With the advancement of computers and access to the Internet, surveys using the web are becoming increasingly popular. Such surveys save money by eliminating postage costs and save time by providing immediate access to returned surveys and improved data processing. This primer describes some considerations and suggestions related to the use of web surveys. In addition to issues of sampling error, question writing, and question sequencing, successful use of web surveys requires determining if a web survey is appropriate, determining the desired level of security and confidentiality, selecting design options that enhance the survey, conducting rigorous pilot testing, and implementing strategies to maximize response rate. J Allied Health 2008; 37:180-186. SURVEY RESEARCH is a frequently used methodology. With the advancement of computer hardware, software, and increasing access to the Internet,1 survey implementation using the World Wide Web (web) is becoming increasingly popular.2 After the necessary hardware and software is acquired, a web survey saves money3 by eliminating postage costs and saves time3 because the survey is immediately available to respondents after it is posted to the Internet and returned surveys are immediately available to the investigators. Additional time savings are realized because data processing is faster with web surveys.

While web surveys offer several advantages, they are not appropriate for every situation. Web surveys add a layer of technology not present in traditional paper surveys and have unique considerations related to design, pilot testing, and response rate. The purposes of this primer are to describe questions to consider when deciding if a web survey is appropriate, issues related to security and confidentiality, and the unique considerations related to design, pilot testing, and response rate inherent in web surveys. Issues related to sampling error, question writing, and question sequencing are beyond the scope of this report; publications addressing these issues are available.4-8

Determining if a Web Survey Is Appropriate

To determine if a web survey is appropriate, the investigators must first answer these questions: "Will the potential respondents have adequate Internet access?" and "Are the potential respondents computer literater2 In 2003, approximately 54% of United States households had Internet access.9 Access was not consistent across groups of people. For example, individuals with lower incomes, individuals with less education, and individuals with disabilities may have less home Internet access.10 In contrast, groups of people such as individuals in business, faculty members, and college students typically have access to the Internet.4

Knowing the potential respondents are likely to have Internet access does not address the adequacy of their access. Because some web survey design elements such as graphics increase download times,11 potential respondents with slower Internet service (i.e., dial-up modems) may become frustrated when attempting to complete a web survey. While broadband Internet service is faster and mitigates concern about download times, access to broadband Internet service is not consistent across the United States. In 2003, approximately 20% of households had broadband access, with less penetration of broadband in rural as compared with urban areas.12 Although limited to one state, a more recent example highlights the continued disparity in Internet access between urban and rural areas. In communities with < 15,000 households, wireline broadband access was 61% compared with 93% in three cities in Missouri (St. Louis, Kansas City, and Springfield).13

If potential respondents do have adequate Internet access, then the second question is "Are the potential respondents computer literate?"2 Computer literacy requires computer skills to access and navigate through the web survey.14 Knowledge of how to use computers, access Internet web sites, and complete web surveys will vary across people. For example, Schleyer and Forrest15 reported that a small number of dental professionals did not know how to return a web survey electronically and instead completed a paper copy of the survey and returned it by fax. Because of the variability in computer literacy, we believe potential web survey respondents should be given the telephone and e-mail contact information for questions related to the survey. Providing both types of contact information ensures that potential respondents with less computer literacy can contact the investigator.

Security and Confidentiality

An important consideration in using web surveys is determining the need for data security and confidentiality (Wolpin S: Personal communication to TM Austin, 20 July 2007).16 Collaboration among the investigators, web survey software providers, and individuals from information technology units at the investigator's institution should occur to ensure the desired level of confidentiality is achieved. Investigators should also ensure that stored data is protected. Examples of data protection may include physical and technical barriers to data access. For example, locked office suites represent physical barriers and computer passwords are technical barriers. Investigators should also establish mechanisms for data recovery in the event of an unforeseen loss of data (Wolpin S: pers. comm.).16

Deciding Among Design Options

Having decided a web survey is appropriate, the focus shifts to survey design. The sheer number of design options available to the web survey designer is a strength and a curse of web surveys. Understanding how design choices influence survey responses maximizes the likelihood of obtaining data that can be used to answer the research question.11,17 Examples of design choices include the following: choosing a question presentation style, displaying one question versus several questions per computer screen, using or not using a progress indicator, using radio buttons versus drop-down lists response formats, checking all that apply in a list versus responding yes or no to each choice, and forcing an answer to all questions versus allowing potential respondents to not answer questions. Each of these is addressed below.

There are multiple ways to present a web survey. One way is to present the web survey on a series of computer screens. With this approach, the respondent does not see the entire survey at once. One screen with one or more questions is delivered, the questions are viewed by the respondent, and responses are submitted to the server before the next computer screen is available.18 These steps are repeated until the survey is completed. This approach, called the interactive or paging approach,18,19 requires repeated contact between the respondent's computer and server hosting the survey and repeated submission of survey responses by the respondent. Using an interactive approach permits the possibility of automatic question routing within the survey. Because routing is controlled by the survey software, relying on the respondent to correctly follow routing directions is eliminated. Because survey responses are repeatedly submitted, investigators can determine how many respondents started but then abandoned the survey.18 For example, a respondent may submit responses to questions on the first two computer screens and then stop. With an interactive approach, the investigator knows the number of people who started but never finished the survey.

Because an interactive approach can deliver one or multiple questions per screen, the web survey designer has to choose which option to use. If the survey questions are related, multiple questions per screen may be appropriate. For example, a scale with four items measuring a construct such as pain could be presented one item per screen or on a single screen. Because the items are related to the same construct, presenting them on a single screen may be beneficial. Couper et al.18 reported that multiple related items on the same screen were completed faster, had less missing data, and had better interitem correlations than single items presented on multiple computer screens.

An alternative to the interactive approach is the scrolling approach.18,19 This approach delivers the entire survey at once. Assuming respondents do not have to answer a survey question to proceed to the next question, they can use the computer mouse to scroll forward or backward through the web survey, similar to paging forward and backward through a paper survey. Dillman4 recommends this approach because the presentation is similar to a paper survey. Routing is possible with the scrolling approach but relies on respondents correctly following routing instructions. The routing instructions can be text based, indicating which question to proceed to, or can be hyperlinked, requiring only a mouse click to move to the appropriate question within the survey. Respondents may not initially understand how to use hyperlinks. Peytchev et al.19 reported that college students were more likely to use hyperlinks that appeared later in the survey, indicating a possible learning effect for using these links. Furthermore, Peytchev et al.19 found limited evidence that hyperlinks for routing may have encouraged respondents to skip items.

Because the scrolling approach delivers the entire survey at one time and survey responses are not repeatedly submitted, potential respondents could begin answering questions but then abandon the survey without submitting any of their responses. In this case, investigators would not have information on the number of respondents who started but then abandoned the survey.18 Finally, using the scrolling approach with a long or graphic-intensive survey may be problematic as this may result in increased download time because the entire survey has to be downloaded before any questions can be completed. The choice between an interactive and a scrolling approach is not necessarily a dichotomy; rather, Peytchev et al.19 describe this choice as a continuum because it is possible to use features of both the interactive and scrolling approach to present a survey. For example, rather than either requiring a submission after each question or delivering the entire survey at once, large sections of the survey could be delivered, allowing the respondent to scroll through all of the questions in an individual section, submit the responses, and then receive the questions for the next section.19 Using this approach, Peytchev et al.19 were able to collect data on abandonment for each of the survey sections.

When choosing the web survey presentation approach, investigators have to decide whether data on abandonment should be collected, whether routing should occur automatically or with instructions, and whether it is important that respondents can view the entire survey at one time. There is likely no single web survey presentation approach suitable for all applications.19

Web survey software often has the option to include a progress indicator with the survey. Progress indicators give respondents an indication of how much of the survey remains to be completed as they answer questions. Progress indicators are not necessary if a scrolling approach is used because respondents can scroll through the survey to see how many items remain.4 Couper et al.18 reported that use of a progress indicator marginally reduced abandonment but increased download time. The increase in download time may have diminished the positive effects the progress indicator had on abandonment. In some cases, a progress indicator may be misleading. For example, when a survey includes both multiple-choice and short- answer questions, a progress indicator showing the percentage of the survey completed may mislead the respondent because these types of questions will vary in the length of time to complete.20

Web survey software programs provide several response option formats. Two that have been studied include radio buttons and drop- down boxes. Each of these has strengths and weaknesses. Dillman4 recommends using response options that most closely resemble traditional paper surveys. Radio buttons, which respondents use by clicking on the button with a mouse to indicate their choice, meet this recommendation because the process is similar to marking a circle to indicate a choice on a paper survey. Another alternative unique to web surveys is drop-down boxes. There are at least two types of drop-down boxes.17 One type of drop-down box initially displays one or more response options and requires a mouse click and scrolling to display the remaining options. The second type of drop- down box displays instructional text such as "click here" and requires a mouse click with potential scrolling to display all the options. Because the computer mouse can be used to scroll through the survey and through drop-down box response options, respondents may inadvertently change their answer in a drop-down box when they meant to scroll to the next survey question. Consequently, accurate selection of a choice with a mouse click and subsequent scrolling to view response options not initially shown may make dropdown boxes difficult to complete.17 Couper et al.17 recommend against drop- down boxes that do not initially display all response options after the initial mouse click. When response options include a well-known constrained list, such as a list of the United States, use of drop- down boxes is acceptable.17

A third response option format involves responding to a list of choices. One possibility is to format the response options so that respondents check all the items that apply among a list of items. A second option is to present each item in the list of items with a yes/no choice. This second option is called a forced-choice format. Recently, Smyth et al.21 compared these response options and found that the forced-choice format resulted in more items being selected. They do, however, note that the results may only apply if the stem wording is similar to the wording used in their study. Specifically, the stem wording in the questions used by Smyth et al.21 included positive and negative categories as part of the question stem. For example, a patient responding to a list of characteristics about his or her hospital stay could be asked in the question stem, "Do you believe that the hospital characteristics listed below do or do not reflect your experience while in the hospital." Smyth et al.21 proposed that the forced-choice format as compared with the check- all-that-apply format resulted in more options being selected because respondents were processing the information at a deeper level.

When responses to a question are presented in a list, the response chosen may be influenced by where it falls in the list of response options. For example, when the survey respondent is asked to make a judgment based on consideration of all possible responses, response options that appear earlier in the list are more likely to be chosen. This is known as the primacy effect.8 Couper et al.17 found a primacy effect with radio buttons and drop-down boxes. The effect was greatest with drop-down boxes that did not display all response options after the initial mouse click. Because response options that were initially visible were more likely to be chosen, Couper et al. call this the visibility principle.17 Smyth et al.21 also found a primacy effect when using the check all response option format.

In a paper survey, respondents are free to leave items blank. Web- based surveys can be designed to force respondents to answer all items or allow respondents to leave items blank. DeRouvray and Couper22 found support for a "middle of the road" approach utilizing a prompt screen when a question was not answered. For example, a question asking about income could have several response options, one of which could be "decline to answer." If respondents chose "decline to answer," the next screen could prompt them to either go back and answer the question or proceed without answering the question. DeRouvray and Couper22 found this approach resulted in less missing data. In addition, this approach maintains a respondent's right not to respond to an individual question.

Conducting Pilot Testing

Typically, pilot testing of paper surveys includes determining if directions are clear, understanding how potential respondents interpret questions, and establishing test-retest reliability of survey items. While these pilot testing procedures are also conducted in web surveys, web surveys require additional pilot testing related to hardware/software interactions,4,15 respondent interactions with the web survey,15,18 and data processing. Each of these is discussed below.

HARDWARE/SOFTWARE INTERACTIONS

Completing a web survey requires communication between the respondent's computer and the server hosting the survey. There are at least two issues related to this communication: server timeout and connection speed. Because of the variability in respondents' computer hardware, server connections, and the time respondents take to complete a survey, it is possible the connection between the respondents' computer and the server hosting the survey will be timed out before the survey is completed or submitted. Schleyer and Forrest15 reported that respondents with slow modem connections experienced server timeout problems before the survey was completed. To test server timeout, the investigators should test their web survey on a variety of computers with varying Internet connection speeds. If timeout problems are observed, a solution to consider is to increase the server timeout setting.

Another pilot testing area related to connection speed is the issue of web survey download time. Regardless whether a scrolling, interactive, or combined approach is used, respondents with slow Internet connections may experience long download times if the survey includes graphics. With the scrolling approach, the initial download time of the survey may be prolonged because the entire survey is being downloaded. Although individual screens are downloaded in the interactive approach, delays may be experienced when particular screens include graphics.11 We suggest testing download times with a variety of dial-up connection speeds and broadband connections to determine if download times seem reasonable. Reasonable times for download will be determined by the investigators' understanding of the potential respondents' tolerance for slow downloads. If the investigators determine that download times are excessive and the survey includes graphics, the investigators must decide which graphics are essential for understanding questions versus those that only serve to make the survey appearance more pleasing. For example, in a web survey of leg pain experienced by collegiate cross-country runners, Reinking et al.23 used pictures of the leg for respondents to identify areas where their pain was present. Reinking et al.23 believed that written explanations alone describing an area of the leg would possibly confuse respondents. Conversely, graphics such as company logos placed on every screen may increase download times without improving the respondent's ability to understand survey questions.

Potential respondents may connect to the Internet through different Internet browsers (e.g., Internet Explorer or Netscape) and versions. In addition, potential respondents may be using a variety of screen settings. Both of these can influence how the survey appears. Dillman4 suggests testing a variety of screen resolutions, text sizes, and Internet browsers to determine if text and critical images are adequately displayed. While the differences may be slight and not have an impact on survey responses, there may be times when the differences affect the survey responses. For example, respondents would likely be confused if question text, response options, or images are cut off at the margins of the computer screen due to display issues. A simple, but perhaps overlooked, pilot testing step is to ensure that the URL hyperlink successfully connects the respondent to the web survey. Likewise, the URL hyperlink should be tested with a variety of Internet service providers (e.g., America Online, AT&T) to ensure that that the survey can be accessed. Schleyer and Forrest15 reported a group of individuals had trouble accessing their survey; most of these individuals used America Online as their Internet service provider. While Schleyer and Forrest15 could not determine why the problem occurred, this finding illustrates the need to test web surveys with a variety of Internet providers.

RESPONDENT INTERACTIONS WITH THE WEB SURVEY

Web surveys can be designed with varying levels of access. For example, a solicitation with the access URL could be posted on a study web site inviting public participation. In this instance, investigators would not know who completed the survey. Conversely, if only certain individuals can access the survey, a user name and password could be required to access the survey. Requiring a user name and password adds steps that a potential respondent has to successfully complete to access the survey and should be tested. Reinking et al.23 used the respondents' e-mail address for their user name. The user name field was unknowingly set to 20 characters. One respondent with an e-mail address that was longer than 20 characters could not access the survey and contacted the investigators. This required the investigators to increase the user name field character length to allow for longer user names. Ambiguous password characters are also problematic. Couper et al.18 found that respondents were less likely to complete the survey if their passwords included ambiguous characters. This finding led Couper et al.18 to recommend against using the letters l (el) and o (oh) and the numbers 1 (one) and 0 (zero) in passwords.

Potential respondents should be give specific directions related to the web survey.4 A solicitation to participate in a web survey will likely include directions on how to access the survey. For example, directions to access the survey may instruct the respondent to double click the URL hyperlink, and if problems are encountered, cut and paste the URL to the web browser's URL address field. In addition, the steps required to progress through a web survey are not as intuitive and easily understood as paging through a paper survey. Directions within the survey may instruct the potential respondent to click on a navigation tool such as a "Next" button to progress through the survey and click on a "Submit" button to submit responses. While these directions may be clear to individuals with higher computer literacy, those with less computer literacy may have difficulty understanding and following the directions. For these reasons, the investigators should pilot test directions on how to access the survey and directions on how to navigate through the survey with individuals who have similar computer literacy skills as the potential respondents.

Web surveys can easily be designed using an array of colors. The use of colors can either enhance or inhibit a respondent's ability to identify question stems and response options or to read the survey items.4 For example, in a single screen with five Likert- type questions, five different background colors could be used for each question. While use of these colors may clearly differentiate each question, respondents may find this use of color distracting. Pilot testing with individuals similar to potential respondents assures that the use of color enhances rather than detracts from the survey.

DATA PROCESSING

During data processing of web surveys, there are at least two processing steps that could lead to inaccuracies between the respondent's intended response and the final data file. The first step occurs when the respondent's responses are recorded in the web survey data file. The second step occurs if the web survey data file is exported to data analysis software. Schleyer and Forrest15 reported data processing problems as their software program incorrectly stored respondents' answers to a Likert scale. They discovered this problem after 130 surveys had been returned, requiring them to discard the incorrectly recorded answers.

To test for data processing inaccuracies, we suggest pilot testing the two data processing steps. Specifically, paper and web versions of the survey should be completed with identical responses. First, responses recorded in the web survey data file are compared with the paper responses for accuracy. If the web survey software has adequate data analysis capabilities, no further pilot testing of data processing is required. However, if the data from the web survey file are imported into a statistical software package, this step must also be tested. Again, the data in the statistical software file are compared with the paper survey. The most rigorous pilot testing of data processing would require testing of each question and each possible response option to ensure that all questions and response options are accurately recorded in the web survey data file and accurately imported into the statistical software package.

Maximizing Response Rate

After determining that a web survey is appropriate, deciding on key design elements for the survey, and pilot testing the survey, the next step is to implement procedures to enhance response rate. A first consideration involves whether or not to use a precontact. Precontacts alert potential respondents that they will receive a solicitation to participate in a web survey. Precontacts could occur via postal mail, e-mail, telephone, or in some instances in person. For example, Reinking et al.23 used the time spent with participants during initial data collection to alert participants that they would be invited to complete a web survey after the cross-country season. Their response rate to the web survey was 74%. Because Reinking et al.23 used other features to improve response rate such as reminders, it is unknown how much the precontact contributed to the response rate. In a meta-analysis, Cook et al.24 found that precontact was positively correlated with response rate (r = 0.255).

Solicitation to participate involves contacting potential respondents requesting their participation in the web survey. How the solicitation letter is constructed may affect response rate. For example, the solicitation letter could be personalized; the respondents could be informed they are part of a select group asked to complete the survey; and a deadline for survey completion could be included in the survey solicitation letter. In a survey of high school students, Porter and Whitcomb25 examined these variables as well as how personalization of sender's e-mail address, authority of e-mail signatory, and profile of requesting office affected response rate. Only a combination of a statement of selectivity and survey deadline in the solicitation e-mail increased response rate. It is unknown if this same effect would be found when surveying other groups such as business owners. Cook et al.24 reported that a personalized contact with potential respondents was associated with higher response rates (r = 0.407).

Although responses to web surveys are often quickly received after the solicitation request,20 follow-up reminders are recommended to maximize response rate.3 The number of follow-up contacts depends on the audience's tolerance for being contacted3 and the criticality of receiving every possible response. Cook et al.24 found that three contacts, one of which was a precontact, resulted in the greatest median response rate. Reinking et al.23 used three follow-up reminders; for each reminder, approximately one third of the remaining nonresponders completed the survey (unpublished data).

Another consideration to enhance response rate is whether or not to offer an incentive. A recent meta-analysis by Goritz26 found an incentive increased response rate by 2.8%. Given this modest effect, the decision to use an incentive depends on the criticality of achieving the highest response rate weighed against the practical difficulties of providing an incentive with a web survey. For example, if the postal address of the respondent is unknown, an offer of an incentive would be limited to an incentive that could be provided electronically, such as an electronic coupon.

In addition to affecting ease of completion, the design options used in a web survey may influence response rate. For example, it is possible to change colors throughout the survey, use bands of color to differentiate rows, and position response answer boxes to either the right or left of the response text. While these options may make the web survey "fancy," they make it less like a paper survey.4 Dillman et al.11 found that a web survey formatted to appear like a traditional paper survey with a white background and black text and answer boxes to the left of the response text had a greater response rate than an identical "fancy" survey that used colors and graphics.

Web Survey Software Options

As with the wide range of choices available for web survey design, there are a variety of web survey software programs. The software programs will vary in available features. Some of these features may minimize the impact of issues described earlier in this commentary. For example, survey software programs may include the ability to automatically randomize the order of response options provided in a list. Randomizing the order of response options may decrease the primacy effect. Software programs may also offer the ability to automatically adjust the appearance of the web survey based on the respondent's screen settings,21 resulting in consistent spacing of response options across screen settings. This is especially important for a horizontally presented Likert scale, which should have proportionally spaced response options. Given the variability among web survey software programs, we recommend that individuals access company web sites to find the most recent features, pricing, service agreements, and information relative to data security because we believe these items can change rapidly. Several web survey software providers with their URL addresses are listed in Table 1 (adapted from Wright27). Conclusions

This primer describes some considerations and suggestions related to the use of web surveys. A summary of these considerations and suggestions is provided in the form of a checklist (Table 2). In addition to issues of sampling error, question writing, and question sequencing, successful use of web surveys requires determining if a web survey is appropriate, determining the desired level of security and confidentiality, selecting design options that enhance the survey, conducting rigorous pilot testing, and implementing strategies to maximize response rate.

TABLE 1. Web Survey Software Providers

Active Websurvey www.activewebsoftwares.com

Apian Software http://www.apian.net/

CreateSurvey http://www.createsurvey.com/

EZSurvey http://www.raosoft.com/

FormSite http://www.formsite.com/

HostedSurvey http://www.hostedsurvey.com/

InfoPoll www.infopoll.net/

KeySurvey http://www.keysurvey.com/

mrlnterview http://www.spss.com/mrinterview/

PollPro http://www.pollpro.com/

Quask http://www.quask.com/

QuestionPro http://www.questionpro.com/

Ridgecrest http://www.ridgecrestsurveys.com/

Snap* http://www.snapsurveys.com/

SumQuest www.sumquest.com/

SuperSurvey http://www.supersurvey.com/

SurveyCrafter http://www.surveycrafter.com/

SurveyMonkey http://www.surveymonkey.com/

SurveySite http://www.surveysite.com/

SurveySolutions http://express.perseus.com

WebSurveyor http://www.websurveyor.com/

Zoomerang http://www.zoomerang.com/

zTelligence http://www.markettools.com

Adapted from Wright,27 with additional information from Crawford28 and Wolpin (pers. comm.). This list does not represent an endorsement by the authors. Website addresses accessed October 2007.

* Authors' personal experience.

TABLE 2. Checklist of Suggestions and Considerations Related to Web Surveys

Appropriateness of a web survey

* Consider potential respondents' Internet access (access may vary among groups10)

* Consider the adequacy of potential respondents' Internet access (broadband vs. dial-up9,10)

* Consider the computer literacy of potential respondents (ability to follow survey and navigation directions4,15)

* Provide multiple ways (e-mail and telephone) for respondents to contact the investigators with questions2,14

Security and confidentiality

* Determine the level of security and confidentiality needed (online security, physical and electronic security) (Wolpin S: pers. comm)16

Design choices

* Consider which type of web survey presentation (interactive, scrolling, or combination) is most appropriate (choice may influence download times, collection of abandonment data, and type of routing directions18,19)

* Consider the number of questions per computer screen to present (may be beneficial to group items related to a similar construct on one screen18)

* Consider use of a progress indicator (determine if the indicator affects download time and accurately reflects progress through the survey18,20)

* Consider possible response option formats

-Radio buttons (similar to a paper survey; may be easier to complete; potential for primacy effect17)

-Drop-down lists (may be more prone to error; useful for constrained lists; potential for visibility principle17)

-Use of forced choice versus check all (forced choice generates more responses; check all may be affected by primacy effect21)

* Consider using a reminder prompt when a question is skipped (maintains respondent's right to refuse to answer and may result in more completed questions22)

Pilot testing

* Pilot test hardware/software interactions

* Pilot test server timeouts (slower dial-up connections may result in server timeout before the survey is completed15)

* Pilot test survey download time with broadband and dial-up connections (survey design options and use of graphics may affect download time11)

* Pilot test the survey with major Internet browsers and service providers4,15

* Pilot test the URL hyperlink to the survey

* Pilot test user name and password (avoid the letters l [el] and o [oh] and 1 [one] and 0 [zero]; character length18,23)

* Pilot test the survey and navigation directions with individuals similar to potential respondents (clarity, understandability)4

* Pilot test the survey design with individuals similar to potential respondents (use of color and graphics)4

* Determine that survey responses are accurately recorded by the software15

* Determine that survey responses are accurately imported to statistical software

Maximizing response rate

* Use a precontact when possible24

* Consider how to solicit potential respondents (when possible, use a personalized solicitation24,25)

* Use follow-up reminders (consider respondents tolerance to repeated contacts3)

* Consider using an incentive (increase in response rate may not warrant the use of an incentive26)

REFERENCES

1. Birnbaum MH: Human research and data collection via the internet. Annu Rev Psychol 2004; 55:803-832.

2. Couper M: Web surveys: a review of issues and approaches. Public Opin Q 2000; 64:464-494.

3. Umbach PD: Web surveys: best practices. New Directions Inst Res 2004; 121:23-38.

4. Dillman D: Mail and Internet Surveys The Tailored Design Method, 2nd ed. New York: Wiley; 2000.

5. Fink A: The Survey Kit, 2nd ed. Thousand Oaks, CA: Sage Publications; 2003.

6. Presser S: Methods for Testing and Evaluating Survey Questionnaires. Hoboken, NJ: Wiley; 2004.

7. Sudman S, Bradburn NM: Asking Questions: A Practical Guide to Questionnaire Design. San Francisco: Jossey-Bass; 1982.

8. Tourangeau R, Rips LJ, Rasinski KA: The Psychology of Survey Response. Cambridge: Cambridge University Press; 2000.

9. US Census Bureau: The 2007 Statistical Abstract The National Data Book. 1142 - Households With Computers and Internet Access by Selected Characteristic: 2003. Available at: http://www.census.gov/ compendia/statab/information_communications/. Accessed 13 Apr 2007.

10. National Telecommunications and Information Administration: Falling through the net: toward digital inclusion. October 2000. Available at: http://search.ntia.doc.gov/pdf/fttn00.pdf. Accessed 9 Feb 2007.

11. Dillman D, Tortora RD, Conradt J, et al: Influence of plain vs. fancy design on response rates for web surveys. Presented at Joint Statistical Meetings, Dallas, TX, Aug 1998.

12. U.S. Department of Commerce, Economics and Statistics Administration National Telecommunications and Information Administration: A Nation Online: Entering the Broadband Age. Sept 2004. Available at: http://www.ntia.doc.gov/reports/anol/ index.html. Accessed 13 Apr 2007.

13. Missouri Public Service Commission: Commissioners' report on Missouri broadband availability. Sept 18, 2007. Available at: http:/ /www.psc.mo.gov. Accessed 1 Oct 2007.

14. Dillman D, Tortora R, Bowker D: Principles for Constructing Web Surveys: An Initial Statement. SESRC Technical Report 98-50. 1998.

15. Schleyer TK, Forrest JL: Methods for the design and administration of web-based surveys. J Am Med Inform Assoc 2000; 7:416-425.

16. Wolpin S, Okada J, Berry D: Authoring tools for web based surveys: a descriptive study. Stud Health Technol Inform 2006; 122:876.

17. Couper MP, Tourangeau R, Conrad FG, et al: What they see is what we get: response options for web surveys. Soc Sci Comput Rev 2004; 22:111-127.

18. Couper MP, Traugott MW, Lamias MJ: Web survey design and administration. Public Opin Q 2001; 65:230-253.

19. Peytchev A, Couper MP, McCabe SE, et al: Web survey design: paging versus scrolling. Public Opin Q 2006; 70:596-607.

20. Crawford SD, Couper MP, Lamias MJ: Web surveys: perceptions of burden. Soc Sci Comput Rev 2001; 19:146-162.

21. Smyth JD, Dillman DA, Christian LM, et al: Comparing check- all and forced-choice question formats in web surveys. Public Opin Q 2006; 70:66-77.

22. DeRouvray C, Couper MP: Designing a strategy for reducing "no opinion" responses in web-based surveys. Soc Sci Comput Rev 2002; 20:3-9.

23. Reinking MF, Austin TM, Hayes AM: Exercise-related leg pain in collegiate cross-country athletes: extrinsic and intrinsic risk factors. J Orthop Sports Phys Ther 2007; 37:670-678.

24. Cook C, Heath F, Thompson RL: A meta-analysis of responserates in web- or internet-based surveys. Educ Psychol Meas 2000; 60:821-836.

25. Porter SR, Whitcomb ME: The impact of contact type on web survey response rates. Public Opin Q 2003; 67:579-588.

26. Goritz A: Incentives in web surveys: methodological issues and a review. Int J Internet Sci 2006; 1:58-70.

27. Wright KB: Researching Internet-based populations: advantages and disadvantages of online survey research, online questionnaire authoring software packages, and web survey services [electronic journal]. J Computer-Mediated Commun 2005; 10:Article 11. http:// jcmc.indiana.edu/vol 10/issue 3/wright.html.

28. Crawford S: Evaluation of web survey data collection systems. Field Methods 2002; 14:307-321. Tricia M. Austin, PhD, PT, ATC

Randy R. Richter, PhD, PT

Mark F. Reinking, PhD, PT, SCS, ATC

Dr. Austin is Assistant Professor, Dr. Richter is Associate Professor, and Dr. Reinking is Associate Professor, Program in Physical Therapy, Doisy College of Health Sciences, Saint Louis University, St. Louis, MO.

Received May 3, 2007; revision accepted February 4, 2008.

Address correspondence and reprint requests to: Tricia M. Austin, PhD, PT, ATC, Program in Physical Therapy, Doisy College of Health Sciences, Saint Louis University, St. Louis, MO 63104. Tel 314-977- 8505; fax 314-977-8513; email austintm@slu.edu.

Copyright Association of Schools of Allied Health Professions Fall 2008

(c) 2008 Journal of Allied Health. Provided by ProQuest LLC. All rights Reserved.