Published
- Central Valley residents
- Clinicians
- Researchers
- Community organizations
Prepared · sources accessed (UTC) · physician review still to be recorded
A map of clinic addresses can look authoritative while leaving the practical access question unanswered. A resident needs a particular service, at a usable time, with an acceptable payment route and a way to get there. No single public directory reliably establishes all of those facts.
This guide is a protocol for a bounded study: one county, one service, one observation period. It presents no completed county findings or provider rankings. The site's research program sets the wider agenda, while Fresno resources and county public-health context help identify questions worth examining. The first deliverable is an honest account of what the available records can answer.
1. Turn an access concern into an answerable question
“Does Fresno County have enough care?” combines many questions and cannot be answered by counting dots. A narrower starting question is: “Among the licensed primary care clinics represented in this annual reporting dataset, what staffing and encounter measures were reported for one specified year?” A second study would be needed to investigate current appointment availability.
Write the population, service, unit of analysis, period, and outcome before downloading data. Specify whether a row represents an organization, physical site, reporting entity, visit, or person. State whether the question is descriptive, a comparison, or an evaluation of a change. Do not let the available column names silently redefine the question after analysis begins.
HCAI's clinic-utilization documentation describes annual reports from licensed primary and specialty clinics, including identification, staffing, financial, and encounter information. That scope does not constitute a census of every clinician's office or every appointment available to a resident. Read the reporting-year documentation and preserve its definitions with the data.
2. Keep the geographic unit visible
A county average cannot describe each neighborhood or every rural community. Decide whether the measure belongs to the facility's location, the patients' residence, or the organization reporting it. A clinic in one area can serve people from elsewhere, so assigning all its encounters to nearby residents would require evidence you may not have.
When adding population context, choose compatible geography and periods. Census guidance on ACS comparisons explains that estimates carry margins of error and that period length and changing geography matter. Its general guidance cautions against treating a five-year estimate as the population characteristic of its final year.
Record the estimate, period, geographic identifier, and uncertainty rather than retaining only the point estimate. Small apparent differences need appropriate statistical assessment before they become a ranking. If your service data describe one year but your neighborhood data cover five years, explain that mismatch and limit the conclusion accordingly.
3. Distinguish a listed location from usable capacity
Use separate variables for site existence, reported staffing, services offered, opening hours, payment acceptance, and appointment access. An address is evidence of location in the source; it does not verify all the other fields.
HRSA's health-center data distinguish administrative sites, service-delivery sites, and look-alike categories. Treat those categories according to their definitions rather than counting every row as an interchangeable patient-care location. Their scope is tied to the Health Center Program and related categories, not all healthcare providers.
Annual encounters describe delivered activity over a reporting period. They do not directly reveal vacant appointment slots now. Staffing measures require their own definitions; a headcount, full-time equivalent, and number of days a clinician works at one site are not interchangeable. If a study needs current availability, design a separate, consistent verification method and state when each observation occurred. Do not present an unanswered call as proof that a service does not exist.
4. Trace an announced program to a documented service
Create a separate record for funding, implementation, and delivery. An award notice may identify an intended service and recipient; a later record may show an operating site; a service report may show activity. The dates and populations should remain attached to each claim.
Use the public decision and money tracker to collect those documents. Decide what would count as confirmation before reviewing favorable announcements. For example, a report of sessions held with a defined reporting period is stronger evidence of operation than a photograph of an opening ceremony, but it still may not show appointment wait time or continuity.
If the relevant report is unavailable, say which milestone is unverified. Do not fill the gap using an optimistic forecast, a repeated press release, or an assumption that all awarded funds became staffing. A missing record is a limitation in the analysis, not automatically evidence of failure or misconduct.
5. Compare travel, staffing, and continuity without inventing a score
These are different dimensions of access. Travel time concerns a particular starting point, mode, time of day, and destination. Staffing concerns personnel and availability. Continuity concerns whether care and information remain connected over time. Combining them into one score requires explicit choices about measurement and weighting that can hide important tradeoffs.
For an initial study, present a small set of separately defined measures. A travel estimate should disclose the route assumptions and should not be labeled the experience of all residents. A staffing table should identify its reporting unit and period. A continuity measure needs a meaningful numerator, denominator, and observation window; a directory alone cannot supply it.
The clinical process measurement guide develops that measurement discipline. For a resident who needs help today, the local service organizer supports direct verification; a research map should not be treated as a booking service.
6. Make missingness and version changes part of the result
Use distinct codes for zero, not reported, not applicable, suppressed, and not verified. Replacing every blank with zero can create false scarcity or false improvement. Document duplicate handling and linkage decisions, especially when organizations operate multiple sites or identifiers change.
HCAI distinguishes preliminary clinic-utilization files from later desk-audited final files and describes replacement of the preliminary release. Preserve the release date and status in your research log instead of treating two versions as two independent observations. The official clinic documentation is the starting point for those checks.
Before publication, review whether tables expose sensitive information, whether comparisons are fair, and whether the wording exceeds the evidence. Aggregate public records do not automatically support causal conclusions about a policy or quality judgments about an individual clinician. The output may appropriately conclude that a proposed question needs different data. That is a useful feasibility finding, not a failed research exercise.
Fictional worked example: twelve rows are not twelve clinics
A fictional researcher receives a file with twelve rows and plans to label it “twelve available clinics.” The source review reveals two versions of one reporting record, one administrative location, and a site with an unverified current service status. Several staffing cells are blank.
She preserves the original file, records her deduplication rule, and separates reporting entities from service locations. She does not convert blanks to zero or describe the remaining sites as accepting new patients. Her feasible first analysis concerns reported annual activity in the defined dataset. A current appointment-access study would need additional observations and a separate protocol. No actual Fresno clinic count is established by this example.
Printable one-question study protocol
Complete this before producing a map or chart.
| Design decision | Record |
|---|---|
| Question and intended reader decision | ________ |
| County, service, population, observation period | ________ |
| Unit of analysis and inclusion/exclusion rules | ________ |
| Source URLs, release dates, preliminary/final status | ________ |
| Variables, units, and original definitions | ________ |
| Geographic linkage and denominator | ________ |
| Duplicate, missing, suppressed, and zero rules | ________ |
| Measures available / measures requiring new data | ________ |
| Uncertainty, privacy, and interpretation review | ________ |
| What this study cannot conclude | ________ |
- [ ] Organization and site records are distinguished.
- [ ] Every estimate retains its period and uncertainty.
- [ ] Announced funding is separate from delivered service.
- [ ] No quality ranking or causal conclusion comes from addresses alone.
- [ ] Source files, transformations, and limitations can be reconstructed.
Questions readers ask
Can a directory tell me whether a clinic can see me this week? Usually that requires direct, current verification of the service, eligibility, and appointment process. A listing alone cannot establish it.
Should missing staffing data be counted as no staff? No. Missing, not applicable, and zero have different meanings and should remain separate.
Is this a completed Central Valley access atlas? No. It is a study method and printable protocol. Empirical findings require collected evidence, documented analysis, and appropriate review.
Sources and scope
Primary sources: HCAI clinic-utilization documentation, HRSA health-center categories, Census comparison guidance, and Census multiyear guidance. Prepared November 5, 2026 Pacific; accessed November 6 UTC. No dataset was analyzed to produce local findings for this guide, and no withdrawn public download is restored or implied.