Clinical decision support: what the system does

This section states exactly what the Nanto Health Session Console does by way of clinical decision support, and what it does not. It is written to be verifiable by a reviewer: every capability described can be demonstrated in the product, and nothing is claimed that cannot.

What the system does

1. Rule-based intradialytic hypotension flagging. During a live session, the system judges each recorded blood pressure by two rules. The absolute rule flags a systolic below 90 mmHg, or below 100 mmHg when the patient’s pre-dialysis systolic was 160 or above. The relative rule flags a systolic drop of 20 mmHg or more from the patient’s own pre-dialysis reading. Either rule alone flags the reading as intradialytic hypotension, and the system states which one fired and on what value. The absolute thresholds follow Flythe JE, Xue H, Lynch KE, Curhan GC, Brunelli SM, Association of Mortality Risk with Various Definitions of Intradialytic Hypotension, J Am Soc Nephrol 2015;26(3):724-734, which found the nadir definition the most strongly associated with mortality. The flag appears on the flowsheet as the reading is entered and raises an alert on the floor board for that chair. It is computed from the observations the nurse records; it is a deterministic rule, not a model.

2. Deterministic readiness gates. Before a claim can proceed, the system runs a fixed set of checks and blocks a claim path that is unfundable or unsafe: SHA cover not confirmed, the session not signed, required evidence missing, or the preauthorisation course exceeded. A blocked claim is held with the specific reason. These gates are rules with a known outcome for a given input, so a clinic and a reviewer can predict and audit them.

3. Structured, checked data entry. Diagnoses are entered as ICD-11 codes rather than free text, flowsheet values are checked against safe bands as they are entered, and lab results are checked for freshness and date-match against the requisition. These reduce error at the point of entry.

4. Allergy check at prescribing. When a medicine is prescribed, from the chart or from order entry, the system compares it with the patient’s recorded allergies by WHO ATC code and drug class, and asks for a reason before a matching medicine is prescribed. The reason is kept with the prescription and in the audit trail. This check is always on.

5. The clinical decision support module (switched on per clinic). A separate module, off unless it is switched on for a clinic, adds cited alerts and references. Every item names the record facts it matched and the published source and version it comes from, so it can be checked against that source:

• Drug interactions at prescribing. A medicine that forms a pair on the ONC High-Priority Drug-Drug Interaction list (Phansalkar et al., J Am Med Inform Assoc 2012) with a medicine already on the patient’s list opens a warning that names both medicines, the list and its version. The list marks these pairs as contraindicated for concurrent use, so prescribing over one needs a reason, which is kept and audited. The prescription is never refused. Only part of the list is configured, and the list itself is short and partial: the absence of an alert is not evidence that a combination is safe.
• Laboratory results. When the latest potassium is outside the reference range this record holds for the test, the chart shows the value, the range, the LOINC code and a reference to the KDIGO Controversies Conference report on potassium in kidney disease (Kidney Int 2020).
• Vital signs. The intradialytic hypotension flag described above carries a reference to the K/DOQI cardiovascular guideline for dialysis patients (Am J Kidney Dis 2005). This adds no new alarm and no new trigger. It does not replace the dialysis machine’s or patient monitor’s alarms, or the nurse’s response at the chair.
• Problem list. Chronic hepatitis B on the problem list brings a reference to the CDC recommendations for chronic haemodialysis patients (MMWR 2001).

A clinician can acknowledge each reference; the acknowledgement is kept, with who and when, and written to the audit trail.

What the system deliberately does not do

The system supports the clinician and does not substitute for clinical judgement. Specifically, it does not:

• recommend or suggest a diagnosis,
• check for drug interactions beyond the cited list above, or claim that the list is complete,
• perform predictive analytics or risk scoring,
• propose a dose or a treatment, or block, cancel or change an order,
• show decision support to patients.

Every clinical decision remains the clinician’s. The flags and gates are prompts and guard-rails computed by fixed rules; they point a trained user at something to look at or stop an unfundable claim, and nothing more.

Why it is scoped this way

A rule a clinician can understand and audit is safer, in a dialysis unit with staff turnover, than a recommendation they cannot see inside. The system earns trust by being predictable: the same input always produces the same flag or gate, and every one can be traced to the observation that caused it.