Pausing the Pandemic: Understanding and Managing Traveler and Community Spread of COVID-19 in Hawaii

Abstract

In the absence of a vaccine, nonpharmaceutical interventions such as social distancing and travel reductions were the only strategies for slowing the spread of the COVID-19 pandemic. Using survey data from Hawaii (n = 22,200) collected in March through May of 2020 at the onset of the pandemic, the differences between traveler spreaders who brought the disease into the state and community spreaders were investigated. In addition to describing the demographic attributes and comparing them with attributes of those who were vulnerable to COVID-19, logit models explaining travel behaviors were developed and tested. Traveler spreaders were likely to be male, younger, and returning students. Community spreaders were more likely to be male, essential workers, first responders, and medical personnel at the highest risk of exposure. Using spatial statistics, clusters and hotspot locations of high-risk individuals were mapped. As transportation researchers are in a position to combine their critical analytical capabilities and experience with relevant databases on mobility and the spread of infectious diseases, this analysis could support efforts to respond to and slow the spread of the pandemic.

Keywords

business continuity community continuity consequence management disaster response recovery emergency management incident management sustainability and resilience transportation systems resilience

A novel coronavirus, COVID-19, has spread throughout the world. At the time of writing this paper, there were more than 2.7 million cases in the United States with no clear end in sight. Without a vaccine, nonpharmaceutical interventions, including travel restrictions, social distancing, and personal hygiene were recommended by the Centers for Disease Control and Prevention (CDC) to pause the spread of COVID-19 ( 1 ). Transportation researchers have studied the effects of illnesses on mobility and how movements of people and goods affect disease spread (2–6). Parr et al. ( 7 ), De Voss ( 8 ), and De Haas et al. ( 9 ) have shown significant reductions in travel and changes in travel behavior because of COVID-19. There have also been shifts in travel modes owing to reduced travel demand and transport operations ( 10 ).

Prior disease outbreaks have informed understanding and responses to COVID-19. Holte et al. examined risk perceptions and protective behaviors during past influenza outbreaks that were relevant to COVID-19 ( 11 ). There are also lessons to be learned from the Ebola ( 12 ) and dengue fever epidemics ( 13 ).

Modern transportation systems have increased efficiencies in moving people, thereby further spreading infectious diseases. However, there are new technologies, applications, and systems for monitoring travelers to better contain the virus. Of particular relevance are actions implemented at airports (14–16), in aircraft ( 17 ), within air networks (18, 19), and in relation to passenger screening (20, 21) to reduce the spread of disease. Understanding transportation behavior across different systems, modalities, users, and technologies therefore is key to slowing the spread of diseases (22–24).

In March 2020, COVID-19 reached Hawaii and the state implemented actions including air travel restrictions, social distancing, and closure of businesses, schools, parks, and restrictions on public gatherings. On March 4, Governor David Ige issued an Emergency Proclamation to combat the disease and on March 21, ordered a mandatory 14-day self-quarantine for travelers entering the state. On March 23, a statewide stay-at-home order was issued and schools were closed, with further orders to use face coverings and increase social distancing. These efforts led to a lowering of reported COVID-19 cases, as shown in Figure 1.

Figure 1.

Weekly reported COVID-19 cases, airline passenger arrivals, and social distancing orders, Hawaii.

Over this period, daily visitor arrivals dropped from approximately 35,000 per day to 300 (Figure 1). The weekly number of confirmed COVID-19 cases went as high as 200 but eventually decreased to single-digit daily counts. During the first week of June, 2020 Hawaii began easing restrictions. Although large gatherings were prohibited, bars and nightclubs were reopened, restaurant dining was allowed, parks and beaches were reopened, and businesses were encouraged to resume activity. As restrictions were lifted in early June (Week 23), clusters of cases emerged. In early July of 2020, according to the CDC, the per capita rate of infection in Hawaii was among the lowest in the nation at 62.9 per 100,000 people, compared with the national average of 849.5, or high-frequency states such as Florida (825), Texas (639.4), and California (627.5) ( 25 ). After reopening in July 2020, that value was exceeded, and as of July 31, 2020, was 132 per 100,000 ( 25 ). As in other cities, there have been public gatherings, protest marches, and holidays that contributed to greater social interaction. This increase in social activities led to spikes in cases.

Several infection models have been developed, which can be used to estimate the effects of the relaxation of social distancing, the reopening of businesses, and increased activity. A widely used model from the Institute for Health Metrics and Evaluation (University of Washington), estimated that, given the rates of infection and easing of restrictions, from July 5 until October 1, 2020, the infections per day in Hawaii were projected to increase 26-fold ( 26 ). This model assumed mask use at the observed rate at the time, with an easing of social distancing restrictions. In Hawaii, according to the COVID ActNow SEIR (Susceptible, Exposed, Infected, and Recovered) model, COVID-19 continued to spread and, on average, each person with COVID-19 was infecting 1.07 people ( 27 ). This SEIR model from the Massachusetts General Hospital, also known as the “COVID-19 simulator,” assumed that over time there would be a 20% increase in mobility as states reopened. For Hawaii, it was estimated that by October 26, 2020, the infection would peak at 370 daily cases ( 28 ). However, the SIR (Susceptible, Infected, and Recovered) model, from the Massachusetts Institute of Technology COVID-19 Policy Alliance, assumed that current interventions would remain in place indefinitely, and estimated that by October 26, 2020, daily cases would peak at 62 ( 29 ). As of July, 2020 these counts were exceeded.

Hawaii, because of its isolation and development of data systems, is an ideal location for studying the impacts of and responses to disasters; for these reasons, it was an excellent site for studying the spread of COVID-19 in response to government orders and social distancing.

As an island state, which conjures images of swaying palms and white sandy beaches, Hawaii is renowned for tourism, with over 10 million visitors in 2019. However, it has many natural hazards and threats, including volcanoes (30, 31), flooding, coastal storms (32, 33), and tsunamis ( 34 ). As a remote island state, it has had to develop systems for monitoring, responding to, and recovering from disasters. Recent research has focused on the critical role and vulnerabilities of its transportation systems (35, 36). Other conditions in Hawaii support this research into transportation and the pandemic, for example, it has a centralized government system with only four counties comprising the local government system, with most of the population residing on the island of O’ahu under the jurisdiction of the City and County of Honolulu. Research also supports the National Disaster Preparedness Training Center, a congressionally authorized partner of the Federal Emergency Management Agency, which develops training courses for first responders and emergency managers. The lessons from this pandemic are relevant to first responders, emergency managers, and also to those involved in reducing health risks and the economic and social impacts of such risks.

This paper focuses on the differences between traveler and community spread of COVID-19 in Hawaii. As the number of travelers from outside the state increases, so too does the risk of infection and the potential for greater community spread. As more people in the community are infected, the rates of infection will also increase. Because of the large number of visitors to Hawaii, the shutdown and restrictions on air travelers, the long period of lockdown, and the activity restrictions in this island state, this represented a unique opportunity to analyze the differences between travelers who brought the disease into the state and others in the community who caught the disease and spread it. The research was based on survey data collected in Hawaii from March 26 to April 14, 2020, combined with public health data and data on government-imposed actions taken over the preceding 6 months. Understanding the characteristics and behaviors of traveler- and community spreaders was key to preventing and slowing the spread of the disease. In addition to describing the demographics and trip-making of traveler- and community spreaders using descriptive statistics, logistic regression, and generalized linear modeling, analyses using geographic information system data were conducted to identify interactions over time and space between spreaders and others. After describing the data, analyses, and models for traveler- and community spreaders, the implications of the findings and results for Hawaii and beyond are described.

Data and Methods

Following pretesting of the instrument and approval of the research protocol by the University of Hawaii Institutional Review Board, a questionnaire was distributed on March 26, 2020, using the Qualtrics online survey system. In addition to distribution through email and social media outlets, the survey received widespread publicity through print, television, and other media outlets. In addition to reaching key community leaders and private sector organizations, the instrument asked respondents to forward the survey link to others. The focus was to determine the initial spread of COVID-19 symptoms among travelers and returning residents and among the general population. By April 14, 2020, a total of 22,201 responses had been received. This was a high number of respondents, reflecting concerns about the disease. The survey results have been shared widely with the media and on publicly accessible blog sites, such as the Hawaii Data Collaborative ( 37 ). These communications with the public encouraged additional respondents to complete the survey.

Survey respondent characteristics were similar to the overall population. Table 1, shows between-state population characteristics based on U.S. Census Bureau data and the sample characteristics from the survey. There were slight differences such as an overrepresentation of females in the survey but, overall, survey respondents were similar to the general population. Other potential biases include those associated with an online survey, that is, more educated, affluent, and digitally connected respondents being more likely to participate than other groups. It should be noted, however, that these are also the groups that are most likely to travel or to be actively engaged in activities associated with the spread of the disease.

Table 1.

Comparison Between Traveler and Community Spreaders and Others

	Census population	Pct (%)	Total		Travel spreaders		Community spreaders		Other
	Census population	Pct (%)	Freq	Pct (%)	Freq	Pct (%)	Freq	Pct (%)	Freq	Pct (%)
Age
<25	419,204	30	4,816	22	70	15	515	9	4,231	27
25–65	727,198	51	14,242	64	352	77	5,202	86	8,688	55
>65	269,470	19	3,142	14	38	8	329	5	2,775	18
Total	1,415,872	100	22,200	100	460	100	6,046	100	15,694	100
Gender
Male	707,855	50	5,004	28	146	35	1,765	31	3,093	26
Female	708,017	50	13,127	72	275	65	3,959	69	8,893	74
Total	1,415,872	100	18,131	100	421	100	5,724	100	11,986	100
Ethnicity
African American	26,923	2	125	1	4	1	32	1	89	1
Asian	547,843	39	6,413	36	133	32	2,411	43	3,869	33
Caucasian	341,211	24	7,099	40	185	45	1,710	31	5,204	44
Native American	6,129	0.4	110	1	2	0	21	0	87	1
Native Hawaiian	90,070	6	1,794	10	35	9	665	12	1,094	9
Pacific Islander	62,531	4	545	3	7	2	244	4	294	3
Other	341,165	24	1,642	9	45	11	521	9	1,076	9
Total	1,415,872	100	17,728	100	411	100	5,604	100	11,713	100
Sector
Banking/Finance	97,900	15	960	5	25	6	374	6	561	5
Construction	67,568	10	459	3	11	3	308	5	140	1
Food service	56,141	8	545	3	12	3	197	3	336	3
Government	21,027	3	1,974	11	39	9	989	17	946	8
Medical	80,662	12	1,583	9	43	10	1,027	18	513	4
Retail	72,550	11	523	3	11	3	205	4	307	3
Self-employed	NA		1,829	10	41	10	297	5	1,491	12
Transportation	43,718	7	212	1	6	1	110	2	96	1
Visitor industry	NA		665	4	12	3	195	3	458	4
Other	232,202	35	9,427*	52	227*	53	2,073*	36	7,127*	60
Total	671,768	100	18,177	100	427	100	5,775	100	11,975	100
Travel mode
Private vehicle	573,951	86	17,918	87	409	83	5,695	87	11,814	87
Transit	43,590	6	939	5	23	5	296	5	620	5
Other	53,369	8	1,727	8	60	12	521	8	1,146	8
Total	670,910	100	20,584	100	492	100	6512	100	13,580	100
Location
Urban Honolulu	345,064	24	3,898	30	88	31	1,220	29	2,589	31
Rest of Oahu	629,499	44	7,652	59	176	61	2,638	62	4,838	57
Neighboring island	441,223	31	1,443	11	24	8	400	9	1,019	12
Total	1,415,786	100	12,993	100	288	100	4,258	100	8,446	100

Note: Pct = %; Freq = frequency; NA = not available

Traveler spreaders are defined as visitors or returning residents with COVID-19 symptoms who traveled to Hawaii from the United States or another country. Community spreaders are those residents with symptoms who did not travel but continued to work after the stay-at-home orders were issued. Because of the lack of testing in March and April, 2020, respondents were asked to report their symptoms of COVID-19 (defined by CDC [ 38 ]), and whether they had been in close contact with symptomatic individuals or were under treatment for the disease. Questions on travel behaviors and compliance with stay-at-home and social distancing actions (i.e., avoiding large gatherings, working from home, and maintaining at least 6 ft from other people) were also included in the survey. In addition to demographic attributes, questions on transportation, travel modes, and trip-making—both locally and out-of-state travel—were included.

In the next section, the findings are summarized. Frequencies using a contingency table (cross tabulation) to summarize the relationships between traveler- and community spreaders and categorical explanatory variables of age, gender, ethnicity, employment sector, mode of transportation, and location of the respondent (i.e., urban Honolulu, rest of O’ahu, neighboring islands) are included. Standard statistical tests (Z-scores) used to compare traveler- and community spreaders with other respondents are also included ( 39 ).

Two logit models were developed and tested using SAS statistical analysis software ( 40 ), which estimated the effects of the explanatory variables on the binary response of being a traveler- or community spreader. A standard logistic regression model was tested with the following:

logit {\Pr (Y = 1 | x)} = \log {\frac{\Pr (Y = 1 | x)}{1 - \Pr (Y = 1 | x)}} = β_{0} + x^{'} β

(1)

where

response Y can take on two possible values, denoted by 1 and 0,

$x = {(x_{1}, . . ., x_{k})}^{'}$ is the vector of explanatory variables,

$β_{0}$ is the intercept parameter, and

$β$ is the vector of slope parameters ( 41 ).

The authors have extensive experience using this multivariate analytical technique for transportation research (42 –45).

The results provided an understanding of traveler spreader characteristics, and then as the infection spread in the community, how others behaved, traveled locally, and spread the disease. As the state at the time was looking to reopen the tourism sector and resume air travel to Hawaii, understanding the traveler-initiated and community spread of COVID-19 was critical.

Spatial analyses were conducted to determine the hotspots for traveler- and community spreaders. Using the Getis-Ord Gi* or Getis statistic, the behaviors of both traveler- and community spreaders were studied. The Getis-Ord Gi* statistic produces a Z-score, and only those significant at the 0.05 confidence level were used to identify and map hot spots. This technique has the advantage of capturing spatial concentrations that are statistically significant: the feature must not only have a high value, but be surrounded by features with high values for the feature to be statistically significant ( 46 ).

The Getis statistic can be represented by

G_{i}^{*} = \frac{\sum_{j = 1}^{n} w_{i, j} x_{j} - \bar{X} \sum_{j = 1}^{n} w_{i, j}}{S \sqrt{\frac{n \sum_{j = 1}^{n} w_{i, j}^{2} - {(\sum_{j = 1}^{n} w_{i, j})}^{2}}{n - 1}}}

(2)

where

$x_{j}$ is the attribute value for feature $j,$

$w_{i, j}$ is the spatial weight between i and j,

n is equal to the total number of features,

$\bar{X}$ is the mean, and

S is the standard deviation.

\bar{X} = \frac{\sum_{j = 1}^{n} x_{j}}{n}

(3)

S = \sqrt{\frac{\sum_{j = 1}^{n} x_{j}^{2}}{n} - {(\bar{X})}^{2}}

(4)

The resulting $G_{i}^{*}$ is a Z-score, and the larger the Z-score, the more intense the clustering (hot spot).

Findings and Discussion

General Survey Results

Most of the 22,201 respondents reported no symptoms (73%). This may be a misleading value, as 20% of infected people may have been asymptomatic ( 47 ), and may have unknowingly spread the disease. Three percent stated that they had been in close contact with an infected individual, and 2% reported that they were living in the same household as someone with symptoms or had been confirmed to have COVID-19. Among the respondents, 26% reported chronic or preexisting medical conditions, and 19% were over 65. Some 51% of the respondents stated that they lived with someone over 65 years of age. Those with chronic conditions or living with older adults indicated a high level of risk and vulnerability to the disease.

The majority of the respondents stated that they or members of their households had not returned from a trip within the last 14 days (90%). Four percent reported that they had traveled interisland, 5% had been to the U.S. mainland, and 1% had traveled internationally. In relation to local trips, the largest percentage of respondents said they had been to grocery and drug stores (62%), followed by 26% reporting they had continued to go to work, although the stay-at-home order was in place. Many of these were classified as essential workers. Overall, 61% of respondents answered that they were staying at home within 5 days of answering the survey.

With the order banning large gatherings, few reported attending conferences (0.1%), concerts (0.03%), theaters (0.1%), sporting events (0.1%), or events with friends or family (0.1%). On March 19, 2020, the state ordered the closure of beaches and parks, and banned dining at restaurants and bars. Nonetheless, 13% reported trips to the beach or park, and 6% reported going to restaurants or bars, presumably for takeout or limited outdoor service.

A large proportion of the respondents reported they traveled by personal vehicle (81%). There has been a reduction in roadway traffic, transit use, and local travel because of the reduction in tourism, the closure of businesses, and movement restrictions. Four percent of the respondents said they take public transit, 0.2% use shuttles, 3% use taxis and rideshare, and another 4% use other means of transportation such as biking or walking.

Table 1 shows the differences between those identified as traveler and community spreaders by age, gender, ethnicity, sector of employment, travel mode, and location of residence for the overall respondents, which are then broken down by traveler spreaders, community spreaders, and those not classified as spreaders (Others). The Other category could also be considered those likely to be infected by the traveler and community spreaders. Z-scores were used to test the statistical significance of the differences between the groups by these characteristics. The representativeness of the sample was found to be quite good because of the high number of respondents (22,201). While there was some bias and overrepresentation of older, educated, and wealthier respondents, there were enough cases to estimate the characteristics of spreaders and whom they might infect. Most of the respondents (64%) were between the ages of 25 and 65, with 460 respondents identified as traveler spreaders and 6,046 as community spreaders. Seventy-seven percent of traveler spreaders and 86% of community spreaders were between the ages of 25 and 65. These represented higher proportions than in the overall sample. A larger percentage of the respondents were female (64%), which may reflect biases associated with online survey completion. Sixty-five percent of those identified as traveler spreaders were female, and in relation to community spreaders, 69% were female. However, notably, the proportion of male spreaders (traveler and community) was higher than the overall response proportions. A larger proportion was of White ethnicity (40%), and the largest proportion of traveler spreaders were of White ethnicity (45%); however, the greatest proportion of community spreaders was of Asian ethnicity (43%). The largest proportion of respondents identified that their sector of employment was Other (52%), followed by Government (11%), and being Self-Employed (10%). The Other category also included students, the unemployed, and retired. Approximately 53% of the traveler spreaders were of the Other employment sector (many of whom were students), but both Government and Self-Employed each made up 10% of the traveler spreaders. For community spreaders, those in the Other sector category made up a large proportion of community spreaders (36%), with those in Medical jobs (18%) and Government (17%) likely to be community spreaders. As for travel mode, the majority of all respondents stated that they use a private vehicle (87%), and with similar proportions for both traveler- (83%) and community spreaders (87%). Traveler spreaders reported higher use of taxis, rideshare, and other transportation services. Of the total, 30% reported that they lived in urban Honolulu, 59% lived outside of urban Honolulu, and 11% of the respondents lived on neighboring islands. A larger proportion of the traveler spreaders (61%) and community spreaders (62%) lived outside of urban Honolulu on the island of O’ahu compared with other parts of the state.

Regression Models

Table 2 shows the logit model for the odds of being a traveler spreader over not being in this category against explanatory variables including age, gender, ethnicity, and travel mode. Those respondents that were younger than 25 (3.87 times more likely), male (1.439 times more likely), Not Asian (1.393 times more likely), Not Hawaiian (1.482 times more likely), and Not Pacific Islander (2.329 times more likely), and those who used taxis or ridesharing (2.082 times more likely) as their means of transportation were more likely to be traveler spreaders.

Table 2.

Logit Regression for Traveler Spreaders

Deviance and Pearson goodness-of-fit statistics
Criterion	Value		df		Value/df			Pr > chi sq
Deviance	33.7882		39		0.8664			0.7062
Pearson	31.6664		39		0.812			0.7917
Analysis of maximum likelihood estimates
Parameter	df	Estimate		Standard error		Wald chi-square	Pr > chi sq		Odds ratio
Intercept	1	−5.3082		0.4561		135.4613	<.0001
Age < 25 versus age > 65	1	0.6425		0.1243		26.7169	<.0001		3.87
Gender (male versus female)	1	0.3639		0.1043		12.1628	0.0005		1.439
Not Asian versus Asian	1	0.3318		0.11		9.0967	0.0026		1.393
Not Hawaiian versus Hawaiian	1	0.3935		0.1877		4.3954	0.036		1.482
Not Pacific Islander versus Other ethnicity	1	0.8456		0.3868		4.7797	0.0288		2.329
Taxi, Uber or Lyft versus Other transportation	1	0.7334		0.1868		15.4194	<.0001		2.082

Note: df = degrees of freedom; Pr = probability; chi sq = chi square.

Table 3 shows the model for community spreaders against explanatory variables age, gender, and ethnicity, and those working in essential jobs (medical/healthcare, finance, construction, government services, and transportation). Community spreaders were most likely to be between the ages of 25 and 65 (3.732 times more likely), to be aged less than 25 (2.543 times more likely), male (1.284 times more likely), essential workers (3.293 times more likely), of Hawaiian ethnicity (1.297 times more likely), or Pacific Islander ethnicity (1.637 times more likely).

Table 3.

Logit Regression of Community Spreaders

Deviance and Pearson goodness-of-fit statistics
Criterion	Value		df			Value/df		Pr > chi sq
Deviance	97.8882		29			3.3755		<.0001
Pearson	93.6862		29			3.2306		<.0001
Analysis of maximum likelihood estimates
Parameter	df	Estimate		Standard error		Wald chi-square	Pr > chi sq		Odds ratio
Intercept	1	−1.6545		0.0369	2,011.216		<.0001		na
Age 25–65 versus age > 65	1	0.5668		0.0344	271.5977		<.0001		3.732
Age < 25 versus age > 65	1	0.1834		0.0548	11.193		0.0008		2.543
Gender (male versus female)	1	0.25		0.0376	44.1958		<.0001		1.284
Essential versus Non-essential workers	1	1.1918		0.035	1,159.0192		<.0001		3.293
Hawaiian versus Other ethnicity	1	0.2602		0.0552	22.2216		<.0001		1.297
Pacific Islander versus Other ethnicity	1	0.4928		0.0945	27.2181		<.0001		1.637

Note: df = degrees of freedom; Pr = probability; chi sq = chi square; na = not applicable.

To investigate the spatial patterns of the spread of COVID-19, respondents were asked to provide the nearest intersection to their residence. Intersections were geocoded, producing latitude and longitude coordinates, then mapped and analyzed. The traveler- and community spreaders were cumulated to a uniform 0.25-mi grid structure from the geocoded data. This approach was taken to mask individual-level information and better analyze the spatial relationships between COVID-19 risk, urbanization, density, land use, and other social and environmental features.

Figures 2 and 3 display the clustering for traveler spreaders and community spreaders, respectively. The Getis-Ord Gi* statistic Z-Scores were calculated to determine the intensity of clusters for traveler- and community spreaders. The of Z-Scores for traveler spreaders ranged from 2.5 to 17.2, whereas for community spreaders they ranged from 2.0 to 22.9. This illustrated that there were more intense clusters of community- as opposed to traveler spreaders. The distribution of traveler spreaders was more dispersed, whereas that of community spreaders was more compact. Spatial references such as mountainous regions (conservation land), coastline, and major roads (where the population reside and work) provided further detail on locations of development and where COVID-19 was likely to spread.

Figure 2.

Map of hot spots of traveler spreaders.

Figure 3.

Map of hot spots of community spreaders.

The data and analyses could be used in different ways. First, it was important to understand and respond to two the different pathways for the spread of the COVID-19. Travelers initially brought the disease into Hawaii, which then was spread by other members of the community. Strategies to contain the disease at the water’s edge were different from efforts to fight the spread within the islands. It was and continues to be important to consider both the similarities and differences between traveler- and community spreaders in the design of community interventions, especially as the disease continues to spread. Second, characteristics such as age, gender, ethnicity, employment sector, and so forth were relevant to the targeting, tracing, and tracking needed to contain the disease and protect others from catching it. Third, the framework presented helps to understand what happened in Hawaii and what is likely to occur once travel to the islands is fully reopened as restrictions remain in place. Finally, longer-term integrated approaches are needed. Infection control is based on an understanding of both carriers of the disease and the pathways for infection among others in the community. In addition to education on social distancing, hygiene, face coverings, and reducing exposure to the disease, it is also important to identify travel modes, especially shared spaces—public transit and ridesharing—as potential contamination pathways. A more holistic approach that considers disease alongside activity centers, trip generators, distribution of risk, mode choice, and network assignment for both spreaders and vulnerable populations will help in the efforts to contain the disease.

Another important finding using statistics such as the Getis-Ord Gi*Z-score was that there were hotspots and clusters where interventions could be targeted. Traveler- and community spreaders, and others susceptible to the virus could be better informed about the risks. Workplaces, businesses, housing, and other activity generators where the disease is likely to reside and spread should be mapped and shared so that individuals, firms, and organizations can take appropriate action. Privacy, confidentiality, and prudent management of sensitive, personal information must be ensured, but with a public health emergency, the public has a right to know. Better communications and integration with media, social media, and public information sources are needed and reinforce the protective actions necessary to reduce risks. The methods developed in this paper, would also support the targeting of messaging, education, and encouraging protective actions in the face of the pandemic.

Conclusions

The COVID-19 pandemic created a new urgency for research and understanding of this complex disaster. Using tools and approaches familiar to transportation researchers, the differences between traveler- and community spreaders of the disease in Hawaii were investigated. There were limitations with the data and the application of the models and statistics resulting from the emergent, messy nature of the disease, including the reliance on self-reported symptoms and exposure. The “lie factor” ( 48 ) or not reporting of conditions or behaviors which are perceived to be socially unacceptable such as not wearing seatbelts, was also a concern with survey research. Improvements in both the understanding and implementation of interventions requiring behavioral change could occur over time. Expanded access and integration of relevant data ( 27 , 29 , 33 , 36 ) on travel behavior, demand modeling, and the pandemic will surely enhance responses. Increased testing, which was limited at the time of the survey, has helped to improve the precision and accuracy of identifying carriers, spreaders, and those susceptible to the disease. There are challenges with protecting privacy, intellectual property, and with accessing and sharing personal data, which need to be addressed so that more focused and targeted protective actions can be employed.

Although this paper focused on traveler- and community spreaders, there needs to be greater attention to protecting vulnerable populations, including the elderly and those with chronic medical conditions. Although there were similarities between traveler- and community spreaders, those most likely to be economically and socially active were also most likely to travel, both bringing the virus into the state but also moving from home to work (e.g., students, essential workers, first responders, and medical personnel), and at higher risk of infection and infecting others. More attention to travelers (14 –16, 21, 24), including returning college students and frequent fliers, as well as new policies on quarantine, isolation, and social distancing ( 49 ) are needed. Building on the research and guidance for transportation pandemic planning and response ( 23 ) with lessons from COVID-19 (49, 50) will have long-term payoffs. Better protection of essential workers, first responders, and others on the frontlines is also needed. Transportation workers are at higher risk of exposure and need additional protection ( 22 ), and the best practices and lessons learned need to be integrated into training and education. Further research will go into the determination and analysis of risk factors, and the development of tools for assessment and management of risks based on both demographic characteristics and travel behaviors. Renewed attention to resilience in the transportation sector (35, 49) will support health, safety, and wellbeing in our communities.

Footnotes

Acknowledgements

The authors acknowledge the support of the National Disaster Preparedness Training Center and Pacific Urban Resilience Lab at the University of Hawaii, the Hawaii Community Foundation, and the Hawaii Data Collaborative. Mark Kimura (Oceanit Industries) provided invaluable technical assistance with geocoding. Many others in the community helped to distribute the questionnaire.

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: K. Kim; data collection: K. Kim, E. Yamashita, and J. Ghimire; analysis and interpretation of results: K. Kim, E. Yamashita, and J. Ghimire; draft manuscript preparation: K. Kim, E. Yamashita, and J. Ghimire. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Karl Kim

Eric Yamashita

Jiwnath Ghimire

References

Centers for Disease Control and Prevention. Coronavirus (COVID-19). https://www.cdc.gov/coronavirus/2019-ncov/index.html. Accessed July 3, 2020.

Aloi

Alonso

Benaavente

Cordera

Echániz

González

Ladisa

, et al. Effects of the COVID-19 Lockdown on Urban Mobility: Empirical Evidence From the City of Santander (Spain). Sustainability, Vol. 12, No. 9, 2020, P. 3870. https://doi.org/10.3390/su12093870.

Kraemer

M. U. G.

Yang

C.-H.

Gutierrez

Klein

Pigott

, et al. Open COVID-19 Data Working Group, The Effect of Human Mobility and Control Measures on the COVID-19 Epidemic in China. Science, Vol. 368, No. 6490, 2020, pp. 493–497. https://doi.org/10.1126/science.abb4218.

Hadjidemetriou

G. M.

Sasidharan

Kouyialis

Parlikad

A. K.

The Impact of Government Measures and Human Mobility Trends on COVID-19 Related Deaths in the U.K.

Transportation Research Interdisciplinary Perspectives, Vol. 6, 2020, pp. 1--6. https://doi.org/10.1016/j.trip.2020.100167.

Ivanov

Predicting the Impacts of Epidemic Outbreaks on Global Supply Chains: A Simulation-Based Analysis on the Coronavirus Outbreak (COVID-19/SARS-CoV-2) Case. Transportation Research Part E: Logistics and Transportation Review, Vol. 136, 2020. https://doi.org/10.1016/j.tre.2020.101922.

Chinazzi

Davis

J. T.

Ajelli

Gioannini

Litvinova

Merler

Piontti

A. P. y

, et al. The Effect of Travel Restrictions on the Spread of 2019 Novel Coronavirus (COVID-19) Outbreak. Science, Vol. 368, No. 6489, 2020, pp. 395–400. https://doi.org/10.1126/science.aba9757.

Parr

Wolshon

Renne

Murray-Tuite

Kim

Traffic Impacts of the COVID-19 Pandemic: Statewide Analysis of Social Separation and Activity Restriction. Natural Hazards Review, Vol. 21, No. 3, 2020. https://doi.org/10.1061/(ASCE)NH.1527-6996.0000409.

De Vos

. The Effect of COVID-19 and Subsequent Social Distancing on Travel Behavior. Transportation Research Interdisciplinary Perspectives, Vol. 5, 2020. https://doi.org/10.1016/j.trip.2020.100121.

De Haas

Faber

Hamersma

How COVID-19 and the Dutch ‘Intelligent Lockdown’ Change Activities, Work and Travel Behaviour: Evidence from Longitudinal Data in the Netherlands. Transportation Research Interdisciplinary Perspectives, Vol. 6, 2020. https://doi.org/10.1016/j.trip.2020.100150.

10.

Bucsky

Modal Share Changes Due to COVID-19: The Case of Budapest. Transportation Research Interdisciplinary Perspectives, Vol. 8, 2020. https://doi.org/10.1016/j.trip.2020.100141.

11.

Hotle

Murray-Tuite

Singh

Influenza Risk Perception and Travel-Related Protection Behavior in the U.S.: Insights for the Aftermath of the COVID-19 Outbreak. Transportation Research Interdisciplinary Perspectives, Vol. 5, 2020. https://doi.org/10.1016/j.trip.2020.100127.

12.

Gendreau

Halting the Travels of Infectious Disease: Lessons Learned From the Ebola Epidemic. T.R. News, Vol. 299, 2015, pp. 40–42.

13.

Gardner

Fajardo

Travis Waller

Wang

Sarkar

Predictive Model for Air-Travel-Associated Dengue Infections in United States and Europe. Proc., 90th Annual Meeting of the Transportation Research Board, Washington, DC, 2011.

14.

National Academy of Sciences, Engineering, and Medicine. Preparing Airports for Communicable Diseases on Arriving Flights. The National Academies Press, Washington, DC, 2017. https://doi.org/10.17226/24880.

15.

National Academy of Sciences, Engineering, and Medicine. Airport Roles in Reducing Transmission of Communicable Diseases. The National Academies Press, Washington, DC, 2019. https://doi.org/10.17226/25367.

16.

National Academies of Sciences, Engineering, and Medicine. Research on the Transmission of Disease in Airports and on Aircraft. The National Academies Press, Washington, DC, 2010. https://doi.org/10.17226/22941.

17.

Sevilla

N. L.

Germs on a Plane: The Transmission and Risks of Airplane-Borne Diseases. Transportation Research Record: Journal of the Transportation Research Board, 2018. 2672(29): 93–102. https://doi.org/10.1177/0361198118799709.

18.

Chen

Gardner

Rey

Bilevel Optimization Model for the Development of Real-Time Strategies to Minimize Epidemic Spreading Risk in Air Traffic Networks. Transportation Research Record: Journal of the Transportation Research Board, 2016. 2569(1): 62–69. https://doi.org/10.3141/2569-07.

19.

Gardner

L. M.

Fajardo

Travis Waller

Inferring Infection-Spreading Links in an Air Traffic Network. Transportation Research Record: Journal of the Transportation Research Board, 2012. 2300(1): 13–21. https://doi.org/10.3141/2300-02.

20.

Zlojutro

Rey

Gardner

A Model to Optimize Air Travel Passengers Screening for Global Outbreak Response and Mitigation. Proc., 98th Annual Meeting of the Transportation Research Board, Washington, DC, 2017.

21.

Gold

Balal

Horak

Cheu

R. L.

Screening of Infectious Disease Among International Travelers and Airports. Proc., 97th Annual Meeting of the Transportation Research Board, Washington, DC, 2018.

22.

Bota

Gardner

L. M.

Khani

Modeling the Spread of Infection in Public Transit Networks: A Decision-Support Tool for Outbreaks Planning and Control. Proc., 96th Annual Meeting of the Transportation Research Board, Washington, DC, 2017.

23.

National Academy of Sciences, Engineering, and Medicine. A Guide for Public Transportation Pandemic Planning and Response. The National Academies Press, Washington, DC, 2014. https://doi.org/10.17226/22414.

24.

Gardner

L. M.

Modeling the Risk of Infectious Disease Spread Through Transport Systems: Emerging Research. T.R. News, Vol. 299, 2015, p. 43.

25.

Centers for Disease Control and Prevention. CDC COVID Data Tracker. https://www.cdc.gov/covid-data-tracker/#cases. Accessed July 4, 2020.

26.

Institute for Health Metrics and Evaluation. COVID-19 Projections. https://covid19.healthdata.org/united-states-of-america/hawaii. Accessed July 4, 2020.

27.

COVID Act Now. America’s COVID Warning System. https://covid19.healthdata.org/united-states-of-america/hawaii. Accessed July 4, 2020.

28.

Massachusetts General Hospital. COVID-19 Simulator. https://analytics-tools.shinyapps.io/covid19simulator/. Accessed July 4, 2020.

29.

Massachusetts Institute of Technology Policy Alliance. COVID-19 Infection Trajectories. covidalliance.com. Accessed July 4, 2020.

30.

Kim

Pant

Yamashita

Managing Uncertainty: Lessons from Volcanic Lava Disruption of Transportation Infrastructure in Puna, Hawaii. Journal of Emergency Management, Vol. 16, No. 1, 2018, pp. 29–40. https://doi.org/10.5055/jem.2018.0351.

31.

Kim

Pant

Yamashita

Ghimire

Analysis of Transportation Disruptions from Recent Flooding and Volcanic Disasters in Hawai’i. Transportation Research Record: Journal of the Transportation Research Board, 2019. 2673(2): 194–208: https://doi.org/10.1177/0361198118825460.

32.

Kim

Pant

Yamashita

Evacuation Planning for Plausible Worst Case Inundation Scenarios in Honolulu, Hawaii. Journal of Emergency Management, Vol. 13, No. 2, 2015, pp. 93–108. https://doi.org/10.5055/jem.2015.0223.

33.

Kim

Pant

Yamashita

Integrating Travel Demand Modeling and Flood Hazard Risk Analysis for Evacuation and Sheltering. International Journal of Disaster Risk Reduction, Vol. 31, 2018, pp. 1177–1186. https://doi.org/10.1016/j.ijdrr.2017.10.025.

34.

Kim

Yamashita

Ghimire

Burke

Morikawa

Kobayashi

Learning From Crisis: Transit Evacuation in Honolulu, Hawaii, After Tsunami Warnings. Transportation Research Record: Journal of the Transportation Research Board, 2013. 2376(1): 56–62: https://doi.org/10.3141/2376-07.

35.

Kim

Francis

Yamashita

Learning to Build Resilience into Transportation Systems. Transportation Research Record: Journal of the Transportation Research Board, 2018. 2672(1): 30–42. https://doi.org/10.1177/0361198118786622.

36.

Kim

Pant

Yamashita

Using National Household Travel Survey Data for the Assessment of Transportation System Vulnerabilities. Transportation Research Record: Journal of the Transportation Research Board, 2013. 2376(1): 71–80. https://doi.org/10.3141/2376-09.

37.

Hawaii Data Collaborative. Hawaii COVID-19 Data Resources. https://www.hawaiidata.org/covid19. Accessed July 4, 2020.

38.

Centers for Disease Control and Prevention. Symptoms of Coronavirus. https://www.cdc.gov/coronavirus/2019-ncov/symptoms-testing/symptoms.html. Accessed July 4, 2020.

39.

Urdan

T. C.

Statistics in Plain English. Lawrence Erlbaum Associates, Inc., Mahwah, NJ, 2005.

40.

Allison

P. D.

Logistic Regression Using SAS: Theory and Application, 2nd ed. SAS Institute, Cary, NC. 2012.

41.

Hosmer

D. W.

Lemeshow

Applied Logistic Regression, 2nd ed. John Wiley & Sons, Inc., New York, NY, 2000.

42.

Kim

Yamashita

Pant

Ghimire

Sustaining Seat Belt Use in a High-Use State. Transportation Research Record: Journal of the Transportation Research Board, 2014. 2425(1): 32–40. https://doi.org/10.3141/2425-05.

43.

Kim

Yamashita

E. Y.

Asleep at the Wheel: Spatial and Temporal Patterns of Fatigue-Related Crashes in Honolulu. Transportation Research Record: Journal of the Transportation Research Board, 2001. 1779(1): 46–53. https://doi.org/10.3141/1779-07.

44.

Kim

Pant

Yamashita

E. Y.

Hit-and-Run Crashes: Use of Rough Set Analysis With Logistic Regression to Capture Critical Attributes and Determinants. Transportation Research Record: Journal of the Transportation Research Board, 2008. 2083(1): 114–121: https://doi.org/10.3141/2083-13.

45.

Kim

Ghimire

Pant

Yamashita

Self-Reported Handheld Device Use While Driving. Accident Analysis and Prevention, Vol. 125, 2019, pp. 106–115. https://doi.org/10.1016/j.aap.2019.01.032.

46.

Getis

Ord

J. K.

The Analysis of Spatial Association by Use of Distance Statistics. In Perspectives on Spatial Data Analysis ( Anselin

Rey

, eds), Springer, Berlin, Heidelberg, 2010. pp. 127–145.

47.

Buitrago-Garcia

Egli-Gany

Counotte

M. J.

Hossmann

Imeri

Ipekci

A. M.

Salanti

Low

Occurrence and Transmission Potential of Asymtomatic and Presymtomatic SARS-CoV-2 Infections: A Living Systematic Review and Meta-Analysis. PLoS Medicine, Vol. 17, No. 9, 2020, P. e1003346. https://doi.org/10.1371/journal.pmed.1003346.

48.

Kim

The Lie Factor in Traffic Safety. Transportation Research Record: Journal of the Transportation Research Board, 1999. 1665(1): 141–146. https://doi.org/10.3141/1665-19.

49.

Kim

. Impacts of COVID-19 on Transportation: Summary and Synthesis of Interdisciplinary Research. Transportation Research Interdisciplinary Research, Vol. 9, 2021, Pp. 100305. https://doi.org/10.1016/j.trip.2021.100305.

50.

Kim

Yamashita

Ghimire

Factors Associated With Differences in Initial Pandemic Preparedness and Response: Findings From a National Survey in the United States. Transportation Research Interdisciplinary Perspectives, Vol. 11, 2021, 100430. https://doi.org/10.1016/j.trip.2021.100430.