Attendance trends in a higher education institution 2014-2025

<i>1. Sample and data collection</i>To examine trends in class attendance before, during, and after the COVID-19 pandemic, we used anonymized administrative data from three faculties of one large university of applied sciences in the Netherlands. Ethical approval for this study was granted by the internal review board of the first author with approval code HVA-1212. Attendance data were obtained through the Academy Attendance application, developed by the educational technology company Your Next Concepts. This digital system allows students to check in to scheduled classes via an institutional app and enables programs to track weekly attendance rates.The dataset contains 8.9 million attendance records from 27,568 unique students attending in 4,896 different courses across multiple programs in three major domains: health (nursing, physiotherapy, nutrition, practice therapy) technology (aviation, built environment, biomedical technology), and business (marketing). The courses in this dataset offer four year bachelor degree programs. Attendance was tracked in all four years of the curriculum level, although year 3 and 4 are less well-represented because a large part of the year 3 and 4 curriculum involves external minor programs and internships which were not included in this dataset. Nearly all classes in the included courses are offered to groups of in between 18 and 40 students, with only a small minority of large traditional lectures with around 100-200 students. Table 1 displays the distribution of the sample across courses, years and periods of time. Each record represents an anonymized observation of a student’s presence in a scheduled class session. For every course, the system registers which students were expected and how many attended. Only programs with consistent attendance tracking across multiple years and at least two era’s (pre, during and post-covid) were included.<br>The data span eleven academic years, from August 2014 to August 2025. This time frame allows for comparison of attendance patterns across three distinct phases of higher education delivery: traditional on-campus instruction (pre-COVID), remote or hybrid teaching (during COVID), and the post-COVID return to in-person learning.<br><i>2. Measurements</i>Each row in the dataset represents an anonymous student-lesson combination. Separate columns identify (1) an anonymized program identifier, (2) the educational program (3) unique course id, (4) event date, (5) attendance (1 or 0), (6) curriculum year (1-4) (7) academic quarter 1-4.For the inferential analyses, the 8.9 million events were aggregated into 36,395 weekly attendance observations, measured in amounts of successes (attending students) and failures (students who were expected but did not attend) within 4,720 courses over the time span of 11 years. We created covariates for the pre-covid, the covid and post-covid era based on the local higher education regulations. Based on this the Covic lockdowns started at 2020-03-15 and ended 2022-02-24. This led to 8,909 observations in the pre-covid time span, 10,141 during Covid and 17,345 after covid.<br><b>Table 1.</b> Distribution of sample across courses, years, and time period in aggregated dataset<b>Domain</b><b>Courses n</b><b>Weekly records n</b><b>Pre-COVID records n</b><b>COVID records n</b><b>Post-COVID records n</b>Technical2,25919,7645,1555,1849,425Business1,2839,4652,2902,4344,741Health1,1787,1661,4642,5233,179Total4,72036,3958,90910,14117,345<br><br>Table 2. Distribution of sample across curriculum years and periods.<b>Year level</b><b>Pre-COVID records n</b><b>COVID records n</b><b>Post-COVID records n</b>Year 11,539,9741,047,8321,709,933Year 21,159,324927,8621,096,110Year 343,396244,32270,389Year 4170,281167,294300,557