Predictive Modeling and Analyses of National Emergency Department Data for Improving Patient Outcomes of Nephrolithiasis

Background: Nephrolithiasis, or kidney stones, imposes a significant burden of disease in the United States and comes with considerable costs, pain, and morbidity. The exact cause of stone formation is undefined, but the formation is a process. Risk factors include environmental, diabetes, obesity, metabolic syndromes, low fluid intake, dehydration, diet, inflammatory bowel disorders, irritable bowel syndrome (IBS), and genetics. Laboratory testing and appropriate diagnostic imaging studies are two key components of assessment and prevention. Methods: This is a retrospective, quantitative study utilizing the Healthcare Cost and Utilization Project’s (HCUP) National Emergency Department Sample (NEDS)'s existing databases from 2012 to 2014 to classify outcomes for nephrolithiasis patients. International Classification of Diseases, Ninth Revision (ICD-9-CM) billing codes related to nephrolithiasis, relevant medical imaging exams, and procedural and surgical billing codes for interventions and procedures were selected. Descriptive statistical analyses as well as multiple regression models were used to analyze frequencies and percentages of variables and the relationship of the data, identify co-linearity among variables, and predict outcomes. Results: The study sample includes a total of 509,192 emergency department (ED) visits for nephrolithiasis from 2012 to 2014 and reveals that IBS patients are two times more likely to require intervention. Stepwise regression models yield P-values of 0.004 for gender, 0.017 and 0.018 for minor diagnostic procedures, 0.006 and 0.001 for minor therapeutic procedures, and 0.000 and 0.001 for major therapeutic procedures when predicting total cost of care, and have a statistically significant impact on patient outcomes of nephrolithiasis. Conclusions: This research offers an investigation of the prevalence of nephrolithiasis based on age, gender, and co-morbidity, specifically IBS, and is the first to report on patient outcomes. This analysis also provides clinicians with recommendations to utilize for a comprehensive assessment of nephrolithiasis patients in the ED.


Introduction
Nephrolithiasis, or kidney stones, imposes a significant burden of disease in the United States. Nearly one in every 11 patients of working age will experience nephrolithiasis and 50% will encounter recurrence [1]. The formation of nephrolithiasis is a process, and not a single event, in which crystallization and crystal aggregation of minerals from supersaturated urine occur. Nephrolithiasis can be composed of calcium oxalate, cystine, uric acid, or struvite. The most common type of stone formed is composed of calcium oxalate. Urine concentration, infection, and urinary pH are inherent factors in this crystallization process and nephrolithiasis formation [2]. nephrolithiasis due to its high sensitivity and specificity by the American College of Radiology (ACR); however, ultrasound is often the first modality utilized at the point of care. Low-dose CT scans are recommended for patients with a body mass index (BMI) of 30 kg/m 2 and a full-dose scan is recommended for those with a greater BMI [4]. Although ultrasound is an acceptable means of imaging nephrolithiasis and a primary means of imaging pediatric and pregnant patients in the first trimester, this modality has its limitations. Often ultrasound over-or underestimates stone size, presenting physicians with incorrect information and potentially delaying proper treatment of the patient.
Upon presentation to the ED, patients with suspected first-time nephrolithiasis or recurrent stones should be evaluated with a complete laboratory and diagnostic imaging workup. Chronic dehydration is easily detected with plasma osmolality and copeptin levels, which can easily be determined with laboratory testing. Certain comorbidities associated with nephrolithiasis produce larger stones, specifically in patients with inflammatory bowel disease/disorders (i.e. IBD and irritable bowel syndrome [IBS]). IBD and IBS patients have a greater risk of nephrolithiasis due to fluid imbalance and dehydration due to chronic and acute gastrointestinal symptoms. Hydration status (detected with plasma osmolality) and comorbidities play an integral role in stone size and these factors should be used to guide the physician in the care of the nephrolithiasis patient, thus improving patient outcomes and decreasing overall healthcare costs.

Materials And Methods
This is a retrospective, quantitative study utilizing the Healthcare Cost and Utilization Project's (HCUP) National Emergency Department Sample (NEDS) databases from 2012 to 2014 to classify outcomes for nephrolithiasis patients. This study analyzed frequencies and percentages of variables, as well as the relationship of data selected with multiple regression models and identification of co-linearity among the variables using IBM SPSS Statistics version 26 (IBM Corp, Armonk, NY). The key objective was to be able to analyze and produce a predictive model for improving patient outcomes of nephrolithiasis. Multiple patient-level variables from the HCUP core datasets including age, gender, and comorbidities of obesity, diabetes, IBS, dehydration, and hyperosmolality, as well as relevant medical imaging exams and procedures in addition to total charges and discharge versus admission were extracted.
Statistical analysis was performed utilizing IBM SPSS Statistics version 26 on an OS X Yosemite operating system. A statistical analysis of the NEDS core data sets from 2012 to 2014 was done using descriptive analytics of frequencies and percentages which provides a detailed summary of the data sets. Regression analyses were performed in SPSS to determine the variables for use and to ensure the model fits the data. A linear, or standard, regression method was utilized, as well as stepwise regression models.   Table 2). Therefore, gender in IBS patients should be considered a key component of nephrolithiasis.  The prevalence of dehydration as a primary diagnosis and/or secondary diagnosis (see Table 3) was investigated as well as hyperosmolality (see Table 4). As the primary recommendation for the prevention of nephrolithiasis is increased fluid intake and based on the process of stone formation, dehydration would likely be present in these patients. Hyperosmolality in patients with nephrolithiasis is worth noting in this study as this is an indicator of dehydration and an alternative method of assessment as plasma osmolality is not identified in ICD-9 coding. As plasma osmolality is a key indicator of chronic dehydration, hyperosmolality is also a key indicator of dehydration. The total number of patients from 2012 to 2014 with a diagnosis of dehydration in this study is 4605 (0.9%). The total number of patients with a diagnosis of hyperosmolality in this study is 204 (0.04%). This low yield results in the question of whether data collection or assessment of hydration status is lacking.

TABLE 4: Prevalence of Hyperosmolality in Nephrolithiasis
Medical imaging is a key component of nephrolithiasis assessment and care. Patients included in the HCUP NEDS 2012, 2013, and 2014 Supplemental Emergency Department Files, which provide "information on procedures that were performed in the ED for treat and release" [5], had ultrasound imaging of the urinary system performed as the principal procedure, more than CT imaging, which the ACR's evidence-based clinical guidelines recommend for initial imaging of suspected nephrolithiasis.
Patients included in the HCUP NEDS Supplemental Inpatient Files, which provide "information on inpatient admissions after ED visits" [5], demonstrate a higher incidence of CT imaging than that of ultrasound demonstrating the need to analyze this data to further support or negate the hypothesis that patients imaged with ultrasound before CT have poorer outcomes and greater overall hospital charges for combined ED and inpatient services.
The intervention of nephrolithiasis is a strong indicator of stone size as the ability of a smaller stone to pass is far more likely than that of a larger stone which typically results in bleeding or infection requiring intervention. The HCUP NEDS datasets include a procedure class data element that "categorizes ICD-9 procedure codes into one of four broad categories" [6,7]. These categories include minor diagnostic, minor therapeutic, major diagnostic, and major therapeutic. Minor categories represent non-operating room procedures and therapies, whereas major categories represent valid operating room procedures and therapies. Of the nephrolithiasis cases found in the datasets, interventions were performed in 9158 (5.6% of this population) inpatients in 2012, 8542 (5.2% of this population) inpatients in 2013, and 9628 (5.3% of this population) inpatients in 2014 (see Table 5) for a total of 27,328 (5.4% of this overall population) interventions. The 2012 interventions include 2038 minor diagnostic procedures, 3978 minor therapeutic procedures, 58 major diagnostic, and 3048 major therapeutic for those admitted to the hospital with nephrolithiasis (see Table 6). The types of interventions related to nephrolithiasis identified in this study include percutaneous nephrostomy without fragmentation, percutaneous nephrostomy with fragmentation, ureteroscopy, cystoscopy, ultrasonic fragmentation of urinary stones, retrograde pyelogram, percutaneous pyelogram, ESWL, and ESWL of the kidney, ureter, or bladder.

Descriptive statistical analysis of IBS
Analyses of the 2012, 2013, and 2014 HCUP NEDS databases have identified a total of 958 ED visits for nephrolithiasis in patients with IBS. Although the prevalence of nephrolithiasis was found to be more common in males than females in the initial analyses, the prevalence is far greater in females (67%) with IBS. Of these patients identified, only 22 patients across the three years analyzed were assessed or diagnosed with CT and none were assessed with ultrasound. Of these 958 IBS patients, 104 (10.9% of this population) underwent diagnostic or therapeutic procedures (see Table 7). This number represents a total of 81 minor diagnostic or therapeutic procedures and 27 major diagnostic or therapeutic procedures. Principal interventions and procedures directly related to nephrolithiasis in patients with IBS include a total of 17 in 2012, 16 in 2013, and 11 in 2014 (see Table 8).

Multiple regression analysis
A standard multiple regression was performed on the HCUP NEDS 2012, 2013, and 2014 datasets for nephrolithiasis patients with IBS. This method allows for all independent, or predictor, variables to be simultaneously entered into the model. Independent variables are each evaluated in terms of their predictive power in comparison to that of the other independent variables. Insight is provided where each of the independent variables may explain unique variances in a dependent variable.
In the 2012 HCUP NEDS dataset linear regression model tested, the primary diagnosis of nephrolithiasis was utilized as the dependent variable, while age, gender, comorbidities, principal and secondary procedures from the ED, as well as principal and secondary procedures from the inpatient discharge record were utilized as independent variables. The P-value for the principal procedure from the inpatient discharge record rejects the null hypothesis that these procedures happen by chance, with a return of 0.001 (see Table 9). This information demonstrates that the principal procedure from the inpatient discharge record represents a statistically significant contribution. Principal procedures include medical imaging modalities such as ultrasound and CT as well as minor and/or major procedures and interventions. Based on the information provided in the descriptive statistical analysis of IBS patients in 2012, ureteroscopy and cystoscopy were the only principal procedures returned related to nephrolithiasis demonstrating the significance of these procedures to nephrolithiasis in IBS patients. While the 2013 dataset does not provide the same results, the 2014 dataset rejects the null hypothesis with a P-value of 0.024 for the principal procedure from the ED (see Table 10). The information provided in the descriptive statistical analysis of IBS patients in 2014 returned CT of the abdomen as the most common procedure in the ED. These results from the linear regression models in 2012 and 2014 require additional regression models of analysis to better understand the data and determine which variables may have higher predictability.   A stepwise regression model was then tested with age as the dependent variable as it has been determined to be a contributing factor of nephrolithiasis and gender, diagnoses, principal procedures from the ED, and principal procedures from the inpatient discharge record as independent variables. This stepwise regression allows SPSS to automatically select variables based on statistical criteria and enter them in the order of the equation. Although this method can be controversial for its false positive errors, in machine learning, stepwise procedures have been utilized with remarkable success.
The 2012 HCUP NEDS dataset yields an adjusted R square of 0.011, demonstrating that 1% of the variance of age (dependent variable) is explained by independent variables. These results determine that age does not influence the principal procedures performed, including ultrasound vs. CT; however, it does not indicate that age does not contribute to nephrolithiasis. The normal probability plot (P-P) for age does not suggest any major deviations from normality demonstrating age as a standardized result. In the 2013 and 2014 HCUP NEDS datasets, age is normalized against the 2012 dataset results with R square (0.061 or 6% in 2013 and 0.026 or 2% in 2014) and P-Plots showing very similar results.
The principal procedure class was then assessed using the stepwise model of regression. This method allowed for the assessment of independent variables such as primary diagnosis and comorbidity to predict outcomes. In the 2012 analysis, procedures play a role in predictability which is seen in values of 0.002 and 0.035 (see Table 11), and in 2013 although not statistically significant, the P-value for the principal procedure from the inpatient discharge record returned 0.054 (see Table 11). In 2014, comorbidity played a role in predictability which can be seen in the P-value as 0.025 (see Table 11). While there is variability in these stepwise regression models, the data provided give insight into the diagnostic and therapeutic procedures for IBS patients, a key component of this study.  Additional stepwise regression models were analyzed with total charges as the dependent variable, while age, gender, and procedure class were independent variables. The 2012 stepwise regression model for total charges for ED services yields a P-value of 0.000 for the principal procedure class for major therapeutic procedures which include interventions requiring the use of the operating room (OR), and a P-value of 0.017 for minor diagnostic procedures which include ultrasound and CT imaging (see Table 12). These P-values reject the null hypothesis that major therapeutic and minor diagnostic procedures and interventions happen by chance and represent statistically significant contributions to total charges for ED visits. The adjusted Rsquare goes up slightly with the addition of minor diagnostic procedures by 0.092.  The 2012 stepwise model with total charges for ED and inpatient services as the dependent variable yields Pvalues of 0.006 and 0.001 for minor therapeutic non-OR procedures, as well as a P-value of 0.018 for minor diagnostic procedures, which again include ultrasound and CT imaging (see Table 13). These P-values reject the null hypothesis that these minor therapeutic and minor diagnostic procedures happen by chance and represent statistically significant contributions to total charges for ED and inpatient services. The adjusted R-square goes up slightly with the addition of minor diagnostic procedures by 0.147.  The 2013 stepwise regression model for total charges for ED services yields a P-value of 0.000 for the principal procedure class for major therapeutic procedures which include interventions requiring the use of the OR, and a P-value of 0.006 for minor therapeutic, or non-OR, procedures (see Table 14). These P-values reject the null hypothesis that major therapeutic and minor therapeutic procedures and interventions happen by chance and represent statistically significant contributions to total charges for ED visits. The adjusted R-square goes up slightly with the addition of minor therapeutic procedures by 0.156.  The 2013 stepwise model with total charges for ED and inpatient services as the dependent variable yields Pvalues of 0.004 for gender as an independent variable and rejects the null hypothesis and represents a statistically significant contribution to total charges for ED and inpatient services (see Table 15).  The 2014 stepwise regression model for total charges for ED services yields a P-value of 0.000 for the principal procedure class for major therapeutic procedures which include interventions requiring the use of the OR (see Table 16). This P-value rejects the null hypothesis and represents a statistically significant contribution to total charges for ED visits.  The 2014 stepwise model with total charges for ED and inpatient services as the dependent variable yields Pvalues of 0.001 for major therapeutic, OR procedures. This P-value rejects the null hypothesis and represents a statistically significant contribution to total charges for ED and inpatient services.
The 2012, 2013, and 2014 stepwise regression analyses with total charges as the dependent variable and gender and procedure class as independent variables contributing to the model confirm the hypothesis that patients with a comorbidity of IBS who are assessed with ultrasound before CT have poorer outcomes and accrue greater healthcare costs, due to additional imaging and the need for intervention.

Discussion
This study analyzed the HCUP NEDS 2012, 2013, and 2014 database variables including age, gender, and comorbidities of obesity, diabetes, and IBS, as well as dehydration and hyperosmolality, relevant medical imaging exams, interventions and procedures, total charges, and discharge versus admission. The study sample included 164,410 ED visits for nephrolithiasis in 2012, 163,254 in 2013, and 181,528 in 2014 for a total of 509,192 visits.
The prevalence of nephrolithiasis is highest among adults between the ages of 30 and 42 and was found to be more common in males than females. While there is a slight prevalence of nephrolithiasis in patients with a comorbidity of obesity (0.57%), there is a higher rate in those with diabetes (4.78%). However, the focus of this study is on patients with IBS, as this population is at greater risk of dehydration due to regular bouts of diarrhea.
The link between nephrolithiasis and dehydration has been clearly outlined. While this link seems to be well known, there is a lack of studies on laboratory values for dehydration, such as plasma osmolality, which is a strong indicator of chronic dehydration. While dehydration is found in patients across the databases utilized in this study, there is a lack of confidence that this is consistently assessed, if at all, upon presentation to the ED and therefore most likely does not accurately represent the prevalence in the IBS population studied, and more importantly stone formers.
Overall data represent ED visits for nephrolithiasis between 2012 and 2014, which often includes medical imaging for assessment and diagnosis. The prevalence of ultrasound and CT exams was reviewed for both ED cases and those admitted to the hospital. Where ultrasound and CT are seen in the same visit, it can be assumed that ultrasound was performed first as this method of assessing patients is more frequently utilized despite the ACR's appropriateness criteria guidelines. Non-contrast CT is the first-line imaging choice for detecting nephrolithiasis with a 95% sensitivity and specificity and is considered the gold standard for its ability to identify the location, size, and composition of stones, and its ability to provide differential diagnoses.
Ultrasound of the urinary system was more commonly utilized in the ED than CT; however, CT was more commonly utilized for patients admitted to the hospital with nephrolithiasis. This is interesting as those that are admitted to the hospital due to nephrolithiasis are due to larger stone size, requiring intervention, blood, or infection.
Analysis of procedure class for inpatient procedures yielded 9158 minor and major diagnostic and therapeutic procedures in 2012, 8542 in 2013, and 9628 in 2014. Total charges for patients with nephrolithiasis presenting to the ED and then being admitted to the hospital were found to be significantly higher than that of patients discharged from the ED. In 2012, the mean total charges for ED services were $21,796 less than the combined ED and inpatient charges; in 2013, the mean total charges for ED services were $23,557 less than the combined ED and inpatient charges; and in 2014, the mean total charges for ED services were $26,555 less than the combined ED and inpatient charges. This is a significant burden of cost to patients and healthcare systems.
IBS was assessed as a comorbidity of nephrolithiasis in this study as dehydration is very common in patients with recurrent bouts of diarrhea. Three hundred and fifty-nine patients were identified in 2012 with nephrolithiasis and IBS, 272 patients in 2013, and 327 patients in 2014 with a predominance of females (70% female versus 30% male) within this population. Somewhat shocking to this study was that within this patient population from 2012 to 2014, zero patients had ultrasound imaging and only 22 patients had a CT. While medical imaging is low in this group, 104 (10.9%) required an intervention and/or procedure, and of the 104, 44 interventions (42%) were directly related to stone removal. The mean total charge for nephrolithiasis patients with IBS admitted to the hospital from the ED is $26,753, which is like that of the non-IBS nephrolithiasis population.
Clinical decision-making is an integral part of patient outcomes and its accuracy is a necessary entity in today's healthcare environment. Several variables in this study were compared to better understand the prevalence of nephrolithiasis and to analyze and investigate a predictive model for improving patient outcomes of nephrolithiasis. Statistical analyses of the HCUP NEDS 2012, 2013, and 2014 datasets were done using descriptive statistics for frequencies and percentages, as well as multiple regression analysis including linear and stepwise regression models.
While descriptive statistics of this study align with the literature review of medical imaging in nephrolithiasis patients, ultrasound being utilized more frequently than CT upon presentation to the ED, which confirms the lack of adherence to the ACR's guidelines, and the lack of laboratory testing and/or results available in the dataset prove difficult to validate adherence to the recommended laboratory studies and metabolic evaluations. Multiple regression models demonstrate that principal procedures, which include medical imaging modalities such as ultrasound and CT, as well as minor and or major procedures and interventions, from the inpatient discharge record represent a statistically significant contribution to nephrolithiasis. Based on the information provided in the descriptive statistical analysis of IBS, a variety of principal procedures returned were related to nephrolithiasis demonstrating the significance of these to nephrolithiasis in IBS patients.
This study shows that although many patients are discharged directly from the ED, there is a significant number of patients admitted to the hospital for nephrolithiasis, many of whom were assessed with ultrasound first. This confirms the hypothesis that patients assessed with ultrasound before CT have poorer outcomes and accrue greater healthcare costs.
Certain comorbidities associated with nephrolithiasis may produce larger stones due to chronic dehydration; however, due to the lack of available data within the HCUP NEDS datasets on plasma osmolality, this theory is difficult to confirm. We can see that obesity, diabetes, and IBS are all contributing factors to nephrolithiasis risk and prove the hypothesis that certain comorbidities are most closely linked to nephrolithiasis. Interventions in IBS patients are often more invasive and fall into the minor diagnostic and therapeutic procedure class categories, as well as a large number in the major therapeutic category. Patients with IBS have a two-fold likelihood of requiring an intervention than that of the general nephrolithiasis population, resulting in poorer outcomes and greater healthcare costs.
Hydration is a significant factor for kidney stone formers, and the literature highlights the need for increased fluids in prevention. However, there is little information available in the HCUP NEDS datasets for dehydration assessment upon presentation to the ED. Prevention of nephrolithiasis would release a significant healthcare burden on the patient and the national healthcare system. Increased fluid intake should increase urine output and ensure mechanical diuresis to prevent the stagnation of organic materials and the formation of stones. Increased water intake has resulted in the attenuation of circulating copeptin, serving as a simple and cost-effective intervention to reduce circulating vasopressin, which conserves water and decreases urine volume [8]. Chronic dehydration can be assessed with plasma osmolality and copeptin levels and should be considered as part of the initial assessment upon presentation to the ED with suspected nephrolithiasis.

Limitations of study
The HCUP NEDS database provided insight and results in this study of nephrolithiasis; however, it is limited due to its retrospective nature. The databases are missing information about ED charges for a variety of weighted percentages, 15% for 2012 and 2013, and 16% for 2014 [5][6][7], and focuses on entry-level versus patient-level records. Therefore, patients who may revisit the ED multiple times each year may be present as separate entries. The NEDS is also limited by direct admission from the ED to the hospital; only one discharge record is included making differentiation of procedures performed in the ED versus part of the inpatient stay unattainable.
There is potential misclassification of variables inherent to administrative data which relies heavily on diagnosis codes entered by individual hospital personnel. Because IBS is not always reported, or diagnosed, the overall data may be incomplete or exposed to bias. Therefore, this exploratory study and analysis may be skewed. Additionally, the lack of ICD-9 coding for plasma osmolality, and perhaps the sheer lack of testing, prevent this study from confirming the hypothesis of plasma osmolality and copeptin levels as a link to nephrolithiasis risk and stone formation.
Despite these limitations, the HCUP NEDS datasets allow for a national perspective of trending in nephrolithiasis patient populations and have provided a great deal of insight.

Conclusions
The HCUP NEDS datasets for 2012, 2013, and 2014 have provided insight into the prevalence of nephrolithiasis based on age, gender, comorbidity, diagnostic assessment, and procedures or interventions required during ED and hospital care. The highest rate of nephrolithiasis occurrence is between 30 and 42 years of age and is more common in the general population of males. However, female patients among the same age group with IBS have a higher prevalence of nephrolithiasis than males. Adherence to medical imaging guidelines and laboratory testing for nephrolithiasis is lacking. This study demonstrates that patients assessed with ultrasound first are more likely to be admitted to the hospital than those assessed following the ACR's recommended low-dose CT imaging and accrue greater healthcare costs overall. Although many studies outline this lack of adherence to the ACR's appropriateness criteria, there are no available reports on patient outcomes or hospital admissions. This research is the first to report on patient outcomes, specifically IBS as a comorbidity, as well as the prevalence of procedures and interventions in nephrolithiasis, and total charges for ED versus inpatient hospital care.

Additional Information Disclosures
Human subjects: Consent was obtained or waived by all participants in this study. Animal subjects: All authors have confirmed that this study did not involve animal subjects or tissue. Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Intellectual property info: This dissertation research is held under copyright by the primary author (Kori L. Stewart). Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.