A treatment result is only as specific as what was measured: in whom, compared with what, and over what period. Reduced craving, improved motivation, attendance and longer-term substance use are different outcomes. Before accepting a claim about “success,” find the actual measure, the number of people included and the follow-up period. A favorable result on one measure does not establish every other benefit.
This guide helps adults and families read research summaries and provider outcome claims. It explains how to frame questions for a qualified professional; it does not evaluate your treatment, rank providers or provide a clinical interpretation of an individual study. Support resources. NIMH’s help guidance.
Start with the outcome’s actual name
Replace a broad word such as “success” with the specific thing being counted or scored. Is it completing a program, reporting less distress, attending appointments, returning to work or reporting a change in substance use? Those can all be meaningful questions, but they are not interchangeable answers.
ClinicalTrials.gov separates each outcome by its name, whether it is primary or secondary, and its measurement timeframe. Its results tables also identify the participants included. That gives you a place to look beyond a headline. The site cautions that interpreting results is complex and should involve expert help.
Copy the measure’s definition into your notes before shortening it. If an advertisement provides no definition, ask for one. “What counted as a successful outcome?” is a more answerable question than “Does this work?”
Find out what the treatment was compared with
A comparison tells you which question the study can address. Treatment added to usual care, two versions of the same activity and care compared with a different intervention are separate questions. Check what every group received, including services shared by all participants.
NIH describes randomized controlled trials as studies that assign participants to treatment or control groups at random to test interventions. Other study designs answer different questions. An interview can explain how a person experienced care without establishing what would have happened to the same person under another treatment.
A before-and-after result alone leaves more possibilities open than a controlled comparison. Ask whether time, other care or changes in who supplied follow-up information could contribute to the reported change. Do not assume that every improvement can be attributed to the highlighted activity.
Keep the measurement date attached to the finding
An immediate response, the end of a program and a later follow-up are different points in a person’s life. Write the interval in ordinary words: “immediately after one session” or “three months after leaving the program.” If the starting point is unclear, leave that ambiguity in the record rather than choosing one.
The Cochrane music-therapy review illustrates why this matters: its conclusion about craving is not a demonstration of lasting recovery. A 2024 trial by Silverman compared three single-session music interventions in 103 detoxification patients using group-relationship measures before and after the session. That study’s outcome and timing do not establish a long-term substance-use result.
Our music feature examines these sources in their own settings. When a provider cites a paper, ask whether its public claim retains the paper’s outcome and observation period. A new publication date does not extend the follow-up of the participants.
Cochrane: music therapy for substance use disorders · Silverman: single-session music interventions and group relationships
Ask who is included in the percentage
A percentage needs a numerator and a denominator: how many people met the stated definition, out of which group? A figure calculated among people who responded to a survey answers a different question from one covering everyone who began treatment.
ClinicalTrials.gov provides participant-flow and outcome-specific participant information. Use those sections to ask who started, who supplied follow-up data and who entered the analysis. A missing response does not reveal whether that person did well or poorly.
For a provider report, ask whether the denominator is all admissions, program completers, people reached after discharge or a selected subgroup. Record exclusions and nonresponses. A large-looking percentage can become much less informative when that information is absent.
| Calculation | What it means |
|---|---|
| 40 people responded out of 100 contacted | The response rate is 40%. Outcomes for the other 60 are unknown in this example. |
| 30 of 40 respondents met the stated criterion | 75% of respondents met it. That is not a measured 75% outcome for everyone contacted. |
| 30 of the original 100 are known to have met it | 30% of the contacted group is documented as meeting the criterion; this is not a complete outcome estimate. |
Separate the size of a change from the way it is presented
Read both the starting value and the changed value. In a fictional comparison, a specified event occurs in 20 of 100 people in one group and 10 of 100 in another during the same period. The difference is 10 people per 100, or 10 percentage points. Relative to the first group’s 20%, the second rate is 50% lower. These are two descriptions of the same invented numbers, not evidence about any treatment.
Do not read “50% lower” as “50 more people out of 100 benefited.” Ask for the baseline and the absolute difference. Also check that the measured event is the outcome you care about and that both groups use the same definition and timeframe.
NCCIH notes that a statistically significant difference can be too small to be clinically useful. A label such as “significant” therefore does not settle whether the change matters to a person’s daily life. Ask a qualified professional to explain the size, uncertainty and relevance of the finding rather than relying on the label alone.
Look for uncertainty and information about harms
Read the study’s limitations with the results. Ask which concerns could change the interpretation: a small sample, incomplete follow-up, a narrow setting or measurements that depend on participants knowing what they received. A precise-looking number does not remove those questions.
NCCIH recommends considering findings alongside earlier research and checking potential financial or reputational conflicts. A single paper rarely closes a question. If several articles reuse the same underlying studies, count that shared evidence once when thinking about how much independent support exists.
Check what was collected about unwanted effects and how it was collected. “No adverse events were reported” and “the study did not collect adverse-event data” are different statements. Neither can be turned into an unrestricted promise that an approach is safe for everyone.
Connect a study to the actual program carefully
Ask the program to identify the study behind its claim and explain the connection. Does it offer the same intervention with comparable staff, participants and duration? Was the named program evaluated, or is it citing research about a broad approach?
Keep a provider’s description, a participant’s experience and a research finding in separate rows. Each can help you understand something about care. None automatically supplies the missing evidence for the others. An appealing testimonial is still one account unless a wider, defined dataset is provided.
Take the claim record to the clinician who is helping you compare options. It is reasonable to leave with an unanswered question instead of a yes-or-no verdict. The purpose is to make the decision better informed, not to turn a family member into a research reviewer.
Turn a broad claim into a specific question
These examples are invented. They show how to ask for evidence without assuming the claim is either true or false. This worksheet is an editorial reading aid, not a validated scoring system.
| Part of the record | Example entry |
|---|---|
| Claim | “Our approach improves recovery.” |
| Definition | Which outcome defines recovery here, and when was it measured? |
| Comparison | What did the other group receive, if there was one? |
| People | Who was included, who was missing, and is this the same setting as the program being considered? |
| Evidence | Can I see the study or report and the limitations attached to that result? |
| Next discussion | Ask the assessing clinician how the evidence relates to the proposed care. |
Download the treatment-outcome claim record (CSV). Open it in a spreadsheet or print it. Keep your completed copy privately; this site does not collect your answers.
Related reading: Music in addiction treatment: findings and limits.
Continue with the next question
- What to ask before choosing a treatment program
- Outpatient, IOP, PHP, residential and inpatient care explained
- How to discuss whether therapy is helping
For public contacts and geographic questions, browse our local care guides.
Sources and editorial scope
This guide combines published source guidance with original preparation examples. No named clinician has reviewed this guide. The worksheets are not diagnostic tests, validated measures or assessments of a provider. No interviews or individual care arrangements were verified.
- NIMH: Finding help for mental illnesses
- ClinicalTrials.gov: how to read study results
- NIH: understanding clinical studies
- Cochrane: music therapy for substance use disorders
- Silverman: single-session music interventions and group relationships
- NCCIH: reading a scientific article’s results
Sources checked September 8, 2026. Read our editorial standards.