Skip to main content
. 2013 May 28;63(611):e370–e377. doi: 10.3399/bjgp13X668159

Table 4.

Predictive value of HIV clinical indicator conditions in the THIN modela

Indicator condition Odds ratio Standard error z P-value 95% CI
Statistically significant conditions
  Bacterial pneumonia 47.7 52.0 3.54 <0.001 5.6 404.2
  Oral candidiasis 29.4 21.8 4.57 <0.001 6.9 125.5
  Herpes zoster 25.4 14.2 5.76 <0.001 8.4 76.1
  Weight loss 13.4 6.7 5.15 <0.001 5.0 36.0
  Non-Hodgkin’s lymphoma 12.6 15.0 2.13 0.033 1.2 129.8
  Lymphadenopathy 11.3 5.3 5.15 <0.001 4.5 28.3
  Sexually transmitted infection 10.8 7.6 3.38 0.001 2.7 43.2
  Pyrexia of unknown origin 7.2 3.5 4.05 <0.001 2.8 18.7
  Blood dyscrasia 5.7 4.0 2.44 0.015 1.4 22.9
  Diarrhoea — one consultation only 3.7 0.9 5.48 <0.001 2.3 6.0
  Diarrhoea — two consultations 4.4 2.3 2.81 0.005 1.6 12.1

Conditions that were not statistically significantb
  Parotitis 8.6 11.0 1.68 0.093 0.7 106.1
  Psoriasis — one consultation only 2.6 1.5 1.69 0.091 0.9 7.9
  Psoriasis — two consultations 3.0 2.5 1.38 0.168 0.6 14.8
a

Also included in conditional logistic regression: Townsend deprivation index and proportion of ward that was Asian and/or Asian British (quintiles).

b

These conditions remained in the stepwise regression model, as their significance was below the 0.2 cut-off threshold for inclusion. Other HIV indicator conditions not shown in the table were excluded during the stepwise process.