The protocol is registered in the International Prospective Register of Systematic Reviews (CRD42018092259). This protocol follows the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols 2015 guideline [25].

We have categorized the inclusion criteria for this systematic review according to study design, study participants, types of intervention, types of technology, and study setting.

We will include parallel randomized controlled trials (RCTs), quasi-RCTs, controlled clinical trials, crossover RCTs, paired repeated measures, cohort and case-control studies, and comparative cross-sectional studies that compare the electronic interviewer-administered survey with paper-based methods.

This review focuses on data quality outcomes from the data collected by data collectors who used paper or electronic data collection modalities during public health surveys. We will also include data collectors, supervisors, or data managers for opinion-, preference-, and usability-related analysis.

Any mobile device data collection tool that was designed to support interviewer-administered data collection processes in public health surveys will be included.

Electronic data collection in our review refers to portable, wireless digital devices usually supported by mobile network or satellite communication infrastructures, such as cell phones, smartphones, personal digital assistants, and tablet computers. The given support includes data capture and instant, stored, and forward transfers to the research center. We will include all apps with technologies that directly support the data collection process by enabling data collectors or interviewers to collect and send data as well as enabling supervisors or data managers to monitor the data collection process.

Our review will include national demographic surveys, demographic and health surveillance systems, and household surveys. We will include all countries and research facilities regardless of the socioeconomic status of the country.

We will exclude the following study types from the review: all studies that compare electronic and paper-based tools in self-administered surveys; studies that are performed in settings other than a house-to-house field survey (eg, electronic medical records and eCRFs); studies not performed on human subjects; studies reported before January 1, 2007; studies that are experience reports, letters, reviews, commentaries, and editorials; and non-English language publications.

The primary outcomes in this review will be data quality indicators and cost-effectiveness evidence. According to Bowling [9], data quality is a vague concept, and it is hard to find any gold standard or framework. Data quality could be defined in terms of survey response rates, questionnaire item response rates, the accuracy of responses, the absence of bias, and completeness of the information obtained from respondents [9]. For this review, we will focus on two data quality indicators: data completeness and data accuracy. Accuracy is hereby defined as the absence of typographical errors, decimal point faults, and illogical values, whereas the completeness of items is inversely proportional to the number of missing responses in the questionnaire. In general, we will compare the proportion of errors or missing items between electronic and paper-based data collection methods [9,11,26,27]. Cost-effectiveness outcomes will be measured using resource costing methods, which include provider perspective direct cost comparing the cost of conducting a survey using electronic and paper-based data collection methods. The secondary outcomes include work efficiency, usability, user experience, and acceptability.

We will conduct a systematic keyword search on electronic databases such as Medical Literature Analysis and Retrieval System Online, PubMed, CINAHL, PsycINFO, Global Health, Trip, ISI Web of Science, and Cochrane Library. In addition, we will screen the reference list citations of included articles. Unpublished and in-progress studies will be identified from the following trial registries: ClinicalTrials.gov, ISRCTN registry, Australian New Zealand Clinical trial Registry, International Clinical Trials Registry Platform.

We will restrict our search to articles published in English from 2007 to mid-2018 (as mobile devices that became available during this time are compatible with the mobile operating system framework that focuses on apps, and most of the EDC apps were tested during this period [11]).

The search strategy will consider 3 categories: the technology or intervention used (eg, mobile device, mobile phone, mHealth, or EDC), area of application (eg, data collection, demographic and health surveys, or large-scale surveys), and the outcome of interest (eg, data quality, missing data, and cost-effectiveness). We will connect all the similar terms in the same group with the Boolean operators “OR” and “AND” (Table 1).

Endnote software (Clarivate Analytics) will be used to import the retrieved literature from all databases to manage duplication and further screening. The Covidence (Veritas Health Innovation) Web-based screening tool will be used to import the set of deduplicated citations and to manage the title and abstract screening process. We will screen the titles and abstracts for the inclusion criteria. Potentially relevant full-text papers will be reviewed, including reference lists of these papers.

Two authors (AAZ and TN) will screen titles and abstracts independently and identify potentially eligible studies based on the eligibility criteria. Review author pairs will then screen the full-text reports and decide whether these meet the inclusion criteria. Any disagreements in each phase will be resolved first by discussion among the review authors using the prespecified inclusion criteria. If the disagreement or uncertainty continues, a consulting third author (FF) will mediate the final decision. We will note the reasons for inclusion and exclusion using a flowchart diagram.

An excel sheet for data extraction will be used based on the inclusion criteria and the objectives of the review. To ensure uniformity across reviewers, we will conduct a pretest standardization exercise before starting the data extraction process. Textbox 1 presents the data items that we will extract.

The primary outcomes of this review are the following data quality indicator parameters: error rates quantified from missing items (number of incomplete records per interview questionnaire) and inaccuracy (mean number of problematic records per interview questionnaire). The secondary outcomes are cost-effectiveness parameters and cost-related outcomes. Moreover, we will consider usability scale and qualitative user satisfaction indices for the secondary outcomes analysis. Multiple definitions of usability, acceptability, user satisfaction, and related terminologies exist [11]. We will extract, summarize, and categorize the definitions in the final filtered full-text papers to include efficiency, acceptability, usability, and user experience outcomes.

The quality of the included studies will be assessed using parameters such as random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases. We will grade each parameter as a low, high, or unclear risk of bias [28]. All assessments of study quality will be performed by at least two reviewers (AAZ and TN), with any disagreement resolved by consensus or mediation with a third reviewer (FF or RR) where necessary.

We will present the analyzed data in a tabular and narrative form. Where possible, meta-analyses will be performed on methodologically comparable studies (comparable particularly with regards to the study design and endpoint measures in the outcomes) reporting primary and secondary outcomes. The choice of statistical tests will depend on the nature of the outcome variable. Where relevant data are missing, we will contact the authors. If we cannot obtain missing data by contacting the authors, we will use an imputation method. If the number of included studies per outcome is sufficient, publication bias will be assessed visually through funnel plots and tested by Egger’s regression test. The Mantel-Haenszel method will be used for the fixed effect model if tests of heterogeneity are not significant. If our data displays statistical heterogeneity, the random effects model will be selected. In cases of significant heterogeneity, we will perform a qualitative narrative summary instead of a meta-analysis.

As only previously published studies are included and reported in the review, no additional formal ethical assessment and no informed consent is required. The findings will be disseminated through publication of a single manuscript in a peer-reviewed journal.