Understanding the travel challenges and gaps for older adults during the COVID-19 outbreak: Insights from the New York City area

The COVID-19 pandemic has greatly impacted lifestyles and travel patterns, revealing existing societal and transportation gaps and introducing new challenges. In the context of an aging population, this study investigated how the travel behaviors of older adults (aged 60+) in New York City were affected by COVID-19, using an online survey and analyzing younger adult (aged 18–59) data for comparative analysis. The purpose of the study is to understand the pandemic's effects on older adults’ travel purpose and frequency, challenges faced during essential trips, and to identify potential policies to enhance their mobility during future crises. Descriptive analysis and Wilcoxon signed-rank tests were used to summarize the changes in employment status, trip purposes, transportation mode usage, and attitude regarding transportation systems before and during the outbreak and after the travel restrictions were lifted. A Natural Language Processing model, Gibbs Sampling Dirichlet Multinomial Mixture, was adopted to open-ended questions due to its advantage in extracting information from short text. The findings show differences between older and younger adults in telework and increased essential-purpose trips (e.g., medical visits) for older adults. The pandemic increased older adults’ concern about health, safety, comfort, prices when choosing travel mode, leading to reduced transit use and walking, increased driving, and limited bike use. To reduce travel burdens and maintain older adults' employment, targeted programs improving digital skills (telework, telehealth, telemedicine) are recommended. Additionally, safe, affordable, and accessible transportation alternatives are necessary to ensure mobility and essential trips for older adults, along with facilitation of walkable communities.


Introduction
The U.S. population is quickly agingaccording to the population estimates from the U.S. Census Bureau, the population aged 65 and older increased by 34.2% over the past decade and reached 55 million in 2020 (Bureau, 2020). Providing viable transportation options for an aging population is crucial. Access to transportation allows older adults to stay active and engaged in their communities, which contributes to their ability to age in place successfully. However, access to transportation is limited or challenging for some older adults. Previous studies have shown that 11.2 million older adults aged 65 and older in the U.S. have difficulty traveling outside of their house due to their disabilities (Brumbaugh, 2018). Additionally, around 80% of older Americans live in car-dependent communities, and 15.5 million older Americans have poor transit access (DeGood, 2011). Furthermore, changes in health status and function may force some older adults to retire from driving and transition to other community mobility options, such as public transit (Edwards et al., 2008). These systemic inequities and mobility disparities existed in transportation systems, and the outbreak of Coronavirus Disease 2019  has dramatically changed people's lives, potentially magnifying these challenges.
With the spread of the coronavirus, nonpharmaceutical interventions (NPIs), such as social distancing, school and non-essential business closures, and stay-at-home orders were introduced in many cities to restrict the movement of people to minimize the possibility of infection. As a result, reduced usage of public transit and private cars was observed, especially in the first few months after NPIs were implemented (Zuo et al., 2020). While NPIs are proven to be effective in decreasing the reproduction number (Rt) of COVID-19 (Bo et al., 2021), limited available options for transportation have made it increasingly challenging for older adults to navigate their community. Multiple studies have shown a more significant decline in public transportation use among older adults (age 65+) than younger adults during the pandemic (Cusack, 2021;Park and Cho, 2021). When public transportation was unavailable, younger adults were able to utilize alternative modes of transportation such as private vehicles, bicycles, and walking (Cusack, 2021). Conversely, older adults had a harder time finding substitute options because some had already ceased driving and were unable to ride bicycles or walk due to health conditions (MacLeod et al., 2014). They had to rely heavily on assistance from friends or family members, or they may have had to forgo travel altogether (Park and Cho, 2021). As a result, older adults experienced negative changes in physical, lifestyle, psychological, and financial aspects during the pandemic.
Some older adults are not only a vulnerable population from a health perspective, but they may also need a caregiver and generally demand more frequent doctor visits or medical needs compared to young adults (Roter, 2000). However, delayed or avoided medical care during the COVID-19 outbreak may have increased the risk of morbidity and mortality associated with non-COVID related chronic and acute health conditions (Czeisler et al., 2020). During the pandemic, the lifestyles of older adults were negatively impacted by the travel restrictions, resulting in decreased physical activity levels, increased sedentary behavior and malnutrition, and accelerated frailty and muscle weakness (Browne et al., 2020;Visser et al., 2020). While older adults who have access to online services (e.g., telework, online shopping) may overcome some mobility barriers (Hülür and Macdonald, 2020), older adults who are frail and are not online may struggle with the double burden of social exclusion (Seifert et al., 2021). In addition, the COVID-19 outbreak has also increased financial hardship among older adults, especially those who are still in the labor force compared to those who have already retired (Li and Mutchler, 2020). Therefore, investigating the behavior change and travel concerns of older adults, including telework's impacts on work trips, is crucial to help the government, transportation agencies, and local communities to understand the mobility needs and provide appropriate conditions for using existing transportation systems and offer alternative transportation modes for older adults. It should be noted that COVID-19 has impacted older adults differently and more significantly, and their behavioral changes and travel concerns should be identified separately from those of younger adults (Browne et al., 2020;Li and Mutchler, 2020;Park and Cho, 2021;Parlapani et al., 2021;Visser et al., 2020). While the literature well documents the effects of a pandemic on older adults' activities, only a few studies have investigated changes in mode preferences (Dadashzadeh et al., 2022) or solely concentrated on limited travel modes, such as changes in subway use (Park and Cho, 2021). More importantly, observations of older adults living in urban areas, such as New York City (NYC), can contribute more insights, as urban areas are often crowded, and are more impacted by travel restrictions (Gimie et al., 2022).
Employing a comparative analysis approach between older and younger adults and using NYC as a case study, this study aims to investigate older adults' travel motivations and mode usage during COVID-19 in an urban area, discern their concerns during essential trips, and identify policy implications to improve their mobility in future pandemics. The study's findings can contribute to understanding the travel challenges and gaps for older adults in urban areas during the COVID-19 outbreak. The paper's contributions are summarized as follows: • The descriptive analysis of the survey reveals differences between older adults and younger adults during COVID-19 in terms of 1) employment status and telework ability, 2) trip purposes, 3) travel mode and frequency, and 4) attitude and comfort regarding transportation systems. Additionally, the Wilcoxon signed-rank test was used to examine within-group differences in travel mode frequency. • A topic modeling model, namely Gibbs Sampling Dirichlet Multinomial Mixture (GSDMM), which targets detecting topics in the shorter text, was adopted to extract key topics from open-ended questions. • The case study of an urban area, NYC, provides a unique opportunity to explore how older adults with relatively more travel options were reacting to the inconveniences due to the COVID-19 pandemic, compared to rural areas which are usually more car-driven. The findings can provide insights for other cities in the world with a dense population and access to multimodal transportation modes. • The findings of this study also provide several policy implications that can help increase older adults' mobility, expand their transportation options, and enhance the connection between multiple policies for older adults, including health, travel, and economic assistance during the COVID-19 outbreak and similar crisis in the future.

Methodology
A multifaceted approach was used to analyze the survey data. Firstly, the quantitative responses to the survey were analyzed using descriptive analysis based on the following aspects: 1) employment status and teleworkability, 2) trip purpose during COVID-19, 3) changes in travel mode and frequency, and 4) attitude and comfort regarding transportation systems. Next, the Wilcoxon signed-rank test (Wilcoxon, 1992) was implemented to compare the frequencies of travel modes prior to the pandemic and following the city's reopening. Finally, a Natural Language Processing (NLP) topic modeling, GSDMM, was applied to cluster two open-ended questions about trip concerns and inconvenience during stay-at-home orders.

Data collection and sampling
After receiving approval by the Institutional Review Board (IRB), an online survey was administered by the C2SMART center at New York University (NYU) between July and October 2020 to assess mobility needs and changes due to COVID-19. The survey contains 36 questions, including 34 multiple-choice, matrix table and rank order questions, as well as two open-ended questions. The survey questions can be viewed at: https://c2smart.engineering.nyu.edu/covid-survey/. A combination of quota and convenience (i.e., snowball sampling) techniques were employed in this study. The survey took about 10 min online and it was distributed via different social media handles, university webpages, newsletters, and outreach as well as to a population group of older populations (aged 60+) through Qualtrics panel data service. A total of 2,022 responses were collected nationwide.
The focus of this paper is investigating older adults' travel behaviors in an urban environment. Therefore, a subset of the survey responses from 699 participants who reside within NYC and one of its close neighbors, Jersey City, was utilized as a case study to ensure similar available travel options for all respondents. This subset comprised 274 older adults aged 60 and above and 425 younger adults aged 18-59. It is important to note that the study primarily focused on older adults, and the use of younger adults' responses was solely for comparison purposes.
The survey was pilot tested by ten student researchers for its flow and clarity.
Besides collecting demographic information such as age, gender, household income, and employment, the stated-preference survey focused on the impact of COVID-19 on a wide range of aspects of respondents' daily life and travel behaviors, including teleworkability, access, usage of 13 different travel modes, travel frequency and purpose, health concern and comfort level of using different travel modes, and changes in caregiver/caretaker trips. Furthermore, travel concerns and inconveniences faced for essential trips were studied through ranking and open-ended questions.
Longitudinally, the survey was divided into three time periods: 1) Period 1 (Before the pandemic): before the pandemic (before March 2020), 2) Period 2 (During NY PAUSE): after the stay-at-home orders were issued and before the reopening of cities (After March 2020 and before June 8, 2020), and 3) Period 3 (After reopening): after reopening of the cities (after June 8, 2020, to October 2020 [the Survey end date]). This separation allows for an understanding of people's perceptions and mobility needs during the initial phases of COVID-19-induced travel restrictions. This is a longitudinal study that follows changes in individuals over time, and the respondents for each period are the same. Fig. 1 shows the geographic distribution of respondents and the number of responses by geographic area.
The two population groups ranged from diverse employment status, household income, and car/bike ownership. The breakdown of each demographic factor is presented in Table 1. However, the overrepresentation of students (39.2%) in the younger adult data may introduce bias, particularly for commute behavior, as the transition to online classes at the start of the pandemic would have eliminated the need to commute almost entirely. To achieve a more balanced and representative data distribution, we employed a stratified sampling strategy, using the random undersampling (RUS) method, to decrease the proportion of student samples. RUS is a non-heuristic method that randomly removes majority class instances until the target class distribution is reached (Batista et al., 2004). It has been demonstrated to enhance the performance of classification and regression tasks compared to not using any data sampling (Branco et al., 2016). Following the resampling process, the percentage of student respondents in the younger adult group aligned with the actual percentage of students in the NYC population (approximately 20.2% for those under 60 years of age in NYC) (Domanico, 2020;U.S. Census Bureau, 2019). The resampled younger adult data was used for all subsequent analyses and comparisons. The breakdown of resampled younger adult group is presented in Table 1.
The gender breakdown in both population groups shows a good representation of our sample when comparing the demographics of our survey to the overall NYC population. Based on the NYC department of health (Greer et al., 2019) and U.S. Census (U.S. Census Bueau, 2022), 40% of older New Yorkers identify as men and 60% as women while about 48% of overall New Yorkers identify as men and 52% as women. The survey results indicate that the employment status of older adults was well-represented, with 40.8% of them earning income from work from the survey, which is comparable to the 38.3% working older adults reported by (Stringer, 2017). The findings from the table also show a much higher percentage of older adults (43.8%) owning a car compared to younger adults (27.0%), but more younger adults purchased a new bike or bike share membership since the start of the pandemic.

Descriptive analysis and the Wilcoxon signed-rank test
Firstly, descriptive analysis is used to effectively showcase the primary features of the findings derived from the survey data. This involves the utilization of summary tables, frequency counts, basic percentages, and various graphical representations such as Sankey charts, bar charts, and radar charts. In conjunction with descriptive analysis, the Wilcoxon signed-rank test (Wilcoxon, 1992) is implemented to compare the frequencies of travel modes prior to the pandemic and following the city's reopening, as reported by the same respondents. This particular method proves advantageous when addressing non-normal distributions or limited sample sizes (Wilcoxon, 1992).
Additionally, our survey data consists of paired observations of travel mode choices, which include many zero differences (pairs of observations with equal values) indicating that individuals did not change their travel frequency for certain modes during periods 1 and 3. This is possibly due to respondents' utilization of only a limited number of modes out of the 13 alternatives provided. To capture the information inherent in the zero differences and enhance the test's statistical power, the z-split method (Pratt, 1959) was applied to segregate positive and negative differences into two sets. The test statistic is calculated as the summation of the ranks of the positive differences minus the summation of the negative differences, divided by the standard deviation of the ranks. The statistical significance of the test statistic is then assessed by comparing it to a critical value derived from a standard normal distribution.

Gibbs sampling Dirichlet Multinomial Mixture (GSDMM)
Dirichlet Multinomial Mixture (DMM) (Nigam et al., 2000) is a popular generative modeling approach for information extraction and classification of documents. It selects a topic according to the topic weight, then generates documents from the selected topic using the conditional probability of the document given the topic. DMM assumes that each document corresponds to a single topic, makes it advantageous when processing short texts. Gibbs Sampling (Geman and Geman, 1984) is an algorithm for continuous sampling of conditional distributions of variables. It defines a hidden states sequence (a Markov chain) in the space of possible variables assignments, and a state sequence can only be transited from a state sequence by changing the state at any one position n, and the distribution over these transitions is following: where o is the observed sequence, s is the sequence of states and s − n means all states except s n . Eq. (1) shows that the transition probability of the hidden states sequence is the conditional distribution of the label at the position given the rest of the sequence. This feature has made the computation easy and fast. A variety of Gibbs sampling algorithms, called collapsed Gibbs sampling, are used to obtain the topic of the document. This algorithm represents the distribution through the observations. Topic assignment of a word is considered a random variable because each word is assumed to be generated from a single topic. Document assignment can be seen as the popularity of each topic in a document. Let the N d,i be the number of words taking topic i in document d. More popular topics will have larger N. Thus, we can estimate the topic distribution by: where α ∕ = 0. The probability of a word in a topic can be estimated as: where V k,w is the times that topic k uses the word w, and λ is the Dirichlet parameter for the topic distribution. By sampling the topic assignment for the model, the topics and per-document topic proportion are integrated. Thus, we have: Given the short nature of our open-ended question responses with an average word count of about five words per response, we selected GSDMM for its ability to analyze short text with limited context data. Compared with other commonly used topic modeling methods such as Latent Dirichlet Allocation (LDA), GSDMM is capable of extracting coherent, interpretable topics and delivers stable results with less sensitivity to hyperparameter choices. More details about GSDMM can be found in (Yin and Wang, 2014). For the document pre-processing and algorithm implementation, the Python packages "Gensim" (Rehurek and Sojka, 2011) and "gsdmm" (Walker, 2021) were used.

Period 1 (Before the pandemic) versus Period 3 (After reopening)
First, survey data from Period 1 and Period 3 were compared to help us understand the changes that occurred before the pandemic and after the reopening of the city when travel restrictions were lifted.

Employment status and teleworkability
Based on the survey results, about 40% of older adults were employed either full-time, part-time, or self-employed. Out of the older respondents, 41% and 19% stated that all or some of their job responsibilities could be performed remotely, respectively, while approximately 40% mentioned that teleworking was not possible for them. Fig. 2 illustrates the changes in their working status before the pandemic and after the reopening of the city.
Though the survey questions did not ask if the respondents lost jobs due to COVID, the evidence of 39% of older adults answering "does not apply" after the outbreak implied possible job loss due to COVID, or a change in work status due to other reasons such as taking early retirement (Faria e Castro, 2021). In Period 3 (after reopening), the percentage of younger adults who worked or attended class remotely or both in person and remotely was higher (67%) compared to older adults (44%) (Fig. 3).
It is worth noting that almost all younger adults (94%) reported that they were able to telework. The disparity in teleworkability between younger adults (94%) and older adults (60%) may be attributed to the differences in digital skills. This observation is consistent with the fact that despite increasing rates of internet use, older adults tend to have lower rates of digital skill utilization compared to other age groups (Hülür and Macdonald, 2020).

Changes in travel mode and frequency
The study investigated the primary mode of transportation used by both population groups for commute and non-commute trips before the COVID-19 pandemic (Table 2). Subway and walk were the top two travel modes for both groups for commute trips before the pandemic. However, the use of private cars (driving alone; 13%) and buses (12%) was higher among older adults compared to younger adults. Albeit small, younger adults also used bikes and bike sharing (4%) as their top choices, which was not observed in old adults. For non-commute trips before the pandemic, bus was the primary travel mode for over one-fourth of the older adult respondents (27%), followed by subway (14%), walk (10%), private car (driving alone, 6%), and private car (with someone else, 6%). Younger adults had a quite different trend in their top travel modes for non-commute trips, where walk ranked the highest (29%), followed by transit (bus 16% and subway 15%), rideshare alone (12%) and bike (7%).
Based on the primary mode of transportation usage before the pandemic, we investigated how the mode of transportation use and frequency changed during the pandemic for all of the transportation modes surveyed. Figs. 4 and 5 show the descriptive analysis on how older respondents shifted in the frequency of the top 5 travel modes for commute and non-commute trips. Table 3 and Table 4 present statistics using the Wilcoxon signed-rank test that examined the differences in all travel mode choices among older adults and younger adults for work and non-work trips. The p-value represents the test's significance level, while 'Avg. change' indicates the average difference between prepandemic and post-reopening choice scores for each traffic mode. A negative value implies a reduction in the mode's frequency, while a positive value denotes an increase.
For commute trips taken by older adults, usage across all the top five modes decreased during the stay-at-home order and after the reopening of the city. Public transit, including subway and bus trips, was significantly reduced in terms of frequency of use due to COVID-19, which is consistent with the observed ridership trends in the C2SMART Mobility Dashboard (Zuo et al., 2020) and MTA day-by-day ridership numbers (MTA, 2020). Walking and private car (with someone else) saw relatively smaller reductions. Among the top five favored travel modes, private car (driving alone) was found to be the least impacted travel mode for older adults.
Similarly, for non-commute trips taken by older adults, ridesharing with someone else, train, subway and bus trips reduced significantly. Although the overall changes in private car (with someone else) usage were not statistically significant (Table 3), the use frequency of this mode slightly increased (from 16% to 20% for users who traveled 'several times a day' and 'few times a week') (Fig. 5), indicating a rising need for care recipient trips during and after the reopening of the city.
Figs. 6 and 7 show how younger adult respondents shifted in terms of the frequency of these travel modes for commute and non-commute trips. For commute trips taken by younger adults, usage across rideshare (with or without someone), transit (subway, train, bus) and walking decreased significantly in period 3 (after reopening), compared with pre-pandemic periods, as would be expected considering the effect of telework, reduced transit services and wariness of the spread of the coronavirus. The only mode that saw a significant increase is personal bike. The same increase trend in personal bike usage was observed for non-commute trips, while the walk, bus, subway, and rideshare saw a reduction in use frequency. A significant increase is private car usage (driving alone) was also observed.
It was expected yet important to find that although cycling became a popular alternative mode during COVID-19 (Cusack, 2021; Lei and  Ozbay, 2021; Shaer and Haghshenas, 2021a, b), and was shown to have a positive impact on the health of older cyclists (Pucher et al., 2010), the percentage of older respondents who purchased a new bike or a bike share membership was lower (~4%) than younger adults (~10%) ( Table 1). Moreover, the survey results also revealed that the increased use of bikes for older adults was not as significant as the younger population (7% vs. 33% for commute trips and 15% vs. 65% for noncommute trips) revealing that there are still certain barriers for riding a bike for older adults living in NYC. Barriers to older adults' bike usage may include health-related limitations (e.g., physical disabilities, fear of falling off bikes, and physical challenges while wearing masks due to low cardiorespiratory endurance) and systemic limitations (e.g., limited bike share coverage, non-discounted costs, and inconvenience due to docked systems providing station-to-station service instead of door-todoor service). These potential barriers are further confirmed by the survey finding that only 2% of the surveyed older adults physically had access to a personal bike, and none of them claimed that they had access to bike sharing systems.

Attitude and comfort regarding transportation systems
Survey respondents were asked to rank five attributes related to attitude (reliability, trip duration, comfort, safety, price, and health) in order of how they affected their choice of travel mode before the pandemic and after the reopening of the city (Fig. 8). Overall, health, safety, and reliability were found to be the top three attributes that affected the choice of travel mode for both population groups after the reopening of the city. Unsurprisingly, health, which was the least concerned attribute for travel before the pandemic, saw the sharpest Table 2 The top 5 travel modes that were used by older adults and younger adults for commute and non-commute trips before the COVID-19 pandemic.  Fig. 4. How often older adults use selected travel modes for commute trips, Period 1 (Before the pandemic) versus Period 3 (After reopening). increase in its importance in determining the choice of travel mode after the period. Younger adults had an increased concern in health, safety, and comfort after the reopening of the city, while they treated reliability, trip duration, and price as almost the same in the before-and-after periods. Older adults had similar increasing changes in health, safety, and comfort, however, unlike younger adults, they also had increasing concern about price. The rationale behind this could be that a certain portion of older adults are also low-income. We further investigated the perceived comfort of the older respondents when they use these travel modes for work and leisure trips post COVID-19 stay-at-home orders. The survey results revealed that older adults are most comfortable with driving alone and walking, followed by driving with someone else, personal bicycle, and ride share alone. Additionally, most of the older respondents found many tasks become more challenging due to COVID-19, including going to get groceries/essentials (67%), visiting the hospital/doctor (70%), eating out (93%), running errands (64%), visiting family/friends (89%), having friends/family over (91%), taking public transit (84%), and flying in an airplane (95%). The only two exceptions are getting medicines and leisure walking. 73% and 54% of the older respondents thought they remained the same compared with pre-pandemic periods, possibly because both are within walkable distances and require less or minimal social interactions.

Period 2 (During NY PAUSE)
Subsequently, we also investigated older adults' purposes of trips, as well as the biggest concerns and inconveniences experienced while traveling during the NY on PAUSE period (Period 2), with the intention of understanding the essential travel requirements of older adults and identifying the distinctions when compared with younger adults. This analysis seeks to help us identify which trips cannot be avoided and where the gaps exist. Fig. 9 illustrates that the top 3 reasons for travel during stay-at-home orders for older adults are: 1) trips to the grocery stores (63%), 2) trips to the pharmacy or drugstore (49%), and 3) medical visits (40%). The percentage of respondents is higher in older adults for medical visits (approximately 1.6 times more than younger adults), trips to the pharmacy, and trips to non-grocery, non-pharmacy shopping. Over 25% of older adults indicated that they need to see the doctor for routine checkups or for health conditions at least once a month, and half of them had to see doctors several times a week or had weekly/bi-weekly doctor appointments. For younger adults, the top 3 reasons for travel during stay-at-home order are 1) trips to the grocery stores, 2) outdoor exercise/recreation, and 3) trips to the pharmacy or drugstore. While other studies (Roe et al., 2021) showed evidence that older adults might take fewer trips during COVID, our survey findings reflected that they may travel for more essential purposes.

Biggest concerns and inconvenience about traveling for essential trips
The GSDMM was used to analyze the open-ended question responses regarding the biggest concerns about traveling during the stay-at-home orders (Period 2, during NY PAUSE). The resulting topic distributions and word distributions for each topic are presented in Fig. 10. For older adults, the most concerning issue was mask wearing and social distancing, followed by concerns about catching and being infected with the virus. The third topic focused on health and safety concerns, especially when riding buses. For younger adults, the identified topics were similar but ranked differently, with the top concerns being virus catching, health and safety concerns and traveling using subway, and mask wearing.
The model results for the second open-ended question asking about the biggest inconvenience while going for essential trips during the time of the pandemic are illustrated in Fig. 11. For older adults, the most Table 3 The Wilcoxon signed-rank test for older adults' commute and non-commute trips, period 1 (before the pandemic) versus period 3 (After reopening). Note: ***p < 0.01, **p < 0.05, *p < 0.1. Note: ***p < 0.01, **p < 0.05, *p < 0.1.
important inconvenience was mask wearing and social distancing, followed by long waiting lines at stores and groceries and concerns about the lack of product in stores. Although the frequency of the third topic is not as high as topic 1 and 2 (Fig. 11), this topic was unique in its relation to work trips (such as not being able to work) and fears about crowds/ travel. For younger adults, their top inconvenience is similar as older adults, which is related to mask wearing and social distancing. The second is long waiting lines at stores and groceries, yet unlike older adults, they didn't worry about the lack of supplies in stores. The third topic is the use of the subway system. These findings are consistent with Fig. 8 on attitude factors affecting the choice of travel mode.

Practice and policy implications
While the COVID-19 pandemic has dramatically changed lifestyles and travel patterns all around the world, it provided a unique opportunity to revisit gaps that existed in our current society and transportation systems as well as understand new and unforeseen challenges that were introduced by the pandemic. Based on the findings from this survey-based study, several practice and policy implications are summarized as follows: First, the digitalization of society (e.g., digitizing social networks) opened an avenue to stay engaged with social lives and other daily   activities (Hülür and Macdonald, 2020), even during a pandemic. Nevertheless, this could be problematic for some older adults because while many older adults are technologically savvy, some have had difficulties in adapting to new technologies such as video conferencing or online shopping (Seifert et al., 2021). Although it may be due to the types of jobs held by survey participants, the disparity in telework for older and younger adults (60% vs. 94%) observed in the survey may provide evidence of a "digital divide" gap for older workers during the pandemic. Providing policies and programs for older adults that target improving their digital skills in telework, telehealth, and telemedicine (e.g., virtual urgent care) may still be beneficial (Pirhonen et al., 2020). A community-based setting that is a viable option for offering hands on training in telehealth and telemedicine platforms is local Older Adult Centers (OAC's, formerly known as senior centers), and community colleges may be a low-cost resource for digital skills training outside the workplace. More importantly, having the ability to choose whether and how much to telework, telehealth, and telemedicine may also be crucial for reducing travel concerns (i.e., commute and essential trip burdens) for older adults during a similar future crisis. The findings from the open-ended questions also emphasize the importance of focusing on training older adults in emergency preparedness, which includes ensuring they have a supply of essential supplies at home.
Secondly, although cycling is often an affordable option and has physical benefits for older adults (Pucher et al., 2010), increasing the use of bikes among older adults should not be suggested blindly without the consideration of two factors: 1) heterogeneity of travel needs in different regions, and 2) adequate practical plans to minimize the physical challenge. For example, survey results from this study indicated that a low percentage of older respondents who purchased a new bike (4.3%) or a bike share membership (4.7%), as well as a low increase in bike use frequency (e.g., 7% increase for commuter trips), which are not consistent with the findings observed from an international study from Iran for older adults that showed 140% increase in bike usage time (Shaer and Haghshenas, 2021b). This inconsistency indicates possible heterogeneity in different regions and policies (i.e., the Iranian government partially restricted private vehicles use during COVID). Moreover, practical plans to overcome systematic limitations of bike sharing systems need to be provided beforehand as cycling may still be physically challenging for some older adults. For instance, the docked bike sharing system in NYC and NJ, Citi Bike, offers station-to-station services  instead of door-to-door services. It may not be an optimal first-/last-mile transport solution for older adults and oftentimes, extra travel may be required to find a vacant bike rack if the nearest rack is all occupied. Increasing the number of bikes and improving infrastructure and road safety for cyclists is necessary to encourage cycling among the growing older adult population. Additionally, unlike other public transportations, Citi Bike does not have a reduced fare rate for older adults and can be twice or over more expensive than transit for a single ride.
However, it is important to note that although the bike usage increase is limited, our survey results also indicated that riding a personal bike is the top 3 travel modes that older adults are comfortable with in terms of coronavirus contraction concerns. Therefore, providing adaptive bikes as a bike sharing option (e.g., power assist electric bicycles) to older adults may be a potential alternative. Another finding is that many older adults used private cars as one of the primary modes of transportation, yet with aging, the number of older adults who cease driving will increase, and accordingly, adequate alternative transport is needed.
Thirdly, while the overall changes in the usage of private cars (with someone else) for non-commute trips did not exhibit statistical significance, the use frequency of this mode for "several times a day" and "few times a week" slightly increased, indicating a rising need for care recipient trips during and after the reopening of the city. Based on the survey results, trip purposes of older adults were relatively more essential (e.g., for food or medical visits). It is therefore important to prioritize strategies for ensuring old adults' trips to medical visits as delayed or avoided medical care can increase morbidity and mortality for even non-COVID related health conditions (Czeisler et al., 2020). Enabling more alternative transportation options for essential travel for older adults should also be considered for a future similar health crisis. For example, older adults have access to discounted taxi fares and taxi coupons in some suburban communities (State of Nevada, 2022), which can be a good alternative in urban environments during a pandemic. Developing a discounted program based on some existing ridesharing services, such as UberHealth (which allows healthcare organizations to arrange rides on behalf of others), can help reduce cost concerns and offer another possible alternative for older adults during a pandemic. Moreover, survey results indicated that tasks with walkable distance (i.e., getting medicines and leisure walking) are relatively less challenging for older adults and walking for non-work purposes was less affected by the pandemic than other modes of transportation. Therefore, integrating the health and social services already available and reducing environmental hazards and barriers to walking outdoors in the community in neighborhoods with high concentrations of older adults (i.e., Naturally Occurring Retirement Communities (NORCs) (Masotti et al., 2006)) to facilitate more walkable communities can be beneficial both in the short-and long run.
Survey findings also suggest that for a similar future crisis like COVID, there is a need to strengthen the connection between multiple policies for older adults, including health, travel, and economic assistance. This is because some older adults may be more financially vulnerable and sensitive to social isolation. This aligns with a previous study that found low-income older adults can be particularly vulnerable to food insecurity attributable to social isolation, lack of transportation, disability, and poor health (Vilar-Compte et al., 2017). Survey results also implied that older adults were concerned more about transportation service prices, and experienced possible unemployment due to COVID. Therefore, combining mobility needs with other health and economic concerns can be beneficial when developing policies for older adults.
Finally, feedback from older adults needs to be collected periodically or after each major change point during different stages of a similar future crisis. For instance, the rapid changes in the COVID-19 pandemic might have reduced the stability of the findings from surveys, and therefore, multiple follow-up surveys might be needed to capture the effect of such changes. Potential communication and outreach gaps need to be considered, as older adults may not be regularly accessing remote communication channels (e.g., social media, web sites, etc.) to provide their feedback.

Conclusions
With a survey-based dataset, this study investigated the impact of COVID-19 on adults in urban areas, especially older adults. Compared to rural areas, which are usually more car-driven, the focus of urban areas provides a unique opportunity to explore how older adults with relatively more travel options were reacting to the inconveniences due to the COVID-19 pandemic, thus providing insights for other densely populated cities worldwide with access to multimodal transportation modes.
The descriptive analysis revealed concerns about changes in employment status (e.g., possible job loss), which may be due to disparities in older adults' ability to telework. During the pandemic, older adults took more essential trips, including grocery store visits, pharmacy or drugstore runs, and medical appointments, than younger adults. Wilcoxon signed-rank tests showed that public transit usage was significantly reduced for both commute and non-commute trips for older adults. The fact that the increased use of bikes among older adults was not as significant as among younger individuals suggests certain transportation barriers in adopting micromobility for older adults living in NYC, potentially due to health-related, physically challenging, and financial limitations.
When choosing a travel mode, both younger and older adults had increased concern about their health, safety, and comfort, especially, older adults also had an increased concern about affordability. Topic modeling results found the three biggest travel concerns for older adults were: mask wearing/social distancing; virus catching/infection; and health and safety concerns while using bus. The three biggest travel inconveniences for older adults were: mask wearing/social distancing; long waiting lines at stores and groceries/lack of supplies in stores; and concerns about not being able to commute to work. These findings can be used to attract attention to several policy solutions, such as offering programs targeted at enhancing digital skills for telework, telehealth, and telemedicine, facilitating walkable communities, enhancing the connection between multiple policies for older adults, including health, travel, and economic assistance, and providing safe, affordable, and accessible transportation alternatives, such as discounted taxis, for older adults to ensure their mobility and essential purpose trips.
The current study has a few limitations. It was conducted before the introduction of vaccinations, and the situation may have changed since then. Additionally, as this study is retrospective, recall bias may influence the results. Although some demographic information was collected, the health and functional status of participants is unknown. Further, self-selection bias may exist by using an online survey as older adults who are less technologically savvy or who do not have access to the internet may be excluded from the sample. Online surveys may also restrict the depth and breadth of the data collected as they typically have a limited number of questions and response options. Future research will focus on a more representative sample, including both online questionnaires and offline interviews, and conduct a follow-up survey with the same respondents to measure work and travel conditions now, as compared to what they were during the first wave.