The implementation of electronic health information systems (HIS) have been pivotal in enhancing healthcare data quality on a global scale[1]. One of the many strengths of implementing HIS is its capability to securely share data between facilities within a network in what is referred to as a health information exchange (HIE). Several studies have reported HIE benefits that includes reductions in laboratory test costs, improvements in record completeness, and timeliness of reporting for disease surveillance[2,3]. A study showed number of days between testing and reporting to health departments reduced from 4 to 11 days to 2 days with HIE implementation[4]. There have also been 54% and 80% reduction in costs for clinical lab tests and diagnostic imaging respectively[5] and improved health outcomes with reduction in all-cause mortality and adverse events with HIE implementation[6].

HIE implementation may be challenged with issues around data quality, infrastructure and data security [7]. Technologies like blockchain and cloud computing, however, can help address some of the infrastructure and security challenges[8-10]. Evidence exists for potential errors that could be observed during the HIE process including wrong dates, inconsistent vocabulary, and invalid observations[11] and missing or incomplete patient demographics in the source system[12]. This study contributes to the body of evidence on the impact of HIE on data quality.

Accurate, reliable, and timely HIV viral load (VL) testing is not only essential for the successful treatment and management of HIV/AIDS[13], but it is also the preferred method of confirming the effectiveness of antiretroviral therapy (ART) for a person living with HIV (PLHIV). Global demand for HIV viral load testing has continued to increase as countries have scaled up their treatment programs towards achieving the joint United Nations Program on HIV/AIDS (UNAIDS) 95,95,95, targets by 2030[14,15]. Nigeria expects 95% of PLHIV receiving ART to achieve viral suppression which implies that VL testing will have to be conducted for more than 1.8 million PLHIV annually. Achieving this goal requires consistent monitoring of the VL testing cascade to identify inefficiencies and develop mitigation plans. In general, the end-to-end VL turnaround time (TAT) is the time between requesting the VL test and receiving the results back for clinical decisions. Measuring TAT is a generally acceptable approach to understanding process inefficiencies in the VL testing life cycle[16]. However, examining TATs is often challenged by the quality of data documented during the life cycle[17,18].

The United States Centers for Disease Control and Prevention (CDC) supports the provision of healthcare services to PLHIV in Nigeria by working with local and international organizations (implementing partners) and the government of Nigeria to provide both technical and financial support to healthcare facilities (n = 956). Implementation of electronic medical record (EMR) systems for HIV programs in Nigeria dates as far back as 2004[19]. As of March 2020, 98% of the CDC-supported healthcare facilities have had an EMR system implemented for their HIV programs. The EMR systems implemented include OpenMRS, an open-sourced EMR system, (98% of the population) and other proprietary EMR systems. 99% of the healthcare facilities use their EMR only for the HIV program. Additionally, Nigeria implements a national laboratory information management system (LIMS) used by the polymerase chain reaction (PCR) laboratories for the management of VL testing data. To improve overall VL TAT for all time points during VL testing process, the President’s Emergency Plan for AIDS Relief (PEPFAR) program in Nigeria began the implementation of an electronic data exchange between the EMR at the treatment facilities and the LIMS at the PCR labs as part of routine program activity in 2021. This study assesses the effect of implementing this electronic health information exchange between the EMR and LIMS on the quality of data available for measuring VL TAT. In this article, we described the implementation of HIE between the EMR and LIMS, examined the completeness and validity of data pre- and post-implementation of HIE, and then compared the results to determine the impact of HIE on data completeness and validity.

We examined the validity and completeness of VL TAT data by comparing data of specimens from healthcare facilities collected between October 2020 and January 2021, as baseline, and those collected between August 2021 and January 2022, as endline. We defined the study intervention as the implementation of automated HIE. Baseline VL TAT data were collected before the implementation of an automated HIE, and endline VL TAT data were collected 6 months postimplementation.

We used purposive sampling to select 3 high volume PCR labs with a functional LIMS that were supported by two CDC Nigeria implementing partners. Purposive sampling was also used to select 30 treatment facilities that: 1) were supported by CDC Nigeria’s implementing partners; 2) had onsite clinical laboratories for specimen collection and processing; and 3) were linked with the selected 3 PCR labs. Consideration was also made to ensure the 30 selected treatment facilities had an adequate representation of a mix of urban, semi-urban and rural sites, as well as ease of access to the locations.

Secondary data (Table 1) from routine VL testing was used for the analysis both at baseline and at endline. Given the EMR-LIMS data exchange was not established at baseline, we performed manual data abstraction from EMR and paper registers at the treatment sites at baseline. At endline, data was extracted from the EMR, and LIMS as shown in table 1. The two datasets from EMR and LIMS were merged using both specimen identifier and patient’s treatment identifier at the treatment facility as linkage keys. All specimen identifier and the patient treatment identifier were removed from the dataset prior to analysis. Informed consent was waived as this was secondary use of data and patients’ identities could not be determined during analysis. We used both Microsoft Excel^®, R studio and OpenEpi[20] for data collation and analyses.

There was a total of 15,226 specimens from 28 healthcare facilities included in the analysis at baseline and 18,022 specimens from 23 facilities included at endline. There were 4 facilities from one partner that were initially included in the baseline analysis that were subsequently excluded from the endline analysis because these facilities did not implement an electronic HIE and opted to implement a different intervention. Records from two healthcare facilities at baseline and three at endline were excluded from the analysis due to missing pertinent information including missing specimen unique identifiers and test types.

This study assessed for improvements in both data completeness and validity of data for VL specimen testing through the entire VL testing lifecycle from the date a VL test was requested, to the date VL testing results were placed in a patient’s folder. Study findings include overall significant improvements in both percentage completeness and validity post-implementation of a HIE. There were also observed decrease in validity and completeness for some TAT segments and data elements respectively. Previous studies have shown similar findings when assessing for improvements in data quality post implementation of a HIE[24,25]. One study that examined the effect of leveraging a HIE to support public health situational awareness in Indiana showed more than 40% increase in data completeness[24]. Another study that accessed the impact of exchange between immunization information system and an EMR system recorded an increased completeness score from 75.0% to 81.6% after 6 months[25].

While this study did find significant improvements in VL TAT data quality overall, there were still opportunities for improvements as demonstrated by decreases in validity and completeness in specific TAT segments three, six, and nine. Several reasons can be attributed to the current existing gaps in record completeness and validity. First, the process workflow in LIMS at the PCR labs has yet to be updated to be in line with the automated data exchange. As the automated data exchange was introduced, the workflow on LIMS was not reviewed to ensure proper documentation of the newly automated steps such as specimen arrival at the PCR lab, automated result update from the PCR machines, and automated result update to the EMR at the facilities. Consequently, this resulted in reduced completeness for data elements including date specimen was resulted, and date results were sent back to facility and reduced percentage validity for TAT3, TAT6, and TAT9. Second, there were data standardization challenges specifically pertaining to inconsistent use of data types. Upon further investigation, we found that LIMS at the different PCR labs reported the same data elements in different data types which was most profound in the date fields. Consequently, some date fields on LIMS were documented as data type datetime while others were documented as data type date. Hence, some TAT segments that occurred on the same calendar day (TAT3 and TAT6) resulted in invalid values.

The next steps include a review of the workflow for HIE and updating the necessary HIS (EMR or LIMS) to align with this workflow. This will ensure that all the necessary steps are properly documented and will result in increases in the percentage of specimens with complete data. Secondly, the standardization of data (common data formats) across all labs is crucial especially for VL TAT data validity. Data standardization activities including the development and implementation of a data dictionary following standard terminologies and vocabulary such as Systematized Nomenclature of Medicine-Clinical Terms (SNOMED-CT) for electronic exchange of clinical health information, Logical Observation Identifiers Names and Codes (LOINC) for identifying medical lab observations, and other available data standards will enable semantic interoperability. This will not only improve on the quality of the data but also support ease of data integration [8,26]. Continuous monitoring and evaluation of data quality is also a very important step to identifying and quickly resolving gaps[27,28]. A standard approach for addressing such issues in data completeness and validity needs to also be implemented. Strategies including a problem solving approach such as the Plan Do Study Act (PDSA) cycle[29], and a root cause analysis[30] can be performed to assist in identifying the sources or root causes of the gaps in data quality and then implementing appropriate solutions to address them.

A limitation of this study was that there were no control group given that the HIE was implemented as a routine program activity. Therefore, only facilities that had implemented the automated health information exchange at endline were used for the analysis. It is important to have had a study control because while we have recorded statistically significant improvement in data completeness and validity for these specimens, there could have been other factors that have contributed to these improvements which the study may not have accounted for. Excluding 2 and 3 healthcare facilities, at baseline and endline respectively, from the analysis due to missing information was on itself a data quality issue that could not be accessed in this study as complete datasets could not be developed for these healthcare facilities. Additionally, facility level clustering effect was not included in the analyses. However, it may be noted that all the significant results have p-values of<0.01, and the conclusions would not change if there was some clustering effect. Specifically, results would hold for any design effect of 2 or less. Finally, this was a small-scale study and may not be a good representation of the population of specimens tested for VL in Nigeria. Hence, it may be difficult to generalize the findings of this study.

Results from this study raise evidence to support automated HIE for improving data completeness and validity. It also highlights key lessons that can help programs that are planning or already implementing automated HIE understand important areas to focus on as they plan and implement their systems. While this study has shown improvement in data quality with the implementation of HIE, it will be important to conduct further analysis to understand how automated HIE impacts VL TAT, which is the goal of such implementation.