How Do You Measure Community Health Worker Program Outcomes?

Which question should your CHW report answer first?

Choose the decision the report will support before selecting measures or building a dashboard.

CDC's 2024 evaluation framework moves from understanding context and describing a program through questions, evidence, conclusions, and action. It also emphasizes collaboration and using findings. The example measures here are our practical suggestions, not a CDC-required indicator set.^[1]

An operations supervisor may need to know which follow-ups are overdue. A funder may ask how many eligible people received a defined service during a reporting period. A program director may want to understand whether an access barrier is becoming less common. These questions need different records, denominators, and interpretation.

Write a sentence for each report: "We will use this information to decide whether to..." If nobody can finish that sentence, reconsider collecting the measure. Include frontline workers and, where appropriate, community representatives in choosing what useful progress looks like. A dashboard should support a decision rather than simply display every available field.

What belongs in a CHW metric dictionary?

For each measure, define the unit, inclusion rules, time period, data source, owner, and known limitations.

Keep the dictionary beside the report specification and record changes. If you change the definition of "reached" halfway through a year, either recalculate earlier periods consistently or label the break in the series. Otherwise an apparent performance improvement may only reflect a new counting rule.

How do you avoid counting the same support more than once?

Separate people, encounters, activities, and note versions, then report each using its own counting rule.

Fictional example: During one month, 40 distinct participants have 72 completed encounters. Those encounters contain 95 recorded activities, and staff create 80 note versions because some notes are corrected. The report can accurately show 40 people, 72 encounters, and 95 activities if those definitions are clear. It should not claim that 80 note versions mean 80 delivered services.

Also define which date controls inclusion. Service date, entry date, and review date answer different questions. If a worker completes a note in the next month, a report based on entry date may move the work into the wrong service period. Choose and document the rule that fits the report's purpose and applicable requirements.

Use a reconciliation sample. Pick a small set of records and explain how each contributes to the totals. Include a correction, a canceled appointment, a group encounter, and a person participating in more than one program if those cases exist in your service model. Investigate discrepancies before sharing the final report.

How should you report referrals without overstating success?

Define the referral cohort and show known results, pending cases, and unknown outcomes separately.

Fictional example: A program sent 20 referrals in a defined cohort. Eight have confirmed service receipt, four are scheduled, three were declined, and five have unknown outcomes. Confirmed receipt is 8 of 20, or 40% of the full cohort. Reporting 8 of the 15 known outcomes, about 53%, answers a different question and should never silently replace the first calculation.

State how long you allowed for follow-up and what counts as confirmation. Participant report and partner confirmation can both be useful information, but they are different evidence sources. Record which was used. If service receipt cannot be verified, say unknown rather than assuming that a referral was successful because staff sent it.

Choose whether the report follows referrals initiated in the month or results confirmed in the month. A referral sent in September and completed in October can appear differently under those approaches. Either approach can be useful when labeled correctly; mixing them produces misleading percentages.

How do you distinguish activity from meaningful change?

Pair service counts with measures tied to the program's intended benefit, and describe what the data can and cannot explain.

Completing more encounters can show additional activity. It does not by itself establish better access, knowledge, confidence, or health. For a program focused on appointment preparation, a useful evaluation question might be whether participants report being better able to ask their intended questions. Select the measure and collection method with the people who will use the finding.

Do not infer that the program caused a change simply because the change occurred after enrollment. Other services, eligibility changes, household circumstances, and differences in who responded may affect the result. Use wording such as "participants reported" when that accurately describes the evidence, and explain the evaluation design before making causal claims.

Show missing information. If only 12 of 40 participants completed a follow-up survey, report that response count alongside the result. Consider whether the respondents differ from people whose follow-up is missing. A high percentage from a small, selected group should not be presented as the experience of everyone served.

Who should review a CHW report before release?

Separate the checks for data, interpretation, and the intended audience.

One person may fill more than one role in a small program, but the questions remain distinct. Keep the approved version and explain later revisions. A corrected total should have a reason that readers can follow.

What does a reconciled monthly CHW report look like?

Start with a small set of defined counts and explain how they relate without treating them as interchangeable outcomes.

Here is a fictional monthly example. A program serves 40 distinct participants, records 68 completed encounters, and makes 22 unsuccessful contact attempts. Six participants receive no completed encounter that month but have a recorded attempt. Define served before publishing the total: if the program uses completed encounters as its service measure, those six people belong in a separate attempted-contact count. A roster of 46 people is not the same as 46 people receiving a completed service.

Suppose the 68 encounters generated 72 note records because four encounters required amendments. The service count remains 68 under this example's definition. The additional records describe the history of documentation, not four additional services. A software export that counts every note version as an encounter would overstate activity. Ask the reporting owner to trace a sample of the total to the actual service events.

Now consider 20 referrals created during the month. At the cutoff, 8 have confirmed service receipt, 5 are accepted with service still pending, 4 are waiting for a response, and 3 were declined or closed for another documented reason. These categories total 20. Under a measure defined as confirmed service receipt among all referrals created, the result is 8 divided by 20, or 40 percent. That calculation should not silently exclude unresolved referrals to produce a larger percentage.

These numbers are original examples, not benchmarks. Your contract may use a different reporting population or definition. State the selected definition, reporting period, cutoff date, and evidence standard beside the measure. A reader should not need to guess whether a result describes all enrolled people, people contacted, or only those with completed follow-up.

How do you compare CHW outcomes across months fairly?

Compare like populations and follow-up windows, and show missing information alongside measured results.

If January participants have had three months to complete a referral but March participants have had only one week, a direct completion-rate comparison is misleading. Decide whether the question concerns activity during a calendar month or outcomes for a group enrolled during that month. Those are different views. Label the report so readers can tell whether it follows a cohort over time or summarizes all activity within a period.

For a fictional goal measure, suppose 30 people agreed to a follow-up review. Twenty completed it, and 12 of those reported achieving the selected goal. Reporting 60 percent alone describes 12 of the 20 respondents. It does not show what happened for all 30 people due for review. Present both the 20-of-30 follow-up completion and the 12-of-20 result, with the remaining outcomes identified as unknown. Do not automatically treat missing responses as either success or failure.

Check whether the participant mix or service changed before interpreting a trend. A new referral partner might send people with different needs. A worker vacancy might change contact frequency. A revised question might change how a goal is recorded. Describe these conditions rather than attributing every change to program effectiveness. Strong causal claims need an appropriate evaluation design, not just a dashboard line moving upward.

How do you turn CHW reporting into a management decision?

Choose a small number of actionable questions, verify the evidence, and assign a follow-up decision.

A monthly meeting can start with three questions: where are people waiting, where is follow-up breaking down, and what change should the team test? Bring counts and a few appropriately handled examples. If referrals are waiting because a partner has no capacity, asking workers to document faster will not solve the delay. Identify the operational cause before choosing an intervention.

Assign a named owner, a specific change, and a review date. For example, the team may test a revised referral form with one partner and compare returned requests before and after the change. Record other changes that could affect the result and avoid claiming proof from a small uncontrolled comparison. The purpose is to learn whether the process improved enough to continue, revise, or stop the test.

Keep externally shared reports within the organization's approved privacy and reporting process. Small groups or detailed stories may identify someone even when a name is removed. Ask the responsible reviewer to decide what can be shared. Use the reporting worksheet to document definitions and limitations before creating a polished chart; presentation should make the evidence clearer, not conceal uncertainty.

What should a CHW monthly report actually contain?

Use a compact report that states the question, definitions, counts, limitations, and decisions supported by the evidence.

The following is an original report outline for a fictional program. Adapt it to your funder's required measures and your organization's approved reporting process. It is not a CDC reporting form. A useful first page names the period, the population included, the cutoff date, the report owner, and the main decision the report supports. Readers should be able to tell what the numbers cover before seeing a performance conclusion.

Populate the activity and referral sections from the reconciled example above. Carry its definitions into the report instead of copying numbers into an unlabeled chart. Add an interpretation such as: eight of twenty referrals have confirmed service receipt at the cutoff; nine remain pending or awaiting response, so their final outcomes are unknown.

End with an action table. Assign a person to investigate the oldest unresolved referrals, identify the partner or process being reviewed, and set the next review date. Record any reporting changes that could affect comparison with prior months. If a definition changed, either recalculate earlier periods consistently or explain the break in the series. A more attractive chart should not obscure a change in what is being counted.

How do you calculate follow-up timeliness without hiding overdue work?

Define the start, endpoint, and eligible records before calculating elapsed time, then report unresolved work separately.

Suppose a fictional program wants to know how long it takes to make the first successful contact after accepting a referral. Define accepted as the recorded date the program takes responsibility for the referral, and successful contact as the first completed conversation under the program's definition. Choose calendar or business days explicitly. Do not change the starting event between months because a different field is easier to export.

For five completed contacts with elapsed times of 1, 2, 2, 5, and 10 calendar days, the mean is 4 days because the total of 20 days is divided by five. The median is 2 days, the middle value in the ordered list. Both calculations are correct, but they describe different aspects of the same set. Report the number of observations and choose the measure that answers the management question.

Now add three accepted referrals with no successful contact by the cutoff. Do not insert zero days for those records. Zero would incorrectly suggest immediate contact. Report their unresolved status and current ages separately, such as 3, 7, and 12 days since acceptance. A measure based only on completed contacts can look favorable while the oldest pending work remains untouched. Review the completed and unresolved groups together before drawing a conclusion.

Document how weekends, missing timestamps, reassigned referrals, and duplicate entries are handled. The rules should come from the question and the data available, not from a desire to produce a better-looking number. If an acceptance date is missing, identify the record as not calculable under this definition and resolve the source issue through the approved process.

Which data checks should happen before a CHW report is shared?

Reconcile totals, inspect impossible or missing values, and trace a small sample back to the underlying records.

Begin with arithmetic checks. Mutually exclusive referral statuses should sum to the intended referral population. A percentage should use the stated numerator and denominator. If categories overlap, say so and do not present their sum as a distinct-person total. The same participant may receive several service types, so adding each type's participant count can double-count people.

Review obvious inconsistencies such as an end time before a start time, a completed result with no supporting record, or a service date outside the selected period. Do not automatically repair questionable data by guessing. Assign the issue to the appropriate owner, preserve the original evidence, and explain any unresolved limitation that affects the report. A clean-looking export is not a substitute for an accurate source record.

Trace a few selected totals through individual examples. Can the reporting lead explain why a record was included, which event it represents, and how a correction is handled? Document the selection method and any limitations rather than calling a small spot check a comprehensive audit. Use repeat findings to improve definitions, forms, or training. Reporting quality improves when the underlying workflow captures the right fact at the right time.

What should you ask software to demonstrate?

Ask the vendor to trace one report total back to the underlying example records and explain the filters and exclusions.

Smores Health has reporting workflows alongside participant records and documentation. Bring a blank funder report or a list of required measures so the walkthrough can establish which fields, filters, and exports fit your program. A custom measure should be confirmed explicitly rather than assumed from a general dashboard demonstration.^[2]

Frequently asked questions

Sources and product pages

Sources support the specific guidance cited above. Program guidance is not a universal requirement. Product pages describe capabilities to verify against your organization’s needs.

1. CDC Program Evaluation Framework: Centers for Disease Control and Prevention. 2024 framework. Checked October 9, 2026. Example metrics and calculations below are Smores Health planning suggestions, not mandated measures.
2. Smores Health: Understand: Smores Health. Product overview. Confirm your program configuration, included services, and rollout scope in a walkthrough and written agreement.
3. Smores Health: Operate: Smores Health. Product overview. Confirm your program configuration, included services, and rollout scope in a walkthrough and written agreement.