2.3. Search strategy
The databases searched included the Korea Education and Research Information Service (RISS), the Korean studies Information Service System (KISS), DBpia, the National Digital Science Library (NSDL), PubMed (Medline), the Cochrane Library, Cumulated Index to Nursing and Allied Health Literature (CINAHL), and Embase. The search protocol was registered in PROSPERO International Prospective Register of Systematic Reviews (registration No. CRD42021231538 available at https://www.crd.york.ac.uk/prospero/# search advanced). As an example, the electronic search strategies for databases are shown in the supplementary file. Search functions, including Medical Subject Headings (MeSH) terminology, text words, and logical operators, were utilized correctly. We modified the search terms and used them according to the characteristics of each database after checking the MeSH term, synonym, and related terms, which were expressed as “aromatherapy or aroma” and “sleep quality or sleep” according to PICO in MeSH. A manual search for papers was conducted based on the references present in the studies included for analysis, and Google Scholar, Google, and Naver search engines were used for the comprehensive search of related research topics to avoid potentially missing literature.
2.4. Data collection
The databases were searched for nurse-led aromatherapy intervention studies between July 2019 and February 2020. One researcher conducted database searches by first excluding duplicate papers by title and abstract and then selecting studies according to the inclusion and exclusion criteria. The information of studies extracted from each search database was managed with Excel and EndNote, and reasons for inclusion and exclusion were recorded. Another researcher checked the search strategy in each database to examine if the search was reliably performed. Two researchers assessed the quality of each included study according to the Joanna Briggs Institute of Critical Appraisal checklist.50 Any disagreement among researchers was resolved through consultations. Articles selected according to the inclusion and exclusion criteria were recorded in the order of precedence. The final studies for inclusion were extracted based on the author, publication year, publishing country, research participants, research design, aromatherapy type, number of trials, intervention period, operation time per session, outcome variables, and methodological quality scores on the coding table. Microsoft Excel and Endnote 20 were used to record data and exclude duplicate studies.
2.5. Statistical analyses
The general characteristics of the studies are presented as frequency, percentage, and mean. The combined effect sizes, homogeneity, heterogeneity, and trim-and-fill method were analyzed statistically using MIX 2.0 Pro (Ver. 2.0.1.6, BiostatXL, 2017). First, the effect size was calculated using Hedges' g statistic; the 95 % confidence interval (CI) and weight of each effect size were used as the inverse of the variance. The overall effect size was calculated using a random-effects model that resets weights by considering the variation between participants from individual studies and the heterogeneity between each study. Second, the homogeneity and heterogeneity of the studies were tested by calculating the Cochran’s Q and Higgin's I2 values.51 Third, the publication bias of the selected studies was evaluated by correcting the effect size using the trim-and-fill method.52 The quality of the included studies (RCTs and quasi-experimental studies) was evaluated using the Joanna Briggs Institute of Critical Appraisal checklist50 to ensure that the scores were matched by a pilot test for each study design.
The quality evaluation for RCTs and quasi-experimental studies consisted of 12 and 18 items, respectively; the score for each evaluation item was assigned 0 (No, Unclear) or 1 (Yes).
