BACKGROUND: Cooling towers containing Legionella spp are a high-risk source of Legionnaires\' disease outbreaks. Manually locating cooling towers from aerial imagery during outbreak investigations requires expertise, is labour intensive, and can be prone to errors. We aimed to train a deep learning computer vision model to automatically detect cooling towers that are aerially visible.
METHODS: Between Jan 1 and 31, 2021, we extracted satellite view images of Philadelphia (PN, USA) and New York state (NY, USA) from Google Maps and annotated cooling towers to create training datasets. We augmented training data with synthetic data and model-assisted labelling of additional cities. Using 2051 images containing 7292 cooling towers, we trained a two-stage model using YOLOv5, a model that detects objects in images, and EfficientNet-b5, a model that classifies images. We assessed the primary outcomes of sensitivity and positive predictive value (PPV) of the model against manual labelling on test datasets of 548 images, including from two cities not seen in training (Boston [MA, USA] and Athens [GA, USA]). We compared the search speed of the model with that of manual searching by four epidemiologists.
RESULTS: The model identified visible cooling towers with 95·1% sensitivity (95% CI 94·0-96·1) and a PPV of 90·1% (95% CI 90·0-90·2) in New York City and Philadelphia. In Boston, sensitivity was 91·6% (89·2-93·7) and PPV was 80·8% (80·5-81·2). In Athens, sensitivity was 86·9% (75·8-94·2) and PPV was 85·5% (84·2-86·7). For an area of New York City encompassing 45 blocks (0·26 square miles), the model searched more than 600 times faster (7·6 s; 351 potential cooling towers identified) than did human investigators (mean 83·75 min [SD 29·5]; mean 310·8 cooling towers [42·2]).
CONCLUSIONS: The model could be used to accelerate investigation and source control during outbreaks of Legionnaires\' disease through the identification of cooling towers from aerial imagery, potentially preventing additional disease spread. The model has already been used by public health teams for outbreak investigations and to initialise cooling tower registries, which are considered best practice for preventing and responding to outbreaks of Legionnaires\' disease.
BACKGROUND: None.

UNASSIGNED: Point-of-sale food messaging can encourage healthier purchases, but no studies have directly compared multiple interventions in the field.
UNASSIGNED: To examine which of 4 food and beverage messages would increase healthier vending machine purchases.
UNASSIGNED: This randomized trial assessed 13 months (February 1, 2019, to February 29, 2020) of vending sales data from 267 machines and 1065 customer purchase assessments from vending machines on government property in Philadelphia, Pennsylvania. Data analysis was performed from March 5, 2020, to November 8, 2022.
UNASSIGNED: Study interventions were 4 food and beverage messaging systems: (1) beverage tax posters encouraging healthy choices because of the Philadelphia tax on sweetened drinks; (2) green labels for healthy products; (3) traffic light labels: green (healthy), yellow (moderately healthy), or red (unhealthy); or (4) physical activity equivalent labels (minutes of activity to metabolize product calories).
UNASSIGNED: Sales data were analyzed separately for beverages and snacks. The main outcomes analyzed at the transaction level were calories sold and the health status (using traffic light criteria) of each item sold. Additional outcomes were analyzed at the monthly machine level: total units sold, calories sold, and units of each health status sold. The customer purchase assessment outcome was calories purchased per vending trip.
UNASSIGNED: Monthly sales data came from 150 beverage and 117 snack vending machines, whereas 1065 customers (558 [52%] male) contributed purchase assessment data. Traffic light labels led to a 30% decrease in the mean monthly number of unhealthy beverages sold (mean ratio [MR], 0.70; 95% CI, 0.55-0.88) compared with beverage tax posters. Physical activity labels led to a 34% (MR, 0.66; 95% CI, 0.51-0.87) reduction in the number of unhealthy beverages sold at the machine level and 35% (MR, 0.65; 95% CI, 0.50-0.86) reduction in mean calories sold. Traffic light labels also led to a 30-calorie reduction (b = -30.46; 95% CI, -49.36 to -11.56) per customer trip in the customer purchase analyses compared to physical activity labels. There were very few significant differences for snack machines.
UNASSIGNED: In this 13-month randomized trial of 267 vending machines, the traffic light and physical activity labels encouraged healthier beverage purchases, but no change in snack sales, compared with a beverage tax poster. Corporations and governments should consider such labeling approaches to promote healthier beverage choices.
UNASSIGNED: ClinicalTrials.gov Identifier: NCT06260176.

UNASSIGNED: Postdischarge outreach from the primary care practice is an important component of transitional care support. The most common method of contact is via telephone call, but calls are labor intensive and therefore limited in scope.
UNASSIGNED: To test whether a 30-day automated texting program to support primary care patients after hospital discharge reduces acute care revisits.
UNASSIGNED: A 2-arm randomized clinical trial was conducted from March 29, 2022, through January 5, 2023, at 30 primary care practices within a single academic health system in Philadelphia, Pennsylvania. Patients were followed up for 60 days after discharge. Investigators were blinded to assignment, but patients and practice staff were not. Participants included established patients of the study practices who were aged 18 years or older, discharged from an acute care hospitalization, and considered medium to high risk for adverse health events by a health system risk score. All analyses were conducted using an intention-to-treat approach.
UNASSIGNED: Patients in the intervention group received automated check-in text messages from their primary care practice on a tapering schedule for 30 days following discharge. Any needs identified by the automated messaging platform were escalated to practice staff for follow-up via an electronic medical record inbox. Patients in the control group received a standard transitional care management telephone call from their practice within 2 business days of discharge.
UNASSIGNED: The primary study outcome was any acute care revisit (readmission or emergency department visit) within 30 days of discharge.
UNASSIGNED: Of the 4736 participants, 2824 (59.6%) were female; the mean (SD) age was 65.4 (16.5) years. The mean (SD) length of index hospital stay was 5.5 (7.9) days. A total of 2352 patients were randomized to the intervention arm and 2384 were randomized to the control arm. There were 557 (23.4%) acute care revisits in the control group and 561 (23.9%) in the intervention group within 30 days of discharge (risk ratio, 1.02; 95% CI, 0.92-1.13). Among the patients in the intervention arm, 79.5% answered at least 1 message and 41.9% had at least 1 need identified.
UNASSIGNED: In this randomized clinical trial of a 30-day postdischarge automated texting program, there was no significant reduction in acute care revisits.
UNASSIGNED: ClinicalTrials.gov Identifier: NCT05245773.

BACKGROUND: Web-based surveys increase access to study participation and improve opportunities to reach diverse populations. However, web-based surveys are vulnerable to data quality threats, including fraudulent entries from automated bots and duplicative submissions. Widely used proprietary tools to identify fraud offer little transparency about the methods used, effectiveness, or representativeness of resulting data sets. Robust, reproducible, and context-specific methods of accurately detecting fraudulent responses are needed to ensure integrity and maximize the value of web-based survey research.
OBJECTIVE: This study aims to describe a multilayered fraud detection system implemented in a large web-based survey about COVID-19 attitudes, beliefs, and behaviors; examine the agreement between this fraud detection system and a proprietary fraud detection system; and compare the resulting study samples from each of the 2 fraud detection methods.
METHODS: The PhillyCEAL Common Survey is a cross-sectional web-based survey that remotely enrolled residents ages 13 years and older to assess how the COVID-19 pandemic impacted individuals, neighborhoods, and communities in Philadelphia, Pennsylvania. Two fraud detection methods are described and compared: (1) a multilayer fraud detection strategy developed by the research team that combined automated validation of response data and real-time verification of study entries by study personnel and (2) the proprietary fraud detection system used by the Qualtrics (Qualtrics) survey platform. Descriptive statistics were computed for the full sample and for responses classified as valid by 2 different fraud detection methods, and classification tables were created to assess agreement between the methods. The impact of fraud detection methods on the distribution of vaccine confidence by racial or ethnic group was assessed.
RESULTS: Of 7950 completed surveys, our multilayer fraud detection system identified 3228 (40.60%) cases as valid, while the Qualtrics fraud detection system identified 4389 (55.21%) cases as valid. The 2 methods showed only \"fair\" or \"minimal\" agreement in their classifications (κ=0.25; 95% CI 0.23-0.27). The choice of fraud detection method impacted the distribution of vaccine confidence by racial or ethnic group.
CONCLUSIONS: The selection of a fraud detection method can affect the study\'s sample composition. The findings of this study, while not conclusive, suggest that a multilayered approach to fraud detection that includes conservative use of automated fraud detection and integration of human review of entries tailored to the study\'s specific context and its participants may be warranted for future survey research.

To end the HIV epidemic in Philadelphia, implementation of evidence-based practices (EBP) to increase viral suppression and retention in HIV care is critical. Managed problem solving (MAPS), an EBP for antiretroviral therapy adherence, follows a problem-solving approach to empower people living with HIV (PWH) to manage their health. To overcome barriers to care experienced by PWH in Philadelphia, the EBP was adapted to include a focus on care retention and delivery by community health workers (CHWs). The adapted intervention is MAPS+. To maximise the clinical impact and reach of the intervention, evaluation of the effectiveness and implementation of MAPS+ is necessary.
This manuscript describes the protocol for a stepped-wedge cluster-randomised type 2 hybrid effectiveness-implementation trial in 10 clinics in Philadelphia. This research incorporates innovative approaches to accomplish three objectives: (1) to evaluate the effectiveness of the CHW-led MAPS+ intervention to improve viral suppression and retention in care 1 year after the individual implementation period (N=390 participants), (2) to examine the effect of the menu of implementation strategies on reach and implementation cost and (3) to examine processes, mechanisms, and sustainment of the implementation strategies for MAPS+ (N=56 participants). Due to various factors (eg, COVID-19), protocol modifications have occurred.
The institutional review board (IRB) at the city of Philadelphia serves as the primary IRB; initial approval was granted on 21 December 2020. The University of Pennsylvania and Northwestern University executed reliance agreements. A safety monitoring committee comprised experts in implementation science, biostatistics and infectious diseases oversee this study. This research will offer insights into achieving the goals to end the HIV epidemic in Philadelphia as well as implementation efforts of MAPS+ and other behavioural interventions aimed at increasing medication adherence and retention in care. Dissemination will include deliverables (eg, peer-reviewed manuscripts and lay publications) to reach multiple constituents.
NCT04560621.

A feature of HIV cure trials is the need to interrupt treatment to test the efficacy of experimental interventions-a process known as analytical treatment interruptions (ATIs).
We report the experiences of participants after they completed an extended ATI.
From April to November 2022, we conducted post-ATI in-depth interviews with BEAT2 clinical trial (NCT03588715) participants who stopped ART while receiving an immunotherapy regimen. We used conventional content analysis to code the data.
We conducted interviews with 11 Black/African American and three White/Caucasian participants (11 males, two females, and one transgender woman). The mean ATI was 38 weeks. Participants noted several significant experiences surrounding the interventions\' side effects, ATI, and returning to medication. Some participants had positive experiences with their ATI. Other participants were nervous during the ATI. Rising viral loads led some to feel a sense of failure. Although trial experiences were heterogeneous, participants unanimously had positive interactions with the clinical trial staff which facilitated their retention in the trial. Participants shared their experiences with the trial, including changes in expectations, experiences with experimental interventions and procedures, compensation as a measure of respect, effort, transportation, and effects of COVID-19 during the trial. Based on these results, we provide considerations for the conduct of future HIV cure-directed clinical trials involving ATIs.
Managing expectations, focusing on participants\' contributions, and providing support to reduce feelings of having failed the research team and/or the HIV community following viral rebound should be part of HIV cure trial design. Discussing the mental health impact of rebound during consent, distinct from risk, is needed. Continued efforts to understand how people with HIV experience ATIs will improve future designs of HIV cure clinical trials.

BACKGROUND: Gun buyback programmes have been popular in the USA since the 1970s. Studies show that they have no effect on citywide gun crime rates, but more microlevel examinations around gun buyback locations have not been conducted. This study tests for local effects of 34 Philadelphia, PA buyback events at 30 locations between 2019 and 2021.
METHODS: We analysed all gun-related crime events and gun-related calls for service attended by the police from 2019 to 2021. Multilevel models with an autoregressive residual structure were estimated on weekly gun crime and call event intensity (inverse distance weighted) totals across a range of distances (4000-8000 feet). Impacts of a gun buyback event were estimated for 1-4 weeks postevent.
RESULTS: Statistically significant weekly increases in gun event intensity are associated with seasonality and after the murder of George Floyd. Gun event intensity was not significantly affected by gun buybacks. Across 20 sensitivity tests of different distances and time periods (4000-8000 feet and between 1 and 4 weeks), gun buybacks were not statistically associated with any localised reduction in the intensity of gun crimes and calls.
CONCLUSIONS: Extant research has failed to uncover any effect of gun buybacks on citywide gun crime rates. The current results now contribute a lack of evidence at the local level to this literature. While gun buybacks remain popular with politicians and the public, this study adds to the ongoing question of whether buyback funds could be better spent more effectively.

Guidelines recommend shared decision-making prior to initiating lung cancer screening (LCS). However, evidence is lacking on how to best implement shared decision-making in clinical practice.
To evaluate the impact of an LCS Decision Tool (LCSDecTool) on the quality of decision-making and LCS uptake.
This randomized clinical trial enrolled participants at Veteran Affairs Medical Centers in Philadelphia, Pennsylvania; Milwaukee, Wisconsin; and West Haven, Connecticut, from March 18, 2019, to September 29, 2021, with follow-up through July 18, 2022. Individuals aged 55 to 80 years with a smoking history of at least 30 pack-years who were current smokers or had quit within the past 15 years were eligible to participate. Individuals with LCS within 15 months were excluded. Of 1047 individuals who were sent a recruitment letter or had referred themselves, 140 were enrolled.
A web-based patient- and clinician-facing LCS decision support tool vs an attention control intervention.
The primary outcome was decisional conflict at 1 month. Secondary outcomes included decisional conflict immediately after intervention and 3 months after intervention, knowledge, decisional regret, and anxiety immediately after intervention and 1 and 3 months after intervention and LCS by 6 months.
Of 140 enrolled participants (median age, 64.0 [IQR, 61.0-69.0] years), 129 (92.1%) were men and 11 (7.9%) were women. Of 137 participants with data available, 75 (53.6%) were African American or Black and 62 (44.3%) were White; 4 participants (2.9%) also reported Hispanic or Latino ethnicity. Mean decisional conflict score at 1 month did not differ between the LCSDecTool and control groups (25.7 [95% CI, 21.4-30.1] vs 29.9 [95% CI, 25.6-34.2], respectively; P = .18). Mean LCS knowledge score was greater in the LCSDecTool group immediately after intervention (7.0 [95% CI, 6.3-7.7] vs 4.9 [95% CI, 4.3-5.5]; P < .001) and remained higher at 1 month (6.3 [95% CI, 5.7-6.8] vs 5.2 [95% CI, 4.5-5.8]; P = .03) and 3 months (6.2 [95% CI, 5.6-6.8] vs 5.1 [95% CI, 4.4-5.8]; P = .01). Uptake of LCS was greater in the LCSDecTool group at 6 months (26 of 69 [37.7%] vs 15 of 71 [21.1%]; P = .04).
In this randomized clinical trial of an LCSDecTool compared with attention control, no effect on decisional conflict occurred at 1 month. The LCSDecTool used in the primary care setting did not yield a significant difference in decisional conflict. The intervention led to greater knowledge and LCS uptake. These findings can inform future implementation strategies and research in LCS shared decision-making.
ClinicalTrials.gov Identifier: NCT02899754.

Over the past decade, the prevalence of food insecurity declined in the United States but curiously climbed in Philadelphia, Pennsylvania, a sizable metropolitan area where many households experience food insecurity and are dependent on programs like SNAP. Therefore, we aimed to determine the burden of food insecurity among populations near Philadelphia Federally Qualified Health Center (FQHC) clinic sites.
This cross-sectional study was conducted in North Philadelphia, a populous and impoverished section of Philadelphia with many zip codes reporting 30-45% or more of the population below the federal poverty line. Students and clinicians affiliated with a local FQHC conducted surveys on residents (n = 379) within 1-mile radiuses of three FQHC sites, using the Hunger Vital Sign™, a validated food security tool. Survey data were collected through door-to-door visits in the summer of 2019. We used simple, age-adjusted bivariable, and multivariable logistic regression models to predict food insecurity with independent variables, including age, sex, language preference, and BMI category.
Food insecurity in North Philadelphia was much higher (36.9%) than previously reported in Philadelphia and nationwide. Food insecurity was inversely associated with age (AOR = 0.98, 95% CI: 0.97, 1.00), overweight (AOR = 0.58, 95% CI: 0.32, 1.06), and obesity (AOR = 0.60, 95% CI: 0.33, 1.09).
In North Philadelphia, the burden of food insecurity is higher than in the greater Philadelphia area, Pennsylvania state, and the rest of the nation and is predicted by age and BMI of residents. These findings demonstrate a need for more locally targeted research and interventions on food insecurity in impoverished urban settings.

To evaluate the effect of a decision aid on decisional conflict scale in patients choosing management for early pregnancy loss.
We conducted a pilot randomized control trial to assess the effect of the Healthwise patient decision aid on decisional conflict scale in patients with early pregnancy loss as compared with a control website. Patients 18years and older were eligible if they had an early pregnancy loss between 5 and 12 completed weeks of gestation. Participants completed surveys at baseline, poststudy intervention, after consultation, and 1week postconsultation. Surveys assessed participant scores on the decisional conflict scale (scale 0-100), knowledge, assessment of shared decision-making, satisfaction, and decision regret. Our primary outcome was the poststudy-intervention decisional conflict scale score.
From July 2020 through March 2021 we randomized 60 participants. After the intervention, the median decisional conflict scale score for the control group was 10 [0-30] and 0 [0-20] for the intervention group (p = 0.17). When assessing the decisional conflict scale subscales postintervention, the informed subscale for the control group was 16.7 [0-33.3] as opposed to 0 [0] for the patient decision aid group (p = 0.003). Knowledge remained significantly higher in the experimental arm from the postintervention to the 1-week follow-up. We found no differences between groups when assessing our other metrics.
Use of a validated decision aid did not result in statistically significant differences in the total decisional conflict scale scores as compared with the control. Participants allocated to the intervention were more informed postintervention and had consistently higher knowledge scores.
Use of a validated decision aid prior to early pregnancy loss management consultation did not affect overall decisional conflict but resulted in improved knowledge.