Project Homeless Connect

Scope and recipient

PHC is a San Francisco service-access project. Its official history places the move to Community Initiatives in July 2020, and its donation page directs checks to Community Initiatives with Project Homeless Connect in the memo. The sponsor EIN is 94-3255070. Earlier records under the San Francisco Public Health Foundation describe a different sponsorship period; they should not be combined with the current sponsor's totals as though they were consecutive PHC stand-alone accounts. Sources: PHC history and giving instructions.

The assessment unit matters more than the label “whole organization.” Every dollar of a gift directed to this project is counted. The analysis does not divide only the money assigned to glasses by the health assigned to glasses and call that an ordinary gift. Conversely, it would be equally wrong to charge all Community Initiatives expenses against PHC patients. The sponsor serves many projects, and an unrestricted sponsor-wide gift has a different destination and counterfactual.

The model is best described as whole project-gift cost, partial health benefits. It is not an annual expense divided by observed annual outcomes. We lack the current project ledger needed for that calculation. A prospective $100,000 gift is instead allocated across a hypothetical portfolio, with explicit delivery costs and a large unquantified remainder. This is a useful decision model if readers understand that its biggest spending inputs are not observed allocations. It is not a substitute for the missing ledger.

Nor does partial health mean the rest of the project is worthless. Navigation, personal documents, mail access, basic supplies, social contact, and referral support can be valuable in their own right. Some may produce substantial health effects. They remain outside the numerical benefit estimate because there is no sufficiently specific chain from an additional PHC gift to completed services and finite incremental health. Keeping them unquantified prevents a convenient but unsupported conversion of every contact into QALYs.

This scope also prevents comparison mistakes. An equipment-only restricted intervention may have a lower reported cost because it omits the surrounding portfolio. A whole-cost model that quantifies more pathways may appear more favorable because it has more measured benefits. Neither difference establishes which ordinary gift is better. The relevant comparison is a common gift boundary with candid disclosure of what is and is not included in health.

Delivery and completion

The current Core Senses description offers prescription glasses, hearing support, and denture pathways. Its dental route uses a waiting list; dentures use applications and a lottery. Named partners include UCSF, University of the Pacific, Native American Health Clinic, and Hearing and Speech Center of Northern California. These are descriptions of a service network, not records of available appointments or completed treatment today.

A particularly useful dated source is the February 10, 2025 Optical Day notice. It describes glasses selection and fitting for eligible Medi-Cal recipients with a valid prescription, and excludes examinations at that event. We therefore distinguish access to a prescription, dispensing a pair, appropriate correction, and continued use. A person arriving at an event has not necessarily completed the whole sequence. The notice also makes public insurance part of the factual baseline rather than an abstract possible alternative.

The current Every Day Connect page describes navigation and practical support, while the homepage describes in-person, off-site, phone, and email access and thanks participants in the 85th Community Day of Service. These establish continuing public-facing operations. They do not establish a September 2026 clinical throughput, waiting time, or verified capacity increase. An undated calendar or copyright year cannot supply those missing dates.

A stronger dated sign of activity is the Elections Department's June 17, 2026 report, which includes PHC among organizations with which outreach staff partnered in May. This is evidence of a continuing partnership after the funding-cut discussion below, not evidence of restored optical, hearing, or denture capacity. It would be inaccurate either to infer complete closure from the 2025 cut or to infer continuity of the old clinical model from this civic-service activity.

The historical FY2018–19 impact preview, published September 27, 2019, usefully separates 771 prescription glasses, 281 vision examinations, 295 dental services, 10 denture sets, and 153 hearing screenings. Those are different units; the hearing number is not aid fittings, and dental services are not necessarily unique completed treatment courses. The same publication reports unduplicated participants separately for Community Days and Every Day Connect, not a deduplicated sum across both. Its self-reported access and satisfaction responses are not randomized counterfactual estimates.

The current history page says more than 6,000 people annually; the undated service-coordinator description says more than 8,000. We do not infer growth from this discrepancy. Neither statement provides a compatible year, cohort definition, or clinical denominator. They may describe different time periods or definitions. A current output table should resolve that before either figure is used to price patient health.

For a prospective donor, the relevant funnel starts with people who have an unmet need, then follows those eligible and reachable, those referred, those attending, those receiving an appropriate device or denture, and those continuing to benefit. PHC may influence several transitions. A clinical partner may influence others. A count at the first transition cannot be substituted for a count at the last. The accepted model starts with a cost per completed episode, so it intentionally folds recruitment and failed appointments into cost rather than multiplying another completion rate after the fact.

This is a model convention, not evidence that current costs actually include those failures. The distinction tells us what to request from financial and service records: the numerator should include unsuccessful attempts and the denominator should be completed appropriate episodes. If a quoted price is merely a pair of lenses, laboratory work, or an event supply charge, it does not meet this definition.

Finances and spending

The sponsor financial index links the latest three available Form 990s. The fiscal dates printed inside the returns take precedence over confusing filenames. These are Community Initiatives totals, not PHC totals:

Original filings: FY2023, prepared January 26, 2024, FY2024, prepared February 5, 2025, and FY2025, prepared March 25, 2026. No PHC-name allocation was found in the reviewed returns. This is a bounded public-record finding, not a claim that the sponsor lacks internal project accounts.

The FY2025 audit, issued January 5, 2026, reports $89,372,099 expense: $76,204,047 program, $8,258,245 management, and $4,909,807 fundraising. Its principal natural categories are $45,610,879 salaries and benefits, $23,835,649 grants, and $6,568,335 professional services. Note 12 identifies $19,976 PHC contributed goods. None is a PHC cash budget. The $229,557 audit/990 expense difference reconciles to $112,370 event costs, $116,887 cost of goods sold, and $300 donated services in Schedule D; it is not unexplained PHC spending.

For context, audited sponsor expense was $68,562,676 in FY2024 and $55,548,613 in FY2023. The FY2023 audit separates program $45,803,808, management $5,609,495, and fundraising $4,135,310. These figures show why mixing Form 990 functional expense with audited gross expense without reconciliation is unsafe. They do not identify which PHC service grew or whether the next donor dollar is needed.

There is also a narrow three-year in-kind trail. The FY2023 audit lists $31,035 clothing and household goods, described as primarily used by PHC, rather than an exclusive PHC allocation. The FY2024 audit explicitly names $64,730 for PHC, followed by $19,976 in FY2025. These are recognized contributed-goods categories, not all donated clinical labor or all project resources. The changing values should not be read as a proportional change in clinical capacity, expenditure, or need. No amount from this series replaces the model's prospective outside-resource priors.

For PHC itself, the annual spending table is therefore necessarily incomplete:

The published standard sponsor fee is 10% of gross receipts and 15% for government funds, with a $50,000 annual fundraising minimum. These are current undated standard terms, not a verified PHC contract. The model already includes allocated sponsor and administrative cost in each completed-care unit and retains the rest of the gift. A second blanket fee deduction would charge the same kind of overhead twice. Conversely, a device-only advertised price should not be mistaken for a unit cost inclusive of this overhead.

The sponsor's aggregate assets and restricted funds cannot be used as PHC reserves. Fiscal sponsorship is precisely a setting where legal-entity totals can be economically misleading for a project-directed donor. A large sponsor balance does not prove PHC has usable cash; a sponsor operating deficit does not prove an urgent PHC shortfall. The missing items are the project's current unrestricted balance, committed revenue, liabilities, and approved spending plan, not another calculation based on the sponsor-wide ratio.

A project-specific historical budget is available and changes the financial picture. The May 2, 2024 HSH grant-amendment packet identifies contract 1000015983, with a $1,460,295 annual budget for FY2023–24 and proposed continuation at that amount in FY2024–25 and FY2025–26. The city's 2024 contracting report subsequently lists the $8,819,447 multiyear ceiling through June 2026. That ceiling includes contingency and is not one year's spending or a remaining grant balance. The later 2025 termination discussion prevents treating its originally scheduled end date as proof of funding continuity.

The packet's FY2024–25 budget narrative supplies real project-associated cost categories:

Operating expense includes $142,280 rental, $37,000 IT, and $35,000 program supplies. The budget gives 4.70 FTE and $274,960 to cross-trained Every Day Connect service coordinators, not separate clinical unit costs. Its city-source classification is General Fund. These facts directly contradict reading the undated private-funding claim as a description of all historical PHC financing.

The same source labels actual city-contract spending of $1,261,088 in FY2020–21, $1,199,277 in FY2021–22, and $1,394,077 in FY2022–23. Those are historical contract expenditures, not the latest three whole-project actual expense statements. They cannot be combined with the sponsor's in-kind line or treated as exhaustive private-plus-public PHC costs without the project ledger. We retain the current project expense field as unknown while preserving these useful narrower observations separately.

These records substantially improve the spending assessment: PHC was not merely a tiny device-purchasing program, and paid coordination, space, and indirect costs were important. They do not identify the allocation of an ordinary private gift after the city funding transition. The 15% indirect budget line must not be added to the model's already inclusive unit costs, nor confused with a verified current private-gift sponsorship contract. A new post-transition budget is the next decisive document.

Clinical evidence and transfer

The three pathways share a plausible clinical premise: appropriate correction can improve daily functioning. The major uncertainty is the size and duration of generic health gain among the additional patients reached by this gift. That differs from asking whether glasses, hearing aids, or dentures can help anyone.

For glasses, Griffiths and colleagues, 2014, observed EQ-5D increasing from .850 to .925 in 41 treated, followed refractive-error/presbyopia patients out of 113 recruited. The study was uncontrolled, used Zimbabwe preference weights, and documented substantial failure to receive prescribed spectacles. The .075 change is therefore not a randomized effect for every recruited patient, still less a PHC estimate. The accepted .0375 utility assumption is a judgmental transfer, not an identified correction for confounding.

The preserved “source attrition glasses” scenario further multiplies .0375 by 41/113. It is a pessimistic transport stress test, not a clean missing-data adjustment. Because that ratio includes nonreceipt of treatment and our unit already means completed care, mechanically imposing it as a delivery probability would risk double discounting. We preserve its numeric output to keep the historical sensitivity auditable while clarifying its interpretation. A future update should distinguish treatment receipt from post-treatment missing outcomes instead of treating every excluded source participant as the same kind of loss.

The glasses health estimate dominates the calculation. It merits more scrutiny than a long list of peripheral benefits. Relevant local information would include presenting visual impairment, whether prescriptions are appropriate, use and breakage, and whether another pair would have been obtained soon. Generic utility may differ substantially between correcting modest refractive error and restoring function in someone unable to carry out daily activities. The model does not observe PHC's mixture of those needs.

For hearing, the accepted source ledger includes Mulrow's randomized hearing-aid comparison and Ye's randomized economic study. These support investigating hearing care; they do not establish 0.02 QALYs per PHC patient-year. Communication scores are not preference utility. The 0.02 QALY figure in the economic abstract is not imported as an annual local effect. This V2 retains the prior evidence boundary rather than extending hearing benefits to cognition, falls, survival, or lifetime functioning.

For dentures, Pearson and colleagues' 2007 randomized trial compared immediate versus delayed domiciliary treatment in 133 older edentate adults and found better oral daily-performance outcomes after treatment. Its outcome instrument is not generic utility. The accepted .03 utility and one-year window are judgments. The earlier denture-material economic study remains a caution about instrument sensitivity, not evidence that no dentures are better than dentures or that all oral improvement becomes EQ-5D gain.

We do not add independent employment, housing, social confidence, and mental-health effects on top of generic utility. Some might already be represented in the preference score; others require separate evidence and causal links. Nor do we infer that providing dentures prevents a specified number of deaths through nutrition. A plausible story is not a calibrated mortality effect.

The one-year cap is an explicit conservative modeling choice, not a measured retention period. It can omit genuine longer-lived benefits while still being optimistic about near-term use. Those possibilities are not contradictory. A device can remain physically intact for years but produce less useful correction than assumed, or be replaced by equivalent care in the counterfactual. The current model exposes those uncertainties with use and alternative-access assumptions rather than claiming durability data it does not have.

Decision and monitoring

Our comparative judgment is cautiously favorable toward the access mechanism, but conditional toward the donor price. We would not dismiss PHC because it collaborates with publicly funded providers: helping a person use an otherwise inaccessible benefit can be valuable. We would not endorse it solely because an advertised device price looks inexpensive. The decisive question is what the next project-directed gift changes in completed, useful care.

The most valuable funding document would be a current PHC project budget with unrestricted cash, committed revenue, sponsor charges, partner payments, and a realistic expansion plan. It should distinguish maintaining existing service from expanding it. Maintaining a program can have additional value if the no-gift alternative is a real reduction, but that counterfactual must be established rather than assuming every recurring gift is incremental growth.

The most valuable service evidence would link the budget to unique patient episodes by modality: referral, attendance, appropriate dispense or fitting, follow-up, use, and reason for noncompletion. A small consistent dataset is preferable to a large mixed contact total. It should identify which patients have equivalent options and the delay to those options, because faster completion may be the principal benefit rather than otherwise impossible treatment.

There is some historical monitoring evidence rather than a complete vacuum. The 2024 HSH packet reports no findings in sponsor fiscal monitoring for FY2021–22 and PHC program monitoring for FY2022–23, including a June 6, 2023 visit reviewing policies, staffing and reports. This is evidence about compliance and documented process, not a causal estimate of health. It neither validates every assumed clinical completion nor establishes the quality of operations after the funding transition. The existing monitoring trail suggests a concrete starting point for a future update: obtain the relevant project reports and reconcile their unit definitions before commissioning a new evaluation.

The most valuable partner evidence would name the capacity being unlocked and its payer. If an optical lab, audiologist, or dental clinic can accept a defined extra cohort, that is different from a referral queue without appointments. If a public benefit pays for devices while PHC supplies navigation, the model should price that pathway accordingly. If a partner's capacity is fixed, the question becomes whether PHC changes allocation among patients or adds any total health at all.

Monitoring should also include adverse outcomes, failed fitting, device loss, repeated appointments, and patient burden. These are not reasons to presume harm; they are reasons to evaluate useful completion rather than delivery alone. For generic utility, patient-reported functioning can help but should not be converted by an invented scale mapping. A validated preference instrument and a credible comparison would materially improve the central health estimate.

For a donor today, the honest route is the official PHC donation page, with project designation preserved. Supporting the broad mission is reasonable if that is the donor's aim. Treating a particular amount as a verified purchase of additional glasses, hearing aids, or dentures is not supported by the reviewed public evidence. Before a substantial health-maximizing gift, request the project ledger and partner-backed completion plan; do not infer a clinical restriction from the general fundraising examples.

No outreach was performed. No current marginal offer, clinical restriction, project reserve balance, or new patient capacity was verified. These are precise remaining information gaps, not evidence that additional support cannot help.

Partner attribution and overlapping health

PHC's hearing pathway explicitly depends on Hearing and Speech Center of Northern California. The same patient might appear in both organizations' service records and in separate charity reports. We cannot assign the entire hearing benefit independently to both gifts and then add them. A donor portfolio supporting both requires a joint counterfactual: which additional patients or stages change when both organizations receive support, compared with neither or one alone?

The current PHC model is a stand-alone gift scenario, not a jointly optimized portfolio of gifts to partners. Its financial additionality assumption is meant to absorb whether PHC's contribution actually changes completed care given the partner baseline. It is not an empirically identified attribution share. The outside-resource allowance prices associated clinical inputs without establishing that they are newly induced. Readers should not confuse charging for resources with solving causal attribution.

Within PHC, the model assigns a primary clinical stream to each modeled patient-year: glasses first, then hearing excluding glasses recipients, then dentures excluding both. Central nonoverlap shares are 1, .8, and .8. These are judgments, not linked records. This is a deliberately simple protection against counting the same person's generic health gain multiple times. It may undercount genuinely distinct improvements, but an additive alternative would need an explicit combined-utility ceiling and evidence about the overlap.

All corresponding costs and procedural harms remain even when benefit is excluded for overlap. This avoids the favorable mistake of removing the cost of duplicate services while keeping their clinical upside elsewhere. It also means a pathway can have a negative net contribution in a cautious scenario without asserting that the real-world intervention is generally harmful.

The same discipline applies to referrals to housing, mental-health, addiction, or legal services. Referral counts are not treatment completions, and completion does not automatically identify PHC's contribution beyond the service provider. No such extra QALYs have been added. A future expansion would need a linked cohort, a finite outcome difference, and a no-PHC pathway through the same publicly funded system.

Model and exact results

The current executable model is phc-independent-health-income-current-2026-10-01. It revises the original health magnitudes rather than keeping their total fixed while adding economic scenarios. Our central estimate is $150K per DALY, versus the prior $77K. Per illustrative $100,000 of project support, Bay health is 0.62 clinical QALYs and net resources contribute 0.05 income-equivalent years, totaling 0.6655164024. This is a normalization, not a claim PHC can accept a $1.5M tranche at that productivity.

Independently reassessed clinical effects

For glasses, the original Los Angeles randomized delayed-correction trial supports vision-specific improvement after three months, but does not provide a generic utility conversion. The 2026 original Hong Kong study finds no significant adjusted generic utility decrement for its correctable distance-impairment group; a mild older cohort does not establish zero benefit for PHC's mixed prescriptions or near vision. Together they weaken reliance on half the uncontrolled Zambia change. We now use 0.02 utility, not 0.0375, as a modest positive judgment—not a measured PHC effect, arithmetic average or statistical confidence bound.

Hearing utility stays 0.02 after checking original randomized evidence of communication benefit and uncertain generic utility. Denture utility falls from 0.03 to 0.015, reflecting condition-specific evidence without a robust generic utility estimate. Local costs, use, additionality and alternatives remain explicit priors because public activity does not identify these inputs.

For each pathway, additional episodes equal the smaller of allocated donor spending divided by inclusive episode cost times financing additionality, or the additional-capacity cap. Distinct recipients and unmet equivalent care are separate factors. Clinical benefit integrates signed utility and use for at most one year after care begins, with delay, 3% discount and a 0.02 mortality hazard. Procedure harms apply to all additional episodes, without positive-benefit filters. Fifty-five percent of the project budget remains in cost without quantified benefit; no extra sponsor fee is added.

Separate net resources and overlap

The THRIVE original randomized trial supports a near-vision earnings mechanism, not a local PHC earnings coefficient. Centrally, 10% of distinct otherwise-unmet glasses recipients are assumed to gain net $1,000 annually for six months after care. This means 4.67 responsive households in the illustrative case. Net resources subtract taxes, foregone earnings, benefit withdrawal, work costs and displacement; they are not gross wages. The $25,000 annual household-resource denominator is a reference assumption, not measured PHC income.

Income-equivalent years use 0.5 × responsive households × independently discounted income duration × log(1 + net gain / baseline resources). Clinical utility and aid-use factors do not reduce that realized economic response again. With baseline resources $50,000, hold the $1,000 gain fixed rather than silently doubling it.

A separate 5% of distinct glasses recipients would otherwise buy equivalent correction; they save an assumed $50 once. They are outside the 70% unmet-care group, so the fractions cannot exceed one. This saving is not PHC's fundraising price or insurance spending. The remaining unmeasured benefit-access effects are not proven zero.

To limit double counting through generic financial distress, centrally exclude the entire positive clinical credit of the 10% earnings-responsive subgroup before adding earnings. This conservative allocation does not mean earners lose vision benefit. It is independently calculated, not old total minus income; actual positive earnings must exist before the reservation applies. Signed clinical harms are never excluded. The coexistence sensitivity retains genuinely distinct functional and income effects together.

Current diagnostic cases

No additional funding gives no additional care or recipient savings. Gift-induced failed applicants can still incur an independently specified resource burden; negative totals have no finite positive price. Portfolio allocation owns shared effects once; independent donor-specific harm persists even if no shared benefit is assigned. Bay 98% and SF 95% are nested residence judgments applied once to health and resources.

The full donor numerator is retained. The gross associated-resource envelope remains $140,000, giving $210K per DALY, not net induced societal cost or public savings. The production model is bounded to positive illustrative gifts no larger than $100,000 pending amount-specific capacity evidence. Original sponsor audits and the city-contract budget do not identify current PHC annual productivity or funding room.

Historical model and exact results

The following earlier coefficients, outputs and economic scenarios are retained to show the previous assessment. They are not the current ranking or current central model.

The prior executable model is phc-health-income-calibration-2026-10-01. Its historical health-only calculation, phc-whole-portfolio-finite-health-v1, remains reproducible without coefficient changes. That prior model added separate net-income components and independently signed sensitivities after re-examining the clinical and financial evidence. It uses no probability weights. The central row is a central scenario; favorable and cautious rows are joint assumption packages, not statistical confidence intervals.

For each pathway, nominal completed episodes equal gift times allocation divided by complete PHC cash cost. Additional episodes are the smaller of nominal episodes times financial additionality and the explicit additional-capacity cap. Distinct patient-years multiply this by the nonoverlap share. Gross health then multiplies by no-equivalent-alternative share, signed utility, and effective discounted years. Procedural harm is charged on every additional completed episode before overlap and alternative filters. Independent harm is subtracted separately.

Effective years integrate from the care delay to delay plus the benefit horizon, discounting at 3% and applying a continuous mortality hazard of .02, then multiplying by effective use. The input named annual_mortality is thus implemented as a hazard, not an exact annual death probability. These values are priors, not a PHC survival study. The horizon is at most one year after care begins, so .25-year delayed care can run to 1.25 years after the gift. No lifetime benefit is claimed.

The total is 1.32 US QALYs, 1.29 Bay QALYs, and 1.25 SF QALYs. Residence shares .98 Bay and .95 SF are nested priors, not observed addresses. Do not add the geographic columns. Donor costs per DALY are $76K US, $77K Bay, and $80K SF.

The gross resource envelope is $140,000: the whole $100,000 gift, $30,000 of associated clinical outside inputs, and $10,000 for outside inputs to other work. It yields $108K per DALY. Outside resources are charged per nominal episode, including activity that the gift might replace rather than create. This is a gross associated-resource sensitivity, not net induced societal cost, not a complete measurement of all donated services, and not a claim of public savings.

The price per DALY is a normalization of the $100,000 scenario, not a literal promise that a $774,408 gift would produce 10.00 Bay QALYs. Holding the other assumptions fixed, the additional-capacity caps begin binding at a $150,000 gift for glasses and hearing and $160,000 for dentures. Beyond those points, modeled health stops scaling with nominal expenditure in the capped pathway. Fixed operating costs and the size of the post-cut funding gap could create other nonlinearities that this simple model does not estimate. A larger donor should therefore request an amount-specific plan, not multiply the normalized ratio by a desired health quantity.

In the preserved joint favorable case, donor Bay cost is $4K per DALY. In the cautious case it is $59.7M. Lower clinical allocation alone gives $310K; lower financial additionality $387K; rapid equivalent alternatives $613K; and limited additional capacity $504K. These results show that plausible structural uncertainty dominates fine numerical precision. They do not describe measured frequencies of success.

No funding additionality and no additional capacity each give zero modeled health. Null utility with no procedural harm also gives zero. Signed adverse utility gives -0.74 Bay QALYs, while an independent one-US-QALY harm with zero clinical activity gives -0.98 Bay QALYs. Positive cost-effectiveness ratios are suppressed for zero and negative health; the gift cost remains. The central small harm inputs are stress-test priors, not documented adverse-event rates at PHC.

Historical retained-price health and income calibration

The previous ranking used a central marginal scenario of $77K per DALY: 1.29 Bay health years plus zero quantified net-income years per $100,000 of modeled project support. Zero is a provisional response to unidentified net household effects, not a measured absence of economic benefits or a statistical best estimate. The cost denominator keeps all project support, including the 55% allocated to otherwise unquantified work. The sponsor's audited annual spending and PHC's historical city-contract budget are not substituted into it.

Better vision can improve earning ability as well as daily functioning. The THRIVE randomized trial studied 824 adults aged 35–65 in Bangladesh, selected for near-vision-intensive work and no prior glasses. Median monthly income was $47.10 with correction versus $35.30 in the control group at eight months, a 33.4% contrast. That supports a real earnings mechanism. It does not identify PHC's employment-responsive share, prescription mix, income, sustained earnings effect or benefit offsets. We therefore do not apply 33.4% to every PHC patient.

Other plausible income effects include avoiding an actually paid clinical purchase, improved benefit access, caregiver time and patient acquisition costs. Insurance payments, donated-device accounting values and fundraising examples are not household disposable income. A recipient who would otherwise forego purchase has no purchase saving. We do not add ordinary lifetime earnings simply because health improves, or relabel a health-utility increase as a second income benefit. The external utility measure may partly capture downstream financial stress; that valuation overlap is unresolved, so we add no separate social-confidence or mental-health bonus.

Income-equivalent years follow the project's Coefficient Giving comparison convention: 0.5 × affected people × ln(1 + net disposable-income change / baseline income), with geography and the appropriate benefit time applied separately. Earnings use the original otherwise-unmet clinical cohort, with eligibility for an earnings response defined before effective use. The one-year flow receives the model's existing delayed, discounted and mortality-adjusted effective-use integral once. A one-off payment saving or acquisition cost is instead adjusted at receipt, without an additional year-duration or wearing filter. These welfare equivalents are not clinical QALYs.

The $50 purchase amount, 5% purchase incidence, 10% earnings-responsive share, 5% earnings change, $6 acquisition burden and $50,000 reference income are sensitivity judgments, not PHC measurements. The earnings change is net disposable income after taxes, benefit substitution and other offsets, not gross productivity. PHC's $50-per-pair fundraising example is not evidence for a household's purchase price. Counterfactual purchasers sit outside the original 70% otherwise-unmet glasses group; their share plus the unmet share cannot exceed one. Do not subtract their health credit twice: the original clinical model already excludes the equivalent-alternative group.

For glasses, effective benefit time is .72810135 years after applying .75 use, a .1-year delay, 3% discount and .02 mortality hazard. The earnings case therefore credits 0.08 income-equivalent years, rather than the .11156684 from an undiscounted full-year illustration. The once-only purchase and acquisition cases are adjusted at .1-year receipt for glasses and .25-year receipt for hearing/dentures. The mortality term is prospective survival-to-receipt incidence, not a financial discount; it should not be applied again if future observations already condition completions on survival.

All health scenarios retain separate income components. No added funding or capacity creates no added clinical recipients and no income from those nonexistent recipients. Null health need not imply null cash, and adverse health and adverse income remain signed; nonpositive totals have no positive price. Zero individual clinical pathways, half utility and half duration are shown separately, not silently stacked or interpreted as confidence intervals.

Annual spending and resource boundary

A whole-PHC annual productivity estimate cannot be reconstructed from the available public records: project expenses and the corresponding annual completed cohort are missing. The sponsor's latest three returns and audits were rechecked October 1. They still describe all sponsored projects, not PHC. The historical $1,460,295 city-contract budget and the $19,976 PHC contributed-goods item are narrower records, not whole-project actual expense. The report's financial tables preserve those distinctions; no sponsor-sized PHC budget or three-year PHC average is fabricated.

The $140,000 gross associated-resource case remains distinct from donor cost; the independently reassessed central now gives $210K per DALY, compared with the historical $108K. It charges external resources per nominal episode, including replacement activity; it is not measured net induced social expenditure. Independent income cases do not transform those outside inputs into household income or public savings.

What changed from the prior assessment

The old glasses-only arithmetic was $100 / (.0375 × .75 × .5), or $7K per DALY. It assumed the money purchased completed glasses pairs under a program-conditional boundary. It did not represent ordinary PHC allocation or associated external resources. We preserve that input and calculation as history, not a competing current headline and not another health stream to add.

The prior project-budget model already corrected the main boundary problem before this review. We have not retuned it to keep PHC within a preferred ranking. New findings clarify the fiscal sponsor's three-year record, a narrow PHC in-kind allocation, a standard fee already included in cost, a dated insured optical pathway, and the limits of current volume and staffing claims. None supplies a defensible replacement for the central allocation, unit cost, or additional-capacity priors.

The city-contract budget and 2025 funding transition are the most material additions. They require correcting any claim that public project funding never existed, and any suggestion that no historical project budget is available. They also make validation of the current operating baseline more urgent. The earlier outputs are preserved as history to show exactly what the prior assumptions imply, not to certify that those assumptions describe the post-cut PHC portfolio.

One substantive evidence clarification is the interpretation of the glasses 41/113 diagnostic. It is not pure outcome attrition. Another is that no verified PHC-specific sponsorship rate remains different from no publicly available standard rate: the latter is now available. The dental baseline can be stated using the current DHCS FAQ's specified 2027 date rather than presenting all official dates as unresolved. These earlier improvements did not change the first recalibration; the latest independent clinical and income reassessment now changes the current price.

The latest assessment independently revises clinical benefits and separates net resource effects, while the local inputs remain uncertain. The model remains useful for identifying what would change the decision. Its apparent precision should not be mistaken for an estimate derived from audited PHC costs and a local causal evaluation.

Funding and counterfactual

There are two separate counterfactual questions. First, does the gift expand or preserve completed care beyond what PHC and its other funders would do? Second, among those additional PHC completions, who would otherwise obtain equivalent care during the modeled period? The model calls these financial additionality and alternative-free share. They must not be interpreted as two arbitrary discounts for the same missed appointment.

Financial additionality is centrally 50%. It is intended to represent replacement of other support and the share of nominal delivery actually changed by the gift, before applying explicit additional-capacity caps. The alternative-free share is centrally .7 for glasses and hearing and .6 for dentures. Despite the variable name, this means without an equivalent alternative, not people receiving free care. The remaining fraction receives no modeled benefit credit because its counterfactual care is treated as equivalent within the year.

The current public baseline is material. Medi-Cal benefits and clinic partners do not disappear in the no-gift world. The dated optical event already targets eligible members. The DHCS dental FAQ currently states July 1, 2027 for the specified existing-member loss of regular dental benefits, while new applicants and exceptions differ. We do not apply a universal 2026 dental-coverage collapse to PHC patients. Individual eligibility and realized access remain different questions.

The model's binary alternative treatment is coarse. A patient obtaining similar care after six months could have a meaningful temporary benefit from PHC, rather than either a full year or zero. The preserved calculation does not explicitly trace such catch-up timing. Its no-alternative fractions must be understood as effective attribution judgments for this simplified boundary. Future patient-level waiting-time data could replace the binary convention without pretending that formal coverage guarantees timely care.

The general donation page describes illustrative support amounts including $100 for two glasses pairs and $1,000 for oral care and dentures. It does not establish the number of additional completed episodes an unrestricted gift will cause. It also describes PHC programs as privately funded; this does not establish that its clinical partners use no public reimbursement, grants, donated labor, or existing clinic capacity.

More strongly, official historical records establish public funding of PHC itself. At the June 5, 2025 Homelessness Oversight Commission meeting, the director's presentation describes a proposed approximately $1.4 million PHC reduction and an intention to seek philanthropic support to sustain the organization. The adopted minutes separately record PHC's executive director reporting a 30-day notice terminating the entire contract effective June 30, 2025. Public comment is not an executed termination instrument; it should be attributed as such. The director's budget description independently confirms the proposed cut, not the exact eventual replacement financing.

This is more decision-relevant than a generic funding-room caveat. A new gift could be replacing lost city support, maintaining a smaller operation, financing a restart, or expanding services after other philanthropists already filled the gap. Those are different counterfactuals. It is not valid to take the old budget's full clinical infrastructure as free and guaranteed while also assuming the private gift has very high additionality because the same infrastructure lost funding. If substantial fixed costs must now be covered, the accepted per-completion costs and allocation may be optimistic even when need is urgent.

Equally, replacement of public money is not automatically zero donor impact. If city support truly ended and useful care would otherwise cease, philanthropic maintenance may be highly additional. The relevant displaced alternative is what patients and the project would actually do without the gift, not the historic public contract that no longer exists. We do not change the 50% prior in either direction from the cut alone; we disclose that current operating continuity and the funded baseline require revalidation.

The official May 2026 partnership evidence rules out assuming no activity simply because a 2025 source warned of closure. It does not resolve the scale or composition of the surviving project. A dated staffing roster, current clinical schedule confirmed by partners, and revenue ledger would distinguish these possibilities. Until then, the old scenario remains inspectable but is not a verified current offer or a reliable post-transition volume forecast.

The sponsor provides a further dated confirmation: its September 5, 2025 project update says PHC lost key funding and launched a Bridge Campaign to continue services. Thus the funding loss is not supported only by a proposed budget or public comment. The linked campaign currently leads to the general donation page, without a dated outstanding target, amount raised, funded staffing plan, or additional clinical tranche. This is a real historical fundraising response, but not verification that the same gap remains in September 2026. It also means the current private-funding statement could describe a changed funding mix; we reject applying it backward, not the possibility that current operations rely on private support.

The service-coordinator listing gives $28–30 per hour and describes documentation, navigation, and partner work. The page is undated. It is not proof of an open unfunded position, a fully loaded annual price, or a causal promise that another coordinator increases clinical completions. It is nevertheless more useful than a generic statement that donations help: it identifies a plausible labor input whose status and workload could be checked.

Waiting lists and lotteries are similarly informative but insufficient. A financial shortage, shortage of donated dentist hours, lack of eligibility, or limited ability to follow up can all produce a queue. The relevant marginal offer would connect a specified increment of money to a feasible additional service plan and identify the remaining limiting resource. No such current offer was verified in this review. Unknown room is not evidence of zero impact, but it is a reason not to market a conditional calculation as a purchasable result.