Encounter Data Validation

Paul Henfield
Managed Care
November 13, 2013

Encounter Data Validation

  • The importance of accurate and complete encounter data can not be overstated, in view of the multiple uses of the data by the DOH. The usefulness of the data has grown over the years, as has its complexity in terms of collection, storage, and reporting. Periodic and timely auditing of the data is essential.
  • Several approaches can be used to validate encounter data. Approaches include, but are not limited to:
    • Medical record validation
    • Creation of encounter data validation reports
    • Calculation of performance measures using encounter data

Medical Record Validation

  • Encounters are compared to documentation in the medical record
    • Considered the gold standard approach; helps to identify areas of under, over and mis-reporting.
    • A drawback to using this approach: discrepancies can be observed due to either between the provider and the health plan, or the health plan and the state warehouse, or both. In order to determine the source of the errors, further drilldown is necessary.
    • There is mounting concern that variation in coding exists, especially in view of automated coding software and the use of EMRs. The literature that exists relates primarily to coding errors at provider and facility levels. Discrepancies in both over and under-coding have been observed; data from various studies indicates that physician accuracy in coding averages somewhere around 55%.

Examples of medical record validation

  • Study #1 Primary Care Encounter Validation
    • This study evaluated diagnosis and procedure coding, and focused on mis- coding
    • Coders reviewed whether the data submitted to MEDS matched the information in the medical records
    • The encounter data sample was stratified into 3 categories; primary care (30 encounters), behavioral health (15 encounters), and cardiology (15 encounters). Sixty (60) encounters across 6 health plans; plus a small oversample to compensate for unavailable medical records . The final sample consisted of 345 encounters.
    • Results (diagnosis coding): 31.6% of the records had a complete diagnosis match, 66.1% of the records contained one or more errors (e.g. at least one diagnosis deleted by an IPRO reviewer, at least one added, at least one code changed, etc), 2.3% of the records had a match at the three digit level (close match).
    • Results (procedure coding): 60.4% of the records had a complete procedure code match, 39.6% of the records contained one or more errors (e.g. at least one procedure deleted by an IPRO reviewer, at least one added, at least one changed, etc)
  • Study 2: Primary care encounter validation
    • Identified under-reporting, over-reporting, and mis-reporting of encounter data
    • Compared diagnosis and procedure code data in enrollee medical records with corresponding encounter data for 60 randomly selected Medicaid managed care enrollees for each of 29 plans
    • Sample was stratified into 2 groups;
      • No Encounter Group: 25 plan members for whom there was no record of an encounter with a primary care provider in MEDS during the study period
      • Primary Care Group: 35 members who had up to 3 primary care encounters during the study period
    • Results /Findings
      • One fifth (20.3%) of the No Encounter group had at least one visit reported in a medical record but no encounter in MEDS. Note: the No Encounter group represented only 4.1% of the total Medicaid managed care population
      • 72.3% of Primary Diagnosis coding and 70.1% of Primary Procedure coding in MEDS is complete and accurate when compared to information contained in the medical records.
      • There was a high rate of underreporting of secondary diagnosis (65.5%) and additional diagnosis (85%) coding in MEDS.
  • Encounter Data Validation reports
    • Edit checking, volume monitoring and reconciliation reports to monitor the encounter data that is reported. Using the results of the reports, IPRO can identify problem areas, e.g. missing/under- reporting of vendor data, problems in interpreting the data dictionary, etc.

IPRO Managed Care Department - MMIS Encounters

Intake Report By Encounter Date∗ Record Count (Includes all claims lines)

. Month/Year Encounter Submitted to
Type Average # Encounters SEPT2013 AUG2013 JUL2013
Inpatient 183,247 206,473 147,109 171,137
Inpatient XOver 35,911 39,726 35,147 57,905
Outpatient 879,560 1,368,977 1,139,319 498,295
Outpatient XOver 175,742 245,039 213,975 119,639
Professional 1,630,023 2,462,157 1,721,324 1,011,489
Professional XOver 287,659 236,927 548,588 255,807
Long Term Care 333 1,650 166 112
Home Health Care 14,326 12,692 12,699 5,898
Total 3,206,800 4,573,641 3,818,327 2,120,282
Dental 187,637 249,862 199,271 118,075
Pharmacy 1,269,559 1,215,270 1,425,420 1,347,672
Total Encounters 4,663,996 6,038,773 5,443,018 3,586,029
Members 719,147 699,199 712,111 715,626

∗ Encounter date is the Month and year the encounters are being submitted to MMIS.
NOTE: Includes all encounters submitted from DSS to IPRO.
Includes paid, denied, adjusted and void encounters.

MMIS Encounter Validation Table
Encounter Detail File Dental Encounters

File Encounter Dates: September 1, 2013 to September 30, 2013

Record Count (Includes all encounter record lines): 249,862

Variable Name # Missing % Missing # Invalid Data % Invalid Data
Billing Provider Key 0 0.00% N/A N/A
Category of Service 0 0.00% N/A N/A
Claim Adj Reason 40,761 16.30% N/A N/A
Claim Adj Void 0 0.00% 0 0.00%
Claim Detail Status 0 0.00% 0 0.00%
First Date of Service 0 0.00% 57 0.00%
ICN Number 0 0.00% N/A N/A
Last Date of Service 0 0.00% 50 0.00%
Place of Service 0 0.00% N/A N/A
Performing Provider Key 249,862 100.00% N/A N/A
Procedure Code 1 0.00% 0 0.00%
Recipient County 1,255 0.50% N/A N/A
Recipient Medicaid ID 52 0.00% 0 0.00%
Recipient Ethnicity 52 0.00% N/A N/A
Recipient Race 52 0.00% N/A N/A
Referring Provider Key 249,862 100.00% N/A N/A
Submitter ID 0 0.00% 0 0.00%
Tooth Number 181,998 72.80% N/A N/A

NOTE: Includes all encounters submitted to IPRO.
Includes paid, denied, adjusted and void encounters

  • Calculation of Performance Measures
    • Using encounter data that the health plan reports, IPRO can write source code to calculate measures that the plan is required to report. Replicating this process will indicate whether the data warehouse is sufficiently robust to produce the same result as produced by the plan.

For more information

Tom LoGalbo
(516) 326-7767 ext. 349

Paul Henfield
(516) 326-7767 ext. 670

1979 Marcus Avenue
Lake Success, NY 11042-1002
20 Corporate Woods Boulevard
Albany, NY 12211-2370