|
 |
|
|||||||
|
|||||||||
Electronic Letters to:
|
|
eLetters: Electronic letters published:
-
![[Read eLetter]](/icons/misc/sectionDOWN.gif)
A combination of conceptual filters and the capture-recapture method in evaluating underreported occ
- John H. Lange, Ronald E. LaPorte, Thomas J. Songer , Yue-Fang Chang (27 May 2003)
|
|
|||
|
John H. Lange, Environmental and Occupational Consultant Envirosafe Training and Consultants, Inc., Ronald E. LaPorte, Thomas J. Songer , Yue-Fang Chang
Send letter to journal:
john.pam.lange{at}worldnet.att.net John H. Lange, et al.
|
Azaroff et al.1, raised the issue as to how much undercounting of occupational injuries actually occur. They proposed a filtering system for “catching” those that fall into this undercount. Undercounting has always been a problem in counting studies in that one never knows how many they really missed. The purpose of this filtering system is to collect and count workers that were not identified on the original reporting. This concept is an excellent idea but in-itself will not provide an accurate count of all those that are reportable within a specific system, facility or type of injury. So the filters primarily become different lists of those being counted. However, this is an excellent way of generating lists that can be used in the capture-recapture method (CRM) of ascertaining the countable number of events2. Azaroff et al.1, does make reference to one study3 in their paper that employed the CRM. Review of the literature reveals that the CRM is not commonly used in injury epidemiology, although its potential is enormous, however, its use is growing. The CRM has been effectively used to count wildlife populations for decades4. This method employs multiple lists to identify the cases and then uses the overlap information among the lists to estimate the total number of cases5,6. The various filters that are illustrated in Figure 1 of Azaroff et al.1 paper can be potential sources for the capture- recapture application to estimate the size of work-related injuries. Each of the filters will have a different efficiency at capturing one segment of the total capturable population of work related injuries. The use of more than two listings will require employment of a CRM log-linear model. The model with three listings is illustrated in Figure 1. Each circle would represent a separate captured segment of the total population that would be obtained from a filter. For example, one filter (Source 1) would represent workers who lost more than 5 days. As indicated, some of these workers would also be represented on a second filter (list) (Source 3) as population A and B. Employment of a simple model, as shown in the Figure, the total ascertainable work related injury population could then be estimated using this contingency table. The number of events estimated by the CRM can be compared to OSHA logs and workers’ compensation records. This will allow comparison for the accuracy of these records and may provide clues to the best filters in capturing injured workers under different events. By combining the proposed filtering for occupational injuries and employment of the CRM there will be a mechanism of potentially locating the best reporting systems for ascertaining underreported injuries. CRM provides an estimate of the number of cases that could have been caught, and to be caught each person needs to have experienced a work related injury. Employment of this methodology has been shown to be effective in previous injury studies that did not use filters per se7,8. This method will not capture those that were never detected by a filter. It does, however, reduce the historical problem of counting in work related injuries, that of ascertainment, and thus produces an ascertainment corrected estimate. The CRM estimate, although not perfect, is most certainly much closer to the truth for methods of counting than for methods that do not control for ascertainment. Many investigators view missing cases of injuries or disease as part of the error measurement, a mistake, sloppiness of the study or just a “sin”. There needs to be a better understanding of the concept of missing cases9 in that they are an “ERROR” and to recognize that when counting populations there will inherently be missed cases. Thus, the “sin” is not missing cases, but instead is not using an appropriate statistical methodology for determining the degree of ascertainment and adjusting for the ascertaiment corrected incidence rates. The CRM is not limited to injury studies or epidemiology, but can be applied effectively in other fields as well10, so why not apply an old method to new problems. References 1. Azaroff LS, Levenstein C, Wegman DH. 2003. Occupational injury and illness surveillance: conceptual filters explain underreporting. Am J Public Health. 2002; 92: 1421-9. 2. McCarty DJ, Tull ES, Moy CS, Kwoh CK, LaPorte RE. Ascertainment corrected rates: applications of the capture-recapture methods. Int J Epidemiol. 1993; 22: 559-65. 3. Morse T, Dillon C, Warren N, Hall C, Hovey D. Capture-recapture estimation of unreported work-related musculoskeletal disorders in Connecticut. Am J Ind Med 2001; 39: 636-42. 4. LaPorte RE, McCarty DJ, Tull ES, Tajima N. 1992. County birds, bees and NCDs. Lancet 339: 494-5. 5. International Working Group for Disease Monitoring and Forecasting. Capture-recapture and multiple-record systems estimation, I. History and Theoretical development. Am J Epidemiology 1995;142-1047-58. 6. International Working Group for Disease Monitoring and Forecasting. Capture-recapture and multiple-record systems estimation,II. Application to human diseases. Am J Epidemiology 1995;142:1059-68. 7. Chiu W-T, Dearwater SR, McCarty DJ, Songer TJ, LaPorte RE. 1993. Establishment of accurate incidence rates for head and spinal cord injuries in developing and developed countries: a capture-recapture approach. J Trama. 35: 206-11. 8. Surkin J, Smith M, Penman A, Currier M, Harkey HL, Chang YF. Spinal cord injury incidence in Mississippi: A capture-recapture approach. J Trama 1998;45:502-504. 9. LaPorte RE. Assessing the human condition: capture-recapture techniques – allows accurate count of those difficult to reach populations. Lancet 1993; 308: 5-6. 10. Lange JH, LaPorte RE, Chang Y-F Exposure to lead and an old way of counting. Environ Health Perspectives (Letter) 2003; (accepted). Total size of the actual number of cases (unknown) This unknown is cases not Identified by any of the lists Source 3 Present Absent Source 2 Source 2 Present Absent Present Absent Present A B E F Source 1 Absent C D G ? Figure 1. Three-source combinations for estimating the number of actual cases |
|||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |