Some test-kits cannot be trusted.

While adequate training reduced uncertainties associated with the test kits, many of our interviewees articulated uncertainties with using some of the test-kits and their associated scripts. Our interviewees preferred some test-kits over others. It was common for interviewees to express which test-kits they preferred or detested. While there was a diversity of views on which tests were ‘good’ or ‘bad’, test 1A, the most commonly used test, was repeatedly mentioned as a trustworthy and efficacious test-kit.

“Test 1B and test 2B would give you a false negative but test 1A continues to be used because we see its efficacy […] Aah test 2B was the worst, it gave out false results.” Female nurse, age 45, urban clinic

While test 1A was by far the most favoured test by a majority of respondents, tests 1B, 2B, and 3 were regarded as most problematic, to the extent that some testers felt anxious or even tried to avoid using them. Both tests 1B and 3 were known to be difficult to interpret if the prescribed time window for reading the results was missed. The concern most often mentioned about using test 3 was a lack of training. A HIV counsellor from a rural mission hospital expressed a lack of confidence in using test 3, the tie-breaker on the 2017 version of the national HIV testing algorithm:

“Test 3 works as a tiebreaker but personally it becomes a challenge for me to use that test…' 'kits like test 3 we haven’t really fully received training on so that may become a problem because you will end up saying if only I could test someone and they test negative so that I don't go through the struggle of using test 3.” Female HIV counsellor, age 40, rural mission hospital

Her lack of training in using test 3, and the fact that she does not use tiebreaker test on a regular basis, may increase the likelihood of her making mistakes when using the test. But it may also amplify her lack of confidence in using the test, which, as she alludes to, could translate into testing bias in order to avoid using test 3 as a tiebreaker. During our observations in one hospital we noted that test 3 was only administered by laboratory staff. Our interviews suggest that tiebreakers were very rare, and we never observed a tiebreaker being performed. While this lack of engagement with test 3 as a tiebreaker suggest non-compliance with the 2017 testing algorithm, the tiebreaker is no longer recommended and has been removed from the testing algorithm. Nonetheless, the message here is that testers’ uncertainty with running new or rarer tests (in the context of limited training) can make them avoid using these tests altogether.

One of the difficulties HIV testers face in interpreting results pertains to the readability of lines. As explained by a 42-year-old nurse working on antenatal care in a rural hospital, “there are times when the line on the test kits is faint.”From our own observations, we were surprised by how faint the control lines generally were, even on 1A test-kits, which were widely regards as producing ‘good’ lines. Different explanations were offered to account for faint or unclear lines on the test-kit. One HIV counsellor explained that they once had a batch of 2B kits that produced random and unintelligible lines. Another counsellor explained that a host of different issues may affect how clear the lines will appear:

“I think it is the test because one could have followed the standard operating procedure on the kit to the dot and still get the faint line. However, at times it is about the client and the anti-bodies in them and then in some cases when one uses too small a blood sample, the test takes forever to read. So in such cases if one doesn’t pay close attention to the time it takes to read the result, one might report a negative result when in actual fact it will be a positive.” Female HIV counsellor, age 43, rural hospital

The same counsellor went on to say that besides faint lines, some tests produce multiple lines shortly after the prescribed time window for reading the results:

“I don't know about the quality of some of the kits because soon after 20 minutes multiple lines will start appearing. We end up seeing that there is a kit which is more accurate than other kits.” Female HIV counsellor, age 43, rural hospital

Again, some test-kits were highlighted as being particularly prone to producing faint lines, while others, including test 1A, were said always to produce clear lines. This emphasis on the test-kits (rather than the tester) as producing unclear results may, like discussed above, lead HIV testers to note a pattern in certain test-kits producing unclear lines. The accumulation of such experiences and their insecurities with using certain test-kits heightens their distrust, which in turn may affect their inclination to use other test-kits or shape how they interpret test results.

Uncertainties associated with new testing algorithms.

Over the past three years, rapid HIV testers in Zimbabwe have endured three changes to the national testing algorithm, and the 2019 WHO guidelines are likely to lead to further iterations. Changes to the national algorithm are spurred by evidence-based recommendations and the pricing and availability of new test-kits. New test-kits are therefore often introduced with changes in the testing algorithm, which contribute to considerable uncertainty and deviation from the script.

“We had an algorithm, which we have stopped using, we now have a new algorithm, but you find that person is following the old algorithm.” Male laboratory technician, age 58, district hospital

“The algorithms change every now and then. Even the kits themselves change every now and then. I explained earlier that test 2A was once our tiebreaker but currently 2A is our second test, so they change every now and then. This makes it challenging.” Female nurse, age 45, rural health clinic

The occasional change in algorithm meant that testers were not always up-to-date or felt confident in applying the new testing algorithm. For example, at one rural hospital we observed that the same test (1A) was used both for the first screening test and as the confirmatory test when a client tested HIV positive. This practice of solely relying on test 1A was not an isolated incident; it was confirmed in interviews and conversations with testers in other hospitals too. At a District General Hospital, for example, a primary counsellor mentioned that some of their satellite clinics only used 1A test kits because they fully trusted this test to produce the right results. At the same time, we observed that here, in contradiction to the national testing algorithm, that test 2A was used as the first screening test. While deviance from the latest algorithm could sometimes be explained by their confidence in old testing practices and uncertainties associated with new testing practices, occasional stock-outs also led testers to tinker with the algorithm. A primary counsellor from a rural facility, for example, explained:

“We haven't had a second test kit since March. We have test 1A, but we are using it as a second confirmatory test because there is a shortage even at the District level. […] If a client tests positive, we won't have a second kit. It's been a month.… The second test is out of stock.” Female counsellor, age 43, rural hospital

The primary counsellor was fully aware of the fact that it would not be possible to re-test the client once the confirmatory test was available again, as they would by then have been initiated on antiretroviral treatment. These pervasive uncertainties and resultant malpractices highlight inadequate levels of training and support.

Inadequate training.

The preceding findings have already alluded to the importance of training when new test-kits or algorithms are introduced. The lab technician participating in our study stated that “the challenge is to train testers and keep them updated.” While all of our respondents had received some level of training in administering HIV test-kits, for some it was more than 10 years ago. New test-kits and algorithms have been introduced since their initial training, and there was agreement amongst our interviewees that refreshers courses were needed. While the Ministry of Health in Zimbabwe continually offers refresher courses, they were described as becoming increasingly rare due to funding constraints. When refreshers courses were organised, it was often for a selected few who, in return, would be expected to pass on their learning to other testers in their health facility. A number of our interviewees drew a direct link between inadequate training and risks of misdiagnosis, as exemplified by one HIV counsellor:

“If you did not go for training then it is difficult to administer the rapid tests. I remember at one point there was a certain individual that did not go for training and they wanted to test someone, and he did it incorrectly. If one goes for training it is very easy because you will be taught the correct procedure of administering these tests and all that is involved to obtain the correct results.” Female HIV counsellor, age 42, rural mission hospital

It is clear from our findings that pervasive uncertainty from HIV testers’ trust in the test-kits, frequent changes in testing algorithm and test-kits, and limited training in using new test-kits and algorithms, introduces biases and a tinkering of recommended testing practices, which are potential sources of misdiagnosis.

Work pressures and distractions during times of testing.

Our interviews repeatedly mentioned the high volumes of people presenting at their health facilities for HIV testing. We also noted that much time was spent on paperwork and report writing, especially for non-governmental organisations (NGOs) that had funded testing activities. The workload fluctuated, with early and late weekdays being particularly busy. Some testers performed up to 50 HIV tests on a busy morning. Given a theoretical maximum of three tests per tester per hour, as each test typically takes 15–20 minutes to read and must be accompanied by pre-and post-test counselling, this is illustrative of a heavy workload–also expressed through accounts of long queues, which put pressure on HIV testers to hurry up, skip lunch, and develop strategies for testing multiple clients at once. In Gutu district for instance, we observed how counsellors routinely tested several people at once, lining the tests up on the table, thus increasing the risk of mixing up results. As a HIV counsellor explains, the burgeoning workload may result in less than perfect HIV testing:

“Yes the workload may become too much since there will be many patients and you do not have time to do everything perfectly. Imagine, the patients will be standing outside in a long queue, you have limited time and the only tester, and the different test kits revealing a different set of results.” Male HIV counsellor, age 50, rural mission hospital

This counsellor also alludes to a dilemma mentioned by a number of participants, namely, the challenge of dealing with the follow-up required from a first positive test result. Subsequent tests would need to be carried out to confirm or disconfirm the first test result, and if positive, the client would require counselling and referral–all of which would cause further delays for the people waiting. This may contribute to a potential bias against testing someone HIV positive. There were also numerous accounts and observations of how HIV testers were distracted or called for urgent matters, disrupting and confusing the testing process. This ranged from nurses, who in the middle of testing and counselling, were called to attend emergency cases, to counsellors being interrupted by phone calls, text messages, or people bursting into the testing room with questions or handing them files to check.

Reading test results too quickly or too late.

Difficult work conditions, and the heavy workload in the particular, may, as explained by one nurse, “cause you to read the results before the actual time has lapsed.” A male HIV counsellor echoed this, linking ‘acting quickly’ with work pressures and failure to follow SOPs:

“A person may be quick, perhaps because of work pressures. I’m just saying what can happen if a person does not follow procedure. They may draw too much blood, when only a certain amount of blood is needed. They may dilute the blood more than necessary or add too little buffer. They may read the results when the time hasn’t lapsed or read them much later, long after the 15 minutes you have to wait.” Male HIV counsellor, age 33, rural hospital

One consequence of reading test results too quickly or too late is misdiagnosis, as explained by a lab technician: “I have seen others read results after a short time, before the line had reached the control line. They read it within some minutes or seconds. What happens then is misdiagnosis.” We made similar observations, noting that in a number of health facilities, testers found it difficult, due to either workloads or disruptions, to read the results in the prescribed time window. Although our interviewees stressed the importance of reading test results after 15 minutes, our observations noted quite lax attitudes to time keeping, even in high-performing facilities. We did not see HIV testers set an alarm to make sure they would keep track of all the different tests they were running at the same time, and sometimes observed testers reading and recording results immediately after administering the test.

Failure to record and document test results accurately.

While many HIV testers felt confident with the procedures for coding and linking clients with test strips and record books and claimed that misclassifications were rare, they acceded it could happen, and “if it happens then it will probably be due to a lapse of concentration or something like that” (Male nurse, age 36, district hospital). Failure to record and document test results accurately was often attributed to ‘being tired’ or the pressures the testers work under. One HIV counsellor explains how misclassification mistakes are more likely to occur when they line up test-kits on a table and bring in groups of people for testing:

“If you are administering tests to a number of people […] there will be a lot of test strips. If you do not label the test kits […] you might get confused when issuing the results.” Female HIV counsellor, age 43, rural hospital

We observed that often only the client’s initials were recorded on the test scripts, which carries a high risk of mixing up results when testing multiple people at once. We also observed occasions when blood was spilled on the table, heightening the potential for contamination of test strips.

A nurse attributed the actual workload associated with the registers as a potential source of misclassification.

“'How easy is it to record results…?' 'As an individual it's not easy […] we will be four, with one person testing, and the other recording in the HTS book and another one recording in the Syphilis book and the other one recording in the patients' cards. […] The registers are too many for a nurse […] you end up having a burnout. It's very difficult like if a person tests positive aaah then you're in big trouble […] then aaaah there's a lot of work.” Female nurse, age 45, urban clinic

This sharing of the recording work between a team of several people was confirmed by our observations.

Stock-outs.

In line with a recent HIV service evaluation in Zimbabwe, which found most health facilities to report at least one test-kit stock-out in the prior year [15], most of our participants spoke about instances of test-kit stock-outs. Earlier we presented the stock-out experiences of a 43-year-old HIV counsellor. Reflecting on supply chain deficiencies, she explained how they had run out of second confirmatory tests for an extended period of time, and how the shortage was district-wide, making it impossible for them to borrow test-kits from neighbouring health facilities. In response to the stock-outs for confirmatory tests, she went on to explain that “the sister in charge… gave us the go ahead to administer the tests [outside of the algorithm] and said that it is better for the client to be initiated on ART.” By using test 1A as both a first and confirmatory test, the stock-out forced them to deviate from the HIV testing algorithm, which may lead to misdiagnoses. The nurse recognises “there is no quality in it, but we don't have any option.”. However, the approach of another hospital to deal with stock-outs was not to tinker with the HIV testing strategy, but simply to send clients home, fully aware of what this might mean for the clients:

“The problem I mentioned earlier on, I think it is important for kits to always be readily available so that there are no times when a client gets turned back due to lack of kits. When people get turned back they go back to their homes and get sick and are vulnerable to opportunistic infections like TB. Some cases the CD4 count would lower so much and the person dies whereas this would not have happened had the kits been available.” Female nurse, age 42, rural hospital

During our observations we noticed this resource-depletion in other ways too, with gloves that kept tearing, the lack of air-conditioning in drug storage rooms, and a general chaotic storage of medicines and equipment.

Tensions between counsellors and nurses.

In Zimbabwe, the relationship between primary counsellors and nurses is complex. Whilst we observed numerous working relationships, we also observed some tensions between them, with nurses occasionally refusing to help counsellors with HIV testing even if the demand was high. The tensions arise from pay differentials and recent task shifting. To meet the demand for HIV testing in Zimbabwe, and with the introduction of point-of-care HIV testing, the task of HIV testing shifted to counsellors, moving them closer to one aspect of the clinical work of nurses. This change in role and responsibility was highly motivating for our participating counsellors, and often accompanied with lucrative employment contracts with NGOs. HIV testers, even within one health facility, were therefore paid very different salaries for the same work, depending on who their employer was. For many registered and state-employed nurses, who are paid less than NGO-employed counsellors despite their higher educational qualifications, this was a source of much discontent and jealousy, affecting their working relationship.

Difficult clients.

We invited our interviewees to reflect on challenges they encountered during the testing process, and sometimes this led to a discussion on clients with “anger issues” who were “intimidating”. Men in particular were described as intimidating, resulting in a fear of issuing a positive test result to certain groups of men.

“What I like the least is when I administer a test to someone and then he tests positive. As soon as I tell him that his result is positive, he just stands up, opens the door and leaves the room. He goes back home and I won’t know if he has any suicidal thoughts or if he is blaming someone for it. So, you will be in suspense […] He won’t be violent as such but you will see that he is angry.” Female HIV counsellor, age 36, mission hospital

Men were also described as clients who wanted the HIV testing to be quickly done and over with, rushing the testers. Although none of our interviewees conceded that the fear of issuing a positive testing result to men, or being rushed by men, affected the quality and authenticity of their HIV testing, it is nonetheless likely to introduce testing biases. HIV testers may for instance be inclined to side with a negative test result if the result is ambiguous.

Similar biases may be introduced during couples testing. A significant number of our interviewees spoke about the challenges of issuing discordant test results to couples, being fully aware that this may lead to break-ups or worse.

“If we test a couple and one person tests positive and the other one tests negative especially in favour of the husband you will observe that the relationship has been affected. The partner will be in fear of getting dumped just because they have tested positive, yet their husband hasn’t tested positive so those are some of the challenges that are happening to us, that some marriages are breaking down.” Female HIV counsellor, age 40, mission hospital

A more generic group of “difficult clients” included anyone who was either not ready to get tested (e.g., expectant mothers, for whom testing is part of antenatal care) or ready to accept their sero-conversion. Men, discordant couples and ‘difficult clients’ required more time and counselling, exacerbating the workload. While none of our interviewees spoke about feeling pressured to interpret results incorrectly, our findings did note social interactions and circumstances that are highly challenging for HIV testers, and which may contribute to testing biases, particularly in busy and resource-depleted health facilities.