Table 2.
Combined Acute Myocardial Infarction, Heart Failure, and Pneumonia Outcomes, Costs, and Value by Hospital Type and Patient Risk on Admission
| Outcome (% unless noted) | Cohort | AMI (N = 43,990 pairs) | HF (N = 84,985 pairs) | PNA (N = 74,947 pairs) | Combined patients (N = 203,922 pairs) | Combined quintiles of risk on admission (Q1 = lowest) | Trend P value* | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Q1 (N = 41,851 pairs) | Q2 (N = 39,207 pairs) | Q3 (N = 39,738 pairs) | Q4 (N = 40,547 pairs) | Q5 (N = 42,579 pairs) | |||||||
| % of pairs | 21.5 | 41.7 | 36.8 | 100.0 | 20.5 | 19.2 | 19.5 | 19.9 | 20.9 | ||
| 30-day mortality (%) | Teaching | 12.1 | 8.3 | 12.7 | 10.7 | 2.5 | 4.8 | 7.9 | 13.1 | 24.6 | 0.0001 |
| Non-teaching | 12.9 | 9.3 | 14.5 | 12.0 | 2.6 | 5.4 | 8.9 | 14.9 | 27.6 | ||
| Difference | − 0.9d | − 1.0d | − 1.8d | − 1.3d | − 0.1 | − 0.6c | − 0.9d | − 1.8d | − 3.0d | ||
| 30-day resource cost ($, m-est.) | Teaching | 23,875 | 19,067 | 16,762 | 19,249 | 16,464 | 18,378 | 19,041 | 20,051 | 22,265 | < 0.0001 |
| Non-teaching | 23,327 | 18,267 | 17,143 | 18,944 | 17,000 | 18,489 | 18,848 | 19,450 | 20,901 | ||
| Difference | 489d | 756d | − 382d | 273d | − 546d | − 140 | 182 | 557d | 1289d | ||
| Value estimate (95% CI) | 563 (138, 989) | 730 (465, 994) | − 208 (− 300, − 116) | 211 (119, 303) | N.C. (N.C.) | − 234 (− 680, 211) | 193 (− 102, 489) | 318 (143, 493) | 427 (309, 546) | ||
| 30-day total payments ($, m-est.) | Teaching | 29,930 | 21,017 | 19,314 | 22,304 | 21,075 | 22,125 | 21,863 | 22,490 | 23,913 | < 0.0001 |
| Non-teaching | 24,060 | 15,677 | 14,789 | 17,141 | 16,703 | 17,420 | 17,044 | 16,899 | 17,624 | ||
| Difference | 5795d | 5206d | 4435d | 5048d | 4277d | 4629d | 4715d | 5431d | 6120d | ||
| Value estimate (95% CI) | 6674 (3488, 9860) | 5022 (3729, 6315) | 2419 (1968, 2870) | 3904 (3336, 4472) | N.C. (N.C.) | 7755 (3818, 11,692) | 5010 (2929, 7091) | 3097 (2257, 3938) | 2030 (1634, 2425) | ||
| 30-day payments without Geo, IME, or DSH ($, m-est.) | Teaching | 23,784 | 16,984 | 15,503 | 17,902 | 16,855 | 17,786 | 17,587 | 18,045 | 19,197 | < 0.0001 |
| Non-teaching | 22,437 | 14,875 | 14,073 | 16,207 | 15,746 | 16,431 | 16,145 | 16,001 | 16,671 | ||
| Difference | 1332d | 2054d | 1408d | 1662d | 1086d | 1349d | 1420d | 1972d | 2440d | ||
| Value estimate (95% CI) | 1534 (760, 2308) | 1981 (1451, 2511) | 768 (612, 925) | 1286 (1080, 1492) | N.C. (N.C.) | 2259 (990, 3529) | 1509 (832, 2185) | 1125 (796, 1453) | 809 (641, 977) | ||
| ICU use (%) | Teaching | 46.8 | 13.8 | 11.1 | 19.9 | 16.6 | 19.6 | 19.8 | 20.7 | 22.7 | 0.0921 |
| Non-teaching | 48.7 | 16.1 | 14.0 | 22.4 | 18.6 | 22.3 | 22.5 | 23.0 | 25.3 | ||
| Difference | − 2.0d | − 2.3d | − 2.9d | − 2.5d | − 2.0d | − 2.7d | − 2.7d | − 2.3d | − 2.6d | ||
| 30-day readmission or death (%) | Teaching | 32.9 | 29.5 | 30.0 | 30.4 | 20.9 | 25.0 | 28.2 | 33.3 | 44.1 | 0.0003 |
| Non-teaching | 36.2 | 30.9 | 31.6 | 32.3 | 23.6 | 27.1 | 29.9 | 34.8 | 45.5 | ||
| Difference | − 3.3d | − 1.4d | − 1.6d | − 1.9d | − 2.8d | − 2.1d | − 1.7d | − 1.5d | − 1.4d | ||
| Length of stay (days, m-est.) | Teaching | 5.3 | 5.5 | 5.7 | 5.5 | 4.5 | 5.0 | 5.4 | 5.9 | 6.8 | 0.6208 |
| Non-teaching | 4.6 | 4.9 | 5.6 | 5.1 | 4.1 | 4.6 | 4.9 | 5.4 | 6.3 | ||
| Difference | 0.7d | 0.6d | 0.1d | 0.4d | 0.4d | 0.4d | 0.4d | 0.4d | 0.5d | ||
Value expressed as increased cost or payment at teaching hospitals per 1% reduction in mortality. Individual tables for AMI, HF, and PNA are provided in the Appendix. P value key is as follows: aP < 0.05, bP < 0.01, cP < 0.001, and dP < 0.0001. Non-significant differences have no letter. P values for binary overall outcomes and within strata were calculated using McNemar’s test. P values and estimates for continuous overall outcomes and within strata were calculated using M-statistics. Value is well defined if there is a significant difference in mortality between groups. Otherwise in this table, values are reported as N.C. (not calculated). For all value estimates and confidence intervals reported, all 1000 jackknives concur in finding a statistically significant difference in mortality
*Trends for binary outcomes were tested using the Mantel test for trend, while trends for continuous outcomes were tested using robust regression