1 Supplementary Material Table S1. Mortality

0 downloads 0 Views 124KB Size Report
0.242-0.287. aAll odds ratios ... mortality and then divided into 25 approximately equal groups based on their relative ranking. The 45- ... 5,202 (0.4). 23,592 (0.4).
Supplementary Material Table S1. Mortality Prediction in the Calibration Subset of the FFS Medicare Dataset

Age 25-34 Age 35-44 Age 45-54 Age 55-64 Age 65-74

Odds Ratioa Reference value 1.178 1.778 2.788 4.321 5.472

Age 75-84

11.773

Age 85-94 Age ≥95 Male (vs female) Raceb: Black

31.199 86.159

Variable Age < 25

Raceb: White Racea: Unknown Racea: Other

1.239 Reference value 1.112

Confidence Intervals

Variable -- 5 HCCs

0.734-1.890 1.132-2.793 1.790-4.343 2.780-6.718 3.521-8.503

6 HCCs 7 HCCs 8 HCCs 9 HCCs 10 HCCs

7.574-18.301 11-12 HCCs 20.067.48.505 Chronic lung disease 55.342-134.136 Congestive heart failure 1.217-1.261 Cardiovascular disease -- Cancer Rheumatoid arthritis, severe 1.077-1.147 hematological disorder, or muscular dystrophy

0.641

Chronic condition

0.797

0.758-0.838 Acute condition

0 HCCc

0.086

0.075-0.098

1 HCC 2 HCCs

0.106 0.153

3 HCCs

0.204

4 HCCs

0.264

Medicare eligibility due to disability 0.097-0.117 Newly eligible for Medicare 0.140-0.167 Dual eligibilityc in prior year Dual eligibility in current 0.187-0.222 year 0.242-0.287

Odds Ratioa

Confidence Intervals

0.324

0.298-0.352

0.394 0.499 0.586 0.697 0.851 Reference value 1.292 1.206

0.362-0.429 0.457-0.545 0.535-0.643 0.632-0.769 0.764-0.948

0.94

0.920-0.961

1.37

1.338-1.402

0.884

0.855-0.914

1.071

0.984-1.165

1.204

1.174-1.234

1.389

1.348-1.431

1.168 1.309

1.098-1.243 1.223-1.400

1.178

1.102-1.259

-1.263-1.321 1.178-1.235

a

All odds ratios were statistically significant. P values were < 0.0001, except in the case of 10 HCCs (P = 0.0034). bRace designations were derived from the Social Security Administrations (SSA). Since the SSA only classified race/ethnicity as White, Black, Other, or Unknown prior to 1980 and since most current Medicare members were born prior to 1980, this study conformed to those categories. cTable 1 in the article maps the medical condition variables to HCCs. cEligiblity for Medicaid in addition to Medicare

1

Figure S1. Model Performance

This figure illustrates model accuracy for individuals of varying mortality risk. Individuals in the testing dataset of the FFS population (n = 1,437,880 person-years) were first ranked according to predicted mortality and then divided into 25 approximately equal groups based on their relative ranking. The 45degree line represents perfect correspondence between actual and predicted mortality, with observations falling above the 45 degree line representing model underestimation. The plots show that the model estimated mortality most accurately for individuals with predicted mortality up to 4.5%. For individuals at greater mortality risk (with predicted rates between 4.5% and 17.0%) the model underestimated mortality risk, while the model overestimated mortality risk for the group of individuals with the very highest predicted mortality rate (> 17.0%). The authors consider these findings to show good model fit, with deviation from the 45-degree line occurring only for individuals at the higher end of predicted mortality risk.

2

Table S2. Population Characteristics Characteristics

FFS Medicare (calibration subset)

Medicare Advantage

1,437,937

5,477,976a

71.5

71.1

631,489 (43.9) 805,137 (56.0)

2,457,861 (44.9) 3,020,115 (55.1)

1.55

1.58

134,427 (9.3) 1,219,178 (84.8) 79,130 (5.5) 5,202 (0.4)

666,270 (12.2) 4,547,516 (83.0) 240,598 (4.4) 23,592 (0.4)

290,300 (20.2) 1,147,637 (79.8)

804,411 (14.7) 4,673,565 (85.3)

264,360 (18.4) 344,695 (24.0) 577,156 (40.1) 251,726 (17.5)

123,257 (2.3) 1,297,761 (23.7) 3,516,702 (64.2) 540,256 (9.9)

n/a

1,378,606 (25.2)

n/a n/a

937,839 (17.1) 766,136 (14.0)

n/a

2,395,395 (43.7)

n (person-years) Age, mean (years) Gender Male, n (%) Female, n (%) HCC Count Raceb, n (%) Black White Other Unknown Income Status, n (%) Dual eligibilityc in prior year No dual eligibility in prior year Geographic Region, n (%) Northeast Midwest South West Plan Type), n (%) Local Preferred Provider Organization (PPO) Regional (PPO) Private Fee for Service HMO (Health Maintenance Organization) a

After a relatively small number of exclusions due to recognizably missing data. bRace designations were derived from the Social Security Administrations (SSA). Since the SSA only classified race/ethnicity as White, Black, Other, or Unknown prior to 1980 and since most current Medicare members were born prior to 1980, this study conformed to those categories. cEligiblity for Medicaid in addition to Medicare.

3