Postoperative Complications After Percutaneous Nephrolithotomy: A Contemporary Analysis by Insurance Status in the United States

Introduction

The association between patients who are not privately insured and poor perioperative outcomes has been documented for several different disciplines of surgery, including otolaryngology, ¹ cardiothoracic surgery, ² neurosurgery, ³ and colorectal surgery. ⁴ Although the mechanism behind this association is largely unknown, an improved understanding of the factors could assist public health officials to more effectively target these areas in future policy making.

No published data to date have assessed the insurance-related disparities among patients undergoing percutaneous nephrolithotomy (PCNL), to our knowledge. In the United States, the prevalence of nephrolithiasis has been estimated at 9%, which represents a major public health concern. ⁵ For patients with a large stone burden, PCNL may represent the best surgical approach. Although data on postoperative complication rates have been published, ⁶ no comprehensive, standardized predictive models of medical comorbidity and perioperative outcomes have been developed in a large, population-based dataset.

We hypothesized that considerable variation exists in the PCNL perioperative complication rates stratified by insurance status, on the basis of findings published in the nonurologic surgical literature. We used the Nationwide Inpatient Sample (NIS), the largest publicly available, all-payer inpatient database in the United States. Because disparities in outcomes may be attributable to differences in overall baseline health, we included a subset of patients with no major medical comorbidity. Furthermore, we developed a multivariable model of medical comorbidity that predicts ≥1 perioperative complication for any patient undergoing PCNL.

Patients and Methods

We assembled data from the NIS between January 1, 1998, and December 31, 2010. This sample is an administrative database published by the Healthcare Cost and Utilization Project, which is sponsored by the Agency for Healthcare, Research, and Quality. ⁷ The NIS contains discharge data from approximately 8 million hospital admissions annually. For example, for 2010, it holds all discharge data from 1051 hospitals in 45 states, encompassing a stratified sample of 20% of U.S. community hospitals. The NIS also has charge information for all patients, regardless of payer, including patients with Medicare, Medicaid, or private insurance coverage and patients with no insurance.

Patient and hospital level variables

To account for case-mix differences, baseline patient characteristics were extracted from the database, including race/ethnicity (white, black, Hispanic, Asian/Pacific Islander, Native American, and other), age, median household income quartiles for patient ZIP code, and sex. A comorbidity score was assigned to every patient through the Elixhauser comorbidity index. ⁸ We also extracted hospital level data, including hospital region (ie, Northeast, Midwest, South, and West), location (rural vs urban), number of beds (small, medium, and large), and teaching status. We excluded patients with missing data on vital demographic characteristics (Fig. 1).

OPEN IN VIEWER

FIG. 1.

Nationwide Inpatient Sample from 1998 through 2010. NIS=National Inpatient Sample; PCNL=percutaneous nephrolithotomy.

To assess the possibility of differences in outcomes attributable to baseline health status, we constructed a dataset composed solely of patients with an Elixhauser comorbidity index of zero, as described previously. ¹ This method was used because isolating the effects of insurance status from those of comorbid conditions is preferable to including comorbidity in the model. The latter approach would necessitate strong assumptions that were unlikely to hold true and would risk undermining any meaningful causal inferences.

Results

Discussion

In this cohort of U.S. patients undergoing PCNL from 1998 through 2010, patients with government insurance (Medicare and Medicaid) had higher overall perioperative complication rates than patients with private insurance and no insurance. This disparity, however, was not present in a subset of healthy patients, a result indicating that this observation can be attributed largely to differences in baseline health status. Furthermore, although primary payer status was not predictive of having ≥1 perioperative complications in the multivariable logistic regression models, female sex and black race/ethnicity were predictive. We also found that several clinical diseases were significantly associated with perioperative morbidity, including pulmonary circulatory disorders, coagulopathy, and deficiency anemias.

Clustering may confound the analysis in that patients with certain demographic characteristics may seek care at higher-volume community or teaching hospitals. To test whether the associations between perioperative morbidity and black race/ethnicity and female sex were explained by a clustering effect, we repeated the analysis in a subset of healthy patients treated in teaching institutions. The associations held true with similar ORs and CIs, indicating that clustering likely had only a minimal, if any, effect on these observations.

Although high-quality, prospective, multi-institutional data examining the overall complication rates after PCNL already exist, the present study is, to our knowledge, the largest population-based study to comprehensively analyze which medical comorbidities are associated with perioperative complications. ⁶ Many of these associations may be self-evident from anecdotal experience (e.g., patients with deficiency anemias are more likely to have perioperative complications, particularly when blood transfusion is defined as a perioperative complication).

Yet, we demonstrated several findings that may be less obvious. For example, on the basis of our experience with complex PCNLs, we hypothesized that patients with paraplegia may be prone to perioperative complications for several reasons, including altered physical anatomy and delayed postoperative activity. The observations from the present study support the notion that complications are more frequent in paraplegic patients. This is an important finding, because clinicians can now begin to define and, perhaps more importantly, quantify this risk for these patients when they are undergoing PCNL.

Likewise, several clinical diseases that would not necessarily be expected to have an association with perioperative morbidity (e.g., depression, rheumatoid diseases, hypothyroidism) were not statistically significant in our model, which helps validate our findings. Of note, hypertension, diabetes mellitus, and obesity were not statistically associated with perioperative morbidity independently, indicating that these diseases may instead be phenotypic markers for more severe comorbidities, such as congestive heart failure, renal failure, and pulmonary disease, which predispose patients to perioperative complications after PCNL.

These strengths notwithstanding, there are several important limitations to recognize. First, our analysis was limited to in-hospital complications. Hence, our dataset does not include several variables that may be of interest, such as second-look PCNL rates, stone burden, stone type, operative technique, and stone-free rates. These data may impact the findings in unmeasured ways. Second, the identification of outcomes of interest was determined primarily through diagnosis and procedure codes. Third, this was a study of U.S. hospitals and may not be generalizable to other regions of the world.

Despite these limitations, we believe that our claims on the perioperative outcomes stratified by insurance provider are relevant given that this is the largest analysis of PCNL complications through a population-based administrative dataset. Furthermore, with these data, clinicians can now define the perioperative risk of complications on the basis of medical comorbidity data.

Conclusion

Perioperative morbidity after PCNL varied significantly in accordance with insurance status, but this variation is mostly explained by differences in overall health. Patients with pulmonary circulatory disorders, coagulopathy, deficiency anemia, and paralysis were most likely to have one or more surgical complications.