Helping professions · Research companion

Eleven intake questions predicted, six months ahead, which Head Start staff would report elevated stress or low intention to remain.

A randomized controlled longitudinal trial of 4,054 Head Start employees (95% women, mean age 43), split into training, development, and test partitions (n = 2,429, 811, and 814). An 11-predictor logistic regression on baseline stress, burnout, anxiety, depressive symptoms, job satisfaction, hope, reappraisal beliefs, and demographics beat the base rate for both elevated stress and low intention to remain.

At a glance

.72AUC, elevated stress95% CI [.67, .77] on the test partition

.70AUC, low intention to remain95% CI [.63, .77]

0.92Calibration slopeElevated stress; 95% CI [0.70, 1.19]

4,054Employees in a randomized trialTraining, development, and test partitions of 2,429, 811, and 814

How well does it rank people

Four-panel figure for the final 11-predictor model on the test partition. Panel a: ROC curve for elevated six-month stress, AUROC .72. Panel b: ROC curve for low six-month intention to remain, AUROC .70. Panel c: precision-recall for stress, AUPRC .52, above a base-rate line near .28. Panel d: precision-recall for intention to remain, AUPRC .26, above a base-rate line near .14. Grey bands are pointwise bootstrap 95% intervals. — **Figure 4.** Test-partition ROC (a, b) and precision–recall (c, d) curves. Stress: n = 462, 129 events; intention to remain: n = 430, 62 events. The further a ROC curve bows above the diagonal, the better the equation ranks staff who later struggled above those who did not.