Mouse Health Monitoring: Records That Hold Up

Every lab does mouse health monitoring. Far fewer can prove it. Ask a colony manager how the tumor-bearing cohort on rack 3 is doing and you will usually get a confident, accurate, entirely verbal answer — one that lives in a single person's head, walks out the door when they graduate, and turns into a blank stare during an IACUC inspection.

The gap is not diligence. Techs walk the room, look at the animals, and notice things. The gap is that noticing is not recording. A health observation nobody wrote down cannot be trended, cannot be handed off, and cannot be defended six months later. This post covers what a mouse health record should actually contain, how often to write one, and how to make the writing fast enough that it happens on a Friday afternoon and not just before a site visit.

Why "the mouse looked fine" is not a health record

A verbal health check fails in three specific ways, and each one costs you something different.

It does not trend. A mouse at 24.1 g is unremarkable. A mouse that was 30.1 g eleven days ago and is 24.1 g today is at a humane endpoint. You cannot see the second thing by looking at the mouse. You can only see it by putting the numbers next to each other, and that requires both numbers to exist somewhere other than memory.

It does not hand off. The person who noticed the hunched posture on Tuesday is at a conference on Thursday. Whoever covers the room inherits a cage card and a guess. This is the same failure mode that breaks health-status handoffs during animal transfers — the information exists, it just never made it into a form the next person can read.

It does not answer "when did you first notice?" That is the question a veterinarian asks, the question an IACUC reviewer asks, and the question you ask yourself when a study arm goes sideways. "A while ago" is not an answer. A timestamped record is.

What a mouse health observation should capture

A useful record has four layers. Most labs capture the fourth one only, as a sentence in a notebook.

Vitals that trend

Three numbers do most of the work:

Weight in grams. The single most predictive measure you can take, and the one that supports the widely used 20% body-weight-loss endpoint. It is only meaningful against a baseline, which is the whole argument for recording it every time rather than only when an animal looks unwell.
Body condition score (BCS), 1–5. A palpation-based score that catches wasting in animals whose weight is masked by tumor burden, ascites, or hydration status. Weight and BCS disagreeing is itself a signal.
Grimace score, 0–10. A pain scale read from facial features. Useful post-surgically and in any protocol where analgesia adequacy is in question.

Clinical signs, coded rather than freehand

Free text is where health monitoring data goes to die. "Looked a bit scruffy" and "poor coat" and "piloerection" are three people describing the same animal, and no query will ever group them. Coded signs solve this: a fixed vocabulary per body system, picked from a list.

A workable standard set covers coat, posture, activity, gait, eyes, nose, respiration, hydration, color, intake, and feces — each with a Normal option and a small number of named abnormals (Ruffled, Hunched, Lethargic, Ataxia, Porphyrin, Labored, Dehydrated, Pallor, Anorexic, Diarrhea, and so on). Add binary flags for the findings that are present or absent rather than graded: masses, lesions, wounds, dermatitis, malocclusion, prolapse.

The payoff is that "show me every animal scored Lethargic in the last 30 days" becomes a question you can answer in seconds instead of a question you stop asking.

Context: who, when, and why you looked

An observation without a trigger is ambiguous. A record that says the animal was checked because of Routine rounds means something different from one triggered by Found sick or Post-op monitoring. Capture the observed-at timestamp, the observer, the protocol the animal sits under, and why the check happened.

Assessment and disposition

The narrative, your assessment, and the plan — then the decision. Disposition should be a short controlled list, not prose: No action, Recheck, Treat, Escalate to vet, Euthanize. If the disposition is Recheck, the record should carry the recheck date, because a recheck nobody scheduled is a recheck that does not happen.

Routine, scheduled, and found-sick are three different jobs

Labs get into trouble by treating health monitoring as one undifferentiated activity. It is three.

Routine rounds. Daily observation of every cage, mostly negative findings. The goal is coverage, not depth. Recording "all normal" has real value — it establishes that the animal was seen.
Scheduled monitoring. Cohort-specific, protocol-driven, on a cadence: daily weights on a tumor study, twice-daily checks for 72 hours post-op, weekly BCS on an aged colony. These are the checks your protocol committed you to, so these are the checks you will be asked to produce.
Found sick. Unscheduled, triggered by something you saw. These need the most detail and the fastest escalation path, and they are the ones most likely to be handled verbally because everyone is in a hurry.

Design your recording workflow around the third case. If logging a found-sick animal takes four minutes at a shared computer down the hall, it will be logged from memory that evening, or not at all.

A practical mouse health monitoring checklist

Use this as the frame for a daily walk-through:

Cage-level first. Food, water, bedding condition, and cage density — overcrowding shows up as fight wounds and barbering before it shows up anywhere else.
Activity at the cage front. Who comes to investigate, who stays flat, who is separated from the huddle.
Coat and posture from outside the cage, before you disturb anyone.
Respiration — labored or shallow/rapid breathing is visible without handling.
Eyes and nose for discharge or porphyrin staining, a reliable early stress marker.
Weigh and palpate any animal on scheduled monitoring, plus anything that caught your eye.
Compare to the last value, not to a vague sense of normal.
Record before you leave the room, including the normals.
Set a recheck date on anything you did not resolve.
Escalate anything meeting a protocol-defined endpoint, same day.

The last three are where most systems break down.

How Moustra records health observations

We built Health records into Moustra so this whole workflow lives next to the animal it describes, rather than in a parallel spreadsheet.

From any animal or cage page, Add record opens a form organized the way you actually observe: Context, Vitals, Clinical signs, Findings, Assessment. Clinical signs are dropdowns backed by a standard clinical vocabulary that ships with every account — coat, posture, activity, gait, eyes, nose, respiration, hydration, color, intake, and feces — so two people scoring the same animal produce comparable data. Findings are checkboxes for masses, lesions, wounds, dermatitis, malocclusion, and prolapse. Assessment captures your narrative, assessment, and plan, then the disposition and a recheck date.

A few details that matter in practice:

Cage-level records. Not every observation belongs to one animal. A record can attach to a cage, which is the right subject for a fight, a leaking water bottle, or a group-housed sentinel check.
A per-animal health timeline. Every record for an animal shows in reverse-chronological order on its page, with weight, BCS, and flagged signs summarized at a glance.
It works on the mobile app. You can log a found-sick animal standing at the rack, on your phone, in the time it takes to write it on a cage card. This is the part that decides whether health monitoring gets recorded or remembered.

Weight and BCS trends catch what a single glance misses

The reason to record normals is that normals are the baseline abnormals are measured against.

Moustra charts weight and body condition score over time for each animal, and draws an Endpoint −20% line on the weight chart based on that animal's baseline. The chart also shows the latest value against baseline and against the previous reading, as percentages. A 6% drop between two visits is easy to shrug off in isolation; three consecutive 6% drops with the endpoint line approaching is not.

This is the argument for coded signs and captured vitals in one sentence: a trend line is something you can act on a week early. A recollection is something you act on a week late.

What IACUC reviewers and journals actually want

Two audiences, one record.

Your IACUC wants to see that the monitoring your protocol described is the monitoring that happened — that scheduled checks occurred at the stated frequency, that defined humane endpoints were applied, and that escalation to the attending veterinarian is documented with dates. An append-only record with timestamps and named observers answers all of that without a scramble. It is the same evidentiary posture that makes the rest of your IACUC compliance work manageable.

Journals and reviewers want the health and welfare details the ARRIVE guidelines ask you to report: monitoring frequency, criteria for exclusion or euthanasia, adverse events, and animals removed from analysis. If those numbers only exist as a reconstruction from memory at manuscript time, they are a reconstruction — and reviewers have gotten better at noticing. The ARRIVE guidelines and OLAW's guidance on humane endpoints are both worth reading against your current monitoring plan.

Make it a two-minute habit

The best health monitoring system is the one your least-enthusiastic team member uses on a Friday. That means a fixed vocabulary instead of a blank text box, a form that opens where you are standing, and a default of recording the normal check rather than only the alarming one.

Start smaller than you think you should. Pick one cohort — the one under active study, with a real endpoint — and record every scheduled check on it for a month, normals included. You will end that month with something you have probably never had: a defensible, queryable picture of how a group of animals actually did, and an early-warning line on a chart instead of a story about a mouse.

A postdoc running a 60-animal tumor study spends maybe two minutes per animal on a monitoring day. Over eight weeks, that is a few hours of work — and the difference between a manuscript methods section you can write from data and one you write from your notebook plus hope. If you want that record to live next to the animal instead of in a spreadsheet nobody else opens, give Moustra a try, or see the docs for a walkthrough of health records and the rest of the colony workflow.