METADATA last updated: 2026-03-23 RT file_name: _context-commentary_pilots-pilot-sites.md category: pilots subcategory: pilot-sites gfile_url: https://docs.google.com/document/d/1UDeyVPHknvMcDDQ-5ZbaSZNiQEDpmh9eSGfizTYAFWA words: 5625 tokens: 7639 CONTENT ## Context This `pilots/pilot-sites` subcategory covers the FloodLAMP pilot programs as site-level implementations and case studies. Quantitative summaries, data-processing notes, and dataset-specific commentary are in `pilots/pilot-data`. ## Commentary #### FLMP - FloodLAMP (includes CRLN and FLSP) The FLMP designation exists because the Carillon (CRLN) and FLSP (FloodLAMP Staff Plus) programs shared a single tenant (instance of the application) in the FloodLAMP app software, and their testing runs were intermingled. Each tube was assigned to either Carillon or FLSP based on whether the participants belonged to or were associated with the Carillon preschool community. For the FloodLAMP founders' family (whose 2 young kids attended Carillon) and our household help, the assignment was date-based: during the dates the Carillon program was active, samples were attributed to Carillon; outside those dates, to FLSP. #### FLSP - FloodLAMP Staff Plus Other Community FLSP served as the catch-all program for FloodLAMP's internal staff testing plus friends, family, neighbors, co-workers, and other community members tested over the course of 2+ years of operations. The program start date was set at December 2020, when data collection began and the test had stabilized on the direct, no-purification protocol we used going forward. Testing was conducted from multiple locations: the main lab at MBC BioLabs in San Carlos, California; a home garage lab in Portola Valley; and later at the 2022 Portola Valley office space. In addition there were several pop-up sites where the mobile kit was used. The internal testing program produced several notable case studies. In multiple instances, routine screening unexpectedly detected positive cases among FloodLAMP staff, and the early detection appeared to contribute to preventing further spread within the company and within household contacts. ##### Case Study 1 — FloodLAMP Founder (December 2021) The founder (Randy True) contracted COVID-19 at a holiday gathering during the Omicron surge ramp-up. He was participating in every-other-day pooled screening with staff and tested negative in the morning; minor symptoms first appeared that evening. A pooled test run 2 days later, early in the morning, was positive. A rapid antigen test confirmed the result immediately. He was able to isolate before anyone else in the household became infected, and no one else in the household tested positive. While this case illustrated the value of early detection through routine screening, it also illustrates a limitation: in some cases, even with every-other-day testing, the more sensitive FloodLAMP test can catch the infection at the same point as a less sensitive rapid antigen test. See: `2021-12-28_Case Report - Randy` ##### Case Study 2 — New Year's Eve (December 31, 2021) This case presented a more striking comparison between FloodLAMP and other FDA authorized commercial tests that missed this infection. This is easiest to review visually. See: `2022-01-01_Case Report - New Years Eve` An individual nasal swab collected at 9 a.m. on December 31st tested positive on the FloodLAMP colorimetric LAMP test, and PCR run on the same TCEP-inactivated sample confirmed the result. However, subsequent testing that same afternoon using several different commercial tests including a rapid antigen test (BinaxNOW), Lucira, and Detect were all negative. A 5 p.m. resampling showed the infection presenting more strongly in a throat swab (Ct 30.2) than the nasal swab (Ct 39.4 on one of three PCR replicates, with the other two undetermined), suggesting low and variable viral load distribution. A saliva sample collected later that evening had a Ct of 28. The following day, New Year's Day, nasal and throat samples were positive on FloodLAMP and PCR but still negative on BinaxNOW antigen. An Accula rapid molecular test (PCR-based, 30-minute cartridge with lateral flow readout) administered at an airport testing site was also negative. It was not until the morning of January 2nd that two different antigen test brands (FlowFlex and BinaxNOW) finally returned positive results, confirming the FloodLAMP detection approximately 48 hours earlier. This case illustrates the variability across FDA-authorized testing modalities, a finding consistent with the FDA's own SARS-CoV-2 Reference Panel results, which showed unexpectedly large differences from the sensitivity figures reported in EUA submissions. See: - `regulatory/fda-policy/_AI_FDA SARS-CoV-2 Reference Panel Report` - `regulatory/fda-policy/FDA SARS-CoV-2 Reference Panel Comparative Data - Complied by Matt McFarlane` - `regulatory/fda-policy/2023-09-23_FDA Response Letter - To Compliant regarding FDA Reference Panel from Dec 2020` FloodLAMP developed its LAMP and PCR tests to run from the same TCEP-inactivated sample, as submitted to the FDA in March 2021. This dual-test capability proved particularly valuable for this case where multiple EUA authorized tests returned false negatives, and could be generally useful for cases requiring additional confirmation (see: `regulatory/fl-fda-submissions`; `various/fl-presentations/BARDA Market Research - FloodLAMP Presentation (2022-03-07).md`, slides 46–50). Over the full span of the FLSP program, staff testing detected four confirmed COVID-19 cases (the founder, one staff member, a domestic partner of a staff member, and the New Year's Eve case). In each instance, the case was caught early and no other FloodLAMP staff tested positive in the period immediately following. The program's primary value was as a longitudinal demonstration that routine molecular screening of a small, closely interacting group can detect infections early enough to stop workplace and even household transmission. #### CRLN - Carillon Pre-School in Portola Valley, CA The Carillon program is documented in detail in FloodLAMP's most comprehensive white paper (see: `various/fl-whitepapers/FloodLAMP Whitepaper - California Preschool Family Pooled Screening Pilot (June 2022)` and `various/fl-whitepapers/_context-commentary_various-fl-whitepapers`). Carillon was a small preschool in Portola Valley, California, attended by the children of FloodLAMP founders Randy True and Theresa Ling. The founders' familiarity with the school's staff facilitated the program launch. As the Omicron surge arrived in December 2021, previous discussions about a testing program turned into concrete plans, and the program was set to begin immediately after the holiday break. The close relationship with the school enabled program features that were likely rare, or possibly unique, among pandemic-era COVID-19 screening programs. Families were onboarded before the first day of testing with signed consent and pooled collection kits (five to ten per family). A dropbox at the school entrance served as both the sample drop-off and kit pickup location. Families registered pooled collections through the FloodLAMP app, identifying all participants including siblings and household members such as nannies or grandparents. The typical daily workflow was straightforward: families collected samples, swabbing themselves and kids, typically at home, the morning of school and then dropped them off at the same time as their children at the school. A FloodLAMP team member collected all kits from the dropbox within minutes of the morning drop-off and drove them home (approximately five minutes) for processing, with results available usually within 2 hours. No extra transportation or trips were required from participating families. A trade-off of this model was that children were already at school when results became available (excluding the first run in early January and after spring break). Follow-up testing and confirmatory information was received for some cases but not all positives. The program demonstrated its value immediately. Before the first day back from the holiday break, families dropped off samples the evening prior. Two pools tested positive that night. Both families were asymptomatic, and both positive results were subsequently confirmed. This was important news and the families were contacted and referred to follow-up diagnostic testing. The kids of those families of course did not come to school, and This early detection may have reduced onward exposure at the school, although this cannot be established from these data. This is a real-world case to understand the problems of the FDA/CMS policy with respect to "surveillance" testing and "individual decision making". See the subcategory `regulatory/surveillance` and the context and commentary file there. The same pre-screening approach was used successfully after spring break. These involved an extra trip to the school for drop-off but significantly increased the effectiveness of the program. Another documented case report describes one asymptomatic family pool detected on the first day back from the holiday break. Individual deconvolution the following day showed both parents positive and both children negative; the children turned antigen-positive the next day. All four family members were confirmed by PCR (results took one to two days). Return testing beginning a week later showed persistent FloodLAMP positivity: 15 days after the initial positive pool result, the family pool was still positive on FloodLAMP while all members tested negative on antigen (see: `2022-01-02_Carillon Case Report of Asymptomatic Positive Family Pool`). Another informative case study involved a teacher who had tested negative on antigen tests for three consecutive days following known household exposure, came to school believing they were clear and tested positive on FloodLAMP the same day. The result was confirmed by a lab PCR test. Either the timing of viral load ramp-up or the increased sensitivity of the molecular test accounted for the detection difference. Regardless, the screening program identified the case before further classroom exposure occurred. In total, the program identified eight positive pools or case clusters. Of the six with deconvolution or confirmatory referral test results, the initially positive individual was the school-aged child in half of cases and a parent or sibling in the other half. This reflects the value of extending coverage in a screening/surveillance program to the entire household. The Carillon program implemented several program features that FloodLAMP considered both valuable and rare: multi-week longitudinal testing, household pooling, fast few-hour turnaround, and deconvolution capability. Operating it within the founders' own school community provided an unusually close perspective on the day-to-day logistics and at the same time, the experience of a participating family. #### FTFC - Eagles/EMS Leadership Conference in Ft. Lauderdale, FL The FTFC conference in June 2021 was FloodLAMP's first significant external testing event and served as the entry point into the EMS and municipal surveillance pilots that followed (Coral Springs, Davie, and Bend). FloodLAMP staff flew from California with all lab materials in two suitcases and set up a pop-up lab in a storage room at the Hard Rock Hotel in Fort Lauderdale. A freezer and refrigerator were available on-site (possibly ordered via Amazon, though this detail is uncertain). A volunteer graduate student was trained to run the test, freeing FloodLAMP staff for sample collection and conference interaction. Testing was offered as a voluntary service and as a chance to learn about our program. We had a table in the hallway at the conference and a curtained tent for private sample collection. Participation was lower than anticipated. We agreed to and started preparing months before, but in June 2021 when the conference happened, prevalence had dropped and that mid-summer was a low point in pandemic prevalence. Many attendees viewed the pandemic as winding down, masks were rarely worn, and Florida maintained a relatively lax COVID-19 posture. FloodLAMP had prepared to run thousands of samples but processed only 61 tubes covering approximately 195 individuals over three days. There were no positive or inconclusive results. The testing was conducted in collaboration with two nonprofits: the Research Aid Network and the National Scientist Volunteer Database (NSVD). FloodLAMP gave a short invited talk to EMS leadership (see: `various/fl-presentations/FTFC EMS Conference - FloodLAMP Talk Slides (6-14-2021)`). The conference was where FloodLAMP trained Dr. Peter Antevy, who became the Program Medical Director for FloodLAMP and was our main contact in the EMS world. Dr. Antevy purchased the testing equipment (approximately $7K), which was subsequently shipped to the Kent Camp pilot. Despite the low testing volume, the FTFC conference yielded valuable EMS leadership contacts and practical experience deploying a pop-up laboratory, both of which directly contributed to FloodLAMP's subsequent pilot programs. #### KENT - Camp Kenmont Youth Camp in Kent, CT The Kent Camp pilot was FloodLAMP's first fully remote deployment, with no FloodLAMP staff traveling to the site. Dr. Peter Antevy trained three graduate student volunteers recruited through the National Scientist Volunteer Database (NSVD), a non-profit co-founded by FloodLAMP's COO. These volunteers performed all on-site testing. A positive COVID-19 case was identified among incoming campers through initial PCR testing. In the period between testing and arrival, other campers had been exposed, creating an immediate need for on-site surveillance screening capability before the camp session began. The FloodLAMP testing was used alongside antigen testing and was reported to be helpful. The known positive sample also served an unplanned validation role as a shipping issue (likely warming during transit) compromised the frozen positive control material that FloodLAMP normally provides, making the known positive case a useful substitute for test verification. Surveillance screening was conducted in two planned sessions, covering two different cohorts of campers. Samples were processed in a single 96-well reaction plate. Besides the known positive, no additional positives were detected. The Kent pilot demonstrated the feasibility of a remote bring-up model without FloodLAMP staff on-site. Its primary contribution was showing that fast-turnaround on-site molecular screening capability could be established remotely with trained volunteer support. #### COSP - Coral Springs City Municipal/EMS in Coral Springs, FL Coral Springs (population ~140,000) in south Florida was FloodLAMP's first major paid pilot and largest testing program by total volume. FloodLAMP staff traveled from California with equipment and consumables, set up a pop-up lab in an empty medical office, ran a successful validation, and trained one firefighter with no prior lab or pipetting experience. The program started in September 2021 with modest volumes: averaging 25 tubes (pools) and 90 participants for the first three months. When the Omicron surge hit in mid-December, the program expanded rapidly to comprehensively cover first responders and municipal staff. At peak, Coral Springs ran 800 tubes per week covering 2,600 participants (includes repeated testing of the same people), with multiple collection points across city buildings, police stations, and fire stations. Two staff members operated two inactivation and amplification stations ("double standard" configuration), running LAMP reactions in 96-well plates. During the December–January Omicron surge, the site ran three to four FloodLAMP processing runs per day to cover three emergency responder shifts, operating seven days a week. One significant limitation in the Coral Springs data is the absence of comparator testing data. They did do extensive antigen testing, including follow-up testing for all FloodLAMP positives, but FloodLAMP was unable to obtain this critical data. We let Coral Springs leadership know how important this data was and gave assurances of privacy protections (FloodLAMP staff had been through CITI training and certification to properly handle PII/PHI), but they did not share the antigen test data. In retrospect, de-identified data publication rights should have been included in the contract and we simply did not consider that they would not share it. The primary tester, however, did share the details of a particular case which was informative. He described a plate with 22 positive results, an unusually high number that he verified by rerunning the samples with identical results. Of those 22, only one was positive on same-day BinaxNOW rapid antigen testing, which was standard protocol to run after a FloodLAMP positive. The remaining 21 antigen-negative individuals returned to work, but two days later "everybody started popping positive," with most having previously tested positive by FloodLAMP. The conclusion after that episode was that BinaxNOW was significantly less sensitive than FloodLAMP for early detection. This observation was consistent with findings from other FloodLAMP pilot sites and is not unexpected given the known sensitivity gap between molecular and antigen testing, i.e. a performance gap at low viral loads. Testing ended abruptly in late February 2022 as the Omicron wave passed through the community. #### DAVI - Town of Davie Fire/EMS in Davie, FL The Town of Davie (population ~110,000) is located next to Coral Springs in south Florida. During the pandemic, the two towns shared the same Medical Director who connected both to FloodLAMP and whom we worked with at the FTFC EMS Leadership Conference (and who also ran the Kent pilot). The firefighter trained at Coral Springs subsequently trained the Davie tester, also a veteran firefighter, with supplementary early-morning support calls from FloodLAMP's founder. This single individual performed all testing at the Davie site. The program timeline paralleled Coral Springs: modest volumes averaging approximately 30 tubes and 80 participants per week from September through November 2021, followed by rapid expansion during the Omicron surge. Testing peaked at approximately 500 tubes per week in early January 2022, with positivity rates reaching 34% during the peak week. Feedback from the site regarding program effectiveness was consistently positive. One reported case illustrated the operational value of pooled screening with deconvolution capability: a leadership member's family pool tested positive, and the tester immediately resampled the family members to deconvolute the pool, running individuals on FloodLAMP to identify positives. The combination of confidence in test performance and operational flexibility was what the site valued most. Program success led to expansion in mid-November 2021, extending FloodLAMP testing from EMS personnel to all Town of Davie municipal employees on a voluntary daily or weekly basis, with six designated collection points and daily pickup by Fire Rescue staff. The Davie site produced the archive's most detailed head-to-head comparison between FloodLAMP and BinaxNOW rapid antigen testing, tracking 51 employee cases through three evolving testing protocols during the Delta and Omicron waves (see: `DAVI_pilot-data/DAVI_pilot-data_referral-antigen-comparison`): - **Group 1** (15 employees, Sep 15 – Dec 23): FloodLAMP as primary test; BinaxNOW used sometimes at home on Day 0. - **Group 2** (22 employees, Dec 24 – Jan 9): Both tests used throughout until return-to-work clearance, providing the most informative direct comparison data. - **Group 3** (14 employees, Jan 10 – Feb 7): FloodLAMP for initial detection; BinaxNOW for return-to-work clearance. Across cases with same-day results, FloodLAMP and BinaxNOW agreed 90% of the time. In the remaining 10% (four employees), FloodLAMP detected the infection while BinaxNOW did not; three of those four were asymptomatic. There were zero cases where antigen testing detected a positive that FloodLAMP missed. During the intensive dual-testing phase (Group 2), the two tests generally tracked closely and cleared on the same day, with an average time to negative of approximately 9 days. The divergence between the BinaxNOW antigen test and FloodLAMP was concentrated at infection onset, particularly in asymptomatic or low-viral-load individuals. This is the detection gap that molecular screening closes. The program reported no false positives or suspected false negatives in available follow-up data, with 98% of FloodLAMP positives confirmed by referral or correspondence. The Davie comparison data, combined with the anecdotal report from Coral Springs (21 of 22 FloodLAMP positives antigen-negative but subsequently confirmed), reinforced a consistent finding across the pilot programs: BinaxNOW rapid antigen tests lagged FloodLAMP molecular detection at infection onset, especially in asymptomatic individuals. See pilot COMB below where the comparison with antigen tests showed significantly more positive cases identified by FloodLAMP but missed by antigen. #### ROSA - ROSA TV Production in Davie, FL The ROSA program was FloodLAMP's only third-party or affiliated pilot, in which another organization ran the FloodLAMP test commercially. "ROSA" is a pseudonym; FloodLAMP does not have permission to disclose the production company name. All samples were run individually (no pooling), collected from TV show staff and actors on the production site, and shuttled to the Davie Fire and Rescue lab for processing. Processing was reportedly done promptly and separately from the Town of Davie staff samples. A dedicated COVID testing program manager at the production site administered the program, handling sample collection, app registration, and results management. FloodLAMP's involvement was limited primarily to digital and app support. The program ran consistently at 50 to 100 samples per week for 11 weeks. No positives were detected until the final week, when one positive was identified and confirmed with a same-day antigen test. FloodLAMP had minimal involvement with this program. #### BEND - Bend Fire and Rescue/EMS in Bend, OR The Bend pilot was a follow-on EMS program resulting from the success of Coral Springs and Davie. Although FloodLAMP received significant interest from other EMS programs nationally, Bend was the only additional EMS site brought online. Unlike Davie and Coral Springs, Bend tested only firefighters and EMS personnel, and all samples were run individually (no pooling). The primary motivation for the FloodLAMP program was dissatisfaction with the cost and turnaround time of PCR testing from a local lab. The bring-up process was conducted entirely remotely, without FloodLAMP staff traveling to Oregon. A local scientist volunteer, recruited through the National Scientist Volunteer Database (NSVD), provided critical on-site support. The program began with a one-hour Zoom kickoff call with fire and emergency management leadership, followed by remote training with the volunteer scientist providing hands-on assistance. The remote training was a bit awkward but proved effective. See the slides presented in the archive file: `various/fl-presentations/Bend Pilot Program Bring-up (12-01-2021)`. Of note is the discussion of contamination procedures and the mental models developed for EMS personnel to adopt (see end of `various/fl-presentations/_context-commentary_various-fl-presentations`). Positivity rates were high throughout January 2022 during the Omicron surge, ranging from 11% to 37%. All FloodLAMP positives were confirmed by the local clinical lab PCR, with no false positives documented in available follow-up data. The program recorded two inconclusive results, one of which was confirmed positive by PCR and the other negative by PCR. Bend purchased a "double mini" system: two small heaters accommodating up to 40 200μl PCR tubes each, run in strip tube format. These min heaters were also used for inactivation, holding 15 1.5mL tubes for the 5 minute heat time. This smaller system was feasible because individual (non-pooled) samples could be collected in 1.5 ml tubes (though these tubes could hold 2 swabs), thus avoiding the 5 ml tubes required for 4-swab pooling (which necessitated water bath inactivation at the Davie and Coral Springs sites). FloodLAMP charged $2,800 for equipment and $6 per reaction for an initial order of 1,000 reactions ($4 for the test, $2 for the collection kit). The initial outlay was $8,800. A subsequent order of 1,000 reactions and a multi-channel pipette brought the total to approximately $15,000 including shipping. The Bend program was a big success - the EMS manager who administered the program described it as a "godsend." #### COMB - Combate TV Show Production in Miami, FL The Combate program was FloodLAMP's only surveillance-screening-as-a-service operation, in which FloodLAMP employees directly performed all testing. An experienced nurse was hired and trained to run the FloodLAMP test. This person administered the program, collected samples, ran testing, and managed an assistant, doing an especially excellent job. The pop-up lab was established in a carpeted hotel room within the hotel that housed the TV program's staff and participants. This was not an ideal laboratory environment and raised initial contamination concerns. With rigorous cleaning protocols and attention to detail, the program operated for over four months without significant contamination problems. Testing was individual (no pooling), using 1.5 ml tubes, with approximately 100 people tested per week. The negotiated price was $35 per test all-in with a minimum of 150 tests per week ($5,250 per week). The initially discussed price was significantly higher but was negotiated downward over the months leading up to the program starting. Given staffing costs, including testing staff salaries and a prorated share of FloodLAMP scientist support time, it is likely that FloodLAMP did not recoup its costs on the contract. ##### Comparison with Antigen Tests (AI generated - Opus 4.6) The testing results were informative for doing a FloodLAMP vs. antigen comparison. The COMB data tells a dramatically different story from DAVI. The key numbers are right in the stats table: **COMB (Combate) -- FL positive vs. antigen comparison:** - 11 FloodLAMP positive/inconclusive cases with same-day antigen referral tests - **6 antigen positive (54.5%)** - **5 antigen negative (45.5%)** Nearly half of FloodLAMP positives were missed by antigen -- compared to DAVI's 10%. The referral-tests-by-person table at the bottom of the file lets us see each case individually: | Employee | FL Result | Antigen Result | Confirmed by | FL+/Ag outcome | |---|---|---|---|---| | COMB170110 | Positive | Positive | -- | Agree | | COMB213226 | Positive | Positive | -- | Agree | | COMB242905 | Positive | **Negative** | Lab PCR Positive | FL caught it | | COMB353066 | Positive | **Negative** | Lab PCR Positive | FL caught it | | COMB413047 | Positive | Positive | -- | Agree | | COMB596496 | Inconclusive | **Negative** | Lab PCR Positive | FL caught it | | COMB704360 | Positive | **Negative** | Lab PCR Positive | FL caught it | | COMB724402 | Inconclusive | Positive | -- | Agree | | COMB770137 | Positive | Positive | -- | Agree | | COMB945952 | Positive | **Negative** | Lab PCR Positive | FL caught it | | COMB638337 | ARF Positive | Positive | -- | Agree | All 5 antigen-negative cases were subsequently confirmed by lab-purified PCR. The COMB protocol was systematic: FL first, then antigen; if antigen disagreed, send for lab PCR. That protocol gives us high confidence in these numbers. **Why so much higher than DAVI?** A few likely factors: 1. **Testing context.** COMB was HCW-collected individual surveillance of a production crew -- likely catching people very early through routine screening before symptoms developed. DAVI had more symptomatic diagnostic testing mixed in (especially Groups 2 and 3, where many employees were already symptomatic at Day 0). 2. **Low positivity rate.** COMB's overall tube positivity was just 0.7% (14 positives out of 1,981 tubes). At low prevalence with routine screening, the cases you do find tend to be earlier in infection with lower viral loads -- exactly where antigen underperforms. 3. **Time period.** COMB ran March--August 2022, a lower-prevalence period compared to DAVI's operation during the Omicron surge. Finding cases during low-prevalence periods often means catching them at the earliest detectable stage. **Side by side:** | Site | Same-day comparisons | Antigen + | Antigen - | % Antigen Negative | |---|---|---|---|---| | DAVI | 42 | 38 | 4 | 10% | | COMB | 11 | 6 | 5 | **45.5%** | The COMB data is a much smaller sample (11 vs. 42), but the signal is strong and directionally consistent with the Coral Springs anecdote (21/22 antigen-negative). Taken together, these suggest that in surveillance contexts catching early/asymptomatic infections, the antigen miss rate can be very high -- far above DAVI's 10%, which was skewed by a large proportion of symptomatic cases who were easier for antigen to detect. #### NDHM - Needham Beth Shalom School in Needham, MA The Needham pilot was a small, late-stage program established through a connection via New England BioLabs, and provided free of charge. FloodLAMP's founder traveled to set up the lab and train the tester, a college student studying a bio-related field with some prior hands-on lab experience. An office room in the school was dedicated to testing. By this point in FloodLAMP's operations, the setup process had matured and was straightforward, using the standardized four-drawer cart configuration. The program also used a recently implemented streamlined web onboarding process. Participation was limited as prevalence was low in that area and at that point in the pandemic. Over the first four to six weeks, the program processed a few dozen samples per week, tapering off substantially afterward. The total volume was low: 80 tubes covering 130 unique individuals. This occurred during a period of declining case rates, which likely contributed to the drop-off in interest and participation. No positive samples were detected during the program. #### ABRM - Abrome K-12 School in Austin, TX Abrome was a small alternative K-12 school in Austin, Texas. The school connected FloodLAMP to the Balvi grant organization and submitted for a Balvi grant that supported their system purchase. Abrome operated the program for nine months (September 5, 2022 through June 2, 2023), running 1,379 initial tubes covering 2,954 participant results across 87 unique individuals. Testing was pooled by household with an average pool level of 2.2 individuals per tube, processed on-site using a mini-amp heater plus water bath configuration. The program detected eight positive tubes corresponding to three distinct cases, with no false positives or suspected false negatives documented in available follow-up data. The program's operational consistency was notable: testing was conducted on 95% of weekdays during 36 active school weeks (171 of 180 weekdays, with an effective uptime of 97% adjusting for holidays). This frequency of testing was unusual outside of significantly higher-budget screening programs. FloodLAMP charged approximately $3,000 for equipment and system, plus $4 to $5 per test kit (reagents, consumables, collection tubes, and swabs). Abrome purchased 3,000 surveillance test kits, bringing the total program cost to approximately $20,000, including roughly $1,500 for in-person training and setup. The testing and administrative lead from the Combate program flew to Texas to set up the lab and conduct in-person training. The head of Abrome provided a written endorsement stating that the school emerged from 3.5 academic years of the pandemic without a known case of in-school transmission, crediting the FloodLAMP program with catching three instances of household infection before the student became infectious at school. A documented case from December 2022 illustrates the detection model: a morning LAMP test flagged a positive household pool (two parents and one student), none of whom had symptoms. The student was immediately separated from the school. Within hours, one parent tested positive on rapid PCR; the student developed symptoms and tested positive two days later. No other Abrome students or staff tested positive, suggesting that the rapid detection and quarantine interrupted transmission to the school community. Abrome's success was supported by conditions that may not be generalizable: a school culture that embraced mandatory daily testing, a small enough community for a single staff member to manage the program, and grant funding covering the full cost. These factors made Abrome a great setting for our final FloodLAMP pilot, but also highlight the financial and participation challenges for such a comprehensive screening program. #### False Positives and False Negatives Across all pilot programs, no session-level false positives or false negatives were documented in available follow-up data. These zeros reflect the structure of the surveillance programs rather than a claim of perfect test performance. For false negatives - defined here as a FloodLAMP negative on a day when a referral test for the same individual had a positive referral test (non-FloodLAMP test) — the number is zero primarily because referral testing was almost always triggered by a positive FloodLAMP result, not collected in parallel. Without simultaneous higher-sensitivity comparator samples, false negatives in a screening context are effectively undetectable. Programs screened the same individuals multiple times per week, so a person will always have a preceding negative test before turning positive, but the viral load trajectory from onset of infection is unknown without paired comparator data. Test-to-return testing of known positives never produced a FloodLAMP negative for a known positive case until the tail-end of their infections (at least as far as the data we have). The Stanford clinical evaluation conducted for the FDA EUA submission (see: `regulatory/fl-fda-submissions/2021-03-22_EUA Submission - FloodLAMP QuickColor COVID-19 Test v1.0.md`) did measure false negatives under controlled conditions and reported 90% positive agreement (4 of 40 positives missed, all at high Ct values indicating low viral load) with 100% negative agreement, at an established limit of detection of 12,500 copies/mL. For false positives — defined here as a FloodLAMP session call of positive for a truly negative individual — the number is zero, but with important context. LAMP reactions are inherently prone to false positives at the individual reaction level (due to the nature of isothermal amplification). This necessitated a robust rerun workflow that resolved many reaction-level false positives into negative or inconclusive final session calls (see: `guides/test-site/Resulting Decision Chart v1.2`; `guides/test-site/Resulting Decision Flow Chart (Single Triplicate Re-run) v1.2`). At the session call level, false positives would be identifiable in the surveillance structure because the referral test for a FloodLAMP positive would come back negative, and program administrators would likely have reported such cases to FloodLAMP. For all FloodLAMP positive tubes with available referral data, every referral test was positive, with the caveat that some FloodLAMP positives lack follow-up data. The closest case to a false positive was a "false inconclusive" in the BEND program: of two inconclusive session calls on separate days, one was confirmed positive and the other confirmed negative by referral PCR, though this is technically distinct from a false positive since the session call was inconclusive rather than positive. In summary, the zeros are consistent with expectations given the program structures but should not be interpreted as evidence of zero false positive or false negative rates as those metrics are normally understood in the testing space. A controlled study with simultaneous comparator testing would be required to establish those rates, and the Stanford clinical evaluation provides the closest available measure of analytical performance under such conditions. #### Organizational Self-Testing as a Pandemic Necessity A fundamental argument for decentralized screening emerges from the pilot experience. During the acute phase of a pandemic, infections rise exponentially, and neither centralized testing services, commercial testing companies, nor device manufacturers can scale fast enough to match that growth. The January 2022 Omicron surge produced the worst testing shortage of the entire pandemic despite two years of investment. One way to address these limitations is to enable organizations to use their own staff and their own resources to test their populations, without depending on external testing infrastructure that cannot keep pace with exponential spread. This is what FloodLAMP's pilot sites did: firefighters with no laboratory background ran the test, expanded coverage during the surge, and validated the test's performance against individuals already known to be positive through clinical testing. Organizations running screening programs are inherently positioned to evaluate whether the test is working. Their administrators know who is sick, see whether the screening catches those cases, and are directly accountable for the program's effectiveness. The regulatory framework should support this modality (see `regulatory/surveillance` and the reform proposals in the paper).