Care cannot be reduced to a checkbox

Every care system eventually faces the same design temptation: if a fact matters, turn it into a field. A field can be required, standardized, searched, counted, and displayed before an appointment. It can trigger a reminder, route a case, support an audit, or help a team see patterns that would otherwise remain hidden. These are significant strengths.

But individualized care is not delivered to a category. It is delivered to a particular cat, in a particular body, environment, relationship, and moment. A Medical Alert icon can make risk visible, but it cannot, by itself, explain which risk is active, how that risk changes handling, what the team should watch for, or when care should be modified. A Caution Cat icon can prompt attention, but it cannot distinguish among sensitivity to touch, sound, restraint, transition, proximity, or a previous adverse experience. A Room Preference icon can identify a choice, but not why that choice matters.

The historical record suggests that staff understood this boundary in practice. They used icons to signal and notes to explain. The pattern is not an argument against structured data. It is evidence about the point at which structure stops carrying enough meaning.

What the historical records show

The source was a 183-page Custom Animal Icons report. Its totals page listed 2,333 icons, while 2,275 visible detail rows could be extracted from the printed pages. Using exact owner name plus animal name as an approximate identity key, the visible rows represented about 1,737 animals. Counts in the analysis are icon records, not appointments, diagnoses, treatments, outcomes, or a deduplicated clinical patient census.

The distribution shows that the system was primarily being used to surface high-attention concerns. Medical Alert accounted for 536 visible rows, Important for 509, Caution Cat for 392, and Senior Cat for 247. Together, those four categories represented 74.0% of the visible detail rows. Medical Alert and Important alone represented 45.9%. The dominant work of the icon system was therefore not cosmetic description; it was operational visibility for risk, care complexity, and handling considerations before service began.

The striking result: categories were rarely left alone

Across all visible icon rows, 90.2% included an explanatory note. In the categories most directly connected to risk and active care planning, note coverage was nearly universal: 99.6% for Medical Alert, 99.4% for Important, and 96.4% for Meds.

The category identified the domain of attention. The note carried the meaning that made the category usable. Staff could have selected an icon and stopped. In more than nine out of ten records, they did not.

The dataset measures the presence of a note, not whether every note was current, complete, accurate, or effective. It also does not establish that notes improved outcomes. Even with those limits, the repeated decision to add narrative at this scale is operationally meaningful. It shows that the structured label was routinely treated as a starting point rather than a complete care instruction.

Structured fields and narrative answer different questions

Structured fields are strongest when the question is whether something is present, how a profile should be grouped, who or what should be alerted, and how frequently a condition or constraint appears across the population. Narrative is strongest when the question is what the factor means for this cat and how it changes the next decision.

A field can indicate Senior Cat. A note can preserve the practical relevance of that status, such as mobility, stamina, positioning, temperature support, or the pace at which care should proceed. A field can indicate Meds. A note can preserve timing, observed response, preparation, or a contingency that matters on the day of service. A field can indicate Caution Cat. A note can describe the earliest readable signs, the approach that has worked, what should be avoided, and what threshold should lead the team to pause or change course.

These examples describe the kind of context a narrative field can hold; the source report did not content-code every note. The design principle is still clear: the field answers the classification question, while the note answers the interpretation question.

Complexity lives between the flags

Most approximate animal profiles carried one icon, but 423 profiles, or 24.4%, carried two or more. Of those, 97 profiles carried three icons and nine carried four. The most common combination was Medical Alert plus Senior Cat, present in 59 profiles. Important plus Medical Alert appeared in 47 profiles, Medical Alert plus Meds in 15, and Important plus Medical Alert plus Senior Cat in 14.

These combinations matter because complexity is not only the number of flags. It lies in how the flags interact. A senior cat with a medical alert and a medication requirement does not have three separate care plans. The profile has one care context, and the team must understand how age, health, medication, environment, timing, and handling shape one another.

An interface that presents icons one at a time can fragment that context. A consolidated pre-appointment view can restore it by assembling every active signal with the latest relevant note, preparation requirement, environmental adaptation, and staffing or scheduling constraint.

The sum of the flags is not the care plan.

A dual-layer model for individualized care records

The answer is not to choose between structure and narrative. It is to give each a disciplined role. A durable care record needs two content layers, surrounded by clear review and accountability rules.

Narrative does not mean unstructured chaos

A case for narrative is not a case for unlimited free text. Notes can become stale, duplicative, contradictory, vague, or difficult to scan. Those failures do not make narrative unnecessary; they make governance necessary.

A useful note should be current, attributable, action-oriented, and connected to the structured signal it explains. It should separate observation from instruction, preserve uncertainty rather than disguising it, and make clear what should prompt a change in plan. Required facts should not be buried only in prose, and narrative should not be discarded after those facts are extracted into fields.

The source report recommends defining the purpose, creation criteria, owner, review cadence, and removal conditions for each icon category. The same discipline should apply to the notes. The latest active context should be easy to find, while outdated or superseded instructions should remain auditable without competing with the current plan.

Implications for automation and AI-assisted documentation

The report recommends moving future processing away from recurring PDF extraction and toward direct structured exports. At an expected scale of roughly 5,000 run cards per year across Beaverton and Powell, that shift is necessary. It also makes preservation of narrative more important, not less.

Automation can extract repeatable facts into standardized fields, assemble multi-icon profiles into a concise briefing, identify missing review dates, and route conflicting or uncertain records to a person. AI-assisted tools can help summarize long histories or suggest structured values. The safest design is additive: it makes narrative easier to retrieve and structured data richer without silently erasing the source note.

Any generated summary should remain linked to the original record, distinguish documented observation from machine inference, and make uncertainty visible. A mature system should not force a choice between a searchable database and a readable care history. It should allow each to strengthen the other.

A practical rule for designing care records

The findings can be reduced to one operational rule: structure what must be counted, narrate what must be interpreted, and review what can become outdated.

In practice, that means standardizing the facts needed for filtering, routing, scheduling, staffing, and reporting; preserving the explanation needed to translate those facts into individualized care; consolidating interacting flags into one appointment-level view; recording who reviewed the plan and when; and eventually linking documentation patterns to service outcomes, modifications, duration, and follow-up needs.

Limitations and next questions

This analysis supports a documentation design argument, not a causal claim. The records count icon rows rather than appointments or diagnoses. Animal profiles were approximated from owner and animal names because stable identifiers were not present in the printed report. The source totals page exceeded the visible detail extraction by 58 records, and 150 visible rows carried at least one extraction review flag. The source also lacked location and outcome fields.

Most importantly for this paper, note presence does not measure note quality. The analysis did not code whether notes were actionable, current, redundant, contradictory, or associated with better service outcomes. Future work can classify the functions of the notes, measure their age and consistency, identify which structured fields can be safely derived from them, and test whether a consolidated pre-appointment briefing changes completion, duration, modifications, escalation, or follow-up.

Conclusion

The custom icon registry succeeded because it compressed critical signals into a form that staff could scan quickly. The notes mattered because compression has a cost. They restored specificity, interaction, sequence, and context.

The practical goal is not to choose between checkboxes and prose. It is to give each a disciplined role. The field identifies the concern. The note explains what the concern means for this cat. The reviewed briefing turns both into coordinated action.

Structured data can make a cat visible to a system. Narrative helps keep the cat recognizable within it.