COMPARISON OF THREE ALGORITHMS FOR MANAGEMENT OF CHRONIC COUGH IN PEOPLE LIVING WITH HIV IN A LOW-INCOME COUNTRY

Purpose: To compare how three algorithmic guidelines about the same topic perform in terms of accuracy, harm and complexity.

Methods: Data were collected prospectively at the University Hospital of Kigali (CHUK) from March to June 2002, in a total of 201 HIV positive hospitalised patients with chronic cough. We virtually and retrospectively injected each patient into three algorithms. The first was locally tailored by clinicians from CHUK, the second and third were drawn form publications by Médecins sans Frontières (MSF) and the World Health Organisation (WHO). Semantic analysis techniques known as Clinical Algorithm Nosology were used to compare these algorithms in terms of complexity and similarity.(1) Further we studied over-all diagnostic accuracy taking as gold standard the final clinical diagnosis at the end of hospitalisation. We computed the hypothetical delay to action in the algorithm compared with the shortest possible way to arrive at the diagnosis. Finally we calculated the expected disutility (harm) of each algorithm, defined as the weight (regret) of false positives and false negatives estimated by the clinicians, compared to natural death.

Results: The main aetiologies were tuberculosis (21%) and pneumocystosis (19%). In respectively 88%, 81% and 64% cases a correct diagnosis would have been made by the CHUK, MSF and WHO algorithm. Sensitivity was respectively 96%, 88% and 70% and specificity 99,6 %, 99% and 99%. Mean delay to action was 1.65 days for the WHO, 1.86 for CHUK and 3.46 for the MSF algorithm. The CHUK algorithm was the most complex with a CASA score of 80, followed by MSF with a score of 73 and WHO with a score of 40.

On average, the weight of omission error was estimated much higher than the weight of commission error (8 versus 4, 5). Total disutility was 92 for CHUK, 240.4 for MSF and 488.3 for the WHO algorithm.

Conclusions: This study confirms our hypothesis that increasing complexity improves accuracy and diminishes expected harm, but on the other hand the degree of complexity becomes so high that it makes the algorithm difficult to use in practice at the third care level.

1) Steven D. Pearson and al., the clinical Algorithm Nosology: a method for comparing Algorithmic guidelines (Medical Decision Making 1992;12:123-131)