How to Audit an Imported Health History for Boundary Gaps

Why the edges deserve a separate check

A health-data import can look complete in the middle and still lose information at its beginning or end. The transfer may start after the source archive's first record, stop before its last, or omit a partial day that falls outside the export window. A continuous-looking chart does not prove that every interval crossed the boundary.

Check the edges before interpreting the middle. Boundary records show whether the transfer starts and ends where expected, whether dates shifted, and whether the destination retained enough context to compare the imported history with current records.

Research on missing wearable-data patterns describes gaps related to timing, device behavior and handling. That makes the location of missing data useful evidence during an audit, not merely a cosmetic issue.

Record the source window first

Before opening the destination chart, write down the source file's earliest and latest timestamps. Include the time zone or offset when available. Also note whether the export covers whole calendar days, rolling periods, or only records that passed a source system's own filters.

This gives the audit a reference point. Without it, a first visible date in the new chart can look plausible even if several records were excluded before the transfer began. Preserve the original export unchanged so you can compare it again if the destination's display changes.

If the source provides a count of observations or a transfer summary, save that too. A file with rows but no documented coverage window may need a second check against the source interface or a separate manifest.

Compare the first imported record

Find the first record in the exported file and the first record in the destination. Check the timestamp, metric name, unit, source label and any status such as measured, estimated or manual. The dates should agree after accounting for a documented time-zone conversion.

A date-only display can conceal a shifted hour. That matters when the underlying event occurred overnight or near midnight. Keep the original timestamp beside the converted one if the destination only shows local calendar dates.

The wearable-data quality review explains that non-wear, user handling and data standards affect what can be inferred from person-generated records. A first-date match is only one part of the check; also confirm that the record actually represents the expected period.

Compare the last imported record

Repeat the same comparison at the other end. Identify the source archive's final observation and the destination's latest imported one. Check whether the export ended naturally, stopped at a cutoff, or contains a final partial period that the destination did not accept.

A recent-looking final date can still be incomplete. If the source recorded only one short interval on that date, a daily chart may present it beside full days without making the difference obvious. Record coverage rather than assuming that each day contains comparable amounts of data.

If the transfer summary says records were skipped, compare the listed rows with the source. A skipped row may reflect an unsupported field, duplicate detection, invalid formatting or a rule in the destination. Do not infer the reason from an empty space alone.

Treat an overnight interval as one event

Sleep and recovery records can cross midnight. One system may label an overnight interval by the date it began; another may use the date on which it ended. A migration can therefore make a night appear to move even when its underlying timestamps are unchanged.

Choose a single rule for the audit. Compare the start and end times, not just the displayed date. If the destination split one interval into two dates or merged adjacent intervals, record the transformation and preserve the original boundaries where possible.

The existing guide to checking time zones before merging overnight health data gives a focused checklist for this edge case. It is especially useful when the first or last imported night sits near a time-zone transition or export cutoff.

Look for a missing interval between two valid records

After checking the endpoints, inspect the days immediately inside the transfer window. A beginning and ending date can match while a middle block is absent. Mark both the duration and the clock-time location of each unexplained interval.

Do not assume every gap has the same cause. The analysis of missing-data mechanisms notes that missingness can vary over time and between signals. A gap in one metric does not prove that every other metric is absent or that the person was inactive.

Keep missing distinct from zero. Some formats use an explicit absent value; others may encode a zero or omit a row. The meaning depends on the metric and source definition. If that definition is unavailable, mark the point as unresolved instead of converting it into a measurement.

Search for duplicate boundary records

Transfers sometimes overlap by a day or more because the source and destination use different cutoff rules. Compare records around the join for matching timestamps, identifiers and metric definitions. Equal values are not enough to establish duplication, and different values are not enough to prove that two independent observations exist.

A useful duplicate check compares source identity, timestamp, metric, unit and original record identifier when available. If a row was transformed, preserve both the original and transformed values and note why the destination created another entry.

Duplicate-looking records can affect daily summaries if they are counted twice. Conversely, deleting one without checking its origin can remove a legitimate repeat measurement. Keep the audit reversible and document any action taken.

Verify time zones and daylight changes

Confirm whether timestamps are stored in UTC, local time or an offset-bearing format. Then compare the source and destination display rules. A day with a repeated or skipped clock hour can expose assumptions that remain hidden on ordinary dates.

The FHIR observation structure treats an observation's effective time as an explicit data element. A personal-data audit can use the same idea: preserve when the observation applied, not only the date on which it was imported.

If a conversion is necessary, keep the raw timestamp and the converted display time. Do not overwrite the original with a guessed local time. A clear conversion rule makes the boundary reproducible later.

Compare coverage, not just row counts

A total record count can hide uneven coverage. One file may contain many short observations in one period and very few in another. When the source exposes duration or sampling intervals, compare them around the boundary and through the first full day after it.

The activity-pattern imputation study found that missing duration and timing can affect calculated patterns, and that one method did not solve every gap. For an audit, note which time intervals are observed, sparse, filled or unknown.

If coverage cannot be measured exactly, describe what the export actually shows. “Records are present on this date” is more precise than “the whole day is complete.”

Check whether values were filled or recalculated

Some destinations interpolate across short gaps, aggregate multiple samples, or calculate a daily summary after import. These operations can make the chart appear continuous. Ask whether the displayed segment contains direct observations or derived values, and whether the transformation is marked.

The wearable time-series study compares approaches to missing data and shows why processing choices belong in the interpretation. A smoothed line is not a record of continuous measurement unless the source confirms that coverage.

Keep imported measurements, estimates and recalculated summaries in separate categories. If the application does not label them, add a companion note that identifies the date range and any known processing.

Audit metric by metric

A complete sleep summary does not prove that continuous heart-rate samples arrived for the same night.

Make one row per metric with the source window, first and last timestamps, visible gaps, duplicates, unit and status. This simple inventory can show whether only one part of the history needs repair or whether the entire export has uncertain coverage.

Research on imputation methods for wearable activity data discusses the assumptions behind filling missing periods. The method that makes sense for one series may not be suitable for another.

Keep the original and the audit notes

Store the original export in a read-only location. Keep a separate copy for any sorting, conversion or comparison. Write down the export date, source period, destination, timezone rule, transformations and unresolved gaps.

This gives a later reviewer a way to distinguish a source omission from a transfer issue or a display rule. It also prevents a future migration from starting with an already altered file whose original boundaries are unknown.

Avoid adding personal stories or assumptions to the audit. A neutral note such as “no source record found from 02:00 to 05:00” is more useful than a guessed explanation.

Decide what can be compared

After the boundary check, group records by compatible definition, unit, source and coverage. A raw measure with the same method may support a cautious before-and-after review. A composite score needs additional information about inputs and calculation rules.

If an interval is incomplete, mark it visually and exclude it from comparisons that require full coverage. You can still inspect neighboring observations as context, but do not present the missing period as measured or infer a smooth pattern across it.

The audit is not a clinical interpretation. It tells you what the transferred record contains and which comparisons remain uncertain. If a health concern exists, discuss it with an appropriate professional rather than relying on a migration check.

FAQ

Which dates should I inspect first after importing wearable data?

Start with the first and last dates in the source export, then check the intervals immediately inside those boundaries. Compare timestamps, time zones and coverage rather than relying on date labels alone.

Can an overnight record move to another date during import?

Yes. Systems can assign an overnight interval to its start date or end date. Compare the underlying start and end timestamps and document the destination's date rule.

Should missing intervals be filled before I compare the history?

Not automatically. Mark the gaps first and identify whether any displayed values are estimated. Only compare records with compatible coverage and clearly understood processing.