Small Area Estimation of Dependency: Final Report. Regression-Adjusted Synthetic Estimates

01/13/1989

Percentages generated with the regression models can be multiplied by corresponding population estimates for specific geographic areas of interest to generate estimates of the number of noninstitutionalized functionally dependent elderly in a given community. Population subgroups, of course, are defined by the explanatory variables included in the model.

As mentioned earlier, as intercensal data are not readily available for the elderly population in small age intervals by race and sex for small areas, we elected to use Medicare Enrollment data for the production of our estimates. Although Medicare data, given its level of detail and recency, are the best available data for our purposes, two adjustments had to be made to it prior to estimation.

First, only 95% of elderly Americans are enrolled in Medicare, thus requiring that we inflate the numbers to be reflective of the total elderly population. As the percent enrolled varies little across sex or family income groups, but does differ across race groups (Ries 1987) adjustments were made which accounted for the race difference. Specifically, the number of white elderly Medicare enrollees was inflated by 4.4%, and the number of nonwhite elderly enrollees was inflated by 13.5%.

Second, because Medicare Enrollment data includes both the noninstitutionalized and institutionalized elderly population, and rates produced with the combined data set (1984 NHIS-SOA and ARF) are applicable for the noninstitutionalized population only, an adjustment had to be made to the data prior to producing the synthetic estimates. The adjustment entailed subtracting the estimated number of institutionalized elderly from the total population in a community. Using the 1985 National Nursing Home Survey and the 1985 National Health Interview Survey, a logistic regression equation was produced to estimate rates of institutionalization among the elderly population at the national level. Candidate explanatory variables for inclusion in the model included those variables available on the merged data set for which corresponding population data existed. Given this constraint, age (in five year intervals from 65 to 85 and over), sex, race (white and nonwhite), and geographic region (northeast, north central, south, and west), as well as their pairwise interactions and transformations were available for use. Region was included in the model as the supply of nursing home beds, and thus rates of institutionalization, are known to vary geographically. The model found to best fit the data included age, age-squared, sex and an indicator variable reflecting whether or not the individual resided in the north central region of the country. Appendix B presents results of the logistic model. Estimates produced from this model were used to deflate the state and county population data to be representative of the noninstitutionalized elderly population.

By applying the rates of dependency generated by the log-linear regression model (which included race, sex, age, age-squared, and the percent of the elderly who reside in poverty) to the adjusted Medicare data, we produced estimates for each state, and the largest county in each state (Table 9 and Table 10).

TABLE 9: 1986 Dependent Noninstitutionalized Elderly Population by State: Regression-Adjusted Synthetic Estimates
State Elderly
  Populations  
Number Dependent Percent Dependent
Total ADL IADL   Total     ADL     IADL  
California 2,685,304   521,891     147,505     374,386   19.4 5.5 13.9
New York 2,169,180 521,526 162,389 359,137 24.0 7.5 16.6
Florida 1,880,487 426,861 133,050 293,811 22..7 7.1 15.6
Texas 1,475,817 409,903 139,822 270,081 27.38 9.5 18.3
Pennsylvania 1,634,088 376,538 115,873 260,665 23.0 7.1 16.0
Illinois 1,278,849 300,132 92,840 207,292 23.5 7.3 16.2
Ohio 1,239,037 284,862 87,738 197,124 23.0 7.1 15.9
Michigan 1,001,901 229,029 70,624 158,405 22.9 7.0 15.8
New Jersey 935,069 215,410 66,269 149,141 23.0 7.1 15.9
North Carolina 699,501 195,934 65,655 130,279 28.0 9.4 18.6
Missouri 640,658 180,495 61,902 118,593 28.2 9.7 18.5
Massachusetts   749,017 177,057 55,050 122,007 23.6 7.3 16.3
Georgia 570,835 163,071 54,646 108,425 28.6 9.6 19.0
Virginia 564,433 158,247 53,509 104,738 28.0 9.5 18.6
Tennessee 551,947 154,822 52,591 102,231 28.1 9.5 18.5
Indiana 614,558 141,704 43,800 97,904 23.1 7.1 15.9
Wisconsin 593,261 136,679 42,885 93,794 23.0 7.2 15.8
Alabama 470,932 136,220 46,280 89,940 28.9 9.8 19.1
Louisiana 426,880 125,130 42,983 82,147 29.3 10.1 19.2
Kentucky 418,922 115,895 39,612 76,283 27.7 9.5 18.2
Minnesota 487,274 114,774 36,478 78,296 23.6 7.5 16.1
Washington 492,373 111,303 34,826 76,477 22.6 7.1 15.5
Oklahoma 383,233 107,760 37,049 70,711 28.1 9.7 18.5
Maryland 439,074 103,331 31,767 71,564 23.5 7.2 16.3
South Carolina 339,007 95,562 31,845 63,717 28.2 9.4 18.8
Connecticut 403,889 92,857 28,771 64,086 23.0 7.1 15.9
Iowa 387,728 91,976 29,141 62,835 23.7 7.5 16.2
Arkansas 321,450 90,470 31,139 59,331 28.1 9.7 18.5
Mississippi 299,024 89,953 31,064 58,889 30.1 10.4 19.7
Arizona 379,578 82,697 25,501 57,196 21.8 6.7 15.1
Oregon 341,834 77,330 24,273 53,057 22.6 7.1 15.5
Kansas 302,189 72,273 22,868 49,405 23.9 7.6 16.3
West Virginia   240,253 65,472 22,317 43,155 27.3 9.3 18.0
Colorado 276,104 63,173 19,751 43,422 22.9 7.2 15.7
Nebraska 199,665 47,939 15,291 32,648 24.0 7.7 16.4
Maine 150,401 41,482 14,291 27,191 27.6 9.5 18.1
New Mexico 135,274 36,127 12,405 23,722 26.7 9.2 17.5
Rhode Island 135,224 31,443 9,707 21,736 23.3 7.2 16.1
Idaho 106,134 27,933 9,641 18,292 26.3 9.1 17.2
Utah 123,388 27,495 8,577 18,918 22.3 7.0 15.3
Hawaii 104,726 27,006 8,674 18,332 25.8 8.3 17.5
New Hampshire 114,532 26,237 8,155 18,082 22.9 7.1 15.8
South Dakota 93,402 26,148 9,186 16,962 28.0 9.8 18.2
District of Co   71,493 23,805 8,134 15,671 33.3 11.4 21.9
North Dakota 82,782 22,777 7,993 14,784 27.5 9.7 17.9
Montana 92,485 20,813 6,567 14,246 22.5 7.1 15.4
Nevada 94,468 19,249 5,845 13,404 20.4 6.2 14.2
Delaware 69,335 15,969 4,920 11,049 23.0 7.1 15.9
Vermont 61,306 14,244 4,470 9,774 23.2 7.3 15.9
Wyoming 39,492 8,928 2,814 6,114 22..6 7.1 15.5
Alaska 17,124 3,676 1,138 2,538 21.5 6.6 14.8

 

TABLE 10: 1986 Dependent Noninstitutionalized Elderly Population by County: Regression-Adjusted Synthetic Estimates
State Elderly
  Populations  
Number Dependent Percent Dependent
Total ADL IADL   Total     ADL     IADL  
Los Angeles, CA   770,182   184,822     57,601     127,221   24.0 7.5 16.5
Cook, IL 566,715 135,009 41,486 93,523 23.8 7.3 16.5
Philadelphia, PA 240,807 71,559 24,182 47,377 29.7 10.0 19.7
Dade, FL 231,893 66,997 23,245 43,752 28.9 10.0 18.9
Queens, NY 253,293 61,693 19,097 42,596 24.4 7.5 16.8
Wayne, MI 247,654 59,915 18,469 41,446 24.2 7.5 16.7
Cuyahoga, OH 203,551 47,719 14,604 33,115 23.4 7.2 16.3
Maricopa, AZ 213,577 46,918 14,419 32,499 22.0 6.8 15.2
Harris, TX 164,488 38,936 11,922 27,014 23.7 7.2 16.4
King, WA 148,252 34,343 10,715 23,628 23.2 7.2 15.9
Middlesex, MA 160,087 32,144 9,062 23,082 20.1 5.7 14.4
Balt. City, MD 103,576 31,189 10,547 20,642 30.1 10.2 19.9
Jefferson, AL 85,278 25,716 8,756 16,960 30.2 10.3 19.9
Hennepin, MN 106,475 25,701 8,071 17,630 24.1 7.6 16.6
New Haven, CT 106,378 24,590 7,607 16,983 23.1 7.2 16.0
Milwaukee, WI 122,379 24,335 6,812 17,523 19.9 5.6 14.3
Shelby, TN 78,581 23,878 8,138 15,740 30.4 10.4 20.0
St. Louis, MO 112,770 21,853 6,088 15,765 19.4 5.4 14.0
Bergen, NJ 114,975 21,774 6,080 15,694 18.9 5.3 13.6
Providence, RI 86,172 20,369 6,300 14,069 23.6 7.3 16.3
Honolulu, HI 76,008 19,656 6,263 13,393 25.9 8..2 17.6
Orleans, LA 61,464 19,335 6,630 12,705 31.5 10.8 20.7
Multnomah, OR 79,847 19,274 6,069 13,205 24.1 7.6 16.5
Jefferson, KY 79,520 19,164 5,889 13,275 24.1 7.4 16.7
Fulton, GA 60,972 18,991 6,416 12,575 31.1 10.5 20.6
Oklahoma, OK 65,010 18,351 6,231 12,120 28.2 9.6 18.6
Marion, IN 61,372 16,774 4,757 12,017 27.3 7.8 19.6
Denver, CO 62,887 15,395 4,838 10,557 24.5 7.7 16.8
Salt Lake, UT 51,619 11,630 3,613 8,017 22.5 7.0 15.5
Clark, NV 54,643 11,017 3,321 7,696 20.2 6.1 14.1
Pulaski, AR 35,744 10,363 3,520 6,843 29.0 9.8 19.1
New Castle, DE 43,351 10,037 3,083 6,954 23.2 7.1 16.0
Mecklenburg, NC   41,666 10,008 3,041 6,967 24.0 7.3 16.7
Douglas, NE 41,539 9,993 3,109 6,884 24.1 7.5 16.6
Greenville, SC 34,633 9,490 3,144 6,346 27.4 9.1 18.3
Sedgwick, KS 40,088 9,250 2,848 6,402 23.1 7.1 16.0
Bernalillo, NM 41,124 9,109 2,800 6,309 22.2 6.8 15.3
Hinds, MS 26,593 8,150 2,787 5,363 30.6 10.5 20.2
Polk, IA 33,580 7,964 2,464 5,500 23.7 7.3 16.4
Hillsborough, NH 31,881 7,336 2,259 5,077 23.0 7.1 15.9
Cumberland, ME 29,557 6,954 2,171 4,783 23.5 7.3 16.2
Kanawha, WV 30,412 6,925 2,121 4,804 22.8 7.0 15.8
Ada, OD 18,190 4,059 1,261 2,798 22.3 6.9 15.4
Minnehaha, SD 12,809 3,007 947 2,060 23.5 7.4 16.1
Henrico, VA 12,621 2,813 949 1,965 22.3 6.7 15.6
Yellowstone, MT 11,424 2,552 794 1,758 22.3 7.0 15.4
Chittenden, VT 9,685 2,276 704 1,572 23.5 7.3 16.2
Cass, ND 8,645 2,056 655 1,401 23.8 7.6 16.2
Laramie, WY 5,667 1,303 411 892 23.0 7.3 15.7
Anchorage, AK 5,786 1,168 346 822 20.2 6.0 14.2

These estimates are based upon three assumptions. First that the race, sex, age, and poverty-specific disability rates from the 1984 NHIS-SOA did not change between 1984 and 1986. Second, that the relationship between dependency and race, sex, age, and the percent of the elderly residing in poverty is the same for a small area as it is for national averages. And third, that race, sex, age, and the percent of the elderly residing in poverty are the only important predictors of functional dependency. Thus, the estimates will err to the extent that the relationship between dependency and race, sex, age, and poverty in a community have changed over time; to the extent that the relationships vary from national averages; and to the extent to which other known or unknown factors which are not in the model strongly influence functional dependency. The latter two reflect phenomena which could occur due to variations in the health of the local aged population from national norms. For example, estimates produced would likely underestimate the prevalence of functional dependency in a community where some disabling disease was highly prevalent, but overestimate the prevalence in a community such as Miami, where there is a large concentration of well elderly.

View full report

Preview
Download

"smareaes.pdf" (pdf, 2.82Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®