This study employed a cross-sectional design conducted from March to June 2018 in New Orleans, Louisiana. The sampling frame consisted of six of the city's thirteen Planning Districts, encompassing 289 of 457 census block groups and 951 sampled properties. Neighborhoods were selected using a risk-based approach informed by documented West Nile virus (WNV) activity, historically high Culex quinquefasciatus counts and their public health importance as identified through routine municipal vector-surveillance operations. Selection criteria incorporated historical mosquito surveillance activities, including routine trap-based monitoring conducted by the New Orleans Mosquito, Termite and Rodent Control Board. Areas with documented WNV activity, elevated Culex quinquefasciatus trap and operational mosquito-control significance were prioritized for inspection. The sampling frame also included a tornado-impacted neighborhood in New Orleans East, where extensive debris, damaged structures and disrupted drainage systems following the February 2017 EF-3 tornado created environmental conditions conducive to mosquito proliferation.

A cross-sectional design was appropriate because the objective was to characterize environmental and structural conditions associated with suitability of mosquito habitat and environmental risk indicators during the onset of peak seasonal activity. This design was appropriate for characterizing environmental and structural conditions associated with mosquito habitat suitability and environmental risk indicators rather than direct measures of mosquito abundance. Because site selection was guided by risk-based criteria and operational feasibility rather than random sampling, the study used a non-probabilistic design and findings should be interpreted as representative of inspected areas rather than the entire city. These combined epidemiologic, ecological and structural considerations provided the rationale for selecting neighborhoods most likely to exhibit conditions associated with mosquito breeding and vector exposure. Only 2018 inspections were included in the analytic dataset because this was the first year in which field data collection was fully standardized using Survey123. Earlier datasets, including those collected during 2016 Zika response activities, utilized different data-collection platforms and workflows that were not methodologically comparable to the 2018 inspection protocol.

This cross-sectional study was conducted from March to June 2018 in New Orleans, Louisiana, and used household inspection data collected by the New Orleans Mosquito, Termite and Rodent Control Board and the New Orleans Health Department as part of routine municipal mosquito-control operations.

The 2018 season represented the first year in which field data collection was fully standardized using mobile-friendly Survey123 forms developed by the City’s GIS Department. Survey123 incorporated structured forms, standardized response options, skip logic and address-validation workflows designed to improve consistency in environmental observations and reduce variability in field documentation. These forms were designed to improve consistency in environmental assessments and were informed by refinements made during the city’s 2016 and 2018 Zika mitigation initiatives, when the data-collection framework shifted from a map-centric to a form-centric system to reduce observer burden and enhance data quality. This transition enabled a more standardized assessment of parcel-level environmental conditions associated with mosquito habitat formation.

Inspection teams recorded observable environmental and structural conditions associated with mosquito habitat formation, including standing water, evidence of mosquito breeding, adult mosquito activity, container presence, yard condition, building condition, air-conditioning status, and other water-holding features. The protocol focused on observable environmental hazards and habitat indicators rather than direct measures of mosquito abundance, reflecting the operational objectives of the municipal inspection program. Because property access varied across neighborhoods, inspections ranged from full parcel assessments to partial visual assessments conducted from the property perimeter or public right-of-way. Property-owner reluctance, resident absence, safety concerns, physical barriers, and limited visibility occasionally restricted access to portions of a parcel. Consequently, observations reflected visible environmental conditions and may have underestimated the prevalence of mosquito habitat features located in backyards, enclosed spaces, or other inaccessible areas.

Of the 951 parcels included in the inspection frame, 634 received an inspection, consisting of 300 full inspections and 334 partial inspections, while 317 parcels were inaccessible, declined participation, or otherwise did not receive a completed inspection (Figure 1). Analyses were restricted to records containing the variables required for model estimation, resulting in differing analytic sample sizes across outcome measures. The inspection protocol focused on habitat indicators rather than entomological sampling. Mosquito specimens were not collected and species-level identification was not performed during parcel inspections. As a result, the study evaluates environmental suitability for mosquito habitat formation rather than species-specific mosquito abundance. Accordingly, findings should be interpreted as indicators of environmental mosquito risk rather than direct measures of vector population density.

Demographic data were obtained from the U.S. Census Bureau 2018 American Community Survey (ACS) 5-year estimates and included total population, racial composition, median household income and housing age. Racial composition was converted to percentages for White, African American, Asian, and a combined “other race” category to align with ACS reporting conventions. Housing age was calculated as the proportion of units built within three ACS-defined construction periods (1940–1969, 1970–1989, and 1990 or later), which reflect major shifts in housing infrastructure and building materials relevant to mosquito ecology. ACS construction-year categories do not distinguish post-Katrina reconstruction from pre-Katrina structures; therefore, homes rebuilt after Hurricane Katrina could not be identified.

Environmental and structural indicators were assessed through direct observation for each inspected parcel, including residential, commercial, paved and vacant lots. Inspection teams used a standardized Survey123 instrument to document observable environmental conditions associated with mosquito habitat formation as part of routine municipal vector-control operations. All environmental variables including standing water, larval habitats, adult mosquito presence, container presence, yard condition, property condition and water-feature maintenance were coded as binary indicators (present = 1, absent = 0) because inspectors documented only whether each condition was observed.

The inspection protocol focused on environmental hazard identification and habitat assessment rather than entomological sampling. Consequently, mosquito specimens were not collected, species-level identifications were not performed, and no direct measures of mosquito abundance were recorded. Findings should therefore be interpreted as indicators of environmental suitability for mosquito habitat formation rather than species-specific measures of mosquito production. Adult mosquito presence was recorded solely through visual confirmation and was treated as an observational indicator rather than a biological measure of abundance. Similarly, evidence of larval habitat was documented through field observation of breeding activity and water-holding environments considered suitable for mosquito development. Larval habitats were infrequently observed, a pattern consistent with ecological expectations that only a minority of water‑holding sites support active mosquito development; this rarity reflects biological processes rather than measurement error.

Property condition was categorized as “good or great” versus “decent or poor” based on visible structural integrity, exterior maintenance, and evidence of deterioration. Yard condition was categorized as “maintained” versus “blighted or unmaintained,” reflecting vegetation overgrowth, debris accumulation, and general upkeep. These categories reflect standard municipal vector-control practice, and inspectors received routine training on their application as part of operational field protocols.

Container presence included any water-holding container, including tires. Examples included buckets, coolers, toys, planters, watering cans, tarps, trash receptacles, lawn furniture, tire piles, and other artificial containers capable of retaining water. Because inspectors documented only the presence or absence of containers rather than individual container counts, analyses focused on the existence of potential breeding habitats rather than container abundance. Water-feature maintenance was coded as maintained or not maintained.

Air-cooling systems were categorized as no air conditioning, window unit, or central air based on external observation. Homeowner presence was recorded to document access and observational completeness, as sidewalk-only assessments occasionally constrained the precision of environmental observations. Homeowner presence was documented because access to private property and inspection completeness were influenced by resident availability and consent. This variable provided additional context for interpreting differences in inspection coverage across parcels. Inspection completeness varied because access to private property was not always possible. Some observations were conducted from the property perimeter or public right-of-way when homeowners were unavailable, access was restricted, or safety concerns limited entry. Consequently, environmental assessments reflect observable conditions and may underestimate the presence of concealed breeding habitats located in backyards or otherwise inaccessible areas. Environmental indicators were observed during inspections and subsequently coded as binary presence/absence variables for analysis.

A total of 951 parcels were included in the field sampling frame. Inspection teams completed 300 full inspections and 334 partial inspections. Partial inspections occurred when access limitations, homeowner absence, physical barriers, or safety concerns restricted observation of portions of a property. An additional 317 parcels could not be fully assessed because of refusal, inaccessibility, or other operational constraints. Figure 1 presents the parcel selection and inspection-completion workflow used in the study.

Data completeness for the 2018 inspection dataset was high because the digital form required inspectors to complete all observational fields before submission. The use of standardized forms, structured response options and built-in validation procedures reduced data-entry errors and improved consistency across inspection teams. Only 0.8% of inspection records had missing environmental indicators, and these were excluded from analysis.

Missingness in ACS variables reflected standard Census suppression patterns and affected fewer than 2% of block-group records. Missing demographic values were handled using listwise deletion in IBM SPSS Statistics version 29 (IBM Corp., Armonk, NY) as missingness was low and not systematically associated with environmental outcomes. No imputation procedures were used.

A total of 951 parcels were included in the study frame. Of these, 634 parcels received an inspection, consisting of 300 full inspections and 334 partial inspections, while 317 parcels were inaccessible, declined participation, or otherwise did not receive a completed inspection. The binary inspection variable identified whether a parcel received any inspection, whereas inspection completeness (full versus partial) was maintained separately to document field coverage. Analytic sample sizes therefore varied across models based on data availability and outcome-specific inclusion criteria. Earlier inspection years (e.g., 2016) were excluded because they were collected using different field tools and workflows and were not methodologically comparable to the standardized 2018 Survey123 dataset. The 2018 season represented the first fully standardized implementation of the municipal inspection protocol and was therefore selected as the primary analytic dataset.

Spatial data management was conducted in ArcGIS 10.5 26. Block‑group demographic attributes were spatially joined to inspection locations using the intersection tool. Parcel inspection locations were geocoded using addresses and GPS coordinates collected during field operations. While standardized address-entry procedures improved spatial accuracy, occasional GPS variability and address verification issues required routine quality-control review during data processing.

For research question 3, residential parcel‑level environmental indicators and ACS sociodemographic variables were aggregated to the planning‑district level, which represented the smallest neighborhood unit consistently available across the 2018 dataset. District‑level aggregation reduced parcel‑level clustering and allowed examination of ecological associations between neighborhood characteristics and standing water prevalence. Because neighborhood demographic characteristics were available only at aggregate geographic scales, district-level analyses were intended to evaluate broader ecological relationships rather than infer individual household-level socioeconomic effects. This approach allowed comparison of parcel-level environmental observations with larger neighborhood patterns relevant to municipal planning and vector-control operations.

All statistical analyses were performed in IBM SPSS Statistics version 29 27. Analyses were structured around three research questions: (1) which property-level environmental and structural conditions were associated with the presence of standing water; (2) which property-level factors predicted the presence of larval habitats; and (3) how neighborhood-level sociodemographic characteristics related to environmental mosquito-risk indicators. Logistic regression was used for each research question because all outcomes were binary. Because standing water and larval habitat observations were relatively uncommon, outcome prevalence was examined prior to model fitting to ensure sufficient events for logistic regression analyses. Categorical predictors, including air-cooling system type, were entered as indicator (dummy-coded) variables using SPSS’s automatic reference-category coding. For logistic regression models, results are presented as odds ratios (ORs) with 95% confidence intervals (CIs).

The first model examined associations between property-level characteristics and the presence of standing water. The logistic regression model took the form:

logit(P(StandingWateri=1))=β0+β1(YardConditioni)+β2(PropertyConditioni)+β3(ContainerPresencei)+β4(WaterFeatureNotMaintainedi)+β5(ACTypei).

The second model evaluated predictors of larval habitat presence and included standing water as a predictor because it is a necessary precursor for larval development:

logit(P(LarvalHabitati=1))=α0+α1(StandingWateri)+α2(ContainerPresencei)+α3(YardConditioni)+α4(PropertyConditioni)+α5(WaterFeatureNotMaintainedi).

Adult mosquito presence was examined descriptively but excluded from regression models because it was an observational outcome rather than a determinant of environmental conditions and lacked standardized measurement. Because species-level mosquito identification was not conducted during inspections, analyses focused on environmental indicators of habitat suitability rather than species-specific mosquito ecology.

The third model assessed whether district-level sociodemographic characteristics were associated with the proportion of parcels containing standing water. Parcel-level environmental indicators and ACS variables were aggregated to the planning-district level, and linear regression was used because the outcome, district-level standing-water prevalence was continuous. Planning districts represented the most consistent administrative geography available across the municipal inspection dataset and aligned with operational mosquito-control planning boundaries. Predictors included median household income, percent Black residents, mean housing-age category, and total population. All predictors were standardized to district-level means prior to analysis. The linear regression model took the form:

StandingWaterPrevalenced=γ0+γ1 (MedianIncomed)+γ2(PercentBlackd)+γ3(HousingAged)+γ4(Populationd) + ∈d

where each predictor represents the district-level mean or total for district d. This ecological analysis was intended to evaluate broader neighborhood patterns and was not used to infer individual household-level relationships. Because district-level analyses were based on six planning districts, findings from these models should be interpreted as exploratory and hypothesis-generating. Model fit was evaluated using R² and adjusted R², and confidence intervals were derived from model-based standard errors.

Model diagnostics were conducted in SPSS to assess multicollinearity, influential observations and overall model fit. Variance inflation factors were examined to evaluate collinearity among predictors. Standardized residuals and Cook’s distance were used to identify influential cases. Model fit was assessed using the Hosmer–Lemeshow goodness-of-fit test and the area under the receiver operating characteristic curve. Statistical significance was defined as p < 0.05.

A total of 951 parcels were included in the study frame, encompassing residential, commercial, paved, and vacant lots. Of these, 634 received an inspection, including 300 full inspections and 334 partial inspections, while 317 parcels were inaccessible, declined participation, or otherwise did not receive a completed inspection (Figure 1). Inspection outcomes were evenly distributed, with 31.5% receiving a full inspection, 35.1% partial inspections, and 33.3% recorded as no inspection or refusal.

Most parcels were residential (57.8%), while commercial properties and paved or vacant lots represented a small proportion of the sample. Air conditioning units were observed at 47.1% of parcels, whereas 52.8% had no visible AC. Property conditions were generally favorable, with 91.3% rated as good or great and nearly all yards (98.8%) were maintained.

Environmental indicators were infrequently observed: standing water (7.2%), containers (12.8%), tire piles (2.1%), adult mosquitoes (2.6%) and larval habitats (4.0%). The relatively low prevalence of larval habitats is consistent with the expectation that only a subset of water-holding environments support active mosquito development at any given time. Consistent with the study design, the sampling frame was concentrated in districts prioritized through municipal mosquito-surveillance activities, including areas with historical West Nile virus activity and elevated Culex quinquefasciatus trap counts.

Parcels were distributed across all planning districts, with the largest share located in Algiers (34.9%) and Uptown/Carrollton (20.2%). Sociodemographic characteristics reflected substantial variation across block groups, with a mean household income of $30,925 (SD = 10,854) and an average population of 831 residents (SD = 524). Table 1 presents the full distribution of parcel characteristics, environmental observations and contextual block group indicators.

Larval habitats were detected on 4.0% of inspected parcels, with strong and statistically significant differences across several environmental indicators (Table 2). Parcels with standing water showed markedly higher larval positivity (45.6%) compared with parcels without standing water (0.8%; p < 0.001). Container presence was also strongly associated with larval habitats, with 27.0% of container‑positive parcels harboring larvae compared with only 0.6% of container‑negative parcels (p < 0.001). Yard condition demonstrated a pronounced gradient: 54.5% of parcels rated as blighted or unmaintained were larval‑positive, compared with 3.4% of maintained parcels (p < 0.001). Property condition showed a similar pattern, with 10.8% of parcels in decent or poor condition exhibiting larval habitats versus 3.3% of parcels in good or great condition (p = 0.002). Water‑feature maintenance was not significantly associated with larval presence in bivariate analysis (p = 0.46). These findings indicate that standing water, container presence, and indicators of poor yard or property maintenance are strongly associated with larval habitat presence and therefore warrant inclusion in multivariable models. Because these observations represent environmental habitat indicators rather than species-specific mosquito measurements, findings should be interpreted as predictors of environmental suitability for mosquito development.

Standing water on a parcel remained the strongest predictor of larval habitat presence after adjustment for other parcel‑level characteristics. Despite the relatively low prevalence of observed larval habitats (4.0%), several parcel-level environmental factors demonstrated strong independent associations with larval habitat presence. Parcels with standing water had 18.45 times higher odds of containing larval habitats (95% CI: 4.86–70.39, p<0.001). Container presence was also a strong independent predictor; parcels with containers had 15.90 times higher odds of larval habitat presence (95% CI: 4.11–61.47, p<0.001). Yard condition showed a borderline association, with blighted or unmaintained parcels exhibiting 7.90 times higher odds of larval habitat presence (95% CI: 0.93–67.10, p=0.058). Property condition and water‑feature maintenance were not statistically significant predictors in the adjusted model. Overall model fit was acceptable (Pseudo‑R² = 0.148), indicating that parcel‑level environmental features meaningfully contributed to predicting larval habitat presence. Standing water and container presence emerged as the dominant environmental predictors, underscoring the importance of source-reduction strategies targeting water-holding containers and unmanaged breeding habitats.

Container presence was the only statistically significant predictor of standing water on residential parcels (Table 4). Parcels with containers had 19.96 times higher odds of having standing water compared to parcels without containers (95% CI: 8.69–45.99, p<0.001). Yard condition, property condition, and water-feature maintenance were not significantly associated with standing water in the adjusted model. The model explained a modest proportion of variance (Pseudo-R² = 0.091), indicating that container presence is the primary parcel-level environmental factor associated with standing water accumulation. The strong association between containers and standing water supports the operational value of container-focused inspection and source-reduction activities within municipal mosquito-control programs.

Planning district‑level sociodemographic characteristics were not significantly associated with the proportion of parcels containing standing water (Table 5). Median household income, percent Black residents, mean housing age, and total population showed no meaningful adjusted associations with standing water prevalence across the six districts included in the 2018 sample (all p>0.49). Although the model explained a substantial proportion of variance (R² = 0.774), the adjusted R² was negative, reflecting the small number of districts and limited statistical power. Overall, neighborhood‑level sociodemographic indicators did not demonstrate clear ecological associations with standing water in this sample. Because analyses were based on only six planning districts, these findings should be interpreted cautiously and viewed as exploratory rather than definitive evidence of an absence of neighborhood-level effects.