Multiple encounters and corresponding diagnoses

sd_ncsu — Mon, 16 May 2022 17:36:43 GMT

Hi. I have data over time that contains the day someone initiated care (StartDate, DATE9.), how long they received care during that encounter (days_in_care, Num type), and the corresponding diagnosis for that visit (dx_1, Num type and the value is either 1 or 0 since only 1 diagnosis is tracked given the patient population). I've created two files ('same' contains everyone with the same diagnosis over time, 'not_same' contains everyone with diagnoses that change over time). I am interested in exploring patterns of how the diagnosis variable changes. Beyond the variables above, the ID is a character variable, and the data are arrayed in a long file--one row per encounter (the most encounters in the 'not_same' file for any individual is 40 and there are ~90 people in the dataset). I am struggling to implement code that:

effectively displays the various visual patterns of 0s and 1s over time in 'not_same', including how many different pattern 'types' there are. For example, some data patterns will look like 0,0,0,0,1,1,1,1,1 and others will look like 1,1,1,0,1,0, etc. Any advice on ways to display multiple patterns of data?
then counts and sums the number of visits and days until a diagnosis changes from 0 to 1 (i.e., want_lag_days and want_lag_encounters).

Sample data:

ID (character)	StartDate (DATE9.)	dx_1 (numeric, 0/1)	days_in_care (numeric, ranges from 0, released in under 24 hours to 352)	want_lag_days	want_lag_encounters
ff52fc4315ddb	11Sep2016	0	15	16	2
ff52fc4315ddb	18Nov2016	0	1	16	2
ff52fc4315ddb	19Dec2016	1	217	16	2
ff52fc4315ddb	16Aug2017	1	7	16	2
593f1e8e2615	16Jul2015	0	174	179	3
593f1e8e2615	9Jan2018	0	1	179	3
593f1e8e2615	9Nov2018	0	4	179	3
593f1e8e2615	8Apr2019	1	7	179	3

Thank you for any suggestions!

Re: Multiple encounters and corresponding diagnoses

JOL — Tue, 07 Jun 2022 20:34:17 GMT

Have your tried transposing the data into a wide structured table as starting point? Based on sample data provided:

/* Sort data by id first, so by can be used in PROC Transpose */

PROC SORT DATA=WORK.BOOK1 OUT=SORT_BOOK1;
BY ID;
RUN;

/* Transpose data so date form the new columns, added a prefix of D_ to create valid column names */

PROC TRANSPOSE DATA=SORT_BOOK1
OUT=WORK.TRNS_BOOK1(DROP=_NAME_)
PREFIX=D_;
BY ID;
ID StartDate;
VAR dx_1;
RUN;

/* Use the Data Step to convert numeric columns to character and concatenate*/

DATA NEW;
SET WORK.TRNS_BOOK1;
ARRAY D{8} _NUMERIC_;
ARRAY CDATE{8} $;
DO I = 1 TO 8;
CDATE(I) = PUT(D(I),3.);

IF CDATE(I)=. THEN
CDATE(I)='';
END;
PATTERN = CATX(',', OF CDATE1-CDATE8);
KEEP CDATE1-CDATE8 PATTERN;
RUN;

/* Create list report */

PROC PRINT DATA=NEW;
RUN;

Result

topic Re: Multiple encounters and corresponding diagnoses in SAS Programming

Multiple encounters and corresponding diagnoses

Re: Multiple encounters and corresponding diagnoses