Why all observations are "on support" on propensity score matching?

[Heebyung Koh]

Hi I'm doing propensity score matching for reducing selection bias in observational study.

There was significant difference between PPI exposure group and PPI non-exposure group in comorbidity.

So I performed propensity score matching, setting the independent variable as a "PPI_exposure" and the other covariates as Age, sex, HTN, DM, Dyslipidemia.

I wanted to PSM 1:1 but there was no off supp, and all observation were on support.

How can I fix this problem?

 psmatch2 PPI_3W Age Sex HTN DM Dyslip , out(CCI) noreplacement

Probit regression                               Number of obs     =      4,557
                                                LR chi2(5)        =      81.84
                                                Prob > chi2       =     0.0000
Log likelihood =   -1434.54                     Pseudo R2         =     0.0277

------------------------------------------------------------------------------
      PPI_3W |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         Age |   .0183717   .0023795     7.72   0.000      .013708    .0230355
         Sex |   .0379768   .0532516     0.71   0.476    -.0663945    .1423481
         HTN |   .0402486   .0552818     0.73   0.467    -.0681018    .1485989
          DM |   .0723273   .0599909     1.21   0.228    -.0452527    .1899072
      Dyslip |   .0814611   .0548364     1.49   0.137    -.0260163    .1889385
       _cons |  -2.601975   .1722587   -15.11   0.000    -2.939596   -2.264354
------------------------------------------------------------------------------
----------------------------------------------------------------------------------------
        Variable     Sample |    Treated     Controls   Difference         S.E.   T-stat
----------------------------+-----------------------------------------------------------
             CCI  Unmatched |  1.1986755   .884015595   .314659902   .046942343     6.70
                        ATT |  1.1986755   1.02207506   .176600442   .067535405     2.61
----------------------------+-----------------------------------------------------------
Note: S.E. does not take into account that the propensity score is estimated.

           | psmatch2:
 psmatch2: |   Common
 Treatment |  support
assignment | On suppor |     Total
-----------+-----------+----------
 Untreated |     4,104 |     4,104 
   Treated |       453 |       453 
-----------+-----------+----------
     Total |     4,557 |     4,557 


.

[Aarushi Dhingra]

Not an expert here, but perhaps you can try other propensity score algorithms such as a nearest neighbor, kernel, stratification ...