HOPE Pacifica

What do they do?

Legal and program boundary

HOPE’s legal identity is distinct from Pacifica Resource Center, which appears as one access location, and from unrelated charities named HOPE. The official mission covers education, support and advocacy as well as overdose prevention. We evaluate the ordinary legal recipient, not a hypothetical naloxone-only affiliate. An unrestricted donation may support events, transport, administration, insurance, memorial work or distribution. Charging only dispenser hardware while assigning the whole nonprofit’s benefits would misstate the donation being evaluated. Official mission

The current IRS-derived record identifies the exact EIN and April 2024 tax-exemption ruling month. That is identity evidence, not an original annual financial statement. The county’s grant record describes founding in March 2024. Founding and exemption dates serve different purposes; neither is an observed start date for the cumulative distribution total. We do not divide that total by elapsed years to create an annual throughput observation.

Current access network

The Narcan locations page, reviewed September 11, 2026, lists Lytt, Longboard, Fog Zone, Pacifica Resource Center, Cameron’s Pub, San Bruno Catholic Worker Hospitality House, Hearts for Humanity, West Bay Alano Club and San Francisco European Collision Repair. The homepage still presents a shorter list. The dedicated page is our current site-count source. Its nine entries establish listed access points, not independently inspected stock or operating hours. Current locations

These locations imply a plausible mix of host organizations, patrons and community contacts. They do not identify the risk profile of people taking medication or the eventual person on whom it is used. Someone may carry naloxone for a family member, obtain it for workplace readiness, or refill supplies after expiration. Such uses can be valuable, but each implies a different conversion from packs to incremental health. The model therefore does not equate visitor, collector, bystander and at-risk beneficiary.

The official homepage advertises late 2026 events and a February 2027 fundraiser. A scheduled event is evidence of planned activity, not verified attendance, revenue or completed overdose prevention. The newsletter index lists December 2025 and April 9, 2026 issues, but accessible text did not expose their contents. We do not claim those issues contain financial or outcome data. These distinctions matter because a current-looking website can coexist with unmeasured delivery outcomes.

From medication to health

The practical chain is: usable stock is available; someone obtains and retains it; the network includes an opioid-overdose risk; the stock is near the emergency; someone recognizes and responds; the response improves the outcome relative to the fastest available alternative; survival or morbidity differs for a finite time. Failure at any step can make another distributed pack have little marginal health value. Success can matter greatly, but the chain cannot be replaced by the statement that naloxone works pharmacologically.

We distinguish preparation from administration. A pack can be use-ready without ever being used, and the absence of a rescue report does not prove that stocking it was pointless. Conversely, a reported administration does not prove death was otherwise certain. The model uses an expected net hazard difference across covered people rather than interpreting every successful response as one life saved. This is a deliberate constraint on the most tempting source-to-model shortcut.

Spending breakdown and financial evidence

What was and was not found

We did not locate an annual Form 990, Form 990-EZ or audited financial statement for HOPE. ProPublica’s IRS-derived record, updated August 19, 2026, identifies the organization but lists no filings. The available records therefore do not verify its annual income, expenses or assets. This does not establish that HOPE failed to file, lacks internal accounts or has no assets; the index’s zero-filled fields are not financial measurements. IRS-derived record

The record’s filing-requirement code 02 needs particular caution. Current IRS instructions associate it with smaller organizations, but also allow assignment when gross receipts cannot be determined. It is not an audited statement that annual expense is below $50,000, and receipts would not be identical to expense anyway. We therefore do not use that code to delete the $100,000 expense world or increase the lean-world probability. IRS administrative instructions

The available records do not support a three-year financial comparison. For 2023, no whole-organization statement was located, and the county describes founding in 2024. For 2024 and 2025, no original annual revenue, expense, asset, liability or functional breakdown was found. For 2026, public operations and funding-source descriptions exist but no reconciled year-to-date accounts were located. “Not located” is not zero, and a grant document is not a substitute annual return.

The installation grant

The county adopted up to $10,000 for installation in August 2025, with a term ending June 30, 2026. The public record identifies approximate hardware prices of $300–$500 and HOPE’s stocking/maintenance responsibility. It is a one-time award, not a promise of renewal. We have no verified closeout, final payment, unspent balance or current replacement request. The end of the term alone does not establish that a new donor must now finance every operating cost. County adopted item

The county memorandum attributes part of a historical county mortality decline to naloxone access. That administrative statement is not a controlled causal estimate and does not identify HOPE’s contribution. We do not use the county death totals to calibrate the model’s hazard reduction. The grant’s proposed accessible siting also does not establish universal around-the-clock availability at every listed business. Current instructions to check hours are a more appropriate limitation.

A stronger current alternative: free hardware

San Mateo County Health now offers a free stand-box program. Its FAQ describes free hardware and shipping, while hosts obtain and replenish medication. Eligibility is for locations in San Mateo County; the listed San Francisco site is not automatically covered by this program. Selection also depends on availability and location criteria; the county offers help with damaged units. Hosts must report restocking and reported reversals monthly. This is a current funded alternative, not proof every application will be approved or every useful site is covered. County FAQ

This evidence weakens a simple private-hardware-purchase story. A donor should first ask whether a proposed location can receive public equipment. It does not eliminate a coordination role: permissions, host recruitment, replenishment, outreach and access monitoring may still need work. The correct residual is the work that would not happen under the public route, not the retail value of all supplies that happen to pass through a nonprofit.

The program’s January 29, 2026 launch notice and current application route are more relevant to today’s comparator than a 2025 statement that few dispensing machines existed. We do not infer that the original grant was unnecessary when awarded. Counterfactuals change over time. A previously additional installation model can become less additional as public provision expands, even if the organization continues doing worthwhile work. Launch notice

Cash cost versus all resources

The modeled expense envelope is intended to cover all ordinary organizational operating costs, including fundraising and support activities. It is not a complete societal resource bill. Medication procurement, volunteer time, host space, transport, emergency response and clinical follow-up may be paid by someone else or donated. Those complements can enable genuine cash leverage, but they do not have zero opportunity cost merely because HOPE does not pay for them.

Nor should every public dollar be mechanically added to the next private gift. The relevant full-resource calculation would identify resources actually changed by the counterfactual donation. Existing stock that would otherwise expire may have a different opportunity cost from scarce stock diverted from another high-risk network. Without quantities and alternatives, assigning a retail price to all public medication would create another unsupported estimate. We leave the full-resource result unresolved and explain what would be needed to estimate it.

Monitoring and information sharing

Units and time periods

The current dedicated page and homepage now both report over 6,000 Narcan distributions. An earlier copy of the page reported 3,000; the current pages no longer conflict on this count. This is a useful correction, but the current term still does not identify devices, paired units, packs, people or annual flow. The numerical base therefore remains a prior of 1,500 annual two-device-pack equivalents rather than an asserted fraction of the cumulative total.

DHCS defines a nasal unit as two 4 mg devices and advises against separating them. It also reports a switch from branded Narcan to generic supply in April 2026. HOPE’s colloquial brand usage does not establish the exact product in every current dispenser. These facts make a consistent stock ledger important; they do not by themselves show that HOPE’s cumulative count should be halved or doubled. DHCS FAQ

A useful annual table would distinguish stock received, stock installed, stock taken, stock expired or damaged, and stock remaining. It would state the unit and date range. Transfers between storage and a dispenser are not necessarily end-user distributions. Restocking is not automatically consumption. Even perfectly reconciled inventory would still not identify how many distinct people gained additional protection, but it would remove a major accounting ambiguity.

Coverage rather than contacts

The central conversion is 1,500 packs × .5 use-ready fraction ÷ 3 repeat packs per network × .8 risk/network factor = 200 prospective at-risk person-equivalents per annual operating budget. Every factor is an analyst judgment. The first reduction concerns usable accessible stock rather than reserve or ineffective placement. The repeat divisor avoids treating several packs supporting one network as several independent people. The final factor accounts jointly for risk relevance and overlapping networks.

The definitions matter. The ready fraction is not an administration probability; administration and timely alternatives enter the net hazard difference. The repeat divisor is not an overdose rate. The network factor is not a measured fraction of collectors who personally use opioids. A bystander may protect someone else, and several bystanders may protect the same person. We do not add an unobserved multiplier because one collector might know several people: without network data, that would preferentially count upside while ignoring overlap.

The small $1,000 gift reaches 2.5 additional person-equivalents in the central cost/delivery world: 200 × 1,000/40,000 × .5 funding response. This is not a claim that two or three identifiable people were newly protected by a donation. It is the expectation of a model relating whole-budget spending to incremental coverage. The distinction prevents the arithmetic from being presented as a purchased treatment slot.

What a useful follow-up could measure

A low-burden monitoring system could use anonymous site/date/unit records and optional follow-up without collecting unnecessary identifying information. Relevant questions include whether the recipient already had usable naloxone, whether this replaced expired stock, whether an alternative was promptly available, and whether the pack remained accessible. Reports of use should state whether the same incident appears in several channels. A program can learn from these data without publishing private health details.

Self-reported reversals would be informative, but not sufficient for causal mortality. Reports may omit unsuccessful events, people who cannot be contacted, or administration by another responder. Reports may also be duplicated by the collector, host and emergency service. A denominator of successful reports is selected on surviving long enough to report. We would not import its apparent survival into a prospective risk cohort or multiply it by an assumed death probability without a separately justified counterfactual.

Geographic attribution

All nine listed locations are in the nine-county Bay region, but the model’s 95% central Bay share remains a judgment. People travel, carry medication elsewhere and obtain it for contacts outside their home county. A site address is not the eventual beneficiary’s residence or the location of all future health gains. Conversely, a San Francisco location does not justify assigning an SF share to the entire network. We retain no separate SF estimate for this packet.

The added sites strengthen evidence that the organization is not limited to two cities. They do not justify multiplying modeled health by 9/5. Site counts differ in throughput, hours, users and alternatives. A newly listed location may replace another point, formalize existing delivery or have little traffic. The right update would use additional useful coverage, not an equal-benefit assumption for every pin on a map.

Causal evidence and finite health

Biological efficacy is not the distribution effect

Naloxone can reverse opioid respiratory depression, but the organization-level comparison is access with an additional gift versus access without it. Another bystander, emergency service or existing pack may already provide timely treatment. Some packs never encounter an emergency, and some overdoses involve other substances or circumstances that naloxone alone does not resolve. The model’s net hazard reduction is meant to include these probabilities, not to be multiplied by a second generic rescue fraction.

Walley and colleagues’ Massachusetts interrupted time-series study found lower community overdose mortality associated with overdose-education/naloxone implementation: adjusted ratios .73 and .54 for lower/higher implementation compared with none. It is observational and from 2002–2009. It supports plausibility but does not identify a modern Bay effect per pack, remove confounding or establish that another unit is valuable where distribution is already extensive. Primary study

The HEALing Communities trial is an important counterweight. Its multicomponent community intervention did not demonstrate a statistically significant primary overdose-mortality reduction: adjusted rate ratio .91, 95% interval .76–1.09. A separate randomized distribution analysis found 79% more naloxone distribution. These results show that increasing distribution does not mechanically establish a detectable mortality change. They do not isolate naloxone’s pharmacology or prove that all targeted access is ineffective. Mortality trial Distribution analysis

The coexistence of plausible efficacy and uncertain marginal delivery is exactly why null worlds remain in the model. A wide trial interval should not be translated into a precise charity probability. Nor should a positive observational estimate be applied to a cumulative website number. The studies constrain our confidence and identify failure modes; they do not provide the local coefficients used in this exploratory calculation.

Avoiding selected-survivor mortality transfer

Historical post-overdose cohorts have high subsequent mortality. Larochelle reported 4.7 all-cause deaths per 100 person-years; Olfson reported 7.783 in a first-year cohort. These are selected survivors of an overdose, not the average person indirectly reached by a dispenser. Age, substance use, treatment, calendar period and selection all affect the transfer. We retain these studies only as historical scale context, not as a measurement of HOPE’s baseline or ten-year prognosis. Larochelle Olfson

The central current-care hazard of .07 per year is therefore a strong prior, not a fact about HOPE recipients. Its avoided hazard is only .002 during one year, after which both arms share .07. The model does not suppose every covered person overdoses, or that every overdose would be fatal. A future empirical calibration should start with the prospective risk network and its actual alternatives, not select people because a rescue succeeded and work backward.

One trajectory per person

During the first year, central survival is exp(−.068t) with improved access and exp(−.070t) under current care. After year one, each year-one survival probability is multiplied by exp(−.07(t−1)). The difference can persist because someone alive at year one may remain alive later, even though there is no continuing program effect. The integral stops at ten years. It does not restart whenever another pack is distributed or another overdose occurs.

This is a finite cohort model, not a repeated-rescue model. Repeats only reduce the estimated distinct covered population. The model’s future survival is also not a forecast that every saved person lives ten years: common subsequent mortality reduces the survival gap continuously. A shorter horizon is still a legitimate conservative boundary, especially when long-run prognosis and the utility of survival are uncertain.

The central discounted survival difference is .0009366601 years during active support plus .0107728637 years afterward per covered person. Multiplying their sum by .65 utility gives .0076111905 QALY before the independent harm. Thus 92.0009% of gross modeled survival benefit comes after year one. This is not a second favorable-scenario weight; it is a temporal decomposition of the central row. A model can be mathematically finite while still depending strongly on unmeasured later life. The report therefore shows short horizons prominently rather than describing the ten-year cutoff as sufficient proof of conservatism.

Harms and omitted morbidity

Independent morbidity harms are subtracted per additional covered person. The central amount .00002 QALY is a judgment, not an observed HOPE adverse-event rate. The adverse world uses zero mortality gain and a larger morbidity decrement. These cases keep the calculation signed without implying that ordinary naloxone distribution usually causes harm.

A CDC comparison of 8 mg and 4 mg products found no observed survival advantage for the higher dose and more withdrawal signs/symptoms. It was not randomized and is not a HOPE evaluation. We use it to reject “more dose is always more benefit,” not to map its event rate to our harm coefficient or suggest withholding an appropriate emergency response. Primary MMWR

No separate brain-injury prevention, treatment engagement, infection prevention or family-survival benefit is added. Such effects might exist, but a common positive story is not sufficient to quantify them. Likewise, grief support is not assigned a generic utility bonus. Keeping those pathways unresolved is more honest than expanding the model until it appears complete.

Decision and reproducibility

Retain HOPE as an exploratory access hypothesis and HOLD an actionable health-based giving recommendation. The new source check improves current operations and reveals a materially stronger public alternative. It does not supply annual finances, unique coverage or a local mortality effect. The result therefore remains prior-driven, with unchanged old/new prices and explicit evidence changes.

The packet preserves the original model, full 18-row signed outputs and all new diagnostic matrices. Tests compare original numerical fields exactly, check the closed-form survival integral independently, verify threshold arithmetic and reject malformed/nonfinite inputs. Passing tests establish computational consistency, not the truth of the hazard, expense or funding priors.

Source dates distinguish dated government actions, undated mutable pages and retrieval times. Failed full-text retrievals are not described as successful reads; inherited primary evidence remains labeled. The current native page supersedes older cached operational text without rewriting that earlier provenance. No organizational interviews or private monitoring records informed this assessment.

Model version: hope-pacifica-depth-v2-unchanged-priors. Read the numerical estimate alongside the full assessment: the public hardware alternative and uncertainty about additional coverage are central to the donation decision.

Qualitative assessment

HOPE’s comparative possibility is local trust and implementation, not proprietary medication or hardware. A network of willing hosts may create access where a government application route alone does not. Community members affected by overdose may recognize barriers that a generic distribution plan misses. These are hypotheses about organizational advantage; the public record does not quantify them or show that they are absent elsewhere.

The strongest immediate alternative for eligible San Mateo County hosts is not “do nothing.” It is using the county’s free equipment route and public medication supply, possibly with another existing partner. That hardware route does not automatically extend to the listed San Francisco site. A fair comparison would ask whether HOPE reaches different people, maintains better uptime, lowers practical barriers or accelerates deployment. If the same useful coverage happens promptly without an extra gift, the incremental health is small even when the service itself is valuable.

Other health interventions may offer more directly measured treatment effects, but cross-charity ranking is not resolved by noting that naloxone can save lives. A low-cost access mechanism with sparse data can be worth investigating; a precise-looking ratio should still receive less confidence than its missing denominator permits. The report keeps the promising possibility and the unfavorable scenarios together rather than forcing a binary good/bad label.

Community healing, education and remembrance should not be dismissed because this health model does not quantify them. They may be reasons some donors value the organization independently of Bay QALYs. They also consume real resources within an ordinary gift. A donor prioritizing those outcomes should assess their quality and importance directly instead of treating them as an unexplained positive residual that rescues a health ranking.

What do you get for your dollar?

Cost, coverage and health equation

For each cost/delivery world, Bay QALYs = gift/whole-expense prior × annual distinct risk-person equivalents × funding response × [utility × discounted survival difference − independent harm] × Bay share. The annual person-equivalent conversion is packs × ready fraction / repeats × risk/network factor. Cost, coverage and clinical effects are not separately observed; their product is an explicit conditional estimate.

The discount rate is 3% annually, implemented continuously as ln(1.03). During active support, the survival integral uses the two hazards. Afterward, it carries the year-one survival gap forward under common mortality. An independent numerical integration test checks this closed form, and a person-time bound prevents gross survival benefit exceeding the finite available utility-weighted lifetime of the additional cohort.

Scenarios and expected value

Three cost worlds—$20,000 at 20%, $40,000 at 50%, $100,000 at 30%—are crossed with six delivery worlds. Delivery weights are 20% funding-null, 20% access-null, 10% harm, 20% cautious, 25% central and 5% favorable. They are subjective judgments, not empirical frequencies or confidence intervals. Independence of expense and delivery is a simplifying assumption, not an observed property of this nonprofit.

The expected Bay health per $1,000 is .013101904653691494 QALY, implying $763,247.8074233638 per 10. The central cost/delivery row is $554,659.5517271358. A central row can be better than the mixture because the mixture includes null and harmful possibilities. The favorable family contributes 65.6865% of net expected Bay health. Removing it and renormalizing the remainder yields $2,113,122.33.

No numerical coefficient was changed to compensate for the new public hardware alternative. The preferred result is a continuity estimate, not a claim that the new evidence leaves confidence unchanged. It strengthens the reason to investigate lower funding response. An arbitrary replacement value would conceal judgment rather than improve the evidence. The next table makes the implication visible without assigning unsupported probabilities.

Decision-relevant diagnostics

The same numerical result for half packs and doubled repeats follows from their algebraic roles, not independent evidence confirming the reduction. Nor should these stresses be stacked automatically: some may describe the same uncertainty. The half-hazard result is slightly more than double the base price because harms remain while positive survival benefit falls. This illustrates why signed models should not be treated as a single positive product under every change.

The cost/delivery dependence diagnostic pairs a low-cost cautious state, central-cost central state and high-cost favorable state. It is not another evaluator expectation: its different weights and absence of the full null/harm family make it unsuitable as a replacement headline. Its purpose is to inspect how favorable delivery can require a larger budget. The complete paired outputs are saved so a reviewer can see the assumptions rather than only a selected favorable ratio.

Threshold interpretation

Keeping every central factor except annual risk coverage fixed, about 111 annual person-equivalents would be needed to reach $1 million per 10 Bay QALYs, compared with the current 200 prior. About 1,109 would be needed to reach $100,000. Alternatively, holding coverage fixed requires funding response about .277 for the first threshold and 2.773 for the second. A response above one is outside the model’s admissible range, so funding response alone cannot deliver the $100,000 target under the other central inputs.

These are conditional thresholds, not estimates of actual coverage or an invitation to choose the parameters that clear them. Their practical use is to structure diligence: can the organization support the relevant order of magnitude after deduplication and alternatives? If not, a high-return conclusion fails without needing a perfect trial. If yes, it remains necessary to examine the clinical and expense assumptions.

Funding and previous grants

The most plausible residual bottlenecks

Public supply does not automatically reach every high-risk network. Someone must identify hosts, obtain authorization, coordinate ordering, keep inventory usable and make locations discoverable. A small nonprofit may be well placed to do that work. A useful donation might support reliable replenishment, extend a neglected location’s availability or sustain an activity whose existing grant has genuinely ended. These are plausible mechanisms, not verified current projects.

The competing explanation is that hosts or county staff would provide equivalent access anyway, or that stock and volunteer attention rather than money are limiting. A gift might replace another donor’s funds or support general events without increasing risk coverage. The funding-null world covers that possibility. Public provision is neither automatically a substitute nor automatically free leverage; the local allocation of responsibility determines which it is.

A specific request should identify the current baseline before describing the proposed change. Existing boxes, stock, committed grants and volunteer commitments belong in the baseline. If the proposed site already qualifies for public hardware, the private budget should explain the remaining cost. If the problem is a replenishment backlog, a credible plan should show its scale and why an additional person-hour or transport expense addresses it.

Spendable marginal room

No current restricted balance, unrestricted cash balance, staffing budget or priced $1,000 tranche was located. The absence of a public offer does not prove the organization cannot use funds well. It means the analysis cannot tell a donor what the next gift will purchase. The model’s small gift ceiling is a safeguard against broad scaling claims, not a certificate that $1,000 is absorbable at the modeled rate.

The official site links a general PayPal donation route. The prior pass did not verify its recipient display at checkout, and this review did not conduct a transaction. No donation URL is added to the structured report solely because a button exists. Even a verified payment route would establish how to send funds, not the counterfactual benefit of sending them.

Three unsent questions

1. What were whole gross expenses, revenue, restricted balances and contributed resources for the latest completed year, and what remains committed after the 2025–26 installation grant? Please distinguish event costs, operating costs and publicly supplied medication or hardware.
2. For a dated twelve-month period, how many paired units were received, stocked, collected and expired; how many distinct risk networks gained usable access; and what evidence separates that access from repeat or otherwise available supply?
3. What specifically would an additional unrestricted $1,000 change now, beyond the county free-box program, NDP medication and existing commitments? Which location or maintenance gap, complementary resources, time period and expected additional coverage would it support?

These questions are deliberately narrower than a request for a perfect impact evaluation. A reconciled budget and an operationally specific plan could materially improve the decision even if local mortality remains too rare to estimate precisely. No question has been sent and no organizational response is implied.