Comparative Analysis of Point-of-Ccare vs Central Lab Diagnostics Accuracy

Thesis Abstract

In many healthcare settings, the accuracy of diagnostic results from point-of-care (POC) testing relative to centralized laboratory (CL) testing remains contested, with implications for clinical decision-making, patient outcomes, and health system efficiency. This study addresses the problem by examining whether POC diagnostics yield results that are clinically comparable to CL assays across routine laboratory measurements, and how factors such as operator training, device calibration, and specimen handling influence agreement and diagnostic decision thresholds. The aim is to quantify concordance and bias between POC and CL results and to identify determinants of discordance that affect clinical interpretation. Specific objectives are (i) to evaluate analytical concordance between POC and CL measurements for a panel of biochemistry and hematology tests (glycated hemoglobin, glucose, creatinine, C-reactive protein, potassium, hemoglobin concentration, and white blood cell count) in adult inpatients and outpatients; (ii) to assess the diagnostic accuracy of POC results using CL results as reference standards, calculating sensitivity, specificity, positive and negative predictive values at clinically relevant cutoffs; (iii) to quantify systematic bias and limits of agreement with Bland-Altman analysis and to model factors associated with discordance using multivariable linear and logistic regression; (iv) to explore user- and device-related determinants of measurement variability through a nested mixed-methods component integrating operator surveys and observational checklists; and (v) to synthesize findings within a theoretical framework anchored in the Technology Acceptance Model (TAM) and the Analytical Quality in Healthcare model to derive implications for practice and policy. Methodologically, the study adopts a cross-sectional diagnostic accuracy design conducted across three tertiary hospitals over 12 months. The population comprises adult patients requiring routine laboratory testing, with a target sample of 1,200 paired measurements, ensuring balanced representation across inpatient and outpatient cohorts, age strata, and disease categories. Sampling uses stratified random sampling to select participants, and consecutive sampling within strata ensures adequate numbers for each analyte. Data collection employs standardized venous and capillary specimen collection protocols, with POC devices including a hematology analyzer, a glucometer, a point-of-care creatinine/urea device, and a CRP handheld immunoassay, alongside a centralized reference laboratory performing certified methods (e.g., enzymatic assays, immunoturbidimetry, and automated hematology analyzers). Instruments include instrument-specific quality control materials, calibration logs, operator proficiency records, and a structured data collection form capturing device type, lot number, operator credentials, and environmental conditions. Validity and reliability are addressed through traceability to reference methods, regular calibration, dual-measurement replication, and inter-device comparison. Statistical analyses commence with descriptive statistics for demographic and clinical variables, followed by Bland-Altman plots to evaluate agreement and Pitman’s test for equality of variances. Correlation coefficients (Pearson/Spearman as appropriate) and Passing-Bablok regression assess proportional and constant biases. Diagnostic performance is quantified via sensitivity, specificity, receiver operating characteristic (ROC) curves, and area under the curve (AUC) calculations for each analyte at clinically relevant cutoffs. Multivariable linear regression identifies factors associated with measurement bias (POC minus CL values), while logistic regression models determinants of discordant classification around clinical thresholds. A mixed-methods component analyzes qualitative data from semi-structured operator interviews and direct observation to triangulate quantitative findings, with thematic analysis guided by the Technology Acceptance Model to interpret adoption barriers and facilitators. Expected findings anticipate moderate to strong concordance for glucose and hemoglobin with clinically acceptable biases under predefined acceptance criteria, but greater discordance for potassium and CRP at extreme values, primarily due to hematocrit interference, sample handling, and device-specific limitations. Operator training and device maintenance are hypothesized to significantly reduce discordance, with TAM indicators (perceived usefulness and ease of use) correlating with reduced measurement variability. The study contributes to knowledge by providing robust, context-specific estimates of POC diagnostic accuracy across commonly used analytes, identifying actionable determinants of measurement discordance, and validating a practical framework for integrating POC testing into existing clinical pathways without compromising patient safety. The main conclusion is that POC diagnostics can achieve clinically acceptable accuracy for select analytes under stringent quality control and proper implementation, but caution is warranted for others where discordance risks misclassification. Recommendations include standardized training programs, regular device calibration, expanded QA/QC protocols, incorporation of decision support that accounts for known biases, and policy guidance on when CL confirmation is required, with prospects for scalable implementation in resource-limited settings.