http://orcid.org/0000-0001-7483-2115
Figures
Abstract
Background
Depression is under-diagnosed and under-treated in most areas of the US. New York City is currently looking to close gaps in identifying and treating depression through the adoption of a screening and collaborative care model deployed throughout the city.
Methods
We examine the cost-effectiveness of universal two-stage screening with the 2- and 9-item Patient Health Questionnaires (PHQ-2 and PHQ-9) in New York City followed by collaborative care for those who screen positive. We conducted microsimulations on hypothetical adult participants between ages 20 and 70.
Results
The incremental cost-effectiveness of the interventions over the average lifespan of a 20-year-old adult in NYC is approximately $1,726/QALY gained (95% plausible interval: cost-saving, $10,594/QALY gained).
Conclusions
Two-stage screening coupled with collaborative care for depression in the clinical setting appears to be significantly less expensive than most clinical preventive interventions, such as HIV screening in high-risk patients. However, effectiveness is dependent on the city’s ability to manage scale up of collaborative care models.
Citation: Jiao B, Rosen Z, Bellanger M, Belkin G, Muennig P (2017) The cost-effectiveness of PHQ screening and collaborative care for depression in New York City. PLoS ONE 12(8): e0184210. https://doi.org/10.1371/journal.pone.0184210
Editor: Benjamin Christopher Kearns, University of Sheffield, UNITED KINGDOM
Received: January 19, 2017; Accepted: August 21, 2017; Published: August 31, 2017
Copyright: © 2017 Jiao et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data used are either within the paper or included in the Supporting Information files.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
In New York City, ranks second, after ischemic heart disease, in terms of Disability Adjusted Life Years (DALYs) (one DALY is defined as one year of healthy life lost to disease) [1]. Moreover, depression takes a large economic toll in New York City, mostly in the form of health care costs and lost productivity in the workplace [1].
Because depression remains largely under-diagnosed and is treatable, improved screening and treatment could reduce the burden of disease due to depression. In New York City, only about half of the cases of adult depression are diagnosed in a clinical setting [2]. The rate of undiagnosed depression is obtained by screening samples of the city population with a sensitive and specific screening instrument, and then asking the participants whether they have ever been diagnosed with clinical depression. Routine primary care screening improves the diagnosis rates of adult depression, and has been recommended by the US Preventive Services Task Force (B recommendation) [3,4].
One inexpensive but also sensitive way to screen for depression is to use the 2-item Patient Health Questionnaire (PHQ-2) followed by the 9-item (PHQ-9) version for people who screen positive on the brief questionnaire. The PHQ is a well-validated screening tool [5]. The PHQ-2, which consists of the first two items of PHQ-9, can be used as a rapid screening test and has a high sensitivity, but low specificity [6]. By administering the remaining 7 questions, the specificity of the test improves, resulting in a reasonable “positive predictive value,” a measure of how often people are correctly diagnosed given the underlying prevalence of disease in the general population [7].
However, even when depression is diagnosed, it is often under-treated [8]. Adequate treatment is defined as receiving at least 8 psychotherapy visits or 4 medication monitoring visits within a year [9]. By this definition, less than half of those diagnosed with depression receive “adequate” treatment in the United States [9]. Collaborative care (CC) models have been promoted to address this problem [10]. In CC, a team approach is used to: 1) monitor the severity of depression, 2) provide proactive follow-up, and 3) ensure regular psychiatric consultation focused on treatment changes for patients who are not improving [11]. Randomized controlled trials have demonstrated that patients with depression who are being treated with CC have significantly greater adherence to treatment, more active adjustment of treatment in response to measured improvement, and better outcomes [12,13]. Of course, in the real world, CC teams probably vary significantly with respect to the quality, efficiency, and effectiveness of the care that they provide.
In 2015, the government of New York City announced ThriveNYC, a comprehensive strategy including 54 initiatives aimed at improving the mental health of New Yorkers [1]. As part of this plan, New York City committed to increasing the availability of screening and implementation of CC models. These models are to be adopted by primary care providers working in the “NYC Mental Health Corps,” an initiative that will engage approximately 400 professionals across the city [1]. At the same time, New York State, through reforms to its Medicaid program, is also promoting the growth of CC through the “1115 Medicaid waiver” which is a policy that in part provides incentives for adoption of this model in large health systems. The goal of the 1115 waiver is for 80% of New Yorkers having access to “advanced primary care”, which includes CC [14,15]. Central to both the city-level and state-level approaches are improved screening with the PHQ and improved treatment with CC. However, the program is a new investment, and is of unknown value.
We therefore project the cost-effectiveness of universal PHQ-2 screening followed by PHQ-9 and paired with CC for those who screen positive in New York City. We compared the interventions to “no screening” plus “usual care” as defined by the current mix of diagnosis and treatment in the community (the status quo).
Method
Overview and definitions
Our model calculates the proportion of true positives, false positives, true negatives, and false negatives using the sensitivity and specificity of the PHQ-2, the sensitivity and specificity of the PHQ-9, and the underlying prevalence of undiagnosed depression in New York City. Costs and outcomes are assigned to both correct and incorrect diagnoses relative to what would be expected to occur under the status quo in the real world (Model inputs can be found in Table 1).
The CC model in New York City is assumed to be equivalent in cost and efficacy to that in the literature, and to include: patient education, an initial two-session treatment with a psychiatrist in the primary clinic, a maximum of 3 months of shared care with the psychiatrist and primary care physician, and monitoring that assesses the need for follow-up visits and the adequacy of medication dose [13] (Model assumptions were all tested in sensitivity analyses, and can be found in Table 2).
To assess the cost-effectiveness of the interventions, we built a Markov model using TreeAge Pro 2016. Our model estimated the costs and outcomes for the average New York City resident in primary care between the ages of 20 and 70. The quality-adjusted life year (QALY), defined as one year of perfect health, was used as an outcome measure. From a societal perspective, we included all costs, including screening costs and treatment costs, as well as indirect costs such as lost productivity. Time and transportation costs were not included, as they were deemed likely to be negligible relative to the screening and treatment costs and benefits. The indirect impact of the interventions cost reductions for other medical conditions was not included. Whether these gains are likely to be significant remains controversial, particularly among those with undiagnosed depression. For example, those with undiagnosed depression likely have less contact with the health care delivery system than those with diagnosed depression. Those who have developed co-morbid conditions are more likely to have had contact with a wide array of providers, and are therefore possibly more likely to have diagnosed depression. The extent to which treating depression reduces co-morbidity also remains controversial [16].
However, to the extent that screening and CC reduces co-morbidity, the health utility gain associated with depression comorbidities was at least partially captured in the measure that we use here, as it was based upon real-world cases of depression and depression remission [17]. To calculate the incremental cost-effectiveness ratio (ICER), we divided the additional money spent in the intervention arm by the additional gains in QALY. A 3% discount rate was used following recommendations of the Panel on Cost-effectiveness in Health and Medicine [18].
Health utility
We used HRQL scores from a study on depression and depression remission that used the EuroQol questionnaire (EQ-5D) (see Table 1) [17]. The EQ-5D is comprised of 5 dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. We assumed that there was no change in HRQL for patients without a remission.
Costs
The monetary costs were adjusted to constant 2015 US dollars by using the Consumer Price Index of New York City [19,20]. The cost values included in our model are listed in Table 1 and the full cost breakdown is presented in S1 Table. The lost productivity associated with depression included the cost of absenteeism from the workplace, $609, and reduced productivity while at work, $1,975 [21]. It has been argued that the EQ-5D implicitly includes a measure of lost productivity [22]. Specifically, it controversially assumes that participants who engage in exercises meant to place a value on health are not only considering the effect of disability on their quality of life, but are also able to implicitly valuate the impact of their disease on their earnings and leisure time. This is especially problematic for a condition for which lost productivity is difficult to fathom [23]. For this reason, we analyzed the data both with and without these secondary measures of lost productivity. Additionally, lost productivity explicitly refers to production of goods and services. While cost-effectiveness standards suggest that lost productivity and leisure time be valued as equivalent, this may not reflect real-world valuations. A sensitivity analysis was therefore performed to apply lost productivity to participants aged 20–64 [24].
To estimate the screening cost, we assumed 1 minute of physician time would be required to review the PHQ-2 score, and 3 minutes for the PHQ-9., We assumed that a nurse would devote 6 minutes to administering this process [25,26]. These values reflect the mean time for screening, whether positive (which takes more time) or negative (which takes less time). We used the mean wages of a general practitioner and a registered nurse in New York City and adjusted for the probability of having positive or negative PHQ-2 test result based on the prevalence of depression and the sensitivity of the test. The mean hourly wages of general practitioners in New York City, $83, and registered nurses, $34, were obtained from United States Department of Labor [27].
The cost of CC includes the cost of antidepressant prescriptions and mental health and intervention visits; the cost of usual care has the same components minus the intervention visits [13].
Probabilities
The age-specific probability of death was derived from a US life table (see S2 Table) [28]. We obtained the 12-month prevalence of depression from the New York City Health and Nutrition Examination Survey (NYC HANES) 2013–2014 (see S2 Table) [29]. It should be noted that the prevalence of depression in primary care was estimated to be 1.74 times higher than in general population of the United States [30,31]. Thus, to estimate the age-specific probabilities of depression, we multiplied the prevalence of depression in each age group by this number.
Additional model probability inputs are listed in Table 1 below. The probability of treatable depression was obtained from a Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial in which a cumulative remission rate of 67% was reported [32].
Scores of 3 and 10 are often recommended as the optimal cut-off points for a depression diagnosis within the PHQ-2 and PHQ-9 respectively [33–35]. Given these cut-off points, the sensitivity and specificity of the PHQ-2 for diagnosing depression were 88% and 88%, respectively; for the PHQ-9, they were 83% and 90%, respectively [33,34].
Decision analysis models
The model incorporated two health states: “alive” and “dead”. Alive patients could be depressed or well, and depressed patients could be treated with CC or untreated [36]. The model diagram is presented in Fig 1, and the key assumptions are listed in Table 2. In our model, we estimate the costs and QALYs associated with the status quo versus universal PHQ-2 screening, followed by PHQ-9 screening for positive screens and CC [36]. The status quo arm includes all costs and benefits currently realized from the availability of primary and psychiatric care in the community. This includes costs and QALYs gained from lost productivity, current care seeking behaviors, and treatment programs. Participants in the status quo arm are exposed to his or her age-specific probability of depression and death, and include patients who are both treated and untreated for depression. Participants who “die” exit the model and simulated participants who “survive” exposure to this probability of death remain in the model.
The universal screening and treatment arm includes an estimate of the increased benefits from having a formal screening program in place alongside a CC model (thereby somewhat improving both the detection and management of clinical depression). Moreover, the probability of depression is reduced by the incremental remission rate of the previous life cycle (both because early treatment may reduce relapse and because CC involves a broader range of modalities for resistant depression).
Sensitivity analyses
We conducted a series of one-way sensitivity analyses along with a Monte Carlo simulation. In the Monte Carlo simulation, a triangular distribution was used for every variable in the model. The random error associated with an estimate or plausible boundaries for the values were included in this distribution.
Results
The results of the fully incremental analysis are presented in Table 3. If $ 40,000 per QALY was chosen as a conservative threshold of willingness-to-pay, two-stage screening followed by CC was more cost-effective than any other strategy such as screening with PHQ-9 only or screening without CC. Implementing two-stage depression screening and CC in primary care in New York City increases the discounted societal costs by $660 per person over a 50-year interval from age 20 to 70. The average 20-year-old would gain 0.38 QALY over this interval, resulting in an ICER of $1,726/QALY at a discount rate of 3% (Table 3, values rounded).
Table 4 lists the effects of a series of one-way sensitivity analyses on ICERs. The treatment cost associated with CC produced more variance in the ICER than any other cost variable. Likewise, the probability of the adequacy of CC treatment was the most influential probability input. If we assumed that the EQ-5D included a measure of lost productivity and withdrew the associated cost, the ICER would increase to $8,840/QALY. If we only applied lost productivity to the working age population (thereby removing monetary valuation of leisure time losses), an ICER of $1,870/QALY would be obtained.
There was an increase in the ICER when lowering the cut-off score for the PHQ-2. Conversely, lowering the cut-off score of PHQ-9 decreased the ICER. In the Monte Carlo analysis, the 95% plausible interval of the ICER ranged from dominance (where treatment actually saves money overall) to $10,594/QALY. (By convention, the word plausible is used to highlight the difference in estimation used between traditional confidence intervals based on random error and error generated by simulation, which includes both random and “plausible” non-random error). The cost-effectiveness acceptability curve and the incremental cost-effectiveness scatter plots are presented in Figs 2 and 3 respectively. Additionally, a graph showing the relationship between the expected value of perfect information versus willingness-to-pay is shown in S1 Fig.
Discussion
Randomized controlled trials (RCTs) have shown that screening for depression in the primary care setting generally increases the diagnosis of depression [3]. Similarly, RCTs have shown that CC improves treatment adequacy [12,13]. Taken together, substantial gains in depression remission rates could be achieved with universal screening and CC in clinical settings. We found that the quality-adjusted life expectancy of those who receive these interventions would increase by about 0.38 QALYs relative to those who do not. This increase is substantial, coming close to the gains associated with smoking cessation [37].
Depression screening itself consumes valuable practitioner time, and the cost of CC is about $860 more costly per patient compared to usual care. Nevertheless, the average incremental net cost across the entire life course (with remissions and exacerbations) was only $660. This is mainly because a considerable amount of lost workplace productivity associated with depression would be saved if remission were achieved. In the United States, the greatest direct economic loss linked to depression is lost productivity [21]. Remission also reduces costs associated with future medical treatments. However, the healthy may be misdiagnosed when there are false positives (a particular problem in low-prevalence settings). Such patients not only spend time in treatment and incur the cost of medication and provider time, they may also experience side effects associated with medications. Similar concerns apply to those who have depression, but who do not respond to treatment.
Indeed, we find that it may be cost-effective to raise the cut-off point of the PHQ-2 so that fewer false positives occur. The PHQ-2 has modest specificity but very high sensitivity [6,34]. Lowering the threshold for detection, and thus classifying more patients as in need of treatment [6,34], will both increase program costs associated with administering the PHQ-9 and costs associated with treating people without depression. Hence, raising the cut-off point of the PHQ-2 seems be a more cost-effective option than doing so for the PHQ-9. Ultimately, though, the cut-off points depend on the policy maker’s view of the trade-off between cost and effectiveness.
This study is prone to a number of important limitations. First, the probability of treatment adequacy in CC and the health utility estimates were derived from randomized trials in Seattle and Northern England respectively, which might not be generalizable to New York City. It is also important to remember that additional various and complex barriers, such as stigma and financial cost, affect many patients’ ability to continue treatment for their depression [38], and these considerations were not included. Another limitation of the study is that we did not calculate the changes in QALYs for those who do not achieve remission but who did respond partially to the treatments. Depressive symptoms may be reduced in this group of patients [32], but there is, to date, inadequate information on partial response rates. Inclusion of such values could increase the cost-effectiveness of the interventions.
In our study, the background 12-month prevalence of depression was assessed using the World Health Organization Composite International Diagnostic Interview (CIDI), which was assumed to be the gold standard. However, the diagnosis of depression is complex, and no single instrument is 100% sensitive and specific. Additionally, our probabilistic sensitivity analysis—a Monte Carlo simulation with triangular distributions—does not account for many nuances in the data. Within each distribution, we included random error in addition to a mutually-agreed upon estimation of non-random error with somewhat arbitrary endpoints. While sampling many triangular distributions will produce a normal distribution, the final distribution is likely to contain much wider confidence intervals than many other approaches. Finally, some inputs of the model, such as the probability of depression diagnosis and treatment, were derived from a general population survey in New York City. It is likely that different probability estimates would be achieved if studied specifically within the primary care settings. For example, there is evidence that white Americans have more access to primary care, and are also more likely to receive treatment for depression than non-white Americans [39,40]. Values may differ depending on the population to which the interventions are applied. Moreover, as the prevalence of undiagnosed depression declines in any given population, the positive predictive value screening declines alongside the ICER.
More medical costs would be saved and more effectiveness would be gained if depression comorbidity were included as costs in our analysis. Even without it, however, our models suggest that PHQ screening coupled with CC appear to be very cost-effective.
We model the two interventions that have received the most attention in the literature. ThriveNYC, an ambitious mental health plan in New York City, consists of a wide variety of initiatives, including extending opportunities for screening, outside of clinical settings altogether. Such approaches could reach even higher risk populations and higher need communities, groups that underutilize primary care [1], but our models do not include such approaches. It is not appropriate to generalize our findings to settings in which the prevalence of undiagnosed depression may be lower than those we observe in clinical settings.
ThriveNYC is also concerned with large scale public messaging and other initiatives to change the current culture around mental health, and specific initiatives have been developed to reframe people’s perception about mental health [1]. Screening and CC can provide even greater value if stigma, a significant barrier to depression treatment, is reduced [41]. However, these broader programs are of unknown efficacy and may increase costs without brining additional benefits.
Nevertheless, by expanding our models to broader, integrated initiatives, it may be possible to find synergies that produce even greater social benefits that we observe here. Our models can support smarter public health approaches for mental health, as starting points for the inclusion of cost-effectiveness goal-setting in policymaking.
Supporting information
S1 Fig. The relationship between expected value of perfect information (EVPI, y-axis) and willingness to pay (x-axis) in our cost-effectiveness model.
https://doi.org/10.1371/journal.pone.0184210.s001
(TIF)
S1 Table. Complete list of costs used in our analysis.
All costs were adjusted to constant 2016 dollars. a Source: Arias E. United States life tables, 2011. Natl Vital Stat Rep [Internet]. 2015 Sep;64(11). Available from: http://www.cdc.gov/nchs/data/nvsr/nvsr64/nvsr64_11.pdf.
b Source: New York City Department of Health and Mental Hygiene. 2013–2014 New York City Health and Nutrition Examination Survey. 2015.
https://doi.org/10.1371/journal.pone.0184210.s002
(DOCX)
S2 Table. Age-specific probabilities of depression and of death used in our analysis.
https://doi.org/10.1371/journal.pone.0184210.s003
(DOCX)
Acknowledgments
We thank the New York City Department of Health and Mental Hygiene for the data support. We also thank Dr. Babak Mohit and Ms. Jinjing Wu for their suggestions.
References
- 1.
Government of New York City. ThriveNYC: A mental health roadmap for all [Internet]. 2015. Available: https://thrivenyc.cityofnewyork.us/wp-content/uploads/2016/03/ThriveNYC.pdf
- 2. Gwynn RC, McQuistion HL, McVeigh KH, Garg RK, Frieden TR, Thorpe LE. Prevalence, diagnosis, and treatment of depression and generalized anxiety disorder in a diverse urban community. Psychiatr Serv Wash DC. 2008;59: 641–647. pmid:18511584
- 3. Pignone MP, Gaynes BN, Rushton JL, Burchell CM, Orleans CT, Mulrow CD, et al. Screening for depression in adults: a summary of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med. 2002;136: 765–776. pmid:12020146
- 4. U.S. Preventive Services Task Force. Screening for depression in adults: U.S. preventive services task force recommendation statement. Ann Intern Med. 2009;151: 784–792. pmid:19949144
- 5. Maurer DM. Screening for depression. Am Fam Physician. 2012;85: 139–144. pmid:22335214
- 6. Arroll B, Goodyear-Smith F, Crengle S, Gunn J, Kerse N, Fishman T, et al. Validation of PHQ-2 and PHQ-9 to screen for major depression in the primary care population. Ann Fam Med. 2010;8: 348–353. pmid:20644190
- 7. Richardson TM, He H, Podgorski C, Tu X, Conwell Y. SCREENING FOR DEPRESSION IN AGING SERVICES CLIENTS. Am J Geriatr Psychiatry Off J Am Assoc Geriatr Psychiatry. 2010;18: 1116–1123. pmid:20808102
- 8. Coventry PA, Hudson JL, Kontopantelis E, Archer J, Richards DA, Gilbody S, et al. Characteristics of effective collaborative care for treatment of depression: a systematic review and meta-regression of 74 randomised controlled trials. PloS One. 2014;9: e108114. pmid:25264616
- 9. Kessler RC, Berglund P, Demler O, Jin R, Koretz D, Merikangas KR, et al. The epidemiology of major depressive disorder: results from the National Comorbidity Survey Replication (NCS-R). JAMA. 2003;289: 3095–3105. pmid:12813115
- 10. Jacob V, Chattopadhyay SK, Sipe TA, Thota AB, Byard GJ, Chapman DP, et al. Economics of collaborative care for management of depressive disorders: a community guide systematic review. Am J Prev Med. 2012;42: 539–549. pmid:22516496
- 11. Whitebird RR, Solberg LI, Jaeckels NA, Pietruszewski PB, Hadzic S, Unützer J, et al. Effective Implementation of collaborative care for depression: what is needed? Am J Manag Care. 2014;20: 699–707. pmid:25365745
- 12. Katon W, Korff M Von, Lin E, Simon G, Walker E, Unützer J, et al. Stepped collaborative care for primary care patients with persistent symptoms of depression: a randomized trial. Arch Gen Psychiatry. 1999;56: 1109–1115. pmid:10591288
- 13. Katon W, Russo J, Korff M Von, Lin E, Simon G, Bush T, et al. Long-term effects of a collaborative care intervention in persistently depressed primary care patients. J Gen Intern Med. 2002;17: 741–748. pmid:12390549
- 14.
New York State Department of Health. A Path toward Value Based Payment-New York State Roadmap For Medicaid Payment Reform [Internet]. 2015. Available: https://www.health.ny.gov/health_care/medicaid/redesign/dsrip/docs/vbp_roadmap_final.pdf
- 15.
New York State Department of Health. Redesigning New York’s Medicaid Program [Internet]. Available: https://www.health.ny.gov/health_care/medicaid/redesign/
- 16. Berkman LF, Blumenthal J, Burg M, Carney RM, Catellier D, Cowan MJ, et al. Effects of treating depression and low perceived social support on clinical events after myocardial infarction: the Enhancing Recovery in Coronary Heart Disease Patients (ENRICHD) Randomized Trial. JAMA. 2003;289: 3106–3116. pmid:12813116
- 17. Mann R, Gilbody S, Richards D. Putting the “Q” in depression QALYs: a comparison of utility measurement using EQ-5D and SF-6D health related quality of life measures. Soc Psychiatry Psychiatr Epidemiol. 2009;44: 569–578. pmid:19011721
- 18.
Gold MR, Siegel JE, Russell LB, Weinstein MC, editors. Cost-Effectiveness in Health and Medicine. 1 edition. New York: Oxford University Press; 1996.
- 19.
United States Department of Labor. Consumer Price Index [Internet]. Available: https://www.bls.gov/cpi/data.htm
- 20.
United States Department of Labor. Databases, Tables & Calculators by Subject (New York-Northern New Jersey-Long Island, NY-NJ-CT-PA) [Internet]. Available: https://data.bls.gov/pdq/SurveyOutputServlet
- 21. Greenberg PE, Fournier A-A, Sisitsky T, Pike CT, Kessler RC. The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry. 2015;76: 155–162. pmid:25742202
- 22. Lensberg BR, Drummond MF, Danchenko N, Despiégel N, François C. Challenges in measuring and valuing productivity costs, and their relevance in mood disorders. Clin Outcomes Res CEOR. 2013;5: 565–573. pmid:24273412
- 23. Tilling C, Krol M, Tsuchiya A, Brazier J, van Exel J, Brouwer W. Does the EQ-5D reflect lost earnings? PharmacoEconomics. 2012;30: 47–61. pmid:22066754
- 24.
The Federal Reserve Bank of St. Louis. Working Age Population: Aged 15–64: All Persons for the United States [Internet]. 2017. Available: https://fred.stlouisfed.org/series/LFWA64TTUSM647S
- 25. Smarr KL, Keefer AL. Measures of depression and depressive symptoms: Beck Depression Inventory-II (BDI-II), Center for Epidemiologic Studies Depression Scale (CES-D), Geriatric Depression Scale (GDS), Hospital Anxiety and Depression Scale (HADS), and Patient Health Questionnaire-9 (PHQ-9). Arthritis Care Res. 2011;63 Suppl 11: S454–466. pmid:22588766
- 26. Valenstein M, Vijan S, Zeber JE, Boehm K, Buttar A. The cost-utility of screening for depression in primary care. Ann Intern Med. 2001;134: 345–360.
- 27.
United States Department of Labor. Occupational employment and wages, May 2015 29–1062 Family and General Practitioners [Internet]. 2015. Available: http://www.bls.gov/oes/current/oes291062.htm
- 28. Arias E. United States life tables, 2011. Natl Vital Stat Rep. 2015;64. Available: http://www.cdc.gov/nchs/data/nvsr/nvsr64/nvsr64_11.pdf
- 29.
New York City Department of Health and Mental Hygiene. 2013–2014 New York City Health and Nutrition Examination Survey. 2015.
- 30. McElroy SL. The Epidemic of Depression With Obesity: (Commentary). J Clin Psychiatry. 2015; e1340–e1342. pmid:26528665
- 31. Mitchell AJ, Rao S, Vaze A. International comparison of clinicians’ ability to identify depression in primary care: meta-analysis and meta-regression of predictors. Br J Gen Pract J R Coll Gen Pract. 2011;61: e72–80. pmid:21276327
- 32. Rush AJ, Trivedi MH, Wisniewski SR, Nierenberg AA, Stewart JW, Warden D, et al. Acute and Longer-Term Outcomes in Depressed Outpatients Requiring One or Several Treatment Steps: A STAR*D Report. Am J Psychiatry. 2006;163: 1905–1917. pmid:17074942
- 33. Kroenke K, Spitzer RL, Williams JBW. The PHQ-9. J Gen Intern Med. 2001;16: 606–613. pmid:11556941
- 34. Kroenke K, Spitzer RL, Williams JBW. The Patient Health Questionnaire-2: validity of a two-item depression screener. Med Care. 2003;41: 1284–1292. pmid:14583691
- 35. Manea L, Gilbody S, McMillan D. Optimal cut-off score for diagnosing depression with the Patient Health Questionnaire (PHQ-9): a meta-analysis. CMAJ Can Med Assoc J J Assoc Medicale Can. 2012;184: E191–196. pmid:22184363
- 36. Pence BW, O’Donnell JK, Gaynes BN. The Depression Treatment Cascade in Primary Care: A Public Health Perspective. Curr Psychiatry Rep. 2012;14: 328–335. pmid:22580833
- 37.
Muennig P, Bounthavong M. Cost-Effectiveness Analysis in Health: A Practical Approach. 3 edition. San Francisco, CA: Jossey-Bass; 2016.
- 38. Mohr DC, Hart SL, Howard I, Julian L, Vella L, Catledge C, et al. Barriers to psychotherapy among depressed and nondepressed primary care patients. Ann Behav Med Publ Soc Behav Med. 2006;32: 254–258. pmid:17107299
- 39. Alegría M, Chatterji P, Wells K, Cao Z, Chen C, Takeuchi D, et al. Disparity in Depression Treatment among Racial and Ethnic Minority Populations in the United States. Psychiatr Serv Wash DC. 2008;59: 1264–1272. pmid:18971402
- 40. Shi L, Chen C-C, Nie X, Zhu J, Hu R. Racial and socioeconomic disparities in access to primary care among people with chronic conditions. J Am Board Fam Med JABFM. 2014;27: 189–198. pmid:24610181
- 41. Interian A, Martinez IE, Guarnaccia PJ, Vega WA, Escobar JI. A qualitative analysis of the perception of stigma among Latinos receiving antidepressants. Psychiatr Serv Wash DC. 2007;58: 1591–1594. pmid:18048562