The Detail
Continuous glucose monitors estimate glucose every few minutes, but a report is not guaranteed to contain every possible reading. A sensor may be warming up, ending its wear period, temporarily disconnected, out of communication range, or unable to produce a value. Some systems can backfill certain gaps and others cannot. An export may also exclude days or readings under product-specific rules. The resulting chart can still draw a smooth-looking summary.
Missing data matter because statistics use observed values. If a five-minute system could theoretically contribute 288 readings in a complete 24-hour day, a day with 144 retained readings represents only half that opportunity. A report may calculate a mean from those 144 values without inserting visible blanks into the headline. That arithmetic can be correct for the observed set while still describing less than the full day.
Not All Gaps Are Equal
A random scattering of short missing intervals may affect a summary differently from one uninterrupted overnight gap. Repeated gaps at the same time each day can remove an entire part of the daily pattern. Missing weekends can change the balance of days. This is called the pattern of missingness: not just how much is absent, but where the absence falls relative to the structure being summarized.
The report rarely proves why each gap occurred. A blank segment may indicate missing sensor values, a display filter, an export problem, or a break between sensors. User-entered events may be absent even when glucose values remain. Label only what is visible. If the legend does not define a gap, do not assume that it represents a particular behavior or device failure.
Find the Denominator
Look for phrases such as sensor active, percent time active, data captured, readings available, days with data, or wear time. These may use different denominators. Percent of calendar time, percent of expected readings, and number of days with at least one value are not equivalent. A report can claim data on 14 days even if several days contain only short segments.
If the report gives a count of readings, compare it only with the sampling schedule documented for that device and report type. Some sensors record at different intervals or store data differently when a receiver is not nearby. Do not impose a five-minute expectation on every product. When the denominator is not stated, record that omission rather than manufacturing a coverage percentage.
How Gaps Change Displays
A line chart usually breaks where values are absent, but smoothing or aggregation can make a gap less obvious. An average daily profile may calculate each clock-time slice from a different number of days. A percentile band can narrow at noon simply because fewer observations contributed, not because the underlying process became more consistent. A heat map may use a neutral color for no data that resembles a midrange value.
Tables can hide gaps too. A daily average may appear for a day with only a few observed hours. A weekly mean may weight each reading equally, which gives well-covered days more influence, or it may average daily means, giving sparse and full days equal influence. Neither rule is inherently the one a reader should assume. Find the definition in the report or product documentation.
A Worked Example
Illustrative data, not patient results.Teaching Export H covers four fictional days from a device scheduled to store one value every five minutes. A complete day would therefore offer 288 possible readings in this simplified exercise. Day 1 has 286 readings, Day 2 has 280, Day 3 has 144, and Day 4 has 72. The total is 782 of 1,152 possible readings, or about 67.9% coverage. This percentage describes the invented export, not a clinical standard.
The daily map shows that Day 3 is missing midnight to noon and Day 4 is missing 6 a.m. to midnight. A single 67.9% figure cannot reveal that pattern. If the retained readings are concentrated at particular hours, the overall mean will describe those hours more strongly. The exercise does not guess what the absent values would have been. It identifies which parts of the window the report cannot show.
| Day | Possible readings | Retained readings | Visible gap |
|---|---|---|---|
| Day 1 | 288 | 286 | Two isolated readings |
| Day 2 | 288 | 280 | Several short intervals |
| Day 3 | 288 | 144 | Midnight to noon |
| Day 4 | 288 | 72 | 6 a.m. to midnight |
| Total | 1,152 | 782 | Coverage is uneven by clock time |
Compare Coverage Before Headline Numbers
When placing two summaries side by side, first align their start and end timestamps, time zones, planned duration, sensor model, and data-coverage definition. Then inspect a daily coverage plot if available. A report with more readings is not automatically more representative; it could still miss the same hours every day. Conversely, a few short random gaps may be visually prominent without dominating the window.
Only after the coverage structures are clear should averages or variation measures be compared. Add a note if one report includes a warm-up period, partial day, sensor change, or manual exclusion. Software updates can also alter backfill or calculation behavior. Preserve the report version and generation date so a later export can be understood.
What It Does Not Tell You
A gap does not reveal the missing glucose values. Filling it with a straight line, the day’s mean, or values from another day creates invented data and can understate uncertainty. The observed summary cannot prove that the missing period was higher, lower, or similar. It also cannot establish why data disappeared or who was responsible for the interruption.
Coverage information does not diagnose a condition, set a personal target, or support a treatment change by itself. A sparse report may still contain information a healthcare professional finds useful, and a nearly complete report still has measurement limits. The reader’s role is to describe the dataset accurately. Questions about safety, device use, and personal care belong with the product instructions and qualified professionals.
A Gap Audit in Five Steps
First, write the exact window and time zone. Second, copy the report’s active-data or coverage statement with its wording. Third, scan each day for long gaps and repeated clock-time gaps. Fourth, read footnotes for warm-up, replacement, exclusions, or synchronization rules. Fifth, note which statistics are calculated from available readings and whether the denominator is defined. Keep unknowns visible.
For a professional discussion, phrase findings as document observations: the report contains no values from midnight to noon on Day 3, or the report states 782 retained readings. Avoid claims about what happened during the blank. If a device repeatedly fails to provide expected information or displays an alert, consult its current labeling and appropriate support or healthcare channels rather than relying on a generic chart explanation.
- Window and time zone
- Coverage metric and denominator
- Location of gaps by day and hour
- Product or export notes
- Statistics affected by available-only data