3.研究步骤
研究步骤（Procedures）需要涵盖研究实施过程中每一步所完成的工作。主要包括：研究对象的招募、数据收集、干预实施（如果有的话）。
（1）研究对象的招募。在符合期刊对于图表数量要求的情况下，建议采用流程图对研究对象的招募进行补充说明。针对实验性研究，可以采用CONSORT流程图，如图2-1所示，从研究对象的纳入（enrollment）和筛选（screening）、随机分组（randomization）、随访（follow-up）到最后纳入统计分析（analysis）。在每个阶段，均需要说明研究对象脱落/失访的人数及相应的原因。针对非实验性研究，可以采用图2-2所示的方式，对研究对象的招募过程进行可视化描述。
（2）数据收集。在介绍此部分时，可从以下几个方面开展：who？when？where？how？what？说明由谁于什么时间段在哪里如何收集什么数据？下面将通过实例，进行详细阐述。示例2-43中，论文介绍了数据收集场所（outpatient and community settings）、时间段（April 2017-December 2018）、人员（Research assistants）。示例2-44中，论文介绍了数据收集方式（online webpage）、收集内容（questionnaire for...）及具体收集步骤。对于纵向研究，还需要说明随访时间点。如示例2-45所示，该研究总计收集了三个时间点的数据。如果对研究对象进行了补偿，需要说明何时提供的什么形式的补偿。如示例2-46所示，该研究在结束时为研究对象提供了50美元的补偿。
图2-1 实验性研究-CONSORT流程图
示例2-43 数据收集举例1
Data collection
Data were collected from outpatient and community settings in Southern and Central Italy between April 2017 and December 2018.Older MCC patients and their care partners were enrolled by trained research assistants，all of whom were registered nurses.
图2-2 非实验性研究-研究对象招募流程图
示例2-44 数据收集举例2
Data collection
The questionnaire was a version of the one already used for the RN4CAST study collected in 2017（Sermeus et al.，2011）...The questionnaire was only available online and data were collected using a secure institutional webpage.Full instructions were given to participants who consented to take part...The online link was accessible for approximately 4 months（September 2017-January 2018）.
示例2-45 数据收集举例3
Clinical and demographic information was collected by research coordinators.Measures at three time points were assessed：①within seven days before chemotherapy（prechemotherapy baseline；assessment 1［Al］）；②within 4 weeks of completion of chemotherapy（post-chemotherapy；assessment 2［A2］）；and③six-months after A2（six month follow-up：assessment 3［A3］）.Controls completed study assessments within the same time windows as the patients with breast cancer.
示例2-46 数据收集举例4
Participants received 1 h of works payment for completing both（TO and T1）measurements.
（3）干预方案（Intervention）。对于实验性研究，还需给出具体的干预方案，包括：干预次数、时长、实施者、形式（面对面或远程网络形式）。如示例2-47所示，该研究中介绍了实验组的干预形式（通过Sleepio Program这一线上干预模块）、干预者（animated virtual therapist）、干预次数（six）、干预时长（15～20 min）。另外，这里的干预措施已经在既往的论文中介绍过。所以，作者在引用该文献的同时，又对核心内容进行了概述。对照组未接受任何干预，只是被告知在研究结束以后，可以访问关于此干预项目的网站。
示例2-47 干预措施
Procedures
dCBT was delivered via the Sleepio program，the efficacy of which has been established in several randomized controlled trials（RCTs）［19，25，26）.Detailed information is available elsewhere［19］but in brief，intervention ingredients，covering key cognitive and behavioral strategies，are introduced by an animated virtual therapist in a personalized but automated manner over six 15-20 min sessions.Participants also have access to a library of sleep-related articles and an online community forum and can participate in weekly live discussions with a sleep expert.Participants randomized to the WLC arm were informed that they would be provided with access to dCBT at the end of the study（24-week follow-up）.
（4）测量工具（Outcomes and Measurement）。撰写测量工具也有章可循，首先要给出研究变量，然后说明相应的研究工具或测量方法。临床研究相关期刊，尤其是护理期刊，对于测量工具介绍方面要求较高。在介绍一个量表时，要说明该量表包含多少个条目和维度、适用的时间范围、每个条目如何计分、量表如何计算总分、总分是否分级、量表得分的含义。最重要的是，要介绍该量表的信度和效度，尤其是在你的研究人群中的内部一致性信度。如示例2-48所示，该论文介绍了研究变量（diabetes distress）、量表（DDS）、适用范围（the past month）、包含条目数（17 items）及维度（four domains）、每个条目的评分方法（6-point Likert scale）及量表得分代表的含义（higher score indicates greater level of distress）。最后，介绍了该量表已知的信度及其在本人群中的内部一致性信度（Cronbach's α）。通过这样一个简短却全面的介绍，使得读者对于该研究工具有了十分清晰的认识。当然，对于研究工具介绍的详细程度，因期刊而异。有时候，受字数限制，无法提供太多的细节。如果某测量工具未得到广泛应用或没有相关的引文支持，则需要着重说明为何要选择此工具。
示例2-48 研究工具的介绍
Diabetes distress
The Diabetes Distress Scale（DDS）［25］was used to measure diabetes distress.The DDS consists of 17 items that measure four domains of diabetes-related emotional distress over the past month：Emotional Burden，physician-related distress，regimen-related distress，and diabetes-related intrapersonal distress.Each item is scored on a 6-point Likert scale.A higher score indicates a greater level of distress.The DDS has high internal consistency reliability（Cronbach's a＞0.87）［25］.In this sample，the Cronbach's a of the DDS was 0.94.
以上是对主观问卷的介绍。同理，对于客观测量方法的介绍，也可参照此模板书写。包括所用检验方法、试剂浓度、试剂盒名称及商家。对于常见的测序方法，可以引用既往研究，如示例2-49所示。
示例2-49 测量方法介绍
Assays and Biomarkers
Urine riboflavin was assayed in the Medical University of South Carolina Clinical Neurobiology Laboratory（directed by R.A.）using standard/calibration curves constructed from known amounts of riboflavin against which unknown amounts of urine riboflavin were calculated by fluorescence detection（448 nm excitation/510 nm emission）.Values greater than 1300 ng/mL indicated adherence.The%dCDT was measured with a reference high-performance liquid chromatography assay.54 Using this international standardized assay，a value greater than 1.7%is close to100%specific for sustained heavy drinking in the weeks prior to testing and can be used to corroborate or independently evaluate drinking in clinical trials.58 Urine ethyl glucuronide（Microgenics Diagnostics）and other blood chemistries，including GGT，were measured with an autoanalyzer.(https://www.daowen.com)
（5）统计分析方法（Statistical Analysis）。本部分，首先需要详细说明使用的统计分析软件，并给出该软件的版本号、公司和地区。在介绍具体的统计分析方法之前，需要对是否存在缺失值、数据的正态性、是否有离群值等进行介绍。如果有缺失值，说明如何处理缺失值。其次，介绍具体的分析方法，包括统计描述和统计推断。如果所采用的统计方法非常普遍（例如，独立样本t检验），则不需进行详细描述。如果采用的统计分析方法比较不常见，需尽量给出引文，说明此方法的出处。最后，需要明确检验水准及检验是双侧还是单侧。对于某些比较复杂的统计分析方法，有些期刊会另外邀请一名统计学专家进行审稿。详见示例2-50、示例2-51和示例2-52。
示例2-50 统计分析软件及检验水准
All statistical tests were two-tailed，and the levels of statistical significance were set at P＜0.05 and P＜0.001.We used SAS 9.3（SAS Institute Inc.，Cary，NC）for all analyses.
示例2-51 检查缺失值
A preliminary data analysis was carried out to identify any inconsistencies or missing data.However，given the fixed responses，the only variable with missing data was one of the demographic open-ended items，the work experience，where 13.2%of the answers were missing.
示例2-52 重点介绍复杂的统计分析方法
Modified intention-to-treat analyses were conducted with the Linear Mixed Models（LMM）procedure，including all nursing staff who participated in one of the questionnaires...The LMM analyses included the fixed effects of group（intervention vs.control），and time as repeated measure（TO vs.T1），and group x time interaction for each well-being outcome measure（i.e.，general well-being，job satisfaction，or work engagement）.
（6）其他重要说明。对于RCT，在撰写研究方法时，建议参考Cochrane Risk of Bias 2 Tool中的条目，这有利于读者对研究质量进行评价。你的论文很可能被其他学者纳入系统综述和meta-analysis中，而在系统综述中，其中一个重要步骤就是对纳入文献进行质量评价。
在对RCT进行质量评价时，Cochrane Risk of Bias 2 Tool是目前应用最广泛的工具。该工具主要从以下5个方面对某研究的质量进行评价，包括：随机过程中的偏倚［随机分组（allocation randomization），分组隐匿（concealment）］、干预过程中的偏倚（是否对研究对象和实施干预的研究人员采取盲法）、缺失值带来的偏倚（失访情况）、测量过程中的偏倚（测量工具是否合适、测量方式是否合适）、汇报中的偏倚（是否根据预先制定的统计分析方法对数据进行分析）。该量表的具体内容及使用手册将在之后的章节进行详细介绍。
根据以上内容，在撰写RCT的研究方法时，需要注意以下几点。
1）研究对象的分组：说明是否进行了随机分组以及是否采用了分组隐匿（allocation concealment）。当然，这些是在研究设计阶段就应该要考虑到的。如示例2-53所示，该论文中明确说明了随机分组方法（按照1∶1的比例，使用软件进行分组）及隐匿方法（通过协调员trial coordinator进行隐匿，科研团队不知道分组情况）。类似地，如示例2-54所示，论文中明确说明了随机分组方法（使用软件进行分组）及分组隐匿方法（进行分组的研究人员不是该研究课题组成员）。
示例2-53 随机分组及隐匿方法举例1
Randomization and masking
Simple randomization with an allocation ratio of 1∶1 was carried out using the randomization function within Qualtrics Survey Software（Qualtrics，Provo，UT）on completion of baseline measures.The research team therefore had no access to future allocations and was unable to influence randomization.Other than the trial coordinator，who emailed participants their allocation，the research team remained blind to allocation.Contact between trial coordinator and participants was limited.
示例2-54 随机分组及隐匿方法举例2
Students were randomly assigned by a faculty member who did not take part in any of the study's stages using the Random system.This software generated a sequence of two numbers，1 and 2：group 1 was the Intervention Group and group 2 was the Control Group.The students were assigned to one of the two groups according to the random sequence as determined by the system.
2）干预的实施：这一部分，要特别说明是否对研究对象和实施干预的人员采取盲法。对于药物干预，多数情况下，可以做到双盲，即：研究对象和干预人员都不知道研究对象接受了哪个干预。如示例2-55所示，该研究是一个药物试验，采用了双盲，研究对象及研究人员都不知道分组情况（药物组或安慰剂组）。对于某行为或心理干预，研究对象不知道自己接受了哪个干预，而干预人员知道研究对象接受了哪个干预。因此，只能做到单盲而无法做到双盲。如示例2-56所示，该研究是一个行为干预，采用了单盲，研究对象不知道分组情况。干预组接受网络认知行为疗法，而对照组（wait-list control）是在研究结束后，才被授权可以获取网络认知行为疗法相关资源。对于多数行为或心理干预，对于研究对象和干预人员，均无法做到盲法。如示例2-57所示，该研究采用了open-label（即unmasked）这一研究设计，论文同时说明了无法对研究对象及研究人员实施盲法的原因。值得一提的是：针对研究是否采用了盲法，也可以放在“研究设计”那一部分进行介绍。
示例2-55 药物试验（双盲）
Design，setting，and participants：This double-blind randomized clinical trial conducted between November 2014 and June 2018 evaluated gabapentin vs placebo in communityrecruited participants screened and treated in an academic outpatient setting over a 16-week treatment period.
示例2-56 心理行为干预（单盲）
The Defining the Impact of Sleep improvement on Cognitive Outcomes（DISCO）study was an online，two-arm，single-blind，randomized clinical trial of dCBT versus WLC.Participants meeting DSM-5 criteria for insomnia disorder were recruited from the community and screened，consented，assessed，and randomized on a 1∶1 basis to intervention or control using the online platform，Qualtrics（www.qualtrics.com）.Participants in the intervention arm received access to the dCBT program，Sleepio（www.sleepio.com）［19］，while the WLC group received access to the same program on completion of study follow-up（24 weeks）.
示例2-57 行为干预-无盲法
Study design and patients：The ISAACC study（Impact of Sleep Apnea syndrome in the evolution of Acute Coronary syndrome.Effect of intervention with CPAP）is a multicenter，open-label，parallel-group，randomized controlled trial of patients with ACS at 15 hospitals across Spain（appendix p 5）.Because of the nature of the intervention，the trial intervention could not be masked to either investigators or patients.
3）数据收集：对于RCT，还需要说明数据收集过程中，是否对数据收集者进行了盲法。如示例2-58和示例2-59所示，数据收集/评估者不知道研究对象的分组情况。当然，有些研究无法在数据收集过程中实施盲法。此时，则无须特别强调。
示例2-58 数据收集者进行了盲法举例1
Data collectors were blinded to participants' gtoup assignment at baseline and follow-up testing.
示例2-59 数据收集者进行了盲法举例2
A committee，whose members were unaware of the study group assignments，adjudicated the major cardiovascular outcomes specified in the protocol.