Reporting of Sepsis Cases for Performance Measurement Versus for Reimbursement in New York State

Abstract

Objectives: Under "Rory's Regulations," New York State Article 28 acute care hospitals were mandated to implement sepsis protocols and report patient-level data. This study sought to determine how well cases reported under state mandate align with discharge records in a statewide administrative database.

Design: Observational cohort study.

Setting: First 27 months of mandated sepsis reporting (April 1, 2014, to June 30, 2016).

Patients: Hospitalizations with sepsis at New York State Article 28 acute care hospitals.

Intervention: Sepsis regulations with mandated reporting.

Measurements and main results: We compared cases reported to the New York State Department of Health Sepsis Clinical Database with discharge records in the Statewide Planning and Research Cooperative System database. We classified discharges as 1) "coded sepsis discharges"-a diagnosis code for severe sepsis or septic shock and 2) "possible sepsis discharges," using Dombrovskiy and Angus criteria. Of 111,816 sepsis cases reported to the New York State Department of Health Sepsis Clinical Database, 105,722 (94.5%) were matched to discharge records in Statewide Planning and Research Cooperative System. The percentage of coded sepsis discharges reported increased from 67.5% in the first quarter to 81.3% in the final quarter of the study period (mean, 77.7%). Accounting for unmatched cases, as many as 82.7% of coded sepsis discharges were potentially reported, whereas at least 17.3% were unreported. Compared with unreported discharges, reported discharges had higher rates of acute organ dysfunction (e.g., cardiovascular dysfunction 63.0% vs 51.8%; p < 0.001) and higher in-hospital mortality (30.2% vs 26.1%; p < 0.001). Hospital characteristics (e.g., number of beds, teaching status, volume of sepsis cases) were similar between hospitals with a higher versus lower percent of discharges reported, p values greater than 0.05 for all. Hospitals' percent of discharges reported was not correlated with risk-adjusted mortality of their submitted cases (Pearson correlation coefficient 0.11; p = 0.17).

Conclusions: Approximately four of five discharges with a diagnosis code of severe sepsis or septic shock in the Statewide Planning and Research Cooperative System data were reported in the New York State Department of Health Sepsis Clinical Database. Incomplete reporting appears to be driven more by underrecognition than attempts to game the system, with minimal bias to risk-adjusted hospital performance measurement.

Figure 1. Data Linkage

This figure depicts linkage between SPARCS and the NYSDOH Sepsis Clinical Database. Key Terms:

1. Coded sepsis discharge: an inpatient or emergency department discharge record in SPARCS with a diagnosis code for severe sepsis or septic shock.

2. Reported Sepsis Discharge: An inpatient or emergency department discharge record in SPARCS that was reported to NYSDOH Sepsis Clinical Database for under mandated sepsis reporting. Only discharge records that were matched to the NYSDOH Sepsis Clinical Database could be confirmed as reported.

3. Unreported Sepsis Discharge: An inpatient or emergency department discharge record in SPARCS with a diagnosis code for severe sepsis or septic shock that was not reported to NYSDOH for performance measurement under mandated sepsis reporting.

4. Matched Sepsis Case: A case submitted to NYSDOH Sepsis Clinical Database and matched to an inpatient or emergency department discharge record in SPARCS.

5. Unmatched Sepsis Case: A case submitted to NYSDOH Sepsis Clinical Database that could not be matched to an inpatient or emergency department discharge record in the SPARCS.

Abbreviations: PPV, positive predictive value; SPARCS, Statewide Planning and Research Cooperative System

Figure 2. Correlation between Hospital-Level Reporting and Risk-Adjusted Mortality Rate

This figure depicts the correlation between individual hospitals' percentage of sepsis cases reports with risk-adjusted in-hospital mortality.