USE OF A GOOGLE-BASED ELECTRONIC DATA COLLECTION SYSTEM FOR MANAGING HAEMODIALYSIS PATIENTS IN A LOW-RESOURCE SETTING

 Collection of accurate data on patient cohorts including demographics, diseases, treatments and outcomes is essential for clinical care, quality improvement and planning. Electronic data systems offer the opportunity to gather and view data of large cohorts of patients from a single point, providing an opportunity for more systematic management and audit, particularly when the case burden per health care worker is very high.  While a few centres in Sri Lanka have developed electronic medical record systems, across most of the country, records remain paper-based. Limitations in expertise, funding and infrastructure to develop and sustain electronic data systems remain a barrier. Thus any form of data collection is labour intensive, time-consuming and often incomplete.  

Maintaining real-time patient data is a fundamental need for evaluating quality and performance in haemodialysis (HD) units. This is challenging using paper-based methods, and has been one of the barriers to developing and maintaining a sustainable national HD registry. A HD unit follows a defined cohort of patients and is an ideal setting for piloting an electronic data system which can then be developed and customized for other settings.  We describe our experience implementing a simple cloud-based data entry platform to capture and visualise HD patient data, and report an audit of its completeness and challenges encountered.

We developed a simple google- based data entry and visualisation system for managing patient clinical and laboratory data in the University  HD unit of the National Hospital of Sri Lanka. Google is a HIPPA complaint platform, and data was only accessible to the investigators via their google accounts. Data entry was using google forms, calculations were performed using multiple interlinked google sheets, and data was visualized using Looker studio. All patients were assigned a unique dialysis unit identification number (ID) and registered in the electronic system.. Data collected included demographics, comorbidity, indication for dialysis (acute kidney injury, acute on chronic kidney disease, and end-stage kidney disease), dialysis sessions, and laboratory tests (haemoglobin, serum creatinine and potassium, blood urea). Serial data entries of each patient were linked by their HD unit ID. Patients who had not received HD in the last 2 weeks were flagged and they were contacted to assess for outcomes. These were manually updated as HD ongoing, HD in another centre, transplanted, transferred to peritoneal dialysis, dialysis withdrawn, deceased or unknown. Data collection began in May 2025. During this period the usual paper-based data collection was supplemented with the electronic system.

We performed a retrospective audit of the pilot electronic data system used in the dialysis unit  from its implementation in May 2025  to Oct 1st 2025. The aim of the audit was to assess completeness and accuracy of data collection and to identify the barriers to using the electronic system reliably. To determine registration status and session documentation completeness, we randomly selected 50 patients with end stage kidney disease and  performed a detailed session-by-session verification against primary source documents (paper documents)  to assess recording accuracy.

339 unique patients had undergone HD during the period according to paper-based clinical logbooks and nursing records, amounting to 2147 HD sessions in total. Of these 285 had been entered into the electronic system.  53 patients (15.63%) had  not been registered in the electronic system. These non-registered patients had undergone HD between May and July, when the data entry was in its early stages.  Of the 2,147 sessions,  1,604 (74.7%) had been documented in the system. In a random sample of 50 patients, 418 of 543 verified dialysis sessions (77.0%) had been recorded electronically. 20 patients had been entered into the electronic system more than once with more than one ID. These duplicates were removed. Duplication occurred due to inconsistencies in entry of name and surnames (spellings, punctuation after initials, first name vs preferred name) and the absence of a unique patient identification number across admissions, which was only partially circumvented by the use of the dialysis unit ID. 6 patients had been assigned a DU ID allocated to another patient. This occurred due to issues on internet connectivity at the time of registration, and slowness of the system to get updated.  In 4 patients HD sessions had been entered under the wrong dialysis unit ID, due to poor legibility of the handwritten notes .  For 90 patients (26.5%) the current status could not be identified, the most recent HD among these patients having been >1 month before.

Data completeness improved with time. Completeness was lowest in May and July. In May this may have been due to the learning curve and in July this was due to network connectivity issues. In this hybrid paper-electronic workflow, nurses preferred to collect data throughout the day and enter it twice daily, contributing to time lag and potential omissions.

Discrepancies in patient registration and HD session recording were primarily caused by data connection issues that prevented real-time database updates.  Significant time and resources were spent identifying and resolving these duplicates. 

This pilot demonstrates the feasibility of a simple cloud-based data collection method developed using minimal IT background and resources. Data completeness improved with time. Continued use of paper-based data entry may have contributed to some incompleteness in data and errors due to illegibility. While the system makes it easier to collect outcomes of patients, manual follow up of those no longer on active HD remains challenging.