Paper-Based Data Collection: Research Summary

October 14, 2026

Purpose and scope

This a summary of an investigation into using optical mark recognition (OMR) to digitize paper data collected at public services. It lists the design options considered, the tools compared, and the published research found.

The working scenario:

• Staff at a public service fill in a paper form for each person served.
• Most fields are tick-box answers that OMR can read.
• Two fields are hard to capture with bubbles: a person ID and a timestamp.

Design options for ID and timestamp

The investigation pursued three approaches:

• machine-readable marks,
• codes printed ahead of time, and
• handwritten fields keyed in by a person.

Each of person ID and timestamp were considered separately for each of the 3 approaches.

Person ID

Error checks considered:

• Parity or count check bubbles. Two bubbles holding the number of filled ID bubbles mod 4 detect any single missed or extra mark.
• Check digit on a written ID, using the Damm or Verhoeff algorithm. These detect every single-digit error and every adjacent swap.
• Roster check against the list of issued IDs.
• Card-status field recording whether a card was presented, lost, or newly issued.

Timestamp

On a log sheet with one row per person, rows are in time order, so each time should be no earlier than the row above. That ordering can be used to flag misreads.

Hybrid log sheet with human keying

The last option considered combines OMR and manual entry:

1. OMR reads all bubble fields.
2. The desktop app uses the template coordinates to crop the handwritten time and ID cells from each row.
3. A person keys the values from the crops, one field type at a time, using the keyboard and number pad.
4. Check digits, roster checks and time ordering validate entries as they are typed.

The estimate given was 3–5 seconds of keying per row, or about 1.5–2 minutes for a 25-row sheet. That figure is not measured.

Tools: OMRChecker and ODK-X Scan

The two tools overlap only in reading bubbles. OMRChecker is a bubble-reading engine. ODK-X Scan is a phone app that also handles manual entry of handwritten fields.

The ODK-X Scan README says the project “has not been updated in a couple of years” and asks for someone to take it over. Its latest release, 2.1.4, is marked alpha, and the documentation describes it as beta. The mainstream ODK project (ODK Collect and ODK Central) is separate and maintained, but has no paper-form scanner.

Research findings

The research found six different approaches to the problem. Most of it comes from low- and middle-income countries and from clinical data collection.

Paper-digital hybrid systems

• ODK Scan reads bubbles and checkboxes on paper forms and shows handwritten fields for manual entry. Field tests were run in Mozambique (health workers’ supply and vaccine forms), Malawi (a register kept by frontline health workers) and Pakistan’s tuberculosis programme. In Pakistan, aggregating and transferring data took less time, while verifying data and filling in forms took more. Only the Pakistan paper’s abstract was read.
• mScan was the phone-camera bubble reader that ODK Scan grew out of.
• Shreddr cut form images into single-field fragments. Software estimated each value, then people verified it. It was piloted in Mali and later became Captricity.
• CAM used camera phones with printed forms that combined barcodes and handwritten fields, for microfinance groups in rural India.
• Dell et al., CSCW studies paper-digital workflows in global development organizations. Its full text was not available.
• Digital pen and paper was evaluated in clinical settings and reported not to add time to data capture.

Accuracy of OMR, handwriting recognition and double entry

A 2020 cohort study at Stavanger University Hospital compared methods on 448 patient questionnaires:

OMR and double entry did not differ significantly (P=.61). Handwriting recognition was significantly worse than double entry (P=.033).

Statistical linkage keys

Australia’s SLK-581 is built from selected letters of the surname and given name, the date of birth and sex.

• In a national nursing-home database, 0.6% of 440,000 people shared a key with someone else.
• Linking on the key alone undercounted matches, for example hospital stays before death. Looser matching rules or probabilistic linkage gave results closer to full linkage.
• One paper asks whether it is time to move on from SLK-581. Another reports that a Bloom filter method gave better privacy and linkage quality.

Paper vs electronic data collection

• A 2021 systematic review in JMIR found electronic collection generally faster and of good quality. It also found the evidence “small in quantity and low in quality”, with only 2 studies linking cost to data quality. Reported problems included data loss, device theft, power surges and poor connectivity.
• A 2019 randomized crossover trial in Ethiopia paired a tablet interviewer with a paper interviewer in each household. Building the database took about 16 hours for paper and half an hour for tablets.
• Another study compared quality and cost in Rufiji, Tanzania, and evaluated an open-source electronic capture system in Africa.

Mobile tools for frontline health workers

• Braun et al. (PLoS ONE 2013) reviewed 25 studies. Collecting data in the field was one of the most common uses of mobile tools by community health workers.
• A separate systematic review assessed whether frontline health workers in developing countries can use mHealth tools feasibly and effectively.
• An evaluation in South Africa compared paper-based monitoring with an mHealth system for the national community health worker programme. Its full text was not available.
• Hartung et al. (ICTD 2010) introduced Open Data Kit for building information services in low-resource settings.

Routine health information systems

• The PRISM framework holds that poor data quality and use come from organizational and behavioural causes as well as technical ones. Its tools compare monthly reports with facility registers and rate how complex and user-friendly registers and forms are.
• DHIS2 grew out of the Health Information Systems Programme (HISP) in South Africa and is coordinated by the University of Oslo. Related work includes the “Networks of Action” paper (MISQ 2004) and a 2022 scoping review of how DHIS2 data is used.

Gaps in the literature

• Setting. Almost all studies found are from low- and middle-income countries or hospitals. Little covers public services in high-income settings such as shelters, food banks and outreach.
• Cost and quality together. The 2021 JMIR review found only 2 studies that linked cost to data quality.
• Evidence quality. The same review described the evidence comparing paper and electronic collection as small in quantity and low in quality.
• Search coverage. This summary comes from a small number of web searches, not a systematic literature search. Several sources were read only as abstracts or search summaries, as noted in each section.

References

Paper-digital hybrid systems

• Dell, Breit, Chaluco, Crawford & Borriello. Digitizing Paper Forms with Mobile Imaging Technologies. ACM DEV 2012 (2nd ACM Symposium on Computing for Development) — mScan.
• Dell, Crawford, Breit, Chaluco, Coelho, McCord et al. Integrating ODK Scan into the Community Health Worker Supply Chain in Mozambique. ICTD 2013. PDF.
• VillageReach. ODK Scan Field Test: Malawi Community Health Worker Register. December 2015.
• VillageReach. ODK Scan Product Recommendations Report. April 2015.
• Ali, Powers, Beorse, Noor, Naureen, Anjum et al. ODK Scan: Digitizing Data Collection and Impacting Data Management Processes in Pakistan’s Tuberculosis Control Program. Future Internet, 2016.
• Chen, Kannan, Yano, Hellerstein & Parikh. Shreddr: Pipelined Paper Digitization for Low-Resource Organizations. ACM DEV 2012.
• Parikh et al. CAM. UW News: Custom cell phones for microfinance co-ops. February 2007.
• Dell, Perrier, Kumar, Lee, Powers, Borriello et al. Paper-Digital Workflows in Global Development Organizations. CSCW 2015.
• Despont-Gros, Landau, Rutschmann, Simon & Lovis. The Digital Pen and Paper: Evaluation and Acceptance of a New Data Acquisition Device in Clinical Settings. Methods of Information in Medicine, 2005.

Accuracy of automated form processing

• Paulsen, Harboe & Dalen. Data Entry Quality of Double Data Entry vs Automated Form Processing Technologies: A Cohort Study Validation of Optical Mark Recognition and Intelligent Character Recognition in a Clinical Setting. Health Science Reports, 2020.

Statistical linkage keys

• Taylor, Irvine, Iannotti, Harchak & Lim. Optimal Strategy for Linkage of Datasets Containing a Statistical Linkage Key and Datasets with Full Personal Identifiers. BMC Medical Informatics and Decision Making, 2014.
• Coulson, Bailey, Reid, Shardey, Williams-Spence, Huckson et al. Linkage of Australian National Registry Data Using a Statistical Linkage Key. BMC Medical Informatics and Decision Making, 2021.
• Randall, Ferrante, Boyd, Brown & Semmens. Limited Privacy Protection and Poor Sensitivity: Is It Time to Move On from the Statistical Linkage Key-581? Health Information Management Journal, 2016.
• Australian Institute of Health and Welfare. SLK-581 Guide for Use.

Paper vs electronic data collection

• Zeleke, Naziyok, Fritz, Christianson & Röhrig. Data Quality and Cost-effectiveness Analyses of Electronic and Paper-Based Interviewer-Administered Public Health Surveys: Systematic Review. Journal of Medical Internet Research, 2021.
• Zeleke, Worku, Demissie, Otto-Sobotka, Wilken, Lipprandt et al. Evaluation of Electronic and Paper-Pen Data Capturing Tools for Data Quality in a Public Health Survey in a Health and Demographic Surveillance Site, Ethiopia. JMIR mHealth and uHealth, 2019.
• Mukasa, Mushi, Maire, Ross & de Savigny. Do Surveys with Paper and Electronic Devices Differ in Quality and Cost? Experience from the Rufiji Health and Demographic Surveillance System in Tanzania. Global Health Action, 2017.
• Dillon, Pirie, Rice, Pomilla, Sandhu, Motala et al. Open-Source Electronic Data Capture System Offered Increased Accuracy and Cost-Effectiveness Compared with Paper Methods in Africa. Journal of Clinical Epidemiology, 2014.

Mobile tools for frontline health workers

• Braun, Catalani, Wimbush & Israelski. Community Health Workers and Mobile Technology: A Systematic Review of the Literature. PLoS ONE, 2013. CHW Central summary.
• Agarwal, Perry, Long & Labrique. Evidence on Feasibility and Effective Use of mHealth Strategies by Frontline Health Workers in Developing Countries: Systematic Review. Tropical Medicine & International Health, 2015.
• Neupane, Odendaal, Friedman, Jassat, Schneider & Doherty. Comparing a Paper Based Monitoring and Evaluation System to a mHealth System to Support the National Community Health Worker Programme, South Africa: An Evaluation. BMC Medical Informatics and Decision Making, 2014.
• Hartung, Lerer, Anokwa, Tseng, Brunette & Borriello. Open Data Kit: Tools to Build Information Services for Developing Regions. ICTD 2010.

Routine health information systems

• Aqil, Lippeveld & Hozumi. PRISM Framework: A Paradigm Shift for Designing, Strengthening and Evaluating Routine Health Information Systems. Health Policy and Planning, 2009.
• MEASURE Evaluation. PRISM: Performance of Routine Information System Management series.
• Braa, Monteiro & Sahay. Networks of Action: Sustainable Health Information Systems Across Developing Countries. MIS Quarterly, 2004.
• Routine Use of DHIS2 Data: A Scoping Review. BMC Health Services Research, 2022.
• DHIS2 research.

Software

• OMRChecker on GitHub.
• ODK-X Scan on GitHub, with releases and documentation.
• ODK-X barcode scanner issue.