You Shouldn't Have to Guess Whether Your Treatment Is Working

Why subjective self-assessment fails, and what objective data actually shows about treatment response.

We built a biometric monitoring platform for therapists. The question that drives the entire product is one that millions of people on psychiatric medication ask themselves constantly but can never definitively answer: is this actually working?

If you bring your hamster to the vet and the vet prescribes medication, they will run labs in two weeks to check whether the medication is working. They will look at numbers. They will compare those numbers to the baseline they took before prescribing. They will make an objective determination.

If you go to a psychiatrist and the psychiatrist prescribes medication, they will ask you how you feel. That is the test.

It is a remarkable question when you think about it. In virtually every other area of medicine, treatment response is measured. Your oncologist does not ask if you feel like the tumor is shrinking. Your endocrinologist does not ask if your blood sugar seems lower. Your cardiologist does not ask if your heart rhythm feels more regular. They measure. They know.

In psychiatry, the standard method for evaluating whether a medication is working is asking the patient how they feel. No blood test. No scan. No continuous monitor. A subjective self-assessment, filtered through mood-congruent memory, delivered in a 15-minute medication check every 4-8 weeks. Approximately 55 million Americans are managing their psychiatric treatment this way.

Why "I Feel Better" Is Not a Reliable Measurement

The problem is not that patients are dishonest. The problem is that human self-assessment of internal states is demonstrably unreliable, and this has been known for decades.

Mood-congruent memory bias means your current emotional state shapes which memories you access. If you feel decent on the day of your appointment, you will disproportionately recall the decent days and underweight the bad ones. If you feel terrible, you will recall the terrible days and forget that Tuesday and Wednesday were actually okay. Research by Stone et al. published in Science (1999) showed that retrospective mood reports correlate poorly with what people reported feeling in the moment when assessed through ecological momentary assessment. People are not good historians of their own internal experience.

There is also the adaptation problem. When a medication causes emotional blunting, reduced motivation, or sexual dysfunction, patients often stop noticing these effects over time because the blunted state becomes their new normal. They report "feeling fine" because they have lost the reference point for how they felt before. This is not improvement. It is adaptation to side effects that may be masking both the benefits and the harms of the medication.

Then there is the placebo response. Meta-analyses of antidepressant clinical trials, including the landmark 2008 study by Turner et al. in the New England Journal of Medicine, have shown that the published literature systematically overstates antidepressant efficacy. When Turner obtained FDA review data for 74 trials of 12 antidepressant drugs involving 12,564 patients, he found that 94% of published trials appeared positive while only 51% of all trials (including unpublished ones) actually were. Overall effect sizes were inflated by 32%. This does not mean the drugs do not work for some people. It means the evidence base for how well they work is distorted, which makes individual assessment even more important and even harder to do subjectively.

What Objective Treatment Response Actually Looks Like

If you move beyond subjective self-report, research has identified several measurable physiological signals that track treatment response independently of how a person feels on any given day.

Sleep architecture normalizes before mood improves. One of the earliest objective signs that a treatment is producing physiological change (whether that treatment is medication, therapy, exercise, or some combination) is improvement in sleep quality. Not just total hours, but the internal structure of sleep: more time in deep sleep and REM sleep, fewer nighttime awakenings, more consistent sleep timing. Research consistently shows that sleep architecture improvements precede subjective mood improvement by days to weeks. A person whose deep sleep percentage increases from 8% to 18% of total sleep over two weeks is showing a measurable physiological shift, even if they do not yet "feel better." Conversely, a person who reports "feeling better" but whose sleep architecture has not changed may be experiencing a placebo response, a temporary mood lift, or hypomania rather than genuine improvement.

Heart rate variability (HRV) trends upward as autonomic regulation improves. HRV reflects the balance between the sympathetic and parasympathetic nervous systems. Reduced HRV is consistently associated with depression, anxiety, and PTSD in hundreds of studies. When treatment is producing real physiological improvement in autonomic regulation, HRV trends upward over weeks and months. A 2019 meta-analysis by Koch et al. in Neuroscience and Biobehavioral Reviews confirmed reduced HRV in major depressive disorder. More importantly for treatment monitoring, the reverse is also true: improving HRV during treatment correlates with clinical improvement. This signal is available from any consumer wearable that tracks heart rate.

Activity levels stabilize. Depression is associated with reduced physical activity and increased sedentary time. Effective treatment, whatever form it takes, tends to produce gradual normalization of daily movement patterns. This is measurable through step counts and active minutes tracked by a wearable device. The key is the trend, not any single day. A person whose 30-day average step count is gradually rising from 2,500 to 5,000 is showing objective improvement in behavioral activation, regardless of what they report feeling.

Circadian rhythm consistency improves. Mood disorders are strongly associated with irregular sleep-wake timing. Effective treatment tends to produce more consistent sleep midpoints (the halfway point between falling asleep and waking up). A person whose sleep midpoint was swinging between 2 AM and 6 AM and has stabilized to a consistent 3 AM is showing improved circadian regulation. This is a measurable, objective signal that does not depend on self-report.

Conversational engagement patterns shift. In AI-powered check-in systems, the content and character of a person's responses over time provide another data stream. Increasing response length, richer emotional vocabulary, more future-oriented language, and higher engagement frequency all correlate with clinical improvement in digital phenotyping research. Declining engagement, shorter responses, and more absolutist language ("never," "always," "nothing") correlate with deterioration.

Why This Matters Regardless of Your Treatment Path

This is not an argument for or against any particular treatment. It is an argument against guessing.

If you are on medication and it is working, objective data confirms that. You have evidence to show your prescriber. You have a reason to stay the course.

If you are on medication and it is not working, objective data shows that too. You have evidence to support a change rather than continuing to wait and hope.

If you are tapering off medication, objective data tracks whether you are stable through the taper. Sleep, HRV, and activity data can distinguish between withdrawal symptoms (temporary, physiological adjustment) and relapse (actual clinical deterioration). This distinction is extremely difficult to make subjectively but often visible in biometric trends.

If you are in therapy without medication, objective data shows whether therapy is producing measurable physiological change. This is empowering for both you and your therapist.

If you are trying lifestyle interventions (exercise, sleep hygiene, stress reduction, supplements), objective data shows whether they are moving the needle.

In every scenario, the answer to "is this working?" stops being a feeling and starts being a fact.

The Current System Is Flying Blind

Consider what the current standard of care actually looks like.

A patient is prescribed an antidepressant. They are told it may take 4-6 weeks to reach full effect. They return for a 15-minute medication check after 4-8 weeks. The psychiatrist asks, "How are you feeling?" The patient says, "I think maybe a little better?" The psychiatrist says, "Let's give it more time" or "Let's increase the dose." The dose is adjusted. Another 4-8 weeks pass. The cycle repeats.

At no point in this process is anything measured. The drug level in the blood is not checked (for most psychiatric medications, therapeutic drug monitoring is not standard). The target symptoms are not quantified against a baseline. The side effects are not objectively assessed. The entire treatment arc, which may last years or decades, is guided by intermittent subjective self-report.

Meanwhile, if you brought your dog in for the same complaint, your vet would have run a CBC and chemistry panel before prescribing, scheduled a recheck with repeat bloodwork in two weeks, and made dosage decisions based on the numbers. Not based on whether the dog seems happier.

This is what open-loop control looks like. Input is applied (medication). Output is never measured. Adjustments are made based on the patient's best guess about their own internal state. In engineering, this is recognized as the least reliable form of system control. In aviation, it would be unconscionable. In chronic disease management for conditions like diabetes or heart disease, it was abandoned decades ago. In mental health, it is still the standard of care.

What Would Change If We Actually Measured

The technology to measure treatment response objectively exists today. Over 100 million Americans wear a wearable device that continuously tracks heart rate variability, sleep stages, activity levels, resting heart rate, and sleep timing. AI systems can conduct daily conversational check-ins and extract patterns from the responses. The data is already being generated. It is sitting on millions of wrists right now, going unused for clinical purposes.

If this data were integrated into clinical care, the "is this working?" question would have an answer. Not a feeling. Not a guess. An answer backed by weeks or months of continuous physiological data, benchmarked against your own personal baseline, interpreted in the context of your treatment timeline.

Your therapist could walk into every session already knowing how your week went. Not from your memory of it. From the data. Your prescriber could see whether a medication change produced measurable physiological improvement or measurable deterioration. You could see it yourself. That is what closing the 167-hour blind spot actually looks like.

Whatever treatment path you are on, you deserve to know whether it is working. Not to guess. To know.