Skip to main content

Blood Pressure Records: Keep Each Pair and Measurement Time Together

Store systolic and diastolic blood pressure from the same measurement event together, with its actual observation time and relevant conditions. A later entry time, a corrected value and an average are different kinds of information. Never create a new BP pair by taking the latest systolic value from one attempt and the latest diastolic value from another; the resulting pair may never have been measured.

By Bractis · Updated · 8 min read

Doctors and clinic teams evaluating BP entry, repeat measurements, paper backfill and longitudinal displays in an Indian OPD.

Treat a blood-pressure pair as one attributable event

A BP record is more than two independent numbers. The pair belongs to a particular attempt, patient, time and set of conditions. If the numbers are saved or displayed independently, a screen can produce a convincing-looking value that no device or observer obtained.

HL7's FHIR R4 Vital Signs profile represents observations with a patient and effective time, and uses components for systolic and diastolic BP. It also provides for missing-value reasons. This is a versioned information model, not a clinical averaging rule or proof that a product implements it correctly.

Use a stable event reference internally so the pair remains connected across the entry form, trend, note and export. The reader need not see a technical identifier everywhere, but the system must be able to explain which observation a displayed pair came from.

Keep units explicit. BP values expressed in mmHg should not be mistaken for pulse rate or a numeric code. If an imported record uses a different representation, resolve its meaning before attaching it as a comparable reading.

Record relevant conditions such as arm, posture, method and source where they affect interpretation. An event measured at home by the patient and an event measured in clinic can both be useful while remaining different sources.

Source: HL7: FHIR R4 Vital Signs profile, v4.0.1

Use a worked example to expose a misleading display

In the original synthetic series below, attempt A at 18:05 is 148/92 mmHg and attempt B at 18:09 is 136/84 mmHg. A third attempt at 18:11 has a recorded systolic value of 144 but no established diastolic value. The partial attempt must remain partial.

A display of 148/84 would combine the systolic value from A with the diastolic value from B. Neither attempt produced that pair. Choosing a lower component from each reading can make the display look reassuring while discarding the underlying observation.

Similarly, filling the third attempt's missing diastolic value with 84 from B invents 144/84. An incomplete record should show the limitation or be resolved from reliable source evidence, not completed using the nearest available number.

The table deliberately includes a derived average as a separate row. That row is a summary of specified inputs, not a fourth cuff measurement. Keeping this distinction visible is necessary when counting observations or reviewing their chronology.

These values illustrate record behaviour only. They do not diagnose hypertension, establish a treatment target or determine the response needed by a particular patient.

On a narrow screen, scroll the table horizontally to read all columns.

Two complete BP events, one partial attempt and a labelled summary
RecordValue and timingMeaning to preserve
Attempt A148/92 mmHg at 18:05One complete pair from this attempt
Attempt B136/84 mmHg at 18:09A second complete pair; does not overwrite A
Attempt CSystolic 144 mmHg at 18:11; diastolic unavailablePartial information with its own limitation
Derived summary of A and B142/88 mmHg across the stated two-event seriesCalculated result, not a new measured pair
Invalid mixed display148/84 mmHgSystolic from A and diastolic from B were never measured together

Preserve observation time, entry time and amendment time

If the two complete readings are copied from paper at 18:15, their observation times remain 18:05 and 18:09. The documentation delays are ten and six minutes respectively. Labelling both readings 18:15 would erase the measurement sequence.

Use an explicit late-entry record when the observation time is known and preserve uncertainty when it is not. A remembered 'around six in the evening' should not become 18:00:00 as if the device supplied exact timing. Precision in a timestamp is not the same as accuracy.

A correction changes the record of an event; a repeat measurement creates a new event. If the source confirms that attempt A's diastolic value was mistyped, amend that attempt with an attributable reason. If the patient was measured again, record another event instead of editing A to the new result.

Do not make the correction time the new observation time. The clinical team may need to know both when the patient was measured and when the chart was corrected, especially if the earlier value influenced a decision.

Review downstream consequences. A corrected pair may affect a displayed average, a printed note or a message already sent. The clinician needs an appropriate way to recognise the change; silently altering the trend while leaving an old shared summary unexplained can create conflicting records.

Label an average and state exactly which readings it uses

For the two complete synthetic attempts, the systolic mean is (148 + 136) ÷ 2 = 142 mmHg and the diastolic mean is (92 + 84) ÷ 2 = 88 mmHg. This calculation uses A and B only. The partial third attempt is excluded from this illustrative two-event summary.

The calculation is an example of transparent reporting, not an instruction that every clinic should average those attempts or exclude a clinically important value. The clinician's measurement protocol and the actual context determine which observations are appropriate for a summary.

Show the number of included events, the time range and the method. A single label such as 'latest BP' should not switch unpredictably between the most recent actual pair and a calculated average.

Keep source readings available. A summary can hide variability, a change in posture or a different arm. A future reviewer should be able to inspect the original events without reconstructing them from a rounded average.

If an included event is corrected or removed from the summary for a documented reason, recalculate transparently and preserve the change. Do not leave a cached mean that no longer matches the displayed inputs.

Test the input form and every place that later shows BP

Use the synthetic A/B/C series in a software demonstration. Enter the pair together, add a partial attempt and inspect the header, note, graph, printed summary and export. Each should make the provenance and limitation understandable.

Test editing only one component of an existing event. That may be a legitimate correction, but the interface must not accidentally pair it with a component from another event. Check what the clinician sees before and after saving.

Rehearse two staff members working around the same visit and a late save returning after the patient context has changed. A system must preserve the correct event and patient; a technically successful response is not enough if it appears in the wrong chart.

In an original nine-case test, six outputs preserve the intended event, one combines components, one loses observation time and one is untested. The total is nine, with three unresolved outcomes. The clinic investigates each failure rather than describing all saved values as reliable.

Check unavailable data as deliberately as complete data. Some devices report an error or incomplete result. A blank diastolic value must not become zero, yesterday's value or a default normal value during export.

Explain the chosen display to clinicians and assistants. A reliable model can still mislead if staff believe a derived mean is the last physical measurement. Use concise labels that identify what the number represents without making the patient-facing document unnecessarily technical.

148/84 is not one of the measured pairs

The original complete readings are 148/92 at 18:05 and 136/84 at 18:09. Their illustrative mean is 142/88, labelled with its two inputs. A display of 148/84 is an invented cross-event pair, while a later partial attempt remains incomplete until its source is resolved.

Test paired blood-pressure records

Use these prompts in your clinic’s own notes. Record the result, responsible person and next action for each item.

Event identity
Keep the two components, unit, patient and observation conditions attached to one attempt.
Three times
Distinguish when the reading was taken, entered and later corrected.
Derived display
Specify which events an average includes, its time range and how corrections update it.
Acceptance series
Test complete, partial, corrected and concurrent entries through headers, notes, graphs and exports.

Questions doctors ask

Can the latest systolic and latest diastolic values be combined?

Only if they belong to the same measurement event. Taking components from different attempts can create a BP pair that was never observed.

What if only one BP component is available?

Preserve the partial result and its limitation. Resolve it from reliable source evidence or obtain an appropriate new measurement; do not borrow the missing component from another event.

What time should an average BP display show?

Identify it as a calculated summary with the included readings and their time range. Do not assign a new measurement time that implies another cuff reading occurred.

Should readings entered together from paper receive the same observation time?

No. Preserve their actual observation times when known and record the later entry separately. If the source time is uncertain, label that uncertainty.

Does using a FHIR profile prove the software handles BP correctly?

No. The profile describes a data model. Test the implemented entry, correction, calculation and output behaviour with concrete cases.

Sources and scope

Updated on . Original clinical-data integrity analysis using a versioned HL7 reference. It does not prescribe a BP measurement schedule, diagnostic threshold or treatment decision. No independent clinical or legal review is claimed.

Try the workflow
with Bractis.

Enter the fictional 148/92 and 136/84 series plus a partial attempt. Verify intact pairs, distinct observation/entry times and an explicitly labelled average across the note, trend and printout.

Request a clinic demo