alive.process.referral@1.0.0

Screening referral and navigation process

Represent referral follow-through as an explicit, inspectable workflow.

Boundary

Contact attempts, booking, reminder, attendance. No clinical results, rescheduling, transport, insurance or eligibility rules.

Each person has a fixed per-attempt reach probability; attempts are independent given it.
A missed appointment closes the referral (no rebooking).
With an intervention, automated outreach is tried first; staff are the fallback.

Completion here means 'attended in the model', not screening performed or any health outcome.

Programs with rebooking loops, multi-step diagnostics or capacity-limited clinics.

Name	Default	Provenance
`max_attempts` Contact attempts before giving up.	6 attempts	synthetic assumption Protocol choice. With the per-attempt Beta(2,8) default, expected reach over 6 attempts is 1 - B(2,14)/B(2,8) = 0.66, inside the cited 0.60-0.94 range.
`contact_success_probability` Chance one navigator attempt reaches the person. A distribution is drawn once per person (heterogeneous reachability).	beta(alpha=2, beta=8) probability	synthetic assumption No per-attempt rate is published; Beta(2,8) (mean 0.20) is anchored on cited cumulative reach. Research grade: very low.
`retry_delay_hours` Wait before the next attempt.	triangular(high=96, low=24, mode=48) hours	synthetic assumption Illustrative placeholder; no cited evidence adopted for this value.
`appointment_lead_days` Booking-to-appointment time.	lognormal(mu=2.65, sigma=0.83) days	public evidence Moment-matched to a 20-day mean with an assumed CV of 1 (median ~14 days). Cited waits are metropolitan; rural waits were not measured.
`no_show_probability` Chance a booked appointment is missed. A distribution is drawn once per person.	beta(alpha=3, beta=17) probability	public evidence Beta(3,17), mean 0.15, spans cited rates of 8% (rural network) to ~23% (systematic review). Definitions of no-show vary across sources.
`adapter_timeout_seconds` How long to wait for the intervention before falling back.	30 seconds	synthetic assumption

nav_contact_cumulative_reach_probevidence: lowthin or conflicting

Probability that a navigator or outreach worker eventually reaches a patient by phone after a full attempt protocol.

Three heterogeneous studies with differing attempt protocols.

nav_contact_per_attempt_success_probevidence: very lowthin or conflicting

Probability that a single outbound phone attempt reaches the patient (live conversation).

Derived from a cumulative figure under strong assumptions.

nav_appointment_lead_time_daysevidence: lowthin or conflicting

Days from scheduling request (or referral) to the appointment date for a new specialist or primary care visit.

Means are verified, but no individual-level distribution was verified.

nav_no_show_rate_baselineevidence: moderatethin or conflicting

Baseline probability that a scheduled outpatient appointment is a no-show (no reminder or navigation effects applied).

A large SLR plus large US cohorts, but very heterogeneous.