GiveBetter x SF

Harm Reduction Therapy Center

Modeled alcohol-focused course within Adult Mobile Behavioral Health

Research time: ~15 min on GPT-5.6 Sol Medium
  • Research — organization and evidence review.
  • Modeling — cost-effectiveness analysis.
  • Historical estimate for research done before time tracking.

Published: 7 September 2026.

Donate

Summary

What do they do? HRTC provides mobile and low-threshold therapy. We model offering five alcohol-focused sessions over 12 weeks to adults experiencing homelessness with alcohol-use disorder, using external HaRT-A evidence. This is an analytic proposal within the adult mobile program, not a claim about the dose or protocol currently delivered. More

Why this approach interests us

  • Low-threshold behavioral care can reach adults whom conventional abstinence-first services do not engage.

Our main reservations

  • The evidence concerns a specific external alcohol intervention. The score-to-utility mapping has not been validated for change scores.

What do you get for your dollar?

Our exploratory best guess is $5.55M per better life (10 QALYs). This is a proposed alcohol-focused slice of Adult Mobile Behavioral Health, not a measured HRTC result or evidence that HRTC uses the exact trial protocol. We would not prioritize it on the current health-only estimate. Inspect the model and assumptions →

our best guess
$5.55Mper 10 incremental QALYs; subjective mapped-score model
positive scenarios
$493K–$444MNot confidence bounds; null benefit remains possible
cost per offer
$1,500Modeled five-session course plus engagement time
funding room
UnverifiedExisting public contracts are not donation capacity

1. What do they do?

HRTC provides mobile and low-threshold therapy. We model offering five alcohol-focused sessions over 12 weeks to adults experiencing homelessness with alcohol-use disorder, using external HaRT-A evidence. This is an analytic proposal within the adult mobile program, not a claim about the dose or protocol currently delivered.

Offer low-threshold care

Engagement and missed-appointment effort are included in the assumed course cost.

Deliver behavioral sessions

The selected external comparison is HaRT-A alone, excluding injection and medication arms.

Credit incremental health time

Allow usual care to catch up; do not turn persistent within-person improvement into a permanent treatment advantage.

Scope of this review. The city budget lists 4,530 staff hours at $198.64 and 500 planned clients for the adult mobile program. Budget targets are not executed costs, completed courses or causal outcomes; the separate transition-age-youth contract is excluded.

2. Monitoring and information sharing

Adults experiencing homelessness with alcohol-use disorder

Randomized behavioral-treatment comparison. HaRT-A alone improved physical SF-12 quality of life (d=0.41); no demonstrated mental-quality-of-life effect. Later active-group outcomes plateaued as usual care improved.

Our assessment. Unblinded treatment, missingness and protocol differences remain. Four physical points is a scenario, not d multiplied by a known trial standard deviation.

Adults experiencing homelessness with alcohol-use disorder

Earlier four-session randomized trial. Quality-of-life differences were inconclusive, despite improvement in alcohol outcomes.

Our assessment. Contrary evidence informs the transfer discount; alcohol outcomes are not added as separate QALYs.

SF-12/SF-36 population datasets

Group-average utility prediction. Hanmer supplies a preference-based SF-6D mapping coefficient for physical scores.

Our assessment. It does not establish the validity of mapping intervention-induced score changes.

3. Qualitative assessment

Low-threshold behavioral care can reach adults whom conventional abstinence-first services do not engage.

Key reservations

  • Current HRTC course completion and preference-based utility outcomes are missing.
  • The mapping and benefit-duration assumptions dominate a small modeled health gain.
  • Even the optimistic positive scenario here exceeds $100K per 10 QALYs.
  • This is not a model of medication, naloxone, overdose mortality or all HRTC services.

Benefits not included in our estimate

  • Medication and naloxone effects
  • Overdose mortality or other survival gains
  • Separate alcohol, depression, employment or housing benefits overlapping the mapped utility
  • Broader youth and training programs
  • Public-payer savings

4. What do you get for your dollar?

Our best estimate: $5.55M per 10 QALYs.

Four physical-component points is an analyst scenario, not a recovered trial endpoint. Multiplying by 0.00781 gives a mapped peak utility change; we halve it for transfer and mapping uncertainty, integrate 18 full-effect-equivalent weeks, and halve again for funding additionality.

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

DOLLARS PER BETTER LIFE: 10 × $1,500 ÷ (4 × 0.00781 × 50% × 18/52 × 50%)
≈ $5.55M per 10 QALYs

Model inputs and assumptions
Donor cost per offered course
1500 (range: 1000 / 1500 / 2500). Five one-hour sessions plus 2.5 hours of engagement, missed appointments and coordination at about $200/hour gives $1,500. Hours and rounding are assumptions. The published $198.64 is a budgeted service-resource benchmark; do not add a second blanket overhead allowance. Explicit analyst assumption.
Peak incremental SF-12 physical score
4 (range: 6 / 4 / 1). Four points is an analyst scenario, compatible in scale with a small-to-medium effect but not recovered from d=0.41. A normative standard deviation is not necessarily the trial's effect-size denominator. Explicit analyst assumption.
Mapping and delivery transfer
0.5 (range: 1 / 0.5 / 0.25). 50% jointly discounts protocol, retention/delivery and population differences, unvalidated change-score mapping and the earlier inconclusive quality-of-life trial. Not a measured causal adjustment. Explicit analyst assumption.
Full-effect-equivalent weeks
18 (range: 30 / 18 / 6). 18 assumes a linear rise over 12 weeks and decline to zero over 24: 12/2 + 24/2. It credits incremental area under the curve, not persistent within-person improvement when usual care catches up. Explicit analyst assumption.
Funding additionality
0.5 (range: 0.75 / 0.5 / 0.25). 50% allows replacement of existing public/private financing or alternative treatment. The existing contract is not available philanthropic capacity. Explicit analyst assumption.

Illustrative $100,000: 66.7 offered courses and 0.180 QALYs

  • Optimistic positive: $493.19K per 10 QALYs. 2.028 QALYs per $100,000; conditional, not available funding room
  • Central analyst estimate: $5.55M per 10 QALYs. 0.180 QALYs per $100,000; conditional, not available funding room
  • Pessimistic positive: $443.88M per 10 QALYs. 0.002 QALYs per $100,000; conditional, not available funding room

Uncertainty. No incremental utility improvement or no additional funded treatment means no finite price. Positive scenarios are not confidence intervals; harms are possible and would require a signed net-benefit model.

A health-score bridge, not measured QALYs

Hanmer's group-average SF-12 to SF-6D equation includes 0.00781 per physical component point. Applying it to changes assumes unchanged demographics and no mental-component change. The source did not validate change-score prediction. No QALYs were measured in the selected HaRT-A trial.

SUBJECTIVE INCREMENTAL TRAJECTORY: 12-week linear rise / 2 + 24-week linear decline / 2
18 full-effect-equivalent weeks, not 36 weeks at peak

QALY conversion assumptions

    The analysis is per course offered. Trial noncompletion is already embedded in treatment-assignment evidence; do not multiply by a completion rate again. The joint transfer factor allows different local delivery without pretending to measure it.

    5. Funding and previous grants

    An additional course, local health trajectory and incremental private-funding plan are not verified.

    The adult mobile contract extends through June 2029 in the 2024 report. A new private gift may substitute for that funding. The budgeted all-in staff rate is only a cost anchor; current incremental staffing capacity must be established.

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

    Donate

    Opens the organization’s giving page. A general donation may not fund the specific activity modeled here; confirm allocation with the recipient.

    6. Sources

    1. November 2024 contract report, printed page 5. SF Health Commission. Adult Mobile Behavioral Health budgeted hours and clients; not marginal course cost. Published: 2024-11-05; retrieved: 2026-09-07.
    2. Mission Mobile team and Gubbio. Harm Reduction Therapy Center. Local delivery description, not causal outcomes. Published: 2025-05; retrieved: 2026-09-07.
    3. Combining extended-release naltrexone and behavioral harm-reduction treatment: randomized clinical trial. Collins et al.. HaRT-A alone comparison only; physical quality of life, not measured QALYs. Published: 2021; retrieved: 2026-09-07.
    4. Randomized controlled trial of harm reduction treatment for alcohol. Collins et al.. Earlier four-session trial; inconclusive quality-of-life effect. Published: 2019; retrieved: 2026-09-07.
    5. Predicting an SF-6D Preference-Based Score Using MCS and PCS Scores from the SF-12 or SF-36. Hanmer, Value in Health. Group-average mapping; change-score validity not established. Published: 2009; retrieved: 2026-09-07.

    Annual expenses: years and sources

    Average annual expenses (three consecutive fiscal years): $2,742,777. Organization size is separate from the modeled cost-effectiveness of a donation.

    Harm Reduction Therapy Center

    Whole legal entity Form 990 Part IX total functional expenses; latest original reconstructed filings checked, fiscal-end year from printed reporting dates. Older fallback rows from IRS extracted API.

    Latest original filings checked; ProPublica extracted API lags these original returns.