APPENDIX D

NATIONAL OSTEOPOROSIS FOUNDATION
OSTEOPOROSIS ATTRIBUTION PROBABILITIES

METHODOLOGY

It has been very difficult to assess the extent of the burden of osteoporosis among different segments of the population because the diagnosis of osteoporosis is rarely listed in conjunction with its resulting fractures. In a study of 1992 hospital discharge records for hip fractures in the United States, for example, only 6% included an associated diagnosis of osteoporosis, although it is widely recognized that the majority of such fractures among the elderly occur because of low bone mass. To address this problem, the National Osteoporosis Foundation convened a panel of six expert clinicians with extensive experience in the treatment of patients with osteoporosis to assess the contribution of this disease to four types of fractures (13). Probabilities were calculated for hip, vertebral, forearm, and all other site fractures among three age groups (45-64; 65-84; 85+), three racial groups (white, black, all others), and both genders. Seventy-two categories resulted from these groupings.

The Delphi method was used to attain a group judgment on fracture probabilities for all 72 different age, gender, and race categories. The process used by the expert panel comprised three stages, or rounds. Round I was conducted by mail prior to the meeting of the panel and involved an initial estimation of osteoporosis attribution probabilities for each category by each panel member, recorded on an Osteoporosis Attribution Probability Response Form. On a separate form, each participant outlined his or her key assumptions used in determining the probabilities. Before Round II, the panelists were provided the modal probabilities and confidence intervals for each category as well as a summary of the assumptions involved. They were also provided with available published data on osteoporosis incidence by age, gender, and ethnicity. At this point, the panel convened and discussed the preliminary estimates. Each panelist then developed a second set of attribution probabilities, again recorded on the Osteoporosis Attribution Probability Form. Modal probabilities and ranges were again determined for each category and distributed to the panel members. Round III involved further discussion on the areas of continued disagreement. The process was completed when each participant assigned a final probability to each of the 72 categories. Median attribution probabilities were calculated and then ranked by the panelists according to a numeric validity scale reflecting the degree of certainty associated with the final probability.

The final attribution probabilities are presented by race for each gender, age group, and fracture site in the following three tables. In addition, the initial assumptions considered by the expert panelists are listed. Because West Virginia has such a small minority population, and because data on race are not included on all hospital discharge records, this study utilized only the probabilities calculated for the white population.

TABLES

Table 1

FINAL OSTEOPOROSIS ATTRIBUTION PROBABILITIES BY FRACTURE TYPE, GENDER, AND AGE
White Population

*Probability can range from 0.00 (no attribution) to 1.00 (100% attribution).

**Validity scores can range from 1 (+5% error) to 4 (more than +20% error).

Table 2

FINAL OSTEOPOROSIS ATTRIBUTION PROBABILITIES BY FRACTURE TYPE, GENDER, AND AGE
Black Population

*Probability can range from 0.00 (no attribution) to 1.00 (100% attribution).

**Validity scores can range from 1 (+5% error) to 4 (more than +20% error).

Table 3

FINAL OSTEOPOROSIS ATTRIBUTION PROBABILITIES BY FRACTURE TYPE, GENDER, AND AGE
Other Race*** Population

*Probability can range from 0.00 (no attribution to 1.00 (100% attribution).
**Validity scores can range from 1 (+5% error) to 4 (more than +20% error).
***Includes Asian/Pacific Islanders, American Indians, and other races.