Symptom screen frequently misses HIV in children and delays treatment, Kenyan study shows

This article is more than 12 years old. Click here for more recent articles on this topic

Clinical diagnosis of HIV infection in infants performed poorly and resulted in delayed diagnosis when compared to virologic testing,  a prospective cohort study conducted in Kenya has shown. The findings emphasise the importance of access to virologic testing for early infant HIV diagnosis in resource-limited settings.

At one month of age, 70 children were already infected with HIV (81% of all infections during the study period) but only one was correctly diagnosed by using the World Health Organization’s (WHO) 2008 Integrated Management of Childhood Illness (IMCI) clinical algorithm, Lara C Diener and colleagues report in this secondary analysis of a four-year prospective cohort study published in the advance online edition of AIDS.

The study evaluated the performance of the WHO IMCI algorithm for HIV diagnosis in infants, in comparison to virologic and serologic tests for HIV diagnosis, in 444 HIV-exposed infants born to mothers with HIV receiving care in Nairobi between 1999 and 2003.



When using a diagnostic test, the probability that a person who does have a medical condition will receive the correct test result (i.e. positive). 

positive predictive value

When using a diagnostic test, the percentage of those testing positive who are correctly diagnosed. This will vary according the prevalence in the local population.


When using a diagnostic test, the probability that a person without a medical condition will receive the correct test result (i.e. negative).

negative predictive value

When using a diagnostic test, the percentage of those testing negative who really don’t have the medical condition. This will vary according to the prevalence in the local population.

prospective study

A type of longitudinal study in which people join the study and information is then collected on them for several weeks, months or years. 

Using virologic testing, 86 infants (19%) were diagnosed with HIV infection during the first year of life.

The overall sensitivity (true positive), specificity (true negative), positive predictive value (PPV – the proportion of positive results that are true positives) and negative predictive value (NPV – the proportion of those with negative results who are without disease) of IMCI compared to virologic testing (the gold standard) were 58, 87, 52 and 90%, respectively.

In other words, only 58% (50/86) were eventually diagnosed by IMCI. While sensitivity increased with age, using IMCI delayed diagnosis by an estimated median time of 5.9 months, p<0.0001.

These findings highlight the limited effectiveness of IMCI as a tool for diagnosis in early infancy. Supporting the development of rapid virologic point-of-care diagnostics for this population, where early diagnosis and prompt treatment are critical to survival, the authors stress.

Without treatment, 50% of HIV-infected infants will die before their second birthday. Current WHO guidelines recommend starting ART at the time of diagnosis in all children under two years of age.  

IMCI, a set of standardised clinical algorithms, is used in settings with limited access to laboratory diagnostics, to diagnose and manage common illnesses in children. The use of a clinical algorithm to identify paediatric HIV infection was first validated in 2003. Updated by WHO in 2006 and 2008, identification is based on the presence of two or more of seven clinical criteria.

In 2010, although the evidence was of low quality. WHO recommended the use of clinical algorithms and serologic testing where virologic testing was unavailable, so acknowledging the potential of IMCI as an important screening tool.

Healthcare workers in resource-poor settings may use IMCI to help make decisions regarding further HIV testing, starting ART, cotrimoxazole prophylaxis and infant feeding.

A wide range of sensitivity (9 to 89%) and specificity (42 to 99%) estimates for HIV clinical algorithms have been reported. Factors affecting this disparity include age distribution, health status, HIV prevalence, health-worker training and the diagnostic test chosen.

Given these findings, together with the pending WHO 2012 review of evidence for the re-evaluation of clinical algorithms among highly vulnerable infant populations, longitudinal studies looking at IMCI performance are timely, note the authors.

In this secondary analysis of a study of HIV-exposed Kenyan infants that looked at HIV transmission and infant immunity, the authors chose to describe the effect of age on IMCI performance during the first year of life and the delay in diagnosis when comparing using IMCI with virologic testing.

Between July 1999 and November 2002, 410 pregnant HIV-positive women were recruited from Nairobi City Council maternal-child health clinics.

In total, 444 infants were included in the secondary analysis. Eligibility included those infants who had survived to at least one month of age and had had, at a minimum, one virologic test and one clinical assessment in the first year of life.

From 1999 to 2003, monthly clinical assessments and quarterly virologic testing were performed. In the first year of life, mothers and their infants had a median of eleven study visits (IQR 8-12).  Archived clinical data were used to compare IMCI sensitivity, specificity, PPV and NPV with virologic testing.

Among the 19% (86) children diagnosed during the first year of life using virologic testing, 28 (33%) were diagnosed at their neonatal visit, 42 (49%) by month one, seven (8%) at month three and the remaining 10% after month three.

In all, 35% (30/86) of HIV-infected infants died between one and twelve months of age; median age at death was 5.4 months (IQR=4.3-6.9).

Forty-two per cent (36) of infected infants were never diagnosed using IMCI; 14 (39%) died and 5 (14%) were lost to follow-up before they were a year old. Thirteen of the 14 deaths were among infants who were diagnosed before the age of one month.

IMCI sensitivity was found to increase with age, consistent with other studies, reflecting increased prevalence of opportunistic infections among HIV-infected infants as they become immunosuppressed. Yet sensitivity remained low throughout early infancy (one month to ten months), ranging from 1.4% to 35%.

IMCI performance was poorest at months one and two, when over 80% of all infections had already happened. This is a critical time for identifying infected infants and getting them onto treatment promptly, with the best prospects for reducing disease and death.

PPV was also lowest at one month of age (14%) and may present a problem where IMCI is used as a diagnostic tool to guide the start of ART, rather than a screening tool. In populations where the maternal HIV status is unknown, PPV could be even lower, leading to children being misdiagnosed and started on ART inappropriately.

The authors believe this to be the first time delay in HIV diagnosis using IMCI has been described. They note that children who met IMCI criteria in their study would have been categorised as “HIV exposed/possible HIV”. The recommendation for appropriate follow-up, the authors note, and sole reliance on these criteria, can result in considerable delays before a “health worker is prompted to recommend confirmatory testing and/or referral for early start of ART”. 

The authors note that otitis and parotitis were not associated with HIV infection in this study and suggest their removal from the algorithm be considered.

Strengths of the study include a prospective longitudinal cohort with a high power to determine time trends and an unbiased assessment of IMCI clinical criteria. 

Sensitivity estimates were probably reduced because of the high death rate among HIV-infected infants who were lost to follow-up or died before presenting with IMCI-specific symptoms at clinic visits.

The authors note they used a “snapshot” approach in which they assessed clinical criteria at a specific visit. They believe this is closer to a real-life situation where most children present to medical officers without access to previous clinical records.

HIV prevalence, healthcare worker experience, length of breastfeeding/HIV exposure and infant health status need to be considered before generalising these findings.

The authors conclude that their findings “underscore the critical need for expanded access to virologic-based rapid diagnostic technologies for infant HIV infection in resource-poor settings. Despite its low sensitivity in infants, IMCI may still prove useful in identifying older children with undiagnosed HIV infection and children who get infected through late breast milk transmission.”


Diener LC et al. Performance of integrated management of childhood illness (IMCI) algorithm for diagnosis of HIV-1 infection among Kenyan infants. Advance online edition AIDS 26, doi: 10.1097/QAD.0b013e3283578bb8, 2012.