There is the short form of Jin Fan's Attention Network Test (ANT; Fan et al., 2002) which can be downloaded here: https://github.com/docksteaderluke/CRSD-ANT. It is a 10 minute computerized test of sustained attention. When completed, you can save a summary of the participant's performance (e.g., accuracy rate, time of attentional shift etc), as well as raw data. I'm using this application these days in a study of the effects of affect state induction on attention and motor performance.

Another, simpler, method to measure sustained attention is the d2 task, in which participants are to identify and cross with two lines the ds in a set of letters (p, b, q). The measure is the time needed to locate all the ds. Being a paper and pencil test, it could also be administrated in a large class by allotting limited time (e.g., 30 seconds) for the task; here, the measure will be the number of ds each participant crossed. Jenssen et al. (2012; JEP:General) compared the d2 task to other tasks.

Both tasks were validated.

Look at the IVA www.braintrain.com (takess 8 minutes on the computer) and tests both auditroy and visual attention with respectable norms

The is also the ANT but it has no norms and TOVA (takes too long)

Thank you very much for your suggestions!! I don't know much about this field and I really appreciate your input.

While I like the idea behind the d2, I would urge caution based upon your sample. It was standardized in Germany with a German speaking sample. There are preliminary norms for English speaking college students, but I have been hesitant to use it. Please consider the normative samples each test uses. If you are in South America, look for a measure that has been standardized for your sample. I am less familiar with measures that have been standardized in other countries.

Most of the measures I am familiar with for attention were not normed to be administered in a group context, so you are now potentially changing standardized administration instructions (something else to be cautious about). The Ruff 2&7 is a visual attention task that is relatively quick (5 minutes) but takes a bit more effort/time to score. The participant crosses out 2s and 7s as quickly as they can in blocks. They switch blocks every 15 seconds.

I would recommend that your primary considerations should be that the test has been normed for your sample (language, preferably country) and type of administration (individual vs in a group setting).

Thanks Holly for your suggestion. Now, my study is a pre/post design conceived to inspect the possible effects that  a given intervention might have on reading comprehension. Because of the nature of this intervention, I was (correctly, I think) suggested to include some attentional measure, but mainly for control purposes. Thus, I naively think that a non-normed  test (or a test normed for a different population) would do inasmuch as I could model it later considering it as a covariant, basically because attention is not the main focus of the study. I guess I should be careful if working with the measure as a grouping factor if based on the test norms, but I could, if needed,try and work with the students' own performance, for instance, based on percentiles....

I dare say you would be ok if using it as a covariate in your statistics. The other thing to be aware of is that many cognitive tests are affected by age, sex, race, and educational level. Standardization adjust for those differences. You can get away with using raw scores and including the variables of race, sex, and years of education as additional covariates to make up for it. It is essentially using your own sample demographics to "norm" it. 

Thank you all for your help.

maybe its work:

HAND BOOK OF Psychoeducational Assessment {Ability, Achievement,and Behavior in Children} by: jac j. W. Andrews