BMI (r =−.11; r = .10, both P < .05), and higher scores on the PHQ-9 and GDS-SF were associated with lower VO_2peak scores (r =−.14; r = .21, both P < .05). The same was true of METs_peak scores (r =−.17; r = .23, both P < .05). Higher PHQ-9 score improvement was also associated with higher METspeak scores (r =−.07, P < .01). No significant correlations were found between number of sessions attended and depression at exit on either the GDS-SF or the PHQ-9 for those who attended ≥ 9 sessions.

COMPARISON OF THE GDS-SF AND PHQ-9

Comparisons of PHQ-9 and GDS-SF identification of ≥ mild/moderate depressive symptoms, as well as improvement and worsening of symptoms, can be seen in Table 4. The GDS-SF classified significantly lower percentages of individuals with ≥ mild depressive symptoms at entry (12%) than the PHQ-9 (32%) (P < .05). The same effect was found within patients ≥ 65 yr of age (11 vs. 29% respectively, P < .001). The PHQ-9 and GDS-SF differed in identifying individuals with ≥ moderate levels of depression (9 vs. 3% respectively, P < .001), and for those ≥ 65 yr (7 vs. 3% respectively, P < .001).

Table 4.

Comparing GDS-SF and PHQ-9 in Depression Level Categorization, Symptom Improvement, and Symptom Worsening Between Age Ranges

Category	Ratio1	%	X2	P Value
Cut Scores for At Least Mild Depression (PHQ-9 ≥5, GDS-SF ≥ 6)
At Least Mild Depressive Symptoms at Entrance Cutoff			180.5	<.001*
PHQ-9	380/1201	31.6
GDS-SF	148/1201	12.3
At Least Mild Depressive Symptoms at Entrance Within Those ≥65 yr			89.3	<.001*
PHQ-9	197/674	29.2
GDS-SF	71/674	10.5
Percent Improving in Symptoms			.4	.44
PHQ-9	279/380	73.4
GDS-SF	114/148	77.0
Percent Improving Within Those ≥65 yr			.2	.63
PHQ-9	150/197	76.1
GDS-SF	52/71	73.2
Percent Worsening in Symptoms			12.2	<.001*
PHQ-9	65/1194	5.4
GDS-SF	31/1182	2.6
Percent Worsening Within Those ≥65 yr			2.8	.09
PHQ-9	31/672	4.6
GDS-SF	19/663	2.9
Cut Scores for At Least Moderate Depression (PHQ-9 ≥ 10; GDS-SF ≥ 10)
At Least Moderate Depressive Symptoms at Entrance			292.8	<.001*
PHQ-9	105/1201	8.7
GDS-SF	41/1201	3.4
At Least Moderate Depressive Symptoms at Entrance Within Those 65+ yr			149.2	<.001*
PHQ-9	48/674	7.1
GDS-SF	17/674	2.5
Percent Improving in Symptoms			5.0	.025*
PHQ-9	78/105	74.3
GDS-SF	33/41	80.5
Percent Improving in Symptoms Within those 65+ yr			7.8	.005*
PHQ-9	38/48	79.2
GDS-SF	12/17	70.6

^*Denotes a significant difference (P<.05) between groups on proportion of patients with at least mild or moderate depressive symptoms at entry.

¹The ratio of those who decreased in depressive symptom severity by at least one category for those who had at least mild or moderate levels of symptoms at entry. Worsening of symptoms was calculated as those who increased by at least one depressive symptom severity category for those who were categorized below the most severe depression category at entry.

The GDS-SF identified a significantly higher percentage of patients who improved from ≥ moderate depression than the PHQ-9 for the general population (81, 74, P < .05), but not within the ≥ mild depression cutoff or in identifying ≥ mild depression in those ≥ 65 yr, although the PHQ-9 identified significantly higher proportions of patients ≥ 65 yr who improved from ≥ moderate depressive symptoms (79 vs 71%, P = .005). The PHQ-9 identified higher proportions of worsening of symptoms for those with ≥ mild depression compared to the GDS-SF in the overall sample (5 vs. 3%, P < .05), but not in patients ≥ 65 yr.

DISCUSSION

CR programs that wish to screen and treat depression have access to an existing framework provided by the AACVPR.⁹ However, screening remains inconsistent and is often performed with measures that are yet to be validated within CR. This made clear the need to establish benchmark values of depression in a large sample of CR patients, and to compare the commonly used screeners by those benchmarks.

Risk factors for having a depression identification were younger age, being female, lower education, higher BMI, and lower cardiorespiratory fitness scores at entry. Meta-analyses have previously found associations between these factors and higher levels of depressive symptoms in CR. ³¹ Approximately three-quarters of patients with depressive symptoms reduced severity by at least one category by exit. These results were consistent across demographic and clinical characteristics, although non-surgical patients were found to improve less than surgical patients.

This study also provided data on risk factors for noncompletion of CR. Patients who completed < 9 sessions were more likely to be younger, female, current smokers, have lower levels of education, higher depressive symptoms, higher BMI, and a nonsurgical diagnosis. Previous studies found similar trends in education, depressive symptoms, and surgical intervention ^{31,32,33,34,35} Although prior findings are mixed, women have also been found at a higher risk of dropout. ^35,36,37 This may be due to transportation barriers, higher age of onset, and aversion to public mixed-gendered exercise.^4,38,39,40

Literature is also inconsistent on the association between age and dropout. In contrast to our results, older age is often inversely associated with completion.^33,41,42 Age likely has a U-shaped association with dropout where the young and very old are at risk for completing fewer sessions.¹⁴

Of note, although those who would go on to complete < 9 sessions were more likely to exhibit cardiac risk factors, there was no significant difference between these patients and completers in entrance VO_2peak. It may be possible that lower age ameliorated risk factor associated differences in VO_2peak between completers and non-completers, as VO_2peak declines with age.⁴³ One large scale study of over 3,800 patients identified a substantial decrease of 3.5 mL/kg/min per decade of age.⁴³ The magnitude of this effect may have been enough to even out VO_2peak differences.

This is the first study to compare the GDS-SF against the PHQ-9 within the same CR sample. Findings indicate that the GDS-SF identified fewer patients than the PHQ-9 with mild or moderate depression, even within older adults for which it was designed. The percent of patients with ≥ mild depression by the GDS-SF (15%) is consistent with previous findings on depression estimates in CR using the same measure (17%).⁴⁴ The GDS-SF was also less likely to identify worsening depression across the full age range.

The difference in benchmarking results between the two measures may be attributable to the low number of meaningful categories in the GDS-SF classifications options but may also be attributable to the constructs the GDS-SF encapsulates. The measure avoids somatic symptoms of depression which may artificially inflate depression estimates in medical populations in favor of items that measure patient quality of life. ^12,45,46

These are certainly potential advantages for using the GDS-SF in CR, however, the percentages identified by the PHQ-9 were more in line with previous research using screeners validated for this population, where 20-40% of patients were identified as having ≥ mild depression by the BDI-II.^1,2 The use of a less conservative measure seems warranted, as even subclinical levels of depression are predictive of increased mortality following MI.⁴⁷ However, it should be remembered that while the PHQ-9 was a more conservative measure in this sample (e.g. identified more patients with symptoms) this does not necessarily mean it is a more accurate measure (e.g. identifying patients with clinically relevant depression).

While these comparisons are useful, screener results were not directly compared to gold standard measures of clinical depression and as such should be interpreted carefully. The GDS-SF was administered to all patients, reflecting how it is commonly used in CR, but which included age ranges it has not been validated for (e.g. <40). This study does not compare CR with non-CR patients, so no statements can be made on the effects of CR on depression. Future research should seek to obtain follow-up data on all CR participants, regardless of the number of sessions completed, and/or include a non-CR control group to parse changes due to CR participation from changes resulting from treatment or spontaneous improvement.

It is also important to note that some argue that screening for depression without a clear treatment plan may be inadvisable and even with treatment depressive symptoms may not improve.^5,19,46 Thus, future research should compare these measures to diagnostic interviews, examine the effects of somatic items on sensitivity and specificity of the PHQ-9, and examine depressive symptom changes in CR compared to non-attending controls.

Despite the limitations, these results provide important benchmarks on how two commonly used depression screeners perform in the same large cohort of CR patients, including proportions of patients with ≥ mild and moderate depressive symptoms, proportions of patients expected to improve or worsen, and which groups may be at risk for having higher depressive symptoms or not improving in symptoms during CR. Understanding the strengths and limitations of chosen depression screeners can better allow for the identification and treatment of at-risk individuals and assess the effectiveness of CR programs on depression.