Discrepancies in Hospital Readmissions Administrative Data

The Centers for Medicare & Medicaid Services recently created a readmission measure that uses a hospital’s administrative data to gauge readmission rates in an initiative to hold hospitals accountable where readmission rates are deemed too high.

A recent study addressed the issue of the accuracy of administrative codes to determine the cause of readmission and evaluated the readmission measure’s ability to identify planned readmissions [JAMA Surg. 2014; DOI:10.101/jamasurg.2014.18]. The study also documented the frequency of unrelated readmissions. Several studies have proposed that administrative claims data correlate poorly with clinic data taken directly from medical records. In turn, the study’s researchers noted that clinical data can potentially offer a more accurate estimate.

This study examined readmission in 3 cases: (1) to determine the accuracy of each readmission in administrative claims data compared with the clinical diagnosis taken from medical records; (2) to test the readmission measure’s capability to exclude planned readmissions; and (3) to determine the frequency of readmissions that were clinically unrelated to the original reason for hospitalization.

The records for all hospital readmissions of general surgery patients within 30 days following discharge were obtained from the University HealthSystem Consortium (UHC). The readmissions data from October 2009 through December 2009 and July 2011 through September 2011 were analyzed. Data were also collected on patient demographics, such as age, sex, race, operation performed during hospitalization, and length of stay. The number of days between discharge and readmission was also recorded.

The researchers recorded discrepancies between administrative and clinical data and determined whether each readmission fell into 1 of 3 categories: (1) secondary diagnosis relevant to readmission but not relevant to the primary reason for readmission; (2) a previous diagnosis given before the readmission; and (3) an inaccurate code that did not match the patient’s medical record. Coding errors also occurred when readmissions were present in the UHC database but a review of the clinical medical record showed that a readmission had not taken place.

The researchers also determined the number of unplanned readmissions that were clinically unrelated to the original hospitalization. Unrelated readmissions were also categorized into 3 cases: (1) acute illness unrelated to the first hospital stay; (2) treatment of a chronic condition; and (3) other social reasons.

During the study, 3788 patients were discharged from surgical services. Within 30 days, 315 (8.3%) were readmitted. Of these readmissions, 4.4% were removed for coding errors and 13.7% were planned readmissions. Of the other 258 readmissions, 72.9% were related to the original cause, whereas 27.1% were unrelated. The readmissions listed in the administrative claims data varied from the clinical diagnosis in 97 readmissions.

The criteria defined by the readmission measure demonstrated that there were 15 (4.8%) planned readmissions. According to the clinical data, 43 were planned, including the 15 identified by the readmission measure (P<.001). The researchers concluded that the readmission measure included one-third of the total planned readmissions. Seventy of the 258 remaining unplanned readmissions were unrelated to the original hospital stay.

The study found that the administrative data did not correctly identify the reason for readmission in one-third of the total cases compared to the readmission measure that failed to identify two-thirds of the planned readmissions. These results demonstrated the restrictions of using administrative claims data to assess a hospital’s readmissions performances. Several related studies have reportedly found similar discrepancies in the data.

Specifically, planned readmissions serve as an important component of patient care and do not reflect on a hospital’s quality. The findings of this study suggest that relying on administrative claims data does not recognize two-thirds of the planned readmissions.

The researchers noted study limitations, including that this was a single-institution study, which makes the generalizability of the results unknown.  

The authors concluded that the use of administrative claims data to evaluate hospital readmissions is inaccurate and does not predict the true number of planned readmissions. The frequency of readmissions that are unrelated to the original stay should not be included in the financial penalties of policymakers. Even if hospitals improve their care quality, unrelated health issues could arise and thus cause a hospital to be penalized, according to the authors.—Jordyn Greenblatt