Pulling data closest to a date

[hanahch]

Hi,

I have a dataset that looks like the below:

id   enc_type  enc_date    mi _date    daysdiff bp_sys bp_dias  ht      wt   bmi

3669 CL        6/8/2011    6/8/2011    0        126    60
3669 CL        6/15/2011   6/8/2011    7
3669 PR        6/17/2011   6/8/2011    9        136    58
3669 PR        6/23/2011   6/8/2011    15       112    58       175.26  195  28.78 -> want this
3695 IM        5/14/2014   5/14/2014    0       125    80
3695 IM        5/21/2014   5/14/2014    7       125    70       160.02            -> want this
3744 PU        1/8/2018    12/14/2017   25      114    60       165.1   174       -> want this
3744 IM        1/8/2018    12/14/2017   25
572 PU         6/8/2011    6/8/2011     0              86       195     208   32.7
572 CL         4/25/2013   3/26/2013    30      124    86       195     207   31.4 -> want this
574 PR         4/3/2013    6/9/2013     63      140    88       172.72  224.5 34.1
575 PR         4/8/2013    3/26/2013    13      158    80       213           32.3 -> want this

I'd like to pull the most complete data on BP, Wt, Ht, and BMI from encounter date that is closest within +/30 from MI_date.  Daysdiff column is the duration in days between the two dates.  If an encounter closest to MI_date has incomplete data then it needs to search for the next record till a record with complete data is found.  

Your help is greatly appreciated! 

[andreas_lds · Accepted solution]

Please post the data in usable form, have look at How to convert datasets to data steps if you don't know how to post data. Are "encounters" identified by "id" alone?

Since you already have the "daysdiff", you could create another variable "non_missing_values", the function N can help to fill it.

[PeterClemmensen]

I don't know how your desired result looks like, so I assume you want as many observations as in your input data with the desired ht, wt, bmi for each obs.

data have;
input id enc_type $ (enc_date mi_date)(:mmddyy10.) daysdiff 30-32 bp_sys 33-35 bp_dias 37-38 ht 40-45 wt 47-52 bmi 54-58;
format enc_date mi_date mmddyy10.;
datalines;
3669 CL 6/8/2011  6/8/2011   0  126 60                     
3669 CL 6/15/2011 6/8/2011   7                             
3669 PR 6/17/2011 6/8/2011   9  136 58                     
3669 PR 6/23/2011 6/8/2011   15 112 58 175.26 195    28.78 
3695 IM 5/14/2014 5/14/2014  0  125 80                     
3695 IM 5/21/2014 5/14/2014  7  125 70        160.02       
3744 PU 1/8/2018  12/14/2017 25 114 60 165.1  174          
3744 IM 1/8/2018  12/14/2017 25                            
572  PU 6/8/2011  6/8/2011   0      86 195    208    32.7  
572  CL 4/25/2013 3/26/2013  30 124 86 195    207    31.4  
574  PR 4/3/2013  6/9/2013   63 140 88 172.72 224.5  34.1  
575  PR 4/8/2013  3/26/2013  13 158 80 213           32.3  
;

data want;
   do _N_ = 1 by 1 until (last.id);
      set have;
      by id notsorted;

      if abs(enc_date-mi_date) <= 30 then do;
         if n(ht, wt, bmi) > n then do;
            n    = n(ht, wt, bmi);
            _ht  = ht;
            _wt  = wt;
            _bmi = bmi;
         end;
      end;
   end;

   ht  = _ht;
   wt  = _wt;
   bmi = _bmi;

   do _N_ = 1 to _N_;
      set have;
      output;
   end;
run;

[hanahch]

The program works great!  Thank so much! 

[hanahch]

Sorry one more thing - How would you change in the program so that it outputs only the encounters that have the most data of ht, wt, and BMI?

Thanks!