Clinical Treatment Protocol

Two scalars and a patient object drive the whole graph: diagnosis names the condition, measurement_value carries the diagnosis-specific reading (HbA1c, systolic pressure, or peak flow), and patient holds age plus yes/no strings for pregnancy, comorbidities, and prior adverse reactions.

Severity is read off diagnosis plus measurement_value, with first hit resolving the bands. 'diabetes_type2' splits at readings below 7.0 ('mild_diabetes', 'lifestyle_modification'), 7.0 to 8.5 ('moderate_diabetes', 'oral_medication'), and above 8.5 ('severe_diabetes', 'insulin_therapy'); 'hypertension' maps [140..160) to 'stage1_hypertension' with 'lifestyle_and_monotherapy' and 160 or higher to 'dual_therapy'; 'asthma' inverts the direction since higher peak flow is better, grading above 80 as mild and below 60 as 'severe_asthma' on 'multiple_controllers'. An unmatched diagnosis lands on 'unclassified' with 'standard_care'.

The bands track real guideline structure: an HbA1c below 7.0 is the common glycemic target in type 2 diabetes care, and escalating from lifestyle change to oral agents to insulin follows standard stepwise therapy. The 140 and 160 systolic cutoffs match the stage 1 and stage 2 boundaries used in international hypertension guidance, where stage 2 typically starts on two agents, and the 80 percent peak-flow line sits where asthma action plans draw their green zone; the 60 percent split below it is a model calibration.

Age pairs with pregnancy status here, most specific row first: a patient of 65 or older who is pregnant maps to a 'high' patient_risk_category with 'reduced_dosage', 65 or older otherwise gets 'moderate' and 'standard_elderly', under 18 becomes 'special' with 'pediatric_dosage', pregnancy at any other age yields 'elevated' with 'monitor_closely', and everyone else passes as 'standard' with 'standard_dosage'.

Using 65 as the geriatric line reflects the conventional cutoff in geriatric pharmacology, where reduced renal and hepatic clearance argues for dose caution, and carving out under-18s acknowledges that pediatric dosing is a separate discipline rather than a scaled-down adult dose. Flagging pregnancy for closer monitoring is standard practice given teratogenicity and physiologic changes; the specific category labels are model vocabulary.

Comorbidity history and prior adverse reactions form a complete two-by-two grid, so no default row is needed: 'yes' on both has_comorbidities and previous_adverse_reaction returns 'critical' with a 'specialist_consult', comorbidities alone return 'high' with 'modified_treatment', a reaction alone returns 'moderate' with 'additional_monitoring', and 'no' on both returns 'low' with 'standard_approach'.

Escalating a documented adverse reaction plus comorbidity to specialist review mirrors pharmacovigilance practice, since re-exposing a complex patient to a drug class that already caused harm is exactly the situation where protocolized care should hand off to a clinician. The exhaustive grid is also good rule hygiene: every combination is stated explicitly instead of leaking into a catch-all.

A single-column lookup translates comorbidity_risk_level into a care intensity: 'critical' becomes 'urgent', 'high' becomes 'intensive', 'moderate' becomes 'standard', 'low' becomes 'routine', and the blank fallback row emits 'regular'.

Keeping this translation in its own table is a maintainability choice rather than a clinical rule: governance can rename or remap intensity tiers without touching the risk logic upstream, and the fallback guarantees the assembled protocol always carries an intensity value.

Visit cadence keys off the same comorbidity_risk_level: 'critical' patients are seen 'weekly', 'high' 'bi-weekly', 'moderate' 'monthly', 'low' 'quarterly', and anything unmatched defaults to 'as_needed'.

Scaling follow-up frequency inversely with risk is the standard shape of chronic-disease management, where stable patients get quarterly reviews and unstable ones weekly contact; the specific cadences here are program-level choices rather than mandated intervals.

Assembly happens last: protocol_name concatenates base_treatment + '_' + patient_risk_category, and treatment_protocol packages a protocol_id built from that name plus protocol_intensity together with the dosage, risk adjustment, follow-up, and severity fields. With passThrough off, callers receive one self-describing plan object instead of the intermediate scoring fields.