Abstract
Introduction:
Homelessness is a public health problem, and persons experiencing homelessness are a vulnerable population. Estimates of the number of persons experiencing homelessness inform funding allocations and services planning and directly determine the ability of a community to intervene effectively in homelessness. The point-in-time (PIT) count presents a logistical problem in large urban areas, particularly those covering a vast geographical area.
Materials and Methods:
Working together, academia, local government, and community organizations improved the methodology for the count. Specific enhancements include use of incident command system (ICS), increased number of staging areas/teams, specialized outreach and Special Weapons and Tactics teams, and day-after surveying to collect demographic information.
Results:
This collaboration and enhanced methodology resulted in a more accurate estimate of the number of persons experiencing homelessness and allowed comparison of findings for 4 years. While initial results showed an increase due to improved counting, the number of persons experiencing homelessness counted for the subsequent years showed significant decrease during the same time period as a “housing first” campaign was implemented. The collaboration also built capacity in each sector: The health department used ICS as a training opportunity; the academics enhanced their community health efforts; the service sector was taught and implemented more rigorous quantitative methods; and the community was exposed to public health as a pragmatic and effective discipline.
Practical Implications:
Improvements made to increase the reliability of the PIT count can be adapted for use in other jurisdictions, leading to improved counts and better evaluation of progress in ending homelessness.
Keywords
Introduction
Estimates of the number of persons experiencing homelessness inform funding allocations and services planning and so directly determine the ability of a community to intervene effectively in homelessness. They also provide evidence of success or failure in these endeavors. These counts are mandated by the Department of Housing and Urban Development (HUD) to be conducted by continuum of care (CoC) grantees annually on sheltered and biannually on unsheltered populations in their geographic area served (Hopper, 1992). They tend to be relegated to volunteers under the general direction of homeless services providers/community development personnel (who often have minimal expertise in statistical models, sampling, and epidemiology) or consultants (who may have quantitative proficiency, but lack an understanding of the exigencies of homelessness). There are also inherent difficulties in getting an accurate assessment of this population (U.S. Department of Housing and Urban Development, 2012), and so point-in-time (PIT) counts across communities are conducted using methods of variable validity and reliability and thus variable accuracy.
Each urban area has its own solution to its unique challenges when counting the unsheltered in the PIT. The jurisdictions of New York City, New Orleans, Washington, DC, and Boston, for example, cover small geographic areas and so send out walking teams to interview all unsheltered persons experiencing homelessness encounters (HOPE 2015 : The NYC Street Survey, 2014; U.S. Interagency Council on Homelessness, 2013). Los Angeles uses a combination approach sending out walking teams and cars, but volunteers who are driving are expected to get out and interview any person experiencing homelessness they see (2015 Greater LA Homeless Count, 2015). Los Angeles also has received special permission from HUD to spread the count over a 3-day period. San Antonio uses both walking and driving teams (Haven for Hope, 2013). Some cities, such as Phoenix, use a sampling method to estimate the population of those experiencing homelessness (Maricopa County, 2014).
To overcome the challenges to conducting the PIT in the Houston area, we created a multisector collaboration between academia, government, and community agencies to employ community health best practices, epidemiologic methods, rigorous training, and stakeholder capacity building. We applied an incident command system (ICS) model (used by police and fire for decades and recently adopted by public health preparedness) to generate a valid, reliable enumeration. A plant capture method of statistically estimating undercounting used in other disciplines such as wildlife management was attempted in 2011 and 2012 but discontinued in 2013 due to logistical problems with the implementation.
The collaboration and processes were found to build capacity in each of the member sectors: The health department used ICS as a training opportunity; the academics enhanced their community health efforts; the service sector was taught and implemented more rigorous quantitative methods; and the community was exposed to public health as a pragmatic and effective discipline. We improved the completeness of the PIT by covering more of the jurisdiction’s large geographical area, more accurately identifying persons experiencing homelessness who were unsheltered, and canvassing the number of patrons at all emergency shelters, not just those using the Homeless Management Information System (HMIS).
Statement of Problem
Jurisdictions that receive funding from the U.S. Department of Housing and Urban Development are required to conduct regular counts of all persons experiencing homelessness, sheltered and unsheltered, and collect certain demographic data on those persons. Although the definition of homeless varies from agency to agency, for the purposes of this PIT count, homeless individuals are those staying in emergency shelter, transitional housing, or safe haven with beds dedicated for those experiencing homelessness or those persons who are unsheltered (i.e., staying in a place not meant for human habitation; U.S. Department of Housing and Urban Development, 2013b). The PIT count must be held during one 24-hour period during the last 10 calendar days of January. In large, geographically dispersed, urban areas such as Houston, Texas, this poses a logistical problem. The City of Houston, Harris County outside the City of Houston, and neighboring Fort Bend County consist of 2,565 square miles (U.S. Department of Commerce, 2014a, 2014b), and Harris County by itself covers a larger area than the state of Rhode Island. Figure 1 shows the challenge inherent in covering this huge geographic area in one night. Because individuals experiencing homelessness cannot be individually interviewed due to time constraints, methods must be carefully devised to eliminate double counting of individuals and to collect requisite demographic information. Methodology that had been used in Houston prior to 2011, although having some appropriate features, did not result in an optimal count, and so new methodology was developed and deployed for the 2011–2014 counts (Houston/Harris County performs a yearly count on both sheltered and unsheltered populations.)
Comparison of Houston geographical area with other major U.S. cities (courtesy of Knudson, LP, www.knudsonlp.com).
New and Enhanced Methodologies Employed
Partnerships
In December 2010, a coalition of organizations was convened to develop a better way to count persons experiencing homelessness in the Houston metropolitan area. These organizations included the lead agency coordinating community responses to the problem of homelessness and responsible for the CoC (Coalition for the Homeless of Houston/Harris County [CFTH]), local government (Houston Department of Health and Human Services [HDHHS]), and academia (the University of Texas School of Public Health [UTSPH]). This coalition of nonprofits, local government, and academia represented the first enhancement in the methodology as each was able to contribute different skills, abilities, and resources to the PIT count. One of us, an epidemiologist at UTSPH, provided oversight to the methodology with input from the other two agencies.
In addition to these major partners, there was a concerted effort to engage other community partners such as CoC housing providers, service providers, schools, faith-based community, students (Rice University, University of Houston, and UTSPH), veterans’ groups, and consumers. Specifically, in regard to consumers, a Consumer Advisory Corps (Coalition for the Homeless, 2013) was formed to provide a voice and a bridge of communication to network and share information between the homeless community, CoC, and CFTH. The Stand Up and Be Counted Corps, a subcommittee of the Consumer Advisory Corps, was formed to assist with the PIT count and funded by a grant from the United Way of Greater Houston and is composed of both formerly homeless persons and persons who are currently experiencing homelessness.
Incident command system
The first thing the oversight committee did was to institute an ICS to organize the PIT count. The ICS has been used by police and fire for decades and has been adopted by homeland security and public health preparedness services (U.S. Department of Labor, 2013). ICS is a scalable, standardized incident management approach that integrates facilities, equipment, personnel, procedures, and communications within a common organized structure (Federal Emergency Management Agency, 2013). One benefit is that the organizational structure allows each participant to have a defined role and similar lexicon. A simplified structure is shown in Figure 2 (U.S. Department of Labor, 2013). The academic partner served as incident commander the first year with the representative from the Coalition for the Homeless taking over in years 2–4, thus contributing to capacity building. The planning and logistics section chiefs came from the local health department, while a Coalition staff member served as finance and administration section chief. Incident command headquarters was housed at the health department. The safety officer for the field was from the Houston Police Department. The ICS structure was very successful, allowing improved organization of resources for the PIT count and contributed to the success of the endeavor. An added benefit was that it provided real-time training in ICS for health department personnel involved in the PIT count, thus keeping their skills up to date in case needed for a public health emergency.
Incident command structure.
Sheltered count
The PIT count includes two populations—those individuals who are experiencing homelessness who are unsheltered and those who are sheltered in emergency shelters. The sheltered count is in a sense easier because it is a static population. While larger shelters use the Homeless Management Information System (HMIS; U.S. Department of Housing and Urban Development, 2013a) to enroll individuals staying in their shelters each night—and thus that data can be used to provide information for the PIT count—smaller, particularly faith-based shelters may not. These faith-based emergency shelters represented approximately 12–15% of the total shelters housing those who are homeless in the jurisdiction—it varied from year to year. These non-HMIS shelters were located by conducting an extensive Internet search for shelters in the area, contacting the shelter, and performing a site visit to determine whether they met the HUD definition of a homeless shelter. To increase the completeness of the emergency sheltered count, all known shelters which did not have information entered into HMIS were contacted, and the number along with demographic information of persons staying at the shelter for the night of PIT count was enumerated. This substantially increased the completeness and accuracy of the sheltered count over previous years.
Unsheltered count
There are not enough emergency shelter beds for all persons experiencing homelessness in Houston/Harris County/Ft. Bend County and even should there be, not all persons chose to enter the shelters. Therefore, to obtain a complete picture of the magnitude of homelessness, it is necessary to count the unsheltered homeless (those sleeping in a place not meant for human habitation such as parks, sidewalks, cars, and abandoned buildings.) While we were confident that we were enumerating all of the individuals staying in shelters the night of the PIT count, counting those who are unsheltered is more difficult. The mild January temperatures usually experienced in southeast Texas means that there are a larger number of unsheltered persons experiencing homelessness than typically found in jurisdictions in colder climates, since the weather does not force them inside.
Staging areas and surface teams
We divided the area to be covered into 9 (10 starting in 2013) geographic areas, an increase of 2–3 over the previous years. In addition, a staging area was added to Fort Bend County which had not had one previously. This shortened the distance volunteers had to drive from their home or work to the staging area and then out in the field. It also lengthened the available time teams could be out in the field and helped provide better attention to Fort Bend County which had not been covered adequately in the past. Just-in-time training was held at the staging areas before deployment so that volunteers did not have to commit to an additional volunteer day. Each area was served by a staging area, a “command central” where teams reported to and received supplies. Each staging area had an assigned captain and cocaptain responsible for activities at that station. Approximately 75 teams, deployed from the staging areas and labeled surface teams, were put together to drive the streets of a section of the larger area to observe and tally individuals who were deemed to be homeless. These teams did not make contact with individuals but stayed in their cars and tallied numbers of those experiencing homelessness seen. Updated forms were created for tallying individuals seen during the observation count. Information included street intersection, age-group, and gender (sometimes volunteers had to guess at these demographic characteristics) and any comments about the observations.
A major change was also made in 2011 to the composition of the surface teams. Volunteers were solicited from the community at large and from student groups at UTSPH and local colleges. They had the option of reporting to a staging area in the part of town with which they were most familiar. Each surface team (i.e., one car) consisted of a dedicated driver and a navigator who had the maps. The navigator directed the driver where to turn and highlighted on the maps where the team had driven, making sure the whole geographic area for which they were responsible was covered but not duplicated. In addition, two other volunteers were in each car—one was in charge of filling out the tally sheet with numbers and basic information about persons experiencing homelessness observed, while the other individual was experiencing or had experienced homelessness. Since the surface teams did not engage with the individual but made an observational count only, it was important to train them on identifying characteristics of persons experiencing homelessness. We consulted with homeless service providers and a community advisory group, the Stand Up and Be Counted Corps (consisting of persons currently or formerly experiencing homelessness), to provide this list of characteristics. Teams were trained to look for individuals who were out at the time of the count (late night on a weeknight), who appeared that they were not going anywhere, who may be congregated in known “hot spots” such as bus terminals (particularly if buses had stopped running), and/or were carrying personal belongings/blankets in bundles or shopping charts. The addition of the volunteers currently or previously experiencing homelessness on each team also helped identify who on the street was likely to be experiencing homelessness and who was not, as these volunteers were experienced at noting telltale signs of persons who are most likely experiencing homelessness, while nontrained community volunteers may not be. The surface teams had to directly observe someone deemed to be experiencing homelessness in order for that person’s data to be added to the tally sheet.
The number of teams assigned to each sector was based on the prevalence of unsheltered persons experiencing homelessness in previous counts. Staging area captains deployed the volunteer surface teams assigned to their staging area into the field starting at approximately 5:30 p.m. on the night of the PIT count. This later start time was done for two reasons. Many community volunteers are not available until after 5 p.m. In addition, by delaying deployment of surface teams until after shelter registration, we greatly reduced the possibility of duplication in our count.
The surface teams counted persons experiencing homelessness found on street corners, parks, parking lots, convenience stores, and other areas where they congregate. Teams were asked to call into their staging area each hour with an updated tally of persons experiencing homelessness observed. If a team did not call in, the captain contacted them. This was done for two reasons. First and most importantly, it provided a safety check to assure that the surface team was secure. Second, it provided a real-time update regarding progress of the count. Updated tallies for each staging area were called into ICS headquarters by each captain. Incident command could then determine whether the numbers were in the same range as seen in the previous year. If it appeared that the counts were very low, extra help was sent out (see “Special Weapons and Tactics [SWAT]” teams).
SWAT teams
Another improvement in the methodology of the PIT count was the use of SWAT teams to augment surface team performance. These teams were convened at incident command headquarters and consisted of HDHHS field workers from the tuberculosis control section. These individuals are very familiar with the geography of the area, as they deliver tuberculosis medications every day to individuals on directly observed therapy—short term, many of whom are experiencing homelessness or are marginally housed. The SWAT teams were deployed into the field on the night of the count if a volunteer team didn’t show up or a staging area captain requested extra assistance with the count in their area. Incident command could also make the decision to send out an SWAT team if the preliminary numbers of those experiencing homelessness counted appeared exceptionally low compared to the previous years’ findings.
Specialized outreach teams
The surface teams covered all of Harris County and the areas of Fort Bend County, where individuals experiencing homelessness were known to congregate. However, not all persons experiencing homelessness are visible from the road. To improve the unsheltered count, we engaged specialized outreach teams which consisted of service providers from the Houston/Harris County area. These teams were not in cars but walked under bridges, into encampments, and into other places where individuals experiencing homelessness may be found but could not be seen by the surface teams. Abandoned buildings, the list of which was provided by the Houston Police Department, were approached with caution. The specialized outreach teams were sent out earlier, around 3 p.m., than the surface teams in order to maximize daylight time. We were not concerned about duplication as individuals in these areas to be canvassed were unlikely to enter a shelter at night. These specialized outreach teams not only enumerated persons experiencing homelessness but also distributed supplies and information to their potential clients and collected demographic information required by HUD that is not available from an observational count using a survey form and directly interviewing the individual if they consented. Teams were not to waken sleeping persons and, unlike the surface teams, they could estimate the number of those in a camp area by observing the number of sleeping bags or other indicators even if the person was not seen directly.
For safety reasons, both surface and specialized outreach teams had to report back to their respective staging area after finishing their allotted geographic area or at the latest by 11:30 p.m. Tally sheets were returned to the staging area captain, and the volunteer currently/previously experiencing homelessness was given a small gift card and bus pass in appreciation of the time spent on the count.
Practice count
We decided to hold two PIT counts, 1 week apart, during the last 10 days of January. This was done for two reasons—to give volunteers an opportunity to get familiar with the process as well as their geographic area and to have a contingency date in case of bad weather (rain, fog, or cold). The higher of the two counts would then be officially reported to HUD. In the first 3 years, the weather cooperated, and the second week’s count was used. In 2014, because we had experienced volunteers returning and felt we had the PIT count protocol well in hand, we expanded the training prior to the PIT count to replace the practice count.
Demographic data
HUD requires assessment of certain demographic information on both the sheltered and the unsheltered populations. The data for the sheltered population are in HMIS and so can be easily ascertained. It is more difficult to obtain for the unsheltered population, however, since the surface teams do not engage the person experiencing homelessness but merely enumerate. However, the specialized outreach teams do engage with individuals and so could administer a short questionnaire to determine those characteristics. For the unsheltered persons experiencing homelessness observed but not contacted, we determined that the most unbiased sampling would be obtained by surveying clients at service and food providers the morning after the PIT count, as recommended in HUD guidance (U.S. Department of Housing and Urban Development, 2010). Individuals were asked where they spent the previous night and were only surveyed if they were considered unsheltered by HUD’s definition. Information was then obtained about age, gender, military status, length, and frequency of homelessness, whether the individual was part of a family which was homeless (defined by HUD as an adult and child [children] under age 18 years) and presence of disabling condition. In addition, chronic homelessness was determined, defined by HUD, as homeless for the past year or longer or four episodes of homelessness in the past 3 years along with presence of a disabling condition which made it hard to find/keep a job. (Hopper, 1992). The survey consisted of 18 questions asked by the volunteer and filled in on a paper survey. The interview lasted less than 5 minutes on average. Metro bus passes were given in appreciation of time spent answering the survey. An example of the survey is shown in Appendix A.
Plant-capture method for estimating undercount
It is extremely likely that the PIT count undercounts unsheltered homeless, as it is logistically very difficult to count all individuals experiencing homelessness within just a few hours in such a large geographic area. Weather can also make a difference, and the presence of fog or rain means that it is much more difficult to observe unsheltered individuals from the street, since visibility is hindered and they are more likely to be in covered areas. HUD allows the application of statistically valid methods to scientifically estimate the number of unsheltered persons experiencing homelessness, but many of these methods require interviewing, something that isn’t possible in this large geographical area in a single night. We chose to implement the “plant-capture” method used in a variety of disciplines to determine percentage undercount (Martin, Laska, Hopper, Meisner, & Wanderling, 1997). Basically, a known number of “plants” are sent into the field, and the percentage observed gives an indication of what percentage of the total population of those experiencing homelessness are being counted (White, Anderson, Burnham, & Otis, 1982). New York City has used this method, sending out college student and service provider volunteers as plants and identifying these volunteers when individuals experiencing homelessness were questioned as part of their daytime direct contact survey (Hopper, Shinn, Laska, Meisner, & Wanderling, 2008). We decided to try this methodology but did not think it would be safe to send out community volunteers to the streets at night with no experience of homelessness in Houston. As an alternative, we recruited volunteers from the Veteran Administration Hospital’s program for veterans who were currently or had formerly experienced homelessness. They are familiar with different parts of town and with the homeless community and social norms. Allocation of plants to geographic area was done on a weighted distribution based on the population density of those experiencing homelessness in the previous year.
Because CFTH does an observational count, we needed some way for surface teams to identify the plants. In the first year, we used flashing orange blinkers as a signal to the surface team that the observed individual was a plant. However, these were problematic as the blinker could only be seen by the surface team if the plant was facing the street. In the second year, we tried orange fluorescent knit hats as the indicator that the person was a plant. These were not readily observed either, and so we did not continue this method during the 2013 count.
Results and Discussion
The PIT count is a federal requirement for all CoCs receiving funding from HUD. Due to Houston’s large population with resultant high numbers of those experiencing homelessness and the vast geographical area to be covered, counting those experiencing homelessness, particularly those unsheltered, is a challenge. However, an accurate count is important because it estimates the extent of the problem and defines trends and service needs as well as federal funding levels. The enhanced methodology to count those experiencing homelessness during the HUD PIT count resulted in a significant increase in observed persons in 2011, the first year the improvements were implemented (Figure 3).
Comparison of point-in-time counts, 2009–2014.
Since the same methodology (with small tweaks) was used for 4 successive years, we can discern trends. We have seen a consistent decrease in the number of emergency sheltered and unsheltered persons experiencing homelessness. At the same time, we’ve seen an increase in those lodged in permanent housing (which consists of rapid rehousing and permanent supportive housing; Figure 4). This increase in lodging is a result of a concerted effort to end homelessness in the Houston area. An action plan with specific goals, strategies, and time frames to prevent and end homelessness in the area has been adopted, concentrating first on veterans and the chronically homeless (more information can be found at www.thewayhomehouston.org). We cannot track individuals and therefore do not have definitive evidence that the decrease in the numbers of those experiencing homelessness is related to the increase in those in permanent housing. However, the decrease seen in one with the increase in the other is suggestive of this. In 2013 and 2014, the total number of those experiencing homelessness plus those in permanent housing remained constant, although the former decreased and the latter increased. Again, this suggests, although does not prove, association between the two, and we believe that we are seeing results of a concentrated campaign to end chronic homelessness in the Houston/Harris County/Fort Bend County area.
Clients in permanent housing 2011–2014.
Although the number of emergency sheltered persons remained fairly constant from 2009 to 2011, the number of unsheltered persons observed rose substantially in 2011. While an actual increase in the number of unsheltered persons experiencing homelessness could account for this rise, there is no evidence that this occurred during the year based on observations of service providers. There was also no increase seen in the number of persons experiencing homelessness nationwide from 2010 to 2011 (236,923 individuals experiencing homelessness counted in 2010 vs. 232,901 in 2011; U.S. Department of Housing and Urban Development, 2014). Should the actual number of unsheltered persons experiencing homelessness be increasing in Houston/Harris County/Fort Bend County, we would expect to see concomitant increases in the sheltered numbers. Instead, the sheltered count, despite increased coverage, decreased. Since neither service providers nor shelters saw an increase in persons experiencing homelessness in 2011, it is unlikely that this is the explanation for the rise in unsheltered persons experiencing homelessness observed. Instead, we believe that, using these enhanced methodologies, we are closer to getting an accurate count of this hard-to-detect population.
To improve the count of sheltered individuals (those in emergency shelters, transitional housing, or safe havens), we used data from the Homeless Management Inventory System (HMIS) when employed by the shelter and reached out by telephone to collect data from other shelters, particularly faith-based shelters, that did not. For 2010, 75% of the emergency shelter projects and 69% of the transitional housing projects reported the number of persons sheltered on the night of the PIT count. For the 2011–2014 counts, due to increased effort, 100% of emergency shelters defined by HUD as housing persons experiencing homelessness for the purposes of the PIT count (n = 65) reported their numbers. Over 94% of sheltered numbers came from HMIS, meeting the HUD requirement of >75%.
The enhanced methodology used to improve estimates of the unsheltered population of those experiencing homelessness (those sleeping on the streets or in places not meant for habitation) consisted of a number of improvements. The first was bringing together a coalition of nonprofits, local government, and academia. This resulted in a sharing of skills and capacity building. An incident command structure was deployed, allowing for better communication and coordination between agencies. This collaboration has added benefits as partners get to know each other, their particular skills and resources, and discover ways they can collaborate in other areas of local public health. Other advantages of using ICS are improved communication with those in the field, addition of SWAT teams to cover areas where undercounts appear to be occurring, and the functions of the planning section in the ICS structure. January can be a rainy/foggy month in Houston, and the planning section is charged with delivering weather reports to those in the field. They have been able to tell the staging area captains what current weather conditions are and when they are scheduled to improve, thus allowing the captains to deploy teams based on these data. Although data from only one night can be submitted to HUD for the PIT count, the addition of another night’s count during the last 10 days of January also worked to improve the level of the count, as it allowed teams to practice and become familiar with their area.
Unsheltered individuals experiencing homelessness were counted using two methods. Surface teams were dispersed out of 10 staging areas to drive the streets and count individuals experiencing homelessness, an increase in both the number of teams and staging areas over previous years. The addition of SWAT teams allowed full coverage of the area. The composition of the teams was improved as well, and the addition of a homeless/formerly homeless volunteer improved the identification of those experiencing homelessness. An added benefit was that community volunteers were exposed to persons they may not interact with in other situations and so were sensitized to the issues involved in homelessness. Many of the volunteers noted that the best part of participating in the PIT count was this opportunity to broaden their understanding of the problem and dispel stereotypes they may have had.
Outreach specialist teams comprised of service providers to the homeless walked under bridges, along the bayous and investigated abandoned houses and other areas where encampments of those experiencing homelessness had been identified. This allowed enumeration of persons experiencing homelessness in areas not visible from the street and so increased the completeness of the unsheltered count.
Requisite information about a jurisdiction’s unsheltered population of persons experiencing homelessness was obtained using an instrument administered in a variety of settings. While we acknowledge that this is not a true random sample and so results may be biased, it was the best solution, given the constraints of the PIT count.
While all of the above-mentioned methods improved the PIT count, one new methodology attempted was not as successful. The plant-capture method, originating as a method to count wildlife, has been successfully used to estimate hard-to-count human populations. However, as other jurisdictions have discovered, there are inherent problems in using this method to count those experiencing homelessness. We had the added difficulty of not directly interviewing the target population and so had to devise a method whereby surface teams could recognize plants from the street. Blinkers one year and fluorescent orange knit caps the next worn by plants were not observed. Additional difficulties were deploying the veterans who were currently or had previously experienced homelessness into the field, given the large geographic area to be covered, and getting them back to the staging areas as local transportation stops running relatively early in the Houston area, as early as 9 p.m. on weeknights (Metropolitan Transit Authority of Harris County, Houston, Texas, 2013). This meant plants had to leave their assigned area before the count officially ended, violating assumptions inherent in the capture–recapture statistical assumptions.
Explaining the nature of the plant-capture method to the volunteers was an additional challenge. We ended up using the term “observer” (even though they weren’t actually observing) because there was some objection to the term “plant.” Plants were supposed to stay within a given small area and not play “hide and seek” but not wave their hands saying “here I am,” either. Whether or not these constraints were adhered to is difficult to ascertain, and this affects the scientific validity of the technique. For these reasons, we decided not to use this method going forward. We remain open to an alternative method to statistically determine the undercount but have not found one that can be implemented in our large geographic area that precludes universal interviewing.
Conclusion
In 2011, Houston/Harris County/Fort Bend County incorporated a number of new and improved methods to enhance the HUD-mandated PIT count. Results indicate that the completeness and accuracy of the PIT count have increased and that we have observed a decrease in the number of persons experiencing homelessness in the area over the past 3 years. New methods implemented may prove useful in other large metropolitan or geographic jurisdictions when conducting this important count of those experiencing homelessness.
Footnotes
Appendix A
Acknowledgment
The authors wish to thank the many community volunteers that assisted with this project.
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.