Skip to main content

Chronic-Care Registries for Clinics: Define the Cohort Before Reading the Dashboard

A useful chronic-care registry starts with an explicit patient cohort, a defined measurement window and clear categories for missing, uncertain and clinically reviewed information. Count people rather than visits, keep the reporting date fixed and assign action to exceptions. A dashboard percentage cannot be interpreted until its denominator and data gaps are visible.

By Bractis · Updated · 9 min read

Indian family physicians, specialists and clinic managers building diabetes, hypertension or other continuing-care patient panels.

Decide which question the registry should answer

A registry can support several different tasks: finding patients with an agreed follow-up need, identifying missing information, monitoring a defined measure or understanding the clinic's active panel. One list should not be assumed to answer all of them.

Write the question in operational terms. 'Which eligible patients lack a usable measurement in the clinician-defined window?' is more precise than 'Who is uncontrolled?' The latter may hide assumptions about diagnosis, measurement source, units, timing and the clinical criterion.

WHO's HEARTS monitoring module uses defined indicators and data-collection tools for primary-care cardiovascular programmes. The relevant lesson is that a measure requires a specification. The fictional clinic calculation below is original and is not a reproduction of a WHO indicator.

The responsible clinician defines the clinical rules and exceptions. This article does not prescribe HbA1c, blood-pressure or other treatment thresholds. A software default should not quietly become the clinic's clinical policy.

Source: WHO HEARTS: Systems for monitoring, 2018

Define who belongs in the cohort on the reporting date

Specify inclusion by person, condition and relationship to the clinic. A confirmed condition, a provisional concern and a family-history entry should not enter the same disease cohort automatically. Review how coding and narrative information are converted into registry membership.

Decide how transfers, deaths, duplicate charts, inactive patients and people seen only for a one-off consultation are handled. These rules should be explicit and appropriately authorised. Removing difficult-to-contact patients merely to improve a percentage would distort the panel.

Use a reporting cutoff and preserve the definition used. A dashboard viewed in November may legitimately differ from an October snapshot because new information arrived. Recomputing an old report with today's data should be labelled as a restatement rather than appearing to be the original result.

Count people once for the stated measure. Multiple visits, phone numbers or uploaded copies of a report do not create additional people. Conversely, several family members sharing a number remain separate patients with separate care needs.

Keep exclusions reviewable. An exclusion should have a reason consistent with the definition, not simply remove a row from view. The clinic may still owe an excluded person a follow-up task under a different workflow.

Define which measurement is eligible before choosing a result

A result needs the right patient, observation date, unit and interpretation context. The upload date is not automatically the date the measurement occurred. A recent upload of an old report should not make the patient appear freshly reviewed.

Specify how the registry handles more than one result in the window. 'Most recent' needs a rule for actual observation time, amended reports and ties. Selecting the most favourable value would answer a different question and can conceal deterioration.

Separate clinic observations, patient-reported home readings and external laboratory results where their meanings differ. An eligible measurement for one indicator may be unsuitable for another. A graph can display several sources while a defined measure uses only the appropriate subset.

Missing, unreadable and clinically uninterpretable data require their own states. They are not automatically normal, abnormal or evidence that the patient did not attend. A patient can attend without the relevant assessment, and a result can exist without a completed clinical review.

Work through the denominator before publishing a percentage

Consider this entirely fictional snapshot. The clinic begins with 320 chart rows. Identity review finds twelve duplicate rows, leaving 308 people. Eighteen people are excluded under the documented cohort definition, leaving 290 eligible people. These figures illustrate reconciliation, not a recommended clinic benchmark.

Of the 290 eligible people, 230 have a usable measurement within the defined window and sixty do not. Among those 230, 150 meet the clinician-approved measure definition and eighty do not. The categories are mutually exclusive for this example and sum to the eligible cohort.

The percentage among measured people is 150 ÷ 230 × 100 = 65.2%. The proportion of the whole eligible cohort with documented evidence of meeting the definition is 150 ÷ 290 × 100 = 51.7%. Both calculations can be correct, but they answer different questions.

Missing eligible measurements account for 60 ÷ 290 × 100 = 20.7%. Do not relabel those sixty people as failing the clinical criterion or silently remove them from all reporting. Show the missing-data category and the action required to understand it.

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

Fictional registry snapshot with explicit denominators
Count or measureCalculationInterpretation
Unique people320 chart rows − 12 duplicate rows = 308Identity reconciliation before cohort selection.
Eligible cohort308 people − 18 defined exclusions = 290People included at the stated cutoff.
Usable measurements230 of 290 peopleMeasurement availability in the defined window.
Meeting definition among measured150 ÷ 230 × 100 = 65.2%Uses measured people as the denominator.
Documented meeting definition in cohort150 ÷ 290 × 100 = 51.7%Uses all eligible people as the denominator.
Missing usable measurement60 ÷ 290 × 100 = 20.7%A data or care-follow-through category requiring review.

Turn exceptions into work without turning the dashboard into triage

The registry should help a responsible clinician or care team decide the next task. It should not independently select treatment or assume that every numerical exception has the same urgency. Current symptoms and circumstances may require assessment outside the routine outreach queue.

In the fictional cohort, the eighty people not meeting the defined measure and the sixty with missing measurements need different first questions. The former may need clinical review of an available result; the latter may need source reconciliation, clarification of outside care or another clinician-selected action.

Assign ownership and track what actually happened. An attempted call, delivered message, booked appointment, attended consultation and completed assessment are distinct outcomes. The registry membership and the outreach task can be connected without treating one as the other.

On shared family phones, use the appropriate identity and communication process before discussing the individual patient's information. An answer from a relative should not automatically close a clinical task or authorise a detailed disclosure.

Keep unsuccessful contact visible with a next action under the clinic's policy. Do not erase the patient from the denominator because the first attempt failed.

Sample the underlying records before trusting a trend

Choose examples from each category and trace them back to source evidence: included and excluded people, measured and missing, qualifying and non-qualifying values. Review the rules as well as the rows. A correct-looking chart can conceal a systematic unit or date error.

Use targeted examples to investigate known risks, but do not call them a representative error rate. If the clinic selects ten suspicious rows and finds four problems, that does not establish that forty percent of the whole registry is wrong.

Compare monthly snapshots with consistent definitions. A change in the measured proportion may reflect new data, altered inclusion rules or corrected duplicates rather than a change in patient health. Annotate definition changes and avoid presenting an incomparable series as a continuous improvement trend.

Protect the distinction between record correction and cohort manipulation. Fixing a wrongly linked report should update the appropriate history and recalculation trail. It should not become an unexplained adjustment made only in the dashboard.

Keep a short specification beside the dashboard

A practical registry needs an owner who can explain its rules and an operational owner who can maintain unresolved work. The same person may hold both roles in a small clinic, but the specification should be understandable to the next clinician or staff member.

Use the original outline below before enabling a new measure. It is deliberately separate from treatment guidance. It documents how the clinic will count and interpret information already governed by the appropriate clinical decisions.

Review the specification when terminology, clinical policy, laboratory mapping, patient membership or software behaviour changes. A registry is useful only while its categories continue to mean what the clinic thinks they mean.

The apparent improvement came from a changed denominator

A fictional clinic reports a higher monthly percentage after excluding patients with no recent measurement. The measured patients' results have not changed. Review shows that the new report answers a different question from the earlier whole-cohort report. The clinic restores labelled denominators, displays the missing-data count and keeps the two series separate. The corrected report supports action without claiming an improvement in clinical outcomes.

Clinic Registry Definition and Reconciliation Sheet

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

Cohort
Reconcile chart rows to unique eligible people and document every exclusion rule.
Measurement
Trace observation dates, units, source and amendments for a sample of included results.
Denominator
Calculate the measure among measured people and among all eligible people; explain the difference.
Action
Separate clinical review, missing-data clarification and outreach completion for named owners.

Questions doctors ask

Can the clinic use the software's default disease threshold?

Only after the responsible clinical team confirms that the rule, context and exceptions are appropriate. A dashboard default is not a substitute for clinical policy or individual assessment.

Should missing measurements count as normal or uncontrolled?

Neither assumption is justified. Keep missing or unusable information explicit and define the appropriate follow-through. The meaning depends on why the data is absent and the clinician's assessment.

Should several visits by one patient increase the registry denominator?

Not in a person-based measure. Count the individual once under the defined cohort rules, while retaining the separate encounters and observations that support the measure.

Does a delivered recall message complete the registry task?

It establishes delivery only. A clinical review, appointment or assessment may still be outstanding. Define the task's real completion event and retain failed or incomplete follow-through.

Which percentage in the worked example is correct: 65.2% or 51.7%?

Both are correct for their stated denominators. The first uses 230 measured people; the second uses all 290 eligible people. Showing the sixty missing measurements prevents the figures from being interpreted as equivalent.

Sources and scope

Updated on . This guide explains registry definitions and data interpretation. Its numbers are fictional and its categories do not prescribe clinical thresholds, treatment or urgency. No independent clinical or legal review is claimed.

  • WHO HEARTS: Systems for monitoring, 2018

    Primary-care monitoring framework with defined indicators. The fictional registry calculation here is original and does not reproduce a WHO indicator or clinical threshold.

Try the workflow
with Bractis.

Reconcile a fictional registry from chart rows to unique people, change one measurement date and inspect the denominator. Trace one missing-data task and one clinical-review task through their distinct completion events.

Request a clinic demo