Collisional excitation of neutral hydrogen atoms can significantly increase the intensity of Balmer lines with respect to pure recombination. If this effect is not taken into account, the abundance analysis of these objects returns biased values, and the bias may be significant if accuracies better than a few percent are required. The most affected objects are young, metal-poor H II regions, due to their comparatively large temperatures. Until now, estimates of collisional enhancement have been based on tailored modeling of individual H II regions. In this contribution, I describe an ongoing effort to develop a general calibration, suitable for application to large samples of objects. Emphasis is placed on the uncertainties affecting the resulting predictions.