Measurement-Informed Psychiatric Care: If Not Now, Then When?

For those of us who regularly use measurements in practice, their utility is unquestionable in empowering and engaging patients, dealing with payers, talking to families, and making more precise, timely, and shared clinical decisions. But many of us remain reluctant to use measures to inform better diagnostic, treatment-delivery, and patient-management decisions.

I first learned about measurement-based care (MBC) from Aaron T. Beck, MD, as a resident (although it wasn’t called MBC at the time). We used the Beck Depression Inventory (BDI) to assess treatment effectiveness and identify cognitions to target in cognitive therapy.

Subsequently, when conducting controlled trials, regularly measuring depressive symptoms helped me make personalized dose adjustment decisions. I found that many patients would tell me, “I’m not getting better,” yet their self-reported depression rating scale was already showing meaningful improvement. Similarly, my patients with bipolar disorder conversationally would say that they were not worsening, but as we went through a mania symptom checklist, we could detect symptomatic worsening that was clinically meaningful and could intervene before things got worse. So I am hooked on measurement because of my training, clinical research, and practice experiences.

Unfortunately, my pro-measurement bias is not yet widely shared. In 2019, only 8% of Veterans Affairs hospital psychiatrists consistently measured outcomes for at least half their patients.¹ Recently, however, Ryland et al² reported that about half of psychiatrists in the United Kingdom are not routinely using outcome measures to guide treatment.

So what is behind measurement reluctance? Does measurement actually improve diagnosis, treatment outcomes, or long-term patient management? Are we missing an opportunity to learn from each other by relying only on notes in our clinical records?

Developing the Habit of Measuring and Then Deciding

In medicine, measurement routinely informs diagnosis, treatment selection and delivery, adverse effect management, and prognostication and long-term management. These measurements, routinely recorded in the electronic health record (EHR), are already being harvested using artificial intelligence (AI) to produce a learning health care system in general medicine.³

Psychiatrists, on the other hand, rely almost exclusively on impressions formed by careful listening, observation, and sometimes information from others who know the patient well, despite clinical practice guideline recommendations to measure outcomes to improve the quality of care since the 1990s. We’re in the habit of trusting our impressions. Often, they are pretty good. But good judgments are not necessarily optimal judgments. Can we kick it up a notch? Maybe some of us are hooked on impression!

Measurements not only make us better clinicians, but they more effectively engage patients in managing their care. Measurements also inform program managers, payers, and clinicians about program effectiveness, patient selection, and resource allocation.

In short, without universally collected, easily accessible, and analyzable outcomes in the EHR, our clinical work cannot serve as a source of learning for other clinicians or support the development of a learning health care system. Having these outcomes available also helps us prepare for the future by taking advantage of AI and biomarker development.

Is It “Measurement-Based Care” or “Measurement-Informed Care”?

MBC typically refers to the regular measurement of symptoms and adverse effects, ideally combined with a plan of action that is implemented based on these measurements to personalize medication delivery.⁴ But measurements also improve diagnostic and management decisions (as noted later in this piece). Further, MBC has also inappropriately come to mean “measurement-directed” care—that is, a scale score is deemed sufficient to recommend or even require a particular action or diagnosis. For example, a friend of mine was told by the health care system that he had to have 80% of all patients with a Patient Health Questionnaire score over 10 on medication.

Thus, measurement-informed care (MIC) might be more accurate because it recognizes that diagnostic, treatment-related, and prognostic and patient-management decisions can all be improved by measurement, and that measurement alone is an insufficient basis for these decisions. Measurements inform; clinicians and patients decide (as in the rest of medicine). Clinicians must consider the available evidence (often more limited than we wish), patient preferences, and individual circumstances (eg, symptom complexity, chronicity, environmental context, measurement performance, medical fragility, comorbid conditions, concomitant medications, prior treatment history, etc) to diagnose, treat, and manage patients and their conditions.

MIC also embraces the sequence of decision-making aligned with routine medical care: history, examination, impression, measurement, decision, action, assessment of results (with measurement), revise accordingly, etc. Impression and clinical experience play a critical role in clinical decisions, but measurements add precision, a degree of certainty about the impressions, and a metric that others can understand. Of course, measurements are more informative for some patients than for others, just as research evidence is more relevant to some than to others.

Does Measurement Improve Diagnosis?

Diagnostic assessment requires an evaluation of a wide range of symptoms, both cross-sectionally at initial presentation and retrospectively to define the nature and course of illness. Systematic approaches to establishing diagnosis, such as structured interviews, have been convincingly shown to outperform routine diagnostic processes (ie, nonstructured interviews). These interviews often detect comorbid conditions not otherwise recognized and confirm or revise the initial diagnostic impressions. Rettew et al,⁵ in a meta-analysis of 38 reports involving more than 15,000 participants, found low to moderate agreement between clinical diagnoses and those from structured interviews.

For example, in a community mental health center, a structured interview conducted by a nurse, combined with a review of the medical records and a 30- to 45-minute diagnostic interview, is defined as the “gold standard” for diagnosis. Compared to this standard, clinical diagnoses had a sensitivity of less than 60% for bipolar I disorder and major depressive disorder.⁶

To determine whether structured interview diagnoses impact practice, a randomized trial compared clinician behavior in those who received or did not receive this diagnostic information (N = 296) (Figure 1). Within 90 days, clinicians with diagnostic feedback changed their diagnoses 73% of the time compared with 16% for clinicians who received no information. Furthermore, within 180 days, medications were more often changed in the feedback group across all diagnoses, including schizophrenia (9% vs 2.2%), bipolar disorder (15% vs 2%), major depressive disorder (21% vs 4%), substance use disorders (41% vs 7%), and anxiety disorders (39% vs 1%). Clinicians who received the structured-interview results were more likely than those who did not to delete (factor of 2.3) or change (factor of 2.8) prescriptions.⁷

Although data from these and other studies support the use of structured interviews, they are time-consuming. Over the past 2 decades, computerized adaptive testing (CAT) has been developed to rapidly acquire sufficient knowledge to estimate a psychiatric diagnosis and severity. CAT uses the computer to choose each new question based on respondents’ previous answers, like an experienced clinician who asks the next question based on what the patient just said.⁸ Diagnostic results have been validated against gold standard structured interviews for a range of conditions.⁹ Given the results from Kashner et al⁷ with the time-intensive structured interview, one would expect similar effects with CAT with far less expense and clinical staff time.

Does Measurement Improve Treatment Delivery and Outcome?

For medication adjustments, MBC entails the systematic measurement of symptoms and adverse effects, often combined with a specific plan to revise dosing based on those measures to personalize treatment.⁴ We initially employed this approach in the Texas Medication Algorithm Project (TMAP)¹⁰ and later in the Sequenced Treatment Alternatives to Relieve Depression study to ensure that each treatment step was well implemented before undertaking the next.

Findings from several randomized controlled studies have shown that MBC-guided medication treatment has better outcomes than routine care delivery without measurement for depression, including depressive symptom outcome and adherence^11,12, as well as strengthening shared decision-making.¹³ Notably, MBC is recommended by many professional organizations. Some would say that the challenge in MBC is no longer determining whether MBC improves outcomes, but how to consistently implement it in clinical practice to facilitate adoption.

To illustrate the magnitude of the effect of MBC on antidepressant treatment, consider 2 randomized controlled trials (RCTs) that compared MBC-guided antidepressant management to treatment-as-usual (TAU) care in depression over 6 months.^14,15 Both studies followed the same protocol. Furthermore, each study provided dose-adjustment guidance to MBC clinicians based on symptoms and adverse effects measured at each visit. The initial treatment was either mirtazapine or paroxetine. If the agent failed, the next step was the alternative. Measurements were taken at baseline and at weeks 2, 4, 8, 12, and 24. By specifying treatment choices in both the MBC and TAU groups, each study isolated the specific effects of MBC vs TAU on dose adjustments and outcome.

In the studies, MBC produced greater and more rapid rates of depressive symptom response and remission at 12 weeks (Figure 2). Specifically, in the single-site RCT (N =120), mean remission rates were 74% (MBC) versus 29% (TAU); response rates were 87% (MBC) and 63% (TAU) after 6 months. Time to response (4.5 vs 8.1 weeks) and to remission (8.4 vs 14.8 weeks) were shorter with MBC.¹⁴

In the multisite RCT (N = 154), response rates at 12 weeks were 91% (MBC) vs 68% (TAU), while remission rates were 72% (MBC) vs 52% (TAU). Times to response and remission were also significantly shorter. The MBC advantages were statistically significant and clinically meaningful (approximating the drug-placebo differences in antidepressant medication registration trials).¹⁵

Two large multisite, multi–step treatment algorithm projects—the German Algorithm Project (GAP)¹⁶ and the TMAP—provide further evidence in psychiatric care systems of the greater efficacy of personalized medication adjustment with MBC than TAU. The GAP, which included inpatients with depression, found all 3 MBC-guided medication algorithms were more effective than TAU, and the length of hospitalization and cost of care were lower with each than with TAU. Similarly, the TMAP, which included public sector outpatients with bipolar disorder, schizophrenia, or major depressive disorder, found that MBC produced better outcomes than TAU (ie, no measurement), with the largest effect in depression.¹⁷

Taken together, the data from these and other studies show that the personalized delivery of medication via MBC (or MIC) is as important as which medication is selected.

Can Measurement Improve Prognostication and Safety?

Simple measurements can also identify patients whose conditions are likely to worsen, providing an early warning system for intervention. In an EHR study, Taquet et al¹⁸ first showed that both instability (waxing/waning of symptoms) and average severity over time, as measured by the Clinical Global Impressions-Severity (CGI-S) scale obtained at every visit during a 6-month period, identified those most likely to be hospitalized within the following 6 months.¹⁹ A subsequent expansion of that study found that the patient’s symptom instability and severity, using the clinician-completed Global Assessment of Functioning (GAF) scale (rated 0 to 100) added comparable predictive validity.²⁰ Together, these 2 ratings (CGI-S and GAF) had even greater predictive value than the CGI-S alone. These predictions were applicable to all diagnoses (eg, psychotic, mood, anxiety, substance abuse, etc),²⁰ suggesting their role as an early warning system regardless of diagnosis.

Measurement Helps Us Learn From Each Other

Global outcome measures, such as the single-item CGI-S or Patient Global Impression of Severity scale, both of which score severity on a 7-point scale, are feasible, informative, and applicable to patients with 1 or more mental health conditions, and they allow us to combine our collective clinical experiences across patients, time, and providers. They can help us make treatment decisions, empower patients, and assess outcomes across various contexts and treatments. Thus, recording this metric at every treatment encounter in the EHR lets us learn from our daily practice in real time and provides information to identify preferred treatment sequences for particular patients, identify patients at risk for adverse events, and personalize treatment selection, dosing, and duration.²¹

Collectively, MBC and/or MIC create a treatment course road map allowing us to more effectively address common challenging issues of patient complexity, comorbidity, chronicity, and treatment-resistance that impact outcome. Simple ratings would tell us when to hold, when to fold, and when to adjust our treatments to optimize response and remission rates, shorten the time to optimal therapeutic benefit, and make treatment more cost-efficient.

As biomarkers arrive to assist in differential diagnosis, treatment decisions, and patient management and prognostication, having these sorts of outcomes as part of our routine EHR would facilitate biomarker evaluation in the real world and their ultimate adoption by years. Being able to compile outcomes within and across practitioners, patients, care systems, and treatment programs will help us learn from our work.

Wrapping Up: Thoughts for Your Consideration

Consistent, evidence-based measurements from the initial evaluation and throughout patient treatment meaningfully improve diagnostic, treatment-delivery, patient-management, prognostic decision-making, and treatment outcomes. Measurement-informed clinical decision-making: (a) breaks a long-standing habit of total reliance on impression; (b) engages patients; (c) improves outcomes; (d) allows us as clinicians to learn from each other; and (e) prepares us to employ AI, evaluate and intelligibly implement biomarkers, and participate in a learning health care system. Most importantly, patient-reported measures enable patients to participate more in managing their conditions and treatments.

Dr Rush is a professor emeritus at Duke-NUS Medical School at the National University of Singapore and an adjunct professor of psychiatry and behavioral sciences at Duke University School of Medicine in Durham, North Carolina.

References

1. Oslin DW, Hoff R, Mignogna J, Resnick SG. Provider attitudes and experience with measurement-based mental health care in the VA Implementation Project. Psychiatr Serv. 2019;70(2):135-138.

2. Ryland H, Bhattacharya R, Richardson J. Use of outcome measures in psychiatry: Royal College of Psychiatrists’ survey of members. BJPsych Bull. Published online January 26, 2026.

3. Mungmode A, Noor N, Weinstock RS, et al. Making diabetes electronic medical record data actionable: promoting benchmarking and population health improvement using the T1D exchange quality improvement portal. Clin Diabetes. 2022;41(1):45-55.

4. Trivedi MH, Rush AJ, Wisniewski SR, et al: STAR*D Study Team. Evaluation of outcomes with citalopram for depression using measurement-based care in STAR*D: implications for clinical practice. Am J Psychiatry. 2006;163(1):28-40.

5. Rettew DC, Lynch AD, Achenbach TM, Dumenci L, Ivanova MY. Meta-analyses of agreement between diagnoses made from clinical evaluations and standardized diagnostic interviews. Int J Methods Psychiatr Res. 2009;18(3):169-184.

6. Ramirez Basco M, Bostic JQ, et al. Methods to improve diagnostic accuracy in a community mental health setting. Am J Psychiatry. 2000;157(10):1599-1605.

7. Kashner TM, Rush AJ, Surís A, et al. Impact of structured clinical interviews on physicians’ practices in community mental health settings. Psychiatr Serv. 2003;54(5):712-718.

8. Gibbons RD, deGruy FV. Without wasting a word: extreme improvements in efficiency and accuracy using computerized adaptive testing for mental health disorders (CAT-MH). Curr Psychiatry Rep. 2019;21(8):67.

9. Gibbons RD, Wang PS. The science of psychiatric measurement. Psychiatr Ann. 2023;53(9):400-404.

10. Rush AJ, Rago WV, Crismon ML, et al. Medication treatment for the severely and persistently mentally ill: the Texas Medication Algorithm Project. J Clin Psychiatry. 1999;60(5):284-291.

11. Fortney JC, Unützer J, Wrenn G, et al. A tipping point for measurement-based care. Psychiatr Serv. 2017;68(2):179-188.

12. Zhu M, Hong RH, Yang T,et al. The efficacy of measurement-based care for depressive disorders: systematic review and meta-analysis of randomized controlled trials. J Clin Psychiatry. 2021;82(5):21r14034.

13. Dey A, Lewis Z, Posel J, Pan RY, Wang K. Quantifying care, qualifying experiences: a systematic review of measurement-based care in psychiatry from patient and provider perspectives. BMJ Ment Health. 2025;28(1):e301663.

14. Guo T, Xiang YT, Xiao L, et al. Measurement-based care versus standard care for major depression: a randomized controlled trial with blind raters. Am J Psychiatry. 2015;172(10):1004-1013.

15. Husain MI, Nigah Z, Ansari SUH, et al. Measurement-based care to enhance antidepressant treatment outcomes in major depressive disorder: a randomized clinical trial. JAMA Netw Open. 2025;8(9):e2529427.

16. Adli M, Wiethoff K, Baghai TC, et al. How effective is algorithm-guided treatment for depressed inpatients? results from the randomized controlled multicenter German Algorithm Project 3 trial. Int J Neuropsychopharmacol. 2017;20(9):721-730.

17. Trivedi MH, Rush AJ, Crismon ML, et al. Clinical results for patients with major depressive disorder in the Texas Medication Algorithm Project. Arch Gen Psychiatry. 2004;61(7):669-680.

18. Taquet M, Griffiths K, Palmer EOC, et al. Early trajectory of clinical global impression as a transdiagnostic predictor of psychiatric hospitalisation: a retrospective cohort study. Lancet Psychiatry. 2023;10(5):334-341.

19. Busner J, Targum SD. The clinical global impressions scale: applying a research tool in clinical practice. Psychiatry (Edgmont). 2007;4(7):28-37.

20. Taquet M, Fazel S, Rush AJ. Transdiagnostic early warning score for psychiatric hospitalisation: development and evaluation of a prediction model. BMJ Ment Health. 2025;28(1):e301622.

21. Rush AJ, Tramontin T. Learning by doing: can our collective experiences as clinicians improve mental health care? J Clin Psychiatry. 2024;85(3):24com15366.