A randomized trial of blood donor recruitment strategies

SAVE OFFLINE

Blackwell Publishing IncMalden, USATRFTransfusion0041-11322006 American Association of Blood BanksJuly 200646710901096Original ArticleBLOOD DONOR RECRUITMENT STRATEGIESREICH ET AL.

BLOOD DONORS AND BLOOD COLLECTION A randomized trial of blood donor recruitment strategies Pascale Reich, Paula Roberts, Nancy Laabs, Artina Chinn, Patrick McEvoy, Nora Hirschler, and Edward L. Murphy

BACKGROUND: Improvement in donor return rates, especially among first-time donors, may significantly improve the blood supply. There are few rigorous studies of the effectiveness of various approaches to donor recruitment, however. STUDY DESIGN AND METHODS: By use of a singleblind, randomized trial design, 6919 post-September 11, 2001, first-time donors were randomly assigned into the following intervention arms: T-shirt incentive versus none, recruitment scripts with a patient story (Script A) versus a complimentary message including the donor’s blood type (Script B), and telephone versus e-mail recruitment. Our primary outcome was a second donation within 6 months. Rate ratios (RRs) and 95 percent confidence intervals (95% CIs) were calculated with Taylor series methods. RESULTS: A total of 1421 (20.5%) first-time donors returned within the 6 months, including 1252 with a second and 169 with both second and third donations. The T-shirt incentive was not effective in increasing returns compared to no incentive (20.5% vs. 20.6%; RR, 0.99; 95% CI, 0.91-1.09). Script A was significantly more effective than Script B (22.2% vs. 18.9%; RR, 1.18; 95% CI, 1.07-1.29). E-mail was substantially less effective than telephone recruitment (13.2% vs. 27.8%; RR, 0.48; 95% CI, 0.40-0.57). CONCLUSIONS: A T-shirt incentive had no apparent effect, but an empathetic message significantly improved the return donation rate. E-mail recruitment was substantially less effective than telephone recruitment, perhaps due to technical problems. The study illustrates the utility of the randomized clinical trial study design for testing donor recruitment strategies.

1090

TRANSFUSION

Volume 46, July 2006

T

here is little scientific evidence to guide decisions relating to blood donor recruitment. Although several comprehensive reviews have been written about donor motivation and behavior,1-4 most decisions regarding recruitment strategies are made on the basis of institutional tradition, personal preference, or anecdotal reports. In a review of the literature, we could find only a few examples of experimental trials testing different approaches to donor recruitment for blood or marrow donation. Hayes and coworkers5 randomly assigned active donors, inactive donors, and nondonors to receive three types of telephone recruitment, namely, standard donor recruitment compared to two types of experimental recruitment scripts inspired by the psychological theory of self-perception.5 They found that a small antecedent request improved compliance with a donation request, but that a large antecedent request did not. A study by Sarason and associates6 examined an intervention based on donor self esteem for its effectiveness in increasing participation of blood donors in a marrow registry. Wholeblood donors were randomly assigned into an intervention group, which complimented them on being blood ABBREVIATIONS: BCP = Blood Centers of the Pacific; RR = rate ratio(s); UBS = United Blood Services.

From the Blood Centers of the Pacific, San Francisco, California; Blood Systems, Inc., Scottsdale, Arizona; the University of California San Francisco, San Francisco, California; and Blood Systems Research Institute, San Francisco, California. Address reprint requests to: Edward L. Murphy, MD, MPH, Departments of Laboratory Medicine and Epidemiology/ Biostatistics, University of California San Francisco, and Blood Systems Research Institute, 270 Masonic Avenue, San Francisco, CA 94118; e-mail: [email protected]. Supported in part by a grant from the Blood Systems Foundation, with salary support to ELM from the National Heart, Lung, and Blood Institute REDS Contract N01-47114 and Grant R01-HL-62235. Received for publication December 2, 2005; revision received December 26, 2005, and accepted March 20, 2005. doi: 10.1111/j.1537-2995.2006.00856.x TRANSFUSION 2006;46:1090-1096.

BLOOD DONOR RECRUITMENT STRATEGIES

donors and asked them to complete a self-descriptive questionnaire, versus two control groups. The intervention group joined the marrow registry at more than two times the rate of the control group receiving no intervention. In 1994 Gimble and coworkers7 performed a randomized assignment of pairs of similar donor workplaces to the distribution 2 weeks before scheduled blood drives of an informational brochure developed according to Consumer Information Processing Theory. The brochure had no effect on recruitment or return behavior of donors at these workplaces, but analysis of covariates revealed a negative association between the number of blood units collected and blood drives held during vacations or during busy times at the workplace and with making announcements over the workplace loudspeaker before the drive. A positive effect on units collected was found following special appeals regarding a specific patient, an accident, or a natural disaster. A couple of studies compared recruitment techniques with a nonrandomized observational design. Ford and Wallace8 found that face-to-face recruiting produced donations in 15 percent of those recruited compared to 5 percent for telephone recruitment, and nondonors preferred donation at a hospital while donors preferred workplace donation. Stroncek and associates9 compared success rates in recruiting potential marrow donors from three groups: apheresis donors, whole-blood donors, and non–blood donors recruited through community appeals. The highest rate of compliance with providing a blood specimen to the marrow program was among wholeblood donors (87%), followed by non–blood donors from the community (78%) and apheresis donors (76%); none of these differences were significant. Many blood banks have experienced recent declines in their donor base, caused in part by deferrals aimed at preventing transfusion-transmitted infections. Concern about blood shortages has focused increased attention on ways to recruit new donors and to increase the return rate of existing donors. The September 11, 2001, terrorist attacks caused a large influx of first-time donors at our centers.10 Historically, only 20 percent of first-time blood donors return to give a second donation within 6 months, and 40 percent ever return.11 We used this opportunity to formally test recruitment strategies aimed at better retention of these first-time blood donors.

MATERIALS AND METHODS

month after September 11, 2001, were randomly assigned to the interventions, and return rates were measured over the subsequent 6 months. These donors were identified with operational computer systems at the two blood centers. The study qualified for a waiver of consent from the University of California San Francisco Committee on Human Research because it was a low-risk study comparing operational recruitment techniques that would have been applied to the donors in any case, except for randomization. If the subjects had been aware that they were receiving different recruitment interventions, they might have changed their return behavior and compromised the results of the study. At the end of the 6-month observation period, we mailed participants an information sheet telling them that they were part of a study and describing its design.

Interventions and endpoints Donors at both blood centers were randomized with a 2by-2 factorial design (see Fig. 1). Each donor had an equal chance of being randomized into groups receiving a Tshirt offer or no incentive and two different recruitment messages. In addition, donors at BCP who had an e-mail address on file were further randomized to receive the recruitment message by e-mail or telephone. Randomization was accomplished by sorting a list of eligible donors according to day and month of birth and then sequentially assigning donors to one of the four initial randomization groups. Donors at BCP with an e-mail address in the computer were further randomly assigned to either e-mail or telephone recruitment within each of the previous four blocks. There were too few UBS donors with e-mail addresses to perform this part of the study at that site. Each randomly assigned donor received an intervention code in his or her operational computer record, which served to indicate the type of recruitment for the recruit-

Script A, No Incentive

Script A, +Incentive

with e-mail: 50% by e-mail 50% by phone

with e-mail: 50% by e-mail 50% by phone

Script B, No Incentive

Script B, +Incentive

with e-mail: 50% by e-mail 50% by phone

with e-mail: 50% by e-mail 50% by phone

Study design and population The study used a randomized clinical trial design to compare three different recruitment interventions, each with an appropriate control. First-time, allogeneic, wholeblood donors at Blood Centers of the Pacific (BCP) in San Francisco, California, and United Blood Services (UBS) in metropolitan Phoenix, Arizona, who donated in the

Fig. 1. Latin squares randomization scheme showing four possible combinations of the incentive and recruitment scripts. At one blood center only, donors with e-mail addresses were further randomized to receive the intervention by e-mail or telephone. Hence, there were a total of eight (2 ¥ 2 ¥ 2) possible intervention arms in the study. Volume 46, July 2006

TRANSFUSION

1091

REICH ET AL.

ment staff for the duration of the 6-month follow-up period through April 2002. Randomly assigned donors were shielded from other routine recruitment campaigns at the blood center for the duration of the study by removing donors with randomization codes from call lists used by operational recruiters. The first intervention consisted of an offer of a T-shirt versus no such offer. Donors were told on the phone or by e-mail that following their next donation, a T-shirt would be mailed to their home address. T-shirts were not distributed at the time of donation to avoid provision of T-shirts (or envy at their distribution) among members of the other randomization arm. The second intervention was the comparison of two different recruitment scripts (see Appendix). Script A, aimed at donor empathy and altruism,12 included a story about a liver transplant patient who required blood transfusion; and Script B, aimed at donor self-esteem,6 mentioned that “special people like you are crucial to our blood supply” and gave them their blood type. Both scripts then asked the donor to return for another donation. Finally, among donors at BCP who provided an e-mail address, we compared e-mail versus telephone delivery of the recruitment message. Telephone recruitment was done by blood center telerecruitment staff following the scripts in the Appendix. An appointment could be made during the call; messages were left on voicemail if the donor did not answer the phone. Email texts were sent by mass e-mail and contained simple text messages together with links to the blood center’s Web site where appointments to donate could be made, and information about donation sites and times and eligibility was available. The primary study endpoint was completion of a second blood donation within 6 months after their first donation at the two blood centers. A secondary endpoint was completion of a third blood donation within 6 months after their first donation, measured only in donors who had completed a second donation. Endpoints were measured by searching for subsequent donations among donors within each of the randomization code groups in the operational computer systems; we could not measure donations at nonaffiliated blood centers. Subsequent donations were checked monthly, and completed donations counted as endpoints. We did not perform additional research monitoring of the penetration of the recruitment incentive offer or message; however, phone messages were delivered according to operational quality control, and e-mail return rates were monitored.

Statistical analysis Sample size calculations were based on a two-sided alpha level of 0.05, power of 0.80, and the ability to detect a 5 percent absolute increase (i.e., from 10% to 15%) in return rate for the 6-month primary endpoint. The calcu1092

TRANSFUSION

Volume 46, July 2006

lated sample size of about 725 per arm was then doubled to 1500 to reflect an approximate 50 percent successful contact rate by either telephone or e-mail. Following best practice in clinical trial data analysis, we used an “intentto-treat” analysis, which assumed that all donors randomized to a particular intervention group received that intervention, irrespective of actual penetration of the message. For each randomization group, 6-month second donation rates were calculated by dividing the number of second donations by the corresponding number of initially randomized subjects. Third donation rates were calculated in the same way, but with the denominator of donors who gave a second donation. To compare return rates between two randomization groups, we calculated rate ratios (RRs) with 95 percent confidence intervals (95% CIs) with computer software (EpiInfo, CDC, Atlanta, GA),13 which utilizes the Taylor series method to compute CIs.

RESULTS We randomly assigned a total of 6919 first-time blood donors who gave their first donation in the 1-month period after September 11, 2001, including 3948 at BCP and 2971 at UBS. Table 1 shows the demographic characteristics of these donors. Most were aged 20 through 49 years; women exceeded men at BCP but not at UBS. Data for race and education were available only at BCP. Minority donors were underrepresented compared to

TABLE 1. Demographic characteristics of the randomly assigned donors (percentage by center)

Characteristic Age (years)