ORIGINAL ARTICLE |
|
Year : 2001 | Volume
: 18
| Issue : 2 | Page : 105-114 |
|
Latent class analysis - A handy tool to evaluate diagnostic tests in the absence of a “Gold Standard“
Hemant R Kulkarni1, Meherbano M Kamal2
1 Department of Preventive and Social Medicine, Govt. Medical College, Nagpur, India; International Clinical Epidemiology Network (INCLEN, Inc.), Philadelphia, USA 2 Department of Pathology; Clinical Epidemiology Unit, Govt. Medical College, Nagpur, India
Correspondence Address:
Hemant R Kulkarni B/1 /1, Ajni Doctors' Colony, Opp. Ajni Railway Station, Ajni, Nagpur - 440 003
 Source of Support: None, Conflict of Interest: None  | Check |

|
|
In many clinical situations it is unrealistic and, indeed, impossible to have a “gold standard“ measurement. Consequently, it is not possible to characterise diagnostic tests. Various strategies can be undertaken to handle this problem. One of the popular methods is to use latent class analysis. This paper describes the principles and logic of latent class analysis. The paper demonstrates the use of latent class modelling on real data. The paper also discusses some limitations of latent class analysis. We recommend that if the assumptions and requirements of latent class modelling are met, this procedure should be used for evaluating the performance of diagnostic tests when three or more diagnostic tests have been simultaneously conducted on all study subjects and when a gold standard measurement does not exist
|
|
|
|
[PDF]* |
|
 |
|