Can Mental Health Be Measured? How We Built the Dzeny Mental Wellbeing Index

We have spent a decade giving the body numbers. A sleep score. A readiness score. Steps, heart-rate variability, glucose curves. In Rock Health's 2025 consumer survey, 57% of US adults own at least one wearable or connected health device, and wearable ownership specifically has climbed from 13% in 2015 to 46% today. Sleep is now the second most tracked metric, after physical activity.

And yet the thing people say they worry about most — their own state of mind — has no number on the wrist. You can wake up and know precisely how you slept. You cannot wake up and see how you actually are.

I'm the Product Owner of Dzeny, and the Mental Wellbeing Index is mine to answer for: which validated instruments go in, how the psychologists and psychiatrists on the team work with them, what that demands technically, and what a person finally sees on the screen. That is the actual job — not any one of those three, but making them agree. A metric that is clinically impeccable and unreadable is a dead metric; a metric that is pleasant and unfounded is a worse one.

This article is how we did it: what science already offers, where it falls short of a product, and what design principles make a number about a person honest.

Why this question stopped being philosophical

Three facts, briefly:

• More than 1 billion people live with a mental health condition, by the WHO's September 2025 data — 359 million with anxiety disorders and 332 million with depressive disorders in 2021. Depression and anxiety alone cost the global economy an estimated $1 trillion a year.
• People are engaging with their mental health at unprecedented scale — therapy and companionship is now the #1 use case of generative AI, ahead of coding and writing.
• The quantification habit is already formed. Half the country checks a number about its body every morning. The question "can the mind have a number too?" is being asked by millions of users — whether or not the industry has a good answer.

So: can you measure mental health? What would 0 even mean? What is 100? Who decides?

The honest starting point: there is no single mechanism

For the body, there is ground truth. A heart rate is a heart rate; a glucose level is a lab value. For the mind, no such single instrument exists — and anyone who tells you their app measures "your mental health" with one universal sensor is selling you a metaphor.

What does exist — and has existed for decades — is a family of validated clinical scales: short, structured self-report questionnaires tested on large populations for reliability and validity. GAD-7 for anxiety (Spitzer et al., 2006). PHQ-9 for depression (Kroenke, Spitzer & Williams, 2001). Rosenberg's self-esteem scale, in continuous use since 1965. The Self-Compassion Scale (Neff, 2003). The WHO-5 Well-Being Index, whose validity is supported by a systematic review covering 213 articles (Topp et al., 2015). Measures of meaning, relationships, mindfulness, post-traumatic growth. Each captures one region of the territory; none claims to be the whole map.

When psychometricians say a scale is "validated," that word carries specific, testable content: internal consistency, test–retest reliability, construct validity, sensitivity to change. The GAD-7 was published with a Cronbach's α of 0.92 — excellent internal consistency (Spitzer et al., 2006). The PHQ-9 at its standard cutoff of 10 detects major depression with 88% sensitivity and 88% specificity (Kroenke et al., 2001). These are not vibes-based quizzes; they are instruments with decades of accumulated evidence across languages and populations.

And there is one more well-established finding: measuring regularly is itself therapeutic. A meta-analysis of randomized controlled trials of measurement-based care — routinely administering outcome scales and reviewing them — found higher response and remission rates and reduced symptom severity compared with usual care (Zhu et al., 2021). A meta-analysis of 24 studies on routine progress feedback in psychotherapy reached the same direction: patients whose progress is measured and fed back do better, with the largest gains among those at risk of deterioration (Lambert, Whipple & Kleinstäuber, 2018). Psychiatry has called the systematic adoption of this practice "a tipping point" for the field (Fortney et al., 2017). The retest is not paperwork around the treatment. It is part of the treatment.

That was our starting material: no universal mechanism, a strong library of partial instruments, and solid evidence that the act of regular measurement helps.

What we built: the Dzeny Mental Wellbeing Index (DMWI)

We anchored the index to the most widely accepted definition there is — the WHO's: mental health is a state of wellbeing in which a person can cope with the stresses of life, realize their abilities, learn and work well, and contribute to their community.

Notice what that definition does: it refuses to reduce mental health to "absence of anxiety." So a number that honors it can't be one-dimensional either. DMWI is built from seven components, and each one stands on a validated instrument rather than on questions we invented:

Together the seven cover the three clauses of the WHO definition: coping with stress (Calm, Energy, Recovery), realizing your abilities (Meaning, Self-acceptance, Self-esteem), functioning and contributing (Relationships, Energy). The index is the WHO definition, operationalized.

We didn't set out to prove the scales right — or wrong

This is the part I find most interesting methodologically.

Clinical scales were built for clinics: a patient, a clinician, an appointment. We run them inside a product where a person also talks — regularly, in their own words, about their own life. That gives us something a paper questionnaire never had: context to check the number against.

Seeing that opportunity at all requires standing in two places at once — knowing what the instrument was validated to do, and knowing what the product actually collects. That is the seam I work in, and it is where the interesting decisions live.

So inside Dzeny, the validated scales are continuously compared with the user's subjective sense of how they're doing, with the context of their sessions, and with behavioral signals — and we watch how all of it correlates with the index. In psychometric terms, this is an ongoing check of convergent validity: independent signals about the same construct should move together, and when they do, the number has earned its place. When they diverge, that divergence is not an error to hide — it is exactly the signal a measurement system is for.

What makes a number honest

A few design principles we treat as non-negotiable — and that I'd suggest as a standard for the category:

The baseline comes only from measurement. Your index starts where your assessment says it starts — not at a flattering default. Point A is yours, not a marketing average.

This is also the answer to the question I opened with: what is 0, what is 100, who decides. Nobody decides on your behalf, because the index is not a ranking of humanity with you placed somewhere along it. The scale is anchored to the validated instruments underneath it, and the only comparison that carries meaning is with yourself — where you were in March against where you are in September. A number that invites you to compare yourself with other people is measuring the wrong thing.

Every value has a date. A number about your state is a photograph, not a portrait. An index that doesn't tell you when it was measured isn't telling you anything.

The retest is built into the rhythm. A single measurement is trivia; the delta is the product. Regular re-measurement under comparable conditions is what turns a score into a trajectory — point A to point B, per goal, visible.

The index reflects attention, not performance. Between measurements, the index lives: it responds when you notice, name, and work with what's happening in you — and it quiets when you lose contact with yourself. It is not a grade to win. It is the dynamics of your inner life, made visible. Growth here doesn't mean "you completed tasks"; it means "you showed up for yourself."

Navigation, not just a total. Under each sphere, the index maps where a pattern lives — in thoughts, in emotions, or in reactions, the classic axes of cognitive-behavioral work. Two people with the same anxiety score can have completely different anxieties; the map matters more than the number.

What the number is not

DMWI is not a diagnosis, and we never present it as one. It doesn't label a person, doesn't screen for disorders, and doesn't replace a clinician's assessment. It is a structured, validated-scale-grounded reflection of wellbeing — built so that a person can see their own movement, and bring a richer picture to a professional if they choose to.

Where we are right now: validation in progress

A measurement instrument earns trust the same way a therapy does — through evidence, staged and public about its own maturity. That's the stage we're in, and I want to be precise about it.

Right now, DMWI is going through structured validation: test groups of experts in psychology and psychiatry review how the index is constructed and how it behaves; and we are collecting quantitative and qualitative responses from our users to test the core question any wellbeing metric must answer — does movement in the index correlate with real, observable changes in a person's life? Not "do users like the number," but does the number track reality: qualitatively, in what people report about their relationships, energy, and state — and quantitatively, in the validated scales themselves.

The first completed piece of that work is already public: a three-month study of regular Dzeny use on five validated clinical scales, published with its methodology and its limitations in full. Burnout fell 23%, anxiety 15%, and the established scales moved together with the index — which is exactly the convergence a composite metric has to show before it deserves to be called one.

This is the path the instruments we build on walked decades ago: publish the construction, test it against reality, refine. We think a wellbeing index deserves nothing less rigorous — and we'd rather be at this stage openly than claim a finished certainty the category hasn't earned yet.

Where this goes

The body's numbers changed behavior at population scale: people genuinely sleep more because a ring shows them a sleep score. We think the mind's number can do the same for the habits that measurably support wellbeing — noticing your state, working with your patterns, coming back to yourself regularly.

Your point A takes a few minutes: the first assessment runs right in onboarding, and each area of the index opens with its own measurement — so from the first week, you're not guessing whether things are moving. You're looking at it. You can start with seven days free, no card required.

P.S. And when a person does talk — how do you build the conversation itself so it helps rather than flatters? That's the first article of this series: [Using ChatGPT as a Therapist: What It Gets Wrong](/blog/ai/chatgpt-therapist).

Natalia Zakrzhevskaia — Product Owner, Dzeny