Physiologic and Exposure Plausibility Checks

Why Plausibility Checks Matter

A dosing system can receive inputs in the wrong units, use a concentration collected at an unexpected time, or calculate an output that conflicts with basic physiology. Those problems can arise even when the data pass ordinary format checks. Conversely, a value far from what is typical may be correct for a particular patient. The central task is therefore not simply to label values “normal” or “abnormal,” but to decide what should be verified before the value is used.

Physiologic and exposure plausibility checks examine whether an input or predicted result is compatible with physical constraints, relevant patient characteristics, dose history, and the model’s assumptions. They complement input validation and out-of-domain assessment: a value can be correctly formatted and still be implausible, while a patient can be unusual without the record being wrong. A plausibility flag is a prompt to investigate, not proof of a data error or a clinical conclusion.

The strength of a check should match the kind of constraint. A negative body weight is physically impossible and can be rejected as an input. An adult record with a very low weight may be inconsistent with the record’s age or height, but might instead reflect a special circumstance or a data-linkage problem. A high laboratory value may be unusual yet genuine. Treating all three situations as equivalent would make the check unreliable.

Build Checks Around Explicit Assumptions

Before interpreting a discrepancy, state what the check assumes. For a physiologic check, this includes the variable’s units, the patient group to which the expected range applies, and whether the range is a physical boundary, a broad plausibility screen, or a clinical reference interval. Reference intervals describe distributions in defined populations; they are not universal limits on what an individual can experience.

For an exposure check, specify the dosing history, sampling time, route, concentration units, and pharmacokinetic model. A predicted concentration is conditional on these assumptions. If the dose time is wrong or an additional dose was given, a mismatch does not establish that the model calculation or concentration measurement is erroneous.

A useful operational distinction is between a hard constraint and a soft screen. Hard constraints include values that violate a mathematical or physical requirement, such as a negative elapsed time. A soft screen identifies a value that warrants review because it is uncommon or inconsistent with context. Soft screens should generally flag rather than overwrite a record. An action threshold should be defined for the intended setting and supported by relevant measurement, model, and population information—not borrowed uncritically from a general reference range.

Check Exposure Against Dose and Time

A concentration-time model gives a structured way to assess whether exposure is internally consistent. For illustration, consider a one-compartment model after an intravenous bolus. Assume instantaneous mixing, linear elimination, constant clearance and volume over the interval, and no additional doses. If the administered dose is \(D\), the apparent volume of distribution is \(V\), clearance is \(CL\), and time since the bolus is \(t\), then the concentration is

This simple model is not appropriate for every drug or regimen. Its value here is that the conditions are explicit: an infusion, absorption phase, another dose, or changing clearance would require a different prediction. Even when this model is a reasonable approximation, a predicted concentration is not an exact boundary. Parameter uncertainty, assay error, and model misspecification all contribute to differences between observed and predicted values.

A short derivation gives a useful serial check. Define \(k=CL/V\). For two times \(t_1\) and \(t_2\) after the same bolus, under the stated model,

The dose and volume terms cancel. When \(CL>0\) and \(t_2>t_1\), this ratio is below 1, so concentrations decline over time under these assumptions. A concentration increase may indicate a timing or dose-history problem, a unit mismatch, or that the assumptions do not fit the patient and regimen. It does not, by itself, tell which explanation is correct.

Distinguish a Data Problem From a Patient Difference

A useful investigation looks for independent evidence. First verify the identity, units, collection time, dose history, and source record. Then compare the value with other relevant patient information and, where appropriate, repeated measurements. If a verified value remains unusual, assess whether it can be explained by a patient-specific characteristic or by limitations of the assumed model. Repetition is informative, but a repeated measurement can also share a systematic error; it is not automatically definitive.

The following example illustrates why a value should not be judged only against a typical parameter value. It uses the same simple model, with an explicitly different clearance. The calculation checks whether the serial pattern is consistent with that clearance; it does not establish the patient’s clearance from concentration data alone.

Diagnostics, Alternatives, and Limitations

When a check flags a value, record what triggered the flag and what was verified. Useful diagnostic questions include:

• Are the units, decimal placement, patient identity, and measurement time consistent across the source record and dosing system?
• Does the recorded dose history include the route, administration time, and any additional or delayed doses needed to interpret the sample?
• Does the discrepancy persist after unit conversion and use of the correct sampling time?
• Is the patient-specific model applicable over the interval, or could a change in physiology or regimen invalidate its assumptions?
• Do other measurements or records support a data problem, or do they support a genuine unusual patient value?

Plausibility checks are not substitutes for other safety methods. A hard bound is transparent and appropriate for a true impossibility, but it cannot distinguish two plausible values. A broad population reference range is easy to implement, but may flag many genuine unusual patients and miss errors that remain inside the range. A model-based exposure check can use dose and timing information, but inherits the model’s assumptions and uncertainty. Clinical review can integrate context, but may be inconsistent if the question and response are not specified. These approaches are complementary; no single screen can establish that a patient or record is correct.

A common failure is to treat a soft threshold as a rule that automatically deletes, caps, or “corrects” a value. For example, a concentration above a typical range may be caused by a wrong unit, but it may also reflect a correct sample in a patient with different disposition. The correction is to flag the discrepancy, check its provenance and assumptions, and retain the distinction between unverified, confirmed data error, and verified unusual value. A plausibility check should also be evaluated for false alarms and missed discrepancies in the setting where it will be used. Thresholds that are too narrow can create alarm fatigue; thresholds that are too broad can fail to catch consequential errors.

Finally, plausibility is not the same as suitability for a dosing decision. These checks assess whether information appears credible enough to interpret under stated assumptions. They do not establish that a recommended dose is appropriate for every clinical circumstance. Keep that scope clear when recording a flag or designing the system’s response.