Medicare Inpatient File, LDS or RIF: Counting Distinct Hospitalizations


Hospitals may submit multiple claims for some hospitalizations. This article provides guidance for counting distinct inpatient hospitalizations and for sequencing claims for each distinct hospitalization found in the Inpatient Research Identifiable File (RIF) and Limited Data Set (LDS) files. This guidance is not applicable to the MedPAR because each record in that file is already a distinct hospitalization.

Current Version Date:

The base segment of the Medicare Inpatient claims files (RIF and LDS) contains a record for each claim submitted by a hospital (facility) during a patient’s hospitalization. More than 99% of the inpatient stays generate only one facility claim[1],[2].  However, for those stays that generate multiple facility claims, ResDAC recommends researchers follow CMS’s method to group distinct hospitalizations[3], which is used to create the MedPAR file. They group claims by each unique combination of three elements:

1)  Patient identifier

  • "Encrypted CCW Beneficiary ID" (BENE_ID) in Identifiable files


  • “DESY SORT KEY” (DSYSRTKY) in Limited Data Set (LDS) files

2)  Admission date

  • “Claim Admission Date” (CLM_ADMSN_DT or ADMSN_DT) in RIF and 2009 and forward LDS files [prior to 2009, LDS files contained QQYY only, so this was not possible]

3)  Provider Number (aka CMS Certification Number [CCN])

  • “Provider Number” (PRVDR_NUM or PROVIDER) in both Identifiable and LDS files

RIF claims within a hospitalization may be sequenced by Claim From Date (CLM_FROM_DT or FROM_DT) and Claim Through Date (CLM_THRU_DT or THRU_DT), and within 2009 and forward LDS claims by Claim Through Date. In the last claim submitted, Claim Through Date will equal Discharge Date (NCH_BENE_DSCHRG_DT or DSCHRGDT).  

[1]Based on STAY_FINL_ACTN_CLM_CNT variable for inpatient stays in the 2011-2012 5% RIF MedPAR file.

[2]Multiple claims may be submitted by physicians, but those will be found in the Carrier claims data, not the IP claims data.

[3]MEDPAR2000_DD_20141022.pdf from Data Administration page of    Accessed January 22, 2016