Table 2. Primary outcome and statistical analysis for health system cluster-randomized trial cohort.
| All (N = 64) |
Cancer (N = 16) |
Cardiovascular disease (N = 18) |
Pulmonary disease (N = 30) |
|
|---|---|---|---|---|
| Primary outcome (s) reported, n (%)& | 60 (93.8) | 15 (93.8) | 18 (100) | 27 (90.0) |
| Patient-reported outcome (PRO), n | 28 (46.7) | 10 (66.7) | 3 (16.7) | 15 (55.6) |
| Clinical outcome, n | 15 (25.0) | 3 (20.0) | 5 (27.8) | 7 (25.9) |
| Process outcome, n | 10 (16.7) | 2 (13.3) | 6 (33.3) | 2 (7.4) |
| Economic outcome, n | 1 (1.7) | 0 | 0 | 1 (3.3) |
| Behavioral outcome, n | 1 (1.7) | 0 | 0 | 1 (3.3) |
| Mortality/survival, n | 8 (13.3) | 0 | 7 (38.9) | 1 (3.3) |
| Statistical test, n (%)& | 61 (95.3) | 14 (87.5) | 17 (94.4) | 30 (100) |
| Tests Accounting for Clustering, n | 50 (81.9) | 13 (92.9) | 14 (82.4) | 23 (76.7) |
| Mixed models | 29 (58.0) | 7 (53.8) | 5 (35.7) | 17 (73.9) |
| Generalized estimating equations | 13 (26.0) | 6 (46.2) | 4 (28.6) | 3 (13.0) |
| Cox model with clustering | 1 (2.0) | 0 | 1 (7.1) | 0 |
| Other models with clustering-CACE, differences 2 time points, ANOVA/ANCOVA, proportional odds model | 7 (14.0) | 0 | 5 (35.7) | 2 (8.7) |
| Tests not Accounting for Clustering, n | 12 (19.6) | 2 (14.3) | 3 (17.6) | 7 (23.3) |
| Comparison of means | 5 (41.7) | 0 | 2 (66.7) | 3 (42.9) |
| Generalized linear model | 5 (41.7) | 1 (50.0) | 1 (33.3) | 3 (42.9) |
| Logistic regression, no clustering | 1 (8.3) | 0 | 0 | 1 (14.3) |
| Cox model, no clustering | 1 (8.3) | 1 (50.0) | 0 | 0 |
| Treatment effect present (primary outcome), n (%)^ | 24 (37.5) | 5 (31.3) | 5 (27.8) | 14 (46.7) |
& Some studies reported more than one primary outcome and more than one statistical test.
*Mixed models include generalized linear mixed model, mixed effect model, multilevel linear regression, linear mixed model, multilevel linear model, logistic regression with clustering, and binomial regression with clustering.
†Demographic variables include primarily individual level demographics- age, sex, race, ethnicity, socioeconomic status (education, income, occupation), or insurance status. Most models included >1 individual-level demographic covariate.
^Treatment effect present refers to statistical significance.