You are writing the needs statement for a rural county. You go to pull the infant mortality rate, or the cancer death rate, or the teen birth rate, and the cell is blank. Suppressed. The measure that would make your strongest argument does not exist for the community that needs it most. This is one of the most common and most frustrating problems in rural grant writing and community health assessment, and it is not you doing something wrong.
It is a structural feature of how public health data is published. Understanding why it happens tells you exactly how to work around it, honestly, without overstating your case or leaving a hole a reviewer will notice.
Why the data disappears
Data suppression is the deliberate withholding of a statistic when the underlying numbers are too small to publish. There are two separate reasons, and they matter because they point to different workarounds.
Reliability.A rate calculated from a handful of events is statistically unstable. If a county has three infant deaths one year and one the next, the “rate” swings wildly for reasons that have nothing to do with real change. Agencies suppress rates below a minimum event count (often around ten or twenty) rather than publish a number that would mislead.
Privacy. In a small population, a published count can identify individuals. If a county of 900 people has one death from a rare cause, naming it in a public table effectively names the person. Suppression protects against that re-identification.
Neither reason is a judgment about your community. Both are mechanical thresholds that trip automatically on small populations, which is why rural counties hit them constantly and metropolitan counties almost never do.
What actually gets suppressed
The pattern is consistent: the smaller the population and the rarer the event, the more likely the measure is gone.
- Mortality by specific cause is the most affected. Death counts for specific causes in small counties routinely fall below reporting thresholds and vanish.
- Infant mortality is a stark example. It is one of the most powerful measures in any maternal, child health, or early childhood application, and it is published for only about a third of U.S. counties in a given year. For the roughly two-thirds that are rural or small, it simply is not there.
- Direct survey estimates (behavioral risk factors from BRFSS, for instance) get wide error bands or disappear because too few residents were surveyed locally.
- Subgroup breakdowns (by race, age, or sex) suppress far faster than the county total, because splitting a small population into groups shrinks each count further.
The good news is that the measures most likely to survive are also some of the most useful for a needs statement: social and economic conditions from the Census American Community Survey (poverty, uninsured rate, housing) are modeled to cover every county, and low birth weight is available for nearly all of them even where infant mortality is not.
Six ways to get a number anyway
None of these is a trick. Each is a legitimate, disclosable method that a careful reviewer will accept, because each is honest about what it is doing.
Combine counties into one service area
This is the single most useful move. A measure suppressed for each of four small counties individually is often publishable for the four combined, because the pooled population clears the reliability threshold. Funders and CHNA frameworks think in service areas anyway, so a combined-area figure usually fits the document better than county-by-county fragments would have. State it plainly: report the measure for the service area, and note which individual counties were suppressed.
Combine years
Mortality tools like CDC WONDER will compute multi-year rates that clear suppression thresholds a single year cannot, because pooling years grows the numerator. A cause of death suppressed for 2024 alone may be reportable for 2020 to 2024 combined. Label the period plainly (“2020–2024 combined”) so a reviewer knows exactly what the number covers.
Use model-based estimates, not direct survey estimates
Direct survey estimates (from BRFSS, for instance) get unreliable or vanish in small counties because too few people were surveyed there. Model-based estimates, most importantly CDC PLACES, are built specifically to produce chronic-disease prevalence for every county and even every census tract, by borrowing statistical strength across geographies. For diabetes, COPD, asthma, depression, and similar measures, PLACES is usually the only consistent small-county source. Cite it as a model-based estimate, which is exactly what reviewers expect for populations this small.
Substitute an available proxy, and disclose it
When the measure you want is suppressed, a related measure in the same domain often is not. Infant mortality is published for only about a third of U.S. counties; low birth weight is available for nearly all of them and is an accepted needs proxy in the same maternal and child health domain. The honest sentence wins points: “county-level infant mortality is suppressed for populations this small; we report low birth weight, which is available for every county in the service area.” That signals you know the data rather than that the data defeated you.
Step up one geography, on purpose and out loud
Sometimes the honest answer is that a reliable local number does not exist, and the defensible move is to report the state or regional figure explicitly labeled as such, paired with whatever you can document locally. “County-specific rates are suppressed; the regional rate is X, and locally we can document Y (uninsured rate, provider shortage designation, poverty).” Reviewers respect a writer who names the limitation instead of disguising a state number as a local one.
Report the rate with its uncertainty, not instead of it
For estimates that exist but are shaky in a small population, publish the confidence interval alongside the point estimate rather than dropping the measure or pretending the point value is precise. “14.1% (95% CI 9.8–18.4%)” is more credible to a careful reviewer than a bare 14.1%, and far more credible than silence. Honesty about precision reads as competence, not weakness.
The disclosure is the point
The thread running through all six is that you name the limitation instead of hiding it. Grant reviewers and CHNA readers read a lot of needs statements, and they can tell the difference between a writer who understands the data and one who is papering over a gap. A sentence like “because the service area's counties are small, several cause-specific rates are suppressed at the county level; we report them for the combined service area and note the data year” does more for your credibility than a clean-looking table that quietly substituted a state number for a local one.
Suppression is not the end of your argument. Handled openly, it can even strengthen it: a community whose data keeps disappearing because its population is too small and too dispersed is, in itself, evidence of the rural access problem most of these grants exist to address.
If you would rather not assemble it by hand
Every U.S. county's environmental, health, access, and social data is free to browse on our county pages, and if you are working across several small counties, the $10 verified data report handles the combined-service-area math for you: it reports measures for the pooled area, which recovers numbers that are suppressed county by county, benchmarks them against state and national rates, and puts a citation on every figure with each claim checked against the source data before the report is assembled. It exists because rural grant writers kept running into exactly this wall.
However you get there, the principle holds. Missing data is not a dead end, and it is not a reason to weaken your case. It is a known problem with known, honest solutions, and knowing them is part of what separates a fundable needs statement from a thin one.