Evaluation data of the PartoMa project Tanzania

The dataset was generated through a prospective cross-sectional survey conducted among healthcare providers attending PartoMa low-dose, high-frequency (LDHF) in-house training seminars in five high-volume urban maternity facilities in Dar es Salaam, Tanzania. Data collection occurred at a single time point during the third round of seminars, ensuring full implementation of the intervention across all sites. The geographical scope is therefore limited to urban public hospitals in Dar es Salaam, while the temporal scope reflects post-intervention assessment within the broader 2019–2024 implementation period. Data were collected using a structured, pre-tested bilingual (English and Swahili) questionnaire administered immediately after seminar participation. The instrument captured sociodemographic characteristics, perceptions of the training and guideline booklet, and logistical information such as travel time, distance, and costs. The final dataset comprises 221 observations (rows), each representing an individual participant, and 39 variables (columns), including categorical, ordinal, and continuous data types.Data processing involved systematic cleaning, coding, and transformation procedures. Raw questionnaire data were entered into Microsoft Excel, where inconsistencies and incomplete entries were reviewed and standardized. Likert-scale responses were rescaled from 1–5 to 0–4 to establish a true zero baseline, and negatively worded items were reverse-coded to ensure interpretive consistency across variables. Composite domain scores for acceptability, appropriateness, and feasibility were calculated by averaging relevant questionnaire items. Statistical analysis was performed using Stata version 17, generating descriptive statistics such as means, standard deviations, and frequency distributions. Measurement units include ordinal Likert scores (0–4), time (hours), distance (kilometers), and cost (Tanzanian Shillings). The dataset is stored in a standard .xlsx format, accessible באמצעות common software such as Microsoft Excel, LibreOffice Calc, and Google Sheets.The dataset demonstrates high completeness, with missing data ranging from 0.5% to 7.7% per variable, primarily due to non-response or challenges in interpreting negatively worded Likert items. Missing values were assumed to be missing at random and handled using complete-case analysis without imputation. Potential sources of error include response bias, social desirability effects, and language comprehension difficulties, particularly for English-language items, as well as occasional misinterpretation of reverse-coded questions. Variability in responses is reflected in moderate standard deviations (approximately 0.7–1.5). No major data corruption or extreme outliers were identified. The dataset file is relatively small (<1 MB), contains a single worksheet, and does not require specialized software for access. Overall, the data provide a structured and reliable representation of healthcare provider perceptions and implementation outcomes related to LDHF training and decision-support tools in a resource-limited urban healthcare context.