Portal de Periódicos da CAPES

Mail Surveys Can Achieve High Response Rates in a Dermatology Patient Population

2004; Elsevier BV; Volume: 122; Issue: 1 Linguagem: Inglês

10.1046/j.0022-202x.2003.22130.x

1523-1747

Jennifer Filip, Michael E. Ming, Ross M. Levy, Ole Hoffstad, David J. Margolis,

Allergic Rhinitis and Sensitization

In order to understand the burden of skin disease on patient populations, researchers need to be able to measure exposures and outcomes of interest in a population-based study. One method of obtaining such information is mailed patient surveys. This method of obtaining information, however, raises concerns regarding possible low response rates, and to the best of our knowledge, this method has not been attempted in large dermatology patient populations with a high response rate. We sought to determine whether using a slightly modified version of the Dillman Total Design Method as a mailed survey protocol would result in a high response rate in a dermatology population. A mail survey was sent using a slightly modified version of Dillman's Total Design Method to 4894 patients seen in the Dermatology Department of the University of Pennsylvania, who were diagnosed with nonmelanoma skin cancer, dermatophytosis, acne rosacea, seborrheic keratosis, or warts; 74.1% of the subjects responded to the mailings; 69.8% (n=3203) of patients returned a completed survey. Response rates (both as overall response and as survey completers only) were high across a wide spectrum of dermatologic illnesses and did not vary significantly by entry diagnosis. Dillman's Total Design Method appears to be an effective tool for researchers studying the burden of skin disease in a large sample of dermatology patients. In order to understand the burden of skin disease on patient populations, researchers need to be able to measure exposures and outcomes of interest in a population-based study. One method of obtaining such information is mailed patient surveys. This method of obtaining information, however, raises concerns regarding possible low response rates, and to the best of our knowledge, this method has not been attempted in large dermatology patient populations with a high response rate. We sought to determine whether using a slightly modified version of the Dillman Total Design Method as a mailed survey protocol would result in a high response rate in a dermatology population. A mail survey was sent using a slightly modified version of Dillman's Total Design Method to 4894 patients seen in the Dermatology Department of the University of Pennsylvania, who were diagnosed with nonmelanoma skin cancer, dermatophytosis, acne rosacea, seborrheic keratosis, or warts; 74.1% of the subjects responded to the mailings; 69.8% (n=3203) of patients returned a completed survey. Response rates (both as overall response and as survey completers only) were high across a wide spectrum of dermatologic illnesses and did not vary significantly by entry diagnosis. Dillman's Total Design Method appears to be an effective tool for researchers studying the burden of skin disease in a large sample of dermatology patients. There is growing concern about our lack of knowledge regarding the burden of skin disease on the population at large, such that in September of 2002 the National Institutes of Arthritis, Musculoskeletal and Skin Diseases convened a conference on this topic. One reason for this lack of awareness is that current data sources are insufficient for a detailed examination of this topic (The National Institutes of Health (NIH), 2002The National Institutes of Health (NIH) NIH conference. National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS).Workshop on the Burden of Skin Diseases. 2002: 4-5〈http://www.niams.nih.gov/ne/reports/sci_wrk/,2002/burden_skin_disease.htm〉Google Scholar). Population-based surveys, whether by in-person interviews, by telephone, or by mail, can be used to develop appropriate data sources. In recent decades, mail surveys have become increasingly popular and have been successfully used as research tools when response rates are high (Fox et al., 1988Fox R.J. Crask M.R. Kim J. Mail survey response rate. A meta-analysis of selected techniques for inducing response.Public Opin Q. 1988; 52: 467-491Crossref Scopus (448) Google Scholar). There is an ever-growing body of research examining possible methods for increasing response rate and reducing nonresponse bias in mail surveys, thereby increasing their overall cost-effectiveness. For example, using multiple mailings to make frequent contacts and the use of financial incentives have been shown to improve response rate in mail surveys (Fox et al., 1988Fox R.J. Crask M.R. Kim J. Mail survey response rate. A meta-analysis of selected techniques for inducing response.Public Opin Q. 1988; 52: 467-491Crossref Scopus (448) Google Scholar;Brennan and Hoek, 1992Brennan M. Hoek J. The behavior or respondents, nonrespondents, and refusers across mail surveys.Public Opin Q. 1992; 56: 530-536Crossref Scopus (41) Google Scholar;Dillman, 2000Dillman D.A. Mail and Internet Surveys: The Tailored Design Method. 2nd edn. Wiley, New York2000Google Scholar;Helgeson et al., 2002Helgeson J.G. Voss K.E. Terpening W.D. Determinants of mail-survey response: Survey design factors and respondent factors.Psychol Marketing. 2002; 19: 303-328Crossref Scopus (71) Google Scholar). Many features that have been shown to increase response rate are integrated into the Dillman Total Design Method (TDM) (Fox et al., 1988Fox R.J. Crask M.R. Kim J. Mail survey response rate. A meta-analysis of selected techniques for inducing response.Public Opin Q. 1988; 52: 467-491Crossref Scopus (448) Google Scholar;Brennan and Hoek, 1992Brennan M. Hoek J. The behavior or respondents, nonrespondents, and refusers across mail surveys.Public Opin Q. 1992; 56: 530-536Crossref Scopus (41) Google Scholar;Sutherland et al., 1996Sutherland H.J. Beaton N. Mazer R. Kriukov V. Boyd N.F. A randomized trial of the total design method for the postal follow-up of women in a cancer prevention trial.Eur J Cancer Prev. 1996; 5: 165-168Crossref PubMed Scopus (18) Google Scholar;Asch et al., 1997Asch D.S. Jedrziewski M.K. Christakis N.A. Response rate to mail surveys published in medical journals.J Clin Epidemiol. 1997; 50: 1129-1136Abstract Full Text PDF PubMed Scopus (1763) Google Scholar;Dillman, 2000Dillman D.A. Mail and Internet Surveys: The Tailored Design Method. 2nd edn. Wiley, New York2000Google Scholar). TDM is a set of techniques designed to increase response rate in mail surveys. These techniques include using a set of timed mailings that are personalized in appearance and tone that attempt to invoke a participant's sense of contribution towards solving a problem. TDM elicits a significantly higher response rate compared with many other mailing procedures (Fox et al., 1988Fox R.J. Crask M.R. Kim J. Mail survey response rate. A meta-analysis of selected techniques for inducing response.Public Opin Q. 1988; 52: 467-491Crossref Scopus (448) Google Scholar;Sutherland et al., 1996Sutherland H.J. Beaton N. Mazer R. Kriukov V. Boyd N.F. A randomized trial of the total design method for the postal follow-up of women in a cancer prevention trial.Eur J Cancer Prev. 1996; 5: 165-168Crossref PubMed Scopus (18) Google Scholar). Medical surveys have been conducted by mail on a number of patient populations for a broad range of specialties, but in general, very few population-based surveys have been conducted on skin diseases. The purpose of this study is to determine whether using a slightly modified version of Dillman's TDM to survey populations with dermatologic illness will result in a high response rate. We show that a carefully planned mail survey is an effective tool for studying skin disease in a large sample of dermatology patients. Of 4894 participants who were originally mailed a survey, 239 subjects had wrong addresses without forwarding information, and 64 subjects were deceased. Thus, we assumed that 4591 surveys were received by the intended subject. Table II represents the completion rate and response rate for the entire sample. Of the 4591 subjects who received a survey (excluding the deceased and incorrectly addressed), 3203 patients were completers (they returned a completed survey), 196 were refusers (they responded, but declined to fill out the survey), and 1192 were survey nonresponders. The overall completion rate was 69.8%, and the overall response rate was 74.1%.Table IICompletion rate and response rate for the entire sample, % (n)Response rateCompletersRefusersOverallNonrespondersDeceasedWrong addressesEntire sample, n=489465.4 (3203)4.0 (196)69.4 (3399)24.3 (1192)1.3 (64)4.9 (239)With deceased and wrong addresses removed n=459169.8 (3203)4.3 (196)74.1 (3399)26 (1192) Open table in a new tab Completers and noncompleters (nonresponders and refusers) were examined with regard to demographic information. Comparison of gender, entry diagnosis, location (living within the Philadelphia Metropolitan region or outside of the region), age, and race Table III found that there were no significant differences between completers and noncompleters with respect to gender and entry diagnosis (all p > 0.05 by analysis using a 2×2 table). A statistically significant difference between the two groups with respect to age, race, and location was noted. Specifically, those younger than 60 y of age or older than 79 y of age, those who were African American, and those who lived in Philadelphia were less likely to respond compared with other groups. Despite being statistically significant, however, the absolute differences in response are generally small, and are thus not likely to be of logistic importance when conducting or analyzing a survey.Table IIIDemographic information for survey completers and non-completers, % (n)CompletersNoncompletersTotaln=459169.8 (3203)30.2 (1388)4591Gender Female70.4 (1703)29.6 (716)2419 Male69.1 (1500)30.9 (672)2172Diagnosis Acne rosacea71.6 (541)28.4 (215)756 Skin cancer70.1 (1181)29.9 (504)1685 Warts70.2 (458)29.8 (194)652 Seborrheic keratoses68.3 (518)31.7 (242)760 Dermatophytes68.4 (505)31.6 (233)738Age 50–5968.8 (956)31.2 (433)1389 60–6972.2 (813)27.8 (313)1126 70–7974.6 (715)25.4 (243)958 80–8966 (303)34 (156)459 90+50 (33)50 (33)66Location Philadelphia Metropolitan area68.5 (2667)31.5 (1225)3892 Outside Metropolitan area76.7 (536)23.3 (163)699Race White71.1 (2569)28.9 (1046)3615 African American57.7 (160)42.3 (118)278 Asian65.4 (17)34.6 (9)26 Indian100 (2)02 Latino85.7 (12)14.3 (2)14 Other54.8 (34)45.2 (28)62 Unknown68.9 (409)31.1 (185)594 Open table in a new tab The results of the demographic analysis comparing responders (those who returned a completed questionnaire or refused by returning it blank) and nonresponders were similar, and are not reported. We found that 65.7% (n=2103) of the 3203 completers were early completers, who did so within the first 5 wk, and most likely in response to the first two mailings. Early and late completers were examined with regard to demographic information. Comparison of gender, diagnosis, location, age, and race variables found that whether participants returned their completed survey within the time frame defined as early (within 5 wk) or late, they were not significantly different from each other (all p>0.05 by analysis using a 2×2 table). We used a slightly modified version of the TDM to survey five groups of different diseases in a population of dermatology patients. Among all diagnoses, the completion rate was 69.8%; when refusers and completers were grouped together as responders, the response rate (i.e., individuals who responded in some fashion to the survey, which was determined by grouping together completers and refusers as responders) was 74.1%. Because of our low number of refusers, the completion rate and response rate for our survey were similar. Demographic characteristics of completers and noncompleters were generally similar, as were characteristics of responders and nonresponders, indicating that nonresponse bias may be of less concern than in some other studies; however, we did not have direct information on the severity of a patient's disease, the number of visits a patient made to his/her doctor (i.e., physician loyalty), or information on their socio-economic status, all which might have affected response rate. In general, our respondents are probably representative of our target population. Printing and assembling almost 5000 surveys is not an easy task. Our use of TDM was enhanced by the hiring of a fulfillment firm that was willing to follow our explicit instructions. Without this assistance, the ability to execute these mailings on time may have been compromised without hiring and training additional personnel. We also discovered an unexpected benefit of using postcards in the series of mailings. Whereas hardly any of the packets containing a survey were returned by the post office, 306 undeliverable postcards were returned from both postcard mailings. These returned postcards provided information on 239 incorrect addresses without forwarding information and 67 postcards with forwarding address information. This information was used to attempt contacting the 67 subjects who moved by mailing them an additional survey to their new address; additional follow-up attempts were not made beyond this single mailing. Thus, the postcards proved to be valuable, for they yielded information that might have otherwise been lost. Gender, diagnosis, living in or outside of the Philadelphia Metropolitan area, age, and race did not have any significant impact on whether those who completed a survey were likely to return it in response to the first two or last two mailings. By our estimates, the third and fourth mailings increased participation rates by an absolute 23.9% (the percentage of participants who responded late), which is an indication that these mailings were important in yielding a high response rate. This suggests that even if target response rates are met early it is worth devoting resources to enhancing response rate at later stages of the study. It has been noted that if initial expectations for response are not met, more intensive follow-up procedures such as telephone calls, certified mail, and additional reminders are effective in increasing overall response (Sudman, 1982Sudman S. Estimating response to follow-ups in mail surveys.Public Opin Q. 1982; 46: 582-584Crossref Scopus (6) Google Scholar;Dillman, 2000Dillman D.A. Mail and Internet Surveys: The Tailored Design Method. 2nd edn. Wiley, New York2000Google Scholar). This study yielded a relatively large response rate from our dermatologic sample in comparison with other published medical surveys. A review of 321 surveys published in medical journals reported that response rates by patient populations were on an average about 60% (SD=21%) (Asch et al., 1997Asch D.S. Jedrziewski M.K. Christakis N.A. Response rate to mail surveys published in medical journals.J Clin Epidemiol. 1997; 50: 1129-1136Abstract Full Text PDF PubMed Scopus (1763) Google Scholar). These investigators found that sending an additional written reminder with a copy of the instrument, and placing an additional phone call to nonresponders shortly after the initial survey mailing, resulted in a 13% higher response rate. Using the TDM to conduct a mail survey is a relatively inexpensive and straightforward method to collect data on a large sample of individuals. It may be possible to even further enhance this method. For example, a large survey in Norway yielded an 89% response rate (Eagan et al., 2002Eagan T.M.L. Eide G.E. Gulsvik A. Bakke P.S. Nonresponse in a community cohort study. Predictors and consequences for exposure–disease associations.J Clin Epidemiol. 2002; 55: 1123-1128Abstract Full Text Full Text PDF Scopus (93) Google Scholar). Researchers attributed their success to a simple questionnaire that covered nonsensitive topics and was advertised in radio and in newspapers. Whereas this may be more difficult to achieve in a more diverse population as the one in the US, such techniques are novel approaches to increasing response rate and are possible avenues to explore in future survey research. Our response rate may have been particularly high because older individuals may be more prone to return postal surveys than younger subjects (Etter and Perneger, 1997Etter J.F. Perneger T.V. Analysis of non-response bias in a mailed health survey.J Clin Epidemiol. 1997; 50: 775-781Abstract Full Text PDF Scopus (268) Google Scholar;Morris et al., 1998Morris C.M. Colditz G.A. Evans D.A. Response to a mail nutritional survey in an older bi-racial community population.J Clin Epidemiol. 1998; 50: 1117-1122Google Scholar). There is also evidence that surveys with university sponsorship elicit higher response rates (Dillman, 1978Dillman D.A. Mail and Telephone Surveys: The Total Design Method. Wiley, New York1978Google Scholar;Fox et al., 1988Fox R.J. Crask M.R. Kim J. Mail survey response rate. A meta-analysis of selected techniques for inducing response.Public Opin Q. 1988; 52: 467-491Crossref Scopus (448) Google Scholar;Groves et al., 1992Groves R.M. Cialdini R.B. Couper M.P. Understanding the decision to participate in a survey.Public Opin Q. 1992; 56: 475-495Crossref Scopus (469) Google Scholar); however, this would need to be researched further before it can be concluded that a large mail survey conducted in a nonacademic setting would not have as equally high a response rate. Furthermore, we feel it is important to mention that a successful response rate was achieved despite asking personal questions and using a nine-page instrument with 74 questions, which might be considered lengthy. It has been noted that in more burdensome surveys, the use of a prepaid incentive was found to significantly enhance response rate, and this may have contributed to a high response rate in this study (Church, 1993Church A.H. Estimating the effect of incentives on mail survey response rates: A meta-analysis.Public Opin Q. 1993; 57: 62-79Crossref Scopus (455) Google Scholar;Singer, 2002Singer E. The use of incentives to reduce nonresponse in household surveys.in: Groves R.M. Dillman D.A. Eltinge J.L. Roderick J.A.L. Survey Nonresponse. John Wiley & Sons, Inc., New York2002: 163-177Google Scholar). Limitations of this study include its generalizability. We evaluated only people in the Philadelphia region and surveyed only people who saw dermatologists in an academic center. We did however, survey subjects from both the main campus and a suburban clinic. Our results indicate there were no differences between these locations on response rate (data not shown). In addition, we did not survey all dermatologic illnesses; however, it appears that whether patients have a more worrisome illness (e.g., skin cancer) or a more routine illness (e.g., warts) does not affect their response rate. Another limitation of using mail surveys to obtain patient information may be the quality of data obtained by self-report methods. Whether mail surveys are a less accurate method than using an interview approach to data gathering is not readily apparent; however, conducting in-person interviews with a large population may be less cost-effective (e.g., requires training staff to conduct interviews) and less convenient for the subjects. Our results are encouraging, and demonstrate the feasibility of using TDM to study a population of dermatology patients. The burden of skin disease can be studied via a large survey approach, and can be applied to studying a spectrum of dermatologic illnesses. Participants were derived from a nested case–control study examining the relationship between atopic dermatitis and nonmelanoma skin cancer. As part of that study, patients were mailed a 74-item questionnaire asking about demographic information, past medical history, exposure to potential environmental risk factors, and medication use. The questionnaire employed either a multiple choice or yes/no format. After obtaining permission from physicians, the department chair, and the Institutional Review Board of the University of Pennsylvania, outpatients seen by a staff dermatologist between 1998 and 2001 through the University of Pennsylvania Health System were contacted to complete a mail survey. The cover letter in the initial mailing asked patients to participate in the study by either returning their completed survey, or to refuse participation by returning their incomplete survey in the provided return envelope. All subjects contacted were seen by a dermatology faculty member either at the main hospital in Philadelphia or in a single suburban office location. The patients who were invited to participate were seen by one of 30 doctors affiliated with the Dermatology Department of the Hospital of the University of Pennsylvania. Patients were randomly sampled from the administrative billing records of eligible patients who were over 45 y old and had diagnostic codes for nonmelanoma skin cancer (i.e., basal cell or squamous cell carcinoma), seborrheic keratoses, acne rosacea, warts, or dermatophyte infection. These patients were selected for having one of these five diagnoses, and each diagnosis was only used once. In total, 4894 patients were asked to participate. Patients were sent up to four mailings using a minor modification of TDM Table I. We modified TDM by using bulk rate instead of first class mail, not mailing participants a pre-notification letter, and sending a postcard as a final reminder and thank you note in place of a phone call or certified letter (Dillman, 2000Dillman D.A. Mail and Internet Surveys: The Tailored Design Method. 2nd edn. Wiley, New York2000Google Scholar). A professional fulfillment firm, under direction from the study's investigators, was hired to execute the mailings in a timely manner. The initial mailing contained a survey and personalized cover letter with a $1 bill attached as a financial incentive. A second mailing of a reminder postcard was sent to all subjects exactly 1 wk after the original mailing.Table IFeatures used in TDMFirst mailingMailing time point:Day 1Contents:Copy of survey with original subject ID number, cover letter, $1, return envelopeFeatures/personalization:White outgoing envelopes with University logo, metered postage, laser-printed addresses (addresses were not hand-written); cover letter addresses patient by name, mentions patient's doctor; uses digital (pre-printed) signatureSecond mailingMailing time point:Day 8Contents:PostcardFeatures/personalization:Reminded patients they should have received a survey a week ago, and asked them to please return the survey.Third mailingMailing time point:Day 28Contents:Second (replacement) survey with new ID number; cover letter with different message; return envelopeFeatures/personalization:White outgoing envelopes with university logo, metered postage, laser-printed addresses (addresses were not hand-written); cover letter addresses patient by name but not patient's doctor; uses digital (pre-printed) signature; sent only to those who had not yet responded.Fourth mailingMailing time point:Day 42Contents:Thank you postcardFeatures/personalization:General reminder to send in their survey or call if they needed an extra copy; thanking them if they already had done so; sent only to those who had not yet responded Open table in a new tab The two subsequent mailings were sent only to those who had not yet responded. The third mailing, sent 4 wk after the first survey was mailed, included a replacement copy of the survey and a personalized cover letter containing a more forceful message urging the recipient to participate. The fourth mailing was a postcard sent 3 wk after the third mailing. This postcard doubled as both a last reminder and a thank you card. Descriptive statistics were used to evaluate the study patients' demographic characteristics. Participants were initially divided into five groups: (1) completers (those who returned an at least partially completed survey); (2) refusers (those who responded to the mailings but did not wish to fill out the survey); (3) nonresponders (those who never responded to the mailings one way or the other); (4) those with wrong addresses and no further forwarding information (indicated by a returned postcard); and (5) those who were deceased (notified by a relative or spouse). Subjects with incorrect addresses or who were deceased were removed from further analysis. The primary analysis consisted of determining the completion rate (number of survey completers divided by the total number of subjects), and of comparing survey completers and noncompleters (refusers plus nonresponders). A secondary analysis was performed to determine the response rate (number of completers plus number of refusers, all divided by the total number of subjects) and of comparing responders (completers and responders) and nonresponders. Although demographic information was requested in the study questionnaire, demographic information was taken solely from the original billing database from which we gathered patient data in order to represent nonrespondents. The effect of repetitive mailings on completion rate and response rate was examined by using the dates participants returned their surveys. In order to determine this effect, we classified subjects into early responders or completers (those who returned the survey within the first 5 wk, likely as a result of the first two mailings) and late responders or completers (those who returned the survey at 5 wk or later, likely as a result of the third and fourth mailings). Analyses were then conducted to see whether there were differences between those who responded to the first two mailings, and those who responded to the last two mailings. The chronologic order and features of each timed mailing is presented in Table I. All analyses were performed using STATA 7.0 (Stata Corporation, College Station, Texas) or Statistical Package for the Social Sciences (SPSS) version 10 for Windows (SPSS, Inc., Chicago, Illinois). Statistical differences within groups were estimated using chi-square analysis and logistic regression and reported as likelihood ratio p-values. This study was approved by the Institutional Review Board of the University of Pennsylvania. This work was partially funded by Novartis Pharmaceuticals to the Trustees of the University of Pennsylvania and partially funded by a National Institutes of Health grant AR-02212.