University of Southern California, Los Angeles, California.
RAND Corporation, Health Division, Los Angeles, California.
J Manipulative Physiol Ther. 2021 Oct;44(8):601-611. doi: 10.1016/j.jmpt.2022.02.004. Epub 2022 Jun 18.
The purpose of this study was to compare the crowdsourcing platform Amazon Mechanical Turk (MTurk) with in-person recruitment and web-based surveys as a method to (1) recruit study participants and (2) obtain low-cost data quickly from chiropractic patients with chronic low back pain in the United States.
In this 2-arm quasi-experimental study, we used in-person clinical sampling and web-based surveys from a separate study (RAND sample, n = 1677, data collected October 2016 to January 2017) compared with MTurk (n = 310, data collected November 2016) as a sampling and data collection tool. We gathered patient-reported health outcomes and other characteristics of adults with chronic low back pain receiving chiropractic care. Parametric and nonparametric tests were run. We assessed statistical and practical differences based on P values and effect sizes, respectively.
Compared with the RAND sample, the MTurk sample was statistically significantly younger (mean age 35.4 years, SD 9.7 vs 48.9, SD 14.8), made less money (24% vs 17% reported less than $30,000 annual income), and reported worst mental health than the RAND sample. Other differences were that the MTurk sample had more men (37% vs 29%), fewer White patients (87% vs 92%), more Hispanic patients (9% vs 5%), fewer people with a college degree (59% vs 68%), and patients were more likely to be working full time (62% vs 58%). The MTurk sample was more likely to have chronic low back pain (78% vs 66%) that differed in pain frequency and duration. The MTurk sample had less disability and better global health scores. In terms of efficiency, the surveys cost $2.50 per participant in incentives for the MTurk sample. Survey development took 2 weeks and data collection took 1 month.
Our results suggest that there may be differences between crowdsourcing and a clinic-based sample. These differences range from small to medium on demographics and self-reported health. The low incentive costs and rapid data collection of MTurk makes it an economically viable method of collecting data from chiropractic patients with low back pain. Further research is needed to explore the utility of MTurk for recruiting clinical samples, such as comparisons to nationally representative samples.
本研究旨在比较众包平台亚马逊土耳其机器人（MTurk）与现场招募和网络调查，作为一种方法：（1）招募研究参与者，（2）从美国慢性下背痛的脊医患者快速获得低成本数据。
在这项 2 臂准实验研究中，我们使用现场临床抽样和来自另一项研究的网络调查（RAND 样本，n=1677，数据收集于 2016 年 10 月至 2017 年 1 月）与 MTurk（n=310，数据收集于 2016 年 11 月）进行比较，作为抽样和数据收集工具。我们收集了接受脊医治疗的慢性下背痛成年人的患者报告的健康结果和其他特征。进行了参数和非参数检验。我们分别根据 P 值和效应大小评估了统计学和实际差异。
与 RAND 样本相比，MTurk 样本在统计学上明显更年轻（平均年龄 35.4 岁，标准差 9.7 vs 48.9，标准差 14.8），收入较低（24%的人报告年收入低于 30000 美元，而 17%的人报告年收入低于 30000 美元），心理健康状况比 RAND 样本差。其他差异包括 MTurk 样本中男性比例较高（37% vs 29%）、白人患者比例较低（87% vs 92%）、西班牙裔患者比例较高（9% vs 5%）、大学学历比例较低（59% vs 68%）和全职工作的患者比例较高（62% vs 58%）。MTurk 样本更有可能患有慢性下背痛（78% vs 66%），疼痛频率和持续时间不同。MTurk 样本的残疾程度和总体健康评分较低。就效率而言，MTurk 样本的每位参与者的调查成本为 2.50 美元。调查开发用时 2 周，数据收集用时 1 个月。
我们的研究结果表明，众包和基于诊所的样本之间可能存在差异。这些差异在人口统计学和自我报告的健康方面从小到中不等。MTurk 的低激励成本和快速数据收集使其成为从下背痛脊医患者收集数据的经济可行方法。需要进一步研究来探索 MTurk 在招募临床样本方面的效用，例如与全国代表性样本进行比较。
