GiveBetter x SF

Face to Face — Sonoma County AIDS Network

Finite unique-person overdose prevention within a whole HIV-support organization

Research time: ~9 min on GPT-6 Astra Light
  • Research — reviewed programs, finances and impact evidence.

Published: 11 September 2026.

Funding limitations

Summary

What do they do? Face to Face provides HIV-related support, housing assistance, testing, prevention supplies, overdose education and referral services in Sonoma County. This narrow health model credits only incremental overdose-survival benefit, while charging the whole organization. More

Why this approach interests us

  • Low-barrier naloxone access can reach a witnessed overdose before an effective alternative. The organization describes ongoing county outreach and hundreds of reversals.

Our main reservations

  • The favorable 10% world contributes 92.6% of signed expected health; central is $3,598,461 per 10. Absolute reporting, recurrence, alternatives and cash response are not verified.

What do you get for your dollar?

Signed prior-weighted estimate: $1,328,757 per 10 Bay QALYs, or 0.075258 per $10,000. This is not measured Face to Face impact. Inspect the model, assumptions and scenarios.

gross annual cost
$2,263,089All programs and netted fundraising costs included.
survival model
One finite trajectory per personA reported rescue is not equated with a death prevented.
subjective scenario mass
10% below $1M per 10 QALYs0% below $100K; judgment weights, not calibrated probabilities.

1. What do they do?

Face to Face provides HIV-related support, housing assistance, testing, prevention supplies, overdose education and referral services in Sonoma County. This narrow health model credits only incremental overdose-survival benefit, while charging the whole organization.

Fund additional access

A small unrestricted gift may expand or maintain delivery beyond existing public/donor support; the fraction is uncertain.

Reach a genuinely different rescue opportunity

Reported events are discounted and converted to unique-person equivalents; alternatives and nonfatal overdoses reduce causal mortality benefit.

Compare finite survival

Integrate with-versus-without survival for each unique person, one support year and a finite common-risk follow-up. Subtract independent morbidity harm.

Scope of this review. No added HIV prevention, PrEP, housing, MOUD referral, drug-checking or other health credit. These omissions may understate true portfolio benefits, but no defensible nonoverlapping local outcomes were identified. Whole-cost dilution remains. Medication, volunteers, EMS and partner resource values are not fully reconciled.

2. Monitoring and information sharing

Unspecified last-year overdose activity

Current official organizational service page. Hundreds of reversals and thousands of kits claimed; exact year, verification, repeated people and numerator definitions absent.

Our assessment. 100/200/500 event equivalents are analyst priors, not observed precision. Reporting validity and unique-person recurrence are separate assumptions.

Mixed harm-reduction/HIV-HCV participants

Provider report to public funder. 425 cumulative unduplicated participants; 121 quarter, with inconsistent term and quarter labels.

Our assessment. Not naloxone-specific people, rescue events, deaths prevented or an extra additive cohort; not used in survival denominator.

Communities with overdose education/naloxone rollout

Massachusetts interrupted time series, 19 communities, 2002–2009. Adjusted mortality ratios .73(low) and .54(high) versus no implementation; acute-care differences nonsignificant.

Our assessment. Observational pre-fentanyl community effect does not identify current Sonoma net rescue probability or donor marginal impact.

67 communities in four states during fentanyl/pandemic period

HEALing Communities cluster-randomized multicomponent intervention. Primary opioid mortality reduction not statistically significant; a RR.91,95% CI.76–1.09. Naloxone distribution increased 79%.

Our assessment. Not proof naloxone pharmacology fails; shows more distribution need not translate into demonstrated marginal mortality reduction. Supports explicit null and alternatives.

Synthetic unique rescued-person cohort

Analyst finite survival model, not a clinical study. Counterfactual excess hazard only during one support year; common post-support mortality; finite 2–8 years and 3% discount.

Our assessment. No source validates the org-specific hazard, utility, recurrence, finite horizon or capacity coefficients. Cohort mathematics prevents repeated rescue double credit but does not make those inputs empirical. During the one-year active interval, survival with rescue is exp(-otherHazard*t), and counterfactual survival is exp(-(otherHazard + repeats*fatality)*t). The repeats parameter serves both as the event-to-person divisor and as a modeled annual Poisson rescue-opportunity rate. Neither that prospective rate nor net fatality is statistically inferred from reported successful reversals; both are explicit analyst transfers from a selected-survivor activity description.

3. Qualitative assessment

Low-barrier naloxone access can reach a witnessed overdose before an effective alternative. The organization describes ongoing county outreach and hundreds of reversals.

Key reservations

  • The primary categorical event claim is weaker than an audited annual rescue ledger.
  • Whole portfolio costs are dated 2024; current prices and program mix can differ.
  • Reported successful rescues are selected survivors, not a prospective at-risk denominator.
  • EMS, peers, pharmacies and state-supported distribution are real alternatives.
  • Unique-person and hazard assumptions are not measured.
  • Favorable world drives 92.6% of expected benefit.
  • No verified donation route is included; donationUrl omitted.

Benefits not included in our estimate

  • HIV transmission prevention and PrEP treatment
  • Housing and non-medical case-management health
  • Referral contacts counted as completed MOUD
  • Drug checking credited independently
  • Unlimited lifetime survival or repeated rescue restart

4. What do you get for your dollar?

$1,328,757 per 10 Bay QALYs

Six correlated judgment worlds, including funding-null, rescue-null and harm, are averaged in signed QALYs before inversion. No coefficient tuned after first output.

A better life is our comparison unit of 10 additional quality-adjusted life years (QALYs), potentially spread across people. These are uncertain estimates, not measured returns or verified donation offers.

How we calculate the estimate

Unique-cohort signed health: gift / gross expense × (reported events × validity / events per person) × funding × [utility × integral(discounted survival with rescue − survival without rescue) − harm] × Bay share
0.07525830883941727 Bay QALYs/$10,000

Model inputs and assumptions
Gross whole annual expense
2263089 (range: 2024 historical basis). Part IX plus netted event costs; all portfolio costs, no public-revenue subtraction. High accounting.
Reported rescue-event equivalent
200 (range: 100–500). Analyst numerical interpretation of undated hundreds, not observed exact annual throughput.100 discounts stale/ambiguous claim; no inflation for unreported rescues. Low.
Credible relevant-event fraction
.5 (range: .3–.8). Discounts uncertain reporting year, duplicate reports of same event, and misclassification; distinct from repeated genuine events in same person. Very low.
Rescue opportunities per unique person
2 (range: 1.5–3; diagnostic 10). Unobserved recurrence; unique cohort=events×validity/repeats. Each person has one finite survival trajectory. Very low.
Otherwise-fatal probability per opportunity
.15 (range: .05–.25 plus zero). Net of equally timely alternative naloxone/EMS, spontaneous recovery and attribution. Not certainty of death after an observed reversal. Very low.
Common mortality hazard in support year
.2/year (range: .1–.3). Includes other competing fatal risks; active incremental hazard removed is repeats×fatality. No org-specific mortality estimate. Very low.
Common post-support mortality hazard
.25/year (range: .15–.5). Both arms face same hazard after one year; accounts for later overdose and other mortality without assuming continuing donor support. Very low.
Finite follow-up
5 years (range: 2–8 years). One year active support, finite subsequent survival credit, 3% discount; favorable 8 years is assumption not measured life expectancy. Very low.
Utility while alive
.7 (range: .6–.8). Judgment, not measured QALYs or valuation of a person; only health-adjusted survival in this model. Very low.
Gift capacity response
.3 (range: 0–.6). Unknown additional service versus funded public/donor baseline, reserves and substitution; no verified marginal tranche. Very low.
Bay health fraction
.98 (range: .95–.99). Sonoma service footprint; rescue location and residence not census-verified. Moderate.
Independent harm QALY per additional person
.001 (range: 0–.01). Judgment for morbidity such as precipitated withdrawal or delay; no measured incidence, no repeat event harm multiplier. Very low.

Small $10,000 diagnostic; no linear $100,000 capacity assurance

  • funding_null: Undefined per 10; weight 0.2. Bay QALYs 0; unique baseline person-equivalents 50; finite Q per unique 0.42882628927998306. Full matrix saved.
  • rescue_null: Undefined per 10; weight 0.2. Bay QALYs 0; unique baseline person-equivalents 50; finite Q per unique 0. Full matrix saved.
  • harm: Undefined per 10; weight 0.1. Bay QALYs-0.0006495546573731746; unique baseline person-equivalents 50; finite Q per unique 0. Full matrix saved.
  • cautious: $300,914,228 per 10; weight 0.2. Bay QALYs 0.00033232061024485073; unique baseline person-equivalents 10; finite Q per unique 0.08116538079141149. Full matrix saved.
  • central: $3,598,461 per 10; weight 0.2. Bay QALYs 0.027789655874849602; unique baseline person-equivalents 50; finite Q per unique 0.42882628927998306. Full matrix saved.
  • favorable: $143,474 per 10; weight 0.1. Bay QALYs 0.696988690081357; unique baseline person-equivalents 266.6666666666667; finite Q per unique 0.9963001500300053. Full matrix saved.
  • half_events: Half categorical event rate. $2,657,514 per 10; Bay QALYs 0.037629154419708634.
  • repeat10: Ten rescue opportunities per person. $2,828,569 per 10; Bay QALYs 0.03535356767227561.
  • one_year: No survival credit after support year. $10,500,290 per 10; Bay QALYs 0.009523546282645678.
  • half_fatality: Half otherwise-fatal probability after alternatives. $2,445,781 per 10; Bay QALYs 0.040886731249627634.
  • half_funding: Half gift response. $2,657,514 per 10; Bay QALYs 0.037629154419708634.
  • full_funding: Full response except funding-null world. $739,977 per 10; Bay QALYs 0.1351393419314912.
  • double_resources: Double whole expense for hypothetical omitted resources. $2,657,514 per 10; Bay QALYs 0.037629154419708634.

Uncertainty. Without favorable world, renormalized cost is $16,188,681 per 10. Prior weights are subjective. Event opportunity counts are modeled as a constant hazard in a closed aligned cohort, not recovered person-time; selection of survivors and actual event timing are unknown. First-year full removal applies only to the net otherwise-fatal component after alternatives, not every overdose. Repeated events never restart follow-up. During the one-year active interval, survival with rescue is exp(-otherHazard*t), and counterfactual survival is exp(-(otherHazard + repeats*fatality)*t). The repeats parameter serves both as the event-to-person divisor and as a modeled annual Poisson rescue-opportunity rate. Neither that prospective rate nor net fatality is statistically inferred from reported successful reversals; both are explicit analyst transfers from a selected-survivor activity description.

5. Funding and previous grants

HOLD giving recommendation. Exploratory publication is defensible; no verified marginal funding offer or measured mortality benefit.

2024 gross cost $2,263,089 includes $59,068 event costs netted from reported revenue and excluded from Part IX. Government grants $132,330, noncash contributions $397,827, net assets $1,482,170 and public investments $920,149 are not a funding gap. State medication supply and city-funded activities already exist. In-kind contributions cannot be subtracted without expense reconciliation. Paid/donated inputs, restricted public contracts, current allocation and complete societal cost remain incompletely identified. A general donation mechanism, if offered, would not establish a cash-priced additional cohort, current expansion capacity, or the modeled funding response.

This review does not establish a verified marginal funding offer or a complete history of grants.

We have not verified a suitable donation route for this reviewed activity. Confirm the legal recipient and intended allocation before donating.

6. Sources

  1. Current overdose prevention services. Face to Face. Primary. Published: Undated; retrieved 2026-09-11; retrieved: 2026-09-11.
  2. Original FY June 2024 Form 990. IRS via ProPublica. Primary. Published: Submitted 2025-05-09; retrieved: 2026-09-11.
  3. Opioid Settlement Funds status report. Face to Face via City of Santa Rosa. Primary. Published: Signed 2025-07-17; retrieved: 2026-09-11.
  4. COPHRI. Face to Face. Primary. Published: 2026-07-22; retrieved: 2026-09-11.
  5. Massachusetts overdose education and naloxone study. Walley et al.. Primary. Published: 2013-01-31; retrieved: 2026-09-11.
  6. HEALing Communities randomized trial results. NIH. Primary. Published: 2024-06-16; retrieved: 2026-09-11.
  7. Exploratory community-level cohort analysis. HEALing Communities investigators. Primary. Published: 2026; retrieved: 2026-09-11.
  8. Naloxone Distribution Project. DHCS. Primary. Published: Current program; retrieved: 2026-09-11.
  9. Syringe services. Face to Face. Primary. Published: Undated; retrieved: 2026-09-11.

Annual expenses: years and sources

Average annual expenses (three consecutive fiscal years): $1,903,012. Organization size is separate from the modeled cost-effectiveness of a donation.

Face To Face Sonoma County Aids Network

Form 990/990-EZ reported whole-entity expenses; includes program, administration and fundraising costs, but excludes any costs netted against revenue.

Latest original return and prior-year comparative combined with earlier filing data.