If your value of 99 means that the data is actually missing, not appropriate, not answered or anything close to those then it is almost certain that you do not want to include them in any sum.
If the 1, 2, 3, 4 and 5 are not actually interval measurements but code values such as on an agreement scale or possibly categories (very good, good, neutral, poor, very poor) then summing values is very questionable in general.
I think you are looking for something similar to:
data want;
set have;
array p (*) promis_pa_scale1 promis_pa_scale3 - promis_pa_scale5;
/* set the 99 to missing*/
do i=1 to dim(p);
if p[i]=99 then p[i]=.;
end;
ptotal = sum(promis_pa_scale1, promis_pa_scale3, promis_pa_scale4, promis_pa_scale5);
/* ugly, long and robust but all I am willing to do with PICTURES of values*/
if ptotal=4 then tscore = 32.3;
else if ptotal = 5 then tscore = 37.3;
else if ptotal = 6 then tscore = 39.5;
/* follow the pattern until*/
else if ptotal =20 then tscore = 68.4;
run;
If you have not seen Array code before the array definition creates a reference to a group of variables of the same type. The array above is named P. If you have a variable named P there will be a conflict as array names cannot duplicate variables in the data set. You reference which value using arrayname(indexnumber). The above array has 4 elements and P[3] would reference the third variable in the list or promis_pa_scale4. The hyphen between names is a sequential list builder.
The DIM function returns the number of elements defined in an array so that you can iterate over only the variables present. So the Do loop removes the 99 values and sets them to missing. Since your Tscore values do not include any values larger than 20 then 99 must be right out.
There a multiple statistic functions such as Sum, Mean, Max, Min, Range, STD, Median and others that will allow you to get such a statistic for variables on a single record. Note that missing values are handled properly in such functions. If one of the values of 4 is missing then 3 are used to sum and 3 would the denominator for Mean calculations.
There are multiple ways to do a look up in SAS. Common ways include Informats created with Proc Format, joining data sets on specified values, If/then/else logic (shown), Select / when (really a shorter way of if/then/else); Update/modify/merge sometimes, Hash code.
