Depression Treatment After a Positive Depression Screen Result

Key Points

Question

Is depression screening associated with appropriate and equitable initial care for patients with elevated depressive symptoms and/or suicidal ideation (SI)?

Findings

In this cohort study including 3980 adults with elevated depressive symptoms, African American or Black and Asian patients, older adults, and those screened via the patient portal had lower odds of initial care than White patients, younger patients, and those with in-person visits. Patients with SI had increased odds of care compared with those without SI.

Meaning

To improve depression care, results of this study suggest that health systems need strategies to increase treatment initiation after screening, particularly for groups at risk for undertreatment.

Abstract

Importance

As primary care practices increase depression screening, it is unknown whether screening is associated with appropriate and equitable treatment.

Objective

To investigate factors associated with initial treatment among patients who screen positive for depression and/or suicidal ideation (SI).

Design, Setting, and Participants

Cohort study using electronic health record data from September 2017 to September 2021, from a large US academic health system. Participants were adult primary care patients with elevated depressive symptoms (Patient Health Questionnaire-9 score ≥10) and/or SI, excluding patients with baseline depression, bipolar disorder, schizophrenia, schizoaffective disorder, or dementia. Data were analyzed from December 30, 2022, to February 17, 2024.

Exposure

Patient characteristics including gender, age, preferred language, and race and ethnicity.

Main Outcomes and Measures

Primary outcome was antidepressant or mental health referral ordered at screening. Secondary outcomes were antidepressant/referral and antidepressant/referral or follow-up visit within 8 weeks.

Results

Of 60 062 patients screened, 3980 (7%) reported elevated depressive symptoms or SI. The cohort was 68.1% women (2711), and the mean (SD) age was 46.5 (17.6) years; 0.8% were 12.4% African American or Black (493), American Indian or Alaska Native (30), 24.8% Asian (988), 14.6% Latino/Latina/Latinx (582), 1.5% Pacific Islander (58), and 36.9% White (1470), and 9.0% other/unknown (359); 5.6% preferred a non-English language (223). Approximately 38% (1518) received antidepressants/referrals when screened (including 44% of 461 patients [203] with SI). By 8 weeks, 2785 patients (70%) received antidepressant/referral and/or follow-up (including 75% of 783 patients with SI). In multivariable logistic regression models adjusting for site and clustered on primary care physician, there were no statistically significant differences in the primary outcome by gender, preferred language, or health insurance. African American or Black and Asian patients had lower estimated probabilities of treatment ordered when screened (34.0% [95% CI, 28.4%-39.6%] for Black patients and 35.4% [95% CI, 31.5%-39.4%] for Asian patients) than White patients (40.5% [95% CI 37.4%-43.5%]). Estimated treatment decreased with increasing age (46.4% [95% CI, 41.2%-51.5%] for patients aged 18-30 years and 17.5% [95% CI, 12.1%-22.9%] for patients aged ≥75 years). Patients with SI had greater estimated treatment than those without SI (43.5% [95% CI, 39.9%-47.1%] vs 35.2% [95% CI, 33.0%-37.5%]), although treatment was overall low for this high-risk group. Secondary outcomes were consistent, although there were no statistically significant differences in follow-up visits for African American or Black and Asian patients compared with White patients.

Conclusions and Relevance

In this cohort study, moderate rates of initial treatment among patients with elevated depressive symptoms and/or SI were found. Targeted interventions are needed for patients at risk of undertreatment, including patients with SI, African American or Black and Asian patients, and older adults.

This cohort study assesses systematic depression screening and recommended initial treatment among those with elevated depressive symptoms and/or suicidal ideation and whether individuals historically at risk for undertreatment received equitable initial care.

Introduction

Many primary care practices began or expanded systematic depression screening of adults in primary care after the United States Preventive Services Task Force (USPSTF) revised guidelines in 2016.^1,2,3,4 In one large health system, the implementation of systematic depression screening increased the screening rate by 48% over 2 years; additionally, initial screening disparities by language, race and ethnicity, age, and insurance were eliminated.² Other health systems reported similar increases, though studies have not consistently reported on screening by patient characteristics.^4,5

However, depression screening alone is not enough; it must be followed by a clinical assessment to confirm the diagnosis and appropriate treatment if indicated. A study of 240 000 patients in 5 health systems found that after a new diagnosis of depression in primary care, only 36% of patients completed a psychotherapy visit or filled a prescription for an antidepressant within 90 days.⁶ Furthermore, Asian, Black, and Latino/Latina/Latinx patients had 30% lower odds of treatment than White patients.⁶ Other studies have found that patients with non-English preferred language and older adults are also less likely to receive depression care than English-preferring patients and younger adults.^{7,8,9,10,11,12,13,14}

In 2023, the USPSTF updated depression screening recommendations, specifying that effective treatment generally includes medications and/or psychotherapy.¹⁵ The USPSTF found adequate evidence that depression screening programs (interventions that included screening plus treatment support) increase treatment and improve patient outcomes.¹⁵ While systematic depression screening has been adopted broadly across the US, it is unknown whether screening is associated with recommended initial treatment or more equitable treatment among those populations at risk for depression undertreatment.

This study objective was to investigate whether systematic depression screening was associated with recommended initial treatment among those with elevated depressive symptoms and/or suicidal ideation (SI) and whether individuals historically at risk for undertreatment received equitable initial care.

Methods

Setting

University of California, San Francisco Health (UCSF Health) is an academic health system with more than 2 million outpatient visits annually. There are 6 adult primary care practices with more than 200 clinicians. Between September 1, 2017, and September 30, 2021, these practices served 66 062 empaneled community-dwelling adults.

UCSF Health implemented adult depression screening in September 2017.² By the end of 2019, high screening rates (approximately 89%) were achieved in this health system. Additionally, early screening disparities, including lower screening rates among patients with non-English language preference, African American or Black individuals, older adults, men, and those with public insurance, were eliminated by 2019. This study was approved by the UCSF institutional review board with a waiver of consent due to minimal risk to participants. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for observational studies.

Study Sample

We constructed an electronic health record (EHR) cohort to examine initial depression treatment and follow-up after screening among patients with elevated depressive symptoms and/or SI. Our dataset included all eligible primary care patients seen in the 6 UCSF adult primary care practices from September 1, 2017, to September 30, 2021. Data were analyzed from December 30, 2022, to February 17, 2024. We included patients with at least one outpatient primary care visit during the study period. We excluded patients with a diagnosis of depression, bipolar disorder, schizophrenia, schizoaffective disorder, or dementia at baseline (based on problem list or encounter International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes).

Medical assistants conducted depression screening at UCSF Health using the Patient Health Questionnaire-2 (PHQ-2)¹⁶ and Patient Health Questionnaire-9 (PHQ-9),¹⁷ which are well-validated and widely used screening measures for current depressive symptoms.^17,18,19,20 If a patient had not completed PHQ screening within the prior 12 months, either electronically or in person, at the time of a visit, the medical assistants administered the PHQ-2 and PHQ-9 verbally or it was completed independently by the patient using a paper form (which was abstracted into the EHR by the medical assistants). Starting in 2019, the health system sent PHQ-2 and PHQ-9 questionnaires electronically to patients via the patient portal before a scheduled visit or without an associated visit (to conduct population-level depression screening) (eMethods in Supplement 1). While the USPSTF concluded that there is insufficient evidence to screen for suicide risk,¹⁵ many health systems use the PHQ-9, which screens for both depression and SI.

If a patient scored a PHQ-2 of 3 or higher (score range 0-6), a PHQ-9 (score range 0-27) was administered. A PHQ-2 score of 3 or higher and a PHQ-9 score of 10 or higher have high sensitivity (85%) and specificity (85%) for depression.¹⁵ Among patients with a confirmed diagnosis of depression, a PHQ-9 score of 10 or higher is often considered an indication for treatment or close monitoring.¹⁶ SI, regardless of PHQ-9 score, should be further assessed, treatment initiated, or close follow-up arranged, if indicated. For this analysis, we therefore limited our sample to individuals with elevated depressive symptoms (PHQ-9 score ≥10) and/or SI (indicated in the PHQ-9). At UCSF Health, if a patient has a PHQ-9 score of 10 or higher or reports SI, a decision support alert appears in the EHR (eMethods in Supplement 1) during the associated visit to prompt a clinical diagnostic evaluation to confirm depression and initiate treatment. This decision support alert is linked to an optional order set with recommended antidepressants and a single mental health referral (for psychiatry or psychotherapy) for all primary care sites. Patients are triaged and referred to short-term collaborative care at UCSF Health or to community mental health clinicians. If a patient reports SI, an additional decision support tool (eMethods in Supplement 1) includes resources to further assess and risk stratify patients, document symptoms, and provide patient information (including emergency hotlines).

Outcomes

We investigated the management of depressive symptoms and SI after a patient’s first positive screen result. We considered initial treatment per USPSTF recommendations: providing an antidepressant prescription and/or placing a behavioral health referral at the time of screening or close monitoring specifically for depressive symptoms. We limited our analysis to treatment ordered at the time of screening or within 8 weeks, rather than 90 days, as used in prior studies,^6,21 to capture physician actions more likely directly related to a single positive depression screen result.

Our primary outcome was an antidepressant prescription or mental health referral ordered at the time of screening. We further investigated 2 secondary outcomes: (1) antidepressant or referral ordered within 8 weeks after a positive screen and (2) antidepressant and/or follow-up visit specifically for depressive symptoms within 8 weeks, defined as an encounter with depression on the problem list or if a repeat PHQ-9 was administered. With our secondary outcomes, we hoped to capture treatment that may have been prompted by the positive screen result. This included individuals who may have screened positive for depressive symptoms but warranted watchful waiting or declined treatment at the time of screening.

Covariates

We evaluated patient characteristics that have been associated with depression screening and treatment in prior studies. Race and ethnicity, gender, and preferred language were obtained from all new patients at the time of the first visit and recorded in the EHR. Additional demographic and patient care characteristics obtained from the EHR included age, baseline Elixhauser comorbidity count,^22,23 number of visits during the study period, insurance (private, Medicare, Medicaid, or other, which includes self-pay, Veterans Affairs health care, and worker’s compensation), visit type (in person, video, or telephone), and whether screening occurred during a visit or via the patient portal. To account for portal screenings that were prompted by an upcoming or recently completed visit, those that occurred within 2 weeks of a clinic visit were assigned to that visit (and the visit was considered the time of screening for treatment outcomes). All other portal screenings were classified as portal only and we assessed the outcomes as of the portal screening date. To evaluate differences in treatment pre–COVID-19 pandemic or post–COVID-19 pandemic, we compared treatment in 3 periods from September 1, 2017, to March 15, 2020, from March 16, 2020, to May 30, 2020, and from June 1, 2020, to September 30, 2021.

Statistical Analysis

Using Stata version 16.1 (StataCorp LP) for all analyses, we evaluated the unadjusted proportion and the adjusted odds of (1) antidepressant/referral ordered at the time of screening, (2) antidepressant/referral ordered within 8 weeks (including at the time of screening), and (3) antidepressant/referral ordered and/or follow-up visit for depressive symptoms within 8 weeks. We used 3 logistic regression models specifying robust SEs to accommodate clustering by primary care physician, and adjusting for primary care site, to evaluate our outcomes by patient demographic and patient care characteristics. We estimated probabilities of treatment for each covariate exposure level while holding every other covariate fixed at its mean value, and they are reported with 95% CIs. We evaluated the unadjusted proportions of all outcomes among those with SI. Given the low number of nonbinary individuals, only women and men were included in the multivariable analysis. We excluded 1 individual who was missing covariates. Statistical significance was set at 2-tailed P < .05.

Results

Of 60 062 patients screened during the study period, most of these patients were insured. A total of 3980 patients (7%) reported elevated depressive symptoms and/or SI (Table 1). The 3980 patients were 68.1% women, had a mean (SD) age of 46.5 (17.6) years, and included 12.4% African American or Black (493), 0.8% American Indian or Alaska Native (30), 24.8% Asian (988), 14.6% Latino/Latina/Latinx (582), 1.5% Pacific Islander (58), and 36.9% White (1470), and 9.0% other/unknown (359), and 5.6% patients with non-English language preference (223). A total of 2942 (74%) reported elevated depressive symptoms only, 87 (2%) reported SI only, and 951 (24%) reported both. SI was most common among men, Asian and American Indian or Alaska Native patients, non-English speakers, both the youngest and oldest patients, and Medicare recipients. A total of 618 patients (15.5%) completed screening via the patient portal.

Table 1. Demographic Characteristics of 3980 Patients Who Screened Positive for Depression or Suicidal Ideation.

Demographic characteristic	No. (%)
Gender
Women	2711 (68.1)
Men	1267 (31.8)
Unknown	2 (0.1)
Race and ethnicitya
African American or Black	493 (12.4)
American Indian or Alaska Native	30 (0.8)
Asian	988 (24.8)
Latino/Latina/Latinx	582 (14.6)
Pacific Islander	58 (1.5)
White	1470 (36.9)
Other/unknown	359 (9.0)
Language preference
English	3757 (94.4)
Chinese	90 (2.3)
Spanish	43 (1.1)
Other	90 (2.3)
Age category, y
18-30	856 (21.5)
31-44	1166 (29.3)
45-54	632 (15.9)
55-64	623 (15.7)
65-74	436 (11.0)
≥75	267 (6.7)
Health insurance
Private	2147 (53.9)
Medicare	920 (23.1)
Medicaid	822 (20.7)
Other (self-pay, Veterans Affairs health care, worker’s compensation)	91 (2.3)
Comorbidities, mean (SD)	2.9 (2.5)
No. of visits per year during study period, mean (SD)	3.6 (3.2)
Reported SI	1038 (26.1)
Screening modality
Office visit	2720 (68.3)
Video visit	619 (15.6)
Telephone visit	23 (0.6)
Patient portal only	618 (15.5)
Screened via the patient portal (n = 1045)
Had a visit within 2 wk (and assigned to that visit)b	427 (40.9)
No visit within 2 wk (patient portal only)	618 (59.1)
Time period of positive PHQ-9 score ≥10
September 1, 2017-March 15, 2020	2480 (62.3)
March 16, 2020-May 30, 2020	148 (3.7)
June 1, 2020-September 30, 2021	1352 (34.0)

Abbreviations: PHQ-9, Patient Health Questionnaire-9; SI, suicidal ideation.

^a Race and ethnicity are presented according to the electronic health records. No further breakdown of the other category is available.

^b The mean (SD) screen was within 4.5 (4.0) days of a visit (median, within 3 days; range, 1-14 days). Among portal screens not associated with a visit within 2 weeks, the mean (SD) screen was within 189 (218) days of a visit (median, within 104 days; range 15-1261 days).

Approximately 38% of patients (1518 of 3980) received antidepressants/referrals when screened (including 44% of 461 patients [203] with SI). Overall, 70% of patients (2785 of 3980) had recommended initial treatment after a positive screen result: 38% (1518 of 3980) had recommended treatment (antidepressant/referral) ordered at the time of screening; 13% (519 of 3980) had antidepressant/referral ordered within 8 weeks; and an additional 19% (738 of 3980) had follow-up for their depressive symptoms within 8 weeks.

For the primary outcome of treatment at the time of screening, adjusted odds ratios (AORs) and estimated probability values are reported in Table 2. In the multivariable models, there was no statistically significant difference in depression treatment at the time of screening by gender, preferred language, health insurance, or by increasing comorbidity count. African American or Black and Asian patients had lower estimated probabilities of treatment ordered when screened (34.0% [95% CI, 28.4%-39.6%] for African American or Black patients and 35.4% [95% CI, 31.5%-39.4%] for Asian patients) than White patients (40.5% [95% CI 37.4%-43.5%]). Treatment decreased with increasing age (46.4% [95% CI, 41.2%-51.5%] for patients aged 18-30 years and 17.5% [95% CI, 12.1%-22.9%] for patients aged ≥75 years). Patients with SI had greater estimated treatment than those without SI (43.5% [95% CI, 39.9%-47.1%] vs 35.2% [95% CI, 33.0%-37.5%]), although overall low for this high-risk group. African American or Black (35.9% treated [AOR, 0.76; 95% CI, 0.59-0.98]) and Asian patients (37.3% treated [AOR, 0.81; 95% CI, 0.65-0.99]) had lower odds of antidepressant/referral being ordered than White patients (39.8% treated) at the time of screening. Patients screened via the patient portal also had markedly lower odds of treatment (13.4%) than patients screened in the context of a visit (42.7%; AOR, 0.20; 95% CI, 0.14-0.27). In contrast, patients with SI had greater odds of an antidepressant/referral being ordered (44.4%) than those who did not (35.9%; AOR, 1.41; 95% CI, 1.21-1.66). Odds of treatment at the time of screening did not vary across the study period.

Table 2. Depression Treatment at the Time of a Positive Screen.

Demographic characteristic	Antidepressant/referral ordered at time of positive screen result
	Treated, %	AOR (95% CI)	Estimated probability (95% CI)
Gender
Women	38.1	1 [Reference]	37.5 (34.8-40.1)
Men	38.2	0.98 (0.83-1.16)	36.9 (33.9-39.9)
Unknown	100	NA	NA
Race and ethnicitya
African American or Black	35.9	0.76 (0.59-0.98)b	34.0 (28.4-39.6)
American Indian or Alaska Native	36.7	0.74 (0.38-1.42)	33.4 (18.6-48.2)
Asian	37.3	0.81 (0.65-0.99)b	35.4 (31.5-39.4)
Latino/Latina/Latinx	36.6	0.81 (0.63-1.05)	35.6 (30.7-40.5)
Pacific Islander	39.7	0.88 (0.47-1.64)	37.5 (23.3-51.6)
White	39.8	1 [Reference]	40.5 (37.4-43.5)
Other/unknown	39.3	0.87 (0.68-1.12)	37.2 (31.7-42.7)
Language preference
English	38.3	1 [Reference]	37.2 (35.1-39.4)
Chinese	33.3	0.87 (0.53-1.44)	34.1 (23.0-45.2)
Spanish	39.5	1.39 (0.64-3.04)	45.2 (26.2-64.3)
Other	37.8	1.09 (0.70-1.70)	39.3 (28.5-50.2)
Age category, y
18-30	41.5	1 [Reference]	46.4 (41.2-51.5)
31-44	36.6	0.82 (0.65-1.03)	41.3 (37.7-45.0)
45-54	39.2	0.71 (0.55-0.91)b	37.9 (33.6-42.2)
55-64	40.0	0.62 (0.49-0.80)b	35.0 (30.9-39.1)
65-74	38.1	0.44 (0.31-0.64)b	27.8 (22.1-33.4)
≥75	27.3	0.24 (0.15-0.39)b	17.5 (12.1-22.9)
Health insurance
Private	37.7	1 [Reference]	36.3 (33.7-38.9)
Medicare	38.4	1.22 (0.94-1.57)	41.0 (35.6-46.3)
Medicaid	38.8	0.96 (0.79-1.15)	35.3 (31.4-39.2)
Other (self-pay, Veterans Affairs health care, worker’s compensation)	40.7	1.26 (0.79-1.99)	41.7 (30.6-52.9)
Comorbidities, mean (SD)	3.1 (2.5)	0.97 (0.94-1.01)	NA
No. of visits per year during study period, mean (SD)	4.3 (3.8)	1.16 (1.12-1.19)b	NA
Reported SI
No	35.9	1 [Reference]	35.2 (33.0-37.5)
Yes	44.4	1.41 (1.21-1.66)b	43.5 (39.9-47.1)
Screening modality
Office visit	42.7	1 [Reference]	43.1 (40.7-45.5)
Video visit	43.1	1.05 (0.80-1.38)	44.3 (38.3-50.3)
Telephone visit	34.8	0.76 (0.29-1.99)	36.5 (14.6-58.4)
Patient portal only	13.4	0.20 (0.14-0.27)b	12.9 (9.7-16.1)
Screened via the patient portal
Had a visit within 2 wk (and assigned that visit type)	39.3	NA	NA
No visit within 2 wk (patient portal only)	13.34	NA	NA
Time period of positive PHQ-9 score ≥10
September 1, 2017-March 15, 2020	40.0	1 [Reference]	37.2 (34.2-40.3)
March 16, 2020-May 30, 2020	29.1	0.69 (0.44-1.09)	29.0 (20.4-37.5)
June 1, 2020-September 30, 2021	35.7	1.05 (0.86-1.28)	38.4 (35.1-41.7)

Abbreviations: AOR, adjusted odds ratio; NA, not applicable; PHQ-9, Patient Health Questionnaire-9; SI, suicidal ideation.

^a Race and ethnicity are presented according to the electronic health records. No further breakdown of the other category is available.

^b Statistically significant result at the 5% level (P < .05).

For the secondary outcomes of treatment within 8 weeks or close follow-up for depressive symptoms, the AORs and estimated probability values are reported in Table 3. There was no statistically significant difference in either secondary outcome by gender, preferred language, or health insurance. African American or Black and Asian patients had lower odds of an antidepressant/referral being ordered within 8 weeks (African American or Black, 47.9% treated [AOR, 0.73; 95% CI, 0.56-0.96] and Asian, 50.3% treated [AOR, 0.80; 95% CI, 0.65-0.99]) than White patients (54.0% treated). However, this association was not found when follow-up visits for depressive symptoms were included in the outcome. As with the primary outcome, the odds of an antidepressant/referral being ordered or a follow-up visit for depressive symptoms within 8 weeks decreased steadily with increasing age (from 54.7% of those aged 18-30 years to 40.1% of those older than 75 years treated). Patients who were screened via the patient portal continued to have lower odds of an antidepressant/referral being ordered (34.5% treated vs 53.3% of those screened in an office visit; AOR, 0.43; 95% CI, 0.34-0.53) and of an antidepressant/referral being ordered or having follow-up (59.7% treated vs 70.8% of those screened in an office visit; AOR, 0.56; 95% CI, 0.45-0.70) within 8 weeks of screening.

Table 3. Depression Treatment Within 8 Weeks of a Positive Screen.

Demographic characteristic	Antidepressant/referral ordered within 8 wk of positive screen result			Antidepressant/referral ordered or had follow-up visit within 8 wk of positive screen result
	Patients treated, %	AOR (95% CI)a	Estimated probability (95% CI)	Patients treated, No. (%)	AOR (95% CI)a	Estimated probability (95% CI)
Gender
Women	51.5	1 [Reference]	53.2 (50.7-55.6)	1904 (70.2)	1 [Reference]	74.0 (72.0-76.0)
Men	51.3	1.03 (0.88-1.21)	53.9 (50.7-57.0)	879 (69.4)	1.00 (0.86-1.17)	74.1 (71.4-76.7)
Unknown	100	NA	NA	2 (100)	NA	NA
Race and ethnicityb
African American or Black	47.9	0.73 (0.56-0.96)c	49.1 (42.8-55.5)	351 (71.2)	1.04 (0.83-1.31)	74.7 (70.1-79.3)
American Indian or Alaska Native	53.3	0.87 (0.41-1.87)	53.5 (34.8-72.3)	21 (70.0)	0.98 (0.43-2.23)	73.6 (57.7-89.4)
Asian	50.3	0.80 (0.65-0.99)c	51.5 (47.8-55.2)	700 (70.9)	1.06 (0.87-1.29)	75.1 (72.1-78.1)
Latino/Latina/Latinx	50.9	0.84 (0.66-1.08)	52.7 (47.7-57.7)	404 (69.4)	0.94 (0.73-1.22)	72.8 (68.8-76.9)
Pacific Islander	51.7	0.91 (0.53-1.55)	54.6 (41.8-67.4)	40 (69.0)	1.05 (0.58-1.88)	74.8 (63.6-86.1)
White	54.0	1 [Reference]	56.9 (53.8-60.0)	1026 (69.8)	1 [Reference]	74.0 (71.4-76.6)
Other/unknown	49.5	0.77 (0.61-0.96)c	50.4 (45.0-55.7)	243 (67.7)	0.90 (0.70-1.15)	71.8 (67.1-76.5)
Language preference
English	51.6	1 [Reference]	53.4 (51.5-55.3)	2632 (70.1)	1 [Reference]	74.1 (72.3-75.9)
Chinese	50.0	1.04 (0.64-1.69)	54.3 (42.4-66.3)	60 (66.7)	0.77 (0.48-1.22)	68.7 (59.1-78.3)
Spanish	48.8	1.16 (0.61-2.21)	57.2 (41.4-72.9)	31 (72.1)	1.30 (0.56-3.03)	78.9 (64.7-93.0)
Other	45.6	0.88 (0.57-1.35)	50.1 (39.2-60.9)	62 (68.9)	0.90 (0.53-1.53)	72.0 (61.3-82.7)
Age category, y
18-30	54.7	1 [Reference]	62.0 (57.7-66.2)	608 (71.0)	1 [Reference]	78.9 (75.6-82.2)
31-44	50.3	0.81 (0.67-0.99)c	57.0 (53.8-60.1)	808 (69.3)	0.86 (0.71-1.04)	76.2 (73.5-79.0)
45-54	54.1	0.76 (0.61-0.95)c	55.4 (51.5-59.2)	451 (71.4)	0.78 (0.59-1.02)	74.4 (70.5-78.2)
55-64	50.7	0.58 (0.45-0.73)c	48.4 (43.8-53.0)	429 (68.9)	0.60 (0.45-0.79)c	69.1 (64.7-73.4)
65-74	52.3	0.51 (0.35-0.74)c	45.3 (37.8-52.7)	315 (72.3)	0.64 (0.43-0.94)c	70.4 (64.1-76.6)
≥75	40.1	0.28 (0.18-0.42)c	31.1 (23.6-38.5)	174 (65.2)	0.44 (0.27-0.71)c	62.2 (52.6-71.7)
Health insurance
Private	51.6	1 [Reference]	52.8 (50.2-55.5)	1476 (68.8)	1 [Reference]	73.5 (71.0-75.9)
Medicare	51.1	1.10 (0.83-1.44)	55.1 (49.6-60.6)	656 (71.3)	1.00 (0.72-1.39)	73.5 (68.6-78.4)
Medicaid	51.8	1.00 (0.83-1.20)	52.8 (48.5-57.1)	589 (71.7)	1.11 (0.90-1.37)	75.5 (71.7-79.3)
Other (self-pay, Veterans Affairs health care, worker’s compensation)	48.4	1.06 (0.69-1.63)	54.4 (43.6-65.1)	64 (70.3)	1.36 (0.83-2.23)	79.0 (70.8-87.3)
Comorbidities, mean (SD)	3.1 (2.5)	0.96 (0.92-0.99)c	NA	3.1 (2.5)	0.99 (0.95-1.03)	NA
No. of visits per year during study period, mean (SD)	4.3 (3.7)	1.24 (1.19-1.29)c	NA	4.1 (3.5)	1.34 (1.28-1.41)c	NA
Reported SI
No	49.0	1 [Reference]	50.9 (48.9-52.8)	2002 (68.1)	1 [Reference]	72.2 (70.2-74.2)
Yes	58.5	1.48 (1.26-1.73)c	60.4 (56.7-64.2)	783 (75.4)	1.43 (1.24-1.65)c	78.8 (76.5-81.1)
Screening modality
Office visit	53.3	1 [Reference]	55.6 (53.2-58.0)	1925 (70.8)	1 [Reference]	75.2 (72.9-77.6)
Video visit	60.1	1.30 (0.99-1.72)	62.0 (56.2-67.8)	474 (76.6)	1.16 (0.89-1.52)	78.0 (74.1-81.8)
Telephone visit	52.2	0.91 (0.36-2.29)	53.2 (30.8-75.7)	17 (73.9)	0.88 (0.31-2.50)	72.9 (52.7-93.1)
Patient portal only	34.5	0.43 (0.34-0.53)c	34.8 (30.6-39.0)	369 (59.7)	0.56 (0.45-0.70)c	63.0 (59.2-66.9)
Screened via the patient portal
Had a visit within 2 wk (and assigned that visit type)a	52.7	NA	NA	306 (71.7)	NA	NA
No visit within 2 weeks (patient portal only)	34.5	NA	NA	369 (59.7)	NA	NA
Time period of positive PHQ-9 score ≥10
September 1, 2017-March 15, 2020	51.8	1 [Reference]	52.7 (49.9-55.4)	1718 (69.3)	1 [Reference]	71.7 (69.4-74.0)
March 16, 2020-May 30, 2020	48.7	0.82 (0.55-1.22)	47.6 (38.4-56.9)	109 (73.7)	1.25 (0.83-1.89)	76.0 (69.2-82.9)
June 1, 2020-September 30, 2021	51.1	1.11 (0.91-1.36)	55.3 (51.6-59.0)	958 (70.9)	1.38 (1.08-1.76)c	77.7 (74.2-81.2)

Abbreviations: AOR, adjusted odds ratio; NA, not applicable; PHQ-9, Patient Health Questionnaire-9; SI, suicidal ideation.

^a Models account for the clustering of patients by physician (random effects); models are adjusted for all covariates presented in the table, as well as primary care sites (fixed effects). Patients with unknown gender and those missing data for any covariates were excluded from the models.

^b Race and ethnicity are presented according to the electronic health records. No further breakdown of the other category is available.

^c Statistically significant result at the 5% level (P < .05).

Patients with SI had greater odds of an antidepressant/referral being ordered (58.5% treated vs 49.0% of those without SI; AOR, 1.48; 95% CI, 1.26-1.73) and of an antidepressant/referral being ordered or having follow-up (75.4% treated vs 68.1% of those without SI; AOR, 1.43; 95% CI, 1.24-1.65) within 8 weeks than those without SI. We evaluated the unadjusted proportions of all outcomes among those with SI; results were similar (eTable Supplement 1). Individuals who were treated from June 1, 2020, to September 30, 2021, had higher odds of antidepressant/referral or having follow-up within 8 weeks (70.9% treated; AOR, 1.38; 95% CI, 1.08-1.76), compared with those treated prepandemic from September 1, 2017, to March 15, 2020 (69.3% treated); no other outcome varied over time.

Discussion

In a large cohort of primary care patients who screened positive for depression and had potential indications for treatment (elevated depressive symptoms and/or SI), we found that 38% of patients received USPSTF-recommended initial treatment (antidepressant, referral) at the screening visit. There were no differences by preferred language, gender, health insurance, or visit type. Consistent with prior studies, we found that African American or Black, Asian, and older patients had lower odds of an antidepressant/referral being ordered at the time of screening. Additionally, patients screened via the patient portal received less treatment at the time of screening. While we found that patients reporting SI were more likely to receive treatment at the time of screening and to be seen for close primary care follow-up than patients without SI, overall treatment was low for this high-risk group. This presents an important opportunity for system-level improvements and could ultimately influence future recommendations for care following a positive suicide screening result.

Treatment with an antidepressant/referral increased by 8 weeks, but initial disparities in treatment persisted. In this study, the inclusion of close primary care follow-up along with treatment eliminated the disparity among African American or Black and Asian patients but not among older adults and those screened via the patient portal.

Historically, depression treatment rates in primary care have been low. For screening to improve depression care and close treatment gaps for groups historically undertreated for depression, screening must be coupled with clinical action (clinical assessment and treatment, if indicated). Prior studies have highlighted physician tendencies to underdetect and undertreat depressive symptoms among African American or Black and Asian patients and to minimize African American or Black patients’ emotional symptoms.^{24,25,26,27,28,29} Both patient and physician stigma and systemic racism may impede treatment, and African American or Black and Asian patients may have additional concerns about pharmacologic treatment, given the mistrust of medical institutions and documented differences in treatment by race and ethnicity.^24,25,30,31 Many African American or Black and Asian patients seek mental health services in primary care rather than in specialized mental health settings³²; increased implementation of the collaborative care model for depression may help close some of these care gaps.^16,33,34,35 In our study, we found lower antidepressant prescription and referral placement for African American or Black and Asian patients after a positive screen result but similar rates of follow-up visits for depressive symptoms. This may reflect differential clinician prescribing practices, patient refusal of treatment or diagnosis, or a patient-centered approach (honoring patient reluctance to start pharmacologic treatment or to accept a mental health referral). Additional efforts to address systemic racism and to increase the trustworthiness of medical institutions, as well as to increase the availability of culturally appropriate treatments and culturally concordant clinicians, may further increase rates of treatment among African American or Black and Asian populations.

Notably, older adults had lower odds of depression care after a positive screen result at the time of screening and within 8 weeks, even when including follow-up visits for depression. With each increasing age category, the odds of treatment decreased steadily. These findings are consistent with prior studies that found that, while depression is less prevalent among older adults, older adults are less likely to have their depressive symptoms recognized and treated.^12,13,14 Physicians may be less likely to treat depression in older adults due to a combination of factors, including underdiagnosis stemming from ageist attitudes (and a belief that depressive symptoms are a normal part of aging), misattribution of somatic symptoms, concerns about adverse effects and medication interactions, stigma surrounding mental illness in older adults, and inadequate professional training in geriatric mental health.^13,36,37 Yet few studies have been conducted among older adults to evaluate depression screening and subsequent treatment.³⁸ Further work must investigate whether physicians clinically assess older adults who screen positive on the PHQ-9 and discuss treatment options.

Similarly, while health systems have used the patient portal as another tool to achieve population-based depression screening, our study found that portal screenings were associated with lower odds of clinical action. This highlights the importance of establishing processes to ensure adequate follow-up of positive patient portal screenings. As telemedicine has become embedded in many primary care practices, screening via the patient portal may play a more prominent role in health systems and could help health systems reach screening targets.^5,39 However, our findings highlight that patient portal screening should not be used in isolation; steps must be taken to allocate sufficient resources to ensure accountability and to proactively establish appropriate workflows for follow-up of positive screen results.

Limitations

Our study has limitations. While we measured physician antidepressant mental health orders, we do not know how many patients declined or did not initiate treatment. Our EHR does not capture patients’ declination of treatment, so our analysis may underestimate physicians’ discussion of or suggestions for depression treatment. However, the inclusion of close primary care follow-up may be a marker for this, and clinicians largely acted on a positive depression screen result. Moreover, it is unknown how many individuals with a positive depression screen result were assessed clinically for depression and determined to fulfill the criteria for a Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) diagnosis of major depressive disorder. While depression screening has the potential to improve detection and treatment of depressive symptoms, it could lead to overtreatment of mild or self-remitting symptoms if not coupled with clinical assessment and accurate diagnosis. There is no way to capture clinician assessment in structured fields in our EHR. However, PHQ-9 scores of 10 or higher have high sensitivity and specificity for depression,³⁸ and a lack of treatment initiation likely reflects missed opportunities. Finally, we cannot account for loss to follow-up in our analysis; patients may have left this health system, changed insurance, or sought care elsewhere for their symptoms, although limiting the analysis to 8 weeks postscreening likely minimized these factors.

Conclusions

To improve depression outcomes with screening, practices must ensure appropriate and equitable treatment of those diagnosed with depression. In this cohort study, we found moderate rates of initial treatment among patients with elevated depressive symptoms and/or SI. While those who reported SI were more likely to have treatment ordered and close follow-up than those without SI, there is considerable room for improvement for this group. However, health systems should track treatment rates and develop approaches to ensure equitable patient-centered treatment and follow-up for all patients, specifically for groups at risk of undertreatment, including African American or Black and Asian patients, older adults, and those screened via the patient portal.

Supplement 1.

eMethods. Timeline of UCSF Health Intervention Components and Implementation Strategies to Implement, Improve, and Sustain Depression Screening and Management

eTable. Depression Treatment Among Patients With at Least One Visit (Between Sep 2017 – Sep 2021) Who Reported Suicidal Ideation (N=1,038)

Supplement 2.

Data Sharing Statement