Hi @mosabbirfardin ,   Your requirements need 3 steps.   In the first step anyone character in '()-,.' is replaced by a SPACE. The second step will look for 2-character word to replace it by a SPACE. The third step looks for a word ending in a NUMBER to replace it by a SPACE.   data _null_; txt = 'ATN1 (atrophin 1) (eg, dentatorubral-pallidoluysian atrophy) gene analysis, evaluation to detect abnormal (eg, expanded) alleles'; txt = translate(txt,' ','()-,.'); wc = countw(txt); do i = 1 to wc; word = scan(txt, i); if length(word) = 2 then txt = transtrn(txt, strip(word), strip(' ')); else if anydigit(word) then txt = transtrn(txt, strip(word), strip(' ')); end; txt = upcase(compbl(txt)); put txt =; run; ... View more

Thanks everyone for your guidance. I kind of figured it out by myself. As you are all aware of the hierarchical structure of HL7 files, it seems one NTE segments or a group of NTE segments will always follow an OBX segment but not all OBX segments are followed by NTE segment(s). In my file the first NTE comes at 18th line and 17th line is an OBX segment.   So what I did was: 1. Get observation numbers of each NTE segments and OBX segments by using ID=_N_ and create 2 separate datasets (OBX and NTE)     NTE dataset:     Partial OBX dataset:     2. Subtract 1 from the ID column for NTE segments. I saw a series of incremental ID from which the lowest number would represent the observation numbers (OBX_ID) of the corresponding OBX segments.         3.I assigned NTE_SK to each NTE  records using the following code so that I can group them together later.       data want; set have; if set_id_NTE=1 then NTE_SK+1; run;     dataset with NTE_SK:         4.Then I transposed the NTE dataset and flattened it. This gave me 2 types of records. One holds the whole comment for the OBX segment and the other holds a bunch of numbers from which the lowest number would represent the observation numbers of the corresponding OBX segments.     proc transpose data=WORK.NTE_WITH_ID out=WORK.TRANSPOSED_NTE prefix=x; by NTE_SK; var COMMENT previous_OBX_ID; run; data WORK.FLATTENED_NTE; set WORK.TRANSPOSED_NTE; LENGTH FULL_COMMENT $3000; FULL_COMMENT = CATX(' ', OF X:); keep NTE_SK _NAME_ FULL_COMMENT previous_OBX_ID; DROP X:; run;     Transposed NTE dataset:     Flattened NTE dataset:       5. Then I split the dataset into 2 datasets using the type of records. One would have just the comments and other would have just the incremental numbers       /* define which libname.member table, and by which column */ %let TABLE=WORK.FLATTENED_NTE; %let COLUMN=_NAME_; proc sql NOPRINT; /* build a mini program for each value */ /* create a table with valid chars from data value */ select distinct cat("DATA NTE_",compress(&COLUMN.,,'kad'), "; set &TABLE.(where=(&COLUMN.='", &COLUMN., "')); run;") length=500 into :allsteps separated by ';' from &TABLE.; quit; /* macro that includes the program we just generated */ %macro runSteps; &allsteps.; %mend; /* and...run the macro when ready */ %runSteps;     Just comments:   Just numbers:     6. I used a SCAN function on the FULL_COMMENT in the numbers' dataset to get the first number which would be the OBX_ID.     7. Then I joined this dataset with comment dataset using NTE_SK and now I have an OBX_ID with every comments to join OBX table with.   Sorry for this huge post. I know I have done a lot of post processing to get where I am now. I would really appreciate if someone shares a faster and more efficient approach to do this. Take care!             ... View more