The response rates to physician postal surveys remain modest. The primary objective of this study was to assess the effect of tracking responses on physician survey response rate (i.e., determining whether each potential participant has responded or not). A secondary objective was to assess the effects of day of mailing (Monday vs. Friday) on physician survey response rate.
We conducted 3 randomized controlled trials. The first 2 trials had a 2×2 factorial design and tested the effect of day of mailing (Monday vs. Friday) and of tracking vs. no tracking responses. The third trial tested the effect of day of mailing (Monday vs. Friday). We meta-analyzed these 3 trials using a random effects model.
The total number of participants in the 3 trials was 1339. The response rate with tracked mailing was not statistically different from that with non-tracked mailing by the time of the first reminder (RR = 1.01 95% CI 0.84, 1.22; I2 = 0%). There was a trend towards lower response rate with tracked mailing by the time of the second reminder (RR = 0.91; 95% CI 0.78, 1.06; I2 = 0%). The response rate with mailing on Mondays was not statistically different from that with Friday mailing by the time of first reminder (RR = 1.01; 95% CI 0.87, 1.17; I2 = 0%), and by the time of the 2nd reminder (RR = 1.08; 95% CI 0.84, 1.39; I2 = 77%).
Citation: Akl EA, Gaddam S, Mustafa R, Wilson MC, Symons A, Grifasi A, et al. (2011) The Effects of Tracking Responses and the Day of Mailing on Physician Survey Response Rate: Three Randomized Trials. PLoS ONE 6(2): e16942. https://doi.org/10.1371/journal.pone.0016942
Editor: Joseph Ross, Yale University School of Medicine, United States of America
Received: October 28, 2010; Accepted: January 10, 2011; Published: February 23, 2011
Copyright: © 2011 Akl et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Research for Health in Erie County, Inc., and the Graduate Medical Education of the University at Buffalo funded the first study. The Lebanese American University funded the second study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Survey researchers employ various methods of questionnaire administration such as postal mail, electronic mail, online surveys, and phone interviews. Postal mail is commonly used as it is more convenient and efficient than phone surveys and provides higher response rates than electronic mail and online surveys . Achieving high response rates is an important goal for generalizing results of surveys to the population it targets. However, the response rate to mail surveys has been trending down threatening the validity of the results . Achieving adequate response rate is even more challenging with surveys of physicians .
Studies have shown that a number of strategies can improve the response rates to postal surveys . These strategies include incentives , , , , , shorter length of the questionnaire , , , listing general questions first in the questionnaire , providing prompts to complete the survey , , and using certified (as opposed to first class) mailing . However, all the above strategies appear to increase responses in small increments suggesting that combining a number of them may be necessary.
One factor that might affect the response rate is the day of the week on which the survey is received. This has been tested indirectly by comparing the day of the week on which the survey is mailed. Olivarius et al. conducted a survey on physicians, dispatching the questionnaire on Thursday versus Saturday and concluded that the probability of response was not influenced by receiving the postal questionnaire just before or just after the week-end . Pressley et al found no statistical significant difference in response rates by mailing the surveys on Monday vs. Friday . However his study was conducted on VIP executives, and not on physicians. Similarly, in a 2001 review of the literature, McColl et al concluded that the response rates are not affected by the day of mailing questionnaire .
Another factor that might affect response rate is tracking of responses, which refers to determining whether each potential participant has responded or not. Tracking is typically used to reduce time and cost expenditures of the survey by identifying and sending reminders to non-responders only. Tracking also helps in differentiating the demographic characteristics of responders from non-responders and in identifying any duplication of responses. Also, it allows record linkage when the same individuals are surveyed more than once over time. However, when the survey is on a sensitive topic, tracking responses might lower the response rates.
Asch surveyed nurses on topic of euthanasia and found that tracking responses (using coded post cards) lowered costs but significantly lowered the response rates . Campbell et al  and McDaniel  surveyed the general public and did not find any significant difference in the response rates of anonymous and non-anonymous groups. McKee surveyed members of an organization and found that the response rate from the coded survey was substantially higher than that of the non-coded survey . We identified no study conducted on physicians.
The primary objective of this study was to assess the effect of tracking responses on physician survey response rate. A secondary objective was to assess the effects of day of mailing (Monday vs. Friday) on physician survey response rate.
We conducted 3 randomized controlled trials. The first 2 trials were conducted respectively among directors of Family Medicine and Internal Medicine residency programs in the United States (US). They used a factorial design to test the effect of day of mailing (Monday vs. Friday) and of tracking vs. no tracking responses. The third trial was conducted with a group of practicing physicians and tested the effect of day of mailing (Monday vs. Friday).
The University at Buffalo Institutional Review Board approved both studies. No informed consent was required and the University at Buffalo Institutional Review Board approved the waiver of consent.
Two trials among program directors
We conducted the two trials in the setting of two national surveys of directors of Family Medicine and Internal Medicine residency programs in the US , . The surveys related to training of residents in the implementation of clinical practice guidelines. The surveys' questionnaire consisted of 2 single sided pages with 15 questions about the curriculum, the characteristics of the program director and the characteristics of the residency program.
We mailed the initial invitation to participate in the surveys in April 2007. We used the following survey methods to maximize response rate , : university sponsorship, personalized cover letter, colored ink, stamped return envelope, first class mailing, follow up mail, including a questionnaire in the follow up mail, non-monetary incentive, and a questionnaire that is interesting, short, user friendly, and with factual questions. The non-monetary incentive was mailed with the initial questionnaire and consisted of a Jeopardy-like game to teach clinical practice guidelines in a Microsoft PowerPoint file format on a CD.
Each trial had a 2×2 factorial design to evaluate tracking responses and the day of mailing. We first randomized subjects to mailing the survey on a Friday versus a Monday. At the same time we randomized subjects to have their responses tracked versus not tracked. In the tracked response group, the second (and last) single sided page of the questionnaire was perforated at 3" from the bottom. Below the perforation were a tracking number and a message that read: “This number is to avoid sending reminders to those who respond. We will separate this number from the responses to keep them anonymous” (Figure 1). Upon the receipt of the response, we detached the lower portion of the page at the perforation line to keep the data anonymous. We used this method to identify responders, but also to reassure participants that we are ensuring anonymity.
Five weeks after the initial invitation, we sent a follow up mail to non-responders in the tracked response group and to all subjects in the non-tracked response group. Nine weeks after the initial invitation we sent faxes in attempt to increase response rate; we do not include data beyond the point of sending the faxes as we did not organize it according to the randomization scheme. We used Microsoft Excel to generate a list of random numbers for the allocation of subjects to one of the four study arms.
Trial among practicing physicians
We surveyed Lebanese medical graduates practicing medicine in the United States regarding their intentions to relocate to Lebanon or Arabic Gulf countries (results of this survey not published yet). The questionnaire included 7 questions about family factors and 3 questions about willingness to relocate. We used the same techniques to improve response rate to this second trial as in the 1st trial except for the use of an incentive.
We conducted the trial to evaluate effect of the day of mailing on response rates. We randomized subjects to mailing the survey on a Friday versus a Monday. We sent the initial invitation to participate in the survey by first class postal mail. Two weeks later, we sent a first reminder by mail to non respondents. For those who still did not respond two weeks later, we sent a second reminder by fax when a fax number was available and attempted a phone contact as a reminder if a phone number was available.
For the tracking comparison, we calculated for each of the 2 related trials (among program directors) the risk ratio for response rates by the time of first reminder and the risk ratio for the response rate by the second reminder. We then conducted a meta-analysis pooling the results of the 2 studies at each of the 2 times using a random effects model. This meta-analysis was not based on an exhaustive systematic review of the literature and thus did not include all potentially eligible studies. We tested results for homogeneity across studies using the I2 test. We followed a similar procedure for the day of the week comparison using the 3 related trials (among program directors and practicing physicians). We calculated the sample size to test the hypothesis of the survey ,  and not specifically to test the hypothesis of the current trial. In other words, the primary outcomes based on which we conducted the power analysis related to the specific subject of the survey and not the surveying trial outcomes.
Concerns have been expressed about the potential for unrecognized interactions between interventions in factorial trials to distort their published results and interpretations . We thus examined interaction between the 2 interventions in the factorial design trials (tracking and day of the week) by generating an “interaction ratio” that compares the effects of each treatment in the presence and absence of the other treatment . The resulting interaction ratio and its 95% confidence interval showed no interaction .
The total number of participants in the 3 trials was 1339. Table 1 shows the number of participants randomized and the number responding for each trial and for each of the intervention arms. The response rates by the end of second reminder (not including fax and telephone reminders) for the 3 trials were 49% (Family Medicine), 39% (Internal Medicine) and 49% (practicing physicians' survey). The overall response rates by the end of the study for the 3 trials were respectively 52% (Family Medicine survey), 51% (Internal Medicine survey) and 57% (practicing physicians' survey).
The risk ratios of response by the time of first reminder for tracked vs. non-tracked mailings for the Family Medicine and Internal Medicine surveys were 1.04 (95% CI 0.82, 1.31) and 0.98 (95% CI 0.73, 1.31) respectively. The pooled risk ratio of response by the time of first reminder was 1.01 (95% CI 0.84, 1.22; I2 = 0%) (Figure 2). The risk ratios of response by the time of second reminder for tracked vs. non-tracked mailings for the Family Medicine and Internal Medicine surveys were 0.92 (95% CI 0.76, 1.11) and 0.89 (95% CI 0.70, 1.15) respectively. The pooled risk ratio of response by the time of second reminder was 0.91 (95% CI 0.78, 1.06; I2 = 0%) (Figure 3).
The risk ratios of response by the time of first reminder for Friday vs. Monday mailings for the Family Medicine, Internal Medicine and Practicing physicians' surveys were 1.09 (95% CI 0.86, 1.38), 0.93 (95% CI 0.69, 1.24), 0.99 (95% CI 0.78, 1.25). The pooled risk ratio of response by the time of first reminder was 1.01 (95% CI 0.87, 1.17; I2 = 0%) (Figure 4). The risk ratios of response by the time of second reminder for Friday vs. Monday mailings for the Family Medicine, Internal Medicine and Practicing physicians' surveys were 1.38 (95% CI 1.13, 1.67), 0.88 (95% CI 0.68, 1.13), 1.02 (95% CI 0.86, 1.22). The pooled risk ratio of response by the time of second reminder was 1.08 (95% CI 0.84, 1.39; I2 = 77%) (Figure 5).
We evaluated the effects of tracking responses and the day of mailing on physician survey response rate in three trials. The response rate with tracked mailing was not statistically different from that with non-tracked mailing by the time of the first reminder. There was a trend towards lower response rate with tracked mailing by the time of the second reminder. The response rate with Monday mailing was not statistically different from that with Friday mailing by the times of first reminder. The results corresponding to the time of the 2nd reminder were inconsistent but suggest no effect.
Cost being one additional consideration with tracking mail, we conducted a post hoc cost analysis. The cost of follow-up mailing for the tracked group vs. non-tracked group was $455 and $649 respectively in the Internal Medicine survey, and $489 and $775 respectively in the Family Medicine survey.
This study has a number of strengths. We conducted three separate trials that we pooled together to improve the precision of our analyses and explore differences between the findings. We believe these are the first trials estimating the effects of tracking responses on response rates to surveys of physicians. We tested the interventions of interest while implementing most of the techniques shown to improve response rates. The heterogeneity across included trials was very low for 3 of 4 analyses which increase our confidence in the results.
One limitation of the study relates to the sample size that did not allow a more accurate estimation of effects despite the use of meta-analysis. Also, the response rates for these surveys were not high (the highest being 50.4 by the time of second reminder). One possibility, although unlikely, is that the increased response rates in the non-tracking group might be due to some respondents answering more than one questionnaire. Finally, the generalizability of these findings might be limited given the studies were conducted among physicians.
The trend toward a lower response rate with tracked mailing is consistent with the finding by Asch et al. that tracking using coded post cards lowered the response rates . That survey was conducted on nurses was on sensitive topic of euthanasia. However, the results were not consistent with the remainder of the literature. Campbell and McDaniel conducted surveys on general population and customers of retail store respectively and did find any effect of anonymity on response rates. McKee et al. surveyed members of a national non-profit professional organization and used coded number at the top of front page of the questionnaire with explanation about the intent to follow-up. They found that coding actually increased the response rates. While these differing results might relate to the sensitivity of the topic, they might also depend on the type of population surveyed.
All the articles we identified on the effect of the day of mailing concluded that there is no statistically significant influence on the response rates , , . In our study, we assumed that surveys mailed on Monday would improve response rate by reaching the participants by mid of the week when they might have more time to respond. Our findings do not support this assumption. In fact, the day of mailing may not be perfectly associated with the day of reading. The latter depends on when the mailroom at the receiving institution actually delivers the letter, when the residency director actually opens the letter, etc.
Implications for conducting surveys
The decision to track responses will depend on the researchers' judgment whether the time and cost savings are worth the potential loss in response rate. One could assume that the less sensitive the survey topic is, the more the balance will be in favor of tracking. Also the survey could cost significantly less when the responses are tracked. As for the day of the week, we suggest that survey researchers not restrict their survey plans to a specific day of the week for the aim of improving response rate.
We thank Ethel Sharp and Karen Devlin for their administrative assistance. We thank Dr. Sameer K Gunukula for his input. We also thank all those who participated in the surveys. EAA was a recipient of the Society of General Internal Medicine Founders Award.
Conceived and designed the experiments: EAA RM MCW HJS. Analyzed the data: EAA SG. Contributed reagents/materials/analysis tools: RM AS AG DM. Wrote the manuscript: EAA SG. Critical revision of the manuscript for important intellectual content: EAA SG RM MCW AS AG DM HJS. Obtained funding: EAA RM HJS.
- 1. Aitken C, Power R, Dwyer R (2008) A very low response rate in an on-line survey of medical practitioners. Australian & New Zealand Journal of Public Health 32: 288–289.C. AitkenR. PowerR. Dwyer2008A very low response rate in an on-line survey of medical practitioners.Australian & New Zealand Journal of Public Health32288289
- 2. Curtis EA, Redmond RA (2009) Survey postal questionnaire: optimising response and dealing with non-response. Nurse Researcher 16: 76–88.EA CurtisRA Redmond2009Survey postal questionnaire: optimising response and dealing with non-response.Nurse Researcher167688
- 3. Field TS, Cadoret CA, Brown ML, Ford M, Greene SM, et al. (2002) Surveying physicians: do components of the “Total Design Approach” to optimizing survey response rates apply to physicians? [Review] [63 refs]. Medical Care 40: 596–605.TS FieldCA CadoretML BrownM. FordSM Greene2002Surveying physicians: do components of the “Total Design Approach” to optimizing survey response rates apply to physicians? [Review] [63 refs].Medical Care40596605
- 4. Edwards PJ, Roberts I, Clarke MJ, Diguiseppi C, Wentz R, et al. (2009) Methods to increase response to postal and electronic questionnaires. Cochrane Database Syst Rev MR000008.PJ EdwardsI. RobertsMJ ClarkeC. DiguiseppiR. Wentz2009Methods to increase response to postal and electronic questionnaires.Cochrane Database Syst RevMR000008
- 5. Beebe TJ, Davern ME, McAlpine DD, Call KT, Rockwood TH (2005) Increasing response rates in a survey of Medicaid enrollees: the effect of a prepaid monetary incentive and mixed modes (mail and telephone). Medical Care 43: 411–414.TJ BeebeME DavernDD McAlpineKT CallTH Rockwood2005Increasing response rates in a survey of Medicaid enrollees: the effect of a prepaid monetary incentive and mixed modes (mail and telephone).Medical Care43411414
- 6. Collins RL, Ellickson PL, Hays RD, McCaffrey DF (2000) Effects of incentive size and timing on response rates to a follow-up wave of a longitudinal mailed survey. Evaluation Review 24: 347–363.RL CollinsPL EllicksonRD HaysDF McCaffrey2000Effects of incentive size and timing on response rates to a follow-up wave of a longitudinal mailed survey.Evaluation Review24347363
- 7. Delnevo CD, Abatemarco DJ, Steinberg MB (2004) Physician response rates to a mail survey by specialty and timing of incentive. American Journal of Preventive Medicine 26: 234–236.CD DelnevoDJ AbatemarcoMB Steinberg2004Physician response rates to a mail survey by specialty and timing of incentive.American Journal of Preventive Medicine26234236
- 8. Harris IA, Khoo OK, Young JM, Solomon MJ, Rae H (2008) Lottery incentives did not improve response rate to a mailed survey: a randomized controlled trial. Journal of Clinical Epidemiology 61: 609–610.IA HarrisOK KhooJM YoungMJ SolomonH. Rae2008Lottery incentives did not improve response rate to a mailed survey: a randomized controlled trial.Journal of Clinical Epidemiology61609610
- 9. John EM, Savitz DA (1994) Effect of a monetary incentive on response to a mail survey. Annals of Epidemiology 4: 231–235.EM JohnDA Savitz1994Effect of a monetary incentive on response to a mail survey.Annals of Epidemiology4231235
- 10. Beebe TJ, Stoner SM, Anderson KJ, Williams AR (2007) Selected questionnaire size and color combinations were significantly related to mailed survey response rates. Journal of Clinical Epidemiology 60: 1184–1189.TJ BeebeSM StonerKJ AndersonAR Williams2007Selected questionnaire size and color combinations were significantly related to mailed survey response rates.Journal of Clinical Epidemiology6011841189
- 11. Jepson C, Asch DA, Hershey JC, Ubel PA (2005) In a mailed physician survey, questionnaire length had a threshold effect on response rate. Journal of Clinical Epidemiology 58: 103–105.C. JepsonDA AschJC HersheyPA Ubel2005In a mailed physician survey, questionnaire length had a threshold effect on response rate.Journal of Clinical Epidemiology58103105
- 12. Kalantar JS, Talley NJ (1999) The effects of lottery incentive and length of questionnaire on health survey response rates: a randomized study. Journal of Clinical Epidemiology 52: 1117–1122.JS KalantarNJ Talley1999The effects of lottery incentive and length of questionnaire on health survey response rates: a randomized study.Journal of Clinical Epidemiology5211171122
- 13. Drummond FJ, Sharp L, Carsin AE, Kelleher T, Comber H (2008) Questionnaire order significantly increased response to a postal survey sent to primary care physicians. Journal of Clinical Epidemiology 61: 177–185.FJ DrummondL. SharpAE CarsinT. KelleherH. Comber2008Questionnaire order significantly increased response to a postal survey sent to primary care physicians.Journal of Clinical Epidemiology61177185
- 14. Pirotta M, Gunn J, Farish S, Karabatsos G (1999) Primer postcard improves postal survey response rates. Australian & New Zealand Journal of Public Health 23: 196–197.M. PirottaJ. GunnS. FarishG. Karabatsos1999Primer postcard improves postal survey response rates.Australian & New Zealand Journal of Public Health23196197
- 15. Rikard-Bell G, Ward J (2000) Maximizing response rates to a survey of dentists: a randomized trial. Australian Dental Journal 45: 46–48.G. Rikard-BellJ. Ward2000Maximizing response rates to a survey of dentists: a randomized trial.Australian Dental Journal454648
- 16. Del Valle ML, Morgenstern H, Rogstad TL, Albright C, Vickrey BG (1997) A randomized trial of the impact of certified mail on response rate to a physician survey, and a cost-effectiveness analysis. Evaluation & the Health Professions 20: 389–406.ML Del ValleH. MorgensternTL RogstadC. AlbrightBG Vickrey1997A randomized trial of the impact of certified mail on response rate to a physician survey, and a cost-effectiveness analysis.Evaluation & the Health Professions20389406
- 17. Olivarius ND, Andreason AH (1995) Day-of-the-week effect on doctors' response to a postal questionnaire. Scand J Prim Health Care 13: ND OlivariusAH Andreason1995Day-of-the-week effect on doctors' response to a postal questionnaire.Scand J Prim Health Care13
- 18. Pressley MM, Dunn MG (1985) A Factor-Interactive Experimental Investigation of Inducing Response to Questionnaires Mailed to Commercial Populations. AMA EDUCATORS PROCEEDINGS 356–361.MM PressleyMG Dunn1985A Factor-Interactive Experimental Investigation of Inducing Response to Questionnaires Mailed to Commercial Populations.AMA EDUCATORS PROCEEDINGS356361
- 19. McColl E, Jacoby A, Thomas L, Soutter J, Bamford C, et al. (2001) Design and use of questionnaires: a review of best practice applicable to surveys of health service staff and patients. Health Technology Assessment. E. McCollA. JacobyL. ThomasJ. SoutterC. Bamford2001Design and use of questionnaires: a review of best practice applicable to surveys of health service staff and patients.Health Technology Assessment
- 20. Asch DA (1996) Use of a coded postcard to maintain anonymity in a highly sensitive mail survey: cost, response rates, and bias. Epidemiology 7: 550–551.DA Asch1996Use of a coded postcard to maintain anonymity in a highly sensitive mail survey: cost, response rates, and bias.Epidemiology7550551
- 21. Campbell MJ, Waters WE (1990) Does anonymity increase response rate in postal questionnaire surveys about sensitive subjects? A randomised trial. Journal of Epidemiology and Community Health 44: 75–76.MJ CampbellWE Waters1990Does anonymity increase response rate in postal questionnaire surveys about sensitive subjects? A randomised trial.Journal of Epidemiology and Community Health447576
- 22. McDaniel S (1981) An Investigation of respondent anonymity's effect on mailed questionnaire response rate and quality. Journal of the Market Research Society 23: 150–160.S. McDaniel1981An Investigation of respondent anonymity's effect on mailed questionnaire response rate and quality.Journal of the Market Research Society23150160
- 23. McKee DO (1992) The effect of using a questionnaire identification code and message about non-response follow-up plans on mail survey response characteristics. Market Research Society. DO McKee1992The effect of using a questionnaire identification code and message about non-response follow-up plans on mail survey response characteristics.Market Research Society
- 24. Akl E, Mustafa R, Wilson M, Symons A, Moheet A, et al. (2009) Curricula for teaching the content of clinical practice guidelines to family medicine and internal medicine residents in the US: a survey study. Implementation Science 4: 59.E. AklR. MustafaM. WilsonA. SymonsA. Moheet2009Curricula for teaching the content of clinical practice guidelines to family medicine and internal medicine residents in the US: a survey study.Implementation Science459
- 25. Akl EA, Gunukula S, Mustafa R, Wilson MC, Symons A, et al. (2010) Support for and aspects of use of educational games in family medicine and internal medicine residency programs in the US: a survey. BMC Med Educ 10: 26.EA AklS. GunukulaR. MustafaMC WilsonA. Symons2010Support for and aspects of use of educational games in family medicine and internal medicine residency programs in the US: a survey.BMC Med Educ1026
- 26. Edwards P, Roberts I, Clarke M, DiGuiseppi C, Pratap S, et al. (2007) Methods to increase response rates to postal questionnaires.[see comment]. Cochrane Database of Systematic Reviews MR000008.P. EdwardsI. RobertsM. ClarkeC. DiGuiseppiS. Pratap2007Methods to increase response rates to postal questionnaires.[see comment].Cochrane Database of Systematic ReviewsMR000008
- 27. Asch DA, Jedrziewski MK, Christakis NA (1997) Response Rates to Mail Surveys Published in Medical Journals. Journal of Clinical Epidemiology 50: 1129–1136.DA AschMK JedrziewskiNA Christakis1997Response Rates to Mail Surveys Published in Medical Journals.Journal of Clinical Epidemiology5011291136
- 28. Lubsen J, Pocock SJ (1994) Factorial trials in cardiology: pros and cons. Eur Heart J 15: 585–588.J. LubsenSJ Pocock1994Factorial trials in cardiology: pros and cons.Eur Heart J15585588
- 29. McAlister FA, Straus SE, Sackett DL, Altman DG (2003) Analysis and Reporting of Factorial Trials: A Systematic Review. JAMA 289: 2545–2553.FA McAlisterSE StrausDL SackettDG Altman2003Analysis and Reporting of Factorial Trials: A Systematic Review.JAMA28925452553