Reading correlations in detail

A deeper look at how to read and interpret the correlation data Tended shows you.

What counts as a correlation

A factor needs to appear on at least 3 flare days and 3 other days, with a meaningful difference between the two groups. This prevents small amounts of data from producing misleading results.

What the numbers mean

Each correlation shows how often you recorded a flare (or a good day) when a given factor was also present. “Flared 5.5x more often with this” describes the difference between those groups in your records; it does not mean the factor caused the flare. When the ratio is below 2x, you’ll see a percentage instead (like “Flared 18% more often with this”).

Confidence

The three small dots next to each correlation summarise the amount of supporting data at a glance.

Low confidence (one dot) means 3–5 entries per group. Growing confidence (two dots) means 6–15 entries. High confidence (three dots) means 15 or more entries with a large observed difference. These labels describe the records available to Tended; none proves cause, clinical significance, or that the pattern will continue.

Tap any correlation to see the exact numbers behind it (like “Based on 9 days with and 18 days without · Growing confidence”). This lets you judge how much weight to give it and discuss it with a healthcare professional.

Delayed effects

Some recorded factors and outcomes happen on different days. For each flare or good day, Tended looks back 1–7 days to see whether a factor was recorded consistently beforehand. When it finds a delayed association, you’ll see a small indicator on the right (like “1d” or “4d”) showing the time window, and the description spells it out (e.g. “Flared 2.2x more often within 4 days of logging this”). This is a timing observation from your journal, not evidence that the factor caused the outcome.

Context days

If you have Major Events, correlation detail views include a context-days control. Use All days for the full picture, Exclude event days to remove days when a Major Event happened, or Exclude recovery windows to also remove recorded recovery periods. This helps separate routine patterns from unusual outliers.

Stable medications

If you take several long-term medications every day, they may create your baseline rather than explain day-to-day changes. Focused medication view leaves quiet baseline medications out unless they are PRN, skipped, tapering, always-watch, or high-confidence medication signals. Chart views let you switch to All meds whenever you want the full medication picture. Flow views keep quiet baseline medications out so the map stays readable.

Dose-specific signals

When your journal contains multiple doses, quantities, or dose times, Tended can keep those details separate from the general medication signal. For example, a PRN medication can appear as the medication overall and as a specific dose pattern when that is the clearer comparison. Duplicate-looking rows are avoided where the specific dose adds no extra meaning.

Medications that show as hurting

If a painkiller shows up under “worse,” it likely means you take it when you already feel bad — not that it’s causing harm.

The chart view

Correlations are shown as a diverging bar chart — factors that may be hurting appear on the left, and factors that may be helping appear on the right. The bar opacity reflects confidence level. Tap any bar to see the sample size and confidence label below the chart, so you can judge data quality at a glance.

Explain this

Tap What this means on an insight view for a plain-English summary of the information visible to you. On individual premium correlations, Explain this finding describes that specific result. Explanations clarify that correlation does not mean causation. With AI enhancements enabled, the same summary can be rewritten in a warmer tone using Apple Intelligence, entirely on your device.

Sharing a correlation

When you expand a correlation, you can tap the share button to export it as an image card. This is useful for sharing a specific finding with your doctor, a support group, or anyone you trust.

Worked example

Say you see Dairy showing “Flared 5.5x more often within 3 days of logging this” with a “3d” indicator and two confidence dots.

The description tells the full story: in the 3 days before each flare day, dairy had been logged — and you flared 5.5 times more often when it was present than when it wasn’t. The “3d” indicator is a quick visual marker for the delay window. Two dots means the pattern is based on 6–15 entries per group — it’s stabilising but not yet conclusive.

Tap the row to see the exact sample size and confidence level (e.g. “Based on 9 days with and 18 days without · Growing confidence”). That’s the raw data — you decide how much weight to give it.