Multi-Month Dispensing (3 bulan)

1 · Yang berubah untuk pasien dan anggaran

2 · Perubahan cascade (vs komparator, akhir horizon)

Mekanisme intervensi: bagaimana distribusi pasien antar-state bergeser dibanding Standard Monthly.

3 · Seberapa yakin hasilnya (uji ketidakpastian)

Model dijalankan ribuan kali dengan parameter yang divariasikan. Kurva di bawah menunjukkan seberapa sering intervensi tetap cost-effective pada berbagai ambang biaya yang dianggap layak.

4 · Analisis sensitivitas deterministik (tornado)

Parameter yang variasi satu-arahnya paling menggerakkan incremental net monetary benefit pada ambang 1× PDB.

5 · Cohort trace & dekomposisi biaya

Okupansi state sepanjang horizon untuk tiap arm, dan dekomposisi selisih biaya menjadi biaya intervensi versus biaya perawatan hilir yang dihindari.

6 · Skenario (9)

7 · Ambang biaya intervensi

ICER saat biaya per-siklus intervensi divariasikan; harga breakeven menjaga cost-effectiveness pada ambang 1× PDB.

8 · Budget impact (5 tahun)

Populasi eligible 2.500.000 · Stable type-2 diabetes patients on chronic oral therapy under JKN eligible for multi-month (Prolanis/PRB) dispensing. Order-of-magnitude: Indonesia adult DM prevalence ~11.7% (SKI 2023) over ~190M adults implies ~22M with DM; ~2.0-2.2% are physician-diagnosed and a stable, on-treatment, JKN-covered subset suitable for 3-month MMD is conservatively ~2.5M.

9 · Validasi model (mass balance)

Massa kohort harus konservatif setiap siklus (jumlah probabilitas state = 1). Toleransi 0.0001, deviasi maksimum 1.0e-5.

10 · Perbandingan dengan rujukan internasional

11 · Adaptasi konteks Indonesia

• Indonesia already has a partial MMD vehicle: the JKN Prolanis (Program Pengelolaan Penyakit Kronis) and PRB (Program Rujuk Balik) chronic-disease refill programs deliver routine DM medication through FKTP, but typically on monthly or short-cycle refills; 3-month dispensing for stable patients is the service-delivery innovation evaluated here.
• Baseline cascade: treatment coverage among physician-diagnosed DM ~92.1% (SKI 2023), but glycaemic control (HbA1c<7%) only ~32.0% and fasting-glucose control ~25.2% (SKI 2023), and Prolanis glycaemic-target attainment declines over 12-18 months, so the binding problem is RETENTION and sustained CONTROL, exactly the lever MMD acts on.
• DM prevalence ~11.7% of adults in SKI 2023 (up from 10.9% in Riskesdas 2018); the eligible MMD population (~2.5M stable on-treatment JKN patients) is the BIA denominator.
• Complication costs use Indonesian cost-of-illness evidence: major-complication episodic cost ~Rp 105 juta vs ~Rp 4.5 juta uncomplicated (Andayani et al. 2021), ~20x, with INA-CBG diabetic-foot/gangrene and CKD-haemodialysis tariffs as the per-state anchor.
• Scorecard, not a lone ICER (per the Service Delivery Pillar design doc): the headline is the retention/control delta (process), with downstream QALY, cost-per-additional-patient-retained, cost-per-QALY, 5-year budget impact and an equity scenario reported alongside, and the cohort-vs-system caveat rendered prominently.

12 · Keterbatasan

• DRAFT_REQUIRES_HEALTH_ECONOMIST_REVIEW.
• COHORT-VS-SYSTEM CAVEAT (mandatory for every service-delivery output): this is a Markov COHORT model. It answers 'for a stable DM patient who reaches the on-treatment cascade step, how does 3-month MMD change their downstream retention, control, complications, QALYs and cost'. It does NOT mechanistically model system-level constraints, pharmacy supply-chain / cold-chain capacity, FKTP and HRH (pharmacist/dispensing) labour availability, or Prolanis enrolment bottlenecks. The cohort result is therefore an UPPER BOUND conditional on the MMD step being deliverable; the realism of scale lives in the bia_config eligible population (how many can actually be reached) and in the rural/low-access scenario_config override, not in the per-patient transition effect.
• The MMD retention effect (RR 0.74 on on-treatment->LTFU) is transported from differentiated-service-delivery / multi-month-dispensing evidence that is predominantly from HIV care (Fatti et al. 2021; AIDS & Behavior 2024). Direct multi-month-dispensing RCT evidence specifically for type-2 diabetes retention in an Indonesian Prolanis setting is sparse; the transported point estimate is deliberately conservative (0.74 vs trial signals of 0.50-0.65) and PSA uses a wide SE so the scorecard spans from a near-null to a trial-strength effect.
• Baseline cascade rates are anchored to national cross-sectional control/coverage figures (SKI 2023, Prolanis evaluations), not to a longitudinal individual-level Indonesian DM retention cohort; the on-treatment->LTFU and control-loss edges are calibrated rather than directly estimated. The ARC BPJS dm-cohort sample (skill arc-bpjs-thematic-report) is the natural next data source to replace these with empirical FKTP retention proxies.
• The single aggregated 'complication' state collapses diabetic nephropathy/dialysis, diabetic foot, retinopathy and cardiovascular events into one annualised cost (~Rp 18 juta) and utility (0.62); a disaggregated complication sub-model would sharpen the downstream DALY/QALY and cost estimates.
• Drug acquisition cost is identical across arms (same molecules); MMD's only cost lever here is the clinic-visit care-cost offset, which is itself uncertain and tested in the no-offset scenario.
• Equity is represented only via a rural/low-access scenario multiplier on LTFU, not via a full wealth-quintile or geographic distributional (ECEA) analysis.