Of the 3549 samples evaluated as part of the quality audit, 54% (n = 1923) were not testable (Table 1). There was variation in this rate across the departments within the laboratory, ranging from 5% of samples for microbiology to 70% of samples sent to the blood bank department. An insufficient sample volume was the most common reason for a sample being deemed untestable (n = 1606). This characterised over 80% of untestable samples (n = 1923) and 45% of all audited samples (n = 3549). Of the high number of samples that were considered untestable because of insufficient blood volume, the vast majority came from the paediatric department, where it is challenging to get sufficient blood from a sick and frequently dehydrated infant. Test order forms were filled out either incorrectly or incompletely in just over 16% (n = 591) of audited samples. The least frequent problems identified were samples that were mixed up, too old, or in the wrong container.

Between July 01, 2010 and June 30, 2011 the KCH laboratory conducted 242 242 tests (Table 2). The number of tests carried out for the parasitology and blood bank departments accounted for over half of the laboratory's total test workload during this time. The most common tests were: malaria parasites; full blood count; blood grouping and cross matches; and CD4 count. There was considerable monthly variation in the number of tests conducted, reflecting the seasonality of diseases such as malaria. The average monthly test load was 20 187 tests, with a standard deviation of 3263 tests. There were several classes of tests, including blood lipids and hormones, which were not conducted at all during this time period because of inoperative equipment and/or lack of reagents.

Figure 1 defines the total testing process workflow at KCH according to the commonly-specified nine stages of laboratory testing, and indicates the staff responsible for the task.^15,25 The process begins and ends on the patient ward with the physician, who makes the initial request for a laboratory test and also determines a course of action based on the interpretation of the test results. The pre-analytical phase is initiated with the ordering of a test and also includes the sample collection and identification or matching of the sample with the patient. These two tasks are both completed by a clinician or nurse and are associated with some of the issues identified in the quality audit, such as incomplete or incorrect labeling and sample mix-ups. Incomplete or illegible labeling result from a number of factors, including insufficient label space and lack of necessary information at time of ordering. The pre-analytical phase then extends beyond the patient ward to include transport of the sample to the laboratory and preparation of the sample for testing by a laboratory technician. The transportation phase represents the transfer of the specimen and associated information from one department to another within the hospital; this task is not assigned to a specific job title in the hospital. It is most likely to be carried out by a nurse, a patient attendant, or a janitor, but there is no established routine for sample transport. Samples waiting for transport to the laboratory are typically stored in a treatment room or at a nursing station, and are taken with varying frequency to the laboratory. It is thus clear that one challenge is a lack of awareness of the time required for sample transport, so issues such as delays or misplaced samples are not proactively addressed. In addition, time is spent during this stage transcribing various information from the test order form, so information is being duplicated. This phase concludes with the preparation phase, in which the laboratory technician uses the test order information to prepare the sample for analysis. Informational gaps may cause delays at this point in the process because, just as the sample transport does not have a systematic workflow, the reporting of sample and test order errors back to the wards is similarly unstructured.

The analytical phase includes a single step, namely, the analysis of the laboratory sample. Information is generated as part of this process and is combined with information supplied during the pre-analytical phase (patient age, gender, etc.) to generate a result. The manual nature of matching the test results with the corresponding patient based on patient name is another informational step that can cause delays and potentially lead to reporting errors. The process of reporting the result back to the physician is the start of the post-analytical phase and, again, is not a formalised process. It could be performed by a variety of personnel at unspecified frequencies. Results are left in the laboratory entryway in cubby-style pigeon holes for pick-up. Often, when someone is sent from a ward to drop off laboratory samples, they will also pick up and deliver any results that are available for that ward. This means that critical results may not be reported to the wards in an expedited manner.

Table 3 presents the duration of the reporting stage for laboratory results from a 24-hour observation of the laboratory. Results were processed for all 25 patients who had test orders sent to the laboratory during the observation period. Test orders were reduced as the haematology instrument was not operational at that time and polymerase chain reaction results for outpatients were delivered through a different mechanism. Additionally, the hospital census was unusually low that day. Of the 25 results, 18 were collected during the 24-hour observation window, whilst seven remained in the pigeon holes. The average duration of the reporting stage was just under six hours, with significant variation. Two of the results were collected immediately because they were related to a critical patient and the laboratory called the ward when the results were available. On the other hand, over one-fifth of results spent more than 16 hours in the laboratory before being collected.

The results present a multi-faceted depiction of laboratory testing in a hospital in a low-resource setting, focusing on both the demand for laboratory testing and the informational challenges in meeting that demand. With more than 240 000 tests conducted at the KCH laboratory during a one-year period, it is clear that laboratory services are very much in demand within the hospital, matching the rhetoric around the importance of accessible laboratory services in low-resource settings. However, the diagnostic process is information-intensive; and the quality audit and workflow analysis suggest that the capture, management, and transfer of this information are a significant barrier to maximising the laboratory's role at KCH.

The quality audit identified informational barriers in the collection and identification of samples. Almost one-sixth of samples were untestable because of incomplete or incorrect paperwork. The quality audit was conducted as a quality improvement effort and, as motivation for the study, laboratory employees articulated several challenges associated with incorrect or incomplete test orders. Firstly, patient details, such as age and gender, affect the interpretation of the results; and test orders that omit these data increase the likelihood of an interpretation error. Secondly, the time and date of the sample is particularly important for tests sent to either the microbiology or the biochemistry departments, so the absence of this information compromises the accuracy of the test. Finally, the current system for results reporting relies on ward information from the test order, so when this information is not included it is common for results to never reach the patient.

These pre-analytical errors and delays are similar to those found in other low-resource laboratory environments.²⁶ Post-analytical delays were also evident from the analysis of results reporting, although the sample size was small and further examination is warranted. These inaccuracies have implications for both the hospital and the patients. Untestable samples equate to wasted resources, including the physical supplies for the sample, such as the syringe and collection vessels, as well as employee time involved in collecting the sample and communicating the error. For patients, these errors amount to delays in care; surgical patients may face delays in scheduled surgeries if laboratory results are not ready, whilst for others it may mean an extra night in the hospital. Efficient laboratory testing is particularly important for patients in critical condition, namely, those who face delays in life-saving care and/or empirical treatments prescribed in the absence of confirmatory laboratory results.

The findings of the quality audit also informed the interpretation of the data around the frequency of laboratory testing. The information about test volume reflected tests conducted by the KCH laboratory and therefore serves as a conservative measure of the demand for laboratory services at KCH. There are at least two reasons that demand for laboratory services may be greater than that shown in Table 1. First, samples deemed untestable may not always be corrected and re-sent to the laboratory, as the process for informing clinicians of untestable samples is unstructured and thus possibly lengthy; and clinicians may choose to pursue diagnosis and treatment without confirmatory laboratory results. Second, material shortages, such as reagents and properly functioning equipment, may prohibit the performance of certain kinds of tests.

The workflow analysis mapped the laboratory testing process at KCH to the standard stages of testing and identified the informational components of each stage. This exercise identified a lack of formalised workflow around the transport of samples from the wards to the laboratory and the reporting of results from the laboratory to the wards. The reporting delays were confirmed in the study of reporting times, although a limitation of this component of the results is the small sample size. These gaps suggest that one unique aspect of the informational challenges faced by laboratory systems is the geographic scale of the testing process. Unlike the paperwork issues identified in the quality audit, the workflow issue involves interactions amongst multiple agents across different departments.

The standard approach to laboratory testing does not necessarily reconcile these multiple players and environments. Information systems tend to have a departmental focus. For example, electronic medical records are patient-centric systems focusing primarily on managing clinical patient data. Whilst this often includes laboratory test results, electronic medical records functionality does not extend into the laboratory. Laboratory information systems, on the other hand, are typically specimen-centric systems, focusing on processes and workflows within the laboratory. In this scenario the generation of test orders and reporting of test results often falls within the gap between the electronic medical records and the laboratory information system.

We propose a system for supporting specimen management, increasing visibility into the status of all orders for both clinical as well as laboratory staff and following a model embraced by courier companies (DHL/Federal Express/UPS) to track and manage packages. This simple model has three main components: (1) the generation of a test order and associated paperwork; (2) the ability to monitor the status of the order, complete with exception alerts when applicable; and (3) electronic results reporting back to the ward. The approach of real-time monitoring of specimens as they move through each stage of the process is novel and is likely to have a higher impact in a low-resource setting, where challenges are arguably greater. In Table 4, we summarise problems identified in the KCH workflow and present proposed informatics interventions for each.

The proposed informatics interventions address problems identified at KCH that would not generally be considered as being within the scope of a traditional laboratory information system implementation.²⁷ This application of informatics to the pre-analytical and post-analytical phases is novel and these stages are natural targets for informatics interventions because they are information-intensive and often where errors arise.¹⁷ Such an approach, which recognises that much of the total testing process takes place outside the laboratory, may also improve the perception of laboratory testing amongst clinicians. Whilst clinicians in many of these settings have a tendency to circumvent laboratory testing in favour of more superficial, less accurate diagnostic signals, they may be more likely to opt for laboratory testing if they perceive it to be quick and accurate.^7,8

We recognise that informatics interventions cannot solve all problems. Whilst our proposed informatics intervention does not prevent insufficient samples from being sent to the laboratory, it provides a formal mechanism for reporting and correcting those errors, potentially saving time in the laboratory testing process and improving the timely delivery of care. A variation of the proposed intervention could include decision support tools that remind the nurse preparing the order of the sample requirements (sample amount, container type, etc.).

The need for investment in laboratory infrastructure for disease prevention and control is recognised in the literature.²⁸ Resource shortages, such as laboratory technicians, microscopes and access to electricity, are commonly cited as limiting factors in improving laboratory services.²⁹ The informational challenges of laboratory testing in low-resource settings, whilst more challenging to identify than the resource limitations, will limit the impact of additional resources if unaddressed. This article synthesised an array of data about laboratory operations in a low-resource hospital setting, calling attention to these informational challenges. Next steps will include a larger-scale effort to document the pre-analytical and post-analytical phases at KCH and other low-resource hospital laboratories. In addition, whilst the proposed interventions are hypothetical at this point, we are in the early stages of modeling such systems.

The results presented represent the actual findings of the analyses described in the research method and design section, without alteration.