Predictive power of socio-demographic attributes of healthcare workers on data quality in Tanzanian regional referral hospitals Authors: Rogate Phinias, Mikidadi Muhanga, Joshua Malago Published: 2025 ABSTRACT BackgroundHealthcare data quality is essential towards effective decision-making. Data of high quality are accurate, complete, consistent, and timely. Any mistake in patient information is likely to result in serious life-threatening issues. Monitoring and evaluation systems have been introduced to address these data quality issues. These systems are affected by human inputs which can subsequently introduce errors that lower the quality of data. This study examined the influence of socio-demographic attributes of healthcare workers on data quality in selected regional referral hospitals in Tanzania.MethodologyA cross-sectional study design was conducted from August to October 2024 across eight regional referral hospitals in Tanzania. Using simple random sampling, 336 healthcare workers involved in regular patient data entry were recruited. The sample size was determined using Cochran’s formula. Data were collected through structured questionnaires, cleaned, and analyzed using IBM SPSS v27, applying both descriptive statistics and multiple linear regression to examine the relationship between socio-demographic factors and healthcare data quality.ResultsThe study identified several significant predictors of data quality. Age had a positive relationship with accuracy (β = 0.241, p <.001), completeness (β = 0.159, p <.001) and timeliness (β = 0.202, p <.001), while sex was positively associated with all outcomes, particularly completeness (β = 0.428, p <.001), and consistency (β = 0.561, p <.001). Education positively impacted accuracy (β = 0.161, p <.01), and timeliness (β = 0.240, p <.01), but negatively affected completeness (β=-0.166, p <.01), and consistency (β=-0.144, p <.01). Marital status significantly predicted completeness (β = 0.139, p =.044), while experience showed a negative relationship with timeliness (β=-0.369, p <.001). All models were statistically significant (all p <.001).ConclusionThis study highlights the significant influence of socio-demographic factors, particularly age and sex, on the accuracy and completeness of healthcare data. These findings suggest that the characteristics of patient data collectors can either enhance or hinder data quality. Therefore, it is crucial to implement targeted training programs for healthcare workers, emphasizing their role in maintaining high-quality healthcare data.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12913-025-13219-1. Introduction In the rapidly evolving healthcare sector, quality data is crucial for improving patient health outcomes, enhancing decision-making, fostering innovation, and delivering effective health services [1, 2]. Healthcare data are used for both primary purposes (direct patient care and clinical decision-making) and secondary purposes (such as research, quality improvement, public health surveillance, and policy-making). Globally, healthcare data quality is assessed across four dimensions: accuracy, completeness, consistency, and timeliness [3]. While all four dimensions support both primary and secondary uses, timeliness is especially critical for real-time clinical decisions, whereas completeness and consistency are foundational for reliable secondary analyses. Accuracy ensures that recorded data correctly reflects a patient’s health status, some minor errors can threaten patient safety. Completeness involves capturing all relevant patient information, some missing data can hinder understanding of a patient’s medical history. Consistency guarantees that data is uniformly provided, preventing issues like delayed treatments and misdiagnoses. Timeliness is essential for real-time decision-making, ensuring that data reflects the most current patient information. In absence of these dimensions, healthcare data can lack significance and quality [4, 5]. Globally, Healthcare data quality remains dynamic, and countries are actively working to enhance its consistency and reliability, these efforts are pioneered by the World Health Organization (WHO), notably the WHO has developed and implemented the healthcare digital strategy to ensure all countries in the world collect, store, process and disseminate quality healthcare data [6]. In low- and middle-income countries (LMICs), a growing body of literature highlights significant challenges related to the healthcare data quality. Such studies indicate that healthcare data in these regions often suffers from various issues, including inaccuracies, incompleteness, and inconsistencies, which undermine the effectiveness of health interventions and policy-making [4, 7]. Similarly, many African nations face similar challenges regarding healthcare data quality [4, 5]. These studies indicate that the completeness, accuracy, and timeliness of health data in these countries vary significantly, with reported rates ranging from 34 to 72%. Such variability suggests that while some regions have made steps to improve their data systems, others continue to struggle with some issues that hinder their ability to collect and use healthcare data effectively. Healthcare Data Quality (HDQ) must be collected and stored at every level of the healthcare system (i.e. Primary, secondary and tertiary) to support informed decision-making [8]. In Tanzania, monitoring and evaluation (M&E) systems have been introduced in all regional referral hospitals (RRH) to ensure the production of quality data [9]. These systems, including Afyacare, Government of Tanzania Health Operations Management Information System (GoTHOMIS), District Health Information System 2 (DHIS2), and Afya Electronic Health Management System (Afya-EHMS) among others, have been designed to collect, store, process, and share patient information. The government has also developed policies, strategies and guidelines to ensure that HDQ is generated through these systems. These include Monitoring and Evaluation Strategic Framework (MESF) 2020–2025, Health Sector Strategic Plan July 2021– June 2026 (HSSP V)-Leaving No One Behind, and Health Information Management System Guidelines (HISG) [10, 12]. Furthermore, the health sector is making good progress in digitalizing its services, it has introduced the Digital Health Strategy July 2019-June 2024 which needs all levels of healthcare systems use and provide digital health services along with digital M&E systems [13, 14]. However, despite these efforts, producing quality healthcare data remains a challenge due to notable inaccuracies, incompleteness, delays, and inconsistencies. Human errors are often cited as a major factor [15, 18]. This is because, healthcare workers may unintendedly enter wrong data or overlook a critical data point in the documentation process. In addressing data quality issues, it is good to investigate the role of human differences in the healthcare context by considering their socio-demographic attributes. Socio-demographic factors such as age, sex, job roles, experience, marital status, and education have been reported to influence health behaviors, performance of healthcare workers, access to healthcare services, and the accuracy of data reporting [19]. Literature indicates that socio-demographic characteristics contribute to performance of healthcare workers hence resulting to inaccuracies, delays, and inconsistencies in healthcare data. For instance, Chebureti and Odhiambo [19] argue that marital status influences routine data quality, noting that married civil servants often prioritize social matters over official tasks, while single staff may work longer hours due to fewer social obligations. Similarly, Belita et al. [20] report that marital status influence absenteeism in various ways due to family responsibilities, these marital challenges leads to delays in data collection/entry, rushing to family matters hence impact in the completeness and accuracy of data. In contrast, Yendemye and Nimubona [21] argue that stable home lives and strong emotional support motivate married staff to value their jobs more, translating into greater attention to data completeness and accuracy. Tuna et al. [22] found that the variability in the type of department affects job satisfaction, while age was reported to associate with increase job stress, ultimately reducing healthcare worker performance. They claim that younger workers are generally more eager and productive; however, this contrasts with Kujala [23], who posited that younger individuals may perform poorly as well. Additionally, variations in education lead to differences in how data is collected and reported [24, 25], further complicating the issue of data quality. Despite the evident need for studies on socio-demographic and healthcare data quality, there is scant empirical evidence from the literature indicating that this aspect has been studied in Tanzania. This study, therefore, addresses this knowledge gap by assessing how the socio-demographic characteristics of healthcare workers predict the production of quality healthcare data in regional referral hospitals in Tanzania. It is important to understand how these factors interplay with data quality to develop strategies to improve healthcare data collection. By identifying the specific challenges faced by different groups within the healthcare workforce, targeted training programs can be designed to address gaps in knowledge and skills. Methods Study setting, period and design The cross-sectional research design was employed in this study from August to October 2024 in eight selected RRH in Tanzania. The country has 28 RRHs, in this study, multi-stage sampling technique was employed to get eight RRH. In the initial stage, all the 28 RRH were identified. In the second stage, the eight administrative zones were recognized, followed by random selection of one RRH from each zone, as shown in Fig. 1., the selected RRH were Mount Meru (Northern zone), Mwalimu Nyerere Memorial (Lake zone), Maweni (Western zone), Singida (Central zone), Morogoro (Eastern zone), Mbeya (South-West highland zone), Sokoine (Southern zone) and Iringa RRH (Southern Highland zone). This was done to ensure randomization and generalizability of the study. The study population comprised all healthcare workers involved in regular patient data entry activities from patients’ arrival at the hospital to their departure. Fig. 1Study area Study area Sample size and sampling technique Cochran’s formula [26] was used to calculate the sample size, based on the following assumptions; z-value of 1.96 from the standard normal distribution with a confidence level of 95%, marginal error of 5%, population size of 2650, and estimated population proportion of 0.05. This resulted into a sample size of 336. The sample size calculations are detailed below: -\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{n}_{0}=\frac{{N\cdot\:Z}^{2}\cdot\:p\cdot\:(1-p)}{{E}^{2}\left(N-1\right)+{Z}^{2}\cdot\:p(1-p)}$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{n}_{0}=\frac{{2650\cdot\:1.96}^{2}\cdot\:0.5\cdot\:(1-0.5)}{{0.05}^{2}\left(2650-1\right)+{1.96}^{2}\cdot\:0.5(1-0.5)}$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{n}_{0}=\frac{2650\cdot\:0.9604}{6.6225+0.9604)}$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{n}_{0}=\frac{2544.66}{7.5829}$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{n}_{0}=335.54$$\end{document} The 336 individuals were identified using simple random sampling technique and proportionately allocated to all the 8 RRH with 39 to 45 respondents in each RRH. To manage attrition, we over-sampled by 10%, inviting 370 healthcare workers via simple random sampling (including alternates) across the eight RRHs. Two reminder visits were made, yielding 324 usable questionnaires (87.6% of those approached). The 12 non-responders were then replaced by their pre-selected alternates to reach a final sample of 336 (90.8% response rate). Study personnel The research team comprised two core members who visited all eight RRHs to coordinate and oversee data collection activities. In accordance with ethical clearance guidelines issued by the Ministry of Health, the team was required to collaborate with staff from each study site during data collection. At every facility, two healthcare workers who were willing to assist were identified and temporarily integrated into the research team. These local collaborators supported questionnaire administration and helped ensure adherence to facility protocols, while the core team-maintained consistency and quality control across all sites. Data collection and measurement The questionnaire used in this study was specifically developed for this research. Data were collected through a self-administered questionnaire designed to assess how socio-demographic attributes such as age, sex, marital status, education, and years of experience, predict the quality of healthcare data. The quality of data was evaluated based on four key factors: accuracy, completeness, consistency, and timeliness of data entry. Table 1 presents the measurement of the dependent variable, while Table 2 outlines the measurement of the independent variables. For additional data on the measurement instruments used in the study, see Supplementary Material. Healthcare data quality level was measured based on percentiles, dividing the data into three distinct quality levels: Low, Moderate, and High. The classification was based on the three percentile ranges as shown in Table 3: Data quality was considered low when it fell below the 33.33rd percentile, it was categorized as moderate when ranged between the 33.33rd and 66.66th percentiles, and a high when it exceeded the 67.00th percentile. Similarly, studies by [27] and [28] employed this approach to categorize levels. Table 1Healthcare data quality and its measurementsType of VariableDimensionIndicatorMeasurementDependent(Healthcare Data Quality)AccuracyConfidence in patient demographic accuracy, encounter diagnostic test discrepancies, medication record accuracy, timely updating of medical histories, data verification processes, peer review and double-checking for accuracy, adherence to clinical note protocols5-point Likert-Scale (1 = strongly disagree, 5 = strongly agree).CompletenessConsistency in recording patient medical histories, completeness of demographics, adequacy of clinical notes, frequency of incomplete data, review and verification processes, resolution of discrepancies in data entry.5-point Likert-Scale (1 = strongly disagree, 5 = strongly agree).TimelinessTimeliness of updating patient demographics, data entry after patient visits, timely review of entries for accuracy and completeness, timely documentation and upload, administrative support for timely entry, impact of workload on data entry.5-point Likert-Scale (1 = strongly disagree, 5 = strongly agree).ConsistencyMaintaining data consistency across interactions and documenting patient medical histories uniformly, entering patient demographics, uniformity of clinical notes, verifying patient identity, satisfaction with current consistency practices.5-point Likert-Scale (1 = strongly disagree, 5 = strongly agree). Table 1Healthcare data quality and its measurementsType of VariableDimensionIndicatorMeasurementDependent(Healthcare Data Quality)AccuracyConfidence in patient demographic accuracy, encounter diagnostic test discrepancies, medication record accuracy, timely updating of medical histories, data verification processes, peer review and double-checking for accuracy, adherence to clinical note protocols5-point Likert-Scale (1 = strongly disagree, 5 = strongly agree).CompletenessConsistency in recording patient medical histories, completeness of demographics, adequacy of clinical notes, frequency of incomplete data, review and verification processes, resolution of discrepancies in data entry.5-point Likert-Scale (1 = strongly disagree, 5 = strongly agree).TimelinessTimeliness of updating patient demographics, data entry after patient visits, timely review of entries for accuracy and completeness, timely documentation and upload, administrative support for timely entry, impact of workload on data entry.5-point Likert-Scale (1 = strongly disagree, 5 = strongly agree).ConsistencyMaintaining data consistency across interactions and documenting patient medical histories uniformly, entering patient demographics, uniformity of clinical notes, verifying patient identity, satisfaction with current consistency practices.5-point Likert-Scale (1 = strongly disagree, 5 = strongly agree). Healthcare data quality and its measurements Dependent (Healthcare Data Quality) Table 2Independent variables and its measurementIndependent VariableTypeMeasurementAgeContinuousNumericalSexDummyTransformation made (1 = Male, 0 = Female; 0 = Male, 1 = Female)Marital statusDummyTransformation made (1 = Married, 0 = Not Married; 0 = Married, 1 = Not Married)Education levelContinuousNumber of years spent schoolingYears of work experienceContinuousNumerical Table 3Summary of healthcare data quality classification criteriaPercentile rangeHealthcare Data Quality LevelLess than 33.33LowBetween 33.33 and 66.66Moderate67.00 and aboveHigh Table 2Independent variables and its measurementIndependent VariableTypeMeasurementAgeContinuousNumericalSexDummyTransformation made (1 = Male, 0 = Female; 0 = Male, 1 = Female)Marital statusDummyTransformation made (1 = Married, 0 = Not Married; 0 = Married, 1 = Not Married)Education levelContinuousNumber of years spent schoolingYears of work experienceContinuousNumerical Independent variables and its measurement Table 3Summary of healthcare data quality classification criteriaPercentile rangeHealthcare Data Quality LevelLess than 33.33LowBetween 33.33 and 66.66Moderate67.00 and aboveHigh Summary of healthcare data quality classification criteria Data processing and analysis Data entry and cleaning were done in Microsoft excel and analyzed using IBM SPSS version 27. Descriptive analysis used frequency, percentage to describe the socio-demographic factors and the dependent variable. Subsequently, Likert scale responses were converted into mean values for inclusion in multiple regression analysis. Age, education, and experience were treated as continuous variables, whereas Sex, and marital status were dummy variables which were transformed using dummy coding. Thereafter, diagnostic tests of dataset and multiple linear regression analysis were done to establish the relationship between the socio-demographic factors and production of high-quality healthcare data. Results Respondent profile Table 4, indicates that 57% of the study participants were males, 56.6% were between 20 and 29 years of age. Most participants (33.5%) had 9 to 12 years of experience. Nearly half (48.3%) of the participants were unmarried, 4.9% married but separated and 2.9% divorced. A large number of participants held a bachelor’s degree (53%), and 34% had a diploma. Table 4Socio-Demographic characteristics of healthcare workers, n = 336CharacteristicsCategoryFrequencyPercentageSexMale19156.6%Female14543.4%Age20–2917953.230–3911935.340–49278.150–59113.4Job experience (in years)Rookie (0–4)6619.7%Novice (5–8)8826.2%Experienced (9–12)11333.5%Veteran (13–16)329.6%Expert/Pro (17+)3710.9%Marital statusSingle16248.3%Married10731.9%Cohabited3410.1%Separated164.9%Divorced102.9%Widow/Widower72.1%Education levelCertificate236.8%Diploma11434.0%Bachelor degree17853.0%Master degree216.2% Table 4Socio-Demographic characteristics of healthcare workers, n = 336CharacteristicsCategoryFrequencyPercentageSexMale19156.6%Female14543.4%Age20–2917953.230–3911935.340–49278.150–59113.4Job experience (in years)Rookie (0–4)6619.7%Novice (5–8)8826.2%Experienced (9–12)11333.5%Veteran (13–16)329.6%Expert/Pro (17+)3710.9%Marital statusSingle16248.3%Married10731.9%Cohabited3410.1%Separated164.9%Divorced102.9%Widow/Widower72.1%Education levelCertificate236.8%Diploma11434.0%Bachelor degree17853.0%Master degree216.2% Socio-Demographic characteristics of healthcare workers, n = 336 Table 5 indicates that significant proportion of professions, comprised of physicians and nurses who were 28.1% and 23.9% respectively, and the outpatient department (OPD) had many representatives (38.7%). Furthermore, the study found 84.7% of respondents used AfyaCare system while 94% were using it daily for data entry. Table 5Other characteristics of respondents, n = 336CharacteristicsCategoryFrequencyPercentageProfessionPhysicians9428.1%Nurses8023.9%Laboratorians144.2%Administrators154.4%Pharmacists5315.8%Finance staff113.4%Health information management staff288.3%Supportive staff (e.g. social welfare workers, attendants etc.)4011.9%Unit/Section/DepartmentOut Patient Department (OPD)13038.7%In-Patient Department (IPD)3510.4%Radiology and Surgical Service Department113.4%Pharmacy Department185.5%Department of Administration and Human Resource Management5115.3%Emergency Service Department5717.1%Labaratory103.1%Pediatric Department92.6%Obstetrics & Gynecology133.9%M&E system (s) currently in useDHIS2, GoTHOMIS, HRHIS, CTC database5115.3AfyaCare28584.7Frequency in using M&E systemDaily31694.3Weekly and monthly1910.7 Table 5Other characteristics of respondents, n = 336CharacteristicsCategoryFrequencyPercentageProfessionPhysicians9428.1%Nurses8023.9%Laboratorians144.2%Administrators154.4%Pharmacists5315.8%Finance staff113.4%Health information management staff288.3%Supportive staff (e.g. social welfare workers, attendants etc.)4011.9%Unit/Section/DepartmentOut Patient Department (OPD)13038.7%In-Patient Department (IPD)3510.4%Radiology and Surgical Service Department113.4%Pharmacy Department185.5%Department of Administration and Human Resource Management5115.3%Emergency Service Department5717.1%Labaratory103.1%Pediatric Department92.6%Obstetrics & Gynecology133.9%M&E system (s) currently in useDHIS2, GoTHOMIS, HRHIS, CTC database5115.3AfyaCare28584.7Frequency in using M&E systemDaily31694.3Weekly and monthly1910.7 Other characteristics of respondents, n = 336 Diagnostic and reliability test of the dataset Table 6 indicates the diagnostic and reliability test of the dataset. Five diagnostic tests were conducted to evaluate the assumptions of multiple linear regression analysis. The Kolmogorov-Smirnov test yielded a P-value of 0.394, indicating that the dataset follows a normal distribution, thus confirming the normality assumption. ANOVA showed insignificant deviations from linearity, leading to the acceptance of the null hypothesis that no deviations exist. Also, Tolerance and Variance Inflation Factor (VIF) metrics confirmed that multicollinearity was not a concern, as Tolerance was above 0.1 and VIF below 10. The Breusch–Pagan test produced a significance value of 0.215, above the acceptable level, confirming no heteroscedasticity in the dataset. The Durbin-Watson results were 1.789 which confirmed the assumption of no auto-correlation in the dataset. In addition, Table 6 illustrates the Cronbach’s alpha results exceeding the common threshold of 0.7. Table 6Diagnostic and reliability tests of the dataset (n = 336)TestVariableResultsNormalityKolmogorov-Smirnov (Sig.)0.394LinearityDeviation from linearity (Sig.)Quality data*age0.617Quality data*Sex0.636Quality data*marital stats0.507Quality data*education0.613Quality data*experience0.484MulticollinearityToleranceVIFAge0.2474.043Sex0.9241.082Marital status0.9491.054Education0.9661.035Years of experience0.2464.062HomoscedasticityBreusch–Pagan (Sig.)0.213AutocorrelationDurbin-Waston1.789ReliabilityCronbach AlphaQuality data0.89 Table 6Diagnostic and reliability tests of the dataset (n = 336)TestVariableResultsNormalityKolmogorov-Smirnov (Sig.)0.394LinearityDeviation from linearity (Sig.)Quality data*age0.617Quality data*Sex0.636Quality data*marital stats0.507Quality data*education0.613Quality data*experience0.484MulticollinearityToleranceVIFAge0.2474.043Sex0.9241.082Marital status0.9491.054Education0.9661.035Years of experience0.2464.062HomoscedasticityBreusch–Pagan (Sig.)0.213AutocorrelationDurbin-Waston1.789ReliabilityCronbach AlphaQuality data0.89 Diagnostic and reliability tests of the dataset (n = 336) Evaluation of healthcare data quality in selected hospitals Table 7 presents an overview of healthcare data quality levels across specific hospitals. The HDQ scale was measured using percentiles (< 33.33rd, between 33.33 and 66.66th and above 67.00th percentile), leading to the classification of data quality into three categories: Low, Moderate, and High HDQ. It was found that, H2 and H8 RRHs were identified as having high data quality levels, while the remaining six regional referral hospitals were rated as having moderate data quality. Notably, there were no hospitals classified with low perceived healthcare data quality. Overall, the assessment indicated that the healthcare data quality in regional referral hospitals is generally perceived as moderate. Table 7Quality of healthcare data in study areas (n = 336)Healthcare data quality levelsTotalLow HDQModerate HDQHigh HDQRRHH12281444H21182140H31311143H41291141H5130940H62321044H73321045H80192039Total11219106336 Table 7Quality of healthcare data in study areas (n = 336)Healthcare data quality levelsTotalLow HDQModerate HDQHigh HDQRRHH12281444H21182140H31311143H41291141H5130940H62321044H73321045H80192039Total11219106336 Quality of healthcare data in study areas (n = 336) Correlation analysis of study variables Table 8 shows the correlation coefficients between the dependent variable (healthcare data quality) and independent variables (socio-demographic characteristics). A strong positive correlation was found with age (0.617), sex (0.636), and education (0.613), while marital status (0.507) and experience (0.484) show moderate and weak positive correlations, respectively. Additionally, the results show the relationship between independent variables, the strongly correlated variables were education and Sex (0.563), age and Sex (0.561), education and marital status (0.561) and age and education level (0.524). Table 8Correlation coefficients of study variables (n = 336)AgeSexEducationMarital statusExperienceData QualityAge1.Sex0.5611.Education0.5240.5631.Marital status0.4560.4840.5611.Experience0.3730.3700.4410.5011.Data Quality0.6170.6360.6130.5070.4841.p <.001 Table 8Correlation coefficients of study variables (n = 336)AgeSexEducationMarital statusExperienceData QualityAge1.Sex0.5611.Education0.5240.5631.Marital status0.4560.4840.5611.Experience0.3730.3700.4410.5011.Data Quality0.6170.6360.6130.5070.4841.p <.001 Correlation coefficients of study variables (n = 336) p <.001 Regression analysis of socio-demographic factors and healthcare data quality The regression analysis was performed to find the association between socio-demographic factors and healthcare data quality. The study used the following regression model: -.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\text{Y}} = \beta _{0} + \beta _{{\text{1}}} X_{{\text{1}}} + \beta _{{\text{2}}} X_{{\text{2}}} + \beta _{{\text{3}}} X_{{\text{3}}} + \beta _{{\text{4}}} X_{{\text{4}}} + \beta _{{\text{5}}} X_{{\text{5}}} + \varepsilon$$\end{document} In the context of this study, Y represented a dependent variable (Healthcare data quality), The letter β0 denoted a constant, illustrating the expected value of Y when all predictor variables (X1– Age, X2–Sex, X3– Education, X4– Marital status and X5–Experience) are set to zero. The coefficients β1, β2, β3, β4, β5 are constant regression coefficients that indicate the relationship between each independent variable and the dependent variable. Additionally, ε represented an extraneous variable which accounted for other factors that affected healthcare data quality but were not captured in this model. Table 9 examines various predictor variables and their influence on multiple data quality outcomes, including data accuracy, completeness, timeliness, consistency, and overall data quality. Each predictor is evaluated using a beta coefficient and p-value, which reflect its impact and statistical significance. Age (20 to 39 years) positively affects all outcome variables, with significant p-values (p <.001). This means that young individuals tend to report better data accuracy, completeness, and timeliness in data entry, although age does not significantly impact consistency. Sex also plays a significant role, especially in accuracy (β = 0.208, p =.000), completeness (β = 0.428, p =.000), and consistency (β = 0.561, p =.000). However, it does not significantly affect timeliness, as shown by a p-value of 0.306. Education shows a mixed effect; it positively influences timeliness (β = 0.240, p =.003) but negatively affects completeness (β=−0.166, p =.002) and consistency (β=−0.144, p =.002). Marital status has limited effects, showing a significant positive relationship only with completeness (β = 0.139, p =.044) and no significant impact on other outcomes. Experience negatively affects accuracy, timeliness and consistency, particularly for timeliness (β=−0.369, p =.000), suggesting that more experience does not necessarily improve data quality in these areas. Overall, the model shows reasonable explanatory power, especially for consistency (R2 = 0.507) and overall healthcare data quality (R2 = 0.516), with F-statistics indicating that the predictors significantly influence the outcome variables. Table 9Regression analysis of socio-demographic factors and healthcare data quality (n = 336)Predictor variablesOutcome VariableAccuracyCompletenessTimelinessConsistencyData QualityBetaP-ValueBetaP-ValueBetaP-ValueBetaP-ValueBetaP-ValueAge0.2410.000***0.1590.000***0.2020.000***0.0110.7690.1130.000***Sex0.2050.000***0.4280.000***0.0430.3060.5610.000***0.1770.000***Education0.1640.011*− 0.1660.002**0.2400.003**− 0.1440.002**0.0460.462Marital status− 0.0960.2390.1390.044*− 0.0670.2600.0580.3240.0500.147Experience− 0.3390.3420.0120.907− 0.3690.000***− 0.0620.4970.0290.512R square0.3960.3990.4430.5070.516F-statistics41.300.000***41.670.000***50.070.000***64.900.000***67.140.000***Significance at the ***P ≤.001, **P ≤.01; *P <.05 Table 9Regression analysis of socio-demographic factors and healthcare data quality (n = 336)Predictor variablesOutcome VariableAccuracyCompletenessTimelinessConsistencyData QualityBetaP-ValueBetaP-ValueBetaP-ValueBetaP-ValueBetaP-ValueAge0.2410.000***0.1590.000***0.2020.000***0.0110.7690.1130.000***Sex0.2050.000***0.4280.000***0.0430.3060.5610.000***0.1770.000***Education0.1640.011*− 0.1660.002**0.2400.003**− 0.1440.002**0.0460.462Marital status− 0.0960.2390.1390.044*− 0.0670.2600.0580.3240.0500.147Experience− 0.3390.3420.0120.907− 0.3690.000***− 0.0620.4970.0290.512R square0.3960.3990.4430.5070.516F-statistics41.300.000***41.670.000***50.070.000***64.900.000***67.140.000***Significance at the ***P ≤.001, **P ≤.01; *P <.05 Regression analysis of socio-demographic factors and healthcare data quality (n = 336) Significance at the ***P ≤.001, **P ≤.01; *P <.05 Discussion This study examined the relationship between various socio-demographic factors of healthcare workers and perceived healthcare data quality from selected eight regional referral hospitals. The findings indicate that all studied socio-demographic factors (age, sex, education, marital status and years of experience) have positive correlation with the overall perceived quality of healthcare data. Notably, age and Sex emerged as significant predictors, demonstrating a strong association with healthcare data quality in healthcare settings, specifically data accuracy and completeness. These findings are similar to a number of empirical literature [19, 22, 29, 30]. The findings indicate that while age does not significantly impact data consistency, it does positively affect data quality. Young individuals tend to report more accurate, complete, and timely data. This finding is consistent with existing literature suggesting that age does significantly affect job-related stress, satisfaction, performance, and ultimately data quality in the healthcare sector [21]. Younger workers often face less stress and have more enthusiasm early in their careers, which boosts their job satisfaction and performance, leading to better data entry. Additionally, young individuals are frequent users of digital technologies even in their personal lives [31, 32], this makes them more familiar with electronic M&E systems too. On the other hand, older healthcare workers may struggle with electronic M&E systems due to less familiarity with technology, which can result in more errors. However, providing adequate training can help bridge the technological gap, enabling healthcare workers of all ages to produce high-quality data [33]. The study also found that sex plays a significant role in the perceived production of quality healthcare data, especially in data accuracy, completeness, and consistency, indicating that male workers influence higher reporting in these areas. Existing literature presents mixed findings on this issue. Some research indicates that female healthcare workers spend more time engaging with electronic health systems, which may enhance the accuracy of data entry [34]. Others argue that this increased time often comes at a cost, as female workers frequently report higher workloads, job burnout, and stress [35]. These contrasting perspectives highlight the complex relationship between Sex and data quality. While the additional time women spend on health information systems can lead to more accurate data entry, the stress and burnout associated with their workloads may ultimately hinder their ability to produce quality data. This calls for efforts by healthcare institutions to address these challenges for optimizing data quality across sex groups. The study found no significant association between marital status and the perceived quality of healthcare data, with only a weak relationship noted in data completeness. This suggests that unmarried healthcare workers do not necessarily produce quality data, which contrasts with the assumption that personal situations like marital status would influence professional performance. In contrast, Yendemye and Nimubona [21] argue that married workers may perform better in enhancing data quality for two key reasons: their stable home life and emotional support, which contribute positively to data management, and the additional family responsibilities that may lead them to value their jobs more, resulting in greater attention to data quality. These factors suggest that, for some workers, marriage may provide both practical and emotional stability, which could improve their focus and performance in tasks such as data entry. On the other hand, these studies [19] and [36] found that marital status does not significantly affect healthcare data quality. They offer three explanations for this: first, family responsibilities can introduce additional stress for married workers, which may reduce their focus on data entry compared to single workers who have fewer personal obligations. Second, married individuals often have fewer flexible schedules, limiting the time they can dedicate to thorough data entry. Third, while marital happiness might increase focus and attention to detail, those experiencing marital difficulties may struggle with concentration, leading to errors in data entry. These differing viewpoints highlight the complexity of the relationship between marital status and work performance. While the current study suggests no strong connection, other research offers sensible reasons for how marital status could either enhance or hinder data quality. These variations in findings could be influenced by cultural, environmental, or individual differences, and future studies may need to explore these factors more deeply. The study also found mixed effects between healthcare workers’ education levels and healthcare data quality, where education level positively affects timeliness of data while negatively affects data completeness and consistency, indicating that higher education may lead to timely reporting but complicate other aspects of data quality. This finding opposes some literature that emphasizes the importance of education level in enhancing data quality. For instance, Solomon et al. [24] argue that higher education correlates with better data outcomes due to increased confidence, similarly Ndishaa et al. [25] suggest that higher education improves skills and attention to detail, leading to better data management. However, other studies, including those by Theofileo [30] and Glèlè et al. [37], argue that education alone is insufficient, instead, factors like work engagement, satisfaction, and organizational support are critical. Ndishaa et al. [25] also propose that organizational culture and support can outweigh individual education levels. Admittedly, despite the fact the current study found high education level is important in the production of data quality, a comprehensive approach that includes training and supportive environments is necessary for improving healthcare data quality. In addition, the association between years of experience and healthcare data quality is complicated. Research often shows that experienced workers know the health information system well and can use it more quickly than those with less experience [38]. This familiarity can lead to more accurate and complete data, fewer mistakes, and better timeliness. However, the current study questions the common belief that years of experience alone determine data quality. There are three main reasons for this. First, a worker with extensive experience might not have received proper training in data management or in using newer system modules. Second, the work environment, including IT infrastructure robustness, system interoperability, and user-support services can significantly moderate performance [39]. For instance, even experienced staff may struggle in facilities with isolated systems that require duplicate entry or lack real-time integration. Third, demographic factors such as digital literacy, age, and educational background interact with these environmental conditions: younger or more digitally literate workers tend to adapt quickly to new interoperability standards, whereas others may experience frustration or make more errors when infrastructure is poor [40, 41]. Indeed, Komba and Lwoga [40] found that in hospitals with high system interoperability, staff with lower baseline digital skills achieved data-entry accuracy rates comparable to their more experienced peers, provided sufficient on-site IT support was available. Similarly, Zheng et al. [41] showed that older healthcare workers’ familiarity with clinical workflows only translated into high-quality data when core infrastructure such as stable networks, and integrated electronic records minimized manual workarounds. Taken together, these findings suggest that enhancing both the technological environment and tailored training rather than relying on tenure alone may be the most effective route to consistently high-quality healthcare data. Implication for Tanzania’s health system Our findings underscore the need for a coordinated national strategy that aligns digital-health investments with workforce diversity. At the policy level, the Ministry of Health should integrate socio-demographic-tailored data-quality training into its Health Information System strategy, ensuring both pre-service curricula and in-service programs address age, sex, and digital-literacy gaps. Resources must be prioritized to upgrade IT infrastructure especially interoperability and real-time integration across all regional referral hospitals to eliminate fragmented systems and reduce redundant data entry. Facility managers should establish on-site IT support teams and flexible staffing models to mitigate burnout and influence the strengths of both younger, tech-adept staff and experienced workers. Finally, embedding regular, sociodemographically sensitive audits and feedback circles will help track progress, fine-tune interventions, and foster a culture of continuous improvement; together, these measures will strengthen Tanzania’s health information backbone and ultimately enhance patient care outcomes. Study limitation and area for further research This study has some limitations. First, it uses a cross-sectional design, which means it collects data at a single point in time. This approach only shows a snapshot of the relationship between the variables. Future research should consider longitudinal studies to track how socio-demographic factors relate to the quality of healthcare data over time. Second, the study focuses on the general socio-demographic characteristics of healthcare workers, giving a general overview of their characteristics related to healthcare data quality. Future studies could use mixed methods and more in-depth analysis to explore these factors further. For example, examining marital status could reveal how being single, cohabiting, separated, divorced, or widowed affects data quality. Similarly, comparing the impact of different education levels, like certificates versus bachelor’s degrees, on data quality could provide valuable insights. This complex relationship might not be fully captured by multiple linear regression, even though it effectively identifies associations. This study did not include measures of occupational stress, excitement, or job satisfaction, focusing solely on socio-demographic predictors. Future research should incorporate these psycho-social factors to more comprehensively understand their influence on healthcare data quality. Even with these limitations, the study methodology is strong, using a solid sampling strategy and trustworthy data collection tools, which backs up the findings. The sample accurately reflects the healthcare workers, making the results more applicable. In general, the findings offer important information that can guide practice and policy, adding to the existing research and pointing out areas for future study. Conclusion This study demonstrated that among healthcare workers in eight Tanzanian regional referral hospitals, age and sex significantly predict data accuracy and completeness, while education, marital status, and experience showed mixed or non-significant effects. All conclusions are drawn strictly from those socio-demographic variables; we did not measure potential confounders such as occupational stress, job satisfaction, or ingrained technology-use habits, so broader inferences should be made with caution. Future research incorporating these psycho-social and behavioral factors will help validate and extend our findings. Ultimately, recognizing and addressing socio-demographic influences on data quality is essential for strengthening Tanzania’s health information system and improving patient care. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1