Hospital volume and outcome in inguinal hernia repair: analysis of routine data of 133,449 patients

The analyses were based on anonymized routine data of the German Local General Sickness Fund (AOK). These included diagnoses and procedures related to hospital care, drug prescriptions as well as insured persons’ master data such as age, gender and survival status.

Included in the analysis were hospital inpatients and outpatients who during the initial hospital stay underwent inguinal hernia repair (Operations and Procedures Key [OPS]: 5-530) between 2013 and 2015 and for whom inguinal hernia (International Statistical Classification of Diseases and Related Health Problems [ICD]-10: K40) was documented as primary diagnosis. Patients under 18 years of age, with simultaneous appendectomy (OPS: 5-470, 5-471), cholecystectomy (OPS: 5-511) or cancer disease (ICD-10: C00–C97, D00–D09, D37–D48; OPS: 8-54) as well as patients who had undergone surgery of the digestive tract (OPS: 5-42 to 5-54) within 365 days prior to hospital admission were excluded (Fig. 1). The applicable criteria had been formulated for the inguinal hernia repair service area by the Abdominal Surgery Expert Panel within the framework of the quality assurance with routine data (QSR) project of the Scientific Institute of the AOK (WIdO) [16].

For each patient a follow-up period of 365 days from discharge after the initial hospital stay was reviewed. Additional inguinal hernia operations within this period were only assessed as outcomes. In compliance with data protection regulations, the AOK data used were anonymized such that the patient’s identity was not known and could not be determined.

The endpoints employed corresponded to the definitions of the quality indicators used for inguinal hernia repair, which had been compiled by the Scientific Institute of the AOK [16]. The endpoints analyzed were 30-day mortality, recurrence operations within 91–365 days, chronic pain requiring treatment within 365 days, and surgical complications within 90 days.

Recurrence operations within 91–365 days were documented on the basis of inguinal hernia repairs (OPS: 5-530) performed on the same side of the body as the primary repair procedure. Reoperations within 90 days were not documented for that indicator since they tended to relate to technical defects or complications. Such events were ascribed to the endpoint “surgical complications within 90 days”.

Chronic pain requiring treatment within 365 days was documented on the basis of continuous drug prescriptions. A new case of continuous pain management was assumed if patients had been prescribed at least 20 defined daily doses [DDD] of analgesics in at least three-out-of-four quarters (Anatomic Therapeutic Chemical classification [ATC]: M01, N02A N02B) and had not been on continuous pain management prior to surgery (≥ 20 DDDs in ≥ 3-out-of-4 quarters before admission).

Surgical complications within 90 days consisted of reoperations in the abdominal and inguinal region, urologic reoperations, wound infections, bleeding requiring reoperation as well as resuscitation, pulmonary embolism and thrombosis, and general complications in association with surgical procedures [16]. The applicable ICD-10 and OPS codes are illustrated in Tables 1 and 2.

Since only data belonging to AOK-insured persons were available, the hospital volume was projected based on the AOK inguinal hernia repair cases described above and on a hospital’s total proportion of AOK cases. Based on each hospital’s annual caseload, the AOK cases were divided into five categories (1–50, 51–75, 76–100, 101–125 and ≥ 126 cases) on an annual basis. In each case descriptive statistics were calculated as a total as well as for the individual volume categories. Trends in respect to the volume categories were verified with the Cuzick test using a significance level of 5%. All key figures given refer in each case to the evaluable caseload. Patients without complete follow-up and who did not experience a complication event within the follow-up period were censored.

The effect of the hospital volume on the endpoints was calculated using multiple logistic regression models. Hospitals with the highest caseloads were used as reference category. The regression models included the hospital volume while also making adjustment for age, gender, surgical technique, comorbidities, and other risk factors such as recurrence status, incarceration, extent of the operation, and preoperative medication. All parameters were defined as dichotomous variables. Age was defined on the basis of dichotomous categorical variables dividing the AOK cases into quintiles. Comorbidities were identified as per the Elixhauser definitions [17]. Cancer diseases were not considered since patients with such disease were not included in the data set. The risk factor obesity was categorized in accordance with the breakdown given in the ICD-10 catalog as grade I (BMI ≥ 30 and < 35), II (BMI ≥ 35 and < 40), and III (BMI ≥ 40) obesity as well as obesity grade unspecified. The surgical techniques were divided into three main groups: open mesh procedure (OPS: 5-530 [3, 7].[0, x]), minimally invasive mesh procedure (OPS: 5-530 [3, 7] [1, 2]), and meshfree procedure (OPS: 5-530.[0, 1, 2, 4, 5, 8, x, y]). Recurrence status, incarceration, extent of operation, and other risk factors were defined in accordance with the AOK’s Quality Assurance of Inpatient Care with Routine Data (QSR) specifications [16]. Model selection was performed using a stepwise backward algorithm based on a model with all adjustment variables. The calculated model was then expanded to include the factors influencing the hospital volume. All models were examined for collinearity by calculating variance inflation factors (VIF). The VIF of a covariate measures the extent of collinearity present [18‐22]. If no collinearity is present for a covariate, its VIF equals 1. As the collinearity increases the VIF increases. In the literature different VIF thresholds are used. Kleinbaum et al. [18] and Montgomery et al. [19] used a value of 10 while Zuur et al. [22] used a more stringent value of 3.

For endpoints significantly impacted by the hospital volume the number needed to treat (NNT) was calculated with the logistic model [23]. If theoretically patients were reassigned from one category to the reference category, the NNT is the number of patients needed to be treated to prevent one complication. The total number of preventable complications of a category is equal to the quotient of the respective caseload and NNT.

All evaluations were performed with the software STATA14.2 (StataCorp, College Station, Texas).

Table 4 illustrates the endpoint frequencies. A surgical complication within 90 days occurred in 4.2% of patients and recurrence operation within 91–365 days was performed in 1.0% of cases. Chronic pain requiring treatment within 365 days was observed in 2.9% of patients. The mortality rate within 30 days of admission was 0.3%. All four endpoints exhibited a significant trend and declined in line with a rising caseload volume. The proportionately greatest decrease of 32.8% was observed in the recurrence operations. Their proportion declined from 1.4% in the lowest caseload to 0.9% in the highest caseload volume category.

The results of logistic regression analysis are presented in Table 5. The hospital volume had a significant influence on the risk-adjusted recurrence rate. The risk of recurrence operation in the two lowest caseload categories (1–50 cases and 51–75 cases per year) compared with the highest caseload category was increased by 53% and 24%, respectively (OR: 1.53 and 1.24). Had these cases been treated in the same way as the highest caseload category 22 (36.1%) and 27 (20.9%), respectively, of recurrence operations could possibly have been prevented in these categories (NNT: 212 and 451, respectively). The decreasing odds ratios in line with increasing caseload point to a linear volume-outcome relationship. Other factors associated with an increasing risk were, e.g., the initial recurrence status, bilateral operation and simultaneous repair of a femoral hernia, drug abuse, and disorders of the water and electrolyte balance as well as of the acid–base balance. The use of a mesh procedure as well as patient age ≥ 76 years reduced the risk of recurrence procedure.

For the endpoints surgical complications within 90 days, chronic pain requiring treatment within 365 days and 30-day mortality, the hospital volume was not found to have any significant influence. Factors that greatly increased the risk of surgical complications were the presence of gangrene, an intestinal procedure, grade III obesity or disorders of the water and electrolyte balance as well as of the acid–base balance. For chronic pain requiring treatment the factors patient age ≥ 76 years, grade III obesity and drug abuse exhibited a strong risk-increasing impact. As regards the 30-day mortality, patient age (≥ 76 and 65–75 years) had by far the greatest influence. Other risk-increasing factors were the presence of gangrene, liver diseases, and disorders of the water and electrolyte balance as well as of the acid–base balance.

We calculated VIFs for all covariates of the presented models. Table 6 shows the mean and maximum VIF for each model. With the exception of one covariate used in the model for surgical complications all VIFs were below 3. These results indicate that multicollinearity is not an issue. Especially not in the case of recurrence procedures. In addition not all covariates listed in Table 3 were present in each model. The footnote of Table 5 indicates that the specific covariate was removed from model due to lack of significance. For example, in the model of recurrence procedures only 12 covariates were included.