Association Between Postop Complication and Readmission
Association Between Postop Complication and Readmission
The study design and procedures were approved by the RAND Health institutional review board. The primary data sources for this study were Medicare inpatient claims and ACS-NSQIP, which have been previously described. Briefly, Medicare data were obtained from the 100% Medicare provider analysis and review file (MedPAR). Medicare is a health insurance program that enrolls people aged 65 years and older, some disabled people under the age of 65, and all people with end-stage renal disease receiving dialysis. Each record in MedPAR represents an inpatient hospital stay for a beneficiary and may include multiple claims. Diagnoses and procedures are recorded by International Classification of Diseases, Ninth Edition (ICD-9) code. Each Medicare beneficiary has a unique identification number allowing for linkage of subsequent hospitalizations without disclosing the patient's identity.
ACS-NSQIP is an institution-based, multispecialty, clinical surgical registry. Hospital participation in ACS-NSQIP is voluntary and requires a dedicated data abstractor who is trained to use strict variable definitions and collection methods. The sampling strategy includes collecting data for the first 40 cases performed within consecutive 8-day cycles, excluding trauma and transplantation. Hospitals are audited to ensure standardized data collection. Data collected include preoperative risk factors, procedures performed by Current Procedural Terminology code, and postoperative complications occurring within 30 days of the index operation.
Eligible patient-level records from ACS-NSQIP, years 2005 to 2008, were linked to Medicare inpatient claim records in MedPAR using indirect patient identifiers and a deterministic linkage algorithm, as previously described. As previously reported, there was an excellent agreement between ACS-NSQIP and MedPAR records on death during the primary hospitalization, supporting the validity of the linkage procedure. Our study population was restricted to patients aged 65 years or older who underwent an inpatient surgical procedure during the years studied, were entered into the ACS-NSQIP database, and for whom we were able to successfully link the ACS-NSQIP record to Medicare claims data. We excluded patients for whom Medicare was not the primary payer and patients with procedures occurring in December 2008 because we lacked a full 30 days of follow-up in the Medicare data. We also excluded patients who did not survive to be discharged from the primary hospitalization (n = 3649) and patients who could not be readmitted within 30 days of the surgery date because they were still hospitalized (n = 2030). Our final sample consisted of 90,932 patients from 214 hospitals.
A third data source, the Nationwide Inpatient Sample (NIS), was used to determine the number of procedures performed in US hospitals on Medicare beneficiaries older than 65 years in 2009. The NIS contains discharge data from a systematic sample of hospitals nationwide. Procedures were identified by ICD-9 code, and national estimates were calculated using sample hospital weights.
Our primary outcomes of interest were 30-day postoperative readmission, defined as any admission to a short stay hospital within 30 days of the index operation, and the cost of the readmission. Both outcome variables were identified from MedPAR. We chose to focus on 30-day postoperative readmission rather than on 30-day postdischarge readmission, because we thought it would provide more relevant and useful information to surgeons and because it is the time frame traditionally used for monitoring postoperative outcomes. Admissions to long-stay hospitals or skilled nursing facilities were not considered readmissions, nor were admissions occurring on the same day as discharge, as these are likely transfers. In addition, admissions for maintenance chemotherapy or radiotherapy were considered to be planned and were not counted as readmissions. The perspective taken for the cost analysis was that of the Medicare program, so the cost of readmission was measured by the dollar amount paid to the hospital by Medicare.
Reason for readmission was identified from the first ICD-9 diagnosis code recorded in MedPAR. ICD-9 codes were grouped into clinically meaningful categories using Clinical Classification Software from the Agency for Healthcare Research and Quality.
Our explanatory variable of interest was a binary variable indicating whether or not a patient had a 30-day postoperative complication. This variable included any occurrence of 20 complications recorded in ACS-NSQIP: surgical site infection (superficial, deep, or organ-space), wound disruption, pneumonia, unplanned intubation, pulmonary embolism, on ventilator for more than 48 hours, progressive renal insufficiency or acute renal failure requiring dialysis, urinary tract infection, stroke, coma, cardiac arrest requiring cardiopulmonary resuscitation, myocardial infarction, bleeding requiring transfusion, deep venous thrombosis requiring therapy, sepsis or septic shock, or unplanned return to the operating room. Complications are recorded in ACS-NSQIP if they occur within 30 days of surgery, except bleeding requiring transfusion, which must occur within the first 72 hours after surgery.
Variables used for risk adjustment (Table 1) and procedures performed were identified from ACS-NSQIP. Procedures were identified by Current Procedural Terminology code and grouped into 73 broad categories, which were used to stratify the analyses.
Methods
Data Sources and Study Sample
The study design and procedures were approved by the RAND Health institutional review board. The primary data sources for this study were Medicare inpatient claims and ACS-NSQIP, which have been previously described. Briefly, Medicare data were obtained from the 100% Medicare provider analysis and review file (MedPAR). Medicare is a health insurance program that enrolls people aged 65 years and older, some disabled people under the age of 65, and all people with end-stage renal disease receiving dialysis. Each record in MedPAR represents an inpatient hospital stay for a beneficiary and may include multiple claims. Diagnoses and procedures are recorded by International Classification of Diseases, Ninth Edition (ICD-9) code. Each Medicare beneficiary has a unique identification number allowing for linkage of subsequent hospitalizations without disclosing the patient's identity.
ACS-NSQIP is an institution-based, multispecialty, clinical surgical registry. Hospital participation in ACS-NSQIP is voluntary and requires a dedicated data abstractor who is trained to use strict variable definitions and collection methods. The sampling strategy includes collecting data for the first 40 cases performed within consecutive 8-day cycles, excluding trauma and transplantation. Hospitals are audited to ensure standardized data collection. Data collected include preoperative risk factors, procedures performed by Current Procedural Terminology code, and postoperative complications occurring within 30 days of the index operation.
Eligible patient-level records from ACS-NSQIP, years 2005 to 2008, were linked to Medicare inpatient claim records in MedPAR using indirect patient identifiers and a deterministic linkage algorithm, as previously described. As previously reported, there was an excellent agreement between ACS-NSQIP and MedPAR records on death during the primary hospitalization, supporting the validity of the linkage procedure. Our study population was restricted to patients aged 65 years or older who underwent an inpatient surgical procedure during the years studied, were entered into the ACS-NSQIP database, and for whom we were able to successfully link the ACS-NSQIP record to Medicare claims data. We excluded patients for whom Medicare was not the primary payer and patients with procedures occurring in December 2008 because we lacked a full 30 days of follow-up in the Medicare data. We also excluded patients who did not survive to be discharged from the primary hospitalization (n = 3649) and patients who could not be readmitted within 30 days of the surgery date because they were still hospitalized (n = 2030). Our final sample consisted of 90,932 patients from 214 hospitals.
A third data source, the Nationwide Inpatient Sample (NIS), was used to determine the number of procedures performed in US hospitals on Medicare beneficiaries older than 65 years in 2009. The NIS contains discharge data from a systematic sample of hospitals nationwide. Procedures were identified by ICD-9 code, and national estimates were calculated using sample hospital weights.
Outcomes of Interest
Our primary outcomes of interest were 30-day postoperative readmission, defined as any admission to a short stay hospital within 30 days of the index operation, and the cost of the readmission. Both outcome variables were identified from MedPAR. We chose to focus on 30-day postoperative readmission rather than on 30-day postdischarge readmission, because we thought it would provide more relevant and useful information to surgeons and because it is the time frame traditionally used for monitoring postoperative outcomes. Admissions to long-stay hospitals or skilled nursing facilities were not considered readmissions, nor were admissions occurring on the same day as discharge, as these are likely transfers. In addition, admissions for maintenance chemotherapy or radiotherapy were considered to be planned and were not counted as readmissions. The perspective taken for the cost analysis was that of the Medicare program, so the cost of readmission was measured by the dollar amount paid to the hospital by Medicare.
Reason for readmission was identified from the first ICD-9 diagnosis code recorded in MedPAR. ICD-9 codes were grouped into clinically meaningful categories using Clinical Classification Software from the Agency for Healthcare Research and Quality.
Other Variables Used for Analyses
Our explanatory variable of interest was a binary variable indicating whether or not a patient had a 30-day postoperative complication. This variable included any occurrence of 20 complications recorded in ACS-NSQIP: surgical site infection (superficial, deep, or organ-space), wound disruption, pneumonia, unplanned intubation, pulmonary embolism, on ventilator for more than 48 hours, progressive renal insufficiency or acute renal failure requiring dialysis, urinary tract infection, stroke, coma, cardiac arrest requiring cardiopulmonary resuscitation, myocardial infarction, bleeding requiring transfusion, deep venous thrombosis requiring therapy, sepsis or septic shock, or unplanned return to the operating room. Complications are recorded in ACS-NSQIP if they occur within 30 days of surgery, except bleeding requiring transfusion, which must occur within the first 72 hours after surgery.
Variables used for risk adjustment (Table 1) and procedures performed were identified from ACS-NSQIP. Procedures were identified by Current Procedural Terminology code and grouped into 73 broad categories, which were used to stratify the analyses.
